Msc-Marc-manual Volome A

Marc 2010
Volume A: Theory and User Information

Corporate
MSC.Software Corporation
2 MacArthur Place
Santa Ana, CA 92707
Telephone: (800) 345-2078
FAX: (714) 784-4056
Europe
MSC.Software GmbH
Am Moosfeld 13
81829 Munich
GERMANY
Telephone: (49) (89) 43 19 87 0
Fax: (49) (89) 43 61 71 6
Asia Pacific
MSC.Software Japan Ltd.
Shinjuku First West 8F
23-7 Nishi Shinjuku
1-Chome, Shinjuku-Ku
Tokyo 160-0023, JAPAN
Telephone: (81) (3)-6911-1200
Fax: (81) (3)-6911-1201
Worldwide Web
www.mscsoftware.com
User Documentation: Copyright  2010 MSC.Software Corporation. Printed in U.S.A. All Rights Reserved.
This document, and the software described in it, are furnished under license and may be used or copied only in accordance with the
terms of such license. Any reproduction or distribution of this document, in whole or in part, without the prior written authorization of
MSC.Software Corporation is strictly prohibited.
MSC.Software Corporation reserves the right to make changes in specifications and other information contained in this document
without prior notice. The concepts, methods, and examples presented in this document are for illustrative and educational purposes
only and are not intended to be exhaustive or to apply to any particular engineering problem or design. THIS DOCUMENT IS
PROVIDED ON AN “AS-IS” BASIS AND ALL EXPRESS AND IMPLIED CONDITIONS, REPRESENTATIONS AND WARRANTIES,
INCLUDING ANY IMPLIED WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, ARE
DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS ARE HELD TO BE LEGALLY INVALID.
MSC.Software logo, MSC, MSC., MD Nastran, Adams, Dytran, Marc, Mentat, and Patran are trademarks or registered trademarks of
MSC.Software Corporation or its subsidiaries in the United States and/or other countries.
NASTRAN is a registered trademark of NASA. Python is a trademark of the Python Software Foundation. LS-DYNA is a trademark of
Livermore Software Technology Corporation. All other trademarks are the property of their respective owners.
This software may contain certain third-party software that is protected by copyright and licensed from MSC.Software suppliers.
METIS is copyrighted by the regents of the University of Minnesota. HP MPI is developed by Hewlett-Packard Development Company,
L.P. MS MPI is developed by Microsoft Corporation. PCGLSS 6.0, copyright © 1992-2005 Computational Applications and System
Integration Inc. MPICH Copyright 1993, University of Chicago and Mississippi State University. MPICH2 copyright © 2002, University
of Chicago.
Use, duplication, or disclosure by the U.S. Government is subject to restrictions as set forth in FAR 12.212 (Commercial Computer
Software) and DFARS 227.7202 (Commercial Computer Software and Commercial Computer Software Documentation),
as applicable.
MA*V2010*Z*Z*Z*DC-VOL-A
Contents
Marc Volume A: Theory and User Information
Contents
Preface
About This Manual 20
Purpose of Volume A 20
Contents of Volume A 20
How to Use This Manual 21
1 The Marc System

Marc Programs 24
Marc for Analysis 24
Marc Mentat or MD Patran for GUI 25
Structure of Marc 25
Procedure Library 25
Material Library 26
Element Library 26
Program Function Library 26
Features and Benefits of Marc 26
2 Program Initiation
Marc Host Systems 30
Workspace Requirements 31
Marc Workspace Requirements 31
File Units 33
Program Initiation 36
Examples of Running Marc Jobs 39

4 Marc Volume A: Theory and User Information
3 Data Entry
Input Conventions 42
Input of List of Items 43
Examples of Lists 45
Table Driven Input 45
Table Input 46
Parameters 50
Model Definition Options 50
History Definition Options 50
REZONE Option 51
4 Introduction to Mesh Definition

Direct Input 54
Element Connectivity Data 54
Nodal Coordinate Data 58
Activate/Deactivate 59
User Subroutine Input 59
MESH2D 59
Block Definition 59
Merging of Nodes 60
Block Types 60
Symmetry, Weighting, and Constraints 63
Additional Options 64
Marc Mentat 64
FXORD Option 65
Major Classes of the FXORD Option 65
Recommendations on Use of the FXORD Option 69
Incremental Mesh Generators 69
Bandwidth Optimization 70
Rezoning 71
Substructure 71
Technical Background 72
Scaling Element Stiffness 74
CONTENTS 5
BEAM SECT Parameter 74

Orientation of the Section in Space 74
Definition of the Section 74
Error Analysis 79
Local Adaptivity 79
Number of Elements Created 79
Boundary Conditions 80
Location of New Nodes 81
Adaptive Criteria 82
Automatic Global Remeshing 85

Remeshing Criteria 90
Remeshing Techniques 91
MD Patran Tetrahedral Mesher 93
5 Structural Procedure Library

Linear Analysis 100
Accuracy 101
Error Estimates 102
Adaptive Meshing 102
Fourier Analysis 102
Nonlinear Analysis 105

Geometric Nonlinearities 109
Eulerian Formulation 118
Arbitrary Eulerian-Lagrangian (AEL) Formulation 119
Nonlinear Boundary Conditions 119
Buckling Analysis 122
Perturbation Analysis 123
Computational Procedures for Elastic-Plastic Analysis 128
Creep 140
Viscoelasticity 145
Viscoplasticity 146
Fracture Mechanics 147

Linear Fracture Mechanics 148
Nonlinear Fracture Mechanics 150
Numerical Evaluation of the J-integral 151
Numerical Evaluation of the Energy Release Rate with the VCCT
Method 154
Automatic Crack Propagation 160
Dynamic Fracture Methodology 163

Modeling Considerations 164
Dynamic Crack Propagation 164
Mesh Splitting 165
Dynamics 168
Modal (Eigenvalue) Analysis 168
Harmonic Response 171
Spectrum Response 173
Transient Analysis 174
Inertia Relief 187

Rigid Body Mode Evaluation 187
Rigid-Plastic Flow 190

Steady State Analysis 191
Superplasticity 193
Soil Analysis 195

Technical Formulation 196
Mechanical Wear 199

Design Sensitivity Analysis 202
Design Optimization 205

Approximation of Response Functions Over the Design Space 206
Improvement of the Approximation 208
The Optimization Algorithm 208
Marc User Interface for Sensitivity Analysis and Optimization 209
Defined Initial State with Result Data from Previous Analysis (including
AXITO3D) 212
Load and Displacement Boundary Conditions Transfer for
PRE STATE Option 213
Steady State Rolling Analysis 213

Kinematics 213
Inertia Effect 215
Rolling Contact 216
Steady State Rolling with Marc 216
CONTENTS 7
Structural Zooming Analysis 216

Element Types Supported 217
Uncoupled Thermal Stress Analysis 219
Cure-Thermal-Mechanically Coupled Analysis 220

Cure Kinetics 221
Cure Shrinkage Strain 223
References 225
6 Nonstructural Procedure Library

Heat Transfer 230
Thermal Contact 231
Convergence Controls 232
Temperature Effects 234
Initial Conditions 235
Surface Energy 240
Thermochemical Ablation and Surface Energy Balance 241
Mathematical Presentation 241
Mechanical Erosion 248
Mechanical Erosion by Other Actions 249
Pyrolysis 249
Coking 253
Presentation of the Energy Equation 257
Ablation 259
Welding 268
Radiation 275
Conrad Gap 289
Channel 290
Output 291
Diffusion 293
Hydrodynamic Bearing 295

Electrostatic Analysis 300

Magnetostatic Analysis 304

Electromagnetic Analysis 316
Piezoelectric Analysis 320

Strain Based Piezoelectric Coupling 323
Acoustic Analysis 324

Rigid Cavity Acoustic Analysis 324
Fluid Mechanics 325

Finite Element Formulation 328
Penalty Method 331
Solid Analysis 332
Solution of Coupled Problems in Fluids 332
Degrees of Freedom 333
Element Types 333
Coupled Analyses 335

Thermal Mechanically Coupled Analysis 337
Coupled Acoustic-Structural Analysis 339
Fluid/Solid Interaction – Added Mass Approach 342
Coupled Electrostatic-Structural Analysis 344
Coupled Magnetostatic-Structural Analysis 346
Coupled Thermal-Electrical Analysis (Joule Heating) 348
Coupled Electrical-Thermal-Mechanical Analysis 352
Coupled Magnetostatic-Thermal Analysis 354
Coupled Electromagnetic-Thermal Analysis 355
References 356
7 Material Library
Linear Elastic Material 360
Composite Material 362

Layered Materials 363
Classical Lamination Theory for Multi-Layered Shells 366
Material Preferred Direction 367
CONTENTS 9
Material Dependent Failure Criteria 371

Interlaminar Shear for Thick Shell, Beam, Solid Shell and 3D Composite Brick
Elements 384
Interlaminar Stresses for Continuum Composite Elements 386
Progressive Composite Failure 386
Mixture Model 389
Gasket 393
Constitutive Model 393
Nonlinear Hypoelastic Material 398
Thermo-Mechanical Shape Memory Model 413

Transformation Induced Deformation 415
Constitutive Theory 417
Phase Transformation Strains 417
Experimental Data Fitting for Thermo-mechanical Shape Memory Alloy 419
Mechanical Shape Memory Model 423

Experimental Data Fitting for Mechanical Shape Memory Alloy 426
Conversion from Thermo-Mechanical to Mechanical SMA 427
Elastomer 429
Updated Lagrange Formulation for Nonlinear Elasticity 438
Time-independent Inelastic Behavior 438

Yield Conditions 441
Mohr-Coulomb Material (Hydrostatic Stress Dependence) 447
Buyukozturk Criterion (Hydrostatic Stress Dependence) 448
Powder Material 449
Obtaining Crush Curve and Shear Failure Parameters by Curve Fitting in
Marc 459
Work or Strain Hardening 463
Flow Rule 470
Constitutive Relations 470
Time-independent Cyclic Plasticity 473
Time-dependent Inelastic Behavior 476

Creep (Maxwell Model) 480
Oak Ridge National Laboratory Laws 485
Swelling 486
Viscoplasticity (Explicit Formulation) 487
Time-dependent Cyclic Plasticity 488
Viscoelastic Material 489
Narayanaswamy Model 500
Temperature Effects and Coefficient of Thermal Expansion 503

Piecewise Linear Representation 505
Temperature-Dependent Creep 506
Coefficient of Thermal Expansion 507
Time-Temperature-Transformation 507
Low Tension Material 510

Uniaxial Cracking Data 510
Low Tension Cracking 511
Tension Softening 511
Crack Closure 512
Crushing 512
Analysis 512
Soil Model 512
Elastic Models 513
Cam-Clay Model 513
Evaluation of Soil Parameters for the Critical State Soil Model 515
Damage Models 524

Ductile Metals 524
Elastomers 526
Cohesive Zone Modeling 529
Nonstructural Materials 537

Heat Transfer Analysis 537
Thermo-Electrical Analysis 538
Coupled Electrical-Thermal-Mechanical Analysis 538
Hydrodynamic Bearing Analysis 538
Fluid/Solid Interaction Analysis – Added Mass Approach 538
Coupled Electrostatic-Structural 539
Fluid Analysis 539
References 539
CONTENTS 11
8 Contact
Definition of Contact Bodies 542
Numbering of Contact Bodies 545
Motion of Bodies 548

Initial Conditions 550
Detection of Contact 550

Shell Contact 552
Neighbor Relations 553
Implementation of Constraints 554
Friction Modeling 556

Coulomb Friction 556
Shear Friction 566
Glue Model 567
Deact Glue 567
Breaking Glue 567
Separation 568
Release 568
Coupled Analysis 569
Thermal Contact 571
Joule Heating 572
Contact in an Electrostatic or Piezoelectric Analysis 573
Contact in Magnetostatic Analysis 573
Contact in Electromagnetic Analysis 573
Contact in Soil Analysis 573
Contact in an Acoustic Analysis 573
Element Considerations 574

2-D Beams 574
3-D Beams 575
Shell Elements 576
Dynamic Impact 577
Global Adaptive Meshing and Rezoning 577

Result Evaluation 578
Tolerance Values 579

Numerical and Mathematical Aspects of Contact 580
Lagrange Multipliers 580
Penalty Methods 581
Hybrid and Mixed Methods 581
Direct Constraints 581
Lagrange Multiplier Procedure 582
Direct Constraint Procedure 582
Solution Strategy for Deformable Contact 589
Iterative Penetration Checking 590
Instabilities 591
Segment-to-Segment Contact 592

Basic Theory 592
Contact Body Definition 594
Contact Detection 595
Force Vector and Stiffness Matrix Contribution 597
Iterative Penetration Checking and Contact Stress Update 599
Separation 600
References 601
9 Boundary Conditions
Loading 605
Loading Types 605
Face ID for Distributed Loads, Fluxes, Charge, Current, Source, Films, and
Foundations 614
Mechanical Loads 618
Fluid Drag and Wave Loads 621
Cavity Pressure Loading 622
Cyclic Loading 627
Thermal Loads 627
Initial Stress and Initial Plastic Strain 628
Heat Fluxes 628
Mass Fluxes and Restrictors 630
Electrical Currents 630
Electrostatic Charges 631
Acoustic Sources 631
CONTENTS 13
Piezoelectric Loads 631

Electrostatic-Structural Loads 632
Magnetostatic Currents 632
Electromagnetic Currents and Charges 632
Kinematic Constraints 632

Transformation of Degree of Freedom 634
Shell Transformation 634
Tying Constraint 637
Rigid Link Constraint 646
Shell-to-Solid Tying 647
Rigid Tying to a Surface Patch 648
Overclosure Tying 650
Insert 654
AUTOMSET 654
Support Conditions 654
Bushings 655
Cyclic Symmetry 658
Nastran RBE2 and RBE3 661
Beam - Shell Offsets 666
Pin Code for Beam Elements 669
Mesh Independent Connection Methods 670
CWELD patch-to-patch connections 673
Parameters for the Projection Process 677
Internal Constraints to Connect Two Surfaces 679
Multiple Surface Connections 679
Coupled Thermo-mechanical Analysis 680
Rules for usage of the GS-node or the GA-, GB-node Pair in the Projection
Process 680
CWELD Properties 682
CWELD Input Styles 685
CWELD Error and Warning Messages 686
CWELD Output 688
Two Dimensional Models 689
CFAST Connections 689
CFAST Properties 689
SWLDPRM, Special CWELD/CFAST Parameters 691
10 Element Library
Truss Elements 700
Membrane Elements 700
Continuum Elements 700
Beam Elements 701
Plate Elements 702

Shell Elements 702
Heat Transfer Elements 702

Coupled Electrostatic-Structural 703
Fluid/Solid Interaction 703
Soil Analysis 704

Fluid Analysis 704
Special Elements 705

Gap-and-Friction Elements 705
Pipe-bend Element 705
Curved-pipe Element 705
Shear Panel Element 705
Cable Element 705
Rebar Elements 706
Interface Elements 706
Incompressible Elements 706

Large Strain Elasticity 707
Large Strain Plasticity 707
Constant Dilatation Elements 707

Reduced Integration Elements 707
Continuum Composite Elements 708

CONTENTS 15
Fourier Elements 708
Semi-infinite Elements 708
Cavity Surface Elements 709

Assumed Strain Formulation 709
Follow Force Stiffness Contribution 709
Explicit Dynamics 709
Adaptive Mesh Refinement 710
11 Solution Procedures for Nonlinear Systems

Considerations for Nonlinear Analysis 712
Behavior of Nonlinear Materials 713
Scaling the Elastic Solution 713
Load Incrementation 713
Selecting Load Increment Size 715
Fixed Load Incrementation 715
Arc Length Method 716
Residual Load Correction 724
Restarting the Analysis 725
Newton-Raphson Method 726
Modified Newton-Raphson Method 727

Strain Correction Method 728
Direct Substitution 729
Arc-length Methods 729
Singularity Ratio 739

AutoSPC 740
Solution of Linear Equations 740

Direct Methods 742
Iterative Methods 742
Mixed Direct-Iterative Solver 742
Preconditioners 743
Storage Methods 743
Nonsymmetric Systems 744

Complex Systems 744
Iterative Solvers 744
Parallel Direct Solvers 745
Basic Theory 745
Flow Diagram 746
Remarks 747
References 748
12 Output Results
Workspace Information 750
Increment Information 754

Summary of Loads 754
Timing Information 754
Convergence 755
Selective Printout 755

Options 755
Grid Force Balance 757
User Subroutines 757
Restart 757
Element Information 758
Mentat Computed Element Results 759
Solid (Continuum) Elements 759
Shell Elements 759
Beam Elements 760
Gap Elements 761
Linear and Nonlinear Springs 761
Joule Heating Analysis 762
Piezoelectric or Electrostatic-Structural Analysis 762
Electromagnetics Analysis 762
CONTENTS 17
Nodal Information 762

Stress Analysis 763
Reaction Forces 763
Residual Loads 763
Dynamic Analysis 763
Joule Heating Analysis 764
Rigid-Plastic Analysis 764
Piezoelectric or Electrostatic-Structural Analysis 764
Supplementary Information 764

Contact Analysis 764
Post File 765

Forming Limit Parameter (FLP) 765
Program Messages 768
Marc HyperMesh Results Interface 769

Marc SDRC I-DEAS Results Interface 769
Marc - ADAMS Results Interface 770
Status File 772
Element Group Information 772
References 773
13 Parallel Processing
Different Types of Machines 776
Supported and Unsupported Features 776
Matrix Solvers 778
Contact 779
Domain Decomposition 779

Running a Parallel Job 780
Domain Decomposition Methods 782
14 Code Coupling Interfaces

Code Coupling 788
Coupling Regions 790

Time Step Control 792
Shell Elements 792
Parallel Processing 793
References 793
A Finite Element Technology in Marc

Governing Equations of Various Structural Procedures 796
System and Element Stiffness Matrices 798
Load Vectors 799
References 800
B Finite Element Analysis of NC Machining Processes

General Description 802
NC Files (Cutter Shape and Cutter Path Definition) 802
Intersection Between Finite Element Mesh and Cutter 804
Deactivation of Elements 804
Adaptive Remeshing 804

Typical Features for Machining 805
Input of Initial Stresses 805
References 806
INDEX
Preface
Preface
About This Manual 20

Purpose of Volume A 20
Contents of Volume A 20
How to Use This Manual 21
About This Manual

This manual is Marc Volume A, the first in a series of five volumes documenting the Marc Finite Element
program. The documentation of Marc is summarized below. You will find references to these documents
throughout this manual.
Marc Documentation
TITLE VOLUME
Theory and User Information Volume A
Element Library Volume B
Program Input Volume C
User Subroutines and Special Routines Volume D
Demonstration Problems Volume E
Purpose of Volume A
The purpose of this volume is:
1. To help you define your finite element problem by describing Marc’s capabilities to model
physical problems.
2. To identify and describe complex engineering problems and introduce Marc’s scope and
capabilities for solving these problems.
3. To assist you in accessing Marc features that are applicable to your particular problems and to
provide you with references to the rest of the Marc literature.
4. To provide you with the theoretical basis of the computational techniques used to solve
the problem.
Contents of Volume A
This volume describes how to use Marc. It explains the capabilities of Marc and gives
pertinent background information. The principal categories of information are found under the
following titles:
Chapter 1 The Marc System

Chapter 2 Program Initiation
Chapter 3 Data Entry
Chapter 4 Introduction to Mesh Definition
Chapter 5 Structural Procedure Library
Chapter 6 Nonstructural Procedure Library
Chapter 7 Material Library
Chapter 8 Contact
21
Preface
Chapter 9 Boundary Conditions

Chapter 10 Element Library
Chapter 11 Solution Procedures for Nonlinear Systems
Chapter 12 Output Results
Chapter 13 Parallel Processing
Chapter 14 Code Coupling Interfaces
Appendix A Finite Element Technology in Marc
Appendix B Finite Element Analysis of NC Machining Processes
The information in this manual is both descriptive and theoretical. You will find engineering mechanics
discussed in some detail. You will also find specific instructions for operating the various options offered
by Marc.
How to Use This Manual

Volume A organizes the features and operations of the Marc program sequentially. This organization
represents a logical approach to problem solving using Finite Element Analysis. First, the database is
entered into the system, as described in Chapter 3. Next, a physical problem is defined in terms of a mesh
overlay. Techniques for mesh definition are described in Chapter 4. Chapters 5 and 6 describe the various
structural analyses that can be performed by Marc, while Chapter 7 describes the material models that
are available in Marc. Chapter 8 describes the contact capabilities. Chapter 9 discusses constraints, in the
form of boundary conditions. Chapter 10 explains the type of elements that can be used to represent
the physical problem. Chapter 11 describes the numerical procedures for solving nonlinear equations.
Chapter 12 describes the results of the analysis in the form of outputs. Chapter 13 describes the use of
multiple processors when performing an analysis. Finally, Chapter 14 describes the coupling interface to
external solvers that is available through user subroutine programming.
This volume is also designed as a reference source. This means that all users will not need to refer to each
section of the manual with the same frequency or in the same sequence.
Chapter 1 The Marc System
1 The Marc System

Marc for Analysis 24
 Marc Mentat or MD Patran for GUI 25

Procedure Library 25

Material Library 26
 Element Library 26

Program Function Library 26
The Marc system contains a series of integrated programs that facilitate analysis of engineering problems
in the fields of structural mechanics, heat transfer, and electromagnetics. The Marc system consists of the
following programs:
• Marc for Analysis
• Marc Mentat or MD Patran for GUI (For a detailed description of the supported functionalities by
MD Patran, refer to the MD Patran Marc Preference Guide.)
These programs work together to:
• Generate geometric information that defines your structure (Marc and Marc Mentat or
MD Patran)
• Analyze your structure (Marc)
• Graphically depict the results (Marc and Marc Mentat or MD Patran)
Figure 1-1 shows the interrelationships among these programs. Marc Programs discusses the Marc
component programs.
MARC AFEA
Preprocessing Marc Mentat or MD Patran
Marc
Analysis
Postprocessing Marc Mentat or MD Patran
Figure 1-1 The Marc System
Marc Programs
Marc for Analysis
You can use Marc to perform linear or nonlinear stress analysis in the static and dynamic regimes, to
perform heat transfer analysis and electromagnetic analysis. The nonlinearities may be due to either
material behavior, large deformation, or boundary conditions. An accurate representation accounts for
these nonlinearities.
Physical problems in one, two, or three dimensions can be modeled using a variety of elements. These
elements include trusses, beams, shells, and solids. Mesh generators, graphics, and postprocessing
capabilities, which assist you in the preparation of input and the interpretation of results, are all available
CHAPTER 1 25
The Marc System
in Marc. The equations governing mechanics and implementation of these equations in the finite element
method are discussed in Chapters 5, 6, 7, 8, and 11.
Marc Mentat or MD Patran for GUI

Marc Mentat is an interactive computer program that prepares and processes data for use with the finite
element method. Interactive computing can significantly reduce the human effort needed for analysis by
the finite element method. Graphical presentation of data further reduces this effort by providing an
effective way to review the large quantity of data typically associated with finite element analysis.
An important aspect of Marc Mentat is that you can interact directly with the program. Marc Mentat
verifies keyboard input and returns recommendations or warnings when it detects questionable input.
Marc Mentat checks the contents of input files and generates warnings about its interpretation of the data
if the program suspects that it may not be processing the data in the manner in which you, the user, have
assumed. Marc Mentat allows you to graphically verify any changes the input generates.
Marc Mentat can process both two- and three-dimensional meshes to do the following:
Generate and display a mesh
Generate and display boundary conditions and loadings
Perform postprocessing to generate contour, deformed shape, and time history plots
The data that is processed includes:
Nodal coordinates
Element connectivity
Nodal boundary conditions
Nodal coordinate systems
Element material properties
Element geometric properties
Element loads
Nodal loads/nonzero boundary conditions
Element and nodal sets
Structure of Marc
Marc has four comprehensive libraries, making the program applicable to a wide range of uses. These
libraries contain structural procedures, materials, elements, and program functions. The contents of each
library are described below.
Procedure Library
The structural procedure library contains procedures such as static, dynamic, creep, buckling, heat
transfer, fluid mechanics, and electromagnetic analysis.
The procedure library conveniently relates these various structural procedures to physical
phenomena while guiding you through modules that allow, for example, nonlinear dynamic and
heat-transfer analyses.
Material Library
The material library includes many material models that represent most engineering materials. Examples
are the inelastic behavior of metals, soils, and rubber material. Many models exhibit nonlinear properties
such as plasticity, viscoelasticity, and hypoelasticity. Linear elasticity is also included. All properties may
depend on temperature.
Element Library
The element library contains over 200 elements. This library lets you describe any geometry under any
linear or nonlinear loading conditions.
Program Function Library

The program functions such as selective assembly, user-supplied subroutines, and restart, are tailored for
user-friendliness and are designed to speed up and simplify analysis work. Marc allows you to combine
any number of components from each of the four libraries and, in doing so, puts at your disposal the tools
to solve almost any structural mechanics problem.
Features and Benefits of Marc

Since the mid-1970s, Marc has been recognized as the premier general purpose program for nonlinear
finite element analysis. The program’s modularity leads to its broad applicability. All components of the
structural procedure, material, and element libraries are available for use, allowing virtually unlimited
flexibility and adaptability.
Marc has helped analyze and influence final design decisions on
Automotive parts Space vehicles

Nuclear reactor housings Electronic components
Biomedical equipment Steam-piping systems
Offshore platform components Engine pistons
Coated fiberglass fabric roof structures Tires
Rocket motor casings Jet engine rotors
Ship hulls Welding, casting, and quenching processes
Elastomeric motor mounts Large strain metal extrusions
CHAPTER 1 27
The Marc System
Marc’s clients gained the following benefits not attainable through other numerical or experimental
techniques. These benefits include:
Accurate results for both linear and nonlinear analysis
Better designs, which result in improved performance and reliability
The ability to model complex structures and to incorporate geometric and material
nonlinear behavior
Documentation, technical support, consulting, and education provided by
MSC.Software Corporation
Availability of Marc on most computers
Efficient operation
Chapter 2 Program Initiation
2 Program Initiation

Marc Host Systems 30
 Workspace Requirements 31

File Units 33

Program Initiation 37
 Examples of Running Marc Jobs 40
Chapter 2 explains how to execute Marc on your computer. Marc runs on many types of machines. All
Marc capabilities are available on each type of machine; however, program execution can vary among
machine types. The allocation of computer memory depends on the hardware restrictions of the machine
you are using.
Marc Host Systems

Marc runs on most computers. Table 2-1 summarizes the types of machines and operating systems on
which Marc currently runs.
Table 2-1 Marc Machines and Operating Systems
FORTRA C
N Versio Also
Vendor OS Hardware Version n Default MPI Works On
HP (64-bit) 4 HPUX 11.11 PA2.0 f90 3.1 A.03.73 HP MPI 2.0
HP (64-bit) 4 HPUX 11.23 Itanium 2 f90 3.3 A.06.20 HP MPI 2.2
IBM (64-bit)4 AIX 5.3 Power 6 xlf 11.1 cc 9.0.0 MPICH1
SGI (Altix 64-bit)2, 4, Linux 2.6.5-7.139- Itanium 2 Intel 10.1 Intel 10.1 SGI MPT
10
sn2 (Propack 4.0) 1.11.1
Sun (64-bit)4, 10 Solaris 10 UltraSparc III f90 8.3 cc 5.9 MPICH1
4, 10
Sun (64-bit) Solaris 10 x86 f90 8.3 cc 5.9 SUN HPC 7.1
Linux (32-bit)8, 9 RedHat AS 4.5 Intel Pentium Intel 10.1 Intel 10.1 HP MPI 2.35
or equiv.
Linux (64 bit)4, 8 RedHat AS 4.5 Itanium 2 Intel 10.1 Intel 10.1 HP MPI
2.2.5.15
Linux (64-bit)4, 8 RedHat AS 4.5 Intel EM64T Intel 10.1 Intel 10.1 Intel MPI 3.16 AMD Opteron, SuSE 10,
RedHat 5
Intel (32-bit)8, 9 Windows XP SP3 Intel Pentium Intel 10.111 Intel 10.1 Intel MPI 3.1 Vista 32, Windows 7,
or equiv. Intel 11.0, Intel 11.112
Intel (64-bit)4, 8, 9 Windows Server Intel EM64T Intel 10.111 Intel 10.1 Intel MPI 3.17 Windows XP 64, Vista
2003 x64 64, Windows 7,
Intel 11.0, Intel 11.112
1
Hardware MPI version also available (via maintain in /tools directory).
2
Supports Solver 6.
3 Supports multi-threading.
4
Supports true 64-bit version.
5
Supports the Intel MPI 3.1
6 Supports the HP MPI 2.3
7
Supports the Microsoft MPI 1.0 (SP1).
8
Supports the PARDISO Solver
9 Supports the MUMPS Solver
10
DMP (network DDM) is not supported
11
Microsoft Visual Studio 2005 must be installed
12 Newer version of Intel Fortran Version 11.1 and Microsoft Visual Studio 2008 is not supported
CHAPTER 2 31
Program Initiation
Workspace Requirements
Computing the amount of workspace required by Marc is a complex function of many variables. The
most efficient method is to use the default values for the allocation. The program dynamically
acquires memory if necessary and if available. In some situations, it is advantageous to initially
allocate an amount of memory as described below. The following sections discuss workspace
requirements for Marc.
Marc Workspace Requirements

The workspace used by Marc is allocated in separate parts. One part is referred to as general memory
and contains items such as element stiffness matrices, assembled global stiffness, and mass matrices and
decomposed operator matrix for certain matrix solvers. The initial amount of memory for this part can
be entered by the user and the program dynamically allocates more memory if necessary. If no initial
memory is specified, it is automatically allocated as needed. Other parts of the dynamically allocated
workspace can not be influenced by the user. This includes data for elements, vectors, tables, sets, contact
bodies, kinematic boundary conditions, transformations, tying, the so-called incremental backup, solver
workspace for certain solvers among other things. These are all allocated separately. The incremental
backup is an extra copy of stress tensors and similar quantities and is used for the Newton-Raphson
iterations in a non-linear analysis. If the cut-back feature is activated, more data is stored in this part to
allow for redoing the increment with a modified time step if a failure occurs.
General Memory stored in common/space/
Basic Data Assembled Stiffness Matrix (2) Decomposed Stiffness Matrix (3)
Second Group of Memory stored in individual dynamically allocated vectors
Element Data (1) Vectors Sets Tables Contact Data Boundary Conditions Transformation Tying Attach
Incremental Backup (4) Decomposed Stiffness Matrix (5)
1. Element Data memory allocation is reduced if ELSTO is used. This also reduces Incremental
Backup memory.
2. Assembled stiffness matrix memory allocation is reduced if out-of-core solver is used.
3. Decomposed Stiffness matrix is for solver type 0, 4, and if Cholesky preconditioner is used,
solver type 2.
4. Incremental Backup memory allocation is reduced if IBOOC is used.
5. Decomposed Stiffness matrix is for solver type 6 and 8.
During the analysis, some of the blocks may grow because of changes in the model. In particular because
of local or global adaptive meshing, the Element Data, Vector Data, Assembled Stiffness Matrix, and
Decomposed Stiffness Matrix may grow in size because of the addition of elements and nodes. The
Assembled Stiffness and Decomposed Stiffness matrices also expand due to changes in the bandwidth of
the system. Changes in the bandwidth occur when deformable-deformable or self-contact occur. These
changes can have a dramatic effect of the amount of memory required. The program
dynamically requests additional memory. If this memory is not available, it activates one or more
of the out-of-core options.
The amount of memory that can be used by an analysis is limited by the hardware employed in the
operating system and by internal restrictions within Marc. On 32 bit systems, it is, in general, not possible
to allocate more than two GB of memory for a process. The exception is Windows 2003 Server (which
allows three GB). Even on this platform, the size of the largest vector is two GB. If this limit is reached,
the memory allocation request by Marc fails. Also, a failure to allocate memory can occur if other
processes are using large amounts of memory. A typical output when a memory request fails is
memory request of 250000000 words failed
This occurs when Marc sends a memory allocation request (using the C function malloc) and the system
refuses the request. There is also an internal restriction in the standard version of Marc. It uses standard
Fortran 32-bit integers as pointers in vectors, and the largest amount of memory that can be addressed
with this is 8 GB on a 64-bit system. This limit applies to each part of the separately allocated memory.
For example, in a job where the general memory requires 4 GB, the solver needs 6 GB and the
incremental backup needs 4 GB. This job can be run provided that the machine has at least 14 GB of
available memory. For a parallel analysis, the 8 GB restriction, as well as the 2 GB limit for 32 bit system,
is for each domain of the job since each domain corresponds to a separate process. In a separate set of
versions of Marc utilizing 64-bit integers, this 8 GB restriction is removed.
The memory allocation for the general memory and matrix solver can be affected by the user. The
ALLOCATE parameter specifies the amount of memory that is initially allocated for this part. This
amount is allocated regardless of whether it is used or not. The default is set to a small value and the
memory grows as needed.
For large problems, you may want to get an estimate of the workspace requirements for running a job
without actually executing the analysis. To do this, insert the STOP parameter to exit the program
normally after the workspace is allocated. Marc prints out a summary of the memory needed. It should
be noted that the reported workspace with this option is only an estimate. Some parts of the memory will
grow during the analysis and there is no way to predict this without actually running the analysis. For
setting the appropriate initial allocation to avoid memory growth of the general memory (which can be
inefficient), one should look at the entry on general memory of the summary printout. Please note that
the ALLOCATE parameter value in a parallel run refer to the complete model; the specified amount is
divided equally among the domains. Chapter 12 Output Results describes the Marc output related to
memory in more detail.
By default, the data is stored in-core. There are three out-of-core storage options in Marc.
• Out-of-core element data storage (the ELSTO parameter also implies IBOOC)
• Out-of-core storage of incremental backup (the IBOOC parameter)
• Out-of-core matrix solution
The out-of-core element data option stores element arrays (strains, stresses, temperatures, etc.) on a
file (Fortran unit 3). Data connected with storage of all element quantities occupy a large amount of
space for the more complex shell or three-dimensional elements. Putting this data out-of-core leads to a
CHAPTER 2 33
Program Initiation
slowdown of the execution (disk access is, in general, slower than memory access) but the effect of this
is usually not severe. The ELSTO parameter is used to set this option. The out-of-core option for
incremental backup stores this data in a file (Fortran unit 29). If the element data is out-of-core, the
incremental backup is automatically out-of-core as well. The slow-down related to incremental backup
out-of-core is much less than the one related to element data out-of-core.
The out-of-core matrix solution has a more severe effect on the execution speed. Not all matrix solvers
support this option. It is not supported by the iterative sparse solver (Marc solver number 2) and the
hardware vendor provided solver by HP (solver 6). The multifrontal direct solver (solver 8) does support
out-of-core matrix solution.
In the case that a memory allocation request fails, Marc automatically switches to out-of-core storage
unless it is too late to do this at the point where the allocation failure occurs. If the memory allocation
for element data or incremental backup fails, this part is put out-of-core. The incremental backup part
can also automatically be put out-of-core to free up memory for the matrix solver. If the default solver is
used (multifrontal, solver 8), there is a special way for checking if out-of-core is needed. The solver
workspace is allocated in two parts. One in general memory for the assembly and one in separate
memory. If the in-core assembly part takes more than 25% of the available memory, the matrix assembly
is done out-of-core. This ensures that this decision is done early enough so there is memory left for other
parts of the job. The actual LDU decomposition of the matrix is done out-of-core if the in-core
decomposition requires more than the available memory to avoid the usage of swap memory during the
decomposition. The available memory is set in the file tools/include (for Windows,
tools\include.bat) in the installation directory as the variable MEMLIMIT. For most platforms
(Unix/Linux except HP PA-RISC), it is set to the amount of physical memory on the machine using a
shell command. This can be set explicitly at the installation or specified with the start-up script
run_marc with the option -ml new_memlimit. For Windows and HP PA-RISC, the automatic setting
of the amount of physical memory is done in the program.
File Units
Marc uses auxiliary files for data storage in various ways. Particular FORTRAN unit numbers are used
for certain program functions (for example, ELSTO, RESTART). Table 2-2 lists these file unit numbers.
Note: On most systems, the files are referenced by file names as well as by the file
Several of these files are necessary for solving most problems. The program input file and program
output file are always required.
Table 2-2 FORTRAN File Units Used by Marc
File name Unit Description File Type
jidname.log 0 Analysis sequence log file sequential access, formatted
jidname.t01 1 Usually contains mesh data random access, formatted
jidname.t02 2 OOC* solver scratch file random access, binary
jidname.t03 3 Element data storage (see ELSTO parameter) random access, binary
*OOC denotes Out-Of-Core solution.
Table 2-2 FORTRAN File Units Used by Marc (continued)

jidname.dat 5 Data input file sequential access, formatted
jidname.out 6 Output file sequential access, formatted
jidname.t08 8 Restart file, written out sequential access, binary
ridname.t08 9 Restart file to be read in from a previous job sequential access, binary
jidname.t11 11 OOC* solver scratch file sequential access, binary
jidname.t16 16 Post file, written out sequential access, binary
ridname.t16 17 Post file to be read in from a previous job sequential access, binary
jidname.t18 18 Mesh optimization correspondence table sequential access, formatted
jidname.fem 18 From Marc to external mesher sequential access, formatted
jidname.t19 19 Post file, written out sequential access, formatted
ridname.t19 20 Post file to be read in from a previous job sequential access, formatted
jidname_j_.dat 21 Temporary input file when cut-back is used. sequential access, formatted
jidname.t22 22 Subspace iteration scratch file random access, binary
jidname.t23 23 Fluid-solid interaction file sequential access, binary
pidname.t19 24 Temperature post file for CHANGE STATE or post sequential access, formatted
file from previous analysis for PRE STATE or
GLOBALLOCAL or MAP TEMP.
pidname.t16 25 Temperature post file for CHANGE STATE or post sequential access, binary
jidname.t29 29 Incremental backup file when ELSTO, IBOOC is sequential access, binary
used, or insufficient memory exists.
sidname.t31 31 Substructure results file sequential access, binary
jidname.t32 32 Secant method file sequential access, binary
jidname.t33 33 Lanczos scratch file sequential access, binary
material.mat 38 Material data base file sequential access, formatted
jidname.g 39 Intergraph post file sequential access, formatted
jidname.unv 40 I-DEAS Universal post file sequential access, formatted
jidname.t41 41 Post file – Domain Decomposition sequential access, binary
ridname.t42 42 Post file – Domain Decomposition sequential access, formatted
jidname.opt 45 Duplicate load case data file during design sequential access, formatted
optimization run
jidname.t46 46 Design optimization scratch file sequential access, binary
jidname.trk 47 New particle tracking file sequential access, formatted
ridname.trk 48 Old particle tracking file sequential access, formatted
CHAPTER 2 35
Program Initiation

jidname.dat 5 Data input file sequential access, formatted
jidname.out 6 Output file sequential access, formatted
jidname.t08 8 Restart file, written out sequential access, binary
ridname.t08 9 Restart file to be read in from a previous job sequential access, binary
jidname.t16 16 Post file, written out sequential access, binary
ridname.t16 17 Post file to be read in from a previous job sequential access, binary
jidname.t18 18 Mesh optimization correspondence table sequential access, formatted
jidname.fem 18 From Marc to external mesher sequential access, formatted
jidname.t19 19 Post file, written out sequential access, formatted
ridname.t19 20 Post file to be read in from a previous job sequential access, formatted
jidname_j_.dat 21 Temporary input file when cut-back is used. sequential access, formatted
jidname.t22 22 Subspace iteration scratch file random access, binary
jidname.t23 23 Fluid-solid interaction file sequential access, binary
pidname.t19 24 Temperature post file for CHANGE STATE or post sequential access, formatted
pidname.t16 25 Temperature post file for CHANGE STATE or post sequential access, binary
jidname.t29 29 Incremental backup file when ELSTO, IBOOC is sequential access, binary
used, or insufficient memory exists.
jidname.t32 32 Secant method file sequential access, binary
jidname.t33 33 Lanczos scratch file sequential access, binary
material.mat 38 Material data base file sequential access, formatted
jidname.g 39 Intergraph post file sequential access, formatted
jidname.unv 40 I-DEAS Universal post file sequential access, formatted
jidname.t41 41 Post file – Domain Decomposition sequential access, binary
ridname.t42 42 Post file – Domain Decomposition sequential access, formatted
jidname.opt 45 Duplicate load case data file during design sequential access, formatted
optimization run
jidname.t46 46 Design optimization scratch file sequential access, binary
jidname.trk 47 New particle tracking file sequential access, formatted
ridname.trk 48 Old particle tracking file sequential access, formatted

userspecified 49 User default file (see Marc Volume C: Program Input, sequential access, formatted
Appendix C: Default File)
jidname.vfs or 50 Viewfactors for cavity number xx sequential access, formatted
jidname_CXX.vfs
jidname.lck 51 Locking of post file sequential access, formatted
jidname.cnt 52 Dynamic control file sequential access, formatted
jidname.mfd 52 rebar - Mentat interface sequential access, formatted
jidname-bbc.mfd 52 beam-beam contact - Mentat interface sequential access, formatted
jidname.seq 53 Sequence option sequential access, formatted
jidname.rst 54 Load case data sequential access, formatted
jidname.mesh 55 User supplied mesh sequential access, formatted
jidname.feb 55 From 3-D mesher to Marc sequential access, formatted
jidname.pass 56 Auto restart command line sequential access, formatted
jidname.rms 57 2-D outline file for remeshing sequential access, formatted
jidname.domesh 59 Lock files indicating meshing status sequential access, formatted
jidname.donemesh
“do mesh” and “done mesh”
jidname.doview 59 Lock file indicating viewfactor calculation status sequential access, formatted
jidname.doneview
jidname.sltrk 60 New streamline tracking file sequential access, formatted
ridname.sltrk 61 Old streamline tracking file sequential access, formatted
jidname.sts 67 Analysis progress reporting file sequential access, formatted
bbctch.noconv 80 beam-beam contact information sequential access, formatted
jidname.t81 81 Multifrontal OOC scratch file random access, binary
jidname.t82 82 Multifrontal OOC scratch file random access, binary
jidname.t83 83 Multifrontal OOC scratch file sequential access, binary
jidname.t84 84 Multifrontal OOC scratch file sequential access, binary
jidname.t85 85 Multifrontal DDM scratch file sequential access, binary
jidname.fld 91 Forming Limit input file sequential access, formatted
jidnamd.stm 91 Output at streamline integration points - sequential access, formatted
PYROLYSIS parameter.
jidname.rec 96 output file of recession surface - sequential access, formatted
ABLATION parameter.
filename.apt 94 APT file - machining option sequential access, file
filename.ccl 95 CL file - machining option sequential access, file
CHAPTER 2 37
Program Initiation

EXITMSG 97 Exit messages sequential access, formatted
USRDEF 98 User global default file (see Marc Volume C: Program sequential access, formatted
Input, Appendix C: Default File)
jidname.t08 99 Base restart file for DDM sequential access, formatted
jidname.grd 103 Grid Force Balance Output sequential access, formatted
user specified 110 - 119 Include files for input sequential access, formatted
jidnam-dmig* 120 - 130 DMIG output files. sequential access, formatted
jidname.hmr N/A Hypermesh results file sequential access, binary C file
jidname.dump N/A Scratch file used during memory reallocation on sequential access, binary C file
Windows if in-core reallocation fails.
Program Initiation
Procedures (shell script) are set up that facilitate the execution of Marc on most computers. These
procedures invoke machine-dependent control or command statements. These statements control files
associated with a job.
This shell script submits a job and automatically takes care of all file assignments. This command must
be executed at the directory where all input and output files concerning this Marc job are available. To
use this shell script, every Marc job should have a unique name qualifier and all Marc output files
connected to that job uses this same qualifier. For restart, post, change state, all default Marc Fortran
units should be used.
To actually submit a Marc job, the following command should be used:
run_marc -prog prog_name -jid job_name -rid rid_name -pid pid_name \
-sid sid_name -queue queue_name -user user_name -back back_value \
-ver verify_value -save save_value -vf view_name -def def_name \
-nprocd number_of_processors (for Single Input file runs, use -nps
number_of_processors) -nthread number_of_threads \
-dir directory_where_job_is_processed -itree message_passing_type \
-host hostfile (for running over the network) -pq queue_priority \
-at date_time -comp compatible_machines_on_network -cpu time_limit \
-nsolver number of processors (MUMPS) -ml memory_limit -mode i4 -mpi intel-mpi
where the \ provides for continuation of the command line.
Table 2-3 Keyword Descriptions*

Keyword Options Description
-jid (-j) job_name Input file (job) name identification.
-prog (-pr) progname Run saved executable progname.marc from a previous job
(see -user and -save).
-user (-u) user_name User subroutine user_name.f is used to generate a new
executable program called user_name.marc (see -save
and -prog).
-save (-sa) no Do not save the new executable program user_name.marc.
yes Save the executable program user_name.marc for a future
time (see -prog and -user).
-rid -(r) restart_name Identification of previous job that created RESTART file.
-pid (-pi) post_name Identification of previous job that created the post file.
-sid (-si) substructure Identify the job that contains the solution to the external
nodes of the superelement.
-back (-b) yes Run Marc in the background.
no Run Marc in the foreground.
-ver (-v) yes Ask for confirmation of these input options before starting
the job.
no
Start the job immediately.
-def (-de) default_file File name containing user defined default data.
-nprocd (-np) number Number of domains for parallel processing.
-nprods (-nps) number Number of domains for parallel processing using a Single
Input file.
-nthread number Number of threads per parallel matrix solver (solver 6, 8, and
11) on hardware that supports this capability.
-nsolver number Number of processes to use for the MUMPS parallel matrix
solver (solver 12).
-dir directory_name Pathname to directory where the job I/O should take place.
Defaults to current directory.
-sdir directory_name Directory where scratch files are placed. Defaults to -dir.
-host (-ho) hostfile Specify the name of the host file for running over a network
(default is execution on one machine only in which case this
option is not needed).
*Default options are shown in bold.
CHAPTER 2 39
Program Initiation
Table 2-3 Keyword Descriptions* (continued)

Keyword Options Description
-comp (-co) yes When machines are compatible in a run over the network.
no When machines are not compatible in a run over the network.
This option is only needed when user subroutines are used.
-ci yes Copy input files automatically to remote hosts for a network
run, if necessary.
no
-cr yes Copy post files automatically from remote hosts used for a
network run, if necessary.
no
-vf viewfactor_name Name of file containing viewfactors for radiation from
previous analysis or from Mentat using either the Monte
Carlo or Hemicube method.
-mo i4 Indicate if the i4 or i8 version is used if both are installed; if
i8 the default unless the run_marc_defaults file is used.
-mpi intel-mpi Select which version of mpi will be used for Domain
ms-mpi Decomposition or the MUMPS solver. If not specified, the
hpmpi
hardware appropriate default shown below will be used:
Alternative
Platform Default MPI MPI
Windows 32 intel-mpi none
Windows 64 intel-mpi ms-mpi
Linux 32 hpmpi intelmpi
Linux 64 hpmpi intelmpi
Linux Itanium hpmpi intelmpi
HP UX hpmpi none
HP UX Itanium hpmpi none
IBM mpich hardware
SGI Altrix hardware none
SGI/Irix mpich hardware
SUN/Solaris/SPARC mpich hardware
SUN/Solaris/8664 hardware none
-ml memlimit Memory limit for deciding if the solver should go out-of-
core. Specified in Mbyte. Defaults to the physical amount of
memory on the machine.
*Default options are shown in bold.
Examples of Running Marc Jobs

Example 1:
run_marc -jid e2x1
This runs the job e2x1 in the background using a single processor.
The input file is e2x1.dat in the current working directory.
Example 2:
run_marc -jid e2x14 -user u2x14 -sav y -nproc 4
This runs the job e2x14 in the background with four processors. The user subroutine is linked with the
Marc library and a new execu module is created as u2x14.marc and saved in the current working
directory after completion of the job.
Example 3:
run_marc -jid e2x14a -prog u2x14 -nproc 4
Use the above saved module u2x14.marc to run the job e2x14a in the background with four processors.
Example 4:
run_marc -jid e3x2a -v no -b no -nproc 2
Run the job e3x2a in the foreground with two processors. The job runs immediately without verifying
any arguments interactively. If there are any input errors in the arguments, the job does not run and the
error message is sent to the screen.
Example 5:
run_marc -jid e3x2b -rid e3x2a
Run the job e3x2b in the background using a single processor. The job uses e3x2a.t08, which is
created from Example 4, as restart file.
Example 6:
run_marc -jid e2x1 -nproc 2
Runs a two processor job on a single parallel machine.
Example 7:
run_marc -jid e2x1 -nproc 2 -host hostfile
Runs a two-processor job over a network. The hosts are specified in the file hostfile.
Example 8:
run_marc -jid e2x1 -nps 2
Runs a two-processor job on a single parallel machine using a single input file.
CHAPTER 2 41
Program Initiation
Example 9:
run_marc -jid e2x1 -nps 2 -host hostfile
Runs a two-processor job over a network using a single input file. The hosts are specified in the
file hostfile.
Example 10:
run_marc -jid e2x10 -mo i8
Runs a job in large integer mode.
Example 11:
run_marc -jid e8x102d -nthread 4
Runs a job using four threads on a machine that supports parallel processing. The Pardiso solver should
also be activated.
Chapter 3 Data Entry
3 Data Entry

Input Conventions 42
 Table Driven Input 45

Parameters 50

Model Definition Options 50
 History Definition Options 50

REZONE Option 51
The input data structure is made up of three logically distinct sections:

1. Parameters describe the problem type and size.
2. Model definition options give a detailed problem description.
3. History definition options define the load history.
Input data is organized in (optional) blocks. Key words identify the data for each optional block. This
form of input enables you to specify only the data for the optional blocks that you need to define your
problem. The various blocks of input are “optional” in the sense that many have built-in default values
which can be used by Marc in the absence of any explicit input from you.
Input Conventions
Marc performs all data conversion internally so that the system does not abort because of data errors
made by you. The program reads all input data options alphanumerically and converts them to integer,
floating point, or keywords, as necessary. Marc issues error messages and displays the illegal option
image if it cannot interpret the option data field according to the specifications given in the manual. When
such errors occur, the program attempts to scan the remainder of the data file and ends the run with an
exit error message at the END OPTION option or at the end of the input file.
Two input format conventions can be used: fixed and free format. You can mix fixed and free format
options within a file, but you can only enter one type of format on a single option.
The syntax rules for fixed fields are as follows:
• You must right-justify integers in their fields. (The right blanks are filled with zeroes).
• Give floating point numbers with or without an exponent. If you give an exponent, it must be
preceded by the character E or D and must be right-justified.
The syntax rules for free fields are as follows:
• Check that each option contains the same number of data items that it would contain under
standard fixed-format control. This syntax rule allows you to mix fixed-field and free-field
options in the data file because the number of options you need to input any data list are the same
in both cases.
• Separate data items on a option with a comma. The comma can be surrounded by any number of
blanks. Within the data item itself, no embedded blanks can appear.
• Give floating point numbers with or without an exponent. If you use an exponent, it must
be preceded by the character E or D and must immediately follow the mantissa (no
embedded blanks).
• Give keywords exactly as they are written in the manual. Embedded blanks do not count as
separators here (for example, BEAM SECT is one word only).
• If a option contains only one free-field data item, follow that item with a comma. For example,
the number “1” must be entered as “1,” if it is the only data item on a option. If the comma is
omitted, the entry is treated as fixed format and may not be properly right-justified.
CHAPTER 3 43
Data Entry
• If the EXTENDED parameter is used, integer data is given using 10 fields as opposed to 5 fields.
This allows very large models to be included in Marc. Additionally, real numbers are entered
using 20 or 30 fields as opposed to 10 or 15. This allows increased accuracy when reading
in data.
• All data can be entered as uppercase or lowercase text.
Input of List of Items

Marc often requests that you enter a list of items in association with certain program functions. As an
example, these items can be a set of elements as in the ISOTROPIC option, or a set of nodes as in the
POINT LOAD option. Fourteen types of items can be requested:
Element numbers Curves ids
Node numbers Surfaces ids
Degree of freedom numbers Body numbers
Integration point numbers Edges pairs
Layer numbers Faces pairs
Increment numbers Oriented curves
Points ids Oriented surfaces
This list can be entered using either the OLD format (compatible with the G, H, and J versions of Marc)
or the NEW format (the K version).
Using the OLD format, you can specify the list of items in three different forms. You can specify:
1. A range of items as:
m n p
which implies items m through n by p. If p is not specified, the program assumes it is 1. Note that
the range can either increase or decrease.
2. A list of items as:
-n a1 a2 a3 ... an
which implies that you should give n items, and they are a ,...a.
3. A set name as:
MYSET
which implies that all items previously specified in the set MYSET are used. Specify the items in
a set using the DEFINE model definition option.
Using the NEW format, you can express the list of items as a combination of one or more sublists.
These sublists can be specified in three different forms. The following operations can be performed
between sublists:
AND
INTERSECT
EXCEPT
When you form a list, subsets are combined in binary operations (from left to right). The following lists
are examples.
1. SUBLIST1 and SUBLIST2
This list implies all items in subsets SUBLIST1 and SUBLIST2. Duplicate items are eliminated
and the resulting list is sorted.
2. SUBLIST1 INTERSECT SUBLIST2
This list implies only those items occurring both in subsets SUBLIST1 and SUBLIST2; the
resulting list is sorted.
3. SUBLIST1 EXCEPT SUBLIST2
This list implies all items in subset SUBLIST1 except those which occur in subset SUBLIST2;
the resulting list is sorted.
4. SUBLIST1 AND SUBLIST2 EXCEPT SUBLIST3 INTERSECT SUBLIST4
This list implies the items in subsets SUBLIST1 and SUBLIST2 minus those items that occur in
subset SUBLIST3. Then, if the remaining items also occur in subset SUBLIST4, they are
included in the list.
Sublists can have several forms. You can specify:
1. A range of items as:
m TO n BY p
or
m THROUGH n BY p
which implies items m through n BY p. If “BY p” is not included, the program assumes “BY
1”. Note that the range can either increase or decrease.
2. A string of items as:
a1 a2 a3 ... an
which implies that n items are to be included. If continuation options are necessary, then either a
C or CONTINUE should be the last item on the option.
3. A set name as:
MYSET
which implies that all items you previously specified to be in the set MYSET are used. You specify
the items in a set using the DEFINE option.
Note: INTERSECT or EXCEPT cannot be used when defining lists of degrees of freedom.
In a list, edges, faces, oriented curves, and oriented surfaces are entered as pairs (i:j) where i is the user
element id and j is the edge id or face id. The edge id/face id for the different element classes is given in
Marc Volume C: Program Input, Chapter 1.
An oriented curve or an oriented surface is used in conjunction with shell elements to indicate if the top
surface or the bottom surface of the shell is to be used.
CHAPTER 3 45
Data Entry
There are two types of edge and face sets; those expressed in Marc convention or the Marc Mentat
convention. The edge/face id in Marc convention is one greater than the Marc Mentat convention.
For example, to specify edge 1 on elements 1 to 20, one would use:
1:1 TO 20:1
Examples of Lists
This section presents some examples of lists and entry formats.
Use the DEFINE model definition option to associate a list of items with a set name with items. Three
sets are defined below: FLOOR, NWALL, and WWALL.
• DEFINE NODE SET FLOOR contains:
1 TO 15(1,2,3,4,5,6,7,8,9,10,11,12,13,14,15)
• DEFINE NODE SET NWALL contains:
5 TO 15 BY 5 AND 20 TO 22(5,10,15,20,21,22)
• DEFINE NODE SET WWALL contains:
11 TO 20(11,12,13,14,15,16,17,18,19,20)
Some possible lists are:
• NWALL AND WWALL, which would contain nodes:
5 10 11 12 13 14 15 16 17 18 19 20 21 22
• NWALL INTERSECT WWALL, which would contain nodes:
15 20
• NWALL AND WWALL EXCEPT FLOOR, which would contain nodes:
16 17 19 20 21 22
Table Driven Input

An alternative approach to defining the data which is based upon defining nonlinear material properties
and variations in the boundary conditions using tables. This input procedure is activated by use of the
TABLE parameter. There are six aspects to this input procedure:
1. The data associated with material properties, boundary conditions, and contact may reference a
function of up to four variables defined in the TABLE model definition option.
2. Boundary conditions and initial conditions are input in a consistent manner and are associated
with a name. They are not applied unless they have been activated by the LOADCASE model and
history definition option.
3. Based upon the use of tables and loadcases, all boundary conditions may be defined in the model
definition section.
4. Mechanical boundary conditions are, in general, entered as total values at the end of the loadcase.
5. Boundary conditions can be applied to either finite element entities (such as nodes, elements,
edges or faces) or they may be applied to geometric entities (such as points, curves, or surfaces).
The boundary conditions are applied to the finite element based upon the current attach
information. This is updated for local adaptive meshing (2-D or 3-D) or global adaptive meshing
in 2-D.
6. Each history definition section provides information on the type of analysis to be performed, the
duration, and the active boundary conditions. The LOADCASE history definition option is used to
perform the latter.
Table Input
Many physical quantities cannot be represented by constants. This includes boundary conditions that are
dependent upon spatial location and/or time, material properties dependent upon temperature or material
damage, contact parameters such as the coefficient of friction and many more. In previous versions, this
data could be input through a variety of options including TEMPERATURE EFFECTS, WORK HARD,
STRAIN RATE, by repetitive input of boundary conditions, or, in the most general case, by
user subroutines.
1 time 28 contact force F 55 normalized arc distance
2 normalized time 56 distance to other contact surface
29 contact body M (near contact only)
3 increment number 30 σn (normal stress) 57 term of series
4 normalized increment 31 voltage 58 hydrostatic stress
5 x coordinate 32 current 59 hydrostatic strain
·
60 B g, p = m g, p ⁄ αm .
6 y coordinate 2
33  -----------------------------------
current radius-
(see THROAT)
 radius of throat
7 z coordinate 34 χ p (pyrolysis damage). ·
61 B g, w = m g, w ⁄ αm .
62 2nd state variable
8 s= 2 2 2 35 φ w (water vapor fraction).
x +y +z
9 θ angle 36 χ c (coking damage). 63 3rd state variable
10 mode number 37 gasket closure distance 64 4th state variable
11 frequency 38 displacement magnitude 65 5th state variable
12 temperature 39 stress rate 66 loadcase number
13 function 40 experimental data 67 degree of cure
14 fourier 41 porosity 68 magnetic field intensity
15 ε p (equivalent plastic strain) 42 void ratio 69 equivalent mechanical strain
16 ε· (equivalent strain rate) 43 · c (equivalent creep strain rate) 70 1st strain invariant
ε
17 B = m · ⁄ αm (normalized 44 minor principal total strain 71 2nd strain invariant
g
mass flow rate).
18 arc length 45 distance from neutral axis 72 3rd strain invariant
(-t/2, +t/2)
19 relative density 46 normalized distance from neutral 73 any strain component
(not available for shells) axis (-1, +1)
CHAPTER 3 47
Data Entry
20 σ (equivalent stress) 47 local x-coordinate of layer point 74 damage

for open or closed section beam
21 magnetic induction 48 local y-coordinate of layer point 75 accumulated crack growth
for open or closed section beam
22 velocity 49 1st isoparametric coordinate 76 relative sliding velocity
(not available in this release)
23 particle diameter 50 2nd isoparametric coordinate -1 to parametric variable 1 to 100
(not available in this release) -100
24 x0 coordinate 51 wavelength (used in spectral
radiation)
25 y0 coordinate 52 creep strain ε e r
26 z0 coordinate 53 pressure or primary quantity in
diffusion
54 equivalent strain rate for
27 s0 = 2 2
x0 + y0 + z0
2
nonNewtonian viscosity
Using the table driven input procedure, virtually all physical data can reference a table/function.
Whenever necessary in the analysis, the table is evaluated based upon the current value of the
independent variables and multiplied with the reference value. A table can have as many as four
independent variables, as long as that independent variable is physically meaningful, For example,
requesting that the Young’s modulus is a function of the equivalent plastic strain is considered a data
error because elasticity does not support this behavior. The list of available independent variable types
is given in the following table.
The function can be input through the TABLE model definition option using either the piecewise linear
mode or the equation mode. In the piecewise linear mode, the result is interpolated between the given
data values. If the independent variable is outside of the range of data entered, the result will be either
extrapolated or the last value will be used.
In the second mode, the function may be entered as a mathematical equation.
A mathematical formula may be either 80 characters or 160 characters long if extended input format is
used. The formula is defined in terms of independent variables v 1 , v 2 , v 3 , and/or v 4 , where the
meaning of those variables is based on the variable type defined in the 3rd data block.
The evaluation is based upon usual mathematical standards moving from left to right with the
conventional rules of the use of parentheses. The following mathematical symbols/operations
are available.
+ addition
- subtraction
* multiplication
/ division
^ exponential
! factor
% mod
Function
F1
v1 x1 x2 x3 x4 v1
Allowing Extrapolation No Extrapolation

Data Entered
x 1, f 1
x 2, f 2
x 3, f 3
x 4, f 4
Given x 0
with extrapolation
f1 – f2
f 0 = f 1 +  ------------------ ⋅ ( x 0 – x 1 )
 x 1 – x 2
without extrapolation
f0 = f1
In addition to v 1 , v 2 , v 3 , and v 4 , the following constants may be used in the equation:
pi π
e exponent
tz offset temperature entered via the PARAMETERS model definition option
q Activation energy entered via MATERIAL DATA model definition option
r Universal gas constant entered via the PARAMETERS model definition option
sb Stefan Boltzman constant entered via the PARAMETERS model definition option
The following mathematical functions may be used in an equation:

cos cosine (x) x in radians
sin sine (x) x in radians
tan tangent (x) x in radians
dcos cosine (x) x in degrees
CHAPTER 3 49
Data Entry
dsin sine (x) x in degrees

dtan tangent (x) x in degrees
acos inverse cosine (x) f in radians
asin inverse sine (x) f in radians
atan inverse tangent (x) f in radians
atan2 inverse tangent (x,y) f in radians
dacos inverse cosine (x) f in degrees
dasin inverse sine (x) f in degrees
datan inverse tangent (x) f in degrees
datan2 inverse tangent (x,y) f in degrees
log log based 10
ln natural log
exp exponent
cosh hyperbolic cosine
sinh hyperbolic sine
tanh hyperbolic tangent
acosh inverse hyperbolic cosine
asinh inverse hyperbolic sine
atanh inverse hyperbolic tangent
sqrt square root
rad convert degrees to radians
deg convert radians to degrees
abs obtain absolute value
int truncates the value to whole
frac take the fractional value
max takes the maximal value
min takes the minimal value
mod return the remainder of x, based on y
mod(x,y) = x - y * int (x/y)
The equation is evaluated based upon the current value of the independent variable. It is your
responsibility to make sure that the equation may be evaluated for the potential values of the independent
values. For example, if the function is 1 ⁄ v 1 , then if v 1 = 0.0 , an error will occur.
Parameters
This group of parameters allocates the necessary working space for the problem and sets up initial
switches to control the flow of the program through the desired analysis, This set of input must be
terminated with an END parameter. The input format for these parameters is described in Marc Volume
C: Program Input.
Model Definition Options

This set of data options enters the initial loading, geometry, and material data of the model and provides
nodal point data, such as boundary conditions. Model definition options are also used to govern the error
control and restart capability. Model definition options can also specify print-out and postprocessing
options. The data you enter on model definition options provides the program with the necessary
information for determining an initial elastic solution (zero increment solution). When boundary
conditions reference time dependent tables, the transient nonlinear behavior can be defined. The transient
period is defined in the history definition section. This group of options must be terminated with the END
OPTION option. The input format for these options is described in Marc Volume C: Program Input.
History Definition Options

This group of options provides the load incrementation and controls the program after the initial elastic
analysis. History definition options also include blocks which allow changes in the initial model
specifications. Each set of load sets must be terminated with a CONTINUE option. This option requests
that the program perform another increment or series of increments if you request the
auto-incrementation features. The input format for these options is described in Marc Volume C:
Program Input.
A typical input file setup for the Marc program is shown below.
• Marc Parameter
Terminated by an END parameter
• Marc Model Definition Options
(Zero Increment)
Terminated by an END OPTION option
• Marc History Definition
Data for the First Increment
Terminated by a CONTINUE option
• (Additional History Definition
Option for the second, third, ..., Increments)
Figure 3-1 is a dimensional representation of the Marc input data file.
CHAPTER 3 51
Data Entry
Proportional
Increment
Load Auto Load
Etc.
Incrementation
Requiring Incrementation
Connectivity
Coordinates
Model Fixed Displacements
Definition Etc.
Linear Analysis
Title
Sizing
Parameter Etc.
Figure 3-1 The Marc Input Data File
REZONE Option
When the REZONE option is inserted into the input file and the manual procedure is used, the program
reads additional data options to control the rezoning steps. These options must immediately follow the
END OPTION option or a CONTINUE history definition option.
You can select as many rezoning steps in one increment as you need. Every rezoning step is defined by
the data, starting with the REZONE option and ending with the CONTINUE option. The END REZONE
option terminates the complete set of rezoning steps that form a complete rezoning increment. Follow
the rezoning input with normal history definition data, or again by rezoning data. The input format for
these options is described in Marc Volume C: Program Input.
Chapter 4 Introduction to Mesh Definition
4 Introduction to Mesh Definition

Direct Input 54
 User Subroutine Input 59

MESH2D 59

Marc Mentat 64
 FXORD Option 65

Incremental Mesh Generators 69

Bandwidth Optimization 70
 Rezoning 71

Substructure 71

BEAM SECT Parameter 74
 Error Analysis 79

Local Adaptivity 79

Automatic Global Remeshing 85
This chapter describes the techniques for mesh definition available internally in Marc. Mesh definition
is the process of converting a physical problem into discrete geometric entities for the purpose of
analysis. Before a body can undergo finite element analysis, it must be modeled into discrete physical
elements. An example of mesh definition is shown in Figure 4-1.
Figure 4-1 Finite Element Mesh of a Brake Rotor (click figure to activate 3D content)
Mesh definition encompasses the placement of geometric coordinates and the grouping of nodes into
elements. For Marc to have a valid mesh definition, the nodes must have geometric coordinates and must
be connected to an element.
First, describe the element by entering the element number, the element type, and the node numbers that
make up the element.
Next, enter the physical coordinates of the nodal points.
Note: You do not need to enter element numbers and node numbers sequentially or consecutively.
Direct Input
You must enter two types of data into Marc for direct mesh definition: connectivity data, which describes
the nodal points for each element, and coordinate data which gives the spatial coordinates of each nodal
point. This section describes how to enter this data.
Element Connectivity Data

You can enter connectivity data from either the input option file (FORTRAN unit 5) or from an auxiliary
file. Several blocks of connectivity can be input. For example, the program can read one block from tape
and subsequently read a block from the input option file. Each block must begin with the word
CONNECTIVITY. In the case of duplicate specification, Marc always uses the data that was input last for
a particular element.
Enter the nodal points of two-dimensional elements in a counterclockwise order. Figure 4-2 illustrates
correct and incorrect numbering of element connectivity data.
CHAPTER 4 55
Introduction to Mesh Definition
4 3 2 3
Y,R
1 2 1 4
X,Z
Correct Numbering Incorrect Numbering

Figure 4-2 Correct/Incorrect Numbering of Two-Dimensional Element Connectivity of 4-Node Elements
When there are eight nodal points on a two-dimensional element, number the corner nodes 1 through 4 in
counterclockwise order. The midside nodes 5 through 8 are subsequently numbered in counterclockwise
order. Figure 4-3 illustrates the correct numbering of element connectivity of 8-node elements.
Y,R 4 7 3
X,Z 8 6
1 5 2
Figure 4-3 Numbering of Two-Dimensional Element Connectivity for 8-Node Quadrilateral Elements
Lower-order triangular elements are numbered using the counterclockwise rule.

X,R 3
X,Z
1 2
Figure 4-4 Numbering of 3-Node Triangular Element
Note that quadrilateral elements can be collapsed into triangular elements by repeating the last node.
The higher order triangular elements have six nodes, the corner nodes are numbered first in a
counterclockwise direction. The midside nodes 4 through 6 are subsequently numbered as shown
in Figure 4-5.
3
6 5
1 4 2
Figure 4-5 Numbering of 6-Node Triangular Element
Number three-dimensional elements in the same order as two-dimensional elements for each plane. Enter
nodes for an 8-node brick in counterclockwise order as viewed from inside the element. First, enter nodes
comprising the base; then enter ceiling nodes as shown in Figure 4-6.
8 7
5 6
Z
4 3
X 1 2
Figure 4-6 Numbering of Element Connectivity for 8-Node Brick
A 20-node brick contains two 8-node planes and four nodes at the midpoints between the two planes.
Nodes 1 through 4 are the corner nodes of one face, given in counterclockwise order as viewed from
within the element. Nodes 5 through 8 are on the opposing face; nodes 9 through 12 are midside nodes
on the first face, while nodes 13 through 16 are their opposing midside nodes. Finally, nodes 17 through
20 lie between the faces with node 17 between 1 and 5. Figure 4-7 illustrates the numbering of element
connectivity for a 20-node brick.
CHAPTER 4 57
8 15 7
16 14
13 20 19
5 6
Z
4 11 3
17 18
12
Y 10
1 9 2
X
Figure 4-7 Numbering of Three-Dimensional Element Connectivity for 20-Node Brick
The four node tetrahedral is shown in Figure 4-8.

4
1 2
Figure 4-8 Numbering of Four-Node Tetrahedral
The ten-node tetrahedral is shown in Figure 4-9. The corner nodes 1-4 are numbered first. The first three
midside nodes occur on the first face. Nodes 8, 9, and 10 are between nodes 1 and 4, 2 and 4, and 3 and
4, respectively.
4
10
8 9 3
7 6
1 5 2
Figure 4-9 Numbering of 10-Node Tetrahedral
The 6-node pentahedral element is shown in Figure 4-10.

5
3
2
Figure 4-10 6-Node Pentahedral
The 15-node pentahedral element is shown in Figure 4-11.

3
15
6
9 8
12 11
1 7 2
13 14
4 10 5
Figure 4-11 15-Node Pentahedral
Nodal Coordinate Data

You can enter nodal coordinates directly from the input option file (FORTRAN unit 5) or from an
auxiliary file. You can enter several blocks of nodal coordinate data in a file. In the case of duplicate
specifications, the program uses data entered last for a particular nodal point in the mesh definition.
Direct nodal input can be used to input local corrections to a previously generated set of coordinates.
These options give the modified nodal coordinates.
The CYLINDRICAL option can be used to transform coordinates given in a cylindrical system to a
Cartesian system.
CHAPTER 4 59
Note: Requires the final coordinate data in terms of a single Cartesian system. Refer to Marc
Volume B: Element Library to determine the required coordinate data for a particular
element type.
Activate/Deactivate
You have the ability to turn on and off elements using this option, which is useful when modeling
ablation or excavation. When you enter the mesh connectivity, the program assumes that all elements are
to be included in the analysis unless they are deactivated. This effectively removes this material from the
model. These elements can be reinstated later by using the ACTIVATE option. If the element is activated,
one can select if the level of stress is to be reinstated or set to zero. The use of these options results in
nonlinear behavior and have an effect upon convergence.
User Subroutine Input

User subroutines can be used to generate or modify the data for mesh definition. User subroutine
UFCONN generates or modifies element connectivity data. The UFCONN model definition option
activates this subroutine. The user subroutine is called once for each element requested. Refer to Volume
D: User Subroutines and Special Routines for a description of the UFCONN user subroutine and
instructions for its use.
The UFXORD user subroutine generates or modifies the nodal coordinates. The UFXORD model
definition option activates this subroutine. The user subroutine is called once for each node requested.
Refer to Volume D: User Subroutines and Special Routines for a description of the UFXORD user
subroutine and instructions for its use.
MESH2D
MESH2D generates a mesh of quadrilateral or triangular elements for a two-dimensional body of any
shape. The generated mesh is written to a separate file and must be read with the CONNECTIVITY,
COORDINATES, and FIXED DISP, etc., options.
Block Definition
In MESH2D, a physical object or domain is divided into quadrilateral and/or triangular parts, called
“blocks”. Quadrilateral blocks are created by Marc by mapping with polynomials of the third order from
a unit square. These blocks can, therefore, be used to approximate curved boundaries. The geometry of
a quadrilateral block is defined by the coordinates on 12 nodes shown in Figure 4-12. If the interior nodes
on an edge of the block are equal to zero or are not specified, the edge of the block is straight.
Triangular blocks have straight edges. The geometry of a triangular block is defined by the coordinates
of the three vertices.
η P3
P9
y P4 P10
4 10 9 3 P8
P11 P7
11 8 P12 P2
2 ξ P6
P1
P5
12 7 x
1 5 6 2 12
x ( ξ ,η ) =  x ( P i )Φ i ( ξ ,η )
i = 1
2
12
y ( ξ ,η ) =  y ( P i )Φ i ( ξ ,η )
i = 1
Figure 4-12 Typical Quadrilateral Block Mapping
Merging of Nodes
Marc creates each block with a unique numbering scheme. The MERGE option fuses all nodes that lie
within a small circle, renumbers the nodes in sequence, and then removes all gaps in the numbering
system. You can select which blocks are to be merged together, or you can request that all blocks be
merged. You must give the closeness distance for which nodes will be merged.
Block Types
MESH2D generates two types of quadrilateral blocks. Block Type 1 is a quadrilateral block that is
covered by a regular grid. The program obtains this grid by dividing the block edge into M by N intervals.
Figure 4-13 illustrates the division of block edges into intervals with M = 4, N = 3.
The P1 P4 face of the block is the 1-4 face of triangular elements and the P1 P2 face of the block becomes
the 1-2 face of quadrilateral elements.
CHAPTER 4 61
η
η
P4 P3
16 17 18 19 20
24 16 17 18 19 20
2/3
17 2/3 9 10 11 12
12 13 14
11 15
N = 3 2/3 16 11 12 13 14 15
ξ
9 N = 3 2/3 5 6 7 8 ξ
7 8 9
6 10
2 4 5 8 6 7 8 9 10
2/3 P2
1 3 6 7 2/3 1 2 3 4
1 2 3 4 5 1 2 3 4 5
P1
1/2 1/2 1/2 1/2 1/2 1/2 1/2 1/2
M=4 M=4
Triangular Quadrilateral
Figure 4-13 Block Type 1
Block Type 2 is a quadrilateral block that allows the transition of a coarse mesh to a finer one. In one
direction, the block is divided into M, 2M, 4M ...; while in the other direction, the block is divided into
N intervals. Figure 4-14 illustrates the division of Block Type 2 edges into intervals.
The P1 P2 face of the block becomes the 1-2 face of quadrilateral elements, and the P2 P3 face of the
block is the 2-3 face of triangular elements.
Block Type 3 is a triangular block. The program obtains the mesh for this block by dividing each side
into N equal intervals. Figure 4-15 illustrates Block Type 3 for triangular and quadrilateral elements.
Block Type 4 is a refine operation about a single node of a block (Figure 4-16). The values of N and M
are not used.
If quadrilateral elements are used in a triangular block, the element near the P2 P3 face of the block is
collapsed by MESH2D in every row. The P1 P2 face of the block is the 1-2 face of the generated elements.
P4 P3
19
18 34 η
2/7
9 17 1/2 1/2 1/2 1/2
7 10 11 12 13 14 15 16
4/7
8 7 8 9 10 11
M=2 5 6 7
4 8
N=3 x 1 3 4 6
1 3 4 6
4 5 6 x
3/7 2
2 5 2 5
P1 1 2 3 P2 1 2 3
1 1
1 1
P3 P3
15 10
13
16
10 6
14 8 9
P1 6 11
8
1 2 7 12 4 5
1 6 7
3 4 8 5
2
9 P1 1 2 3
3 P2
7
4 P2
1 2 3 4
N=4 5
5 6 7
4 3
4
3
1 2
1 2

CHAPTER 4 63
Symmetry, Weighting, and Constraints

MESH2D contains several features that facilitate the generation of a mesh: use of symmetries, generation
of weighted meshes, and constraints. These features are discussed below.
MESH2D can use symmetries in physical bodies during block generation. An axis of symmetry is
defined by the coordinates of one nodal point, and the component of a vector on the axis. One block
can be reflected across many axes to form the domain. Figure 4-17 illustrates the symmetry features
of MESH2D.
2
1 1
Original Block One Symmetry Axis
3 2 3 2
4 1 4 1
5 8
6 7
Two Symmetry Axes Three Symmetry Axes

Figure 4-17 Symmetry Option Example
A weighted mesh is generated by the program by spacing the two intermediate points along the length
of a boundary. This technique biases the mesh in a way that is similar to the weighting of the boundary
points. This is performed according to the third order isoparametric mapping function.
Note: If a weighted mesh is to be generated, be cautious not to move the interior boundary points
excessively. If the points are moved more than 1/6 of the block length from the 1/3
positions, the generated elements can turn inside-out.
The CONSTRAINT option generates boundary condition restraints for a particular degree of freedom for
all nodes on one side of a block. The option then writes the constraints into the file after it writes the
coordinate data. The FIXED DISP, etc., option must be used to read the boundary conditions generated
from the file.
Additional Options
Occasionally, you might want to position nodes at specific locations. The coordinates of these nodes are
entered explicitly and substituted for the coordinates calculated by the program. This is performed using
the SPECIFIED NODES option. Some additional options in MESH2D are:
• MESH2D can be used several times within one input file.
• The START NUMBER option gives starting node and element numbers.
• The CONNECT option allows forced connections and/or disconnections with other blocks. This
option is useful when the final mesh has cracks, tying, or gaps between two parts.
• The MANY TYPES option specifies different element types.
Marc Mentat
Marc Mentat is an interactive program which facilitates mesh definition by generating element
connectivity and nodal coordinates. Some of the Marc Mentat capabilities relevant to mesh generation
are listed below.
• Prompts you for connectivity information and nodal coordinates. Accepts input from a keyboard
or mouse.
• Accepts coordinates in several coordinate systems (Cartesian, cylindrical, or spherical).
• Translates and rotates (partial) meshes.
• Combines several pre-formulated meshes.
• Duplicates a mesh to a different physical location.
• Generates a mirror image of a mesh.
• Subdivides a mesh into a finer mesh.
• Automatic mesh generation in two- and three-dimensions.
• Imports geometric and finite element data from CAD systems.
• Smooths nodal point coordinates to form a regular mesh
• Converts geometric surfaces to meshes.
• Refines a mesh about a point or line.
• Expands line mesh into a surface mesh, or a surface mesh into a solid mesh.
• Calculates the intersection of meshes.
• Maps nodal point coordinates onto prescribed surfaces.
• Writes input data file for connectivity and coordinates in Marc format for use in future analyses.
• Apply boundary conditions to nodes and elements.
• Define material properties.
• Submit Marc jobs.
CHAPTER 4 65
FXORD Option
The FXORD model definition option (Volume C: Program Input) generates doubly curved shell elements
of element type 4, 8, or 24 for the geometries most frequently found in shell analysis. Since the
mathematical form of the surface is well-defined, the program can generate the 11 or 14 nodal
coordinates needed by element type 8, 24, or 4 to fit a doubly curved surface from a reduced set of
coordinates. For example, you can generate an axisymmetric shell by entering only four coordinates per
node. The FXORD option automatically generates the complete set of coordinates required by the
elements in the program from the mathematical form of the surface.
A rotation and translation option is available for all components of the surface to give complete
generality to the surface generation. The input to FXORD consists of the reduced set of coordinates given
in a local coordinate system and a set of coordinates which orient the local system with respect to the
global system used in the analysis. The program uses these two sets of coordinates to generate a structure
made up of several shell components for analysis.
The FXORD option allows for the generation of several types of geometries. Because you may need to
analyze shells with well-defined surfaces not available in this option, you can use the UFXORD user
subroutine to perform your own coordinate generation (Volume D: User Subroutines and Special
Routines). The FXORD option can also be used to convert cylindrical coordinates or spherical
coordinates to Cartesian coordinates for continuum elements.
Major Classes of the FXORD Option

The following cases are considered:
• Shallow Shell (Type I)
• Axisymmetric Shell (Type 2)
• Cylindrical Shell Panel (Type 3)
• Circular Cylinder (Type 4)
• Plate (Type 5)
• Curved Circular Cylinder (Type 6)
• Convert Cylindrical to Cartesian (Type 7)
• Convert Spherical to Cartesian (Type 8)
Shallow Shell (Type I)
Type 1 is a shallow shell with
θ 1 = x 1, θ2 = x2 (4-1)
The middle surface of Figure 4-18 (Type 1) is defined by an equation of the form
x 3 = x 3 ( x 1, x 2 ) (4-2)
and the surface is determined when the following information is given at each node.
∂x 3 ∂x 3 ∂ 2 x 3
x 1, x 2, x 3, ---------, --------- , ------------------ (4-3)
∂x 1 ∂x 2 ∂x 1 ∂x 2
The last coordinate is only necessary for Element Type 4.

X3
X3
R
f
X2
X2
q
X1
Type 1 Type 2
X3
X2
R
X2 φ
3
X1
X3
X1
Type 4
X3
Type 3
Figure 4-18 Classification of Shells
Axisymmetric Shell (Type 2)

The middle surface symmetric to the x 3 axis (Figure 4-18, Type 2) is defined as:
x 1 = R ( φ ) cos φ cos θ
x 2 = R ( φ ) cos φ sin θ
(4-4)
x 3 = R ( φ ) sin φ
where φ and θ are the angles shown in Figure 4-18. In this case, the surface is defined by
dR
θ, φ, R, ------- (4-5)
dφ
The angles θ and φ are given in degrees.
CHAPTER 4 67
Cylindrical Shell Panel (Type 3)

The middle surface is the cylinder defined by Figure 4-18.
x1 = x1 ( s )
x2 = x2 ( s ) (4-6)
x3 = x3
The nodal geometric data required is
dx 1 dx 2
s, x 3, x 1, x 2, ---------, --------- (4-7)
ds ds
Circular Cylinder (Type 4)

This is the particular case of Type 3 where the curve
x 1 ( s ), x 2 ( s ) (4-8)
is the circle given by Figure 4-18 (Type 4).
x 1 = R cos θ
(4-9)
x 2 = R sin θ
The only nodal information is now
θ, x 3, R (4-10)
Note that θ is given in degrees and, because R is constant, it needs to be given for the first nodal
point only.
Plate (Type 5)
The shell is degenerated into the plate
x3 = 0 (4-11)
The data is reduced to
x 1, x 2 (4-12)
Curved Circular Cylinder (Type 6)

Figure 4-19 illustrates this type of geometry.
X3
Shell Middle Surface
q1
q2 R
f
X2
q
X1
Type 6
Figure 4-19 Curved Circular Cylinder
The middle surface of the shell is defined by the equations
x 1 = r cos θ
x 2 = r sin θ cos φ + R ( 1 – cos φ ) (4-13)
x 3 = ( R + – r sin θ r sin φ ) sin φ
The Gaussian coordinates on the surface are
θ 1 = rθ
(4-14)
θ 2 = Rφ
and form an orthonormal coordinate system. The nodal point information is
θ, φ, r, R (4-15)
θ and φ in degrees. You need to specify the radii r and R only for the first nodal point.
CHAPTER 4 69
Convert Cylindrical to Cartesian (Type 7)

Type 7 allows you to enter the coordinates for continuum elements in cylindrical coordinates, which are
converted by Marc to Cartesian coordinates. In this way, you can enter R , θ , Z and obtain x, y, z where
θ is given in degrees and
x = R cos θ
y = R sin θ (4-16)
z = Z
Convert Spherical to Cartesian (Type 8)

Type 8 allows you to enter the coordinates for continuum elements in spherical coordinates, which are
converted by Marc to Cartesian coordinates. In this way, you can enter R , θ , φ and obtain x, y, z where
θ and φ are given in degrees and
x = R cos θ sin φ
y = R sin θ cos φ (4-17)
z = R cos φ
Recommendations on Use of the FXORD Option

When a continuous surface has a line of discontinuity, for example, a complete cylinder at
θ = 0° = 360° , you must place two nodes at each nodal location on the line to allow the distinct
coordinate to be input. You must use tying element type 100 to join the degrees of freedom. Generally,
when different surfaces come together, you must use the intersecting shell tyings.
The FXORD option cannot precede the COORDINATES option, because it uses input from that option.
Incremental Mesh Generators

Incremental mesh generators are a collection of options available in Marc to assist you in generating the
mesh. Incremental mesh generators generate connectivity lists by repeating patterns and generate nodal
coordinates by interpolation. Use these options directly during the model definition phase of the input.
During the model definition phase, you can often divide the structure into regions, or blocks, for which
a particular mesh pattern can be easily generated. This mesh pattern is established for each region and is
associated with a single element connectivity list. Use the CONNECTIVITY option to input this element
connectivity list. The incremental mesh generators then generate the remainder of the connectivity lists.
Critical nodes define the outline of the regions to be analyzed. Use the COORDINATES option to
enter the critical nodes. The incremental mesh generators complete the rest and join the regions by
merging nodes.
A special connectivity interpolator option generates midside nodes for elements where these nodes have
not been specified in the original connectivity. A separate mesh generation run is sometimes required to
determine the position of these nodes. This run can be followed by mesh display plotting.
The incremental mesh generators are listed below:

Element Connectivity Generator – The CONN GENER option repeats the pattern of the
connectivity data for previously defined master elements. One element can be removed for each
series of elements, allowing the program to generate a tapered mesh. Two elements can be
removed for each series with triangular elements.
Element Connectivity Interpolator – The CONN FILL option completes the connectivity list by
generating midside nodes. You first generate the simpler quadrilateral or brick elements without
the midside nodes. You can then fill in the midside nodes with this option.
Coordinate Generator – The NODE GENER option creates a new set of nodes by copying the
spacing of another specified set of nodes.
Coordinate Interpolator – The NODE FILL generates intermediate nodes on a line defined by two
end nodes. The spaces between the nodes can be varied according to a geometric progression.
Coordinate Generation for Circular Arcs – The NODE CIRCLE option generates the coordinates
for a series of nodes which lie on a circular arc.
Bandwidth Optimization
Marc can minimize the nodal bandwidth of a structure in several ways. The amount of storage is directly
related to the size of the bandwidth, and the computation time increases in proportion to the square of the
average bandwidth.
The OPTIMIZE option allows you to choose from several bandwidth optimization algorithms. The
minimum degree algorithm should only be used if the direct sparse solver is used.
The four available OPTIMIZE options are listed in Table 4-1.
Note: This option creates an internal node numbering that is different from your node numbering.
Use your node numbering for all inputs. All output appears with your node numbering. The
occurrence of gap or Herrmann elements can change the internal node numbers. On
occasion, this change can result in a non optimal node numbering system, but this system
is necessary for successful solutions.
Table 4-1 Bandwidth Optimization Options
Option
Number Remarks Solver
2 Cuthill-McKee algorithm Profile (0)
9 Sloan (Recommended) Profile (0)
10 Minimum Degree Algorithm Sparse Direct (4) or Multifrontal (8)
11 METIS (Recommended)* Multifrontal (8)
*Reference “A Fast and Highly Quality Multilevel Scheme for Partitioning Irregular
Graphs”, George Karypis and Vipin Kumar, SIAM Journal on Scientific Computing, Vol.
20, No. 1, pp. 359-392, 1999. METIS documentation may be downloaded from
http://glaros.dtc.umn.edu/gkhome/views/metis
CHAPTER 4 71
The nodal correspondence obtained through this process can be saved and then used in subsequent
analyses. This eliminates the need to go through the optimization step in later analyses. The
correspondence is used to relate the user-defined node (external) numbers to the program-optimized
(internal) node numbers and vice versa.
Rezoning
The REZONING parameter defines a new mesh and transfers the state of the old mesh to the new mesh.
Elements or nodes can be either added to or subtracted from the new mesh. This procedure requires a sub
increment to perform the definition of the new mesh. The rezoning capability can be used for two- and
three-dimensional continuum elements and for shell elements 22, 75, 138, 139, and 140. See Figure 4-20
for an example of rezoning.
Rezoning or remeshing can be carried automatically using the mesh generators. This feature is described
in detail in the Automatic Global Remeshing section.
Before Rezoning After Rezoning

Figure 4-20 Mesh Rezoning
Substructure
Marc is capable of multilevel substructuring that includes:
Generation of superelements
Use of superelements in subsequent Marc analyses
Recovery of solutions (displacements, stresses, and strains) in the individual substructures
The Marc multilevel procedure allows superelements to be used. One self-descriptive database stores all
data needed during the complete analysis. You only have to ensure this database is saved after every step
of the analysis.
The advantages of substructuring are the following:
• Separates linear and nonlinear parts of the model
• Allows repetition of symmetrical or identical parts of the model for linear elastic analysis
• Separates large models into multiple, moderate-size models
• Separates fixed model parts from parts of the model that may undergo design changes
A disadvantage of substructuring is the large amount of data that must be stored on the database.
Three steps are involved in a substructuring run.
• The superelement generation step is done for every superelement at a certain level.
• The use of superelements in subsequent Marc runs is done at the highest level, or is incorporated
into Step 1 for the intermediate levels.
• Recovery of solutions within a certain superelement can or cannot be done for
every superelement.
Substructuring in Marc is only possible for static analysis. Nonlinearities are not allowed with a
superelement. You, as a user, must ensure that nonlinearities are not present.
The maximum number of levels in a complete analysis is 26. The maximum number of substructures in
the complete analysis is 676.
Each step can be done in an individual run, or an unlimited number of steps can be combined into a
single run.
During superelement generation in Marc, you can generate a complete new superelement or you can copy
a previously defined superelement with identical or newly defined external load conditions. Any number
of superelements can be formed in a generation run.
Marc offers flexibility in the use of superelements by allowing rotation or mirroring of a superelement.
If a run is nonlinear, superelements are treated as linear elastic parts. At every increment, you can perform
a detailed analysis of certain substructures by descending down to the desired superelements. Use the
normal Marc control algorithm (AUTO INCREMENT, AUTO LOAD, PROPORTIONAL INCREMENT) to
control the load on the superelements.
Technical Background
The system of equations for a linear static structure is
Ku = P (4-18)
When local degrees of freedom (subscripted l ) and external degrees of freedom (subscripted e ) are
considered, this can be rewritten as
K l l K e l  u l   P 
 l 
= (4-19)
K l e K e e  u e   P 
 e 
To obtain both the stiffness matrix and the load vector of the substructure, it is necessary to eliminate u 1
and rewrite the above system with u e as the only unknown splitting the above equation.
CHAPTER 4 73
Kl l ul + Ke l ue = Pl
and (4-20)
Kl e ul + Ke e ue = Pe
The first equation can be written as
u l = – K l–l1 • K e l u e + K l–l1 P l (4-21)
Substituting this equation into the second
– K l e K l–l1 K e l u e + K l e K l–l1 P l + K e e u e = P e (4-22)
This can be rewritten as
K e*e u e = P e* (4-23)
where
–1
K e*e = K e e – K l e K l l K e l (4-24)
and
–1
P e* = P e – K l e K l l P l (4-25)
K e*e and P e* are solved by the triangularization of K l l , the forward and backward substitution of K e l
and P l , respectively, and premultiplication with K l e , K e*e , and P e* are used in the next part of the
analysis with other substructures or with another element mesh. That analysis results in the calculation
of ue.
You can now calculate the local degrees of freedom and/or the stresses using the following procedure:
Kl l • ul = Pl – Ke l ue (4-26)
which can be written as

–1
u l = K l l • P l* (4-27)
where
P l* = P e – K e l u e (4-28)
The displacement of the substructure is, therefore, known, and stresses and strains can be calculated in
the normal way.
4 Scaling Element Stiffness

Introductio Occasionally, it is desirable to perform a scalar multiplication of the stiffness, mass, and load matrix to
n to Mesh represent a selective duplication of the finite element mesh.
Definition
The STIFSCALE option can be used to enter the scaling factor for each element. In this case, the global
stiffness, mass, and load matrices are formed as follows:
K g = Σs i K ie l, M g = Σs i M ie l, and F g = Σs i f ie l (4-29)
Note that no transformation of the stiffness matrix occurs and that point loads are not scaled.
BEAM SECT Parameter

The BEAM SECT parameter inputs data to define the sectional properties for three-dimensional beam
elements. Include this option if you are using:
element types 13, 77, or 79
element types 14, 25, 76, or 78 with a noncircular section
element types 52 or 98 and torsional and shear stiffness must be defined independently
element types 52 or 98 with numerical integrated solid cross-section.
The convention adopted for the local (beam) coordinate system is: the first and second director (local X
and Y) at a point are normal to the beam axis; the third director (local Z) is tangent to the beam axis and
is in the direction of increasing distances along the beam. The director set forms a right-handed system.
Orientation of the Section in Space

The beam axis in an element is interpolated from the two nodes of the element.
dx dy dz
x, y, z, ------, ------, ------ (4-30)
ds ds ds
where the last three coordinates are only used for element 13.
The beam section orientation in an element is defined by the direction of the first director (local X) at a
point, and this direction is specified via the coordinates of an additional node or through the GEOMETRY
option (see Marc Volume C: Program Input).
Definition of the Section

You can include any number of different beam sections in any problem. Data options following the BEAM
SECT parameter of the Marc input (see Marc Volume C: Program Input) define each section. The
program numbers the sections in the order they are entered. To use a particular section for a beam
element, set EGEOM2 (GEOMETRY option, Option 2, Columns 11-20) to the floating-point value of the
section number, for example, 1, 2, or 3. The program uses the default circular section for the closed
section beam elements (14, 25, 76, 78) if EGEOM1 is nonzero. The program uses the default solid
rectangular cross section for elements 52 or 98 if EGEOM1 is nonzero. Figure 4-21 shows how the
thin-walled section is defined using the input data.
CHAPTER 4 75
Section 1 Section 2
10 xl 8
xl
5 4 6 3 4
yl
16
yl
20
3 1 2
2 1
8
Branch Divisions Thickness Branch Divisions Thickness

1-2 8 1.0 1-2 10 1.0
2-3 4 0.0 2-3 10 1.0
3-4 8 0.3 3-4 10 1.0
4-5 4 0.0
5-6 8 1.0
1 6 Section 3
Branch Divisions Thickness
1-2 8 0.5
10
2-3 4 0.5
3-4 8 0.5
4-5 4 0.5
5-6 8 0.5
2 60° 5
R=5
17.85
3 4
xl
yl
Figure 4-21 Beam Section Definition Examples for Thin-walled Sections
The rules and conventions for defining a thin-walled section are listed below:
1 1 1
1. An x – y coordinate system defines the section, with x the first director at a point of the
1 1
beam. The origin of the x – y system represents the location of the node with respect to
the section.
2. Enter the section as a series of branches. Branches can have different geometries, but they must
form a complete traverse of the section in the input sequence so that the endpoint of one branch
is the start of the next branch. It is often necessary for the traverse of the section to double back on
itself. To cause the traverse to do this, specify a branch with zero thickness.
3. You must divide each branch into segments. The stress points of the section are the branch
division points. The stress points are the points used for numerical integration of a section’s
stiffness and for output for stress results. Branch endpoints are always stress points. There must
always be an even number of divisions (nonzero) in any branch. Not counting branches of zero
thickness, you can use a maximum of 31 stress points (30 divisions) in a complete section.
4. Branch thickness varies linearly between the values given for branch endpoint thickness. The
thickness can be discontinuous between branches. A branch is assumed to be of constant thickness
equal to the thickness given at the beginning of the branch if the thickness at the end of the branch
is given as an exact zero.
1 1
5. The shape of a branch is interpolated as a cubic based on the values of x and y and their
directions, in relation to distance along the branch. The data is input at the two ends of the branch.
1 1
If both dx ⁄ ds and dy ⁄ ds are given as exact zeros at both ends of the branch, the branch is
assumed to be straight. The section can have a discontinuous slope at the branch ends. The
beginning point of one branch must coincide with the endpoint of the previous branch. As a result,
1 1
x and y for the beginning of a branch need to be given only for the first branch of a section.
6. Stress points are merged into one point if they are separated by a distance less than t ⁄ 10 , where
t is a thickness at one of these points.
Figure 4-21 shows three sections of a beam. Notice the use of zero thickness branches in the traverse of
the I section. The program provides the following data: the location of each stress point in the section,
the thickness at that point, the weight associated with each point (for numerical integration of the section
stiffness), and the warping function at each section.
Solid sections can be defined by using one of the standard sections and specifying its typical dimensions
or by using quadrilateral segments and specifying the coordinates of the four corner points of each
quadrilateral segment in the section. The local axes of a standard section are always the symmetry axes.
For the standard sections, the symmetry axes are also the principal axes. The principal axes for each
section are shown in Figure 4-22. For the more general sections, it is not required to enter the coordinates
of the corner points of the quadrilateral segments with respect to the principal axes. The section is
automatically realigned and you can make specifications about how this realignment should be carried
out. The details about these orientation methods are described in the Cross-section Orientation and
Location of the Local-section Axis sections in Marc Volume B: Element Library.
Figure 4-22 shows the standard solid sections, the dimensions needed to specify them, and the orientation
of their local axes. If the dimension b is omitted for an elliptical section, the section will be circular. If
the dimension b is omitted for a rectangular section, the section will be square. If the dimension c is
omitted for a trapezoidal or a hexagonal section, the section will degenerate to a triangle or a diamond.
For the elliptical section, the center point is the first integration point. The second is located on the
negative y-axis and the points are numbered radially outward and then counterclockwise. For the
rectangular, trapezoidal and hexagonal sections the first integration point is nearest to the lower left
corner and the last integration point is nearest to the upper right corner. They are numbered from left to
CHAPTER 4 77
right and then from bottom to top. The default integration scheme for all standard solid sections uses 25
integration points.
y y
x x
b b
a a
Elliptical Section Rectangular Section
Circular (set b = 0) Square (set b = 0)
y y
c c
[2(a+c)b]/[3(a+c)]
x
b b
x
a a
Trapezoidal Section Hexagonal Section
Figure 4-22 Standard Solid Section Types with Typical Dimensions
The rules and conventions for defining a general section solid are listed below:
1. Each quadrilateral segment is defined by entering the coordinates of the four corner points in the
local xy-plane. The corner points must be entered in a counterclockwise sense.
2. The quadrilateral segments must describe a simply connected region; i.e., there must be no holes
in the section, and the section cannot be composed of unconnected “islands” and should not be
connected through a single point.
3. The quadrilateral segments do not have to match at the corners. Internally, the edges must, at
least, partly match.
4. The segments must not overlap each other (i.e., the sum of the areas of each segment must equal
the total area of the segment).
5. Each section can define its own integration scheme and this scheme applies to all segments in
the section.
6. A section cannot have more than 100 integration points in total. This limits each section to, at
most, 100 segments that, by necessity, will use single point integration. If less than 100 segments
are present in a section, they may use higher order integration schemes, as long as the total
number of points in the section does not exceed 100.
7. Integration points on matching edges are not merged, should they coincide.
This method allows you to define an arbitrary solid section in a versatile way, but you must make sure
that the conditions of simply connectedness and nonoverlap are met. The assumptions underlying the
solid sections are not accurate enough to model sections that are thin-walled in nature. It is not
recommended to use them as an alternative to thin-walled sections. Solid sections do not account for any
warping of the section and, therefore, overestimate the torsion stiffness as it is predicted by the Saint-
Venant theory of torsion.
The first integration point in each quadrilateral segment is nearest to its first corner point and the last
integration point is nearest to its third corner point. They are numbered going from first to second corner
point and then from second to third corner point. The segments are numbered consecutively in the order
in which they have been entered and the first integration point number in a new segment simply continues
from the highest number in the previous segment. There is no special order requirement for the
quadrilateral segments. They must only meet the previously outlined geometric requirements. For each
section, the program provides the location of the integration points with respect to the principal axes and
the integration weight factors of each point. Furthermore, it computes the coordinates of the center of
gravity and the principal directions with respect to the input coordinate system. It also computes the area
and the principal second moments of area of the cross section. For each solid section, all of this
information is written to the output file.
All solid sections can be used in a pre-integrated fashion. In that case, the area A and the principal
moments of area Ixx and Iyy are calculated by numerical integration. The torsional stiffness is always
computed from the polar moment of inertia J=Ixx+Iyy. The shear areas are set equal to the section area A.
The stiffness behavior may be altered by employing any of the stiffness factors. In a pre-integrated
section, no layer information is computed and is, therefore, not available for output. Pre-integrated
sections can only use elastic material behavior and cannot account for any inelasticity (e.g. plasticity).
For pre-integrated sections there is no limit on the number of segments to define the section.
Figure 4-23 shows a general solid section built up from three quadrilateral segments using a 2 x 2 Gauss
integration scheme. It also shows the numbering conventions adopted in a quadrilateral segment and its
mapping onto the parametric space. For each segment in the example, the first corner point is the
lower-left corner and the corner points have been entered in a counterclockwise sense. The segment
numbers and the resulting numbering of the integration points are shown in the figure. Gauss integration
schemes do not provide integration points in the corner points of the section. If this is desired, you
can choose other integration schemes like Simpson or Newton-Cotes schemes. In general, a 2 x 2 Gauss
or a 3 x 3 Simpson scheme in each segment suffices to guarantee exact section integration of linear
elastic behavior.
(x3,y3)
y η
11 12 (x4,y4) (-1,1) (1,1)
3
9 10
7 8 x 3 4
2 ξ
5 6
1 2
(x2,y2)
3 4
1
1 2 (-1,-1) (1,-1)
(x1,y1)
Figure 4-23 Beam Section Definition for Solid Section plus Numbering Conventions Adopted
CHAPTER 4 79
Error Analysis
You can determine the quality of the analysis by using the ERROR ESTIMATE option.
The ERROR ESTIMATE option can be used to determine the mesh quality (aspect ratio, and skewness),
and how they change with deformation. While all of the Marc elements satisfy the patch test, the
accuracy of the solution often depends on having regular elements. In analyses where the updated
Lagrangian method is used, the mesh often becomes highly distorted during the deformation process.
This option tells you when it would be beneficial to perform a rezoning step.
This option can also be used to examine the stress discontinuity in the analysis. This is a measure of the
meshes ability to represent the stress gradients in the problem. Large stress discontinuities are an
indication that the mesh is not of sufficient quality. This can be resolved by increasing the number of
elements or choosing a higher order element or using local adaptive meshing.
Local Adaptivity
The adaptive mesh generation capability increases the number of elements and nodes to improve the
accuracy of the solution. The capability is applicable for both linear elastic analysis and for nonlinear
analysis. The capability can be used for lower-order elements, 3-node triangular solids and shells, 4-node
quadrilateral solids and shells, 4-node tetrahedrals, 8-node hexahedral elements, and the 8-node
solid-shell elements. Local adaptive meshing cannot be used with the triangular and tetrahedral
Herrmann elements, type 155 - 157.
When used in conjunction with the ELASTIC parameter for linear analysis, a steady-state heat transfer,
electrostatic, or magnetostatic, the mesh is adapted and the analysis repeated until the adaptive criteria
is satisfied. When used in a nonlinear analysis, an increment is performed. If necessary, this increment
is followed by a mesh adjustment which is followed by the analysis of the next increment in time. While
this can result in some error, as long as the mesh is not overly coarse, it should be adequate.
Local adaptivity is available in a parallel analysis. In this case, the new elements remain in the same
domain as the parent element.
Number of Elements Created

The adaptive meshing procedure works by dividing an element and internally tying nodes to insure
compatibility. Figure 4-24 shows the process for a single quadrilateral element.
Original Element Level 1 Refinement Level 2 Refinement Level 3 Refinement

Figure 4-24 Single Quadrilateral Element Process
A similar process occurs for the triangles, tetrahedrons, and hexahedrons elements. You can observe that
for quadrilaterals the number of elements expands by four with each subdivision; similarly, the number
of elements increases by eight for hexahedrals. For the 8-node solid-shell element, the subdivision is
similar to 4-node quadrilaterals; there is no refinement through the thickness. If full refinement occurs,
( level x 2 ) ( level x 3 )
you observe that the number of elements is 2 for quadrilaterals and 2 for hexahedrons
elements. The number of levels also limits the amount of subdivisions that may occur.
Number of Elements
Level Quadrilaterals Hexahedrals
0 1 1
1 4 8
2 16 64
3 64 512
4 256 4096
For this reason, it is felt that the number of levels should, in general, be limited to three.
When adaptive meshing occurs, you can observe that discontinuities are created in the mesh as
shown below:
E
A
B C
To ensure compatibility node B is effectively tied to nodes A and C, and node D is effectively tied to
nodes C and E. All of this occurs internally and does not conflict with other user-defined ties or contact.
Boundary Conditions
When mesh refinement occurs, boundary conditions are automatically adjusted to reflect the change in
mesh. The rules listed below are followed:
1. Fixed Displacement
For both 2-D and 3-D, if both corner nodes on an edge have identical boundary conditions, the
new node created on that edge has the same boundary conditions. For 3-D, if all four nodes on a
face have identical boundary conditions, the new node created in the center of the face has the
same boundary conditions. Note that identical here means the same in the first degree of freedom,
second degree of freedom, etc. independently of one another.
CHAPTER 4 81
2. Point Loads
The point loads remain unchanged on the original node number.
3. Distributed Loads
Distributed loads are automatically placed on the new elements. Caution should be used when
using the FORCEM user subroutine as the element numbers can be changed due to the new
mesh process.
4. Contact
The new nodes generated on the exterior of a body are automatically treated as potential contact
nodes. The elements in a deformable body are expanded to include the new elements created.
After the new mesh is created, the new nodes are checked to determine if they are in contact.
It should be noted that new nodes on shells that are completely tied to the corresponding edge
nodes are not checked for contact.
Caution: None of the nodes of an element being subdivided should have a local coordinate system
defined through the TRANSFORMATION or COORD SYSTEM option.
Location of New Nodes

When an element is refined, the default is that the new node on an edge is midside to the two corner
nodes. As an alternative, the POINTS, CURVES, SURFACES, ATTACH EDGE and ATTACH FACE
options or the UCOORD user subroutine can be used. The CURVES and SURFACES options can be used
to describe the mathematical form of the curve or surface. If the corner nodes of an edge are attached to
the surface, the new node is placed upon the actual surface. This option to attach new nodes is available
both in a linear as well as a nonlinear analysis.
This is illustrated in Figure 4-25 and Figure 4-26, where initially a single element is used to represent a
circle. The circle is defined with the CURVES option and the original four nodes are placed on it using
the ATTACH EDGE option. Notice that the new nodes are placed on the circle.
Figure 4-25 Original Mesh and Surface

Level 1 Refinement Level 2 Refinement
Level 2 and 3 Refinement

Figure 4-26 Levels of Refinement
Adaptive Criteria
The adaptive meshing subdivision occurs when a particular adaptive criterion is satisfied. Multiple
adaptive criteria can be selected using the ADAPTIVE model definition option. These include:
Mean Strain Energy Criterion

The element is refined if the strain energy of the element is greater than the average strain energy in a
chosen set of elements times a given factor, f 1 .
total strain energy

element strain energy > ----------------------------------------------- * f 1 (4-31)
number of elements
Zienkiewicz-Zhu Criterion
The error norm is defined as either
2 2
π
2
=
 ( σ∗ – σ ) dV
------------------------------------------------------------- γ
2
=
 ( E∗ – E ) dV
------------------------------------------------------------- (4-32)
2 2 2 2
 σ dV +  ( σ∗ – σ ) dV  E dV +  ( E∗ – E ) dV
The stress error and strain energy errors are
2 2
X =  ( σ∗ – σ ) dV and Y =  ( E∗ – E ) dV (4-33)
CHAPTER 4 83
where σ* is the smoothed stress and σ is the calculated stress. Similarly, E is for energy.
An element is subdivided if
π > f 1 and (4-34)
X e l > f 2 * X/NUMEL + f 3 * X * f 1 ⁄ π ⁄ NUMEL (4-35)
or
γ > f 1 and (4-36)
Y e l > f 4 * Y/NUMEL + f 5 * Y * f 1 ⁄ γ ⁄ NUMEL (4-37)
where NUMEL is the number of elements in the mesh. If f 2 , f 3 , f 4 , and f 5 are input as zero, then
f 2 = 1.0 .
Zienkiewicz – Zhu Plastic Strain Criterion

p p 2
The plastic strain error norm is defined as α
2
=
 ( ε * – ε ) dV
--------------------------------------------------------------- .
p2 p p 2
ε dV +  ( ε * – ε ) dV
p* p 2
The plastic strain error is A = (ε – ε ) dV . The allowable element plastic strain error is
AEPS = f 2 * A ⁄ NUMEL + f 3 * A * f 1 ⁄ α ⁄ NUMEL . The element will be subdivided when
α > f 1 and A e l > AEPS . NUMEL is the number of elements in the mesh.
Zienkiewicz-Zhu Creep Strain Criterion

c c 2
Zienkiewicz-Zhu creep strain error norm is defined as β
2
=
 ( ε * – ε ) dV
-------------------------------------------------------------- . The creep
c2 c c 2
ε dV +  ( ε * – ε ) dV
c* c 2
strain error is B =  (ε – ε ) dV . The allowable element creep strain error is
AECS = f 2 * B ⁄ NUMEL + f 3 * B * f 1 ⁄ β ⁄ NUMEL .
The element will be subdivided when β > f 1 and B e l > AECS . NUMEL is the number of elements in
the mesh.
Equivalent Values Criterion

This method is based upon either relative or absolute testing using either the equivalent von Mises stress,
the equivalent strain, equivalent plastic strain or equivalent creep strain. An element is subdivided if the
current element value is a given fraction of the maximum (relative) or above a given absolute value.
m ax
σ vm > f 1 σ or σ
vm > f 2
vm
m ax
ε vm > f 3 σ or ε
vm > f 4
vm
Node Within A Box, Cylinder, or Sphere Criterion

An element is subdivided if it falls within the specified box, cylinder, or sphere, respectively. If all of the
nodes of the subdivided elements move outside the box, the elements are optionally merged back
together.
The location of this box can be repositioned using the UADAPBOX user subroutine.
The criteria may also be used in conjunction with the WELD FLUX model definition option. In such cases,
the motion of the weld source controls the location of the box.
Nodes In Contact Criterion

An element is subdivided if one of its nodes is associated with a new contact condition. In the case of a
deformable-to-rigid contact, this implies that the node has touched a rigid surface. For deformable-to-
deformable contact, the node can be either a tied or retained node. Note that if chattering occurs, there
can be an excessive number of elements generated. Use the level option to reduce this problem.
Temperature Gradient Criterion

An element is subdivided if the temperature gradient in the element is greater than a given fraction of the
maximum gradient in the solution. This is the recommended method for heat transfer.
Pressure Gradient Criterion

An element is subdivided if the pressure gradient in the element is greater than a given fraction of the
maximum gradient. This is the recommended method for diffusion analysis.
Electrical Potential Criterion

An element is subdivided if the electrical potential in the element is greater than a given fraction of the
maximum gradient. This is the recommended method for electrostatic analysis.
Magnetostatic Potential Criterion

An element is subdivided if the magnetic potential in the element is greater than a given fraction of the
maximum gradient. This is the recommended method for magnetostatic analysis.
CHAPTER 4 85
User-defined Criterion
The UADAP user subroutine can be used to prescribe a user-defined adaptive criteria. The user subroutine
UADAP2 can be used to prescribe a criterion for merging elements back (unsubdivide).
Previously Refined Mesh Criterion

Use the refined mesh from a previous analysis as the starting point to this analysis. The information from
the previous adapted analysis is read in.
Angle Between Shell Elements

An element is refined if the change in angle between neighboring shell elements is larger than the given
value. For each node, a normal is calculated by averaging shell element normals for elements connected
to the node. If an element is to be subdivided, the angle used for checking is two times the angle between
the averaged nodal normal and the element normal. Only the change in the normals from the undeformed
shape is used to avoid that elements originally connected at an angle get subdivided immediately.
Automatic Global Remeshing

In the analysis of metal or rubber, the materials may be deformed from some initial (maybe simple) shape
to a final, very often, complex shape. During the process, the deformation can be so large that the mesh
used to model the materials become highly distorted, and the analysis cannot go any further without
using some special techniques (see Figure 4-27). Remeshing/rezoning in Marc is a useful feature to
overcome the difficulties. Although global remeshing can be done manually using the REZONE option,
the automatic global remeshing procedure is recommended.
Figure 4-27 Mesh too Distorted and Simulation Stops (Left); Using Global Remeshing, the Simulation
Completes (click to play video)
Global remeshing can only be carried out on a contact body. Therefore, contact bodies are expected in the
analysis with global remeshing. The basic steps, which are automatic in global remeshing, are as follows:
1. Analysis checks remeshing criteria at the end of each increment. When one of the remeshing
criteria is met, analysis starts the remeshing procedure.
2. The deformed shape of the contact body is extracted. A new mesh is created by calling a
stand-alone mesher or an internal mesher.
3. The new mesh is checked and corrected to avoid any penetration or contact loss to other
contact bodies.
4. A data mapping is performed to transfer necessary data from the old, deformed mesh to the new
mesh.
5. The contact tolerance is recalculated (if not specified by you) and the contact conditions are
redefined.
6. Boundary conditions, if any, are transferred to the new mesh.
7. The analysis completes global remeshing and continues its computation based on the new mesh.
Global remeshing/rezoning can be used in two- or three-dimensional solid body or a three-dimensional
shell body. Most of the linear element types are supported. For shell, element type 75, 138, 139, and 140
are supported. Figure 4-28 shows a simple 3-D rubber seal remeshing.
When remeshing/rezoning in 2-D, Marc finds the outline of the body to be rezoned and repairs the outline
to remove possible penetration. Marc then calls the mesher to create a new mesh based on the clean
outline.
CHAPTER 4 87
2-D Global Remeshing of Rubber Seal Insertion
3-D Global Remeshing of Connecting Rod Forging
Deep Drawing of a Box with Shell Remeshing
Figure 4-28 3-D Automatic Remeshing and Rezoning of a Rubber Seal
When remeshing in 3-D with tetrahedral elements, Marc extracts and outputs the surface information
together with other contact surface information and the meshing control parameters. A stand-alone 3-D
mesher is then called to recreate 3-D surface mesh and the volume mesh. Contact check and volume
checks are used to ensure no penetration and loss of volume in the new mesh. In 3-D hexahedral
remeshing, hexahedral elements are build from the interior toward the surface. Projections are used to
project nodes to the surface.
When remeshing with 3-D shell elements, the 3-D mesher is called and surface mesh is created with
triangular or quadrilateral elements Contact check is also performed to remove possible penetration.
The automatic remeshing/rezoning feature uses the Updated Lagrange formulation by default.
For global remeshing controls, the following commands are required:
1. REZONING – parameter to activate global remeshing controls.
2. ADAPT GLOBAL – used in both model definition and history definition options to specify global
remeshing criteria, remeshing body, mesh generator, and meshing parameters.
The following are the supported features and limitations:
1. Analysis type:
Mechanical analysis, thermal-mechanical, thermal Joule mechanical, and electrostatic-structural
coupled analysis are supported.
2. Element types:
Low order, continuum element types are supported. In 2-D, these include both lower order
quadrilateral and triangular elements. In 3-D, only lower order tetrahedral element types are
supported including Herrmann type element 157. For 3-D shell elements, triangle element 138
and quadrilateral element 75, 139, and 140 are tested and supported.
3. Contact analysis:
The remeshing body is a meshed contact body. Contact information, including boundary
conditions defined through contact definitions, is re-determined based on the new mesh. Meshes
that are not defined in contact bodies are not supported.
• Support Rigid-Deformable contact
• Support Deformable-Deformable contact
• Support 2-D Self-contact
4. Boundary conditions:
Boundary conditions are allowed in a remeshing body. These boundary conditions are transferred
to the new mesh after remeshing. The new table style input format is required. These boundary
conditions include:
• Distributed load, flux, and current
• Point load, flux, and current
• Fixed displacement, temperature, and potential
The above boundary conditions defined in the interior of a contact body are not supported.
5. Local adaptivity meshing is not supported.
6. Parallel computation is not supported.
7. Other analysis options:
• PRE STATE model definition option is supported
• CYCLIC SYMMETRY model definition option is supported
• FLOW LINE and TRACK model definition options are supported
CHAPTER 4 89
Other supported global remeshing features include:

1. Remeshing procedures:
• pause and continue: a standalone mesher is called within the analysis process.
• stop and continue: analysis is stopped while a standalone mesher is called. This is to save
memory usage during the remeshing process. Analysis is automatically resumed after
remeshing is complete.
2. Forced remeshing control:
• When global remeshing is selected with automatic time stepping scheme, remeshing will be
forced when there is an element distortion during the analysis. This is carried out before the
end of the increment.
3. Remeshing criteria:
• Element distortion check
• Increment frequency check
• Contact penetration check
• Equivalent strain change check
• Immediate meshing
4. Meshing control:
• Based on input element size
• Based on input number of elements
• Change element types to triangle (2-D) or tetrahedral elements (3-D)
• Using Overlay mesher – quadrilateral elements
• Using advancing front mesher – quadrilateral or triangular elements
• Using Delauney triangulation mesher – triangular elements
• Using Hybrid MD Patran mesher – 3-D tetrahedral elements
• Using MD Patran MOM mesher - 3-D shell elements
• Using 3-D Overlay mesher – hexahedral elements (SuperForm only)
• Read user defined mesh files (.mesh)– also supported with RESTART and REAUTO options
5. Refinement and Coarsening control:
• Local refinement based on surface curvatures
• Local refinement based on thin section (2-D only)
• Local refinement based on refinement boxes
• Local refinement based on the USIZEOUTL user subroutine – for 2-D advancing front
mesher and Delauney triangulation mesher
• Interior coarsening using 2-D overlay mesher
• Interior coarsening using 3-D Tetrahedral mesher or Hexahedral mesher
6. Special Features:
• Support trimming load case to remove part of mesh in a contact body
• Support 2-D contact body split using the USPLIT user subroutine.
• Support user-defined meshing using the UMAKENET user subroutine. (3-D shell remeshing
not supported)
Remeshing Criteria
It is possible to choose any remeshing criteria simultaneously. How these criteria work is shown below.
Note: In general, frequent remeshing should be avoided for an effective and computationally
efficient analysis. Also, since each remeshing and subsequent rezoning step involves
interpolation and extrapolation of element variables, a possibility of error accumulation
exists as the analysis progresses when remeshing occurs too frequently.
Increment
Remeshing occurs at specified increment frequency.
Element Distortion
The identified body is remeshed when the distortion in the elements becomes large.
For 2-D analysis, the distortion check is based on corner angles. Remeshing is performed if the following
conditions are met:
• Any inner angle is greater than 175° or less than 5°
• Any inner angle change is greater than the user input data
For 3-D analysis, a volume ratio is measured to determine if remeshing is required. A volume ratio is
calculated based on each corner node and its connecting nodes. If v is the volume of a tetrahedron
formed by nodes 1, 2, 3, and 4 and s is the triangle area of nodes 2, 3, and 4, then the ratio:
h
r = ---
l
l
or if h > l , r = ---
h
Where h is the height and l is the equivalent length of the triangle. They can be calculated
respectively by
3v
h = ------
s
l = s
The default control ratio is 0.01. Any volume ratio of each corresponding corner node smaller than this
value forces the analysis to perform remeshing. Users can change this number to control the remeshing.
Contact Penetration
The identified body is remeshed when the curvature of the contact body is such that the current mesh
cannot accurately detect penetration.
CHAPTER 4 91
For 2-D analysis, the penetration remeshing criteria is based upon examining the distance between the
edge of an element and the contacted body.
For 3-D analysis, the penetration is measured from the center of each boundary element face to the
contacting surface.
By default, remeshing is carried out if the penetration is greater than twice
the contact tolerance and less than the target element size, where the
contact tolerance is 0.05 of the smallest element length and the target
element size is the element size for remeshing. This check does not apply
to the self-contact situation. Also, this penetration limit can be given by
user input to avoid too many remeshing operations.
Remeshing is activated when the penetration distance reaches or exceeds the given penetration tolerance.
Immediate
The identified body is remeshed before performing any analysis. This control can also be used to change
element types from quadrilateral elements to triangular elements or from hexahedral elements to
tetrahedral elements. For example, it is possible to use this control to change an initially defined mesh
using element type 7 (an 8 node-hexahedral element) to one using element type 157 (5 node-tetrahedral
element) in order to use the tetrahedral remeshing capability.
Strain Change
Equivalent strain measures element deformation. This criterion keeps a record of the strain change after
remeshing for each element. When any element of the body has a strain change greater than the control
limit, the remeshing will start.
Remeshing Techniques
For 2-D analysis, the remeshing techniques include outline extraction and repair and the mesh
generation. After the outline is extracted and repaired, the mesh generator is called to create a mesh. For
2-D remeshing, the new mesh is created either through the built-in mesh generator using the overlay
method or through a standalone mesh generator. When a standalone mesh generator is called, the
program pauses while waiting for the mesh generator to create the new mesh. This can be memory
intensive as both program and mesher are using the memory. Alternatively, the program can be stopped
automatically while the mesh is being created, freeing the memory for the mesh generation. The program
automatically resumes after the meshing is complete. In Marc, this is accomplished by using the AUTO
RESTART option, -autorst, through the command line parameter. In Marc Mentat, this is instructed
through JOB→JOB PARAMETERS→REMESHING CONTROL→STOP AND RESTART.
For 3-D analysis, the outer surface of the contact body is extracted to a data file. This data file also
contains remeshing control information. For tetrahedral meshing, a standalone 3-D mesher is called to
create the surface mesh with triangles and then mesh the body with the tetrahedral elements. MD Patran
mesher is used for the meshing. The new mesh nodes are adjusted to ensure that there is no penetration
on the contact surfaces.
Mesh Generation
There are various 2-D and 3-D mesh generators available in Marc based on Advancing front, Overlay,
and Delauney Triangulation techniques.
Advancing front mesher

This 2-D mesher creates either triangular, quadrilateral, or mixed triangular and quadrilateral mesh. For
a given outline boundary, it starts by creating the elements along the boundary. The new boundary front
is then formed when the layer of elements is created. This front advances inward until the complete
region is meshed. Some smoothing technique is used to improve the quality of the elements. In general,
this mesher works with any enclosed geometry and for geometry that has holes inside. The element size
can be changed gradually from the boundary to the interior allowing smaller elements near the boundary
with no tying constraints necessary (see Figure 4-29).
Figure 4-29 Advancing Front Meshing
Overlay meshing
This is a quadrilateral mesh generator. The 2-D overlay mesher is included within Marc. It creates a
quadrilateral mesh by forming a regular grid covering the center area of a body. A projection is then used
to project all boundary nodes onto the real surface and form the outer layer elements. For the surface that
is not in contact with other bodies, a cubic spline line is used to make the outline points smoother. This
mesher also allows up to two level refinements on boundary where finer edges are needed to capture the
geometry detail, and one level of coarsening in the interior where small elements are not necessary. This
refinement and coarsening are performed using the tying constraints (see Figure 4-30).
Figure 4-30 Local Refinement and Coarsening

CHAPTER 4 93
In general, overlay meshing produces good quality elements. However, it does not take geometry with
holes inside. It may not create a good mesh with geometry that has a very thin region or very irregular
shape. Also, because the regular grid is created based on the global coordinate system, it may create a
poor mesh if the geometry is not aligned with the global coordinate system.
Delauney Triangulation
This mesher creates only the triangular mesh. All the triangles satisfy the Delauney triangulation
property. It takes all the seed points on the improved outlines as initial triangulation points. The
triangulation is implemented by sequential insertion of new points into the triangulation until all the
triangles satisfy the local density and quality requirement. Delauney triangulation algorithm assures the
triangular mesh created has the best quality possible for the given set of points. The mesher also allows
geometry to have holes inside the body and a variation of the elements with different sizes (Figure 4-31).
Figure 4-31 Meshing with Delauney Triangulation
MD Patran Tetrahedral Mesher

MOM (Meshing-On-Mesh) Surface Mesher
This is a 3-D surface mesher that creates a surface mesh based on an input mesh. The Meshing-on-mesh
technology allows users to input a mesh, which is not good for the analysis, such as a distorted mesh or
a mesh from the STL file.
Tetrahedral Hybrid Mesher

This 3-D mesher is based on Delaunay triangulation and advancing front technology. It generates
tetrahedral elements based on an enclosed triangular surface mesh.
2-D Outline Extraction and Repair

The outline consists of all the boundary edges of the contact body. Once the outline is extracted from the
mesh, it is checked against other contacting bodies. The penetration from other contacting bodies is
marked and the outline is corrected according to the penetration. If the outline is to be used for the
advancing front or Delaunay mesher, some refinement and corrections are required to obtain better
outline points. These outline points become boundary nodes in the new mesh and cannot be altered
during the meshing process. The following procedures are taken by the program to prepare the new
outline for the remeshing:
Step 1: Marking the hard points: The hard points are those points that represent important features
of the original outline. Hard points are points that mark the beginning or end of the boundary
portion of the mesh that is in contact or assigned with distributed boundary conditions, and
the points that represent a sharp corner (such as 90° angle) or nodal boundary conditions.
Step 2: Marking the points with target element size and minimum element size: The outline points
are placed based on the target element size. The refinements and the user-defined outline
points are allowed to change this control. However, the minimum element size is used to
make sure the outline segment is not too small for the mesh generation.
Step 3: Marking the points with curvature consideration: The curvature control is used to allow
small outline segments to be used on the boundary where curvature radii are small. Three
neighboring outline points are used to calculate the curvature radius. The associate
curvature circle can then be formed by a number of line segments. With the same number of
line segments used to approximate a circle, the curve with a smaller curvature circle gets a
smaller line segment (Figure 4-31). Thus, we have a good variation of mesh size based on
the curvature.
2πR
r l = -----------
n
R
Figure 4-32 Curvature Consideration
Step 4: Marking the points with thin region consideration: In the area where a thin region is formed,
small elements are preferred. This can be done by detecting the thin region and using smaller
outline segments in the area. The segment length used for the thin area is to allow at least
three elements to be presented across the thin area.
Step 5: Smoothing the outline points: Smoothing is required on the outline so that the segment
length can gradually vary.
Step 6: Interpolations: Interpolation is the actual process to create a new outline based on the
extracted outline and the marking of the outline points. Linear interpolation is used on the
contact area to prevent penetration into the other contact bodies. Cubic spline line
interpolation is used for the free surfaces.
3-D Surface Extraction and Meshing

Similar to the 2-D outline extraction, 3-D surface faces are extracted before calling the mesh generator.
If the original mesh is a hexahedral mesh, the boundary element faces are converted into triangles. The
contact information on each element face is also extracted and output to a data file. A standalone mesher
reads the information and perform the following steps:
CHAPTER 4 95
Step 1: Construct surface information for contact projection and volume check.
Step 2: Generate triangle surface mesh using MD Patran surface mesher (MOM). Surface elements
are created observing the basic geometry features and contact conditions. Local refinement
will be carried out based on the surface curvatures and user defined refinement boxes (see
Figure 4-33).
Small Elements
Large Elements
Figure 4-33 Local Refinement
Step 3: Check contact penetration of new nodes and adjust coordinates if penetration is found.
Step 4: Generate tetrahedral element mesh with MD Patran Hybrid mesher. By default, a coarsening
factor is adopted to enlarge interior elements (see Figure 4-34).
Step 5: Output the mesh to a data file.
Figure 4-34 Interior Coarsening
Remeshing Based on the Target Number of Elements

Instead of giving the element size, users can give the target number of elements for the remeshing. A
uniform mesh assumption is used to compute the element size with the target number of elements. If the
target number of elements is not provided, the number of elements in the current mesh is used. For 2-D
analysis, the number of elements in the new mesh can also be controlled by using a percentage tolerance
to ensure that the new mesh does not have too many or too few elements. However, this tolerance control
requires remeshing trials and it cannot be used with the AUTO RESTART option.
Remeshing with Boundary Conditions

There are two types of boundary conditions that global remeshing needs to transfer from old mesh to the
new one:
• Contact boundary conditions:
This type of boundary conditions includes friction, heat convection, radiation, symmetry or cyclic
symmetry conditions. After a new mesh is created, these contact boundary conditions are
automatically updated based on the new contact detection using the new mesh.
• User defined boundary conditions:
If the remeshing body has user defined boundary conditions, such as, point loads, distributed
loads or fixed nodal displacements, the boundary conditions are transferred to the new mesh with
two different approaches:
a. Defined Set Approach
This approach allows users to directly apply boundary conditions to the mesh entities, such as,
element faces, edges, or nodes. Each boundary condition is arranged in a defined set with a set
name and a set type. Currently, the set type can be either a node or an element edge or face set.
The element edge sets in 2-D and face sets in 3-D are typically those of distributed boundary
conditions. Edges or faces in the set can be continuous or discontinuous. During the remeshing
stage, the boundaries of these edges or faces are marked and preserved. This allows the new
edges or faces to be created coincident with the old edges or faces in the set. After remeshing,
the element edges or faces in the set are replaced with the new element edges or faces (see
Figure 4-35). The node ids in nodal sets are automatically replaced with the new node ids (see
Figure 4-36).
Figure 4-35 Distributed Loads Transfer

CHAPTER 4 97
Figure 4-36 Point Loads Transfer
The nodal displacement or temperature conditions applying to all the nodes of an element edge
or face are treated as an element edge or face boundary condition. This means the new nodes
created on the same element edge or face will have the same displacement or temperature
boundary conditions. In 2-D, an element edge or face can take up to four boundary condition
sets for the remeshing. In 3-D, the number of sets on the same element face is two. There is no
limitation on the node set.
b. Geometry Attachment Approach
Boundary conditions can be assigned to geometry entities, such as points, curves and surfaces.
In this approach, these geometry entities are attached to the mesh entities of a remeshing body,
such as, a point attached to a node and a curve attached to a set of element edges. During
remeshing, mesh entities associated with these geometry entities are marked and preserved.
After remeshing, the geometry entities are re-attached to the new mesh entities. By doing so,
boundary conditions assigned to these geometry entities are automatically applied to the new
mesh (see Figure 4-37). In 3-D, the attached surface ID number cannot exceed 99 and only one
surface can be attached to an element face. In 3-D, curves can also be attached to the element
edges but only nodal displacement conditions are supported (see Figure 4-38). Geometry
attachments are shown.
Figure 4-37 Distributed Load with Curve Attachment

Figure 4-38 Geometry Attachment Boundary Conditions in 3-D
History Data Mapping Technique

History data such as stress, strain or temperature need to be transferred to the new mesh (dashed red),
from the old mesh (solid black). In general, this is carried out in the following steps (see Figure 4-39):
Figure 4-39 Locate New Nodes in Old Mesh
• Store nodal data at the nodal position based on the old mesh
• If old mesh is formed of quadrilateral or hexahedral elements, subdivide into triangles or
tetrahedrals, respectively.
• Extrapolate data from the integration points to the nodal position
• If old mesh is formed of quadrilateral or hexahedral elements, determine values at extra nodes
• Compute weighted averaged nodal data based on the contributions from different elements
• Locate new nodes in the old triangular/tetrahedral region
• Determine the value at the new node based upon linear interpolation within triangular/tetrahedral
region
• Data in the new integration points can be computed based on interpolating from the nodal data in
the new mesh
After data mapping, new equilibrium is achieved at the end of the new increment.
Chapter 5 Structural Procedure Library
5 Structural Procedure Library

Linear Analysis 100
 Nonlinear Analysis 105

Fracture Mechanics 147

Dynamics 168
 Inertia Relief 186


Superplasticity 192
 Soil Analysis 195

Mechanical Wear 199

Design Sensitivity Analysis 202
 Design Optimization 205

Defined Initial State with Result Data from Previous Analysis
(including AXITO3D) 212

Steady State Rolling Analysis 213
 Structural Zooming Analysis 216

Cure-Thermal-Mechanically Coupled Analysis 220

References 225
This chapter describes the analysis procedures in Marc applicable to structural problems. These
procedures range from simple linear elastic analysis to complex nonlinear analysis. A large number of
options are available, but you need to consider only those capabilities that are applicable to your physical
problem. This chapter provides technical background information as well as usage information about
these capabilities.
Linear Analysis
Linear analysis is the type of stress analysis performed on linear elastic structures. Because linear
analysis is simple and inexpensive to perform and generally gives satisfactory results, it is the most
commonly used structural analysis. Nonlinearities due to material, geometry, or boundary conditions are
not included in this type of analysis. The behavior of an isotropic, linear, elastic material can be defined
by two material constants: Young’s modulus E , and Poisson’s ratio v .
Marc allows you to perform linear elastic analysis using any element type in the program. Various
kinematic constraints and loadings can be prescribed to the structure being analyzed; the problem can
include both isotropic and anisotropic elastic materials.
The principle of superposition holds under conditions of linearity. Therefore, several individual solutions
can be superimposed (summed) to obtain a total solution to a problem.
Linear analysis does not require storing as many quantities as does nonlinear analysis; therefore, it uses
the core memory more sparingly. The ELASTIC parameter uses the assembled and decomposed stiffness
matrices to arrive at repeated solutions for different loads.
Note: Linear analysis is always the default analysis type in the Marc program.
Linear analysis in Marc requires only the basic input. Table 5-1 shows a subset of the Marc options and
parameters which are often used for linear analysis.
Table 5-1 Basic Input
Type Name
Parameter TITLE
SIZING
ELEMENTS
ELASTIC
ALL POINTS
CENTROID
ADAPTIVE
FOURIER
END
CHAPTER 5 101
Structural Procedure Library
Table 5-1 Basic Input (continued)

Type Name
Model Definition CONNECTIVITY
COORDINATES
GEOMETRY
ISOTROPIC
FIXED DISP
DIST LOADS
POINT LOAD
CASE COMBIN
END OPTION
More complex linear analyses require additional data blocks.

1. The ELASTIC parameter allows solutions for the same structural system with different loadings
(multiple loading analysis). When using the ELASTIC parameter, you must apply total loads,
rather than incremental quantities (for example., total force, total moment, total temperature) in
subsequent increments.
2. The RESTART option, used with the ELASTIC parameter and/or CASE COMBIN option, stores
individual load cases in a restart file.
3. The CASE COMBIN model definition option combines the results obtained from different loading
cases previously stored in a restart file.
4. The ADAPTIVE option can be used to improve the accuracy of the analysis.
5. The J-INTEGRAL option allows the study of problems of linear fracture mechanics.
6. The FOURIER option allows the analysis of axisymmetric structures subjected to
arbitrary loadings.
7. The ORTHOTROPIC or ANISOTROPIC model definition option activates the anisotropic
behavior option. In addition, the ANELAS, HOOKLW, ANEXP, and ORIENT user subroutines
define the mechanical and thermal anisotropy and the preferred orientations.
8. You can use both the linear SPRINGS and FOUNDATION options in a linear stress analysis.
Accuracy
It is difficult to predict the accuracy of linear elastic analysis without employing special error
estimation techniques. An inaccurate solution usually exhibits itself through one or more of the
following phenomena:
• Strong discontinuities in stresses between elements
• Strong variation in stresses within an element
• Stresses that oscillate from element to element
Error Estimates
The ERROR ESTIMATE option can also be used to obtain an indication of the quality of the results. You
can have the program evaluate the geometric quality of the mesh by reporting the aspect ratios and
skewness of the elements. In a large deformation updated Lagrange analysis, you can also observe how
these change during the analysis, which indicates mesh distortion. When the mesh distortion is large, it
is a good idea to do a rezoning step.
The ERROR ESTIMATE option can also be used to evaluate the stress discontinuity between elements.
Marc first calculates a nodal stress based upon the extrapolated integration point values. These nodal
values are compared between adjacent elements and reported. Large discrepancies indicate an inability
of the mesh to capture high stress gradients, in which case you should refine the mesh and rerun the
analysis or use local adaptive meshing.
The ERROR ESTIMATE option can be used for either linear or nonlinear analysis.
Adaptive Meshing
The ADAPTIVE option can be used to insure that a certain level of accuracy is achieved. The elastic
analysis is repeated with a new mesh until the level of accuracy requested is achieved.
Fourier Analysis
Through Fourier expansion, Marc analyzes axisymmetric structures that are subjected to arbitrary
loading. The FOURIER option is available only for linear analysis.
During Fourier analysis, a three-dimensional analysis decouples into a series of independent two-
dimensional analyses, where the circumferential distribution of displacements and forces are expressed
in terms of the Fourier series. Both mechanical and thermal loads can vary arbitrarily in the
circumferential direction. You can determine the structure’s total response from the sum of the
Fourier components.
The Fourier formulation is restricted to axisymmetric structures with linear elastic material behavior and
small strains and displacements. Therefore, conditions of linearity are essential and material properties
must remain constant in the circumferential direction.
To use Fourier expansion analysis in Marc, the input must include the following information:
• The FOURIER parameter allocates storage for the series expansion.
• Fourier model definition blocks for as many series as are needed to describe tractions, thermal
loading, and boundary conditions. Number the series sequentially in the order they occur during
the FOURIER model definition input. Three ways to describe the series are listed below:
• Specify coefficients a 0, a 1, b 1 … on the Fourier model definition blocks.
• Describe F ( θ ) (where θ is the angle in degrees about the circumference) in point-wise
fashion with an arbitrary number of pairs [ θ, F ( θ ) ] given on the blocks. Marc forms the
corresponding series coefficients.
• Generate an arbitrary number of [ θ, F ( θ ) ] pairs using the UFOUR user subroutine and let
the program calculate the series coefficients.
CHAPTER 5 103
You can obtain the total solution at any position around the circumference by superposing the
components already calculated after completion of all increments required by the analysis. The CASE
COMBIN option calculates this total solution by summing the individual harmonics which are stored in
the restart file.
The number of steps or increments needed for analysis depends on the number of harmonics that
are chosen. For a full analysis with symmetric and antisymmetric load cases, the total number of
increments equals twice the number of harmonics. Table 5-2 shows which Fourier coefficients are used
for a given increment.
Table 5-2 Fourier Coefficients – Increment Number
LOAD TERMS
INC. 1st DOF,Z 2nd DOF,R 3rd DOF,θ
0 a0 a0 0
1 0 0 a0
2 a1 a1 b1
3 b1 b1 a1
. . . .
. . . .
. . . .
2n an an bn
2n + 1 bn bn an
The magnitude of concentrated forces should correspond to the value of the ring load integrated around
the circumference. Therefore, if the Fourier coefficients for a varying ring load p ( θ ) are found from the
[ θ, p ( θ ) ] distribution, where p ( θ ) has the units of force per unit length, the force magnitude given in
the point load block should equal the circumference of the loaded ring. If p ( θ ) is in units of force per
radian, the point load magnitude should be 2π .
The Fourier series can be found for varying pressure loading from [ θ, p ( θ ) ] input with p expressed in
force per unit area. Marc calculates the equivalent nodal forces and integrates them around the
circumference. The distributed load magnitude in the distributed loads block should be 1.0.
Table 5-3 shows the elements in the program that can be used for Fourier analysis.
Table 5-3 Elements Used for Fourier Analysis
Element Type Description

62 8-node
73 8-node with reduced integration
63 8-node for incompressible behavior
74 8-node for incompressible behavior with reduced integration
90 3-node shell
The general form of the Fourier series expansion of the function F ( θ ) is shown in the equation below.
∞
F ( θ ) = a0 +  ( a n cos nθ + b n sin nθ ) (5-1)
n = 1
This expression expands the displacement function in terms of sine and cosine terms. A symmetric and
an antisymmetric problem are formulated for each value of n .
The displacements for the symmetric case, expressed in terms of their nodal values, are
u n = [ N 1, N 2, … ] cos nθ { u n } e
v n = [ N 1, N 2, … ] cos nθ { v n } e (5-2)
wn = [ N 1, N 2 , … ] sin n θ { w n } e
Nodal forces are

n
Z = Z0 +  Z n cos nθ
1
n
(5-3)
R = R0 +  R n cos nθ
1
n
T = T0 +  T n sin nθ
1
The value n = 0 is a special case in Fourier analysis. If only the symmetric expansion terms are used,
the formulation defaults to the fully axisymmetric two-dimensional analysis. The antisymmetric case for
n = 0 yields a solution for the variable θ that corresponds to loading in the tangential direction.
Analyze axisymmetric solids under pure torsion in this way.
CHAPTER 5 105
Modal Shapes and Buckling Load Estimations During a Fourier Analysis

During a Fourier analysis, Marc can be asked to estimate both the modal shapes and buckling loads for
each harmonic in the analysis. In either case, the program performs a Fourier analysis first and then
estimates the modal shapes/buckling load at prescribed harmonic numbers. In addition to the input data
required for a Fourier analysis (FOURIER parameter and FOURIER model definition option), the
following must also be added: DYNAMIC parameter and MODAL INCREMENT model definition option;
BUCKLE parameter and BUCKLE INCREMENT model definition option, for Fourier modal shape and
Fourier buckling load estimations, respectively.
In the Fourier modal analysis, the mass matrix in the eigenvalue equation is a constant matrix. The
stiffness matrix in the eigenvalue equation is the one associated with a prescribed harmonic of the Fourier
analysis. The expression of the eigenvalue equation is:
[ K m ]φ – ω 2 [ M 0 ]φ = 0 (5-4)
m 0
where [ K ] is the stiffness matrix associated with the mth harmonic of the Fourier analysis and [ M ]
is a constant matrix. Multiple modes for each harmonic can be extracted.
Similarly, in a Fourier buckling analysis, the stiffness matrices in the eigenvalue equation are
(respectively); the linear elastic stiffness matrix and the geometric stiffness matrix associated with the
prescribed harmonic of the Fourier analysis.
The eigenvalue equation is expressed
m m
[ K ]φ – λ [ K g ]φ = 0 (5-5)
m m
where [ K ] is the linear elastic stiffness matrix and [ K g ] is the geometric stiffness matrix, associated
with the mth harmonic of the Fourier analysis. The stresses used in the calculation of the geometric
stiffness matrix are those associated with the symmetric load case, m = 0 . Multiple buckling load
estimations for each harmonic are also available.
Nonlinear Analysis
The finite element method can be used for nonlinear, as well as linear, problems. Early development of
nonlinear finite element technology was mostly influenced by the nuclear and aerospace industries. In
the nuclear industry, nonlinearities are mainly due to the nonlinear, high-temperature behavior of
materials. Nonlinearities in the aerospace industry are mainly geometric in nature and range from simple
linear buckling to complicated post-bifurcation behavior. Nonlinear finite element techniques have
become popular in metal forming manufacturing processes, fluid-solid interaction, and fluid flow.
In recent years, the areas of biomechanics and electromagnetics have seen an increasing use of
finite elements.
A problem is nonlinear if the force-displacement relationship depends on the current state (that is, current
displacement, force, and stress-strain relations). Let u be a generalized displacement vector, P a
generalized force vector, and K the stiffness matrix. The expression of the force-displacement relation
for a nonlinear problem is
P = K ( P, u )u (5-6)
Linear problems form a subset of nonlinear problems. For example, in classical linear elastostatics, this
relation can be written in the form
P = Ku (5-7)
where the stiffness matrix K is independent of both u and P . If the matrix K depends on other state
variables that do not depend on displacement or loads (such as temperature, radiation, moisture content,
etc.), the problem is still linear.
Similarly, if the mass matrix is a constant matrix, the following undamped dynamic problem is
also linear:
P = Mu·· + Ku (5-8)
There are three sources of nonlinearity: material, geometric, and nonlinear boundary conditions. Material
nonlinearity results from the nonlinear relationship between stresses and strains. Considerable progress
has been made in attempts to derive the continuum or macroscopic behavior of materials from
microscopic backgrounds, but, up to now, commonly accepted constitutive laws are phenomenological.
Difficulty in obtaining experimental data is usually a stumbling block in mathematical modeling of
material behavior. A plethora of models exist for more commonly available materials like elastomers and
metals. Other material model of considerable practical importance are: composites, viscoplastics, creep,
soils, concrete, powder, and foams. Figure 5-1 shows the elastoplastic, elasto-viscoplasticity, and creep.
Although the situation of strain hardening is more commonly encountered, strain softening and
localization has gained considerable importance in recent times.
σ σ
ε ε
Elasto-Plastic Behavior Elasto-Viscoplastic Behavior
ε c
t
Creep Behavior
Figure 5-1 Material Nonlinearity
CHAPTER 5 107
Geometric nonlinearity results from the nonlinear relationship between strains and displacements on the
one hand and the nonlinear relation between stresses and forces on the other hand. If the stress measure
is conjugate to the strain measure, both sources of nonlinearity have the same form. This type of
nonlinearity is mathematically well defined, but often difficult to treat numerically. Two important types
of geometric nonlinearity occur:
a. The analysis of buckling and snap-through problems (see Figure 5-2 and Figure 5-3).
P Linear
P S
u Pc Neutral
Uns
u
Figure 5-2 Buckling
u
Figure 5-3 Snap-Through
b. Large strain problems such as manufacturing, crash, and impact problems. In such problems,
due to large strain kinematics, the mathematical separation into geometric and material
nonlinearity is nonunique.
Boundary conditions and/or loads can also cause nonlinearity. Contact and friction problems lead to
nonlinear boundary conditions. This type of nonlinearity manifests itself in several real life situations;
for example, metal forming, gears, interference of mechanical components, pneumatic tire contact, and
crash (see Figure 5-4). Loads on a structure cause nonlinearity if they vary with the displacements of the
structure. These loads can be conservative, as in the case of a centrifugal force field (see Figure 5-5);
they can also be nonconservative, as in the case of a follower force on a cantilever beam (see Figure 5-6).
Also, such a follower force can be locally nonconservative, but represent a conservative loading system
when integrated over the structure. A pressurized cylinder (see Figure 5-7) is an example of this.
Figure 5-4 Contact and Friction Problem
Figure 5-5 Centrifugal Load Problem (Conservative)

CHAPTER 5 109
Figure 5-6 Follower Force Problem (Nonconservative)
Figure 5-7 Pressurized Cylinder (Globally Conservative)
The three types of nonlinearities are described in detail in the following sections.
Geometric Nonlinearities
Geometric nonlinearity leads to two types of phenomena: change in structural behavior and loss of
structural stability.
There are two natural classes of large deformation problems: the large displacement, small strain
problem and the large displacement, large strain problem. For the large displacement, small strain
problem, changes in the stress-strain law can be neglected, but the contributions from the nonlinear terms
in the strain displacement relations cannot be neglected. For the large displacement, large strain problem,
the constitutive relation must be defined in the correct frame of reference and is transformed from this
frame of reference to the one in which the equilibrium equations are written.
The collapse load of a structure can be predicted by performing an eigenvalue analysis. If performed after
the linear solution (increment zero), the Euler buckling estimate is obtained. An eigenvalue problem can
be formulated after each increment of load; this procedure can be considered a nonlinear buckling
analysis even though a linearized eigenvalue analysis is used at each stage.
The kinematics of deformation can be described by the following approaches:
A. Lagrangian Formulation
B. Eulerian Formulation
C. Arbitrary Eulerian-Lagrangian (AEL) Formulation
The choice of one over another can be dictated by the convenience of modeling physics of the problem,
rezoning requirements, and integration of constitutive equations.
Lagrangian Formulation
In the Lagrangian method, the finite element mesh is attached to the material and moves through space
along with the material. In this case, there is no difficulty in establishing stress or strain histories at a
particular material point and the treatment of free surfaces is natural and straightforward.
The Lagrangian approach also naturally describes the deformation of structural elements; that is, shells
and beams, and transient problems, such as the indentation problem shown in Figure 5-8.
sz
Δu
Figure 5-8 Indentation Problem with Pressure Distribution on Tool
This method can also analyze steady-state processes such as extrusion and rolling. Shortcomings of the
Lagrangian method are that flow problems are difficult to model and that the mesh distortion is as severe
as the deformation of the object. Severe mesh degeneration is shown in Figure 5-9b. However, recent
advances in adaptive meshing and rezoning have alleviated the problems of premature termination of the
analysis due to mesh distortions as shown in Figure 5-9c.
The Lagrangian approach can be classified in two categories: the total Lagrangian method and the
updated Lagrangian method. In the total Lagrangian approach, the equilibrium is expressed with the
original undeformed state as the reference; in the updated Lagrangian approach, the current configuration
acts as the reference state. The kinematics of deformation and the description of motion is given in
Figure 5-10 and Table 5-4.
CHAPTER 5 111
(a) Original (b) Deformed Mesh

(Undeformed Mesh) Before Rezoning
(c) Deformed Mesh

After Rezoning
Figure 5-9 Rezoning Example
Previous f
t=n Δu
Current
t=n+1
Fn
un + 1
un
F
Reference
t=0
Fn+1 = fFn
Figure 5-10 Description of Motion
Depending on which option you use, the stress and strain results are given in different form as discussed
below. If the LARGE DISP or LARGE STRAIN parameters are not used, the program uses and prints
“engineering” stress and strain measures. These measures are suitable only for analyses without large
incremental or total rotation or large incremental or total strains.
Using the LARGE DISP parameter, Marc uses the total Lagrangian method. The program uses and prints
second Piola-Kirchhoff stress and Green-Lagrange strain. These measures are suitable for analysis with
large incremental rotations and large incremental strains.
Table 5-4 Kinematics and Stress-Strain Measures in Large Deformation
Configuration Measures Reference (t = 0 or n) Current (t = n + 1)

Coordinates X x
Deformation Tensor C (Right Cauchy-Green) b (Left Cauchy-Green)
Strain Measure E (Green-Lagrange) e (Logarithmic)
F (Deformation Gradient)
Stress Measure S (second Piola-Kirchhoff) σ (Cauchy)
P (first Piola-Kirchhoff)
With the LARGE STRAIN, Marc uses Cauchy stresses and true strains. This is suitable for analyses with
large elastic and plastic strains. Stress and strain components are printed with respect to the current state.
Theoretically and numerically, if formulated mathematically correct, the two formulations yield exactly
the same results. However, integration of constitutive equations for certain types of material behavior (for
example, plasticity) make the implementation of the total Lagrange formulation inconvenient. If the
constitutive equations are convected back to the original configuration and proper transformations are
applied, then both formulations are equivalent. However, for deformations involving excess distortions,
ease of rezoning favors the updated Lagrangian formulation. This is reflected in the fact that a rezoned
mesh in the current state is mapped back to excessively distorted mesh leading to negative Jacobian in
the total Lagrangian formulation.
The terminology total and updated Lagrangian has been used with some vagueness [1, 2]. In this
document, for a sequence of incremental motions at t = 0, 1, 2, …n and n + 1 , the total Lagrangian
formulation entails the use of t = 0 configuration as reference; while in the updated Lagrangian
configuration, the t = n + 1 (unequilibriated) configuration is the reference.
Total Lagrangian Procedure

The total Lagrangian procedure can be used for linear or nonlinear materials, in conjunction with static
or dynamic analysis. Although this formulation is based on the initial element geometry, the incremental
stiffness matrices are formed to account for previously developed stress and changes in geometry.
This method is suitable for the analysis of nonlinear elastic problems (for instance, with the Mooney,
Ogden, or NLELAST material behavior or the HYPELA2 user subroutine). The total Lagrangian approach
is also useful for problems in plasticity and creep, where moderately large rotations but small strains
occur. A case typical in problems of beam or shell bending. However, this is only due to the
approximations involved.
To activate the large displacement (total Lagrangian approach) option in Marc, use the LARGE DISP
parameter. Include the FOLLOW FOR parameter for follower force (for example, centrifugal or pressure
load) problems. This parameter forms all distributed loads on the basis of the current geometry. Do not
use the CENTROID parameter with this parameter. Always use residual load corrections with this
parameter. To input control tolerances for large displacement analysis, use CONTROL model
definition option.
CHAPTER 5 113
In the total Lagrangian approach, the equilibrium can be expressed by the principle of virtual work as:
0 0
 S i j δE i j dV =  b i δη i dV +  t i δη i dA (5-9)
V V A
0 0 0
0
Here S i j is the symmetric second Piola-Kirchhoff stress tensor, E i j , is the Green-Lagrange strain, b i is
0
the body force in the reference configuration, t i is the traction vector in the reference configuration, and
η i is the virtual displacements. Integrations are carried out in the original configuration at t = 0 . The
strains are decomposed in total strains for equilibrated configurations and the incremental strains
between t = n and t = n + 1 as:
n+1 n
Ei j = E i j + ΔE i j (5-10)
n
while the incremental strains are further decomposed into linear, ΔE i j and nonlinear, ΔE i j parts as:
n
ΔE i j = ΔEi j + ΔE i j
where ΔE is the linear part of the incremental strain expressed as:
n n
1 ∂Δu i ∂Δu j 1  ∂u k   ∂Δu k  ∂u k   ∂Δu k
ΔE = --
- -----------
- + -----------
- + 
--
- --------
-  ------------
- +  - -------------
-------- (5-11)
2 ∂X j ∂X i 2  ∂X i  ∂X j   ∂X j  ∂X i 
n
The second term in the bracket in Equation (5-11) is the initial displacement effect. ΔE is the nonlinear
part of the incremental strain expressed as:
1 ∂Δu k ∂Δu k
= ---  -------------  -------------
n
ΔE (5-12)
2  ∂X i   ∂X j 
Linearization of equilibrium of Equation (5-9) yields:
{ K 0 + K 1 + K 2 }δu = F – R (5-13)
where K 0 is the small displacement stiffness matrix defined as
0 0
( K0 )i j =  β i m n D m n p q β p q j dV
V0
K 1 is the initial displacement stiffness matrix defined as
u u u 0 u u
( K1 )i j =  { β i m n D m n p q β p q j + β i m n D m n p q β p qj + β i m n D m n p q β p q j } dv
V
0
0 u
in the above equations, β i m n and β i m n are the constant and displacement dependent symmetric shape
function gradient matrices, respectively, and D m n p q is the material tangent,
and K 2 is the initial stress stiffness matrix
( K2 )i j =  N i , k N j , l S k l dV
V
0
in which S kl is the second Piola-Kirchhoff stresses and N i, k is the shape function gradient matrix.
Also, δu is the correction displacement vector. Refer to Chapter 11 in this manual for more details on
the solution procedures. F and R are the external and internal forces, respectively.
This Lagrangian formulation can be applied to problems if the undeformed configuration is known so
that integrals can be evaluated, and if the second Piola-Kirchhoff stress is a known function of the strain.
The first condition is not usually met for fluids, because the deformation history is usually unknown. For
solids, however, each analysis usually starts in the stress-free undeformed state, and the integrations can
be carried out without any difficulty.
For viscoelastic fluids and elastic-plastic and viscoplastic solids, the constitutive equations usually
supply an expression for the rate of stress in terms of deformation rate, stress, deformation, and
sometimes other (internal) material parameters. The relevant quantity for the constitutive equations is the
rate of stress at a given material point.
It, therefore, seems most obvious to differentiate the Lagrangian virtual work equation with respect to
time. The rate of virtual work is readily found as
· ∂v k ∂δη k · ·
 S i j δE i j + S i j --------- ------------- dV =
∂X i ∂X j  b i δη i dV +  t i δη i dA (5-14)
V0 V0 A0
This formulation is adequate for most materials, because the rate of the second Piola-Kirchhoff stress can
be written as
· · ·
S i j = S i j ( E k l , S m n, E p q ) (5-15)
For many materials, the stress rate is even a linear function of the strain rate
· ·
S i j = D i j k l ( S m n, E p q )E k l (5-16)
CHAPTER 5 115
Equation (5-14) supplies a set of linear relations in terms of the velocity field. The velocity field can be
solved non iteratively and the displacement can be obtained by time integration of the velocities.
The second Piola-Kirchhoff stress for elastic and hyperelastic materials is a function of the Green-
Lagrange strain defined below:
Si j = Si j ( Ek l ) (5-17)
If the stress is a linear function of the strain (linear elasticity)
Si j = Di j k l Ek l (5-18)
the resulting set of equations is still nonlinear because the strain is a nonlinear function of displacement.
Updated Lagrangian Procedure

The Updated Lagrange formulation takes the reference configuration at t = n + 1 . True or Cauchy
stress and an energetically conjugate strain measure, namely the true strain, are used in the
constitutive relationship.
The updated Lagrangian approach is useful in:
a. analysis of shell and beam structures in which rotations are large so that the nonlinear terms in
the curvature expressions may no longer be neglected, and
b. large strain elasticity and plasticity analysis.
In general, this approach can be used to analyze structures where inelastic behavior (for example,
plasticity, viscoplasticity, or creep) causes the large deformations. The (initial) Lagrangian coordinate
frame has little physical significance in these analyses since the inelastic deformations are, by
definition, permanent.
The LARGE STRAIN parameter specifies large strain analysis within the framework of Updated
Lagrange formulation. It signals Marc to calculate a geometric stiffness matrix and the initial stress
stiffness matrix based on the current deformed configuration.
Note: Because large strain analysis involves nonlinearity, the CENTROID parameter must not be
used with this option.
For large strain analysis for rubber-like materials with incompressibility (such as materials defined with
MOONEY, OGDEN, GENT, and ARRUDBOYCE model definition options), Marc uses a mixed
formulation, in which both the displacement and the hydrostatic pressure are independent variables, to
overcome the numerical difficulties resulting from the volumetric constraints. For compressible
hyperelastic materials defined with FOAM model definition option, Marc uses conventional
displacement formulation.
For large strain elastic-plastic analysis, the default procedure in Marc uses a procedure based on an
additive decomposition of incremental strain into an elastic part and a plastic part, together with a mean
normal return-mapping algorithm. In this case, volumetric strain in a lower-order plane strain,
axisymmetric or 3-D brick element is assumed to be constant for von Mises plasticity to overcome
volumetric locking because of the possible large and incompressible plastic deformation.
Marc can also use a procedure (LARGE STRAIN parameter with option 2) based on a multiplicative
decomposition of deformation gradient into an elastic part and a plastic part together with a radial return-
mapping algorithm for large strain elastic-plastic analysis. A mixed formulation is used to deal with the
problem associated with volume constraints. This procedure is only available for continuum elements.
Because Herrmann elements do not support additive plasticity, Marc internally switches to the
multiplicative procedure. Herrmann elements have additional pressure degrees of freedom which
increase numerical costs; hence, it is generally more efficient to use displacement-based elements.
Marc uses Cauchy stress (true stress) and logarithmic strain with Updated Lagrange formulation.
It is instructive to derive the stiffness matrices for the updated Lagrangian formulation starting from the
virtual work principle in Equation (5-9).
Direct linearization of the left-hand side of Equation (5-9) yields:
 S i j ( d( δE i j ) ) dV =  ∇η i k σ k j ∇Δu i j dv (5-19)
V0 Vn + 1
where Δu and η are actual incremental and virtual displacements respectively, and σ k j is Cauchy
stress tensor.
s s
 dS i j δE i j dV =  ∇ η i j L i j k l ∇ ( Δu k l ) dv
(5-20)
V0 Vn + 1
s
∇ denotes the symmetric part of ∇ , which represents the gradient operator in the current configuration.
Also, in Equation (5-19) and Equation (5-20), three identities are used:
1
σ i j = --- F i m S m n F j n
J
s
δE i j = F m i ∇ η m n F n j (5-21)
and
1
L i j k l = --- F i m F j n F k p F l q D m n p q
J
in which D m n p q represents the material moduli tensor in the reference configuration which is convected
to the current configuration, L i j k . This yields:
{ K 1 + K 2 }δu = F – R (5-22)
CHAPTER 5 117
where K 1 is the material stiffness matrix written as
( K1 )i j =  βi m n Lm n p q βp q j
V
n+1
in which β i m n is the symmetric gradient operator-evaluated in the current configuration and σ k l is the
Cauchy stresses
and K 2 is the geometric stiffness matrix written as
( K2 )i j =  σ k l N i, k N j, l dv
Vn + 1
while F and R are the external and internal forces, respectively.

Keeping in view that the reference state is the current state, a rate formulation analogous to
Equation (5-14) can be obtained by setting:
∂ ∂
F i j = δ i j, δE i j = δd i j, --------- = --------, Si j = σi j (5-23)
∂X i ∂x i
where F is the deformation tensor, and d is the rate of deformation. Hence,
∇ ∂v k ∂δη k ·
σ t·i δη i da
 i j δd i j + σ i j --------- ------------- dv =
∂x i ∂x j  b i δη i dv +  (5-24)
Vn + 1 Vn + 1 An + 1
in which b i and t i is the body force and surface traction, respectively, in the current configuration
∇
In this equation, σ i j is the Truesdell rate of Cauchy stress which is essentially a Lie derivative of Cauchy
stress obtained as:
·
∇ –1 –1
σi j = F i n ( JF n k σ k l F m l ) F m j (5-25)
The Truesdell rate of Cauchy stress is materially objective implying that if a rigid rotation is imposed on
the material, the Truesdell rate vanishes, whereas the usual material rate does not vanish. This fact has
important consequences in the large deformation problems where large rotations are involved. The
constitutive equations can be formulated in terms of the Truesdell rate of Cauchy stress as:
∇
σ i j = Li j k dk (5-26)
Eulerian Formulation
In analysis of fluid flow processes, the Lagrangian approach results in highly distorted meshes since the
mesh convects with the material. Hence, an alternative formulation, namely Eulerian, is used to describe
the motion of the body. In this method, the finite element mesh is fixed in space and the material flows
through the mesh. This approach is particularly suitable for the analysis of steady-state processes, such
as the steady-state extrusion or rolling processes shown in Figure 5-11.
Figure 5-11 Rolling Analysis
The governing differential equations of equilibrium for fluid flow through an enclosed volume are now
written as:
D ( ρv i ) ∂σ
- = ρb i + ---------i-j
----------------- (5-27)
Dt ∂x j
D
where, ------ is the material time derivative of a quantity and v is the velocity of the particle flowing
Dt
through the mesh. For an incompressible fluid, Equation (5-27) along with continuity equation (mass
conservation) yields:
∂v i ∂v i ∂σ i j
ρ -------- + ρv j -------- = ρb i + ---------- (5-28)
∂t ∂x j ∂x j
The left-hand side of Equation (5-28) represents the local rate of change augmented by the convection
effects. The same principle can be called to physically explain the material time derivative of Cauchy
stress; that is, Truesdell rate of Cauchy stress. It can be seen from Equation (5-25) that:
∇ · ∂v i ∂v j ∂v k
σ i j = σ i j – --------- σ k j – σ i k --------- + σ i j --------- (5-29)
∂x k ∂x k ∂x k
The second and third terms on the right-hand side represent the convection effects. The last term vanishes
for a completely incompressible material; a condition enforced in the rigid-plastic flow of solids.
CHAPTER 5 119
5 Nonlinear Boundary Conditions

Structur There are three types of problems associated with nonlinear boundary conditions: contact, nonlinear
al support, and nonlinear loading. The contact problem might be solved through the use of special gap
Procedu elements or the CONTACT option. Nonlinear support might involve nonlinear springs and/or
re foundations. Sometimes nonlinearities due to rigid links that become activated or deactivated during an
Library analysis can be modeled through adaptive linear constraints. Nonlinear loading is present if the loading
system is nonconservative, as is the case with follower forces or frictional slip effects.
Discontinuities are inherent in the nature of many of these nonlinearities, making the solution by means
of incremental linear approximations difficult. Some of the most severe nonlinearities in mechanics are
introduced by nonlinear boundary conditions. It is, therefore, very important to be aware of potential
problem areas and to have a good understanding of the underlying principles. This awareness and
understanding enables you to validate numerical answers and to take alternative approaches if an initial
attempt fails.
Arbitrary Eulerian-Lagrangian (AEL) Formulation

In the AEL formulation referential system, the grid moves independently from the material, yet in a way
that is spans the material at any time. Hence, a relationship between derivative with respect to the
material and grid derivative is expressed as:
·
( ) = ( * ) + c i ( ), i (5-30)
p m
where c i is the relative velocity between the material particle, v and the mesh velocity, v ; for
i i
example,
p m
ci = vi – vi (5-31)
The second or latter term represents the convective effect between the grid and the material. Note that
p m
for v = v i , a purely Eulerian formulation is obtained. The equation of momentum, for instance, can
i
be represented as:
* p m ∂v i ∂σ i j
ρv i + ρ ( v j – v j ) ∂x = ∂x + b i
-------- ---------- (5-32)
j j
Due to its strong resemblance to the pure Eulerian formulation, AEL is also called
quasi-Eulerian formulation.
Nonlinear Boundary Conditions

There are three types of problems associated with nonlinear boundary conditions: contact, nonlinear
support, and nonlinear loading. The contact problem might be solved through the use of special gap
elements of the CONTACT option. Nonlinear support might involve nonlinear springs and/or
foundations. Sometimes nonlinearities due to rigid links that become activated or deactivated during an
analysis can be modeled through adaptive linear constraints. Nonlinear loading is present if the loading
system is nonconservative, as is the case with follower forces or frictional slip effects.
Discontinuities are inherent in the nature of many of these nonlinearities, making the solution by means
of incremental linear approximations difficult. Some of the most severe nonlinearities in mechanics are
introduced by nonlinear boundary conditions. It is, therefore, very important to be aware of potential
problem areas and to have a good understanding of the underlying principles. This awareness and
understanding enables you to validate numerical answers and to take alternative approaches if an initial
attempt fails.
Contact Problems
Contact problems are commonly encountered in physical systems. Some examples of contact problems
are the interface between the metal workpiece and the die in metal forming processes, pipe whip in piping
systems, and crash simulation in automobile designs.
Contact problems are characterized by two important phenomena: gap opening and closing and friction.
As shown in Figure 5-12, the gap describes the contact (gap closed) and separation (gap open) conditions
of two objects (structures). Friction influences the interface relations of the objects after they are in
contact. The gap condition is dependent on the movement (displacement) of the objects, and friction is
dependent on the contact force as well as the coefficient of (Coulomb) friction at contact surfaces. The
analysis involving gap and friction must be carried out incrementally. Iterations can also be required in
each (load/time) increment to stabilize the gap-friction behavior.
B
A n
Figure 5-12 Normal Gap Between Potentially Contacting Bodies
Two options are available in Marc for the simulation of a contact problem. A detailed description of these
options (gap-friction element and the CONTACT option) is given in Chapter 8 Contact of this manual.
Nonlinear Support
Marc provides two options for the modeling of support conditions: springs and elastic foundations. Both
linear and nonlinear springs can be specified in the input. In a nonlinear problem, the spring stiffness and
the equivalent spring stiffness of the elastic foundation can also be modified through a user subroutine.
In the nonlinear spring option, the incremental force in the spring is
ΔF = K ( Δu 2 – Δu 1 ) (5-33)
CHAPTER 5 121
where K is the spring stiffness, Δu 2 is the displacement increment of the degree of freedom at the
second end of the spring, and Δu 1 is the displacement increment of the degree of freedom at the first
end of the spring.
Use the SPRINGS model definition option for the input of linear and nonlinear spring data. The
USPRNG user subroutine may also be used to specify the value of K based on the amount of previous
deformation for nonlinear springs. In dynamic analysis, the SPRINGS option can also be used to define
a dashpot. In thermal analysis or electrical analysis (heat transfer, Joule heating, heat transfer pass of a
coupled analysis), the SPRINGS option can be used to define a thermal or electrical link.
In the elastic nonlinear FOUNDATION option, the elements in Marc can be specified as being supported
on a frictionless (nonlinear) foundation. The foundation supports the structure with an increment force
per unit area given by
ΔP n = K ( u n )Δu n (5-34)
where K is the equivalent spring stiffness of the foundation (per unit surface area), and Δu n is the
incremental displacement of the surface at a point in the same direction as ΔP n .
To input nonlinear foundation data, use the FOUNDATION model definition option.
To specify the value of K for the nonlinear equivalent spring stiffness based on the amount of previous
deformation of the foundation, use the USPRNG user subroutine.
Nonlinear Loading
When the structure is deformed, the directions and the areas of the surface loads are changed. For most
deformed structures, such changes are so small that the effect on the equilibrium equation can be ignored.
But for some structures such as flexible shell structure with large pressure loads, the effects on the results
can be quite significant so that the surface load effects have to be included in the finite element equations.
Marc forms both pressure stiffness and pressure terms based on current deformed configuration with the
FOLLOW FOR parameter. The FOLLOW FOR parameter should be used with the LARGE DISP or
LARGE STRAIN parameters. The CENTROID parameter should not be included due to the use of the
residual load correction.
Follower force point loads may be applied by either specifying this in the POINT LOAD option or by
applying a transformation which rotates with the displacements of the nodes specified in the COORD
SYSTEM option.
A special case of nonlinear loading is the resultant pressure due to a gas in an enclosed cavity. In such
problems, the pressure changes as the volume changes based upon the ideal gas law. This is discussed in
Chapter 9, in the Cavity Pressure Loading section of this manual.
Buckling Analysis
Buckling analysis allows you to determine at what load the structure will collapse. You can detect the
buckling of a structure when the structure’s stiffness matrix approaches a singular value. You can extract
the eigenvalue in a linear analyses to obtain the linear buckling load. You can also perform eigenvalue
analysis for buckling load in a nonlinear problem based on the incremental stiffness matrices.
The buckling option estimates the maximum load that can be applied to a geometrically nonlinear
structure before instability sets in. To activate the buckling option in the program, use the BUCKLE
parameter. If a nonlinear buckling analysis is performed, also use the LARGE DISP parameter.
Use the BUCKLE history definition option to input control tolerances for buckling load estimation
(eigenvalue extraction by a power sweep or Lanczos method). You can estimate the buckling load after
every load increment. The BUCKLE INCREMENT option can be used if a collapse load calculation is
required at multiple increments.
The linear buckling load analysis is correct when you take a very small load step in increment zero, or
make sure the solution has converged before buckling load analysis (if multiple increments are taken).
Linear buckling (after increment zero) can be done without using the LARGE DISP parameter, in which
case the restriction on the load step size no longer applies. This value should be used with caution, as it
is not conservative in predicting the actual collapse of structures.
In a buckling problem that involves material nonlinearity (for example, plasticity), the nonlinear problem
must be solved incrementally. During the analysis, a failure to converge in the iteration process or
nonpositive definite stiffness signals the plastic collapse.
For extremely nonlinear problems, the BUCKLE option cannot produce accurate results. In that case, the
AUTO INCREMENT history definition option allows automatic load stepping in a quasi-static fashion for
both geometric large displacement and material (elastic-plastic) nonlinear problems. The option can
handle elastic-plastic snap-through phenomena. Therefore, the post-buckling behavior of structures can
be analyzed.
The buckling option solves the following eigenvalue problem by the inverse power sweep method:
[ K + λΔK G ( Δu, u, Δσ ) ]φ = 0 (5-35)
where ΔK G is assumed to be a linear function of the load increment ΔP to cause buckling.
The geometric stiffness ΔK G used for the buckling load calculation is based on the stress and
displacement state change at the start of the last increment. However, the stress and strain states are not
updated during the buckling analysis. The buckling load is therefore estimated by:
P ( beginning ) + λΔP (5-36)
where for increments greater than 1, P ( beginning ) is the load applied at the beginning of the increment
prior to the buckling analyses, and λ is the value obtained by the power sweep or Lanczos method.
The control tolerances for the inverse power sweep method are the maximum number of iterations in the
power sweep and the convergence tolerance. The power sweep terminates when the difference between
the eigenvalues in two consecutive sweeps divided by the eigenvalue is less than the tolerance. The
CHAPTER 5 123
Lanczos method concludes when the normalized difference between all eigenvalues satisfies the
tolerance. The maximum number of iterations and the tolerance are specified through the BUCKLE
Perturbation Analysis
The buckling mode can be used to perform a perturbation analysis of the structure. In the manual mode,
a buckling increment is performed upon request and the coordinates are perturbed by a fraction of the
buckling mode or eigenvector. You can enter an individual eigenvector number and the fraction or can
request that a combination of modes be used. In the subsequent increments, the coordinates are:
φ φi
X = X + f ------ or X = X +
φ  fi -------
φi
- (5-37)
The manual mode can be activated by using the BUCKLE INCREMENT model definition, or BUCKLE
load incrementation option. In the automatic mode, the program checks for a nonpositive definite system
during the solution phase. When this occurs, it automatically performs a buckle analysis during the next
increment and updates the coordinates. The automatic mode can be activated by using the BUCKLE
INCREMENT option. Also, be sure to force the solution of the nonpositive definite system through the
CONTROL option or PRINT parameter.
Material Nonlinearities
In a large strain analysis, it is usually difficult to separate the kinematics from the material description.
Table 5-5 lists the characteristics of some common materials.
Table 5-5 Common Material Characteristics
Material Characteristics Examples Marc Models
Composites Anisotropic: Bearings, aircraft Composite continuum
1) layered, ds i j = C i j k dε k panels elements
21 constants
2) Fiber reinforced, Tires, glass/epoxy Rebars
E t
S = --- ( T CT – 1 )
2
one-dimensional strain
in fibers
Creep Strains increasing with time under Metals at high ORNL
constant load. temperatures, Norton
Stresses decreasing with time polymide films, Maxwell
under constant deformations. semiconductor
materials
Creep strains are
non-instantaneous.
Table 5-5 Common Material Characteristics (continued)

Material Characteristics Examples Marc Models
Elastic Stress functions of instantaneous Small deformation
strain only. Linear load- (below yield) for most
displacement relation. materials: metals, glass,
wood Hookes Law
Elasto- Yield condition flow rule and Metals von Mises Isotropic
plasticity hardening rule necessary to Soils Cam -Clay
calculate stress, plastic strain. Snow
Permanent deformation Wood Hill’s Anisotropic
upon unloading. Generalized Mohr-
Coulomb
Hyperelastic Stress function of instantaneous Rubber Mooney
strain. Nonlinear load- Ogden
displacement relation. Unloading Arruda-Boyce
path same as loading. Gent
NLELAST
Hypoelastic Rate form of stress-strain law Concrete Buyukozturk
Viscoelastic Time dependence of stresses in Rubber, Simo Model
elastic material under loads. Full Glass, industrial Narayanaswamy
recovery after unloading. plastics
Viscoplastic Combined plasticity and Metals Power law
creep phenomenon Powder Shima Model
Shape Superelastic and shape memory Biomedical stents, Aurrichio, Thermo-
Memory effect with phase transformations. Satellite antennae mechanical
A complete description of the material types mentioned is given in Chapter 7 of this manual. However,
some no characteristics and procedural considerations of some commonly encountered materials
behavior are listed next.
Inaccuracies in experimental data, misinterpretation of material model parameters and errors in user-
defined material law are some common sources of error in the analysis from the materials viewpoint. It
is useful to check the material behavior by running a single element test with prescribed displacement
and load boundary conditions in uniaxial tension and shear.
Large Strain Elasticity

Structures composed of elastomers, such as tires and bushings, are typically subjected to large
deformation and large strain. An elastomer is a polymer, such as rubber, which shows a nonlinear elastic
stress-strain behavior. The large strain elasticity capability in Marc deals primarily with elastomeric
materials. These materials are characterized by the form of their elastic strain energy function. For a more
detailed description of elastomeric material, see Elastomer in Chapter 7 on this manual.
CHAPTER 5 125
For the finite element analysis of elastomers, there are some special considerations which do not apply
for linear elastic analysis. These considerations, discussed below, include:
• Large Deformations
• Incompressible Behavior
• Instabilities
• Existence of Multiple Solutions
Large Deformations
The formulation is complete for arbitrarily large displacements and strains.
When extremely large deformations occur, the element mesh should be designed so that it can follow
these deformations without complete degeneration of elements. For problems involving extreme
distortions, rezoning must be done. Rezoning can be used with the formulation in the updated Lagrangian
framework using conventional displacement based elements.
Incompressible Behavior
One of the most frequent causes of problems analyzing elastomers is the incompressible material
behavior. Lagrangian multipliers (pressure variables) are used to apply the incompressibility constraint.
The result is that the volume is kept constant in a generalized sense, over an element.
Both the total, as well as updated Lagrange formulations, are implemented with appropriate constraint
ratios for lower- and higher-order elements in 2-D and 3-D. For many practical analysis, the LBB
(Ladyszhenskaya-Babuska-Brezzi) condition does not have to be satisfied in the strictest sense; for
example, four node quadrilateral based on Herrmann principle.
For elements that satisfy the LBB condition, error estimates of the following form can be established
h h m i n { k, + 1 }
u –u 1 + p –p 0 = O(h ) (5-38)
where k and are the orders of displacements and pressure interpolations, respectively. If
K = min { k, + 1 } , the rate of convergence is said to be optimal, and elements satisfying the LBB
condition would not lock.
The large strain elasticity formulation may also be used with conventional plane stress, membrane, and
shell elements. Because of the plane stress conditions, the incompressibility constraint can be satisfied
without the use of Lagrange multipliers.
Instabilities
Under some circumstances, materials can become unstable. This instability can be real or can be due to
the mathematical formulation used in calculation.
Instability can also result from the approximate satisfaction of incompressibility constraints. If the
number of Lagrangian multipliers is insufficient, local volume changes can occur. Under some
circumstances, these volume changes can be associated with a decrease in total energy. This type of
instability usually occurs only if there is a large tensile hydrostatic stress. Similarly, overconstraints give
rise to mesh locking and inordinate increase in total energy under large compressive stresses.
Existence of Multiple Solutions

It is possible that more than one stable solution exists (due to nonlinearity) for a given set of boundary
conditions. An example of such multiple solutions is a hollow hemisphere with zero prescribed loads.
Two equilibrium solutions exist: the undeformed stress-free state and the inverted self-equilibrating state.
An example of these solutions is shown in Figure 5-13 and Figure 5-14. If the equilibrium solution remains
stable, no problems should occur; however, if the equilibrium becomes unstable at some point in the
analysis, problems can occur.
x
Figure 5-13 Rubber Hemisphere
Figure 5-14 Inverted Rubber Hemisphere
When incompressible material is being modeled, the basic linearized incremental procedure is used in
conjunction with mixed variational principles similar in form to the Herrmann incompressible elastic
formulation. These formulations are incorporated in plane strain, axisymmetric, generalized plane strain,
and three-dimensional elements. These mixed elements may be used in combination with other elements
in the library (suitable tying may be necessary) and with each other. Where different materials are joined,
the pressure variable at the corner nodes must be uncoupled to allow for mean pressure discontinuity.
Tying must be used to couple the displacements only.
CHAPTER 5 127
Large Strain Plasticity

In recent years there has been a tremendous growth in the analysis of metal forming problems by the
finite element method. Although an Eulerian flow-type approach has been used for steady-state and
transient problems, the updated Lagrangian procedure, pioneered by McMeeking and Rice, is most
suitable for analysis of large strain plasticity problems. The main reasons for this are: (a) its ability to
trace free boundaries, and (b) the flexibility of taking elasticity and history effects into account. Also,
residual stresses can be accurately calculated.
The large strain plasticity capability in Marc allows you to analyze problems of large-strain, elastic-
plastic material behavior. These problems can include manufacturing processes such as forging,
upsetting, extension or deep drawing, and/or large deformation of structures that occur during plastic
collapse. The analysis involves both material, geometric and boundary nonlinearities.
In addition to the options required for plasticity analysis, the LARGE STRAIN parameter is needed for
large strain plasticity analysis.
In performing finite deformation elastic-plastic analysis, there are some special considerations which do
not apply for linear elastic analysis include:
• Choice of Finite Element Types
• Nearly Incompressible Behavior
• Treatment of Boundary Conditions
• Severe Mesh Distortion
• Instabilities
Choice of Finite Element Types

Accurate calculation of large strain plasticity problems depends on the selection of adequate finite
element types. In addition to the usual criteria for selection, two aspects need to be given special
consideration: the element types selected need to be insensitive to (strong) distortion; for plane strain,
axisymmetric, and three-dimensional problems, the element mesh must be able to represent
nondilatational (incompressible) deformation modes.
Nearly Incompressible Behavior

Most finite element types tend to lock during fully plastic (incompressible) material behavior. A remedy
is to introduce a modified variational principle which effectively reduces the number of independent
dilatational modes (constraints) in the mesh. This procedure is successful for plasticity problems in the
conventional “small” strain formulation. Zienkiewicz pointed out the positive effect of reduced
integration for this type of problem and demonstrates the similarity between modified variational
procedures and reduced integration. The lower-order elements, invoking the constant volumetric strain
or the lower-order elements, using reduced integration and hourglass control, behave well for nearly
incompressible materials. Higher order elements in Marc are formulated to be used in large strain
analysis including contact.
Treatment of Boundary Conditions

In many large strain plasticity problems, specifically in the analysis of manufacturing processes, the
material slides with or without friction over curved surfaces. This results in a severely nonlinear
boundary condition. The Marc gap-friction element and CONTACT option can model such sliding
boundary conditions.
Severe Mesh Distortion

Because the mesh is attached to the deforming material, severe distortion of the element mesh often
occurs, which leads to a degeneration of the results in many problems. The ERROR ESTIMATE option
can be used to monitor this distortion. To avoid this degeneration, generate a new finite element mesh
for the problem and then transfer the current deformation state to the new finite element mesh. The
global adaptive and rezoning procedure in the program is specifically designed for this purpose.
Instabilities
Elastic-plastic structures are often unstable due to necking phenomena. Consider a rod of a rigid-plastic
·
incompressible workhardening material. With ε the current true uniaxial strain rate and h the current
workhardening, the rate of true uniaxial stress σ is equal to
· ·
σ = Hε (5-39)
The applied force is equal to F = σA , where A is the current area of the rod. The rate of the force is
therefore equal to
·
F· = σ A + σA· (5-40)
On the other hand, conservation of volume requires that
·
Aε + A· = 0 (5-41)
Hence, the force rate can be calculated as
·
F· = ( H – σ )Aε (5-42)
Instability clearly occurs if σ > H . For applied loads (as opposed to applied boundary conditions), the
stiffness matrix becomes singular (nonpositive definite).
For the large strain plasticity option, the workhardening slope for plasticity is the rate of true stress versus
the true plastic strain rate. The workhardening curve must, therefore, be entered as the true stress versus
the logarithmic plastic strain in a uniaxial tension test.
Computational Procedures for Elastic-Plastic Analysis

Three basic procedures for plasticity exist in Marc. In this section, the variational form of equilibrium
equations and constitutive relations, and incompressibility are summarized. Issues regarding return
mapping procedures for stress calculation in three-dimensional and plane-stress conditions are
also discussed.
CHAPTER 5 129
For notational purpose, three configurations are considered at any point, original ( t = 0 ), previous
( t = n ) and current ( t = n + 1 ). An iterative procedure, full or modified Newton-Raphson, secant,
or arc-length is used to solve for the equilibrium at t = n + 1 .
1. Small Strain Plasticity (reference configuration: t = 0 ):
In this approach, the basis of variational formulation are 2nd Piola-Kirchhoff stress, S and
Green-Lagrange strain, E . Equilibrium of the current state can thus be represented by the
following virtual work principle:
 S : δE dV =  t : δη dA +  b : δη dV (5-43)
V A V
n n n
During the increment, all state variables are defined with respect to the state at t = n and are
updated at the end of increment upon convergence.
The linearized form of constitutive equations is given as:
ep
dS = L : dE – σ.dE – dE.σ (5-44)
ep
in which S is the second Piola-Kirchhoff stress, σ is the Cauchy stress, and L is the elasto-
plastic moduli. This linearization has the advantage that it is fully independent of the rotation
increment, but the disadvantage is that the linearization causes errors equal to the square of the
strain increment. Moreover, imposing the incompressibility condition in terms of the trace of
Green-Lagrange strains leads to errors in the form of fictitious volume changes in fully developed
plastic flow.
The above procedure works well in the context of small strain plasticity. However, in many large
deformation problems including metal forming processes, the plastic strain increments can be
very large and the above procedure can lead to large errors in the results. The two finite strain
plasticity formulations to model the large inelastic strains are: rate based (hypoelastic) and total
(hyperelastic).
2. Finite Strain Plasticity with additive decomposition of strain rates (reference configuration:
t = n + 1 ).
This formulation is based on the integration of the constitutive equations in the current
configuration. To maintain objectivity, the notion of rotation neutralized stress and strain
measures is introduced. All objective stress rates, are manifestation of Lie derivative:
·
–1 –T T
L v ( Σ ) = Φ ( Φ ΣΦ ) Φ (5-45)
where Φ is a deformation measure; for example, deformation gradient or rotation tensor, R

while Σ is a stress measure in the current configuration. The general form for an objective stress
rate is:
∇ · T
σ = σ – σΩ – Ωσ + ασtr ( d ) (5-46)
From Table 5-6, it can be seen that there is a possibility of a number of stress rates. It can be
observed that while all the above stress rates are objective, the Truesdell and Durban-Baruch rates
would not yield symmetric matrices.
Table 5-6 Objective Stress Rates
Stress Rates Ω α
Truesdell L 1
Cotter-Rivlin L 0
Oldroyd -LT 0
Jaumann-Zaremba-Noll W 0
Green-McInnis-Nagdhi –1 0
R· R
Durban Baruch 1 1
--- ( D + W )
2
Marc’s implementation of rate formulation involves the use of Jaumann rate of Cauchy stress
which is obtained as an average of the Oldroyd and Cotter-Rivlin stress rates. Thus, the Jaumann
rate can be written as:
∇ ·
σ = σ – Wσ + σW (5-47)
∇ · T
σ = σ + Ω ⋅ σ + σ ⋅ Ω + tr ( dε )σ (5-48)
∇
where ( · ) is the ordinary rate and ( ) is the objective rate with W is the spin or the antisymmetric
part of the velocity gradient, L . The last term is neglected because of the incompressible nature
of plasticity.
The equilibrium in the current state is given by the virtual work principle at t = n + 1 :
 σ : δε dv =  t . δηda +  b . δη dv (5-49)
V A V
n+1 n+1 n+1
Linearization of the above form leads to the variational statement:
T
 [ dσ : δε – 2 ( dε ⋅ σ ) : δε + tr ( dε )σ : δε + σ : { ( ∇ u ) ⋅ ( ∇ η ) } ] dv =
V
n+1 (5-50)
 
d   b . δη dv +  t . δη da –  σ : δε dv
V A
 V
n+1 n+1 n+1
CHAPTER 5 131
where,
RN T
dσ = R ⋅ dσ ⋅R (5-51)
RN
Within the context of rotation neutralized form of constitutive relations, dσ is defined as:
RN RN RN
dσ = L : dε (5-52)
RN T
where, dε = R ⋅ dε ⋅ R . Also, the rotation neutralized strain can be calculated by:
RN –T MID –1
ε = (U + I) ( F + I )ε (F + I)(U + I )
MID
with ε being the mid-increment strain, a good approximation for incremental logarithmic
strain measure.
Where e = U – I is the engineering strain and U is the stretch tensor obtained by the polar
decomposition of F . Admitting an error of the order of e 2 in the approximation of the
logarithmic strain, one obtains:
ep
dσ = L : dε (5-53)
ep RN T T
where L = R ⋅ (R ⋅ L ⋅R )⋅R (5-54)
During computations, the third term of Equation (5-50), tr ( dε ) ⋅ σ : δε , is neglected due to its
nonsymmetric nature.
In a fully developed plastic flow, the volumetric part of the energy can become extremely large
and lead to volumetric locking. Hence, a special treatment of incompressibility is done to relax
the volumetric constraint in an assumed strain format. The volumetric part of deformation
gradient is modified such that, the assumed deformation gradient is:
1 1
--- – ---
3 3 (5-55)
F = J J F
Linearization of F in the original state relates it to the displacement gradients in the current
state as:
DF = ∇ ( Δu ) F (5-56)
where
1 1
∇ ( Δu ) = ( Δu ) i, j – --- div ( Δu ) δ i j + --- div ( Δu ) δ i j (5-57)
3 3
where the first two terms are evaluated at each integration point and the last term is averaged over
the element.
For a lower order element, the procedure leads to mean or constant dilatation approach.
Considering Equations (5-50) to (5-57) the resulting system can be expressed as:
s ep s s s T
 [∇ η : L : ∇ ( Δu ) – 2∇ η ⋅ σ : ∇ ( Δu ) + σ : ∇ ( Δu ) ⋅ ( ∇η ) ] dv =
V
n+1
  s T (5-58)
d   b . δη dv +  t . δη da –  [ ∇ η ] : σ dv
V A
 V
n+1 n+1 n+1
In the event, the elastic strains become large in an elastic-plastic analysis the rate based
constitutive equations do not accurately model the material response. This results from the fact
that the elasticity matrix are assumed to have constant coefficients in the current
deformed configuration.
In the next formulation, the rate of deformation tensor is decomposed multiplicatively into the
elastic and plastic parts to resolve these problems.
3. Finite strain plasticity with multiplicative decomposition of deformation gradient.
An alternative formulation, based on the multiplicative decomposition of the deformation
gradient has been implemented in Marc, namely:
e θ p
F = F F F (5-59)
e θ p
where F , F , and F are (elastic, thermal, and plastic) deformation gradients, respectively. The
thermo-mechanical coupling is implemented using the staggered approach.
The above decomposition has a physical basis to it as the stresses are derived from quadratic -
logarithmic strain energy density function:
e 2
W = --- λ ( ln J ) + μ tr  --- ln b 
1 e 2 1
(5-60)
2  2 
This function has been chosen due to the availability of the material coefficients from material
testing in a small strain case.
In the metal forming applications, the elastic strains are negligible and the rate-based (or
hypoelastic) as well as the total (or hyperelastic) form of constitutive equations give virtually the
same results. However, many polymers, metals subjected to large hydrostatic pressures, high
velocity impact loading of metals and processes, where shape changes after deformation need to
be evaluated precisely, the hyperelastic formulation yields physically more meaningful results.
The return mapping procedure for the calculation of stresses is based on the radial return
procedure. With the use of exponential mapping algorithm, the incompressibility condition is
imposed exactly:
p
det F = 1 (5-61)
CHAPTER 5 133
The strain energy is separated into deviatoric and volumetric parts in the framework of mixed
formulation. The general form of three-field variational principle is:
e e e
π ( u, p , J ) =  [W ( b ) + U ( J ) + p ( J – J )] dV (5-62)
V
0
2
where, – --- (5-63)
3
b = J b
e
W ( b ) and U ( J ) are the deviatoric and elastic volumetric parts of the free energy, p is the
e e
pressure, J and J are elastic pointwise and average elastic Jacobian of the element,
respectively. The volumetric free energy can be of any form in Equation (5-61). However, if the
incompressibility is enforced in a pointwise fashion and the volumetric free energy is assumed to
be of the form:
1 2
---
e 9 e 3
U ( J ) = --- K ( J ) – 1 (5-64)
2
the perturbed Lagrangian form of the variational principle can be cast in a two-field framework
as:
 1--- 
 e 3  P2
π ( u, P ) =  W ( b ) + 3P  ( J ) – 1  – ------- dV (5-65)
V0   2K
 
2
---
e 3
Note that P = p ( J ) , hence, P is not the true pressure in Equation (5-65). Choice of the
volumetric free energy in Equation (5-64) is not arbitrary and is described in more detail in
Chapter 7 in the Elastomer section of this manual.
Linearization of the equilibrium condition arising from the stationary of the variational principle
Equation (5-65), yields:
  1
--- 
1 
∇ η :  --- C d e v + P ( J ) --- 1 ⊗ 1 – 2I   : ∇ u +
s e 31  s
 J  3 
V   
n+1  
1
---
dP e 3T (5-66)
σ : { ( ∇ ( Δu ) ). ( ∇η ) } + ------ ( J ) 1 : ∇η dv =
J
  s T
d   b . δη dv +  t . δη da –  ( ∇ η ) : σ dv
V A n+1
 V
n+1 n+1
The linearization of the constraint equation defines the term dP in Equation (5-66).
The derivation of the spatial tangent C d e v follows by a push forward of the material tangent into
the current configuration:
–1 d e v –T
Hence, S = F τ F (5-67)
where S is the pull back of the deviatoric Kirchhoff stress tensor. Assuming the entire
incremental deformation to be elastic, a symmetric stress tensor, Sˆ can be defined with respect
p
to a fixed plastic intermediate configuration F n as:
t r –1 d e v t r –T
Sˆ = ( F e ) τ ( Fe ) (5-68)
also noting that,
3
βA
Sˆ =  S A N A ⊗ N A where Sˆ A = ----------------- (5-69)
tr 2
A = 1 ( λA e )
3 3 3
dSˆ =  dSˆ A N A ⊗ N A +   ( Sˆ B – Sˆ A ) Ω B A N A ⊗ N B (5-70)
A = 1 A = 1B ≠ A
1  ∂β A  tr tr
where dSˆ A = ----------------------------------- - – 2β A δ A B λ B e dλ B e
 ----------- (5-71)
tr 2 tr 2
( λA e ) ( λB e )  ∂ε B etr

and dN A = Ω A B N B (5-72)
CHAPTER 5 135
3
tr T tr tr tr
similarly, Cˆ = ( F e ) F e with F e =  λA e nA ⊗ NA (5-73)
A = 1
3
tr tr
dCˆ =  2λ B e dλ B e N B ⊗ N B +
B = 1
(5-74)
3 3
 tr 2 tr 2 
   ( λB e ) – ( λB e )  ΩB A NA ⊗ NB
 
A = 1 B≠A
The algorithmic elasto-plastic tangent can be obtained as:
∂S T T ∂Ŝ
C d e v = 2F  F ------- F  F = 2 ( F etr ) ( F etr ) ------- ( F etr ) ( F etr )
T T
(5-75)
 ∂C  ∂Ĉ
Combining Equations (5-70), (5-75), and (5-76), the spatial form of deviatoric part of the
tangent as:
3 3  ∂β A 
C dev =   - – 2β A δ A B n A ⊗ n A ⊗ n B ⊗ n B
 -----------
1 B ≠ A  ∂ε B e 
tr
A =
tr 2 tr 2 (5-76)
3 3 ( λA e ) βB – ( λB e ) βA
+   -------------------------------------------------------- ( n A ⊗ n B ⊗ n A ⊗ n B +
tr 2 tr 2
A = 1B ≠ A ( λB e ) – ( λA e )
nA ⊗ nB ⊗ nB ⊗ nA )
The singularity for two- or three-equal stretch ratios is removed by repeated application of
L’Hospital’s rule. Due to separate interpolation of the pressure, a larger system of equations needs
to be solved. A static condensation of the element variables is carried out before going in the
solver. This step renders the singularity of the system to be inversely proportional to the bulk
modulus of the material. Due to the multiplicative split of the deformation gradient and an
additive split of the free energy into deviatoric and volumetric parts, this framework is eminently
suitable for implementation of general nonlinear elastic as well as inelastic material models.
Besides, compressible and nearly incompressible material response can be naturally handled.
Finally, the stress calculation is done via return mapping to the yield surface. There are several
methods to return the stress. Among them, the popular ones include:
Closest point projection (which reduces to radial return for von Mises), mean normal
approach, midpoint rule, trapezoidal rule, tangent cutting plane algorithm, and
multistage method.
Krieg and Krieg (1977) investigated the radial return algorithm which has been proposed by
Wilkins (1964). The third algorithm analyzed was the so-called secant stiffness method
proposed by Rice and Tracy (1973) which represented a special case of the generalized
Trapezoidal rule for nonhardening materials algorithm in contrast to the fully implicit radial
return algorithm and the fully explicit tangent stiffness approach. The radial return method
(shown in Figure 5-15) was found to be the most accurate among the three analyzed,
particularly when large strain increments were used. The radial return and mean normal
method are available in Marc and are described next.
′
σ1 Δσ
el
σ fin al σ*
′ ′
σ2 σ3
Figure 5-15 Radial Return Method
1. Closest Point Projection procedure:

The closest point projection or the backward Euler algorithm reduces to a radial return scheme for
cases where there is no anisotropy or plane stress condition. The trial stress is determined as:
d d
σ r r = σ + 2GΔε d (5-77)
The final stress state is obtained by simply scaling the trial stress by a scale factor, β
2GΔλ
where, β = 1 – --------------- (5-78)
d
σr r
the plastic consistency parameter Δλ , is obtained by solving for consistency condition iteratively.
An explicit form of the material tangent for the isotropic, von Mises yield surface can be given as:
= 2G β  I – --- 1 ⊗ 1 – 2G γ N
ep 1 ˆ ⊗N
ˆ
L (5-79)
 3 
1
where, γ = ------------------ – 1 + β (5-80)
H
1 + -------
3G
CHAPTER 5 137
ˆ
Here N is the return direction normal to the current yield surface. For the plane stress case, the
zero normal stress condition is explicitly imposed in the yield condition. Since the yield surface
is not a circle in the plane, the return direction is not radial anymore.
2. Mean-Normal method:
Assuming the entire deformation to be elastic in nature, the trial stress is evaluated as:
d d d d
σ m n = σ + 2αGΔε + ( 1 – α )GΔε (5-81)
Due to the use of pressure independent yield function, only the deviatoric stresses are evaluated.
As shown in Figure 5-16, ( 1 – α ) represents the fraction of strain increment for which the plastic
flow occurs.
Δσ e l
σ* + σ o + Δσ e l
d
σ1
σ f i n al
σ*
σo
d
σ2 σ3
d
Figure 5-16 Mean-Normal Method
d
σ includes the deviatoric stresses at the previous increment scaled such that it satisfies the yield
condition. The yield criterion might not be exactly satisfied due to temperature effects, numerical
integration of elastic-plastic relationships, or accumulated numerical inaccuracy.
F ( σ ) = 0 indicates that the stress state is exactly on the yield surface. If F ( σ ) > 0 , scale σ by
a factor λ such that:
F ( λσ ) = 0 0<λ<1 (5-82)
The equivalent plastic strain, plastic strain tensor and the elasto-plastic moduli are obtained as:
e ˆ
p (C : N ) : Δε̂ ( 1 – α )
Δε = ----------------------------------------------------- (5-83)
ˆ e ˆ
N:C :N +H
p p ˆ
Δε = Δε N (5-84)
e ˆ e ˆ
ep e (C : N ) ⊗ (C : N )
L = C – ( 1 – α ) --------------------------------------------------- (5-85)
ˆ :C :N e ˆ +H
N
ˆ is the mean-normal direction to the yield surface, H is the hardening modulus and C e
where, N
is the elastic tangent moduli. For nonhardening materials, the mean normal algorithm is a special
case of trapezoidal rule.
Marc plasticity algorithms are unconditionally stable and accurate for moderate strain increments.
However, overall global stability of the system can be dictated by other considerations like
contact, buckling which then guide the selection of appropriate load increment size. It should also
be recognized that the algorithmically consistent tangent in the closest point projection algorithm
is based on the current state, and a lack of embedded directionality (unlike secant methods) in the
tangent can lead to divergent solutions unless line search or automatic time stepping algorithms
are used. However, when it converges it shows quadratic convergence as compared to linear
convergence for the mean-normal scheme.
3. Multistage Return Mapping:
The return mapping scheme described in this section is used for isotropic and anisotropic
plasticity when LARGE STRAIN is used to describe the Updated Lagrange Procedure. Marc uses
this scheme for the Hill and Barlat yield criteria. This scheme is not available for pressure
dependent yield criteria like Mohr Coulomb. In a large strain analysis, Marc automatically
chooses the return mapping scheme that is most suited for the given material.
The increment of Cauchy stress for elasto-plasticity becomes
e e e p
Δσ̂ = Cˆ : Δε̂ = Cˆ : ( Δε̂ – Δε̂ ) . (5-86)
In Equation (5-86), superscript ‘ ˆ ’ means a materially embedded co-rotational coordinate

system. For the numerical implementation of Equation (5-86), the general midpoint rule can be
expressed as follows:
eˆ
σ̂ n + 1 = σ̂ nT + 1 – λĈ N n+α (5-87)
ˆe
where σ̂ nT + 1 = σ̂ n + C Δε̂
(Case 1) if F ( σ̂ nT + 1 ) ≤ 0 , λ = 0 , σ̂ n + 1 = σ̂ nT + 1
(Case 2) if F ( σ̂ nT + 1 ) ≤ 0 , λ = λ such that F ( σ̂ nT + 1 ) = 0
where superscript T stands for a trial state. For Case 2, the condition stating that the updated stress
p
stays on the strain-hardening curve ( σ = ρ ( ε ) ) provides the following condition:
eˆ
F ( λ ) = σ ( σ̂ nT + 1 – λĈ N n + α ) – ρ ( εn + λ ) = 0 .
p (5-88)
Equation (5-88) is a nonlinear equation to solve for λ .

CHAPTER 5 139
5+ If the strain increment is not small enough, it is difficult to obtain the numerical solution of
Equation (5-88) even though it has a mathematical solution. Therefore, an iterative scheme,
which utilizes the control of the potential residual, is introduced. The method is applicable to a
nonquadratic yield function and a general hardening law. For the implementation of the
algorithm, Equation (5-88) has been modified to the following relationship with α = 1 ; that is,
Structura ˆ eN
F ( λ ( k ) ) = σ ( σ̂ T – λ ( k ) C ˆ ) – ρ(ε p + λ ) = F (5-89)
(k) n (k) (k)
l
Procedur where
e Library F ( λ( 0 ) = 0 ) = F( 0 ) , { F( k ) F( 0 ) > F( 1 ) > F( k ) > … > F( N ) ( F( N ) = 0 ) , k = 0 ∼ N } ,
ΔF = ( F k – 1 – F k ) < σ Y ( = ( σ ( ε p = 0 ) ) ) and F k = 1∼N–1 are prescribed values.
ˆ
As shown in Equation (5-89) and Figure 5-17, the direction of the first substep N ( 1 ) is guessed
ˆ
( 0 ) , which is normal to the yield surface at the trial stress σ̂ . Then, the exact
from the direction N T
ˆ
direction N can be obtained from the first substep nonlinear solution based on Euler backward
(1)
ˆ
method. In general, the new direction for the second substep N ( i + 1 ) is guessed from the
ˆ
direction N ( i ) based on the previous substep stress σ̂ ( i ) . This procedure is completed when
F = F ( N ) ( = 0 ) . In fact, the equation for the Nth step is equivalent to Equation (5-89) and
σ̂ n + 1 = σ̂ ( N ) . Finally, the proportional logarithmic plastic strain remains normal to the yield
surface at the final stress σ̂ n + 1 ; that is,
∂σ ˆ
Δε̂ p = Δε p ------- ( σ̂ n + 1 ) ( = λ ( N ) N (N)) . (5-90)
∂σ̂
T
σ ( σ̂ )
σ ( σ̂ ( 1 ) )
ˆ
σ ( σ̂ ( 2 ) ) N( 1 )
ˆ
N( 2 ) σ̂ T
ˆ
N( N )
σ ( σ̂ ( n + 1 ) ) p
Δε̂ ( N )
σ ( σ̂ ( n ) )
Δσ̂ p
Δε̂
σ̂ n + 1 p (2)
Δε̂ ( 2 )
σ̂ n
Figure 5-17 Schematic View for Multistage Return Mapping Method

Therefore, the normality condition of the incremental deformation theory is satisfied at the current
state ( n + 1 ) for α = 1 .
In order to solve Equation (5-89), the Euler Backward method is employed to solve the iteration
(i)
procedure for the kth substep. At each iteration, Δλ (at ( k ) substep and ( i ) iteration) becomes
(k)
(i) (i) (i) (i) (i)

g1  λ  – N ˆ E – 1g λ  + g λ  H
(i)  ( k ) (k) (k) 2  ( k ) 3  ( k )
Δλ = -------------------------------------------------------------------------------------------------------------------- (5-91)
(k)
N ˆ (i) E (i) – 1N ˆ (i) + H
(k) (k) ( k)
where
(i) ( i ) ∂N̂ –1
E –1 = Cˆ e – 1 + λ ------- ,
(k) ( k ) ∂σ̂
(i) (i)
g 1  λ  = σ ( σ̂ ) – ρ  ε np + λ  – F ( k ) ,
 ( k )  ( k )
(i) (i) ˆ
g 2  λ  = Cˆ e – 1 ( σ̂ – σ̂ T ) + λ N ,
 ( k ) (k)
(i) (i)
g3  λ  = H ( ρ – ρn ) – λ ,
–1
 ( k ) (k)
and H is the hardening modulus in stress-strain curve.

The detailed derivations of Equation (5-91) are shown in the work of Yoon at al. [Ref. 32].
In order to solve the equilibrium equation iteratively, the elasto-plastic tangent modulus consistent
with the current return mapping method is obtained as follows:
dσˆ j = dσˆ n + 1 = L e p dεˆ n + 1 (5-92)
ˆ ˆ –1
 CN n + α ⊗ CN n + α ∂N̂ n + α
where Lep =  C – ------------------------------------------------
ˆ ˆ
- and C = Cˆ e – 1 + λ ------------------- .
 N n + α CN n + α + h′ ∂σ̂ n + 1
In Equation (5-92), h′ is instantaneous slope and α = 1 is used.
Creep
Creep is a time-dependent inelastic behavior that can occur at any stress level, either below or above the
yield stress of a material. Creep is an important factor at elevated temperatures. In many cases, creep is
also accompanied by plasticity, which occurs above the yield stress of the material.
CHAPTER 5 141
Marc offers two schemes for modeling creep in conjunction with plasticity:
a. treating creep strains and plastic strains separately using an explicit procedure (where the creep
is treated explicitly) or an implicit procedure (where both creep and plasticity are treated
implicitly). These procedures are available with standard options via data input or with user-
specified options via user subroutines. More details are provided below.
b. modeling creep strains and plastic strains in a unified fashion (viscoplasticity). Both explicit
and implicit procedures are again available for modeling unified viscoplasticity. More details
are provided in the section titled Viscoplasticity in this chapter.
The options offered by Marc for modeling creep are as follows:
• Creep data can be entered directly through data input or user subroutine. For explicit creep, the
CRPLAW user subroutine is to be used, and for implicit creep, the UCRPLW user subroutine is to
be used.
• An automatic time stepping scheme can be used to maximize the time step size in the analysis.
• Eigenvalues can be extracted for the estimation of creep buckling time.
In addition, for explicit creep, the following additional options can be used:
• Creep behavior can be either isotropic or anisotropic.
• The Oak Ridge National Laboratory (ORNL) rules on creep can be activated.
The creep analysis option is activated in Marc through the CREEP parameter. The creep time period and
control tolerance information are input through the AUTO CREEP history definition option. This option
can be used repeatedly to define a new creep time period and new tolerances. These tolerances are
defined in the section on Creep Control Tolerances. Alternatively, a fixed time step can be specified
through the CREEP INCREMENT history definition option. In this case, no additional tolerances are
checked for controlling the time step.
Creep analysis is often carried out in several runs using the RESTART option. Save restart files for
continued analysis. The REAUTO option allows you to reset the parameters defined in the AUTO CREEP
option upon restart.
Adaptive Time Control

The AUTO CREEP option takes advantage of the diffusive characteristics of most creep solutions.
Specifically, this option controls the transient creep analysis. You specify a period of creep time and a
suggested time increment. The program automatically selects the largest possible time increment that is
consistent with the tolerance set on stress and strain increments (see Creep Control Tolerances in
this chapter).
The algorithm is: for a given time step Δt , a solution is obtained. Marc then finds the largest values
cr el
of stress change per stress, Δσ ⁄ σ , and creep strain change per elastic strain, Δε ⁄ ε . It compares
these values to the tolerance values, T s (stress change tolerance) and T e (strain change tolerance), for
this period.
The value p is calculated as the larger of
( Δσ ⁄ σ ) ⁄ T σ
or (5-93)
cr
( Δε ⁄ εe l ) ⁄ Tε
If p > 1 , the program resets the time step as
Δt n e w = 0.8 Δt o l d ⁄ p (5-94)
The time increment is repeated until convergence is obtained or the maximum recycles control is
exceeded. In the latter case, the run is ended.
If the first repeat does not satisfy tolerances, the possible causes are:
• excessive residual load correction
• strong additional nonlinearities such as creep buckling-creep collapse
• incorrect coding in the CRPLAW, VSWELL, or UVSCPL user subroutine.
Appropriate action should be taken before the solution is restarted.
If p < 1 , the solution is stepped forward to t + Δt and the next step is begun. The time step used in the
next increment is chosen as
Δt n e w = Δt o l d if 0.8 < p < 1 (5-95)
Δt n e w = 1.25 Δt o l d if 0.65 < p< 0.8 (5-96)
Δt n e w = 1.5 Δt o l d if p < 0.65 (5-97)
Since the time increment is adjusted to satisfy the tolerances, it is impossible to predetermine the total
number of time increments for a given total creep time.
Creep Control Tolerances

Marc performs a creep analysis under constant load or displacement conditions on the basis of a set of
tolerances and controls you provide.These are as follows:
1. Stress change tolerance – This tolerance controls the allowable stress change per time step during
the creep solution, as a fraction of the total stress at a point. Stress change tolerance governs the
accuracy of the transient creep response. If you need accurate tracking of the transient response,
specify a tight tolerance of 1 percent or 2 percent stress change per time step. If you need only the
steady-state solution, supply a relatively loose tolerance of 10-20 percent. It is also possible to
check the absolute rather than the relative stress.
CHAPTER 5 143
2. Creep strain increment per elastic strain – Marc uses either explicit or implicit integration of the
creep rate equation. When the explicit procedure is used, the creep strain increment per elastic
strain is used to control stability. In almost all cases, the default of 50 percent represents the
stability limit, so that you need not provide any entry for this value. It is also possible to check
the absolute rather than the relative strain.
3. Maximum number of recycles for satisfaction of tolerances – During AUTO CREEP, Marc
chooses its own time step. In some cases, the program recycles to choose a time step that satisfies
tolerances, but recycling rarely occurs more than once per step. Excessive recycling can be caused
by physical problems such as creep buckling, poor coding of the CRPLAW, VSWELL, or UVSCPL
user subroutine or excessive residual load correction that can occur when the creep solution
begins from a state that is not in equilibrium.
The maximum number of recycles allows you to avoid wasting machine time under such
circumstances. If there is no satisfaction of tolerances after the attempts at stepping forward, the
program stops. The default of five recycles is conservative in most cases.
4. Low stress cut-off – Low stress cut-off avoids excessive iteration and small time steps caused by
tolerance checks that are based on small (round off) stress states. A simple example is a beam in
pure bending. The stress on the neutral axis is a very small roundoff-number, so that automatic
time stepping scheme should not base time step choices on tolerance satisfaction at such points.
The default of five percent of the maximum stress in the structure is satisfactory for most cases.
5. Choice of element for tolerance checking – Creep tolerance checking occurs as a default for all
integration points in all elements. You might wish to check tolerances in only 1 element or in up
to 14 elements of your choice. Usually, the most highly stressed element is chosen.
When you enter the tolerances and controls, the following conventions apply:
• All stress and strain measures in tolerance checks are second invariants of the deviatoric state
(that is, equivalent von Mises uniaxial values).
• You can reset all tolerances and controls upon the completion of one AUTO CREEP sequence.
Background Information
Creep behavior is based on a von Mises creep potential with isotropic behavior described by the
equivalent creep law:
· cr
ε = f ( σ, ε c r, T, t ) (5-98)
The material behavior is therefore described by:
· c r ∂σ
Δε c r = ε --------- (5-99)
∂σ d
∂σ
where --------- is the outward normal to the current von Mises stress surface and ε· c r is the equivalent creep
∂σ d
strain rate.
There are two numerical procedures used in implementing creep behavior. The default is an explicit
procedure in which the above relationship is implemented in the program by an initial strain technique.
In other words, a pseudo-load vector due to the creep strain increment is added to the right-hand side of
the stiffness equation.
KΔu = ΔP +  β T DΔε c r dv (5-100)

V
where K is the stiffness matrix, and Δu and ΔP are incremental displacement and incremental nodal
force vectors, respectively. The integral
 β T DΔε c r dv (5-101)
V
is the pseudo-load vector due to the creep strain increment in which β is the strain displacement
relation and D is the stress-strain relation. When plasticity is also specified through a suitably
defined yield criterion and yield stress in Marc, the plasticity is treated implicitly while the creep is
treated explicitly.
As an alternative, an implicit creep procedure can be requested with the CREEP parameter. In this case,
the inelastic strain rate has an influence on the stiffness matrix. Using this technique, significantly larger
steps in strain space can be used. In Marc, this option is only to be used for isotropic materials with the
creep strain rate defined by
· cr n cr
ε = Aσ f ( ε )g ( T )h ( t )
where A is the creep constant that can be defined through input data or through the UCRPLW user
subroutine; power law expression is always to be used for the effective stress with the coefficient
provided through the input data or the UCRPLW user subroutine; and power law coefficients or more
general expressions can be provided for the creep strain, temperature, and time through the input data or
the UCRPLW user subroutine, respectively.
When plasticity is also specified through a suitably defined yield stress in conjunction with the von Mises
yield criterion, a sub-iterative scheme within each Newton-Raphson cycle is used to determine the plastic
strains needed to keep the stress state on the yield surface and the creep strains that develop due to the
equivalent stress being greater than a user-defined back stress. The yield stress for the plastic component
can be varied as a function of the equivalent plastic strain, temperature and spatial coordinates. Similarly,
the back stress for the creep component can be varied as a function of the equivalent creep strain,
temperature and spatial coordinates.
Creep Buckling
Marc also predicts the creep time to buckling due to stress redistribution under given load or repeated
cyclic load.
The buckling option solves the following equation for the eigenvalue
( K + λK G )Φ = 0 (5-102)
CHAPTER 5 145
The geometric stiffness matrix, K G , is a function of the increments of stress and displacement. These
increments are calculated during the last creep time step increment. To determine the creep time to
buckle, perform a buckle step after a converged creep increment. Note that the incremental time must be
scaled by the calculated eigenvalue, and added to the total (current) time to get an estimate as to when
buckling occurs.
AUTO THERM CREEP (Automatic Thermally Loaded

Elastic-Creep/Elastic-Plastic-Creep Stress Analysis)
The AUTO THERM CREEP option is intended to allow automatic, thermally loaded elastic-creep/elastic-
plastic-creep stress analysis, based on a set of temperatures defined throughout the mesh as a function of
time. The temperatures and transient times are presented to the program through the CHANGE STATE
option, using input option 3 (post file), and the program creates its own set of temperature steps
(increments) based on a temperature change tolerance provided in this option. The times at all
temperature steps are calculated by the program for creep analyses.
At each temperature step (increment), an elastic/elastic-plastic analysis is carried out first to establish
stress levels in the structure. A creep analysis is performed next on the structure for the time period
between current and previous temperature steps (increments). Both the elastic/elastic-plastic stress and
the creep analyses are repeated until the total creep time provided in this option is reached. Convergence
controls are provided on the CONTROL option for elastic-plastic analysis and in the AUTO THERM
CREEP option for creep analysis. The analysis can be restarted at temperature steps (increments) or at
creep steps (subincrements). The results can be saved on a post file (POST option) for postprocessing.
If no DIST LOADS, POINT LOAD, or PROPORTIONAL INCREMENT option appears with the AUTO
THERM CREEP set, all mechanical loads and kinematic boundary conditions are held constant during
the AUTO THERM CREEP. However, DIST LOADS, POINT LOAD, PROPORTIONAL INCREMENT, or
DISP CHANGE can be included in the set – the mechanical loads and kinematic boundary conditions
which are then defined are assumed to change in proportion to the time scale of the temperature history
defined by the CHANGE STATE option and are applied accordingly. This is based on the fact that the
increments of load and displacement correspond to the end of the transient time of the AUTO THERM
CREEP input.
Viscoelasticity
In a certain class of problems, structural materials exhibit viscoelastic behavior. Two examples of these
problems are quenching of glass structures and time-dependent deformation of polymeric materials. The
viscoelastic material retains linearity between load and deformation; however, this linear relationship
depends on time. Consequently, the current state of deformation must be determined from the entire
history of loading. Different models consisting of elastic elements (spring) and viscous elements
(dashpot) can be used to simulate the viscoelastic material behavior described in Chapter 7. A special
class of temperature dependence known as the Thermo-Rheologically Simple behavior (TRS) is also
applicable to a variety of thermal viscoelastic problems. Both the equation of state and the hereditary
integral approaches can be used for viscoelastic analysis.
To model the thermo-rheologically simple material behavior, the SHIFT FUNCTION model definition
option can be used to choose the Williams-Landel-Ferry equation or the power series expression or
Narayanaswamy model.
In Marc, two options are available for small strain viscoelastic analysis. The first option uses the equation
of state approach and represents a Kelvin model. The second option is based on the hereditary integral
approach and allows the selection of a generalized Maxwell model. The thermo-rheologically simple
behavior is also available in the second option for thermal viscoelastic analysis. The Time-independent
Inelastic Behavior section in Chapter 7 discusses these models in detail. Automatic time stepping
schemes AUTO CREEP and AUTO STEP can be used in a viscoelastic analysis for first and second
options, respectively.
The first option for viscoelastic analysis uses the Kelvin model. To activate the generalized Kelvin
model in Marc, use the VISCO ELAS or CREEP parameter. To input the matrices [A] and [B] for the
Kelvin strain rate computations, use the CRPVIS user subroutine. To input creep time period
and the tolerance control for the maximum strain in an increment, use the AUTO CREEP history
definition option.
The Simo model for large strain viscoelasticity can be used in conjunction with the damage and
hyperelastic Mooney or Ogden material model. The large strain viscoelastic material behavior can be
simulated by incorporating the VISCELMOON, VISCELOGDEN, or VISCELFOAM model definition
option. Viscoelasticity for Mooney and Ogden materials is available in both the total and updated
Lagrangian framework. Viscoelasticity for foam materials is available only in the updated Lagrangian
framework.
Nonlinear structural relaxation behavior of materials can be modeled by the Narayanaswamy model
which accounts for memory effect. This model allows simulation of evolution of physical properties of
glass subjected to complex time temperature histories. The thermal expansion behavior for the
Narayanaswamy model is controlled via the VISCEL EXP model definition option.
Viscoplasticity
There are two procedures in Marc for viscoplastic analysis: explicit and implicit. A brief description of
each procedure follows:
Explicit Method
The elasto-viscoplasticity model in Marc is a modified creep model to which a plastic element is added.
The plastic element is inactive when the stress is less than the yield stress of the material. You can use
the elasto-viscoplasticity model to solve time-dependent plasticity and creep as well as plasticity
problems with a nonassociated flow law.
The CREEP option in Marc has been modified to enable solving problems with viscoplasticity. The
method is modified to allow solving elastic-plastic problems with nonassociated flow rules which result
in nonsymmetric stress-strain relations if the tangent modulus method is used.
The requirements for solving the viscoplastic problem are:
• CREEP parameter and creep controls
• Load incrementation immediately followed by a series of creep increments specified by
AUTO CREEP
CHAPTER 5 147
• Use of the CRPLAW user subroutine and/or the NASSOC user subroutine
The following load incrementation procedure enables a viscoplastic problem to be solved:
1. Apply an elastic load increment that exceeds the steady-state yield stress.
2. Relieve the high yield stresses by turning on the AUTO CREEP option.
You may repeat steps 1 and 2 as many times as necessary to achieve the required load history.
Note: The size of the load increments are not altered during the AUTO CREEP process
so that further load increments can be effected by using the PROPORTIONAL INCREMENT
option.
The viscoplastic approach converts an iterative elastic-plastic method to one where a fraction of the
initial force vector is applied at each increment with the time step controls. The success of the method
depends on the proper use of the automatic creep time step controls. This means that it is necessary to
select an initial time step that will satisfy the tolerances placed on the allowable stress change.
allowable stress change x 0.7
The initial time step Δt =
Maximum viscoplastic strain rate x Young’s modulus
The allowable stress change is specified in the creep controls. The most highly stressed element usually
yields the maximum strain rate. It is also important to select a total time that gives sufficient number of
increments to work off the effects of the initial force vector. A total time of 30 times the estimated Δt is
usually sufficient.
Marc does not distinguish between viscoplastic and creep strains. A NASSOC user subroutine allows you
to specify a nonassociated flow rule for use with the equivalent creep strains (viscoplastic) that are
calculated by the CRPLAW user subroutine. A flag is set in the CREEP parameter in order to use the
viscoplastic option with a nonassociated flow rule.
The viscoplasticity feature can be used to implement very general constitutive relations with the aid of
the ZERO and YIEL user subroutines.
Since the viscoplasticity model in Marc is a modified creep model, you should familiarize yourself with
the creep analysis procedure (see Nonlinear Analysis at the beginning of this chapter).
Implicit Method
A general unified viscoplastic material law can be implemented through the UVSCPL. user subroutine
When using this method, you are responsible for defining the inelastic strain increment and the
current stress.
Fracture Mechanics
The fracture mechanics capabilities in Marc covers the evaluation of energy release rate and J-integral
including automatic crack propagation. Two methods are offered for the evaluation: the mode separation
method through the LORENZI option and the Virtual Crack Closure Technique (VCCT) with the VCCT
option. The VCCT option also supports automatic crack propagation.
Linear Fracture Mechanics

Linear fracture mechanics presupposes existence of a crack and examines the conditions under which
crack growth occurs. In particular, it determines the length at which a crack propagates rapidly for
specified load and boundary conditions. The concept of linear fracture mechanics stems from Griffith’s
work on purely brittle materials. Griffith stated that, for crack propagation, the rate of elastic energy
release should at least equal the rate of energy needed for creation of a new crack surface. This concept
was extended by Irwin to include limited amounts of ductility. In Irwin’s considerations, the inelastic
deformations are confined to a very small zone near the tip of a crack.
The basic concept presented by Griffith and Irwin is an energy balance between the strain energy in the
structure and the work needed to create a new crack surface. This energy balance can be expressed using
the energy release rate G as
G = Gc (5-103)
G is defined as
dΠ
G = – -------- (5-104)
da
where Π is the strain energy and a is the crack length. G depends on the geometry of the structure and
the current loading. G c is called the fracture toughness of the material. It is a material property which is
determined from experiments. Note that the energy release rate is not a time derivative but a rate of
change in potential energy with crack length. An important feature of Equation (5-103) is that it can be
used as a fracture criterion; a crack starts to grow when G reaches the critical value G c .
The stress and strain fields near the tip of a crack are singular for a linear elastic material model. The
stresses and strains have the principal form
K
σ = ------ f ( θ )
r
(5-105)
K
ε = ------ g ( θ )
r
in a polar coordinate system centered at the crack tip. Thus, a linear elastic material is said to have a
1 ⁄ r singularity near a crack tip. It is easy to demonstrate that in both Griffith’s and Irwin’s
considerations, the elastic energy release rate is determined by a single parameter: the strength of the
singularity in the elastic stress field at the crack tip. This is the so-called stress intensity factor, and is
usually denoted by capital K . The magnitude of K depends on the crack length, the distribution and
intensity of applied loads, and the geometry of the structure. Crack propagation occurs when any
combination of these factors causes a stress intensity factor K to be equal to or greater than the
experimentally determined material property K c , which is equivalent to Equation (5-103). Hence, the
objective of linear fracture mechanics calculations is to determine the value of K .
CHAPTER 5 149
There are three possible modes of crack extension in linear elastic fracture mechanics: the opening mode,
sliding mode, and tearing mode (see Figure 5-18).
y
y
x
x
z
z
(a) Mode I: Opening (b) Mode II: Sliding

y
x
z
(c) Mode III: Tearing

Figure 5-18 Irwin’s Three Basic Modes of Crack Extension
The opening mode (see Figure 5-18a), Mode I, is characterized by the symmetric separation of the crack
surfaces with respect to the plane, prior to extension (symmetric with respect to the X-Y and X-Z planes).
The sliding mode, Mode II, is characterized by displacements in which the crack surfaces slide over one
another perpendicular to the leading edge of the crack (symmetric with respect to the X-Y plane and
skew-symmetric with respect to the X-Z plane).
The tearing mode, Mode III, finds the crack surfaces sliding with respect to one another parallel to the
leading edge (skew-symmetric with respect to the X-Y and X-Z planes).
It is customary to associate a stress intensity factor with each of these mode: K I , K I I , and K I I I . There
is also an associated fracture toughness associated with each mode: K I c , K I I c , and K I I I c . The most
critical mode is usually mode I and in many cases the other modes are not considered. The connection
between the energy release rate and the stress intensity factors is given by
2 2
KI KI I 1 + ν 2
G = ------- + -------- + ------------- K I I I (5-106)
E' E' E
where
E' = E for plane stress (5-107)

and
E
E' = --------------- for plane strain. (5-108)
1 – ν2
Marc uses the so-called J-integral for evaluating the energy release rate, see below. The J-integral is
similar to G but is more general and is also used for nonlinear applications. J is equivalent to G when a
linear elastic material model is used.
Nonlinear Fracture Mechanics

Nonlinear fracture mechanics is concerned with determining under which conditions crack propagation
(growth) occurs. In this sense, nonlinear fracture mechanics is similar to linear fracture mechanics.
However, there are additional questions addressed in nonlinear fracture mechanics. Is the crack
propagation stable or unstable? If it is stable, at which speed does it occur? After some propagation, is
the crack arrested?
There also exists a singularity at the crack tip in fracture mechanics problems with nonlinear elastic-
plastic material behavior, though the singularity is of a different nature. If one takes an exponential
hardening law of the form
σ ε 1/n
------ =  ----- (5-109)
σ0  ε 0
it can be shown that the singularities in the strain and the stresses at the crack tip are of the form
ε = f ( θ )r – n ⁄ ( n + 1 )
(5-110)
σ = g ( θ )r – 1 ⁄ ( n + 1 )
If n approaches infinity, the material behavior becomes perfectly plastic and the singularity in the stresses
vanishes. The singularity in the strains, however, takes the form of
ε = f ( θ )r – 1 (5-111)
It has not been possible to establish that the strength of the singularity is the only factor that influences
initiation of crack propagation for nonlinear situations. In fact, it is doubted that initiation of crack
propagation is dependent on only a single factor. The J-integral probably offers the best chance to have
a single parameter to relate to the initiation of crack propagation.
CHAPTER 5 151
The J-integral was introduced by Rice as a path-independent contour integral for the analysis of cracks.
As previously mentioned, it is equivalent to the energy release rate for a linear elastic material model. It
is defined in two dimensions as
∂u j
 
J =   ( W + T )n 1 – σ i j n i -------
∂x 1
- dΓ (5-112)
Γ
where W is the strain energy density, T is the kinetic energy density, σ i j is the stress tensor and u i is
the displacement vector. The x1 direction is the same as the x direction in the local crack tip system in
Figure 5-19. The integration path Γ is a curve surrounding the crack tip, see Figure 5-19.The J-integral
is independent of the path Γ as long as it starts and ends at the two sides of the crack face and no other
singularities are present within the path. This is an important feature for the numerical evaluation since
the integral can be evaluated using results away from the crack tip.
ni
y
x
Γ
Figure 5-19 Definition of the J-integral
Numerical Evaluation of the J-integral

The J-integral evaluation in Marc is based upon the domain integration method as described in [Ref. 30].
It is available as the LORENZI model definition option. A direct evaluation of Equation (5-112) is not
very practical in a finite element analysis due to the difficulties in defining the integration path Γ . In the
domain integration method for two dimensions, the line integral is converted into an area integration over
the area inside the path Γ . This conversion is exact for the linear elastic case and also for the nonlinear
case if the loading is proportional, that is, if no unloading occurs. By choosing this area as a set of
elements, the integration is straightforward using the finite element solution. In two dimensions, the
converted expression is
∂u j δq 1
 
J =   σ i j -------
∂x 1
- – Wδ 1 i -------- dA
 δx i
(5-113)
A
for the simplified case of no thermal strains, body forces or pressure on the crack faces. For the general
expression, see [Ref. 1]. A is the area inside Γ and q 1 is a function introduced in the conversion into an
area integral. The function q 1 can be chosen fairly generally, as long it is equal to one at the crack tip
and zero on Γ . The form of the function chosen in Marc is that it has the constant value of one at all nodes
inside Γ , and decreases to zero over the outermost ring of elements in A . It can be interpreted as a rigid
translation of the nodes inside Γ while the nodes on Γ remain fixed. Thus, the contribution to Equation
(5-113) comes only from the elements in a ring away from the crack tip. This interpretation is that of
virtual crack extension and this method can be seen as a variant of such a technique, although it is
extended with the effects of thermal strains, body forces, and pressure on the crack faces.The set of nodes
moved rigidly is referred to as the rigid region and the function q 1 as the shift function or shift vector.
For the evaluation of the J-integral the direction of the shift vector is simply the x axis in the local crack
tip system.
In three dimensions, the line integral becomes an area integral where the area is surrounding a part of the
crack front. In this case, the selection of the area is even more cumbersome than in two dimension. The
converted integral becomes a volume integral which is evaluated over a set of elements. The rigid region
is a set of nodes which contains a part of the crack front, and the contribution to the integral comes from
the elements which have at least one but not all its nodes in the rigid region.
Determination of the Rigid Region and the Shift Vector

For the evaluation of the J-integral, Marc requires the nodes along the crack front, the shift vector, and
the nodes of the rigid region. Marc allows three ways of defining the rigid region. The first is direct input;
the nodes or elements in the rigid region are listed explicitly. With this variant, the shift vector is also
specified directly. The second and third variants use an automatic search for the nodes of the rigid region.
The second variant is based on the mesh topology (connectivity) where a number of regions of increasing
size are found by Marc. The first region for two dimensions consists of the nodes of all elements
connected to the crack tip node. The second region consists of all nodes in the first region and the nodes
of all elements connected to any node in the first region and so on for a given number of regions. This
way, contours of increasing size are determined. In the third way of determining the rigid region, a radius
is given and all nodes within that radius are part of the rigid region. For the automatic search methods the
shift vector can also be determined automatically. It is then determined using the first element edge on
the crack face (for meshes with notches see below). Symmetry at the crack face is automatically detected
by Marc, also for the case that the symmetry condition is applied by means of a rigid contact body.
In three dimensions, it is a bit more complicated. The crack front is defined by an unsorted list of nodes.
In order to obtain the variation of the J-integral along the crack front, a disk of nodes with the normal of
the disk directed along the tangent to the crack front can be determined. This type of disk can readily be
defined if the mesh around the crack front consists of brick elements in a regular mesh. This mesh is
typically created from a two-dimensional mesh which is extruded along the crack front. The topology
based determination of the rigid region assumes such a mesh and creates disks of increasing size at each
crack front node from element faces. The shift vector at each crack front node is automatically
determined to be perpendicular to both the tangent to the crack front and the normal to the crack face at
each crack front node. At the first and last crack front node, where a free surface is assumed to exist, the
shift direction is projected to the tangent to the free surface. This is important since the shift must not
CHAPTER 5 153
change the outer boundary of the model. The geometry based search method here works with a cylinder
with the axis aligned with the tangent of the crack front and centered at the crack front node. The length
of the cylinder is given as a fraction of the distance to the neighboring crack front nodes. All nodes within
the cylinder are part of the rigid region. This method is useful if, for instance, the mesh around the crack
front is created with an automatic mesh generator. The shift vector is determined in the same way as for
the topology based search method.
In elasto-plastic analyses, it is often advantageous to use a mesh where there are several (multiple) nodes
at the crack tip. If collapsed elements are used at the crack tip, the nodes are kept separate and a notch is
formed as the crack tip deforms. For this kind of mesh, a “multiple nodes” distance is given, which
should be smaller than the smallest element at the crack tip. All nodes within that distance are then
considered part of the crack tip, and the first contour for the topology based search will consist of all
elements connected to any of these nodes (in three-dimensions of element faces connected to any of these
nodes). The normal to the crack face used for determining the shift vector is taken as the first face outside
the crack tip nodes. Thus, it is also possible to model the crack with an initial notch.
Mode Separation
For the linear elastic case, Marc automatically calculates and prints the stress intensity factors KI , KI I ,
and K I I I for the three modes I, II, and III. This evaluation is done using analytical functions for the stress
field near the crack tip. The implementation is based upon the procedure outlined in [Ref. 33]. Mode
separation is not calculated when nonlinearities like large deformations, nonlinear material, or general
contact are present. However, the use of glued contact is allowed to connect different parts of the mesh
near the crack front. The material is also not allowed to be temperature dependent. Orthotropic materials
are allowed. The calculation of the stress intensity factors is automatically suppressed if unsupported
features are used.
The stress intensity factors are defined in the local crack tip system, see Figure 5-19. K I will be positive
when the crack is opening and negative when it is overlapping. Similarly, K I I is positive when the shear
stress is positive ahead of the crack in the local system.
Supported Features
The LORENZI J-integral option supports the following features:
Linear and nonlinear materials.
Large deformations. Both total and updated Lagrangian formulation are supported. The crack tip
system can undergo finite rotations and translations.
Thermal loads.
Transient dynamics.
Contact between crack faces including friction.
Loads on the crack faces.
Mode separation for linear elasticity and no loads in the crack region.
Numerical Evaluation of the Energy Release Rate with the

VCCT Method
The VCCT option offers a simpler but more general way for obtaining the energy release rate. The
implementation follows the description in R. Krueger [Ref. 34]. Consider again Equation (5-104).
dΠ
G = – -------- (5-114)
da
It states that G is the change in potential energy by a change in crack length. Now, consider the simple
finite element model in Figure 5-20. The models A and B are the same, except that in B, the crack has
grown by one element edge of length a. Suppose we do one analysis for each one of the two models. We
can now calculate the energy release rate G as:
Fu
G = ------ (5-115)
2a
a
u
Model A Model B
Figure 5-20 Mesh for Illustrating the CCT Method
Here, F is the force (obtained from Model A) that keeps the crack together, and the crack opening, u , is
obtained from Model B. In order to obtain these quantities, we would need to perform two analyses and
this method is often referred to as CCT (Crack Closure Technique). In the virtual crack closure technique,
we only do the analysis with a closed crack (Model A) and use the opening displacement at the closest
nodes to the crack tip. Figure 5-21 shows the case of pure mode I. The other modes are treated similarly
and separately. The displacements and reactions are transformed into the local crack tip system for
this evaluation.
CHAPTER 5 155
F
u
Figure 5-21 The VCCT Method
With x. y, and z denoting the coordinate directions in the local crack tip system (see Figure 5-18),
we obtain:
F y uy Fx ux Fz uz
G I = ------------ G I I = ------------ G I I I = ----------- (5-116)
2a 2a 2a
and the total energy release rate as:
Gt o t = GI + GI I + GI I I (5-117)
For higher-order elements, we need to include the contributions from the midside nodes (see
Figure 5-22).
F 1 u1 + F2 u2
G I = ------------------------------- (5-118)
2a
a
F1
F2
u1
u2
Figure 5-22 Higher-order Elements

For the case that the midside nodes are not at the middle of the element edges (for example, using the 1/4
point position for increased accuracy, the displacements for these nodes are interpolated to the
appropriate locations.
For 3-D solids, we have a situation as shown in Figure 5-23. The situation is similar to the 2-D case, and
the evaluation is done separately for each node along the crack front. The area is given by the shaded part
in Figure 5-23.
For the case of higher-order elements, we obtain the following by using the notation in Figure 5-24.
1
F 1 u 1 + F 2 u 2 + --- ( F 3 u 3 + F 4 u 4 )
2
G = ------------------------------------------------------------------------------- (5-119)
2a
u is Crack Opening
Displacement At Crack Face
Crack Front
Figure 5-23 3-D Mesh for VCCT

CHAPTER 5 157
u1 u3
u4
F1 F3
u2
F4 F2
Figure 5-24 3-D Mesh for VCCT for Higher-order Elements
The above figures show a regular mesh of hexahedral elements. While a mesh designed like this is
advantageous for accuracy, it is not strictly necessary. It is also possible to use a general tetrahedral mesh
or a hexahedral mesh with the crack front as defined in Figure 5-25. The latter is typically obtained in
the case of crack propagation as described below. The program will find the appropriate nodes to use for
the forces and crack opening displacement and calculate an area a. In the current release, the case of
higher order tetrahedral elements is not supported.
Figure 5-25 Example of Crack Front Not Following Element Edges

The definition of the data involved in the VCCT calculation is done automatically. The user only specifies
the crack tip or crack front. In order to check what the program finds there is debug option available. If
the PRINT parameter is used with a value of 43, the program will print out the nodes it finds, the
corresponding forces, crack opening displacements, and crack areas.
Symmetry is automatically detected and accounted for. The symmetry condition can be enforced by
boundary conditions or by rigid contact.
The mesh in the crack region can contain user tyings or contact. The program automatically detects if the
crack tip node is connected to a node, and the node it is connected to is considered part of the crack tip.
The supported connections are glued contact, user typing, RBE2 and RROD. For the case that the
connection is done node-to-node, it does not matter if the tied or retained node of the tying is selected as
the crack node. With glued contact, it is not necessary to connect the parts node-to-node. If the meshes
do not match up, the crack will be treated similarly as a case of symmetry. This non-matching glued case
is automatically detected and for this case, it is necessary to select the touching node as the crack tip node.
A useful option when defining a crack using glued contact is DEACT GLUE. Suppose two parts are glued
together and the crack is defined as part of the interface between the bodies. One can then use DEACT
GLUE to identify the nodes that should be on the crack faces (see Figure 5-26). These nodes will have
regular contact but not be glued.
Crack Tip
Glued Contact
Nodes Identified
with DEACT GLUE
Contact Body 1
Contact Body 2
Figure 5-26 Using DEACT GLUE for Defining a Crack
The calculation of VCCT is done in the current geometry in case of large deformations. The updated
crack coordinate system is calculated at each increment and the calculations are done in this system.
Thus, arbitrary rotations and deformations are allowed. The current crack tip system is available on the
post file. The three coordinate axes can be plotted as vectors for verification.
The program calculates an estimated crack growth direction for each crack front node. Four methods for
defining this are supported: the maximum principal stress criterion ([Ref. 35]), along the pure mode with
largest G i – G i c (where i is mode I, II or III), along mode I and along a specified vector. The maximum
principal stress criterion states that a crack will grow in the direction normal to the direction of greatest
CHAPTER 5 159
tension. The crack growth direction normal to the direction of greatest tension. The crack growth
direction ( d x ,d y ) in the local crack tip system can be expressed in terms of the stress intensity factors
in mode I and II as:
2 2 2
KI I K II KI I KI I KI I
3 -------- + 1 + 8 -------- 3 -------- – 3 -------- 1 + 8 --------
2 2 KI KI 2
KI KI KI
dx = ---------------------------------------------- dy = ---------------------------------------------------------- (5-120)
2 2
KI I KI I
1 + 9 -------- 1 + 9 --------
2 2
KI KI
For the 3-D case, mode III is assumed to have the same effect as mode I so K I + K I I I is used instead of
K I in the above equation. It is, thus, assumed that crack growth occurs perpendicular to the crack front
tangent. The user can also specify the crack growth direction through the user subroutine
UCRACKGROW.
Figure 5-27 shows some examples of supported crack configurations. There is a large flexibility in how
parts can be tied or glued together to form a crack.
Crack Tip
Crack Front
Line Crack in Shell Shell Glued to Shell 3-D Solid
Shell Glued to Solid (Line Crack) Shell Glued to Solid (Surface Crack)
Figure 5-27 Examples of Supported Crack Types
Remeshing is supported for 2-D solids and shells. A structure containing a crack can be remeshed and
the new crack tip is automatically found after remeshing. The crack opening displacement is saved
before the remesh, and applied afterwards to make sure that the results are also correct after remeshing
takes place. Three-dimensional solids with cracks cannot currently be remeshed. Remeshing can also be
used for automatic crack propagation as described below.
Automatic Crack Propagation

The VCCT option allows automatic crack propagation to be performed. There are two types of crack
propagation analysis types available: fatigue growth and direct growth. There are three methods
supported for how the actual crack growth is performed: growth during remeshing, growth by releasing
tyings or glued contact, and growth along element edges. These options are described in the following.
Fatigue Crack Growth

Here, the user specifies a fatigue time period. The analysis is performed without crack growth during this
period, which can be repeated any number of times. During the time period, it records the maximum and
minimum energy release rate and the estimated crack growth direction at the maximum. At the end of
each fatigue time period, the crack growth is applied. For the growth, we need to know the distance and
direction the crack should grow. The distance is either user specified (possibly scaled by a table) or
obtained by Paris’ law. With Paris’ law we use the following expression:
a = C [ ( Gm a x – Gm i n ) m – Gt hm ] (5-121)
This formula is from T. L. Anderson [Ref. 31], modified to use G instead of the stress intensity K.
No growth occurs if G m a x – G m i n < G t h or a < a m i n . The ucrack_paris.f user subroutine can
be used for specifying the growth increment. The crack growth increment is first calculated with
Equation (5-121) and then the user subroutine is called. See UCRACKGROW in Volume D: User
Subroutine, Special Routines, and Utility Routines for details.
If growth by remeshing is used, the specified or calculated crack growth increment and estimated crack
growth direction are used for defining the location of the new crack tip. When release of glued contact
or tyings are used, then as many element edges are released as needed for growing the crack according
to the specified or calculated growth increment. If there are multiple choices available for which element
edge to release, then the one closest to the crack growth direction is chosen.
Direct Crack Growth

For this case, a standard transient analysis is performed. The user specifies a crack growth resistance
(fracture toughness) for each crack. When the crack growth criterion is fulfilled, a propagation is done.
This takes place within an increment; the increment iterates until no further crack growth occurs. For
unstable crack growth, this may have the effect that the crack grows through the whole structure within
one increment. This ensures that an increment does not show convergence when a crack should grow
which would violate the crack growth criterion.
With remeshing based growth, the user specifies the crack growth increment to use and the current
estimated crack growth direction is used. Repeated remeshing may be used in an increment if more
growth is triggered. For the growth by releasing tyings or glued contact, one element edge at the time is
released until no more growth occurs. For both methods, the growth only occurs when the increment is
otherwise converged.
CHAPTER 5 161
Remeshing Based Growth

Crack growth by means of remeshing is available for 2-D solid elements. The model must be set up to
do global remeshing, but no specific remeshing criterion needs to be set for the crack propagation. Any
specified remeshing criterion will be treated independently from the remeshings due to crack growth. It
is, for example, allowed to do remeshings during the fatigue load cycle.
When a remeshing occurs in a 2-D analysis, the program first finds the outline of the remeshing body.
When a crack is growing, this outline is extended to form a new crack tip location. In order to assure that
we obtain a mesh with an element edge in the x direction of the local crack tip system, we extend the
outline a little further than to the new crack tip. The new mesh is modified such that the new mesh is
correct. This way we avoid meshes as shown in Figure 5-28a.
If a crack reaches the boundary of the model, the remeshing body automatically splits. Thus, a crack can
grow into another crack, an internal hole of the body, or the external boundary. Due to the outline
extension mentioned above, the crack will reach the boundary slightly earlier than one would
otherwise expect.
The setting of target element length after remeshing will be overridden for the case that it is larger than
the growth increment. This ensures that the mesh resolution is large enough to model the new crack tip.
The mesh around the crack tip is automatically refined when growth occurs. An example of a remeshed
crack is shown in Figure 5-28b.
a) poor mesh which is avoided during b) refined mesh at current crack tip
crack growth
Figure 5-28 Meshes after Remeshing.
Growth by Releasing Tyings or Glued Contact

For this option, the user needs to specify the crack growth path. The crack can only grow in the direction
where there are tyings or glued contact to release. All crack configurations are supported: 2-D, shells
and 3-D.
For the case of user tyings, there are three types of tyings supported: user tying type 100, RBE2 and
RROD. Here, the nodes need to be aligned on both sides of the crack faces. The user specifies only the
crack tip node; the program automatically finds the tyings to release. It does not matter if the tied or
retained node is selected as the crack tip node or crack front nodes.
For glued contact, the meshes do not need to match up. As mentioned in the previous section, a special
non-matching glue symmetry case is automatically used if the nodes do not match up. For this case,
the tied nodes must be selected as the crack tip nodes. When nodes in glued contact are released due
to crack growth, they switch from glued contact to regular contact. The DEACT GLUE option is here
used internally. Thus, a node which is released can still contact the new crack face, but it will not glue
back again.
If higher-order elements are used, the midside node will be released when the corresponding corner node
is released.
If a crack reaches a model boundary, the crack evaluation for this crack will be turned off and the analysis
can continue. The program avoids that a single node is tied at a boundary; if the last element edge is
released, then the last node will also be released.
If the mesh is designed such that there are multiple choices for which element edge to release, then the
one closest to the estimated crack growth direction is chosen.
For the 3-D case, it is assumed that the area where the crack grows has a regular mesh. The crack should
not grow into areas of irregular mesh. See Figure 5-29 for an illustration. If the crack reaches the point
indicated at the top part of the figure, the crack front would have to be extended, which is not supported
at this point. Within the regular growth zone the crack can grow irregularly of course.
Crack Front Branch
Original Crack Front Largest Growth Zone

Figure 5-29 Possible Growth Zone for 3-D Crack Propagation
CHAPTER 5 163
Growth Along Element Edges

For this option, the crack can grow along element edges for 2-D and shell elements. The element edge
closest to the crack growth direction is used. New nodes are automatically inserted and the element
connectivity is changed for elements around the crack tip in order to grow the crack. The new nodes
inherit the properties of the respective original node.
Crack Growth Criteria and Direction
There are currently four different crack growth criteria available:
Total G. The crack will grow if G > G c , where G c is the crack growth resistance (also called
fracture toughness).
Separately for each mode. Crack growth will take place if either of the following is true: G I > G Ic ,
G I I > G I I c or G I I I > G I I I c .
GI n1 G I I n2 G I II n 3
Mixed mode criterion 1  -------- +  ---------- +  ------------- > 1
 G I c  G I I c  G II I c
Mixed mode criterion 1

G I + G I I + G II I
---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- >1
GI I + GI I I n1 G I I I  G I I + G I II  n 1

G I c + ( G I Ic – G I c ) ---------------------------------------  + ( G I I I c – G I I c ) -------------------------  ---------------------------------------
 G I + G I I + G I I I GI I + GI I I  G + G + G 
I II II I
Calculation of Crack Growth Direction

Four methods are available for the calculation of the crack growth direction are the same as
described above.
Dynamic Fracture Methodology

In complete similarity with static fracture mechanics concepts, it is assumed that dynamic crack growth
processes for linear materials are governed by the following condition:
K I ( t ) = R I D ( a·, T, B ) a· ≥ 0 (5-122)
where K I ( t ) is the dynamic stress intensity factor for mode I, a· is the crack velocity, and R I D is the
dynamic crack propagation toughness, which is assumed to be a material parameter that in general
depends on crack velocity a· , temperature T , and specimen thickness B .
The dynamic stress intensity factor depends on crack length ( a ), applied loading ( σ ), time ( t ), specimen
dimensions ( D ), temperature ( T ), and initial stress fields ( σ i ) caused by residual stresses or by an initial
strain field. The prediction of the crack propagation history and crack arrest event demands complete
knowledge of the R I D vs. a· relation. The Dynamic Fracture Methodology procedure consists of the
following two phases:
1. Generation phase – in this phase, a crack arrest experiment is performed yielding a crack
propagation-versus-time curve. In addition, a numerical simulation of the experiment is carried
out by using the measured crack propagation curve. This is used as input for the numerical model.
This allows the calculation of dynamic stress intensity factors as a function of time. Combination
of the latter relation with the measured crack propagation curve results in a curve, which can be
considered the dynamic crack propagation toughness-versus-crack velocity relation.
2. Application phase – in order to predict the crack growth and possible crack arrest point in a
structural component, the inverse problem is solved. Now, the actual stress intensity factors are
calculated for the structural component, that is subjected to a particular loading history, by means
of a dynamic finite element analysis. These calculated values are compared to the fracture
toughness curve obtained during the generation phase, Equation (5-122), and from this the crack
growth is predicted.
Modeling Considerations
The main difficulty in the finite element analysis of linear elastic fracture mechanics is representing the
solution near the crack front. Typically, a focused mesh is used with the elements in a spider web shape.
This gives a mesh with rings of elements in a radial fashion and is particularly useful for studying the
variation of the values of J with contour radius. If higher order elements are used for an elastic analysis,
it is advantageous to use so-called quarter point elements at the crack tip. These are regular elements
where the midside nodes at element sides for which one node is part of the crack tip are shifted towards
the crack tip to the quarter point location along the side. This gives a more accurate description of the
singularity at the crack tip and is important to use if a coarse mesh is used.
In three-dimensions, the mesh along the crack front is typically created from a planar mesh which is
extruded along the crack front region as mentioned above. This allows the rigid regions to be created in
a regular manner which is advantageous for accuracy. Note that Marc allows rigid contact surfaces to be
used for applying symmetry conditions. For a three-dimensional example, see the Marc User’s Guide,
Chapter 6.2.
Dynamic Crack Propagation

The concepts of fracture mechanics discussed in previous sections have been applied to the prediction of
crack initiation, as well as slow stable crack growth of statically loaded structures and for the prediction
of fatigue crack growth in cyclically loaded structures. In problems where inertial effects cannot be
ignored, application of quasi-static fracture mechanics techniques can lead to erroneous conclusions. The
use of dynamic fracture mechanics concepts for these problems is clearly of necessity. The main
CHAPTER 5 165
emphasis on dynamic fracture mechanics is to predict the initiation of stationary cracks in structures,
which are subjected to impact loading. It also focuses on the conditions for the continuous growth of fast
propagating cracks, and on the conditions under which a crack is arrested.
The problem of predicting the growth rate and the possible crack arrest point is quite complicated. It is
often treated by means of a so-called dynamic fracture methodology, which requires the combined use
of experimental measurements and of detailed finite element analyses. An essential step in this approach
is formed by the numerical simulation of propagating cracks by means of the finite element method.
The J-integral as implemented in Marc takes into account the effect of inertial and body forces, thermal
and mechanical loading and initial strains.
The use of the DYNAMIC parameter and the LORENZI model definition option allows for the calculation
of dynamic energy release rates for cracked bodies which are subjected to arbitrary thermal and
mechanical loadings including initial stresses.
Mesh Splitting
Marc has the capability to split up the mesh automatically during the analysis. The mesh split can be done
at nodes and along element edges (2-D and shells) or element faces (3-D). A new node is created where
the split is done. This node inherits the current properties of the original node. The elements at one side
of the split are modified to use the new node to create the opening in the mesh.
The way the mesh splitting works is illustrated in the following. Suppose we have the 2-D mesh shown
in Figure 5-30a. The edges shown in red are targeted for mesh split and this results in the mesh shown
in Figure 5-30b. Figure 5-31 illustrates some different cases and how the mesh is then split. In general,
we need to specify edges for the split so we know how to redefine the element connectivity in order to
split the mesh. The examples are shown in 2-D for simplicity of illustration. The mesh split also applies
to shells and 3-D solids. For 3-D solids, it is based upon faces instead of edges.
a b
Figure 5-30 Mesh Split at Red Edges
No split possible
Figure 5-31 Examples of Mesh Split
The mesh splitting technique is used in a number of features in Marc. The VCCT crack propagation along
element edges (see Direct Crack Growth) makes use of this feature.
The mesh can also be split by means of the USPLIT_MESH user subroutine (see Marc Volume D: User
Subroutines and Special Routines for details).
The DELAMIN model definition option allows the specification of mesh splitting between regions and
within a region using a stress criterion. The mesh is split if the following criterion is fulfilled:
σ m σ n
 -----n- +  -----t > 1 if σ ≥ 0 (5-123)
 S n  S t n
σt n
otherwise  -------- > 1 . (5-124)
 St 
CHAPTER 5 167
Here σ n is the normal stress, σ t is the tangential stress and S n , S t , m and n are user-defined parameters.
Hence, no splitting takes place due to compressive stresses. If S n or S t is zero, the respective term is set
to zero. It is also possible to replace the default stress criterion with a user defined criterion via the
UDELAM user subroutine (see Marc Volume D: User Subroutines and Special Routines for details).
The region where the split is considered can be a material, a composite definition or a PSHELL definition.
With the option to split between regions, the program checks all element edges (faces for 3-D solids) in
the interface between the specified regions. The stresses are extrapolated to the nodes and transformed
to a coordinate system aligned with the region interface. See Figure 5-32. The normal direction is
perpendicular to the region interface. With the option to split within a region, it checks all edges/faces
inside a region. The normal direction is perpendicular to each element edge or face.
region 1
region 2
σt
σn
σt
Figure 5-32 Coordinate System for Delamination between Regions
For layered elements like shells and solid composites, the delamination criterion is calculated separately
for each layer. A mesh split is triggered if all layers fulfill the delamination criterion. The same criterion
is applied to each individual layer. An alternative approach to model material failure in each layer is the
Progressive Failure Analysis approach (PFA) available through the FAIL DATA option (also see
Progressive Composite Failure in Chapter 7 of this manual).
If contact is used, the program recalculates the contact boundary after mesh splitting takes place. This is
important in order to avoid penetration between the newly generated free surfaces.
Nodal post codes are available for post processing the separate terms in Equation (5-123) as well as
the sum. The first has a nodal post quantity labeled Delamination Index (Normal), the second is called
Delamination Index (Tangential) and the sum Delamination Index.
An option is available in the DELAMINATION option for inserting interface elements using a cohesive
zone material model where the mesh is split. This is automatically done whenever a split takes place. The
new elements use the nodes of the surrounding elements and the appropriate type of the interface element
is automatically selected. For example, if a split occurs between two 10-noded tetrahedral elements, then
a 15-noded pentahedral element is inserted. If the split-up zone grows, more interface elements are
added. Collapsed interface elements are used where needed, and these elements change to regular
elements if the split zone grows further. There is no corresponding interface element for shells connected
edge-to-edge, so for this case no interface elements are added.The cohesive material properties of the
interface elements must be defined in the model. The ELEMENTS parameter defining the element type
for the interface elements must be included.
The option is supported for lower- and higher-order 2-D solid, shell and 3-D solid elements. For the case
of higher-order elements, only the corner nodes are considered, and when corner nodes are split, so are
the involved midside nodes. The midside nodes are assigned the average delamination index value of the
corner nodes for post processing.
Dynamics
Marc’s dynamic analysis capability allows you to perform the following calculations:
• Modal (Eigenvalue) Analysis
• Harmonic Response
• Spectrum Response
• Transient Analysis
The program contains two methods for eigenvalue extraction and three time integration operators.
Nonlinear effects, including material nonlinearity, geometric nonlinearity, and boundary nonlinearity,
can be incorporated.
Linear problems can be analyzed using modal superposition or direct integration. All nonlinear problems
should be analyzed using direct integration methods.
In addition to distributed mass, you can also attach concentrated masses associated with each degree of
freedom of the system. You can include damping in either the modal superposition or the direct
integration methods. You can also include (nonuniform) displacement and/or velocity as an initial
condition, and apply time-dependent forces and/or displacements as boundary conditions.
Modal (Eigenvalue) Analysis

Marc uses either the inverse power sweep method or the Lanczos method to extract eigenvalues (natural
frequencies) and eigenvectors (mode shapes). The DYNAMIC parameter is used to determine which
procedure to use, and how many modes are to be extracted. The inverse power sweep method is typically
used for extracting a few modes while the Lanczos method is optimal for several modes. After the modes
are extracted, they can be used in a transient analysis or spectrum response calculation.
In dynamic eigenvalue analysis, we find the solution to an undamped linear dynamics problem:
( K – ω 2 M )φ = 0
where K is the stiffness matrix, M is the mass matrix, ω are the eigenvalues (frequencies) and φ are
the eigenvectors. In Marc, if the extraction is performed after increment zero, K is the tangent stiffness
matrix, which can include material and geometrically nonlinear contributions. The mass matrix is formed
from both distributed mass and point masses.
CHAPTER 5 169
Inverse Power Sweep

Marc creates an initial trial vector. To obtain a new vector, the program multiplies the initial vector by
the mass matrix and the inverse (factorized) stiffness matrix. This process is repeated until convergence
is reached according to either of the following criteria: single eigenvalue convergence or double
eigenvalue convergence. In single eigenvalue convergence, the program computes an eigenvalue at each
iteration. Convergence is assumed when the values of two successive iterations are within a prescribed
tolerance. In double eigenvalue convergence, the program assumes that the trial vector is a linear
combination of two eigenvectors.
Using the three latest vectors, the program calculates two eigenvalues. It compares these two values
with the two values calculated in the previous step; convergence is assumed if they are within the
prescribed tolerance.
When an eigenvalue has been calculated, the program either exits from the extraction loop (if a sufficient
number of vectors has been extracted) or it creates a new trial vector for the next calculation. If a single
eigenvalue was obtained, Marc uses the double eigenvalue routine to obtain the best trial vector for the
next eigenvalue. If two eigenvalues were obtained, the program creates an arbitrary trial vector
orthogonal to the previously obtained vectors.
After Marc has calculated the first eigenvalue, it orthogonalizes the trial vector at each iteration to
previously extracted vectors (using the Gram-Schmidt orthogonalization procedure). Note that the power
shift procedure is available with the inverse power sweep method.
To select the power shift, set the following parameters:
• Initial shift frequency – This is normally set to zero (unless the structure has rigid body modes,
preventing a decomposition around the zeroth frequency).
• Number of modes to be extracted between each shift – A value smaller than five is probably not
economical because a shift requires a new decomposition of the stiffness matrix.
• Auto shift parameter – When you decide to do a shift, the new shift point is set to
Highest frequency2 + scalar x (highest frequency - next highest frequency)2
You can define the value of the scalar through the MODAL SHAPE option.
The Lanczos Method

The Lanczos algorithm converts the original eigenvalue problem into the determination of the
eigenvalues of a tri-diagonal matrix. The method can be used either for the determination of all modes
or for the calculation of a small number of modes. For the latter case, the Lanczos method is the most
efficient eigenvalue extraction algorithm. A simple description of the algorithm is as follows.
Consider the eigenvalue problem:
–ω 2 M u + K u = 0 (5-125)
Equation (5-125) can be rewritten as:
1
------ M u = M K – 1 M u (5-126)
ω2
Consider the transformation:

u = Q η (5-127)
T
Substituting Equation (5-127) into Equation (5-126) and premultiplying by the matrix Q on both sides
of the equation, we have
1
------ Q T M Q η = Q T M K – 1 M Q η (5-128)
ω2
The Lanczos algorithm results in a transformation matrix Q such that:
QT M Q = I (5-129)
Q T M K – 1 MQ = T (5-130)
where the matrix T is a symmetrical tri-diagonal matrix of the form:
α1 β2 0
β2 α 2 β3
T = (5-131)
βm
0 βm αm
Consequently, the original eigenvalue problem, Equation (5-126), is reduced to the following new
eigenvalue problem:
1
------ η = T η (5-132)
ω2
The eigenvalues in Equation (5-132) can be calculated by the standard QL-method.
Through the MODAL SHAPE option, you can either select the number of modes to be extracted, or a range
of modes to be extracted. The Sturm sequence check can be used to verify that all of the required
eigenvalues have been found. In addition, you can select the lowest frequency to be extracted to be
greater than zero.
The Lanczos procedure also allows you to restart the analysis at a later time and extract additional roots.
It is unnecessary to recalculate previously obtained roots using this option.
Convergence Controls
Eigenvalue extraction is controlled by:
a. The maximum number of iterations per mode in the power sweep method; or the maximum
number of iterations for all modes in the Lanczos iteration method,
b. an eigenvalue has converged when the difference between the eigenvalues in two consecutive
sweeps divided by the eigenvalue is less than the tolerance, and
c. the Lanczos iteration method has converged when the normalized difference between all
eigenvalues satisfies the tolerance.
CHAPTER 5 171
5 Modal Stresses and Reactions

After the modal shapes (and frequencies) are extracted, the RECOVER history definition option allows
for the recovery of stresses and reactions at a specified mode. This option can be repeated for any of the
extracted modes. The stresses are computed from the modal displacement vector φ ; the nodal reactions
2
are calculated from F = Kφ – ω Mφ . The RECOVER option is also used to place eigenvectors on the
post file.
Harmonic Response
Harmonic response analysis allows you to analyze structures vibrating around an equilibrium state. This
equilibrium state can be unstressed or statically prestressed. Statically prestressed equilibrium states can
include material and/or geometric nonlinearities. You can compute the damped response for prestressed
structures at various states.
In many practical applications, components are dynamically excited. These dynamic excitations are
often harmonic and usually cause only small amplitude vibrations. Marc linearizes the problem around
the equilibrium state. If the equilibrium state is a nonlinear, statically prestressed situation, Marc
considers all effects of the nonlinear deformation on the dynamic solution. These effects include
the following:
• initial stress
• change of geometry
• influence on constitutive law
The vibration problem can be solved as a linear problem using complex arithmetic.
The analytical procedure consists of the following steps:
where
De l are element damping matrices
Dd are damper contributions
α mass damping coefficient
β stiffness damping coefficient
γ numerical damping coefficient
i = –1
ω = excitation frequencies
u = u r e + iu i m complex response vector
P = P re + iP i m complex load vector

If all external loads and forced displacements are in phase and the system is undamped, this equation
reduces to
( K – ω 2 M )u r e = P r e (5-133)
which could be solved without activating the complex arithmetic on the HARMONIC parameter.
The values of the damping coefficients ( α , β , γ ) are entered via the DAMPING model definition option.
The spring and damper contributions are entered in either the SPRINGS, PFAST, or PBUSH model
definition options and mass points are specified in the MASSES, CONM1, or CONM2 model
definition options.
The element damping matrix ( D e l ) can be obtained for any material with the use of a material damping
matrix which is specified in the UCOMPL user subroutine. You specify the material response with the
constitutive equation.
·
σ = Bε + Cε (5-134)
where B and C can be functions of deformation and/or frequency.

The global damping matrix is formed by the integrated triple product. The following equation is used:
T
D =   β C dV e l (5-135)
el v
el
where β is the strain-displacement relation.
Similarly, the stiffness matrix K is based on the elastic material matrix B . The program calculates the
response of the system by solving the complex equations:
[ K + iωD – ω 2 M ]u = P (5-136)
where u now is the complex response vector
u = u r e + iu i m (5-137)
A special application of the harmonic excitation capability involves the use of the elastomeric analysis
capability in Marc. Here, the Mooney formulation (used in conjunction with the various Herrmann
elements) is used to model the stress-strain behavior of the elastomeric compound. In Marc, the behavior
is derived from the third order invariant form of the strain energy density function.
W ( I 1, I 2 ) = C 10 ( I 1 – 3 ) + C 01 ( I 2 – 3 ) + C 11 ( I 1 – 3 ) ( I 2 – 3 )
(5-138)
+ C 20 ( I 1 – 3 ) 2 + C 30 ( I 1 – 3 ) 3
with the incompressibility constraint
I3 = 1 (5-139)
CHAPTER 5 173
where I 1 , I 2 , and I 3 are the invariants of the deformation. For the harmonic excitation, the constitutive
equation has the specific form
ΔS i j = [ D i j k l + 2iωφ i j k l ]ΔE k l (5-140)
with D i j k l as the quasi-static moduli following form the Mooney strain energy density function and
φ i j k l = φ 0 [ C i–k1 C j–l1 + C i–l1 C j–k1 ] + φ 1 [ δ i k C j–l1 + C i–l1 δ j k ] + (5-141)

φ 2 [ C i k C j–l1 + C i–l1 C j k ] + φ 10 δ i j C k–l1 + φ 11 δ i j δ k l +
φ 12 δ i j C k l + φ 20 C i j C k–l1 + φ 21 C i j δ k l + φ 22 C i j C k l
The output of Marc consists of stresses, strains, displacements and reaction forces, all of which may be
complex quantities. The strains are given by
ε = β u (5-142)
and the stresses by
·
σ = Bε + C ε (5-143)
The reaction forces are calculated with
R = Ku – ω 2 Mu + iωΣD e l u + iωΣD d u (5-144)
2
where – ω Mu is only included if requested on the HARMONIC parameter.
The printout of the nodal values consists of the real and imaginary parts of the complex values, but you
can request that the amplitude and phase angle be printed. You do this with the PRINT CHOICE model
definition option.
Spectrum Response
The spectrum response capability allows you to obtain maximum response of a structure subjected to
known spectral base excitation response. This is of particular importance in earthquake analysis and
random vibration studies. You can use the spectrum response option at any point in a nonlinear analysis
and, therefore, ascertain the influence of material nonlinearity or initial stress.
The spectrum response capability technique operates on the eigenmodes previously extracted to
obtain the maximum nodal displacements, velocities, accelerations, and reaction forces. You can
choose a subset of the total modes extracted by either specifying the lowest n modes or by selecting
a range of frequencies.
Enter the displacement response spectrum S D ( ω ) for a particular digitized value of damping through
the USSD user subroutine. Marc performs the spectrum analysis based on the latest set of modes
extracted. The program lumps the mass matrix to produce M . It then obtains the projection of the inertia
forces onto the mode φ j
Pj = M φj (5-145)
The spectral displacement response for the jth mode is
αj = SD ( ωj ) Pj (5-146)
Marc then calculates the square roots of the sum of the squares as
1⁄2
u =  ( αjφj ) 2 DISPLACEMENT (5-147)
j
1⁄2
v =  ( αj ω j φj ) 2 VELOCITY (5-148)
j
1⁄2
a =  ( α j ω j2 φ j ) 2 ACCELERATION (5-149)
j
1⁄2
f =  ( α j ω j2 Mφ j ) 2 FORCE (5-150)
j
The internal forces given by Equation 5-150 are identified as reaction forces on the post file. The force
transmitted by the structure to the supporting medium (also referred to as base shear) is only reported in
the out file and is given by
2
tf =  S D ( ω j )ω j2 P j TRANSMITTED FORCE (5-151)
j
Transient Analysis
Transient dynamic analysis deals with an initial-boundary value problem. In order to solve the equations
of motion of a structural system, it is important to specify proper initial and boundary conditions. You
obtain the solution to the equations of motion by using either modal superposition (for linear systems) or
direct integration (for linear or nonlinear systems). In direct integration, selecting a proper time step is
very important. For both methods, you can include damping in the system.
The following sections discuss the seven aspects of transient analysis listed below.
• Modal Superposition
• Direct Integration
• Time Step Definition
• Initial Conditions
• Time-Dependent Boundary Conditions
• Mass Matrix
• Damping
CHAPTER 5 175
Modal Superposition
The modal superposition method predicts the dynamic response of a linear structural system. In
using this method, we assume that the dynamic response of the system can be expressed as a
linear combination of the mode shapes of the system. For the principle of superposition to be valid,
the structural system must be linear. Damping can be applied to each mode used in the
superposition procedure.
To select the eigenvalue extraction method and the number of modal shapes, use the DYNAMIC
parameter. To select a fraction of critical damping for each mode, use the DAMPING model definition
option. The DYNAMIC CHANGE history definition option can be used to select the time step.
Marc obtains the transient response on the basis of eigenmodes extracted. The number of modes
extracted (rather than the choice of time step) governs the accuracy of the solution.
Structur Consider the general linear undamped problem
al
Procedu M u·· + K u = f ( t ) (5-152)
re
Library and suppose that n eigenvectors φ 1, …, φ n are known.
u = Σφ i u i ( t ) (5-153)
The eigenmodes are orthogonal with respect to the M and K matrices. After substitution in
Equation (5-152), we find a set of uncoupled dynamic equations
m i u·· i + k i u i = f i ( t ) (5-154)
where:
m i = φ iT Mφ i (5-155)
k i = φ iT Kφ i (5-156)
f i ( t ) = φ iT f ( t ) (5-157)
We can introduce damping on each mode as a fraction of critical damping ( μ i ) for that mode. We can
rewrite Equation (5-154) in the form
m i u·· i + 2m i ω i μ i u· i + k i u i = f i ( t ) (5-158)
or if we normalize Φ i such that m i = 1 :
u·· i + 2ω i μ i u· i + ω i2 u i = f i ( t ) (5-159)
The response of a particular mode is then given by the solution of Equation (5-159) which is:
t
ui ( t ) =  f i ( τ )h ( t – τ ) dτ (5-160)
0
with:
1 – μi ωi t
h ( t ) = ------- e sin ( ω id t ) t≥0
ω id
h(t) = 0 t<0
ω id = ( 1 – μ i2 ) ⋅ ω i
The evaluation of the Duhamel integral, Equation (5-160), can be performed exactly if the load changes
linearly in a specific time increment. Hence, if the load changes rapidly in a specific time period, small
load steps have to be taken.
In case initial displacements u 0 or initial velocities u· 0 are given a transformation to the reduced modal
system is needed for those conditions:
u i0 = Φ iT ⋅ M ⋅ u 0 and u· i0 = Φ iT ⋅ M ⋅ u· 0
The solution of these initial conditions which must be added to the response given in Equation (5-160) is:
–μ ω t u· i0 + u i0 μ i ω i
ui ( t ) = e i i ------------------------------- ⋅ sin ( ω id ⋅ t ) + u i0 ⋅ cos ( ω id ⋅ t ) (5-161)
ω id
The initial accelerations due to given initial displacements u 0 and initial velocities u· 0 can be obtained
by differentiation of Equation (5-161):
u·· i0 = – ( ω i2 ⋅ u i0 ) – ( 2 ⋅ μ i ⋅ ω i ⋅ u· i0 )
Direct Integration
Direct integration is a numerical method for solving the equations of motion of a dynamic system. It is
used for both linear and nonlinear problems. In nonlinear problems, the nonlinear effects can include
geometric, material, and boundary nonlinearities. For transient analysis, Marc offers four direct
integration operators listed below.
• Newmark-beta Operator
• Houbolt Operator
• Generalized-Alpha Operator
• Central Difference Operator
CHAPTER 5 177
To select the direct integration operator, use the DYNAMIC parameter. Specify the time step size through
the DYNAMIC CHANGE or AUTO STEP option. Direct integration techniques are imprecise; this is true
regardless of which technique you use. Each technique exhibits at least one of the following problems:
conditional stability, artificial damping, and phase errors.
Newmark-beta Operator
This operator is probably the most popular direct integration method used in finite element analysis. For
linear problems, it is unconditionally stable and exhibits no numerical damping. The Newmark-beta
operator can effectively obtain solutions for linear and nonlinear problems for a wide range of loadings.
The procedure allows for change of time step, so it can be used in problems where sudden impact makes
a reduction of time step desirable. This operator can be used with adaptive time step control. Although
this method is stable for linear problems, instability can develop if nonlinearities occur. By reducing the
time step and/or adding (stiffness) damping, you can overcome these problems.
Houbolt Operator
This operator has the same unconditional stability as the Newmark-beta operator. In addition, it has
strong numerical damping characteristics, particularly for higher frequencies. This strong damping
makes the method very stable for nonlinear problems as well. In fact, stability increases with the time
step size. The drawback of this high damping is that the solution can become inaccurate for large time
steps. Hence, the results obtained with the Houbolt operator usually have a smooth appearance, but are
not necessarily accurate. The Houbolt integration operator, implemented in Marc as a fixed time step
procedure, is particularly useful in obtaining a rough scoping solution to the problem.
Single Step Houbolt Operator

Two computational drawbacks of the Houbolt operator are the requirement of a special starting
procedure and the restriction to fixed time steps. In [Ref. 29], a Single Step Houbolt procedure has been
presented, being unconditionally stable, second order accurate and asymptotically annihilating. In this
way, the algorithm is computationally more convenient compared to the standard Houbolt method, but
because of its damping properties, the time steps have to be chosen carefully. This algorithm is
recommended for implicit dynamic contact analyses.
Generalized Alpha Operator

One of the drawbacks of the existing implicit operators is the inability to easily control the numerical
dissipation. While the Newmark-Beta method has no dissipation and works well for regular vibration
problems, the Single-Step Houbolt method has numerical dissipation and works well for implicit
dynamic contact problems. A single scheme that easily allows zero/small dissipation for regular
structural dynamic problems and high-frequency numerical dissipation for dynamic contact problems is
desirable. In [Ref. 36], a Generalized-alpha method has been presented as an unconditionally stable,
second-order algorithm that allows user-controllable numerical dissipation. The dissipation is controlled
by choosing either the spectral radius S of the operator or alternatively, two parameters αf and αm. The
choice of the parameters provides a family of time integration algorithms that encompasses the
Newmark-Beta, Single-Step Houbolt and the Hilber-Hughes-Taylor time integration methods as
special cases.
Central Difference Operators

These explicit operators for IDYN = 4 and IDYN = 5 are only conditionally stable. The program
automatically calculates the maximum allowable time step. This method is not very useful for shell or
beam structures because the high frequencies result in a very small stability limit. This method is
particularly useful for analysis of shock-type phenomena. In this procedure, since the operator matrix is
a diagonal mass matrix, no inverse of operator matrix is needed. However, this fact also implies that you
cannot use this method in problems having degrees of freedom with zero mass. This restriction precludes
use of the Herrmann elements, gap-friction elements, the pipe bend element, shell elements 72 and 89
and beam elements 76 and 77. These shell and beam elements are precluded because they have a
rotational degree of freedom, which do not have an associated mass. The mass is updated only if the
LARGE STRAIN or UPDATE parameter or the CONTACT option is used. The elastomer capability can be
used with explicit dynamics in an updated Lagrange framework where the pressure variables are
condensed out before going into the solver.
Consider the equations of motion of a structural system:
Ma + Cv + Ku = F (5-162)
where M , C , and K are mass, damping, and stiffness matrices, respectively, and a , v , u , and F are
acceleration, velocity, displacement, and force vectors. Various direct integration operators can be used
to integrate the equations of motion to obtain the dynamic response of the structural system. The
technical background of the three direct integration operators available in Marc is described below.
Newmark-beta Operator
The generalized form of the Newmark-beta operator is
u n + 1 = u n + Δtv n + ( 1 ⁄ 2 – β )Δt 2 a n + βΔt 2 a n + 1 (5-163)
v n + 1 = v n + ( 1 – γ )Δta n + γΔta n + 1 (5-164)
n
where superscript denotes a value at the nth time step and u , v , and a take on their usual meanings.
The particular form of the dynamic equations corresponding to the trapezoidal rule
γ = 1⁄2, β = 1⁄4
results in
 -------
4
- M + ----- C + K Δu = F n + 1 – R + M  a n + ----- v n + Cv n
2 n 4
(5-165)
 Δt 2 Δt   Δt 
where the internal force R is
R =  β T σdv (5-166)
V
CHAPTER 5 179
Equation (5-165) allows implicit solution of the system
u n + 1 = u n + Δu (5-167)
Notice that the operator matrix includes K , the tangent stiffness matrix. Hence, any nonlinearity results
in a reformulation of the operator matrix. Additionally, if the time step changes, this matrix must be
recalculated because the operator matrix also depends on the time step. It is possible to change the values
of γ and β through the PARAMETERS option.
Houbolt Operator
The Houbolt operator is based on the use of a cubic fitted through three previous points and the current
(unknown) in time. This results in the equations
v n + 1 =  ------ u n + 1 – 3u n + --- u n – 1 – --- u n – 2  ⁄ Δt

11 3 1
(5-168)
6 2 3 
and
a n + 1 = ( 2u n + 1 – 5u n + 4u n – 1 – u n – 2 ) ⁄ Δt 2 (5-169)
Substituting this into the equation of motion results in
 -------
2 11  1
 Δt 2 M + 6Δt C + K Δu = F
- --------- - ( 3u n – 4u n – 1 + u n – 2 )M +
n + 1 – R n + -------
Δt 2
(5-170)
-----  --- u n – --- u n – 1 + --- u n – 2 C
1 7 3 1
Δt  6 2 3 
This equation provides an implicit solution scheme. By solving Equation (5-167) for Δu , you obtain
n+1 n+1
Equation (5-171), and so obtain v and a .
u n + 1 = u n + Δu (5-171)
Equation (5-170) is based on uniform time steps – errors occur when the time step is changed. Also, a
n–1 n–2
special starting procedure is necessary since u and u appear in Equation (5-170).
Single Step Houbolt Operator

The Single Step Houbolt operator according to [Ref. 29] starts with the following equilibrium equation
and expressions for the velocity and acceleration:
m1 n+1 c1 n+1 k1 n+1 m n c n k n
α Ma + α Cv +α Ku + α Ma + α Cv + α Ku =
(5-172)
f1 n + 1 f n
α F +aF
n+1 n n 2 n 1 2 n+1
u = u + Δtv + βΔt a + β Δt a (5-173)
n+1 n n 1 n+1
v = v + γΔta + γ Δta (5-174)
Notice that in contrast to the Newmark and the standard Houbolt method, the equilibrium equation also
contains terms corresponding to the beginning of the increment. Without loss of generality, the parameter
m1
α can be set to 1. Based on asymptotic annihilation and second order accuracy, the remaining
parameters can be shown to fulfill:
k 1 1 m k1 1
α = 0, β = γ, β = γ+γ ,α = –1 ⁄ 2 , α = 1 ⁄ 2β ,
c 1 12 c1 1 12 f k f1 k1
α = – ( 2β + β ) ⁄ 4β ,α = ( 2β + 3β ) ⁄ 4β ,α = α ,α = α
In this way, the number of unknown parameters has been reduced to two. Based on a Taylor series
1 1
expansion of the displacement about the nth time step, β and β should be related by β + β = 1 ⁄ 2,
1 1
which finally yields γ = 1 ⁄ 2 ( 1 ⁄ 2 – γ ) . According to [Ref. 29], γ should be set to 3/2 (with
γ = – 1 ⁄ 2 ) to minimize the velocity error and to 1/2 (with γ = 0 ) to avoid velocity overshoot. The
1 1
default values in Marc are γ = 3 ⁄ 2 and γ = – 1 ⁄ 2 , but the user can modify γ and γ using the
PARAMETERS model definition and history definition option.
Substitution of the velocity and acceleration into the equilibrium equation results in:
c1 1
 1 α γ  n+1 n
- M + ---------------------- C + K Δu = F
 ----------------------- – Ku + (5-175)
1 2 k1 1 k1
 β Δt α β Δtα 
m
1 n 2 n α n
------------------------ M { Δtv + βΔt a } – --------- Ma –
1 2 k1 k1
β Δt α α
αc 1  n n
1
γ  2  α
c
n
- C  v + γΔta – ------------  Δtv n + βΔt a n   – --------- Cv
--------
k1 1 k1
α  β Δt   α
Generalized Alpha Operator

From [Ref. 36], the equilibrium equations for the generalized alpha method can be expressed in the form
n + 1 + αm n + 1 + αf n + 1 + αf n + 1 + αf
Ma + Cv + Ku = F (5-176)
where
n + 1 + αf n+1
u = ( 1 + α f )u – αf un (5-177)
CHAPTER 5 181
n+1+α
f n+1 n
v = ( 1 + α f )v – αf v (5-178)
n+1+α
m n+1 n
a = ( 1 + α m )a – αm a (5-179)
The displacement and velocity updates are identical to those of the Newmark algorithm
u n + 1 = u n + Δtv n + ( 1 ⁄ 2 – β )Δt 2 a n + βΔt 2 a n + 1 (5-180)
v n + 1 = v n + ( 1 – γ )Δta n + γΔta n + 1
where, as shown in [Ref. 36], optimal values of the parameters β and γ are related to α f and α m by
1 2
β = --- ( 1 + α m – α f ) (5-181)
4
1
γ = --- + α m – α f (5-182)
2
In Equation (5-166), α f = α m =0 gives rise to the Newmark-beta method, α m =0, – 0.33 ≤ α f < 0 gives
rise to the HHT - a method [Ref. 37], and α f =0, 0 < α m ≤ 1 gives rise to the WBZ - a method [Ref. 38].
It is seen that the α f and α m parameters can be used to control the numerical dissipation of the operator.
A simpler measure is the spectral radius S. This is also a measure of the numerical dissipation; a smaller
spectral radius value corresponds to greater numerical dissipation. The spectral radius of the generalized
alpha operator can be related to the α f and α m parameters as follows
S
α f = – ------------- (5-183)
1+S
1 – 2S
α m = ---------------- (5-184)
1+S
S varies between 0 and 1. Accordingly, the ranges for the α f and α m parameters are given by
– 0.5 ≤ α f ≤ 0.0 and – 0.5 ≤ α m ≤ 1 . α = -0.5, α m = -0.5 corresponds to a spectral radius of 1.0 and
f
α f = 0, α m = 1 corresponds to a spectral radius of 0.0. By substituting the values in equation (5-145)
and comparing with Equation (5-162), it can be noted that the special case of S = 0 ( α f = 0, α m = 1) for
1
the generalized alpha operator is identical to the default single step houbolt operator with γ = 3⁄2
and γ = – 1 ⁄ 2 . It can also be noted that the case of S = 1 has no dissipation and corresponds to a mid-
increment Newmark- beta operator.
In the Marc input file, the generalized alpha operator is flagged through the option 8 on the second field
of the DYNAMIC parameter. Options to define the associated generalized alpha parameters are available
through the DYNAMIC parameter and the PARAMETERS model definition and history definition option.
On the DYNAMIC parameter, a default value of ‘0’ on the 8th field indicates contact-optimized
parameters ( α f = 0, α m = 1) and a ‘1’ on the 8th field indicates non-contact optimized parameters ( α f
= -0.05, α m = 0). ‘0’ would be recommended for dynamic contact problems or for any problem where
high-frequency numerical dissipation is desired. ‘1’ would be recommended for noncontact structural
dynamic problems and free-vibration problems where a small amount of numerical dissipation is desired.
On the PARAMETERS, the 4th, 5th, and 6th fields of the 5th data block are reserved for α f , α m and S,
respectively. S is the main control parameter: 0 ≤ S ≤ 1 indicates that Equations (5-173) and (5-174) are
used to calculate α f and α m ; S = -1 indicates that α f and α m remain unchanged from previously set
values; S = -2 indicates contact-optimized parameters; S = -3 indicates noncontact optimized parameters;
S = -4 indicates that the 4th and 5th fields will be checked and if valid, used to set α f and α m directly.
Central Difference Operator

The central difference operator assumes a quadratic variation in the displacement with respect to time.
a n = ( v n + 1 ⁄ 2 – v n – 1 ⁄ 2 ) ⁄ ( Δt ) (5-185)
v n = ( u n + 1 ⁄ 2 – u n – 1 ⁄ 2 ) ⁄ ( Δt ) (5-186)
so that
a n = ( Δu n + 1 – Δu n ) ⁄ ( Δt 2 ) (5-187)
where
Δu n = u n – u n – 1 (5-188)
for IDYN=4:
M- n + 1 M
------- Δu = F n – R n + -------- Δu n (5-189)
Δt 2 Δt 2
for IDYN=5:
1
n – ---
M n+1 n n M n 2
-------- Δu = F – R + -------- Δu – Cv (5-190)
2 2
Δt Δt
Since the mass matrix is diagonal, no inverse of the operator matrix is needed. Also, since the operator
is only conditionally s, the critical time step is evaluated at the beginning of the analysis. For IDYN = 4,
the critical time step is computed by a power sweep for the highest mode in the system only at the
beginning of the analysis. For IDYN = 4, no damping is included. For IDYN = 5, an approximated
method based on element geometry is used to compute the highest eigenvalue. The critical time step is
CHAPTER 5 183
calculated at a user-specified frequency or every 100 steps. The variable time step can be used only for
IDYN = 5. Unless there is significant distortion in an element or material nonlinearity, the change of
critical time step is not significant.
Time Step Definition

In a transient dynamic analysis, time step parameters are required for integration in time. The DYNAMIC
CHANGE option can be used for either the modal superposition or the direct integration procedure. The
AUTO STEP option can be used for the Newmark-beta, the Single Step Houbolt, and the Generalized-
alpha operator to invoke the adaptive time control. Enter parameters to specify the time step size and
period of time for this set of boundary conditions.
When using the Newmark-beta operator, decide which frequencies are important to the response. The
time step in this method should not exceed 10 percent of the period of the highest relevant frequency in
the structure. Otherwise, large phase errors will occur. The phenomenon usually associated with too large
a time step is strong oscillatory accelerations. With even larger time steps, the velocities start oscillating.
With still larger steps, the displacement eventually oscillates. In nonlinear problems, instability usually
follows oscillation. When using adaptive dynamics, you should prescribe a maximum time step.
As in the Newmark-beta operator, the time step in Houbolt integration should not exceed 10 percent of
the period of the highest frequency of interest. However, the Houbolt method not only causes phase
errors, it also causes strong artificial damping. Therefore, high frequencies are damped out quickly and
no obvious oscillations occur. It is, therefore, completely up to the engineer to determine whether the
time step was adequate.
For the Generalized-alpha operator, depending on the chosen parameters, the integration scheme can
vary between the Newmark-beta operator and the Single-step houbolt operator. For spectral radii < 1,
there is artificial damping in the system. Depending on the type of problem, the Generalized-alpha
parameters and the associated time step should be carefully chosen to reduce phase errors and effects of
artificial damping.
In nonlinear problems, the mode shapes and frequencies are strong functions of time because of plasticity
and large displacement effects, so that the above guidelines can be only a coarse approximation. To
obtain a more accurate estimate, repeat the analysis with a significantly different time step (1/5 to 1/10
of the original) and compare responses.
The central difference integration method is only conditionally stable; the program automatically
calculates the stable time step. This step size yields accurate results for all practical problems.
Initial Conditions
In a transient dynamic analysis, you can specify initial conditions such as nodal displacements and/or
nodal velocities. To enter initial conditions, use the following model definition options: INITIAL DISP for
specified nodal displacements, and INITIAL VEL for specified nodal velocities. As an alternative, you can
use the USINC user subroutine.
Dynami Time-Dependent Boundary Conditions
cs Simple time-dependent load or displacement histories can be entered on data blocks. However, in general
cases with complex load histories, it is often more convenient to enter the history through a user
subroutine. Marc allows the use of the FORCDT and FORCEM user subroutines for boundary conditions.
the FORCDT user subroutine allows you to specify the time-dependent incremental point loads and
incremental displacements. The FORCEM user subroutine allows you to specify the time-dependent
magnitude of the distributed load.
Mass Matrix
The mass matrix is a discrete representation of the system mass. System mass can be defined through
either distributed masses and concentrated masses.
Distributed masses are defined for elements through the mass density material property. Marc offers both
consistent and lumped element mass matrices. The consistent mass matrix is given by
T
[m] =  ρ[N] [ N ] dV
where ρ is the mass density and [ N ] is the shape function matrix. The lumped mass matrix is flagged
through the LUMP parameter. Marc uses the Hinton, Rock, Zienkiewicz lumping scheme to produce a
diagonal mass matrix. The salient features of this scheme are as follows: The diagonal coefficients of
the consistent mass matrix are computed. The total mass of the element m is also computed. A scale
factor s is computed by adding the translational diagonal coefficients that are mutually parallel and in the
same direction. All diagonal coefficients are then scaled by m ⁄ s , thereby preserving the total mass of
the element.
Concentrated masses are defined through the MASSES, CONM1, and/or CONM2 model definition
options. The MASSES option allows the placement of particle masses m i at degree of freedom of i at
specific nodes. Both translational and rotary masses are allowed.
The CONM1 option allows the specification of a diagonal or 6 x 6 symmetric mass matrix at a node in
the global cartesian or in a local coordinate system. The local system is specified through a coordinate
system ID. The fully coupled 3-D form is given below
M 11 … … … … …
M 21 M 22 … sym … …
M 31 M 32 M 33 … … …
[ m conm1 ] =
M 41 M 42 M 43 M 44 … …
M 51 M 52 M 53 M 54 M 55 …
M 61 M 62 M 63 M 64 M 65 M 66
Each mass term can be varied using tables. Independent variables allowed include time, increment
number, coordinates, temperature. When the diagonal form is specified, only the M 11 to M 66 diagonal
terms need to be specified. In 2-D, the mass matrix reduces to 3 x 3 and only the M 11 , M 12 …M 33 terms
need to be specified. If the CONM1 mass matrix is specified in a local system, then it is transformed to
the global system before assembly.
CHAPTER 5 185
The CONM2 option allows a concentrated diagonal mass and the mass moments of inertia with respect
to the global cartesian or a local coordinate system. The local system is specified through a coordinate
system ID. The mass terms for each degree of freedom at the node are a function of the concentrated
mass, the moments of inertia and the distance of the center of mass from the node. The offset distance
from the node to the center of mass is given by the vector { V 1 V 2 V 3 } . The 3-D form of the CONM2
matrix is given below:
M … … … … …
0 M … sym … …
0 0 M … … …
[ m conm2 ] = 0 – MV 3 MV 2
a
I 11 … …
a a
MV 3 0 – MV 1 I 21 I 22 …
a a a
– MV 2 MV 1 0 I 31 I 32 I 33
where the combined moments of inertia are given as follows:

a 2 2 a 2 2 a 2 2
I 11 = I 11 + M ( V 2 + V 3 ) ; I 22 = I 22 + M ( V 1 + V 3 ) ; I 33 = I 33 + M ( V 1 + V 2 ) ;
a a a
I 21 = – I 21 – MV 1 V 2 ; I 31 = – I 31 – MV 1 V 3 ; I 32 = – I 32 – MV 2 V 3
Each mass coefficient can be varied using tables. Independent variables allowed include time, increment
number, coordinates, temperature. In 2-D, the mass matrix reduces to 3 x 3 and only M , I 33 , V 1 , and
V 2 are required. If the CONM2 mass matrix is specified in a local system, then it is converted to the
global system before assembly.
Damping
In a transient dynamic analysis, damping represents the dissipation of energy in the structural system. It
also retards the response of the structural system.
Marc allows you to enter two types of damping in a transient dynamic analysis: modal damping and
Rayleigh damping. Use modal damping for the modal superposition method and Rayleigh damping for
the direct integration method. For modal superposition, you can include damping associated with each
mode. To do this, use the DAMPING option to enter the fraction of critical damping to be used with
each mode.
During time integration, Marc associates the corresponding damping fraction with each mode.
The program bases integration on the usual assumption that the damping matrix of the system is a
linear combination of the mass and stiffness matrices, so that damping does not change the modes of
the system.
For direct integration damping, you can specify the damping matrix as a linear combination of the mass
and stiffness matrices of the system. You can specify damping coefficients on an element basis.
Stiffness damping should not be applied to either Herrmann elements or gap elements because of the
presence of Lagrangian multipliers.
Numerical damping is used to damp out unwanted high-frequency chatter in the structure. If the time step
is decreased (stiffness damping might cause too much damping), use the numerical damping option to
make the damping (stiffness) coefficient proportional to the time step. Thus, if the time step decreases,
high-frequency response can still be accurately represented. This type of damping is particularly useful
in problems where the characteristics of the model and/or the response change strongly during analysis
(for example, problems involving opening or closing gaps).
Element damping uses coefficients on the element matrices and is represented by the equation:
n
  Δt 
C =   α i M i +  β i + γ i ----
π  i 
- K (5-191)
i = 1
where
C is the global damping matrix
Mi is the mass matrix of ith element
Ki is the stiffness matrix of the ith element
αi is the mass damping coefficient on the ith element
βi is the usual stiffness damping coefficient on the ith element
γi is the numerical damping coefficient on the ith element
Δt is the time increment
If the same damping coefficients are used throughout the structure, Equation (5-191) is equivalent to
Rayleigh damping.
The damping associated with springs and mass points can be controlled via the springs and masses input
options. The damping on elastic foundations is the same as the damping on the element on which the
foundation is applied. For springs, a dashpot can be added for nonlinear analysis.
Inertia Relief
Inertia relief refers to an analysis procedure that allows unconstrained systems to be subjected to a quasi-
static analysis by taking rigid body inertia forces into account. Examples of such systems include an
aircraft in a steady turn, an underwater structure in equilibrium under gravity and buoyancy.
Conventional static analysis cannot be performed for such systems since, in the absence of constraints,
the stiffness matrix is singular. Inertia relief analysis is applicable to such free bodies. The response is
measured relative to a steady accelerating frame induced by the external loads.
CHAPTER 5 187
Two important steps are followed by the solver in an analysis with inertia relief:
1. Rigid Body Mode evaluation: This is the central step in an inertia relief analysis since the rigid
body modes are extensively used in the load computations. The Support Method is currently
provided in Marc to evaluate the rigid body modes.
2. Inertia Relief Loads evaluation: The loads associated with the rigid body motion of the system
are calculated using the rigid body modes and the global lumped mass matrix. The loads are
calculated such that they balance the current external load vector in the system.
Next, more details on each of the steps are provided.
Rigid Body Mode Evaluation

Support Method
The Support Method requires the manual specification of nodal degrees of freedom which can be used
to evaluate the rigid body modes of the system. Any number of nodal entries can be provided, each
with degrees of freedom ranging from 1 to the maximum possible degrees of freedom at each node.
These degrees of freedom are referred to as the r-degrees of freedom set. It is the user’s responsibility to
ensure that the r-degrees of freedom set is necessary and sufficient to evaluate the rigid body modes of
the system.
The rigid body modes are evaluated by sequentially setting each r-degrees of freedom to 1 with all other
r-degrees of freedom set to 0 and solving for the system of equations
[ Kl l ] [ Kl r ]  [ φl ] 
  = 0 (5-192)
[ K r l ] [ K rr ]  [ I ] 
where [ I ] is the identity matrix for the r-degrees of freedom and [ φ l ] are the modal amplitudes for the
remaining degrees of freedom. Once the rigid body modes are evaluated, the r-degrees of freedom are
set to 0 and used as fixed displacement constraints. Other displacement boundary conditions, specified
through normal options like FIXED DISP, DISP CHANGE, etc. can be used in conjunction with the rigid
body degrees of freedom constraints.
The Support Method is computationally inexpensive since the provided Support degrees of freedom
are automatically constrained and no additional ties/constraints are needed while calculating the
response. It is available for single-processor and multi-processor runs (both single-input and
multi-input file formats).
The Support Method requires manual input to evaluate the rigid body modes of the system. For direct
solvers, the rigid body modes are calculated through repeated back-solves of the factorized matrix. The
computation of rigid body modes is more expensive for the conjugate gradient iterative solver. Finally,
the Support Method cannot be used in contact problems with dynamically changing rigid body modes.
Rigid Body Mode Update

In a small displacement analysis, the rigid body modes are evaluated at the first cycle of the first
increment of the inertia relief loadcase and are retained through the loadcase. In a large displacement
analysis, the rigid body modes are updated at the beginning of each increment by using the current
stiffness matrix while solving Equation (5-192).
Inertia Relief Loads Evaluation

A general derivation for the inertia relief load vector in a nonlinear analysis is presented here. At the i th
iteration of the ( n + 1 ) th increment, the total response is given by
{ u tn, i+ 1 } = { u rn } + { Δu rn, +i –1 1 } + { δu rn, +i 1 } + { u n } + { Δu in–+11 } + { δu in + 1 } (5-193)
where { u t } is the total response, { u r } is the rigid body response and { u } is the flexible response.
Neglecting accelerations associated with the flexible response, the governing equations are given as
n+1 n+1 n+1 n+1 n+1 n+1
[ M i – 1 ] { δu·· r, i } + [ K i – 1 ] { δu i } = { P n + 1 } – { I i – 1 } – { P φ, i – 1 } (5-194)
n+1 n+1
where [ M i – 1 ] and [ K i – 1 ] are the mass and stiffness matrices at the ( i – 1 ) th iteration of ( n + 1 ) th
n+1 n+1
increment, { P n + 1 } is the external load vector at the ( n + 1 ) th increment, I i – 1 and { P φ, i – 1 } are
the internal load vector and the inertia relief load vector at the ( i – 1 ) th iteration of the ( n + 1 ) th
increment, respectively.
The change in the rigid body acceleration is computed as follows:
{ δu·· rn, +i 1 } = [ Φ in–+11 ] { δA φn, +i 1 } (5-195)
where [ Φ in–+11 ] is the matrix of rigid body modes at the ( i – 1 ) th iteration of the ( n + 1 ) th increment
n+1
and { δA φ, i } represents the iterative change in the modal acceleration vector.
n+1
Pre-multiplying Equation 5-184 by individual rigid body modes, { φ k, i – 1 } T , and using the fact that the
work done by the rigid body modes is 0, one gets
n+1 n+1 n+1
[ C i – 1 ] { δA φ, i } = { δR i } (5-196)
n+1
where the kl th term of matrix [ C i – 1 ] is given by
n+1 n+1 n+1 n+1

C i – 1, k l = { φ k, i – 1 } T [ M i – 1 ] { φ l, i – 1 }
n+1
and the k th term of { δR i } is given by
CHAPTER 5 189
n+1 n+1 T n+1 n+1 n+1

δR i, k = { φ k, i – 1 } [ { P } – { I i – 1 } – { P φ, i – 1 } ] (5-197)
n+1
Note that if the rigid body modes are mass ortho-normalized, matrix [ C i – 1 ] is an identity matrix.
n+1
Solving Equation (5-196) for { δA φ, i } , the iterative change in the inertia relief load vector is given by
n+1 n+1 n+1 n+1

{ δP φ, i } = [ M i – 1 ] [ Φ i – 1 ] { δA φ, i } (5-198)
The governing equations in Equation 5-184 can then be rewritten in the form,
n+1 n+1 n+1 n+1 n+1
[ K i – 1 ] { δu i } = { Pi } – { I i – 1 } – { P φ, i }
n+1
where the inertia relief load vector at the i th iteration of the ( n + 1 ) th increment, { P φ, i } , is given by
n+1 n+1 n+1

{ P φ, i } = { P φ, i – 1 } + { δP φ, i }
Additional items to be kept in consideration for inertia relief analysis in Marc are as follows:
Convergence Checking
For systems that are unconstrained or just constrained against rigid body motion (statically determinate
structures), the reaction forces at nodes are 0 or very close to 0. In this case, using relative residual force
checking may lead to unnecessary recycling and possible non-convergence. It would be more
appropriate to use displacement checking or absolute residual force checking in such situations.
Time Stepping
Inertia relief can be used in conjunction with all time stepping procedures used for static analyses. Time
step cutback schemes are supported. Inputs specified in either table or non-table formats are also
supported.
Parallel Processing
The Support Method is available for serial and parallel runs. For the multi-input file parallel mode, the
Support nodes should be available in all the domains. For the single-input file parallel mode, the inertia
relief option should be present in the model definition section itself in order to allow the Support nodes
to be added to domains they are not naturally present in.
Deactivating Inertia Relief

When inertia relief is active during an analysis, and is subsequently turned off, the user has three choices
for treating the pre-existing inertia relief load vector.
1. The inertia relief load vector can be retained. No new inertia relief loads are computed, but pre-
existing inertia relief loads are included in the right-hand side.
2. The inertia relief load vector can be removed instantaneously. Note that this is the option used by
Marc if the INERTIA RELIEF option is present in one loadcase and is absent in the next loadcase.
This could lead to convergence difficulties at the beginning of the second loadcase due to the large
change in load vectors.
3. The inertia relief load vector can be removed gradually. This allows the inertia relief load vector
to be removed over the course of the second loadcase.
Inertia Relief Ouput Options

The nodal inertia relief load and moment vectors are written to the post file by default using nodal post
codes 51 and 52, respectively. It should be noted that the inertia relief loads are always written out in the
global cartesian system. Also, the command INER can be used on the PRINT NODE option to obtain the
inertia relief loads at a particular node. Also, when the NO PRINT option is revoked, a summary of the
total inertia relief load is printed in the out file.
Limitations for Inertia Relief Analyses

1. All inertia relief analyses provide the uninteresting solution of zero displacements when a uniform
mesh is subjected to uniform gravity loads or when a uniform shell mesh is subjected to uniform
distributed loads. This is as expected because the inertia relief equations place an inertia load on
each node that is equal and opposite to the external load.
2. The Support Method is currently not supported for the CASI iterative solver.
3. The Support Method is capable of dealing with material and geometric non-linearities. However,
boundary condition non-linearities like contact between multiple bodies cause the number of rigid
body modes to change in the system. These require relative rigid body positions of the bodies to
be continuously updated. These cannot be handled by the Support Method.
Rigid-Plastic Flow
The rigid-plastic flow analysis is an approach to large deformation analysis which can be used for metal
forming problems. Two formulations are available: an Eulerian (steady state) and Lagrangian (transient)
approach. The effects of elasticity are not included. If these effects are important, this option should not
be used.
In the steady state approach, the velocity field (and stress field) is obtained as the solution of a steady-
state flow analysis. The time period is considered as 1.0 and, hence, the velocity is equal to the
deformation. In the transient formulation, the incremental displacement is calculated.
The R-P FLOW parameter invokes the rigid-plastic procedure. This procedure needs to enforce
the incompressibility condition, which is inherent to the strictly plastic type of material response
being considered.
Incompressibility can be imposed in three ways:
1. by means of Lagrange multipliers. Such procedure requires Herrmann elements which have a
pressure variable as the Lagrange multiplier.
2. by means of penalty functions. This procedure uses regular solid elements, and adds penalty terms
to any volumetric strain rate that develops. It is highly recommended that the constant dilatation
formulation be used – by entering a nonzero value in the second field of the GEOMETRY model
definition option.
CHAPTER 5 191
Penalty factor can be treated as constant or variable through the R-P FLOW parameter. The
penalty value is entered through the PARAMETERS option.
3. In plane stress analysis (shell and membrane elements), the incompressibility constraint is
satisfied exactly by updating the thickness. This capability is not available for steady
state analysis.
In R-P flow analysis, several iterations are required at any given increment, the greatest number
occurring in the first increment. Subsequent increments require fewer iterations, since the initial iteration
can make use of the solution from the previous increment.
Due to the simplicity of the rigid-plastic formulation, it is possible to bypass stress recovery for all
iterations but the last in each increment, provided that displacement control is used. In such cases,
considerable savings in execution time are achieved. If nodal based friction is used in a contact analysis,
then a stress recovery is always performed after each iteration.
Steady State Analysis

The steady state R-P flow formulation is based on an Eulerian reference system. For problems in which
a steady-state solution is not appropriate, an alternative method is available to update the coordinates.
The UPNOD user subroutine is used to update the nodal coordinates at the end of a step according to
the relationship.
n
x in = x in – 1 + v Δt (5-199)
where n refers to the step number, v n is the nodal velocity components, and Δt is an arbitrary time step.
Δt is selected in such a way as to allow only a reasonable change in mesh shape while ensuring stability
with each step.
Updating the mesh requires judicious selection of a time step. This requires some knowledge of the
magnitude of the nodal velocities that will be encountered. The time step should be selected such that the
strain increment is never more than one percent for any given increment.
The quantities under the title of ENGSTN in the printouts actually refer to the strain rate at an element
integration point. The reaction forces output by the program gives the limit loads on the structure.
Transient Analysis
In the transient procedure, there is an automatic updating of the mesh at the end of each increment.
During the analysis, the updated mesh can exhibit severe distortion and the solution might be unable to
converge. Global adaptive meshing or manual mesh rezoning can be used to overcome this difficulty.
The rigid-plastic flow capability is based on iteration for the velocity field in an incompressible, non-
Newtonian fluid. The normal flow condition for a nonzero strain rate can be expressed as:
 2 σ · · ·
σ′ i j =  --- --·- ε i j = μ ( ε ) ε i j (5-200)
 ε
3
where
· 2· ·
ε = --- ε i j ε i j (5-201)
3
is the equivalent strain rate, σ is the yield stress (which may be rate-dependent) and
1
σ′ i j = σ i j – --- δ i j σ k k (5-202)
3
gives the deviatoric stress. The effective viscosity is evaluated as:
2 σ
μ = --- --·- (5-203)
3 ε
·
Note that as ε → 0, μ → ∞ . A cutoff value of strain rate is used in the program to avoid this difficulty.
·
An initial value for ε is necessary to start the iterations. These values can be specified in the
PARAMETERS option. The default cut-off value is 10 – 6 , and the default initial strain rate value is 10-4.
The value of the flow stress is dependent upon both the equivalent strain, the equivalent strain rate, and
the temperature. This dependence can be given through the WORK HARD, STRAIN RATE,
TEMPERATURE EFFECTS, or TABLE options, respectively. For steady state analysis, the UNEWTN user
subroutine can be used to define a viscosity. In this manner, a non-Newtonian flow analysis can be
performed. For the transient procedure, the URPFLO user subroutine can be used to define the
flow stress.
Superplasticity
Superplasticity is the ability of a material to undergo extensive deformation such as strains of 1000%
without necking. Superplastic behavior has been reported in numerous metal, alloys and ceramics. Every
instance of superplasticity is associated with: (1) A fine grain size ( 1 – 10 μm ), (2) deformation
temperatures > 0.4T m , and (3) a strain-rate sensitivity factor m > 0.3 .
Using finite element analysis to simulate superplastic fabrication of complex parts used in the aerospace
and automotive industries requires this material behavior and contact with friction. Furthermore, the
process pressure needs to be automatically adjusted to keep the material within a target strain rate. The
CHAPTER 5 193
simulation can be used to predict thinning, forming time, areas of void formation, and can ultimately be
employed in shape optimization; thus, reducing the number of prototypes of forming trials required to
product an acceptable part.
Three mechanisms have been proposed to account for the high strain-rate sensitivity found in
superplastic materials: (1) Vacancy creep, (2) creep by grain boundary diffusion, and (3) grain boundary
sliding. According to Ghosh and Hamilton, the strain-rate sensitivity of metals arises from the viscous
nature of the deformation process. The viscosity is a result of the resistance offered by internal obstacles
within the material. In dislocation glide and climb processes, the obstacles are a fine dispersion of second
phase particles within the grain interior, between which the dislocations are bent around and moved. At
high homologous temperatures ( T > 0.4T m ; where T m is the melting temperature), the high
diffusivities around grain boundary regions can lead to grain boundary sliding. The overall rate
sensitivity of a material is then a result of the rate sensitivities of the grain boundary and the grain
interior. The more the material behaves as a viscous liquid, the greater its superplasticity.
The superplastic behavior is characterized by the dependence of the flow stress upon the strain-rate,
which is usually depicted by the logarithmic relationship shown in Figure 5-33.
As indicated in Figure 5-33, the stress-strain rate behavior of a superplastic material can be divided into
three regions. Values of strain-rate sensitivity, m (the slope of flow stress versus strain-rate curve) which
is a measure of resistance to localized necking, are relatively low in both the low stress-low strain rate
region I and the high stress-high strain rate region III and superplasticity is not manifested. Rather,
superplasticity is found only in region II, a transition region in which stress increases rapidly with
increasing strain-rate. As temperature increases and/or grain size decreases, region II is displaced to
higher strain-rates. Moreover, the maximum observed values of m increase with similar changes in
these parameters.
Certainly the forming process innovations evoked will need to be carefully studied and developed.
Forming times are slow, and there will be a critical need for optimizing forming pressures, stress strain-
rate and deflection in sheet forming. Based on the schematic flow stress-strain rate relationships given
above, it is apparent that high values of m are requisite for superplastic materials. Since, for a given
material and forming temperature, m , usually varies with strain-rate, it is desirable to control strain-rate
during forming so that optimum or at least adequate strain-rate sensitivity is exhibited. Ductility is also
dependent upon forming temperature, which must lie within a narrow range. If forming temperatures and
pressure cycle are optimum, then unlike conventional ductile materials, superplastic materials are much
less susceptible to localized necking. Additionally, under such conditions, the flow stress occurring
during forming is much lower than the mechanical yield stress.
(a) Flow Stress

Region III
ln σ
Region II
Decreasing grain size or

Region I increasing temperature
.
ln ε
(b) Strain-rate
Sensitivity Decreasing grain size or
increasing temperature
d ln σ
m  = ------------·- Region I
 
d ln ε
Region II
Region III
.
ln ε
Figure 5-33 Function of Strain-rate
Thus, the superplastic materials may be viewed as exhibiting time-dependent inelastic behavior with the
yield stress as a function of time, temperature, strain-rate, total stress and total strain. Typical materials
used in commercial superplastic forming applications include Ti-6A1-4V titanium alloy and 5083
aluminium alloy.
The form of constitutive equation used to simulate superplasticity is given as:
·m
σy = ε (5-204)
The form in Equation (5-204) can be recovered by using appropriate constants in the ISOTROPIC model
definition option to define power law or rate power law. Thus,
m ·n
Power Law Model: σ y = A ( ε o + ε ) + Bε (5-205)
m ·n
Rate Power Law Model: σ y = Aε ε + B (5-206)
CHAPTER 5 195
The superplastic forming process requires the use of the SPFLOW parameter. The use of this parameter
automatically activates the FOLLOW FOR and PROCESSOR parameters. The process parameters are
controlled by the use of the SUPERPLASTIC history definition option.
Typical outputs that are available from the superplastic forming simulation are the thickness distribution
for the part, the equivalent plastic strain rate, and the history of the process pressure.
The process pressure is automatically calculated during the analysis. The pressure magnitude is adjusted
such that the equivalent strain rate in the part is at or close to the user-specified target strain rate. The
equivalent strain rate in the part is an average value calculated by sampling a suitable subset of elements.
The recommended scheme is one in which elements with a strain rate greater than a cut-off factor times
the maximum element strain rate are sampled. This maximum strain rate is based on a smoothing
algorithm described below. The cut-off factor can vary between 0 (all elements below the maximum are
sampled) and 1 (only the elements with maximum strain rate are sampled). The recommended value for
the cut-off factor is 0.7 to 0.9 (default value is 0.8). To reduce undesirable oscillations in the pressure-
time history, a pressure smoothing algorithm is incorporated. The basis for this algorithm is a smoothing
of the maximum strain rate in the mesh based on the fact that the maximum strain rate should be typically
representative of a few elements in the mesh, rather than an isolated individual value. The peak strain
rates in a few elements are calculated. The number of elements that are used in this calculation varies
with the cut-off factor (for a cut-off factor of 1, only 1 element is used; for the default of 0.8, 10 elements
are used; for a cut-off factor of 0, 50 elements are used). The strain rate values are successively
disregarded in descending order if the difference from the highest strain rate to the lowest differs by more
than 10 percent from the mean.
The value of the cut-off factor has significant influence on the maximum strain rate control and on the
smoothness of the pressure-time curve. Larger the factor (that is, 0.9 or higher) provides more control on
the maximum strain rate, but may potentially cause oscillations in the pressure history. Smaller the factor
(that is, 0.7 or lower) provides less control on the maximum strain rate, but causes smoother pressure-
time curves. The default of 0.8 should work in most cases - in situations where physically realistic
localized strain rates occur and one desires good control on these localized values, a higher value could
be used.
Soil Analysis
This section has the solution procedure for fluid-soil analysis. In the current formulation, it is assumed
that the fluid is of a single phase, and only slightly compressible. This formulation will not be adequate
if steam-fluid-solid analysis is required. The dry soil can be modeled using one of the three models:
linear elasticity, nonlinear elasticity and the modified Cam-Clay model.
There are three types of soil analysis available in Marc. In the first type, you perform an analysis to
calculate the fluid pressure in a porous medium. In such analyses, heat transfer elements 41, 42, or 44
are used. The SOIL model definition option is used to define the permeability of the solid and the bulk
modulus and dynamic viscosity of the fluid. The porosity is given either through the INITIAL POROSITY
or the INITIAL VOID RATIO options and does not change with time. The prescribed pressures can be
defined using the FIXED PRESSURE option, while input mass flow rates are given using either the
POINT FLUX or DIST FLUXES option.
5 In the third type of soil analysis, a fully-coupled approach is used. Element types 32, 33, or 35 are
available. These elements are “Herrmann” elements, which are conventionally used for incompressible
analysis. In this case, the extra variable represents the fluid pressure. The SOIL option is now used to
define both the soil and fluid properties. The porosity is given through the INITIAL POROSITY or the
INITIAL VOID RATIO option. The prescribed nodal loads and mass flow rates are given through the
POINT LOAD option, while distributed loads and distributed mass flow rates are given through the DIST
LOADS option. The FIXED DISP option is used to prescribe either nodal displacements or pore pressures.
Struct Technical Formulation

ural In soil mechanics, it is convenient to decompose the total stress σ into a pore pressure component p and
Proce the deviatoric or effective stress σ d .
dure
σ = σ d – pI (5-207)
Librar
Note the sign convention used; a positive pore pressure results in a compressive stress. The momentum
y balance (equilibrium) equations are with respect to the total stresses in the system.
∇σ + f = ρu·· (5-208)
where ρ is the density, and f , u·· are the body force and the acceleration. The equilibrium equation can
then be expressed as
∇σ d – ∇p + f = ρu·· (5-209)
The fluid flow behavior can be modeled using D’Arcy’s law, which states that the fluid’s velocity, relative
to the soil’s skeleton, is proportional to the total pressure gradient.
–K
u· f = ------- ( ∇p + ρ f g ) (5-210)
μ
where
u· f is the fluid’s bulk velocity
K is the soil permeability

μ is the fluid viscosity
ρf is the fluid density
g is the gravity vector.
The fluid is assumed to be slightly compressible.

·
ρf
p· = K ----- (5-211)
f ρf
CHAPTER 5 197
where K f is the bulk modulus of the fluid.
However, the compressibility is assumed small enough such that the following holds:
K K
∇ ⋅ ρ f ---- ( ∇p + ρ f g ) ≈ ρ f ∇ ⋅ ---- ( ∇p + ρ f g ) (5-212)
μ μ
It is also assumed that the bulk modulus of the fluid is constant, introducing the fluid’s
compressibility β f .
βf = 1 ⁄ Kf (5-213)
The solid grains are assumed to be incompressible. Under these assumptions, the governing equations
for fluid flow is
K ·
∇ ⋅ ---- ( ∇p + ρ f g ) = φβ f p + ∇u· (5-214)
μ
where σ is the medium’s porosity.

It is important to note that the medium’s porosity is only dependent upon the original porosity and the
total strains. Letting V f and V s stand for the fluid and solid’s volume
φ = dV f ⁄ ( dV f + dV s ) = 1 – J – 1 ( 1 – φ 0 ) (5-215)
where J is the determinant of the deformation gradient and φ 0 is the original porosity, both with respect
to the reference configuration.
Introducing the weighting function η u and η p , the weak form, which is the basis for the finite element
system, then becomes
 [ η u f – ρη u u·· – ∇η u σ + ∇η u p ]dv +  ηu t dA = 0 (5-216)

Vn + 1 A
where V and A are the conventional volumes and surface area and t is the applied tractions. Note that
the applied tractions is the combined tractions from both the effective stress σ d and the pore pressure p .
and
K
 ∇η p ---- ( ∇p + ρ f g ) + φβ f η p p· + η p ( ∇u· ) dv –
μ  η p q n dA = 0 (5-217)
V A
n+1 n+1
where the normal volumetric inflow, q n , is:
k
q n = --- ( ∇p + ρ f g ) ⋅ n (5-218)
μ
The weak form of equilibrium can be written as:
ru =  [∇ ⋅ η u p ] dv (5-219)
V
n+1
rp =  ∇η p ⋅ K
---- ( ∇p + ρ f g ) + φβ f η p p· + η p ∇ ⋅ u· dv
μ
(5-220)
Vn + 1
Application of the directional derivative formula yields:
d
D u ( r u ) ⋅ δu = ------ r u ( u + εδu ) (5-221)
dε ε = 0
Hence,
D u ( r u ) ⋅ δu =  [∇ ⋅ Δu ( ∇ ⋅ η u p ) + ∇η u
T
⋅ ∇Δup ] dv (5-222)
Vn + 1
Similarly,
D p ( r u ) ⋅ δp =  ∇ ⋅ η u Δp dv (5-223)
Vn + 1
D u ( r p ) ⋅ δu = 

 ∇η p ⋅ K
---- ( ∇p + ρ f g ) + φβ f η p p· + η p ∇
μ
⋅ u· ∇ ⋅ Δu – η p ∇u·
T
⋅
V 
n+1
(5-224)
∇Δu + η p ∇ ⋅ Δu· – ∇Δu∇η p ⋅ K

---- ( ∇p + ρ f g ) – ∇η p
μ
⋅ K
---- ( ∇Δu )∇p }dv
μ
D p ( r p ) ⋅ δp =  ∇η p ⋅ K
---- ∇Δp + φβ f η p Δp· dv
μ
(5-225)
Vn + 1
CHAPTER 5 199
with the displacement and pressures interpolated independently as:
Δu =  N i Δu i
and (5-226)
p =  Nj pj
we get a linearized system of equilibrium equation,
In the second type of soil analysis, the pore (fluid) pressure is directly defined, and the structural
analysis is performed. Element types 27, 28, or 21 are available. In such analyses, the pore pressure
is prescribed using the INITIAL PORE and CHANGE PORE options. The characteristics of the soil
material are defined using the SOIL option. If an elastic model is used, the Young’s moduli and Poison’s
ratio are important. If the Cam-Clay model is used, the compression ratios and the slope of the critical
state line is important. For the Cam-Clay model, the preconsolidation pressure is defined using the
INITIAL PC option. For this model, it is also important to define an initial (compressive stress) to ensure
a stable model.
K u u K u p  δu   Ru 
  = F–  (5-227)
K p u K p p  δp   Rp 
The resulting system of equations is highly nonlinear and nonsymmetric, and is solved by full Newton-
Raphson solution scheme.
Note that it is assumed that the permeability, porosity, viscosity, and the bulk modulus of the fluid are
considered independent of the state variables u and p . It is evident that, in general, this is not the case;
however, in the analysis that follows, these dependencies are ignored for tangent purposes. Note that they
are included in the calculation of the residuals R u and R p ; hence, convergence is always achieved at the
true solution.
Three types of analyses can be performed. The simplest is a solution for only the fluid pressure based
upon the porosity of the soil. In this case, a simple Poisson type analysis is performed and element types
41, 42, or 44 are used. In the second type of analysis, the pore pressures are explicitly defined and the
structural analysis is performed. In this case, the element types 21, 27, and 28 should be used. Finally, a
fully-coupled analysis is performed; in which case, you should use element types 32, 33, or 35. Of
course, the soil can be combined with any other element types, material properties to represent the
structure, such as the pilings.
Mechanical Wear
Mechanical wear is an important physical phenomenon in any structure subjected to repeated loadings.
Often, this behavior is modeled by determining the stress on the surface and using it in a subsequent
fatigue calculation. For certain applications, including manufacturing, disk brakes, bearings, gears, tires,
and seals, it is important to know the amount of wear and possibly the change in geometry which would
influence the behavior. It should be noted that for some applications, such as polishing, surface wear is
the positive objective. Wear, which is defined as the removal of material from the surface, may be due to
mechanical processes or chemical processes. The latter are not considered but the UWEARINDEX user
subroutine may always be used to incorporate these effects. The wear due to mechanical behavior can be
classified as:
1. Asperity deformation and removal
2. Plowing of the surface
3. Delamination
4. Adhesive
5. Abrasion
6. Fretting
7. Solid particle impingement
These processes lead to either mild wear or severe wear (rough, torn surfaces). The wear model
implemented in Marc, based upon Archard equation, is not applicable for severe wear.
There are several wear models available beginning with Archard’s equation:
Gt
w = KF ⋅ -----
H
where
K is the wear coefficient
F is the normal force
Gt is the sliding distance
H is the hardness
This is converted into an incremental form. As the wear is calculated as a local quantity, we use normal
stress instead of the normal force (except for beam elements as discussed below). This leads to a series
of models that are written as:
K Simple Archard
w· = ---- ⋅ σ ⋅ V r e l
H
K m n
w· = ---- ⋅ σ ⋅ V r e l Bayer exponential form
H
–T0 ⁄ T
K m n
w· = ---- σ V r e l exp Bayer exponential form with thermal activation
H
where
CHAPTER 5 201
w· is the rate of change of wear in the direction normal to the surface
σ is the normal stress
V r e l is the relative sliding velocity
T0 is the activation temperature
T is the current temperature
A table may be associated with the coefficient to allow temperature dependence or other effects. The
following quantities can be used as independent variables for the wear coefficients: time, increment,
current and original coordinate position, normal stress and relative sliding velocity.
The incremental wear calculation is performed at the nodal points that are in contact. Hence, the
CONTACT option is required. Although friction is typically used in a wear analysis, it is not needed for
the wear calculation itself since it only uses the normal stress and the relative sliding velocity. The wear
model and data is specified for each contact body that should do wear calculation. The wear can occur
on one or both sides of the bodies in contact. Each side uses the wear model and properties of the
respective body. A scaling factor can be used on the contact table allowing the wear to be different for
different contact combinations.
The incremental wear is calculated as:
w· Δt
and the wear is accumulated as:

w n + 1 = w n + w· Δt .
The total accumulated wear can be interpreted as the distance a node would move due to wear and
indicates a volume loss. The total volume loss due to wear is calculated per contact body and printed to
the output file. The total accumulated wear w t o t is available for post processing as nodal post code 77,
Wear Index. The current wear rate w· is available as a vector in the normal direction of the contact point
as nodal post code 78, Wear Rate.
For continuum elements, the nodal coordinates can be updated using
·
x n + 1 = x n + fw Δt ⋅ n
where n is the normal to the surface and f is a scale factor. If large steps are taken such that the
incremental wear becomes large one might get inaccurate results due to nodes ending up outside the
contact tolerance. New contact will be found again, but small steps are often needed for accurate results.
If the total amount of wear becomes large, then remeshing may be required.
Wear calculation is supported for all continuum, shell, and beam elements. In addition, for
thermo-mechanically coupled analysis, wear calculation is supported on rigid (meshed) bodies with heat
transfer. These also support coordinate update due to wear. The coordinate update option is not supported
for shell elements (since the shell thickness should be changed, not the coordinate positions, but shell
thickness update due to wear is not supported).
Beam elements can as mentioned be used in a wear calculation. This covers both the case that a beam
element is touching a solid or shell element and the case of beam-to-beam contact. Since we do not have
a contact area available in beam contact, we use the contact normal force instead of the contact normal
stress in the wear calculation. This needs to be taken into account when using wear on beam elements.
The volume loss due to wear is calculated for each element as an area obtained by the length of the
element and the length offset by the amount of wear, and this area is multiplied with twice the contact
distance radius.
Design Sensitivity Analysis

5 Design sensitivity analysis is used to obtain the sensitivity of various aspects of a design model with
respect to changes in design parameters in order to facilitate structural modifications. The design
parameters that are amenable to change are called “design variables”. The two major aspects of the
design model for which design sensitivity is considered herein are:
a. Design objective
b. Design model response
As a result, the design sensitivity analysis capability in Marc is currently limited to finite element models
of structures with linear response in the computation of
1. Gradients for
a. An objective function (or the design objective), if one is defined by you (for example,
minimizing the material volume in the model).
b. Various types of design model responses under multiple cases of static mechanical loading, or
free vibration, in linear behavior.
2. Element contributions to the responses of the model.
The gradient of the objective function or of a response function is simply the set of derivatives of such a
function with respect to each of the design variables, at a given point in the design space (that is, for a
given design).
For sensitivity analysis to proceed, the design model, the analysis requirements, the design variables, and
the functions for which the gradients are to be found have to be specified by you.
The existing design sensitivity analysis capability in Marc can be applied in one of two ways:
1. As a stand-alone feature, where you are concerned only with obtaining sensitivity analysis results.
Such an application is completed with the output of the sensitivity analysis results.
CHAPTER 5 203
2. Within a design optimization process, where you are concerned mainly with the optimization of
a design objective related to a finite element model. This type of an application of sensitivity
analysis is transparent to you. The design optimization process is completed with the output of
design optimization related data, such as the optimized objective function, related values of the
design variables, and the analysis results for the optimized design.
These two procedures are described below.
1. Sensitivity analysis as a stand-alone feature
This type of solution usually aims at obtaining the derivatives of user prescribed response
quantities at a given design, with respect to each of multiple design variables specified by the
user. This set of derivatives is therefore the gradient of the response function at the given design
in the design variable space (or, in short, in the design space). For example, for a prescribed
response function r, given the design variables x 1 , x 2 , and x 3 , the gradient is defined as
dr dr dr T
∇r = --------- --------- --------- (5-228)
˜ dx 1 dx 2 dx 3
The number of response quantities for which gradients are computed is limited either by the
program default or by a user-specified number. If you are interested in obtaining the sensitivity
analysis results in order of criticality, the option to sort them in this order is also available. The
responses are currently prescribed as constraints with user-defined bounds. If sorting is not
required, the bounds can be mostly arbitrary, although they still have to conform to the type of
constraint prescribed. However, if sorting is required and is to be meaningful, it is important for
you to give realistic bounds on the response.
Element contributions to each response quantity are obtained as a by-product of the type of
response sensitivity analysis capability in Marc. Thus, the response r can be represented as a sum
of these element contributions:
N
r =  re + C (5-229)
e = 1
where the second term, C , involves work done elsewhere, such as in support settlement, if any.
This is helpful for a visual understanding of which regions of the model contribute in what
manner to each of the response quantities at the given design, since it can be plotted in a manner
similar to, say stress contours.
Finally, as an option, if you also prescribe an objective function, the gradient of the objective
function with respect to the design variables is also computed at the given design. Thus, for the
objective function W , Marc obtains
dW dW dW T
∇W = --------- --------- … --------- (5-230)
˜ dx 1 x 2 dx n
2. Sensitivity analysis within a design optimization process:

The design optimization algorithm in Marc requires the utilization of gradients of the objective
function and of the constraint functions, which are very closely related to the response functions.
The current algorithm, described under “design optimization”, ignores your initial prescribed
design, but instead begins by generating other designs within the prescribed bounds for the design
variables. Once the optimization algorithm is completed and the optimized design is available, if
a sensitivity analysis is required at the optimized state, it will be necessary for you to modify the
model accordingly and to use sensitivity analysis as a stand alone feature.
During design optimization, sensitivity analysis is performed for a maximum number of
constraints either indicated by the program default or prescribed by you.
Theoretical Considerations
The method currently employed in Marc for response sensitivity analysis is the “virtual load” method.
For sensitivity analysis of the objective function, finite differencing on the design variables is
performed directly.
In the virtual load method, first a design is analyzed for the user-prescribed load cases, and, if also
prescribed, for eigenfrequency response. The response of the structure having been evaluated for each of
these analyses, the response quantities for which sensitivity analysis is to be performed are then decided
upon and collected in a database. In sequence, a virtual load case is generated for each such response
quantity. Reanalysis for a virtual load case leads to virtual displacements.
The principle of virtual work is then invoked. This defines the element contributed part of the response
quantity, for which the virtual load was applied, as a dot product of the structural displacement vector,
for the actual load case with which the response is associated, and the virtual load vector itself.
The jth response r j can be expressed as the dot product of the actual load vector with the virtual
displacement vector as
T
rj = p uv (5-231)
By differentiation of this expression, you can show that the derivative of the response r j with respect to
a given design variable x i is given by
T
dr dp v
- u + u v  -------- – -------- u
T dp dK
--------j = ---------
dx i dx i  dx i dx i 
where K is the stiffness matrix of the structure. The response derivative above is evaluated on the
element basis as:
T
dr dp v
- u + u v  -------- – -------- u
T dp dK
--------j =  --------- (5-232)
dx i dx i  dx i dx i 
e e
CHAPTER 5 205
where the vectors u and p v are the vector of element nodal displacements due to the actual load case
and the vector of element nodal forces due to the virtual load, respectively. The case of eigenfrequencies
follows the same logic except that an explicit solution for the virtual load case is not necessary. The
derivatives are now evaluated at the element level via finite differencing. This is known as the
semi-analytical approach.
Note that the derivative expression for the virtual load method is quite similar to that for the Adjoint
Variable method. In fact, while they are conceptually different approaches, for certain cases they reduce
to exactly the same expressions. However, for certain other cases, the terms take on different meanings
although the end result is the same.
Design Optimization
Design optimization refers to the process of attempting to arrive at certain ideal design parameters,
which, when used within the model, satisfy prescribed conditions regarding the performance of the
design and at the same time minimize (or maximize) a measurable aspect of the design. In Marc, you can
ask to minimize
1. total material volume
2. total material mass
3. total material cost.
When there is more than one material in the finite element model, the specification of different mass
densities and unit costs are taken into account in the computation of the objective function.
The performance requirements might not necessarily have to be related to response, but also to different
concepts such as packaging, design envelope, even maintenance. The current capability is based on
optimization with constraints on response. Also, the lower and upper bounds on the design variables
themselves define the limits of design modifications. Hence, the design optimization problem can be
posed mathematically in the following format:
Minimize W
Subject to c j ≥ 0.0 j = 1, …, m
where W is the objective function, and c j is the jth constraint function, either specified as an inequality
related to a response quantity or as a limit on a design variable. For example, to limit the x-direction
translation component at a node k, the constraint can first be written as
( u x ) k ≤ u x*k
Assuming that the displacement is constrained for positive values, the normalized constraint expression
c (dropping the subscript j ) becomes:
( u∗ – u ) ⁄ u∗ ≥ 0.0 (5-233)
with its derivatives as:
dc – 1 du
------ = ------ ------ (5-234)
dx u∗ dx
Within Marc, the constraints can be imposed on strain, stress, displacement, and eigenfrequency response
quantities. For stresses and strains, the constraints are defined as being on the elements, and for
displacements, the constraints are defined as being at nodes. Stress and strain components, as well as
various functions of these components (the von Mises equivalent stress and principal stresses, stresses on
prescribed planes) and generalized stress quantities can be constrained. Similarly, translation and rotation
components of displacement, resultant and directed displacements as well as relative displacements
between nodes can be constrained. For free vibration response, constraints can be placed on the modal
frequencies as well as on differences between modal frequencies. A full listing of such constraints are
given in Marc Volume C: Program Input.
The upper and lower bounds on the design variables are posed as
xi ≤ xu and xi ≥ xl (5-235)
after which they can be transformed into expressions similar to Equation (5-233).
The response quantities associated with the model are implicit functions of the design variables. Analyses
at sample design points are used to build explicit approximations to the actual functions over the design
space. This approach minimizes the number of full scale analyses for problems which require long
analysis times such as for nonlinear behavior. This method is summarized next.
Approximation of Response Functions Over the Design Space

The design space for the optimization problem is bounded by limits on the design variables of a model
to be optimized. The simplest case is that of a single design variable, where the design space is a
straight line, bounded at the two ends. For higher number of design variables, say n, the design space can
2
be visualized as a bounded hyperprism with n vertices. For such a construct, you can build
approximations to the constraint functions by way of analyses conducted at the vertices. However, this
2
requires n analyses.
We now note that the minimum geometrical construct spanning n-dimensional space is a simplex with
n + 1 vertices. Thus, an approximation based on analyses at vertices requires only ( n + 1 ) analyses.
The simplex has already been used for first order response surface fitting based on only function values
[Ref. 3]. However, the use here involves higher order response functions.
Like the hyperprism, in one-dimension, the simplex degenerates into a straight line. However, in two
dimensions it is a triangle, and in general it is a hyper-tetrahedron. Figure 5-34 compares the simplex to
the hyperprism in normalized two-dimensional design space.
CHAPTER 5 207
n=2
Hyperprism: 22 = 4 vertices
Simplex : 2 + 1 = 3 vertices
Figure 5-34 Comparison of the Simplex to the Hyperprism in Two-dimensions
While the orientation of the simplex in the design space appears to be a relatively arbitrary matter, once
an origin and the size of the simplex are prescribed, a simple formula will locate all vertices of a simplex
in n-dimensions [Ref. 4]. The response gradient information at the simplex vertices is combined with the
function values to achieve enhanced accuracy. Thus, an analysis at a vertex can be utilized to yield
both response function values and, by way of sensitivity analysis, the response gradients at that vertex.
The virtual load method employed in Marc for obtaining the response gradients is discussed under
Design Sensitivity Analysis. The response gradients can then easily be converted to constraint gradients
for use in an optimization algorithm.
As a result, the results of an analysis at a vertex of the simplex can be summarized as the vector of
constraint function values c j , and the gradient, ∇c j , of each constraint function ( j ) at that vertex.
For the case of a one variable problem, the results of analyses at the two vertices are symbolized in
Figure 5-35, for a hypothetical constraint c j .
dc j1
------------ E1
dx
Possible Actual Function
c j1
2
1 c j2
E2 dc j2
x ------------
dx
Figure 5-35 Vertex Results for One-dimensional Design Space
From Figure 5-35, it appears almost natural to fit a cubic function to the four end conditions (two function
values and two slopes) depicted in the figure. However, this approach is too rigid, and is not easily
generalizable to higher dimensions. On the other hand, the use of two quadratics, which are then merged
in a weighted manner gives higher flexibility and potential for increased accuracy.
It can be seen that the two equations, E1 and E2 , are designed such that they both satisfy the function
values at the two vertices, but E1 satisfies the slope at vertex 1 only, and E2 satisfies the slope at vertex
2 only. Finally, at any design point x , the response function c j can be approximated as
c j = ( W1 E1 + W2 E1 ) ⁄ ( W1 + W2 ) (5-236)
where the weight functions W1 and W2 are normally functions of x and possibly some
other parameters.
This type of approach has the further advantage that it is consistent with the use of the simplex for
determining analysis points and approximating constraint functions in the higher dimensional cases.
Therefore, for an n-dimensional problem, the simplex having n + 1 vertices, each equation needs to have
( n + 1 ) + n = 2n + 1 unknown parameters. The general quadratic polynomial without the cross-
coupling terms satisfies this condition for all n .
Improvement of the Approximation

When approximations are used, the results of an optimization process need to be checked by means of a
detailed analysis. As a result, the approximations can be adjusted and the optimization algorithm can be
reapplied. Depending on how accurate the approximations prove to be and how many more detailed
analyses are acceptable to you, this process can be applied for a number of cycles in order to improve
upon the results.
The Optimization Algorithm

The optimization algorithm implemented in Marc is the Sequential Quadratic Programming method
[Ref. 5]. This method is employed using the approximation described above. By obtaining response
function and gradient values from the approximate equations, the need for full scale analyses is reduced.
The method is summarized below.
The quadratic programming technique is based on the approximation of the objective function by a
quadratic function. When nonlinear constraints exist, as is the case in most practical design optimization
problems, the second order approximation concept is extended to the Lagrangian which is a linear
combination of the objective function and the constraint functions. The solution method for a quadratic
programming problem with nonlinear constraints can be summarized as the following:
At each step, modify the design variables vector x by adding a scaled vector
x ← x + qs (5-237)
CHAPTER 5 209
where s is a search direction and q is the scale factor along the search direction. If the search direction
has been determined, the scale factor can be found by some type of line search along the search direction.
The determination of the search direction constitutes the major undertaking in the quadratic
programming method. If H is the Hessian of the Lagrangian and g is the gradient of the objective
function, then the search direction s is that which minimizes the function
T T
g s + ( s Hs ) ⁄ 2 (5-238)
subject to the linearized constraints
Js ≥ – c (5-239)
where J is the Jacobian matrix of the constraints and c is the vector of constraint functions at the current
iteration. Due to lack of knowledge about which constraints will be active at the optimum, the Hessian
of the Lagrangian is not always readily available. Thus, some iterations take the form of a gradient
projection step.
The coefficients of the constraint functions in the Lagrangian are the Lagrange multipliers which are
unknown before solution has started. At the optimum these multipliers are zero for inactive constraints.
Normally, the above problem is solved using only those constraints which appear to be active at or close
to the current design point, with the assumption that these constraints will be active at the optimum. The
selection of these active constraints is done within the framework of an active set strategy, the set being
modified appropriately with the progression of the iterations. Similarly, the arrays g , H , and J are also
modified as the iterations proceed.
Marc User Interface for Sensitivity Analysis and Optimization

This is discussed under Input, Output, and Postprocessing in the following paragraphs.
Input
Input features related to design sensitivity and design optimization are similar. However, they differ in
the parameter data blocks and in the optional specification of an objective function for the case of
sensitivity analysis. Therefore, other than these two items, all discussion of input should be construed to
refer to both procedures equally. All design sensitivity and design optimization related parameters and
options in a Marc input file start with the word DESIGN.
All load cases and any eigenfrequency analysis request associated with sensitivity analysis or
optimization should be input as load increments in the history definition section after the END OPTION
line. The maximum number of nodes with point loads or the maximum number of distributed load cases
should be defined in the DIST LOADS parameter.
The type of solution requested from Marc can simply be indicated by the parameter input DESIGN
SENSITIVITY or DESIGN OPTIMIZATION. Only one of these lines can appear in the input. These
parameters also let you change the sorting of constraints, the maximum number of constraints in the
active set, and the maximum number of cycles of design optimization. The first two items are discussed
in previous sections. A cycle of design optimization refers to a complete solution employing the
sequential quadratic programming technique followed by a detailed finite element analysis to accurately
evaluate the new design point reached by way of approximate solution. If required, the approximation is
improved using the results of the last finite element analysis, and a new cycle is started.
The specification of an objective function, being optional for design sensitivity, is made by use of the
model definition option DESIGN OBJECTIVE. This allows you to choose from one of the available
design objectives.
The DESIGN VARIABLES option allows you to specify the design variables associated with the finite
element model. You have a choice of three major types of design variables:
1. Element cross-sectional properties as given by way of the GEOMETRY option.
2. Material properties as given by ISOTROPIC or ORTHOTROPIC options.
3. Composite element properties of layer thickness and ply angle, as given by the
COMPOSITE option.
The properties which are supported are listed under the DESIGN VARIABLES option. The relevant
elements, for which cross-sectional properties can be specified as design variables, each has a section in
Marc Volume B: Element Library, describing which properties, if any, can be posed as design variables
for that element.
Similarly, the list of material properties (currently all related to linear behavior) that can be design
variables is given under the DESIGN VARIABLES option. For composite groups, the layer thicknesses
and ply angles can be given as design variables for composite groups.
Design variables can be linked across finite element entities such that a given design variable controls
several entities. An example is the linking of the thicknesses of several plane stress elements by means
of a single design variable. Thus, when this variable changes, the thicknesses of all linked elements
reflect this change. On the other hand, for the unlinked case, all thicknesses are associated with separate
design variables. This feature is controlled by the LINKED and UNLINKED commands.
Response quantities, or constraints on response quantities, are specified by means of the DESIGN
DISPLACEMENT CONSTRAINTS, DESIGN STRAIN CONSTRAINTS, DESIGN STRESS
CONSTRAINTS, and DESIGN FREQUENCY CONSTRAINTS options. There is no limit to the number of
constraints. Displacement constraints are posed at nodes or groups of nodes, while the strain and stress
constraints are posed over elements or groups of elements, and frequency constraints are posed for free
vibration modes. A complete list can be found under the above mentioned options in Marc Volume C:
Program Input.
When strain or stress constraints are prescribed, it is useful to know that the program evaluates such
constraints at all integration points of all layers of an element and proceeds to consider the most critical
integration point at the related layer for the element. Thus, a strain or stress constraint on an element
normally refers to the most critical value the constraint can attain within the element. During
optimization, the most critical location within the element may change and any necessary adjustment
takes place internally.
For certain responses, the limiting values can be the same in absolute value for both the positive and
negative values of the response. For constraints on such response functions, you have the choice of
prescribing either separate constraints on the positive and negative values, or a combined constraint on
the absolute value. The first approach is more accurate albeit at a higher computational cost.
CHAPTER 5 211
Output
For sensitivity analysis, the output file contains the following information:
• Echo of input, any warnings or error messages.
• A of numbers and definitions for your prescribed design variables. It is important to note that the
variable values are always output in the internal numbering sequence which is defined in this.
Search for the words ‘Design variable definitions’ to reach this in the output file.
• Analysis results for your prescribed design.
• The value of the objective function and its gradient with respect to the design variables
prescribed by you.
• Sensitivity analysis results for the response functions in the default or user-defined set, sorted in
order of criticalness (if specified). It should be noted that although the responses are sorted across
multiple load cases, the sensitivity results are output for each load case. The related output file
consists of a check on the actual response value (obtained by sensitivity analysis) and the
gradient of the response with respect to the design variables. For eigenfrequency results, the
check values on the response can be somewhat more accurate than the results from
eigenfrequency analysis since the latter is iterative, but the check uses the Rayleigh quotient on
top of the iteration results. A constraint reference number allows you to track the sensitivity
analysis results plots when postprocessing. Search for the word ‘Sensitivity’ to reach this output.
For design optimization, the following is written into the output file:
• Echo of input, any warnings or error messages.
• Certain indications that some analyses are being done, but no analysis results except for the ‘best’
design reached during optimization.
• The objective function values and the vector of design variable values:
• At each of the simplex vertices.
• For the starting simplex vertex (from this point on, the information also includes whether the
design is feasible or not).
• At the end of each cycle of optimization.
• For the ‘best’ design found.
• Analysis results for the ‘best’ design.
Postprocessing
This requires that you ask the program to create a post file. The following plots can be obtained by way
of postprocessing.
For sensitivity analysis:
• Bar charts for gradients of response quantities with respect to the design variables.
• Contour plots of element contributions to response quantities. Thus, a finite element model
contour plot gives the element contributions to a specific response quantity which was posed in
the form of a constraint in the input file.
The increment number of the sensitivity analysis results is the highest increment number available in the
post file. The information for each response quantity is written out as for a subincrement. The zeroth
subincrement corresponds to the objective function information. Numbers for the other subincrements
correspond to the constraint reference number(s) in the output file.
For design optimization:
• Path plots showing the variation of the objective function and of the design variables over the
history of the optimization cycles. The best design (feasible or not) is not necessarily the last
design point in the plot. The values at the starting vertex are considered as belonging to the zeroth
subincrement. Each optimization cycle is then another subincrement. The increment number
corresponding to these subincrements is taken as zero. The analysis results for the ‘best’ design
start from increment one.
• Bar charts where each chart gives the values of the design variables at the optimization cycle
corresponding to that bar chart.
Defined Initial State with Result Data from Previous Analysis

(including AXITO3D)
In many cases, it is necessary to analyze a nonlinear process in several stages. Each stage may involve
different contact bodies and boundary conditions, but history data such as temperature, displacement,
stress and strain have to be carried over for contact bodies to be passed from one stage to another. In
Marc, it is possible to start a numerical simulation as a two-dimensional axisymmetric or plane strain
problem and switch to a full three-dimensional analysis in subsequent stages. This makes multiple stage
analysis more efficient. The PRE STATE option is designed to read result data from Marc and use the
data as initial conditions in the new analysis. For contact analysis, it allows the selection of contact
bodies by names. The PRE STATE option includes the capability of the AXITO3D option available in
the 2003 release.
The methodology of the PRE STATE option can be outlined as the following:
• Scan through the result file for the desired time step or time instant.
• Collect the list of nodes and elements corresponding to the contact body transfer list.
• Collect node and element data as required (see the PRE STATE option in Marc Volume C:
Program Input for the data type).
• Copy data to the new model as the initial conditions with 2-D to 3-D expansion if applicable.
The PRE STATE option takes several steps:
• Run first stage analysis.
• Open the result file.
• Extract and expand (if 2-D to 3-D is required) the mesh from the previous result file at the
increment. The expansions of boundary conditions can also be done through Marc Mentat.
• Build second stage model based on the mesh extracted and add other contact bodies and
conditions.
• Specify the PRE STATE option under initial condition menu.
• Run 3-D analysis.
CHAPTER 5 213
Most quadrilateral or hexahedral elements (including Herrmann elements) as well as commonly most
available materials (such as metal and rubber) can be used in the PRE STATE option. For large
deformation problems, either total or updated Lagrangian formulation can be used. Thermal and
dynamic effects can also be included.
For more detailed information on this feature, see the PRE STATE model definition option, in Marc
Volume C: Program Input.
Load and Displacement Boundary Conditions Transfer for

PRE STATE Option
In general, the PRE STATE option can only transfer history data from the result file as the initial
conditions in the new model. Users are responsible for reapplying the boundary and load conditions as
well as the material properties. If users use Marc Mentat, the boundary conditions defined in the previous
2-D analysis automatically expand to 3-D when the mesh is expanded from 2-D to 3-D. However, if a
load function is used and is required in the new model, curve shifting is required. That is, if, in the
previous model, a load is applied from 0 to 10 in the new model based on the last step of the previous
analysis, the load is applied starting from 10. For prescribed displacement, as nodal positions are updated
to the final configuration in the previous analysis, the load function should start from 0 in the new model.
Steady State Rolling Analysis

Rolling contact analysis of a cylindrical deformable body in a Lagrangian framework can be
computationally expensive because it may require not only time-dependent transient process but also a
fine mesh in the entire body to accurately capture contact characteristics. However, some problems
involve only fully axisymmetric structures with constant moving/spinning velocities. These problems
can be considered steady state if a reference configuration, which moves with the body but does not spin
around the rolling axis, is used.
Marc provides the capability of steady state rolling analysis. The feature is characterized by a mixed
Eulerian/Lagrangian formulation with inertia effects in spinning/cornering deformable bodies. Using a
non-spinning reference frame attached to the wheel axel, the analysis becomes purely space dependent.
It presents a better alternative to the unnecessary computational burden of arriving at a steady state
condition through a transient analysis. Furthermore, a finer mesh only needs to be used in the contact
region as opposed to the entire rolling surface.
Kinematics
We consider the axisymmetric body shown in Figure 5-36. The body spins at an angular velocity ω s
around the axisymmetric axis T s at point P s and, simultaneously rotates with a cornering angular
velocity ω c around an axis T c at point P c . Assume that a particle in the body has a location P 0 at time
t = 0 . At time t , its motion contains three parts:
1. from P 0 to a location X because of the spinning
2. from X to Y because of a deformation Y = D ( X ) , where D is time independent function

resulting from the steady state condition
3. from Y to Z because of the cornering.
Ps
Ts
ωs
Tc
ωc
Pc
Figure 5-36 Kinematics
The three motions can be described as
X = Rs ⋅ ( Po – Ps ) + Ps (5-240)
Y = D(X) (5-241)
Z = Rc ⋅ ( Y – Pc ) + Pc (5-242)
with
R s = exp ( ω s t ) and R c = exp ( ω c t ) (5-243)
In Equation (5-243), ω̃ s and ω c are the skew-symmetric tensors associated with the rotation vectors
ω s T s and ω c T c , respectively, with
ω s ⋅ r = ω s T c × r and ω c ⋅ r = ω c T c × r (5-244)
for any vector r .

Time derivative of Equation (5-243) gives
· ·
R s = ω s ⋅ R s and R c = ω c ⋅ R c (5-245)
Making use of Equations (5-240), (5-241), (5-244), and (5-245), the velocity of the particle can be
obtained by the first time derivative of Equation (5-242) as
· ∂D
Z = R c ⋅ ω c T c × ( Y – P c ) + ω s ------- (5-246)
∂α
where α = ω s t , is the spinning angle.
Similarly, the acceleration is obtained by the second derivative of Equation (5-242) with respect to time:
CHAPTER 5 215
·· ∂D ∂2D
Z = R c ⋅ ω c2 ( T c ⊗ T c – 1 ) ⋅ ( Y – P c ) + 2ω s ω c T c × ------- + ω s2 ---------- (5-247)
∂α ∂α 2
where ⊗ denotes the tensor product and 1 is the unit tensor.
Transformation of Equations (5-246) and (5-247) into the reference configuration defined by X by
premultiplying R cT , gives
∂D
v = ω c T c × ( Y – P c ) + ω s ------- (5-248)
∂α
∂D ∂2D
a = ω c2 ( T c ⊗ T c – 1 ) ⋅ ( Y – P c ) + 2ω s ω c T c × ------- + ω s2 ---------- (5-249)
∂α ∂α 2
where v and a are velocity and acceleration of the particle with respect to the reference frame.
Inertia Effect
The contribution of inertia effect into the right-hand-side of the system equation can be calculated using
the weak form
δπ =  ρa ⋅ δu dv
v
∂D
= – ρω c2  ( T c ⊗ T c – 1 ) ⋅ ( Y – P c ) ⋅ δu dv – 2ρω s ω c  T c × ------- ⋅ δu dv (5-250)
∂α
V V
∂D ∂δu
+ ρω s2  ------- ⋅ --------- dv
∂α ∂α
V
where ρ is the density, v is the volume, and u is the displacement.

Linearization of Equation (5-250) gives
∂Δu
Δδπ = – ρω c2  Δu ⋅ ( T c ⊗ T c – 1 ) ⋅ δu dv – 2ρω s ω c  T c × ---------- ⋅ δu dv
∂α
v V
(5-251)
∂Δu ∂δu
+ ρω s2  ---------- ⋅ --------- dv
∂α ∂α
v
which can be used to calculate the contribution of the inertia effect to stiffness matrix.
Rolling Contact
To take into account the rolling effect in contact, the velocity vector in Equation (5-248) is decomposed
into a normal and a tangential component, with respect to the contact surface, for all nodes in contact with
ground. The normal component is then forced to become zero because of the contact conditions. The
relative slipping velocity used in friction calculation is the difference between the tangential component
and the ground moving velocity.
Steady State Rolling with Marc

The capability of steady state rolling analysis in Marc has taken into account the effects of rolling
frictions and the inertia effects resulting from both spinning and cornering. The deformable rolling
body can contact with multiple, flat or nonflat rigid surfaces. A typical example is a tire model which
is in contact with a rigid rim and a rigid road surface. The spinning effects are involved only for
selected contact body pair (for example, tire/road) which is defined with the SS-ROLLING history model
definition option. The steady state rolling analysis can follow a static stress analysis with various
loadcases and can also combine with any steady state loads. The feature is available for 3-D analysis only.
The element types supported for steady state rolling analysis include 7, 9, 18, 21, 35, 57, 61, 84, 117, 120,
146, 147, and 148.
The friction type supported in contact for steady state rolling analysis, based on nodal force, is Coulomb
friction for rolling.
Because the system matrix becomes nonsymmetric in steady state rolling analysis, it is recommended
that the multifrontal direct sparse solver be used.
The feature requires the so-called streamline along the circumferential direction of the spinning body.
Therefore, the 3-D mesh must be generated by revolving its corresponding axisymmetric mesh. For a 3-D
brick element, the circumferential direction must be from the element face defined by nodes 1-2-3-4 to
the element face defined by nodes 5-6-7-8.
Below is a list of options required in a steady state analysis, in addition to the others in a standard static
stress analysis.
SS-ROLLING: Parameter activating the steady state rolling analysis
ROTATION A: Model definition option defining the spinning axis
CORNERING AXIS: Model definition option defining the cornering axis
SS-ROLLING: History definition option defining the spinning body, ground body,
spinning body motion (spinning/cornering/moving relative to the ground)
and others. All velocities/forces defined in the option are total values.
Structural Zooming Analysis

Local variations, such as the changes in model geometries or in the degrees of finite element refinement
to achieve a better evaluation of the local gradients in the solution, often need a complete re-analysis of
the entire model. However, in cases that these local changes have negligible influence on the solution
a certain distance away from the changes, it is computationally more efficient to model only the part
CHAPTER 5 217
with the local changes. It can be realized by applying the existing loads or/and boundary conditions in
the local model along with properly defined kinematic conditions to the local boundaries connecting to
the global model.
A typical structural zooming analysis contains two steps:
1. Global run to obtain a post file containing global results.
2. Local run to define kinematic boundary conditions in the local model and to obtain refined results
in the local model.
This procedure can be repeated as many times as desired. Any local analysis can be the global analysis
of next level refinement.
The GLOBALLOCAL option (see Marc Volume C: Program Input for details) is used in the input of the
local run to define the list of nodes connecting to the global model. Marc calculates the deformation
(temperature) history of these nodes, based on their locations in the global model and on the solution of
the global analysis. The obtained deformation (temperature) history is then applied to the nodes as
prescribed kinematic boundary conditions. The detailed steps include:
a. Reading in the GLOBALLOCAL option to get the list of connecting nodes.
b. Reading in the global model (element types, node coordinates, element connectivity, thickness
if shell elements) and solution of the model from the generated post file in global run.
c. Finding out the locations of local connecting nodes in the global model (the element each node
is associated with and its isoparametric location within the element).
d. Calculation of the deformation (temperature) of each connecting node for every increment
available in the global post file, based on its location in the global elements associated using
the interpolation techniques, and storage of the deformation (temperature) history in the format
of time-dependent tables.
e. Applying the deformation (temperature) history of the connecting nodes to the local model as
prescribed kinematic boundary conditions.
All the steps above are performed automatically inside Marc once the GLOBALLOCAL option is used in
the input file of the local run.
If the global adaptive meshing is used in the first analysis, then the GLK model definition should also be
used.
Element Types Supported

The global to local modeling can be used in the following four cases:
Global Model Local Model
2-D Solid 2-D Solid
3-D Solid 3-D Solid
3-D Shell/Membrane 3-D Shell/Membrane
3-D Shell/Membrane 3-D Solid
One can use composite continuum elements in the local model if the global model had composite
continuum elements with the same number of layers or was composed of shell elements
See Figure 5-37 for examples from 2-D solid elements to 2-D solid elements and from 3-D shell elements
to 3-D solid elements.
global model
local model
cracks
Cylinder
(a) 2-D Solid to 2-D Solid (b) 3-D Shell to 3-D Solid
Figure 5-37 Structural Zooming

CHAPTER 5 219
Uncoupled Thermal Stress Analysis

5 A typical chaining analysis in Marc for thermal stress problems includes two steps:
• Heat transfer analysis to generate a post file containing temperature distribution on the model and
history over the time period considered.
• Thermal stress analysis based on the temperature distribution/history obtained from the previous
heat transfer analysis
Structur To perform a chaining analysis for a thermal stress problem two options can be used in Marc. They are
al the CHANGE STATE option and the MAP TEMP option.
Procedu
Using the CHANGE STATE option, the mesh in stress analysis has to be the same as that used in the heat
re
transfer analysis. The structural element type has to correspond with the heat transfer element type.
Library
Please see Marc Volume B: Element Library for this correspondence. The basic idea of the option is to
read in the integration point temperatures for each element from the post file of the heat transfer analysis
and use them directly in the stress analysis. No spatial data mapping is required. Since this option is
integration point based, INITIAL STATE should be used if one needs to specify initial temperatures in the
stress analysis. This option can be used with AUTO THERM, AUTO THERM CREEP, and AUTO STEP.
The data mapping used with the MAP TEMP option contains the following steps:
1. Read in the MAP TEMP option to get the list of structural nodes. The temperature history of the
nodes is expected to be obtained from the heat transfer post file.
2. Read in heat transfer model (element types, node coordinates, element connectivity, etc.), and
solution of the model in the heat transfer analysis.
3. Find the location of the structural nodes in the heat transfer model (the element each node
associated and its isoparametric location within the element).
4. Calculate the temperature of each structural node for every increment available in the post file of
the heat transfer analysis, based on its location in the heat transfer element associated. Store the
temperature history in the format of a time-dependent table.
5. Apply the temperature history to the structural node in stress analysis.
Since this option is node-based, INITIAL TEMP should be used if one needs to specify initial temperatures
in the stress analysis. This option cannot be used with AUTO THERM and AUTO THERM CREEP. It can
be used with AUTO STEP.
Compared to the CHANGE STATE option, the MAP TEMP option provides more flexibility. It allows a
temperature distribution/history obtained from a previously heat transfer simulation to be used in the
stress analysis even though the meshes are dissimilar. The element types or even element classes used in
heat transfer and stress analyses can be different. The model or the time period covered in heat transfer
and stress analyses can also be different. For time period outside the range in heat transfer analysis, a
time shift or end values or extrapolation option can be used. For the part of the structural model or time
period not covered by the post file of the heat transfer analysis, POINT TEMP can always be used to
directly input nodal temperatures.
Cure-Thermal-Mechanically Coupled Analysis

When resin is used in the manufacturing of composite parts, the curing process during and after the
forming processes has a significant effect on the thermal and mechanical behavior of the formed part.
The shrinkage caused by curing may severely distort the final geometry of the part. Marc has
implemented a capability that allows you to perform cure-thermal-mechanical coupled analysis to predict
this behavior.
The curing analysis is incorporated into the existing staggered coupled thermal-mechanical analysis
procedure (see Thermal Mechanically Coupled Analysis in Chapter 6: Nonstructural and Coupled
Procedure Library). Before the heat transfer analysis pass takes place, the curing analysis is performed
based on the estimated temperatures at the beginning and end of the increment. The cure rate is then
calculated according to the cure kinetics of the resin materials. Using the cure rate, a heat flux due to cure
reaction heat generation is calculated and added into the heat transfer system of equations. Volumetric
curing reaction heat flux is calculated according to the cure rate
dα
Q c = ------- ( 1 – V f )ρ r H r (5-252)
dt
where:
Hr – resin cure reaction heat
ρr – resin density
α – resin degree of cure

Vf – fiber volume fraction
The governing matrix equation for cure-thermal coupled analysis can be expressed as:
C ( T )T· + K ( T )T = Q + Q I + Q F + Q C (5-253)
In Equation (5-253), C ( T ) and K ( T ) are the temperature-dependent heat capacity and thermal
conductivity matrices, respectively. T is the nodal temperature vector. T· is the time derivative of the
temperature vector, Q is the external heat flux vector, Q I is the heat flux vector due to plastic work, Q F
represents the heat generated due to friction, and Q C is the heat generated due to curing.
With the CURING parameter included in the input data, Marc first activates the pass to calculate the
degree of cure and curing reaction heat flux at the beginning of each cycle of heat transfer analysis.
Then, Marc considers the cure shrinkage strain in the mechanical pass of the thermal-mechanical coupled
analysis.
To include curing into a heat transfer or thermal-mechanical coupled analysis, it is necessary to add the
CURING parameter into the input data. In addition, the CURE RATE model definition option is also
needed to define the curing properties, usually called cure kinetics, of resin materials as one kind of
property for heat transfer analysis.
CHAPTER 5 221
To include cure shrinkage strain into the mechanical pass of a thermal-mechanical coupled analysis, in
addition to having both the CURING parameter and the CURE RATE model definition option in the input
data, the CURE SHRINKAGE model definition must be included to define the cure shrinkage properties
of the resin material. Note that the cure shrinkage effect is calculated based on the degree of cure, so the
cure shrinkage effect is ignored if the degree of cure is not calculated during the heat transfer pass.
Cure Kinetics
In Marc, the cure kinetics defines the cure rate as a function of the degree of cure and the temperature of
resin materials. There are different approaches to define the cure kinetics of resin materials: embedded
cure kinetics models, table definition and user subroutine.
In a curing analysis, the cure rate is calculated for each time step. Assuming that the cure rate is defined
as the function of the degree of cure and temperature of:
dα
------- = f ( α ,T ) (5-254)
dt
The time integration of the degree of cure is done using backward Euler method:
α ni = Δt ⋅ f ( α ni – 1 ,T ) + α n – 1 (5-255)
where
i – iteration number for cure
n – current increment number
n–1 – previous increment number
f – function defined by cure kinetics model
Δt – time step size of the increment
The four cure kinetics models implemented in Marc are:

Cure kinetics model 1: by Lee, Loos and Springer (1982) [Ref. 43];
Cure kinetics model 2: by Scott(1991) [Ref. 50];
Cure kinetics model 3: by Lee, Chiu, and Lin (1992) [Ref. 44];
Cure kinetics model 4: by Johnston and Hubert (1996) [Ref. 52].
Tables 5-7 and 5-8 summarize the details of these cure kinetics models and the definition of the
required parameters.
Table 5-7 Cure Kinetics Models Embedded in Marc

Model Equations Parameters
Model 1 dα A 1, A 2, A 3, ΔE 1, ΔE 2,
------- = ( K 1 + K 2 α ) ( 1 – α ) ( B – α )
Lee, Loos and Springer dt ΔE 3, B, α C, H R
(1982) [Ref. 43] α ≤ αc
dα
------- = K 3 ( 1 – α ) α > αc
dt
–Δ E ⁄ ( R T )
Ki = Aie i
(Included in Model 2) dα A, ΔE, m, n, H R

------- = Kα m ( 1 – α ) n
Lee, Chiu, and Lin dt
(1992)[Ref. 44]; White and
K = Ae – Δ E ⁄ ( R T )
Hahn (1992) [Ref. 48]
(Included in Model 2) dα A, ΔE, n, H R
------- = K ( 1 – α ) n
Kenny (1992) [Ref. 49]; dt
Scott (1991)[Ref. 50]
K = Ae – Δ E ⁄ ( R T )
Model 2 (combined Model) dα A 1, A 2, ΔE 1, ΔE 2, m,
------- = ( K 1 + K 2 α m ) ( 1 – α ) n
Scott (1991)[Ref. 50] dt n, H R
–Δ E ⁄ ( R T )
Ki = Aie i
Model 3 dα A 1, A 2, ΔE 1, ΔE 2, l,
------- = K 1 ( 1 – α ) l + K 2 α m ( 1 – α ) n
Lee, Chiu, and Lin dt m, n, H R
(1992)[Ref. 44] –Δ E ⁄ ( R T )
Ki = Aie i
Model 4 dα Kα m ( 1 – α ) n A, ΔE, m, n, C, α C 0,
------- = ---------------------------------------------------------------
Johnston and Hubert dt C {α – (α
C0
+α
CT
T )} α C T, H R
1+e
(1995)[Ref. 52]
–Δ E ⁄ ( R T )
Ki = Aie i
CHAPTER 5 223
Table 5-8 Parameters used in the Embedded Resin Cure Kinetics Models
Variable Description Units
α Resin degree of cure. -
T Resin temperature K or R
HR Total resin heat of reaction (α = 0 to 1) J/kg or BTU/lbm
Ai Pre-exponential factor. /s
ΔE i Activation energy. J/mol or BTU/mol
l Equation superscript. -
m Equation superscript. -
n Equation superscript. -
R Gas constant J/(mol K) or BTU/(mol R)
C Diffusion Constant -
αC 0 Critical Resin Degree of Cure -
αC T The Increase in Critical Resin Degree of Cure with

-
Temperature
In Marc, the table definition allows the user to define the cure rate as a function of the degree of cure and
temperature if the resin kinetics is not defined by the four embedded models.
In addition, the degree of cure and cure rate can also be calculated through the UCURE user subroutine.
Cure Shrinkage Strain

The cure shrinkage strain is calculated according to the volumetric shrinkage due to curing process. The
S
resin degree of cure shrinkage is defined as the ratio of volumetric cure shrinkage ( V r ) and maximum
S
S∞ Vr
volumetric cure shrinkage ( V r ) of the resin material, as α S = ----------- . Equation (5-256) then
S∞
Vr
calculates the cure shrinkage strain:
S 1⁄3
ε rS = ( 1 + V r ) –1 (5-256)
Considering the anisotropic shrinkage behavior of composite with resin, the strain components of
the composite are calculated by using the directional Cure Shrinkage Coefficient (CSC) matrix by
Equation 5-257.
ε iSj = CS ( C i j ⋅ ε rS ) i, j = 1 ,2 ,3 (5-257)
S
The calculation of volumetric cure shrinkage, V r , is conducted based on the cure shrinkage models
defined by users. Marc has two cure shrinkage models available. Note that a positive cure shrinkage
strain indicates a reduction of volume.
Tables 5-9 and 5-10 summarize the details of the cure kinetics models and the definition of the requires
parameters.
Table 5-9 Resin Cure Shrinkage Models
Model Equations Parameters
Model 1 S α < αC 1 S
V r = 0.0 V r , α C 1, α C 2, A
Bogetti and
S S∞ 2
Gillespie Vr = A * αS + ( Vr – A ) * αS αC 1 ≤ α < αC 2
(1992) S S∞
[Ref. 42] Vr = Vr
α ≥ αC 2
α – αC 1
α S – ---------------------------
αC 2 – αC 1
Model 2 B ( α – αc )
V r = V rS ∞ *10 S∞
S
α ≤ αC V r , α C, B
White and
Hahn 1992 S S∞ α > αC
Vr = Vr
[Ref. 48]
Table 5-10 Parameters used in the Resin Cure Shrinkage Models

Variable Description Units
S -
Vr Resin volumetric cure shrinkage
S∞ -
Vr Total volumetric resin shrinkage from = 0 to 1.
αs The degree of cure shrinkage. -
αC 1 Degree of cure after which the resin shrinkage begins (model 1). -
αC 2 Degree of cure after which the resin shrinkage stops (model 1). -
A Linear cure shrinkage coefficient -
αC Degree of cure after which the resin shrinkage begins (model 2). -
B Cure shrinkage model superscript. -

CHAPTER 5 225
References
1. Key, S. W. and R. D. Krieg, 1982, “On the Numerical Implementation of Inelastic
Time-Dependent, Finite Strain Constitutive Equations in Structural Mechanics”, Computer
Methods in Applied Mechanics in Engineering, V.33, pp. 439–452, 1982.
2. Bathe, K. J., E. Ramm, and E. L. Wilson. “Finite Element formulation for Large Deformation
Dynamic Analyses”, International Journal for Numerical Methods in Engineering, V. 9, pp. 353-
386, 1975.
3. Montgomery, D. C., Design and Analysis of Experiments, (2nd ed.) John Wiley and Sons, 1984.
4. Spendley, W., Hext, G. R., Himsworth, F. R., “Sequential Application of Simplex Designs in
Optimisation and Evolutionary Operation”, Technometrics, Vol. 4, No. 4,
pp. 441-461 (1962).
5. Gill, P. E., Murray, W., Saunders, M. A., Wright, M. H., “Sequential Quadratic Programming
Methods for Nonlinear Programming”, NATO-NSF-ARO, Advanced Study Inst. on Computer
Aided Analysis and Optimization of Mechanical System Dynamics, Iowa City, IA, Aug. 1-2,
1983, pp. 679-700.
6. Bathe, K. J. Finite Element Procedures Prentice-Hall, 1996.
7. Wilkinson, J. H. The Algebraic Eigenvalue Problem. Oxford: Clarendon Press, 1965.
8. Zienkiewicz, O. C., and Taylor, L. C. The Finite Element Method. Fourth Ed., Vol. 1 & 2.
London: McGraw-Hill, 1989.
9. Barsoum, R. S. “On the Use of Isoparametric Finite Elements in Linear Fracture Mechanics.” Int.
J. Num. Methods in Engr. 10, 1976.
10. Dunham, R. S. and Nickell, R. E. “Finite Element Analysis of Axisymmetric Solids with
Arbitrary Loadings.” No. 67-6. Structural Engineering Laboratory, University of California at
Berkeley, June, 1967.
11. Hibbitt, H. D., Marcal, P. V. and Rice, J. R. “A Finite Element Formulation for Problems of Large
Strain and Large Displacement.” Int. J. Solids Structures 6 1069-1086, 1970.
12. Houbolt, J. C. “A Recurrence Matrix Solution for the Dynamic Response of Elastic Aircraft.” J.
Aero. Sci. 17, 540-550, 1950.
13. McMeeking, R. M., and Rice, J. R. “Finite-Element Formulations for Problems of Large Elastic-
Plastic Deformation.” Int. J. Solids Structures 11, 601-616, 1975.
14. MARC Update. “The Inverse Power Sweep Method in MARC.” U.S. Ed., vol. 3, no. 1, February,
1984.
15. Simo, J. C., Taylor, R. L., and Pister, K. S. “Variational and Projection Methods for the Volume
Constraint in Finite Deformation Elasto-Plasticity,” Comp. Meth. in App. Mech. Engg., 51, 1985.
16. Simo, J. C. and Taylor, R. L. “A Return Mapping Algorithm for Plane Stress Elasto-Plasticity,”
Int. J. of Num. Meth. Engg., V. 22, 1986.
17. Wriggers, P., Eberlein, R., and Reese, S. “A Comparison of Three-Dimensional Continuum and
Shell Elements for Finite Plasticity,” Int. J. Solids & Structures, V. 33, N. 20-22, 1996.
18. Simo, J. C. and Taylor, R. L. “Quasi-Incompressible Finite Elasticity in Principal Stretches,

Continuum Basis and Numerical Algorithms,” Comp. Meth. App. Mech. Engg., 85, 1991.
19. Marcal, P. V. “Finite Element Analysis of Combined Problems of Nonlinear Material and
Geometric Behavior.” in Proceedings of the ASME Computer Conference, Computational
Approaches in Applied Mechanics, Chicago, 1969.
20. Melosh, R. J., and Marcal, P. V. “An Energy Basis for Mesh Refinement of Structural Continua.”
Int. J. Num. Meth. Eng. 11, 1083-1091, 1971.
21. Morman, K. N., Jr., Kao, B. G., and Nagtegaal, J. C. “Finite Element Analysis of Viscoelastic
Elastomeric Structures Vibrating about Nonlinear Statically Stressed Configurations.” SAE
Technical Papers Series 811309, presented at 4th Int. Conference on Vehicle Structural
Mechanics, Detroit, November 18-20, 1981.
22. Morman, K. N., Jr., and Nagtegaal, J. C. “Finite Element Analysis of Small Amplitude Vibrations
in Pre-Stressed Nonlinear Viscoelastic Solids.” Int. J. Num. Meth. Engng, 1983.
23. Nagtegaal, J. C. “Introduction in Geometrically Nonlinear Analysis.” Int. Seminar on New
Developments in the Finite Element Method, Santa Marherita Ligure, Italy, 1980.
24. Nagtegaal, J. C., and de Jong, J. E. “Some Computational Aspects of Elastic-Plastic Large Strain
Analysis.” in Computational Methods in Nonlinear Mechanics, edited by J. T. Oden. North-
Holland Publishing Company, 1980.
25. Newmark, N. M. “A Method of Computation for Structural Dynamics.” ASCE of Eng. Mech. 85,
67-94, 1959.
26. Parks, D. M. “A Stiffness Derivative Finite Element Technique for Determination of Elastic
Crack Tip Stress Intensity Factors.” International Journal of Fractures 10, (4), 487-502,
December 1974.
27. Timoshenko, S., Young, D. H., and Weaver, Jr., W. Vibration Problems in Engineering. John
Wiley, New York: 1979.
28. Wilkinson, J. H. The Algebraic Eigenvalue Problem. Oxford: Clarendon Press, 1965.
29. Chung, J. and Hulbert, G.M., “A family of single-step Houbolt time integration algorithms for
structural dynamics”, Comp. Meth. in App. Mech. Engg., 118, 1994.
30. Shih, C.F., Moran B. and Nakamura K., “Energy release rate along a three-dimensional crack
front in a thermally stressed body”, International Journal of Fracture, vol. 30, pp. 79–102, 1986.
31. Anderson, T.L., Fracture Mechanics: Fundamentals and Applications (2nd ed.) CRC Press, 1995.
32. Yoon, J.W., Yang, D.Y. and Chung, K., “Elasto-plastic finite element method based on
incremental deformation theory and continuum based shell elements for planar anisotropic sheet
materials”, Comp. Methods Appl. Mech. Eng., 174, 23 (1999).
33. Shih, C.F. and Asaro, R.J., “Elastic-Plastic Analysis of Cracks on Bimaterial Interfaces: Part I –
Small Scale Yielding”, Journal of Applied Mechanics, Vol. 110, pp. 299–316, June 1988.
34. Krueger, R., “Virtual Crack Closure Technique: History, Approach and Applications”, Appl.
Mech. Rev., Vol. 57:2, pp. 109–143, March 2004.
35. Knops, B., “Numerical Simulation of Crack Growth in Pressurized Fuselages”, Doctoral thesis,
Faculty of Aerospace Engineering, Delft University of Technology, The Netherlands, 1994.
CHAPTER 5 227
36. Chung, J. and Hulbert, G.M., “A time integration algorithm for structural dynamics with
improved numerical dissipation: The Generalized-α Method”, Journal of Applied Mechanics,
Vol. 60, pp. 371 - 375, June 1993.
37. Hilber, H.M., Hughes, T.J.R., and Taylor, R.L., “Improved Numerical Dissipation for Time
Integration Algorithms in Structural Dynamics”, Earthquake Engineering and Structural
Dynamics, Vol. 5, pp. 283-292, 1977
38. Wood, W.L., Bossak, M., and Zienkiewicz, O.C., “An Alpha Modification of Newmark’s
Method”, International Journal for Numerical Methods in Engineering, Vol., 15, pp. 1562-1566,
1981.
39. J. Reeder, 3D Mixed-Mode Delamination Fracture Criteria - An Experimentalist's Perspective.
Proceedings of American Society for Composites, 21st Annual Technical Conference, Dearborn,
Michigan, 2006.
40. M.L. Benzeggagh and M. Kenane, Measurement of Mixed-Mode Delamination Fracture
Toughness of Unidirectional Glass/Epoxy Composites with Mixed-Mode Bending Apparatus,
Composites Science and Technology, vol. 56, pp. 439-449, 1996.
41. Yih-Farn Chen, Modeling Analysis of Defect Formation during Thick Laminate Compression
Molding Process, Presented at the American Helicopter Society 58th Annual Forum, Montreal,
Canada, June 11-13, 2002.
42. Bogetti, T. A., Gillespie, J. W., 1992 Process-Induced Stress and Deformaton in Thick Section
thermoset Composite Laminates, Journal of Composite Materials, 26, 626-660.
43. Lee, W.I., A. Loos and G.S. Springer. 1982 Cure Kinetics and Viscosity of Fiberite 976 Resin,
Journal of Composite Materials, 21; 243-261.
44. S.N. Lee, M.T. Chiu, and H.S. Lin, 1992 Kinetic Model for the Curing Reaction of a
Tetraglycidyl Diamino Diphenyl Methane/Diamino Diphenyl Sulfone (TGDDM/DDS) Epoxy
Resin System, Polymer Engineering and Science 32 (15); pp. 1037-1046.
45. T. G. Gutowski, et. al. The consolidation of laminate composites, Journal of Composite
Materials, Vol. 21, February 1987, pp. 172 – 187.
46. W. B. Young, Consolidation and cure simulations for laminated composites, Polymer
Composites, February 1996, Vol. 17, No. 1, pp. 142 – 148
47. Min Li and Charles Tucker, Optimal Curing for Thermoset Matrix Composites : Thermochemical
and Consolidation Considerations, Polymer Composites, Aug. 30, 2001
48. S. R. White and H.T. Hahn, 1992 Process Modeling of Composite Materials: Residual Stress
Development during Cure. Part I. Model Formulation, Journal of Composite Materials, 26, 2402-
2422.
49. J.M. Kenny, 1992 Integration of Process Models with Control and Optimization of Polymer
Composites Fabrication, Proceedings of the Third Conference on Computer Aided Design in
Composite Materials Technology, pp. 530-544.
50. E.P. Scott, 1991 Determination of Kinetic Parameters Associated with the Curing of Thermoset
Resins Using Dielectric and DSC Data, Composites: Design, Manufacture, and Application,
ICCM/VIII, Honolulu, 1991, pp. 10-0- 1-10.
51. George S. Springer, Resin Flow During the Cure of Fiber Reinforced Composites, Journal of
Composite Materials 1982 16: 400-410.
52. P. Hubert, 1996 Aspects of Flow and Compaction of Laminated Composite Shapes During Cure,
Ph.D. Thesis, The University of British Columbia, Vancouver, B.C.
Chapter 6 Nonstructural Procedure Library
6 Nonstructural Procedure
Library
 Heat Transfer 230


Diffusion 293

Hydrodynamic Bearing 295
 Electrostatic Analysis 300


 Piezoelectric Analysis 320


Fluid Mechanics 325
 Coupled Analyses 335

References 356
This chapter describes the nonstructural analysis procedures available in Marc. These are comprised of
several areas including heat transfer, hydrodynamic bearing, electrostatic, magnetostatic,
electromagnetic, piezoelectric, fluid mechanics, and coupled analysis. This chapter provides the
technical background information as well as usage information about these capabilities.
Heat Transfer
Marc contains a solid body heat transfer capability for one-, two-, and three-dimensional, steady-state
and transient analyses. This capability allows you to obtain temperature distributions in a structure for
linear and nonlinear heat transfer problems. The nonlinearities in the problem may include temperature-
dependent properties, latent heat (phase change) effect, heat convection in the flow direction, and
nonlinear boundary conditions (convection and radiation). The temperature distributions can, in turn, be
used to generate thermal loads in a stress analysis.
Marc can be applied to solve the full range of two- and three-dimensional transient and steady-state heat
conduction and heat convection problems. Marc provides heat transfer elements that are compatible with
stress elements. Consequently, the same mesh can be used for both the heat transfer and stress analyses.
Transient heat transfer is an initial- boundary value problem, so proper initial and boundary conditions
must be prescribed to the problem in order to obtain a realistic solution. Marc accepts nonuniform nodal
temperature distribution as the initial condition, and can handle temperature/time-dependent boundary
conditions. The thermal conductivity can be isotropic, orthotropic, or anisotropic. Both the thermal
conductivity and the specific heat in the problem can be dependent on temperature; however, for
conventional heat transfer, the mass density remains constant at all times. Latent heat effects (solid-to-
solid, solid-to-liquid phase changes) can be included in the analysis. A time-stepping procedure is
available for transient heat transfer analysis. Temperature histories can be stored on a post file and used
directly as thermal loads in subsequent stress analysis. User subroutines are available for complex
boundary conditions such as nonlinear heat flux, directional heat flux, convection, and radiation.
A summary of Marc capabilities for transient and steady-state analysis is given below.
• Selection of the following elements that are compatible with stress analysis:
1-D: three-dimensional link (2-node, 3-node)
2-D: planar and axisymmetric element (3-, 4-, and 8-node)
2-D: axisymmetric shells
3-D: solid elements (4-, 6-, 8-, 10-, 15-, and 20-node)
3-D: membrane elements (3-,4-, 6-, and 8-node)
3-D: shell elements (4- and 8-node)
• Specification of temperature-dependent materials (including latent heat effects) is performed with
the ISOTROPIC, ORTHOTROPIC, ANISOTROPIC, TEMPERATURE EFFECTS, ORTHO TEMP,
TABLE, and LATENT HEAT model definition options.
• Selection of initial conditions is done using the INITIAL TEMP option.
• Selection of the temperature and time-dependent boundary conditions (prescribed temperature
history, volumetric flux, surface flux, film coefficients, radiation, change of prescribed
temperature boundary conditions during analyses) is done using the FIXED TEMPERATURE,
TEMP CHANGE, DIST FLUXES, POINT FLUX, QVECT, and FILMS options. Moving heat sources
due to welding can be specified using the WELD FLUX, WELD PATH, and WELD FILL options.
CHAPTER 6 231
Nonstructural Procedure Library
• Import of viewfactors calculated by Marc Mentat for radiation analyses.

• Selection of time steps using the TRANSIENT or AUTO STEP history definition option.
• Application of a tying constraints on nodal temperatures using the TYING model
definition option.
• Generation of a thermal load (temperature) file using the POST option, which can be directly
interfaced with the stress analysis using the CHANGE STATE model definition option.
• Use of the ANKOND user subroutines for anisotropic thermal conductivity, FILM or UFILM
for convective and radiative boundary conditions or FLUX or QVECT for heat flux
• Selection of nodal velocity vectors for heat convection is done using VELOCITY and VELOCITY
CHANGE options.
• Use of the UVELOC user subroutines for heat convection.
In addition, a number of thermal contact gap and fluid channel elements are available in Marc. These
elements can be used for heat transfer problems involving thermal contact gap and fluid channel
conditions.
The CONRAD GAP option, used in conjunction with 4- and 8-node 2-D continuum elements or 8- and
20-node 3-D continuum elements, provides a mechanism for perfect conduction or radiation/convection
between surfaces, depending on the surface temperatures. The perfect conduction capability in the
elements allows for the enforcement of equal temperatures at nodal pairs and the radiation/convection
capability allows for nonlinear heat conduction between surfaces, depending on film coefficient and
emissivity. The perfect conduction is simulated by applying a tying constraint on temperatures of the
corresponding nodal points. An automatic tying procedure has been developed for such elements. The
radiation/convection capabilities in the elements are modeled by one-dimensional heat transfer in the
thickness direction of the elements with variable thermal conductivity.
For the purpose of cooling, the CHANNEL option allows coolant to flow through passage ways that often
appear in the solid. The fluid channel elements are designed for the simulation of one-dimensional
fluid/solid convection conditions based on the following assumptions:
1. Heat conduction in the flow direction can be neglected compared to heat convection;
2. The heat flux associated with transient effects in the fluid (changes in fluid temperature at a fixed
point in space) can be neglected.
For high velocity air flow, these assumptions are reasonable and the following approach is used:
Each cooling channel is modeled using channel elements. On the sides of the channel elements,
convection is applied automatically. The film coefficient is equal to the film coefficient between
fluid and solid, whereas the sink temperature represents the temperature of the fluid. A two-step
staggered solution procedure is used to solve for the weakly coupled fluid and solid temperatures.
Thermal Contact
The CONTACT and THERMAL CONTACT options may also be used to define the fluxes entering the
surfaces. For more information, see Chapter 8 in this volume.
Use the CONTROL model definition option to input convergence controls for heat transfer analysis.
These options are:
1. Maximum allowable nodal temperature change. This is used only for transient heat transfer
analysis in conjunction with an automatic time stepping scheme. It controls the time step size.
2. Maximum allowable nodal temperature change before properties are re-evaluated and matrices
reassembled. This is used only for transient heat transfer analysis in conjunction with an automatic
time stepping scheme. For mildly nonlinear problems, if the time step remains constant, the
operator matrix is not reassembled until this value is reached.
3. Maximum error in temperature estimate used for property evaluation. This control provides a
recycling capability to improve accuracy in highly nonlinear heat transfer problems (for instance,
for latent heat or radiation boundary conditions). When nonzero, this control is used for both
steady state and transient heat transfer analysis when either the fixed or adaptive stepping
procedures are used.

For steady state problems, use the STEADY STATE history definition option. If the problem is nonlinear,
use the tolerance for temperature estimate error on the CONTROL model definition set to obtain an
accurate solution.
You must distinguish between two cases of nonlinearity in steady-state solutions: mild nonlinearities and
severe nonlinearities. In the case of mild nonlinearities, variations are small in properties, film
coefficients, etc., with respect to temperatures. The steady-state solution can be obtained by iteration.
After a small number of iterations, the solution should converge.
The technique described above is not suitable for severe nonlinearities. Examples of severe nonlinearities
are radiation boundary conditions and internal phase-change boundaries. In these cases, you must track
the transient with a sufficiently small time step ( Δt ) to retain stability until the steady-state solution is
reached. Clearly, the choice of Δt is dependent on the severity of the nonlinearity. The number of steps
necessary to obtain the steady-state solution can often be reduced by judicious choice of initial
conditions. The closer the initial temperatures are to the steady-state, the fewer the number of increments
necessary to reach steady-state.
Transient Analysis
Both fixed stepping and adaptive stepping schemes are available for transient heat transfer analysis. The
fixed stepping scheme is TRANSIENT NON AUTO. The adaptive stepping schemes are TRANSIENT and
AUTO STEP.
In the fixed stepping scheme, the program is forced to step through the transient with a fixed time step
that is user specified. Only the error in temperature estimate (convergence control 3) is used with the
fixed stepping scheme and the first two convergence controls are not checked.
CHAPTER 6 233
The scheme for the TRANSIENT option is as follows:

1. After the program obtains a solution for a step, it calculates the maximum temperature change in
the step and checks this value against the allowable temperature change per step given on the
CONTROL option (convergence control 1).
2. If the actual maximum change exceeds the specified value, the program repeats the step with a
smaller time step and continues repeating this step until the maximum temperature change is
smaller than the specified value or until the maximum number of recycles given on the control
option is reached (in which case the program stops).
3. If the actual maximum temperature change is:
a. between 80 percent and 100 percent of the specified value, the program goes on to the next
step, using the same time step.
b. between 65 percent and 80 percent of the specified value, the program tries the next step with
a time step of 1.25 times the current step.If the actual maximum is below 65 percent of the
specified value, the program tries the next step with a time step that is 1.5 times the
current step.
The objective of the scheme is to increase the time step as the analysis proceeds.
The TRANSIENT option can be used for both heat transfer as well as thermo-mechanical coupled
analysis. The time step for coupled analyses is only based on the convergence characteristics of the heat
transfer pass.
The AUTO STEP option is a unified time-stepping scheme that is available for thermal, mechanical and
thermo-mechanically coupled analysis. The time step for coupled analyses is based on the convergence
characteristics of each of the heat transfer and structural passes. Details of the AUTO STEP scheme are
provided in Chapter 11.
Let the temperature T ( x ) within an element be interpolated from the nodal values T of the element
through the interpolation functions N ( x ) ,
T ( x ) = N ( x )T (6-1)
The governing equation of the heat transfer problem is
dt
C ( T ) ----- + K ( T )T = Q (6-2)
dt
In Equation (6-2), C ( T ) and K ( T ) are the temperature-dependent heat capacity and thermal
conductivity matrices, respectively, T is the nodal temperature vector, T· is the time derivative of the
temperature vector, and Q is the heat flux vector.
The selection of the backward difference scheme for the discretization of the time variable in
Equation (6-2) yields the following expression:
1- 1
---- C ( T ) + K ( T ) T n = Q n + ----- C ( T )T n – 1 (6-3)
Δt Δt
Equation (6-3) computes nodal temperatures for each time increment Δ ( t ) .

For the evaluation of temperature-dependent matrices, the temperatures at two previous steps provide a
linear (extrapolated) temperature description over the desired interval
τ
T ( τ ) = T ( t – Δt ) + ----- ( T ( t – Δt ) – T ( t – 2Δt ) ) (6-4)
Δt
This temperature is then used to obtain an average property of the material f over the interval to be used
in Equation (6-3), such that
t
1
f = -----
Δt  f [ T ( τ ) ]dτ (6-5)
t – Δt
During iteration, the average property is obtained based on the results of the previous iteration:
τ
T ( τ ) = T ( t – Δt ) + ----- ( T* ( t ) – T ( t – Δt ) ) (6-6)
Δt
T* ( t ) are the results of the previous iteration.
Temperature Effects
The thermal conductivity, specific heat, and emissivity in a heat transfer analysis can depend on
temperatures; however, the mass density remains constant. Specify reference temperature values of
thermal conductivity, specific heat, mass density, and emissivity with the ISOTROPIC option. Enter
temperature variations of both the thermal conductivity and specific heat using the TEMPERATURE
EFFECTS or the TABLE option. This option also allows input of latent heat information. The
temperature-dependent data can be entered using either the slope-break point representation or the
property versus temperature representation. During analysis, an extrapolated/interpolated averaging
procedure is used for the evaluation of temperature-dependent properties.
Latent heat can be induced because of a phase change that can be characterized as solid-to-solid, solid-
to-fluid, fluid-to-solid, or a combination of the above, depending on the nature of the process.
Phase change is a complex material behavior. Thus, a detailed modeling of this change of material
characteristic is generally very difficult. The use of numerical models to simulate these important
phenomena is possible; several major factors associated with phase change of certain materials have been
studied numerically.
The basic assumption of the latent heat option in Marc is that the latent heat is uniformly released in a
temperature range between solidus and liquidus temperatures of the materials (see Figure 6-1). Marc uses
a modified specific heat to model the latent heat effect. If the experimental data is sufficient and available,
CHAPTER 6 235
a direct input of the temperature-dependent specific heat data (see Figure 6-2) can be used. Results of
both approaches are comparable if the temperature increments are relatively small.
Latent Heat Uniformly Release
Between ST and LT
Specific Heat
ST – Solidus Temperature
LT – Liquidus Temperature
ST LT
Temperature
Figure 6-1 Modeling Phase Changes with the Latent Heat Option
Specific Heat
Temperature
Figure 6-2 Modeling Phase Changes with the Specific Heat Option
Initial Conditions
In a transient heat transfer analysis, Marc accepts nonuniform nodal temperature distribution as the initial
condition. Enter the initial condition through the INITIAL TEMP model definition option or through the
USINC user subroutine. Initial conditions are not required in steady-state heat transfer analysis, even
though they can improve convergence when temperature-dependent properties are included.
Boundary Conditions
There are two types of boundary conditions in transient/steady-state heat transfer analysis: prescribed
nodal temperatures and nodal/element heat fluxes. These boundary conditions are entered directly
through input or through user subroutines.
Prescribe nodal temperatures using the FIXED TEMPERATURE model definition option. These
prescribed temperatures are constant with time unless a time dependent table is referenced, the
FORCDT user subroutine is invoked, or a TEMP CHANGE history definition option is encountered.
The POINT FLUX model definition option allows you to enter concentrated (nodal) heat fluxes such as
heat source and heat sink. These applied fluxes are constant with time unless a time dependent table is
referenced, the FORCDT user subroutine or a POINT FLUX history definition option is encountered.
The DIST FLUXES or QVECT model definition option allow the definition of distributed (surface or
volumetric) heat flux. These applied fluxes may be spatially varying by an invoked spatial table or by the
use of the FLUX or UQVECT user subroutine. These applied fluxes are constant with time unless a time
dependent table is referenced, the FLUX or UQVECT user subroutine is invoked, or a DIST FLUXES
history definition option is encountered. A special case is the volumetric heat generated by inelastic
loads. This will be calculated by the program when the distributed flux type is 101. It may also be
necessary to include the CONVERT option. For spatially varying fluxes associated with a welding heat
source, it is more convenient to use the WELD FLUX model definition option. Specialized welding related
options are discussed in the next section.
Notes: In heat transfer analysis, you must always specify total values; for example, total
temperature boundary conditions or total fluxes. This specification is to be used
consistently for the heat transfer portion of analysis on coupled thermal-mechanical
(thermal-solid), fluid-thermal, and fluid-thermal-solid.
The time variation used for thermal boundary conditions is a function of the stepping
procedure that is used. If adaptive stepping (either TRANSIENT or AUTO STEP) is used,
thermal boundary conditions are applied instantaneously. If fixed stepping (TRANSIENT
NON AUTO) is used, the magnitudes of the thermal boundary conditions are applied in
accordance with the tables used to define them.
Use the FILMS model definition option to input the constant film coefficient and the ambient temperature
associated with the convective boundary conditions. Use the FILM user subroutine for time/temperature-
dependent convective boundary conditions. When not using the table input option, the expression of the
convective boundary condition is
q = H ( Ts – T∞ ) (6-7)
where q , H , T s , and T ∞ are heat flux, film coefficient, unknown surface temperature, and ambient
temperature, respectively.
The radiative boundary condition can be expressed as
q = σε ( T sa
4 – T4 )
∞a (6-8)
where q is the heat flux, σ is the Stefan-Boltzmann coefficient, ε is emissivity, and T s a and T ∞ a are
unknown surface and ambient temperatures, respectively. The radiative boundary condition can be
rewritten as
q = σε ( T sa sa ∞ a + T s a T ∞ a + T ∞ a ) ( T s – T ∞ )
3 + T2 T 2 3
(6-9)
= H ( σ, ε, T s, T ∞ ) ( T s – T ∞ )
CHAPTER 6 237
This shows that the radiative boundary condition is equivalent to a nonlinear convective boundary
condition, in which the equivalent film coefficient H ( σ, ε, T s, T ∞ ) depends on the unknown surface
temperature T s . When using the nontable driven input, this case requires the FILM user subroutine.
When using the table driven input, the distributed thermal flux into an element may be defined as:
α 4 4
q = H ( T s – T ∞ ) + H n v ( T s – T ∞ ) + e v f ⋅ εσ ( T s a – T ∞ a ) + f (6-10)
where:
H is the conventional film coefficient
Ts is the unknown surface temperature
T∞ is the sink temperature
T s a is the unknown surface temperature in absolute units
T ∞ a is the sink temperature in absolute units
H n v is the natural convection coefficient
α is the natural convection exponent

ev f in the effective view factor to the environment; usually equal to 1.0
ε is the emissivity
σ is the Stefan Boltzman constant
f is the distributed flux from other sources
The absolute temperature is the temperature in user units plus the offset, which is defined in the
PARAMETERS model definition option.
A surface flux may be entered on the FILMS model definition option using the table driven input or the
DIST FLUXES model definition option. If both convective behavior and other terms contributing to the
flux are present it is computationally efficient to do it here.
When using table driven input, all coefficients given above may be temperature dependent by
referencing a table. If more complex physical boundary conditions exist, then the UFLIM user subroutine
should be used.
When using the table driven input, an alternative method is available for defining the sink temperature
through the use of the SINK POINTS model definition option. This allows the user to interface to exterior
fluid/thermal programs to input spatially varying environment temperatures. In Figure 6-3, temperatures
may be known at the points marked with an X. Using the SINK POINTS model definition option when
evaluating the flux into an element, the program will first determine the closest sink point to the surface
integration point, and then use this sink point temperature in the calculation.
x
x x x x
x x x
Figure 6-3 Temperature Points
If the heat flux defined in the FLUX or FORCDT user subroutine is temperature dependent, then the
convergence behavior of the solution process of the set of nonlinear equations can be improved by
defining not only the current flux value, but also the derivative of the flux with respect to temperature.
Based on this derivative, Marc adapts the conductivity matrix K ( T ) as well as the heat flux vector Q .
This is done by rewriting the heat flux using a Taylor series expansion. For a temperature dependent point
flux on node i , the flux during iteration n is thus given by:
i
q n = q n – 1 +  -------
i i dq i i
( T – Tn – 1 ) (6-11)
 dT  n – 1 n
i
Since all the terms in Equation (6-11) except for T n are known from iteration n – 1 , the corresponding
entries in the conductivity matrix and heat vector are updated as:
i
K n → K n –  -------
ii ii dq
(6-12)
dT n – 1
i
Q n → Q n –  -------
i i dq i
T (6-13)
 dT  n – 1 n – 1
The temperature dependent distributed fluxes are treated in a similar way and require the FILM
user subroutine.
If a film coefficient is temperature dependent, then, similar to temperature dependent fluxes, defining the
derivative of the film coefficient with respect to temperature may speed up the convergence of the
iterative process. This derivative should be defined in the FILM user subroutine.
The QVECT option may be used to define a directed flux into the surface. The user specifies a magnitude
q 0 and an orientation n . The magnitude of the flux applied is then Q = – q 0 ⋅ αA ⋅ n ⋅ n s where A is
˜ ˜
the area of the surface, α is the face absorptivity, and n s in the outward unit normal into the surface. The
absorptivity is defined through the EMISSIVITY option. Note that when using heat transfer membranes
or shell element, the normal is based upon the right-hand rule.
For both the FILMS and QVECT option using table driven input, the coefficients used in the expressions
may be temperature dependent or dependent upon other quantities.
For FILMS, this lead to:
q = H ( Te v a l ) ⋅ ( Ts – T∞ )
CHAPTER 6 239
and for QVECT, this leads to:

q = q0 ( Te v a l ) ⋅ α ( Te v a l ) ⋅ A ( n ⋅ ns )
˜
The user can specify that the evaluation temperature be either at the surface (TEMTYP=0), the average
of the surface and environment temperature (TEMPTYP=10), or the environment temperature
(TEMPTYP=20).
Furthermore, for the FILMS option, one can specify that the environment temperature at a control node;
in which case,
q = H ( Te v a l ) ⋅ ( T s – T ∞ ( T c n t r l n d ) )
Control Systems
Often in heat transfer, one desires that a boundary condition is dependent upon the temperature at another
point. In HVAC systems, this is driven by the temperature at the thermostat (multiple thermostats). If
multiple thermostats are involved, an effective temperature should be obtained by using a Servo Link.
If a control node is specified, then the following boundary conditions are applied:
Point Flux
q = Q ( Tc n t r l n d ) FLUXTYPE=6
q = Q ⋅ Tc n t r l n d FLUXTYPE=4
Distributed Flux
q = qa ( Tc n t r l n d ) ⋅ A FLUXTYPE=6
q = q a ⋅ T c n t rl n d ⋅ A FLUXTYPE=4
Film
q = H ( Te v a l ) ⋅ ( T s – T ∞ ( T c n t r l n d ) ) FLUXTYPE=6
q = H ⋅ ( Te v a l ) ⋅ Tc n t r l n d ⋅ ( Ts – T∞ ) FLUXTYPE=4
QVECT
q = qo ( Tc n t r l n d ) ⋅ α ( Te v a l ) ⋅ A ⋅ ( n – ns ) FLUXTYPE=6
˜ ˜
q = qo ⋅ Tc n t r l n d ⋅ α ( Te v a l ) ⋅ A ⋅ ( n – ns ) FLUXTYPE=4
˜ ˜
The FLUXTYPE is specified in the respective options. FLUXTYPE=4 is compatible with Nastran
definition of control node.
Velocity and Pressure Dependent Convection

In many applications, the convective coefficients are dependent upon the external fluid boundary
conditions. In such cases, one can express:
q = H ( T e v a l ,V ,P ) ⋅ ( T s – T ∞ )
where P and V are the pressure and the velocity of the flow field. This simulation can be done by using
the VELOCITY option to define the velocity and the FIXED PRESSURE option to define the pressure at
the nodal points on the boundary. This data is then interpolated to the integration points where it is used
to evaluate the function/table describing H. Furthermore, in a transient simulation where the velocity and
pressure are a function of time, the VELOCITY and FIXED PRESSURE option may be used to define the
temporal variation. In the most general case, one then has:
q = H ( T e v a l ,V ( t ) ,P ( t ) ,t ) ⋅ ( T s – T ∞ ( t ) )
If spatial variation is also required, then the UFILM user subroutine is required.
Surface Energy
The SURFACE ENERGY option calculates contributions to the thermal boundary conditions, including
the recession due to thermochemical ablation at the surface of a material which is subjected to very high
thermal fluxes. This includes the effects of convective heat flux with a possible blowing effect by mass
injection, an enthalpy flux due to molecular diffusion, a thermochemical ablation due to heterogeneous
chemical reactions with gases, a possible thermal internal decomposition, a thermochemical ablation and
a mechanical erosion by liquid or solid particles impacts. Surface energy allows one to obtain the
recession due to thermochemical ablation by gases or particles. The mechanical erosion must be
calculated by use of a UFLUXMEC user subroutine, and added to the thermochemical ablation in order to
obtain the total surface recession. Surface energy is coupled with the heat transfer into the material by
the conductive heat flux, and possibly with the mass flow rate towards the surface due to water
evaporation and/or thermal decomposition gases generated inside the material.
The SURFACE ENERGY option may be used in any heat transfer analysis, but depending on the model,
some terms could be null. When used in conjunction with the THERMO-PORE option for materials
undergoing pyrolysis, most of the terms are nonzero.
The PRINT,15 parameter allows the printing of each contribution to the surface energy.
CHAPTER 6 241
Thermochemical Ablation and Surface Energy Balance

Physical Presentation
particles
radiative blowing impact
convection diffusion
balance
flow
surface
wall
conduction decomposition ablation by ablation by

gases particles
Figure 6-4 Schema of the Surface Energy Balance
Surface Energy Balance

The surface energy balance is summarized by this equation:
Convection + Enthalpy flux due to external chemical species diffusion effects + Enthalpy flux due
to thermochemical ablation by gases + Enthalpy flux due to internal thermal decomposition
products flow – Enthalpy flux due to blowing of gaseous products + Enthalpy flux due to
thermochemical ablation by impacting particles + Thermal effects of mechanical erosion by
impacting particles – Enthalpy flux due to mechanical removal of liquid phases at the surface +
Radiative heat transfer balance – Thermal conduction = 0
Mathematical Presentation
The equation below is noted
Φ c o n v ( x ,t ,T s ) + Φ d i ff ( x ,t ,T s ) + Φ s ,t h ,g ( x ,t ,T s ,m· s ,t h ) + Φ g ( x ,t ,T s ,m· g ,p ,m· g ,w )
– Φ b l o w ( x ,t ,T s ,m· g ,p ,m· g ,w ,m· s ,t h ) – Φ l i q ( x ,t ,T s ,m· l ,j ) + Φ s ,t h ,p ( x ,t ,T s ,m· p ,j ,… )
+ Φ p a r t ,k i n ( x ,t ,T s ,m· p ,j ,… ) + Φ r a d ( x ,t ,T s ) – Φ c o n d ( x ,t ,T s ) = 0
It is homogeneous with a power per surface unit [ W.m – 2 ] .
Details on Terms:
Convection: enthalpy flux at the surface due to processes of heat conduction in the
boundary layer.
Diffusion: enthalpy flux at the surface due to processes of chemical species
diffusion in the boundary layer.
Thermochemical ablation enthalpy flow associated to mass flow of thermochemically ablated

by gases: material by gases.
Decomposition products enthalpy flow from inside of the material and ejected to the boundary
flow: layer without change of phase (generally pyrolysis gases and water).
Blowing: enthalpy flux due to evacuation of gases freed both by thermochemical
ablation (by both gaseous products and chemical effects of particle
impacts) at the surface, and internal thermal decomposition (including
possibly water evaporation).
Liquid phases removal: enthalpy flow associated with the blowing of liquid phases possibly
appearing at the surface, by external actions (flow, …)
Thermochemical ablation enthalpic flow associated with mass flow of thermochemically ablated
by external particles: material by surface impacting particles.
Thermal effects of enthalpy flow associated with energy given to the surface by impacting
mechanical erosion by particles (fraction of conversed kinetic energy).
impacts of particles:
Radiative balance: balance of radiative heat flux, resulting from absorption and emission.
Conduction: apparent conductive flow within the material.
Each time that a upper-case H is used for an enthalpy, it means that is the total enthalpy, including both
the formation enthalpy and the so-called sensible enthalpy. If we refer only to the sensible part, a
lower-case h is used. In all cases, we refer to specific enthalpies J.kg – 1 .
Equation of Each Term of the Energy Balance:
Convection: α H  H r e c – H e ;T s
 
with:
α H = α H ( x ,t ,T s ) heat transfer coefficient (homogeneous with surface mass flow rate
[ kg.m – 2 .s – 1 ] /specific enthalpy [ J.kg – 1 ] ). In some papers, α H is
sometimes written as ρ e u e C H , with C H being the Stanton number for
heat transfer, and ρ e u e the mass flow rate of the external flow parallel to
the surface at the edge e of the boundary layer.
CHAPTER 6 243
H r e c = H r e c ( x ,t ) specific recovery enthalpy of the external flow ( [ J.kg – 1 ] ) . This quantity

depends also implicitly on the external pressure. But this dependence is
hidden behind the time dependence (because the pressure is itself a
function of the time and the space variable).
Ts Ts specific enthalpy of the external flow, calculated for the frozen chemical
He ; = H e ; ( x ,t ,T s ) composition existing at the edge e of the boundary layer, but evaluated at
the surface temperature T s . This term is calculated by neglecting the
variation of the chemical composition of the external flow between the
edge of the boundary layer and the wall, due to both chemical reactions in
the gas phase, and molecular diffusion of chemical species in the
boundary layer with possibly unequal mass diffusion coefficients for the
species. Other diffusion effects of the second-order are neglected (Soret
effect, Dufour effect). The non-neglected diffusion effects are taken into
account in the diffusion term. This quantity depends also implicitly on the
external pressure. But this dependence is hidden behind the time
dependence (because the pressure is itself a function of the time and the
space variable).
Often, a 'blowing correction' modifies the heat transfer coefficient. In fact, there are two possibilities:
1. the CFD software allows one to take into account the blowing in some manner, for instance by
using a simplified subroutine representing the material behavior as a boundary condition. In that
case, the coefficient α H ( x ,t ,T s ) is obtained directly.
2. the CFD software allows one to calculate only a heat transfer coefficient α H 0 for 'inert' wall (no
blowing). So, this coefficient must be corrected in the thermal software. The usual correction is:
2κB' ; Ln  1 + 2κB' ; 
CH αH v0  v
---------- = ---------- = --------------------------------------------- = -----------------------------------------
CH αH
0 0 exp  2κB' ; – 1 2κB' ;
 v0  v
with:
( ρv ) w ( ρv ) w
B' ; ≡ --------------- B' ; ≡ ---------------
v0 αH v αH
0
κ is the ‘transpiration factor’. Usual values are κ = 0.5 for laminar flow and κ = 0.4 for
turbulent flow.
Marc has two options to enter the heat transfer coefficient: either directly α H or α H with the above
0
correction. For the latter, there is both the standard form above with κ as an entry in the data deck, and
a UFAH user subroutine allowing one to enter another kind of correction.
Ts
Diffusion: α M   Z i ;* – Z i ;* H
 e s i
i
with:
α M = α M ( x ,t ,T s ) mass transfer coefficient (homogeneous with surface mass flow rate
[ kg.m – 2 .s – 1 ] /specific enthalpy [ J.kg – 1 ] ). In some papers, α M is
sometimes written as ρ e u e C M ; C M being the Stanton number for mass
transfer and ρ e u e the mass flow rate of the external flow parallel to the
surface at the edge e of the boundary layer. Very often, C M (or α M ) is
supposed to be correlated to C H (or α H ) by C M ≈ C H ( Le ) n , Le being the

Lewis number for the frozen external flow. This Lewis number is often
supposed to be equal to 1, so that C M ≈ C H and n is meaningless Otherwise,
n depends on the configuration of the external flow, a value n = 2 ⁄ 3 is
generally accepted. For the sake of generality, one can enter separate entry
tables for α H and α M .
T T specific enthalpy of the chemical component i in the external flow, evaluated
H i s = H i s ( x ,t ,T s )
at the surface temperature T s ( [ J.kg – 1 ] ) .
Z i ;*e = Z i ;*e ( x ,t ) special fraction characterizing the component i in the chemical composition
of the external flow at the edge e of the boundary layer. Z i ;*e has no
dimension (no unit). It is a rather complicated algebraic combination of the
mole fractions and the mass fractions of all the chemical species existing in
the flow. These quantities depend also implicitly on the external pressure and
on the temperature. But these dependencies are hidden behind the time
dependence (because the pressure and the temperature of the flow are
themselves functions of the time and the space variable).
Z i ;*s = Z i ;*s ( x ,t ) special fraction characterizing the component i in the chemical composition
of the external flow near the surface s of the material. The difference between
Z i ;*e and Z i ;*s characterizes the diffusion effects across the boundary layer.
These quantities depend also implicitly on the external pressure and on the
temperature. But these dependencies are hidden behind the time dependence
(because the pressure and the temperature of the flow are themselves
functions of the time and the space variable).
Ts Ts
The quantities  Z i ;*e H i ( x ,t ,T s ) and  Z i ;*s H i ( x ,t ,T s ) are tabulated as a whole.
i i
CHAPTER 6 245
Thermochemical Ablation by External Gases: m· s ,t h ,g H s
m· s ,t h ,g = m· s ,t h ,g ( x ,t ,m· g ,T s ) surface mass flow rate of thermochemically ablated surface material

by the action of external gases ( [ kg.m – 2 .s – 1 ] ) .
H = H (T )
s s s
specific enthalpy of the solid material of the surface, evaluated at the
surface temperature T s ( [ J.kg – 1 ] ) .
A table gives H s as a function of T s , for each material possibly present at the surface.
Decomposition Products Flow: m· g H g
m· g = m· g ( x ,t ) surface mass flow rate of gaseous products produced internally by thermal

decomposition (both pyrolysis and water evaporation, [ kg.m – 2 .s – 1 ] ). Note that
if there is no pyrolysis in the model defined by the THERMO-PORE option, this
term is zero.
Hg = Hg ( Ts ) specific enthalpy of the gases taken into account in, evaluated at the surface
temperature T s ( [ J.kg – 1 ] ) . This quantity is tabulated.
Blowing: ( ρv ) w H w
with:
( ρv ) w = ( ρv ) w ( m· g ,m· s ,t h ,g m· s ,t h ,p ) surface mass flow rate of all gaseous products leaving the
surface of the material (including gas coming from internal
thermal decomposition [pyrolysis, water, …],
thermochemical ablation by external gases, thermochemical
ablation by impacting particles) [ kg.m – 2 .s – 1 ] .
H w = H w x ,t ,( T s, m· g m· s ,t h ,g m· s ,t h ,p ) specific enthalpy of the mix of all gaseous products existing
at the surface, including blowed gases and chemical species
from the external flow, evaluated at the surface temperature
T s ( [ J.kg – 1 ] ) .
For each point x and time t (discrete values), a table gives H w as a function of the mass flow rate of
gaseous products coming from the inside (pyrolysis, water evaporation), the total mass flow rate coming
from thermochemical ablation, and the surface temperature T s .
Thermochemical Ablation by Impacting Particles: m· s ,t h ,p = f t h ,p ( T s )  G t h ,p ,j m· p ,j ΔH r ,p ,j
with:
m· p ,j m· p ,g ( x ,t ) surface mass flow rate of particles for the j family
( [ kg.m – 2 .s – 1 ] ) .
G t h ,p ,j = G t h ,p ,j ( V p ,j ( x ,t ) ,D p ,j ,α p ,j ( x ,t ) ,… ) empiric law for thermochemical ablation by

impacting particles (without unit)
ΔH r ,p ,j = ΔH r ,p ,j ( T s ,J ) specific enthalpy of reaction for the interaction
between the surface material and the j family of
particles ( [ J.kg – 1 ] ) .
f t h ,p = f t h ,p ( T s ) empiric correction for the effect of surface
temperature (without unit)
The summation is based on families of particles. Each family j is characterized by an impacting mass
flow rate, a velocity of impact V p ,j , an angle of impact α p ,j , and the mean diameter of the family D p ,j .
The three first quantities are generally dependent on time, and on the location on the surface. The mass
flow rate of the surface material ablated thermochemically by the family j is generally given by a
correlation G t h ,p ,j ( V p ,j ,D p ,j ,α p ,j ) of these parameters (often called G-Law):
m· s ,t h ,p ,j ( x ,t ) = G t h ,p ,j ( V p ,j ( x ,t ) ,D p ,j ,α p ,j ( x ,t ) )m· p ,j ( x ,t )
The correlation is established by experiments. The G-Law can be entered by a UGLAW user subroutine
or a table.The total ablated mass flow rate m· s ,t h ,p is obtained by summing the above formula for the j
families.For the sake of generality, all the families are not obliged to be composed by particles of the
same chemical species. So it is more general to consider an enthalpy of reaction different for each family
j : ΔH r ,p ,j .
The enthalpy of reaction depends on the temperature of the surface T s This enthalpy of reaction is a
property of both the surface material (which can be different depending on the location on the part to be
calculated), and the family of particles p under consideration. It is attached to the material, and the family
of particle j is considered as a parameter.
This model does not take into account the effect of the temperature of the surface, as far as the ablated
mass flow rate is concerned. An empiric correction has been proposed by several authors. The value
given above (after summation on j ) must be corrected by multiplying it by a function of the surface
temperature. Several functions have been proposed:
 Ts – T1 
f 1 ( T s ) =  0 if T s < T 1 ; ------------------- if T 1 < T s < T 2 ; 1 if T 2 < T s 
 T 2 – T 1 
Ts – T0
f 2 ( T s ) = --- tan h  ------------------- + 1
1
2  ΔT 
1
with T 0 = --- ( T 1 + T 2 ) and ΔT = n ( T 2 – T 1 ) .
2
These two corrections are implemented in standard input, and an alternate UFTHP user subroutine allows
one to enter other kinds of corrections.
CHAPTER 6 247
A last point is that m· p ,j ( x ,t ) generally comes from an aerothermal analysis (performed with a CFD
software) at discrete times and are calculated on a mesh which does not match with the thermal mesh.
The analysis provides as well other parameters such as V p ,j ( x ,t ) , D p ,j , α p ,j ( x ,t ) . For 3-D analysis, the
proper angle α p ,j ( x ,t ) will be calculated within the CFD software (angle between the normal direction
at the surface and the velocity vector, in the common plan defined by these two vectors).
Thermal Effects of Particle Impacts: Φ p ,k
with:
φ p ,k = φ p ,k ( m· p ,j ( x ,t )V p ,j ( x ,t )D p ,j α p ,j ( x ,t ) ,… ) This term takes into account a possible energy
deposit by the particles impacting the surface,
besides the phenomena of thermochemical
ablation and mechanical erosion. For instance, it
could be due to a partial conversion into heat of
the kinetic energy of the particles, during
inelastic collisions. There does not seem to exist
any general accepted law or formulation for this
term. This effect is even neglected in many
ablation models. So this term can only be
introduced by a UTIMP user-subroutine.
Liquid Phases Removal: f t h ,p  m· 1 H 1
· surface mass flow rate of liquid products formed at the surface of the
m 1 = m· 1 ( x ,t )
material, and blowed out by the action of the external flow or other
mechanisms ( [ kg.m – 2 .s – 1 ] ) .
H1 = H1 ( Ts ) specific enthalpy of the liquid phase 'l' taken into account in, evaluated at
the surface temperature T s ( [ J.kg – 1 ] ) .
This term takes into account possible liquid phases appearing at the surface of the ablating material, by
several mechanisms (phase change of a component of the material [silica, …], oxidation at the surface
of gaseous products coming from the inside forming liquid phase [ SiO g → SiO 2 ] , for instance), and
the blowing of these phases (gravity, shear stress, …). Moreover, this kind of phenomena would require
a more complex modeling, for example with the explicit modeling of the liquid film running at the
surface. Finally, there could exist a liquid film at the surface produced by other mechanisms such as
impacts of many liquid particles.
These quantities are tabulated, and a user-subroutine can be used to provide this flux as a whole.
Radiative Balance: Φ r – Φr
ab em
This is defined through the FILMS or RAD-CAVITY option.

Mass Conservation Equation:

Mass flow blown (injected in the outflow) + mass flow of removed liquid phases = mass flow of
internally produced gaseous products + mass flow of the gaseous products of thermochemical ablation
reactions with external gases + mass flow of gaseous products of thermochemical ablation reactions with
impacting particles.
( ρv ) w +  m· 1 = m· s ,t h ,g + m· g + m· s ,t h ,p = m· s ,t h ,g + m· g + f t h ,g  G t h ,p ,j m· p ,j
1
Velocity of Surface Recession Due to Thermochemical Ablation:
The surface recession S· t h due to thermochemical ablation (and by extension due to other possible
'physical-chemical' mechanisms such as phase change or blowing of liquid phase) is only one part of the
total surface recession S· . This part is calculated from the SEB. To obtain the total surface recession, it
must be added the contribution of mechanical erosion S· m e c :
S· = S· t h + S· m e c
The contribution S· m e c must be calculated by other means.
Velocity of surface thermochemical recession S· t h velocity of chemical thermochemical ablation by

gases + velocity of thermochemical ablation by impacting particles (unit: m s-1)
S· t h = [ m· s ,m e c ,p + m· s ,h ,p ] ⁄ ρ̂ s
with m· s ,t h ,p = f t h ,p  G t h ,p ,j m· p ,j
ρ̂ s is the density of the solid, when pyrolysis is included through the THERMO-PORE option, this is the
current density.
Mechanical Erosion
The mechanical erosion is currently considered to be due both to impacts of particles, and to other
external actions such as the shear stress of the flow, the vibrations of the part, …
S· m e c = [ m· s ,m e c ,p + m· s ,m e c ,o t h ] ⁄ ρ̂ s
Mechanical Erosion by Particles

The expression is similar to the one for thermochemical ablation by particles described above:
m· s ,m e c ,p = f m e c ,p  G m e c ,p ,j m· p ,j
CHAPTER 6 249
m· s ,m e c ,p unit: kg m – 2 s – 1
f m e c ,p without unit
G m e c ,p ,j without unit
with here a mechanical 'G-Law' G m e c ,p ,j , and a similar temperature correction f m e c ,p . These two
quantities may be defined through the UGMEC and UFMEC user subroutines, respectively.
Alternatively, a UTIMP user-subroutine is provided for calculating m· s ,m e c ,p .
Mechanical Erosion by Other Actions

There is no generally accepted expression for. And there are many possibilities depending on the kind of
material and the environment. The UFLUXMEC user-subroutine is provided for calculating m· s ,m e c ,o t h .
Pyrolysis
Pyrolysis involves the decomposition of materials due to thermal processes. There are two aspects of
these simulations: the thermal-chemical decomposition of the material and the creation of gas and
transport of this gas. The creation of the gas is based upon the conservation of mass. The material models
used in these simulations have increasing levels of complexity. There are two models used to simulate
the transport of the gas known as the streamline model and the D’Arcy fluid model. The choice of these
two models is made on the PYROLYSIS parameter.
In the streamline model, the gas moves along streamlines that are aligned with the mesh, as shown in
Figure 6-5. This procedure is available for planar, axisymmetric and solid elements using either linear or
quadratic interpolation functions. But, because the need for regular meshes, triangular and tetrahedral
elements are not available. The user must identify which element contain streamlines.
The streamline method introduces two new terms: streamlines and streamline integration points (SIP).
For lower order elements, the SIP are at the intersection of the streamlines and the element edges. For
higher order elements, an additional SIP exists midway between the two. The pyrolysis calculation
occurs at the SIP. An extrapolation and interpolation procedure is used to move data from the SIP to
conventional integration points (CIP). The pyrolysis gas m· g moves from SIPi to SIPi+1. If the
temperature becomes hotter in the interior, such that flow of gases would reverse, an error would occur.
When using the D’Arcy law model, any solid heat transfer element may be used. When this method is
used, the pyrolysis calculation occurs at the CIP. The pyrolysis gas may move in any direction.
SFIP
node
SIPi+1
SIPi
CIP
Regular Mesh
SIP Streamline Integration Point
CIP Conventional Integration Point
SFIP Surface Integration Point
Figure 6-5 Streamlines
Presentation of the Mass Equation

The conservation of mass equation for the streamline model yields:
∂ρ̂ s ,p * ∂ρ̂ l ∂ρ̂ s ,c ,c *

∇.m· g = – --------------- – -------- – -------------------
∂t ∂t ∂t
where the first term is the conventional generation (source) term due to a change in density of the solid,
the second term represent the change in density of initial water vapor, and the final term is due to change
in density due to coking.
m· g mass flow rate of the gases of decomposition = ρ̂ g υ g .
ρ̂ g mass density of the pyrolysis gas.
ρ̂ s ,p * mass density of the solid undergoing pyrolysis.
ρ̂ l mass density of the liquid vapor.
ρ̂ s ,c ,c * mass density of the coked solid.
υg velocity of the pyrolysis gas.

CHAPTER 6 251
∂ρ̂ s ,p *
– --------------- source term of decomposition.
∂t
∂ρ̂ s ,c ,c *
– ------------------- source term of carbon deposit.
∂t
∂ρ̂ l
– -------- source term of water drying.
∂t
The drying of the initial water vapor occurs before the pyrolysis occurs so:
∂ρ̂ l
∇.m· g ,w = – --------
∂t
The global mass equation is defined as the sum of these. The global mass flow rate m· g is equal to
m· g ,p + m· g ,w .
Principle of Pyrolysis
When pyrolysis occurs, the solid material is degraded due to the effect of the temperature. This
degradation is done in steps (Figure 6-6), according to the temperature. We represent these steps on the
drawing below. With each step a loss of density corresponds: Δρ̂ s ,p ,j .
ρ̂ s
ρ̂ s ,p ,v
Δρ̂ s ,p ,1
Δρ̂ s ,p ,3
ρ̂ s ,p ,c
Temperature
Figure 6-6 Degradation Steps
The density of material ρ̂ s thus varies from ρ̂ s ,p ,v , density of virgin material to ρ̂ s ,p ,c , the density of
charred (entirely pyrolyzed material). If coking occurs the density increases because carbon is
redeposited in the solid
We define a dimensionless variable ϕ j which varies from 1 to 0 during pyrolysis (it represents the
advancement of pyrolysis).
It is calculated by the law of Arrhenius:
∂ϕ j – T a ,j ψ j
-------- = – B j exp  -----------
- ϕ .
∂t  Ts  j
The coefficients B j , T a ,j , and ψ j are determined by thermogravimetry (TGA). The material properties
associated with each phase (virgin, charred, coked, liquid, gas) are entered through the ISOTROPIC or
ORTHOTROPIC options. The THERMO-PORE option is used to reference these phases and provide
additional material data including the terms of the Arrhenius series.
∂ρ̂ s ,p *
This variable ϕ j facilitates the calculation of the variation in solid density due to pyrolysis, --------------- ,
∂t
which is the sum of the preceding ϕ j law of Arrhenius:
Nd Nd
∂ρ̂ s ,p * ∂ϕ j – T a ,j ψj
 
--------------- =
∂t  ( Δρ̂ ) s ,p ,j -------- = –
∂t  ( Δρ̂ ) s ,p ,j B j exp  -----------
Ts  j
- ϕ
j = 1 j–1
or
Nd Nd
∂ρ̂ s ,p * ∂ϕ j – E a ,j ψ
--------------- =  ( Δρ̂ ) s ,p ,j -------- = –  ( Δρ̂ ) s ,p ,j B j exp  ------------ ϕ j j
∂t ∂t  R*T
j = 1 j = 1
where
Nd is the number of terms in the Arrhenius series.
T a ,j is the temperature of activation of the reactions of decomposition
ψj is the order of the reaction
ϕj dimensionless variable of decomposition varying from 1 to 0 during pyrolysis
Bj is a factor to the exponential
E a ,j is the energy of activation of the reaction j of decomposition,
( Δρ̂ ) s ,p ,j represents the reduction of density (counted in positive quantity)

Ts is the local temperature of the solid phase.
R is the perfect gas constant.
It is assumed that at a point thermal equilibrium occurs and the temperature of the solid and the gas are
the same.
CHAPTER 6 253
Alternative Formulation
We see in the literature an alternative writing of the Arrhenius law know as the rho law:
Nd ψ
δρ̂ s ,p T a ,j  ρ̂ s ,p ,j – ρ̂ s, c ,j j
------------ =  – Γ j B' j exp  – --------- ρ̂ s ,v ,j  --------------------------------
δt  T   ρ̂ 
s ,v ,j
j = 1
where:
ρ̂ s ,v ,j is the virgin j material density.
ρ̂ s ,c ,j is the charred j material density.
ρ̂ s ,p ,j is the current j material density.
The user can enter the input for Arrhenius law for the both laws. If he chooses the rho law, then he must
give: Nd, Γ j, B' j, E a ,j ψ j, ρ̂ s ,v ,j, ρ̂ s ,c ,j . If he chooses the ϕ j law, then he must give:
Nd, Δρ̂ s ,p ,j, B j, E a ,j, ψ j .
One can define an alternate model for pyrolysis using the UPYROLSL user subroutine.
Coking
The processes considered when the material is heated are very complex. For instance, we can consider
the primary pyrolysis chemical reactions producing gases, and the subsequent secondary chemical
reactions, between gases, between gases and solid, and between constituents in the solid phase. We focus
here on one class of secondary chemical reactions, namely the heterogeneous reactions between gases
and porous solid leading to a carbon deposit, known as coking phenomenon.
The processes mentioned above are governed by complex sets of kinetically controlled chemical
reactions, which are neither well known, nor the constants associated with them.
One Term Arrhenius Model

˜
In this model, we consider the total mass fraction of atomic element carbon K c g in the gaseous phase,
independently of the molecular configuration (chemical specie) in which it is present. The initial value
˜
K c g 0 , corresponding to the gases generated by the initial primary pyrolysis chemical reactions, is data
that can be determined experimentally by calculating the difference in carbon content between the
˜
virgin material and the one of the char before coking. We should note that K c g m· is the mass flow rate g
of atomic element carbon independently of the molecular configuration.
This model is inspired by a paper written by R.A. Rindal [Ref. 20].
During the subsequent process of secondary gas/gas chemical reactions, this quantity does not change,
because only the molecular configurations are changing in the gaseous phase, but not the content in any
˜ ˜
atomic element: K c g = K c g 0 .
It is known that the gaseous phase at lower temperature contains more carbon than it would contain in
the case of chemical equilibrium. When the gases flow towards the heated surface, the temperature is
rising; the rate of reactions is rising too, and the composition of gases approaches the one corresponding
˜
to chemical equilibrium. It results in a deposit of solid carbon (coking), and then the quantity K c g is
changing in the zone where coking occurs.
From the considerations above, it seems reasonable to assume that the potential of coking is driven by
˜ ˜
the difference between the local value K c g and the value of K c g E at chemical equilibrium, and that the
rate with which chemical equilibrium is approached is of Arrhenius kind. Furthermore, the homogenous
chemical reactions in the gaseous phase are generally not equimolar, so that a pressure dependence and
an order of reaction different from the unity are assumed. So, the following kinetic equation for the
˜
evolution of the quantity K c g is proposed:
∂K̃ c g –Ea c
- = – k c exp  -----------
- ( P ) c ( K
n ˜ ˜ nc
------------
  c g – Kc g E )
∂t RT
with:
k c, E a c, n c : kinetic coefficients for coking reactions.
˜ total mass fraction of the atomic element carbon in the pyrolysis gas.
Kc g :
˜ total mass fraction of the atomic element carbon in the pyrolysis gas when chemical
Kc g E :
equilibrium is achieved.
˜
The quantity K c g E can be calculated using standard thermochemical calculation software.
The streamline model does not give access to the value of the local pressure P . There are two
assumptions that can be made:
• either neglect the pressure dependence, by setting P equals to 1
• assume that P is equal to the external pressure.
In the current implementation the pressure ( P ) = 1.0

When using the D’Arcy law model, the pressure to be considered will be the calculated pore pressure.
Remarks
˜ ˜ ˜
The quantity K c g will always be between K c g 0 and K c g E , and its variation rate will approach zero
˜ ˜
when K c g becomes close to K c g E .
CHAPTER 6 255
Based upon the assumption made in this analysis that the coking only begins when the material is
completely pyrolysed, we control the coking area by the advancement of the pyrolysis. Coking occurs if
the advancement variable of pyrolysis at this location is bigger than a critical value, named “pyrmax”. It
is usually just smaller than 1., which means that pyrolysis is over. This may be defined through the
THERMO-PORE option; the default is 0.96.
Beside this Arrhenius model to calculate the mass fraction of carbon in the pyrolysis gas, a linear model
is proposed.
Linear Model
This model determines the mass fraction of carbon in the pyrolysis gas as a function of the temperature.
The evolution is linear from a low temperature where the pyrolysis gas has an excess of carbon to a
higher temperature where it is in equilibrium:
˜ ˜
1. For temperatures below the low temperature T l , there is no coking: K c g = K c g ( T l ) .
2. For temperatures between the low temperature and the equilibrium one:
˜
• If the temperature is increasing, K c g is a linear function of temperature. ((2))
˜
• If the temperature is decreasing, the coking is stopped and K c g retains its last value.
This forbids the inverse coking reaction that would lead to an increase of carbon in the
pyrolysis gas.
For temperatures larger than the equilibrium temperature, there is no more chemical reaction:
˜ ˜
K c g = K c g E . Figure 6-7 illustrates this:
The user can also specify a coking model through the UCOKSL user subroutine.
Kcg
~
Kcg(Tl)= K cg 0
KcgE
Tlow Thigh=TE Temperature

Figure 6-7 Calculation of Mass Fraction of Carbon in Pyrolysis Gas by Linear Function of Temperature
Water Drying Model

The water drying model chosen for the streamline model is based upon the same principle than the
pyrolysis one, namely an Arrhenius law.
Written in function of the liquid mass density variable ρ̂ l , the Arrhenius law for water drying is:
ψ
∂ρ̂ l Ew  ρ̂ l  w
-------- = – B w * exp  – ----------
- *ρˆ l ,0  --------
-
∂t  R*T  ρ̂ l ,0
ρ̂ l
If we set ϕ w = 1 – --------- , we have:
ρ̂ l ,0
∂ ( 1 – ϕw ) Ew ψ
ρ̂ l ,0 ------------------------- = – B* exp  – ----------- *ρˆ l ,0 ( 1 – ϕ w ) w ,
∂t  R*T
So the temporal gradient of ϕ w is:
∂ϕ w Ew ψ
---------- = B w * exp  – ----------
- * ( 1 – ϕ w ) w
∂t  R*T 
where:
ρ̂ 1
ϕw is defined by ϕ w = 1 – ---------- , and represents the drying state. It varies from 0 to 1
ρ̂ 1 ,0
during evaporation,
Bw is a pre-exponential factor,
Ew is the energy of activation of the reaction of water drying,
ψw is the order of the reaction,
T is the local temperature of the solid phase.

R is the perfect gas constant.
Moreover, we have:
∂ρ̂ l ∂ϕ
-------- = – ρ̂ l ,0 * ---------w-
∂t ∂t
where ρ̂ l ,0 is the mass density of liquid at the beginning of the analysis.
∂ρ̂ l
The user can define its own law to calculate the term -------- via a UWATERSL user subroutine.
∂t
CHAPTER 6 257
Presentation of the Energy Equation

To take into account pyrolysis and coking and water evaporation in the calculation of the temperature, a
convective and three enthalpic terms are added to the standard equation. The phenomena is included in:
• the convective term by the mass flow rate m· g . The latter decreases during coking because of the
mass loss of gaseous carbon to solid carbon.
∂ρ̂ s ,p *
• an enthalpic term --------------- ( H g ,p – H s ,p ,v c ) that represents the energy consumed by the
∂t
pyrolysis reaction.
∂ρ̂ s ,c ,c *
• an enthalpic term – ------------------- ( H c * – H g ,p ) that represents the energy consumed by the
∂t
coking reaction.
∂ρ̂ l
• an enthalpic term -------- ( H v – H l ) that represents the energy consumed by the evaporation of the
∂t
water vapor.
The equation becomes:
∂T
( ρ̂ s ,i * c p i + ρ̂ s ,p * c s ,p * + ρ̂ s ,c ,c * c c * ) ------- + c p g ,p m· g .ΔT =
∂t
∂ρ̂ s ,p * ∂ρ̂ s ,c ,c * ∂ρ̂ l
∇. ( λ*∇T ) + --------------- ( H g ,p – H s ,p ,v c ) – ------------------- ( H c * – H g ,p ) + -------- ( H v – H l )
∂t ∂t ∂t
where:
( ρ̂c p ) e f f is effective heat capacity of the material during the analysis.
c p ,g is the effective specific heat of the gas (pyrolysis and vapor).
λ* is the effective conductivity of the material during the analysis.

H g ,p is the enthalpy of the gas, including only gas of decomposition.
H s ,p ,v c is the enthalpy of the solid in course of pyrolysis (no liquid influence).
Hv is the enthalpy of the water vapor.
Hl is the enthalpy of the liquid water.
The effective specific heat c p ,g is based upon a mixture law between heat capacity of pyrolysis gas
( ρ̂ g ,p c p ,g ,p ) and the one of vapor ρ̂ g ,w c p ,g ,w :
We have ρ̂ g c p ,g = ρ̂ g ,p c p ,g ,p + ρ̂ g ,w c p ,g ,w
m· g ,p c p ,g ,p + m· g ,w c p ,g ,w
And so c p ,g = -----------------------------------------------------------
-
m· g
To take into account the effect of water drying in the material properties, we use the drying state ϕ w
ρ̂ l
defined above by ϕ w = 1 – --------- .
ρ̂ l ,0
The effective heat capacity is defined by:
( ρ̂c p ) e f f = ( 1 – ξ p ) ( ρ̂c p ) v + ξ p ( 1 – ξ c ) ( ρ̂c p ) c + ξ p ξ c ( ρ̂c p ) c d + ( 1 – ϕ w )ρ̂ l ,0 c p ,l .
The effective conductivity λ* = ( 1 – ξ p )λ v + ξ p ( 1 – ξ c )λ c + ξ p ξ c λ c d
The heat capacity of the virgin material, charred material, deposit by coking effect and the liquid part
( ρ̂c p ) v, ( ρ̂c p ) c, ( ρ̂c p ) c d, ( ρ̂ c )
l ,0 p ,l
The conductivity of the virgin material, charred material, deposit by coking effect λ v, λ c, λ c d
We can note that:

• the vapor is not taking into account, which is consistent with the fact that we neglect also the
effect of the pyrolysis gas,
• the conductivity does not depend on the liquid water. Indeed, we can consider that the liquid
belongs to small porosity, and that water evaporates before the beginning of the pyrolysis.
We have, respectively:
• the rate of pyrolysis, coking ξ p, ξ c
• the drying state ϕ w
∂ξ 1 ∂ρ̂ s ,c ,c *
--------c = ----------------------------------- ------------------
-
∂t ρ̂ c ,d * – ρ̂ s ,p ,c * ∂t
The enthalpy of material in the course of pyrolysis function of the densities and enthalpies of virgin and
ρ̂ s ,p ,v * H s ,p ,v * – ρ̂ s ,p ,c * H s ,p ,c *
charred material: H s ,p ,v c = --------------------------------------------------------------------------- .
ρ̂ s ,p ,v * – ρ̂ s ,p ,c *
The enthalpies, written in upper case, are the absolute enthalpies. They are equal to:
T
H = ΔH f0 + h = ΔH f0 + T c p dT .
0
CHAPTER 6 259
ΔH f0 is the enthalpy of formation, is an input of the program.
T0 is the temperature of reference for enthalpy of formation, is an input of the program.
h is the sensible enthalpy.
Ablation
In many mechanical and thermal processes, there is the loss or removal of surface material by an erosive
process such as melting or vaporization. Mechanical processes, such as particle impact or friction, may
be involved as well. This process is particularly significant in high temperature applications, such as
spacecraft reentry, where the intentional removal of material contributes to the thermodynamic cooling.
For high temperature applications, special ablative materials have been created which may be defined
using the THERMO-PORE option. In the current release, ablation has been designed to be used in thermal
analyses only. The capability exists for 2-D, axisymmetric and 3-D solid elements, but the 3-D capability
is in pre-release stage.
There are seven aspects that control the receding surface calculation in Marc.
1. Specifying which surfaces are to undergo recession.
The RECEDING SURFACE option is used to specify which surface is to be subjected to recession.
For nodes that reside on elements edges (2-D) of faces (3-D) specified by the RECEDING
SURFACE, the incremental displacement will be calculated during the recovery phase.
2. Specifying the magnitude of the recession rate.
The recession rate is based upon either:
a. Simple model, that is either a constant, or evaluated by a table.
b. Based upon the data given in the SURFACE ENERGY model definition block.
c. Specified in UABLATE user subroutine.
3. Specifying the direction of the movement of the nodes on the surface.
The direction of the recession is based upon the normal evaluated at the node. If the ABLATION,1
parameter is used, then the normal is based upon applying a unit distributed load to all external
surfaces. The magnitude of the equivalent force is the effective area, and the direction is the
normal. This procedure is based upon lower order elements. The result is that at corners of the
mesh the calculated normal is not perpendicular to either surface, but a weighted average (based
upon the area of the edges).
This procedure may lead to more physically reasonable results, as corners tend to be rounded off,
but leads to more meshing problems.
If the ABLATION,2 parameter is used, then the normal is based upon applying a unit distributed
load to those element edges/faces specified by the ABLATION option. This results in the
movement of the node lying at the corner to be perpendicular to the surface.
If the ABLATION,3 parameter is used, then the direction of recession is based upon the
direction of the streamline, assuming that the streamline method is specified on the
PYROLYSIS parameter.
If the ABLATION,4 parameter is used, then the direction of recession is based upon the direction
of the streamline, assuming that the streamline method is specified on the PYROLYSIS parameter.
The magnitude of the recession is adjusted based upon the difference between the true normal and
this direction.
4. Specifying when remeshing is to occur.
Remeshing is based upon three criteria specified via the ADAPT GLOBAL option. The tolerance
values are relevant, but not critical unless the recession rate is very high, the time step is large, or
the elements are small. The two key issues are:
a. The element should not be so distorted at the beginning of the increment that the recession
during the increment would lead to an inside-out element. This may lead to an Exit 1005.
b. The ablation criteria should only find at most one element along each streamline that should be
shaved off. This may lead to an Exit 1111.
With this in mind, the following remeshing criteria have been developed:
a. Element size reduction – two ratios are calculated. The first is the ratio of the current element
edge length divided by the original element edge length. The second ratio is the ratio of the
projected current element edge length with the surface normal divided by the projected original
element edge length with the surface normal. For a 2-D quadrilateral mesh for each element on
the receding surface, two edges are checked; while for 3-D hexahedral mesh, four edges are
checked. If any edge satisfies the criteria, remeshing occurs. The recommended value of the
tolerance is 0.4.
b. Motion of a node. The magnitude of the displacement since the last remeshing is compared
with the user-specified tolerance. This method is not recommended.
c. The program also uses the previous displacements to predict if in the next increment the
element will go inside out. As long as the next increment is similar to the previous increment
this is a reasonable procedure. If may fail if the either the recession rate suddenly increases, or
if the time step is increased. The user does not need to provide any tolerance value. While in
theory, this method could be used exclusively, without the ones mentioned above, this is not
recommended, because it leads to very thin elements. Two procedures are used to determine if
an element has gone inside out. The first is based upon the Jacobian of the isoparametric shape
functions. This method was determined not to be conservative. Hence additionally the area of
four triangles (2-D) or volume of eight tetrahedrals is used.
5. Specifying the mesher used to create a new mesh
Currently the user needs to specify whether the shaver mesher, relax mesher, or stretch mesher is
to be used. These meshers may be used with quadrilateral and hexahedral elements.
a. The shaver (type 9) is advantageous because rezoning is required in only a smaller region of
the mesh. Elements are only modified if they have failed any of the criteria specified above, or
they are adjacent to a modified element. The shaved mesher is also advantageous because the
original biased mesh is retained during the process. The shaver mesher results in degenerated
quadrilateral elements (triangles), or degenerated hexahedral elements (wedges), which may
not be esthetically pleasing.
CHAPTER 6 261
b. The relax mesher modifies ever element in a body, as soon as any one of them fails the mesh
criteria specified above. The relax mesher itself is simpler than the shaver mesher. The relax
mesher also requires the use of the SPLINE option to identify edges that are to be fixed.
c. The stretch mesher modifies ever element in a body, as soon as any one of them fails the
mesh criteria specified above. The stretch mesher is similar to the relax mesher, but it only
works in one direction, and it reasonably assures that all elements have a uniform thickness
in this direction.
6. Which element type is being used and control of midside nodes
Either mesher may be used with either lower order or higher order elements for both 2-D and 3-D
problems. When using higher order elements, two considerations are made.
a. During the recession itself the midside node on the surface is positioned such that the surface
geometry is linear within and element, bi-linear for 3-D geometries. While this is not strictly
consistent with the idea of higher order elements, it resulted in less distortion problems.
Temperature fields are still quadratic. The midside nodes on the edges perpendicular to the
surface that are in the interior of the element are adjusted such that they are mid-edge. This was
done to make sure the quadratic shape functions did not break down.
b. The remeshing process itself is based upon using lower order elements, and the midside nodes
are then repositioned to the midpoints.
7. Region geometry.
For the streamline model the element edges normal to the surface in theory should follow the
streamline behavior; i.e., be normal to the surface. This is also advantageous from a surface
recession perspective. When the recession, which by default is normal to the surface, is not
aligned with the element edges, addition mesh distortion occurs. To overcome this problem a
projection procedure can be requested. This is only available in conjunction with the shaver
mesher, and is still not as good as creating a good mesh to begin with. It is highly advantageous
to have initially a good mesh.
Examples
1. Axisymmetric analysis with edges not perpendicular to surface). This is an example of a
nonoptimal meshing technique because the element edges are not perpendicular to the surface.
Figure 6-8 Original Mesh – Note vertical edges are not normal to the surface
Figure 6-9 Time=44.0 seconds
Figure 6-10 Close Up Showing Effect of Edges Not Aligned Perpendicular to Surface
2. Axisymmetric analysis of a nozzle using four node element and shaver mesher. Here the mesh
is well aligned with the geometry, which leads to the ability to ablate further, and improves
the remeshing.
Figure 6-11 Initial Mesh

CHAPTER 6 263
Figure 6-12 Time = 220. seconds
Figure 6-13 Close Up at Leading Edge
Figure 6-14 Final Temperatures

3. Example of nozzle using 2-D quadratic elements tolerance = 0.2
Figure 6-15 Initial Mesh, Showing Nodes
Figure 6-16 Time = 220 Seconds
Figure 6-17 is a close-up showing nodes, including those shaved off. Note that outer edges remain
straight, and that the midside nodes for the elements on the receding surface also remain at the
midside of the edges. The element distortion is due to the fact that the element edges are not
aligned perpendicular to the surface.
Figure 6-17 Time = 220 Seconds

CHAPTER 6 265
4. Example of 3-D nozzle.
Figure 6-18 Initial Mesh has 6480 Elements, 7657 8-node Bricks
Figure 6-19 Time = 44.0 seconds

Figure 6-20 Close Up of Leading Edge
5. Example using linear elements and relax mesher.
Figure 6-21 Final Configuration
Note that the number of elements using the relaxed mesher does not change so elements become
thinner during the ablation process.
CHAPTER 6 267
6. Example – 3-D – Linear hexahedral elements – Relax mesher – extruded nozzle.
Figure 6-22 Final Configuration - Linear 3-D Elements
7. Example – 3-D – Quadratic hexahedral elements – Relax mesher – extruded nozzle.
Figure 6-23 Time = 44.0 Seconds

Note that in Figure 6-24, the outer edges remain straight and that the midside nodes for the
elements on the receding surface also remain at the midside of the edges.
Figure 6-24 Close Up of Leading Edge – Showing nodes
Welding
Welding is a thermal process with specialized boundary conditions. These boundary conditions are
specified through the WELD FLUX model definition option in conjunction with the WELD PATH and
WELD FILL model definition options. Using these options, the welding heat input can be specified
through two different techniques:
• Modeling the heat input as a spatially varying distributed flux through the WELD FLUX model
definition option applied to the base metal and filler elements. No temperatures are specified on
the WELD FILL model definition option in this case.
• Modeling the heat input as a spatially varying temperature boundary condition through the WELD
FILL model definition option applied to the nodes of filler elements. No flux value (zero power) is
specified on the WELD FLUX model definition option in this case.
It is also possible to combine a nonzero flux applied to base metal elements through the WELD FLUX
model definition option with a nonzero temperature applied to filler element nodes through the WELD
FILL model definition option. In all these cases, the heat input path for the welding heat source or for the
filler element deposition is defined by the WELD PATH model definition option.
Weld Flux
A disc shaped surface/edge heating source or a double ellipsoidal shaped volume heating source can be
specified using the WELD FLUX model definition option. For more complex shapes, the heating source
can be specified through the UWELDFLUX user subroutine. The associated weld path (defined through
the WELD PATH model definition option) and the associated filler element set (defined through the
WELD FILL model definition option) are also referred to through the WELD FLUX model
definition option.
CHAPTER 6 269
Pavelic’s disc shaped weld heat source is expressed as [Ref. 16].
– 3x 2 ( – 3z 2 )
q ( x, y, z ) = -------- exp  ------------ exp -----------------
3Q
(6-14)
πr 2  r 2  r2
where q is the heat flux rate per unit area; Q = ηVI is the applied power with η = efficiency, V =
Applied Voltage and I = Applied Current; r is the radius of the disc; z is the local coordinate along the
weld path and x is the local coordinate along the tangent to the weld path. The disc shaped model can
be used for specifying edge fluxes in 2-D and surface fluxes in 3-D. It is particularly useful for situations
where the depth of weld penetration is not significant.
Goldak’s double ellipsoidal shaped weld heat source can be used to specify volume fluxes in 2-D and
3-D [Ref. 16]:
6 3f f Q – 3x 2 – 3y 2 – 3z 2
q f ( x, y, z ) = ----------------------- exp  ------------ exp  ------------ exp  ------------
abc f π π  a 2   b 2   c2 
6 3f r Q (6-15)
– 3x 2 – 3y 2 – 3z 2
q r ( x, y, z ) = ----------------------- exp  ------------ exp  ------------ exp  ------------
abc r π π  a 2   b 2   c2 
where q f and q r are the weld flux rates per unit volume in the front and rear weld pools respectively;
Q = ηVI is the applied power; a is the weld width along the tangent direction x; b is the weld
penetration depth along the arc direction y; c f and c r are the forward and rear weld pool lengths in the
weld path direction z; f f and f r are dimensionless factors given by
2
f f = -----------------------------
( 1 + cr ⁄ cf )
(6-16)
2
f r = -----------------------------
( 1 + cf ⁄ cr )
The double ellipsoidal source is also shown below in Figure 6-25.
cr cf z
a
x
b
y
Figure 6-25 Double Ellipsoidal Weld Flux showing the Local Coordinate Dimensions
The weld pool dimensions in Equations (6-14) and (6-15) are to be estimated and provided by the user.
The width, depth, forward length and rear length are used for the volume weld flux. The surface radius
is used for the edge or face weld flux. The dimensions are mandatory when the standard disc or ellipsoidal
models are used and are optional when the UWELDFLUX user subroutine is used. The weld pool
dimensions are used in three different ways by the program:
– for defining the weld flux rate,
– for defining a bounding box for filler element detection, and
– for defining box dimensions when a box adaptive meshing criterion is used.
The weld pool dimensions can be optionally varied with the help of tables. The tables can be functions
of time or arc length. The arc length is measured along the associated weld path, from the first point of
the path to the current position of the weld flux.
The weld flux rate in Equations (6-14) and (6-15) can be further scaled by a factor. This scale factor can
be automatically calculated by the program or optionally set by the user. For 3-D problems, the automated
scale factor is set to 1. For 2-D problems (both planar and axisymmetric), the automated scale factor s
is calculated by equating the integral of the flux rate over the entire x-y plane and the out-of plane
thickness t to the applied power Q .
s   q ( x, y, 0 )tdxdy = Q
For Equations (6-14) and (6-15), the scale factor can be shown to be
π r
s = --- -
3 t
(6-17)
π ( cr + cf )
s = --- ---------------------
3 2t
It should be noted that the automated scale factor only ensures that the applied fluxes in the 2-D case
matches the 3-D case. In general, the temperature solution comparisons between 2-D and 3-D also
depend on the velocity of the weld source, the thermal properties, other thermal and mechanical
boundary conditions, and heating effects before and after the heat source has crossed the 2-D plane.
Weld Path
The path followed by the welding heat source and the orientation of the weld arc are specified by the
WELD PATH model definition option. This information is used to determine a local coordinate system
for the moving source, as shown in Figure 6-26. The weld path vector is taken as the local z axis. The arc
orientation vector is taken as the local y axis. The tangent vector that is perpendicular to both y and z axes
is taken as the local x axis.
It is possible to specify the weld path and orientation information through the Marc Input, Text Input or
the UWELDPATH user subroutine.
CHAPTER 6 271
Y (weld path) z
X
x (weld width)
Z
y (weld orientation)
Figure 6-26 Local Coordinate System for Moving Source obtained by Translation and Rotation of Global
Coordinate System
Marc Input
Two options are available for specifying the weld path – Ordered List of Nodes or Point Coordinates of
Polyline Curves. The weld path vector at point N is defined as the vector from point N to point N+1. For
a closed weld path, the first and last node should be identical, or, the first and last point should have the
same coordinates. An update path option is provided for the nodes option for a coupled analysis which
allows the path to be updated based on the current displacements of the nodes along the path. This option
is used in conjunction with the FOLLOWER FOR parameter. When the latter flag is used, all distributed
fluxes are evaluated on the updated geometry and the path followed by the weld source is also updated
in this case.
A variety of options are available for specifying the arc orientation: Ordered List of Nodes, Point
coordinates of Polyline Curves, Vector Components, or Euler Angles. When nodes or points are used,
the arc vector at point N is defined as the vector from the weld path point at N to the arc orientation point
at N. It is further possible to rotate the arc vector through an angle specified in degrees about the weld
path axis. When vector components are used, they can be specified at distinct points and can be
optionally interpolated as a function of arc length (length of the path from the first point) or of position
coordinates along the path. When Euler angles are used, a unit vector in the global X direction is rotated
through the global X-axis, Y-axis, and Z-axis by the specified angles to yield the arc orientation vector.
The rotation angle, arc orientation vectors or Euler angles can be further varied along the weld path using
tables as a function of arc length or point coordinates. For a point that is in between two given path
points, Marc can automatically interpolate to find out the current arc orientation.
Text Input
In many industrial applications, weld paths for robots are specified via text files. A generic text file
capability is provided in Marc in order to allow the specification of the weld path and the weld
orientation. The weld path is specified through point coordinates in the columns 1-3 of the text file. The
weld path vector at point N is defined as the vector from point N to point N+1. The arc orientation,
defined in columns 4 - 6 of the text file for each point N, can be defined via point coordinates, vector
components or Euler angles. The entry in each column is a real number of width 10. The columns can be
in free or fixed format, with commas being used to separate the columns in the free format mode.
User Routine Input

The weld path can also be defined using the UWELDPATH user subroutine. The final position of
the associated weld flux, the weld vector at that point, and the orientation vector at that point have to
be defined in the subroutine. Refer to Marc Volume D: User Subroutines and Special Routines for
more details.
In some cases, it is possible that the given/computed local y axis is not perpendicular to the local z axis.
It is however important for purpose of weld flux computation to have a set of mutually perpendicular
local axes. Since weld flux dimensions are typically provided in directions along and perpendicular to
the weld path, the weld path vector (z axis) is taken as fixed and an equivalent set of perpendicular x and
y axes are computed from the given information.
The actual position of a weld heat source can be offset by a user specified amount from the associated
weld path. The offsets are specified in the local x and y directions. This offset allows the same weld path
to be used for multiple weld heat sources and is particularly useful for multi-pass welding where the weld
heat source repeatedly moves in a similar pattern, but offset from the original path by a small amount.
Weld Filler
The dynamic creation of filler elements and the generation of associated boundary conditions is
defined by the WELD FILL model definition option. Two methods are supported for the creation of
filler elements:
– the quiet element method and
– the deactivated element method.
In the quiet element method, filler elements are initially used with scaled down material properties. The
scale factor is 10-5 by default. The coefficient of thermal expansion is set to zero and all other material
properties (except yield stress, specific heat, and thermal mass density) are scaled down. The scaling is
applied to reference temperature material properties and temperature dependencies are ignored. When
temperature dependent material properties have been entered through tables, the reference temperature
properties are evaluated at the lowest temperature (i.e, the first table value). Weld fluxes and other
point/distributed thermal loads on the quiet elements are ignored. When the filler elements are physically
created by the moving heat source, the thermal properties are fully restored while the filler elements are
still quiet on the mechanical side. If the thermal activation time is zero, the mechanical properties are also
fully restored in the next increment. If the thermal activation time is nonzero, the filler elements remain
mechanically quiet till the thermal activation time duration is passed. During this period, all stresses and
strains in the elements are reset to zero. This disjoint thermal and mechanical treatment of the weld filler
prevents the build up of mechanical strains and stresses during the temperature ramp up period. The quiet
element method allows the elements to move with the model. This is particularly useful for
accommodating large deformations during the welding process. However, the quiet element method is
prone to ill-conditioning due to the large discrepancy in material stiffnesses.
In the deactivated element method, filler elements are initially deactivated in the analysis and are not
shown on the post file. When the elements are physically created by the moving heat source, they are
activated in the model and appear on the post file. The filler elements are only thermally activated
initially and remain inactive on the mechanical side. If the thermal activation time is zero, the filler
elements are activated on the mechanical side in the next increment. If the thermal activation time is
nonzero, the filler elements remain mechanically inactive until the thermal activation time is passed. This
CHAPTER 6 273
disjoint thermal and mechanical activation of the weld filler prevents the build up of mechanical strains
and stresses during the temperature ramp up period. The deactivated element method does not suffer
from any ill-conditioning problems, but the interior nodes of the deactivated elements do not move
with the rest of the model. This can cause distorted filler elements when used in a large-deformation
welding analysis.
The initial temperatures for the deposited filler elements can be directly specified on the WELD FILL
model definition option. Alternately, heat input from the weld torch can be modeled as distributed fluxes
acting on for the base metal and deposited weld filler through the WELD FLUX model definition option.
When the temperature option is used, special thermal boundary conditions are dynamically specified on
the filler elements as they are created. These melting point temperatures are directly specified on the
nodes of the created filler elements as long as the latter remain in the weld pool. Once the heat source
moves on, the nodal temperature boundary conditions are removed and the filler elements are allowed to
cool. By default, the melting point temperatures are applied instantaneously. This sudden boundary
condition can cause convergence difficulties, especially with the AUTO STEP time stepping algorithm
which bases the time step on the allowable temperature increase. A small ramp time (thermal activation
time) can be optionally provided by the user to alleviate this problem. The temperature of the filler
elements is increased from the current value to the melting point value over the ramp time specified by
the user. If the ramp time is left at 0, the melting point temperature is applied instantaneously.
The identification of the filler elements to be created in any particular increment is given by the following
algorithm: Let X g i , Y g i , Z g i be the global position of the weld heat source origin at the beginning of
the increment. Let X g f , Y g f , Z g f be the global position of the weld heat source origin at the end of the
increment. The global coordinates of the final position and of the filler element nodal points are
transformed to local coordinates using the local welding coordinate system based on the weld path and
arc orientation. Let X l f , Y l f , Z l f be the local coordinates of the final source position and X l n , Y l n ,
Z l n be the local coordinates of the nth node of a filler element. A bounding box in the local coordinate
system is then used to determine if the nodal point falls within the weld pool. Let the dimensions of the
bounding box be defined as X b , Y b , Z b f and Z b r , where X b is the dimension in the weld width
direction, Y b is the dimension in the weld depth direction, Z b f is the dimension in the forward weld
path direction and Z b r is the dimension in the rear weld path direction. If not specified by the user, X b
is taken as 1.5 times the weld pool width, Y b is taken as 2 times the weld pool width, Z b f is taken as
the forward weld pool length and Z b r is taken as the rear weld pool length.
Nodal point n is said to be inside the weld pool if the following relations are satisfied:
X ln – X b < 0 Y ln – Y b < 0
(6-18)
Z ln + Z br < 0 Z ln – Z lf – Z bf < 0
If all nodes of a particular element are inside the weld pool, then the element is activated. It should be
noted that in case the weld heat source moves across multiple segments of the weld path in a given
increment, a number of sub-steps are used within the increment and the above algorithm is used for a
linear segment in each substep.
Filler elements can be set up as a homogeneous mesh with the base structure or as separate contact bodies.
In the former case, thermal boundary conditions are automatically created at the filler-base interface. In
the latter case, an appropriate contact heat transfer coefficient needs to be specified to ensure that the heat
from the filler is transferred to the base structure.
Additional items to facilitate welding analysis in Marc are:
Time-Stepping and Convergence Checking

Fixed stepping or adaptive stepping procedures can be used for welding problems. The fixed stepping
scheme that should be used is TRANSIENT NON AUTO. The adaptive stepping scheme that should be
used is AUTO STEP.
For fixed stepping, the only temperature control that can be used is the temperature error in estimate. For
adaptive stepping, the temperature controls that can be used include the temperature error in estimate and
the allowable temperature change per increment. The default temperature change per increment is 20.
Time-step cutback is used in the AUTO STEP algorithm if the temperature controls are not satisfied.
The TRANSIENT time stepping scheme is obsolete and is not recommended for welding problems since
time step cutback is not available with that scheme when the time step is reduced. When the time step is
reduced, time step cutback is necessary to ensure that the weld motion, filler element activation, etc.
work properly.
Adaptive Meshing
Two of the most useful adaptive meshing criteria for welding are:
– Temperature Gradient Based criterion and
– the Node in Box criterion.
The general Box criterion requires the UADAPBOX user subroutine. However, for welding, the box
definition has been automated in the program. The current location of the weld source and the weld pool
dimensions are used to define the box. Nodes that fall in the box and the associated elements are adapted.
An unrefine capability is also provided so that when the box moves away, the subdivided element mesh
become coarse. It is important that the user provides weld pool dimensions through the WELD FLUX
model definition option for the box criterion to work properly. It is also important to note that if the weld
filler elements are identified as candidates for adaptive meshing, a suitable upper bound for the number
of weld filler elements should be provided on the WELDING parameter.
Material Properties
Nonlinear temperature dependence for mechanical and thermal material properties is automatically taken
into account by the program. Latent heat of solidification can be modeled by specifying the
solidus/liquidus temperatures or by varying the specific heat as a highly nonlinear function of
temperature. An appropriate temperature error in estimate control should be used to ensure that the
material properties at the current temperatures are used. It should be noted that a general solid-solid phase
transformation capability is not available in the current version of Marc.
Parallel Computation
The welding capability is fully supported in DDM. Each data file should contain the complete weld path.
When nodes are used to define the weld path/weld orientation, this may require nodes belonging to other
domains to be written out to the domain file. Also, if a weld flux belonging to domain A refers to a weld
CHAPTER 6 275
filler F, the elements of which are in domain B, then weld filler F with an empty set of elements needs
to be available in domain A.
Radiation
There are six approaches to solve radiation problems in Marc with different levels of sophistication.
They include:
1. Viewfactor calculation by direct adaptive integral method.
2. View factor calculation by Monte Carlo method.
3. Viewfactor calculation by Pixel Based Modified Hemi-cube method.
4. Radiation to Space using the FILMS model definition option.
5. Radiation to Space using any of the CONTACT or THERMAL CONTACT options.
6. Radiation into the body using the QVECT option.
There are several aspects in performing a radiation viewfactor calculation including:
1. Defining the edges or faces involved in the viewfactor calculation and determining if the region
(cavity) is open or closed.
2. Calculation of the viewfactors.
3. If the region is open defining the environment temperature. This temperature may be constant or
varying with time.
4. Definition of the surface emissivity and absorptivity, including temperature dependence and
frequency dependence (spectral behavior). By default, the absorptivity is equal to the emissivity.
5. Redefinition of the viewfactors due to large deformation or other phenomena.
6. Redefinition of the viewfactors due to either local or global adaptive meshing.
7. Radiation between surfaces results in increasing the size of the operator (stiffness) matrix, which
results in greater memory requirements and increased computational times.
All of these aspects are discussed in the following paragraphs.
In the radiation calculations in Marc, there are several assumptions made:
• Each surface is a diffuse emitter and reflector; i.e., the thermal behavior is independent of the
orientation of the radiation.
• Each surface is black; i.e., is a perfect absorber for all incident radiation.
• The surfaces are isothermal.
The third assumption requires either that an “adaptive” procedure is used to insure accuracy or that the
finite elements are sufficiently small for each surface to be assumed to be isothermal.
Using modern mesh generation techniques, there is a tendency to over-refine the finite element mesh, so
the need for these adaptive techniques may be less significant.
In theory, the viewfactors would form a symmetric matrix, the size of which is dependent upon the
number of radiating surfaces. If one has a closed cavity, in theory, the summation of all viewfactors
emitting from a surface should equal one. If desired, the numerically evaluated viewfactors can be scaled
such that the sum is one.
Computational Approaches
There are several methods available for the viewfactor calculation in Marc, namely Direct Adaptive
Integration, Monte Carlo, and Pixel Based Modified Hemi-Cube Method. The choice is made on the
RADIATION parameter. While this is the historical order that they have been implemented, the Hemi-
Cube method runs fastest, gives the most accurate viewfactors, and is the most flexible, and hence, is
discussed first.
1. Defining the Radiating Surfaces
Depending upon which technique is used for calculating the viewfactors, this can be summarized
as follows:
a. Pixel Based Modified Hemi-Cube
The CAVITY DEFINITION option is used to define the edges (2-D) or faces (3-D). Using Marc
Mentat, this is in the MODELING TOOLS CAVITIES menu. Shells may be included in the
model, in which case, one needs to indicate if it is the top or bottom or both surfaces. Symmetry
surfaces may be defined. If physically there are multiple cavities in the problem, they should
be placed in separate cavities to reduce the computational time.
b. Monte Carlo
The definition of the edges (2-D) or faces (3-D) is done via Marc Mentat or MD Patran; see
BOUNDARY CONDITIONS>THERMAL>EDGE RADIATION OR FACE RADIATION menu. The
calculation of the viewfactors is also performed in Marc Mentat or MD Patran via the
BOUNDARY CONDITIONS>THERMAL>COMPUTE RADIATION menu.
c. Direct Integration
In Marc, this is only available for axisymmetric solid elements. The geometry is defined using
the RADIATING CAVITY option by specifying the node numbers around the cavity. It is not
supported by Marc Mentat.
2. Radiation Viewfactors
In many analyses, the radiative transfer of heat between surfaces plays a significant role. To
properly model this effect, it is necessary to compute the proportion of one surface which is visible
from a second surface known as the formfactor or viewfactor. The viewfactor, defined by a fourth
order integral, presents many difficulties in its computation. Primary among these is the large
amount of computing power needed, especially when shadowing effects are included.
The radiative flow of hear from surface 1 to surface 2 is given by:
4 4
q 12 = σF 12 ( T 1 – T 2 ) (6-19)
in which, F 12 is the viewfactor and is calculated as:
1 cos φ 1 cos φ 2
F 12 = -------
A1   -----------------------------
2
- dA 2 dA 1
A1 A2 πr
CHAPTER 6 277
Pixel Based Modified Hemi-Cube Method

The foundation of radiation is based upon radiation being received from multiple directions and/or being
emitted in multiple directions. This leads to the concept of a hemisphere, where one considers an element
being projected onto the hemisphere as shown in Figure 6-27.
This geometric projection can be “easily” calculated for 2-D or axisymmetric, but is problematic in 3-D.
To overcome this problem of exact projections, one can consider dividing the hemisphere into pixels as
shown in Figure 6-28, and the simple “count” the pixels. This procedure is easier to implement in 3-D,
though it is still expensive. As an alternative one can approximate the hemisphere with a Hemi-Cube as
shown in Figure 6-29.
Figure 6-27 Element Projected onto Hemisphere
Figure 6-28 Hemisphere Divided into Pixels

Figure 6-29 Hemisphere with Hemi-Cube
Using the Hemi-Cube method is less expensive, but is still difficult around the edges, and especially the
corners in 3-D. One way to reduce this problem is to change the dimensions of the box, such that there
is a smaller possibility that leakage out of the sides is important and then simply neglect this. This is
shown in Figure 6-30, where a Hemi-Box now replaces the Hemi-Cube. To retain accuracy, one must
select a good dimension of the box, and as the space is now biased, an equally spaced pixel grid will lead
to inaccuracies. To overcome this problem, a Hemi-Plane will be used with non-equally spaced pixels,
such that the weight of each pixel is the same. This is shown in Figure 6-31
Figure 6-30 Hemisphere with Hemi-Box

CHAPTER 6 279
Figure 6-31 Hemisphere with Hemi-Plane
The basic algorithm for the viewfactor program is as follows:

a. Read in file created by the analysis program containing the geometry.
b. For axisymmetric geometries, radiating faces are swept over a user controlled angle.
c. In 3-D, subdivide quadrilateral faces into triangles.
d. Create a map of nonuniform pixels and associate a weight with them.
e. Convert coordinates to local system.
f. Loop over receiving surface, lines in 2-D, triangles in 3-D.
g. In 2-D, project line unto line; in 3-D, project triangle onto plane.
h. Identify pixels (number and location).
i. Sum the pixel weights.
j. Output the view factor for this surface.
When symmetry planes are present, the receiving faces are doubled for each symmetry plane. For cyclic
symmetry, the receiving faces are doubled n times.
This method does not guarantee a symmetric Radiation Exchange Matrix ( A j ⋅ F i j ) because all F i j are
calculated separately and independently based upon the above algorithm. There exists a possibility to
make the radiation Exchange Matrix symmetric. For large models, this may be time consuming. The
method is based upon an iterative averaging and column normalization, and for large problems, requires
a large amount of memory or disk space.
Monte Carlo Method

In this method, the idea is to randomly emit rays from the surface in question. The percentage of these
rays which hit another surface is the formfactor between the surfaces. The Monte Carlo method
computes N formfactors at one time, providing linear scaling. In fact, the larger the number of surfaces,
the faster the formfactors are computed compared to the Direct Adaptive Integration and Adaptive
Contour Integration. Hence, this method is adopted for the viewfactor calculations.
Once the viewfactors are calculated by Marc Mentat; the resulting output file containing the viewfactors
can be read in with the inclusion of the RADIATION parameter and the VIEW FACTOR model
definition option.
Some of the features of the viewfactor calculation in Marc are:
a. You are not required to specify blocking elements. This is embedded into the algorithm
completely and, hence, done automatically. This is specially useful in three-dimensional
analysis since, for complex geometries, it is impractical to predict what surfaces are blocking
other surfaces.
b. The cost of calculation is nearly linearly proportional to the number of elements which means
that, for big problems, the cost does not increase significantly.
c. The methodology in Marc guarantees that the sum of viewfactors is always 1.0. This is in
contrast to the direct integration approach which has the requirement of normalization to avoid
artificial heat gain and loss, because the sum of the viewfactors for any one surface is unlikely
to add up to 1.0 due to errors in the computation of individual viewfactors.
d. Shadowing effects (due to two surfaces being hidden from one another by other surfaces) can
be modeled.
For a surface to participate in the computations, it must participate in the following operations:
a. Ray Emission: The surface should be able to randomly emit a ray from its surface. The origin
of the ray should be randomly distributed over the area of the surface (see Figure 6-32). The
direction of the ray should be distributed according to the cosine of the angle between the ray
and the normal to the surface at the origin of the ray, thereby emitting more rays normal to the
surface than tangent to it.
b. Ray Intersection: Given an origin, direction, and length of a ray, the surface should be able to
determine whether it is hit by that ray, and if hit, the length of the ray at the point of intersection.
Determination of viewfactors involves consideration of several desired properties like speed, accuracy,
shadowing, translucence, absorption, nonuniform emission, reciprocity, and efficiency. In light of the
properties listed above, the Monte Carlo approach is adapted in conjunction with the ray tracing and
boxing algorithms.
Figure 6-33 depicts the formation of shadows which essentially involves the computation of incident
light by tracing rays from the light source to the point of incidence of the eye rays. These rays do not
reflect or refract. Shadows are formed when light source rays are obstructed, either partially or fully, from
reaching an object. Refraction and reflection of light from light sources is not computed.
Figure 6-32 Random Ray Emission in Monte Carlo Method

CHAPTER 6 281
Point Light
Eye Ray
Light Source Rays
Shadow
Figure 6-33 Shadowing Effects in Ray Tracing
Finally, to compute the viewfactors, it is not necessary to check the intersection of incident rays with all
objects under consideration. Such a method would be prohibitively computationally expensive and
preclude a large scale three dimensional analysis. An effective technique for fast calculation of
intersections is employed.
The method relies on sorting the objects before any intersections are computed. The information
computed is used to eliminate most of the intersection computations. The technique requires that each
object to be sorted have a bounding box which entirely encloses it.
The sorting relies on creation of a binary tree of bounding boxes. Thus, a bounding box is computed for
all objects. The objects are then sorted by the coordinate of the longest dimension of the bounding box.
This list of objects is then divided into two sets, each having an equal number of objects. This process is
repeated recursively until each set contains no more than a given maximum number of objects. This
recursive sorting process is depicted in Figure 6-34.
Ray intersections are performed by searching the binary bounding box tree which involves determining
whether the ray hits the bounding box of all the objects. If the intersection with one bounding box is not
found, then another bounding box is considered. In the event an intersection does exist, each object in
the node is intersected with the ray if the intersection is a bottom node of the tree. Otherwise, the
bounding boxes of both subtrees are intersected with the search in the closest subtree conducted first. The
process continues until all rays are exhausted.
1 2
3 4 5
6 7 8
9
Figure 6-34 Fast Intersection Technique for Sorting Objects
Direct Adaptive Integration

This approach computes viewfactors by directly computing a fourth order integral between every pair of
surfaces. This means that there are N squared integrations to be performed, a quadratic scaling.
CHAPTER 6 283
Besides being quadratic, to achieve reasonable accuracy, this approach requires a huge number
computations for surfaces which are close together; a situation which frequently occurs.
Using the direct integration approach, Marc calculates the viewfactors automatically for you in each
cavity of an axisymmetric body involving radiative heat transfer. This capability is only available for
axisymmetric bodies.
You must subdivide the radiative boundary in this heat transfer problem into one or more unconnected
cavities. For each cavity, you define the outline of the cavity in terms of an ordered sequence of nodes.
Usually, the nodes coincide with the nodes of the finite element mesh. You can add extra nodes, provided
you also give the appropriate boundary conditions.
The nodes must be given in counterclockwise order with respect to an axis orthogonal to the plane of the
figure and pointing to the viewer (see Figure 6-35). If the cavity is not closed, the program adds the last
side by connecting the last node with the first one. This side is treated as a black body as far as radiation
is concerned; its temperature is taken as the average between the temperatures of the adjacent nodes.
7
8
6
9
5
10 4
R
Visibility
11 ϑV 3
Z 1 2
Figure 6-35 Radiating Cavity (Approach II: Valid for Axisymmetric Capability Only)
Marc internally computes the viewfactor between every side of the cavity and all other sides. The matrix
with the viewfactors can be stored into a file, and read in again during a subsequent analysis, thus
avoiding a new computation. During a transient heat transfer analysis, for every time step, the program
estimates the temperature reached at the end of the step. From the estimated temperature, the emissivity
(temperature dependent) is computed. In addition, the radiating heat fluxes are computed. The
temperatures at the end of the step are computed by solving the finite element equations.
At every node, the difference between estimated and computed temperature is obtained. If the tolerance
allowed by the CONTROL model definition option is exceeded, iterations within the time step take place.
Otherwise, the computation of the step is concluded.
The cavity is defined by its boundary defined by a list of nodes ordered counterclockwise.
Insulated boundary condition (for example, symmetry boundaries) requires that the sum of the heat
fluxes at a node be zero. This requirement is satisfied automatically. Therefore, no input is required for
this type of boundary condition.
As previously mentioned, in a heat transfer analysis of axisymmetric body involving radiative boundary
conditions, Marc automatically calculates viewfactors for radiation. A description of the viewfactor
calculation follows.
The amount of radiation exchanged between two surfaces will depend upon what fraction of the radiation
from each surface impinges the other surfaces. Referring to Figure 6-35, the radiation propagating from
surface i to surface j will be:
cos φ i cos φ j A j
q i j = J i A i  ---------------------------------- = J i A i F i j (6-20)
 r2 
It is noted that the viewfactor, F i j , is solely geometrical in nature. From the definition of F i j , we see that
Aj Fj i = A i Fi j (6-21)
We are now ready to derive the heat transfer radiation equation. Steady-state (equilibrium) energy
conservation requires:
qi = Ai ( Ji – Gi ) (6-22)
Two independent expressions for G i can be formed.

a. The incoming radiation on a surface must equal the radiation emitted by all other surfaces
which strike this surface,
N N
Ai Gi =  Jj Aj F j i =  Jj Ai Fi j
j = 1 j = 1
or (6-23)
N
Gi =  Jj Fi j
j = 1
where N is the number of surfaces involved in the computations.
b. The other expression for G i is
Ji – εi En Ji – εi En
i i
G i = ----------------------- = ----------------------- (6-24)
ρi 1 – εi
where
En is the emissive power

CHAPTER 6 285
ε is the emissivity
ρ is the reflectance.
Substituting Equation (6-24) into Equation (6-22) and rearranging it gives
1 – εi
J i = E n – -------------- q i (6-25)
i Ai εi
This expression for J i can be inserted into Equation (6-23). After regrouping terms, you get
the governing equation for gray body diffuse radiation problems
A i δ i j – ( 1 – ε j )A i F i j  
---------------------------------------------------- { q j } = [ A i δ i j – A i F i j ]  E n  (6-26)
A j εj  i 
For the problem of black bodies, that is, ε = 1, we have,
 
{ qj } = [ Ai δi j – A i Fi j ]  En  (6-27)
 i
This equation states the obvious; net radiation heat flow from a black surface is the difference
between radiation given off and received; that is, there is no reflection.
Aj
Ai
r θj
θi
Figure 6-36 Viewfactor Definition
The heat flux radiating from A i to A j is computed as
A j cos θ j
q i j = J i ( A i cos θ i )  -------------------- (6-28)
 πr 2 
where J i is the power radiating from A i , the first term within parentheses is the projection of
A i normal to the connecting line, and the second term is the solid angle under which A j is seen
from the center of A i defining the viewfactor:
cos θ i cos θ j
F i j = ---------------------------- A j
2 (6-29)
πr
qi j = J i Ai Fi j
3. Environment Temperature
If the sum of the viewfactors from an emitting face is not equal to one, then there is a radiation
contribution to the environment. Hence, If the cavity is closed, there is no radiation to the
environment and the environment temperature is not used.
Including a PRINT-30 parameter, one will obtain the sum of the viewfactors from each emitting
surface. For a closed, cavity, comparing this to one will be a measure of the accuracy of the
viewfactor calculation.
The environment temperature may either be defined by creating an extra node and associating
it with a cavity on the CAVITY DEFINITION model definition option. The temperature of the
node is defined using FIXED TEMPERATURE model definition option. As an alternative one
can specify the environment temperature on the RAD-CAVITY model definition option.
b. Monte Carlo
A constant environment temperature is defined in Marc Mentat or MD Patran. This information
is passed into Marc in the vfs file.
c. Direct Integration
The cavity is assumed to be closed; there is no radiation to the environment.
4. Definition of the Surface Emissivity
The emissivity may be entered on an element edge or face. Hence, a particular element such as
a brick could have different emissivities because of surface conditions. The emissivity could
be temperature and frequency dependent. This is defined using the EMISSIVITY model
definition option and the TABLE model definition option if required. Is is also possible to define
the emissivity via the ISOTROPIC, ORTHOTROPIC, or other material options.
The emissivity of materials as a function of wavelength is discontinuous in nature due to the
discrete nature of particle physics.
Frequency dependent emissivity (spectral) is defined by three types of bands: zero value,
narrow, and wide. The evaluation is also dependent on whether the band is at low or high
frequency as shown in Figure 6-37. The value of emissivity at the endpoints of each band may
be a function of temperature. Based upon the table input procedure, it is possible to enter the
emissivity as a function of temperature and wavelength. The emissivity is assumed to have a
linear variation within each band.
CHAPTER 6 287
Figure 6-37 Band Frequency
Besides entering the emissivity, if wavelength dependent data is used then one also needs to
enter Planck’s second constant and the speed of light. This is done via the PARAMETERS
model definition option. The unit of wavelength used in the table should be consistent with the
unit of length used to define the speed of light.
Wavelength dependent emissivity is based upon:
• Narrow bands - use average reciprocal wavelength
• Wide bands at short wavelength - use exponential expansion
• Wide band at long wavelength - use power series in reciprocal wavelength
b. Monte Carlo or Direct Integration
The emissivity is defined via the ISOTROPIC, ORTHOTROPIC, or other material options.
Temperature dependent emissivity is entered through the TEMPERATURE EFFECTS or
TABLE model definition options.
5. Redefinition of Viewfactors due to Large Deformation
If the distance between the updated coordinates and the coordinates at which the last viewfactor
calculation was performed is greater than user-specified value, then the viewfactors will be re-
evaluated. Note that the motion of nodes on the cavity is only an approximate measure of the
inaccuracies of the previously generated viewfactors. The theoretically correct way would be to
continuously re-evaluate the viewfactors, and then measure the difference between the two sets
of viewfactors. This, of course, would defeat the objective of minimizing calculations. The
critical distance to re-evaluate is specified on the RAD-CAVITY model definition option. This
method is only available if the LARGE STRAIN parameter is used. It is also possible to specify
that the viewfactors should be re-evaluated based upon increment numbers.
Viewfactors are only recalculated if the Pixel Based Modified Hemi-Cube method is used. In the
current release, this is only available if there is one cavity in the model.
6. Redefinition of Viewfactors due to Adaptive Meshing

If the geometry of the cavity has been defined using geometric entities (curves and surfaces) in
the CAVITY DEFINITION model definition option, then when local adaptive meshing (2-D or 3-
D) or global adaptive meshing (2-D) occurs, the viewfactors will be re-evaluated. This is
based upon that after remeshing the new edges or faces are automatically attached to the curves
or surfaces. The program will determine the new list of element edges or faces in the cavity and
recalculate the viewfactors.
This capability is only available with the Pixel Based Modified Hemi-Cube method.
7. Computational Costs
Regardless of the method used to calculate the viewfactors, there are two consequences:
a. The number of viewfactors becomes large, growing quadratically with the number of
radiating surfaces.
b. The subsequent calculation in the analysis program is dependent upon the number of
viewfactors calculated.
The number of viewfactors is reduced because of shadowing effects, but depending upon the
procedure used to calculate the viewfactors, this may lead to higher computational costs in
calculating them. Because self-shadowing or third surface shadowing is likely, sparse storage
techniques are beneficial in storing the view factors. As the data structures required for the view
factor calculations are different than what is required for traditional finite element or finite
difference calculations, the computation is often put into stand-alone programs.
Using finite difference, contour integral, Gaussian integration or direct integration, the number of
operations increases quadratically with the number of radiating surfaces. All of these procedures
(combined with adaptive approaches or refined meshes) can produce accurate results, but at a
clear computational cost. With the Monte Carlo method, the computational cost increases linearly
with the number of radiating surfaces and increases linearly with the number of emitting rays. This
clearly is beneficial for large models. The problem with the Monte Carlo method is that a large
number of rays may be necessary to achieve accurate results (for example, if thin cavities exist).
The advantage of the Pixel Based Modified Hemi-Cube method is that the computational costs
are linearly dependent on the number of radiating surfaces and linearly dependent upon the
number of pixels. The accuracy is dependent upon the number of pixels chosen, and the size/shape
of the cube. A very large number of pixels has the ramification that the amount of memory
required is increased and the amount of calculations is increased.
The number of viewfactors increases the cost of the analysis calculation, because the viewfactor
file is larger, and because the bandwidth of the system increases. There are two ways to control
the computational costs:
a. One can set a tolerance and any viewfactor less than this value will be ignored. This results in
radiation being neglected.
CHAPTER 6 289
Nonstructural and Coupled Procedure Library
6\ b. One can set a second tolerance and viewfactors below this number will be treated explicitly as
opposed to implicitly. This means that the radiation contribution will be on the right-hand side
and not in the operator matrix. This may result in more iterations, but lead to a significant
reduction in the size of the operator matrix. This leads to a smaller bandwidth and a reduction
in the computational costs to solve the system of equations.
Nonstru Graphical Visualization

ctural Using the Pixel Based Modified Hemi-Cube method, is possible to visualize the relative magnitude of
and the viewfactors associated with an emitting surface. This will create a post file, jid_cxx_vfs.t19,
Coupled where xx is the cavity id. Each increment will represent the viewfactors from one surface. Creating a
Procedu contour plot with Marc Mentat, one will observe a triangle (2-D) or pyramid (3-D) indicating which face
re is emitting and the values of the viewfactors to the other faces.
Library
Data, Data Flow, and Data Constraints
The geometric information is created in Marc and written to a file, jid_cxx_t18, where jid
is the job id and xx is the cavity number.
The stand-alone program then reads this file and calculates the view factors. This is then
written to a file, jid_cxx.vfs.
The analysis program then reads this file and performs the thermal analysis.
If graphical visualization of the viewfactors is requested, this will be written to file,
jid_cxx_vfs.t19.
b. Limitations
A radiating cavity must be fully contained within one domain if DDM is used.
There is currently a limitation of 99 cavities.
Conrad Gap
For the thermal contact gap element, in the gap open condition, two surface temperatures T a and T b at
the centroid of the surfaces of a thermal contact element are obtained by interpolation from the nodal
temperatures. These two surface temperatures are used for the computation of an equivalent conductivity
for the radiation/convection link. The expression of the equivalent thermal conductivity k 1 is:
2 2
k 1 = εσL ( T k a + T k b ) ( T k a + T k b ) + H L (6-30)
where ε is the emissivity, σ is the Stefan-Boltzmann constant, L is the length of the element (distance
between a and b), T k a , T k b are absolute temperatures at a and b converted from T a and T b ; and H is
the film coefficient. The film coefficient can be a function of the temperature by referencing a table.
The equivalent thermal conductivity k 1 for the thermal contact element is assumed to be in the gap
direction. The thermal conductivities in other two local directions are all set to zero. A coordinate
transformation from the local to the global coordinate system allows the generation of the thermal
conductivity matrix of the thermal contact element in the global system for assembly.
Similarly, in the gap closed condition, tying constraints are automatically generated by the program for
thermal contact elements. The constraint equation for each pair of nodes can be expressed as:
TI + TJ if ( T g a p > T c l o se ) (6-31)
where
1
T g a p = --- ( T I + T J )
2
T I, T J = nodal temperatures at nodes I and J
T close = gap closure temperature.
Channel
For the fluid channel element, the one-dimensional, steady-state, convective heat transfer in the fluid
channel can be expressed as:
∂T f
m· c -------- + Γh ( T f – T s ) = 0
∂s (6-32)
Tf ( 0 ) = Ti n l e t
where m· is mass flow rate, c is specific heat, T f is fluid temperature, T s is solid temperature, s is
streamline coordinate, Γ is circumference of channel, h is film coefficient, and T inlet is
inlet temperature. The film coefficient may be a function of the temperature, and the inlet mass flow rate
and temperature may be a function of time.
Similarly, the conductive heat transfer in the solid region is governed by the following equation:
·
CT s + KT s = Q (6-33)
subjected to given initial condition and fixed temperature and/or flux boundary conditions. At the
interface between the fluid and solid, the heat flux estimated from convective heat transfer is
q = h ( Ts – Tf ) (6-34)
In Equation (6-33), C is the heat capacity matrix, K is the conductivity matrix and Q is the heat flux
vector. Equations (6-33) and (6-34) are coupled equations. The coupling is due to the unknown solid
temperature T s appearing in Equation (6-32) and unknown fluid temperature T f in Equation (6-34) for
the solution of Equation (6-33).
CHAPTER 6 291
The solutions for Equations (6-32) and (6-33) are obtained from the introduction of a backward
difference for the discretization of time variable in Equation (6-33) and of streamline distance in
Equation (6-32). Let
· i i–1
T s = [ T s – T s ] ⁄ ( Δt ) (6-35)
we obtain
 ----
1
-  i i 1
-----
i–1
 Δt C + K T s = Q + Δt CT s (6-36)
where Δt = time-step in transient analysis. Similarly, let
dT j–1
---------f = [ T f – T f ] ⁄ ( Δs ) (6-37)
ds
we obtain
j j–1
T f = [ Δsβ + T f ] ⁄ ( 1 + Δsα ) (6-38)
where Δs is the streamline increment,
j–1
α = Γh ⁄ ( m· c ) ; and β = ΓhT s ⁄ ( m· c ) (6-39)
Output
Marc prints out both the nodal temperatures and the temperatures at the element centroid when the
CENTROID parameter is used, or at the integration points if the ALL POINTS parameter is invoked. You
can also indicate on the HEAT parameter for the program to print out the temperature gradients and the
resulting nodal fluxes.
To create a file of element and nodal point temperatures, use the POST model definition option. This file
can be used as temperature input for performing a thermal stress analysis. This file is processed using the
CHANGE STATE option in the subsequent thermal stress analysis. This post file can also interface with
Marc Mentat or MD Patran plot temperature as a function of time.
Heat Transfer with Convection

Marc has the capability to perform heat transfer with convection if the velocity field is known. The
numerical solutions of the convection-diffusion equation have been developed in recent years.
The streamline-upwind Petro-Galerkin (SUPG) method has been implemented into the Marc heat
transfer capability.
The elements which are available are described in Table 6-1.
Table 6-1 Heat Transfer Convection Elements

36, 65 2-, 3-node link
37, 39, 41, 69, 131 3-, 4-, 6-, 8-node planar
38, 40, 42, 70, 87, 88, 122, 132 3-, 4-, 8-node axisymmetric
43, 44, 71, 123, 133 8-, 20-node hexahedral
135. 133 4-, 10-node tetrahedral
To activate the convection contribution, use the HEAT parameter and set the fifth field to 2. Due to the
nonsymmetric nature of the convection term, the nonsymmetric solver is used automatically. Specify the
nodal velocity vectors using the VELOCITY option. To change velocity, use VELOCITY CHANGE. If
nonuniform velocity vectors are required, the UVELOC user subroutine is used. This capability can be
used in conjunction with the Rigid-Plastic Flow section in Chapter 5 of this volume to perform a coupled
analysis, in which the velocity fields are obtained.
The general convection-diffusion equation is:
∂T
ρ * c  ------- + v ⋅ ∇T = ∇ ⋅ ( κ∇T ) + Q (6-40)
 ∂t 
The perturbation weighting functions are introduced as:
W = N + α  ---------- ( v ⋅ ∇N ) + β  ---------- Δt ( v ⋅ ∇N )
h h
2 v  4 v 
N is the standard interpolation function in Equation (6-1). The upwinding parameter, α , is the weighting
used to eliminate artificial diffusion of the solution; while the beta term, β , is to avoid numerical
dispersion. v is the magnitude of local velocity vectors. T is the temperature, κ is the diffusion tensor.
Q is the source term and Δt is the time increment.
The optimal choice for α and β are:
α = coth ( Peclet ⁄ 2 ) – ( 2 ⁄ Peclet )

β = C ⁄ 3 – ( 2 ⁄ Peclet ) * ( α ⁄ C ) (6-41)
where Peclet is the local Peclet number in the local element and C is the local Courant number:
Peclet = density ∗ specific heat ∗ characteristic length ∗ magnitude of the fluid
velocity/conductivity
Peclet = ρ*c*h* v ⁄ k
and
C = v * ( Δt ) ⁄ h where ( Δt ) is the time increment.
CHAPTER 6 293
The characteristic length h is defined in [Ref. 11] where C ≤ 1 is required for numerical stability. When
C > 1 , the β is set to be zero and a large time step is recommended to avoid numerical dispersion.
Note: The interpolation function N is not the time-space functions defined in [Ref. 6], so that most
Marc heat transfer elements can be used. The convection contribution of heat transfer
shell elements is limited due to the definitions of the perturbation weighting function and the
interpolation function.
Diffusion
Marc has a capability to perform diffusion analyses using two approaches. One is to use the capability
using the PORE parameter which was developed for poro-elasticity/soil analysis, and the one described
here which is activated using the DIFFUSION parameter. The basic equations governing the behavior is
D’Arcys Law, which is complemented by a compressibility model for the fluid/gas. This will be
described later.
The capability is available for either steady-state or transient behavior. The capability is available for
one-, two-, or three-dimensional, including shell analysis using the standard heat transfer elements.
Marc computes and prints the following information at the element integration points: pressure, gradient
or pressure, and the mass flow rate. The nodal point data consists of the pressure and the equivalent nodal
mass flux.
The material permeability should be entered on either the ISOTROPIC, ORTHOTROPIC, or
ANISOTROPIC model definition options. Dependency on the pressure, or the temperature should be
defined by referencing a table. The porosity or the void ratio should be entered using the INITIAL
POROSITY or INITIAL VOID RATIO model definition options. In a pure diffusion analysis, those values
do not change with time. The fluid/gas data required is the fluid viscosity, the fluid mass density, and the
fluid bulk modulus. This data is also entered in the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC
model definition options.
It should be noted that for a purely incompressible fluid, the rate term is not significant, and a steady-state
solution is immediately obtained.
The boundary conditions associated with diffusion analysis consists of prescribed pressures defined with
the FIXED PRESSURE model definition option. Unless shell elements are used, there is always only one
degree of freedom per node. For transient analyses, initial pressure should also be defined using the
INITIAL PRESSURE model definition option.
Input nodal point mass flux using the POINT MASS model definition option. Specify distributed mass
fluxes in the DIST MASS model definition option. Nonuniform or time dependent distributed mas fluxes
may be defined by the FLUX user subroutine or referencing a table.
The diffusion capability is based upon an implementation of a modified D’Arcys law for the calculation
of the fluid/gas pressure in the solid. The results in an equation for the conservation of mass. D’Arcys
law require that the Reynolds number governing the fluid velocity be small. Additionally the current
formulation is appropriate for a single fluid.
1. D’Arcys Law
The fundamental equation of D’Arcys law is expressed as:
V D = – K ⁄ μ∇P
where
VD is the D’Arcys velocity or the seepage velocity or filtration velocity.
K is the absolute permeability
μ is the dynamic viscosity
P is the pressure
The D’Arcys velocity can be related to the real velocity V g by the porosity, φ .
V D = φV g
The porosity is a dimensionless quantity relating the volume of the pores to the total volume,
which is bounded by 0 < φ < 1 , where a homogeneous material would have the value of zero. A
related quantity is the void ratio, e , which is a dimensionless quantity that is the ratio of the
volume of pore to the volume of solid, which is an unbounded number, 0 < e < ∞ . They are
related by:
φ = e ⁄ (1 + e)
The next issue is the permeability, which in the literature is given the same word and symbol to
mean different things. The permeability in D’Arcys law is the “absolute Permeability” and has
units of l 2 . It is related to the D’Arcys coefficient of permeability, k , by:
K = kμ ⁄ γ
Note that the D’Arcys coefficient of permeability has units of l ⁄ t and γ is the specific weight,
γ = ρg .
The dynamic viscosity has the units of m ⁄ lt and is related to the kinematic viscosity v which
has the units of l 2 ⁄ t by:
v = μ⁄ρ
where ρ is the real mass density (units of m ⁄ l 3 ).

Note that in the literature the dynamic viscosity often is given the symbol of ν and the kinematic
viscosity the symbol η .
CHAPTER 6 295
2. Extension for Compressibility

D’Arcys law is extended by assuming that the fluid is slightly compressible, following the
approach used in soil mechanics as:
· ·
P = Kf ρf ⁄ ρf ,
where K f is the bulk modulus of the fluid, and ρ f is the density of the fluid. β f the inverse of
the bulk modulus is often introduced as the fluid “compressibility”.
βf = 1 ⁄ Kf
The governing equation for the fluid flow becomes

·
∇ ( K ⁄ μφ )∇P = φβ f P + ∇V .
It is often more meaningful to give the mass flow rate across a boundary, so the complete
expression is multiplied by the fluid density.
3. Initial Values
To complement the initial values in the thermal problem, it is necessary to provide initial values
for the pressure problem as well.
Hydrodynamic Bearing
Marc has a hydrodynamic bearing analysis capability, which enables you to solve lubrication problems.
This capability makes it possible to model a broad range of practical bearing geometries and to calculate
various bearing characteristics such as load carrying capability, stiffness, and damping properties. It can
also be used to analyze elasto-hydrodynamic problems.
The lubricant flow in hydrodynamic bearing is governed by the Reynolds equation. The bearing analysis
capability has been implemented into Marc to determine the pressure distribution and mass flow in
bearing systems.
Marc is capable of solving steady-state lubrication problems; the incremental procedure analyzes a
sequence of different lubricant film profiles. Marc also can be used to solve coupled elasto-
hydrodynamic problems. This analysis requires a step-by-step solution for both the lubrication and the
stress problems using separate runs. Because the finite element meshes for each problem are different,
the program does not contain an automated coupling feature.
Only one-dimensional or two-dimensional lubricant flow needs to be modeled, since no pressure
gradient exists across the film height. This modeling is done with the available heat transfer elements.
The library elements listed in Table 6-2 can be used for this purpose.
Table 6-2 Hydrodynamic Bearing Elements

Element Description
36 2-node, three dimensional link
37 3-node, planar triangle
39 4-node, bilinear quadrilateral
41 8-node, planar biquadratic quadrilateral
65 3-node, three-dimensional link
69 8-node, biquadratic quadrilateral with reduced integration
121 4-node bilinear quadrilateral with reduced integration
131 6-node triangle
Marc computes and prints the following elemental quantities: lubricant thickness, pressure, pressure
gradient components, and mass flux components. Each of these is printed at the element integration point.
The nodal point data consists of pressures, equivalent nodal mass flux at fixed boundary points, or
residuals at points where no boundary conditions are applied. In addition, the program automatically
integrates the calculated pressure distribution over the entire region to obtain consistent equivalent nodal
forces. This integration is only performed in regions where the pressure exceeds the cavitation pressure.
The output includes the load carrying capacity (the total force on the bearing). This capacity is calculated
by a vectorial summation of the nodal reaction forces. In addition, the bearing moment components with
respect to the origin of the finite element mesh can be calculated and printed.
To activate the bearing analysis option, use the BEARING parameter. If the analysis requires modeling of
flow restrictors, also include the RESTRICTOR parameter.
The values of the viscosity, mass density, and cavitation pressure must be defined in the ISOTROPIC
option. Specify temperature-dependent viscosity values via TEMPERATURE EFFECTS or TABLE. If
thermal effects are included, the STATE VARS parameter is also required. In hydrodynamic bearing
analyses, temperature is the second state variable. Pressure is the first state variable.
The fluid thickness field can be strongly position-dependent. A flexible specification of the film profile
is allowed by using either the nodal thickness or elemental thickness option. Define nodal thickness
values in the THICKNESS option. You may also redefine the specified values via the UTHICK user
subroutine. Elemental values of lubricant thicknesses can be defined in the GEOMETRY option.
Marc also enables the treatment of grooves. Constant groove depth magnitudes can be specified in the
GEOMETRY option. If the groove depth is position-dependent, the contribution to the thickness field can
be defined in the UGROOV user subroutine.
The relative velocity of the moving surfaces is defined on a nodal basis in the VELOCITY option. In
addition, you can redefine the specified nodal velocity components in the UVELOC user subroutine.
Specify prescribed nodal pressure values in the FIXED PRESSURE option. Define restrictor type
boundary conditions in the RESTRICTOR option. To specify nonuniform restrictor coefficients, use the
URESTR user subroutine.
CHAPTER 6 297
Input nodal point mass fluxes using the POINT MASS option. Specify distributed mass fluxes in the DIST
MASS option. If nonuniform fluxes are necessary, apply this via the FLUX user subroutine. The TABLE
option may also be used to define spatially varying boundary conditions.
Define variations of the previously specified lubrication film thickness field through the THICKNS
CHANGE option. The program adds this variation to the current thickness values and solves the
lubrication problem.
Activate the calculation of bearing characteristics (that is, damping and stiffness properties) through the
DAMPING COMPONENTS or STIFFNS COMPONENTS options. Marc evaluates these properties based
on the specified change in film thickness. This evaluation requires the formation of a new right-hand-
side, together with a matrix back substitution. This is performed within so-called subincrements. The
bearing force components calculated within these subincrements represent the bearing characteristics
(that is, the change in load carrying capacity for the specified thickness change or thickness rate). The
previously specified total thickness is not updated within subincrements. The calculated bearing
characteristics are passed through to the UTHICK user subroutine. This allows you to define an
incremental thickness change as a function of the previously calculated damping and/or stiffness
properties. This procedure can be applied when analyzing the dynamic behavior of a bearing structure.
Mechanical problems can often be represented by simple mass-damper-spring systems if the bearing
structure is nondeformable. A thickness increment can be derived based on the current damping and
stiffness properties by investigating the mechanical equilibrium at each point in time.
The bearing analysis capability deals with only steady-state solution and does not include the analysis of
transient lubrication phenomena. Note that the incrementation procedure is only meant to analyze a
sequence of film profiles. No nonlinearities are involved; each increment is solved in a single step
without iteration.
To calculate the reaction forces that act on the bearing structure, Marc requires information about the
spatial orientation of the lubricant. This information is not contained in the finite element model because
of the planar representation of the lubricant. Therefore, it is necessary to define the direction cosines of
the unit normal vector that is perpendicular to the lubricant on a nodal basis in the UBEAR user
subroutine. The resulting nodal reaction forces are printed.
Marc requires a step-by-step solution of both the lubrication problem and the stress problem in separate
runs. The thickness changes need to be defined within the lubrication analysis based upon the
displacements calculated in the stress analysis. The stress analysis post file and the UTHICK user
subroutine can be used for this purpose. The tractions to be applied in the stress analysis can be read from
the bearing analysis post file in the FORCDT user subroutine.
The flow of a lubricant between two surfaces that move relative to each other is governed by the
Reynolds equation
ρh 3 ∂ ( ρh ) 1
∇ ⋅  ---------- ∇p – --------------- – --- ∇ ( ρhu ) + M = 0 (6-42)
 12η  ∂t 2
where:
p is lubricant pressure
ρ is mass density
h is film thickness
η is viscosity
t is time
u is the relative velocity vector between moving surfaces
M is the mass flux per unit area added to the lubricant
The following assumptions are involved in the derivation of this equation:

• The lubricant is a Newtonian fluid; that is, the viscosity is constant.
• There is no pressure gradient across the film height.
• There is laminar flow.
• Inertial effects are negligible.
• The lubricant is incompressible; that is, mass density is constant.
• Thermal effects are absent.
By introducing the film constant
ρh 3
λ = ---------- (6-43)
12η
Equation (6-43) can be written as
∇ ⋅ ( λ∇p ) + M r = 0 (6-44)
r
where M is the reduced mass flux given by
∂ ( ρh ) 1
M r = M – --------------- – --- ∇ ( ρhu ) (6-45)
∂t 2
In case of a stationary bearing, the transient term in Equation (6-45) vanishes.
Three different kinds of boundary conditions can be distinguished for the lubrication problem:
prescribed pressure on boundary, prescribed mass flux normal to the boundary, and mass flux
proportional to pressure.
Prescribed pressure on boundary is specified as
p = p (6-46)
where p is the value of the prescribed pressure.

CHAPTER 6 299
Prescribed mass flux normal to the boundary has the form
∂p 1
– λ ------ = m n – --- ρhu n = m nr (6-47)
∂n 2
where m nr is the reduced inward mass flux. Here, n refers to the inward normal on the boundary,
and m n and u n are the inward components of total mass flux and relative velocity, respectively.
If a restrictor is used (as in hydrostatic bearings), the total mass flux is a linear function of the pressure
on the boundary. This condition is specified as
∂p 1
m n = – λ ------ + --- ρhu n = c ( p – p ) (6-48)
∂n 2
or, written in a slightly different form
∂p
– λ ------ = c ( p r – p ) (6-49)
∂n
where c is the restriction coefficient and
ρhu n
p r = p – ------------ (6-50)
2c
is the reduced pressure.
The differential Equation (6-42), together with the boundary conditions (Equations (6-47), (6-48), and
(6-50)) completely describe the lubrication problem. This is analogous to a heat conduction problem as
shown in Table 6-3.
Table 6-3 Comparison of Lubrication and Heat Conduction
Lubrication Heat Conduction
Pressure p Temperature T
Film constant λ Conductivity k
Reduced body mass flux r Body heat flux Q

M
Reduced boundary mass flux mn Boundary heat flux qn
Restriction coefficient c Film coefficient h
Reduced reference pressure r Reference temperature Tr

p
Electrostatic Analysis
Marc has the capability to perform electrostatic analysis. This allows the program to evaluate the electric
fields in a body or media, where electrical charges are present. This can be solved for one-, two-, or three-
dimensional fields. Semi-infinite elements can be used to represent an infinite domain. The electrostatic
problem is governed by the Poisson equation for a scalar potential. This analysis is purely linear and has
been implemented in Marc analogously to the steady state heat transfer problem. The available elements
are described in Table 6-4 below.
Table 6-4 Element Types for Electrostatic Analysis
37, 39, 131, 41 3-, 4-, 6-, 8-node planar
69 8-node reduced integration planar
101, 103 6-, 9-node semi-infinite planar
38, 40, 132, 42 3-, 4-, 6-, 8-node axisymmetric
70 8-node reduced integration axisymmetric
43, 44 8-, 20-node 3-dimensional brick
71 20-node reduced integration brick
105, 106 12-, 27-node semi-infinite brick
123 8-node reduced integration brick
135, 133 4-, 10-node tetrahedral
50, 85, 86 3-, 4-, 8-node shell
87, 88 2-, 3-node axisymmetric shell
137, 203 6-, 15-node pentrahedral
Marc computes and prints the following quantities: electric potential field vector ( E ) and electric
displacement vector ( D ) at the element integration points. The nodal point data consists of the potential
φ and the charge Q .
To activate the electrostatic option, use the ELECTRO parameter. The value of the isotropic permittivity
constant is given in the ISOTROPIC option, orthotropic constants can be specified using the
ORTHOTROPIC option. Optionally, the UEPS user subroutine can be used.
Specify nodal constraints using the FIXED POTENTIAL option. Input nodal charges using the POINT
CHARGE option. Specify distributed charges by using the DIST CHARGES option. If nonuniform
charges are required, the FLUX user subroutine can be used for distributed charges and the FORCDT user
subroutine for point charges. The TABLE option may also be used to define spatially varying
CHAPTER 6 301
The electrostatic capability deals with linear, steady-state problems only. The STEADY STATE option is
used to begin the analysis. The resultant quantities can be stored on the post file for processing with
Marc Mentat.
The Maxwell equations to govern electrostatics are written in terms of the electric displacement vector
D and the electric field vector E such that
∇⋅D = ρ (6-51)
and
∇×E = 0 (6-52)
where ρ is a given volume charge density.

The constitutive law is typically given in a form as:
D = εE (6-53)
where ε is the dielectric constant.
Introducing the scalar potential φ such that
E = – ∇φ (6-54)
which satisfied the constraint Equation (6-52) exactly.
Denoting the virtual scalar potential by ψ , the variational formulation is
 ∇ψ ⋅ ( ε∇φ )dV =  ψρdV +  ψ ⋅ ( ε∇φ – n )dA (6-55)

V V Γ
The natural boundary condition is applied through the surface integral in terms of the normal electric
displacement using:
– ε∇φ ⋅ n = D ⋅ n = D n (6-56)
Consider a material interface Γ 12 , separating two materials 1 and 2, and as φ is continuous over the
material interface, the tangential electric field constraint is automatically satisfied.
n × ( E1 – E2 ) = 0 on Γ 12 (6-57)
If charges are present on the interface, these are applied as distributed loads as follows:
n ⋅ ( D1 – D2 ) = ρs on Γ 12 (6-58)
Using the usual finite element interpolation functions N and their derivatives β , we obtain
ψ = NΨ φ = NΦ
K =  βTε βdV (6-59)

V
F =  N T ρNdV +  N T ND n dA +  N T Nρ s dA (6-60)
V Γ Γ
12
and finally
KΦ = F (6-61)
Capacitance
A capacitor is a device that can store electrostatic charge and hence electrostatic potential energy. This
energy can be thought as the energy stored in the electric field created by the charge on the
capacitor.Capacitance is always associated with conductors that can store charge. A problem can contain
a number of conductors each of which will store charge. This gives rise to self and cross capacitance due
to interaction between the conductors. This is expressed by the capacitance matrix.
All conductors in the problem are assumed to be bodies and defined as a set of elements. This is done
using the THERMAL CONTACT model definition option. No two conductors can touch each other or
overlap. It is required to model the inside of the conductor bodies. A sub-set of the conductor bodies can
be considered for capacitance computation. This subset is specified using the EMCAPAC history
definition in the THERMAL CONTACT model definition option. Marc performs capacitance computation
only if the above definitions are specified in the input file.
The electrostatic analysis is used for computation of the capacitance matrix. The computation of the
capacitance matrix results from the usual electrostatic analysis with some constraints on boundary
conditions:
• The boundary of the problem must satisfy far-field conditions and the electric potential on this
boundary must be zero.
• At problem boundaries where far field is not satisfied, the homogenous Neumann boundary
condition must apply.
• The electrostatic charge boundary condition must not be used on the problem boundary.
• Any boundary condition applied on an individual conductor body is ignored.
When dissimilar meshes are used to model the problem, then the conductors and dielectrics are distinct
contact bodies. In this case it is best that the conductors are the contact bodies which are being touched
and the dielectrics are the touching contact bodies.
Consider a single conducting body placed in an infinite homogenous dielectric medium. This conductor
can be charged by putting a charge on it or applying an electric potential. The charge redistributes on the
CHAPTER 6 303
conductor such that the electric field inside the conductor is zero. This indicates that these charges reside
on the outside surface of the conductor and there is no charge inside the conductor. A normal electric
field is present on the surface of the conductor. The value of the normal electric field is equal to:
σ
E n = --- (6-62)
ε
where:
σ is the surface charge density
ε is the permittivity of the medium surrounding the conductor
The normal electric displacement is
D n = εE n = σ (6-63)
The total charge on the conductor is:
Q =
° ° σdS (6-64)
S
where S defines the surface of the conductor.

The capacitance C is then defined as:
Q
C = ---- (6-65)
V
where V is the constant potential on the conductor.
Capacitance Matrix
For multiple conductors the capacitance computation is expressed as a capacitance matrix. For n
multiple conductors, the capacitance matrix is:
Q1 C 11 C 12 . . C 1 n V1
Q2 C 21 C 22 . . C 2 n V2
. = . . . . . . (6-66)
. · ,
. . . . .
Qn Cn 1 Cn 2 . , Cn n Vn
where
Ci j is the self capacitance of conductor i , if i = j
Ci j is the mutual capacitance between conductor i and j , if i ≠ j
To find the values of C i j , the analysis has to be repeated n times.
To find C i j for any column j and i = 1 ,2 ,…n , the procedure is as follows:
1. Specify V j = 1 volt and V 1 = V 2 = …… = V i ≠ j = … = V n = 0 .

2. For this run the values of Q i ( i = 1 ,2…n ) are computed and the capacitance values of column
j are calculated.
Magnetostatic Analysis
Marc has the capability to perform magnetostatic analysis. This allows Marc to calculate the magnetic
field in a media subjected to steady electrical currents. This can be solved for two- or three-dimensional
fields. Semi-infinite elements can be used to represent an infinite domain.
The magnetostatic analysis for two-dimensional analysis is solved using a scalar potential, while for
three-dimensional problems, a full vector potential is used. The magnetic permeability can be a function
of the magnetic field, hence, creating a nonlinear problem. Only steady-state analyses are performed. The
available elements which are described in Table 6-5.
Table 6-5 Elements Types for Magnetostatic Analysis

37, 39, 131, 41 3-, 4-, 6-, 8-node planar
102, 104 6-, 9-node semi-infinite axisymmetric
109 8-node brick
110 12-node semi-infinite brick
181 4-node tetrahedron
183 2-node 3-D link
204 6-node pentahedral
205 15-node pentahedral
206 20-node brick
Marc computes and prints magnetic induction ( B ), the magnetic field vector ( H ), and the vector
potential at the element integration points. The nodal point data consist of the potential ( A ) and the
current ( J ).
CHAPTER 6 305
To activate the magnetostatic option, use the MAGNETO parameter. The value of the isotropic
permeability (μ) is given on the ISOTROPIC option; orthotropic constants can be specified using the
ORTHOTROPIC option. Optionally, the UEPS user subroutine can be used. Often, it is easier to specify
(1/μ), which is also available through these options. A nonlinear permeability can be defined using the
B-H relation.
Specify nodal constraints using the FIXED POTENTIAL option. Input nodal currents using the POINT
CURRENT option. Specify distributed currents by using the DIST CURRENT option. Permanent magnets
can be introduced by using the PERMANENT option for isotropic materials, or by entering a remanence
vector via the B-H RELATION option for orthotropic materials. In addition, the FLUX user subroutine can
be used for nonuniform distributed current and the FORCDT user subroutine for fixed nodal potentials
or point current. For convenience line element 183 can be used to define an external loading. The current
on this element will be converted to point currents on the nodes, pointing in the direction of this line
element. The TABLE option may also be used to define spatially varying boundary conditions.
The magnetostatic capability is linear unless a nonlinear B-H relation is defined. In such problems,
convergence is reached when the residual current satisfies the tolerance defined in the CONTROL option.
The STEADY STATE option is used to begin the analysis. The resultant quantities can be stored on the
post file for processing with Marc Mentat.
The Maxwell equations for magnetostatics are written in terms of the magnetic flux density vector B
such that:
∇×H = J (6-67)
and
∇⋅B = 0 (6-68)
where J is the current density vectors.
For magnetic materials, the following relation between B , H , and M , the magnetization vector, holds:
B = μ0 ( H + M ) (6-69)
with μ 0 being the permeability of vacuum. Denoting the magnetic susceptibility by χ m and the
permanent magnet vector by M 0 , we have
M = χm H + M0 (6-70)
which can be substituted into Equation (6-69) to yield:
B = μ H + Br (6-71)
in which μ is the permeability, given by:
μ = μ0 ( 1 + χm ) (6-72)
and Br is the remanence, given by:
Br = μ0 M0 (6-73)
Notice that ( 1 + χ m ) is usually called the relative permeability μ r .

B
Br
H
Figure 6-38 Nonlinear B-H Relationship
For isotropic linear material, χ m and M 0 are scalar constants. If the material is orthotropic, χ m and M
change into tensors. For real ferromagnetic material, χ m and μ are never constant. Instead, they depend
on the strength of the magnetic field. Usually this type of material nonlinearity is characterized by a so-
called magnetization curve or B-H relation specifying the magnitude of (a component of) B as a function
of (a component of) H.
In Marc, the magnetization curve can be entered via the B-H RELATION option. For isotropic material,
only one set of data points, representing the magnitude of the magnetic induction, B , as a function of
H , the magnitude H, needs to be given. For orthotropic material, multiple curves are needed with each
curve relating a component of B to the corresponding component of H.
When the table driven input is used, different ways of entering a magnetization curve are available. Then
the ISOTROPIC or ORTHOTROPIC material option contains either the permeability, the inverse
permeability, the H-B relation, or the B-H relation. For ORTHOTROPIC, this can be different for each
component. For the H-B relation and the B-H relation, a table has to be given where for the H-B relation,
B is the independent variable, and for the B-H relation, H is the independent variable. Permeability and
inverse permeability can also be controlled by a table where the independent variable can be H, B, or any
other variable. A table can be either a set of data points or a function.
From Equations (6-71) and (6-73), it can be seen that for orthotropic materials, a permanent
magnetization or remanence can be entered through the B-H RELATION option, by putting in a nonzero
B value for H = 0. For isotropic material, this does not work since the direction of the remanence vector
is still indeterminate. Therefore, in the isotropic case, the only possibility is to supply the magnetization
vector through the PERMANENT option. Any offset of the B ( H ) -curve, implying
B ≠ 0 at ( H = 0 ) is disregarded in this case. For orthotropic material, it is not allowed to use the
PERMANENT option. In this case, the magnetization can exclusively be specified through the B-H
RELATION option or using the table driven input.
CHAPTER 6 307
It is emphasized that the magnetization curve specified in the B-H relation or H-B relation must be
monotonic and uniquely defined.
Introducing the vector potential A such that:
∇×A = B (6-74)
which automatically satisfies the constraint, Equation (6-68), and we then have the final form:
–1
∇ × μ–1 ( ∇ × A ) = J + μ Br (6-75)
The vector potential ( A ) is not uniquely defined by Equation (6-74). In 2-D magnetostatic simulations,
this indeterminacy is removed by the reduction of A to a scalar quantity. In 3-D situations, Marc uses
the Coulomb gauge for this purpose:
∇•A = 0 (6-76)
Denote the virtual potential by W ; then, the variational formulation is:
–1 –1
μ ( ∇ × W ) • ( ∇ × A )dV =  W • JdV +  W • ( ∇ × μ B r )dV +  W • ( H × n )dA
V V V Γ (6-77)
where n is the outward normal to V at the boundary Γ .
In the three-dimensional case, the Coulomb gauge, Equation (6-76), is enforced with a penalty
formulation. The resulting term added to the variational formulation, Equation (6-77) reads:
–1 –1
μ ( ∇ × W ) • ( ∇ × A )dV =  W • JdV +  W • ( ∇ × μ B r )dV +  W • ( H × n )dA
V V V Γ
(6-78)
+  r ( ∇ • W ) ( ∇ • A ) dV
V
The default value used for r is:
–4 –1
r = 10 μ (6-79)
Using the usual finite element interpolation functions N , the discrete curl operator G , and the weighting
function W = N .
G = ∂N ⁄ ∂y for two-dimensional problems

∂N ⁄ ∂x
(6-80)
∂N ⁄ ∂y – ∂N ⁄ ∂z
G = ∂N ⁄ ∂z – ∂N ⁄ ∂x for three-dimensional problems
∂N ⁄ ∂x – ∂N ⁄ ∂y
Leads to the resulting system of algebraic equations
Ku = F (6-81)
where
T –1
K = G μ G dV (6-82)
V
T T –1 T
F = N N ( J ) dV + G M B r dV +  N N ( H × n ) dA (6-83)
V V Γ
The Coulomb gauge is based upon the principle of conservation of electric charge which in its steady
state form reads:
∇•J = 0 (6-84)
In Marc, it is up to you to specify the current distribution. When doing so, it is recommended to ensure
that this distribution satisfies Equation (6-83). Otherwise, condition Equation (6-76) could be violated.
As a consequence, the results could become less reliable.
From the third term on the right-hand side of Equation (6-83), it becomes clear that the natural boundary
condition in this magnetostatic formulation is H × n , the tangential component of the magnetic field
intensity. Consequently, if no other condition is specified by you, by default a zero tangential magnetic
field intensity at the boundaries is assumed.
Besides, when there are no currents present on a Γ 12 material interface separating two materials 1 and
2, the tangential magnetic field intensity is assumed to be continuous:
n × ( H 1 – H 2 ) = 0 on Γ 12 (6-85)
With H 1 and H 2 , the magnetic field intensities in material 1 and 2, respectively. A discontinuous
tangential magnetic field intensity can be modeled by assigning an appropriate distributed “shear”
current density to the interface. This (surface) current density J is related to H 1 and H 2 by:
n × ( H 1 – H 2 ) = J on Γ 12 (6-86)
Edge Effects at Interface Between Materials of Different Magnetic Permeabilities

At the interface between materials of different permeabilities, there is an abrupt change of permeability
value. In absence of a surface current at the interface, the following holds
B n 1 = B n 2 and H t 1 = H t 2 (6-87)
where
H is the magnetic field intensity.
B is the magnetic flux density.
CHAPTER 6 309
n corresponds to the normal component

t corresponds to the tangential component at interface
1 and 2 corresponds to either side of interface
A surface sheet current can flow at the interface only if one of the materials is a perfect conductor
( σ → ∞ ). Many magnetic materials are conducting too but are not perfect conductors.
For any material we have:
B = μH (6-88)
where
μ is the permeability of the material.
Hence,
Bt 1 Bt 2
-------- = -------
- (6-89)
μ1 μ2
Now,
B = ∇×A (6-90)
where
A is the magnetic vector potential.
The conditions (equations 6-87, (6-89), and (6-90) are satisfied if the normal component of the magnetic
vector potential A is allowed to be discontinuous along the interface normal. The tangential component
of magnetic vector potential A is continuous.
For 2-D problems, only the scalar Z component of the magnetic vector potential A is considered. This
component is always tangent to any material interfaces. Hence, the edge effect is automatically satisfied
for these problems. The edge effects are required only for three-dimensional magnetostatic problems.
In order to obtain accurate results for the magnetic vector potential A, the edge effect requires the
imposition of the Coulomb’s gauge. The Coulomb’s gauge is imposed using a penalty factor. The value
of the correct value of the penalty factor depends on the following factors:
1. Type of element: tetrahedral or hexahedral
2. Order of the element: first or second order.
3. Finite element mesh density and on the uniformity or non uniformity of the mesh. A uniform
mesh requires a lesser value.
The value of penalty factor is quite stable only for a linear hexahedral element and this value is 1.0E-04.
For all other elements, this value lies in different ranges and it is difficult to know the correct value for
any given problem. It is, hence, advisable to use linear hexahedral element with a default penalty value
of 1.0E-04.
This interface conditions is implemented in Marc through the THERMAL CONTACT option, where
materials on either side of the interface have to be defined as electromagnetic contact bodies. There is
now a choice between glued and touching contact (CONTACT TABLE or CONTACT TABLE with
TABLES). With glued contact, all components of the magnetic vector potential are tied and this is
recommended when two dissimilar meshes containing the same material properties need to be connected.
With touching contact, only the tangential components of the magnetic vector potential are tied and thus
satisfying the above interface conditions. This is recommended at interfaces with different material
properties. The glued or touching contact interface condition should be given on the CONTACT TABLE
option. Note that for convenience, Mentat has a special tool to create a disconnected mesh from a uniform
mesh which is called matching boundaries.
Electrical Windings and Coil Current Boundary Condition:

In many electrical problems, a magnetic circuit is excited by electrical windings. A winding is a set of
multi-turn coils. A set of multi-turn coils end at two termination points called terminals. Each coil has the
same cross section with respect to shape and size. Usually, a coil has a rectangular or circular shape. Other
shapes are rarely used. Each coil has a thin insulation coating. The multi-turns are bound together by
some insulation material. The set of multi-turn itself defines a cross section. This is the winding cross
section. A winding cross section is usually rectangular, sometimes circular. A winding cross section
contains the coil cross section and required insulation. The current in each coil is multiplied by the
number of turns (this is called Ampere-turns). It is then assumed that this Ampere-turns flows through
the full winding cross-section.
Electrical windings are specified in the following.
1. Windings:
a. The winding path is defined by a contiguous set of nodes or a contiguous set of polylines.
b. Orientation of the winding cross section.
c. Winding cross-section type: rectangular or circular.
d. Length and breadth for rectangular type and radius for circular type.
e. Number of turns of the coil.
f. Coil cross section type: rectangular or circular.
g. Length and breadth for rectangular type and radius for circular type.
h. Direction of the current with respect to the defined path.
i. Associated electromagnetic contact body identification number. This is used for calculation of
inductance of the winding.
2. Coil current boundary condition
This defines the actual current in each coil which requires the following data.
a. The winding to which this boundary condition applies.
b. The magnitude of current in each coil. For harmonic analysis, the coil current has a
real/imaginary or magnitude/phase component.
c. The geometric type and the list of these geometric entities.
CHAPTER 6 311
The complete winding data is entered in the windings modeling tools. The coil current and the winding
association is entered in the COIL CURRENT boundary condition. To activate this option, use the
EMWINDING and COIL CURRENT model definition options.
Inductance Matrix Computation

An inductor is a device that can store magnetic potential energy. This energy can be thought as the energy
stored in the magnetic field created by the current flowing in the windings.
Self-inductance
The self inductance of a coil is defined as:
NΦ
L = --------- (6-91)
I
where
Φ is the flux linking the coil
I is the current in the coil that produces the flux
N is the number of coil turns
Mutual-inductance
Two coils, C 1 and C 2 , are of the same dimensions and are glued to each other, but separated by an
insulator. Coils C 1 and C 2 have N 1 and N 2 turns, respectively. A current I 1 flows through C 1 , while
the current I 2 in coil C 2 is zero. I 1 produces a magnetic field intensity (H) and flux density (B) and
magnetic energy is stored in this field. The magnetic field intensity (H) depends on the current I 1 , the
coil dimensions and the material surrounding the coil. The magnetic flux produced by coil C 1 and also
passing through the coil C 2 is obtained from
Φ1 = Φ2 =   B • dS (6-92)
coil area
The self inductance of each coil is given by:
N1 Φ1
L 11 = --------------
I1
(6-93)
N2 Φ2
L 22 = --------------
I2
Then mutual-inductance M 21 of coil C 2 with respect to coil C 1 is given by
N2 Φ2
M 21 = -------------- (6-94)
I1
If the roles of the two coils are reversed, then
N1 Φ1
M 12 = -------------- (6-95)
I2
where M 12 is the mutual-inductance of coil C 1 with respect to coil C 2
In this case, the self and mutual inductance are related by

L 11 = M 21 = L 22 = M 12
Hence,
L 11 I L 22 I
- = ---2- and ---------
--------- - = ---1- (6-96)
M 12 I1 M 21 I2
and since
M 12 = M 21
(6-97)
M 12 = L 11 L 22
If the two coils are loosely coupled; i.e., there is leakage flux, then
N1 Φ1 N1 k1 Φ1
L 11 = -------------- and M 12 = -------------------- (6-98)
I1 I2
L 11 I2 k1 < 1
---------
- = ---------
- (6-99)
M 12 k1 I2
Similarly, for the second coil,
L 22 I1 k2 < 1
---------
- = ---------
- (6-100)
21 k2 I2
M
and since M 12 = M 21
M 12 = −
+ k L 11 L 22
(6-101)
2
k = k1 k2 0<k<1
k is called the coupling factor of the coils.

CHAPTER 6 313
The Inductance Matrix:

For ‘n’ coils C 1 , C 2 , C 3 , … C n , each coil has self inductance and mutual inductance with the
remaining coils. For multiple conductors, it is better to express inductance as an inductance matrix. For
‘n’ multiple conductors, the inductance matrix is:
Φ1 L 11 L 12 . . L n I1
Φ2 L 21 L 22 . . L 2 n I2
. = . . . . . . (6-102)
. . . . . . .
Φn Ln 1 Ln 2 . . Ln n In
where
Li j is the self inductance of conductor ‘i’, if i = j
Li j is the mutual inductance between conductor ‘i’ and ‘j’, if i ≠ j .
Note that the inductance matrix is symmetric and L i j = L j i . The terms L i j are also written as M i j . The
magnetic stored energy due to a currents in ‘n’ coils C 1 , C 2 , C 3 , … C n is:
1 2 2 2
W t = --- [ L 11 I 1 + L 22 I 1 + L 33 I 1 + … + L n n I n2 + 2M 12 I 1 I 2 + 2M 23 I 2 I 3 + 2M 34 I 3
2 (6-103)
+ 2M 13 I 1 I 3 + … ]
The total magnetic stored energy can also be written in terms of B and H as,
1
W t = ---    B • H dV (6-104)
2
V
where B is the magnetic induction, H is the magnetic field, V is the total domain, and W is the total
stored magnetic energy
For inductance calculations the two expressions for the stored magnetic energy are equivalent, and Marc
uses this to compute the inductance matrix.
The inductance matrix computation requires the windings feature. The current flowing through each
inductor has to be defined by the EMWINDING model definition and COIL CURRENT boundary
condition. To achieve this, the inductors must be defined as THERMAL CONTACT (electromagnetic
contact) bodies. Each winding has to refer to one of these contact bodies using the EMWINDING model
definition option. The inductors for which the inductance matrix will be computed are selected in the
EMINDUC history definition option.
Differential Inductance Computation

The differential inductance corresponds to the change in inductance due to perturbation in the exciting
currents. For the set of n coils described above, the flux linkage of any coil C 1 is the surface integral of
the magnetic flux density multiplied by the number of turns N 1 . The change in inductance can occur with
current variation for nonlinear magnetic materials. Then, the inductance matrix can be described by the
flux linkage as follows:
[ Φ ] = { L ( [ I ] } [ I ] + [ Φ0 ]
where
[Φ] is the vector of flux linkage in coils.
[I] is the coil current vector.
[ Φ 0 ] is the flux linkage vector when the coil currents are zero and accounts for permanent magnet.
The terminal coil voltage vector is defined by:
∂[Φ] d[L] ∂[I] ∂[ I]

[ V ] = ------------- =  ------------ [ I ] + [ L ] ---------- = { [ L d ] [ I ] } ----------
∂t  d[I]  ∂t ∂t
where [ L d ] is called the differential inductance matrix. The equation shows that the induced voltage is
directly proportional to the differential inductance matrix and the time derivative of the coil current.
Losses in Ferromagnetic Lamination Cores

Materials consisting of multiple layers of ferromagnetic material separated by isolators (laminations) are
always used in all power industry devices; i.e., alternators, synchronous generators, DC generators and
motors, transformers, induction motors, relays and switch gear. Eddy currents and hysteresis is produced
due to time varying magnetic flux density (B) in laminations. For steady state operations, the time
variation is assumed sinusoidal.
These laminations are used to drastically reduce eddy current and hysteresis losses in ferromagnetic
materials. These losses reduce the efficiency of the device, cause undesirable heating and require
additional devices like cooling etc. and it is required to predict them accurately. Accurate prediction goes
a long way in reducing manufacturing and prototyping costs.
Core Losses
Core loss in a magnetic material occurs when the material is subjected to a time varying magnetic flux.
Traditionally, core loss, P c , consists of two components: hysteresis loss, P h , and eddy current loss, P e .
According to the Steinmetz equation, measurement and calculation of core losses are normally made with
sinusoidal flux density of varying magnitude and frequency. These measurements and calculations are
based on the standard coil and frequency modeled by a two term function of the form
CHAPTER 6 315
Pe = Ph + Pe
, (6-105)
= k h fB n + k e f 2 B 2
where f is the frequency of the external magnetic field, B is the flux density, k h , k e , and n are the
coefficients, which depend on the lamination material, thickness, conductivity, as well as other factors.
However, this formula is only applicable under the assumption that the maximum magnetic flux density
of 1.0 Tesla is not exceeded and the hysteresis loop is under the static situation, which is not practical in
electrical machines. When the magnetic flux density is over 1.0 Tesla or the field frequency becomes
high, there is a big discrepancy between the calculation based on (6.97) and experimental results.
Core loss plays an increasingly important role both in the improvement of the quality of electrical steels
at the production stage and in the optimization of their operating conditions. According to the statistical
loss theory, the core loss under sinusoidal flux condition is given by
Pe = Ph + Pe + Pa
2) (6-106)
= k h fB ( a + b B + c B + k e f 2 B 2 + k a f 1.5 B 1.5
where P a is referred to as the excess loss and k a is related to the material thickness, cross-sectional area,
conductivity and a parameter which describes the material microstructure. The coefficients k h , k e , and
k a vary slightly with the magnetic flux density B and frequency f and this variation must be considered
to obtain better accuracy. The coefficients a, b, and c depend on the frequency and this must be
considered in the loss calculations.
In the present procedure, the loss coefficients are evaluated from the original loss data supplied by
manufacturers. The coefficients are obtained mathematically using polynomial curves.
In practice, the lamination steel manufacturers do not present the loss coefficients mentioned above.
They only provide the loss curves and tables in watts per kilogram (W/kg) or watts per pound (W/lb.)
versus the flux density or frequency to indicate the combination of hysteresis loss, eddy loss, and excess
loss. The classical estimation of core loss (6-105) is systematically lower than the measured values.
Ratios of two or more have been observed. Thus, for the analysis and calculation of core loss, the
accuracy of the computed loss coefficients is essential.
To compute lamination losses the materials in question are defined as THERMAL CONTACT
(electromagnetic contact) bodies. Each such contact body is a stack of lamination held together by bolts,
clamps etc. The lamination loss curve is provided by the manufacturer. The loss curve is actually a set
of curves for different operating frequencies. There must be at least three curves in this set in order to
compute lamination losses in Marc. Each curve in this set shows the variation of the loss with the
magnetic induction B. To achieve good accuracy a sufficient number of data points must be extracted
from the curves. The data points must also cover all values of magnetic induction B for the problem in
question. The data points are entered by the TABLE model definition option. For each lamination body,
the loss curve table should be given on the THERMAL CONTACT option. It is given as the table reference
of the scaling factor of the loss curve. Needless to say that this option only works with the table driven
input. The default multiplying factor for the loss curve is 1.
To activate this feature and to set the operating frequency use the EMLAMIN history definition option in
the history definition section.
Electromagnetic Analysis
Marc has the capability to perform both transient (dynamic) and harmonic fully coupled electromagnetic
analysis. This allows Marc to calculate the electrical and magnetic fields subjected to external excitation.
This can be solved for both two- or three-dimensional fields. A vector potential for the magnetic field is
augmented with a scalar potential for the electrical field. If a transient analysis is performed, the magnetic
permeability can be a function of the magnetic field; hence, a nonlinear problem. The elements available
for electromagnetic analysis are described in Table 6-6.
Table 6-6 Element Types for Electromagnetic Analysis
Element
Type Description
111 4-node planar
112 4-node axisymmetric
113 8-node brick
Marc prints the magnetic flux density ( B ), the magnetic field vector ( H ), electric flux density ( D ), and
the electrical field intensity at the integration points. In a harmonic analysis, these have real and
imaginary components. The nodal point data consists of the vector potential A , the scalar potential V ,
the charge Q , and current I .
To activate the electromagnetic option, use the EL-MA parameter. The values of the isotropic permittivity
( ε ), permeability ( μ ), and conductivity ( σ ) are given in the ISOTROPIC option. Orthotropic constants
can be specified using the ORTHOTROPIC option. Optionally, the UEPS, UMU, and USIGMA user
subroutines can be used. A nonlinear permeability can be defined using the B-H relation.
Specify nodal constraints using the FIXED POTENTIAL option. Input nodal currents and charge using the
POINT CURRENT option. Specify distributed currents by using the DIST CURRENT option and
distributed charges by using the DIST CHARGE option. Nonuniform distributed currents and charges can
also be specified by the FORCEM user subroutine.
The electromagnetic capability is linear, unless a nonlinear B-H relation is defined. In such problems,
convergence is reached when the residual satisfies the tolerance defined in the CONTROL option. The
transient capability is only available with a fixed time step; use the DYNAMIC CHANGE option to activate
this option. The resultant quantities can be stored on the post file for processing with Marc Mentat.
In electromagnetic analysis, you can enter the current and/or the charge. In a harmonic analysis, you can
enter both the real and imaginary components or the magnitude and the phase if the table driven input
format is used.
CHAPTER 6 317
Table 6-7 Input Options for Electromagnetic Analysis
Input Options
Load Description Model Definition History Definition User Subroutine
Nodal Current POINT CURRENT POINT CURRENT FORCDT
Nodal Charge POINT CHARGE POINT CHARGE
Distributed Current DIST CURRENT DIST CURRENT FORCEM
Distributed Charge DIST CHARGE DIST CHARGE FORCEM
Technical Formulation
Electromagnetic analysis is based upon the well-known Maxwell’s equations. This has been
implemented in Marc for both transient and harmonic analyses.
Transient Formulation
The Maxwell’s equations are:
·
∇×E+B = 0 (6-107)
·
∇ × H – εE – σ E = 0 (6-108)
·
∇ ⋅ ( εE + σE ) = 0 (6-109)
∇⋅B = 0 (6-110)
where the constitutive relations are
D = εE
B = μ0( H + M )
J = σE
and
E is the electric field intensity
D is the electric flux density
H is the magnetic field intensity
B is the magnetic flux density
J is the current density
M is the magnetization
and
ε is the permittivity
μ is the permeability
σ is the conductivity
μ0 is the permeability of free space.
·
Additionally, we have the conservation of charge: ρ + ∇ ⋅ J = 0 where ρ is the charge density. We
assume that the magnetization vector is given by:
M = χH + M 0 (6-111)
where M 0 is the strength of the permanent magnet and χ is the susceptibility. The magnetic field can be
defined as:
B = μ H + Br (6-112)
in which μ is the permeability, given by:
μ = μ0 ( 1 + χm ) (6-113)
and B r is the remanence, given by:
Br = μ0 M0 (6-114)
Notice that ( 1 + χ m ) is usually called the relative permeability μ r .
A vector magnetic potential A and a scalar potential V are introduced, such that
B = ∇×A (6-115)
E = – ( ∇V + A· ) (6-116)
Note that since only the curl of A is required, an arbitrary specification of the divergence can be made.
The Coulomb gauge is then introduced.
∇⋅A = 0 (6-117)
This is implemented using a penalty condition. It is important to note that E depends not only on the
scalar potential, but also upon the vector potential. Hence, interpretation of V as the usual voltage can
lead to erroneous results.
Substituting into Maxwell’s equations results in:
∇ × [ μ – 1 ( ∇ × A – μ 0 M 0 ) ] + ε ( ∇V· + A·· ) + σ ( ∇V + A· ) = 0 (6-118)
– ∇ ⋅ [ ε ( ∇V· + A·· ) + σ ( ∇V + A· ) ] = 0 (6-119)

CHAPTER 6 319
·
6 It has been assumed that ε = 0 ; in that, the permittivity has a zero time derivative.
Two time stepping schemes have been implemented in Marc for transient electromagnetics. The default
scheme uses the Newmark-Beta algorithm. This discretizes the second-order hyperbolic equations given
in Equations (6-118) and (6-119) and is preferred in mid-to-high frequency situations where the
permittivity-based terms in these equations have a significant influence in the transient response. For
low-frequency systems where the permittivity-based terms can be insignificant, the Newmark-Beta
scheme is sometimes found to produce spurious oscillations in the solutions for the potential, E and D.
In such situations, a second scheme that uses the Backward-Euler algorithm to discretize first-order
equations obtained by dropping the permittivity-based terms in Equations (6-118) and (6-119) is
preferred.
For the Newmark-Beta scheme, the general form is given by:
A n + 1 = A n + ΔtA· n +  --- – β Δt 2 A·· n + βΔt 2 A n + 1

1
Nonstru (6-120)
2 
ctural
and
Coupled A· n + 1 = A· n + ( 1 – γ )ΔtA·· n + γΔtA·· n + 1 (6-121)
Procedu The particular form of the dynamic equations corresponding to the trapezoidal rule:
re
Library 1 1
γ = --- β = --- (6-122)
2 4
results in a symmetric formulation, which is unconditionally stable for linear systems.
For the Backward-Euler scheme, the governing equations are rewritten as
∇ × [ μ – 1 ( ∇ × A – μ 0 M 0 ) ] + σ ( ∇V + A· ) = 0 (6-123)
– ∇ ⋅ [ σ ( ∇V + A· ) ] = 0 (6-124)
The derivative term A· in Equations (6-123) and (6-124) is discretized as
n
An + 1 – A
A· n + 1 = ----------------------------- (6-125)
Δt
In the current formulation, a fixed time step procedure must be used. The time step is defined through
the DYNAMIC CHANGE option.
Harmonic Formulation
In harmonic analysis, it is assumed that the excitation is a sinusoidal function, and the resultant also has
a sinusoidal behavior. This results in the solution of a complex system of equations. In this case,
Maxwell’s equations become:
∇ × E + iωB = 0 (6-126)
∇ × H – iεωE – σE = 0 (6-127)
∇⋅D–ρ = 0 (6-128)
∇⋅B = 0 (6-129)
where ω is the excitation angular frequency and i = –1 .

Additionally, we have the conservation of charge:
iωρ + ∇ ⋅ J = 0 (6-130)
where all symbols are the same as in the discussion above regarding transient behavior. Again, a vector
potential A and a scalar potential V are introduced. In this case, these are complex potentials.
Substituting into the Maxwell’s equations results in:
∇ × [ μ – 1 ( ∇ × A – μ o M o ) ] + σ̃ ( ∇V + iωA ) = 0 (6-131)
– ∇ ⋅ [ σ̃ ( ∇V + iωA ) ] = 0 (6-132)
where σ̃ = σ + iωε
With a little manipulation, a symmetric complex formulation may be obtained. The excitation frequency
is prescribed using the HARMONIC option.
Note that the capability to extract the natural frequencies of a complex system by modal analysis does
not exist in Marc. The harmonic formulation is assumed to be linear; therefore, you should not include
the B-H RELATION option.
Piezoelectric Analysis
The piezoelectric effect is the coupling of stress and electric field in a material. An electric field in the
material causes the material to strain and vice versa. Marc has a fully coupled implementation of
piezoelectric analysis, thus simultaneously solving for the nodal displacements and electric potential.
The elements available for piezoelectric analysis are described in Table 6-8. They can be used in static,
transient dynamic, harmonic, and eigenvalue analysis. The analysis can be geometrically nonlinear but
is materially linear. The piezoelectric elements have their equivalent heat transfer elements, so that they
can also be used in a coupled thermal-piezoelectric analysis. Such a coupled analysis is weakly coupled,
and solved using a staggered approach. When piezoelectric elements are used in a contact analysis with
a node of the piezoelectric element touching a segment of another piezoelectric element, then a
multipoint constraint relation is set up for the nodal displacements as well as for the electric potential.
Table 6-8 Element Types for Piezoelectric Analysis
160 4-node plane stress
161 4-node plane strain
162 4-node axisymmetric
CHAPTER 6 321
Table 6-8 Element Types for Piezoelectric Analysis (continued)

163 8-node brick
Marc prints the stresses ( σ ), the strains ( ε ), the electric displacement ( D ) and the electric field intensity
( E ) at the integration points. The nodal data points consists of the displacements u , forces f , the
potential ϕ , and the charge Q .
To perform a piezoelectric analysis, use the PIEZO parameter. The values of the isotropic, orthotropic
and anisotropic mechanical properties are given in the ISOTROPIC, ORTHOTROPIC, and
ANISOTROPIC model definition option, respectively. The electric constants, and the constants defining
the coupling between the mechanical and electric part can be specified with the PIEZOELECTRIC model
definition option.
Specify nodal constraints using the FIXED DISP option or FIXED POTENTIAL option. Input nodal loads
using the POINT LOAD option or nodal charges using the POINT CHARGE option. Specify distributed
loads with the DIST LOADS option and distributed charges with the DIST CHARGE option. Fixed nodal
displacements and potentials, or nodal forces and charges can also be specified by the FORCDT user
subroutine, nonuniform distributed loads can be specified by the FORCEM user subroutine, and
nonuniform distributed charges can be specified by the FLUX user subroutine. The TABLE option may
also be used to specify spatially and temporarily varying boundary conditions.
The mechanical equilibrium equation for the piezoelectric effect is:
 σ:δε dV =  t ⋅ δu dA +  f ⋅ δu dV (6-133)
V Γ V
and the electrostatic equilibrium equation is (see also Equation (6-55):
 D ⋅ δE dV =  δϕD ⋅ n dA +  ρ V δϕ dV (6-134)
V Γ V
where
σ is the stress tensor
ε is the strain tensor
t is the traction at a point on the surface
u is the displacement
f is the body force per unit volume
D is the electric displacement vector

E is the electric field vector
ρ is the volume charge density
∂
δE = – δϕ is the virtual electric field corresponding to the virtual potential δϕ .
∂x
The constitutive equations to govern piezoelectricity are written for the mechanical behavior as:
E
σ = L :ε – e ⋅ E (6-135)
and for the electrostatic behavior as:
T ε
D = e :ε + ξ ⋅ E (6-136)
where
L is the elastic stiffness
e is the piezoelectric matrix (stress based)
ξ is the permittivity
The superscripts E and ε represent coefficients measured at constant electric field, and constant strain,
respectively. The term with e gives the electro-mechanical coupling in the two consitutive Equations
(6-135) and (6-136).
We approximate the displacements and electrical potential within a finite element as
u = NU
and
ϕ = NΦ
where N contains the shape functions and U and Φ contain the nodal degrees of freedom. The body
forces and charges, as well as the distributed loads and distributed charges, are interpolated in a similar
manner.
The strains and the electric field are given as:
ε = Bu U (6-137)
and
E = Bφ Φ (6-138)
where B u and B φ contain the gradients of N .

CHAPTER 6 323
Denoting the virtual displacement by δU and the virtual potential by δΦ the variational formulations
can be obtained by substituting Equations (6-135) and (6-137) into Equation (6-133):
T T T
 δε σ dV =  δU B u σ dV =  tδU dA +  fδU dV
V V Γ V
T T T T (6-139)
 δU B u LB u U dV +  δU B u eB ϕ Φ dV =  tδU dA +  fδU dV
V V Γ V
T T T
δU K u u U + δU K u ϕ Φ = δU F u
and similarly by substituting Equations (6-136) and (6-138) into Equation (6-134):
T T T
 δE D dV =  –δ Φ B ϕ D dV =  δΦDn dA +  ρ V δΦ dV
V V Γ V
T T T T (6-140)
– δΦ B ϕ eB u U dV +  δΦ B ϕ ξB ϕ Φ dV =  δΦDn dA +  ρ V δΦ dV
V V Γ V
T T T
– δΦ K ϕu U + δΦ K ϕ ϕ Φ = δΦ ρ ϕ
The final set of equations in matrix form is then:
Ku u Ku ϕ u = Fu
(6-141)
–K ϕ u Kϕ ϕ Φ ρϕ
Strain Based Piezoelectric Coupling

It is also possible to apply strain based coefficients for the piezoelectric coupling matrix. Then the
consitutive equations are for the mechanical behavior:
ε = C:σ + d ⋅ E (6-142)
and for the electrostatic behavior:
D = d:σ + ξ∗ ⋅ E (6-143)
where
C is the elastic compliance
d is the piezoelectric matrix (strain based)
ξ∗ is the permittivity (strain based)
When d and ξ∗ are given, Marc converts this into stress based properties, where e = L:d and
T
ξ = ξ∗ – e ⋅ d .
Acoustic Analysis
Marc has the capability to perform acoustic analysis in a rigid as well as a deformable cavity. This allows
the program to calculate the fundamental frequencies of the cavity, as well as the pressure distribution in
the cavity. This can be solved for two- or three-dimensional fields.
Rigid Cavity Acoustic Analysis

The acoustic problem with rigid reflecting boundaries is a purely linear problem analogous to dynamic
analysis, but using the heat transfer elements.
The elements which are available for acoustic analysis of a rigid cavity are described in Table 6-9.
Marc computes and prints the following quantities: pressure and pressure gradient at the integration
points. The nodal point data consists of the pressure and the source.
Table 6-9 Element Types for Rigid Cavity Acoustic Analysis
37, 39, 131, 41 3-, 4-, 6-, 8-node planar
133 10-node tetrahedral
The wave equation in an inviscid fluid can be expressed in terms of the pressure p as:
2
∂--------p-
= c2∇2p (6-144)
2
∂t
where c is the sonic velocity:
c = K⁄ρ (6-145)
where K is the bulk modulus and ρ is the density.

CHAPTER 6 325
Equation (6-144) can be rewritten as:

2
2 ∂ p
K∇ p – ρ --------- = 0 (6-146)
2
∂t
Where the source terms are neglected, note that this is analogous to the dynamic equation of motion.
The modeling of rigid reflecting boundaries can be done as follows.
Mathematically, a reflecting boundary is described by:
∂p
------ = 0 (6-147)
∂n
∂p
where ------ is the pressure gradient normal to the reflecting surface.
∂n
This is analogous to an insulated boundary in heat transfer. Hence, a reflecting boundary can be
modeled by a set of nodes at the outer surface of the area which are not connected to another part of the
medium. A reflecting plate in the middle of an acoustic medium can be modeled by double nodes at the
same location.
Note: Where there are no boundary conditions applied, there is a zero-valued eigenvalue,
corresponding to a constant pressure mode. Hence, you need to have a nonzero initial
shift point.
To activate the acoustic option, use the ACOUSTIC parameter. The number of modes to be extracted
should also be included on this parameter. The bulk modulus and the density of the fluid are given in the
ISOTROPIC option.
Specify nodal constraints using the FIXED PRESSURE option. Input nodal sources using the POINT
SOURCE option. Specify distributed sources using the DIST SOURCES option. If nonuniform sources
are required, apply these via the FLUX user subroutine or use the TABLE option.
To obtain the fundamental frequencies, use the MODAL SHAPE option after the END OPTION. The
nodes can be used in a transient analysis by invoking the DYNAMIC CHANGE option. The point and
distributed sources could be a function of time.
Fluid Mechanics
Marc has the capability to perform fluid flow analysis. Marc solves the Navier-Stokes equations in the
fluid under the restrictions that the fluid is considered to be nonreactive, incompressible, single phases,
and laminar. The capabilities in Marc can be applied to four different problems listed below:
Fluid behavior only
Fluid-thermal coupled behavior
Fluid-solid coupled behavior
Fluid-thermal-solid coupled behavior
Mass Conservation
The principle of mass conservation for a single-phase fluid can be expressed in differential form as:
Dρ
-------- + ρ∇ ⋅ v = 0 (6-148)
Dt
where ρ is the mass density, v is the Eulerian fluid velocity, and t is the time.
Momentum Conservation
The principle of conservation of linear momentum results in:
∂v
ρ  ------ + v ⋅ ∇v = ρf + ∇ ⋅ σ (6-149)
 ∂t 
where σ is the stress tensor and f is the body force per unit mass.
In general, the stress tensor can be written as the sum of the hydrostatic stress and the deviatoric stress as.
tr ( σ ) 1
------------- = --- σ i j δ i j (6-150)
3 3
1 tr ( σ )
σ d = σ i j – --- σ i j δ i j = σ i j – ------------- δ i j (6-151)
3 3
In fluid mechanics, the fluid pressure is introduced as the negative hydrostatic pressure:
trσ
p = – -------- (6-152)
3
Energy Conservation
For incompressible fluids, the principle of conservation of thermal energy is expressed by:
∂T ∂T ∂q i
ρC p  ------- + v i -------- = – -------- + H (6-153)
 ∂t ∂x i ∂x i
where T is the temperature, C p is the specific heat at constant pressure, q i is the thermal flux, and H
is the internal heat generated.
Equation of State
The equation of state for a homogeneous, single-phase gas is:
mRT
pV = ------------- (6-154)
M
CHAPTER 6 327
m is the mass of the fluid, M is the molecular weight of the gas, R is the universal gas constant, V is
the volume that the gas occupies.
The gas constant is often expressed as R = R ⁄ M for each substance. The equation of state is then
written as:
P
ρ = -------- (6-155)
RT
In Marc, it is assumed that the fluid is incompressible. In such cases, the density is constant ρ = ρ o
Constitutive Relations
The shear strain rate tensor is defined:
1 ∂v i ∂v j
ε i j = ---  -------- + -------- (6-156)
2  ∂x j ∂x i
For viscous incompressible fluids, it is assumed that there is an expression:
d · ·
σ i j = 2με i j = μγ i j (6-157)
where μ is the dynamic viscosity.

If the viscosity is not a function of the strain rate, the material is considered Newtonian. Viscous fluid
flow is usually characterized by the Reynolds number Re :
ρLv
Re = ---------- (6-158)
μ
where v is a typical velocity and L is the characteristic length.

Viscosity for common fluids like air and water is fairly constant over a broad range of temperatures.
However, many materials have viscosity which is strongly dependent on the shear strain rate. These
materials include glass, concrete, oil, paint, and food products. There are several non-Newtonian fluid
models in Marc to describe the viscosity as described below.
Piecewise Non-Newtonian Flow

· · 2· · 1 ⁄ 2
ε is the equivalent strain rate = ε =  --- ε i j ε i j (6-159)
3 
·
μ = μ ( ε ) , where μ is a piecewise linear function.
Bingham Fluid
A Bingham fluid behaves as a “rigid” fluid if the stress is below a certain level, labeled the yield stress,
and behaves in a nonlinear manner at higher stress levels.
d · · ·
σ i j = μ o γ i j + gγ i j ⁄ γ if σ ≥ g (6-160)
·
γ i j = 0 if σ < g (6-161)
1
---
· · 1· · 2
The effective viscosity is μ = μ o + g ⁄ γ and γ =  --- γ i j γ i j
2 
The materials that can be simulated with this model are cement, slurries, and pastes.
Power Law Fluid

The fluid is represented as:
d · n – 1·
σi j = μo K ( γ ) γi j (6-162)
where K is a nondimensional constant.

This model is useful for simulating flow of rubber solutions, adhesives, and biological fluids.
Carreau Model
This model alleviates the difficulties associated with the power law model and accounts for the lower and
upper limiting viscosities for an extreme value of the equivalent shear strain rate.
2 · 2 (n – 1) ⁄ 2
μ = μ∞ + ( μo – μ∞ ) ( 1 + τ γ ) (6-163)
where μ o and μ ∞ are the viscosity at time equals zero and infinity, respectively.
The thermal flux is governed by the Fourier law.

∂T
q i = k i j -------- (6-164)
∂x j
Finite Element Formulation

In the procedures developed, the fluid-thermal coupling can be treated as weak or strong. In the weak
formulation, the solution of the fluid and thermal equilibrium equations are solved in a staggered manner.
In the strongly coupled approach, a simultaneous solution is obtained.
This section cannot cover all of the details of the finite element formulation. It does discuss the final form
of the linearized set of equations and some of the consequences. Beginning with degrees of freedom in a
system – namely, v, p, and T for the mixed method – we utilize the traditional finite element interpolation
functions to relate the values within the element to the nodal values.
Because the pressure stabilizing Petrov-Galerkin (PSPG) method is employed in Marc, equal order
interpolation functions can be used for the velocity and the pressure. The method of weighted residuals
is used to solve the coupled Navier-Stokes equations. Based upon the conservation laws of momentum
(mass and energy), we obtain first order differential equations in the form:
CHAPTER 6 329
Mv· + A ( v )v + K ( T, v )v – Cp + B ( T )T = F ( t ) (6-165)
T
C v = 0 (6-166)
NT· + D ( v )T + L ( T )T = Q ( v, t ) (6-167)
where the first equation is obtained from momentum, the second from mass, and the third from energy
conservation, respectively. This is expressed in matrix form as:
M 0 0  v·  A + K B –C  v 
 
 F 
 
0 N 0  T +
· 0 D+L 0  T  =  Q  (6-168)
  T    
0 0 0  p·  –C 0 0  p   0 
if the penalty method is employed

–1 T
p = χM P C v (6-169)
resulting in
M 0  v·  + A + K + χCM P C  v   F 
–1 T
B
 ·    =   (6-170)
0 N  T  0 D+L  T   Q 
The introduction of the streamline upwinding technique (SUPG) developed by Brooks and Hughes is a
major improvement for the stability of the fluid equations. This procedure controls the velocity
oscillations induced by the advection terms. Effectively, this procedure adds artificial viscosity to the true
viscosity. The second stabilization method, PSPG, allows equal order velocity and pressure interpolation
functions to be used without inducing oscillations in the pressure field. The PSPG term is not included
when using the penalty formulation. The details of these stabilization terms are not provided here, but
note that the magnitude of the contributions are dependent upon the element size, viscosity, and the time
step. Including these contributions, the semi-discrete set of equations takes the form:
M + Mδ 0 0  A + K + K δ B –( C + Cδ )
v·   v   F + Fδ 
     
Mε N 0  T·  + Kε D+L 0  T  =  Q + Qε  (6-171)
     
0 0 0  p·  –C
T
0 Cε  p   0 
The terms K d is included for SUPG and C e and G e for PSPG. The other terms are neglected,
leading to:
M 0 0  v·   v   F 
A + K + Kδ B –C
   
0 N 0  T·  + 0 D+L 0  T  =  Q  (6-172)
 ·  T    
0 0 0  p  –C 0 Cε  p   0 
These submatrices can be interpreted as follows:

A represents advection of momentum
D represents advection of energy
K represents diffusion of momentum or viscosity matrix
L represents diffusion of energy or conductance matrix
M represents mass
N represents heat capacitance
B represents buoyancy
Kδ represents SPUG stabilization matrix
C represents gradient matrix

CT represents divergence matrix
Ce represents PSPG stabilization matrix
F represents externally applied forces

Q represents externally applied fluxes
Note that typically D , K , B , L , Q , and F are dependent upon the temperature. If a non-Newtonian
fluid is used, A , K , D , and G are dependent upon the rate of strain. If the fluid is subjected to large
motions, F can also depend on the total displacement.
In terms of physical parameter, excluding geometry, observe:
A = A ( ρo )
D = D ( ρ o, C p ) = D ( ρ o, C p, T )
K = K ( μ ) = K ( μ, T )
L = L ( K ) = L ( K, T )
M = M ( ρo )
CHAPTER 6 331
N = N ( ρ o, C p )
B = B ( ρ 0, g , B )
where:
ρo is the initial density
Cp is the specific heat
K is the thermal conductivity

g is the gravity
B is the coefficient of thermal expansion
Penalty Method
The penalty method is an alternative method to satisfy the incompressibility constraints. The objective
is to add another term to the operator matrix (viscosity matrix) such that incompressibility is satisfied.
Effectively, K is replaced with K c where:
K c = K + χK p (6-173)
where χ is a large number and K p the penalty matrix. K p can be written as:
–1 T
K p = CM p C (6-174)
where C and M p are functions of the geometry and shape functions only. The value of χ is typically
between 10 5 to 10 9 .
The penalty method should not be used for three-node planar elements, or four-node
tetrahedral elements.

It is possible to simplify the equations governing fluid flow by assuming that the time derivatives of
the velocity and temperature are zero. This is achieved by using the STEADY STATE option. For
thermally coupled, non-Newtonian flows, you still have a highly nonlinear system and multiple
iterations are required.
Transient Analysis
The equations discussed above are still differential equations because of the presence of time derivatives
of velocity and temperature. The conversion of the equations to fully discrete linear equations requires
an assumption of the behavior during the increment. The approximations inevitably result in problems
with accuracy, artificial damping, and/or stability problems.
Marc uses the first order backward Euler procedure such that:
v· n + 1 = ( v n + 1 – v n ) ⁄ Δt (6-175)
This is substituted into the equations to yield:
A + K + K δ + M ⁄ Δt B –C  v   F – M ⁄ Δt ⋅ v n 
   
0 D + L + N ⁄ Δt 0  T  =  Q + Q – N ⁄ Δt ⋅ T  (6-176)
ε
T    n

–C 0 Cε  p   0 
Marc allows either fixed time step or adaptive time step procedure.
The equations of fluid flow are highly nonlinear even for Newtonian behavior because of the inclusion
of the advection terms. The Newton-Raphson and direct substitution procedures are available to solve
these problems.
Solid Analysis
Solid can be modeled in two ways:
• The first is to model them as a fluid with a very large viscosity. In such cases, an Eulerian
procedure is used throughout the model. The constitutive laws are limited to the fluid
relationships.
• The second is to model them as true solids. In such cases, a Lagrangian or updated Lagrangian
approach is used in the solid.
The complete Marc constitutive routines are available for representing the solid behavior.
An important consideration is the interface between the fluid and solid. Large structural deformation or
fluid motion are not supported in the coupled Fluid-Solid capability. For small changes in the structural
deformation, the nodes at the interface are updated. The pressure is transmitted between the fluid and
the solid.
Solution of Coupled Problems in Fluids

The coupled fluid-thermal problem can be solved using either a tightly coupled procedure as shown
above or in a staggered manner. Problems such as free convection are inherently coupled and are best
solved using the tightly coupled procedure.
The fluid-solid interaction problem is solved in a weakly coupled manner. Every attempt has been made
to maximize the computational efficiency. It should be noted that, while the staggered procedure can
result in more global iterations, there are several motivating factors for solving the system in this manner.
• In solving fluid problems, especially in three dimensions, very large systems of equations are
obtained. Any procedure that reduces the number of equations (for example, excluding the solid)
is potentially beneficial.
CHAPTER 6 333
• The fluid flow solution always requires the solution of a nonsymmetric system. In a weakly
staggered analysis, the structural problem can still be solved with a symmetric solver.
• When using an iterative solver, the solution of the strongly coupled system will result in an
ill-conditioned system, which results in poor convergence. This is because the terms associated
with the fluid, thermal, and structural operator are of several different orders of magnitude.
The problem can be divided into two regions (fluid and solid) to accommodate these requirements. In
each region, a different solver can be invoked; different nodal optimizers can be used; and a different
memory allocation can be performed. Often, the fluid uses an out-of-core nonsymmetric solver while the
structure uses an in-core symmetric solver.
Degrees of Freedom
The degrees of freedom in a fluid analysis are the velocities or, for the mixed method, the velocity and
the pressure. When the pressure is not included explicitly as a degree of freedom, the incompressibility
constraint is imposed using a penalty approach.
If a strongly coupled fluid-thermal problem is solved, the degrees of freedom are either the velocities and
temperatures or the velocities, pressure, and temperature. For a three dimensional problem, the number
of degrees of freedom could be 5.
The degrees of freedom associated with the solid are the conventional displacements or the
displacements and temperatures.
Element Types
The fluid region can be represented using the conventional displacement elements in Marc. When using
the mixed method, the pressure has the same order and interpolation functions as the velocity. In a
coupled fluid-thermal analysis, the temperature also has the same order as the velocity. Hence, the
element types available are:
Planar
Type
4-node isoparametric bilinear 11
4-node isoparametric bilinear reduced integration 115
6-node isoparametric triangle 125
8-node isoparametric biquadratic 27
8-node isoparametric biquadratic reduced integration 54
Axisymmetric
Type
4-node isoparametric bilinear 10
4-node isoparametric bilinear reduced integration 116
Type
6-node isoparametric triangle 126
8-node isoparametric biquadratic 28
8-node isoparametric biquadratic reduced integration 55
Three Dimensional
Type
8-node trilinear brick 7
8-node trilinear brick with reduced integration 117
20-node brick 21
20-node brick with reduced integration 57
The three-node triangular elements or the four-node tetrahedral element give poor results when used with
the penalty formulation.
The shape functions used are identical to those used for structural analyses and can be found in any finite
element textbook.
All of the mesh generation capabilities in Mentat can be used to generate the fluid mesh. Furthermore,
Mentat can be used to visualize the results in a manner consistent with structural analyses.
Marc’s implementation of the Navier-Stokes equations utilizes the natural boundary condition.
At nodal points, you can prescribe the time dependent values of the velocity and temperature. This can
be achieved by using the FIXED VELOCITY, FIXED TEMPERATURE, VELOCITY CHANGE, and/or
TEMP CHANGE options. The FORCDT user subroutine can be used for time-dependent behavior or the
TABLE option.
Gravity and centrifugal loads can also be applied.
For coupled fluid-solid interaction problems, the pressure of the fluid is automatically applied to the
structure. The resultant deformation of the structure is applied to the fluid boundaries. In the current
release, only small deformations of the solid are permitted.
The viscosity, mass density, conductivity, and specific head are defined through the ISOTROPIC option.
The STRAIN RATE or TABLE option is used to define non-Newtonian viscous behavior. The penalty
parameter can be entered through the PARAMETERS option.
Note: In fluid analysis, data for POINT LOAD and DIST LOADS should be prescribed as total
rather than incremental quantity (as used in mechanical analysis). Similarly, POINT FLUX
and DIST FLUXES for heat transfer analysis are also given as total quantity. This
specification is to be used consistently for the fluid and/or heat transfer portion of analysis
in coupled fluid-solid, fluid-thermal, and fluid-thermal-solid.
CHAPTER 6 335
Coupled Analyses
The definition of coupled systems includes the multiple domains and independent or dependent variables
describing different physical systems. In the situation with multiple domains, the solution for both
domains is obtained simultaneously. Similarly, the dependent variables cannot be condensed out of the
equilibrium equations explicitly.
Coupled systems can be classified into two categories:
1. Interface variables coupling: In this class of problems, the coupling occurs through the interfaces
of the domain. The domains can be physically different (for example, fluid-solid interaction) or
physically the same but with different discretization (for example, mesh partition with
explicit/implicit procedures in different domains).
Figure 6-39 Fluid-structure Interaction (Physically Different Domains)
2. Field variables coupling: In this, the domain can be the same or different. The coupling occurs
through the governing differential equations describing different physical phenomenon; for
example, coupled thermo-mechanical problems.
Figure 6-40 Metal Extrusion with Plastic Flow Coupled with Thermal Field
Marc can solve the following types of coupled problems: fluid-solid, fluid-thermal, fluid-solid-thermal
interaction, piezoelectric, electrostatic-structural, thermo-electrical (Joule heating), thermo-mechanical,
electrical-thermal-mechanical (Joule Mechanical), fluid-soil (pore pressure), electromagnetic, and
electromagnetic-thermal. The type of coupling is summarized in Table 6-10.
6
Nonstructural and Coupled Procedure
Library
Table 6-10 Summary of Coupled Procedures
Interaction Category Coupling
Thermal- Mechanical Field Weak
Fluid-Solid
added mass approach Interface Weak
general approach Interface Weak
Fluid-Thermal
channel approach Interface Weak
general approach Field Strong or Weak
Fluid-Thermal-Solid Field/Interface Strong or Weak/Weak
Piezoelectric Field Strong
Electrostatic-Structural Field Weak
Magnetostatic-Structural Field Weak
Thermal-Electric (Joule) Field Weak
Electrical-Thermal- Mechanical Field Weak
Magnetostatic-Thermal Field Weak
Fluid-Pore Field Strong
Electromagnetic Field Strong
Electromagnetic-Thermal Field Weak
There are two approaches for solving fluid-solid interaction problems. In the first approach, the fluid is
assumed to be inviscid and incompressible. The effect of the fluid is to augment the mass matrix of a
structure. Modal shapes can be obtained for a fluid/solid system; the modal superposition procedure
predicts the dynamic behavior of the coupled fluid/solid system. This prediction is based on extracted
modal shapes. This method is discussed later.
The general fluid-solid capability models nonlinear transient behavior and is discussed in the fluid
mechanics section.
There are two approaches for solving fluid-thermal interaction problems. The first is for a fluid
constricted to move through thin areas and is implemented using the CHANNEL option. This is discussed
earlier in this chapter. The second approach solves the coupled Navier-Stokes equations and is discussed
in the fluid mechanics section.
The fluid-thermal-mechanical capability is described in more detail in the fluid mechanics section.
In the coupled thermo-electrical problem, the coupling takes place through the temperature-dependent
electrical conductivity in the electrical problem and the internal heat generation caused by electrical flow
in the thermal problem. The program solves for the voltage and temperature distribution.
Similarly, the coupling between the thermal and mechanical problems takes place through the
temperature-dependent material properties in the mechanical (stress) problem and the internal heat
CHAPTER 6 337
generation in the mechanical problem caused by plastic work, which serves as input for the heat transfer
problem. The temperature distribution and displacements are obtained.
In each of the coupled problems described above, two analyses are performed in each load/time
increment. Iterations can also be carried out within each increment to improve the convergence of the
coupled thermo-electrical and thermo-mechanical solutions.
In the coupled fluid-soil model, the fluid is assumed to be inviscid and incompressible. The effect of
the fluid is to augment the stress in the soil material to satisfy equilibrium, and to influence the soil’s
material behavior.
In the electromagnetic analysis, the fully coupled Maxwell’s equations are solved.
In the latter two analyses, the equations (fluid flow/structural or electrical/magnetic) are
solved simultaneously.
Thermal Mechanically Coupled Analysis

Many operations performed in the metal forming industry (such as casting, extrusion, sheet rolling, and
stamping) can require a coupled thermo-mechanical analysis. The observed physical phenomena must
be modeled by a coupled analysis if the following conditions pertain:
• The body undergoes large deformations such that there is a change in the boundary conditions
associated with the heat transfer problem.
• Deformation converts mechanical work into heat through an irreversible process which is large
relative to other heat sources.
In either case, a change in the temperature distribution contributes to the deformation of the body through
thermal strains and influences the material properties.
Marc has a capability that allows you to perform mechanically coupled analysis. This capability is
available for all stress elements and for small displacement, total Lagrangian, updated Lagrangian, or
rigid plastic analysis.
The COUPLE parameter is used to invoke this option. When defining the mesh, if you specify the
element as a stress type through the CONNECTIVITY option, Marc generates an associated heat transfer
element, if possible. The region having an associated heat transfer element has coupled behavior. If you
specify the element as a heat transfer type through the CONNECTIVITY option, that region is considered
rigid. Only heat transfer is performed in that region.
Marc uses a staggered solution procedure in coupled thermo-mechanical analysis. It first performs a heat
transfer analysis, then a stress analysis. Use the CONTROL option to enter the control tolerances used in
the analysis.
Depending on the type of stress analysis performed, Marc can perform three different types of
coupled analysis:
• Quasi-static coupled analysis: This comprises of a transient heat transfer pass and a static
mechanical pass. Fixed stepping can be specified using TRANSIENT NON AUTO and
adaptive stepping can use TRANSIENT or AUTO STEP. AUTO STEP is preferred over the
TRANSIENT option.
• Creep coupled analysis: This comprises of a transient heat transfer pass and a creep mechanical
pass. Fixed stepping can be specified using CREEP INCREMENT and adaptive stepping can use
AUTO CREEP or AUTO STEP
• Dynamic coupled analysis: This comprises of a transient heat transfer pass and a dynamic
mechanical pass. Fixed stepping can be specified using DYNAMIC CHANGE and adaptive
stepping can use AUTO STEP.
Load control and time step control can be specified in either of two ways:
• A fixed time step/load size can be specified by using the TRANSIENT NON AUTO option, the
CREEP INCREMENT option, or the DYNAMIC CHANGE option. In these cases, mechanical loads
and deformations are incremental quantities that are applied to each step. Fluxes are total
quantities. Time variations for the mechanical and thermal loads can be specified using
appropriate s.
• An adaptive time step/load size can be specified by using the AUTO CREEP option or the AUTO
STEP option. In this case, mechanical loads and deformations are entered as the total quantities
that are applied over the load set. Fluxes are total quantities. By default, mechanical loads are
linearly increased while thermal loads are applied instantaneously.
• Exercise caution when applying boundary conditions. Use the FIXED DISP, FIXED
TEMPERATURE, DISP CHANGE, TEMP CHANGE, or TABLE options for mechanical or thermal
boundary conditions, respectively.
There are two primary causes of coupling. First, coupling occurs when deformations result in a change
in the associated heat transfer problem. Such a change can be due to either large deformation or contact.
Large deformation effects are coupled into the heat transfer problem only if the LARGE STRAIN or
UPDATE parameter is invoked. The gap element in Marc (Type 12) has been modified so that if no
contact occurs, the gap element acts as a perfect insulator. When contact does occur, the gap element acts
as a perfect conductor.
The second cause of coupling is heat generated due to inelastic deformation. The irreversibility of plastic
flow causes an increase in the amount of entropy in the body. This can be expressed as:
·p
TS· = fW (6-177)
· p is the fraction of the rate of plastic work dissipated into heat. Farren and Taylor measure f as
where fW
approximately 0.9 for most metals. Using the mechanical equivalent of heat ( M ), the rate of specific
volumetric flux is:
·p⁄ρ
R = MfW (6-178)
Use the CONVERT model definition option to define Mf .

Of course, all mechanical and thermal material properties can be temperature dependent. The governing
matrix equations can be expressed as:
Mu·· + Du· + K ( T ,u ,t )u = F (6-179)

CHAPTER 6 339
C T ( T )T· + K T ( T )T = Q + Q I + Q F (6-180)
I F
In Equation (6-180), Q represents the amount of heat generated due to plastic work and Q represents
the amount of heat generated due to friction. The specific heat matrix C T and conductivity matrix K T
can be evaluated in the current configuration if the updated Lagrange option is used.
Note: All terms except M can be temperature dependent.
Coupled Acoustic-Structural Analysis

In a coupled acoustic-structural analysis, both the acoustic medium and the structure are modeled. In this
way, the effect of the acoustic medium on the dynamic response of the structure and of the structure on
the dynamic response of the acoustic medium can be taken into account. Such a coupled analysis is
especially important when the natural frequencies of the acoustic medium and the structure are in the
same range. In Marc, only a harmonic coupled acoustic-structural analysis can be performed. Since the
interface between the acoustic medium and the structure is determined automatically by Marc based on
the CONTACT option, setting up the finite element model is relatively easy since the meshes do not need
to be identical at the interface.
The acoustic medium will be called the fluid, although it might also be a gas, and is considered to be
inviscid and compressible. The equilibrium equation is given by:
∇p + ρ f u·· = 0 (6-181)
in which p is the pressure, ρ f is the fluid density, u is the displacement vector and a superposed dot
indicates differentiation with respect to time.
Using the bulk modulus of the fluid K f , the constitutive behavior can be written as:
p = – K f ∇u (6-182)
The Equations (6-181) and (6-182) can be combined to:
1 1
------ p·· – ----- ∇2p = 0 (6-183)
Kf ρf
It is also possible to add a damping term ru· to Equation (6-181), with r the resistivity or fluid drag:
∇p + ru· + ρ f u·· = 0 (6-184)
Since we restrict the discussion to harmonic analyses, we can write u· = iωu , with ω the excitation
frequency and i the imaginary unit, so that Equation (6-184) reduces to:
1 1
------ p·· – ----- ∇2p = 0 (6-185)
Kf ρf
in which the complex density ρ f is given by:
ρ f =  ρ f – ----
ir
(6-186)
 ω
The weak form of Equation (6-185) is obtained in a standard way by introducing the variational field δp
and integrating over the fluid volume V :
 1 ·· 1 
 δp  K
------ p – ----- ∇2p dV
f ρ 
= 0 (6-187)
V f
Applying the Green’s divergence theorem to this expression yields:
 1 ·· 1  1 ∂p
  δp -----
Kf
- p + ----- ∇δp ∇p dV +  ----- δp ----- dA
ρ  ρ ∂n
= 0 (6-188)
V f Γ f
in which Γ represents the boundary of the fluid with an inward normal n .

Along the boundary, various conditions can occur:
• The pressure p is prescribed on the boundary Γ p .
• The normal acceleration is prescribed to be a n on the accelerating boundary Γ a .
• The normal acceleration of the fluid equals the normal acceleration of the solid, u·· f n = u·· s n ,
on the fluid-structure interface Γ f s .
• Nonreflecting or reactive boundary conditions are introduced using a spring-dashpot analogy on

1 ∂p
Γ as --- ----- = p· ⁄ c + p·· ⁄ k . In this way, a spring and a dashpot are placed in series between
fn ρ ∂n I I
the acoustic medium and its boundary, with k I the spring and c I the dashpot parameter, both per
unit area. Some classical conditions can be modeled by this expression:
∂p
1. Free surfaces waves in a gravity field: ----- = p·· ⁄ g , with g the gravity acceleration;
∂n
∂p
2. The plane wave radiation boundary condition: ----- = p· ⁄ c , with c the wave velocity;
∂n
3. Using the complex admittance 1 ⁄ z ( ω ) of the boundary, with z ( ω ) the impedance, the normal
velocity can be related to the pressure by u· f n = p ⁄ z ( ω ) = ( 1 ⁄ c I + iω ⁄ k I )p , so that a given
z ( ω ) can be modeled by setting 1 ⁄ c I and 1 ⁄ k I .
Upon discretizing the fluid and the solid using finite elements, the fluid pressure and the displacements
of the solid can be approximated in a standard way, so that Equation (6-188) yields:
CHAPTER 6 341
 T p 1 pT T p 1 pT  T p T p uT
  δp N ------ N p·· + δp ( ∇N ) ----- ( ∇N )p dV –
Kf ρf 
 δp N a n dA –  δp N N Gu·· dA
V Γ Γ
a fs
p 1 pT · p 1 p T ··
(6-189)
δp  N ---- N p + N ---- N p dA = 0
T
+   cI kI 
Γ
fn
where p and u contain the pressure and displacement degrees of freedom, N contains the interpolation
functions and G is a transformation matrix to relate the displacements in normal direction to the global
displacement components. Since Equation (6-189) must be valid for arbitrary admissible values of δp ,
· ·· 2 ·· 2
evaluating the various integrals and using p = iωp , p = – ω p and u = – ω u , results in:
 2  2
 K f + iωC f – ω M f p + ω S f s u = F f (6-190)
 
In order to end up with a set of coupled equations, we need to consider the solid as well. The effect of
the fluid on the solid originates from the fluid pressure at the fluid-solid interface. This can be easily
included by evaluating the virtual work corresponding to the fluid pressure:
p T T u pT T T
δW =  – δ u n s p dA =  –δ u G N N p dA = – δu S f s p (6-191)
Γ Γ
fs fs
Notice that the minus sign has been used to express that u ns is positive in the outward direction of the
solid. Including Equation (6-191), the solid behavior for a harmonic analysis is governed by:
 2  T
 K s + iωC s – ω M s u + S f s p = F s (6-192)
 
with K s the stiffness matrix, which can include stress-stiffening, C s is the damping matrix, M s the mass
matrix, and F s the external load vector, except for the fluid pressure.
–2
Premultiplying Equation (6-190) by ω and combining the result with Equation (6-192) gives the
desired coupled complex equation system:
–2
ω Af Sf s p
–2
ω Ff
= (6-193)
T u Fs
Sf s As
in which A f = K f + iωC f and A s = K s + iωC s .

The procedure to perform a coupled acoustic-structural is as follows. The acoustic medium and the
structure are modeled separately; the acoustic structure using heat transfer elements with acoustic
material properties and the structure using conventional stress elements. The elements representing the
acoustic medium are assigned to an acoustic contact body and the elements representing the solid to a
deformable contact body. If the acoustic medium is not surrounded completely by a deformable structure
and one wants to, for example, model radiation boundary conditions along this part of the acoustic
medium, then one can define rigid contact bodies in this area. Like deformable bodies, rigid bodies can
be used to define the spring-dashpot analogy defined before. A harmonic analysis can be performed
based on initial contact, so using the undeformed configuration of the contact bodies, but also after a
preload of the deformable contact bodies. In the latter case, it might be necessary to remesh the acoustic
and/or deformable bodies before the harmonic analysis can be performed, since during the static loading
of the deformable bodies the deformation of the acoustic bodies is not taken into account.
The elements which are available for coupled-structural acoustic analysis are described in Table 6-11.
Table 6-11 Element Types for Coupled-Structural Acoustic Analysis

37, 39 3-, 4-node planar
38, 40 3-, 4-node axisymmetric
Marc computes and prints the following quantities: pressure and pressure gradient at the integration
points. The nodal point data consists of the pressure and the source.
The ACOUSTIC parameter is used to indicate that a coupled acoustic-structural analysis is performed. In
addition to the CONTACT option, the ACOUSTIC and REGION model definition options are used to
define the material properties of the acoustic medium and to set which elements correspond to the solid
and the fluid region. Typical boundary conditions for the acoustic medium can be entered using
the FIXED PRESSURE model definition and PRESS CHANGE history definition option. The DIST
SOURCES and POINT SOURCE model definition and DIST SOURCES and POINT SOURCE history
definition options are used to define the load on the acoustic medium. The HARMONIC history definition
option is used to define the excitation frequencies. Spatially and frequency dependent boundary
conditions may be defined by referencing the TABLE option.
Fluid/Solid Interaction – Added Mass Approach

The fluid/solid interaction procedure investigates structures that are either immersed in, or contain a
fluid. Examples of problems that make use of this feature are vibration of dams, ship hulls, and tanks
containing liquids.
Marc is capable of predicting the dynamic behavior of a structural system that is subject to the pressure
loading of fluid. The fluid is assumed to be inviscid and incompressible; for example, water.
CHAPTER 6 343
In Marc, the fluid is modeled with heat transfer elements (potential theory) and the structure is modeled
with normal stress or displacement elements. The element choice must ensure that the interface between
the structural and fluid models has compatible interpolation; that is, both solid and fluid elements are
either first order or second order. The TYING option can be used to achieve compatibility if necessary.
To identify the interfaces between the fluid and the structure, the FLUID SOLID model definition set
is necessary.
During increment zero, Marc calculates the stiffness matrix for the structure and the mass matrix for the
structure augmented by the fluid effect. Marc then extracts the eigenvalues of the coupled system using
the MODAL SHAPE option. The modal superposition procedure can then be used to predict the time
response of the coupled system. DYNAMIC CHANGE can be used to perform modal superposition.
To input properties of solids and the mass density of the fluid elements, use the ISOTROPIC model
definition option.
The calculation of the structural mass augmentation requires triangularization of the fluid potential
matrix: this matrix is singular, unless the fluid pressure is fixed at least at one point with the
FIXED DISP option.
In a fluid/solid interaction problem, the equations of motion can be expressed as
1 T
M s a + Ku = ----- S p (6-194)
ρf
The pressure vector p can be calculated from:
– Sa = Hp (6-195)
The matrices in Equations (6-194) and (6-195) are defined as
β α
Ms =  Ni Ni ρ s dV (6-196)
V
β α
K =  β i j D i j k l β k l dV (6-197)
V
β α
S =  R ρ f n i N i dA (6-198)
T
A
β α
∂R ∂R
H =  ---------- ---------- dV
∂x i ∂x i
(6-199)
f
V
where
ρ f and ρ s are mass densities of the fluid and solid, respectively
a is the acceleration vector

u is the displacement vector
βi j is the strain displacement relation
Di j k l is the material constitutive relation
Ni is the displacement interpolation function
R is the pressure interpolation function

ni is the outward normal to the surface with fluid pressure p
T is the surface on which the fluid pressure acts
A
In the present case, the fluid is assumed to be incompressible and inviscid. Only infinitesimal
displacements are considered during the fluid vibration, so that the Eulerian and material
coordinates coincide.
Substituting Equation (6-195) into Equation (6-194), we obtain:
( M s + S T H – 1 S ⁄ ρ f )a + Ku = 0
or (6-200)
Ma + Ku = 0
This equation now allows the modes and frequencies of the solid structure immersed in the fluid to be
obtained by conventional eigenvalue methods.
Coupled Electrostatic-Structural Analysis

In a coupled electrostatic-structural analysis, both the electrostatic field and the structure are modeled. In
this analysis type, the Coulomb force (the force between charged bodies) links the electrostatic part to
the structural part, and, in turn, the deformation influences the electrostatic field. This is different from a
piezoelectric analysis where a change in the electrostatic field causes the deformation which causes a
change in the electrostatic field (movement of charge), but Coulomb forces are not taken into account.
The electrostatic-structural coupling is a weak coupling, where in the first pass the electrostatic field is
computed, then the corresponding Coulomb forces are transferred to the structural pass, and finally the
structural pass is evaluated. In a next cycle, the newly deformed state is used in the electrostatic field
calculation. Since the electrostatic solution is a steady-state solution, a time dependent problem will be
considered as quasi static for the electrostatic part. In general, in an electrostatic analysis, we can
distinguish conductors and insulators. For example, in a capacitor, the plates carrying a charge are the
conductors, and the dielectric or air between the plates is the insulator. When a potential difference is
CHAPTER 6 345
applied between the plates, the surfaces of these plates are charged. In Marc, the calculation of the charge
on the different bodies is performed using the concept of contact bodies. In the case of the capacitor, the
plates and the dielectric are separate contact bodies. The charge is then computed at the interface of two
contact bodies. Note that it is necessary that the insulating contact body is touching the conducting
contact body. In other words, the contact bodies on which a Coulomb force is acting are the ones being
touched.
Coulomb Force
Two methods are available to compute the Coulomb force in a coupled electrostatic-structural analysis.
The default method uses the following equation:
qE
F C o u l o m b = ------- (6-201)
2
where q is the charge at the node where the Coulomb force is calculated, and E is the electric field at
the insulator side of the interface. Note that in principle the Coulomb force gets half the contribution
from both sides of the interface, but it is assumed that one side is the conductor where E ≈ 0 , so this
contribution is neglected. This method works well if the charged bodies are relatively close to each other
(within 20 to 40 element edge lengths) such as the capacitor shown in Figure 6-41. When the gradient in
the electric field becomes too large, the computation of the Coulomb force becomes less accurate. This
is especially true when two charged bodies are located far away from each other. In such cases, the
contribution to the Coulomb force from other bodies is small relative to the self contribution. For this
situation, the following equation is available to compute the Coulomb force
n
1 Qi Qj
Fi =  ------------ -----------
4πε 0 r 2
- (6-202)
j = 1, i ≠ j ij
where Q i and Q j are the charge on two nodes, r i j the distance between the two charged nodes, and ε 0
the permittivity of the medium between the two nodes. This means that for each node carrying a charge
the interaction with all the other nodes carrying a charge is calculated. However, the relevant nodes
carrying a charge are at the surface of contact bodies, which reduces the number of calculations. The total
charge computed on the surface of a contact body will balance the external and reaction charge on other
nodes in the contact body. This procedure is activated by the ELECTRO,2 parameter.
Figure 6-41 Capacitor with Two Closely Spaced Charged Bodies

Equation (6-202) is valid in a 3-D analysis. For an axisymmetric analysis, the charge carrying nodes are
extended to circles. Equation (6-202) is used for a number of points on the circle. For a planar analysis,
the nodal charge is a line charge and the following equation then holds for the Coulomb force:
n
1 QiQj
Fi =  ------------ -----------
2πε 0 r i j
- (6-203)
j = 1, i ≠ j
Analysis Procedure
The procedure to perform a coupled electrostatic-structural analysis is as follows. The different bodies
(conductors and insulators) that can be distinguished must be selected as contact bodies. The way the
contact bodies are touching is important for the computation of the Coulomb Force. The contact bodies
on which a Coulomb force is acting are the ones being touched, so in general a conductor is being
touched, and air or a dielectric is the touching contact body. Both electrostatic and structural element
types are available. If an electrostatic element type is selected, then it is not active in the structural pass.
This means that this body does not deform, unless remeshing is carried out. When a structural element is
selected the corresponding electrostatic element will be automatically used in the electrostatic pass.
The ELECTRO parameter together with the STRUCTURAL parameter is used to indicate the coupled
electrostatic-structural analysis. In addition the to the CONTACT option, the ISOTROPIC or
ORTHOTROPIC model definition option is used to define the material properties for both the
electrostatic and structural pass. Specify nodal constraints using the FIXED DISP or FIXED POTENTIAL
option. Input nodal loads using the POINT LOAD option or nodal charges using the POINT CHARGE
option. Specify distributed loads with the DIST LOADS option and distributed charges with the DIST
CHARGE option. Fixed nodal displacements and potentials, or nodal forces and charges can also be
specified by the FORCDT user subroutine, nonuniform distributed loads can be specified by the
FORCEM user subroutine, and nonuniform distributed charges can be specified by the FLUX user
subroutine. For the history definition, the AUTO LOAD and AUTO STEP options can be used.
Coupled Magnetostatic-Structural Analysis

In a coupled magnetostatic-structural analysis, both the magnetostatic field and the structure are
modeled. In this analysis type, the Lorentz force links the magnetostatic part to the structural part, and in
turn, the deformation influences the magnetostatic field. Two methods to obtain the Lorentz force are
available, the Virtual Work Method (VWM), and the Maxwell Stress Tensor Method (MST). The
magnetostatic-structural coupling is a weak coupling, where in the first pass the magnetostatic field is
computed, the Lorentz force is obtained and transferred to the structural pass, and finally the structural
pass is evaluated. In a next cycle, the newly deformed state is used in the magnetostatic field calculation.
Since the magnetostatic solution is a steady-state solution, a time dependent problem will be considered
and quasi static for the magnetostatic part.
CHAPTER 6 347
Virtual Work Method

With the virtual work method the Lorentz force is derived from the variation of the magnetic energy,
caused by small displacements of a considered body. The magnetic energy is obtained from
1
W = ---    B ⋅ H dV , (6-204)
2
V
with W the magnetic energy, B the magnetic induction, H the magnetic field intensity and V the total
volume of the system. The Lorentz force is then
F L ⋅ dr = – dW . (6-205)
A group of elements needs to be selected onto which the Lorentz force is calculated. This group of
elements is then displaced in the x-, y-, and for 3-D also in the z-direction. In Marc, this group of
elements must be defined as a contact body. For correct results, this body must be surrounded by air; to
improve accuracy, Marc finds a number of layers of elements surrounding the body, and these elements
together with the body get the delta displacements to obtain the Lorentz force. The computed force is a
total force on the particular body, and is applied in the structural pass as a global load on this body.
Maxwell Stress Tensor

The general expression for the Maxwell Stress Tensor is
T = ------  BB – --- B I ,
1 1 2
(6-206)
μ0  2 
with B the magnetic induction, and I the identity tensor. The force equation at any regular point then
becomes,
fL = ∇ ⋅ T . (6-207)
The total force is the integration of this force on an arbitrary closed surface as,
FΩ =
° ( n ⋅ T )dS , (6-208)
S
with S any closed surface lying entirely in vacuum bounding the region Ω , and n the normal to the
surface. Note that, although nodal forces are calculated, only the summation of these forces is a
representation of the force on the specific body or region. In Marc, the region must be defined by a
contact body, where the surface of this body will be the surface S onto which the nodal forces are
calculated.
The accuracy of this method depends on the choice of the surface S and the accuracy of the magnetic
field B . Best results are obtained when the body onto which the force is calculated is surrounded by a
number of layers of elements having material properties of vacuum/air.
Analysis Procedure
The procedure to perform a coupled magnetostatic-structural analysis is as follows. Setup the model as
usual, structural elements are the default. For structural elements, the corresponding magnetostatic
elements are automatically selected in the magnetostatic pass. If, for certain regions, like the outer
surrounding air needed for the magnetic field, magnetostatic elements are chosen; then these elements
are not active in the structural pass. This means that this region does not deform, so elements at the
interface could get heavily distorted or even go inside out. This can be overcome by remeshing those
regions. Either the VWM or the MST method must be selected. The VWM method only works for one
body; for the MST, method multiple bodies can be selected.
The MAGNETO parameter together with the STRUCTURAL parameter is used to indicate the coupled
magnetostatic-structural analysis. In addition to the CONTACT option to define the different bodies, the
ISOTROPIC or ORTHOTROPIC model definition option is used to define the material properties for both
the magnetostatic and structural pass. Specify nodal constraints using the FIXED DISP or FIXED
POTENTIAL option. Input nodal loads using the POINT LOAD option or nodal currents using the
POINT CURRENT option. Specify distributed loads with the DIST LOADS option and distributed
currents with the DIST CURRENT option. To create a coil, the COIL CURRENT and EMWINDING can be
used to define its geometry, the location, and the magnitude of the current in this coil. Fixed nodal
displacements and potentials or nodal forces and currents can also be specified by the FORCDT user
subroutine; nonuniform distributed loads or nonuniform distributed currents can be specified by the
FORCEM user subroutine. For the history definition, the AUTO LOAD and AUTO STEP options can be
used. Use the FORCE MAGNETOSTATIC option to specify which method for the force calculation should
be used.
Coupled Thermal-Electrical Analysis (Joule Heating)

The coupled thermal-electrical analysis procedure can be used to analyze electric heating problems. The
coupling between the electrical problem and the thermal problem in a Joule heating analysis is due to the
fact that the resistance in the electric problem is dependent on temperatures, and the internal heat
generation in the thermal problem is a function of the electrical flow.
Marc analyzes coupled thermo-electrical (Joule heating) problems. Use the JOULE parameter to initiate
the coupled thermo-electrical analysis. This capability includes the analysis of the electrical problem, the
associated thermal problem, and the coupling between these two problems.
The electrical problem is a steady-state analysis and can involve current and/or voltage boundary
conditions as well as temperature-dependent resistivity.
The thermal analysis is generally a transient analysis with temperature-dependent thermal properties and
time/temperature-dependent boundary conditions.
Use the ISOTROPIC, ORTHOTROPIC, TEMPERATURE EFFECTS, ORTHO TEMP, or TABLE model
definition options to input reference values of thermal conductivity, specific heat, mass density, and
electrical resistivity, as well as their variations with temperatures. The mass density must remain constant
throughout the analysis.
CHAPTER 6 349
Use the FIXED VOLTAGE model definition option for nodal voltage boundary conditions, and the POINT
CURRENT and/or DIST CURRENT model definition options for current boundary conditions. No initial
condition is required for the electrical problem, since a steady-state solution is obtained.
In the thermal problem, you can use the INITIAL TEMP, FILMS, POINT FLUX, DIST FLUXES, and FIXED
TEMPERATURE model definition options to prescribe the initial conditions and boundary conditions.
Use the FILM and FLUX user subroutines or tables for complex convective and flux boundary conditions.
To enter the unit conversion factor between the electrical and thermal problems, use the JOULE model
definition option. Marc uses this conversion factor to compute heat flux generated from the current flow
in the structure.
Use the STEADY STATE, TRANSIENT, POINT CURRENT, DIST CURRENT, VOLTAGE CHANGE,
POINT FLUX, DIST FLUXES, and TEMP CHANGE history definition options for the incrementation and
change of boundary conditions.
A weak coupling between the electrical and thermal problems is assumed in the coupled
thermo-electrical analysis, such that the distributions of the voltages and the temperatures of the structure
can be solved separately within a time increment. A steady-state solution of the electrical problem (in
terms of nodal voltages) is calculated first within each time step.
The heat generation due to electrical flow is included in the thermal analysis as an additional heat
input The temperature distribution of the structure (obtained from the thermal analysis) is used to
evaluate the temperature-dependent resistivity, which in turn is used for the electrical analysis in the
next time increment.
For output, voltage, current density, and heat generation are available as integration point values. Note
that current density must not be confused with Ohmic current. Current density is the electric current per
unit area of cross section, while Ohmic current is the current going through the total area. So the latter is
a global quantity.
In the coupled thermo-electrical analysis, the matrix equation of the electrical problem can be
expressed as;
E
K ( T )V = I (6-209)
and the governing equation of the thermal problem is:
C T ( T )T· + K T ( T )T = Q + Q
E
(6-210)
In Equations (6-209) and (6-210):
E is the temperature-dependent electrical conductivity matrix
K
I is the nodal current vector
V is the nodal voltage vector
C T ( T ) and K T ( T ) are the temperature-dependent heat capacity and thermal conductivity

matrices, respectively
T is the nodal temperature vector
T· is the time derivative of the temperature vector
Q is the heat flux vector

E is the internal heat generation vector caused by the current flow.
Q
E E
The coupling between the electrical and thermal problems is through terms K and Q in Equations
(6-209) and (6-210).
The selection of the backward difference scheme for the discretization of the time variable in Equation
(6-210) yields the following expression:
1 1
----- [ C T ( T ) ] + K T ( T )T n = Q n + Q nE + ----- C T ( T )T n – 1 (6-211)
Δt Δt
Equation (6-211) is used for the computation of nodal temperatures in each time increment Δt .
E
The internal heat generation vector Q is computed from:
E T E
Q = B q dV (6-212)
V
E 2
q = i R (6-213)
where i and R are the current and electrical resistance, respectively.

The controls for the heat transfer option allow input of parameters that govern the convergence solution
and accuracy of heat transfer analysis.
Resistance
A resistor is a device that can dissipate electric energy. For a pure resistor, this usually happens in the
form of heat. This energy per unit time (or power) depends on the current in the resistor and its resistance.
Resistance is always associated with all conductors allowing flow of electric current. A typical model can
contain a number of conductors allowing flow of electric current, which gives rise to a resistance value
for each conductor.
All conductors in a model are assumed to be bodies and defined as a set of elements. This is done using
the THERMAL CONTACT model definition option. Two conductors can touch each other, but cannot
overlap. If two conductors touch, it implies flow of current between the two. A subset of the conductor
bodies can be considered for a resistance computation. This subset is specified using the EMRESIS
history definition option, which refers to the bodies defined on the THERMAL CONTACT model
definition option. Marc performs a resistance computation only if the EMRESIS history definition option
is specified in the input file.
A Joule heating analysis is used to compute the resistance values. All insulators in the problem are
neglected for modeling. The computation of the resistance values results from the usual Joule heating
CHAPTER 6 351
analysis with some constraints on boundary conditions: there must be at least one electric potential
boundary condition on any node. If no boundary condition is specified on a portion of the problem
boundary, it is assumed to satisfy homogeneous Neumann condition.
The resistivity of each material is specified as per the usual Joule heating analysis. If the resistivity of
any material is much higher than those of other materials, then that material is to be treated as an insulator
and should not be included in the model.
Ohms law for conductors is:
V
R = ---- (6-214)
I
V2
W E = I 2 R = ------- = VI (6-215)
R
where:
I is the current through the resistor
R is the resistance
V is the voltage across it
The conductance G is given by:
1
G = ---- (6-216)
R
Consider a single conducting body placed in an infinite homogenous insulating medium. The insulating
medium can be assumed to have negligible conductivity, in which case the current through it is
negligible. All current flows in the conductor. The current density J is then defined as the normal current
passing through a unit cross-sectional area. This means that:
I =   J ⋅ dS (6-217)
S
where S is the cross-sectional surface of the conductor.
Conduction current I is conserved; hence, I is same for any conductor cross section.
The electric field E is related to current density J in the conductor by:
J = E (6-218)
where σ is the conductivity of the conductor.
The power W E in a conductor is also given by:

1 1 1
W E = ---    J ⋅ E dv = ---    --σ- J ⋅ J dv (6-219)
2 2
V V
c c
where V c is the conductor domain.
The electric potential drop V c across the conductor is obtained from the finite element analysis. The
resistance is then calculated using Equation (6-212).
Coupled Electrical-Thermal-Mechanical Analysis

Coupled electrical-thermal-mechanical analysis (Joule-mechanical) basically combines electrical-
thermal analysis (Joule heating) with thermal-mechanical analysis. Coupled electrical-thermal-
mechanical analysis is handled using a staggered solution procedure. Using this approach, the electrical
problem is solved first for the nodal voltages. Next, the thermal problem is solved to obtain the nodal
temperatures. The mechanical problem is solved last for the nodal displacements.
Coupling between the electrical and thermal problems is mainly because of heat generation due to
electrical flow Q E (Joule heating). The thermal and mechanical problems are coupled through thermal
strain loads F T and heat generation due to inelastic deformation Q I and friction Q F . Additional
coupling may be introduced in case of temperature-dependant electrical conductivity K E and
mechanical stiffness K M . Nonlinearities may arise in the thermal problem due to convection, radiation,
and temperature-dependant thermal conductivity and specific heat. The mechanical problem may involve
geometric and material nonlinearities. Contact is another source of nonlinearity. If contact occurs
between deformable bodies or deformable and rigid bodies in the mechanical problem, boundary
conditions of the electrical and thermal problems are updated to reflect the new contact conditions.
E T M
Q F ,K
Electrical Thermal Mechanical
E I F
K Q,Q
The matrix equations governing the electrical, thermal and mechanical problems can be expressed as:
K E ( T )V = I
C T ( T )T· + K T ( T )T = Q + Q E + Q I + Q F
Mu·· + Du· + K M ( T ,u ,t )u = F + F T
where
V is the nodal voltage vector
T is the nodal temperature vector
CHAPTER 6 353
u is the nodal displacement vector
KE ( T ) is the temperature-dependent electrical conductivity matrix
I is the nodal current vector
CT( T ) is the temperature-dependent heat capacity matrix
KT ( T ) is the temperature-dependent thermal conductivity matrix
Q is the heat flux vector
QE is the heat generation due to electrical flow vector
QI is the heat generation due to inelastic deformation vector
QF is the heat generation due to friction vector
M is the mass matrix

D is the damping matrix
K M ( T ,u ,t ) is the temperature, deformation and time-dependent stiffness matrix
F is the externally applied force vector
FT is the force due to thermal strain vector
The JOULE and COUPLE parameters are needed to initiate the coupled electrical-thermal-mechanical
analysis. When defining the mesh through the CONNECTIVITY model definition option, specify the
elements as stress type. Marc internally switches to the associated heat transfer element in the electrical
and thermal passes.
The FIXED VOLTAGE model definition option can be used for nodal voltage boundary conditions, and
POINT CURRENT and/or DIST CURRENT model definition options for current boundary conditions. No
initial condition is required for the electrical problem, since a steady-state solution is obtained.
In the thermal problem, INITIAL TEMP, FILMS, POINT FLUX, DIST FLUXES, and FIXED
TEMPERATURE model definition options can be used to prescribe the initial conditions and boundary
conditions. The FILM and FLUX user subroutines can be used for complex convective and flux boundary
conditions. To enter the unit conversion factor between the electrical and thermal problems, use the
JOULE model definition option. Marc uses this conversion factor to compute heat flux generated from
the current flow in the structure.
In the mechanical problem, the usual structural model definition options such as FIXED DISP, POINT
LOAD, and DIST LOADS are used to prescribe the boundary conditions.
The history definition options (VOLTAGE CHANGE, POINT CURRENT, DIST CURRENT, TEMP
CHANGE, POINT FLUX, DIST FLUXES, DISP CHANGE, POINT LOAD, and DIST LOADS) for the
incrementation and change of boundary conditions.
JOULE and CONVERT model definition options are used to enter the conversion factor for the electrical-
thermal and the thermal-mechanical problems.
Coupled Magnetostatic-Thermal Analysis

A coupled magnetostatic thermal analysis couples a magnetostatic analysis with a thermal analysis. With
this analysis type, heat generation in magnetostatic devices can be processed and simulated. The
implementation in Marc follows a staggered approach. First, a magnetostatic analysis is performed
followed by a thermal analysis. In magnetostatic devices, the source of heat generation is the applied
currents flowing through conductors. The lamination loss computation is available with the
magnetostatic analysis and the heat generated due to these losses is also considered here. It should be
noted that lamination loss occurs due to a sinusoidally time-varying operation and the losses are
expressed as RMS (root-mean-squared) quantities. Hence, the heat generated due to lamination loss gives
a correct picture of the subsequent temperature distribution. The applied currents flowing through
conductors are assumed to be steady time-invariant and will not contain skin effect. The lamination loss
is considered here, since the magnetostatic analysis can handle a non-linear B-H curve, which is not
possible in harmonic electromagnetic analysis.
The total generated heat is then used in the thermal analysis. Temperature dependency for material data
can also be taken into account, which is especially important for the permeability of metals. The relative
permeability drops to one when the Curie temperature is reached, which can result in a significant change
in the generated heat flux.
The heat generated in magnetostatic analysis is handled in the same way as Joule heating analysis.
To initiate a coupled magnetostatic thermal analysis, add the MAGNETO and HEAT parameters. The
Magnetostatic pass is a static magnetostatic pass; the thermal part can be either a steady state or a
transient analysis. When defining the mesh, if you specify the element as a magnetostatic type through
the CONNECTIVITY option, Marc generates an associated heat transfer element. The region having an
associated heat transfer element has coupled behavior. If you specify the element as a heat transfer type
through the CONNECTIVITY option, that region is only active in the thermal pass. Use the ISOTROPIC
or ORTHOTROPIC model definition options to input values for the thermal conductivity, specific heat,
mass density, and emissivity, as well as permeability and electric conductivity. Temperature dependency
for these material properties can be added when the table driven input option is used.
For the magnetostatic pass, boundary conditions can be applied. Use FIXED POTENTIAL to prescribe the
magnetostatic potential on nodes. To apply nodal currents use POINT CURRENT. Distributed currents
can be applied using DIST CURRENT. To create a coil, the COIL CURRENT and EMWINDING can be
used to define its geometry, the location, and the magnitude of the current in this coil. The FORCDT and
FORCEM user subroutines can be used to manually control the size and/or direction of the different
vectors.
For the thermal pass, use the INITIAL TEMP, FIXED TEMPERATURE, DIST FLUXES, and FILMS model
definition options and the TEMP CHANGE and POINT FLUX history definition options to prescribe the
initial conditions and boundary conditions. The FILM and FLUX user subroutines can be used for complex
convective and flux boundary conditions.
The operational frequency for the lamination losses is given using the EMLAMIN history
definition option.
CHAPTER 6 355
The thermal pass can be STEADY STATE or time dependent using the TRANSIENT or AUTO STEP
history definition options.
If noncompatible units are used in the magnetostatic and thermal analysis, a unit conversion factor
between the current density and the generated heat can be used. It has to be ensured that the same factor
applies to the lamination losses. Use the CONVERT model definition option for this.
Coupled Electromagnetic-Thermal Analysis

A coupled electromagnetic thermal analysis couples a harmonic electromagnetic analysis with a thermal
analysis. With this analysis type, induction heating processes can be simulated. The implementation in
Marc follows a staggered approach. First, a harmonic electromagnetic analysis is performed followed by
a thermal analysis. The harmonic electromagnetic field generates induced currents in the model. Current
is also induced in conductors that carry applied currents. The net current in these conductors is the sum
of the applied and induced current. The net current in the remaining model is the induced current. These
net currents generate heat and a heat flux is computed which is then used in the thermal analysis.
Temperature dependency for material data can also be taken into account, which is especially important
for the permeability of metals. The relative permeability drops to one when the Curie temperature is
reached, which can result in a significant change in the generated heat flux. The depth of the material
induction currents generated is an important factor in the heating process. This so-called skin depth is
–1
defined as the depth at which the magnitude of the induced current density drops to e of the magnitude
at the surface
1
δ = ------------- , (6-220)
πfσμ
where f is the frequency, σ the electrical conductivity, and μ the permeability. Hence, the penetration
depth depends on the frequency. Low frequencies are used for more uniform preheating, and higher
frequencies are used for surface heating.
The heat generated in induction heating analysis is similar to a Joule heating analysis, however now in
the harmonic electromagnetic analysis the computed current density is a complex vector instead of a
scalar. The local heat generation is (see [Ref. 17])
E 1 2 1
q = --- J = --- J ⋅ J∗ , (6-221)
σ σ
with σ the electric conductivity, and J the complex current density.

To initiate a coupled electromagnetic thermal analysis, add the EL-MA and HEAT parameters. The
electromagnetic pass is a harmonic pass; the thermal part can be either a steady state or a transient
analysis. When defining the mesh, if you specify the element as an electromagnetic type through the
CONNECTIVITY option, Marc generates an associated heat transfer element. The region having an
associated heat transfer element has coupled behavior. If you specify the element as a heat transfer type
through the CONNECTIVITY option, that region is only active in the thermal pass. Use the ISOTROPIC
or ORTHOTROPIC model definition options to input values for the thermal conductivity, specific heat,
mass density, and emissivity, as well as permeability, permittivity, and electric conductivity. Temperature
dependency for these material properties can be added when the table driven input option is used.
For the electromagnetic pass, boundary conditions can be applied. Use FIXED POTENTIAL to prescribe
the magnetostatic or electrostatic potential on nodes, and POTENTIAL CHANGE in the history definition
option. To apply nodal currents or nodal charges used POINT CURRENT and POINT CHARGE
respectively. Distributed currents or distributed charges can be applied using DIST CURRENT or DIST
CHARGE, respectively. The FORCDF and FORCEM user subroutines can be used to manually control
the size and/or direction of the different vectors.
For the thermal pass INITIAL TEMP, FIXED TEMPERATURE, TEMP CHANGE, DIST FLUXES, POINT
FLUX, and FILMS can be used to prescribe the initial conditions and boundary conditions. The FILM and
FLUX user subroutines can be used for complex convective and flux boundary conditions.
The excitation frequency has to be given for the harmonic electromagnetic pass using the HARMONIC
The thermal pass can be STEADY STATE or time dependent using the TRANSIENT or AUTO STEP
history definition options.
If noncompatible units are used in the electromagnetic and thermal analysis, a unit conversion factor
between the current density and the generated heat can be used. Use the CONVERT model definition
option for this.
References
1. T. J. R. Hughes, L. P. France, and M. Becestra, “A New Finite Element Formulation for
Computational Fluid Dynamics, V. Circumventing the Babuska-Brezzi Conduction, A S Petrov-
Galerkin Formulation of the Stokes Problem Accommodating Equal-Order Interpolations”,
Comp. Meth. in Applied Mech. and Eng., p. 85-90, Vol. 59, 1986.
2. G. Hauke and T. J. R. Hughes, “A unified approach to compressible and incompressible flow”,
Comput. Methods Appl. Mech. Eng., p. 389-395, 113, (1994).
3. T. E. Tezduyar, S. Mittal, S. E. Ray, and R. Shih, “Incompressible Flow computations with
stabilized bilinear and linear equal order-interpolation velocity-pressure elements”, Comput.
Methods. Appl. Mech. Eng., p. 221-242, 95, (1992).
4. B. Ramaswaymy, T. C. Jue, “Some Recent Trends and Developments in Finite Element Analysis
for Incompressible Thermal Flows”, Int. J. Num. Meth. Eng., p. 671-707, 3, (1992).
5. A. N. Brooks and T. J. R. Hughes, “Streamline upwind/Petrov-Galerkin formulations for
convection dominated flows with particular emphasis on the incompressible Navier-Stokes
equations”, Comp. Meth. Appl. Eng., 30, (1982).
6. Codina, R. “Finite element formulation for the numerical solution of the convection-diffusion
equation” 1993.
7. Cornfield, G. C., and Johnson, R. H. “Theoretical Predictions of Plastic Flow in Hot Rolling
Including the Effect of Various Temperature Distributions.” Journal of Iron and Steel Institute
211, pp. 567-573, 1973.
CHAPTER 6 357
8. Hsu, M. B. “Modeling of Coupled Thermo-Electrical Problems by the Finite Element Method.”

Third International Symposium on Numerical Methods for Engineering, Paris, March, 1983.
9. Peeters, F. J. H. “Finite Element Analysis of Elasto-Hydrodynamic Lubrication Problems.” in
Proceedings of the XIth Int. Finite Element Kongress, edited by IKOSS GmbH. Baden- Baden,
Germany, Nov. 15-16, 1982.
10. Yu, C C. and Heinrich, J. C. “Petro-Galerkin methods for the time-dependent convective
transport equation.” Int. J. Numer. Meth. Engrg., Vol. 23 (1986), 883-901.
11. Yu, C C. and Heinrich, J. C. “Petro-Galerkin methods for multidimensional time-dependent
convective transport equation.” Int. J. Numer. Meth. Engrg., Vol. 24 (1987), 2201-2215.
12. Zienkiewicz, O. C. The Finite Element Method in Engineering Science. Third Ed. London:
McGraw-Hill, 1978.
13. Zienkiewicz, O. C., and Godbole, P. N. “A Penalty Function Approach to Problems of Plastic
Flow of Metals with Large Surface Deformations.” Journal of Strain Analysis 10, 180-183, 1975.
14. Zienkiewicz, O. C., and Godbole, P. N. “Flow of Plastic and viscoPlastic Solids with Special
Reference to Extrusion and Forming Processes.” Int. Num. Methods in Eng. 8, 1974.
15. Zienkiewicz, O. C., Loehner, R., Morgan, K., and Nakazawa, S. Finite Elements in Fluid
Mechanics – A Decade of Progress, John Wiley & Sons Limited, 1984.
16. J. Goldak, A. Chakravarti, and M. Bibby, “A New Finite Element Model for Welding Heat
Sources”, Metallurgical Transactions B., Volume 15B, June 1984, pp. 299 - 305
17. S. Clain, J. Rappaz, M. Swierkosz, and R. Touzani, “Numerical Modelling of Induction Heating
for two-Dimensional Geometries”,Math. Models Methods Appl. Sci., Vol 3 no 6, 805-822, 1993
18. C. Chaboudez, S. Clain, R. Glardon, J. Rappaz, M Swierkosz, and R. Touzani, “Numerical
Modelling of Induction Heating of Long Workpieces”, IEEE Trans. Magn.,Vol 30, 5026-5037,
1994
19. C. Chaboudez, S. Clain, R. Glardon, D.Mari, J. Rappaz, and M Swierkosz, “Numerical Modeling
of Induction Heating of Axisymmetric Geometries”, IEEE Trans. Magn.,Vol 33, 739-745, 1997
20. R.A. Rindal, “An analysis of the coupled chemically reacting boundary layer and charring
ablator, Part IV: An approach for characterizing charring ablator response with in-depth coking
reactions.”, NASA CR-1065, June 1968.
CHAPTER 7 359
Material Library
Chapter 7 Material Library
7 Material Library

Linear Elastic Material 360

Composite Material 362
 Gasket 393

Nonlinear Hypoelastic Material 398

Thermo-Mechanical Shape Memory Model 413
 Mechanical Shape Memory Model 423

Elastomer 429

Time-independent Inelastic Behavior 438
 Time-dependent Inelastic Behavior 476

Temperature Effects and Coefficient of Thermal Expansion 503

Time-Temperature-Transformation 507
 Low Tension Material 510

Soil Model 512

Damage Models 524
 Nonstructural Materials 537

References 539
This chapter describes the material models available in Marc. The models range from simple linear
elastic materials to complex time- and temperature-dependent materials. This chapter provides basic
information on the behavior of various types of engineering materials and specifies the data required by
the program for each material. For example, to characterize the behavior of an isotropic linear elastic
material at constant temperatures, you need only specify Young's modulus and Poisson's ratio. However,
much more data is required to simulate the behavior of material that has either temperature or rate effects.
References to more detailed information are cited in this chapter.
Data for the materials is entered into Marc either directly through the input file or by user subroutines.
Each section of this chapter discusses various options for organizing material data for input. Each section
also discusses the constitutive (stress-strain) relation and graphic representation of the models and
includes recommendations and cautions concerning the use of the models.
Linear Elastic Material

Marc is capable of handling problems with either isotropic linear elastic material behavior or anisotropic
linear elastic material behavior.
The linear elastic model is the model most commonly used to represent engineering materials. This
model, which has a linear relationship between stresses and strains, is represented by Hooke’s Law.
Figure 7-1 shows that stress is proportional to strain in a uniaxial tension test. The ratio of stress to strain
is the familiar definition of modulus of elasticity (Young’s modulus) of the material.
E (modulus of elasticity) = (axial stress)/(axial strain) (7-1)
Stress
Strain
Figure 7-1 Uniaxial Stress-Strain Relation of Linear Elastic Material
Experiments show that axial elongation is always accompanied by lateral contraction of the bar. The ratio
for a linear elastic material is:
v = (lateral contraction)/(axial elongation) (7-2)

This is known as Poisson’s ratio. Similarly, the shear modulus (modulus of rigidity) is defined as:
G (shear modulus) = (shear stress)/(shear strain) (7-3)

CHAPTER 7 361
Material Library
It can be shown that for an isotropic material
G = E ⁄ ( 2( 1 + v) ) (7-4)
The shear modulus G can be easily calculated if the modulus of elasticity E and Poisson’s ratio v
are known.
Most linear elastic materials are assumed to be isotropic (their elastic properties are the same in all
directions). Anisotropic material exhibits different elastic properties in different directions. The
significant directions of the material are labeled as preferred directions, and it is easiest to express the
material behavior with respect to these directions.
The stress-strain relationship for an isotropic linear elastic method is expressed as
σ i j = λδ i j ε k k + 2Gε i j (7-5)
where λ is the Lame constant and G (the shear modulus) is expressed as
λ = νE ⁄ ( ( 1 + ν ) ( 1 – 2ν ) ) and (7-6)
G = E ⁄ ( 2( 1 + ν) )
The material behavior can be completely defined by the two material constants E and v .
Use the ISOTROPIC model definition option for the input of isotropic linear elastic material constants
E (Young’s modulus) and v (Poisson’s ratio).
The effects of these parameters on the design can be determined by using the DESIGN SENSITIVITY
parameter. The optimal value of the elastic properties for linear elastic analysis can be determined using
the DESIGN OPTIMIZATION parameter.
The stress-strain relationship for an anisotropic linear elastic material can be expressed as
σi j = Ci j k l εk l (7-7)
The values of C i j k l (the stress-strain relation) and the preferred directions (if necessary) must be defined
for an anisotropic material. For example, the orthotropic stress-strain relationship for a plane stress
element is
E1 ν 21 E 1 0
1
C = ------------------------------- ν 12 E 2 E2 0 (7-8)
( 1 – ν 12 ν 21 )
0 0 ( 1 – ν 12 ν 21 )G
There are only four independent constants in Equation (7-8).

To input anisotropic stress-strain relations, use the ORTHOTROPIC or ANISOTROPIC model definition
option and the ANELAS or HOOKLW user subroutine. The ORTHOTROPIC option allows as many as 9
elastic constants to be defined. The ANISOTROPIC option allows as many as 21 elastic constants to be
defined. If the anisotropic material has a preferred direction, use the ORIENTATION model definition
option or the ORIENT user subroutine to input a transformation matrix.
A Poisson’s ratio of 0.5, which would be appropriate for an incompressible material, can be used for the
following elements: Herrmann, plane stress, shell, truss, or beam. A Poisson’s ratio which is close (but
not equal) to 0.5 can be used for constant dilation elements and reduced integration elements in situations
which do not include other severe kinematic constraints. Using a Poisson’s ratio close to 0.5 for all other
elements usually leads to behavior that is too stiff. A Poisson’s ratio of 0.5 can also be used with the
updated Lagrangian formulation in the multiplicative decomposition framework using the standard
displacement elements. In these elements, the treatment for incompressibility is transparent to you.
Composite Material
Composite materials are composed of layers of different materials (or layers of the same anisotropic
material) with various layer thicknesses and different orientations. The material in each layer may be
either linear or nonlinear. Tightly bonded layers (layered materials) are often stacked in the thickness
direction of beam, plate, shell structures, or solids.
Figure 7-2 identifies the locations of integration points through the thickness of beam and shell elements
with/without the COMPOSITE option.
Note that when the COMPOSITE option is used, as shown on the left, the layer points are positioned
midway through each layer. When the COMPOSITE option is not used, the layer points are equidistantly
spaced between the top and bottom surfaces. Marc forms a stress-strain law by performing numerical
integration through the thickness. If the COMPOSITE option is used, the trapezoidal method is employed;
otherwise, Simpson’s rule is used.
*
* *
* *
* *
* *
Beams or Shells with Beams or Shells without
Composite Option Composite Option
Figure 7-2 Integration Points through the Thickness of Beam and Shell Elements
Figure 7-3 shows the location of integration points through the thickness of composite continuum
elements. Marc forms the element stiffness matrix by performing numerical integration based on the
standard isoparametric concept.
CHAPTER 7 363
Material Library
* *
* *
* *
* *
Figure 7-3 Integration Points through the Thickness of Continuum Composite Elements
Layered Materials
To model layered materials including plates, shells, beams, and solids with Marc, use the COMPOSITE
option. In this option, three quantities are specified on a layer-by-layer basis: material identification
number, layer thickness, and ply angle. The entire set of data (a “composite group”) is then associated
with a list of elements. For each individual layer, various constitutive laws can be used. The layer
thickness can be constant or variable (in the case of variable total thickness elements), and the ply angle
can change from one layer to the next. The orientation of the 0o ply angle within each element is defined
in the ORIENTATION option. The ply thickness and the ply angle can be used as design variables in a
design sensitivity analysis. The optimal values can be determined using the design optimization
capability for linear elastic analysis.
The material identification number specified in the COMPOSITE option, is cross-referenced with the
material identification number supplied in the ISOTROPIC, ORTHOTROPIC, ANISOTROPIC,
NLELAST, TEMPERATURE EFFECTS, ORTHO TEMP, WORK HARD, and STRAIN RATE options. The
ISOTROPIC, ORTHOTROPIC, and ANISOTROPIC model definition options allow you to input material
constants such as Young’s modulus, Poisson’s ratio, shear modulus, etc. The TEMPERATURE EFFECTS
and ORTHO TEMP options allow for input of temperature dependency of these material constants.
Material constants for a typical layer are as follows:
ti thickness of the ith layer
Young’s moduli E x x, E y y, E z z
Poisson’s ratios v x y, v y z, v z x
Shear moduli G x y, G y z, G z x
ρ density
α x x, α y y coefficients of thermal expansion
σy yield stress
Mat material identifier associated with temperature-dependent properties and
workhardening data
The ANELAS, HOOKLW, ANEXP, and ANPLAS user subroutines can be used for the anisotropic behavior
of elastic constants, coefficient of thermal expansion, and yield condition.
In models where there is ply drop-off, it is possible to specify a user ply layer id to simplify
postprocessing. In this way, the top, middle, and bottom layers may all have the same id even if the
elements have a different number of layers.
There are eight given classes of strain-stress relations. The class of a particular element depends on the
number of direct (NDI) and shear (NSHEAR) components of stress. Table 7-1 lists the eight classes of
elements.
Table 7-1 Classes of Stress-Strain Relations
Class 1 NDI = 1,NSHEAR = 0
Beam Elements 5, 8, 13, 16, 23, 46, 47, 48, 52, 64, 77, 79 and Rebar Elements
{ ε } = [ 1 ⁄ Ex x }{ σ }
Axisymmetric Shells 15 and 17
 εx x  1 ⁄ Ex x –νy x ⁄ Ey y  σx x 
  =  
 εy y  –νx y ⁄ Ex x 1. ⁄ E y y  σ y y 
νy x = νx y Ey y ⁄ Ex x
Beam Elements 14, 45, 76, 78
ε  1 ⁄ Ex x 0 σ 
  =  
γ  0 1 ⁄ Gx y τ 
Plane Stress, Plates and Thin Shells 49 and 72
 εx x  1 ⁄ Ex x –νy x ⁄ Ey y 0  σx x 
   
 εy y  = –ν x y ⁄ Ex x 1 ⁄ Ey y 0  σy y 
   
 γx y  0 0 1 ⁄ Gx y  τx y 
νy x = νx y ( Ey y ⁄ Ex x )
CHAPTER 7 365
Material Library
Table 7-1 Classes of Stress-Strain Relations (continued)

Thick Axisymmetric Shells 1 and 89
 εm m  1 ⁄ Em m –νθ m ⁄ Eθ θ 0  σm m 
   
 εθ θ  = –νm θ ⁄ Em m 1 ⁄ Eθ θ 0  σθ θ 
   
 γT  0 0 1 ⁄ Gm θ  τT 
Plane Strain, Axisymmetric with No Twist, Elements 151-154.
 εx x  1 ⁄ Ex x –νy x ⁄ Ey y –νz x ⁄ Ez z 0  σx x 
   
 εy y  –νx y ⁄ Ex x 1 ⁄ Ey y –νz y ⁄ Ez z 0  σy y 
  =  
 εz z  –νx z ⁄ Ex x –νy z ⁄ Ey y 1 ⁄ Ez z 0  σz z 
   
 γx y  0 0 0 1 ⁄ Gx y  τx y 
νy x = νx y Ey y ⁄ Ex x νz y = νy z Ez z ⁄ Ey y νx z = νz x Ex x ⁄ Ez z
Thick Shell, Elements 22, 75, and 140
 εx x  1 ⁄ Ex x –νy x ⁄ Ey y 0 0 0  σx x 
   
 εy y  –νx y ⁄ Ex x 1 ⁄ Ey y 0 0 0  σy y 
   
 γx y  = 0 0 1 ⁄ Gx y 0 0  τx y 
   
 γy z  0 0 0 1 ⁄ Gyz 0  τy z 
   
 γz x  0 0 0 0 1 ⁄ Gz x  τz x 
νy x = νx y Ey y ⁄ Ex x
Three-Dimensional Brick Elements, Elements 149, 150
 εx x  1 ⁄ Ex x –νy x ⁄ Ey y –νz x ⁄ Ez z 0 0 0  σx x 
ε  σ 
 yy  –νx y ⁄ Ex x 1 ⁄ Ey y –νz y ⁄ Ez z 0 0 0  yy 
   
 εz z  –νx z ⁄ Ex x –νy z ⁄ Ey y 1. ⁄ E z z 0 0 0  σz z 
  =  
 γx y  0 0 0 1 ⁄ Gx y 0 0  τx y 
   
 γy z  0 0 0 0 1 ⁄ Gyz 0  τy z 
   
 γz x  0 0 0 0 0 1 ⁄ Gz x  τz x 
Classical Lamination Theory for Multi-Layered Shells

Basic CLT Theory
The stress tensor at a point inside a shell element is defined as
σ = G ( ε – zχ ) (7-9)
where G is the tangent matrix connecting stress and strain tensors; z is the coordinate of the
point in the thickness direction; ε is the strain tensor on the middle surface of the shell and χ is
the tensor of curvature.
The membrane forces is given by
f =  σ dz (7-10)
The bending moments is given by
m =  ( – z )σ dz (7-11)
Substituting Equations (7-9) into Equations (7-10) and (7-11) and use the definitions
hG 1 =  G dz (7-12)
h 2 G4 =  ( – z )G dz (7-13)
IG 2 =  z 2 G dz (7-14)
we obtain
 f  hG 1 h 2 G 4  ε 
  =   (7-15)
 m  h 2 G 4 IG 2  χ 
In Equations (7-12) though (7-15), h is the shell thickness and I = h 3 ⁄ 12 .

In the case where the composite material remains linear elastic, it is possible to choose alternative
computational procedures to improve computational time. This maybe activated using either the SHELL
SECT parameter or the COMPOSITE option.
The first procedure is for linear elastic composite materials where no temperature dependent
materials are present and thermal strains occur. This substantially reduces the computational time
and the memory requirements.
The second procedure is for linear elastic composite material, but either temperature dependent
materials or thermal strains are present. This substantially reduces the computational time.
CHAPTER 7 367
Material Library
PSHELL Option
PSHELL option is based on the classical lamination theory (also known as equivalent stiffness method).
This option allows you to define the membrane, bending, transverse shear, and coupling properties of the
shell elements independently.
Shell made of homogeneous materials can be modeled with the PSHELL option by simply using the same
material properties for stiffness calculation of membrane, bending, and transverse shear deformations,
and using nothing for the coupling part if there is no offset of the shell middle surface. However, for
homogeneous shell structures, it is more efficient to use the standard shell technique which is relatively
easy and inexpensive, generally more accurate, capable to deal with nonlinear material behavior.
Shell structures with layered composite materials can be solved using a full integration technique or the
PSHELL option. The full layer integration technique is very general. It can simulate material behavior
ranging from a simple linear material to a very complex material with nonlinearity and failure
mechanism. The use of the PSHELL option is limited to linear-elastic behavior. By use of this
option, composite layers are converted into one layer with equivalent stiffness behaviors. This way,
only one integration point is needed across the thickness which makes PSHELL very useful for shell
structures with layered composite materials. It is particularly attractive when the number of composite
layers is large, because analysis of these smeared shell structures uses less computer time and
storage space.
The smeared material matrices G 1 , G 2 , and G 4 are defined in Equations (7-12), (7-13), and (7-14)
representing membrane, bending and coupling stiffness, respectively.
G 1 in Marc contains both membrane and transverse shear parts. Unless you want to adjust transverse
shear stiffness, there is no need to define new types of materials.
Generally G 1 , G 2 , and G 4 can be described by the ANISOTROPIC option.
Material Preferred Direction

Every element type in Marc has a default orientation (that is, a default coordinate system) within which
element stress-strain calculations take place. This system is also assumed to be the coordinate system of
material symmetry. This is especially important for non-isotropic materials (orthotropic, anisotropic,
nlelast, or composite materials). With the ORIENTATION option, you specify the orientation of the
material axes of symmetry (relationship between the element coordinate system and the global
coordinate system, or the 0o ply angle line, if composite) in one of seven different ways:
1. as a specific angle offset from an element edge,
2. as a specific angle offset from the line created by two intersecting planes,
3. as a particular coordinate system specified by user-supplied unit vectors,
4. as specified by the ORIENT2 user subroutine,
5. by referencing a coordinate system defined by the COORD SYSTEM option.
6. by specifying rotations around the local element system,
7. by defining NURBS curves and obtaining the orientation as the tangent of the closest point on
a curve.
For the first option (EDGE I-J orientation type), the intersecting plane is defined by the surface normal
vector and a vector parallel to the vector pointing from element node I to element node J. The intersection
of this plane with the surface tangent plane defines the 0o orientation axis. (See Figure 7-4.) The
orientation angle is measured in the tangent plane positive about the surface normal.
n = Normal to Surface Tangent Plane

Node I
Vector Parallel to Edge
I-J Projected onto
Surface Tangent Plane
Integration Point
Ω α
Node J
00
Ply Angle Direction 1 of
preferred coordinate
system (fiber direction
in the ply)
Element Surface
Z
Ω = Orientation Angle (Positive

Y Right-hand Rotation About n)
X α = Ply angle (if COMPOSITE)
Figure 7-4 Edge I-J Orientation Type
For the second option (global plane orientation type), the intersecting plane is the chosen global
coordinate plane. The intersection of this plane with the surface tangent plane defines the 0o orientation
axis. (See Figure 7-5.)
The third option (user-defined plane orientation type) makes use of one or two user-defined vectors to
define the intersecting plane. Using a single vector, the intersecting plane is that plane which contains the
user vector and the chosen coordinate axis. Using two user vectors, the intersecting plane is that plane
which contains both of them. (See Figure 7-6.)
Orientation type 3-D ANISO also makes use of two user-defined vectors, but in this case, the first vector
defines the first (1) principal direction and the second vector defines the second (2) principal direction.
(See Figure 7-7.)
In the fourth option, the ORIENT2 user subroutine must be used for the definition of the orientation of
the material axis of symmetry.
In the fifth option, if continuum elements are used, the material orientation will be as defined by the
COORD SYSTEM option. If shell elements are used, the local x-axis is projected on the shell as shown
in Figure 7-8.
CHAPTER 7 369
Material Library
n - Normal to Surface Tangent Plane

Global ZX Plane
Surface α
Tangent Ω
Plane Direction 1
of preferred
coordinate
ce
In system fa
pr ters
ur
efe ec
tS
rre tion
en
dp o
lan f
em
e Ω = Orientation Angle (Positive

El
Right-hand Rotation About n)

Y
X
α = Ply angle (if COMPOSITE)
Figure 7-5 Global ZX Plane Orientation Type
n = Normal to Surface Tangent Plane

Tangent Plane
u = user-defined vector
Surface Tangent
Plane
Ω α
global
X Direction 1 of
Preferred
Intersection of Coordinate
Two Planes System
Element Surface
Z
Ω = Orientation Angle (Positive

Y Right-hand Rotation About n)
X α = Ply angle (if COMPOSITE)

Figure 7-6 User Defined XU Plane Orientation Type
U2 = User Vector 2
U3 = U1 x U2
U1 = User Vector 1
Z
U1 = Direction 1 of Preferred Coordinate System

Y
U2 = Direction 2 of Preferred Coordinate System
X
Figure 7-7 3-D ANISO Orientation Type
x
MCID Coordinate System
z
G2
y G3
ymaterial
xmaterial
G4
G1
Figure 7-8 COORD SYSTEM when Shell Elements are used
The sixth option is available for hexahedral elements. The material orientation is defined as rotations
around the three axes of the element coordinate system. See Figure 7-9. The first preferred direction joins
the centroids of faces 4-1-5-8 and 3-2-6-7. A second vector joins faces 1-2-6-5 and 4-3-7-8. The third
preferred direction is given by the cross product of the first preferred direction and this vector. The second
preferred direction is given by the cross product of the third and first preferred directions. This system is
then rotated around the three local axes by the three given angles.
CHAPTER 7 371
Material Library
8
7
5 6
4 3
2
Figure 7-9 Element Coordinate System Definition of Hexahedral Element
The seventh option method uses projection onto a curve. One or more NURBS curves are used for
defining the preferred system. A list of curves are given as input. These curves must be defined with the
CURVES model definition option and only the NURBS variant is allowed. Using the centroid of the
element, the closest point on any of the given curves is found. The first preferred direction is given by
the tangent vector at this point. For 2-D elements, the second preferred direction is given by the cross
product of the global z direction and the first preferred direction. For 3-D elements, this option is only
supported for solid shell elements and solid composite elements. The third preferred direction is given
by the thickness direction and the second preferred direction by the cross product between the third and
first preferred direction. The first preferred direction is recalculated as the cross product between the
second and third preferred directions to insure that we have an orthogonal system.
Post codes 691 and 694 can be used to obtain the first and second orientation vectors of the material
coordinate system. It should be noted that these vectors are element based quantities and only account
for the zero degree ply orientation. For specific layer orientations, these vectors are further modified by
the layer angle given through post code 697. Marc Mentat can be used to plot both the element
orientations (zero degree ply angle) and the layer orientations. These postprocessing features can be used
to verify if the initial orientations are setup correctly (especially for the ORIENT2 user subroutine or the
COORD SYSTEM option) and to evaluate the evolution of the orientation vectors as the structure
deforms. More details on these orientation vectors are given under the POST option in Marc Volume C:
Program Input.
Material Dependent Failure Criteria

Calculations of user specified failure criteria on a layer by layer basis are available in Marc. The available
criteria are: maximum stress (MX STRESS), maximum strain (MX STRAIN), TSAI-WU, HOFFMAN,
HILL, HASHIN, HASHIN FABRIC, HASHIN TAPE, PUCK, and the UFAIL user subroutine. During
each analysis, up to three failure criteria can be selected; failure indices and strength ratios are calculated
and printed for every integration point. The FAIL DATA model definition option is used for the input of
failure criteria data. When the table driven input format is used, the failure material parameters may
reference a table to introduce temperature dependent behavior.
A simple description of these failure criteria is given below:

1. Maximum Stress Criterion
At each integration point, Marc calculates six failure indices FI and six strength ratios SR: The six
failure indices are given by:
 σ 1
 ------- if σ1 > 0
 X t
1. (7-16)
 σ 1
 – ------- if σ1 < 0
 X c
 σ 2
 ------ if σ2 > 0
 Yt 
2. (7-17)
 σ 2
 – ------ if σ2 < 0
 Y c
σ
 -----3- if σ3 > 0
 Zt 
3. (7-18)
σ
 – -----3- if σ3 < 0
 Z c
 σ 12  (7-19)
4.  --------
S 12 
-
 σ 23  (7-20)
5.  -------
S 23 
-
 σ 31  (7-21)
6.  -------
S 31 
-
where
X t, X c are the maximum allowable stresses in the 1-direction in tension and compression.
Y t, Y c are maximum allowable stresses in the 2-direction in tension and compression.
Z t, Z c are maximum allowed stresses in the 3-direction in tension and compression.

CHAPTER 7 373
Material Library
S 12 maximum allowable in-plane shear stress.
S 23 maximum allowable 23 shear stress.
S 31 maximum allowable 31 shear stress.
For the Maximum Stress Failure Criterion, the strength ratios SR are the reciprocals of the
corresponding failure indices FI.
1.0
SR = ---------
FI
For example, the sixth strength ratio for the Maximum Stress Failure Criterion is ( S 31 ⁄ σ 31 ) .
Note that if the maximum allowable stresses are not defined or if the actual stresses are 0, then
the strength ratio is set to 100.
2. Maximum Strain Failure Criterion
At each integration point, Marc calculates six failure indices and six strength ratios. The six
failure indices are given by:
ε1
 ------
- if ε1 > 0
 e 1 t
1. (7-22)
ε1
 – ------- if ε1 < 0
 e 1 c
ε2
 ------
- if ε2 > 0
 e 2 t
2. (7-23)
ε2
 – ------- if ε2 < 0
 e 2 c
ε3
 ------
- if ε3 > 0
 e 3 t
3. (7-24)
ε3
 – ------- if ε3 < 0
 e 3 c
γ 12 
 -------
4.  g - (7-25)
12
γ2 3 
 ------- (7-26)
5.  g -
23
 γ 31  (7-27)
6.  -------
g 31 
-
where
e 1 t, e 1 c are the maximum allowable strains in the 1 direction in tension and compression.
g 12 is the maximum allowable shear strain in the 12 plane.
For the Maximum Strain Failure Criterion, the strength ratios at the integration points are the
reciprocals of the corresponding failure indices.
1.0
SR = ---------
FI
For example, the sixth strength ratio for the Maximum Strain Failure Criterion is ( g 31 ⁄ γ 31 ) .
Note that if the maximum allowable strains are not defined or if the actual strains are 0, then the
strength ratio is set to 100.
3. Hill Failure Criterion
Assumptions:
a. Orthotropic materials only
b. Incompressibility during plastic deformation
c. Tensile and compressive behavior are identical
At each integration point, Marc calculates the failure index FI as follows:
σ2 σ2 σ2
------1- + ------2- + -----3- –  ------- + ------- – ------ σ 1 σ 2 –  ------- + ------ – ------- σ 1 σ 3
1 1 1 1 1 1
2 2 2  2 2 2  2 2 2
X Y Z X Y Z X Z Y
(7-28)
σ 122 σ 132 σ 232
–  ------- + ------ – ------- σ 2 σ 3 + --------- + --------- + ---------
1 1 1
 2 2 2 S 122 S 132 S 232
Y Z X
CHAPTER 7 375
Material Library
For plane stress condition, it becomes

2 2 2
 σ 1 σ 1 σ 2 σ 2 σ 12 
 ------2- – ------------
- + ------- + --------- (7-29)
X X2 Y 2 S 12 2

where
X is the maximum allowable stress in the 1 direction
Y is the maximum allowable stress in the 2 direction
Z is the maximum allowable stress in the 3 direction
S 12, S 23, S 31 are as before
F failure index scale factor
For the Hill Criterion, the strength ratio at each integration point is given by:
1.0
SR = ----------
FI
Note that if the failure index is 0 (either because no allowable stresses are prescribed or because
actual stresses are 0), the strength ratio is set to 100.
4. Hoffman Failure Criterion
Note: Hoffman criterion is essentially Hill criterion modified to allow unequal maximum
allowable stresses in tension and compression.
At each integration point, Marc the failure index FI calculates:

2 2 2
[ C1 ( σ2 – σ 3 ) + C2 ( σ3 – σ1 ) + C3 ( σ 1 – σ2 ) + C4 σ1 + C5 σ2 (7-30)
+ C 6 σ 3 + C 7 σ 23
2 + C σ2 + C σ2 ]
8 13 9 12
with
C 1 = ---  ----------- + ------------- – -------------

1 1 1 1
2  Z t Z c Y t Y c X t X c
C 2 = ---  ------------- + ----------- – -------------

1 1 1 1
2 X t X c Z t Z c Y t Y c

C 3 = ---  ------------- + ------------- – -----------

1 1 1 1
2  X t X c Y t Y c Z t Z c
1 1
C 4 = ----- – ------
Xt Xc
1 1
C 5 = ----- – ------
Yt Yc (7-31)
1 1
C6 = ----- – ------
Zt Zc
1
C 7 = --------
S 232
1
C 8 = --------
S 132
1
C 9 = --------
S 122

2 2 2
 1 σ1 σ2 σ 12 σ 1 σ 2 
----- – ------ σ 1 +  ----- – ------ σ 2 + ------------
1 1 1
X X   Y Y  X X
- + ------------
Y Y 2
- – ------------- 
- + --------
Xt Xc 
(7-32)
 t c t c t c t c S 12
where: X t, X c, Y t, Y c, Z t, Z c, S 12, S 23, S 31, F are as before.
1 σ
Note: For small ratios of, for example, ------ , the Hoffman criterion can become negative due
Xt
to the presence of the linear terms.
For the Hoffman Failure Criterion, the strength ratio SR is obtained by taking the smaller of the
absolute values of the two roots obtained by solving the quadratic equation:
A ( SR ) 2 + B ( SR ) + C = 0
where for the general 3-D case:

2 2 2
C 1 ( σ 2 – σ 3 ) + C 2 ( σ 3 – σ 1 ) + C 3 ( σ 1 – σ 2 ) + C 7 σ 23 2 + C σ2 + C σ2
8 13 9 12
A = -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ,
F
CHAPTER 7 377
Material Library
C4 σ1 + C5 σ2 + C6 σ 3
B = ------------------------------------------------------- , and C = – 1.0 .
F
Note that if the allowable stresses are not defined or if the actual stresses are 0, then the strength
ratio is set to 100.
5. Tsai-Wu Failure Criterion
Tsai-Wu is a tensor polynomial failure criterion. At each integration point, Marc calculates the
failure index FI as follows:
σ 12 σ 22 σ 32
 -----
1
– ------ σ 1 +  ----- – ------ σ 2 +  ----- – ------ σ 3 + ------------- + ------------- + -----------
1 1 1 1 1
 X t X c  Y t Y c  Z t Z c Xt Xc Yt Yc Zt Zc
(7-33)
τ 122 τ 232 τ 132
+ -------- + -------- + -------- + 2F 12 σ 1 σ 2 + 2F 23 σ 2 σ 3 + 2F 13 σ 1 σ 3 ]
S 122 S 232 S 132
where X t, X c, Y t, Y c, Z t, Z c, S 12, S 23, S 31, F are as before.
F 12 Interactive strength constant for the 12 plane

2 2 2
 1 σ1 σ2 σ1 2 
– ------ σ 1 +  ------- – ------ σ 2 + ------------- + ------------- + --------
1 1 1
  ----- - + 2F 12 σ 1 σ 2  (7-34)
X X   Y Y  X X Y Y 2
 t c 1 c t c t c S 12 
Note: In order for the Tsai-Wu failure surface to be closed,

1
2 < ------------ 1 1 1 1 1
F 12 - • ------------- 2 < ------------
F 23 - • ----------- F 31
2 < ------------
- • -----------
Xt Xc Yt Yc Yt Yc Zt Zc Xt Xc Zt Zc
See Wu, R.Y. and Stachurski, 2, “Evaluation of the Normal Stress Interaction
Parameter in the Tensor Polynomial Strength Theory for Anisotropic Materials”,
Journal of Composite Materials, Vol. 18, Sept. 1984, pp. 456-463.
For the Tsai-Wu failure criterion, the strength ratio SR is obtained by taking the smaller of the
absolute values of the two roots obtained by solving the quadratic equation:
A ( SR ) 2 + B ( SR ) + C = 0
where for the general 3-D case:
σ 12 σ 22 σ 32 2
τ 12 2
τ 23 τ 132
------------
- + ------------ - + ----------- + ------- - + ------- - + -------- + 2F 12 σ 1 σ 2 + 2F 23 σ 2 σ 3 + 2F 13 σ 1 σ 3
Xt Xc Yt Yc Zt Zc S 2 S 232
S 132
12
A = ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
F
 -----
1
– ------ σ +  ----- – ------ σ +  ----- – ------ σ
1 1 1 1 1
 X t X c 1  Y t Y c 2  Z t Z c 3 , and C = – 1.0 .
B = ---------------------------------------------------------------------------------------------------------------
F
6. User-defined Failure Criteria
Using the UFAIL user subroutine, you can evaluate your own failure criterion as a function of
stresses and strains at each integration point. Two quantities can be returned by the routine:
A user-defined failure index
A user-defined strength ratio.
7. Hashin Failure Criterion
The Hashin failure criterion distinguishes between fiber failure and matrix failure. At each
integration point, Marc calculates the failure index FI for each mode as follows:
Tension fiber mode,
σ 11 2
 -------
- + ------ ( σ 12
1 2 + σ 2 ) = 1 or σ
 Xt  13 11 = X T (7-35)
S2
Compressive fiber mode, σ 1 < 0
σ 11
-----------
- = 1 (7-36)
Xc
Tensile matrix mode, σ 2 + σ 3 > 0
1 1 1
----- ( σ + σ 3 ) 2 + -------- ( σ 23 - ( σ 12
2 – σ σ ) + ------- 2 + σ2 ) (7-37)
Yt 2 S 232 2 3 S 12
13
Compressive matrix mode, σ 2 + σ 3 < 0
Yc 2
------   ----------- – 1 ( σ + σ ) + -----------
1 1 1
- ( σ 2 + σ 3 ) 2 + -------- ( σ 23
2 – σ σ ) (7-38)

Y c 2S 23
  2 3
4S 23 2 S 232 2 3
1
+ -------- ( σ 12
2 + σ2 )
13
S 122
CHAPTER 7 379
Material Library
For the Hashin failure criterion, the strength ratio SR for each mode is calculated as follows:
1.0
Tension fiber mode: SR = ---------- where FI is the failure index for the tension fiber mode
FI
1.0
Compressive fiber mode: SR = --------- where FI is the failure index for the compressive
FI
fiber mode
1.0
Tensile matrix mode: SR = ---------- where FI is the failure index for the tensile matrix mode
FI
Compressive matrix mode: A ( SR ) 2 + B ( SR ) + C = 0 where
1 1 2 1 2 2
- ( σ 2 + σ 3 ) 2 + -------- ( σ 23
A = ----------- – σ 2 σ 3 ) + -------
- ( σ 12 + σ 13 )
2 2 2
4S 23 S 23 S 12
Yc 2
B = ------   ----------- – 1 ( σ 2 + σ 3 ) , and C = – 1.0 .
1
Y c 2S 23
A sixth failure index is available for postprocessing. It is the maximum of the first five failure
criteria. Similarly, the minimum strength ratio is available as the seventh strength ratio.
8. Hashin Fabric Failure Criterion
The Hashin Fabric failure criterion is a variant of the Hashin criterion adapted for fabric
type materials. The 1 direction is the first fiber direction, the 2 direction is the second fiber
direction, and the 3 direction is through the thickness direction. At each integration point,
Marc calculates the failure indices as follows:
Tensile fiber 1 mode, σ 1 > 0
σ 2 σ1 2 2 σ 13 2
 -----1- +  --------
- +  -------
- (7-39)
 X t  S 12   S 13
Compressive fiber 1 mode, σ 1 < 0
σ 2 σ1 2 2 σ 13 2
 -----1- +  --------
- +  -------
- (7-40)
 X c  S 12   S 13
Tensile fiber 2 mode, σ 2 > 0
σ 2 σ1 2 2 σ 13 2
 -----2- +  --------
- +  -------
- (7-41)
 Y t  S 12   S 13
Compressive fiber 2 mode, σ 2 < 0
σ 2 σ1 2 2 σ 13 2
 -----2- +  --------
- +  -------
- (7-42)
 Y c  S 12   S 23
Tensile matrix mode, σ 3 > 0
σ 2 σ1 2 2 σ 13 2 σ 23 2
 -----3- +  --------
- +  -------
- +  -------
- (7-43)
 Zt   S 12   S 13  S 23
Compressive matrix mode, σ 3 < 0
σ 2 σ1 2 2 σ 13 2 σ 23 2
 -----3- +  --------
- +  -------
- +  -------
- (7-44)
 Z c  S 12   S 13  S 23
where: X t, X c, Y t, Y c, Z t, Z c, S 12, S 23, S 13 are as before.
For the Hashin Fabric Failure Criterion, the strength ratio SR for each of the six modes is
1.0
calculated as: SR = ---------
FI
where FI is the failure index for the corresponding mode. Note that if the maximum allowable
stresses are not defined or if the actual stresses are 0, then the associated strength ratios are set
to 100.
9. Hashin Tape Failure Criterion
The Hashin Tape failure criterion is a variant of the Hashin criterion adapted for tape type of
materials. The 1 direction is in the tape fiber direction, the 2 direction is perpendicular to the fiber
direction in the plane of the tape, and the 3 direction is through the thickness direction. At each
integration point, Marc calculates the failure indices as follows:
Tensile fiber mode, σ 1 > 0
σ 2 σ1 2 2 σ 13 2
 -----1- +  --------
- +  -------
- (7-45)
 X t  S 12   S 13
σ 2 σ1 2 2 σ 13 2
 -----1- +  --------
- +  -------
- (7-46)
 X c  S 12   S 13
Tensile matrix mode, σ 2 + σ 3 > 0
σ 2 + σ 3 2 σ 22 σ 33 σ1 2 2 σ 13 2 σ 23 2
 ------------------- – ----------------- +  --------
- +  -------
- +  -------
- (7-47)
 Yt  S 23 2  S 12   S 13   S 23
CHAPTER 7 381
Material Library
Compressive matrix mode, σ 2 + σ 3 < 0
Yc 2 σ2 + σ3 σ2 + σ3 2 σ2 σ3 σ1 2 2 σ 13 2 σ 23 2 σ 2
  ----------- – 1  ------------------- +  ------------------- – ------------- +  --------
- +  -------
- +  -------
- + A 1  -----1-
  2S 23   Y c   2S 23  S 232  S 12   S 13  S 23  S 1
(7-48)
where: X t, X c, Y t, Y c, S 12, S 23, S 13 are as before. A 1 is equal to zero or one and is used to
determine if the last term should be used. S 1 is the maximum fiber stress for matrix compression.
For the Hashin Tape Failure Criterion, the strength ratio SR for the first three modes is calculated
1.0
as SR = ---------- where FI is the failure index for the corresponding mode.
FI
For the compressive matrix mode, the strength ratio is calculated as follows:
A ( SR ) 2 + B ( SR ) + C = 0
where
σ2 + σ3 2 σ2 σ3 σ 12 2 σ 13 2 σ 23 2 σ 2
A =  ------------------- – ------------
- =  -------
- +  -------
- +  -------
- + A 1  -----1- ,
 2S 23  S 232  S 12  S 13  S 23  S 1
Yc 2
B = ------   ----------- – 1 ( σ 2 + σ 3 ) , and C = – 1.0 .
1
Y c   2S 23 
Note that if the maximum allowable stresses are not defined or if the actual stresses are 0, then
the associated strength ratios are set to 100.
10. Puck Failure Criterion
The Puck failure criterion distinguishes just as Hashin between fiber failure and matrix failure.
The main difference is in the matrix failure where the concept of a fracture failure angle is used.
This angle is the angle at which the fracture due to matrix failure will occur. See Figure 7-10 for
the definition on how it is defined with respect to the preferred coordinates system.
3 θfp
t n
1
Figure 7-10 Definition of the Fracture Failure Angle θ fp
The Puck failure criterion uses the following input parameters as defined above:
X t, X c, Y t, Y c, S 12 . In addition, four parameters are used for describing the slopes of the failure
(–) , p
envelope: p 12 c = p ⊥|| (+) (–) (+)
12 t = p ⊥ || , p 23 c = p ⊥⊥ , and p 23 t = p ⊥⊥ . The notation
p 12 c etc. is used in Marc Volume C: Program Input and in the Marc output while the other
notation is from Puck [Ref. 25]. The following parameters are calculated from the input
parameters and used in the following
Yc S 12  Yc 
R A = ----------------------------- = --------------  1 + 2p 12 c -------- – 1 (7-49)
2 ( 1 + p 23 c ) 2p 12 c  S 12 
σ 21 c = S 12 1 + 2p 23 c (7-50)
The first two failure indices are given by

Tensile fiber mode, σ 1 > 0
σ1
------ (7-51)
Xt
σ1
--------- (7-52)
Xc
The fracture failure angle θ fp can be calculated analytically for the plane stress case provided that
the following relation between two of the failure envelope slopes is enforced:
RA
p 23 c = p 12 c -------- (7-53)
S 12
It is recommended that either p 12 c or p 23 c is specified in the input. The other is by default

calculated using Equations (7-49) and (7-53). This default setting is also used for other cases than
plane stress although then this relationship is not necessary.
For the plane stress case, we have three failure indices, denoted modes A, B, and C:
Mode A, σ 2 > 0 , θ fp = 0
σ 12 2 Y 2 σ 2 σ2
- +  1 – p 12 t -------t-  -----2- + p 12 t -------
 ------- - (7-54)
 S 12  S 12   Yt  S 12
σ RA
Mode B, σ 2 < 0 and 0 ≤ -------2- ≤ ----------- , θ fp = 0
σ 12 σ 21 c
CHAPTER 7 383
Material Library
1
-------- ( σ 12
2 + (p
12 c σ 2 ) + p 12 c σ 2 )
2 (7-55)
S 12
σ1 2 σ 21 c
Mode C, σ 2 < 0 and 0 ≤ --------
- ≤ ----------
-.
σ2 RA
σ 12 2 σ 2 Yc
  -------------------------------------
- +  -----2-  --------- (7-56)
  2 ( 1 + p 23c S 12 )  Y c  σ 2
For this plane stress case, the fracture failure angle is calculated from the following formula
2
 σ 12 2  R A  
= -----------------------------   --------  -------- + 1
1
cos2 θ (7-57)
fp 2 ( 1 + p 23 c )   σ 2   S 12 
It is available as the sixth failure index but only for postprocessing.
For other than plane stress, the failure angle is not determined analytically. Instead, a numerical
procedure is used as outlined below.
The stresses in a failure plane defined by 1, n and t directions in Figure 7-10 are
σ n = σ 2 cos2 θ + σ 3 sin2 θ + 2σ 23 sin θ cos θ (7-58)
σ n t = ( σ 3 – σ 2 ) sin θ cos θ + σ 23 ( cos2 θ – sin2 θ ) (7-59)
σ n 1 = σ 31 sin θ + σ 12 cos θ (7-60)
The failure index as a function of θ is now formulated as follows, σ n ≥ 0 :
σn t 2 σn 1 2
f(θ ) =  -----
1
– p 1 σ n2 +  -------- +  --------- + p 1 σ n
2
(7-61)
 Yt   R A  S 12 
and for σ n < 0
σn t 2 σn 1 2
f(θ ) =  -------
- +  --------
- + ( p 2 σ n ) 2 + p 2 σ n (7-62)
 R A  S 12 
where
p 23 t σ n2t p 12 t σ n21
p 1 = --------- ------------------------- + --------- ------------------------- (7-63)
R A σ n2t + σ n21 S 12 σ n2t + σ n21
p 23 c σ n2t p 12 c σ n21
p 2 = ---------- ------------------------- + ---------- ------------------------- (7-64)
R A σ n2t + σ n21 S 12 σ 2 + σ 2
nt n1
The critical failure angle θ fp and the corresponding failure index is obtained by evaluating
equations (7-58) through (7-64) with a sufficient number of angles between -90 and 90 degrees.
For positive normal stress, it is saved as the 3rd failure index and as the 4th failure index. The 5th
failure index is not used in this case. The critical failure angle is stored as the 6th failure index for
post processing just as for plane stress.
For the Puck Failure Criterion, the strength ratios for each mode are the reciprocals of the
1.0
corresponding failure indices, SR = --------- . Note that if the maximum allowable stresses are not
FI
defined or if the actual stresses are 0, then the associated strength ratios are set to 100.
A seventh failure criterion is available for postprocessing. It is the maximum of the first five
failure criteria. Similarly, the minimum strength ratio is available as the seventh strength ratio.
Interlaminar Shear for Thick Shell, Beam, Solid Shell and 3D

Composite Brick Elements
Another addition made for composite analysis is the calculation of interlaminar shears. These
interlaminar shears are printed in the local coordinate system above and below each layer selected for
printing by PRINT CHOICE or PRINT ELEMENT. These values are also available for postprocessing.
The TSHEAR parameter must be used for activating the parabolic shear distribution calculations.
In Marc, the distribution of transverse shear strains through the thickness for thick shell and beam
elements was assumed to be constant. From basic strength of materials and the equilibrium of a beam
cross section, it is known that the actual distribution is more parabolic in nature. As an additional option,
the formulations for elements 1, 22, 45, 75, 89, 140, 149, 150, and 185 (TSHEAR will be switched off
for elements 185, 149 and 150 when the elements are stacked) have been modified to include a parabolic
distribution of transverse shear strain. The formulation is exact for beam element 45, but is approximate
for the other thick shell elements. Nevertheless, the approximation is expected to give improved results
from the previous constant shear distribution. Furthermore, interlaminar shear stresses for composite
beams and shells can now be easily calculated.
The generalized stiffness matrix for the complete section excluding transverse shear terms is given by:
F 11 ε 11
F 22 ε 22
F 12
= X Y ⋅ ε 12 (7-65)
M 11 Y Z κ 11
M 22 κ 22
M 12 κ 12
where X, Y, and Z are 3 x 3 matrices

F = section forces
M = section moments
CHAPTER 7 385
Material Library
ε = strain at mid plane of section

κ = curvature
1,2 are in-plane directions
A unique X, Y direction in the plane of the section is defined by a rotation around the element normal
which maximizes the value of X 11 in the above equation.
We then assume that the stresses in the X and Y direction are uncoupled, this gives:
       
 Fx x  Xx x Yx x  ε   Fy y  Xy y Yy y  ε 
  = ⋅  xx  and   = ⋅  yy  (7-66)
 Mx x  Yx x Zx x  κx x   My y  Yy y Zy y  κy y 
       
If we assume only bending and transverse shear in the section, all section forces are zero and inverting
the above equation gives:
   
 εx x  H1 x  ε  H1 y
  = ⋅ { M x x } and  y y  = ⋅ { My y } (7-67)
 κx x  M2 x  κy y  M2 y
   
For a point in the section, we can now define the stresses as:
σ x x ( z ) = E ( z ) ⋅ ε x x ( z ) = E ( z ) ⋅ { ε x x + κ x x ⋅ z } = E ( z ) ⋅ { H 1 x + H 2 x ⋅ z }M x x and
(7-68)
σ y y ( z ) = E ( z ) ⋅ ε y y ( z ) = E ( z ) ⋅ { ε y y + κ y y ⋅ z } = E ( z ) ⋅ { H 1 y + H 2 y ⋅ z }M y y
Since we assumed that all stresses in the X and Y direction are uncoupled, the equilibrium conditions
through the thickness are given by:
∂τ z x ∂σ x x ( z ) ∂τ z y ∂σ y y ( z )
----------- + -------------------- = 0 and ----------
- + -------------------- = 0 (7-69)
∂z ∂x ∂z ∂y
where τ z y and τ z x are the transverse shear stresses.
From beam theory, we have:
∂M x x ∂M y y
V x + -------------- = 0 and V y + -------------- = 0 (7-70)
∂x ∂y
where M is the bending moments and V is the shear forces.

Combining the Equations (7-68), (7-69), and (7-70) gives:
∂τ z x ∂τ z y
----------- = E ( z ) ⋅ { H 1 x + H 2 x ⋅ z } ⋅ V x and ----------
- = E ( z ) ⋅ { H1 y + H2 y ⋅ z } ⋅ Vy (7-71)
∂z ∂z
We can integrate this through the thickness giving:

t⁄2 y
τz x ( z ) =  E ( z ) { H 1 x + H 2 x ⋅ z }V x dz and τ z y ( z ) =  E ( z ) { H 1 y + H 2 y ⋅ z }V y dz (7-72)
–t ⁄ 2 –t ⁄ 2
with the boundary conditions that the shear stresses at the top and the bottom of the shell are zero.
We can now define the transverse shear stiffness by matching the shear strain energy over the section
obtained with the transverse shear stresses given in (7-72). This yields the flexibility matrix S:
t⁄2
1   Vx  1   Gx x ( z ) Gx y ( z )  τz x ( z ) 
---  V V [ S ]   = ---   τz x ( z ) τz y ( z )  ⋅ ⋅ dz (7-73)
2 x y   y
V 2   Gy x ( z ) Gy y ( z )  τz y ( z ) 
–t ⁄ 2
where G is the transverse shear flexibilities of the material through the thickness.
Inversion of the flexibility matrix S gives the transverse shear stiffness of the section.
Interlaminar Stresses for Continuum Composite Elements

In Marc, the interlaminar shear and normal stresses are calculated by averaging the stresses in the stacked
two layers. The stresses are transformed into a component tangent to the interface and a component
normal to the interface. The two components, considered as shear stress and normal stress, respectively,
are printed out in the output file.
By using POST code 501 or 511 (see Marc Volume C: Program Input) representing interlaminar normal
and shear stresses respectively, the interlaminar normal or shear stress can be written into a post file in
the form of a stress tensor defined in the global coordinate directions. Marc Mentat can be used to plot
the principal directions of the stress tensor which show the magnitude and the direction of the stress, and
the changes based on deformation.
Progressive Composite Failure

Marc supports progressive failure analysis for composites and other elastic materials. The material is
assumed to be linear elastic up to the point of failure. Failure is indicated by the failure criteria described
in the previous section. When failure occurs, the element stiffness is degraded. Marc offers three different
methods for the material degradation as described below. This is flagged through the FAIL DATA model
definition option. Failure occurs when any of the up to three different failure criteria is satisfied. The
material will not heal; the damaged elements keep the degraded properties after unloading.
Model 1 – Selective Gradual Degradation

This model uses a selective degradation of the moduli depending on failure mode. The moduli are
decreased gradually when failure occurs. Within an increment, it attempts to keep the highest failure
index less than or equal to one. Whenever a failure index F larger than one occurs, stiffness reduction
factors ri are calculated based upon the value of the failure indices. The incremental contribution to the
total reduction factor is calculated as
CHAPTER 7 387
Material Library
Δr i = – ( 1 – e 1 – F ) (7-74)
This is done differently for different failure criteria as described below. Six such reduction factors are
stored and updated. They are then used for scaling the respective material modulus according to
new orig
E 11 = r 1 E 11
new orig
E 22 = r 2 E 22
new orig
E 33 = r 3 E 33
new orig
G 12 = r 4 G 12
new orig
G 23 = r 5 G 23
new orig
G 31 = r 6 G 31
The Poisson’s ratios are scaled in the same way as the corresponding shear modulus.
For the maximum stress and maximum strain criteria the reduction factors are calculated separately from
each separate failure index: r1 is calculated from the first failure index as given by Equation 7-16 above,
r2 is calculated from the second failure index from Equation 7-17 etc. Thus, there is no coupling of the
different failure modes for these criteria.
For the failure criteria which only have one failure index: Tsai-Wu, Hoffman and Hill, all six reduction
factors are decreased in the same way, using the smallest of the ri:s.
For the criteria which distinguish between fiber and matrix failure (Hashin, Hashin-tape and Puck) there
is a more complex coupling between the failure modes. There is a default behavior which can be
influenced by a number of input parameters. The default is as follows.
• r1 depends on fiber failure (first and second failure index)
• r2 depends on matrix failure (third, fourth and fifth failure index)
• r3 behaves the same as r1 ( r 3 = r 1 )
• r4 behaves the same as r2 ( r 4 = r 2 )
• r5 and r6 behave the same as r4 ( r 5 = r 6 = r 4 )
In the FAIL DATA option, there are five parameters available for controlling the way the moduli are
reduced. With the exception of a1, they are only used for the Hashin variants and Puck.
a1 – Residual stiffness factor: The stiffness is never reduced to less than this factor. The default is 0.01.
a2 – Matrix compression factor. With this factor, r2 can reduce less due to failure in matrix
compression. Experiments show that certain materials show less degradation of the matrix properties in
compression than in tension. See for example [Ref. 27]. For the case that Fmc indicates failure in matrix
compression, Equation 7-74 is modified into
1–F
Δr 2 = – ( 1 – a 2 ) ( 1 – e mc) (7-75)
a2 defaults to 0; so the default is full reduction

a3 – Shear stiffness factor. This factor is used for taking into account the effect that the shear stiffness
G12 can reduce less than the matrix stiffness E2. See [Ref. 28] for experimental data on this subject. With
Fm indicating a matrix failure we have
1–F
Δr 4 = – ( 1 – a 3 ) ( 1 – e m) (7-76)
The combined effect of a2 and a3 on the shear stiffness reduction for the case of matrix compression
failure is then
1–F
Δr 4 = – ( 1 – a 2 ) ( 1 – a 3 ) ( 1 – e mc)
a4 – E33 reduction from fiber failure: This factor controls the reduction of E33 due to fiber and matrix
failure. The default is as mentioned above that E33 reduces due to fiber failure. With this factor this can
be changed to vary linearly with fiber and matrix failure. With Ff indicating a fiber failure and Fm a matrix
failure we have
1–F 1–F
Δr 3 = – ( 1 – a 4 ) ( 1 – e f) – a4 ( 1 – e m) (7-77)
a5 – Shear reduction from fiber failure: With this factor it is possible to control the reduction of the
shear stiffness due to fiber failure. By default it only reduces due to matrix failure. With Ff and Fm as in
the previous we have
1 – Fm 1 – Ff
Δr 4 = – ( 1 – a 5 ) ( 1 – e ) – a5 ( 1 – e ) (7-78)
The Hashin-fabric failure criterion reduces the first three reduction factors from the respective failure
index. The three shear reduction factors are taken from the worst of the first three factors. The factors a2
through a5 are not used for this criterion.
In addition, it is also possible to use the UPROGFAIL user subroutine to explicitly define the reduction
factors r1 through r6.
Model 2 – Selective Immediate Degradation

This model uses selective degradation just as Model 1, but the stiffness is abruptly decreased. As soon as
failure is indicated, the stiffnesses are set to a1 – the residual stiffness factor. The same rules as in Model
1 for how the different factors are defined depending on the type of failure is applied here.
CHAPTER 7 389
Material Library
Model 3 – Original Marc Method

This is the original method in Marc.
1. Upon failure, the material moduli for orthotropic materials at the integration points are set to the
smallest of the original moduli, and the smallest is set to 10% of the original.
2. Upon failure, for isotropic materials, the failed moduli are taken as 10% of the original moduli.
3. If there is only one modulus, such as in a beam or truss problem, the failed modulus is taken as
10% of the original one.
The different options are flagged through the FAIL DATA model definition option.
Mixture Model
The mixture models allow the user to define a composite material that consists of multiple components
for all element types. Effective material properties are formed based upon the volume fraction of each
material component. There are three variations of the mixture model which have consequences on the
type of material that may be used in each component. The component material is defined using the
conventional ISOTROPIC, ORTHOTROPIC, ANISOTROPIC, MOONEY, etc. options. The mixture model
acts as a continuum in the sense that debonding between the components is not considered in any of
the models.
Mixture Model 1
This model is applicable to linear elastic materials and heat transfer. An effective material property is
formed as
N
eff i
E =  Vf ⋅ E
i
i = 1
i
where E is the value of the ith component, V f is the volume fraction of the ith component, and N is
i
the number of components. The same summation is done for all properties. If an orthotropic component
is used, it is done for all nine material properties.
Note that for orthotropic or anisotropic properties, is assumed that all components have the same
preferred directions.
If the component has temperature dependent properties, then the temperature dependence is evaluated
for each component which can be expressed as
N
eff · i
E (T) =  Vfi ⋅ E ( T )
:
The stresses obtained are an effective stress associated with the mixture material and not the stress in
each component.
Mixture Model 2
This model is applicable to linear elastic materials and heat transfer. An effective material stress-strain
relationship is formed as
N
e ff i
D =  Vf D
i
i = 1
where
i i
σ = Dε
eff eff
σ = D ε
For heat transfer, analogous equations are used:

i i
q = D ∇T
eff eff
q = D ∇T
Again, temperature dependent properties are dealt with on a component basis, and orthotropic or
anisotropic mixtures must have the same preferred orientations.
Mixture Model 3
This model is applicable to nonlinear material behavior. This requires additional allocation of memory to
store the associated state variables (plastic strain, shift tensors, etc.) for each component. It is similar to
model 2 in that an effective material stress-strain relationship is formed as
N
e ff i
D =  Vf D
i
i = 1
i
but now, D includes material nonlinear effects. for the case of elastic-plastic materials, we now have
two types of quantities which are not completely consistent with one another. On a component level:
i i i i
Δσ = D e l ( Δε – Δε t h – Δε p l )
where again the i indicates the component number which is the correct expression. We also calculate
effective quantities. These quantities are accumulated in the usual manner.
i i i
σ n + 1 = σ n + Δσ
pl(i) pl(i) pl(i)

εn + 1 = εn + Δε
CHAPTER 7 391
Material Library
There are also effective quantities calculated as:

N
eff
σ =  Vf σi
i
i = 1
N
pl(eff) pl(i)
Δε =  V f Δε
i
i = 1
N
pl(eff) pl(i)
Δε =  V f Δε
i
i = 1
which are also accumulated.

pl(eff) pl(eff) p l ( e ff )
εn + 1 = εn + Δε
pl(eff) pl(eff) p l ( e ff )
εn + 1 = εn + Δε
pl
where the ε is the equivalent plastic strain. Note that the effective plastic strain is not used in a
subsequent calculation and is only provided for output.
Given a mixture of two identical nonlinear materials at a 50%-50% mix, the output of the effective
equivalent plastic strain will not be the same as for a solitary material.
Mixture model 3 can be used with most nonlinear material behavior, but there are a few restrictions in
this release which include:
All damage models introduced via the DAMAGE option
Gasket material
Soils
User-defined generalized stress-strain law.
ORNL
Rigid Plastic
Viscoelasticity
Cohesive
Grain size effects
Thermo-Pore
Mass Density
The effective mass density for all of the mixture models is calculated as
N
eff i
ρ =  Vf ρ
i
i = 1
Specific Heat
The effective specific heat for all of the mixture models is calculated as
N
eff i
Cp =  Vf Cp
i
i = 1
Coefficient of Thermal Expansion

There are two ways to calculate the coefficient of thermal expansion; the first directly uses the rule
of mixtures.
N
eff i
α =  α
i = 1
The second formulation:

i i
eff E ⋅ Vf ⋅ α
i
α = ------------------------------------
i
 E ⋅ Vf i
where, again, if temperature dependent material properties are used, these are applied at the
component level.
i i
eff  E ( T ) ⋅ V fi ⋅ α ( T )
α ( τ ) = ------------------------------------------------------
i
 E ( T ) ⋅ Vf i
Mixtures and Composites

It is possible to have a shell element or a composite brick element where the layer of a composite material
may be a mixture model that consists of multiple components.
Restrictions
If the VOID CHANGE or POROSITY CHANGE is used to define a state variable, and the material
properties are a function of the void ratio, then all components will use the same void ratio.
A mixture material cannot have as a component another mixture material.
Mixture materials should not be used with the PSHELL option.
Mixture materials cannot be used with design sensitivity or design optimization where the design variable
is an elastic property in the mixture material or one of the components.
CHAPTER 7 393
Material Library
Gasket
Engine gaskets are used to seal the metal parts of the engine to prevent steam or gas from escaping. They
are complex (often multi-layer) components, usually rather thin and typically made of several different
materials of varying thickness. The gaskets are carefully designed to have a specific behavior in the
thickness direction. This is to ensure that the joints remain sealed when the metal parts are loaded by
thermal or mechanical loads. The through-thickness behavior, usually expressed as a relation between
the pressure on the gasket and the closure distance of the gasket, is highly nonlinear, often involves large
plastic deformations, and is difficult to capture with a standard material model. The alternative of
modeling the gasket in detail by taking every individual material into account in the finite element model
of the engine is not feasible. It requires a lot of elements which makes the model unacceptably large.
Also, determining the material properties of the individual materials might be cumbersome.
The gasket material model addresses these problems by allowing gaskets to be modeled with only one
element through the thickness, while the experimentally or analytically determined complex pressure-
closure relationship in that direction can be used directly as input for the material model. The material
must be used together with the first-order solid composite element types 149 (three-dimensional solid
element), 151 (two-dimensional plane strain element) or 152 (two-dimensional axisymmetric element).
In that case, these elements consists of one layer and have only one integration point in the thickness
direction of the element.
Gaskets can be used in mechanical, thermal or thermo-mechanically coupled analyses. The usual
staggered scheme of a heat transfer pass followed by a structural pass is used for coupled analyses. For
the heat transfer part, the elements used to model the gaskets are type 175 (three-dimensional first-order
solid element), type 177 (two-dimensional first-order planar element), or type 178 (two-dimensional
first-order axisymmetric element).
Constitutive Model
The behavior in the thickness direction, the transverse shear behavior, and the membrane behavior are
fully uncoupled in the gasket material model. In subsequent sections, these three deformation modes are
discussed.
Local Coordinate System

The material model is most conveniently described in terms of a local coordinate system in the
integration points of the element (see Figure 7-11). For three-dimensional elements, the first and second
directions of the coordinate system are tangential to the midsurface of the element at the integration
point. The third direction is the thickness direction of the gasket and is perpendicular to the midsurface.
For two-dimensional elements, the first direction of the coordinate system is the direction of the
midsurface at the integration point, the second direction is the thickness direction of the gasket and is
perpendicular to the midsurface, and the third direction coincides with the global 3 direction.
2 3
1
1
Midsurface
Integration Point Midsurface
Integration Point
Figure 7-11 The Location of the Integration Points and the Local Coordinate Systems in Two- and
Three-dimensional Gasket Elements
In a total Lagrange formulation, the orientation of the local coordinate system is determined in the
undeformed configuration and is fixed. In an updated Lagrange formulation, the orientation is
determined in the current configuration and is updated during the analysis.
Thickness Direction - Compression

In the thickness direction, the material exhibits the typical gasket behavior in compression, as depicted
in Figure 7-12. After an initial nonlinear elastic response (section AB), the gasket starts to yield if the
pressure p on the gasket exceeds the initial yield pressure py0. Upon further loading, plastic deformation
increases, accompanied by (possibly nonlinear) hardening, until the gasket is fully compressed (section
BD). Unloading occurs in this stage along nonlinear elastic paths (section FG, for example). When the
gasket is fully compressed, loading and unloading occurs along a new nonlinear elastic path (section
CDE), while retaining the permanent deformation built up during compression. No additional plastic
deformation is developed once the gasket is fully compressed.
The loading and unloading paths of the gasket are usually established experimentally by compressing
the gasket, unloading it again, and repeating this cycle a number of times for increasing pressures. The
resulting pressure-closure data can be used as input for the material model. If the pressure also varies with
temperature, then pressure-closure data at different temperatures can be provided. The user must supply
the loading path and may specify up to ten unloading paths. In addition, the initial yield pressure py0 must
be given. The initial yield pressure can also be varied with temperature and spatial coordinates. The
loading path should consist of both the elastic part of the loading path and the hardening part, if present.
If no unloading paths are supplied or if the yield pressure is not reached by the loading path, the gasket is
assumed to be elastic. In that case, loading and unloading occurs along the loading path.
CHAPTER 7 395
Material Library
loading path
py1 D
py G
py0 B
Gasket Pressure p
unloading path
A F C
cp0 cp cy0 cp1 cy cy1
Gasket Closure Distance c
Figure 7-12 Pressure-closure Relation of a Gasket
The loading and unloading paths must be defined using the TABLE model definition option (see Marc
Volume C: Program Input) and must relate the pressure on the gasket to the gasket closure. Optionally,
at each closure value, the loading and unloading paths can also be defined as functions of temperature
and spatial coordinates using multi-variate TABLEs. The unloading paths specify the elastic unloading
of the gasket at different amounts of plastic deformation; the closure at zero pressure is taken as the
plastic closure on the unloading path. If unloading occurs at an amount of plastic deformation for which
no path has been specified, the unloading path is constructed automatically by linear interpolation
between the two nearest user supplied paths. The unloading path, supplied by the user, with the largest
amount of plastic deformation is taken as the elastic path at full compression of the gasket.
For example, in Figure 7-12, the loading path is given by the sections AB (elastic part) and BD
(hardening part) and the initial yield pressure is the pressure at point B. The (single) unloading path is
curve CDE. The latter is also the elastic path at full compression of the gasket. The amount of plastic
closure on the unloading path is cp1. The dashed curve FG is the unloading path at a certain plastic
closure cp that is constructed by interpolation from the elastic part of the loading path (section AB) and
the unloading path CD.
The compressive behavior in the thickness direction is implemented by decomposing the incremental
gasket closure into an elastic and a plastic part:
Δc = Δc e + Δc p (7-79)
Of these two parts, only the elastic part contributes to the pressure. The constitutive equation is given by
the following equation:
p
Δp = D c Δc e = D c ( Δc – Δc ) .
Here, D c is the consistent tangent to the pressure-closure curve.

7 Plastic defomation develops when the pressure p equals the current yield pressure p y . The latter is a
function of the amount of plastic deformation developed so far and is given by the hardening part of the
loading path (section BD in Figure 7-12).
Material Initial Gap

Library The thickness of a gasket can vary considerably throughout the sealing region. Since the gasket is
modeled with only one element through the thickness, this can lead to meshing difficulties at the
boundaries between thick regions and thin regions. The initial gap parameter can be used to solve this.
The parameter basically shifts the loading and unloading curves in the positive closure direction. As long
as the closure distance of the gasket elements is smaller than the initial gap, no pressure is built up in the
gasket. The sealing region can thus be modeled as a flat sheet of uniform thickness and the initial gap
parameter can be set for those regions where the gasket is actually thinner than the elements of the finite
element mesh used to model it. The initial gap can optionally be varied as a function of spatial coordinates
by using a table.
Thickness Direction - Tension

The tensile behavior of the gasket in the thickness direction is linear elastic and is governed by a tensile
modulus D t . The latter is defined as a pressure per unit closure distance (that is, length). D t can
optionally be varied as a function of temperature and spatial coordinates by using a multi-variate table.
Transverse Shear and Membrane Behavior

The transverse shear is defined in the 2-3 and 3-1 planes of the local coordinate system (for three-
dimensional elements) or the 1-2 plane (for two-dimensional elements). It is linear elastic and
characterized by a transverse shear modulus G t . G t can optionally be varied as a function of temperature
and spatial coordinates by using a multi-variate table.
The membrane behavior is defined in the local 1-2 plane (for three-dimensional elements) or the local 3-
1 plane (for two-dimensional elements) and is linear elastic and isotropic. Young’s modulus E m and
Poisson’s ratio ν m that govern the membrane behavior are taken from an existing material that must be
defined using the ISOTROPIC model definition option. Multiple gasket material can refer to the same
isotropic material for their membrane properties (see also the GASKET model definition option in Marc
Volume C: Program Input).
Thermal Expansion
The thermal expansion of the gasket material is isotropic and the thermal expansion coefficient is taken
from the isotropic material that also describes the membrane behavior.
Constitutive Equations
As mentioned above, the behavior in the thickness direction of the gasket is formulated as a relation
between the pressure p on the gasket and the gasket closure distance c. In order to formulate the
constitutive equations of the gasket material, this relation (Equation (7-80)) must first be written in terms
of stresses and strains. This depends heavily on the stress and strain tensor employed in the analysis. For
CHAPTER 7 397
Material Library
small strain analyses, for example, the engineering stress and strain are used. In that case, the incremental
gasket closure and the incremental pressure are related to the incremental strain and the incremental
stress by
Δc = – hΔε and Δp = – Δσ (7-80)

in which h is the thickness of the gasket.
The resulting constitutive equation for three-dimensional elements, expressed in the local coordinate
system of the integration points, now reads
Em νm Em
----------------
- ----------------
- 0 0 0 0
2 2
1 – νm 1 – νm Δε 11
Δσ 11
νm Em Em Δε 22
Δσ 22 ----------------
- ----------------
- 0 0 0 0
2 2
Δσ 33 1 – νm 1 – νm p
Δε 33 – Δε 33
= (7-81)
Δσ 12 0 0 C 0 0 0 Δγ 12
Em
Δσ 23 0 0 0 -------------------------- 0 0 Δγ 23
2 ( 1 + νm )
Δσ 31 Δγ 31
0 0 0 0 Gt 0
0 0 0 0 0 Gt
in which C = hD c . For two-dimensional elements, the equation is given by
Em νm Em
----------------
- 0 ----------------
- 0
Δσ 11 1 – νm
2
1 – νm
2 Δε 11
Δσ 22 0 C 0 0 Δε 22 – Δε 22
p
= (7-82)
Δσ 33 νm Em Em Δε 33
----------------- 0 ----------------- 0
Δσ 12 1 – νm
2
1 – νm
2
Δγ 12
0 0 0 Gt
For large deformations in a total Lagrange formulation, in which the Green-Lagrange strains and the
second Piola-Kirchhoff stresses are employed, as well as in an updated Lagrange environment, in
which the logarithmic strains and Cauchy stresses are being used, similar but more complex relations
can be derived.
Nonlinear Hypoelastic Material

The hypoelastic model is able to represent a nonlinear elastic (reversible) material behavior. For this
constitutive theory, Marc assumes that
σ· i j = L i j k l ε· k l + g i j (7-83)
where L is a function of the mechanical strain and g is a function of the temperature.

The stress and strains are true stresses and logarithmic strains, respectively, when used in conjunction
with the LARGE STRAIN.
When used in conjunction with the LARGE DISP option only, Equation (7-83) is expressed as
· ·
Si j = Li j k l Ek l + gi j (7-84)
where E, S are the Green-Lagrangian strain and second Piola-Kirchhoff stress, respectively.
A HYPOELASTIC or NLELAST model definition option is necessary to invoke this model. These models
can be used with any stress element, including Herrmann formulation elements.
The tensors L and g are defined by you in the HYPELA2 user subroutine or via the NLELAST model
definition option.
HYPELA2 User Subroutine

In order to provide an accurate solution, L should be a tangent stiffness evaluated at the beginning of the
iteration. In addition, the total stress should be defined as its exact value at the end of the increment. This
allows the residual load correction to work effectively.
In HYPELA2, additional information is available regarding the kinematics of deformation. In particular,
the deformation gradient ( F ), rotation tensor ( R ), and the eigenvalues ( λ ) and eigenvectors ( N ) to form
the stretch tensor ( U ) are also provided. This information is available only for the continuum elements
namely: plane strain, generalized plane strain, plane stress, axisymmetric, axisymmetric with twist, and
three-dimensional cases.
For more information on the use of user subroutines, see Marc Volume D: User Subroutines and
Special Routines.
NLELAST Model Definition

The NLELAST model definition in Marc supports the Nastran NLELAST capability and simplified
nonlinear elasticity models. These models all have the form that a strain energy function does not exist
but have been found to be useful in solving engineering simulations, where only a limited amount of test
data is available. The theory and algorithms are adequate to trace the stress-strain curve accurately for
the uniaxial loading cases. However, the theory is not based on the classical theory of finite elasticity.
Consequently, some of the constitutive relations may be violated in the multiaxial stress state and, as a
result, limit the use of this model to small strain.
CHAPTER 7 399
Material Library
Six models are implemented in Marc to support nonlinear elasticity based upon the input of a uniaxial
stress strain law. These models are summarized as:
1. Nastran like NLELAST model based upon MATS1 input option.
2. Hypoelastic strain invariant model.
3. Model based upon working in principal strain space which results in induced
anisotropic behavior.
4. Linear elasticity with tension and/or compression cut-off.
5. Bi-modulus elasticity with tension and/or compression cut-off.
6. Nonlinear orthotropic elasticity based upon non-rotating preferred directions.
The HYPELA2 user subroutine can be used if special material behavior is required.
Basic assumptions and definitions

A material model is called elastic if the stress is a function of the strain. This implies two important
properties: the material is reversible and the stress at a point in the material depends only on some
measure of strain at that point. Elastic materials implemented with this option are in the hypoelastic
framework.
This group of nonlinear elastic materials is therefore grouped in Mentat under Materials ->
HYPOELASTIC.
The nonlinearity in the material models introduced here results from the definition of the stress as a
function of the strain path. The models are anticipated to behave correctly in the small strain region only
(less than 5%).
Input of uniaxial stress-strain data

In the next sections the theory and input is described for each of the different material models. As a basis
it is assumed that the user has a set of engineering stresses and engineering strains available. This means
that in the uniaxial test:
• The measured elongations are compared to the original length of the test specimen, resulting in:
ε engineering = ΔL ⁄ L 0
• The measured forces are compared to the original cross-section of the test specimen, resulting in:
σ engineering = F ⁄ A 0
For analysis without large incremental or total rotation or large incremental or total strains, the
engineering values can be used as is. In some models, certain characteristics still have to be derived. This
is described in the sections per material type.
For analysis where the LARGE DISP, UPDATE, and FINITE, or PLASTICITY,3, or LARGE STRAIN
parameters are used, the uniaxial test data cannot be used ‘as is’. Engineering strain and stress are
measures that are invalid once the strain in the analysis model is no longer ‘small’ (approximately 5%).
For these large strain simulations, the true stress and strain definition have been adopted in Marc. This
nonlinear strain measure is dependent upon the final length of the model. Since the NLELAST material
models are only valid in this small range it is not necessarily required to convert the engineering data.
Converting engineering strain/stress to true strain/stress

True strain is defined as:
ε true =  ΔL ⁄ L
By integration, it can be shown that this directly defines the true strain as:
ε true = ln ( 1 + ε engineering ) (7-85)
Similarly, the uniaxial engineering stress data can be converted to true stress (also known as Cauchy
stress) for incompressible materials by means of:
σ true = σ engineering ( 1 + ε engineering ) (7-86)
and, in general, by σ t ru e = σ engineering ⁄ ( 1 – v ⋅ ε engineering ) 2
By using Equations (7-85) and (7-86), the uniaxial test data can quickly be transformed into true
stress/strain data. Figure 7-13 shows the deviation of engineering stress/strain compared to the
engineering definition. In the remainder of this section on nonlinear elasticity the engineering stresses
and strains can be mutually exchanged with their true counterpart. The tables can be either engineering
or true data depending on the options activated in the analysis.
Comparison between engineering and true stress/strain
200
150
100
true or engineering stress
True Stress
Engineering Stress
50
0
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
-50
-100
-150
-200
true or engineering strain
Figure 7-13 Comparison Between Engineering and True Stress/Strain

CHAPTER 7 401
Material Library
Type 1: Nastran like nonlinear elasticity model

The stress curve is supplied in a table with independent variable mechanical strain (id=69). An additional
independent variable in the table may be temperature (id=12). Note that a table with independent variable
mechanical strain is a requirement for the analysis.
It is optional to define Poisson’s ratio as a function of temperature. The initial Young’s modulus is the
slope of the equivalent stress-strain curve is derived by Marc from the table at ε = 0. This imposes the
continuous definition of the slope at ε = 0.
Theoretical Basis
The nonlinear elastic capability in Nastran was designed to satisfy the equivalence of the deformation
work per unit volume in the simple tension to the strain energy per unit volume (conservation of energy),
while the work done for deformation may be defined by a stress-strain curve in simple tension, i.e.,
 σ dε =  <σ> { dε }
It is further assumed that the effective strain ε may be defined by:
1 2 1
--- Eε = --- <ε> [ D e ] { dε }
2 2
From the total differential of the above equation, we obtain

1
dε = ------ <ε> [ D e ] { ε }
Eε
Substituting the latter in the first equation above, where stresses may be expressed in terms of total
strains, i.e.
σ
{ σ } = ------ [ D e ] { ε }
Eε
The tangential matrix for such material may be obtained by differentiating the latter equation, i.e.
∂{ σ} σ α ∂σ σ
[ D n e ] = ------------- = ------ [ D e ] + ------  ------- – --- [ D e ] { ε }<ε> [ D e ] T
∂{ ε} Eε Eε  ∂ε ε 
In the original Nastran, implementation α is set to 1. In the Marc implementation, testing proved that
setting this value to 0.0 (default) improves convergence significantly.
Solution Algorithm
The user specifies the nonlinear stress-strain curve, σ ( ε ) in a TABLE parameter along with a NLELAST
model definition. The NLELAST entry does require a table for Young’s modulus; whereas, strain
dependency for Poisson’s ratio is optional.
The NLELAST material interfaces with the element routines with the following data:
• Input: { σ } old , { ε } o l d , { Δε } , E , ν
• Output: { σ } new , { ε } new , [ D n e ]
The computational procedure is described below:
Step 1: Upon entry the new strain state is computed by:

{ ε } new = { ε } old + { Δε }
Step 2: The effective strain is computed based on { ε } new by

1
ε 2 = --- <ε> { σ e }
E
where
{ σe } = [ De ] { ε }
D e is simply formed using the Hookean law.
Step 3: The effective stress ( σ ) is determined by looking-up the user-specified stress-

strain curve for . ε .
Step 4: The new stress state is determined by scaling the stress-strain law:
σ
{ σ } new = ------ { σ e }
Eε
Step 5: The tangential matrix is determined by

σ α ∂σ σ
[ D n e ] = ------ [ D e ] + ---------  ------- – --- { σ e } { σ e } T
Eε Eε 2  ∂ε ε 
∂σ
Furthermore ------- is the slope at the current strain.
∂ε
The tables for NLELAST must be supplied in the new table format.
Combining Uniaxial Tension and Compression Stress-strain Curves

Some materials exhibit appreciably different behavior in compression from that in tension even in the
small strain range. For uniaxial loading the magnitude of the strain in that direction becomes the effective
strain, i.e.,
ε = ε x for uniaxial tension in x
ε = – ε x for uniaxial compression in x

CHAPTER 7 403
Material Library
In order to remain compatible with the Nastran implementation of this material model, the user has the
option to ignore the uniaxial compression data, even if it is supplied on input.
In case of this unsymmetric material behavior, we need to be able to distinguish between a state of
compression and a state of tension. There are two known data points for one effective strain ε , namely
the effective stress for uniaxial tension ( σ t ) and the effective stress for uniaxial compression ( σ c ) .
Some method of interpolation or extrapolation is required to predict the effective stress for the general
stress state using two known data points.
The first stress invariant ( I 1 ) has been adopted for interpolation/extrapolation as follows,
I1 = σx + σy + σz
Considering that the pure shear is in the midway between simple tension and simple compression, it
seems appropriate to use the first stress invariant. Hydrostatic tension and compression cases will impose
lower- and upper-bounds for extrapolation, i.e.,
I 1 = σ x for uniaxial tension/compression
I 1 = 0 for pure shear
I1 = 3 p for hydrostatic pressure
∂σ
The instantaneous modulus  ------- should be interpolated or extrapolated in the same manner.
 ∂ε 
Solution Algorithm for Bilateral Stress-strain Relations

The new stress state is proportional in magnitude to the effective stress σ , which should be determined
as follows:
1. Compute the effective stress ( σ e ) based on { σ e } ; i.e.,
{ σe } = [ De ] { ε }
1
σ = --- [ ( σ x – σ y ) 2 + ( σ y – σ z ) 2 + ( σ z – σ x ) 2 ] + 3 ( τ x2y + τ y2z + τ z2x ) for 3-D
2
2 2 2
σ = σ x – σ x σ y + σ y + 3τ x y for plane stress
1
σ = --- [ ( σ x – σ y ) 2 + ( σ y – σ z ) 2 + ( σ z – σ x ) 2 ] + 3τ x2y for plane strain
2
2. Compute the first invariant of I = σ x + σ y + σ z :

where σ z = 0 for plane stress
3. Determine the ratio ( r ) by normalizing I I by σ c , i.e.,

I1
r = ------
σe
where r signifies the relative distance from the midpoint of σ c and σ t at ε . It would be
implausible to process a large value of r (such is the case with a hydrostatic load). Therefore, r ,
will be confined to a plausible range, – 1 ≤ r ≤ 1 . The value will be reset to the limit ( r = ± 1 ) if
r lies outside the range.
4. Look up the user specified stress-strain curve in the TABLE entry and determine σ t and σ c ; i.e.,
σt = σ ( ε )
σc = σ ( ε )
5. Determine based on σ t , σ c . and r ; i.e.,
σt + σc σt – σc
σ = ------------------ + r ------------------
2 2
∂σ
For the tangent matrix the instantaneous modulus  ------- should be determined using the same ratio (r)
 ∂ε 
as follows:
1. Compute the instantaneous slope at ε for tension, i.e.,
yi + 1 – yi
 ∂σ
------- = -----------------------
- for x i ≤ ε ≤ x i + 1
 ∂ε  t xi + 1 – xi
where ( x i, y i ) is the i-th data point in the TABLE entry
2. Compute the instantaneous slope at – ε for compression; i.e.,

yj + 1 – yj
 ∂σ
------- = -----------------------
- for x i ≤ ε ≤ x i + 1
 ∂ε  c xj + 1 – xj
∂σ ∂σ  ∂σ
3. Determine ------- based on  ------- , ------- and r , i.e.,
∂ε  ∂ε  t  ∂ε  c
∂σ 1 ∂σ ∂σ  ∂σ ∂σ
------- = ---  ------- +  ------- + --- ------- –  -------
r
∂ε 2  ∂ε  t  ∂ε  c 2  ∂ε  t  ∂ε  c
Subsequently the values can now be used as in the tension-only algorithm.

CHAPTER 7 405
Material Library
Type 2: Strain Invariant Model

This model, like others in Marc is based upon the incremental theory where
dσ = C e l dε e l
And C e l is a Hooke’s law based upon the current values of the Young’s modulus and Poisson’s ratio. In
this model these quantities are a function of the strain invariants I 1 , I 2 , I 3 .
For this model, Marc computes the three strain invariants in each integration point of the element. The
three strain invariants are independent variables in the Young’s modulus vs strain-invariant table that is
supplied on input. In principle a table with all three strain invariants can be supplied if necessary,
however, when converting the uniaxial data into Young vs strain-invariant tables it is usually sufficient
to choose one. The choice may depend on the element type used and the need to model non-symmetric
behavior in compression and tension. This is explained further below.
Marc looks up Young’s modulus in the user-supplied table and forms a Hookean stress-strain law.
In order to arrive at proper input, additional data conversion of the uni-axial data needs to be done.
Starting from uniaxial (engineering or true) stress strain data, the elasticity modulus (Young) needs to be
derived for a number of engineering or true strain data points. See Figure 7-14.
Uniaxial stress & Young vs Strain
160 10000
140
8000
Young's modulus
120
engineering strain
100 6000
Stress
Young's modulus
80
60 4000
40
2000
20
0 0
0 0.01 0.02 0.03 0.04 0.05
Strain
Figure 7-14 Uniaxial Stress and Young vs Strain
The definition of these invariants in Marc is as follows:

I 1 = trace [ ε ] = ε x x + ε y y + ε z z
2 2 2 2
I 2 = 1 ⁄ 2 { trace ( [ ε ] ⋅ [ ε ] – I 1 ) } = ε x y + ε y z + ε z x ( ε x x ε y y + ε y y ε z z + ε z z ε x x )
2 2 2
I 3 = det [ ε ] = ε x x ε y y ε z z + 2ε x y ε y z ε z x – ε y y ε y z + ε y y ε z x + ε z z ε x y
Because only two elastic moduli (Young’s modulus and Poisson ratio) which are functions of the
invarients are defined, the material will behave isotropically in the simulation. In the input table, the three
strain invariants are indicated by means of id’s 70, 71, and 72.
For uniaxial and plane stress elements, a special treatment is followed in Marc. The out-of-plane strains
are not computed internally. Therefore incompressibility (poisson = 0.5, see Figure 7-15) is assumed,
such that I 1 = 0 : the first invariant is very closely related to the hydrostatic strain ( I 1 ⁄ 3 ) . The other
invariants automatically follow from this assumption.
Influence of incompressibility on first strain invariant
0.02
0.015
First strain invariant
(poisson = 0.3) 0.01

0.005
(poisson = 0.5)
0
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
-0.005
-0.01
-0.015
-0.02
Engineering strain
Figure 7-15 Influence of Incompressibility on First Strain Invariant
For a uniaxial element:

ε x y = ε y z = 0 and ε y y = ε z z
1
I 1 = 0 = >ε y y = ε z z = – --- ε x x
2
3 2
I 2 = --- ε x x
4
1 3
I 3 = --- ε x x
4
For a plane stress element:

εy z = εz x = 0
CHAPTER 7 407
Material Library
I 1 = 0 = >ε z z = – ( ε x x + ε y y )
2 2 2
I2 = εx y + εx x + εy y + εx x εy y
2
I3 = ( εx x + εy y ) ( εx y – εx x εy y )
It is important to realize that the incompressibility assumption for different elements has a direct impact
on the material data supplied in the input deck. For different element types, the input data derived from
the uniaxial test data will be different. We compare the behavior of a plain stress element (e.g., element
3), a hexahedral element (e.g., element 7), a plane strain element (e.g., element 11) and a beam element
(e.g., element 52).
For the materials, we define a Poisson’s ratio of 0.3 and we prescribe a fixed displacement which brings
us into the nonlinear part of the stress-strain curve. Note that Poisson’s ratio may also vary with the
uniaxially measured strain, in this case this has to be accounted for in the 2-D and 3-D elements.
3-D Element – Three Direct Components of Strain

For the hexahedral element, we may choose any of the invariants as independent variable. In order to
get a corresponding behavior with the uniaxial test, we have to supply the first invariant in the
following form:
ε x x = ε engineering
ε y y = ε z z = – υε engineering
I 1 = ε x x + ε y y + ε z z = ( 1 – 2υ )ε engineering ( or: I 1 = ( 1 – 2ν )ε true )
2-D Elements – Two Direct Components of Strain (= Plane Stress)

Although the plane stress elements in Marc have a third strain component unequal to zero, Marc does
not compute it. It has been chosen that the user input for the stress-strain data has to follow the
incompressibility assumption ( I 1 = 0 ) . In this case, we may use the definition of the second invariant
in order to define a dependency on strain, supposing a uniaxial loading:
ε x y = ε y z = ε z x = 0 (uniaxiality)
ε y y = – υε engineering
from I 1 = 0 , we deduce (see above):
2 2
I2 = εx x + εy y + εx x εy y
2
I 2 = ( υ 2 – υ + 1 )ε engineering
2-D Elements – Three Direct Components of Strain (= Plane Strain)

For the plane strain elements, the third direct strain term is zero. In this case, incompressibility is not
assumed and the strain invariants can be easily derived from the conditions: σ y y = 0 and ε x x = 0
under uniaxial loading:
ε x y = ε y z = ε z x = ε z z = 0 (uniaxiality and plain strain)
ε y y = υ ⁄ ( υ – 1 ) ε engineering
I 1 = ε x x + ε y y + ε z z = { 1 + ν ⁄ ( ν1 ) }ε engineering
I2 = εx x εy y
2
I 2 = υ ⁄ ( υ – 1 ) ε engineering
1-D Elements – One Direct Components of Strain

For the beam and truss elements, we cannot use the first invariant as independent variable either, since
I 1 = 0 . We may quickly derive that:
2 2
I 2 = 3 ⁄ 4ε engineering (or I 2 = 3 ⁄ 4ε true )
Loading Under Tension/Compression

It must be noted that for both beam and plane stress elements a nonsymmetric behavior in compression
cannot be described by means of the second invariant because I 2 ≥ 0 (see Figure 7-16).
The third invariant can be used for this purpose; e.g., for the 1-D elements:
3
I 3 = 1 ⁄ 4ε engineering 3
(or I 3 = 1 ⁄ 4ε true )
And finally for the 2-D plane stress elements:

ε x y = ε y z = ε z x = 0 (uniaxiality)
ε y y = – υε engineering
CHAPTER 7 409
Material Library
Second strain invariant (> 0)
1.4000E-03
1.2000E-03
Second strain invariant
1.0000E-03
8.0000E-04
6.0000E-04
4.0000E-04
2.0000E-04
0.0000E+00
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
Engineering strain
Figure 7-16 Second Strain Invariant (>0)
from I 1 = 0 , we deduce
2
I3 = ( εx x + εy y ) ( εx y – εx x εy y )
3
I 3 ν ( υ – 1 )ε engineering
Data Approximation
Finally it is important to realize that the second and third invariants are nonlinear functions of the strain.
As a result, even a linear stress-strain curve will form a polynomial of the second or third order for the
second and third strain invariant respectively. It is, therefore, wise to apply the load in a sufficient number
of increments, even though the actual behavior is only linear. This is shown in Figure 7-17.
Type 3: Principal Strain Space Model

In the principal strain space theory, the strain tensor’s principal values are now computed. The
compliance matrix is formed using a total strain approach; i.e., the secant modulus is computed for each
principal strain by looking up the stress in the user defined table. This results in three values for the
modulus. These values are then used to form an orthotropic stress-strain law together with poisson’s ratio
and a shear modulus, which are also user defined.
The definition of the orthotropic stress-strain law is described in Volume C (chapter 3). It must be noted
that the constants have to fulfill certain requirements in order to guarantee stability. These criteria are
checked by Marc. However, if stability cannot be guaranteed, Marc will continue to solve the increment
at hand, and moreover that it will continue if the increment converges. From this point on, the solution
may not be trustworthy anymore and care should be taken.
Influence of # of datapoints on third strain invariant
2.0000E-05
1.5000E-05
1.0000E-05
Third strain invariant
5.0000E-06
0.0000E+00
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05
-5.0000E-06
-1.0000E-05
-1.5000E-05
-2.0000E-05
Engineering strain
Figure 7-17 Influence of Number of Data Points on the Third Strain Invariant
Theoretical Background
The Principal strain space theory can be defined as the incremental stress strain law which is based upon
an orthotropic Hooke’s law expressed with respect to a set of axis aligned with the principal axes based
upon the strain tensor. The moduli ( E , ν , G ) are evaluated at the current values of the principal strains.
Consider E to be the reference material system of the element. For a continuum element, this is the global
system, while for a shell element this is the ν 1 , ν 2 , ν 3 system. Now, consider the total strain at the
beginning of the increment ε n . One can obtain the principal strains and the eigenvectors of this system.
The principal strains are ε 1 , ε 2 , ε 3 and the eigenvectors are n 1 , n 2 , n 3 , The matrix that transforms the
E system to the principal system is R .
The state is then rotated into the principal, this includes the stresses, strains, thermal strains and
creep strains.
σ n + 1 = R T σR , etc.
Based upon the principal strains, the material properties for the current value of the principal strains
are determined.
The evaluated material constants are tested to make sure stability is not violated, also positive
definiteness of the compliance matrix is checked.
Then the constitutive law is evaluated based upon the current Young’s moduli, Poisson ratios, and
shear modulus.
CHAPTER 7 411
Material Library
Based upon either the estimated or calculated incremental strain, the incremental stress in principal space
is determined. Subsequently the total stress in principal space is calculated and the quantities are rotated
back to the E system.
p
σ n + 1 = Rσ n + 1 R T = R ( σ p + Δσ p )R T
The constitutive law is also rotated back into the E system.
C = RC p R T
This is then used in the tangent stiffness matrix calculation.
Type 4: (Bimodulus) Elasticity Model

This is conventional Hooke’s law, where the Young’s modulus and Poisson ratio may be function of
temperature or position, but not strain or stress. Further below, a description is given of the treatment
regarding tension and compression limits.
Linear Elasticity Bimodulus Elasticity
Type 5: (Bimodulus) Elasticity Model with Either No Tension, Limited Tension, No

Compression, or Limited Compression Enhancements
In this model, the hydrostatic strain will be calculated and if the value is positive, the tensile values of
Young’s modulus and Poisson’s ratio will be used. If the hydrostatic strain is 0, as in the very beginning
of the analysis, the tensile values will be used.
For Model 4
Given total strains, the total stresses are calculated based upon tensile properties. This results in C and
σn + 1 ,
For Model 5
To distinguish a tension or compression stress state the hydrostatic strain is computed.
If the hydrostatic strain is positive the total stresses are calculated based upon tensile properties.
Resulting in C ten and σ n + 1
If the hydrostatic strain is negative, the total stresses are calculated based upon compressive properties.
This results in C com and σ n + 1 ,
No Tension Material No Compression Material
Bimodulus with both Tension and Compression Cut-off
For Either Model 4 or 5

The treatment of the cut-offs works as follows, if they are defined in the simulation.
The principal stresses σ 1 , σ 2 , σ 3 and the eigenvectors are n 1 , n 2 , n 3 are obtained.
If σ I > Tensile stress limit, then principal direction i will be reduced.
If σ I < Compressive stress limit, then principal direction i will be reduced.
The reduction implies that a new secant modulus is computed for the current reduced direction. The
maximum secant modulus in one of the reduced directions is used to form a new stress-strain law.
The transformation matrix: R e p is obtained between Element system E and Principal system P .
CHAPTER 7 413
Material Library
A stiffness matrix is recreated using the secant modulus of the current limit stress vs strain state.This is
also done to improve stability.
Cz = RepCpRe pT
Rotate stresses to principal system and maximize the stress to the limit in ith component, and then rotate
back into element system:
p
σn + 1 = R e p T σn + 1 R e p
p–z
σ nz + 1 = R e p σ n + 1 R e p T
Note that there is no need to form C p because C is from isotropic elasticity, and R is a proper
orthogonal matrix. Therefore the stiffness matrix C can be formed in the global system with the limited
secant modulus.
Type 6: Nonlinear Orthotropic Elasticity Model

This model is based upon orthotropic Hooke’s law, written with respect to a fixed preferred material
orientation. The difference is that the nine constants, E 11 , E 22 , E 33 , ν 12 , ν 23 , ν 31 , G 12 , G 23 , G 31
may be a function of the spatial position, temperature, and the strain in the associated direction. This can
be expressed as:
E 11 = E 11 ( X ,T ,ε 11 ) , etc.
ν 12 = ν 12 ( x ,T ,γ 12 ) , etc.
G 12 = G 12 ( x ,T ,γ 12 ) , etc.
The material properties are evaluated based upon the current value of the independent variables, and the
Hooke’s law is used to create the current tangent modulus. The incremental stress follows from these
moduli and the incremental strains.
The evaluated material constants are evaluated to make sure stability is not violated, also it is determined
if the compliance matrix is positive definite.
As the material orientations do not change, the usual procedure for defining them is used.
Thermo-Mechanical Shape Memory Model

NiTi alloys with near-equiatomic composition exhibit a reversible, thermoelastic transformation
between a high-temperature, ordered cubic (B2) austenitic phase and a low-temperature, monoclinic
(B19) martensitic phase. The density change and thus the volumetric are small and on the order of 0.003.
The transformation strains are, thus mainly deviatoric, of the order of 0.07-0.085. However, these small
dilational strains do not necessarily lead to a lack of pressure sensitivity in the phenomenology. The
behavior of nitinol is different depending on whether the materials are subjected to hydrostatic tension
or compression.
Typical phenomenology is shown in Figure 7-18 taken from Miyazaki et al. (1981). The curves indicate
that upon cooling, the material transformation from austenite to martensite begins once the M s
temperature is reached. Upon further cooling, the volume fraction on martensite is a given function of
temperature; the volume fraction becomes 100% martensite when the M f temperature is reached. Upon
heating, transformation from martensite to austenite begins only after A s temperature is reached. This
re-transformation is complete when the A f temperature is reached. Finally, note that the four
transformation temperatures are stress dependent. The experimental data indicate the M s , M f , A s , and
A f may be approximated from their stress-free values, M s0 , M f0 , A s0 , and A f0 by
σe q
M s = M s0 + -------- ,
Cm
σe q
M f = M f0 + -------- ; and
Cm
σe q
A s = A s0 + -------- ,
Ca
σe q
A f = A f0 + -------- .
Ca
where σ e q is the von Mises equivalent stress. At a sufficiently high temperature, often called the M d
temperature, transformation to martensite does not occur at any level of stress.
The transformation characteristics such as the transformation temperatures depend sensitively on alloy
composition and heat treatment.
CHAPTER 7 415
Material Library
Mf As
1.0
Austenite to martensite &
0.9 martensite to austentie
decomposition
0.8
Stress = 0
0.7
Martensite Volume Fraction
Note:
0.6 600
After partial
Tensile Stress (MPa)

transformation,
0.5 decomposition
begins at As. 400
0.4
200
0.3
0
0.2 77 150 Ms 200 Af 250 300
Temperature (K)
0.1
MS Af
0.0
0 10 20 30 40 50 60 70 80 90 100
Temperature
Figure 7-18 Austenite to Martensite and Martensite to Austenite Decomposition
Transformation Induced Deformation

For the discussion of the thermo-mechanical response of NiTi, the data of Miyazaki et al. (1981) is shown
in Figure 7-19. Following this thermal history, it is observed that, when unstrained specimens with fully
austenitic microstructures are cooled, the transformation to martensite begins at a temperature of 190K;
the transformation is complete at 128K. This established the so-called martensite start ( M s0 ) and
martensite finish ( M f0 ) temperatures at 190K and 128K, respectively. With the imposition of an applied
uniaxial tensile stress, the low temperature martensite is favored and the M s0 and M f0 temperatures
increase. Upon heating a specimen with fully martensitic microstructure, the reverse transformation is
observed to begin at a temperature of 188K and to be complete at 221K. These define the austenite start
( A s0 ) and austenite finish ( A f0 ) temperatures, respectively. Uniaxial tension tests are carried out in
temperature ranges where T < M s , M s < T < A f , and A f < T < T c where T c is defined as the
temperature above which the yield strength of the austenitic phase is lower than the stress required to
induce the austenite-to-martensite transformation.
(a) 77K (b) 153K (c) 164K

300
200
100
0 0 0
400
(d) 224K (e) 232K (f) 241K
Tensile Stress (MPa)
300
200
100
0 0 0
(h) 273K (i) 276K
600 (g) 263K
400
200 Ms = 190K
AF = 221K
0 2 4 0 2 4 0 2 4
Strain (%)
Figure 7-19 Thermal History
In the temperature range where T < M f , the microstructures are all martensitic. The stress versus strain
curves display a smooth parabolic type of behavior which is consistent with deformation caused by the
movement of defects such as twin boundaries and the boundaries between variants. Note that unloading
occurs nearly elastically and that the accumulated deformation, caused by the reorientation of the existing
martensite and the transformation of any pre-existing austenite, remains after the specimen is completely
unloaded. Note also that the accumulated deformation is entirely due to oriented martensite and this
would be recoverable upon heating to temperatures above the ( A s – A f ) range. This would, then, display
the shape memory effect.
Pseudoelastic behavior is displayed in the temperature range A f < T < T c . In this range, the initial
microstructures are essentially all austenitic, and stress induced martensite is formed, along with the
associated deformation; upon unloading, however, the martensite is unstable and reverts to austenite
thereby undoing the accumulated deformation. Note that, as expected, the stress levels rise with
increasing temperature. In this range, the transformation induced deformation is nearly all reversible
upon unloading.
At temperatures where T > T c , plastic deformation appears to precede the formation of stress induced
martensite. The unloading part of the stress versus strain behavior displays nonlinearity and the
unloading is now associated with permanent (plastic) deformation. Permanent deformation, due to plastic
deformation of the martensite, is non-recoverable and, if such deformation is large, shape memory
behavior is lost.
CHAPTER 7 417
Material Library
Constitutive Theory
The model formulated below is based on the kinematics of small strains, although the extension to large
strain is straightforward. Accordingly, the incremental strain, Δε , is simply the sum of the following
contributions:
Δε = Δε e l + Δε t h + Δε p l + Δε p h (7-87)
In Equation (7-87), Δε e l is the incremental elastic, or lattice, strain rate; Δε t h is the incremental
thermal strain, Δε p l is the incremental visco-plastic strain, and Δε p h is the incremental strain associated
with thermoelastic phase transformations. The incremental elastic strain is taken to be simply related to
a set of elastic moduli, L , and the incremental Cauchy stress rate, Δσ , as
Δσ = LΔε e l (7-88)
To calculate the coefficient of thermal expansion of the composite, the rule of mixtures is used as
α = ( 1 – f )α A + fα M .
In the above equations, the superscripts A and M refer to the austenite and martensite values, and f is
the volume fraction of martensite.
Phase Transformation Strains

As noted earlier, the phase transformation induced strains are a result of the formation of oriented, stress
induced, martensite and the reorientation of randomly oriented thermally induced martensite. To account
for this, Δε P h is expressed as
Δε P h = Δε T R I P + Δε T W I N
where
3 σ'
Δε T R I P = Δf ( + ) g ( σ e q )ε eTq --- -------- + Δf ( + ) ε νT I + Δf ( - ) ε P h . (7-89)
2 σe q
and
3σ'
Δε T W I N = fΔg ( σ e q )ε eTq ------------ { σ e q } { σ e q – σ egf f } . (7-90)
2σ e q
where Δf = Δf ( + ) + Δf ( - ) and { } represents McCauley’s bracket where { x } = ---  ---------------- ,

1 x+ x
2 x 
x ≠ o.
In Equation (7-89), Δf ( + ) represents the rate at which martensite is formed, ε eTq is the magnitude of the
deviatoric part of the transformation, and ε νT is the volumetric part of the transformation strain. The
function g ( σ e q ) is schematically depicted in Figure 7-20, and is a measure of the extent to which the
martensite transformation strains are aligned with the deviatoric stress. σ e q is the equivalent stress
defined as:
3 d
σe q = --- σ : σ d where σ d is the deviatoric stress.
2
1.1
0.9
0.7
0.5
g
0.3
0.1
-1.0
0.0 0.5 1.0 1.5 2.0
stress/g0
Figure 7-20 Function g ( σ e q )
The first two terms in Equation (7-89) describe the development of transformation induced strains due to
the formation of stress induced (partially oriented) martensite. Δf ( - ) is the change of formation of
austenite; for example, the rate at which the volume fraction of martensite decreases. The last term in
Equation (7-89), therefore, represents the recovery of the accumulated phase transformation strain.
Note that there is no dilatational contribution to Δε T W I N since f is fixed. Note that the twinning strain
rate is zero when σ e q is less than σ egf f , or when the magnitude of the stress change is negative
( Δσ e q < 0 ). Hence, σ egf f can be considered as a stress below which no twinning is possible.
The function g represents the extent to which the transformation strains are coaxial with the applied
deviatoric stress. This function can be calibrated with the experimental data. Note for uniaxial stress-
strain curves performed below the martensite finish temperature, the material starts as 100% martensite,
and that other than elastic strains, the deformation is dominated by the “twinning” of the randomly
oriented martensite.
CHAPTER 7 419
Material Library
A functional form that leads to sufficient fit to most experimental data has been implemented in Marc.
g g g
σe q b σe q d σe q f
g ( σ e q ) = 1 – exp g a  -------- + g c  -------- + g e  --------
g0 g0 g0
In most cases, the first term is sufficient, and a value of g a < 0 and g b = 2 yields the best results. g 0
is a stress level used to non-dimensionalize the constants, and can be chosen such that g → 1 when the
σ e q → g 0 . In some cases, it is necessary to include the higher powers of equivalent stress for better
experimental fits. In these, cases suggested values for g d = 2.55 or 2.75 and g e = 3 . However,
depending on the values of g c and g e , this could lead to maxima or minima values of g in the range of
interest. Note that 0 ≤ g ≤ 1 and it should be a monotonically increasing function (an increase in the
stress level should lead to an increase in the increment of the phase strains). Thus, cut off values of g are
provided in Marc, such that when g reaches its maximum value g = g m a x at the stress level
σe q = σm
g
a x g 0 , it is held constant at the value g m a x . For the proper selection of g 0 , see the
following section.
Experimental Data Fitting for Thermo-mechanical Shape

Memory Alloy
The properties and transformation/retransformation behavior of Nititol depend upon alloy chemistry,
microstructure and the thermal processing applied to the specimens and eventually to the components
built from them. Every time any of the above change, it might be necessary to redo the calibration.
Calibration of Nitinol experimental data works best if, in fact, all specimens can be initially rendered as
100% austenite. The list of properties that require calibration is given as follows:
• The “unstressed transformation temperatures”, M s0 ,M f0 ,A s0 ,A f0 .
• The coefficients C M , and C A that provide the stress dependence of the

transformation temperatures.
• The elastic constants ( E M ,E A ,v M and v A ).
• Coefficients of thermal expansion, α M ,α A .

• The calibration of the detwinning function, g ( σ e q ) that provides the description of the degree to
which martensite is co-axial with the deviatoric stress state.
M and σ A ), and their strain
• The yield stress of the pure martensite and austenite phase ( σ Y Y
hardening properties.
• The calibration of the transformation strains, ε eTq and ε vT .
Transformation Temperatures and Their Stress Dependence ( M s0 ,M f0 ,A s0 ,A f0 ,C M ,

and C A )
For almost any use of shape memory alloy, it is highly desirable that one knows the Transformation
Temperatures (TTRs) of the alloy. The TTRs are those temperatures at which the alloy changes from the
higher temperature austenite to the lower temperature martensite or vice versa. There are in common use
with NiTi alloys to provide helpful data to product designers – Constant Load, DSC and Active A f . The
detail procedures to obtain TTRs using the above methods are shown in website (www.sma-inc.com). It
is recommended that combined dilatometry and DSC tests be performed on unstressed specimens of
thermally processed material to establish both the unstressed transformation temperatures and the
thermal expansion coefficients. Those tests would provide a baseline set of values for M s0 ,M f0 ,A s0 , and
A f0 . Note that the TTRs are stress dependent parameters, but it is difficult, in practice, to prepare totally
unstressed samples. In order to determine the TTRs at zero stress, experimental data must be obtained at
two or more stress levels. The particular transformation point of interest can then be extrapolated to zero
stress. The estimations of TTRs, C M , and C A are shown in Figure 7-21. The typical range of TTRs is –
200 to 100°C. So, it is difficult to recommend the default values. As references, there are examples for
two different SMA materials below.
SMA 1)
M s0 : – 50°C , M f0 : – 100°C , A s0 : 5°C , A f0 : 20°C , C M : 6.0Mpa/°C , C A : 8.0Mpa/°C
SMA 2)
M s0 : 190K , M f0 : 128K , A s0 : 188K , A f0 : 221K , C M : 5.33Mpa/K , C A : 6.25Mpa/K
Stress
CM
CA
M f0 M s0 A s0 A f0
Temperature
Figure 7-21 Typical Stress vs. Temperature Curve Showing the Stress Dependence of Martensite and
Austenite Start and Finish Temperature
Elastic Constants ( E M ,E A ,v M , and v A )

Literature estimates for the elastic moduli of martensite and austenite are typically in the range of
E M = 28000-41000 Mpa , E A = 60000-83000Mpa , v M = v A = 0.33 . However, most
experimental data appears to be significantly different than these. It is, therefore, suggested that estimates
CHAPTER 7 421
Material Library
of these moduli should be made using actual experimental data for the materials being calibrated. Initial
loading from a state corresponding to 100% austenite produces a linear elastic response from which E A
can be readily estimated as in Figure 7-22. The modulus of martensite ( E M ) is also estimated for the
unloading line, again as illustrated in Figure 7-22. In this figure, the loading should be performed to
produce 100% martensite and thus the unloading occurs with the elastic response of martensite.
Typical pseudoelastic
response
T = T1°C
EA
Stress
EM
Strain
Figure 7-22 Typical Stress-strain Curves in the Pseudo-elastic Regime, Depicting the Elastic Moduli
Thermal expansion coefficients ( α M ,α A )

A recommended method for measuring thermal expansion coefficients is through the use of dilatometry
whereby carefully controlled cycles of temperature can be made. An alternative to this type of precise
calibration, is to use literature values that have been shown to be consistent with values measured on
actual specimens.
These values are given as follows (See for example, TiNi Smart Sheet)
α M = 6.6 × 10 – 6 ⁄ °C = 3.67 × 10 – 6 ⁄ °F ; and
α A = 11.0 × 10 – 6 ⁄ °C = 6.11 × 10 – 6 ⁄ °F
Detwinning and the calibration of the g function g ( σ e q )

The phenomenology of the NiTi phase transformation is such that the alignment of the martensite varies
with the prevailing deviatoric stress. This intensity is measured via the von Mises equivalent stress, σ e q .
As shown in Equations 7-89 and 7-90, the scaling function that provides the description of the degree to
which the martensite is aligned is g ( σ e q ) . The most direct path to calibrating this g function is to fit it
to the uniaxial stress vs. strain curve for pure, randomly oriented martensite conducted at a temperature
below the M f0 temperature. Such a curve is shown as Figure 7-23. The solid curve shown in Figure 7-23
is the actual measured record of uniaxial stress vs. total strain for a specimen of 100% martensite tested
at a temperature sufficiently low to ensure it remains 100% martensite. The dot curve is simply a
convenient fit to it. The parameter ε eTq is the “equivalent deviatoric transformation strain”. Note that the
function g is defined as it relates to the development of deviatoric strain due to the alignment of
martensite variants. As mentioned in the previous section, in general, the variables g a < 0 ,
g b = 2f = 2.0 , g c ≥ 0 , g d = 2.25 and 2.75 , g e ≤ 0 and g f = 3.0 yield a good match to many
experimental results. It is often observed that there exists a threshold equivalent stress level below which
detwinning does not occur; this stress is referred to as σ egf f . The value of g at this stress is
g e f f = g ( σ fgf ) . Note that from Equation (7-90), twinning strain is zero when σ e q < σ egf f . In addition,
it is also found, in practice, that the function g tends to approach unity at a finite equivalent stress level,
g g
called σ 0 . By definition, g ( σ 0 ) = 1 . Also, g a should be chosen to match the general shape of the
function. Since the ratio σ e q ⁄ g 0 is less than one, the higher powers take effect later, and thus g c can be
added to lower the middle slope of the curve and g e to fix the final slope of the curve. However,
depending on the relative values of g a , g b and g c , this curve might reach a maximum in the range of
interest, and therefore, it should be cut-off at its maximum value g m a x . The value of g m a x which is
reach at a stress value σ e q ⁄ g 0 = σ m
g
a x are also supplied as an input to Marc. Usually
g g
g 0 = 2σ e f f ∼ 10σ e ff is a good approximation. But, the selection of g 0 depends on the
experimental measurement.
g0
ε T W I N = g ( σ e q )ε eTq Experimental
Model
Stress
g0
Strain
Figure 7-23 Typical Stress-Strain Curve of 100% Martensite Tested Below M f0 Temperature
Others
M
Yield stresses of the pure martensite and austenite phases for NiTi: σ Y and σ Y
A.
M
σY = 70-140 Mpa
A
σY = 195 – 690 Mpa
CHAPTER 7 423
Material Library
The calibration of the transformation strains for NiTi:
ε eTq (deviatoric transformation strain): 0.05-0.085
ε vT (volumetric transformation strain): 0 – 0.003.
σ egf f (detwinning stress): 100-150 Mpa
g0 g
2σ e f f ∼ 10σ egf f
Note: The current model uses a nonsymmetric Jacobian matrix. It is recommended that the
nonsymmetric solver be used to improve convergence.
Mechanical Shape Memory Model

Shape-memory alloys can undergo reversible changes in the crystallographic symmetry-point-group.
Such changes can be interpreted as martensitic phase transformations, that is, as solid-solid diffusion-
less phase transformations between a crystallographically more-ordered phase (the austenite or parent
phase) and a crystallographically less-ordered phase (the martensite). Typically, the austenite is stable at
high temperatures and high values of the stress. For a stress-free state, we indicate with the temperature
above which only the austenite is stable and with the temperature below which only the martensite is
stable.
The phase transformations between austenite and martensite are the key to explain the superelasticity
effect. For the simple case of uniaxial tensile stress, a brief explanation follows (Figure 7-24). Consider
a specimen in the austenitic state and at a temperature greater than A; accordingly, only the austenite is
stable at zero stress. If the specimen is loaded, while keeping the temperature constant, the material
presents a nonlinear behavior (ABC) due to stress-induced conversion of austenite into martensite. Upon
unloading, while again keeping the temperature constant, a reverse transformation from martensite to
austenite occurs (CDA) as a result of the instability of the martensite at zero stress. At the end of the
loading/unloading process, no permanent strains are present and the stress-strain path is a closed
hysteresis loop.
σ C
D
A ε
Figure 7-24 Superelasticity

At the crystallographic level, if there is no preferred direction for the occurrence of the transformation,
the martensite takes advantage of the existence of different possible habit plates (the contact plane
between the austenite and the martensite during a single-crystal transformation), forming a series of
crystallographically equivalent variants. The product phase is then termed multiple-variant martensite
and it is characterized by a twinned structure. However, if there is a preferred direction for the occurrence
of the transformation (often associated with a state of stress), all the martensite crystals trend to be formed
on the most favorable habit plane. The product is then termed single-variant martensite and is
characterized by a detwinned structure. According to the existence of different types of single-variant
martensite species, the conversion of each single-variant martensite into different single variants is
possible. Such a process, known as a reorientation process, can be interpreted as a family of martensite
phase transformations and is associated with changes in the parameters governing the single-variant
martensite production (hence, it is often associated to nonproportional change of stress).
In addition to the thermo-mechanical shape memory model, a superelastic shape memory alloy model is
also implemented in Marc based on the work of Auricchio [Ref. 1] and [Ref. 2]. This work has been
enhanced to allow different elastic properties for the Austenite and Martensite phases.
The superelastic shape memory model has been implemented in Marc in the framework of multiplicative
decomposition. We assume the deformation gradient, F as the control variable, and the martensite
fraction, ξ S as the only scalar internal variable. We also introduce a multiplicative decomposition of F
in the form:
F = Fe Ft r
where F e is the elastic part and F t r is the phase transition part.
Assuming an isotropic elastic response, the Kirchhoff stress τ and the elastic left Cauchy-Green tensor
b e , defined as: b e = F e F e T , share the same principal directions. Accordingly, the following spectral
decompositions can be introduced:
3
τ =  τA nA ⊗ nB
A = 1

d
d = σA nA ⊗ nB
A = 1
3
be =  ( λA
e )2 nA ⊗ nB
A = 1
with λ A
e the elastic principal stretches and σ d the deviatoric part, according to the relation:
τ = pI + σ d (7-91)
CHAPTER 7 425
Material Library
where I is the second-order identity tensor, p is the pressure, defined as p = tr ( τ ) ⁄ 3 , and tr ( . ) is

the trace operator. We can write Equation (7-91) with the following component form:
d
τA = p + σ A (7-92)
with
d
p = Kθ e , σ A = 2Ge A
e .
Phase Transformations and Activation Conditions

We consider two phase transformations: the conversion of austenite into martensite (A→ S) and the
conversion of martensite into austenite (S→ A). To model the possible phase-transformation pressure-
dependence, we introduce a Drucker-Prager-type loading function:
F( τ) = σ d + 3αp (7-93)
where α is a material parameter and

. indicates the Euclidean norm, such that:
3 1⁄2

d 2
σd = ( σA ) .
A = 1
Indicating variants in time with a superposed dot, we assume the following linear forms for the evolution
of ξ S :
·
· F
ξ S = H A S ( 1 – ξ S ) --------------------- for (A→ S) (7-94)
F – R fA S
·
· F
ξ S = H S A ξ S --------------------- for (S→ A) (7-95)
F – R fS A
where
σ fAS  --- + α , R fS A = σ fSA  --- + α

2 2
R fA S =
 3   3 
SA
with σ sA S , σ fA S , σ sS A , and σ fS A material constants. The scalar quantities H AS and H embed the
plastic-transformation activation condition – hence, allowing a choice between Equations (7-94) and
(7-95) – and they are defined by the relations:
H AS = 1 , if R sA S < F < R fA S , or F· > 0 . Otherwise, H AS = 0 .
H SA = 1 , if R fS A < F < R sS A , or F· < 0 . Otherwise, H SA = 0 .

where
σ sAS  --- + α , R sS A = σ sSA  --- + α .

2 2
R sA S =
 3   3 
Time-discrete Model
The time-discrete model is obtained by integrating the time-continuous model over the time interval
[ t n, t ]. In particular, we use a backward-Euler integration formula for the rate-equations evaluating all
the nonrate equations at time t . Written in residual form and clearing fraction from Equations (7-94) and
(7-95), the time-discrete evolutionary equations specialize to:
R A S = ( F – R fAS )λ s – H A S ( 1 – ξ S ) ( F – F n ) = 0 (7-96)
R S A = ( F – R fSA )λ s – H S A ξ S ( F – F n ) = 0 (7-97)
where
t
·
λS =  ξ dt = ξ S – ξ S ,n (7-98)
tn
The quantity λ S in Equation (7-98) can be computed expressing F as a function of λ S and requiring the
satisfaction of the discrete equation relative to the corresponding active phase transition.
The detailed solution algorithm for stress update and consistent tangent modulus are given in the work
of Auricchio [Ref. 2].
In the enhanced version of this model, the user enters different elastic constants for the two phases. In
this case, the effective elastic moduli are taken as:
A M
E = E ( 1 – ξS ) + E ξS
A M
ν = ν ( 1 – ξS ) + ν ξS
Experimental Data Fitting for Mechanical Shape Memory Alloy

The experiment for mechanical shape memory alloy is quite simple.
1. To determine the transformation stresses ( σ SA S ,σ fAS ,σ SS A ,σ fS A ):
σ SA S Initial Stress for Austenite to Martensite
σ fA S Final Stress for Austenite to Martensite

CHAPTER 7 427
Material Library
σ SS A Initial Stress for Martensite to Austenite
σ fS A Initial Stress for Martensite to Austenite
Uniaxial tension test is performed at the same temperature at which the simulation is desired.
Here is one example set for SMA materials.
σ SA S = 500Mpa , σ fA S = 600Mpa , σ SA S = 300Mpa , σ fS A = 200Mpa
2. α :It is measured from the difference between the response in tension and compression.
Case 1) if the behavior in tension and compression are the same, the value is set to 0.
Case 2) if the behavior in tension and compression have a difference as in the classical case of
SMA, the value is usually set to 0.1 if there is no compression data for the phase
transformation. One value for the phase transformation in compression, say σ sA S–
(sigAS_s_compression) is available, α is calculated as follows:
α = sqrt ( 2 ⁄ 3 ) ( σ SA S, – – σ SA S ) ⁄ ( σ SA S, – + σ SAS )
3. ε L : epsL is a scalar parameter representing the maximum deformation obtainable only by

detwinning of the multiple-variant martensite (or maximum strain obtainable by variant
orientation). Classical values for epsL are in the range 0.005 and 0.10. Marc sets the default
value as 0.07.
Note: The mechanical shape memory model only supports ndi = 3 case (3-D, plane-strain and
axisymmetric elements). It does not support either ndi = 1 or ndi = 2 cases (1-D and
plane-stress elements).
Conversion from Thermo-Mechanical to Mechanical SMA

Table 7-2 Conversion Table
Thermo-Mechanical SMA Mechanical SMA Enhanced Mechanical SMA
E
A E = 0.5 ( E A + E M ) E
A
ν
A ν = 0.5 ( ν A + ν M ) ν
A
E
M E = 0.5 ( E A + E M ) E
M
ν
M ν = 0.5υ ( ν A + ν M ) ν
M
Table 7-2 Conversion Table (continued)

Thermo-Mechanical SMA Mechanical SMA Enhanced Mechanical SMA
ε L = sqrt  --- ε e q ε L = sqrt  --- ε e q

εe q
T 3 T 3 T
 2  2
CM CM CM
CA CA CA
To To To
Table 7-3 Prediction from Linear Algebra

The Relationship between Mechanical Model
and Thermo-Mechanical Model
AS 0
σs = ( T o – M s )C m
AS 0
σf = ( T o – M f )C m
0
σ sS A = ( T o – A s )C a
0
σ fS A = ( T o – A f )C a
σ AS
σf
AS
σs
CA
SA
σs
SA
σf
CM
T
Mf Ms As Af To
CHAPTER 7 429
Material Library
Elastomer
An elastomer is a polymer which shows nonlinear elastic stress-strain behavior. The term elastomer is
often used to refer to materials which show a rubber-like behavior, even though no rubbers exist which
show a purely elastic behavior. Depending upon the type of rubber, elastomers show a more or less
strongly pronounced viscoelastic behavior. Marc considers both the viscous effects and the elastic
aspects of the materials behavior. These materials are characterized by their elastic strain
energy function.
Elastomeric materials are elastic in the classical sense. Upon unloading, the stress-strain curve is retraced
and there is no permanent deformation. Elastomeric materials are initially isotropic. Figure 7-25 shows
a typical stress-strain curve for an elastomeric material.
σ, Stress
100%
e, Strain
Figure 7-25 A Typical Stress-Strain Curve for an Elastomeric Material
Calculations of stresses in an elastomeric material requires an existence of a strain energy function which
is usually defined in terms of invariants or stretch ratios. Significance and calculation of these kinematic
quantities is discussed next.
In the rectangular block in Figure 7-26, λ 1 , λ 2 , and λ 3 are the principal stretch ratios along the edges
of the block defined by
λi = ( Li + ui ) ⁄ Li (7-99)
L3 λ3L3
Undeformed
Deformed
λ1L1 λ2L2
L2
L1
Figure 7-26 Rectangular Rubber Block
In practice, the material behavior is (approximately) incompressible, leading to the constraint equation
λ1 λ2 λ3 = 1 (7-100)
the strain invariants are defined as
2 2 2
I1 = λ + λ2 + λ3
1
2 2 2 2 2 2
I2 = λ1 λ + λ λ +λ λ (7-101)
2 2 3 3 1
2 2 2
I3 = λ1 λ2 λ3
Depending on the choice of configurations, for example, reference (at t = 0 ) or current ( t = n + 1 ),

you obtain total or updated Lagrange formulations for elasticity. The kinematic measures for the two
formulations are discussed next.
A. Total Lagrangian Formulation
The strain measure is the Green-Lagrange strain defined as:
1
E i j = --- ( C i j – δ i j ) (7-102)
2
where C i j is the right Cauchy-Green deformation tensor defined as:
Ci j = Fk i Fk j (7-103)
in which F kj is the deformation gradient (a two-point tensor) written as:
∂x k
F kj = --------- (7-104)
∂X j
The Jacobian J is defined as:
1---
2
J = λ 1 λ 2 λ 3 = ( det C i j ) (7-105)
Thus, the invariants can be written as:
I1 = Ci i (implied sum on i)
2
( Ci j Ci j – ( Ci i ) )
I 2 = ------------------------------------------ (7-106)
2
1
I 3 = --- e i j k e p q r C i p C j q C k r = det ( C i j )
6
(7-107)
CHAPTER 7 431
Material Library
7 in which e i j k is the permutation tensor. Also, using spectral decomposition theorem,
2 A A
Material Ci j = λA Ni Nj (7-108)
Library
2
in which the stretches λ A are the eigenvalues of the right Cauchy-Green deformation tensor, C i j
A
and the eigenvectors are N i .
B. Updated Lagrange Formulation

The strain measure is the true or logarithmic measure defined as:
1
ε i j = --- l n b i j (7-109)
2
where the left Cauchy-Green or finger tensor b i j is defined as:
bi j = F i k Fj k (7-110)
Thus, using the spectral decomposition theorem, the true strains are written as:
3
A A
εi j =  ln ( λ A ) n i n j (7-111)
A = 1
A
where n i is
the eigenvectors in the current configuration. It is noted that the true strains can
also be approximated using first Padé approximation, which is a rational expansion of the
tensor, as:
–1
εi j = 2 ( V i j – δ i j ) ( Vi j + δi j ) (7-112)
where a polar decomposition of the deformation gradient F i j is done into the left stretch tensor
V i j and rotation tensor R i j as:
Fi j = Vi k Rk j (7-113)
The Jacobian J is defined as:

1---
2
J = λ 1 λ 2 λ 3 = ( det b i j ) (7-114)
and the invariants are now defined as:
I1 = bi i
1 2
I 2 = --- ( b i j b i j – ( b i i ) )
2 (7-115)
1
and I 3 = --- e i j k e p q r b i p b j q b k r = det ( b i j )
6
It is noted that either Equation (7-106) or Equation (7-115) gives the same strain energy since
it is scalar and invariant. Also, to account for the incompressibility condition, in both
formulations, the strain energy is split into deviatoric and volumetric parts as:
W = W deviatoric + W volumetric (7-116)
Thus, the generalized Mooney-Rivlin (gmr) and the Ogden models for nearly-incompressible
elastomeric materials are written as:
N N
gmr m n
Wd e v i a t o r i c =   Cm n ( I 1 – 3 ) ( I2 – 3 ) (7-117)
m = 1 n = 1
where I 1 and I 2 are the first and second deviatoric invariants.
A particular form of the generalized Mooney-Rivlin model, namely the third order deformation (tod)
model, is implemented in Marc.
tod
W devratoric = C 10 ( I 1 – 3 ) + C 01 ( I 2 – 3 ) + C 11 ( I 1 – 3 ) ( I 2 – 3 ) + C 20 ( I 1 – 3 ) 2 +
3 (7-118)
C 30 ( I 1 – 3 )
where
tod
W deviatoric is the deviatoric third order deformation form strain energy function,
C 10, C 01, C 11, C 20, C 30 are material constants obtained from experimental data
Simpler and popular forms of the above strain energy function are obtained as:
nh
W deviatoric = C 10 ( I 1 – 3 ) Neo-Hookean
(7-119)
mr
W deviatoric = C 10 ( I 1 – 3 ) + C 01 ( I 2 – 3 ) Mooney-Rivlin
Use the MOONEY model definition option to activate the elastomeric material option in Marc and enter
the material constants C 10, C 01, C 11, C 20, C 30 .
The TEMPERATURE EFFECTS model definition option can be used to input the temperature
dependency of the constants C 10 and C 01 . The UMOONY user subroutine can be used to modify all five
constants C 01 , C 10 , C 11 , C 20 , and C 30 . Using the table driven input, all material parameters can
reference tables to define temperature dependent behavior. For viscoelastic, the additional
VISCELMOON model definition option must be included.
CHAPTER 7 433
Material Library
The form of strain energy for the Ogden model in Marc is,
N
μ α α α
-----k-  λ 1 + λ 2 + λ 3 – 3
ogden k k k
Wd e v i a t o r i c =  αk  
(7-120)
k = 1
α
k
α – ------- α
k 3 k
where λ i = J λi are the deviatoric stretch ratios while C m n , μ k , and α k are the
material constants obtained from the curve fitting of experimental data. This capability is available in
Mentat.
If no bulk modulus is given, it is taken to be virtually incompressible. This model is different from the
Mooney model in several respects. The Mooney material model is with respect to the invariants of the
right or left Cauchy-Green strain tensor and implicitly assumes that the material is incompressible. The
Ogden formulation is with respect to the eigenvalues of the right or left Cauchy-Green strain, and the
presence of the bulk modulus implies some compressibility. Using a two-term series results in identical
behavior as the Mooney mode if:
μ 1 = 2C 10 and α 1 = 2 and μ 2 = – 2C 01 and α 2 = – 2
The material data is given through the OGDEN model definition option or the UOGDEN user subroutine.
For viscoelastic behavior, the additional VISCELOGDEN model definition option must be included.
In the Arruda-Boyce strain energy model, the underlying molecular structure of elastomer is represented
by an eight-chain model to simulate the non-Gaussian behavior of individual chains in the network. The
two parameters, nkΘ and N ( n is the chain density, k is the Boltzmann constant, Θ is the temperature,
and N is the number of statistical links of length l in the chain between chemical crosslinks) representing
initial modules and limiting chain extensibility and are related to the molecular chain orientation thus
representing the physics of network deformation.
As evident in most models describing rubber deformation, the strain energy function constructed by
fitting experiment data obtained from one state of deformation to another fails to accurately describe that
deformation mode. The Arruda-Boyce model ameliorates this defect and is unique since the standard
tensile test data provides sufficient accuracy for multiple modes of deformation.
The model is constructed using the eight chain network as follows [Ref. 3]:
Consider a cube of dimension α 0 with an unstretched network including eight chains of length
r0 = Nl , where the fully extended chain has an approximate length of Nl. A chain vector from the
center of the cube to a corner can be expressed as:
α0 α0 α0
C 1 = ------ λ 1 i + ------ λ 2 j + ------ λ 3 k (7-121)
2 2 2
Using geometrical considerations, the chain vector length can be written as:
1 1⁄2
r chain = ------- Nl ( λ 12 + λ 22 + λ 32 ) (7-122)
3
and
r chain 1 1⁄2
λ chain = ------------ = ------- ( I 1 ) (7-123)
r0 3
j
λ2 α0
C1
λ3 α0
λ1 α0
k
Figure 7-27 Eight Chain Network in Stretched Configuration
Using statistical mechanics considerations, the work of deformation is proportional to the entropy change
on stretching the chains from the unstretched state and may be written in terms of the chain length as:
r chain β
W = nkΘN  ------------ β + ln --------------  – ΘĈ (7-124)
 Nl sinh β 
where n is the chain density and Cˆ is a constant. β is an inverse Langevin function correctly accounts
for the limiting chain extensibility and is defined as:
r chain
β = L – 1  ------------ (7-125)
 Nl 
where Langevin is defined as:
1
ℑ ( β ) = coth β – --- (7-126)
β
With Equations (7-123) through (7-126), the Arruda-Boyce model can be written
Arruda-Boyce 1 1 11
W dev = nkΘ --- ( I 1 – 3 ) + ---------- ( I 12 – 9 ) + ------------------- ( I 13 – 27 )
2 20N 2
1050N
19 519 (7-127)
+ ------------------- ( I 14 – 81 ) + ------------------------- ( I 15 – 243 ) ]
3 4
7000N 673750N
CHAPTER 7 435
Material Library
Also, using the notion of limiting chain extensibility, Gent [Ref. 5] proposed the following
constitutive relation:
Gent – EI m  I 1* 
W dev = ------------- log  1 – ------ (7-128)
6  I m
where
I 1* = I 1 – 3 (7-129)
The constant EI m is independent of molecular length and, hence, of degree of cross linking. The model
is attractive due to its simplicity, but yet captures the main behavior of a network of extensible molecules
over the entire range of possible strains.
The Arruda-Boyce and Gent model can be invoked by using the ARRUDBOYCE and GENT model
definition options, respectively.
The volumetric part of the strain energy is for all the rubber models in Marc:
1 2
9K  3 
---
W volumetric = -------  J – 1 (7-130)
2  
when K is the bulk modulus.

It can be noted that the particular form of volumetric strain energy is chosen such that:
1. The constraint condition is satisfied for incompressible deformations only; for example:

 > 0 if I 3 > 0

f ( I 3 )  = 0 if I 3 = 1 (7-131)

 < 0 if I 3 < 0

2. The constraint condition does not contribute to the dilatational stiffness.
This yields the constraint function as:
1
 --6- 
f ( I 3 ) = 3  I – 1 (7-132)
 3

upon substitution of Equation (7-134) in Equation (7-130) and taking the first variation of the
variational principle, you obtain the pressure variable as:
1
 --3- 
p = 3K  J – 1 (7-133)
 
The equation has a physical significance in that for small deformations, the pressure is linearly
related to the volumetric strains by the bulk modulus K .
The discontinuous or continuous damage models discussed in the models section on damage can be
included with the generalized Mooney-Rivlin, Ogden, Arruda-Boyce, and Gent models to simulate
Mullins effect or fatigue of elastomers when using the updated Lagrangian approach. In the total
Lagrangian framework however, this is available for the Ogden model only.
The rubber foam model which is based on Ogden formulation has a strain energy form as follows:
N N
μ α α α μn β
------  1 – J 
n
W =  -----n- ( λ 1 n + λ 2 n + λ 3 n – 3 ) +  (7-134)
αn βn  
n = 1 n = 1
where μ n , α n , β n are material constants. The model reduces to incompressible Ogden model when β n
equals zero.
You can define any other invariant based models through the use of the UENERG user subroutine when
using the MOONEY option in model definition for elasticity in total or Updated Lagrangian framework.
A more general and easy-to-use the UELASTOMER user subroutine (uelastomer.f) can be used to
define a general strain energy function in the Updated Lagrangian framework.
Once the strain energy function is defined, the stresses and material tangent can be evaluated for the total
and Updated Lagrangian formulations as:
A. Total Lagrangian Formulation:
The stress measure in the total Lagrangian formulation is the symmetric second
Piola-Kirchhoff stress S i j , calculated as:
∂W ∂W
S i j = ---------- = 2 ---------- (7-135)
∂E i j ∂C i j
The material elasticity tangent is:

2 2
∂ W ∂ W
D i j k l = ---------------------- = 4 ----------------------- (7-136)
∂E i j ∂E k l ∂C i j ∂C k l
B. Updated Lagrangian Formulation:

The stress measure in the Updated Lagrangian formulation is the Cauchy or true stress
calculated as:
2 ∂W
σ i j = --- ---------- b k j (7-137)
J ∂b i k
The spatial elasticity tangent is:

2
4 ∂ W
L i j k l = --- b i m ------------------------ b n l (7-138)
J ∂b m j ∂b k n
The material constants for the Mooney-Rivlin form can be obtained from experimental data. The
Mooney-Rivlin form of the strain energy density function is
mooney-rivlin
W deviatoric = C 10 ( I 1 – 3 ) + C 01 ( I 2 – 3 ) (7-139)
CHAPTER 7 437
Material Library
For the Mooney-Rivlin model, the force and deformation for a uniaxial test specimen can be related as
P = 2A 0  1 – ------ ( λ 1 C 10 + C 01 )
1
(7-140)
 3
λ1
which can be written in the form:
P C 01
------------------------------------ = C 10 + --------
- (7-141)
λ1
2A 0  λ 1 – ------
1
 2
λ1
where P is the force of the specimen, A 0 is the original area of the specimen, and λ 1 is the uniaxial
stretch ratio. This equation provides a simple way to determine the Mooney-Rivlin constants. The
Mooney-Rivlin constitutive equation is applicable if the plot of P ⁄ 2A 0  λ 1 – ------ versus ----- should
1 1
 2 λ1
λ1
1
yield a straight line of slope C 01 and intercept ( C 01 + C 10 ) on the vertical axis ----- = 1 as shown in
λ1
Figure 7-28.
0.4
G
0.3
σ/2(λ-1/λ2) (N/mn2)
E
D
C
A
0.2
B
0.1
0.5 0.6 0.7 0.8 0.9 1.0
1/λ
Figure 7-28 Plots for Various Rubbers in Simple Extension for Mooney-Rivlin Model
If only the Young’s modulus E is supplied, and full uniaxial data are not available then
C 01 ≅ 0.25C 10 (7-142)
is a reasonable assumption. The constants then follow from the relation:
6 ( C 10 + C 01 ) ≅ E (7-143)
The material coefficients for the models can be obtained from Mentat. This allows you to select which
model is most appropriate for your data.
Updated Lagrange Formulation for Nonlinear Elasticity

The total Lagrange nonlinear elasticity models in Marc have been augmented with a formulation in an
Updated Lagrange framework. Hence, Rezoning can be used for elastomeric materials based upon the
current configuration. This is specially useful in large deformation analysis since typically excessive
element distortion in elastomeric materials can lead to premature termination of analysis. The new
formulation accommodates the generalized Mooney-Rivlin and Ogden material models preserving the
same format and strain energy functions as the total Lagrange formulation. In addition, the Arruda-Boyce
and the Gent models are only available in the updated Lagrange framework.
The updated Lagrangian rubber elasticity capability can be used in conjunction with both continuous as
well as discontinuous damage models. Thermal, as well as viscoelastic, effects can be modeled with the
current formulation. Using the table driven input, all of the material properties can reference tables,
which facilitates the use of temperature dependent material properties. The singularity ratio of the system
is inversely proportional to the order of bulk modulus of the material due to the condensation procedure.
A consistent linearization has been carried out to obtain the tangent modulus. The singularity for the case
of two- or three-equal stretch ratios is analytically removed by application of L’Hospital’s rule. The
current framework with an exact implementation of the finite strain kinematics along with the split of
strain energy to handle compressible and nearly incompressible response is eminently suitable for
implementation of any nonlinear elastic as well as inelastic material models. In fact, the finite
e θ p
deformation plasticity model based on the multiplicative decomposition, F = F F F is implemented
in the same framework.
To simulate elastomeric materials, incompressible element(s) are used for plane strain, axisymmetric,
and three-dimensional problems for elasticity in total Lagrangian framework. These elements can be used
with each other or in combination with other elements in the library. For plane stress, beam, plate or shell
analysis, conventional elements can be used. For updated Lagrangian elasticity, both conventional
elements (as well as Herrmann elements) can be used for plane strain, axisymmetric, and three-
dimensional problems.
Time-independent Inelastic Behavior

In uniaxial tension tests of most metals (and many other materials), the following phenomena can be
observed. If the stress in the specimen is below the yield stress of the material, the material behaves
elastically and the stress in the specimen is proportional to the strain. If the stress in the specimen is
greater than the yield stress, the material no longer exhibits elastic behavior, and the stress-strain
relationship becomes nonlinear. Figure 7-29 shows a typical uniaxial stress-strain curve. Both the elastic
and inelastic regions are indicated.
CHAPTER 7 439
Material Library
Stress Inelastic
Region
Yield
Stress
Strain
Elastic Region
Note: Stress and strain are total quantities.
Figure 7-29 Typical Uniaxial Stress-Strain Curve (Uniaxial Test)
Within the elastic region, the stress-strain relationship is unique. As illustrated in Figure 7-30, if the stress
in the specimen is increased (loading) from zero (point 0) to σ 1 (point 1), and then decreased
(unloading) to zero, the strain in the specimen is also increased from zero to ε 1 , and then returned to
zero. The elastic strain is completely recovered upon the release of stress in the specimen.
The loading-unloading situation in the inelastic region is different from the elastic behavior. If the
specimen is loaded beyond yield to point 2, where the stress in the specimen is σ 2 and the total strain is
e
ε 2 , upon release of the stress in the specimen the elastic strain, ε , is completely recovered. However,
2
p
the inelastic (plastic) strain, ε 2 , remains in the specimen. Figure 7-30 illustrates this relationship.
Total Strain = Strain and Elastic Strain
Stress
σ3 3
σ2 p
2
Δε 3
Yield Stress σ y
σ1 1
0 Strain
ε1 ε2 ε3
p e
ε
p
ε 2e ε2 = ε2 + ε2
2 p e
ε3 = ε3 + ε3
p e
ε ε
3 3
Figure 7-30 Schematic of Simple Loading - Unloading (Uniaxial Test)
Similarly, if the specimen is loaded to point 3 and then unloaded to zero stress state, the plastic strain
p p p
ε 3 remains in the specimen. It is obvious that ε is not equal to ε . We can conclude that in the
2 3
inelastic region:
• Plastic strain permanently remains in the specimen upon removal of stress.
• The amount of plastic strain remaining in the specimen is dependent upon the stress level at
which the unloading starts (path-dependent behavior).
The uniaxial stress-strain curve is usually plotted for total quantities (total stress versus total strain). The
total stress-strain curve shown in Figure 7-29 can be replotted as a total stress versus plastic strain curve,
as shown in Figure 7-31. The slope of the total stress versus plastic strain curve is defined as the
workhardening slope (H) of the material. The workhardening slope is a function of plastic strain.
Total Stress
σ θ
Plastic Strain
εp
H = tan θ (Workhardening Slope)
= dσ/dεp
Figure 7-31 Definition of Workhardening Slope (Uniaxial Test)
The stress-strain curve shown in Figure 7-29 is directly plotted from experimental data. It can be
simplified for the purpose of numerical modeling. A few simplifications are shown in Figure 7-32 and
are listed below:
1. Bilinear representation – constant workhardening slope
2. Elastic perfectly-plastic material – no workhardening
3. Perfectly-plastic material – no workhardening and no elastic response
4. Piecewise linear representation – multiple constant workhardening slopes
5. Strain-softening material – negative workhardening slope
In addition to elastic material constants (Young’s modulus and Poisson’s ratio), it is essential to include
yield stress and workhardening slopes when dealing with inelastic (plastic) material behavior. These
quantities can vary with parameters such as temperature and strain rate. Since the yield stress is generally
measured from uniaxial tests, and the stresses in real structures are usually multiaxial, the yield condition
of a multiaxial stress state must be considered. The conditions of subsequent yield (workhardening rules)
must also be studied.
CHAPTER 7 441
Material Library
σ σ
ε ε
(1) Bilinear Representation (2) Elastic-Perfectly Plastic
σ σ
ε ε
(3) Perfectly Plastic (4) Piecewise Linear Representation
ε
(5) Strain Softening
Figure 7-32 Simplified Stress-Strain Curves (Uniaxial Test)
Yield Conditions
The yield stress of a material is a measured stress level that separates the elastic and inelastic behavior
of the material. The magnitude of the yield stress is generally obtained from a uniaxial test. However,
the stresses in a structure are usually multiaxial. A measurement of yielding for the multiaxial state of
stress is called the yield condition. Depending on how the multiaxial state of stress is represented, there
can be many forms of yield conditions. For example, the yield condition can be dependent on all stress
components, on shear components only, or on hydrostatic stress. A number of yield conditions are
available in Marc, and are discussed in this section.
von Mises Yield Condition

Although many forms of yield conditions are available, the von Mises criterion is the most widely used.
The success of the von Mises criterion is due to the continuous nature of the function that defines this
criterion and its agreement with observed behavior for the commonly encountered ductile materials. The
von Mises criterion states that yield occurs when the effective (or equivalent) stress (σ) equals the yield
stress (σy) as measured in a uniaxial test. Figure 7-33 shows the von Mises yield surface in two-
dimensional and three-dimensional stress space.
For an isotropic material:
σ = [ ( σ1 – σ2 ) 2 + ( σ2 – σ3 ) 2 + ( σ3 – σ1 ) 2 ]1 ⁄ 2 ⁄ 2 (7-144)
where σ1, σ2, and σ3 are the principal Cauchy stresses.

d
σ2 σ3
Yield
Surface
Yield
Surface
Elastic
Region
σ1
d d
Elastic
Region
σ1 σ2
(a) Two-dimensional Stress Space (b) p-Plane
Figure 7-33 von Mises Yield Surface
σ can also be expressed in terms of nonprincipal Cauchy stresses.
( σ = [ ( σ x – σ y ) 2 + ( σ y – σ z ) 2 + ( σ z – σ x ) 2 + 6 ( τ x2y + τ y2z + τ z2x ) ] 1 ⁄ 2 ) ⁄ 2 (7-145)
The yield condition can also be expressed in terms of the deviatoric stresses as:
3 d d
σ = --- σ i j σ i j (7-146)
2
d
where σ i j is the deviatoric Cauchy stress expressed as
d 1
σ i j = σ i j – --- σ k k δ i j (7-147)
3
For isotropic material, the von Mises yield condition is the default condition in Marc. The initial yield
stress σ y is defined in the ISOTROPIC and ORTHOTROPIC options. A user-defined finite strain,
isotropic plasticity material model can be implemented through the UFINITE user subroutine. In this
case, the finite strain kinematics is taken care of in Marc. You have to do the small strain return mapping
only. See Marc Volume D: User Subroutines and Special Routines for more details.
Hill’s [1948] Yield Function

The anisotropic yield function (of Hill) and stress potential are assumed as
2
σ = [ a 1 ( σ y – σ z ) 2 + a 2 ( σ z – σ x ) 2 + a 3 ( σ x – σ y ) 2 + 3a 4 τ z x +
(7-148)
2 1⁄2
+ 3a 5 τ y2z + 3a 6 τ x y ] ⁄ 2
where σ is the equivalent tensile yield stress for isotropic behavior.

CHAPTER 7 443
Material Library
Ratios of actual to isotropic yield (in the preferred orientation) are defined in the array YRDIR for direct
tension yielding, and in YRSHR for yield in a shear (the ratio of actual shear yield to σ ⁄ 3 isotropic
shear yield). Then the a 1 above are defined by:
1 1 1
a 1 = --------------------------------------- + --------------------------------------- – --------------------------------------- (7-149)
YRDIR ( 2 )**2 YRDIR ( 3 )**2 YRDIR ( 1 )**2
1 1 1
a 2 = --------------------------------------- + --------------------------------------- – --------------------------------------- (7-150)
1 1 1
a 3 = --------------------------------------- + -------------------------------------- – -------------------------------------- (7-151)
2
a 4 = ---------------------------------------- (7-152)
YRSHR ( 3 )**2
2
a 5 = --------------------------------------- (7-153)
YRSHR ( 2 )**2
2
a 6 = ---------------------------------------- (7-154)
YRSHR ( 1 )**2
For anisotropic material, use the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition
options to indicate the anisotropy. Use the ORTHOTROPIC option or the ANPLAS user subroutine for
the specification of anisotropic yield condition (constants a1 through a6, as defined above), and the
ORIENTATION model definition option or the ORIENT user subroutine, if necessary, to specify preferred
orientations.
Hill’s (1948) Yield Criterion has been extensively used in sheet metal forming, especially for steel. The
experimental data can be related to the Marc input for the Hill’s Yield Criterion for the shells or plane
stress case as given below.
N (Thickness Direction) R (Rolling Direction)
T (Transverse Direction)
Figure 7-34 Axes of Anisotropy

The sample of tensile coupon cut from a sheet in the three directions, θ = 0 (rolling), 45° and 90°
(transverse) is tested to obtain σ = σ 0 , σ 45 , and σ 90 , respectively. Similarly, the anisotropy
parameter defined as:
ε width
r = -------------------- (7-155)
ε thickness
is obtained for the 0°, 45°, and 90° directions. The yield stress in the third (thickness) direction can be
written as:
r 90 ( 1 + r 0 ) r 0 ( 1 + r 90 )
- = σ 90 --------------------------
σ N = σ 0 -------------------------- - (7-156)
r 0 + r 90 r 0 + r 90
The direct stress coefficients are now:

σ0
YRDIR ( 1 ) = -------- (7-157)
σa v
σ 90
YRDIR ( 2 ) = -------- (7-158)
σa v
σN
YRDIR ( 3 ) = -------- (7-159)
σa v
where σ a v is the initial yield stress on the stress-strain curve used. If the stress-strain curve is averaged
σ 0 + 2σ 45 + σ 90
from all directions, σ a v is defined as σ a v = ----------------------------------------- in orthotropic plasticity. Similarly, the
4
shear coefficients can be evaluated at:
3
YRSHR ( 1 ) = YRDIR ( 3 ) -------------------- (7-160)
2r 45 + 1
YRSHR ( 2 ) = YRSHR ( 3 ) = 1.0 (7-161)

It is noted that the transverse direction along the thickness is usually considered isotropic which is
reasonable on physical grounds. Also, notice that for complete isotropy, σ 0 = σ 45 = σ 90 and
r 0 = r 45 = r 90 = 1 , which yields the von Mises yield criterion.
Barlat’s (1991) Yield Function

Barlat et al. [Ref. 6] proposed a general criterion for planar anisotropy that is particularly suitable for
aluminum alloy sheets. This criterion has been shown to be consistent with polycrystal-based yield
surfaces which often exhibit small radii of curvature near uniaxial and balanced biaxial tension stress
CHAPTER 7 445
Material Library
states. An advantage of this criterion is that its formulation is relatively simple as compared with the
formulation for polycrystalline modeling and, therefore, it can be easily incorporated into finite element
codes for the analysis of metal forming problems.
For three dimensional deformation, the yield function f is defined as (Barlat et al. [Ref. 6])
m m m m
f = S1 – S2 + S2 – S3 + S3 – S1 = 2σ (7-162)
where S i = 1 , 2, 3 are principal values of a symmetric matrix S i j defined with respect to the components
of the Cauchy stress as
C 3 ( σ xx – σ yy ) – C 2 ( σ zz – σ xx )
-------------------------------------------------------------------------------
- C6 σ x y C 5 σ zx
3
S = C 1 ( σ y y – σ zz ) – C 3 ( σ x x – σ y y ) (7-163)
C 6 σ xy -------------------------------------------------------------------------------- C 4 σ zy
3
C 2 ( σ z z – σ x x ) – C 2 ( σ y y – σ zz )
C 5 σ zx C4 σz y ------------------------------------------------------------------------------
-
3
In Equation (7-163), the symmetry axes (x, y, z), which represent the mutually orthogonal axes of
anisotropy, are aligned with the initial rolling, transverse, and normal directions of the sheet. During
deformation, the anisotropic yield surface of each material element rotates so that the symmetry axes are
all in different directions during deformation. Therefore, it is necessary to trace the rotation of the yield
surface during deformation in order to calculate the plastic strain increment properly. The rotation of the
anisotropy axes is carried out based on the polar decomposition method.
The material coefficients, C i = 1∼6 in Equation (7-163) represent anisotropic properties. When
Ci = 1∼6 = 1 , the material is isotropic and Equation (7-162) reduces to the Tresca yield condition for
m = 1 or ∞ , and the von Mises yield criterion for m = 2 or 4. The exponent “ m ” is mainly
associated with the crystal structure of the material. A higher “ m ” value has the effect of decreasing the
radius of curvature of rounded vertices near the uniaxial and balanced biaxial tension ranges of the yield
surface, in agreement with polycrystal models. Values of m = 8 for FCC materials (like aluminum)
and m = 6 for BCC materials (like steel) are recommended. The yield surface has been proven to be
convex for m ≥ 1 . Figure 7-35 shows the yield surfaces obtained from von-Mises, Hill and Barlat yield
functions for Aluminum 2008-T4 alloy.
1.5
σyy
Mises
σ
1.0
Hill
(1948)
0.5
Barlat’s 6D
σxx
σ
0.0
-0.5
-1.0
-1.0 -0.5 0.0 0.5 1.0 1.5
Figure 7-35 Comparison of Yield Surfaces Obtained from von Mises, Hill and Barlat Yield Functions
p
Utilizing the normality rule, the associated plastic strain increment Δε i j is obtained from the yield
function f as
∂f
Δε ipj = λ ---------- (7-164)
∂σ i j
∂f
where λ is a scalar function. The calculation of ---------- in Equation (7-164) is lengthy but straightforward.
∂σ i j
p ∂f
The stress integration to obtain Δε i j along with the calculation of ------------- are shown in the works by
∂σ α β
Chung and Shah, [Ref. 7], and Yoon et al,.[Ref. 8].
In order to obtain four unknown independent coefficients ( C 1, C 2, C 3, C 6 ) with the assumption of
C 4 = C 5 = 1 (isotropic properties for transverse directions), it is necessary to use four stresses from
the experimental data ( σ 0, σ 45, σ 90, σ b ) , where σ 0, σ 45, σ 90 are the tensile yield stresses at 0o, 45o,
and 90o from the rolling direction, and σ b is the balanced biaxial yield stress measured from bulge test.
A detailed procedure to calculate the coefficients of Barlat’s yield function are summarized in the work
of Yoon at el., [Ref. 9]. In Marc, Barlat’s coefficients are automatically calculated from user inputs for
σ 0, σ 45, σ 90, σ b in Marc Mentat. If biaxial data is not available, generally Marc assumes σ b = σ 0 .
Barlat’s yield function can be accessed from ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model
definition options and can be used in conjunction with the ORIENTATION option.
CHAPTER 7 447
Material Library
Mohr-Coulomb Material (Hydrostatic Stress Dependence)

Marc includes options for elastic-plastic behavior based on a yield surface that exhibits hydrostatic stress
dependence. Such behavior is observed in a wide class of soil and rock-like materials. These materials
are generally classified as Mohr-Coulomb materials (generalized von Mises materials). Ice is also
thought to be a Mohr-Coulomb material. The generalized Mohr-Coulomb model developed by Drucker
and Prager is implemented in Marc. There are two types of Mohr-Coulomb materials: linear and
parabolic. Each is discussed on the following pages.
Linear Mohr-Coulomb Material

The deviatoric yield function, as shown in Figure 7-36, is assumed to be a linear function of the
hydrostatic stress.
σ
f = αJ 1 + J 21 ⁄ 2 – ------- = 0 (7-165)
3
where (7-166)
J1 = σi i (7-167)
1 d d
J 2 = --- σ i j σ i j (7-168)
2
τ
Yield Envelope
R
c φ
σ
σx + σy
-------------------
2
Figure 7-36 Yield Envelope of Plane Strain (Linear Mohr-Coulomb Material)
Analysis of linear Mohr-Coulomb material based on the constitutive description above is available in
Marc through the ISOTROPIC model definition option. Through the ISOTROPIC option, the values of
σ and α are entered. Note that, throughout the program, the convention that the tensile direct stress is
positive is maintained, contrary to its use in many soil mechanics texts.
The constants α and σ can be related to c and φ by
σ 3α
c = ---------------------------------------------- ; ---------------------------------- = sin φ (7-169)
1⁄2 ( 1 – 3α 2 ) 1 ⁄ 2
[ 3 ( 1 – 12α 2 ) ]
where c is the cohesion and φ is the angle of friction.
Parabolic Mohr-Coulomb Material

The hydrostatic dependence is generalized to give a yield envelope which is parabolic in the case of plane
strain (see Figure 7-37).
f = ( 3J 2 + 3βσJ 1 ) 1 ⁄ 2 – σ = 0 (7-170)
The parabolic yield surface is obtained in Marc through the ISOTROPIC model definition option. Enter
the values σ and β through the ISOTROPIC model definition option.
2 α
2 α
= 3  c – ------ β = ---------------------------------------------
2
σ (7-171)
 3 ( 3 ( 3c 2 – α 2 ) ) 1 ⁄ 2
where c is the cohesion.

τ
R c
σ
σx + σy c2
------------------- -----
2 α
Figure 7-37 Resultant Yield Condition of Plane Strain (Parabolic Mohr-Coulomb Material)
Buyukozturk Criterion (Hydrostatic Stress Dependence)

This yield criterion [Ref. 4], which originally has been proposed as a failure criterion, has the
general form:
2 2
f = β 3σJ 1 + γJ 1 + 3J 2 – σ (7-172)
CHAPTER 7 449
Material Library
Through the ISOTROPIC model definition option, the user has to define σ and the factor β , where γ has
a fixed value of 0.2 . The Buyukozturk criterion reduces to the parabolic Mohr-Coulomb criterion if
γ = 0.
Powder Material
Some materials, during certain stages of manufacturing are granular in nature. In particular, powder
metals are often used in certain forging operations and during hot isostatic pressing (HIP). These material
properties are functions of both the temperature and the densification. It should be noted that the soil
model discussed in this chapter also exhibits some of these characteristics.
There are two models for representing granular materials such as powders. The Shima-Oyane model and
the exponential cap model may be used to represent these types of materials. Both of these models are
dependent upon the hydrostatic stress and the relative density. It is felt that the newer exponential Cap
model better represents the physical phenomena as shown in the figure below.
Figure 1-38 Comparison with Experiments [Ref. 1]

In the Shima-Oyane model incorporated into Marc, a unified viscoplastic approach is used. The yield
function is
p2 1 ⁄ 2
F = ---  --- σ d σ d + ------
1 3
– σy (7-173)
γ 2  β 2
where σ y is the uniaxial yield stress, σ d is the deviatoric stress tensor, and p is the hydrostatic pressure.
γ , β are material parameters.
σ y can be a function of temperature and relative density, γ , β are functions only of relative density.
Typically, we allow:
q q
γ = ( q1 + q2 ρ 3) 4
b
(7-174)
b
β = ( b1 + b2 ρ 3) 4
where ρ is the relative density.
As the powder becomes more dense, ρ approaches 1 and the classical von Mises model is recovered.
It should be noted that the elastic properties are also functions of relative density. In particular, as the
material becomes fully dense, the Poisson’s ratio approaches 0.5.
As most processes involving powder materials are both pressure and thermally driven, it can be necessary
to perform a coupled analysis. Marc also allows you to specify density effects for the thermal properties,
conductivity and specific heat. The basic input data is entered through the POWDER option. In addition
to the TEMPERATURE EFFECTS option, there is a DENSITY EFFECTS option. The initial relative
density is entered through the RELATIVE DENSITY option.
------------------
2 2 2
β γ σy
2
p
σ d :σ d
------------------
2 2 2
--- γ σ y
3
Figure 1-39 Yield Function of Shima Model
Exponential Cap Powder Model

The exponential cap powder model implemented into Marc is based upon the Sandia Geomodel [Ref. 2]
and [Ref. 3], a generalized plasticity model, wherein the combined effect of the two underlying
mechanisms, porosity and microcracks, is represented through a unified yield function given as,
ξ
f ( σ ,α ,κ ) = J 2 Γ 2 – [ F s ( I 1 ) – N ] 2 F c ( I 1 ,κ )
CHAPTER 7 451
Material Library
where
σ is the stress tensor
α is the deviatoric backstress tensor
κ is a internal state variable determining the position of the cap

ξ is the second invariant of the shifted stress tensor ξ = dev ( σ ) – α
J2
N is the maximum allowable translation of the yield surface when kinematic hardening is enabled
Γ is the Gudehus Lode angle function given as
J3 3 3 ⁄ 2 1  J 3 3 3 ⁄ 2
Γ ( J 2 J 3 ) = --- 1 – -----  ----- + ----  1 + -----  -----
1

2 2  J 2 ψ 2  J 2 
where J 2 ,J 3 are the second and third invariants of the stress tensor respectively and ψ is the
strength ratio.
Shear Failure
Surface Fs
Fs Fc
J 2 = ----------------
Γ
Strain hardening
cap Fc
Direction of
plastic strain
on cap
I1 X X0 X0 κ κ0
Cap Nose
Point Cap Branch
Point
Shear Failure Surface

The shear failure surface F s ( I 1 ) is represented as,
A I
Fs ( I1 ) = A1 – A3 e 2 1 – A4 I1
where, A 1 ,A 2 ,A 3 ,A 4 are all non negative material parameters.
fs
A1 – A3
I1
Cap Function
The cap function F c ( I 1 ,κ ) can be written as
F c ( I 1 ,κ ) = 1 if I 1 > κ
I1 – κ 2
= 1 –  -------------- otherwise
 X – κ
where, X ( κ ) = κ – RF s ( κ ) and R is the cap eccentricity parameter.
fc
1.0
I1 X κ
CHAPTER 7 453
Material Library
The volumetric plastic strain ε vp is determined from the expression
D ( X – X ) – D ( X – X )2
ε vp = W e 1 0 2 0 –1
where W , D 1 , D 2 , and X 0 are all material constants.
Evolution equations
The flow rule,
· ∂f ·
ε p = ------- λ
∂σ
·
where λ is the consistency parameter.
Evolution of κ is governed by,

· ·
κ = hκ λ
∂f I1 – κ
3 ------- --------------
∂σ R
where h κ = min -----------------
- , ------------------
∂ε vp ∂X ∂ε vp ∂X
--------- ------- --------- -------
∂X ∂κ ∂X ∂κ
When kinematic hardening is enabled, the backstress α evolves as

· ·
α = Hαλ
 J 2α ∂f
where, H α = H  1 – ---------- dev  ------
 N  ∂ξ
Kinematic Hardening:
Kinematic hardening can be introduced by use of a shifted stress tensor ξ = dev ( σ ) – α
The backstress tensor α is initialized to zero. Upon onset of yielding, the backstress α evolves in
proportion to the deviatoric part of the plastic strain rate.
In order to turn on kinematic hardening, the user needs to specify N , the maximum allowable translation
of the yield surface, and a material parameter H , which determines the speed at which the yield surface
approaches the limit surface. To suppress kinematic hardening, set both H and N to zero.
Non-associative flow rule

The exponential cap model supports non-associatively by allowing the user to specify a flow function φ
which has the same form as the yield function f . The flow potential can be made to differ from the yield
PF PF PF PF
function by assigning values to A 2 , A 4 , R , and ψ that differ from their counterpart parameters
( A 2 , A 4 , R , and ψ ) used to define the yield surface. To set an associative flow rule, set all potential
PF PF PF PF
flow coefficients ( A 2 , A 4 , R , and ψ ) to zero.
Simplification to standard models

The exponential cap powder model can be simplified to standard models such as von Mises, Linear
Mohr-Coulomb etc. by setting some of the parameters to zero. The following simplifications are
documented here:
1. von Mises
σy
The von Mises yield surface f ( J 2 ) = J 2 – ------- can be obtained by the following
3
simplifications:
• Set A 2 and A 4 to 0
σy
• Set A 1 – A 3 = -------
3
• Set κ 0 = – 1e30 (cap at infinity)
• Set R , W , D 1 ,and D 2 to 0
• Set ψ = 1
2. Linear Mohr-Coulomb
σy
The Linear Mohr-Coulomb yield surface f ( J 2 ) = αI 1 + J 2 – ------- can be obtained by the
3
following simplifications:
• Set A 2 = 0
• Set A 4 = α
σy
• Set A 1 – A 3 = -------
3
• Set κ 0 = – 1e30 (cap at infinity)
• Set R , W , D 1 ,and D 2 to 0
• Set ψ = 1
CHAPTER 7 455
Material Library
User Input Parameters
Symbol Description SI Units

E Young’s modulus, supports table input Pa
ν Poisson’s ratio, supports table input dimensionless
ρ Density, supports table input Kg/m3
Ct h Coefficient of thermal expansion, to be used for thermal analysis 1/K
A1 First constant term in the function defining the shear failure surface Pa
A1 > 0
A2 Second constant term in the function defining the shear failure surface 1/Pa
A2 ≥ 0
Set A 2 and A 4 to 0 for simplification to von Mises yield surface
A3 Third constant term in the function defining the shear failure surface Pa
A3 ≥ 0
σy
Set A 1 – A 3 = ------- to mimic the von Mises yield surface
3
f ( J2 ) = J2 – σy
A4 Fourth constant term in the function defining the shear failure surface dimensionless
A4 ≥ 0
PF Potential function parameter for setting non-associative flow rule. 1/Pa

A2
PF
A2 = A 2 for associative flow. Set to zero in input file for
associative flow.
PF Potential function parameter for setting non-associative flow rule. Set dimensionless
A4
PF
A4 ≠ A 2 for non-associative flow. Set to zero in input file for
associative flow.
H Kinematic hardening parameter which governs how quickly the yield dimensionless
surface evolves towards the limit surface. Set both H and N to zero to
suppress kinematic hardening.
N Kinematic hardening parameter which sets the maximum allowable Pa
translation of the yield surface. Set both H and N to zero to suppress
kinematic hardening.
N≥0

R Cap eccentricity parameter. It is the ratio of the width divided by the dimensionless
height of the elliptical cap function. This parameter affects the stress
level at which dilatation will occur in triaxial compression.
R>0
If not entered, R = 1 .
κ0 This is the initial value of I 1 at which the meridional yield profile Pa

branches away from the shear failure surface and morphs into the cap
function. κ 0 must be negative. Set κ 0 = – 1e30 (cap at infinity) to
simulate a von Mises response.
W Limit value of the absolute plastic volumetric strain. This parameter is dimensionless
approximately equal to the initial porosity of the material and can be
obtained from the hydrostatic crush data.
D1 One third of the slope of the porosity versus pressure crush curve at the 1/Pa
elastic limit.
D2 Extra fitting parameter for the hydrostatic crush curve data. Used only 1/Pa2
when the crush curve has an inflection point. Else set to zero.
ψ The triaxial extension to compression strength ratio (TXE/TXC) dimensionless
7 9
Convexity of the yield surface requires --- ≤ ψ ≤ --- .
9 7
If not entered, ψ = 1 .
PF Flow potential analog of the yield surface parameter R . Used only for dimensionless
R
non-associative flow.
PF Flow potential analog of the yield surface parameter S . Used only for dimensionless
ψ
non-associative flow.
Plotable output parameters

Cap Softening Flag Set this flag to 0 to disallow contraction of the cap. Set to 1 Flag
to allow the powder cap to soften.
κ The value of I 1 at which the meridional yield profile Pa
branches away from the shear failure surface and morphs
into the cap function. This variable is referred to as “Cap
Branch Point” in Mentat.
CHAPTER 7 457
Material Library
X The value of I 1 at which the cap surface intersects the x-axis. Pa

This variable is referred to as “Cap Nose Point” in Mentat.
ρv o l Current relative density based on total volumetric strain. dimensionless
ρ vpo l Stress free relative density based on the plastic volumetric

dimensionless
strain.
Material Parameters
The fully generalized exponential cap powder model requires the user to specify 16 different material
parameters. However, simplifications of the exponential model can be run using as few as two material
parameters. Material parameters for the exponential cap powder model can be inferred by curve fitting
the data obtained through a series of experiments. The experiments needed and the method of curve
fitting is detailed below.
Crush Curve Parameters

Step 1: Obtain hydrostatic pressure vs. total volumetric strain data by conducting experiments run to
the point of total pore collapse.
ε vtotal
W
Step 2: Use the elastic unloading curve to determine a shift distance that must be applied at any
pressure to remove the elastic part of the strain. This will give you a plot of pressure vs. plastic
volumetric strain.
X
p=
3
ε vp
W
Step 3: Use a curve fitting program to fit the above data to the following curve
p D1 ( X – X0 ) – D2 ( X – X0 ) 2
εv = W e –1
where
X 0 = κ 0 – Rf s ( κ 0 ) and R are the cap eccentricity parameter
W is approximately equal to the initial porosity of the material.
Shear Failure Parameters

Step 1: Conduct load-to-failure triaxial compression tests on the material and record all peak stress
states. Each individual experiment has precisely one stress state at which the second stress
variant J 2 achieves a peak value. Characterizing the shear failure surface requires numerous
such experiments.
Step 2: For every available load-to-failure experiment, find the stress state at which J 2 is larger than
peak
any of the other stress states in that experiment. Construct a table of data ( I 1 , J 2 ) where
peak
I1 is the value of I 1 at the stress state at which J 2 is at its peak value.
Step 3: Fit the data to the exponential shear failure function using a curve fitting program
A I
fs ( I1 ) = A 1 – A3 e 2 1 – A4 I1
CHAPTER 7 459
Material Library
More details on obtaining material parameters for the exponential cap powder model can be found in the
following reference:
A.F. Fossum and R.M. Brannon, “The Sandia GEOMODEL: Theory and User's Guide,” Technical
Report SAND2004-3226, Sandia Nat'l Laboratories, Albuquerque, N.M., 2004.
Obtaining Crush Curve and Shear Failure Parameters by

Curve Fitting in Marc
Marc now supports curve fitting for the crush curve and shear failure parameters through the existing
Mentat interface. The user can input the experimental data via tables and Marc carries out the curve
fitting and obtains both the shear failure and the crush curve parameters. The following steps are used
for curve fitting in Marc.
Crush Curve Parameters: ( κ 0 ,W ,D 1 ,D 2 )
1. Set the value of W to – 1 in Mentat to indicate that you will be using curve fitting for the crush
curve parameters. The rest of the parameters can be set to 0.
2. Assign a table ID for W and include a table of pressure vs. the plastic volumetric strain (X vs.
p
ε v ). Data should be sorted by X. Sample set of data and the corresponding plot are shown below:
Data p
Point X = 3*p εv
1 -6.00E+09 -0.079988058
2 -5.00E+09 -0.079948061
3 -4.00E+09 -0.079774105
4 -3.00E+09 -0.079017528
5 -2.00E+09 -0.075726998
6 -1.50E+09 -0.071088731
7 -1.00E+09 -0.061415709
8 -8.00E+08 -0.055063898
9 -7.00E+08 -0.051115167
10 -6.00E+08 -0.04654114
11 -5.00E+08 -0.041242797
12 -4.00E+08 -0.03510544
13 -3.00E+08, -0.027996208
14 -2.00E+08 -0.019761202
15 -7.00E+07 -0.007076085
16 -7.00E+06 0
3. Marc internally uses nonlinear regression analysis to fit the data and to obtain the crush curve
parameters ( κ 0 ,W ,D 1 ,D 2 ).
Shear Failure Parameters: ( A 1 ,A 2 ,A 3 ,A 4 )

1. Set the value of A 1 to – 1 in Mentat to indicate that you will be using curve fitting for the shear
failure parameters. The rest of the shear failure parameters can be set to 0.
2. Assign a table ID for A 1 and include a table of ( I 1 vs. J 2 ). Data should be sorted by I 1 . Sample
set of data and the corresponding plot are shown:
Data
Point I1 J2
1 -600000000 631520775
2 -440000000 574163050
3 -400000000 553409562
4 -320000000 500886644
5 -280000000 467829871
6 -220000000 407501319
7 -190000000 371605878
8 -160000000 331244305
9 -130000000 285856240
10 -100000000 234811012
11 -80000000 197291022
12 -60000000 156711243
13 -40000000 112820812
14 -20000000 65348293
15 0 14000000
16 10000000 -13225871
3. Marc internally uses nonlinear regression analysis to fit the data and to obtain the shear failure
parameters ( A 1 ,A 2 ,A 3 ,A 4 )
Limitations:
The following limitations should be noted.
1. The implementation treats the material as initially isotropic. Kinematic hardening is the only way
to simulate deformation induced anisotropy.
2. The elasticity model is hypoelastic rather than hyperelastic.
3. While the exponential shear failure curve is quite general, only an elliptic cap function is available
for modeling shear effects on pore collapse and other plastic volume reducing mechanisms.
CHAPTER 7 461
Material Library
4. The model is not extensively tested for tensile applications.

5. The triaxial extension/compression ratio is presumed constant. Neither does it change with
pressure, nor does it evolve with time.
6. Temperature dependent plasticity is not incorporated in the material model.
7. While the elastic parameters (Young’s modulus, Poisson’s ratio) can be made to vary with
temperature and relative density using tables, the same is not supported for the other material
parameters.
8. Curve fitting programs used to fit experimental data to the required material parameters are not
provided.
Tips and Tricks:

1. Getting the complete set of 16 material parameters for the exponential cap powder model is a
daunting task which requires extensive experimentation and curve fitting. But do remember that
in most cases one can make do with much lesser material parameters by running simplifications
of the exponential cap model. Also the Sandia report, referenced in the material parameters
section above, has a handy set of material data for several geomaterials.
2. Make sure that the material properties satisfy the following constraints:
A 1 – A 3 – ( N ≥ 0 ) (Unloaded virgin material must be below yield)
A2 A3 + A4 ≥ 0 (Non-negative slope at low pressures)

A2 ≥ 0 (Positive Lode radius)
A3 ≥ 0 (Convexity condition)
A4 ≥ 0 (Non-negative slope at low pressures)
κ0 ≤ 0 (Initial cap position should be negative)
R>0 (Positive cap radius)
7⁄9≤ψ≤9⁄7
On violation of any of these conditions, an error message is displayed in the output file. Check
output file to see which condition has been violated.
3. If the cap radius R and the strength ratio ψ are not specified, then they are each set to a default
value of 1.
PF PF
4. The exponential cap model assumes a associative flow rule as default when A 2 and A 4 are
PF PF
not specified (or specified to be 0) by internally setting A 2 = A 2 and A 4 = A 4 . To set a
PF PF
non-associative flow rule set values of A 2 and A 4 different than values of A 2 and A 4 .
5. Many times, powders are subjected to hydrostatic loading or close to hydrostatic loading. In such
cases, J 2 is close to zero and the yield stress is return mapped back to the nose of the cap. When
running problems where such hydrostatic loading is likely, better convergence can be achieved by
using only the contribution of the deviatoric stress to the initial stress stiffness matrix. This option
can be found under LOADCASE > STATIC > SOLUTION CONTROL.
6. The exponential cap model has not been extensively tested under tension as most powder
materials are subjected to compression in HIP/CIP processes. The shear failure function and its
derivatives become very large as the exponent A 2 ⋅ I 1 increases in the positive regime. We have
internally set the limiting value of A 2 ⋅ I 1 to be 10 and issue the warning of “large hydrostatic
tension in element” whenever this condition is violated. Small amounts of tension are acceptable
as A 2 is typically small. Only when one gets into large amounts of tension and I 1 becomes really
large the above condition will be violated which may make the return mapping numerically
unstable. In such cases a warning is printed in the output files and the simulation continues for as
long as possible.
Oak Ridge National Laboratory Options

In Marc, the ORNL options are based on the definitions of ORNL-TM-3602 [Ref. 10] for stainless steels
and ORNL recommendations [Ref. 16] for 2 1/4 Cr-1 Mo steel.
The initial yield stress should be used for the initial inelastic loading calculations for both the stainless
steels and 2 1/4 Cr-1 Mo steel. The 10th-cycle yield stress should be used for the hardened material. The
100th-cycle yield stress must be used in the following circumstances:
1. To accommodate cyclic softening of 2 1/4 Cr-1 Mo steel after many load cycles
2. After a long period of high temperature exposure
3. After the occurrence of creep strain
To enter initial and 10th-cycle yield stresses, use the ISOTROPIC or ORTHOTROPIC model
definition option.
Effects of Temperature and Strain Rates on Yield Stress - not using table driven input
Marc allows you to input a temperature-dependent yield stress. To enter the yield stress at a reference
temperature, use the ISOTROPIC or ORTHOTROPIC model definition option. To enter variations of
yield stress with temperatures, use the TEMPERATURE EFFECTS and ORTHO TEMP model
definition options. Repeat the TEMPERATURE EFFECTS and ORTHO TEMP model definition
options for each material, as necessary. The effect of temperatures on yielding is discussed further in
Constitutive Relations.
CHAPTER 7 463
Material Library
Marc allows you to enter a strain rate dependent yield stress, for use in dynamic and flow (for example,
extrusion) problems. To use the strain rate dependent yield stress in static analysis, enter a fictitious time
using the TIME STEP option. The zero-strain-rate yield stress is given on the ISOTROPIC or
ORTHOTROPIC options. Repeat the STRAIN RATE model definition option for each different material
where strain rate data are necessary. Refer to Constitutive Relations for more information on the strain-
rate effect on yielding.
Work or Strain Hardening

In a uniaxial test, the workhardening slope is defined as the slope of the stress-plastic strain curve. The
workhardening slope relates the incremental stress to incremental plastic strain in the inelastic region and
dictates the conditions of subsequent yielding. The yield stress and the workhardening data must be
compatible with the procedure used in the analysis. For small strain analyses, the engineering stress and
engineering strain are appropriate. If only the LARGE DISP parameter is used, the yield stress should be
entered as the second Piola-Kirchhoff stress, and the workhard data be given with respect to plastic
Green-Lagrange strains. If the LARGE STRAIN parameter is used, the yield stress must be defined as a
true or Cauchy stress, and the workhardening data with respect to logarithmic plastic strains.
Work hardening can be specified in Marc in the different ways listed below.
1. Table driven input
2. WORK HARD option
3. Chaboche
4. Power Law
5. Rate Power Law
6. Johnson-Cook
7. Kumar
8. User subroutine (WKSLP)
The first two methods enables the user to define the yield stress as a function of independent variables
(like equivalent plastic strain, temperature, and strain rate) through user-specified data. Apart from the
above methods, Marc provides commonly used hardening laws available in the literature like Chaboche,
Power law, Rate Power Law, Johnson-Cook, and Kumar models. The user can also customize the
hardening law through the WKSLP user subroutine. Brief explanations of all the above mentioned classes
of models are given below.
Flow Stress Definition - using table driven input

When using the table driven input format, the yield (flow) stress dependence on the temperature,
equivalent plastic strain, and the strain rate is defined simultaneously through a single table. As this table
may be either a piecewise linear description or a mathematical equation based upon these independent
variables, it allows a very general definition. The value of the yield stress is obtained by evaluating the
table, including the reference value given on the ISOTROPIC or ORTHOTROPIC option. The strain
hardening slope is obtained when necessary by numerically differentiating the values given.
7 Work Hardening Definition - not using table input

Material When using the nontable input format, the yield (flow) stress dependence on the equivalent plastic strain,
Library temperature, and strain rate are defined through the WORK HARD, TEMPERATURE EFFECTS, and
STRAIN RATE options, respectively. The latter two options were discussed in the previous section. The
WORK HARD option is defined in more detail in this section.
There are two methods to enter this information, using the WORK HARD option. In the first method, you
must enter workhardening slopes for uniaxial stress data as a change in stress per unit of plastic strain
(see Figure 7-40) and the plastic strain at which these slopes become effective (breakpoint).
Stress
Δσ3
Δσ2
Δσ1
σ
E E E E
p p p
Strain
Δε1 Δε 2 Δε 3
Figure 7-40 Workhardening Slopes
Slope Breakpoint
Δσ 1
---------- 0.0
p
Δε1
Δσ 2 p
----------
p
Δε1
Δε 2
Δσ 3 p p
----------
p
Δε1 + Δε 2
Δε 3
and so on.
Note: The slopes of the workhardening curves should be based on a plot of the stress versus plastic
strain for a tensile test. The elastic strain components of the stress-strain curve should not be
included. The first breakpoint of the workhardening curve should be 0.0.
In the second method, you enter a of yield stress, plastic strain points. This option is flagged by adding
the word DATA to the work hard statement.
CHAPTER 7 465
Material Library
Note: The data points should be based on a plot of the stress versus plastic strain for a tensile test.
The elastic strain components should not be included. The first plastic strain should equal 0.0
and the first stress should agree with that given as the yield stress in the ISOTROPIC or
ORTHOTROPIC options.
Chaboche Model
The Chaboche model is described in detail in the Time-independent Cyclic Plasticity section.
Power Law Model

m ·n
σy = A ( ε0 + ε ) + Bε
where σ y is the yield stress, ε 0 is the strain corresponding to initial yield stress, ε is equivalent strain,
·
ε is rate of equivalent strain rate and A , B , m , and n are material parameters.
Rate Power Law Model

·n
σ y = Aε m ε + B
where the parameters are same quantities as that of Power law.
Johnson-Cook Model
n   ε·    T – Tr o o m m
 -----------------------------------
σ y = ( A + Bε p )  1 + C ln  ----
· -   1 –  T - 
  ε 0  melt – T room
 
· ·
where σ y is the yield stress, ε p is the equivalent plastic strain, ε is the current equivalent strain rate, ε 0
is strain rate of material characterization and A , B , C , m and n are material parameters. T , T r o o m ,
T m e l t are, respectively, the current, room, and melting temperatures of the material in absolute scale.
The following conditions should be noted for the Johnson-Cook model.

• T should be between T r o o m and T m e l t . If T < T r o o m , T is set to T r o o m . If T > T m e l t ,
T = T m e l t – 0.01 .
· · ·
• ε 0 > 0 and ε ≥ ε 0 . If either condition is violated, the middle term in the above expression is set
to 1.0.
Kumar Model
· 1⁄n
ε
 ---
σy = B0 * sin h – 1 - eQ ⁄ (n R T)
 A
·
where σ y is the yield stress, ε is equivalent strain rate, Q is the activation energy, T is temperature, and
A , B 0 , and n are material parameters.
Yada Law
Used for grain size predictions, is a generic variation of Kumar model. If the constants A and n in the
Kumar model are related to strain and the material constants B 1 , B 2 , B 3 , B 4 , B 5 , and B 6 as given
below, we get Yada Law.
WKSLP User Subroutine

The user-subroutine option can be employed to specific requirement of hardening law. The yield stress
can be a function of temperature, strain, strain-rate, etc. The user is required to return the work hardening
slope and, optionally, the yield stress at each integration point. If the latter is not defined, then the current
yield stress is calculated internally. WKSLP can be called multiple times within each cycle to capture the
dependence of the slope and the yield stress on temperature, strain, strain-rate, etc.
Work Hardening Rules

A number of workhardening rules (isotropic, kinematic, and combined) are available in Marc. A
description of these workhardening rules is given below.
Isotropic Hardening
The isotropic workhardening rule assumes that the center of the yield surface remains stationary in the
stress space, but that the size (radius) of the yield surface expands, due to workhardening. The change of
the von Mises yield surface is plotted in Figure 7-41b.
A review of the load path of a uniaxial test that involves both the loading and unloading of a specimen
will assist in describing the isotropic workhardening rule. The specimen is first loaded from stress free
(point 0) to initial yield at point 1, as shown in Figure 7-41a. It is then continuously loaded to point 2.
Then, unloading from 2 to 3 following the elastic slope E (Young’s modulus) and then elastic reloading
from 3 to 2 takes place. Finally, the specimen is plastically loaded again from 2 to 4 and elastically
unloaded from 4 to 5. Reverse plastic loading occurs between 5 and 6.
It is obvious that the stress at 1 is equal to the initial yield stress σ y and stresses at points 2 and 4 are
larger than σ y , due to workhardening. During unloading, the stress state can remain elastic (for example,
point 3), or it can reach a subsequent (reversed) yield point (for example, point 5). The isotropic
workhardening rule states that the reverse yield occurs at current stress level in the reversed direction.
Let σ 4 be the stress level at point 4. Then, the reverse yield can only take place at a stress level of – σ 4
(point 5).
CHAPTER 7 467
Material Library
The isotropic workhardening model (with a work slope of 0) is the default option in Marc.
For many materials, the isotropic workhardening model is inaccurate if unloading occurs (as in cyclic
loading problems). For these problems, the kinematic hardening model or the combined hardening
model represents the material better.
σ σ′
3
4
2
1
σy E
E +σ4 5 6
E
3
0 0
3
−σ4 21
4
5 σ′ σ′
1 2
6
(a) Loading Path (b) von Mises Yield Surface
Figure 7-41 Schematic of Isotropic Hardening Rule (Uniaxial Test)
Kinematic Hardening
Under the kinematic hardening rule, the von Mises yield surface does not change in size or shape, but
the center of the yield surface can move in stress space. Figure 7-42b illustrates this condition. Prager’s
law is used to define the translation of the yield surface in the stress space.
The loading path of a uniaxial test is shown in Figure 7-42a. The specimen is loaded in the following
order: from stress free (point 0) to initial yield (point 1), 2 (loading), 3 (unloading), 2 (reloading), 4
(loading), 5 and 6 (unloading). As in isotropic hardening, stress at 1 is equal to the initial yield stress σ y ,
and stresses at 2 and 4 are higher than σ y , due to workhardening. Point 3 is elastic, and reverse yield
takes place at point 5. Under the kinematic hardening rule, the reverse yield occurs at the level of
σ 5 = ( σ 4 – 2σ y ) , rather than at the stress level of – σ 4 . Similarly, if the specimen is loaded to a higher
stress level σ 7 (point 7), and then unloaded to the subsequent yield point 8, the stress at point 8 is
σ 8 = ( σ 7 – 2σ y ) . If the specimen is unloaded from a (tensile) stress state (such as point 4 and 7), the
reverse yield can occur at a stress state in either the reverse (point 5) or the same (point 8) direction.
To invoke the kinematic hardening in Marc, use the model definition options ISOTROPIC,
ORTHOTROPIC, or ANISOTROPIC. To input workhardening slope data, use the TABLE or WORK
HARD option or the WKSLP user subroutine.
For many materials, the kinematic hardening model gives a better representation of loading/unloading
behavior than the isotropic hardening model. For cyclic loading, however, the kinematic hardening
model can represent neither cyclic hardening nor cyclic softening.
σ σ′3
7 σ7
4 σ4 2σy
2
σy 1
3 2σy σ8
8
ε
0 5 σ5
6
(a) Loading Path σ′1

σ′2
(b) von Mises Yield Surface
Figure 7-42 Schematic of Kinematic Hardening Rule (Uniaxial Test)
Combined Hardening
Figure 7-43 shows a material with highly nonlinear hardening. Here, the initial hardening is assumed to
be almost entirely isotropic, but after some plastic straining, the elastic range attains an essentially
constant value (that is, pure kinematic hardening). The basic assumption of the combined hardening
model is that such behavior is reasonably approximated by a classical constant kinematic hardening
constraint, with the superposition of initial isotropic hardening. The isotropic hardening rate eventually
decays to zero as a function of the equivalent plastic strain measured by
·p p p 1⁄2
  --3- ε· i j ε· i j
2
εp = ε dt = dt (7-175)
This implies a constant shift of the center of the elastic domain, with a growth of elastic domain around
this center until pure kinematic hardening is attained. In this model, there is a variable proportion between
the isotropic and kinematic contributions that depends on the extent of plastic deformation (as measured
p
by ε ).
Use the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition option to activate the
combined workhardening option in Marc. Use the TABLE or WORK HARD option or the WKSLP user
subroutine to input workhardening slope data.
There are two formulations in the Marc program. The older one occurs whenever PLASTICITY,4 is not
included as a parameter. The workhardening data at small strains governs the isotropic behavior, and the
data at large strains ( ε p > 1000 ) governs the kinematic hardening behavior. If the last workhardening
slope is zero, the behavior is the same as the isotropic hardening model.
CHAPTER 7 469
Material Library
σ
Initial Combined Fully Hardened
Elastic Hardening Pure Kinematic
Range Range Range
Stress
Initial
Yield
One-half Current
Elastic Range
3 dα
Kinematic Slope, --- --------
2 d εp
ε
Strain
Figure 7-43 Basic Uniaxial Tension Behavior of the Combined Hardening Model
The newer formulation, which is invoked by including a PLASTICITY,4 parameter allows greater
generality by introducing the fractional contribution ( f h ) to kinematic hardening as a user input. As
shown in Figure 7-44, f is the value between 0 and 1 .
Work Hardening Curve Isotropic Hardening Kinematic Hardening
σ σα α
σy σy
εp εp εp
σα = σy + ( 1 – fh ) * ( σ – σy ) α = f( σ – σy )
Figure 7-44 Schematic Explanation for Kinematic Hardening Fraction
Isotropic hardening fh = 0
Kinematic hardening fh = 1
Combined hardening 0 < fh < 1 (default f h = 0.5 )
Combined hardening model utilizing kinematic fraction factor is available for Hill and Barlat models in
ISOTROPIC, ORTHOTROPIC, and ANISOTROPIC options.
Flow Rule
Yield stress and workhardening rules are two experimentally related phenomena that characterize plastic
material behavior. The flow rule is also essential in establishing the incremental stress-strain relations for
p
plastic material. The flow rule describes the differential changes in the plastic strain components dε as
a function of the current stress state.
The Prandtl-Reuss representation of the flow rule is available in Marc. In conjunction with the von Mises
yield function, this can be represented as:
p ∂σ
dε i j = dε p ---------- (7-176)
∂σ i j
where dε p and σ are equivalent plastic strain increment and equivalent stress, respectively.
The significance of this representation is illustrated in Figure 7-45. This figure illustrates the “stress-
space” for the two-dimensional case. The solid curve gives the yield surface (locus of all stress states
causing yield) as defined by the von Mises criterion.
Equation (7-176) expresses the condition that the direction of inelastic straining is normal to the yield
surface. This condition is called either the normality condition or the associated flow rule.
If the von Mises yield surface is used, then the normal is equal to the deviatoric stress.
σ2
dεp p
dε2
dε1p
σ1
Yield Surface
Figure 7-45 Yield Surface and Normality Criterion 2-D Stress Space
Constitutive Relations
This section presents the constitutive relation that describes the incremental stress-strain relation for an
elastic-plastic material. The material behavior is governed by the incremental theory of plasticity, the von
Mises yield criterion, and the isotropic hardening rule.
Let the workhardening coefficient H be expressed as:
H = dσ ⁄ dε p (7-177)
CHAPTER 7 471
Material Library
and the flow rule be expressed as:
p ∂σ
dε = dε p : ∇σ where ∇σ = ---------- (7-178)
∂σ i j
Consider the differential form of the familiar stress-strain law, with the plastic strains interpreted as
initial strains
dσ = C : dε – C : dε p (7-179)
where C is the elasticity matrix defined by Hooke’s law and “:” denotes the tensor contraction.
After substitution of Equation (7-178), this becomes
dσ = C : dε – C : ∇σdε p (7-180)
Contracting Equation (7-180) by ∇σ
∇σ : dσ = ∇σ : C : dε – ∇σ : C : ∇σdε p (7-181)
and recognizing that
dσ = ∇σ : dσ (7-182)
with use of Equation (7-177) in place of the left-hand side,
Hdε p = ∇σ : C : dε – ∇σ : C : ∇σdε p (7-183)

By rearrangement
∇σ : C : dε
dε p = ------------------------------------------- (7-184)
H + ∇σ : C : ∇σ
Finally, by substitution of this expression into Equation (7-180), we obtain
dσ = L e p :dε (7-185)
where L e p is the elasto-plastic, small strain tangent moduli expressed as:
ep ( C : ∇σ ) ⊗ ( C : ∇σ )
L = C – ------------------------------------------------------ (7-186)
H + ∇σ : C : ∇σ
The case of perfect plasticity, where H = 0 , causes no difficulty.
Temperature Effects
This section discusses the effects of temperature-dependent plasticity on the constitutive relation.
The following constitutive relations for thermo-plasticity were developed by Naghdi. Temperature
effects are discussed using the isotropic hardening model and the von Mises yield condition.
The stress rate can be expressed in the form

· ·
σ i j = L i j k l ε k l + h i j T· (7-187)
For elastic-plastic behavior, the moduli L i j k l are
∂σ ∂σ
L i j k l = C i j k l –  C i j m n ------------- ------------ C p q k l ⁄ D (7-188)
 ∂σ m n ∂σ p q 
and for purely elastic response
Li j k l = Ci j k l (7-189)
The term that relates the stress increment to the increment of temperature for elastic-plastic behavior is
∂σ 2 ∂σ
h i j = X i j – C i j k l α k l –  C i j k l -----------  σ p q X p q – --- σ -------  ⁄ D (7-190)
 ∂σ k l  3 ∂T 
and for purely elastic response
Hi j = Xi j – Ci j k l αk l (7-191)
where
4 ∂σ ∂σ ∂σ
D = --- σ 2 -------- + ---------- C i j k l ----------- (7-192)
9 ∂ε p ∂σ i j ∂σ k l
and
∂C i j k l e
X i j = --------------- ε (7-193)
∂T k l
and α k l are the coefficients of thermal expansion.
Strain Rate Effects

This section discusses the influence of strain rate on the elastic-plastic constitutive relation.
Strain rate effects cause the structural response of a body to change because they influence the
material properties of the body. These material changes lead to an instantaneous change in the
strength of the material. Strain rate effects become more pronounced for temperatures greater than
half the melting temperature ( T m ). The following discussion explains the effect of strain rate on the
size of the yield surface.
Using the von Mises yield condition and normality rule, we obtain an expression for the stress rate of
the form
· · ·· p
σi j = Li j k l εk l + ri j ε (7-194)
CHAPTER 7 473
Material Library
For elastic-plastic response

∂σ ∂σ
L i j k l = C i j k l –  C i j m n ------------- ------------ C p q k l ⁄ D (7-195)
 ∂σ m n ∂σ p q 
and
∂σ 2 ∂σ
r i j = C i j m n ------------- --- σ -------- ⁄ D
∂σ m n 3 ·p (7-196)
∂ε
where
4 ∂σ ∂σ ∂σ
D = --- σ 2 -------- + ---------- C i j k l ----------- (7-197)
9 ∂ε p ∂σ i j ∂σ k l
Time-independent Cyclic Plasticity

The cyclic plasticity model is based on the work of Chaboche [Ref. 23]. The current version of Marc
consists only of the basic model and plastic-strain-range memorization. The associated time-dependent
model is described on Time-dependent Cyclic Plasticity.
The model combines the isotropic hardening rule, to describe the cyclic hardening (Figure 7-46a) or
softening, and the nonlinear kinematic hardening to capture the proper characteristic of cyclic plasticity
like Bauschinger (Figure 7-46b), ratchetting (Figure 7-46c), and mean-stress relaxation (Figure 7-46d)
effect. The influence of the plastic strain range on the stabilized cyclic response is taken into account by
introducing the plastic-strain-range memorization variable (Figure 7-46e).
The von Mises yield function is now defined as follows:
f = σ – (R + k)
1
---
3S i j S i j 2
where σ =  ----------------- , s i j = σ' i j – --- δ i j σ' k k and σ' = σ – X
1
 2  3
X is the back stress tensor representing the center of the yield surface in stress space.
σ σ
–ε
ε
ε ε
(a) Cyclic Hardening under (b) Bauschinger Effect (c) Ratchetting

Multiple Cyclic Loading
σ
–ε
ε
(d) Mean Stress Relaxation (e) Cyclic Hardening

Figure 7-46 Typical Behavior of Material that can be Simulated with Cyclic Plasticity Model
Isotropic Hardening/Softening
The isotropic hardening/softening determines the size of the elastic region during the plastic loading. In
this model, it is controlled by parameter R and k . The initial conditions of cyclic hardening are given as
k = σ y and R = 0 , while a cyclic softening is initially described by k = σ y – R 0 and R = R 0 .
The evolution equation for the variable R is described as follows:
· ·
R = b ( R ∞ – R )λ
where b and R ∞ are material constants. R ∞ represents the limit of the isotropic hardening/softening. In
–b e
case of hardening, then R = R ∞  1 – e
p s
.
 
CHAPTER 7 475
Material Library
Nonlinear Kinematic Hardening

The nonlinear kinematic hardening is defined from the linear-Ziegler rule by adding the recall term as
shown in the evolution of the back stress tensor below:
· C - ·
X = ------------- ( σ – X ) – γX λ
R+k
where C and γ are two material constants. γ = 0 stands for linear-kinematic rule.
Plastic-strain-range Memorization
Several experimental observations show that the asymptotic stress value of cyclic hardening can depend
on the prior history. The influence of plastic-strain range on the stabilized cyclic response is evident from
the comparison between the different histories of loading used to obtained the cyclic curve. Therefore,
an introduction of new internal variables that memorize the prior maximum plastic range is introduced
by defining a “memory” surface in the plastic strain space as follows:
2
F = --- ε e ( ε p – ζ ) – ρ
3
The evolution of the state variables are as follows
· * ·
ρ = ηH ( F )  n ⋅ n  λ
· * *·
ζ = 3 ⁄ 2 ( 1 – η )H ( F )  n ⋅ n  n λ
*
where n and n are the unit normal to the yield surface f = 0 and to the memory surface F = 0
defined as follows:
2 εp * ε p– ζ
2 -----------
n = --- ----
·- and n = --- -
3 λ 3 ρ
The coefficient η is introduced in order to induce a progressive memory. For η = 0.5 then, the
memorization is instantaneous and stabilization occurs after one cycle. A progressive memory is given
by η < 0.5 .
The dependency between cyclic plastic flow and the plastic strain range is introduced by considering an
asymptotic isotropic state as follows:
–2 μ ρ
R ∞ = Q M + ( Q 0 – Q M )e
where Q M , Q 0 and μ are material constants.

Plastic Evolution Process and Elasto-plastic “Classical” Modular Matrix

·
The plastic evolution process must conform to the consistency condition, f = 0 . From this condition
the plastic “rate” multiplier can be derived as follows:
T
· a L ·
λ = -------------------------------------------------------------------------------------------------------------------- ε (7-198)
T C T T
a La + -------------- a ( σ – X ) – γa X + b ( R ∞ – R )
R+k
∂f 3 s
where a = ------ = --- -------------- .
∂σ 2R + k
Since the process involves nonlinear equation, iteration process using predictor-corrector technique is
used. The predictor is calculated using the “trial” elastic stresses as follows:
σ B = σ A + L Δσ
and then calculate f B based on σ B and hardening parameter on A. Using the Taylor expansion at B, then
fB
δλ = ---------------------------------------------------------------------------------------------------------------------
C
a T La + -------------- a T ( σ – X ) – γa T X + b ( R ∞ – R )
R+k
Having the global plasticity iteration converged, then the iteration to satisfy the plastic strain
memorization is started. If both iterations are converged, then the total plasticity iteration is considered
completed.
Inserting Equation (7-200) into Equation (7-199) and using Equation (7-198), the elasto-plastic
“classical” tangent modular matrix can be derived as follows:
 T 
= L  I – --------------------------------------------------------------------------------------------------------------------
EP aa L
L (7-199)
 T C T T 
 a La + -------------- a ( σ – X ) – γa X + b ( R ∞ – R )
R+k
Time-dependent Inelastic Behavior

Force-displacement relationships vary in different material models. A perfectly elastic material and a
perfectly viscous material can be represented by a spring and a dashpot, respectively (as shown in
Figure 7-47). In a perfectly elastic material, the deformation is proportional to the applied load. In a
perfectly viscous material, the rate of change of the deformation over time is proportional to the load.
In the class of viscoelastic and creeping materials, the application of a constant load is followed by a
deformation, which can be made up of an instantaneous deformation (elastic effect) followed by a
continual deformation with time (viscous effect). Eventually, it can become pure viscous flow. Continued
deformation under constant load is termed creep (see Figure 7-48).
CHAPTER 7 477
Material Library
SPRING DASHPOT
f = ku f = ηu
f = force f = Force
u = Displacement u = Velocity (Time Rate of Change of Displacement
k = Spring Stiffness η = Viscosity of the Dashpot
Figure 7-47 Perfectly Elastic (Spring) and Viscous (Dashpot) Materials
ε (Strain)
0 t (Time)
OA – Instantaneous Elastic Effect
AB – Delayed Elastic Effect
BC – Viscous Flow
Figure 7-48 The Creep Curve
A viscoelastic material can be subjected to sudden application of a constant deformation. This results in
an instantaneous proportional load (elastic effect), followed by a gradual reduction of the required load
with time, until a limiting value of the load is attained. The decreasing of load for a constant deformation,
is termed relaxation (see Figure 7-49).
Viscoelastic and creeping materials can be represented by models consisting of both springs and
dashpots because the material displays both elastic effects and viscous effects. This implies that the
material either continues to flow for a given stress, or the stress decreases with time for a given strain.
The measured relation between stress and strain is generally very complex.
σ (Stress)
0 t (Time)
Figure 7-49 The Relaxation Curve
Two models that are commonly used to relate stress and strain are the Maxwell and Kelvin (Voigt or
Kelvin-Voigt) models. A description of these models is given below.
The mathematical relation which holds for the Maxwell solid is
· ·
ε = ασ + βσ (7-200)
In the one-dimensional case for normal stress
1
α = ---
E
(7-201)
1
β = ---
η
This relation can be depicted as a spring and dashpot in series, as shown in Figure 7-50. The integration
of Equation (7-200) yields
σ σ
ε = --- +  --- dt (7-202)
E η
σ σ
Figure 7-50 Maxwell Solid
The strain and stress responses of the Maxwell Solid model are shown in Figure 7-51 and
Figure 7-52, respectively.
σ ε
σ
-----0-
E
t t
Constant Stress Applied Response to Constant Stress
Figure 7-51 Strain Response to Applied Constant Stress (Maxwell Solid)

CHAPTER 7 479
Material Library
ε σ
t t
τ
Constant Strain Applied Response to Constant Strain
Figure 7-52 Stress Response to Applied Constant Strain (Maxwell Solid)
The mathematical relation which holds for the Kelvin (Voigt or Kelvin-Voigt) solid is
·
σ = αε + βε (7-203)
This equation is depicted as a spring and dashpot in parallel. (See Figure 7-53). When β = 0
(no dashpot), the system is a linearly elastic system in which α = E , the elastic modulus.
When E = 0 (no spring), the “solid” obeys Newton’s equation for a viscous fluid and β = η , the
viscous coefficient. Thus, we can rewrite Equation (7-203) in the form
·
σ = Eε + ηε (7-204)
In the above relation, we have considered one-dimensional normal stress and strain. The relation holds
equally well for shear stress τ and shear strain γ in which α = G , the shear modulus, and β = η ,
the viscous coefficient. Equation (7-203) can be rewritten as
·
τ = Gγ + ηγ (7-205)
σ σ
Figure 7-53 Kelvin (Voigt or Kelvin-Voigt) Solid
The strain responses of the Kelvin Solid model are depicted in Figure 7-54. For multiaxial situations,
these equations can be generalized to tensor quantities.
To invoke the Maxwell model, use the CREEP parameter. The creep strain can be specified as either
deviatoric creep strain (conventional creep) or dilatational creep strain (swelling). To invoke the Kelvin
model, also use the CREEP parameter and CRPVIS user subroutine.
σ0
t
t1
(a) Stress Pulse
ε
σ0 t
------
E
t
(b) Strain Response to Stress of Infinite Domain
ε
t1
t
(c) Strain Response to Stress Pulse of Finite Length
Figure 7-54 Strain Response to Applied Stress (Kelvin Solid)
Creep (Maxwell Model)

Creep is an important factor in elevated-temperature stress analysis. In Marc, creep is represented by a
Maxwell model. Creep is a time-dependent, inelastic behavior, and can occur at any stress level (that is,
either below or above the yield stress of a material). In many cases, creep is accompanied by plasticity
which occurs above the yield stress of the material. The creep behavior can be characterized as primary,
secondary, and tertiary creep, as shown in Figure 7-55. Engineering analysis is often limited to the
primary and secondary creep regions. Tertiary creep in a uniaxial specimen is usually associated with
geometric instabilities, such as necking. The major difference between the primary and secondary creep
is that the creep strain rate is much larger in the primary creep region than it is in the secondary creep
region. The creep strain rate is the slope of the creep strain-time curve. The creep strain rate is generally
dependent on stress, temperature, and time.
The creep data can be specified in either an exponent form or in a piecewise linear curve. To specify creep
data, use the CREEP model definition option. The CRPLAW user subroutine allows alternative forms of
creep behavior to be programmed directly.
CHAPTER 7 481
Material Library
·c dε c
ε = -------- (7-206)
dt
Creep Strain
εC
Tertiary
Creep
Secondary
Creep
Primary
Creep
Time (t)
Note: Primary Creep: Fast decrease in creep strain rate
Secondary Creep: Slow decrease in creep strain rate
Tertiary Creep: Fast increase in creep strain rate
Figure 7-55 Creep Strain Versus Time (Uniaxial Test at Constant Stress and Temperature)
Marc offers two schemes for modeling creep in conjunction with plasticity: (a) treating creep strains and
plastic strains separately; and (b) modeling creep strains and plastic strains in a unified fashion
(viscoplasticity). Both schemes can be treated using two different procedures: explicit and implicit.
Creep (Explicit Formulation)

There are six possible modes of input for creep constitutive data.
1. Express the dependence of equivalent creep strain rate on any independent parameter through a
piecewise linear relationship. The equivalent creep strain rate is then assumed to be a piecewise
linear approximation to
·c dk ( t )
ε = A ⋅ f ( σ ) ⋅ g ( ε c ) ⋅ h ( T ) ⋅ ------------- (7-207)
dt
·c c
where A is a constant; ε is equivalent creep strain rate; and σ , ε , T , and t are equivalent
stress, equivalent creep strain, temperature and time, respectively. The functions f , g , h , and k
are piecewise linear and entered in the form as either slope-break point data or function-variable
data. This representation is shown in Figure 7-56. Enter functions f , g , h , and k through the
CREEP model definition option. (Any of the functions f , g , h , or k can be set to unity by setting
the number of piecewise linear slopes for that relation to zero on the input data.)
2. The dependence of equivalent creep strain rate on any independent parameter can be given
directly in power law form by the appropriate exponent. The equivalent creep strain rate is
· n
ε c = Aσ m ⋅ ( ε c ) ⋅ T p ⋅ ( qt q – 1 )
Enter the constants A , m , n , p , and q directly through the CREEP model definition option. This
is often adequate for engineering metals at constant temperature where Norton’s rule is a good
approximation.
·c
ε = A σn (7-208)
3. input procedure, one can define a table such that
·c
ε = A ⋅ f ( v 1 ,v 2 ,v 3 ,v 4 ) (7-209)
where v 1 ,v 2 ,v 3 ,v 4 are four variables that could include equivalent stress, temperature, time,
strain, position, or one of 30 variables. In this case, the function is piecewise linear.
4. Using the table driven input procedure, one can define the strain rate such that
· cr
ε = A ⋅ equation ( v 1 ,v 2 ,v 3 ,v 4 ) (7-210)
For example, one could enter the equation
Aσ n e – Q R ⁄ T (7-211)
The equation can take form as long as there is no conditional logic required. If this is not the case,
the CRPLAW user subroutine should be used.
5. Define the equivalent creep strain rate directly with the CRPLAW user subroutine.
6. Use the ISOTROPIC option to activate the ORNL (Oak Ridge National Laboratory rules)
capability of the program.
Isotropic creep behavior is based on a von Mises creep potential described by the equivalent creep law
·
ε = f ( σ , ε c , T, t ) (7-212)
The material creep behavior is described by
· · c  ∂σ 
ε icj = ε  ----------  (7-213)
 ∂σ i j 
During creep, the creep strain rate usually decreases. This effect is called creep hardening and can be a
function of time or creep strain. The following section discusses the difference between these two types
of hardening.
Consider a simple power law that illustrates the difference between time and strain-hardening rules for
the calculation of the creep strain rate.
CHAPTER 7 483
Material Library
c
ε = βt n (7-214)
F4
F3
S3
Function F (X)
[for example, t ( σ ) , F2 S2
c
g ( ε ) , h (T),
k (t)]
S1
F1
X1 X2 X3 X4
Variable X (Such as σ, εC, T, t)
(1) Slope-Break Point Data

Slope Break Point
S1 X1
S2 X2
S3 X3
(2) Function-Variable Data

Function Variable
F1 X1
F2 X2
F3 X3
F4 X4
Figure 7-56 Piecewise Linear Representation of Creep Data
c
where ε is the creep strain, β and n are values obtained from experiments and t is time. The creep rate
c
can be obtained by taking the derivative ε with respect to time
c
·c dε
ε = -------- = nβt n – 1 (7-215)
dt
However, t being greater than 0, we can compute the time t as
c 1/n
ε
t =  ----- (7-216)
 β
Substituting Equation (7-211) into Equation (7-210) we have
·c c (n – 1) ⁄ n
ε = nβt n – 1 = n ( β 1 ⁄ n ( ε ) ) (7-217)
Equation (7-211) shows that the creep strain rate is a function of time (time hardening). Equation (7-217)
indicates that the creep strain rate is dependent on the creep strain (strain hardening). The creep strain
rates calculated from these two hardening rules generally are different. The selection of a hardening rule
in creep analysis must be based on data obtained from experimental results. Figure 7-57 and Figure 7-58
show time and strain hardening rules in a variable state of stress. It is assumed that the stress in a structure
varies from σ 1 to σ 2 to σ 3 ; depending upon the model chosen, different creep strain rates are
calculated accordingly at points 1, 2, 3, and 4. Obviously, creep strain rates obtained from the time
hardening rule are quite different from those obtained by the strain hardening rule.
εc
σ1
σ2
3
σ3
1
4
2
0
t
Figure 7-57 Time Hardening

CHAPTER 7 485
Material Library
εc
σ1
σ2
3
σ3
1 4
2
0
t
Figure 7-58 Strain Hardening
Oak Ridge National Laboratory Laws

Material Oak Ridge National Laboratory (ORNL) has performed a large number of creep tests on stainless and
Library other alloy steels. It has also set certain rules that characterize creep behavior for application in nuclear
structures. A summary of the ORNL rules on creep is given below. The references listed at the end of
this section offer a more detailed discussion of the ORNL rules.
1. Auxiliary Rules for Applying Strain-Hardening to Situations Involving Stress Reversals
The Blackburn Creep Law is required as the CRPLAW user subroutine. The parameter EQCP
(first parameter in CRPLAW) is defined as
c 1/2
ε c =  --- ΣΔε i j ΣΔε i j 
2 c
(7-218)
3 
when the ORNL constitutive option is flagged through use of the ISOTROPIC option. In all other
cases, the definition
c 1/2
ε c = Σ  --- Δε icj Δε i j 
2
(7-219)
3 
is retained. The equivalent primary creep strain passes into CRPLAW in EQCPNC, the second
parameter. The second parameter must be redefined in that routine as the equivalent (total) creep
strain increment. The first parameter (EQCP) must be redefined as the equivalent primary creep
strain increment when the ORNL constitutive option is flagged. During analysis with the ORNL
option, equivalent creep strain stores the distance between the two shifted origins in creep strain
space ( ε in ORNL-TM-3602). The sign on this value indicates which origin is currently active,
so that a negative sign indicates use of the “negative” origin ( – ε i j ).
2. Plasticity Effect on Creep

The effect of plastic strains on creep must be accommodated for the time-dependent creep
behavior of 2 1/4 Cr -1 Mo Steel. Since plastic strains in one direction reduce the prior creep strain
hardening accumulated in the reverse direction, ORNL recommends that the softening influence
due to plastic strains be treated much the same as when reversed creep strain occurs. The
following quantities are defined:
+ + +
N i j = ( ε iIj – ε i j ) ⁄ G (7-220)
– –
Ni j = ( ε iIj – ε i–j ) ⁄ G (7-221)
where ε iIj is instantaneous creep strain components
ε i+j ;ε i–j = positive and negative strain origins (7-222)
and
+
G = G ( ε iIj – ε i+j ) = [ 2 ⁄ 3 ( ε iIj – ε i+j ) ( ε iIj – ε i+j ) ] 1 ⁄ 2 (7-223)
G– = G ( ε iIj – ε i–j ) = [ 2 ⁄ 3 ( ε iIj – ε i–j ) ε iIj – ε i–j ] 1 ⁄ 2 (7-224)
Swelling
Marc allows pure swelling (dilatational creep) effect in a creep analysis. To use the swelling option,
perform a regular creep analysis as discussed earlier. Use the VSWELL user subroutine to define the
ΔV
increment of volumetric swelling  -------- . The increment of volumetric swelling is generally a function
 V
of neutron flux, time, and temperatures.
For example, radiation-induced swelling strain model for 20% C. W. Stainless Steel 316 can be
expressed as:
ΔV R 1 + exp ( α ( τ – φt ) )
-------- = Rφt + ---- ln ------------------------------------------------- (7-225)
V α 1 + expτ
where R , t , and α are functions of temperature, φ is neutron flux, and t is time.
Creep (Implicit Formulation)

This formulation, as opposed to that described in the previous section, is fully implicit. A fully implicit
formulation is unconditionally stable for any choice of time step size; hence, allowing a larger time step
than permissible using the explicit method. Additionally, this method is more accurate than the explicit
method. The disadvantage is that each increment may be more computationally expensive. This model
is activated using the CREEP parameter. There are two methods for defining the inelastic strain rate. The
CHAPTER 7 487
Material Library
CREEP model definition option can be used to define a Maxwell creep model. The back stress must be
specified through the field normally reserved for the yield stress in the ISOTROPIC or ORTHOTROPIC
options. The yield stress must be specified through the field normally reserved for the 10th cycle yield
stress in the ISOTROPIC option. There is no plastic strain when the stress is less than the yield stress and
there is no creep strain when the stress is less than the back stress.
The equivalent creep strain increment is expressed as
·
ε c = Aσ m ⋅ ( ε c ) n ⋅ T p ⋅ ( qt q – 1 ) (7-226)
Enter the constants A , m , n , p , and q directly through the CREEP model definition option. A more
general expression for the equivalent creep strain rate is given by:
·c dk ( t )
ε = A ⋅ σ m ⋅ g ( ε c ) ⋅ h ( T ) ⋅ ------------- (7-227)
dt
Enter the terms A and m and the functions g , h , and k through the UCRPLW user subroutine.
The creep strain components are given by:
d
3 σi j
Δε i j = --- Δε -------
2 σ
d
where σ i j is the deviatoric stress at the end of the increment. A is a function of temperature, time, etc.
An algorithmic tangent is used to form the stiffness matrix.
Based on a parameter defined in the CREEP parameter, one of three tangent matrices is formed. The first
is using an elastic tangent, which requires more iterations, but can be computationally efficient because
re-assembly might not be required. The second is an algorithmic tangent that provides the best behavior
for small strain power law creep. The third is a secant (approximate) tangent that gives the best behavior
for general viscoplastic models. When creep is specified in conjunction with plasticity, the elastic tangent
option is not available.
Viscoplasticity (Explicit Formulation)

The creep (Maxwell) model can be modified to include a plastic element (as shown in Figure 7-59). This
plastic element is inactive when the stress ( σ ) is less than the yield stress ( σ y ) of the material. The
modified model is an elasto-viscoplasticity model and is capable of producing some observed effects of
creep and plasticity. In addition, the viscoplastic model can be used to generate time-independent
plasticity solutions when stationary conditions are reached. At the other extreme, the viscoplastic model
can reproduce standard creep phenomena. The model allows the treatment of nonassociated flow rules
and strain softening which present difficulties in conventional (tangent modulus) plasticity analyses.
The viscoplasticity option can be used to implement very general constitutive relations with the aid of
the following user subroutines: ZERO, YIEL, NASSOC, and CRPLAW. See Nonlinear Analysis in
Chapter 5 for details on how to use these procedures.
εe
εvp
p vp
ε = ε Plastic Element
Inactive if σ < σy
Figure 7-59 Uniaxial Representation of Viscoplastic Material
Viscoplasticity (Implicit Formulation)

To allow for the implementation of general unified creep-plasticity or viscoplastic models, the UVSCPL
user subroutine is available. This routine requires you to define only the inelastic strain rate. The program
automatically calculates a tangent stiffness matrix (only elastic tangent or secant tangent can be used).
This option is activated by indicating that the material is VISCO PLAS in the ISOTROPIC or
ORTHOTROPIC option.
Time-dependent Cyclic Plasticity

The time-dependent effect of the model described in “Time-independent Cyclic Plasticity” on page 473
is modeled using the unified viscoplastic framework. The viscoplastic potential is based on “overstress”
quantity as follows:
K f n+1
Ω = -------------  ----
n+1 K
The viscoplastic strain rate is defined as follows:
· ∂Ω 3 · σ' – X' 3·
εv p = = --- λ ----------------- = --- λ ∇f (7-228)
∂σ 2 σ̃ e 2
where
· f n
λ =  ---- (7-229)
K
1
---
·n
In this case, the viscoplastic stress is σ v p = Kλ .
The hardening rules are chosen to be identical to the time-independent case.

CHAPTER 7 489
Material Library
Viscoplastic Evolution Process and Visco-plastic “Classical” Modular Matrix

The iteration procedure, starting from the trial elastic stress as the predictor, is based on the implicit
integration of Equation (7-229) that can be expressed as follows:
Δλ f n
r λ = – ------ +  ----  (7-230)
Δt K
Using Newton iteration scheme, the iterative value of Δλ (that is, δλ ) can be expressed as follows:
i
Δt r n
δλ = ---------------------------------------------------------------------------------------------------------------------------------------
n f n – 1 T
a La + --- Ca a – γa X + b ( R ∞ – R )
2 T T
1 + ----  ----
K K  3 
i
where r λ is the i-th iteration of the residual of Equation (7-230).
The tangent modular matrix is based on the assumption that δr λ = 0 . Therefore the classical visco-
plastic modular matrix can be expressed as follows:
 
 
vp  aa L
T 
L = L  I – ------------------------------------------------------------------------------------------------------------------------------------
 a La + --- Ca a – γa X + b ( R ∞ – R ) + --------------------------------
T 2 T T 1
 3 n f n–1 
 ----  ----  Δt 
K K
This matrix is in line with the consistent model derived in [Ref. 20].
Viscoelastic Material
Marc has two models that represent viscoelastic materials. The first can be defined as a Kelvin-Voigt
model. The latter is a general hereditary integral approach.
Kelvin-Voigt Model
The Kelvin model allows the rate of change of the inelastic strain to be a function of the total stress and
previous strain. To activate the Kelvin model in Marc, use the CREEP parameter.
k
The Kelvin material behavior (viscoelasticity) is modeled by assuming an additional creep strain ε ,
ij
governed by
d
----- ε k = A i j k l σ kdl – B i j k l ε k (7-231)
dt i j kl
where A and B are defined in the CRPVIS user subroutine and the total strain is
ε i j = ε iej + ε ipj + ε icj + ε ikj + εti hj (7-232)

εti hj = thermal strain components (7-233)
ε iej = elastic strain components (instantaneous response) (7-234)
ε ipj = plastic strain components (7-235)
ε icj = creep strains defined via the CRPLAW and VSWELL user subroutines (7-236)
ε ikj = Kelvin model strain components as defined above (7-237)
The CRPVIS user subroutine is called at each integration point of each element when the Kelvin model
is used.
Use the AUTO CREEP option to define the time step and to set the tolerance control for the maximum
strain in any increment.
The CREEP option allows Maxwell models to be included in series with the Kelvin model.
Hereditary Integral Model

The stress-strain equations in viscoelasticity are not only dependent on the current stress and strain state
(as represented in the Kelvin model), but also on the entire history of development of these states. This
constitutive behavior is most readily expressed in terms of hereditary or Duhamel integrals. These
integrals are formed by considering the stress or strain build-up at successive times. Two equivalent
integral forms exist: the stress relaxation form and the creep function form. In Marc, the stress relaxation
form is used.
The viscoelasticity option in Marc can be used for both the small strain and large strain Mooney or Ogden
material stress-relaxation problems for total Lagrange formulation. It can also be used with all
hyperelastic models; i.e., Mooney, Ogden, Gent, Arruda-Boyce, Foam and generalized hyperelastic
materials if the Updated Lagragian formulation is used. A description of these models is as follows:
Small Strain Viscoelasticity

In the stress relaxation form, the constitutive relation can be written as a hereditary integral formulation
t
dε k l ( τ )
σi j ( t ) =  G i j k l ( t – τ ) -----------------
dτ
- dτ + G i j k l ( t )ε k l ( 0 ) (7-238)
0
The functions G i j k l are called stress relaxation functions. They represent the response to a unit applied
strain and have characteristic relaxation times associated with them. The relaxation functions for
materials with a fading memory can be expressed in terms of Prony or exponential series.
N
∞ n n
Gi j k l ( t ) = Gi j k l +  G i j k l exp ( – t ⁄ λ ) (7-239)
n = 1
CHAPTER 7 491
Material Library
n n
in which G i j k l is a tensor of amplitudes and λ is a positive time constant (relaxation time). In the
∞
current implementation, it is assumed that the time constant is isotropic. In Equation (7-239), G i j k l
represents the long term modulus of the material.
The short term moduli (describing the instantaneous elastic effect) are then given by
N
0 ∞ n
Gi j k l = Gi j k l ( 0 ) = Gi j k l +  Gi j k l (7-240)
n = 1
The stress can now be considered as the summation of the stresses in a generalized Maxwell model
(Figure 7-60)
N
∞ n
σi j ( t ) = σi j ( t ) +  σi j ( t ) (7-241)
n = 1
where
∞ ∞
σi j = Gi j k l εk l ( t ) (7-242)
t
n n n dε k l ( τ )
σi j =  G i j kl exp [ – ( t – τ ) ⁄ λ ] ------------------ dτ
dτ
(7-243)
0
1 2 3 N
E E E E
∞
E
1 2 3 N
η η η η
i i i
τ = η ⁄E
Figure 7-60 The Generalized Maxwell or Stress Relaxation Form
For integration of the constitutive equation, the total time interval is subdivided into a number of
subintervals ( t m – 1, t m ) with time-step Δt = t m – t m – 1 . A recursive relation can now be derived
expressing the stress increment in terms of the values of the internal stresses σ inj at the start of the
interval. With the assumption that the strain varies linearly during the time interval Δt , we obtain the
increment stress-strain relation as
N N
∞ n n n n
Δσ i j ( t m ) = Gi j k l +  β ( Δt )G i j k l Δε k l –  α ( Δt )σ i j ( t m – Δt ) (7-244)
n = 1 n = 1
where
n
α n ( Δt ) = 1 – exp ( – ( Δt ) ⁄ λ ) (7-245)
and
n n n
β ( Δt ) = α ( Δt )λ ⁄ ( Δt ) (7-246)
In Marc, the incremental equation for the total stress is expressed in terms of the short term moduli (See
Equation (7-240)).
N N
0 n n n n
Δσ i j ( t m ) = Gi j k l –  {1 – β ( Δt ) }G i j k l Δε k l ( t m ) –  α ( Δt )σ i j ( t m – Δt ) (7-247)
n = 1 n = 1
In this way, the instantaneous elastic moduli can be specified through the ISOTROPIC or
ORTHOTROPIC options. Moreover, since the TEMPERATURE EFFECTS or TABLE option acts on the
instantaneous elastic moduli, it is more straightforward to use the short term values instead of the long
term ones. Note that the set of equations given by Equation 7-247 can directly be used for both
anisotropic and isotropic materials.
Isotropic Viscoelastic Material

For an isotropic viscoelastic material, Marc assumes that the deviatoric and volumetric behavior are fully
decoupled and that the behavior can be described by a time dependent shear and bulk modules. The bulk
moduli is generally assumed to be time independent; however, this is an unnecessary restriction of the
general theory.
Both the shear and bulk moduli can be expressed in a Prony series
N
∞ n n
G(t) = G +  G exp ( – t ⁄ λ d ) (7-248)
n = 1
N
∞ n n
K(t) = K +  K exp ( – t ⁄ λ v ) (7-249)
n = 1
with short term values given by

N
G0 = G∞ +  Gn (7-250)
n = 1
CHAPTER 7 493
Material Library
N
K0 = K∞ +  Kn (7-251)
n = 1
Let the deviatoric and volumetric component matrices π d and π v be given by
4 ⁄ 3 –2 ⁄ 3 –2 ⁄ 3 0 0 0
–2 ⁄ 3 4 ⁄ 3 –2 ⁄ 3 0 0 0
–2 ⁄ 3 –2 ⁄ 3 4⁄3 0 0 0
πd =
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1 (7-252)
1 1 1 0 0 0
1 1 1 0 0 0
πv = 1 1 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
The increment set of equations is then given by
 Nd 
 0 n n
Δσ ( t m ) =  G –

 [1 – β d ( Δt ) ]G π d Δε ( t m )

 n = 1 
 N
v  (7-253)
 0 n n
K –  [1 – β v ( Δt ) ]K π v Δε ( t m )
 
 n = 1 
Nd Nv
n n n n
–  α d ( Δt )σ d ( t m – Δt ) –  α v ( Δt )σ v ( t m – Δt )
n = 1 n = 1
and
n n n n n
Δσ d ( t m ) = β d ( Δt )G π d Δε ( t m ) – α d ( Δt )σ d ( t m – Δt )
n n n n n (7-254)
Δσ v ( t m ) = β v ( Δt )K π v Δε ( t m ) – α v ( Δt )σ v ( t m – Δt )
Note that the deviatoric and volumetric response are fully decoupled.
The instantaneous moduli need to be given in the ISOTROPIC option. Time dependent values (shear
n n n
moduli G n and time constants λ d ; bulk moduli K and time constants λ v ) need to be entered in the
VISCELPROP option.
Time-stepping is performed using the TIME STEP or AUTO STEP option in the history definition block.
Note that the algorithm is exact for linear variations of the strain during the increment. The algorithm is
implicit; hence, for each change in time-step, a new assembly of the stiffness matrix is required.
Anisotropic Viscoelastic Material

Equation 7-244 can be used for the analysis of anisotropic viscoelastic materials. The tensor of
0 n
amplitudes G i j k l must be entered through the ORTHOTROPIC option. However, both the G i j k l and the
n
λ must be entered using the VISCELORTH option.
Alternatively, a complete set of moduli (21 components) can be specified in the HOOKVI user subroutine.
The ORIENTATION option or ORIENT user subroutine can be used to define a preferred orientation both
0 n
for the short time moduli G i j k l and the amplitude functions G i j kl .
Incompressible Isotropic Viscoelastic Materials

Incompressible elements in Marc allow the analysis of incompressible and nearly incompressible
materials in plane strain, axisymmetric and three-dimensional problems. The incompressibility of the
element is simulated through the use of an perturbed Lagrangian variational principle based on the
Herrmann formulation.
The constitutive equation for a material with no time dependence in the volumetric behavior can be
expressed as
 N 
 0 n n  1
Δσ i j ( t m ) = 2  G i j kl –

 [1 – β ( Δt ) ]G i j k l 

Δε k l ( t m ) – --- Δε p p ( t m )δ k l
3
 n = 1  (7-255)
N
n n 1
–  α ( Δt ) ( σ′i j ) ( t m ) + --- σ k k δ i j
3
n = 1
0
Δσ p p ( t m ) = 3K Δε p p ( t m ) (7-256)
The hydrostatic pressure term is used as an independent variable in the variational principle. The
Herrmann pressure variable is now defined in the same way as in the formulation for time independent
elastic materials.
CHAPTER 7 495
Material Library
σp p
H = -------------------------------- (7-257)
2G 0 ( 1 + ν 0 )
The constitutive Equations (7-255) and (7-256) can then be rewritten
N
e d n
Δσ i j ( t m ) = 2G ( Δε i j + ν∗ Hδ i j ) – 
n
α ( Δt ) ( σ i j ) ( t m – Δt ) (7-258)
n = 1
where
N
Ge = G0 –  [ 1 – β n ( Δt )G n ] (7-259)
n = 1
0 0 e 0
G ( 1 + ν ) – G ( 1 – 2ν )
ν∗ = ------------------------------------------------------------------- (7-260)
e
3G
Thermo-Rheologically Simple Behavior

The rate processes in many viscoelastic materials are known to be highly sensitive to temperature
changes. Such temperature-dependent properties cannot be neglected in the presence of any appreciable
temperature variation. For example, there is a large class of polymers which are adequately represented
by linear viscoelastic laws at uniform temperature. These polymers exhibit an approximate translational
shift of all the characteristic response functions with a change of temperature, along a logarithmic time
axis. This shift occurs without a change of shape. These temperature-sensitive viscoelastic materials are
characterized as Thermo-Rheologically Simple.
A “reduced” or “pseudo” time can be defined for the materials of this type and for a given temperature
field. This new parameter is a function of both time and space variables. The viscoelastic law has the
same form as one at constant temperature in real time. If the shifted time is used, however, the
transformed viscoelastic equilibrium and compatibility equations are not equivalent to the corresponding
elastic equations.
In the case where the temperature varies with time, the extended constitutive law implies a nonlinear
dependence of the instantaneous stress state at each material point of the body upon the entire
local temperature history. In other words, the functionals are linear in the strains but nonlinear in
the temperature.
The time scale of experimental data is extended for Thermo-Rheologically Simple materials. All
characteristic functions of the material must obey the same property. The shift function is a basic
property of the material and must be determined experimentally. As a consequence of the shifting of the
mechanical properties data parallel to the time axis (see Figure 7-61), the values of the zero and infinite
frequency complex moduli do not change due to shifting. Hence, elastic materials with
temperature-dependent characteristics neither belong to nor are consistent with the above hypothesis for
the class of Thermo-Rheologically Simple viscoelastic solids.
T0
f(T1)
T1
T2 f(T2)
GT
ln t
Figure 7-61 Relaxation Modulus vs. Time at Different Temperatures
Let E ( ln t ) be the relaxation modulus as a function of ln t at uniform temperature, T . Then
E T ( ln t ) = E T [ ln t+f ( T ) ]* ( ρT ⁄ ρ 0 T 0 ) (7-261)
0
where f ( T ) is measured relative to some arbitrary temperature T . The modulus curve shifts towards
shorter times with an increase of temperature; f ( T ) is a positive increasing function for T > T 0 . If
G T ( t ) denotes the relaxation modulus as a function of time at uniform temperature T , so that,
G T ( t ) = E T ( ln t ) (7-262)
then
GT ( t ) = GT ( ξ ) (7-263)
0
The relaxation modulus (and the other characteristic functions) at an arbitrary uniform temperature
is thus expressed by the base temperature behavior related to a new time scale that depends on
that temperature.
There is some mapping of the time coordinate for nonuniform, nonconstant temperature, T ( x, t ) , which
depends on the position
For a nonuniform, nonconstant temperature, the shift function is a ( T ( x, t ) ) and the rate of change of
reduced time can be written as:
dξ = a T [ T ( x ,t ) ]dt
Marc offers two explicit forms for entering the shift function. The first is based on the familiar Williams-
Landel-Ferry (WLF) equation. Rewriting the above expression for reduced time as
t
dt′
ξ ( x, t ) =  α-------------------------------
[ T ( x, t′ ) ]
(7-264)
T
0
then the WLF form state that

CHAPTER 7 497
Material Library
– C 1 ( T – T0 )
log 10aT ( T ) = ------------------------------------ = – h ( T ) (7-265)
C2 + ( T – T 0 )
and
t
h[T(t′)]
ξ( t) =  10 dt′ (7-266)
0
Typically, the glassy transition point is taken as the reference temperature in the above relation. The
logarithmic shift can also be expressed in a polynomial expansion about the arbitrary reference point as
m
i
log 10 A T ( T ) =  αi ( T – T 0 ) (7-267)
i = 0
Enter the shift function parameters associated with Thermo-Rheologically Simple behavior through the
SHIFT FUNCTION model definition option. As an alternative to the WLF function, Marc allows use of
series expansion or specification via the TRSFAC user subroutine.
In addition to the Thermo-Rheologically Simple material behavior variations of initial stress-strain
0
moduli G i j k l , the temperature of the other mechanical properties (coefficient of thermal expansion, etc.)
due to changes in temperature can be specified via the TEMPERATURE EFFECTS option.
Note, however, that only the instantaneous moduli are effected by the TEMPERATURE EFFECTS
option. Hence, the long term moduli given by
N
∞ 0 n
Gi j k l = Gi j k l ( t ) –  Gi j k l (7-268)
n = 1
can easily become negative if the temperature effects are not defined properly.
Large Strain Viscoelasticity

For an elastomeric time independent material, the constitutive equation is expressed in terms of an
energy function W . For a large strain viscoelastic material, Simo generalized the small strain
viscoelasticity material behavior to a large strain viscoelastic material. The energy functional
then becomes
N N
n 0 n n n
ψ ( Ei j Qi j ) = ψ ( Ei j ) –  Qi j Ei j +  ψI ( Qi j ) (7-269)
n = 1 n = 1
n 0
where E i j are the components of the Green-Lagrange strain tensor, Q i j internal variables and ψ the
0
elastic strain energy density for instantaneous deformations. In Marc, it is assumed that ψ = W,
meaning that the energy density for instantaneous deformations is given by the third order James-Green-
Simpson form or the Ogden form. When used with Updated Lagrange, one can also use the Arruda-
Boyce or Gent model.
The components of the second Piola-Kirchhoff stress then follow from
N
∂ψ ∂ψ 0 n
Si j = ---------- = ---------- –
∂E i j ∂E i j  Qi j (7-270)
n = 1
The energy function can also be written in terms of the long term moduli resulting in a different set of
n
internal variables T i j
N
n n
ψ ( E i j, T i j ) = ψ∞(E ij) +  Ti j Ei j (7-271)
n = 1
∞
where ψ is the elastic strain energy for long term deformations. Using this energy definition, the
stresses are obtained from
N
∂ψ ∞ ( E ) n
Si j = -------------------- +
∂E i j  Ti j (7-272)
n = 1
Let the total strain energy be expressed as a Prony series expansion

N
ψ = ψ∞ +  ψ n exp ( – t ⁄ λ n ) (7-273)
n = 1
If, in the energy function, each term in the series expansion has a similar form, Equation (7-273) can be
rewritten as
N
∞
ψ = ψ +  δ n ψ 0 exp ( – t ⁄ λ n ) (7-274)
n = 1
n
where δ is a scalar multiplier for the energy function based on the short term values.
The stress-strain relation is now given by
N
∞ n
Si j = Si j +  Ti j (7-275)
n = 1
CHAPTER 7 499
Material Library
 N 
∞ ∂ψ ∞ ∂ψ 0
S i j = ----------- =  1 – n ---------
  ∂E i-j
δ (7-276)
∂E i j 
 n = 1 
Observing the similarity with the equations for small strain viscoelasticity, the internal variables can be
obtained from a convolution expression
t
n n ·0 n
Ti j =  S i j ( τ )exp [ – ( t – τ ) ⁄ λ ]dτ
δ (7-277)
0
Analogue to the derivation for small strain viscoelasticity, a recursive relation can be derived expressing
the stress increment in terms of values of the internal stresses at the start of the increment.
In Marc, the instantaneous values of the energy function are always given on the MOONEY, OGDEN, or
FOAM option, the equations are reformulated in terms of the short time values of the energy function
 N 
ΔS i j ( t m ) = 1 –  [ 1 – β n ( Δt ) ] δ n { S i j ( t m ) – S i j t m – Δt }
 0 0
 
 n=1 
N
(7-278)

n
– α n ( Δt )T i j ( t m – Δt )
n = 1
n 0 0 n
ΔT i j ( t m ) = β n ( Δt )δ n [ S i j ( t m ) – S i j ( t m – Δt ) ] – α n ( Δt )T i j ( t m – Δt ) (7-279)
It is assumed that the viscoelastic behavior in Marc acts only on the deviatoric behavior. The
incompressible behavior is taken into account using special Herrmann elements.
Large strain viscoelasticity is invoked by use of the VISCELMOON, VISCELOGDEN, or VISCELFOAM
n
option in the constitutive option of the model definition block. The time dependent multipliers δ and
n
associated relaxation times λ as defined by Equation 7-266 are given in the VISCELMOON,
VISCELOGDEN, or VISCELFOAM option. For the Ogden model, both deviatoric and dilatational
relaxation behavior is allowed.
Viscoelasticity can be modeled with Arruda-Boyce and Gent models using VISCELMOON option.
Time-stepping can be performed using the TIME STEP with AUTO LOAD or AUTO STEP option of the
history definition block.
The free energy function versus time data being used for large strain viscoelasticity can be generated by
fitting the experimental data in Marc Mentat or MD Patran provided the following two tests are done:
1. Standard quasi-static tests (tensile, planar-shear, simple-shear, equi-biaxial tension, volumertic)
0
to determine the elastomer free energy W constants.
2. Standard relaxation tests to obtain stress versus time.
Narayanaswamy Model
The annealing of flat glass requires that the residual stresses be of an acceptable magnitude, while the
specification for optical glass components usually includes a homogenous refractive index. The design
of heat treated processes (for example, annealing) can be accomplished using the Narayanaswamy
model. This allows you to study the time dependence of physical properties (for example, volumes) of
glass subjected to a change in temperature.
The glass transition is a region of temperature in which molecular rearrangements occur on a scale of
minutes or hours, so that the properties of a liquid change at a rate that is easily observed. Below the glass
transition temperature T g , the material is extremely viscous and a solidus state exists. Above T g , the
equilibrium structure is arrived at easily and the material is in liquidus state. Hence, the glass transition
is revealed by a change in the temperature dependence of some property of a liquid during cooling. If a
mechanical stress is applied to a liquid in the transition region, a time-dependent change in dimensions
results due to the phenomenon of visco-elasticity.
If a liquid in the transition region is subjected to a sudden change in temperature, a time-dependent
change in volume occurs as shown in Figure 7-62. The latter process is called structural relaxation.
Hence, structural relaxation governs the time-dependent response of a liquid to a change of temperature.
T1
T(t)
T2
t0 t
(a) Step Input for Temperature
V(0,T1)
αg(T2-T1)
V(0,T2) αl(T2-T1)
V(∞,T2)
t0 t
(b) Volume Change as Function of Temperature
Figure 7-62 Structural Relaxation Phenomenon
Suppose a glass is equilibrated at temperature T 1 , and suddenly cooled to T 2 at t 0 . The instantaneous

change in volume is α g ( T 2 – T 1 ) , followed by relaxation towards the equilibrium value V ( ∞, T 2 ) . The
total change in volume due to the temperature change is α 1 ( T 2 – T 1 ) as shown in Figure 7-62b. The
rate of volume change depends on a characteristic time called the relaxation time.
CHAPTER 7 501
Material Library
The slope of dV ⁄ dT changes from the high value characteristic of the fluid α 1 to the low characteristic
of the glass α g as shown in Figure 7-63. The glass transition temperature T g is a point in the center of
the transition region. The low-temperature slope α g represents the change in volume V caused by
vibration of the atoms in their potential wells. In the (glassy) temperature range, the atoms are frozen into
a particular configuration. As the temperature T increases, the atoms acquire enough energy to break
bonds and rearrange into new structures. That allows the volume to increase more rapidly, so α 1 > α g .
The difference α = α 1 – α g represents the structural contribution to the volume.
V(T)
V(T0) αl
Liquid
State
V(T1)
αg Transition Range
Solidus
State
T0
T2 T1 Tg Tf(T1) Tf (T1): Fictive Temperature
Figure 7-63 Property (Volume) – Temperature Plot
When a liquid is cooled and reheated, a hysteresis is observed as shown in Figure 7-64.
V
Equilibrium
Nonequilibrium
Tg T
Figure 7-64 Volume Change During Cyclic Temperature History
Unfortunately, the notion of a glass transition temperature is insufficient as real glassy materials
generally exhibit a temperature regime, called a transition range, across which their bulk properties
gradually change from being solid-like to liquid-like in nature.
As discussed earlier, properties have a time dependence in the transition range. An explanation for the
strong time dependence lies in that the material resides at a nonequilibrium temperature which lags
behind the applied temperature during the heating-cooling cycle. The nonequilibrium temperature is
called the fictive temperature, T f , as shown in Figure 7-63. The fictive temperature at T 1, T f ( T 1 ) is
found by extrapolating a line from V ( T 1 ) with slope α g to intersect a line extrapolated from V ( T 0 )
with slope α 1 (see Figure 7-63). For T ≤ T 2 (well below the glass transition), T f reaches a limiting
value that is called T g . If the material were equilibrated at T f ( T 1 ) , then instantaneously cooled to T 1 ,
it would change along the line with slope α g because no structural rearrangement could occur.
Therefore, it would have the same volume as the continuously cooled sample.
The response of the volume change can be described by:
V ( T 2 ,t ) = V ( T 1 ,∞ ) + α g ( T 2 – T 1 ) + ( α l – α g ) ( T f ( t ) – T 2 ) (7-280)
where T f ( t ) is the current value of the fictive temperature. The response function, M v , which dictates
the value of the fictive temperature is assumed to be linear in its argument and governs both the value of
the fictive temperature as well as the material property of interest.
V(t ) – V( ∞) Tf ( t ) – T ( ∞ )
--------------------------------- = M v ( ξ ( t ) ) = --------------------------------
- (7-281)
V(0 ) – V( ∞) T(0 ) – T( ∞)
By virtue of its linearity, Boltzmann’s superposition principle can be invoked to calculate the fictive
temperature at any time:
t
d
Tf ( t ) = T ( t ) –  M v ( ξ ( t ) – ξ ( t′ ) ) -------(T ( t′ )) ( dt' )
dt′
(7-282)
–∞
The concept of reduced time, ξ ( t ) , is introduced in the spirit of Thermo-Rheologically Simple materials
to capture the disparate nonlinear response curves on a single master curve.
The reduced time used in Marc is given by the following expression:
t
τ re f
ξ( t) =  --------------------
τ ( T ( t′ ) )
- dt′ (7-283)
–∞
Here τ ref is the reference relaxation time of the material evaluated at a suitable reference temperature,
T ref . The relaxation time τ at the given time and temperature can be represented as:
x (1 – x)
τ = τ re f ⋅ exp  – ---- ---------- – --- – ------------------ 
H 1
(7-284)
 R Tr e f T Tf 
The parameter x allows you to dictate how much of the fictive temperature participates in the
prescription of the relaxation time, and must, therefore, range between 0 and 1.
CHAPTER 7 503
Material Library
H is the activation energy for the particular process and R is the gas constant. A typical response
function is:
ξ
M v ( ξ ) = exp  – -- (7-285)
 τ
Multiple structural relaxation times can exist. The response function has, therefore, been
implemented as:
n
ξ
Mv ( ξ ) =  ( W g ) i ⋅ exp  – ----
 τ i
i = 1
For a complete description of the model, it is necessary to prescribe the following:
1. The weight ( W g ) i for each term in the series (usually  ( W g ) i ≈ 1 ).

2. The reference relaxation times τ i, ref .
3. The fraction parameter x and the activation energy-gas constant ratio.
4. The solid and liquid coefficients of thermal of expansion, α g and α 1 through the
VISCEL EXP option.
A stable algorithm is employed to calculate the convolution integrals. For improved accuracy it is
recommended that the time steps used during the simulation be sufficiently small.
In Figure 7-65, the volume of cube of material, which is allowed to contract freely and is experiencing a
100oC quench, is displayed.
ΔT=100
Volume
αgΔT
αlΔT
Temperature
e
Tim
Figure 7-65 Volume-Temperature-Time Plot
Temperature Effects and Coefficient of Thermal Expansion

Experimental results indicate that a large number of material properties vary with temperatures. In Marc,
almost all material parameters may be a function of temperatures when using the table driven input
procedure. A subset of which are shown in Table 7-4 for different types of analysis.
Please note that T is temperature in the above expressions. With the exception of heat transfer, Joule
heating, or coupled thermal-mechanical analysis, the temperature is a state variable.
CHAPTER 7 505
Material Library
7 Material Library
Table 7-4 Temperature-Dependent Material Properties

Analysis Type Material Properties
Stress Analysis Modulus of elasticity (Young’s Modulus) E ( T )
Poisson’s Ratio ν( T)
Yield Stress σy ( T )
Workhardening Slope h ( T )
Coefficient of Thermal Expansion α ( T )
Mooney Constants C 01 ( T ), C 10 ( T )
C 11 ( T ), C 20 ( T )
C 30 ( T )
Heat Transfer Analysis Thermal Conductivity K ( T )
Specific Heat C(T)
Emissivity ε(T)
Couples Thermo-Electrical Electric Resistivity π(T)
(Joule Heating Analysis)
Hydrodynamic Heating Viscosity μ(T )
Piecewise Linear Representation

In Marc, the temperature variation of a material constant F ( T ) may be entered as a piecewise linear
function of temperature as shown in Figure 7-66 or as an equation. The base value or reference value
is entered in the ISOTROPIC, ORTHOTROPIC, ANISOTROPIC, MOONEY, OGDEN, etc. model
definition options. Using the nontable input format, these should be at the lowest temperatures. The
variation with temperature is entered on the TEMPERATURE EFFECTS and ORTHO TEMP or TABLE
model definition options.
Temperature
Dependent
Property
F(T) F4
F3
S3
F2
S2
F1 S1
Base
Value
F1
T4
T1 T2 T3 Temperature (T)
(1) Slope-Break Point Data

Slope Break Point
S1 = (F2 - F1)/(T2 - T1) T1
S2 = (F3 - F2)/(T3 - T2) T2
S3 = (F4 - F3)/(T4 - T3) T3
(2) Function-Variable Data
Function Variable
F1 T1
F2 T2
F3 T3
F4 T4
Figure 7-66 Piecewise Linear Representation of Temperature-Dependent Material Properties
Temperature-Dependent Creep
In Marc, the temperature dependency of creep strain can be entered in two ways. The creep strain rate
·
may be entered as a piecewise linear function. If the creep strain ε c can be expressed in the form of a
power law
·c
ε = AT m (7-286)
where A and m are two experimental constants, input the experimental constants through the CREEP
model definition option.
CHAPTER 7 507
Material Library
For other temperature dependency, you may use an equation or use the CRPLAW user subroutine for
explicit creep and UCRPLW user subroutine for implicit creep to input the variation of creep strain
with temperature.
Coefficient of Thermal Expansion

Marc always uses an instantaneous thermal expansion coefficient definition
th
dε i j = α i j dT in general (7-287)
or
th
dε i j = α dT δ i j for the isotropic case (7-288)
In many cases, the thermal expansion data is given with respect to a reference temperature
εth = α( T – T0 ) (7-289)
where a is a function of temperature:
α = α( T) (7-290)
Clearly, in this case
dα
dε t h = α + ------- ( T – T 0 ) dT (7-291)
dT
so the necessary conversion procedure is:
1. Compute and plot Equation (7-289) in the form Equation (7-291)
dα
α = α + ------- ( T – T 0 ) (7-292)
dT
as a function of temperature.
2. Model Equation (7-292) in the ANEXP user subroutine, or with piecewise linear slopes and
breakpoints in the TEMPERATURE EFFECTS or TABLE option.
The anisotropic coefficient of thermal expansion can be input through either the ORTHOTROPIC
model definition option or the ANEXP user subroutine.
Time-Temperature-Transformation
Certain materials, such as carbon steel, exhibit a change in mechanical or thermal properties when
quenched or air cooled from a sufficiently high temperature. At any stage during the cooling process,
these properties are dependent on both the current temperature and the previous thermal history. The
properties are influenced by the internal microstructure of the material, which in turn depends on the rate
at which the temperature changes. Only in instances where the temperature is changed very gradually
does the material respond in equilibrium, where properties are simply a function of the current
temperature. In addition, during the cooling process certain solid-solid phase transformations can occur.
These transformations represent another form of change in the material microstructure which can
influence the mechanical or thermal properties. These transformations can be accompanied by changes
in volume.
The occurrence of phase change is also dependent on the rate of cooling of the material. This relationship
is shown in a typical cooling diagram (see Figure 7-67). The curves A, B, and C in Figure 7-67 represent
the temperature history of a structure that has been subjected to a different cooling rate. It is obvious that
the structural material experiences phase changes at different times and temperatures, depending on the
rate of cooling. Under cooling rate A, the material changes from phase 1 to phase 4 directly. The material
undergoes three phase changes (phase 1 to phase 2 to phase 3 to phase 4) for both cooling rates B and C.
However, the phase changes take place at different times and temperatures.
The Time-Temperature-Transformation (TIME-TEMP) option allows you to account for the
time-temperature-transformation interrelationships of certain materials during quenching or
casting analyses.
Use the T-T-T parameter to invoke the time-temperature-transformation. Input all the numerical data
required for this option through the TIME-TEMP model definition option.
In a transient heat transfer analysis, the thermal properties which can be defined as a function of time and
temperature are the thermal conductivity and the specific heat per unit reference mass. Here, the effects
of latent heat or phase transformation can be included through the definition of the specific heat.
In a thermal stress analysis, the mechanical properties which can be defined as a function of time and
temperature are the Young’s modulus, Poisson’s ratio, yield stress, workhardening slope, and coefficient
of thermal expansion. The effects of volumetric change due to phase transformation can be included
through the definition of the coefficient of thermal expansion.
A T1 T1
B C
T4 Line of Phase Change –

Change in Volume
Temperature (T)
T5
T3 T3
T2 T2
t1 t2 Time (t)
Figure 7-67 Simplified Cooling Transformation Diagram

CHAPTER 7 509
Material Library
Test data must be available in a tabular form for each property of each material group. For a given cooling
rate, the value of a property must be known at discrete points over a range of temperatures. There can be
several sets of these discrete points corresponding to measurements at several different cooling rates. The
cooling tests must be of a specific type known as Newton Cooling; that is, the temperature change in the
material is controlled such that
T ( t ) = A exp ( – at ) + B (7-293)
In addition, a minimum and a maximum temperature that bracket the range over which the TIME-TEMP
option is meant to apply must also be given.
For the simulation of the cooling rate effect in finite element analysis, material properties of a structure
can be assumed as a function of two variables: time and temperature. Two-dimensional interpolation
schemes are used for the interpolation of properties.
Interpolation is based on making the time variable discrete. Stress analysis is carried out incrementally
at discrete time stations and material properties are assumed to vary piecewise linearly with temperature
at any given time. These temperature-dependent material properties are updated at each increment in the
analysis. For illustration, at time t 1 , the material is characterized by the phase 1 and phase 4 behaviors
at temperature ranges T 1 to T 3 , and T 3 to T 2 , respectively (see Figure 7-68). Similarly, at time t 2 , the
material behavior must be characterized by all four phases, each in a different temperature range (that is,
phase 1, T 1 to T 4 ; phase 2, T 4 to T 5 ; phase 3, T 5 to T 3 ; phase 4, T 3 to T 2 ). The selection of an
interpolation scheme is generally dependent on the form of the experimental data. A linear interpolation
procedure can be effectively used where the properties are expressed as a tabulated function of time
and temperature.
During time-temperature-transformation, the change in volume in a stress analysis is assumed to take
place in a temperature range ΔT . The change in volume is also assumed to be uniform in space, such
that the effect of the volume change can be represented by a modification of the coefficient of thermal
expansion. For a triangular distribution of α ( T ) in the temperature range ΔT , the value of the modified
coefficient of thermal expansion is
2
α m = ------- [ 3 1 + φ – 1 ] (7-294)
ΔT
where φ is the change in volume. A schematic of the modified α ( T ) is shown in Figure 7-68.
ΔT
Coefficient of thermal Expansion (α)
Temperature (T)
αm
Figure 7-68 Modified Coefficient of Thermal Expansion for Short-Time Change in Volume
Low Tension Material

Marc can handle concrete and other low tension material. The CRACK DATA option assists in predicting
crack initiation and in simulating tension softening, plastic yielding and crushing. This option can be used
for the following:
• Elements with a one-dimensional stress-strain relation (beam and truss elements)
• Elements with a two-dimensional stress-strain relation (plane stress, plane strain, axisymmetric,
and shell elements)
• Three-dimensional elements (bricks)
Analytical procedures that accurately determine stress and deformation states in concrete structures are
complicated by several factors. Two such factors are the following:
• The low strength of concrete in tension that results in progressive cracking under increasing loads
• The nonlinear load-deformation response of concrete under multiaxial compression
Because concrete is mostly used in conjunction with steel reinforcement, an accurate analysis requires
consideration of the components forming the composite structure. Steel reinforcement bars are
introduced as rebar elements.
Each rebar element must be input with a separate element number. The REBAR model definition option
or REBAR user subroutine is used to define the orientation of the reinforcement rods.
Uniaxial Cracking Data

The cracking option is accessed through the ISOTROPIC option. Uniaxial cracking data can be specified
using the CRACK DATA option or the UCRACK user subroutine. When the CRACK DATA option is used
to specify uniaxial cracking data, the following must be specified: the critical cracking stress, the
modulus of the linear strain softening behavior, and the strain at which crushing occurs. Material
properties, such as Young’s modulus and Poisson’s ratio, are entered using the ISOTROPIC option and
CHAPTER 7 511
Material Library
the WORK HARD option. This model is for a material which is initially isotropic; if the model is initially
orthotropic, see FAIL DATA for an alternative cracking model. A typical uniaxial stress-strain diagram is
shown in Figure 7-69.
σcr
Es
εcrush
ε
E Young’s Modulus
Es Tension-Softening Modulus
σy σy Yield Stress
σcr Critical Cracking Stress
Workhardening εcrush Crushing Strain
Figure 7-69 Uniaxial Stress-Strain Diagram
Low Tension Cracking

A crack develops in a material perpendicular to the direction of the maximum principal stress if the
maximum principal stress in the material exceeds a certain value (see Figure 7-70). After an initial crack
has formed at a material point, a second crack can form perpendicular to the first. Likewise, a third crack
can form perpendicular to the first two. The material loses all load-carrying capacity across the crack
unless tension softening is included.
Tension Softening
If tension softening is included, the stress in the direction of maximum stress does not go immediately
to zero; instead the material softens until there is no stress across the crack. At this point, no load-carrying
capacity exists in tension (see Figure 7-70). The softening behavior is characterized by a descending
branch in the tensile stress-strain diagram, and it may be dependent upon the element size.
σ2
y
σ1
θ
σ1
σ2
Figure 7-70 Crack Development
Crack Closure
After a crack forms, the loading can be reversed; therefore, the opening distance of a crack must be
considered. In this case, the crack can close again, and partial mending occurs. When mending occurs, it
is assumed that the crack has full compressive stress-carrying capability and that shear stresses are
transmitted over the crack surface, but with a reduced shear modulus.
Crushing
As the compressive stress level increases, the material eventually loses its integrity, and all load-carrying
capability is lost; this is referred to as crushing. Crushing behavior is best described in a multiaxial stress
state by a crushing surface having the same shape as the yield surface. The failure criterion can be used
for a two-dimensional stress state with reasonable accuracy. For many materials, experiments indicate
that the crushing surface is roughly three times larger than the initial yield surface.
Analysis
The evolution of cracks in a structure results in the reduction of the load carrying capacity. The internal
stresses need to be redistributed through regions that have not failed. This is a highly nonlinear problem
and can result in the ultimate failure of the structure. The AUTO INCREMENT or AUTO STEP option
should be used to control the applied load on the structure.
Soil Model
Soil material modeling is considerably more difficult than conventional metals, because of the
nonhomogeneous characteristics of soil materials. Soil material usually consists of a large amount of
random particles. Soils show unique properties when tested. The bulk modulus of soil increases upon
pressing. Also, when the preconsolidation stress is exceeded, the stiffness reduces dramatically while
the stiffness increases upon unloading. At failure, there is no resistance to shear, and stiff clays or dense
sands are dilatant. Over the years, many formulations have been used, including linear elastic, nonlinear
elastic, Drucker-Prager or Mohr Coulomb, and Cam-Clay and variations thereof. In Marc, the material
models for available soil modeling are linear elasticity, nonlinear elasticity, and the Cam-Clay model.
CHAPTER 7 513
Material Library
Elastic Models
Linear elasticity is defined in the conventional manner, defining the Young’s moduli and the Poisson’s
ratio in the SOIL option. The nonlinear elasticity model is implemented through the NLELAST option or
the HYPELA2 user subroutine. Some of the simplest models include the bilinear elasticity, where a
different moduli is used during the loading and unloading path, or to represent total failure when a critical
stress is obtained.
A more sophisticated elastic law is the hyperbolic model, where six constants are used. In this model, the
tangent moduli are
R f ( 1 – sin φ ) ( σ 1 – σ 3 ) 2 σ n
E = 1 – ---------------------------------------------------------  -----3- (7-295)
2c cos φ + 2σ 3 sin φ  pa
Two of the elastic models are the E-ν and K-G variable elastic models. In the E-ν model, Poisson’s ratio
is considered constant and
E = E0 + αE p + βE τ (7-296)
while, in the K-G model
K = K0 + αK p (7-297)
G = G0 + αG p + βG τ (7-298)
The key difficulty with the elastic models is that dilatancy cannot be represented.
Cam-Clay Model
The Cam-Clay model was originally developed by Roscoe, and then evolved into the modified Cam-
Clay model of Roscoe and Burland. This model, which is also called the critical state model, is
implemented in Marc.
The yield surface is an ellipse in the p,τ plane as shown in Figure 7-71, and is defined by
τ2
F = ---------- – 2pp c + p 2 = 0 (7-299)
M c2s
where p c is the preconsolidation pressure, and M c s is the slope of the critical state line.
The Cam-Clay model has the following no properties. At the intersection of the critical state line and the
ellipse, the normal to the ellipse is vertical. Because an associated flow rule is used, all plastic strain at
failure is distortional; the soil deforms at constant volume (Figure 7-71). The strain hardening and
softening behavior are shown in Figure 7-72. Also, if the preconsolidation pressure is large, the soil
remains elastic for large stresses. The evolution of the preconsolidation pressure is
·
p· c = – θp c tr ( ε p l ) (7-300)
where
1+e
θ = ------------- (7-301)
λ–κ
where
e is the void ratio
λ is the virgin compression index (see Figure 7-73)
κ is the recompression index (see Figure 7-73)
τ
(εq)
Critical State Line
Strain Hardening
Ceases
Strain Softening
Ceases 2
1
3
Region
originally
4 elastic
P ( ε v )p
CO
Figure 7-71 Modified Cam-Clay Yield Surface
q q
1 3
q2
4
q4
εq εq
Figure 7-72 Strain Hardening and Softening Behavior

CHAPTER 7 515
Material Library
1 κ 1
λ
ln ( – p )
Figure 7-73 Response of Idealized Soil to Hydrostatic Pressure
The void ratio and the porosity are related by the expression
e
φ = ------------ (7-302)
1+e
In the modified Cam-Clay model, it is also assumed that the behavior is nonlinear elastic with a constant
Poisson’s ratio, the bulk modulus behave as:
1+e
K = – ------------ p (7-303)
κ
Note that this implies that at zero hydrostatic stress, the bulk modulus is also zero. To avoid
computational difficulties, a cutoff pressure of one percent of the preconsolidation pressure is used.
This constitutive law is implemented in Marc using a radial return procedure. It is available for either
small or large strain analysis.When large displacements are anticipated, you should use the LARGE
STRAIN parameter.
The necessary parameters for the Cam-Clay soil model can be obtained by the following experiments:
1. Hydrostatic test: determines volumetric elastic bulk modulus, yield stress, and the virgin and
recompression ratio of soils.
2. Shear-box test: determines the slope of critical state line and shear modulus of the soil. However,
the tests have to be calibrated with numerical simulations to get the necessary constants.
3. Triaxial shear test: the most comprehensive experimental information and obviates the need for
the first two tests and obtains all the necessary constants listed above.
Evaluation of Soil Parameters for the Critical State Soil Model

To illustrate how to extract soil parameters, namely M , κ , and λ , a hypothetical data set for a normally
consolidated clay in presented in Figures 7-74 to 7-80. Now we shall use these data to show the
procedure of determination of parameters for the critical state model. The data presented here pertain to
conventional triaxial conditions.
Figures 7-74 and 7-75 show the stress-strain relations for constant pressure tests under different initial
conditions p0=10 (69), 20 (139), and 30 (207) psi (kPa), respectively. Here the mean effective pressure
of the soil sample is kept constant throughout the test. The last data point shown on the deviatoric stress
(q) versus axial strain (ε1) plot for any test is considered as the ultimate condition for that test. The void
ratio values at the beginning and the end of each test are given in Table 7-5.
Figures 7-77 to 7-79 show stress-strain relation plots for three fully drained tests performed at initial
pressures 10, 20, and 30 psi. Here, there is no pore pressure development, and the effective mean pressure
increases during the test. The void ratio values at the beginning and end of each test are given in
Table 7-5.
Figures 7-80 to 7-82 show stress-strain behavior under undrained conditions. Here there is no change in
volume of the sample. However, fluid pore pressure are developed during the test, thereby reducing the
effective stresses until the ultimate (failure) state is reached.
Figure 7-83 shows the variation of void ratio with mean pressure which is obtained from a hydrostatic
compression (HC) test with three load reversals (that is, unloading-reloading cycles). In this figure, void
ratio is plotted to a linear scale while the hydrostatic stress is plotted to a logarithmic (base 10) scale. In
fact, this is the usual way of presenting one-dimensional consolidation or hydrostatic test results in
geotechnical engineering practices.
Determination of parameter, M: The parameter M is the slope of the critical state line on a p-q plot. To
determine its value, the values of p and q at ultimate conditions for each test are plotted as shown in
Figure 7-84; the ultimate condition for each test is assumed to be the last point plotted in Figures 7-74 to
7-82. At the ultimate conditions, the sample undergoes excessive deformations under constant deviatoric
stress, and hence it could be taken as the asymptotic stress to the curve. The slope of the critical state line
(Figure 7-84) is calculated as 1.0 for the soil above. That is, M = 1.0 .
Table 7-5 Test Values for Cam Clay Model

Initial Final
Pressure, p0 Initial Pressure, pf Final
(effective) void Ratio, (effective) Void Ratio,
Test (psi) e0 (psi) e1
p-constant 10 1.080 10 0.980
p-constant 20 0.959 20 0.860
p-constant 30 0.889 30 0.787
Drained 10 1.080 15 0.908
Drained 20 0.959 30 0.787
Drained 30 0.889 45 0.716
Undrained 10 1.080 05.55 1.080
Undrained 20 0.959 11.09 0.959
Undrained 30 0.889 16.64 0.889
CHAPTER 7 517
Material Library
Deviatoric stress,q = σ 1 – σ 3 (psi)
10
0 0.10 0.20
Axial strain, ε1
0.10 0.20
0.01
Volumetric strain, ε ν
0.02
0.03
0.04
0.05
Figure 7-74 Constant Pressure Test: p0 = 10 psi

Deviatoric stress,q = σ 1 – σ 3(psi)
25
20
15
10
0 0.10 0.20
Axial strain, ε1
0.10 0.20
0.01
Volumetric strain, εv
0.02
0.03
0.04
0.05

Deviatoric stress, q = σ 1 – σ 3 (psi)
30
25
20
15
10
0.10 0.20
Axial strain, ε1
0.10 0.20
0.01
εν Volumetric strain,
0.02
0.03
0.04
0.05

16
14
12
10
8
6
4
2
0.10 0.20
Axial strain, ε1
0.10 0.20
ν
Volumetric strain, ε
0.02
0.04
0.06
0.08
0.10
Figure 7-77 Drained Test: p0 = 10 psi

CHAPTER 7 519
Material Library
Deviatoric stress, q = σ – σ (psi)

3
30
1
25
20
15
10
5
0.10 0.20
Axial strain, ε1
0.10 0.20
εν
0.02
Volumetric strain,
0.04
0.06
0.08
0.10

Deviatoric stress, q = σ 1 – σ 3(psi)
50
40
30
20
10
0.10 0.20
Axial strain, ε1
0.10 0.20
0.02
εν
Volumetric strain,
0.04
0.06
0.08
0.10

0.01 0.02 0.03 0.04

Axial strain, ε1
Pore pressure, u (psi)
0.01 0.02 0.03 0.04

Axial strain, ε1
Figure 7-80 Undrained Test: p0 = 10 psi

12
10
8
6
4
2
0.01 0.02 0.03 0.04

Axial strain, ε1
10
8
6
4
2
0.01 0.02 0.03 0.04

Axial strain, ε1

CHAPTER 7 521
Material Library
Deviatoric stress, q = σ – σ (psi)
16
3
14
1
12
10
8
6
4
2
0
0.01 0.02 0.03 0.04
Axial strain, ε1
16
14
12
10
8
6
4
2
0.01 0.02 0.03 0.04

Axial strain, ε1
1.2
1.1 A
Void ratio, e
1.0 Cc
0.9
C B
0.8
Cs
0.7
10 20 30 40 50
Pressure, p (psi) (logarithmic scale with base 10)
Figure 7-83 Hydrostatic Compression Test: Cc = 0.04, Cs = 0.06

45
40
Deviatoric stress, q (psi)
30
Drained
Constant pressure
Undrained
20
10 Slope = M = 1.0
0 M
10 20 30 40 50
Mean Pressure, p (psi)
Figure 7-84 Critical State Line in q-p (psi)
Determination of parameters λ and κ: The values of gamma and kappa can be related to the commonly
known quantities such as compression index ( C c ) and swelling index ( C s ). The compression index, C c ,
is defined as the slope of virgin loading line on e-log10p plot while the swelling index, C s , is defined as
the unloading-reloading curves on the same plot. Usually, the compression index and swelling index are
defined with respect to a one-dimensional consolidation test. However, it can be shown that the e-ln(p)
curve for any constant stress ratio test, that is, for constant q/p ratio, is parallel to that obtained from a
hydrostatic test, (Figure 7-85).
1.2
e0 vs. p0
ef vs. pf
1.1
Void ratio, e
1.0
Hydrostatic loading
0.9
0.8 At critical state
0.7
20 30 40 50 60 70 80 90
Pressure (psi) (logarithmic scale with base 10)
Figure 7-85 Isotropic Consolidation (Data from 5-8)
In fact, one-dimensional hydrostatic test is parallel to that obtained under critical state conditions. The
values of gamma and kappa can be related to C c and C s as follows. The virgin compression line can be
expressed as
CHAPTER 7 523
Material Library
e – e 0 = C c log 10  -----
p
(7-304)
 p 0
or
e – e 0 = λ ln  -----
p
(7-305)
 p 0
and the swelling line (unloading-reloading) can be expressed as
e – e 0 = C s log 10  -----
p
(7-306)
 p 0
or
e – e 0 = κ ln  -----
p
(7-307)
 p 0
Therefore, comparing Equations (7-304) and (7-305), we have

Cc Cc
λ = ------------- = ------------- (7-308)
ln 10 2.303
and comparing Equations (7-306) and (7-307) yields
Cs Cs
κ = ------------- = ------------- (7-309)
ln 10 2.303
The value of C c can be computed by considering two points, A and B, in Figure 7-83 as
eA – eB
C c = --------------------------------------
log p B – log p A
(7-310)
1.08 – 0.08
= -------------------------------------
log 50 – log 10
= 0.40
Here the subscript denotes the value at that point. The swelling index, C s , can be computed by
considering points B and C of the same figures
eC – eB
C s = --------------------------------------
log p B – log p C
0.842 – 0.80 (7-311)
= -------------------------------------
log 50 – log 10
= 0.06
Hence, the values of gamma and kappa can be computed from Equations (7-308) and (7-309) as
0.40
λ = ------------- = 0.174 (7-312)
2.303
Damage Models
In many structural applications, the finite element method is used to predict failure. This is often
performed by comparing the calculated solution to some failure criteria, or by using classical fracture
mechanics. Previously, we discussed two models where the actual material model changed due to some
failure, see “Progressive Composite Failure” on page 386 and the previous section on “Low Tension
Material” on page 510. In this section, the damage models appropriate for ductile metals and elastomeric
materials will be discussed.
Ductile Metals
In ductile materials given the appropriate loading conditions, voids will form in the material, grow, then
coalesce, leading to crack formation and potentially, failure. Experimental studies have shown that these
processes are strongly influenced by hydrostatic stress. Gurson studied microscopic voids in materials
and derived a set of modified constitutive equations for elastic-plastic materials. Tvergaard and
Needleman modified the model with respect to the behavior for small void volume fractions and for
void coalescence.
In the modified Gurson model, the amount of damage is indicated with a scalar parameter called the void
volume fraction f. The yield criterion for the macroscopic assembly of voids and matrix material is given
by:
σ 2  q 2 σ k k
F =  ------ + 2q 1 f∗ cosh  --------------- – [ 1 + ( q 1 f∗ ) 2 ] = 0 (7-313)
 σ y  2σ y 
as seen in Figure 7-86.
σe ⁄ σM
1.0
·
f* = 0
·
0.5 f* ⁄ f u* = 0.01
0.1
0.3
0.6
0.9
0
0 1 2 3 4 σ k k ⁄ 3σ M
Figure 7-86 Plot of Yield Surfaces in Gurson Model
CHAPTER 7 525
Material Library
7 The parameter q 1 was introduced by Tvergaard to improve the Gurson model at small values of the void
volume fraction. For solids with periodically spaced voids, numerical studies [Ref. 10] showed that the
values of q 1 = 1.5 and q 2 = 1 were quite accurate.
Material The evolution of damage as measured by the void volume fraction is due to void nucleation and growth.
Library Void nucleation occurs by debonding of second phase particles. The strain for nucleation depends on the
particle sizes. Assuming a normal distribution of particle sizes, the nucleation of voids is itself modeled
as a normal distribution in the strains, if nucleation is strain controlled. If void nucleation is assumed to
be stress controlled in the matrix, a normal distribution is assumed in the stresses. The original Gurson
model predicts that ultimate failure occurs when the void volume fraction f, reaches unity. This is too
high a value and, hence, the void volume fraction f is replaced by the modified void volume fraction f∗
in the yield function.
The parameter f∗ is introduced to model the rapid decrease in load carrying capacity if void
coalescence occurs.
f∗ = f if f ≤ fc
 f u* – f c (7-314)
f∗ = f c +  ---------------- ( f – f c ) if f > fc
 f F – f c
where f c is the critical void volume fraction, and f F is the void volume at failure, and f u* = 1 ⁄ q 1 . A
safe choice for f F would be a value greater than ( 1 ⁄ q 1 ) namely, f F = 1.1 ⁄ q 1 . Hence, you can
control the void volume fraction, f F , at which the solid loses all stress carrying capability.
Numerical studies show that plasticity starts to localize between voids at void volume fractions as low
as 0.1 to 0.2. You can control the void volume fraction f c , beyond which void-void interaction is
modeled by Marc. Based on the classical studies, a value of f c = 0.2 can be chosen.
The existing value of the void volume fraction changes due to the growth of existing voids and due to
the nucleation of new voids.
· · ·
f = f g r o wt h + f n u c l e a t i o n (7-315)
The growth of voids can be determined based upon compressibility of the matrix material surrounding
the void.
· ·p
f g r o w t h = ( 1 – f )ε k k (7-316)
As mentioned earlier, the nucleation of new voids can be defined as either strain or stress controlled. Both
follow a normal distribution about a mean value.
In the case of strain controlled nucleation, this is given by
1 εm – εn
fN p 2
exp – ---  ------------------- ε m
· ·p
f n u c l e a t i o n = -------------- (7-317)
S 2π 2 S 
where f N is the volume fraction of void forming particles, ε n the mean strain for void nucleation and S
the standard deviation.
In the case of stress controlled nucleation, the rate of nucleation is given by:
fN 1 Σ – σn 2
f n u c l e a t i o n = -------------- Σ exp – ---  ----------------
·
(7-318)
S 2π 2 S 
1
where Σ = σ + --- σ k k .
3
If the second phase particle sizes in the solid are widely varied in size, the standard deviation would be
larger than in the case when the particle sizes are more uniform. The Marc user can also input the volume
fraction of the nucleating second phase void nucleating particles in the input deck, as the variable f N .
A typical set of values for an engineering alloy is given by Tvergaard for strain controlled nucleation as
ε n = 0.30 ; f N = 0.04 ; S = 0.01 . (7-319)
It must be remarked that the determination of the three above constants from experiments is extremely
difficult. The modeling of the debonding process must itself be studied including the effect of differing
particle sizes in a matrix. It is safe to say that such an experimental study is not possible. The above three
constants must necessarily be obtained by intuition keeping in mind the meaning of the terms.
When the material reaches 90 percent of f F , the material is considered to be failed. At this point, the
stiffness and the stress at this element are reduced to zero.
Elastomers
Under repeated application of loads, elastomers undergo damage by mechanisms involving chain
breakage, multi-chain damage, micro-void formation, and micro-structural degradation due to
detachment of filler particles from the network entanglement. Two types of phenomenological models
namely, discontinuous and continuous, exists to simulate the phenomenon of damage.
1. Discontinuous Damage:
The discontinuous damage model simulates the “Mullins’ effect” as shown in Figure 7-87.
CHAPTER 7 527
Material Library
Figure 7-87 Discontinuous Damage
This involves a loss of stiffness below the previously attained maximum strain. The higher the
maximum attained strain, the larger is the loss of stiffness. Upon reloading, the uniaxial stress-
strain curve remains insensitive to prior behavior at strains above the previously attained
maximum in a cyclic test. Hence, there is a progressive stiffness loss with increasing maximum
strain amplitude. Also, most of the stiffness loss takes place in the few earliest cycles provided
the maximum strain level is not increased. This phenomenon is found in both filled as well as
natural rubber although the higher levels of carbon black particles increase the hysteresis and the
loss of stiffness. The free energy, W , can be written as:
0
W = K ( α, β )W (7-320)
0
where W is the nominal strain energy function, and
0
α = max ( W ) (7-321)
determines the evolution of the discontinuous damage. The reduced form of Clausius-Duhem
dissipation inequality yields the stress as:
0
∂W
S = 2K ( α ,β ) ----------- (7-322)
∂C
where C is the deformation tensor.
Mathematically, the discontinuous damage model has a structure very similar to that of strain
space plasticity. Hence, if a damage surface is defined as:
Φ = W–α≤0 (7-323)
The loading condition for damage can be expressed in terms of the Kuhn-Tucker conditions:
· ·
Φ≤0 α≥0 αΦ = 0 (7-324)
The consistent tangent can be derived as:

2 0 0 0
∂ W ∂K ∂W ∂W
D = 4 K --------------- + ----------- ----------- ⊗ ----------- (7-325)
∂C∂C ∂W 0 ∂C ∂C
2. Continuous Damage:
The continuous damage model can simulate the damage accumulation for strain cycles for which
the values of effective energy is below the maximum attained value of the past history as shown
in Figure 7-88.
This model can be used to simulate fatigue behavior. More realistic modeling of fatigue would
require a departure from the phenomenological approach to damage. The evolution of continuous
damage parameter is governed by the arc length, s , of the effective strain energy as:
t
∂ 0
β =  ----- W ( s ) ds
∂s
(7-326)
0
Hence, β accumulates continuously within the deformation process.
Figure 7-88 Continuous Damage
The Kachanov factor K ( α, β ) is implemented in Marc through both an additive as well as a

multiplicative decomposition of these two effects as:
2 2
α β
exp  – ------ + d n exp  – -----
∞ α β
K ( α, β ) = d +  dn
 η n   λ n
(7-327)
n = 1 n = 1
2
α + δnβ
d n exp  – --------------------
∞
K ( α, β ) = d +   ηn 
(7-328)
n = 1
CHAPTER 7 529
Material Library
α β ∞ ∞
You specify the phenomenological parameters d n , d n , η n, λ n, d n, δ n and d . If d is not
defined, it is automatically determined such that, at zero values of α and β , the Kachanov factor
K = 1 . If, according to Equation (7-326) or Equation (7-327) the value of K exceeds 1, K is
set back to 1.
The above damage model is available for deviatoric behavior and is flagged by means of the OGDEN
and DAMAGE model definition options. If, in addition, viscoelastic behavior is desired, the
VISCELOGDEN option can be included. Finally, the UELDAM user subroutine can be used to define
damage functions different from Equations (7-324) to (7-327).
The parameters required for the continuous or discontinuous damage model can be obtained using the
experimental data fitting option in Marc Mentat and MD Patran.
Cohesive Zone Modeling

Marc has a library of so-called interface elements (186, 187, 188, 189, 190, 191, 192,and 193), which
can be used to simulate the onset and progress of delamination. The constitutive behavior of these
elements is expressed in terms of tractions versus relative displacements between the top and bottom
edge/surface of the elements (see Figure 7-89).
8
top face
4
v1 7
˜ v3
˜ 3
5
v2 bottom face
1 ˜
6
2
Figure 7-89 3-D Linear Interface Element
Considering a 3-D interface element, the relative displacement components are given by one normal and
two shear components, expressed with respect to the local element system (see Marc Volume B: Element
Library for the definition of the local element systems):
top bottom
vn = u1 – u1
top bottom (7-329)
vs = u2 – u2
t op bottom
vt = u3 – u3
Based on the relative displacement components, the effective opening displacement is defined as:
2 2 2
v = vn + vs + vt (7-330)
Later on, some modifications of this definition will be discussed.

The effective traction t is introduced as a function of the effective opening displacement and is
characterized by an initial reversible response followed by an irreversible response as soon as a critical
effective opening displacement v c has been reached. The irreversible part is characterized by increasing
damage ranging from 0 (onset of delamination) to 1 (full delamination).
Three standard functions are currently available; namely, a bilinear, an exponential, and a linear-
exponential function (see Figure 7-90):
2G c v if 0 ≤ v ≤ vc
t = ---------- -----
vm vc
2G c  v m – v  if vc < v ≤ vm
t = ----------  ------------------- Bilinear (7-331)
v m  v m – v c
t = 0 if v > vm
v – v ⁄ vc
t = G c ----- e Exponential (7-332)
2
vc
2qG c v if 0 < v ≤ vc
t = ------------------------ -----
vc ( q + 2 ) vc
Linear-exponential (7-333)
2qG c q ( 1 – v ⁄ v c ) if v > vc
t = ------------------------ e
vc ( q + 2 )
in which G c is the energy release rate (cohesive energy), v m is the maximum effective opening
displacement (which is only used by the bilinear model), and q is the exponential decay factor (which is
only used by the linear-exponential model).
t t t
v v v
vc vm vc vc
Figure 7-90 Bilinear (left), Exponential (middle), and Linear-exponential (right)
Cohesive Material Model
CHAPTER 7 531
Material Library
It can easily be verified that the maximum effective traction t c , corresponding to the critical effective
opening displacement v c is given by:
2G c
t c = ---------- Bilinear (7-334)
vm
Gc
t c = -------- Exponential (7-335)
ev c
2qG c
t c = ------------------------ Linear-exponential (7-336)
vc ( q + 2 )
So if the maximum effective traction is known, the critical or maximum effective opening displacement
can be determined by:
2G c
v m = ---------- Bilinear (7-337)
tc
Gc
v c = ------- Exponential (7-338)
et c
2qG c
v c = ---------------------- Linear-exponential (7-339)
tc ( q + 2 )
Note that for the bilinear model, the critical effective opening displacement does neither affect the
cohesive energy nor the maximum effective traction.
Until now, the behavior in the normal and shear direction is treated similarly. However, sometimes the
behavior of an interface material may be different in tension and shear. The first method to include such
differences is incorporated by the shear-normal stress ratio β 1 , which defines the ratio of the maximum
stress in shear and the maximum stress in tension [Ref. 26]. This ratio is used to redefine the effective
opening displacement according to:
2 2 2 2 2
v = vn + β1 v s + β1 v t
The effect of β 1 = 0.5 is depicted in Figure 7-91 for the bilinear model.
t t
tension shear
only only
Gc Gc
vc vm v vc vm v
Figure 7-91 Response in Tension and Shear for a Shear-normal Stress Ratio β 1 = 0.5
(Bilinear Model)
Although the use of the shear-normal stress ratio offers some flexibility, it assumes that the cohesive
energy in tension and shear is the same. If one wants to define a different value of the cohesive energy in
shear than in tension, the shear-normal energy ratio β 2 can be used. In a general state of deformation,
when β 2 ≠ 1 , the curve defining the effective traction versus the effective opening displacement is
defined as a linear combination of the response in pure tension and pure shear. Using β 1 = 0.5 and
β 2 = 0.75 , Figure 7-92 shows the response in tension and shear for the bilinear model.
t t
tension shear
only only
Gc 0.75G c
vc vm v vc vm v
Figure 7-92 Response in Tension and Shear for a Shear-normal Stress Ratio β 1 = 0.5 and a
Shear-normal Energy Ratio β 2 = 0.75 (Bilinear Model)
In order to avoid convergence problems in a finite element simulation of delamination, one may activate
so-called viscous energy dissipation. The basic idea of the dissipation model is that when delamination
starts, the rate of deformation may suddenly increase. This increase is used to augment the constitutive
behavior with a viscous contribution being equivalent to this rate of deformation:
ζt c v·
t v i s = ---------- (7-340)
v· r
CHAPTER 7 533
Material Library
in which ζ is the viscous energy factor, v· is effective opening displacement rate and v· 0 is the reference
value of the effective opening displacement rate. This reference value can either be user-defined or
calculated by the program. In the latter case, the reference value is given by the maximum effective
opening displacement rate in any interface element, as long as the response in all the interface elements
is reversible. The viscous energy dissipation model does not directly have a physical background, but is
basically numerical in nature.
In the equations discussed above, no distinction has been made between tensile and compressive loading
in the normal direction. Assuming that in compression the behavior will remain reversible, Equation
(7-341) will be adapted as:
2 2 2
v = [ max ( v n, 0 ) ] + v s + v t (7-341)
Since, irrespective of the damage level, the interface elements should be able to sustain ongoing loading
in compression (so that inter-penetration is prohibited), it is possible to make the stiffness in compression
a function of the corresponding (negative) opening displacement. By default, the stiffness in
compression is constant and given by the slope of the traction versus opening displacement curve at the
origin. If a non-default value of the stiffening factor in compression F is given, the stiffness at v = – v c
is given by:
∂t ∂t n
-------n- = F --------
∂v n –v c ∂v n 0
So far, the constitutive behavior has been discussed in terms of an effective traction versus an effective
opening displacement. The traction components follow from the effective traction according to:
∂v ∂v
t n = t --------- ; t s ,t = t ----------- (7-342)
∂v n ∂v s ,t
As an alternative to the above mentioned standard models, the UCOHESIVE user subroutine can be used
to enter a user-defined material behavior.
There is a corresponding thermal cohesive model to what was discussed above which uses heat transfer
cohesive elements. In this model, the thermal conductivity can be a function of the displacement
opening. The thermal gradient through the thickness is based upon the temperatures along the top and
bottom surfaces. The UCOHESIVET user subroutine may be used to provide a more complex thermal
conductivity.
Lemaitre Model
The Lemaitre damage model is a phenomenological approach to ductile damage in ferrous materials that
are subject to large plastic deformations as they occur in the manufacturing processes. The model is
based on the thermodynamic dissipation potential of the material where ductile damage is considered as
a specific energy that is released when macroscopic fracture occurs. Without going further into detail the
mathematical derivation can be found in: Lemaitre, J.: A Course on Damage Mechanics, 2nd Ed.,
Springer Verlag, Berlin, 1996. In the following paragraphs, some basic formulas are given that serve for
the calculation and the interpretation of the damage values. The material parameters can be derived by
uniaxial tensile tests for the assumed forming conditions (strain rate, temperature). Some standard values
are given in this context, more information can be found in the mentioned literature.
The Lemaitre damage model calculates three damage values which have different meanings.
Macroscopic damage is characterized by plastic deformation that leads to pore growth, pore coalescence
and final rupture of the material matrix. The damage growth begins approximately after an equivalent
plastic strain threshold, ε d . This is the first material parameter to be defined in an experiment. For mild
steels, it is assumed to be between 0.1 and 0.2. The so-called absolute damage D represents the ductile
damage growth in the material. The incremental damage law is given as follows:
2
f(η) ⋅ σ
dD = ------------------------------------------- dε p , 0≤D≤1
2E ⋅ S ⋅ ( 1 – D ) 2
where the triaxiality function f ( η ) contains information about the state of stresses and is defined as
follows:
2 σm
f ( η ) = --- ( 1 + v ) + 3 ( 1 – 2v )η 2 η = -------
3 σ
σ is the von Mises stress, E the Young’s modulus, v the Poisson’s ratio, σ m the mean normal stress, D
the current integrated value of the absolute damage at that material point and dε p the effective plastic
strain increment. For most steels one can assume a maximum value of D from 0.15 to 0.4 at fracture.
Copper might even reach D = 0.9 . The more ductile the material, the higher D becomes. S is called
the damage resistance factor, a material parameter to be determined from tensile tests. S is from 1 to 8
according to the ductility of the material (1 low, 8 high). This parameter influences mostly the growth of
D and can also be determined by data correlation (for example, simulate the material test then derive the
correct material parameter).
The critical damage D c is used to compare the ductile damage D with the “state” of the material; such
as, whether the actual conditions (stresses, strains, state of stresses, already reached damage, etc.) might
be critical for macroscopic failure:
2
σU 2
Dc = D1 c -(1 – D)
-------------------------- 1 ≥ Dc ≥ 0
2
( σ f( η))
D 1 c is the critical damage in the uniaxial loadcase, the third material parameter to be determined by
tensile tests. Most steels show a D 1 c from 0.15 to 0.4. σ v is the ultimate stress during the tensile test
(before necking begins). The lower D c , the more likely a material damage is. Note that D c behaves
contrary to D : D c = 1 for the “safe” material and low for possible damaged regions.
CHAPTER 7 535
Material Library
In fact the comparison between D and D c is necessary to identify critical forming zones: as long as D
is (much) lower than D c , the forming operation is safe. When D reaches D c , the damage probability
tends to be 100%. The comparison is done by the relative damage value D r e l (reflected in the
postprocessing):
D
D r e l = ------ 0 ≤ Dr e l ≤ 1
Dc
When D r e l approaches 1 in a specific region, fracture is highly possible whereas small values indicate
a “safe” region.
Simplified Model
For elastic, elastic-plastic, or rigid-plastic materials, there is the option for you to define a simplified
damage model. You define the damage factor (df) in the UDAMAG user subroutine.
If model 9 is used, then:
p ·p
σ y = σ y ( ε , ε , T )* ( 1.0 – df )
If model 10 is used, then:

p ·p
σ y = σ y ( ε , ε , T )* ( 1.0 – df ) and E = E ( T )* ( 1.0 – df )
The normal data for a specific material are defined with the ISOTROPIC, WORK HARD, and OGDEN
options. Cross-reference to this material is made with the material number.
Cockroft-Latham Damage Indicator

Cockroft-Latham damage indicator does not affect the yield stress. It is a postprocessing value to indicate
a possible damage area. It can also be used to initiate crack by removing elements in the area.
σm a x ·
 ------------
σ
- ε dt ≥ C
·
where σ m a x is the maximum principal stress, σ is the effective von Mises stress and ε is the effective
plastic strain rate. C is material constant threshold for damage.
Principal-tension Damage Indicator

This damage indicator does not affect the yield stress.
σm a x
 ------------
σ
- dt ≥ C
where σ m a x is the maximum principal stress and σ is the effective von Mises stress. C is material
constant threshold for damage.
Figure 7-93 shows the application of the Cockroft-Latham damage criterion used to predict cracks known
as Chevron cracks in the drawing process.
Oyane Damage Indicator

σm
 ·
  ------
σ
- + B ε dt ≥ C

Similar to Cockraft-Latham, Oyane damage indicates a possible damage area where σ m is the mean or
average stress, B and C are both material constants.
The UDAMAGE_INDICATOR user subroutine can be used to define unsupported damage criteria. The
UACTIVE user subroutine can be used for element removal.
Figure 7-94 show the application of the Oyane damage criterion used to predict cracks known as Chevron
cracks in the extrusion process.
Figure 7-93 Chevron Crack Prediction with Cockroft-Latham Damage Criterion

CHAPTER 7 537
Material Library
Figure 7-94 Chevron Crack Prediction with Oyane Damage Criterion
Nonstructural Materials
In addition to stress analysis, Marc can be used for heat transfer, coupled thermo-electrical heating (Joule
heating), coupled electrical-thermal-mechanical analysis (Joule-mechanical), hydrodynamic bearing,
fluid/solid interaction, electrostatic, magnetostatic, electromagnetic, piezoelectric, acoustic, and fluid
problems. Material properties associated with these analyses and Marc options that control these
analyses are described below.
Heat Transfer Analysis

In heat transfer analysis, use the ISOTROPIC, ORTHOTROPIC, or ANISOTROPIC model definition
options to input values of thermal conductivity, specific heat, and mass density. If the latent heat effect
is to be included in the analysis, the value of latent heat and associated solidus and liquidus temperatures
must be entered through the LATENT HEAT or TEMPERATURE EFFECTS model definition option. Both
the thermal conductivity and specific heat can be dependent on temperatures. The mass density must be
constant throughout conventional heat transfer analysis. In addition, the ANKOND user subroutine can
be used for the input of anisotropic thermal conductivity. For radiation analysis, you must enter the
Stefan-Boltzmann constant in the RADIATION parameter and the emissivity through the EMISSIVITY or
ISOTROPIC option. The emissivity can be temperature dependent.
For high temperature analysis where pyrolysis occurs, the THERMO-PORE option is used to define the
advanced material behavior.
In a piezoelectric analysis, the material properties contain a mechanical part, an electrostatic part, and a
part connecting these two. The material properties for the mechanical part are the same as for an elastic
stress analysis. The material properties for the electrostatic part, the permittivity, and the material
properties for the piezoelectric coupling can be defined through the PIEZOELECTRIC option. The
coefficients for the piezoelectric coupling can be either stress based or strain based.
Thermo-Electrical Analysis
In addition to thermal conductivity, specific heat and mass density, the value of electric resistivity must
be entered using the ISOTROPIC and ORTHOTROPIC model definition options in a coupled thermo-
electrical analysis. Input the variation of electric resistivity with temperature through the
TEMPERATURE EFFECTS option.
Coupled Electrical-Thermal-Mechanical Analysis

Material properties for coupled electrical-thermal-mechanical analysis are the same for stress analysis in
addition to those of Joule heating analysis (thermal conductivity, specific heat and mass density and
electric resistivity).
Hydrodynamic Bearing Analysis

In a hydrodynamic bearing analysis, the ISOTROPIC and ORTHOTROPIC model definition options can
be used for entering both the viscosity and specific mass. Define the temperature dependency of viscosity
through the TEMPERATURE EFFECTS option.
Fluid/Solid Interaction Analysis – Added Mass Approach

In a fluid/solid analysis, the density of a fluid is given in the first field of the ISOTROPIC option. The
fluid is assumed to be nonviscous and incompressible.
In an electrostatic analysis, the permittivity can be defined through either the ISOTROPIC or
ORTHOTROPIC options. In addition, the UEPS user subroutine can be used for the input of
anisotropic permittivity.
In a magnetostatic analysis, the permeability can be defined through either the ISOTROPIC or
ORTHOTROPIC options. A nonlinear relationship can be entered via the B-H RELATION option. The
UMU user subroutine can be used for the input of isotropic permeability.
In an electromagnetic analysis, the permittivity, permeability and conductivity can be defined using
either the ISOTROPIC or ORTHOTROPIC options. An assumption made is that the permittivity is a
constant (does not vary with time) in the analysis. A nonlinear permeability can be entered via the B-H
RELATION option. The UEPS, UMU, and USIGMA user subroutines can be used for the input of
anisotropic permittivity, permeability, and conductivity, respectively. Both the dependent and harmonic
analysis can be performed.
CHAPTER 7 539
Material Library
Coupled Electrostatic-Structural
Material properties for a coupled electrostatic-structural are the same as for a stress analysis in addition
to those for an electrostatic analysis (permittivity).
Acoustic Analysis
In an acoustic analysis of a cavity with rigid boundaries, the bulk modulus and the relative density of the
medium can be entered through the ISOTROPIC option.
The ACOUSTIC parameter is used to indicate that a coupled acoustic-structural analysis is performed. In
addition to the CONTACT option, the ACOUSTIC and REGION model definition options are used to
define the material properties of the acoustic medium and to set which elements correspond to the solid
and the fluid region.
Fluid Analysis
In the fluid analysis, viscosity and density can be defined using the ISOTROPIC option. In addition,
conductivity and specific heat are defined for coupled fluid-thermal analysis. Non-Newtonian fluid
behavior can be defined using the STRAIN RATE option. The TEMPERATURE EFFECTS option can be
used to define the temperature dependence of the properties above.
References
1. Auricchio, F. and Taylor, R.L., “Shape-memory alloy: modeling and numerical simulations of the
finite-strain superelastic behavior”, Comput. Methods Appl. Mech. Engrg., Vol. 143, pp.175-194
(1997).
2. Auricchio, F., “A robust integration-algorithm for a finite-strain shape-memory-alloy
superelastic model”, Int. J. Plasticity, Vol.17, pp.971-990 (2001).
3. Arruda, E. M. and Boyce, M. C. “A three-dimensional constitutive model for the large stretch
behaviour of rubber elastic materials”, J. Mech. Phys. Solids, Vol.41, No. 2, 1993.
4. Buyukozturk, O., “Nonlinear Analysis of Reinforced Concrete Structures”, Computers and
Structures, 7, 149-156 (1977).
5. Gent, A. N., “A new constitutive relation for rubber”, Rubber Chem. Tech., 69, 1996.
6. Barlat, F., Lege, D.J. and Brem, J.C., “A six-component yield function for anisotropic metals”,
Int. J. Plasticity, 7, 693-712 (1991).
7. Chung, K. and Shah, K., “Finite element simulation of sheet metal forming for planar anisotropic
metals”, Int. J. Plasticity, 8, 453-476 (1992).
8. Yoon, J.W., Yang, D.Y. and Chung, K. and Barlat. F., “A general elasto-plastic finite element
formulation based on incremental deformation theory for planar anisotropy and its application to
sheet metal forming”, Int. J. Plasticity, 15, 35-67 (1999).
9. Yoon, J.W., Barlat, F., Chung, K., Pourboghrat, F. and Yang, D.Y., “Earing predictions based on
asymmetric nonquadratic yield function”, Int. J. Plasticity, 16, 1075-1104 (2000).
10. “Current Recommended Constitutive Equations for Inelastic Design Analysis of FFTF
Components.” ORLN-TM-360Z, October 1971.
11. “Finite Element Calculation of Stresses in Glass Parts Undergoing Viscous Relaxation”,
J.Am.Ceram.Soc. Vol. 70[2], pp. 90-95, 1987.
12. Mooney, M. J. Appl Phys., Vol. II, p. 582, 1940.
13. Naghdi, P. M. “Stress-Strain Relations in Plasticity and Thermoplasticity.” In Plasticity,
Proceedings of the Second Symposium on Naval Structural Mechanics, edited by E. H. Lee and
P. S. Symonds. Pergamon Press, 1960.
14. Narayanaswamy, O.S., “A Model of Structural Relaxation in Glass”, J.Am.Ceram.Soc., Vol.
54[10], pp. 491-498, 1971.
15. Prager, W. Introduction to Mechanics to Continua. New York: Dover Press, 1961.
16. “Proposed Modification to RDT Standard F9-5T Inelastic Analysis Guidelines.” ORNL, October
1978.
17. Rivlin, R. S. Phil Trans Roy Soc (A), Vol. 240, 459, 1948.
18. Rivlin, R. S. Phil Trans Roy Soc (A), Vol. 241, 3-79, 1948.
19. Timoshenko, S. P., and J. N. Goodier. Theory of Elasticity. Third Ed. New York: McGraw-Hill,
1970.
20. Simo, J. C. “On a Fully Three-dimensional Finite Strain Viscoelastic Damage Model:
Formulation and Computational Aspects,” Computer Methods in Applied Mechanics and
Engineering, Vol. 60, 1987, pp. 153-173.
21. Tvergaard, V., “Influence of Voids on Shear Band Instabilities under Plane Strain Conditions”,
Int. J. Fracture, Vol. 17, pp. 389-407, 1981.
22. Tvergaard, V., “Material Failure by Void Coalescence in Localized Shear Bands”, in Int. J. Solids
Struct., Vol. 18, No. 8, pp. 652-672, 1982.
23. Chaboche, J. L., “Constitutive Equations for Cyclic Plasticity and Cyclic Viscoplasticity”,
International Journal of Plasticity, Vol. 5, pp. 247-302, 1989
24. van den Boogard, A.H., “Implicit integration of the Perzyna viscoplastic material model”, TNO
Building and Construction Research, The Netherlands, TNO-report, 95-NM-R711, 1995
25. Puck, A. and Schürmann, H., “Failure Analysis of FRP Laminates by Means of Physically Based
Phenomenological Models”; Composite Science and Technology, 58, pp. 1045-1067, 1998.
26. Camacho, G.T. and Ortiz, M., “Computational modelling of impact damage in brittle materials”,
Int. J. Solids Struct., Vol. 33, pp 2899-2938, 1996.
27. Greve, L. and Picket, A. K., “Modelling Damage and Failure in Carbon/Epoxy Non-crimp Fabric
Composites Including Effects of Fabric Pre-shear”, Composites: Part A, Vol. 37, pp. 1983-2001,
2006.
28. Knops, M. and Bögle, C. “Gradual Failure in Fibre/Polymer Laminates”, Composites Science and
Technology, Vol. 66, pp. 616-625, 2006.
Chapter 8 Contact
8 Contact

Definition of Contact Bodies 542
 Numbering of Contact Bodies 545

Motion of Bodies 548

Detection of Contact 550
 Implementation of Constraints 554

Friction Modeling 556

Separation 568
 Coupled Analysis 569

Thermal Contact 571

Element Considerations 574
 Dynamic Impact 577

Global Adaptive Meshing and Rezoning 577

 Result Evaluation 578

Tolerance Values 579

Numerical and Mathematical Aspects of Contact 580
 References 601
The simulation of many physical problems requires the ability to model the contact phenomena. This
includes analysis of interference fits, rubber seals, tires, crash, and manufacturing processes among
others. The analysis of contact behavior is complex because of the requirement to accurately track the
motion of multiple geometric bodies, and the motion due to the interaction of these bodies after contact
occurs. This includes representing the friction between surfaces and heat transfer between the bodies if
required. The numerical objective is to detect the motion of the bodies, apply a constraint to avoid
penetration, and apply appropriate boundary conditions to simulate the frictional behavior and heat
transfer. Several procedures have been developed to treat these problems including the use of Perturbed
or Augmented Lagrangian methods, penalty methods, and direct constraints. Furthermore, contact
simulation has often required the use of special contact or gap elements. Marc allows contact analysis to
be performed automatically without the use of special contact elements. A robust numerical procedure to
simulate these complex physical problems has been implemented in Marc.
Definition of Contact Bodies

There are two types of contact bodies in Marc – deformable and rigid. Deformable bodies are simply a
collection of finite elements as shown below.
Figure 8-1 Deformable Body
This body has three key aspects to it:

1. The elements which make up the body.
2. The nodes on the external surfaces which might contact another body or itself. These nodes are
treated as potential contact nodes.
3. The edges (2-D) or faces (3-D) which describe the outer surface which a node on another body
(or the same body) might contact. These edges/faces are treated as potential contact segments.
Note that a body can be multiply connected (have holes in itself). It is also possible for a body to be
composed of both triangular elements and quadrilateral elements in 2-D or tetrahedral elements and brick
elements in 3-D. Beam elements and shells are also available for contact. Both lower-order (linear) and
higher-order (quadratic) elements may be used. One should not mix continuum elements, shells, and/or
beams in the same contact body.
Each node and element should be in, at most, one body. The elements in a body are defined using the
CONTACT option. It is not necessary to identify the nodes on the exterior surfaces as this is done
automatically. The algorithm used is based on the fact that nodes on the boundary are on element edges
or faces that belong to only one element. Each node on the exterior surface is treated as a potential contact
node. In many problems, it is known that certain nodes never come into contact; in such cases, the
CHAPTER 8 543
Contact
CONTACT NODE option can be used to identify the relevant nodes. As all nodes on free surfaces are
considered contact nodes, if there is an error in the mesh generation such that internal holes or slits exist,
undesirable results can occur.
Note: These problems can be visualized by using Marc Mentat to create an outline plot and fixed
by using the sweep command in Marc Mentat.
The potential segments composed of edges or faces are treated in potentially two ways. The default is
that they are considered as piece-wise linear (PWL). As an alternative, a cubic spline (2-D) or a Coons
surface (3-D) can be placed through them. The SPLINE option is used to activate this procedure. This
improves the accuracy of the calculation of the normal to the surface.
Rigid bodies are composed of curves (2-D) or surfaces (3-D) or meshes with only thermal elements in
coupled problems. The most significant aspect of rigid bodies is that they do not deform. Deformable
bodies can contact rigid bodies, but contact between rigid bodies is not considered.
They can be created either in CAD systems and transferred through Marc Mentat into Marc, created
within Marc Mentat, or created directly through the Marc input. There are several different types of
curves and surfaces that can be entered including:
Table 8-1 Geometrical Entities Used in Modeling Contact
2-D 3-D
line 4-node patch
circular arc ruled surface
spline surface of revolution
NURBS Bezier
poly-surface
cylinder
sphere
NURBS
trimmed NURBS
Within the Marc Mentat GUI, all curves or surfaces are mathematically treated as NURBS curves or
surfaces. This allows the greatest level of generality. Within the analysis, these rigid curves or surfaces
can be treated in two ways – discrete piecewise linear lines (2-D) or patches (3-D), or as analytical
NURBS curves or surfaces. When the discrete approach is used, all geometric primitives are subdivided
into straight line segments or bilinear patches. You have control over the density of these subdivisions to
approximate a curved line or surface within a desired degree of accuracy. This subdivision is also
relevant when the corner conditions (see “Detection of Contact” on page 550) are determined. Too few
subdivisions might lead to a very bad surface description and too many subdivisions might lead to a more
expensive analysis due to more complex contact checking. When using the analytical description, the
number of subdivisions given is less relevant. It should be large enough to roughly visualize the shape of
the NURBS. In 3-D, a too large number is automatically reduced so that the analysis cost is not
influenced too much. The treatment of the rigid bodies as NURBS surfaces is advantageous because it
leads to greater accuracy in the representation of the geometry and a more accurate calculation of the
surface normal. Additionally, the variation of the surface normal is continuous over the body which leads
to a better calculation of the friction behavior and better convergence.
To create a rigid body, you can either read in the curve and surface geometry created from a CAD system
or create the geometry in Marc Mentat or directly enter it in Marc. You then use the CONTACT option to
select which geometric entities are to be a part of the rigid body. An important consideration for a rigid
body is the definition of the interior side and the exterior side. For two-dimensional analysis, the interior
side is formed by the right-hand rule when moving along the body.
This is easily visualized with Marc Mentat by activating ID CONTACT.
Deformable top/front
Rigid top/front/inside
bottom/back/outside
Y Y
Z X X
Z
1 1
Figure 8-2 Orientation of Rigid Body Segments
For three-dimensional analysis, the interior side is formed by the right-hand rule along a patch. The
interior side is visualized in Marc Mentat as the front surface (pink); whereas, the exterior side is
visualized in Marc Mentat as the back surface (gold).
It is not necessary for rigid bodies to define the complete body. Only the bounding surface needs to be
specified. You should take care, however, that the deforming body cannot slide out of the boundary curve
in 2-D (Figure 8-3). This means that it must always be possible to decompose the displacement increment
into a component normal and a component tangential to the rigid surface.
Incorrect Correct
Figure 8-3 Deformable Surface Sliding Out of Rigid Surface
CHAPTER 8 545
Contact
Numbering of Contact Bodies

When defining contact bodies for a deformable-to-deformable analysis, it is important to define them in
the proper order. As a general rule, a body with a finer mesh should be defined before a body with a
coarser mesh.
Note: For problems involving adaptive meshing or automated remeshing, care must be taken to
satisfy this rule before as well as after the mesh change.
If one has defined a body numbering which violates the general rule, or if the rule is violated upon
remeshing, then a CONTACT TABLE model or history definition option can be used to modify the order
in which contact is established. This order can be directly user-defined or decided by the program. In the
latter case, the order is based on the rule that if two deformable bodies might come into contact, searching
is done such that the contacting nodes are on the body having the smallest element edge length and the
contacted segments are on the body having the coarser mesh. It should be noted that this implies
single-sided contact for this body combination, as opposed to the default double-sided contact.
For deformable-deformable contact, the CONTACT option offers the following three global contact
search controls:
• SINGLE-SIDED
• DOUBLE_SIDED
• OPTIMIZE CONTACT CONSTRAINT EQUATIONS
If flexible region A comes into contact with flexible region B, the following rules will be satisfied by the
program to determine the constraint equations:
1. Region A and region B belong to the same contact body:
a. If OPTIMIZE CONTACT CONSTRAINT EQUATIONS is active, the finer meshed region
contacts the coarser meshed region.
b. If OPTIMIZE CONTACT CONSTRAINT EQUATIONS is not active, the region containing
the lower node number coming into contact contacts the other region.
2. Region A and region B do not belong to the same contact body:
a. If SINGLE-SIDED contact is active, the region belonging to the lower contact body number
contacts the region belonging to the higher contact body number.
b. If DOUBLE_SIDED contact is active, the region containing the lower node number
coming into contact contacts the other region. Because of the search order for contact is
from lower body numbers to higher body numbers, this often leads to the same solution as
mentioned above.
c. If OPTIMIZE CONTACT CONSTRAINT EQUATIONS is active, the region with the softer
material contacts the region with the harder material, or the region with the finer mesh contacts
the region with the coarser mesh (soft-hard criterion has priority over mesh-density criterion).
d. All the above settings can be overruled by the CONTACT TABLE model or history definition
option where you can specify that a certain body AA should contact body BB and not vice-
verse, or that the body with the smallest element edge should contact the other body.
As an example, two identical bodies shown in Figure 8-4, but created using two different
approaches so that the node numbering order is different. Each U-Section is one contact body.
In particular, for the model on the left, the left leg of the U-section has lower node ids that the
right leg; while for the model on the right, the opposite is true.
Figure 8-4 Node Numbers on Models
A uniform distributed load is placed on the right leg, such that self-contact occurs. Without
using optimized contact, the behavior is shown in Figure 8-5.
Figure 8-5 Contact Based upon Original Model.
The dark squares indicate nodes that are in contact based upon the Contact Status. Observe that
the model on the left gives good behavior, while that on the right does not. This is because the
lower node numbers on the left leg detect contact with the right leg before the corner node on
CHAPTER 8 547
Contact
the right leg (which has a higher id) detects contact with the left leg. Then, activate the
Optimized Contact Constraint option, and the results are shown in Figure 8-6. The node
number IDs are no longer controlling the solution.
Figure 8-6 Contact using Optimized Contact Constraints
Each model is then subdivided into two contact bodies as shown in Figure 8-7, the contact
table option is not used and the optimized procedure is not used.
Figure 8-7 Contact Based upon Contact Body Numbering
Observe now that in the left model, the left leg is a lower body number, so the nodes on this
body try to contact the segments on the right leg. In the model on the right then, the right leg
has the lower body number, and so the nodes on this leg contact (successfully) with the
segments on the left leg.
When either the Optimized Contact Constraint Procedure or a well-selected Contact Table
utilizing Single-Sided Contact, the consistent results may be obtained for either model as
shown in Figure 8-8
Figure 8-8 Contact Using either Optimized Contact Constraint or Contact Table.
Motion of Bodies
The motion of deformable bodies is prescribed using the conventional methods of applying
displacements, forces, or distributed loads to the bodies. It is advantageous to not apply displacements or
point loads to nodes which might come into contact with other rigid bodies. If a prescribed displacement
is to be imposed, it is better to introduce another rigid body and apply the motion to the rigid body.
Symmetry surfaces are treated as a special type of bodies which have the property of being frictionless
and where the nodes are not allowed to separate. If a distributed load is applied to an edge or face, it is
possible for this load to be deactivated if all nodes on the edge or face are in contact with another body.
There are four ways to prescribe the motion of rigid surfaces:
Prescribed velocity
Prescribed position
Prescribed load
Prescribed scaling
Associated with the rigid body is a point labeled the centroid. When the first two methods are chosen,
you define the translational motion of this point, and the angular motion about an axis through this point.
The direction of the axis can be defined for three-dimensional problems. For two-dimensional problems,
it is a line normal to the plane. For complex time-dependent behavior, the MOTION user subroutine can
be used to prescribe the motion as an alternative to the input. The motion during a time increment is
considered to be linear. The position is determined by an explicit, forward integration of the velocities
based upon the current time step. A time increment must always be defined even if a static, rate-
independent analysis is performed.
When load controlled rigid bodies are used, two additional nodes, called the control nodes, are associated
with each rigid body. In 2-D problems, the first node has two translational degrees of freedom
(corresponding to the global x- and y-direction) and the second node has one rotational degree of freedom
(corresponding to the global z-direction). In 3-D problems, the first node has three translational degrees
of freedom (corresponding to global x-, y-, and z-direction) and the second node has three rotational
degrees of freedom (corresponding to the global x-, y-, and z-direction). In this way, both forces and
moments can be applied to a body by using the POINT LOAD option for the control nodes. Alternatively,
one may prescribe one or more degrees of freedom of the control nodes by using the FIXED DISP or DISP
CHAPTER 8 549
Contact
CHANGE options. Generally speaking, load-controlled bodies can be considered as rigid bodies with
three (in 2-D) or six (in 3-D) degrees of freedom. These degrees of freedom may be in a user-defined
system of either the TRANSFORMATION or COORD SYSTEM option is used. The prescribed position
and prescribed velocity methods (see Figure 8-9) have less computational costs than the prescribed load
method (see Figure 8-10).
2
Centroid
1 3
V ω
2
1
Figure 8-9 Velocity Controlled Rigid Surface
Fy
Mz
Fx
Extra Node
Figure 8-10 Load Controlled Rigid Surface
If the second control node is not specified, the rotation of the body is prescribed to be zero.
It should be noted that the nodal coordinates of the first control node define the center of rotation for the
body. The nodal coordinates of the second control node are arbitrary.
It is also possible to either expand or contract the size of the rigid body through the use of specified scale
factors multiplied by a table to make the scaling time dependent. Alternately, the UGROWRIGID user
subroutine may be used to provide scale factors. The scale factors must be initialized to 1.0 but can
change afterwards. Abrupt or discontinuous changes in the scale factors should be avoided while the
rigid surface is in contact with other bodies. If no rotations are applied, the scale factors can be different
in the x-, y-, and z-directions.
Initial Conditions
At the beginning of the analysis, bodies should either be separated from one another or in contact. Bodies
should not penetrate one another unless the objective is to perform an interference fit calculation. Rigid
body profiles are often complex, making it difficult for you to determine exactly where the first contact
is located. Before the analysis begins (increment zero), if a rigid body has a nonzero motion associated
with it, the initialization procedure brings it into first contact with a deformable body. No motion or
distortion occurs in the deformable bodies during this process. In a coupled thermal mechanical analysis,
no heat transfer occurs during this process. If more than one rigid body exists in the analysis, each one
with a nonzero initial velocity is moved until it comes into contact. Because increment zero is used to
bring the rigid bodies into contact only, you should not prescribe any loads (distributed or point) or
prescribed displacements initially. For multistage contact analysis (often needed to simulate
manufacturing processes), the APPROACH and SYNCHRONIZED options in conjunction with the
CONTACT TABLE and MOTION CHANGE options allow you to model contact bodies so that they just
come into contact with the workpiece.
In assembly analysis, it is possible that multiple bodies are initially in contact. Because of mesh
discretization, it is possible that the contact is not perfect, which would result in inducing stresses due to
overclosure. The CONTACT TABLE option may be used to specify that the initial contact should be stress
free. In such cases, the coordinates of the nodes are relocated on the surface so that initial stresses do
not occur.
Detection of Contact
During the incremental procedure, each potential contact node is first checked to see whether it is near a
contact segment. The contact segments are either edges of other 2-D deformable bodies, faces of 3-D
deformable bodies, or segments from rigid bodies. By default, each node could contact any other segment
including segments on the body that it belongs to. This allows a body to contact itself. To simplify the
computation, it is possible to use the CONTACT TABLE option to indicate that a particular body will or
will not contact another body. This is often used to indicate that a body will not contact itself. During the
iteration process, the motion of the node is checked to see whether it has penetrated a surface by
determining whether it has crossed a segment.
Because there can be a large number of nodes and segments, efficient algorithms have been developed
to expedite this process. A bounding box algorithm is used so that it is quickly determined whether a node
is near a segment. If the node falls within the bounding box, more sophisticated techniques are used to
determine the exact status of the node.
During the contact process, it is unlikely that a node exactly contacts the surface. For this reason, a
contact tolerance (Figure 8-11) is associated with each surface.
CHAPTER 8 551
Contact
rance
2 x Tole
Figure 8-11 Contact Tolerance
If a node is within the contact tolerance, it is considered to be in contact with the segment. The default
contact tolerance is calculated by the program (see Tolerance Values in this chapter). It is also possible
to define the contact tolerance through the input, either globally via the CONTACT option or per pair of
contact bodies via the CONTACT TABLE option.
(t) ( t+ Δ ) ( t+ Δ )
During an increment, if node A moves from A to A , where A is beyond the contact
tolerance, the node is considered to have penetrated (Figure 8-12). In such a case, a special procedure is
invoked to avoid this penetration. More details are discussed in Numerical and Mathematical Aspects of
Contact in this chapter.
A(t)
A(t + δt)
Figure 8-12 Trial Displacement with Penetration
The size of the contact tolerance has a significant impact on the computational costs and the accuracy of
the solution. If the contact tolerance is too small, detection of contact and penetration is difficult which
leads to higher costs. Penetration of a node happens in a shorter time period leading to more recycles due
to iterative penetration checking or to more increment splitting and increases the computational costs. If
the contact tolerance is too large, nodes are considered in contact prematurely, resulting in a loss of
accuracy or more recycling due to separation. Furthermore, the accepted solution might have nodes that
“penetrate” the surface less than the error tolerance, but more than desired by the user. The default error
tolerance is recommended.
Many times, areas exist in the model where nodes are almost touching a surface (for example, in rolling
analysis close to the entry and exit of the rolls). In such cases, the use of a biased tolerance area with a
smaller distance on the outside and a larger distance on the inside is advised. This avoids the close nodes
from coming into contact and separating again and is accomplished by entering a bias factor. The bias
factor should be given a value between 0.0 and 0.99. The default is 0.0 or no bias. Also, in analyses
involving frictional contact, a bias factor for the contact tolerance is recommended. The outside contact
area is (1. - bias) times the contact tolerance on the inside contact area (1. + bias) times the contact
tolerance (Figure 8-13). The bias factor recommended value is 0.95.
In some instances, you might wish to influence the decision regarding the deformable segment a node
contacts (or does not contact). This can be done using the EXCLUDE option.
(1 - Bias)∗ tolerance
(1 + Bias)∗ tolerance
Figure 8-13 Biased Contact Tolerance
Shell Contact
A node on a shell makes contact when the position of the node plus or minus half the thickness projected
with the normal comes into contact with another segment (Figure 8-14). In 2-D, this can be shown as:
x1 = A + n t ⁄ 2
x2 = A – n t ⁄ 2
Shell
Midsurface
2
x A t
anc e
2 x toler
x
Figure 8-14 Default Shell Contact
If point x or y falls within the contact tolerance distance of segment S, node A is considered in contact
with the segment S. Here x 1 and x 2 are the position vectors of a point on the surfaces 1 and 2 on the
shell, A is the position vector of a point (node in a discretized model) on the midsurface of the shell, n
is the normal to the midsurface, and t is the shell thickness.
CHAPTER 8 553
Contact
As the shell has finite thickness, the node (depending on the direction of motion) can physically contact
either the top surface, bottom surface, or mathematically contact can be based upon the midsurface. You
can control whether detection occurs with either both surfaces, the top surface, the bottom surface, or the
middle surface. In such cases, either two or one segment will be created at the appropriate physical
location. Note that these segments will be dependent, not only on the motion of the shell, but also the
current shell thickness (Figure 8-15).
Note: Shell elements should be oriented consistently. Also, a shell segment cannot be in contact
simultaneously on both the Top and Bottom sides. If simultaneous contact is required, use
the solid shell element.
S1
n n S1
2
S2
2
1
1
Include Both Segments Top Segment Only
2
n
S1
S2 2
1
1
Bottom Segments Only Ignore Shell Thickness
Figure 8-15 Selective Shell Contact
S 1, S 2 are segments associated with shell consisting of node 1 and 2.
Neighbor Relations
When a node is in contact with a rigid surface, it tends to slide from one segment to another. In 2-D,
the segments are always continuous and so are the segment numbers. Hence, a node in contact with
segment n slides to segment n – 1 or to segment n + 1 (Figure 8-16). This simplifies the
implementation of contact.
n-1
n
n+1
Figure 8-16 Neighbor Relationship (2-D)
In 3-D, the segments are often discontinuous (Figure 8-17). This can be due to the subdivision
of matching surfaces or, more likely, the CAD definition of the under lying surface geometry.
Hole Physically
Discontinuous
Nonmatching
Segments
Continuous Surface Segments Discontinuous Surface Geometry
Figure 8-17 Neighbor Relationship (3-D)
Continuous surface geometry is highly advantageous as a node can slide from one segment to the next
with no interference (assuming the corner conditions are satisfied). Discontinuous surface geometry
results in additional operations when a node slides off a patch and cannot find an adjacent segment.
Hence, it is advantageous to use geometry clean-up tools to eliminate small sliver surfaces and make the
surfaces both physically continuous and topologically contiguous.
Implementation of Constraints
For contact between a deformable body and a rigid surface, the constraint associated with no penetration
is implemented by transforming the degrees of freedom of the contact node and applying a boundary
condition to the normal displacement. This can be considered solving the problem:
K â â K â b  u â   f â 
  =  
K b â K b b  u b   fb 
where â represents the nodes in contact which have a local transformation, and b represents the nodes
not in contact and, hence, not transformed. Of the nodes transformed, the displacement in the normal
direction is then constrained such that δu â n is equal to the incremental normal displacement of the rigid
body at the contact point.
t
P
n
Figure 8-18 Transformed System (2-D)
CHAPTER 8 555
Contact
As a rigid body can be represented as either a piecewise linear or as an analytical (NURBS) surface, two
procedures are used. For piecewise linear representations, the normal is constant until node P comes to
the corner of two segments as shown in Figure 8-19. During the iteration process, one of three
circumstances occur. If the angle α is small ( – α smooth < α < α smooth ) , the node P slides to the next
segment. In such a case, the normal is updated based upon the new segment. If the angle α is large
( α > α smooth or α < – α smooth ) the node separates from the surface if it is a convex corner, or sticks if
it is a concave corner. The value of α smooth is important in controlling the computational costs. A larger
value of α smooth reduces the computational costs, but might lead to inaccuracies. The default values are
8.625° for 2-D and 20° for 3-D. These can be reset using the PARAMETERS option in the model
definition or history definition section.
α
α
Π
Convex Corner Concave Corner

Figure 8-19 Corner Conditions (2-D)
In 3-D, these corner conditions are more complex. A node (P) on patch A slides freely until it reaches
the intersection between the segments. If it is concave, the node first tries to slide along the line of
intersection before moving to segment B. This is the natural (lower energy state) of motion.
These corner conditions also exist for deformable-to-deformable contact analysis. Because the bodies
are continuously changing in shape, the corner conditions (sharp convex, smooth or sharp concave) are
continuously being re-evaluated.
When a rigid body is represented as an analytical surface, the normal is recalculated at each iteration
based upon the current position. This leads to a more accurate solution, but can be more costly because
of the NURBS evaluation.
B
P
Figure 8-20 Corner Conditions (3-D)

When a node of a deformable body contacts a deformable body, a multipoint constraint (called tying) is
automatically imposed. Recalling that the exterior edges (2-D) or faces (3-D) of the other deformable
bodies are known, a constraint expression is formed. For 2-D analysis using lower-order elements, the
number of retained nodes is three – two from the edge and the contacting node itself. For 3-D analysis,
the number of retained nodes is five – four from the patch and the contacting node itself. When using
higher-order elements and true quadratic contact, the number of retained nodes for 2-D becomes four, for
3-D, (hexahedrals) nine, and for 3-D (tetrahedrals) seven. The constraint equation is such that the
contacting node should be able to slide on the contacted segment, subject to the current friction
conditions. This leads to a nonhomogeneous, nonlinear constraint equation. In this way, a contacting
node is forced to be on the contacted segment. This might introduce undesired stress changes, since a
small gap or overlap between the node and the contacted segment will be closed (note that using the
Single Step Houbolt or generalized alpha/HHT operator in dynamics, forcing a node to be on the
contacted segment is switched off by default, but can be activated via the PARAMETERS option). During
initial detection of contact (increment 0), the stress-free projection option avoids those stress changes for
deformable contact by adapting the coordinates of the contacting nodes such that they are positioned on
the contacted segment. This stress-free projection can be activated using CONTACT TABLE. A similar
option exists for glued contact; however, in this case, overlap will not be removed.
During the iteration procedure, a node can slide from one segment to another, changing the retained nodes
associated with the constraint. A recalculation of the bandwidth is automatically made. Because the
bandwidth can radically change, the bandwidth optimization is also automatically performed.
A node is considered sliding off a contacted segment if is passes the end of the segment over a distance
more than the contact tolerance. As mentioned earlier, the node separates from the contacted body if this
happens at a convex corner. For deformable contact, this tangential tolerance at convex corners can be
enlarged by using the delayed sliding off option activated via CONTACT TABLE.
Friction Modeling
Friction is a complex physical phenomenon that involves the characteristics of the surface such as surface
roughness, temperature, normal stress, and relative velocity. An example of this complexity is that quite
often in contact problems neutral lines develop. This means that along a contact surface, the material
flows in one direction in part of the surface and in the opposite direction in another part of the surface.
Such neutral lines are, in general, not known a priori.
The actual physics of friction and its numerical representation continue to be topics of research.
Currently, in Marc the modeling of friction has basically been simplified to two idealistic models. The
most popular friction model is the Coulomb friction model. This model is used for most applications with
the exception of bulk forming as encountered in e.g. forging processes. For such applications the shear
friction model is more appropriate.
Coulomb Friction
The Coulomb model can be characterized by:
σ t < μσ n (stick) and σ t = – μ σ n ⋅ t (slip)
CHAPTER 8 557
Contact
where
σt is the tangential (friction) stress
σ n is the normal stress
μ is the friction coefficient

t is the tangential vector in the direction of the relative velocity:
vr
t = ----------- , in which v r is the relative sliding velocity.
vr
Similarly, the Coulomb model can also be written in terms of nodal forces instead of stresses:
f t < μf n (stick) and f t = – μ f n ⋅ t (slip)
where
ft is the tangential (friction) force
fn is the normal force
When Coulomb friction is used with the stress-based model, the integration point stresses are first
extrapolated to the nodal points and then transformed, so that a direct component is normal to the
contacted surface. Given this normal stress and the relative sliding velocity, the tangential stress is then
evaluated and a consistent nodal force is calculated. For shell elements, the nodal force based Coulomb
model must be used, since a reduced stress state is assumed with σ n = 0 . For continuum elements, both
the stress-based and force-based Coulomb model can be used.
For a given normal stress or normal force, the friction stress or force has a step function behavior based
upon the value of the relative sliding velocity v r or the tangential relative incremental displacement
Δu t , as outlined in Figure 8-21 for a 2-D case, where the relative velocity and incremental displacement
are scalar values.
σ t or f t
Stick
v r or Δu
t
Slip
Figure 8-21 Coulomb Friction Model
Since this discontinuity in the friction value may easily cause numerical difficulties, different
approximations of the step function have been implemented. They are graphically represented in
Figure 8-22 and they will be successively discussed.
σ t or f ft ft
t
vr Δu t Δu t
arctangent model stick-slip (modified step bilinear model

function) model
Figure 8-22 Different Approximations for the Coulomb Friction Model
Arctangent Model
The first procedure is based on a continuously differentiable function in terms of the relative
sliding velocity:
vr
σ t = – μ σ n --- arctan  ------------------------- ⋅ t
2
π  RVCNST
for the friction stress, and:

vr
f t = – μ f n --- arctan  ------------------------- ⋅ t
2
π  RVCNST
for the friction force.

Physically, the value of RVCNST can be seen as the value of the relative velocity below which sticking
occurs. The value of RVCNST is important in determining how closely the mathematical model
represents the step function, as shown in Figure 8-23. A very large value of RVCNST results in a
reduced value of the effective friction. A very small value may result in poor convergence. It is
recommended that the value of RVCNST be 1% to 10% of a typical relative sliding velocity, v r .
The friction implementation not only affects the external force vector, but also the stiffness matrix of the
final set of equations to be solved on a global level. This stiffness matrix contribution follows from:
∂f t ∂v r
i k
K i j = – ---------- -------------
∂v r ∂Δu t
k j
CHAPTER 8 559
Contact
This later contribution, if fully implemented, would lead to a non-symmetric system. Because of the
additional computational costs, both in terms of memory and CPU costs, the contribution to the stiffness
matrix is symmetrized.
ft
μf n
vr
– 10 10
RVCNST = 10
RVCNST = 1
–μ fn RVCNST = 0.1
Figure 8-23 Step Function Approximation for Different Values of RVCNST (-10<vr<10)
(i)
During iteration ( i ) of the Newton-Raphson process, the change of the friction force δf t is related to
(i)
the change of the relative sliding velocity δv r . Since the latter can be written as:
(i) (i – 1) (i)
(i) Δu t – Δu t δu t
δv r = ----------------------------------------- = ------------
Δt Δt
the resulting equation can still be expressed in displacement degrees of freedom. During the very first
iteration of an increment, it turns out that the convergence can be significantly improved by using:
(1) ( previous increment ) (1)
(1) Δu t Δu t δu t ( previous increment )
δv r = -------------- – ------------------------------------------------ = ------------- – v r
Δt Δt
( previous increment ) Δt
( previous increment )
where v r results in a contribution to the right hand side of the global set of equations.
As a result of the smoothing procedure, a node in contact always has some slipping. Besides the
numerical reasons, this ‘ever slipping node’ model has a physical basis. Oden and Pires pointed out that
for metals, there is an elasto-plastic deformation of asperities at the microscopic level (termed as ‘cold
weld’) which leads to a nonlocal and nonlinear frictional contact behavior. The arctangent representation
of the friction model is a mathematical idealization of this nonlinear friction behavior.
A complicating aspect of the arctangent model is that it may be difficult to estimate a priori the typical
relative sliding velocity. This is particularly true in quasi-static deformable contact analyses (where
the velocity is not an input quantity) and in cases where the sliding velocity varies strongly during
the analysis.
Stick-slip (Modified Step Function) Model

The second procedure is based on a slightly modified step function and can be used to simulate true stick-
slip behavior. In this procedure, each node in contact gets a friction status, being either stick or slip.
Depending on this status, different constraints are applied and after each iteration in the iterative solution
process, the correctness of the friction status is checked and if necessary adapted.
The typical parameters used (see also Figure 8-24) are the friction coefficient multiplier α , the slip to
stick transition region β and the friction force tolerance e .
The multiplier α (with a default value of 1.05 ) can be seen as an overshoot parameter used to simulate
the difference between static and dynamic friction.
–6
The slip-to-stick transition region β (with a default value of 1 ×10 ) can be seen as a tolerance on the
friction solution. If a node is in slipping mode and moves in the direction of the friction force (which is
physically incorrect), but the corresponding relative displacement magnitude is within the slip-to-stick
transition region, then this will not cause the increment to be restarted with modified friction conditions.
–6
Related to the parameter β , a reduction factor ε = 1.0 ×10 is used. This fixed factor causes the
product of ε and β to be very small, so that it can be used as a criterion to estimate if a node is initially
in slipping or sticking mode. This initial assumption serves as the starting point of the above mentioned
process of checking and adapting the friction status, which, for a 2-D analysis, is given in Figure 8-25
(note that Δu t ≈ 0 implies Δu t < εβ ).
ft α : friction coefficient multiplier
β : slip to stick transition region

–6
β ε : reduction factor ( 1 ×10 )
αμf n μf n
εβ Δu t
Figure 8-24 Stick-slip or Modified Step Function Friction Parameters

CHAPTER 8 561
Contact
Initial Contact
No Yes
Δu t ≈ 0
Assume Slipping Assume Sticking

Mode Mode
Determine Solution
of Next Iteration
Remain in Slipping Mode if: Remain in Sticking Mode if:

f t • Δu t < 0 and Δu t > β f t ≤ αμf n
Change to Sticking Mode if:

f t • Δu t > 0 and Δu t > β Change to Slipping Mode if:
f t > αμf n
or if Δu t ≈ 0
Figure 8-25 Flow Diagram for the Stick-slip or Modified Step Function Friction Model
Contact When a node is in slipping mode, the contribution to the stiffness matrix is:
∂f t
i
Ki j = – -------------
∂Δu t
j
while in sticking mode constraints are applied which exactly enforce a zero relative tangential
displacement. It should be noted that in slipping mode the current friction force is not only a function of
the current normal force and relative displacement, but also of the friction force during the previous
iteration. This generally improves the stability of the nonlinear solution process.
As opposed to the procedure based on the arctangent model, this procedure requires additional friction-
based testing. Changes in the friction status as well as too large changes in the friction forces compared
to the previous iteration may cause the program to perform an additional iteration. In 2-D, the latter
implies that the friction force should fulfil1:
ft
1 – e ≤ ----p- ≤ 1 + e
ft
where f p is the friction force of the previous iteration and e is the above mentioned friction force
t
tolerance, with a default value of 0.05. In 3-D, in addition to this requirement on the magnitude of the
friction force, a check is performed on the change in direction of the friction force vector:
1 – e ≤ cos ( ϕ f ) ≤ 1 + e
in which ϕ f is the angle between the friction vector in the current and the previous iteration.
In order to limit the amount of friction-related iterations, the above mentioned requirements will not
cause extra iterations if the number of nodes not fulfilling the friction requirements compared to the total
number of nodes in contact is small. This can be expressed in the friction force tolerance as:
number of nodes which do not fulfil friction tolerances
----------------------------------------------------------------------------------------------------------------------------------- < e
total number of nodes in contact
and must be true for all contact bodies in the model.

If the maximum number of iterations as defined on the CONTROL model definition option has been
reached, additional friction iterations will not be enforced and the analysis will continue with the
next increment.
The stick-slip model is always based on nodal forces. This model is not recommended for quadratic
elements in 3-D, because of the varying magnitude and sign of nodal forces of such elements.
When using the PRINT parameter to get extra contact-related information in the output file, the program
will echo the contact status, the ratio of the magnitude of the current and previous friction force vector
(and in 3-D also the cosine of the angle between the current and previous friction force vector), as well
as the total number of nodes with a friction contribution and the number of nodes with a friction
contribution which do not fulfil the friction tolerance criteria.
Bilinear Model
The third procedure is called the bilinear model. Similar to the modified step function model, it is based
on relative tangential displacements. Instead of defining special constraints to enforce sticking
conditions, the bilinear model assumes that the stick and slip conditions correspond to reversible (elastic)
and permanent (plastic) relative displacements, respectively. The clear resemblance with the theory of
elasto-plasticity will be used to derive the governing equations.
First, Coulomb’s law for friction is expressed by a slip surface φ :
φ = f t – μf n
The stick or elastic domain is given by φ < 0 , while φ > 0 is physically impossible.
Next, the rate of the relative tangential displacement vector is split into an elastic (stick) and a plastic
(slip) contribution according to:
e p
u· t = u· t + u· t (8-1)
CHAPTER 8 563
Contact
and the rate of change of friction force vector is related to the elastic tangential displacement by:
f·t = Du· t
e
(8-2)
in which matrix D is given by:
μf n
-------- 0
D = δ
μf
0 -------n-
δ
with δ the slip threshold or relative sliding displacement below which sticking is simulated (see
Figure 8-26). The value of δ is by default determined by the program as 0.0025 times the average edge
length of the finite elements defining the deformable contact bodies.
Now attention is paid to the case that, given a tangential displacement vector, the evolution of f t would
result in a physically impossible situation, so φ > 0 . This implies that the plastic or slip contribution
must be determined.
Equations (8-1) and (8-2) yield:
· · ·p
ft = D ( ut – ut )
ft
μf n
δ Δu t
δ slip threshold
Figure 8-26 Bilinear Model
It is assumed that the direction of the slip displacement rate is given by the normal to the slip flow
potential ψ , given by:
ψ = ft
·
So that, by indicating the slip displacement rate magnitude as λ:
·p · ∂ψ
u t = λ -------
∂f t
with the slip surface, φ , different from the slip flow potential, ψ an analogy to nonassociative plasticity
can be observed.
Since a ‘force point’ can never be outside the slip surface, it is required that:
∂φ T ·
φ =  ------ f t = 0
·
 ∂f t
In this way, the magnitude of the slip rate can be determined. To this end, the equations above can be
combined to:
∂φ  T  ·
 ----- ∂ψ
- D u t – λ· ------- = 0
 ∂f t  ∂f t 
or:
∂φ  T ·
 ----- - Du t
 ∂f t
·
λ = ----------------------------
∂φ  T ∂ψ
 ----- - D -------
 ∂f t ∂f t
Utilizing this result, the final set of rate equations reads:
 ∂ψ ∂φ T 
 D -------  ------ D
∂f t  ∂f t
f t =  D – --------------------------------- u t = ( D – D )u t
· · * ·
 ∂ φ T ∂ψ 
 ------ D ------- 
  ∂f t
 ∂f t 
*
Similar to non-associative plasticity, matrix D will generally be non-symmetric. In Marc, a special
procedure has been used which results in a symmetric matrix, while maintaining sufficient numerical
stability and rate of convergence.
The bilinear model also uses an additional check on the convergence of the friction forces, which has
been achieved if the following equation is fulfilled:
p
Ft – Ft
-≤e
--------------------------------
Ft
CHAPTER 8 565
Contact
p
where F t is the current total friction force vector (the collection of all nodal contributions), F t is the
total friction force vector of the previous iteration and e is the friction force tolerance, which has a
default value of 0.05.
In the output file one can find, per deformable contact body, information about the current total
friction force, the current total normal force and the total friction force change, compared to the
previous iteration.
If a node comes into contact and the structure is still stress-free, then the friction stiffness matrix
according to the derivation above will still be zero. This could result in an ill-conditioned system during
the next solution of the global set of equations. To avoid this problem, the initial friction stiffness will be
based on the average contact body stiffness (following from the trace of the matrix defining the material
behavior), which is determined during increment 0 of the analysis. Alternatively, the user can define this
initial friction stiffness via the PARAMETERS model definition or history definition option.
Limitations of the Coulomb Model

When the normal force or stress becomes large, the Coulomb friction model may not correlate well with
experimental observations. This is caused by the fact that the Coulomb model predicts that the frictional
shear stresses increase to a level that can exceed the flow stress or the failure stress of the material. As
this is not physically possible, a different friction calculation should be applied. The choices are either
to have a nonlinear coefficient of friction, or to introduce the friction stress limit in the bilinear model or
to use the shear based friction model.
σt
Linear Coulomb Model
μ
Observed Behavior
σn
Figure 8-27 Linear Coulomb Model versus Observed Behavior
A nonlinear friction coefficient can be defined via UFRIC user subroutine or via table driven input.
limit
The friction stress limit for the bilinear model ( σ t ) can be entered via the CONTACT TABLE option
and is used to bound the maximum friction stress, based on the assumption that the extrapolated and
averaged shear (friction) stress in a node is proportional to the applied shear (friction) force (see
Figure 8-28). If the shear stress reaches the limit value, then the applied friction force is reduced, so that
limit
the maximum shear stress is given by min ( μσ n, σ t ).
Finally, instead of using an adapted Coulomb friction model, one may use the shear friction model, which
is the second idealistic friction model in Marc.
σt
Friction Stress Limit
σn
Figure 8-28 Friction Stress Limit used by Bilinear Model
Shear Friction
The shear based model states that the frictional stress is a fraction of the equivalent stress σ in
the material:
σ σ
σ t < m ------- (stick) and σ t = – m ------- ⋅ t (slip)
3 3
where m is the friction factor.

Given the similarity with the Coulomb model, this shear model is implemented using two approximations
of the theoretical step function. First, the arctangent function to smooth out the step function is applied:
σ 2 vr
σ t = – m ------- --- arctan  ------------------------- ⋅ t
3π  RVCNST
This model is available for all elements using the distributed load approach.
Second, the bilinear model has been adapted based on the assumption that the shear (friction) stress in a
node is proportional to the applied shear (friction) force. Similar to the friction stress limit, the shear
stress due to friction is limited by:
σ
σ t = min  mσ n, m -------
 3
Friction Coefficient
When a node contacts a rigid body, the coefficient of friction associated with the rigid body is used. When
a node contacts a deformable body, the average of the coefficients for the two bodies are used. The
CONTACT TABLE option can be used if complex situations occur.
Recalling that friction is a complex physical phenomenon, due to variations in surface conditions,
lubricant distribution, and lubricant behavior, relative sliding, temperature, geometry, and so on, it was
decided to implement the above two friction models (Coulomb and shear), and to allow you to extend
them, if necessary, by means of the UFRIC user subroutine or by referencing a table. In such a routine,
you provide the friction coefficient or the friction factor as
μ = μ ( x, f n, T, v r, σ )
CHAPTER 8 567
Contact
or
m = m ( x, f n, T, v r, σ )
where
x is the position of the point at which friction is being calculated
fn is the normal force at the point at which friction is being calculated
T is the temperature at the point at which friction is being calculated
vr is the relative sliding velocity between point at which friction is being calculated and surface
σy is the equivalent stress at the point at which friction is being calculated
Glue Model
A special type of friction model is the glue option, which imposes that there is no relative tangential
motion. The glue motion is activated through the CONTACT TABLE option.
A novel application of contact is to join two dissimilar meshes. In such a case, by specifying that the glue
motion is activated, the constraint equations are automatically written between the two meshes. Unless
a flag has been set, nodes in contact via the glue option are not allowed to separate.
When a node of a shell or beam element is glued to a load controlled rigid body or to the face of a shell
or solid element, the rotation of the shell or beam can be suppressed. This allows a true moment carrying
glued connection. For the case that a connection is made to the face of a solid element, the rotations of
the touching node are connected to the translations of the nodes of the contacted patch by a constraint
relation based upon the large rotation RBE3 theory. The moment carrying feature is optional and is
activated through the CONTACT TABLE option.
Deact Glue
Another option related to the glue option id DEACT GLUE. Suppose two bodies are glued together and
a part of the interface should have standard contact (in order to model, for example, a crack), the nodes
in the region that should have standard contact are then identified with DEACT GLUE. Without this
option, one would have to split up the contact bodies in order to model this case.
Breaking Glue
There is a capability on the CONTACT TABLE option to break up the glued connection using a stress
criterion. When the following criterion is fulfilled at a node, the glued contact is released:
σ m σ n
 -----n- +  -----t > 1 (8-3)
 S n  S t
Here σ n is the contact normal stress, σ t the contact tangential stress, and S n , S t , m, and n are
user-defined parameters. A user-defined criterion can be given with the UBREAKGLUE user subroutine.
When a node is released due to breaking glued, it changes status from being glued to standard contact
permitting separation and friction.
The contact stresses are calculated using extrapolated stresses for solid elements and as contact force
divided by equivalent area for shell elements.
Nodal post codes are available for postprocessing the separate terms in Equation (8-3) as well as the sum.
A nodal post code for the deact glue status is also available.
Separation
After a node comes into contact with a surface, it is possible for it to separate in a subsequent iteration or
increment. Mathematically, a node should separate when the reaction force between the node and surface
becomes tensile or positive. Physically, you could consider that a node should separate when the tensile
force or normal stress exceeds the surface tension. Rather than use an exact mathematical definition, you
can enter the force or stress required to cause separation.
Separation can be based upon either the nodal forces or the nodal stresses. The use of the nodal stress
method is recommended as the influence of element size is eliminated.
When using higher-order elements and true quadratic contact, the separation criteria must be based upon
the stresses as the equivalent nodal forces oscillate between the corner and midside nodes.
In many analysis, contact occurs but the contact forces are small; for example, laying a piece of paper
on a desk. Because of the finite element procedure, this could result in numerical chattering. Marc has
some additional contact control parameters that can be used to minimize this problem. As separation
results in additional iterations (which leads to higher costs), the appropriate choice of parameters can be
very beneficial.
When contact occurs, a reaction force associated with the node in contact balances the internal stress of
the elements adjacent to this node. When separation occurs, this reaction force behaves as a residual force
(as the force on a free node should be zero). This requires that the internal stresses in the deformable body
be redistributed. Depending on the magnitude of the force, this might require several iterations.
You should note that in static analysis, if a deformable body is constrained only by other bodies (no
explicit boundary conditions) and the body subsequently separates from all other bodies, it would then
have rigid body motion. For static analysis, this would result in a singular or nonpositive definite system.
This problem can be avoided by appropriate boundary conditions. When the direct solvers are used, the
AUTOSPC parameter may also be used to suppress singular modes.
Release
A special case of separation is the intentional release of all nodes from a rigid body. This is often used in
manufacturing analysis to simulate the removal of the workpiece from the tools. After the release occurs
in such an analysis, there might be a large redistribution of the loads. It is possible to gradually reduce
the residual force to zero, which improves the stability, and reduces the number of iterations required.
The RELEASE history definition option allows the release (separation) of all the nodes in contact with a
CHAPTER 8 569
Contact
particular surface at the beginning of the increment. The rigid body should be moved away using the
MOTION CHANGE option or deactivated using the CONTACT TABLE option to ensure that the nodes do
not inadvertently recontact the surface they were released from.
Coupled Analysis
Contact analysis has several consequences on performing a coupled analysis. When performing a
coupled contact analysis between multiple deformable bodies, each body deforms due to mechanical and
thermal loads and undergoes heat conduction. When the bodies come into contact, there is heat flux
across the surfaces. You need to provide a coefficient of heat transfer between the surfaces. This is often
quite significant as a hot workpiece might come into contact with a cold tool set. Additionally, if friction
is present, heat is generated based upon:
q = ff r ⋅ v r ⋅ Me q
where M e q is the mechanical equivalent of heat. Half of the heat generated due to friction is contributed
to each body. If the model allows a body to be rigid, two options are available. In the first case, the rigid
body is considered to have a constant temperature (heat source or heat sink). Heat flux is exchanged
between the rigid body and the deformable body. In the second case, it is required to perform a heat
transfer analysis in the rigid body. In this case, the rigid body must be constructed of finite elements, but
they are chosen as heat transfer types. Deformation does not occur in this body, but a heat transfer
analysis is performed. This is computationally more efficient than performing a coupled analyses where
all bodies are deformable. Symmetry bodies have different characteristics as they are frictionless (hence,
no heat generated) and because there is no heat flux across a symmetry plane.
Heat fluxes (FILMS) are automatically created on all the boundaries of the deformable-bodies. Three
types of heat fluxes can be distinguished, where each flux has its own physical characteristic. The type
of flux Marc uses, depends on the distance between the contacting bodies (see Figure 8-29). We define
the distance, d c o n t a c t , as the distance below which the two bodies are touching each other. For a distance
smaller than d c o n t a c t , the first flux type is used. The second flux type is used when the distance is
between d c o n t a c t and d n e a r , where d n e a r is called the near contact distance. This distance represents
a small gap between the contact bodies, and should not be chosen larger than the smallest element size.
If this distance is not set, this type of flux is disregarded. The third flux type is used when the distance is
larger then d n e a r .
T1
d
T2
Figure 8-29 Contact Distance

If d < d c o n t a c t , the bodies touch and conduction takes place between the two contacting bodies. The
heat flux is written as:
q = HT C ( T2 – T1 )
where
T1 is the surface temperature,
HT C is the film coefficient between the two surfaces,
T2 is the temperature of the same contact location, as obtained from interpolation of nodal
temperatures of the body being contacted.
If d c o n t a c t < d < d n e a r , a small gap between the bodies exists. The heat flux that exists through this gap
can be decomposed in the following physical processes: convection, natural convection, radiation, and a
distance dependent term. This flux is given by:
BN C 4 4
q = HC V ( T2 – T1 ) + HN C ( T2 – T1 ) + σεf ( T A 2 – T A 1 )
 d 
+  H C T  1 – ------------- + H B L ------------- ( T 2 – T 1 )
d
 dn e a r  dn e a r 

where
HC V is the convection coefficient for near field behavior.
HN C is the natural convection coefficient for near field behavior.
BC N is the exponent associated with natural convection.
σ, ε, and f are the coefficients for radiation, the Stefan-Boltzman coefficient, the emissivity, and
the viewfactor, respectively. The Stefan-Boltzman constant is given via the
PARAMETERS option.
HB L is the separation distance dependent heat transfer coefficient.
TA 2 T 2 + T A B S R E F is the temperature in absolute scale. The offset between the user

temperature and the absolute temperature is given in the PARAMETERS option.
If d > d n e a r , or d n e a r has not been defined and d > d c o n t a c t , the bodies are not touching and a heat
flux to the environment exists in the following form,
4 4
q = H C T VE ( T 2 – T S I NK ) + σεf ( T A 2 – T A 1 )
CHAPTER 8 571
Contact
where
HC T V E is the heat transfer coefficient to the environment
T S I NK is the environment sink temperature
When a deformable-body contacts a rigid-body, the coefficients associated with the rigid-body are used.
When two deformable bodies are in contact, the average value of the film coefficient specified for the
bodies is used. This only works for the coefficients discussed in the first and third heat flux type. A better
way to prescribe the coefficients is to use the CONTACT TABLE option. With this option, all the
previously discussed coefficients can be set for each interface separately. For the second heat flux type,
this is the only way to specify the coefficients. As with all other coupled problems, heat generated by
plastic deformation can be calculated and applied as a volumetric flux. The heat generated by friction is
also calculated and applied as a surface flux.
Three user subroutines are available to facilitate the creation of more sophisticated boundary flux
definitions (such as radiation and convections with variations in space, temperature, pressure, etc.).
UHTCON allows you to specify a film coefficient while the surface is in contact ( d < d c o n t a c t ),
UHTNRC allows you to specify a film coefficient while the surface is almost in contact
( d c o n t a c t < d < d n e a r ), and UHTCOE allows you to specify a film coefficient while the surface is
free ( d > d n e a r ).
Thermal Contact
A thermal contact analysis is performed when no structural pass is present, this results in a
computationally more efficient analysis. Possible analyses types are Heat Transfer, Coupled Thermal-
Electrical, Magnetostatics and Electrostatics. For Heat Transfer and Coupled Thermal-Electrical the
analysis is comparable to a coupled analysis. The two types of rigid bodies that are described in the
coupled analysis are allowed. So one is a rigid which is considered to have a constant temperature (heat
source or heat sink), and the other is a finite element body with thermal properties. The three types of
heat fluxes as described in Coupled Analysis can also be used here.
The coefficients are prescribed in the same way as in a Coupled Analysis where it is advised to use the
CONTACT TABLE option.
The three user subroutines UHTCON, UHTNRC, and UHTCOE as previously discussed in Coupled
Analysis can also be used in a Heat Transfer and Coupled Thermal-Electrical analysis.
Joule Heating
In a Joule heating analysis similar conditions as thermal contact apply regarding the flow of current
between contacting bodies.
If d < d c o n t a c t , the bodies touch and electrical conduction takes place between the two contacting
bodies. The current is written as:
i = H E C ( V 2 – V 1 ) or H E C ( V b o d y – V 1 )
where
V1 is the surface voltage,
HE C is the electrical contact conductivity,
V2 is the voltage of the same contact location, as obtained from interpolation of nodal voltages
of the body being contacted.
VB o d y is the voltage of the rigid body being touched.
If d c o n t a c t < d < d n e a r , a small gap between the bodies exists. The current that exists through this gap
can be decomposed into:
 d 
i = H NE ( V 2 – V 1 ) +  H E C  1 – ------------- + H B L E ------------- ( V 2 – V 1 )
d
 d  d
 near near 
where
HN E is the coefficient for near field behavior.
HB L E is the separation distance dependent conductivity coefficient.
If d > d n e a r , or d n e a r has not been defined and d > d c o n t a c t , the bodies are not touching and a current
leakage to the environment exists in the following form,
q = H V E ( V SI N K – V 1 )
where
HV E is the coefficient to the environment
VS I N K is the environment sink voltage
Electrical contact between bodies does contribute a term to the heat generation.
Note that near electrical contact for Joule heating does not represent electrical arcing between bodies,
which is a much more complex phenomena.
CHAPTER 8 573
Contact
In a Joule heating analysis, the THERMAL CONTACT model definition option is also used to identify
elements which are in a conducting body to perform a resistance calculation using EMRESIS history
definition option.
Three user subroutines are available to facilitate the creation of more sophisticated boundary definitions.
UVTCON allows you to specify a film coefficient while the surface is in contact ( d < d c o n t a c t ),
UVTNRC allows you to specify a film coefficient while the surface is almost in contact
( d c o n t a c t < d < d n e a r ), and UVTCOE allows you to specify a film coefficient while the surface is
free ( d > d n e a r ).
Contact in an Electrostatic or Piezoelectric Analysis

In an electrostatic analysis, if two bodies are in contact, the scalar potential is tied between the bodies
composed of elements. For an electrostatic analysis the THERMAL CONTACT option is used. When a
node is in contact with a symmetry surface, it will be treated as a free surface, that is no current
transmitted through the symmetry surface. The THERMAL CONTACT option is also used to identify the
elements which are in a conducting body to perform a capacitance calculation using EMCAPAC history
definition option. In such cases, the conducting bodies should not be in contact. Cyclic symmetry may
not be used in a pure electrostatic analysis in this release.
Contact in Magnetostatic Analysis

For magnetostatics, all the degrees of freedom of the magnetic vector potential are tied for nodes in
contact. When a node is in contact with a symmetry surface, a constraint is imposed such that only the
normal component of the potential is zero. For a magnetostatic analysis, the THERMAL CONTACT
option is used. Cyclic symmetry may not be used in a pure magnetostatic analysis in this release.
Contact in Electromagnetic Analysis

Contact is not currently supported in harmonic or transient electromagnetic analysis.
Contact in Soil Analysis

When two deformable bodies are in contact, the displacements are treated in the conventional manner.
The pressure degree of freedom associated with the fluid hydrostatic pressure is tied. If a node contacts
a rigid body, nothing is done with the pressure degree of freedom.
Contact in an Acoustic Analysis

In an acoustic analysis with flexible and/or rigid walls composed of finite elements and/or rigid bodies,
a coupling interface between the acoustic medium and the walls is created to connect the fluid pressure
to the wall movement taking into account accelerations and reactive boundary conditions.
Element Considerations
Marc allows contact with almost all of the available elements, but the use of certain elements has a
consequence on the analysis procedure. Contact analysis can be performed with all of the structural
continuum elements, either lower order or higher order, including those of the Herrmann
(incompressible) formulation, except axisymmetric elements with twist. Friction modeling is available
in all of these elements except the semi-infinite elements. Higher-order isoparametric elements use shape
functions which, when the elements are loaded by a (for example) uniform pressure, lead to equivalent
nodal loads that oscillate between the corner and midside nodes. This has a detrimental effect on
determining contact separation and two procedures have been implemented to eliminate this problem:
1. Linearized Contact: The midside nodes on the exterior surface are automatically tied to the corner
nodes in this case. This effectively results in a linear variation of both the geometry and the
displacement on the exterior element edges. Hence, the elements with edges on the exterior do not
behave as full bi-quadratic (2-D) or tri-quadratic (3-D) elements.
All elements in the interior of the body behave in the conventional higher-order manner, but the
constraints on the exterior easily cause the behavior of the complete structure to be too stiff; while
in the area of contact, the stress distribution might be irregular.
2. True Quadratic Contact: No special constraints introduced on the exterior surface other than
coming from contact in this case. Both the midside and the corner nodes may come into contact
and when contact is established with another deformable body consisting of quadratic elements,
a constraint equation corresponding to the complete quadratic shape function is automatically
incorporated. Since the above mentioned oscillating nodal loads cannot be used for separation, the
decision whether or not a node should separate is based on the contact normal stress rather than
the contact normal force.
In many manufacturing and rubber analyses, the lower-order elements behave better than the
higher-order elements because of their ability to represent the large distortion; hence, these lower-order
elements are recommended.
The constraints imposed on the nodal degrees of freedom are dependent upon the type of element.
1. When a node of a continuum element comes into contact, the translational degrees of freedom
are constrained.
2. When a node of a shell element comes into contact, the translational degrees of freedom are
constrained and no constraint is places on the rotational degrees of freedom. The exception to this
is when a shell contacts a symmetry surface. In this case, the rotation about the element edge is
also constrained.
Additionally, beams and shells contact is governed by the rules outlined below.
2-D Beams
All nodes on beams are potential contact nodes. Beam elements can be used in contact in two modes.
1. The two-dimensional beams can come into contact with rigid bodies composed of curves in the
same x-y plane. The normal is based upon the normal of the rigid surface.
2. The two-dimensional beams can come into contact with deformable bodies either of continuum
elements or other beam elements. As the beams are in two dimensions, they do not intersect
one another.
CHAPTER 8 575
Contact
3-D Beams
Three-dimensional beam elements can be used in contact in three methods.
1. The nodes of the beams can come into contact with rigid bodies composed of surfaces. The
normal is based upon the normal of the rigid surface.
2. The nodes of the beam and truss elements can also come into contact with the faces of three-
dimensional continuum elements or shell elements. The normal is based upon the normal of the
element face.
3. The three-dimensional beam and truss elements can also come into contact with other beam or
truss elements (beam-to-beam contact).
The first two methods are activated by default if contact bodies consisting of beam or truss elements are
defined using the CONTACT option. The third method must be activated explicitly by additionally
switching on the beam-to-beam contact flag on the CONTACT option.
In the beam-to-beam contact model, a beam or truss element is viewed as a conical surface with a circular
cross-section. The radius of the cross-section can vary linearly between the start and end node of a beam
element. For each beam or truss element of a contact body, a contact radius must be entered via the
GEOMETRY option. The contact radius at a node follows from the average contact radius of the elements
sharing that node. Hence, the start and end node of an element may have different contact radii.
Contact is detected between two beam or truss elements if the associated conical surfaces touch each
other; that is, if the distance d between the closest points on the conical surfaces is smaller than the
distance below which bodies are considered touching each other. This is outlined in Figure 8-30, where
beam elements and their contact body representation are given. It should be emphasized that the
contacting points are points on the conical surfaces and not nodes of the finite element model.
X Z
Figure 8-30 Beam-to-Beam Contact: Finite Element Model (top) and Contact Body Representation
(bottom)
If two beam or truss elements are in contact, a multipoint constraint equation (tying) is automatically
imposed to ensure that the conical surfaces will not interpenetrate. This constraint equation involves the
displacements of the begin and end nodes of both elements. The tied node in that equation is
automatically selected, taking into account the location of the contacting points with respect to the
elements, any boundary conditions applied to the nodes and any contact between the nodes and rigid
surfaces or faces of continuum or shell elements.
During the iteration procedure, the contacting points of two beam or truss elements can change if the
elements slide with respect to each other. In addition, the points in contact can move from one element
to another. In that case, the nodes involved in the multipoint constraint equations are automatically
updated. During sliding, friction may be taken into account. Since for beam or truss elements the normal
stress in the contact points is not available, only Coulomb friction based on nodal forces (either the
arctangent or the bilinear model) is supported for beam-to-beam contact. The glue model, which imposes
that there is no relative tangential motion, is also available.
A limitation of the beam-to-beam contact model is that a contact body cannot contain branches, that is,
every element in the contact body must have a unique successor and predecessor.
Shell Elements
All nodes on shell elements are potential contact nodes. As the midside nodes of shell elements are
automatically tied, the high-order shell element (type 22) has no benefit. Shell elements can contact either
rigid bodies, continuum elements, or other shell elements. Shell-shell contact involves a more complex
analysis because it is necessary to determine which side of the shell contact occurs.
By default, the top and bottom face of a shell element (which follow from the mid face by using the
thickness offset vector) are separately taken into account. This implies that a node may be found to be in
contact based on the top or the bottom face of an element. However, due to the nature of the shell
formulation, it is not possible to simultaneously use contact conditions following from the top and the
bottom face. Touching a shell element may also occur at its top or bottom face.
The user has the option to set per pair of contact bodies on the CONTACT TABLE option how the
geometry of the shell elements should be handled. This means that both for the contacting and the
contacted body it can be defined that the contact description will be based on one of the following
options: both top and bottom (default), top only (with or without the thickness offset vector) or bottom
only (with or without the thickness update vector). For glued contact it is possible to use the top and
bottom face, but without the thickness update vector.
The beam-to-beam contact model discussed above can also be used to model shell edge-to-edge contact.
This requires switching on the beam-to-beam contact flag on the CONTACT option and, for a shell
contact body combination, the edge-to-edge contact flag on the CONTACT TABLE option. For edge-to-
edge contact, half of the thickness of the shell is used to set the contact radius.
CHAPTER 8 577
Contact
Dynamic Impact
The capability is available for both implicit time integration via the Newmark-beta or Houbolt operators,
and the explicit time integration via the (fast) central difference operator.
The Newmark-beta, Single Step Houbolt and Generalized Alpha procedure have the capability to allow
variable time steps and, when using the user-defined fixed time step procedure, the time step is split by
the algorithm to satisfy the contact conditions. Since the central difference operator is constrained to
small time steps by the stability requirement, time splitting is not used by the contact algorithm.
For most dynamic impact problems, the Single Step Houbolt method is recommended, as this procedure
possesses high-frequency dissipation. This is often necessary to avoid numerical problems by contact-
induced high-frequency oscillations. If the other dynamic operators are used, it is recommended that
numerical damping is used during the analysis, or, when using the Generalized Alpha method, that a
small spectral radius is defined. Note that for a zero spectral radius, the Generalized Alpha method is
1
identical to the Single Step Houbolt method with default parameters ( γ = 1.5 , γ = – 0.5 ).
In dynamic analysis, the requirement of energy conservation is supplemented with the requirement of
momentum conservation. In addition to the constraints placed upon the displacements, additional
constraints are placed on the velocity and acceleration of the nodal points in contact, except for the
Single Step Houbolt and Generalized Alpha method.
When a node contacts a rigid surface, it is given the velocity and acceleration of the rigid surface in the
normal direction. The rigid surfaces are treated as if they have infinite mass, hence, infinite momentum.
Global Adaptive Meshing and Rezoning

In manufacturing simulations the objective is to deform the workpiece from some initial (simple) shape
to a final, often complex shape. This deformation of the material, results in mesh distortion. For this
reason, it is often required to perform a rezoning/remeshing step. At this point, a new mesh is created;
the current state of deformation, strains, stresses, etc. is transferred to the new mesh, and the analysis is
continued. This has several consequences for contact analysis.
1. When manual rezoning is used, it is necessary to redefine the elements associated with the
deformable body that is being rezoned. This is done through the CONTACT CHANGE option.
2. If a contact tolerance was not defined by you, a new contact tolerance is calculated based upon
the new mesh. This distance can be smaller than the previously calculated tolerance, leading to
more iterations.
3. After the new mesh is created, the program redetermines the potential contact nodes. The memory
is automatically allocated to accommodate the contact data in the new mesh.
4. At the first increment after remeshing, Marc determines which nodes are in contact. If the new
mesh does not reflect the exterior surface of the old model, it is possible that a region that was
previously in contact is no longer in contact. This is usually not so serious; as most likely, the
node subsequently comes into contact. A more significant problem is if a node in the new mesh
is created which, in fact, has penetrated another body. If the newly created node is beyond another
body’s surface by more than the contact tolerance, it is possible that a poor contact analysis
results. With automatic global remeshing, steps are taken to made sure contact conditions are
preserved and incorrect penetrations are removed in the new mesh. This is achieved by correcting
the nodal position after a new mesh is created.
If rezoning occurs, rigid surfaces must be kept the same and cannot be changed.
Marc auto-rezoning feature, activated by the parameter REZONING,1 for 2-D problems and REZONING,2
for 3-D problems, automatically creates new meshes and performs rezoning to resolve mesh distortion
problems. In such cases, points (1) and (4) are treated automatically.
Adaptive Meshing
Contact between a deformable-body and a rigid body is insured such that the nodes do not penetrate the
rigid surface. It is possible that an edge of an element penetrates a rigid surface, especially where high
curvature is present because of the finite element discretization. The use of the adaptive mesh generation
procedure can be used to reduce these problems.
In addition to the traditional error criteria such as Zienkiewicz-Zhu or maximum stress, etc., you can
request that the mesh be adaptively refined due to contact. In such a case, when a node comes into
contact, the elements associated with that node are refined. This results in a greater number of elements
and nodes on the exterior region where contact occurs. This can lead to a substantial improvement in the
accuracy of the solution.
Penetration of Element Edge due to Mesh Discretization Result of Adaptive Meshing

Figure 8-31 Contact Closure Condition
The adaptive meshing procedure has consequences to the contact procedure similar to the
rezoning procedure.
If a contact tolerance was not defined by you, a new contact tolerance is calculated based upon the new
mesh. This distance can be smaller than the previously calculated tolerance, leading to more iterations.
Result Evaluation
The Marc post file contains the results for both the deformable bodies and the rigid bodies. In performing
a contact analysis, you can obtain three types of results. The first is the conventional results from the
deformable body. This includes the deformation, strains, stresses, and measures of inelastic behavior
CHAPTER 8 579
Contact
such as plastic and creep strains. In addition to reaction forces at conventional boundary conditions, you
can obtain the contact forces and friction forces imparted on the body by rigid or other deformable
bodies. By examining the location of these forces, you can observe where contact has occurred, but Marc
also allows you to select the contact status as a post file variable:
A value of 0 means that a node is not in contact.
A value of 0.5 means the node is in near thermal contact.
A value of 1 means that a node is in contact.
A value of 2 means the node is on a cyclic symmetry boundary.
It is also possible to obtain the resultant force following from contact on the deformable bodies and the
resultant force and moment on the rigid bodies. The moment is taken about the user-defined centroid of
the rigid body. The time history of these resultant forces are of significant issues in many engineering
analysis. Of course, if there is no resultant force on a rigid body, it implies that body is not in contact with
any deformable body.
Finally, if the additional print is requested via the PRINT parameter, the output file reflects information
on when a node comes into contact, what rigid body/segment is contacted, when separation occurs, when
a node contacts a sharp corner, the displacement in the local coordinate system, and the contact force in
the local coordinate system. For large problems, this can result in a significant amount of output.
The motion of the rigid bodies can be displayed as well as the deformable bodies. Rigid bodies which
are modeled using the piecewise linear approach are displayed as line segments for flat patches. When
the rigid surfaces are modeled as analytical surfaces, the visualization appears as piecewise linear in
older versions (pre-K7/Mentat 3.1) or as trimmed NURBS in subsequent versions.
Tolerance Values
On the third data block for contact, five tolerances can be set for determination of the contact behavior.
Not entering any values here means that Marc calculates values based on the problem specification.
Relative Sliding Velocity Between Surfaces Below Which Friction Forces Drop
As discussed in the Friction Modeling section, the equations of friction are smoothed internally in the
program to avoid numerical instabilities. The equations are inequalities whenever two contacting
surfaces stick to each other and equalities whenever the surfaces slide (or slip). Thus, the character of
contact constraints change depending on whether there is sticking or slipping. The smoothing procedure
consist of modifying it in such a way, that there is always slip; the amount is a function of the relative
velocity and a constant RVCNST. The value of this constant must be specified here. It actually means,
that if we specify a small value in comparison to the relative velocity, the jump behavior is better
approximated, but numerical instabilities can be expected. A large value means, that we need a large
relative velocity before we get the force at which the slip occurs.
It is suggested to use values between 0.1 and 0.01 times a typical surface velocity.
Distance Below Which a Node is Considered Touching a Surface

In each step, it is checked whether a (new) node is in contact with other surfaces. This is determined by
the distance between the nodes and the surfaces. Since the distance is a calculated number, there are
always roundoff errors involved. Therefore, a contact tolerance is provided such that if the distance
calculated is below this tolerance, a node is considered in contact. A too large value means that a high
number of body nodes are considered to be in contact with the surface and are consequently all moved
to the surface, which can be unrealistic in some applications. A too small value of this number means that
the applied deformation increment is split into a high number of increments, thus increasing the cost
of computation.
The tolerance must be provided by the analyst or can be calculated by Marc. In general, the contact
tolerance should be a small number compared to the geometrical features of the configuration being
analyzed. The value calculated by Marc is determined as 1/20 of the smallest element size of all elements
in any contact body or 1/4 of the smallest thickness of all shell or beam elements in any contact body,
whichever is the smallest. Note that for a continuum element the element size is defined here as the
smallest edge of the surrounding rectangle (2D) or box (3D), set up in the global coordinate system, while
for a shell element it is the smallest element edge and for a beam element it is the element length.
Tolerance on Nodal Reaction Force or Nodal Stress Before Separation Occurs

If a tensile force occurs at a node which is in contact with a surface, the node should separate from the
surface. Rather than using any positive value, a threshold value can be specified. This number should
theoretically be zero. However, because a small positive reaction might be due only to errors in
equilibrium, this threshold value avoids unnecessary separations. A too small value of this force results
in alternating separation and contact between the node and the surface. A too large value, of course,
results in unrealistic contact behavior.
Marc calculates this value as the maximum residual force in the structure. Note that the maximum
residual force can be specified in the CONTROL option. Default for this value, a 10 percent of the
maximum reaction force is used. Consequently, if locally high reaction forces at a particular point are
present, the separation force is large as well. In most cases, however, the default value is a good measure.
If you indicate that separation is to be based upon stresses, a value of the separation stress should be
entered. The default value is the maximum residual force at a node divided by the contact area of node n.
Numerical and Mathematical Aspects of Contact

Lagrange Multipliers
In performing contact analysis, you are solving a constrained minimization problem where the constraint
is the ‘no penetration’ constraint. The Lagrange multiplier technique is the most elegant procedure to
apply mathematical constraints to a system. Using this procedure, if the constraints are properly written,
overclosure or penetration does not occur. Unfortunately, Lagrange multipliers lead to numerical
difficulties with the computational procedure as their inclusion results in a nonpositive definite
mathematical system. This requires additional operations to insure an accurate, stable solution which
leads to high computational costs. Another problem with this method is that there is no mass associated
with the Lagrange multiplier degree of freedom. This results in a global mass matrix which cannot be
CHAPTER 8 581
Contact
inverted. This precludes the used of Lagrange multiplier techniques in explicit dynamic calculations
which are often used in crash simulations. The Lagrange multiplier technique has often been
implemented in contact procedures using special interface elements such as the Marc gap element (type
12). This facilitates the correct numerical procedure, but puts a restriction on the amount of relative
motion that can occur between bodies. The use of interface elements requires an apriori knowledge of
where contact occurs. This is unachievable in many physical problems such as crash analysis or
manufacturing simulation.
Penalty Methods
The penalty method or its extension, the Augmented Lagrangian method, is an alternative procedure to
numerically implement the contact constraints. Effectively, the penalty procedure constrains the motion
by applying a penalty to the amount of penetration that occurs. The penalty approach can be considered
as analogous to a nonlinear spring between the two bodies. Using the penalty approach, some penetration
occurs with the amount being determined by the penalty constant or function. The choice of the penalty
value can also have a detrimental effect on the numerical stability of the global solution procedure. The
penalty method is relatively easy to implement and has been extensively used in explicit dynamic
analysis although it can result in an overly stiff system for deformable-to-deformable contact since the
contact pressure is assumed to be proportional to the pointwise penetration. The pressure distribution is
generally oscillatory.
Hybrid and Mixed Methods

In the hybrid method, the contact element is derived from a complementary energy principle by
introducing the continuity on the contact surface as a constraint and treating the contact forces as
additional elements. Mixed methods, based on perturbed Lagrange formulation, usually consist of
pressure distribution interpolation which is an order less than the displacement field, have also been used
to alleviate the difficulties associated with the pure Lagrange method.
Direct Constraints
Another method for the solution of contact problems is the direct constraint method. In this procedure,
the motion of the bodies is tracked, and when contact occurs, direct constraints are placed on the motion
using boundary conditions – both kinematic constraints on transformed degrees of freedom and nodal
forces. This procedure can be very accurate if the program can predict when contact occurs. This is the
procedure that is implemented in Marc through the CONTACT option. No special interference elements
are required in this procedure and complex changing contact conditions can be simulated since no apriori
knowledge of where contact occurs is necessary. The details of the implementation are presented later.
Contact can be defined as finding the displacement of points A and B such that
( u A – u B ) ⋅ n < TOL
where A is on one body and B is on another body, n is the direction cosine of a vector between the two
points, and TOL is the closure distance.
B
A n
Figure 8-32 Normal Gap Between Potentially Contacting Bodies
Lagrange Multiplier Procedure

In the Lagrange Multiplier Procedure, the variational procedure is augmented by the constraint. The
virtual work principle leads to the traditional systems equation:
Ku = f
The constraint conditions can be expressed as

Cu = 0
Through the minimization of the augmented functional

1 T T T
Ψ = --- u Ku – u f + λ Cu
2
you obtain
K C  u  =  f 
T
C 0 λ  0 
This equation can be solved simultaneously for both the displacement ( u ) and the Lagrange multiplier
( λ ). Note that the introduction of Lagrange multipliers results in a zero on the diagonal. Hence, even well
posed physical problems no longer have positive definite systems. This often results in numerical
difficulties and restricts the type of linear equation solver that can be used.
Direct Constraint Procedure

Marc divides contact problems into two domains; the first is when a deformable body makes contact with
a rigid surface and the second is when a deformable body makes contact with another deformable body
or itself.
Two-dimensional or three-dimensional bodies are treated conceptually in the same manner.
CHAPTER 8 583
Contact
Deformable-Rigid Contact
In such a problem, a target node on the deformable body has no constraint while contact does not occur.
Once contact is detected, the degrees of freedom are transformed to a local system and a constraint is
imposed such that
Δu n o r m a l = v ⋅ n
where v is the prescribed velocity of the rigid surface. This local transformation is continuously updated
to reflect sliding of point A along the rigid surface.
If the glue option is activated, an additional displacement constraint is formed as
Δu tangential = v ⋅ t
The determination of when contact occurs and the calculation of the normal vector are critical to the
numerical simulation. Historically in Marc, three procedures are used for detecting penetration and
imposing contact, but now two of them are preferred and are discussed here. The first (and most general
one) is the Iterative Penetration Checking Procedure. The second one (Time Step Reduction Procedure)
is recommended only for a transient dynamic analysis with automatic time stepping.
In Figure 8-34, node A is not in contact with the rigid body at the beginning of an iteration. After time
step Δt , node A penetrates rigid surface if no constraint is imposed during the iteration which is not an
acceptable solution.
B B
A
A
t y
x
n
Before Contact After Contact
z
Figure 8-33 Contact Coordinate System
A A
Start of Iteration, No Contact End of Iteration, If No Constraint

Figure 8-34 Contact Constraint Requirement
The Iterative Penetration Checking Procedure checks for every iteration in the Newton-Raphson process
if the displacement solution found would cause nodes to penetrate. If this is the case, the iterative
displacement solution is scaled such that at the start of the subsequent iteration, new contact will be
established. Some mathematical details can be found at the end of this chapter.
When in a transient dynamic analysis, the adaptive time stepping procedure AUTO STEP is used, Marc
reduces the time step of the current increment such that nodes, which would penetrate based on the
original time step, will just come into contact. In the current increment, no contact constraints are applied
to these nodes. In the next increment, the nodes are given the appropriate contact constraints and a new
time step is determined based upon the convergence criteria.
Analytical Contact
Utilizing the analytical representation of rigid surfaces influences the numerical procedure in
several ways:
1. The rigid surface is represented by a nonuniform rational B-spline surface (NURBS) which
provides a precise mathematical form capable of representing the common analytical shapes –
circle, conic curve, free-form curves, surfaces of revolution, and sculptured surfaces.
The surface can be expressed in simple mathematical form to model complex multiple surfaces
with advantage of continuity in first and second derivative. Those analytical forms are expressed
in four-dimensional homogeneous coordinate space by:
n+1 m+1
  B i, j h i, j N i, k ( u )M j, l ( v )
P ( u, v ) = i----------------------------------------------------------------------------------------
= 1
n+1
j = 1
m+1
-
  h i, j N i, k ( u )M j, l ( v )
i = 1 j = 1
For curves, it simplifies to:

n+1
 B i h i N i, k ( u )
P ( u ) = i-----------------------------------------
= 1
n+1
-
 h i N i, k ( u )
i = 1
where the B are 4-D homogeneous defining polygon vertices, N i, k and M j, l are nonrational
B-spline basis functions, hi,j is homogeneous coordinates. For given parameters u and v in local
system, the location (x, y, z) in three-dimensional space, first derivative and second derivative (if
required), are calculated. Given a point with x, y and z Cartesian coordinates, there is no explicit
mathematical form to find out the parameters u and v in the local space, so they are obtained using
an iterative procedure.
CHAPTER 8 585
Contact
8 2. The location of the closest point on the contact surface corresponding to node A needs to be
determined. This process, determined analytically when using flat patches, requires an iterative
approach for rigid bodies modeled with NURBS.
Contact
Figure 8-35 Closest Point Projection Algorithm
3. Once the point P is known, the normal is calculated based upon the NURBS. This normal is used
in a manner consistent with the PWL procedure.
4. Because the normal can change from iteration to iteration, the imposition of the kinematic
boundary conditions is continuously changing.
In the PWL approach:
0
( Δu n ) = v⋅n
i
( δu n ) = 0
In the analytical approach:

0
( Δu n ) = v ⋅ n0
i
( δu n ) = v ⋅ n 1 – v ⋅ n 0
where the subscripts represent the iteration number.

5. In the PWL approach, when a node slides from one segment to another, the corner condition logic
is activated. The advantages of analytical surfaces is that this logic is not necessary until you
come to the end of the spline. This reduces the amount of iterations required.
6. The friction calculation is dependent upon the surface normal and tangent. When using the
analytical approach, the friction calculation includes the effect of changes in the direction of the
normal vector from iteration to iteration. This improves the accuracy and convergence behavior.
7. Because the normal is continuous, the calculation of nodal forces is more accurate. This allows
for the determination of nodal separation to occur beginning with the second iteration as opposed
to when equilibrium convergence has been achieved using the PWL approach.
Solution Strategy for Rigid Contact

Step 1: At the start of an increment, all boundary nodes are checked for contact with surfaces and
flagged if so. If the Iterative Penetration Checking Procedure has detected potential
penetration, this is repeated at the start of an iteration.
Step 2: Next, transformations and imposed displacements are determined for each touching node.
With the knowledge of the time increment and surface velocity, the configuration of the
surface at the end of the increment is found. Then, the distance from the node starting position
to the current surface segment is determined and is the probable normal displacement to be
imposed. If there is a previous solution for displacement increments, these are used to estimate
the tangential displacement increment (Figure 8-36). If this estimate still puts the node at the
end of the increment in the same surface segment, then this one is used to determine the
transformation matrix between local and global coordinates, and the normal displacement is
accepted. Otherwise, the procedure is repeated for the adjacent segment. If a concave corner
is hit, the boundary node is fixed to such corner. If the node goes out of a sharp convex corner,
it is immediately released from the surface.
SEGB SEGB - Die segment at beginning
P
of increment
Δu SEGE - Die segment at end of increment
Δu - Displacement increment of
previous increment
Δun Δun - Imposed normal displacement
Δutest - Estimate of tangential
displacement
SEGE
β - Angle used in
Δutest transformation matrix
β
Figure 8-36 Node P Already In Contact (2-D)
Step 3: One increment of the problem is solved iteratively. During each iteration, the Iterative
Penetration Checking Procedure checks if the current iterative displacement solution must be
scaled to avoid penetration (Figure 8-37). If needed, Step 1 is repeated to establish new contact
constraints. Step 2 is repeated at each iteration for possible changes in surface segment.
SEGB
A SEGE
B SEGB - Die segment at beginning of iteration
SEGE - Die segment at end of iteration
Δu - Iterative Displacement calculated
Δu PB
-------- - Factor to scale iteration
PA
P
Figure 8-37 New Node in Contact P
Step 4: Once convergence is obtained, nodal forces are checked. Whenever a positive normal force (or
stress) is detected which is larger than the tolerance, a node is released from the surface
segment at which such force (or stress) corresponds and Step 3: is repeated. This tolerance can
be modified using the SEPFOR or SEPSTR user subroutine.
CHAPTER 8 587
Contact
Step 5: Go to next increment (Step 1:).
PATB
Δu
Δun
Δutest
PATE
3
1
2
PATB - Patch at beginning of increment

PATE - Patch at end of increment
Δu - Previous displacement increment of Node P
Δun - Imposed normal displacement increment (local
direction 3)
Δutest - Estimated tangential displacement
Figure 8-38 Node P Already in Contact (3-D)
Deformable-Deformable Contact
When a node contacts a deformable body, a tying relation is formed between the contacting nodes and
the nodes of the contacted segment on the other body. This constraint relationship uses information
regarding the normal vector to the segment and the closest point projection of the contacting node on the
contacted segment. For lower-order elements, the normal vector and the projection are calculated based
upon the piecewise linear representation of the element. This has the consequence that the constraint
relation may not be accurate because the normal vector is constant over the complete segment, whereas
the actual structure may be curved with a varying normal vector. When a node slides from one segment
to another, there may be a discontinuity of the normal vector, which leads to potential numerical
difficulties or an inaccurate solution. This is especially important for contact analyses where the
deformations are small.
When higher-order elements are used, the normal vector and projection are created based upon the true
quadratic element description. Provided that the midside nodes are defined to be on the actual curved
geometry, this improves the accuracy, but does not necessarily result in continuity of the normal vector
at nodes shared by multiple segments.
Actual Geometry Finite Element Approximation Cubic Spline Representation
: Nodes with a normal vector discontinuity
Figure 8-39 2-D Deformable Contact; Piece wise Linear FE Description and
Cubic Spline Representation
Both for lower- and higher-order elements, you can fit a smooth curve (2-D) or surface (3-D) through the
finite element segments. For 2-D, a cubic spline is used (see Figure 8-39). This spline is calculated based
upon the tangent and position vectors of the nodes on the segment. For lower-order elements, each
segment is internally replaced by one cubic spline; where for higher-order elements, each segment is
replaced by two cubic splines. This gives a more accurate representation of the actual physical geometry
and a more accurate calculation of the normal vector. You must identify where actual discontinuities
exist. This is done by defining nodes where the normal vector is discontinuous or by activating the option
that the program will determine such nodes based on the angle spanned by the normal vectors to adjacent
segments. You can use the SPLINE option to define deformable bodies in 2-D or 3-D where the
smoothness is required.
For 3-D problems, the concept of a cubic spline can be expanded to generate a Coons surface to improve
the geometric representation of a segment (see Figure 8-40). When calculating the closest point
projection, it is advantageous to replace the Coons surface description by a cubic serendipity surface. For
2-D problems, you have to define nodes where the actual normal vector has a discontinuity; for 3-D
problems, you have to specify edges where the normal vector is discontinuous. The SPLINE option
allows you to enter a list of nodes (where each edge is defined by its start and end node) or to activate the
option that the program will determine such edges based on the angle spanned by the normal vectors to
adjacent segments. At edges with a discontinuous normal vector, the smooth surface description of
adjacent segments can be set up independently of each other, which usually results in C0-discontinuity.
Alternatively, the smooth surface description of adjacent segments can be forced to include information
from the interface they have in common, which results in C0-continuity. You can force C0-continuity in
3-D also on the SPLINE option.
If the SPLINE option is used for curved contact bodies defined by higher-order elements, the coordinates
of the midside nodes can be recalculated based on the cubic splines defined by the position and tangent
vectors of the corner nodes. This can be advantageous if the structure to be modeled is available only
based on lower-order elements, so that a change from lower- to higher-order elements yields midside
nodes which are not on the curved boundary, but on the straight edges between the corner nodes.
CHAPTER 8 589
Contact
C0-discontinuous Coons
Surface Representation
Actual Geometry Finite Element Approximation
C0-continuous Coons
Surface Representation
Figure 8-40 3-D Deformable Contact; Piece wise Linear FE Description and
Coons Surface Representations
Solution Strategy for Deformable Contact

1. Upon initiation, the program determines the boundary nodes on each body. For 2-D, these nodes
are ordered in sequence in order to describe the boundary in a counterclockwise manner. Linear
geometrical entities are then created out of subsequent pairs of nodes, to define a surface profile
as in a rigid surface. If higher-order elements are used in full quadratic mode, quadratic entities
are created from the three nodes. For 3-D using lower-order elements, 4-node patches are created.
For 3-D higher-order hexahedral elements, a quadratic surface is created.
2. Once contact between a node and a deformable surface is detected, a tie is activated. The tying
matrix is such that the contacting node follows the shape of the surface (Figures 8-41 and 8-42);
it can slide along it or be stuck according to the general contact conditions.
3. Contact occurs between all deformable bodies unless the CONTACT TABLE option is used.
In deformable contact, there is no master or slave body; each body is checked against every
other body, except single-sided contact or the CONTACT TABLE option is used.
4. During the iteration process of finding an incremental solution, a node can slide from one segment
to another, changing the retained nodes of its tie. A recalculation of the bandwidth is, therefore,
made every iteration. From increment (or subincrement) to increment, the number of nodes in
contact can change. In such cases, an optimization of the bandwidth is automatically available to
cope with the possible drastic changes in bandwidth that a new tie produces.
When the glue option is used between two deformable bodies, a simpler tying relationship is
formed, such that no relative motion occurs.
Tied Node A
A Retained Nodes A B C
segt
B
ΔUB ΔUA ΔUC
segt+Δt
Figure 8-41 Tyings in Deformable Contact (2-D) using Lower-order Elements
R2
R3 T/R1
R5
t Tied Node T
Retained Nodes T, R2, R3,R4,R5
R4
t+ Δt
Figure 8-42 Tying in Deformable Contact (3-D) using Lower-order Elements
Iterative Penetration Checking

Except for a transient dynamic analysis with automatic time stepping where the Time Step Reduction
Procedure is used, the commonly used method to prevent penetration in a contact analysis is the Iterative
Penetration Checking Procedure. Using this procedure, the iteration process is done simultaneously to
satisfy both the contact constraints and global equilibrium using the Newton-Raphson procedure. This
procedure is accurate and stable. This procedure is automatically activated when the AUTO STEP
procedure is used in static analysis or when beam-beam contact is present.
In a conventional iteration process, the finite element system calculates for each iteration:
K T δu i = R i – 1
where K T is the tangent stiffness matrix and R i – 1 are the residuals based on displacements from the
previous iteration.
CHAPTER 8 591
Contact
Using the iterative penetration method K T is now based upon the contact status at this iteration, both
from a constraint and friction perspective. Furthermore, after the solution for δu i is obtained, the
contact procedure is used to determine if new penetration will occur. If, at least, one node penetrates a
contact surface, a scale factor is applied to the change in displacements such that the penetrating nodes
are moved back to the contact surface. This procedure can be considered a type of line search. Given
that s is the fraction of δu i such that no new penetration occurs, the displacement increment then
becomes
ΔU i = ΔU i – 1 + sδu i
and the total displacement is
U n = ΔU n – 1 + ΔU i
The strains, stresses, and residuals are based upon these quantities. Separation is based upon these
values. When global equilibrium is achieved, based upon the user’s criteria, the solution proceeds to the
next increment. Because the procedure can reduce the change in displacements, it may require more
iterations to complete an increment. It is important to ensure that the maximum allowable number of
iterations to complete an increment is set to a sufficiently large value.
Instabilities
In some analyses, because of instabilities such as buckling or a loss of contact, large displacement
increments may occur. In such cases, it is possible to specify a maximum iterative displacement
component allowed δu a l l o w e d . If the magnitude of any iterative displacement component turns out to
be larger than this maximum value, the iterative displacement solution is scaled using a factor s , such
that:
s δu i ≤ δu a l l o w e d
for any node in the model. If δu a l l o w e d is not specified, the value used defaults to the average of the
maximum dimension of the global rectangular boxes enveloping the contact segments. In case of fine
meshes, this default value can be small and scaling of the solution vector can be reduced by specifying
a more physically meaningful length scale.
Finally, if the solution is still not able to converge and the cutback feature is activated, the time step is
reduced in magnitude and the increment is repeated.
Segment-to-Segment Contact
The contact algorithm discussed above is based on nodes being in contact with a segment (a curve,
surface, element edge, or element face), and can thus be called a node-to-segment algorithm. This contact
algorithm has matured and has been successfully applied to a large variety of contact problems. However,
there are a few weaknesses of this algorithm especially for deformable contact, which call for the
development of an alternative contact algorithm:
• When contact has been detected, the nonpenetration constraint is enforced on a nodal basis.
Because of this point-wise application of constraints, the node-to-segment algorithm does not
generally maintain stress continuity across the contact interface of deformable contact bodies.
This can be easily illustrated using the contact patch test, where a uniform stress distribution
should be obtained in two deformable contact bodies with dissimilar meshes (see [Ref. 7]).
• Since the nonpenetration constraints are enforced using multi-point constraint equations, there is
a potential dependency of the solution on the selection of the master and the slave nodes. In other
words, the solution depends on which nodes are touching and which nodes correspond to a
touched contact segment. Although there are various options to optimize the multi-point
constraint equations, it is not always possible to achieve the best results everywhere in the model
and to completely eliminate the dependency on the contact body numbering.
• If there would be contact detected at the top and the bottom face of a shell element, it is not
possible to only use the nodes of the shell elements to apply both multi-point constraint equations.
This implies that there are limited modeling options for so-called double sided shell contact. A
similar problem occurs when the edge of a shell element is in contact with the face of another
element. If this happens, then the contact constraints are based on either the shell top or the shell
bottom face. Consequently, the “footprint” of the shell edge is not directly related to the shell
thickness.
In Marc 2010, a new contact algorithm has been introduced based on a segment-to-segment algorithm,
where the nonpenetration constraints are enforced using augmented Lagrangians. In this section, the
basic aspects of this algorithm will be discussed. It should be noted that currently the segment-to-segment
algorithm has the following limitations:
• Mechanical contact only; no thermal or coupled contact;
• Small sliding for deformable-deformable contact (large sliding is allowed for deformable-rigid
contact);
• No friction (glued contact is allowed).
Basic Theory
For a set of deformable contact bodies, the virtual work equation commonly used as the basis for the
derivation of the standard set of equations to be solved within a finite element analysis can symbolically
be written as:
G ( u, δu ) = 0 (8-4)
where u is the displacement field and δu the kinematically admissible variation of this field.
CHAPTER 8 593
Contact
Emphasis will first be on frictionless mechanical contact. The corresponding contact conditions can be
added to the virtual work equation as follows:
G ( u, δu ) +  ( δg n λ n + g n δλ n ) dΓ = 0 (8-5)
Γ
Here Γ is that part of boundary of the bodies being in contact and the subscript n is used to indicate the
normal direction to the contact boundary. The function g n is called a gap function and expresses the
distance between a point and its closest point projection on the contact boundary. If g n > 0 , a point is
outside the contact boundary, if g n = 0 , a point is on the contact boundary and if g n < 0 , a point would
be beyond the contact boundary, which is physically inadmissible. The Lagrange multiplier λ n
represents the contact normal stress.
The solution procedure will be based on an augmented Lagrangian approach. To this end, the previously
introduced Lagrange multiplier λ n will be rewritten as:
λn = pn + En gn (8-6)
while it is assumed that:
δλ n = 0 (8-7)
In equation (8-6), p n is a fixed estimate of the Lagrange multiplier λ n and E n is the penalty
regularization factor. It can easily be seen that if a contact condition is exactly fulfilled, then the gap
function g n = 0 and p n = λ n . The default value of the penalty factor is derived from the contact body
properties (this will be discussed later on), but can also be user defined.
The gap function g n depends on the geometry of the body (defining the normal vectors) and the
displacement field u . Neglecting the variation of the normal vectors, the contact part of equation (8-5)
can be expressed as:
∂g n
 δg n ( p n + E n g n ) dΓ =  -------
∂u
- δu ( p n + E n g n ) dΓ (8-8)
Γ Γ
Since equation (8-5) has to be valid for all kinematically admissible variations δu , the normal contact
contribution to the equilibrium equation is given by:
∂g n
 ( p n + E n g n ) -------
∂u
- dΓ (8-9)
Γ
Within the Newton-Raphson process to solve the global system of equations, the first term in equation
(8-9) contributes to the right hand side or force vector, where the second term contributes to the stiffness
matrix.
The evolution of the contact normal stress during the Newton-Raphson process is illustrated in
0
Figure 8-43. The first displacement solution is obtained using p n = p n (which initially can be zero, but
in an arbitrary increment is given by the value at the end of the previous increment). Given the
1
displacement solution, the gap function g n is calculated and the contact normal stress update is given by
1 0 1
p n = p n + E n g n . This new contact normal stress is used in a subsequent iteration. If the augmentation
is not performed, then the method reduces to a penalty method with a constant penalty factor E n .
In the paragraphs below, it will be shown how equation (8-9) will be numerically integrated in a 2-D or
3-D contact analysis.
p
n
exact value of p n
2
pn
1
pn
atan ( E n )
0
p
n 2 1
gn gn gn
Figure 8-43 Updating Contact Normal Stress according to the Augmented Lagrange Procedure
Contact Body Definition

The definition of contact bodies for the segment-to-segment algorithm is identical to that of the node-to-
segment algorithm. This implies that one can define deformable contact bodies consisting of finite
elements, and rigid bodies consisting of curves (2-D) or surfaces (3-D). Rigid bodies can be load, velocity
or displacement controlled; load controlled rigid bodies have one (translation only) or two (translation
and rotation) control nodes.
Based on the finite elements assigned to a contact body, the program will automatically set up the outer
boundary of the deformable bodies. This boundary is defined by element edges (2-D) or element faces
(3-D). Figure 8-44 shows the basic concept for 2-D continuum and 2-D shell/beam elements. For 2-D
continuum elements, the contact segments coincide with the element edges. For 2-D shell/beam
elements, not only the top and the bottom are separately included in the contact body description, but also
the free ends of the finite element model are converted into contact segments. The same concept is
applied to 3-D continuum elements and 3-D shell elements, where for the shell elements the free edges
CHAPTER 8 595
Contact
of the finite element model are converted into contact segments. The user can also decide to ignore the
shell thickness, in which case the top and the bottom segments have the same position, but an opposite
orientation. The contact segments of the free ends will always include the shell thickness.
: nodal points
: auxiliary points
Finite Elements Contact Segments Auxiliary Versus Nodal Points

Figure 8-44 2-D Contact Bodies: Continuum Elements (Upper); Shell/beam Elements (Lower)
The SPLINE option, as discussed before for node-to-segment contact, is available for segment-to-
segment contact. When this option is selected for a shell element or a 2-D beam element, then the
transition from the top and bottom contact segments to the segment at the free ends is automatically
marked to have a normal vector discontinuity.
Contact Detection
As also shown in Figure 8-44, each contact segment will get a number of auxiliary points. These
auxiliary points are only used during the contact detection phase of the analysis and they are defined at
fixed positions on the contact segments. Since the auxiliary points are defined on a per segment basis,
there are multiple points with possibly different normal vectors at adjacent contact segments. Especially
at sharp corners, this allows for multi-body contact in such a way that each (part of a) segment can touch
only one other segment.
The actual contact detection phase consists of two passes. During the first pass, there is a check whether
an auxiliary point is sufficiently close to a contact segment and if the corresponding normal vectors are
“almost opposite”. Hence there is a distance check followed by a direction check (see also Figure 8-45).
During the distance check, the distance between an auxiliary point and its closest point projection on a
contact segment is calculated and compared with a reference value. This reference value is determined
by the contact tolerance ε and the bias factor B , where 0 ≤ B < 1 . The default contact tolerance is
calculated by the program as the smaller of 5% of the smallest element side or 25% of the smallest (beam
or shell) element thickness, based on all the elements defined in any contact body. Alternatively, the user
can define the contact tolerance. The default bias factor is zero. A non-zero bias factor, entered by the
user, implies that the outside zone becomes smaller and the inside zone larger and usually avoids
unnecessary iterations following from physically unrealistic contact detection. If the position of an
auxiliary point is x a and that of its closest point projection on a segment s of contact body is x s , then
the distance check is passed if x a – x s ≤ ( 1 – B )ε when the auxiliary point lies outside the contact
body, and if x a – x s ≤ ( 1 + B )ε , when the auxiliary point lies inside the contact body. If the distance
check is passed, then the direction check is performed. This direction check is passed if the angle between
the normal vector at the auxiliary point and the normal vector at the potentially contacted segment is
larger than a threshold value α , so ∠( n a, n s ) ≥ α . The default value of α is 120° , but it can be user
defined using the PARAMETERS option.The angle between the normal vectors is defined according to
Figure 8-46. Once both the distance and the direction check are passed, the corresponding segments are
marked as a pair of potentially contacting segments and for each of these pairs, the projection of one
segment on the other is determined. In 2-D, this projection defines a polyline, in 3-D it defines a polygon.
In Figure 8-47, a 2-D and a 3-D example is given. Each polyline/polygon has a number of
polygon/polygon points, which will serve as the basis for the numerical integration of equation (8-9). It
should be noted that the polyline/polygon points are not the same as the auxiliary points introduced
before. The auxiliary points only serve to find potential contact pairs, the polyline/polygon points
characterize the actual contact area between such contact pairs.
: closest point projection
na 1
( 1 + B )ε
( 1 – B )ε
na 2
ns ns
Distance Check And Direction Check Passed Distance Check Passed, Direction Check Failed
Figure 8-45 Contact Detection, Phase 1: Distance And Direction Check
na
∠( n a, n s )
ns
Figure 8-46 Definition of the Angle Between Normal Vectors

CHAPTER 8 597
Contact
: polyline points
Figure 8-47 Common Areas of Contact Segments: Polylines (Left) And Polygons (Right)
During the second pass of the contact detection phase, for each polyline/polygon point it is checked
whether or not the segments are locally in contact. This check is equivalent to the check which has been
described above, so a distance check followed by a direction check. At the end of this second pass, there
is a set of polyline/polygon points being in contact and representing the contact area.
Force Vector and Stiffness Matrix Contribution

The polyline/polygon points discussed above basically define a local connection between two contact
segments. Indicating such a point by a subscript i , the gap function g n i can be evaluated as the normal
displacement difference between a point and its closest point projection on the contacted segment. The
displacement of a point on a contact segment is a function of the displacements and (in case of shell
elements) rotations of the nodes corresponding to that segment. By collecting all the nodal displacement
and rotation degrees of freedom in a vector U , the gap function can be written as:
T T
g n i = ( G i – G' i )U (8-10)
Here, vector G i expresses the dependency of the normal displacement of a point on a contact segment
on the total set of degrees of freedom. In a similar way, G' i is used for the closest point projection of this
point on the contacted segment. It should be noted that, if applicable, G i and G' i also include the effect
of the shell thickness and shell offset vectors. Care has been taken to define G i and G' i in such a way
that they are able to represent rigid body modes of structures being in contact without introducing
spurious energies.
Equation (8-10) can be used to numerically evaluate equation (8-9). First, the contribution to the force
vector is given by:
N
∂g n
 p n -------
∂u
- dΓ =  ( G i – G' i )p n i ΔΓ i (8-11)
Γ i = 1
in which N is the total number of polyline/polygon points, p n i is the current contact normal stress in
polyline/polygon point i and ΔΓ i is the area corresponding to this point.
Next, the contribution to the global stiffness matrix is:

N
∂g n T T
 E n g n -------
∂u
- dΓ =  ( G i – G' i )E n i ( G i – G' i )ΔΓ i (8-12)
Γ i = 1
The right choice of the penalty factor is important to get accurate results in an acceptable number of
iterations. If the penalty factor is too small, then there will be a relatively large amount of penetration,
which has to be removed by the augmentation process outlined above. However, if the penalty factor is
small, then the first order correction process illustrated in Figure 8-43 will converge slowly. On the other
hand, if the penalty factor is too large, the overall system of equations may be ill-conditioned and a non-
smooth contact stress field may be found.
The default value of the penalty factor E n i depends on the body stiffness of the two contact bodies
involved and a characteristic length (note that the dimension of the penalty factor is force per cubic
length). The body stiffness is either defined by the average trace of the initial stress-strain law of the
elements defining the two contact bodies or by the average bulk modulus for (nearly) incompressible
rubber materials, whichever of the two is the largest. For continuum elements, the characteristic length
is given by one half of the average length of all the edges being part of the contact boundary. For shell
elements, the characteristic length is given by half of the average thickness of all the shell elements being
part of a contact body. When there is contact between a solid and a shell element, then the characteristic
length is defined by the shell element. If polyline/polygon point i defines contact between the
deformable bodies k and l , with body stuffiness S k and S l , and the characteristic length of the model
is L , then the default value of E n i is given by:
0.5 ( S k + S l )
E n i = ------------------------------- (8-13)
L
In case of contact with a rigid body, the equations shown before can be applied with minor modifications.
First, the displacements G' i U of the closest point projection directly follow from the motion of the rigid
body. Second, the issue of an ill-conditioned system does not easily occur and the default penalty factor
can be chosen to be larger. Since there is no body stiffness associated with a rigid body, the default value
of E n i is related to the deformable body only and given by:
CHAPTER 8 599
Contact
1000S k
E n i = ------------------ (8-14)
L
As mentioned before, the user can define the penalty factor, either globally via the CONTACT option or
per body combination via the CONTACT TABLE option.
Iterative Penetration Checking and Contact Stress Update

After each solution in the global Newton-Raphson process, there is a check if the iterative displacement
field would cause penetration of contact bodies. This check involves first the auxiliary points on the
contact segments not yet being marked as contact pairs and next the polyline/polygon points not yet
being in contact. The maximum amount of penetration at the auxiliary and polyline/polygon points is
computed and if this penetration is larger than the inside contact tolerance distance ( 1 + B )ε , then the
entire iterative displacement solution is scaled back to avoid this penetration.
i
Once the iterative displacement solution ΔU is known, the gap function can be updated with
i T T i i i–1 i
Δg n i = ( G i – G' i )ΔU and g n i = g n i + Δg n i , where the superscript i indicates the iteration
number. The update of the contact normal stress is given by:
i i–1 i i
pn i = pn i + E n i Δg n i + E n i g n i (8-15)
Note that the last term in equation (8-15) results from the augmentation procedure.
Since there are in general more polyline/polygon points on a contact segment than there are nodes on
such a segment, enforcing a zero gap function everywhere would in general result in an over-constrained
system. To this end, a reduction of the penetration field has to be applied. The following options are
available:
• No augmentation. This results in a pure penalty method and is recommended for most analyses,
since it gives reasonably accurate results in a relative small number of iterations.
• Augmentation based on a constant penetration field. This is the recommended procedure for
linear finite elements.
• Augmentation based on a (bi-)linear penetration field. This procedure should be used only for
quadratic finite elements.
It is also possible to let the program decide which of the augmentation methods should be applied
(constant or linear). Then the choice is made based on the combination of elements corresponding to the
contacting segments. It is noted that in case of contact with a rigid body, the augmentation procedure is
always applied and the restriction of small sliding does not apply.
In case of augmentation based on a constant penetration field, the average gap function per contact
segment is determined and applied to each polyline/polygon point corresponding to that segment.
In case of augmentation based on a (bi-)linear penetration field, a least squares method is used to
determine the values of the gap function in the corner nodes of the segment. Based on the position of a
polyline/polygon point on this segment, a (bi-)linear interpolation between the corner values then gives
the local value in the point.
Since an exact fulfillment of a zero gap function is difficult to achieve, a threshold value is needed to
determine whether or not an extra iteration due to the augmentation procedure is needed. This penetration
distance beyond which an augmentation will be applied can be calculated by the program or entered by
the user. The default penetration distance beyond which an augmentation will be applied is defined as:
ε a u g = 0.001L (8-16)
where the characteristic length L is the same as used in equation (8-13). Similar to the penalty factor, the
user can define the augmentation distance either globally for the analysis on the CONTACT option or per
body pair on the CONTACT TABLE option.
If an augmentation procedure is selected and the maximum penetration is beyond ε a u g , then an extra
Newton-Raphson iteration will take place, as long as the maximum number of iterations has not been
exceeded. If the maximum number of iterations has been reached, then the program will continue with
the next increment if the usual convergence criteria have been fulfilled; so the analysis will not stop
because of the augmentation procedure.
Separation
Since in each polyline/polygon point the contact normal stress is calculated, separation will always be
based on stresses. By default, if the contact normal stress is in tension, then the corresponding
polyline/polygon point will separate, which implies that this point will no longer contribute to the global
stiffness matrix and force vector.
In order to speed up the separation process on a structural level, there is an extra check performed on a
contact segment basis. If there are points on a segment which separate, then the total contact normal force
F nt due to points on this segment being in tension (negative) is compared with the total contact normal
force F nc due to points on this segment being in compression (positive). If F nc < 0.75 F n t , then all
the points on this segment will separate. This extra check is only active if the separation stress is default
(zero). As soon as the user enters a non-zero separation stress, either globally for the analysis on the
CONTACT option or per body pair on the CONTACT TABLE option, this extra check will be skipped.
Postprocessing
The following contact related results are available for postprocessing:
• Total forces and moments on contact bodies.
• Motion of rigid bodies.
• Contact status. This is a nodal quantity and set to one as soon as there is contact in a
polyline/polygon point on a contact segment to which this node belongs. So for segment-to-
segment contact, the contact status is shown on all bodies that are in contact. This simplifies the
identification of segments that are in contact and is independent of the body ordering.
• Contact normal force. This is a nodal vector directly following from the integration of the contact
normal stress in the polyline/polygon points.
• Contact normal stress. This is a nodal vector representing the normal stress in the
CHAPTER 8 601
Contact
polyline/polygon points. This nodal representation is based on a (bi-)linear field following from
the polyline/polygon values using a least squares approximation. Especially when some of the
polyline/polygon points involved are not in contact, then this nodal contact stress vector may
show large fluctuations. It should be noted that this is a post processing issue only.
The discussion above on normal contact can be readily extended to tangential glued contact. For glued
contact, the contact friction force and friction stress vectors are available for post processing.
References
1. Oden, J. T. and Pires, E. B. “Nonlocal and Nonlinear Friction Laws and Variational Principles for
Contact Problems in Elasticity,” J. of Applied Mechanics, V. 50, 1983.
2. Ju, J. W. and Taylor, R. L. “A perturbed Lagrangian formulation for the finite element solution
of nonlinear frictional contact problems,” J. De Mechanique Theorique et Appliquee, Special
issue, Supplement, 7, 1988.
3. Simo, J. C. and Laursen, T. A. “An Augmented Lagranian treatment of contact problems
involving friction,” Computers and Structures, 42, 1992.
4. Peric, D. J. and Owen, D. R. J. “Computational Model for 3-D contact problems with friction
based on the Penalty Method,” Int. J. of Meth. Engg., V. 35, 1992.
5. Taylor, R. L., Carpenter, N. J., and Katona, M. G. “Lagrange constraints for transient finite
element surface contact,” Int. J. Num. Meth. Engg., 32, 1991.
6. Wertheimer, T. B., “Numerical Simulation Metal Sheet Forming Processes,” VDI BERICHET,
Zurich, Switzerland, 1991.
7. Feng, Q. and Prinja, N.K., NAFEMS Benchmark Tests for Finite Element Modelling of Contact,
Gapping and Sliding, NAFEMS, 2001.
Chapter 9 Boundary Conditions
9 Boundary Conditions

Loading 605
 Kinematic Constraints 632

Mesh Independent Connection Methods 670
Marc is based on the stiffness method and deals primarily with force-displacement relations. In a linear
elastic system, force and displacement are related through the constant stiffness of the system; the
governing equation of such a system can be expressed as
Ku = F (9-1)
where K is the stiffness matrix and u and F are nodal displacement and nodal force vectors,
respectively. Equation (9-1) can be solved either for unknown displacements subjected to prescribed
forces or for unknown forces (reactions) subjected to prescribed displacements. In general, the system
is subjected to mixed (prescribed displacement and force) boundary conditions, and Marc computes both
the unknown displacements and reactions. Obviously, at any nodal point, the nodal forces and nodal
displacements cannot be simultaneously prescribed as boundary conditions for the same degree of
freedom.
Note: You must prescribe at least a minimum number of boundary conditions to insure that rigid
body motion does not occur.
The prescribed force boundary conditions are often referred to as loads and the prescribed displacement
boundary conditions as boundary conditions.
Note: Boundary conditions can be prescribed in either the global or a local coordinate system. A
nodal transformation between the global and the local coordinate systems must be carried
out if the boundary condition is prescribed in a local system.
In a nonlinear stress analysis problem, Marc carries out the analysis incrementally and expresses the
governing equation in terms of the incremental displacement vector δu and the incremental force vector
δF .
Kδu = δf (9-2)
Consequently, you must also define both the loads and the prescribed nodal displacements incrementally.
In addition to the prescribed displacement boundary conditions, constraint relations can exist among the
nodal displacements. For example, the first degree of freedom of node i is equal to that of node j at all
times. The expression of this constraint relation is
ui = uj (9-3)
Generally, a homogenous linear constraint equation can be expressed as
ut = a 1 u1 + a 2 u2 + … + an un (9-4)
where u represents the degrees of freedom to be constrained, u 1, …, u n are other retained degrees of
freedom in the structure, and a 1, …, a n are constants.
CHAPTER 9 605
Boundary Conditions
You can enter constraints through either the TYING, SERVO LINK, RBE2, or RBE3 options.
You can use linear/nonlinear springs and foundations to provide special support to the structure and the
gap and friction element (or CONTACT option) to simulate the contact problem.
Loading
Different types of analyses require different kinds of loading. For example, the loads in stress analysis
are forces; those in heat transfer analysis are heat fluxes. Force is a vector quantity defined by magnitude
and direction; heat flux is a scalar quantity defined by magnitude only. Loading can be time invariant
(constant value) or time dependent.
Loading Types
You can categorize a particular type of load as either a point (concentrated) load or surface/volumetric
(distributed) load, depending on application conditions. The spatial distribution of the load can be
uniform or nonuniform. Special loading types also exist in various analyses. For example, centrifugal
loading exists in stress analysis, and convection and radiation exist in heat transfer analysis. You can add
point loads directly to the nodal force vector, but equivalent nodal forces first must be calculated by Marc
from distributed loads and then added to the nodal force vector. These distinguishing features are
described below.
A point (or nodal) load of either a vector (force, moment) or a scalar (heat flux) quantity is a concentrated
load that is applied directly to a nodal point (see Figure 9-1). Mechanical point loads can be defined as
fixed direction forces or as follower forces. In a global or local coordinate system, a force vector must
be defined in terms of its vector components (see Figure 9-2). If the force vector is defined in a local
coordinate system, then a global-to-local coordinate transformation matrix must be defined for the nodal
point (see Figure 9-3 and Figure 9-4). For axisymmetric elements, the magnitude of the point load must
correspond to the ring load integrated around the circumferences.
Q F
y y
x x
Point (a) Point (b)
Heat Flux Q (Scalar) Force F (Vector)
Figure 9-1 Schematic of a Point Load
y’
Fy F F
Fy’
Fx Fx’
y y x’
x x
Figure 9-2 Force Components
y” x”
Fy” = 0
y F = Fx”
x
Special Selection of Local (x”, y”) Coordinate System Force Components: F
Figure 9-3
y'' = 0
Surface/volumetric loads, such as pressure, distributed heat flux, and body force, are distributed loads that
are applied to the surface (volume) of various elements (see Figure 9-4). A surface/volumetric load is
characterized by the distribution (uniform/nonuniform) and the magnitude of the load, as well as the
surface to which the load is applied (surface/volume identification). The total load applied to the surface
(volume) is, therefore, dependent on the area (interior) of the surface (volume).
Equivalent nodal forces first must be calculated from surface/volumetric loads and then added to the
nodal force vector. Marc carries out this computation through numerical integration. (See Marc Volume
B: Element Library for the numbers and locations of these integration points for different elements).
The calculated equivalent nodal forces for lower-order elements are the same as those obtained by
equally lumping the uniformly distributed loads onto the nodes. However, for high-order elements, the
lumping is no longer simple (see Figure 9-5). As a result, the surface/volumetric loads should not be
lumped arbitrarily.
CHAPTER 9 607
Boundary Conditions
P1 P2 P1 P2
4 3 4 3
Px
Q
Py
1 2 1 2
y y
P q
x x
(a) Distributed Mechanical Load (b) Distributed Heat Flux
Surface 2-3:
Uniform Normal Pressure p Uniform Heat flux q
Surface 3-4:
Nonuniform Normal Pressure p1 - p2 Nonuniform Heat flux q1 = q2
Whole Volume:
Volumetric Loads Px’Py Volumetric Heat Flux Q
Figure 9-4 Schematic of Surface/Volumetric Load
1/2 1/2 1/4

1/4 1/4 1/4
1/4
1/4
1/4 1/4 1/6 2/3 1/6 -1/12

-1/12 1/3 -1/12 1/3
-1/12 1/3
-1/12
1/3 1/3
1/3
1/3
-1/12
-1/12 1/3 -1/12
Figure 9-5 Allocation of a Uniform Body Force to Nodes for a Rectangular Element Family
Using the table driven input format, boundary conditions may be applied to finite element entities, nodes,
elements, edges or faces, or geometric entities such as points, curves, and surfaces. The latter is
advantageous because it improves compatibility with CAD and because it facilitates adaptive meshing.
The application of boundary conditions on geometric entities is only effective if finite elements have
been associated with these geometric entities using the ATTACH NODE, ATTACH EDGE, and
ATTACH FACE model definition options. As shown in Figure 9-6, Marc Mentat shows a 2-D model
consisting of four curves around the perimeter and a curve where the circle is. The mesh was created
using the automatic mesh generator which automatically attaches the edges to the curves. Those nodes
which are attached to points are shown using the circle symbol as opposed to the square symbol. Now, if
a boundary condition, such as an internal pressure, is applied on a curve during the analysis, it will
automatically be applied to those edges attached to the curve as shown in Figures 9-8 and 9-9. In a similar
manner, if a displacement is applied to a point, then it will be applied to the node. For 3-D, the analogous
situation exists; namely, a distributed load applied to a surface will be applied to those elements attached
to this face as shown in Figure 9-10. And nodal boundary conditions will be attached to nodes attached
to face.
Figure 9-6 Geometric Representation with Five Curves
Figure 9-7 Original Mesh, Red Edges Indicate that they are Attached to Curve
CHAPTER 9 609
Boundary Conditions
Figure 9-8 Boundary Conditions Applied to Curves
Figure 9-9 Boundary Conditions Applied to Attached Element Edges

Figure 9-10 Distributed Load Applied to Surfaces. In Marc Mentat, Faces Attached to Surfaces are
Indicated by a Darker Color (usually Blue and Midnight Blue)
Now, if global or local adaptive meshing is used, then, as shown in Figures 9-11 and 9-12, the load
automatically is applied to the new element edges which are attached to the curve. If a point load of
prescribed displacement is applied to a geometric point, the mesh generator will make sure that a
point/node is in this exact location after remeshing occurs. This allows all boundary conditions to be used
in conjunction with global or local adaptive meshing in 2-D. In 3-D, one can apply boundary conditions
to geometric entities as well, but they are not updated with global adaptive meshing in this release.
For the case of shells, this is complicated because, for certain types of load, there is a difference between
application to the top and bottom surfaces. In mechanical analysis, a positive pressure on the top surface
is equivalent to a negative pressure applied to the bottom surface. This is because the degrees of freedom
of a mechanical shell are considered to be at the neutral surface. For thermal analysis, distributed fluxes,
convective boundary conditions and radiation, there is a difference between application of thermal
boundary conditions on the top or bottom surfaces. This is particularly relevant for radiation, as the
choice of surfaces will strongly influence the viewfactor calculation. Figure 9-13, shows a surface, with
the top (yellow) and bottom (blue) surfaces identified. To select either the top or bottom surface, use the
FILTER command. A boundary condition applied to the “top” and “bottom” surface is shown in Figures
9-14 and 9-15, respectively. If the shell mesh is automatically generated from the surface, the top face of
the element is associated with the top surface of the shell. It is also possible to display the shell elements
to indicate the thickness as shown in Figure 9-16, in which case, one can easily select the top or bottom
faces for which the boundary condition is to be applied.
CHAPTER 9 611
Boundary Conditions
Figure 9-11 Distributed Load on Top Surface after Global Adaptive Remeshing
Figure 9-12 Distributed Load to Top Surface after Local Adaptive Remeshing
Figure 9-13 Identification of Top and Bottom Surface of a Shell

Figure 9-14 Boundary Condition on Top Surface
Figure 9-15 Boundary Condition on Bottom Surface
Figure 9-16 Shell Elements shown in Expanded Mode
Figure 9-17 is an example of what can go wrong if either geometric entities or elements are not
consistently defined. Two curves are used to represent an axisymmetric shell, but as shown by the
orientation vector, they are not continuous in direction.
CHAPTER 9 613
Boundary Conditions
Figure 9-17 Oriented Curves
Figure 9-18 shows the finite element mesh obtained from the convert option, which does not indicate any
potential problem. If one displays the shell in expanded mode and identifies backfaces, then one will see
an inconsistency between the top (blue) and the bottom (brown) surfaces as shown in Figure 9-19.
Figure 9-18 Finite Element Mesh
Figure 9-20 shows the boundary condition applied to the “top” curve, which is not what is desired. To
get a consistent applied boundary condition, one needs to flip both the curves and elements of one of
the sections.
Figure 9-19 Shell Elements Expanded showing Top (Blue) and Bottom (Brown) Surfaces
Figure 9-20 Incorrect Boundary Condition on Top Surface
Face ID for Distributed Loads, Fluxes, Charge, Current,

Source, Films, and Foundations
When using the table driven input procedure, there are three ways to identify the type of load and which
edge or face it is applied to.
In the first method. an IBODY, as defined for each element type defined in Marc Volume B: Element
Library, is associated with each distributed load list. This IBODY effectively defines the type (normal
pressure, shear, gravity, pressure, centrifugal, Coriolis) and the edge or face to which the load is applied.
For the second method, one needs to enter the type of load and the location separately, but the location
of the load is consistent for all element types of the same class. This is called the face ID and is shown in
the following figures. It should be noted that when input files are created using Marc Mentat, the
convention is to use the same ID as used in the Marc convention minus 1.
1-D 2-Node Elements

y FACE ID NODES
2
1 1–2
1
1-D 3-Node Elements

3 FACE ID NODES
1 1–2–3
2
1
CHAPTER 9 615
Boundary Conditions
2-D 4-Node Quadrilateral Elements

4 3 FACE ID NODES
1 1–2
2 2–3
3 3–4
4 4–1
1 2
Load shown on FACE ID 1
2-D 8-Node Quadrilateral Elements

4 7 3 FACE ID NODES
1 1–5–2
2 2–6–3
8 6 3 3–7–4
4 4–8–1
1 5 2
2-D 3-Node Triangle

3 FACE ID NODES
1 1–2
2 2–3
3 3–1
1 2
2-D 6-Node Triangle

3 FACE ID NODES
1 1–4–2
2 2–5–3
6 5 3 3–6–1
1 4 2
3-D 3-Node Shell

z 3 FACE ID NODES
1 1–2–3 (top)
2 1–3–2 (bottom)
1
y
x 2
3-D 4-Node Shell/Membrane

4 P FACE ID NODES
1 1–2–3–4 (top)
3 2 1–4–3–2 (bottom)
3-D 6-Node Shell

P 3 FACE ID NODES
1 1–2–3–4–5–6 (top)
2 1–3–2–6–5–4 (bottom)
6 5
1 4 2
3-D 8-Node Shell

4 7 3 FACE ID NODES
1 1–2–3–4–5–6–7–8 (top)
2 1–4–3–2–8–7–6–5 (bottom)
8 6
1 5 2
3-D 4-Node Tetrahedral

4 FACE ID NODES
3 1 1–2–4
2 2–3–4
3 3–1–4
4 1–2–3
1
2
CHAPTER 9 617
Boundary Conditions
3-D 6-Node Pentahedral

6 FACE ID NODES
4 1 1–2–5–4
2 2–3–6–5
3 3–1–4–6
5 4 1–2–3
5 4 – 6– 5
3
1
3-D 15-Node Pentahedral

3 FACE ID NODES
15
1 1 – 2 – 5 – 4 – 7 – 14 – 10 – 13
6 2 2 – 3 – 6 – 5 – 8 – 15 – 11 – 14
3 3 – 1 – 4 – 6 – 9 – 13 – 12 – 15
4 3–2–1–8–7–9
8
9 5 4 – 5– 6 – 10 – 11 – 12
12 11
1 7 2
13 14
4 10 5
3-D 8-Node Brick

8 FACE ID NODES
1 1–2–6–5
5 7
2 2–3–7–6
3 3–4–8–7
6
4 4–1–5–8
4 5 1–2–3–4
6 6–5–8–7
1 3
2
3-D 10-Node Tetrahedral

4 FACE ID NODES
10
1 1–2–4–5– 9– 8
3
8 9 2 2 – 3 – 4 – 6 – 10 – 9
3 3 – 1 – 4 – 7 – 8 – 10
7
4 1–2–3–5– 6– 7
6
1
5
2
3-D 20-Node Brick

8 FACE ID NODES
16 15
1 1 – 2 – 6 – 5 – 9 – 18 – 13 – 17
5 7 2 2 – 3 – 7 – 6 – 10 – 19 – 14 – 18
13 20
14 3 3 – 4 – 8 – 7 – 11 – 20 – 15 – 19
6 4 4 – 1 – 5 – 8 – 12 – 17 – 16 – 20
17 19
4 5 1 – 2 – 3 – 4 – 9 – 10 – 11 – 12
12 11 6 6 – 5 – 8 – 7 – 13 – 16 – 15 – 14
18
1 3
9 10
2
For the third method, the element edges or faces are attached to a curve or surface, respectively. In this
case, one only needs to give the type of load as the location is inferred from the ATTACH EDGE and
ATTACH FACE data.
Mechanical Loads
Marc allows you to enter mechanical loads in various forms for stress analysis. These loads can be
concentrated forces and moments, uniformly and nonuniformly distributed pressures, body forces,
gravity or centrifugal loads. Table 9-1 lists input options for mechanical loads.
CHAPTER 9 619
Boundary Conditions
Table 9-1 Input Options for Mechanical Loads

Input Options
Point Load: POINT LOAD POINT LOAD FORCDT
Concentrated
Force/Moment
Surface Load Pressure DIST LOADS DIST LOADS FORCEM
Shearing Forces, and
Distributed Moment
(Uniform/Nonuniform
Distribution)
Volumetric Load Body DIST LOADS DIST LOADS FORCEM
Forces and
Acceleration Forces
Centrifugal Loading DIST LOADS DIST LOADS
ROTATION A
Coriolis Loading DIST LOADS DIST LOADS
ROTATION A
Fluid Loading DIST LOADS
FLUID DRAG
Application of centrifugal and Coriolis loadings require the ROTATION A model definition option,
which defines the data corresponding to the axis of rotation. If multiple parts are rotating, then different
rotation axes may be defined. The actual load can be invoked by specifying an IBODY load types 100,
103,104, or 105 for centrifugal and Coriolis loadings, respectively. For load types 100 or 103, the square
2
of rotation speed, ω , is entered in radians per time, for the magnitude of the distributed load. For load
types 104 or 105, the angular velocity is entered as cycles per time. The mass density must also be
defined in the ISOTROPIC, ORTHOTROPIC, or other material options.
Application of gravity load (load per unit mass) is achieved by using IBODY load type 102. The mass
density must be defined in the ISOTROPIC, ORTHOTROPIC, or other material options. The acceleration
can be given independently in the x, y, and z direction through the DIST LOADS option.
Volumetric loads (load per unit volume) may be entered by using IBODY type 106 or 107.
The FOLLOW FOR parameter is used when the direction of Distributed and/or Point Loads are not fixed
and need to follow the geometry. FOLLOW FOR automatically invokes the LOAD COR parameter, so
stresses should be stored at all integration points. Separate flags under this parameter are used to control
follower forces for distributed loads and point loads respectively. When the FOLLOW FOR parameter is
flagged for distributed loads, equivalent nodal loads due to pressures are calculated based on the current
geometry. Any change of surface area or orientation results in a change of load. You can also specify that
the follower force stiffness matrix due to distributed loads can be included. The FOLLOW FOR parameter
for distributed loads is typically used when a shell structure, which can undergo large deformation and
rotations, is subjected to a pressure load.
A separate flag on the FOLLOW FOR parameter allows follower forces to be specified for point loads.
When this parameter option is flagged, the user still has the choice of specifying individual point loads
as fixed direction load vectors or as follower forces. Fixed direction load vectors are specified by the
usual vector components. When defined as a follower force, there are two techniques available for
specifying the load:
• The first technique is similar to the option available in Nastran. The scalar magnitudes of the load
and/or moment are specified instead of the vector components. The direction of the force is then
specified through 2 or 4 independent nodes in a manner similar to the FORCE1/MOMENT1 or
FORCE2/MOMENT2 options in Nastran. In the former, two nodes A and B are to be specified by
the user and the direction of the load follows the vector directed from node A to node B, as shown
in Figure 9-21(a). In the latter, four nodes A, B and C, D are specified by the user. The direction
of the load follows the cross product of the vector directed from node A to node B and the vector
directed from node C to node D, as shown in Figure 9-21(b).
P (Load Magnitude)
P (Load
A Magnitude)
Load
Direction Load D
Direction
B
A,C
B
(a) FORCE1 Style (b) FORCE2 Style
Figure 9-21 Nastran Style Follower Force Definition for Point Load
• The second technique offers an automated way of defining the follower force. The load/moment
is specified through vector components as usual. The initial orientation of the load with respect to
an optimal mesh location is noted. Then, as the structure deforms, the load direction is updated
such that the relative orientation of the load with respect to the mesh is maintained, as shown in
Figure 9-22. Optimal nodes in the mesh are automatically chosen for calculating and maintaining
the relative orientation. One or two optimal nodes can be used. The optimal nodes are chosen
based on the following criteria: The nodes should belong to the same element as the loaded node.
If the initial vector from the loaded node to any node matches the initial load direction, then that
node is chosen as the optimal node. Else, a node whose vector from the loaded node has a
component in the load direction is chosen. If this also does not provide a suitable node, the closest
node to the loaded node is chosen. For 3-D shells, membranes, and 3-D solids, two optimal nodes
are chosen if the cross product of the vectors of the loaded node to these two nodes matches the
load direction. Else, two nodes whose vectors from the loaded node have a cross product with a
component in the load direction are chosen. If none of the above rules provide suitable nodes, the
two closest nodes to the loaded node are chosen.
CHAPTER 9 621
Boundary Conditions
Initial Load
Direction
α
α Updated Load
Optimal Direction
Node
Figure 9-22 Geometry Based Automated Follower Force Definition for Point Load
The second technique is quite convenient and can be generally applied to 2-D and 3-D beams, shells and
solids. In situations where the loaded node does not belong to an element (for example, load controlled
rigid dies, etc.), the first technique should be used. It should be noted that a follower force stiffness matrix
is not currently available for the point loads. Also, the follower force capability is not supported for a
point load specified through the FORCDT user subroutine.
A number of history definition options are available for input of multiple load increments. For example,
the AUTO LOAD option generates a specified number of increments, all having the same load increment,
and is useful for nonlinear analysis with proportional loads; the PROPORTIONAL INCREMENT option
allows the previous load increment to be scaled up or down for use in the current load increment. The
AUTO INCREMENT option allows automatic load stepping in a quasi-static analysis and is useful for
both geometrically and materially nonlinear problems.
Fluid Drag and Wave Loads

Marc provides a fluid drag and wave load capabilities that can be applied on beam type structures that
are partially or fully submerged in fluid (see Figure 9-23). Morison’s equation is used to evaluate the
fluid drag loads that are associated with steady currents. Only distributed drag and buoyancy effects are
considered. Marc employs Airy wave theory to evaluate wave velocities that can be invoked for dynamic
analysis option.
Fluid drag and wave loads are invoked using the DIST LOADS model definition with an IBODY load type
of 11. These loads also require the FLUID DRAG model definition to input the relevant information
regarding the fluid elevation and its flow. Table 9-2 lists input options for fluid drag and wave loads.
Table 9-2 Input Options for Fluid Drag and Wave Loads
Input Options
Fluid Drag Load DIST LOADS DIST LOADS
FLUID DRAG
Wave Load DIST LOADS DIST LOADS
FLUID DRAG
Steady Velocity
Wave Load
Flow Velocity Pipe

Distribution
Buoyancy
Force
Velocity Outside
Gradient Fluid
Fluid Drag
Force
Inside
Fluid
Sea Bed
Figure 9-23 Fluid Drag and Wave Loads
Cavity Pressure Loading

Marc allows the modeling of structures enclosing cavities by updating the cavity internal pressure as the
cavity volume change. For ideal gas-filled cavities, the equation of state relating the cavity pressure and
volume can be written as:
pV = nR o T (9-5)
where p is the cavity total pressure, V is the cavity volume, n is the number of molecules of the gas,
R o is the Universal Gas Constant ( R o = 8.31447 J/(mol°K) = 1545.35 ft.lbf/(mol°R) ), and T is the
absolute gas temperature. The gas mass, M , is related to the number of molecules of the gas by:
μ = M⁄n (9-6)
where μ is the molar mass of the gas. Substituting by n from Equation (9-6) into Equation (9-5) gives:
pV = MRT (9-7)
where R = R 0 ⁄ μ is the Specific Gas Constant. With the gas density, ρ , defined as:
ρ = M⁄V (9-8)
CHAPTER 9 623
Boundary Conditions
the equation of state of an ideal gas can be finally written as:
p = ρRT (9-9)
The Specific Gas Constant is calculated from:
R = pr ⁄ ρr Tr (9-10)
where p r is the gas reference pressure, T r is the gas reference temperature, and ρ r is the gas reference
density. The user must ensure that the values entered for the gas reference pressure, temperature and
density are consistent. The cavity total pressure is given by:
p = pa + pg (9-11)
where p a is the ambient pressure and p g is the cavity gage pressure. Only the cavity gage pressure is
applied to the structure forming the cavity.
The following loading scenarios are available for cavities:
Closed Cavity
If the cavity is closed, the mass of the gas is preserved
M = Mo (9-12)
where M o is the cavity mass from the previous increment. The gas is assumed to undergo a general
polytropic process represented by:
γ
pV = constant (9-13)
where γ is the polytropic exponent. The gas pressure can thus be updated using
V γ
----- =  ------o-
p
(9-14)
po  V
where p o and V o are the cavity pressure and volume from the previous increment, respectively. Using
Equation (9-7) and Equation (9-14), the gas temperature can be updated using
V γ–1
------ =  ------o-
T
(9-15)
To  V
where T o is the cavity temperature from the previous increment. The gas density is calculated from:
ρ = M⁄V (9-16)
Closed cavity processes can be set to occur:
• at constant pressure, isobaric, using γ = 0 .
• at constant temperature, isothermal, using γ = 1 .

• with no heat transfer to the surroundings, adiabatic, using γ = k , where k is the adiabatic
exponent. For ideal gases, k is a constant that depends only on the number of atoms in the gas
molecule (monoatomic gases: k = 1.67 , diatomic gases: k = 1.4 , triatomic gases:
k = 1.33 ).
Applied Pressure
In this case the new cavity pressure is updated using
p = p o + Δp (9-17)
The gas temperature is assumed to be constant
T = To (9-18)
The gas density is updated using
ρ = p ⁄ RT (9-19)
and the gas mass is recalculated as
M = ρV (9-20)
Applied Mass
In this case, the new cavity mass is updated using
M = M o + ΔM (9-21)
If ΔM ≠ 0 , gas is pumped in or out of the cavity. The gas density is first updated as
ρ = M⁄V (9-22)
the gas temperature is assumed to be constant
T = To (9-23)
and the cavity pressure is calculated as
p = ρRT (9-24)
If ΔM = 0 , the cavity is assumed to be closed and Equations 9-12 to 9-16 apply with γ = 1 .
User Defined Loading

The UCAV user routine allows the user to define and control the cavity pressure for loading scenarios
other than the ones specified above.
CHAPTER 9 625
Boundary Conditions
Cavity Modeling
The CAVITY parameter is used to enter the number of cavities in the model (maximum 1000), an upper
bound to the number of segments in each cavity and an upper bound to the number of nodes per segment
of cavity boundary.
The CAVITY model definition option is used to enter the ambient pressure, the polytropic process
exponent, and the reference pressure, temperature and density for each cavity.
The applied pressure or mass is entered through the DIST LOADS model definition and history definition
options. The distributed load type, entered on the DIST LOADS option of the elements forming the cavity,
is modified for cavity loading according to the following relation
ibody_cavity = icavity * 10000 + icavity_type * 1000 + ibody (9-25)

where
ibody_cavity is the cavity-modified value for the distributed load type.
icavity is the cavity id.
icavity_type is the cavity load type:
0: cavity is closed.
1: cavity is loaded with an applied pressure.
2: cavity is loaded with an applied mass.
9: cavity load is defined by the UCAV user subroutine
ibody is the original value for the distributed load type (see library element description in Marc
Volume B: Element Library.)
In the first increment of the analysis, increment zero, the cavity temperature is calculated by averaging
the temperatures of the elements forming the cavity. If there are no nodal temperatures defined, the cavity
temperature is taken equal to the cavity reference temperature. For AXITO3D (or PRE STATE) analysis,
the cavity pressure, mass, and temperature for increment zero are read from the post file of the
corresponding axisymmetric problem. If the cavity is closed in increment zero, the cavity is assumed to
be initially at ambient conditions. In this case, the initial gage pressure is equal to zero and the initial
mass is based on the ambient pressure and initial volume. Moreover, if the ambient pressure is equal to
zero, the cavity is assumed to be initially unloaded. Tables 9-3 and 9-4 summarize the cavity functioning
for the different cavity load types and incremental load values during the initial and subsequent load
increments, respectively.
Table 9-3 Cavity Functioning for Increment Zero
Cavity Load Type Incremental Load Value Function
0 Ignored Cavity is initially at ambient conditions
1 Δp = 0.0 Cavity is initially empty
1 Δp ≠ 0.0 Cavity is initially at the applied pressure
2 ΔM = 0.0 Cavity is initially empty
2 ΔM ≠ 0.0 Cavity initially contains the applied mass
9 Passed to UCAV Call UCAV
Table 9-4 Cavity Functioning for Subsequent Increments

Cavity Load Type Incremental Load Value Function
0 Ignored Cavity is closed, constant mass,
polytropic process
1 Δp = 0.0 Cavity is open, constant pressure and
temperature
1 Δp ≠ 0.0 Cavity is open, pressure is applied, constant
temperature
2 ΔM = 0.0 Cavity is closed, constant mass, isothermal
process
2 ΔM ≠ 0.0 Cavity is open, mass is applied, constant
temperature
9 Passed to UCAV Call UCAV
In general, standard structural elements are used to define the boundaries of cavities and no extra
elements are required. However, to model the boundaries of cavities in regions where standard finite
elements are not present (for example, along rigid boundaries) cavity surface elements (elements 171-
174) can be used. These elements can also be glued to moving rigid surfaces. They are for volume
calculation purposes only and do not contribute to the stiffness equations of the model.
For 2-D problems, the volume of the cavity is calculated as the sum of the areas of all the triangles formed
by the cavity segments as bases and the coordinate system origin as the apex multiplied by the cavity
thickness. Elements forming the cavity are assumed to be of equal thickness. For 3-D problems, the
volume of the cavity is calculated as the sum of the volumes of all the tetrahedrons formed by the cavity
patches as bases and the coordinate system origin as the apex. For axisymmetric problems, the volume
of the cavity is calculated as the sum of the volumes of all the cone frustums formed by the cavity
segment as the cone slant height and two parallel base circles and with an axis along the axis of symmetry
of the problem. In the axisymmetric case, it is not necessary to use cavity surface elements along lines
that are perpendicular to the axis of symmetry to close cavities.
In coupled thermo-mechanical analysis, heat transfer between the gas inside the cavity and the
surrounding structure is not supported.
Cavities are not allowed to split or to join due to deformation or self-contact. If self-contact occurs within
a cavity, the resulting cavities are still treated as a single cavity.
If the cavity is formed out of membrane and/or shell elements, the user must ensure that all elements
defining the cavity are aligned; that is, the cavity surface is defined by either all top or all bottom
element faces.
CHAPTER 9 627
Boundary Conditions
Cyclic Loading
Application of cyclic loading is important when investigating fatigue issues in a structure. There are two
ways of easily doing this in Marc. In the first method, one associates a table that is time dependent with
each boundary condition. This table can be either a sawtooth function or a sinusoidal function. The
boundary condition will be repeated for as long as one has a time period assigned. note that the AUTO
STEP option can either exactly reach the peaks or not be constrained to reach the peaks.
The alternative is to use the BEGIN SEQUENCE and END SEQUENCE options. All history definition
options placed between these two options will be repeated as often as requested. One should make sure
that the value of the boundary conditions at the beginning of the sequence is the same as the value at the
end of the sequence. Multiple loadcases may be repeated.
Thermal Loads
Element integration point temperatures are used in a thermal stress analysis to generate thermal load. The
AUTO THERM history definition option allows automatic application of temperature increments based
on a set of temperatures defined throughout the mesh as a function of time. The CHANGE STATE option
presents the temperatures to Marc which then creates its own set of temperature steps based on a
temperature change tolerance provided through this option. Table 9-5 lists input options for thermal
loads. You can input either the incremental temperature or the total temperatures as thermal loads.
Table 9-5 Input Options for Thermal Loads
Input Options
User
Load Description Parameter Model Definition History Definition Subroutine
Incremental THERMAL THERMAL LOADS** THERMAL LOADS** CREDE
Temperatures*
Total Temperatures* THERMAL CHANGE STATE CHANGE STATE NEWSV
Initial Temperature* INITIAL STATE INITSV
Total Nodal POINT TEMP POINT TEMP
Temperatures
Initial Nodal INITIAL TEMP USINC
Temperature
Total Nodal Temperature MAP TEMP
Alternative Mesh
* Temperatures must be specified at each integration point (or at the centroid if the CENTROID
parameter is used) of each element in the mesh.
** The THERMAL LOADS option and CREDE user subroutine should not be used with the table
driven input procedure.
9 The thermal strain increment is defined as

th
Δε =  αΔT (9-26)
th
where Δε is thermal strain increment, α is the coefficient of thermal expansion, and ΔT is the
temperature increment. Equivalent nodal forces are calculated from the thermal strain increment and then
added to the nodal force vector for the solution of the problem.
You can input the coefficient of thermal expansion through the ISOTROPIC, ORTHOTROPIC, or other
material options and the temperature increment through various options.
Initial Stress and Initial Plastic Strain

Marc allows you to enter a set of initial stresses that simulate the stress state in the structure at the
beginning of an analysis. A typical example is prestress in a tensioned fabric roof. The set of initial
stresses must be self-equilibrating and should not exceed the yield stress of the material. Table 9-6 shows
the input options for initial stress.
Table 9-6 Input Options for Initial Stress and Initial Plastic Strain
Input Options
Load Description Parameter Model Definition User Subroutine
Initial Stress Prestress ISTRESS INIT STRESS UINSTR
Initial Plastic Strain INITIAL PLASTIC STRAIN INITPL
Marc also provides various ways of initializing the equivalent plastic strain throughout the model. This
is useful in metal forming analysis in which the previous amount of equivalent plastic strain is often
required. This history dependent variable represents the amount of plastic deformation that the model
was subjected to, and is used in the work (strain) hardening model. The input option is shown in
Table 9-5.
Boundar Heat Fluxes

y In a heat transfer analysis, you can enter heat fluxes in various forms. Heat transfer analysis requires
Conditio entering the total values of flux. Table 9-7 lists the input options for heat fluxes.
ns
CHAPTER 9 629
Boundary Conditions
Table 9-7 Input Options for Heat Fluxes

Input Options
Point Heat Flux POINT FLUX POINT FLUX FORCDT
(Sink or Source)
Surface Heat Flux, Convection, DIST FLUXES DIST FLUXES FLUX*
Radiation QVECT QVECT UQVECT
FILMS FILMS FILM*
SURFACE ENERGY
Volumetric Flux Load DIST FLUXES DIST FLUXES FLUX*
Body Flux FILMS
* Can be used for complicated flux loadings, convection, and radiation, allowing the input of
nonuniform temperature- and time-dependent boundary conditions.
There are three special heat flux conditions representing insulation, convection, and radiation as
discussed in Chapter 6: Nonstructural and Coupled Procedure Library in the Boundary
Conditions section.
1. Insulation
q = 0 (9-27)
No input is required for the insulated case.
2. Convection
Nondirected:
q = H ( Ts – T ) (9-28)
You must enter the film coefficient H and ambient temperature T through the FILMS model
definition option or the FILM user subroutine. You can also directly input the heat flux q using
the FLUX user subroutine.
Directed
q = –q0 ⋅ α ⋅ n ⋅ ns (9-29)
˜
where n is the orientation defined in the QVECT option and n s is the outward unit normal.
˜
3. Radiation
4 4
q = σ ⋅ ε ( Ts – T∞ ) (9-30)
You must enter either the heat flux q using the FLUX user subroutine or the temperature
dependent film coefficient H ( T s, T, σ, ε ) and ambient temperature T using the FILM user
subroutine. These relationships are shown in Equation (9-31). Using the table driven input format,
you can directly specify radiation to the environment through the FILMS model definition option.
The use of the FILM user subroutine is recommended to ensure a stable solution.
4 4
q = σε ( T s – T ∞ )
3 2 2 3 (9-31)
= σ ⋅ ε ( Ts + Ts T∞ + Ts T∞ + T∞ ) ( Ts – T∞ )
= H ( T s, T ∞ ) ( T s – T ∞ )
where σ is the Stefan-Boltzmann constant, ε is emissivity, T s and T ∞ are unknown surface and
ambient temperatures, respectively.
As an alternative for radiation between surfaces, Marc will calculate the viewfactors using the
CAVITY DEFINITION and RAD-CAVITY model definition options or the VIEW FACTOR model
definition option can be used to read in viewfactors calculated by Marc Mentat. The RADIATION
parameter is also required.
Mass Fluxes and Restrictors

In a hydrodynamic bearing analysis, you can enter mass fluxes, restrictors, and pump pressures as loads.
Table 9-8 lists input options for these quantities.
Table 9-8 Input Options for Mass Fluxes and Restrictors
Input Options
Model History User
Load Description Parameter Definition Definition Subroutine
Nodal Mass Fluxes BEARING POINT MASS POINT MASS FORCDT
Distributed Mass Fluxes BEARING DIST MASS DIST MASS FLUX*
Restrictors BEARING RESTRICTOR URESTR**
RESTRICTOR
* Can be used for nonuniform mass fluxes
** Can be used for nonuniform restrictions or pump pressures.
Electrical Currents
In coupled thermo-electrical (Joule heating) analysis and coupled electrical-thermal-mechanical analysis
(Joule-mechanical), you can prescribe electrical currents as loads for the calculation of unknown nodal
voltages. In magnetostatic analysis, you can also define the current. In such cases, as a steady state
analysis is performed, there is no time variation of the currents. Table 9-9 lists input options for
electrical currents.
CHAPTER 9 631
Boundary Conditions
Table 9-9 Input Options for Electrical Currents

Input Options
Surface and Body Currents DIST CURRENT DIST CURRENT FLUX
Electrostatic Charges
In an electrostatic analysis, the charge can be entered, noting that a steady state analysis is performed
so there is no time variation of charge. In a coupled electrostatic structural analysis there can be a
variation of charge in time, but this variation is considered to be quasi-static. Table 9-10 summarized the
input options.
Table 9-10 Input Options for Electrostatic Charge
Input Options
Nodal Charge POINT CHARGE POINT CHARGE FORCDT
Distributed and Body Charges DIST CHARGES DIST CHARGE FLUX
Acoustic Sources
In acoustic analysis, you can enter a source pressure if a transient analysis by modal superposition is
being performed. Table 9-11 summarized the input options.
Table 9-11 Input Options for Acoustic Sources
Input Options
Nodal Source POINT SOURCE POINT SOURCE FORCDT
Distributed Source DIST SOURCES DIST SOURCES FLUX
Nodal Source FIXED PRESSURE PRESS CHANGE FORCDT
Piezoelectric Loads
In a piezoelectric analysis, both mechanical loads and electrostatic charges can be entered. These values
can have time variation if a transient analysis is performed or a harmonic excitation can be applied.
Table 9-1 gives a summery of the mechanical input options, and Table 9-10 gives a summary of the
electrostatic input options.
Electrostatic-Structural Loads
In a coupled electrostatic-structural analysis, both mechanical loads and electrostatic charges can be
entered. Table 9-1 gives a summery of the mechanical input options, and Table 9-10 gives a summary of
the electrostatic input options.
Magnetostatic Currents
In a magnetostatic analysis, the current can be entered, noting that a steady state analysis is performed so
there is no time variation of current. Table 9-12 summarizes the input options.
Table 9-12 Input Options for Magnetostatic Current
Input Options
History
Load Description Model Definition Definition User Subroutine
Distributed and Body Current 2-D DIST CURRENT DIST CURRENT FLUX
Distributed and Body Current 3-D DIST CURRENT DIST CURRENT FORCEM
Electromagnetic Currents and Charges

In an electromagnetic analysis, the current can be entered. These values can have time variation if a
transient analysis is performed or a harmonic excitation can be applied. Table 9-13 summarizes the
input options.
Table 9-13 Input Options for Currents and Charges
Input Options
Distributed and Body Current DIST CURRENT DIST CURRENT FLUX
Nodal Charge POINT CURRENT POINT CURRENT FORCDT
POINT CHARGE
Distributed and Body Charge DIST CHARGE DIST CHARGE
Kinematic Constraints
Marc allows you to input kinematic constraints through various options that include
• Boundary Conditions (prescribed nodal values)
• Transformation of Degree of Freedom
• Shell Transformation
• Tying Constraint
• Rigid Link Constraint
• Shell-to-Solid Tying
CHAPTER 9 633
Boundary Conditions
• Insert
• AUTOMSET
• Support Conditions
• Bushings
• Cyclic Symmetry
• Nastran RBE2 and RBE3
• Beam - Shell Offsets
• Pin Code for Beam Elements
Boundary Conditions
Marc allows you to specify the nodal value for a particular degree of freedom. If you do not give a nodal
value when you specify the boundary condition, Marc sets the fixed nodal value to zero. An option
allows boundary conditions to be specified at the time of two-dimensional mesh generation with
MESH2D. You can apply a different set of boundary conditions for each load increment.
Table 9-14 gives the nodal values for the various analyses, and Table 9-15 lists the input options for
boundary conditions in different analyses.
Table 9-14 Analyses with Corresponding Nodal Values
Analysis Nodal Values
Acoustics Pressure
Coupled Fluid Thermal Velocity, Pressure, and Temperature
Coupled Thermo-electrical Voltage and Temperature
Coupled Thermo-mechanical Displacement and Temperature
Coupled Electrical-thermo-mechanical Voltage, Temperature, and Displacement
Electrostatic Potential
Piezoelectric Displacement and Potential
Fluid Velocity and Pressure
Heat transfer Temperature
Hydrodynamic Bearing Pressure
Magnetostatic Potential
Rigid Plastic Flow Velocity
Stress Displacements
Table 9-15 Input Options for Boundary Conditions

Input Options
User
Load Description Model Definition History Definition Subroutine
Displacement FIXED DISP DISP CHANGE FORCDT
Temperature FIXED TEMPERATURE TEMP CHANGE FORCDT
Voltage FIXED VOLTAGE VOLTAGE CHANGE FORCDT
Pressure FIXED PRESSURE PRESS CHANGE FORCDT
Potential (Electrostatic) FIXED EL-POT FORCDT
FIXED POTENTIAL
Potential (Magnetostatics) FIXED MG-POT
FIXED POTENTIAL
Velocity FIXED VELOCITY VELOCITY CHANGE FORCDT
Transformation of Degree of Freedom

Marc allows transformation of individual nodal degrees of freedom from the global direction to a local
direction through an orthogonal transformation that facilitates the application of boundary conditions and
the tying together of shell and solid elements.
Transformations are assumed to be orthogonal. Once you invoke a transformation on a node, you must
input all loads and kinematic conditions for the node in the transformed system.Nodal output is in the
transformed system. This option is invoked using the TRANSFORMATION option.
The UTRANFORM option allows transformations to be entered via the UTRANS user subroutine. This
allows you to transform the degrees of freedom at an individual node from global directions to a local
direction through an orthogonal transformation. UTRANFORM allows you to change the transformation
with each increment. When you invoke this option, the nodal output is in both the local and the
global system.
Shell Transformation
The SHELL TRANSFORMATION option allows you to transform the global degree of freedom of doubly
curved shells to shell degrees of freedom in order to facilitate application of forces in the shell directions,
edge moments, and clamped or simply supported boundary conditions. There are four types of shell
transformations.
The SHELL TRANSFORMATION model definition option specifies information on the shell
transformation. For Types 1 and 3, only the node number has to be specified. For Types 2 and 4, a
boundary direction (the direction cosine of t as shown in Figure 9-24) also has to be specified in the
θ 1, θ 2 surface.
CHAPTER 9 635
Boundary Conditions
θ2
θ1
Figure 9-24 Boundary Directions in Shell Transformation
After transformation, the following definitions apply.

Type 1: Transformation for two-dimensional beams and shell nodes (Element Types 15, 16, and
17). The transformation defines the degrees of freedom with respect to a local coordinate
system (s, n) (see Figure 9-32). The four degrees of freedom after transformation are:
1 = us tangential displacement
2 = u n normal displacement
3= φ rotation
4= ε Meridional stretch
x2
n s
Positive
Direction
x1
Figure 9-25 Type 1: Shell Transformation
Type 2: Transformations for doubly curved shell nodes with nine degrees of freedom (Element
Type 4 corner nodes of Element Type 24). The transformation defines a local coordinate
system ( t, s, n ). (See Figure 9-26).
The nine degrees of freedom after transformation are:
1 = ut displacement in specified (boundary) direction
2 = us displacement normal to (boundary) direction but tangential to shell surface
3 = un displacement normal to shell surface

4 = φt rotation of shell around boundary
5 = φs rotation of shell around normal to boundary tangential to the shell surface
6 = φn rotation of boundary around normal to shell surface
7 = εt stretch tangential to specified (boundary) direction
8 = εs stretch normal to specified (boundary) direction
9 = γt s shear stretch in t-s direction
n θ2
t
s
θ1
x3
x2
x1
Figure 9-26 Types 2 and 4: Shell Transformations
Type 3: Transformations at midside nodes for doubly curved shell nodes with three degrees of
freedom (Element Type 24). The transformation defines a local coordinate system
( t, s, n ), (see Figure 9-27). The degrees of freedom after transformation are:
1 = ∂u t ⁄ ∂s rotation of normal to boundary around normal to shell
2 = ∂u s ⁄ ∂s stretch normal to boundary
3 = ∂u n ⁄ ∂s rotation around boundary
n t
x3 s
x2
x1
Figure 9-27 Type 3: Shell Transformation

CHAPTER 9 637
Boundary Conditions
Type 4: Transformation for doubly curved nodes with 12 degrees of freedom (Element Type 4).
The transformation defines a local coordinate system ( t, s, n ). The first 9 degrees of
freedom after transformation are the same as 1 through 9 in Type 2 (Figure 9-26), and
the remaining three are:
2
10 = ∂ u t ⁄ ∂Θ 1 ∂Θ 2 variation of g along the boundary (9-32)
2
11 = ∂ u s ⁄ ∂Θ 1 ∂Θ 2 variation of e along the boundary (9-33)
2
12 = ∂ u n ⁄ ∂Θ 1 ∂Θ 2 variation of f along the boundary (9-34)
When using the SHELL TRANSFORMATION option, the displacement increments and reaction
forces are output in the local directions immediately after solution of the equations. At the end of the
increment, you can print out the global displacement increments and total displacements in the global
coordinate directions.
If you invoke the FOLLOW FOR parameter, the SHELL TRANSFORMATION option defines a local
coordinate system in the current (updated) geometry of the structure. This additional feature is especially
useful with the LARGE STRAIN parameter, because you can then specify edge moments and/or large
edge rotations of shells and beams.
Caution: If you apply a shell transformation to a node, do not apply a standard transformation or shell
tying type to that node.
Tying Constraint
Marc contains a generalized tying (constraint) condition option. Any constraint involving linear
dependence of nodal degrees of freedom can be included in the stiffness equations.
A tying constraint involves one tied node and one or more retained nodes, and a tying (constraint)
condition between the tied and retained nodes. The degrees of freedom (for example, displacements,
temperatures) of the tied node are dependent on the degrees of freedom of the retained nodes through the
tying condition. In some special tying conditions, the tied node can also be a retained node. The tying
condition can be represented by a tying (constraint) matrix. Note that if the tying constraint involves only
one retained node, the choice of which node is to be tied to retained is arbitrary.
As a simple example, impose the constraint that the first degree of freedom of node I be equal to that of
node J at all times (see Figure 9-28).
As a second example, the simulation of a sliding boundary condition requires the input of both the
boundary conditions and the tying constraints (see Figure 9-29).
I UI
Constraint Equation:
UI = UJ
2,V
J UJ
1,U Tied node I, retained node J, or

Tied node J, retained node I
Figure 9-28 Simple Tying Constraint
Local axes
X
Y (v) i Y
j
k
l
X (u)
Figure 9-29 Tying Constraint Illustration (Sliding Boundary Conditions)
The example illustrated in Figure 9-29 enforces rigid sliding on the boundary in the local coordinates
defined above.
vi = vj = vk = v1 = 0 (9-35)
ui = uj = uk = u1 (9-36)
The first equation is a set of fixed boundary conditions. The second equation is a constraint equation and
can be rewritten as three constraint equations:
ui = uj (9-37)
uk = uj (9-38)
u1 = uj (9-39)
These equations express all the u displacements in terms of u j . In this example, node j is chosen to be
the retained node; nodes i, k, and l are tied nodes. You can use the TYING option to enter this information.
Marc has a number of standard tying constraints that can be used for mesh refinement, shell-to-shell,
shell-to-beam, beam-to-beam, and shell-to-solid intersections. Tables 9-16 through 9-19 describe these
options. Tables Table 9-21 through Table 9-23 show the tying constraints for pipe elements, shell
stiffeners, and nodal degrees of freedom. Table 9-24 summarizes the rigid link constraint.
CHAPTER 9 639
Boundary Conditions
The SERVO LINK option uses the homogeneous linear constraint capability (tying) to input simple
constraints of the form
ut = a1 ur 1 + a2 ur 2 + … (9-40)
where u t is a degree of freedom to be constrained; u r 1, u r 2, … are the other retained degrees of freedom
in this structure; and a 1, a 2, … are constants provided in this option.
You can use the TYING or SERVO LINK model definition option to enter standard tying constraint
information, and the TYING CHANGE option to change tying constraints during load incrementation.
The UFORMSN user subroutine is a powerful method to specify a user-defined constraint equation. This
constraint can be nonlinear; for example, it can be dependent on time or previous deformation.
Table 9-16 Tying Constraints for Mesh Refinement
Number of
Tying Retained
Code Nodes Purpose Remarks
31 2 Refine mesh of first order Tie interior nodes on refined side
(linear displacement) elements in to corner nodes in coarse side (see
two dimensions Figure 9-30)
32 3 Refine mesh of second order Tie interior nodes on refined side
(quadratic displacement in to edge of element of coarse side
two dimensions (see Figure 9-31)
33 4 Refine mesh of 8-node bricks Tie interior node on refined side to
four corner nodes of an element
face on coarse side (see
Figure 9-32)
34 8 Refine mesh of 20-node bricks Tie interior nodes on refined side
to eight (four corner, four midside)
nodes of element face on coarse
side (see Figure 9-33)
T – Tied Node
T R – Retained Node
R
Figure 9-30 Mesh Refinement for 4-Node Quad
T
R T – Tied Node
R – Retained Node
T
R
Figure 9-31 Mesh Refinement for 8-Node Quad
T
R
R
Figure 9-32 Mesh Refinement for 8-Node Brick
R
T
R
T
R
T
R
R
Figure 9-33 Mesh Refinement for 20-Node Brick
CHAPTER 9 641
Boundary Conditions
Table 9-17 Tying Constraints for Shell-to-Shell Intersection

Tying Number of
Code Retained Nodes Purpose Remarks
22 2 Join intersecting shells, Element Tied node is also the second
Type 4, 8, or 24; fully moment- retained node*
carrying joint
Type 4, 8, or 24; fully moment- retained node
carrying joint
Type 4, 8, or 24; pinned point retained node
24** 2 Join intersecting shells or beams, Tied node is also the second
Element Types 15-17 retained node
* Thickness vector must be specified at tied nodes and retained nodes.
** See Figure 9-34.
A
B
S
S
1
Figure 9-34 Standard Tying Type 24, Tie Shell-to-Shell or Beam-to-Beam; Moment-Carrying
Table 9-18 Tying Constraints for Beam-to-Beam Intersection

Number of
Tying Retained
13 2 Join Two Elements Type 13 under Tied node also the second
an arbitrary angle; full moment- retained node
carrying joint
52 1 Pin joint for Beam Type 14, 25,
52, 76-79, 98
53 1 Full moment-carrying joint for Tie interior node on refined side to
Beam Types 14, 25, 52, 76-79, 98 four corner nodes of an element
face on coarse side
(see Figure 9-32)
Table 9-19 Tying Constraints for Shell-to-Solid Intersections

Number of
Tying Retained
23 1 Tie axisymmetric solid node to Tied and retained nodes must be
axisymmetric shell transformed to local system and
(Element Type 1 or 89) node TRANSFORMATION option
invoked (see Figure 9-35).
25 2 Join solid mesh to beam (Type 16) Tied node also second retained
Join solid mesh to axisymmetric node. (see Figure 9-36)
shell (Type 15)
26 2 Tie axisymmetric solid node to Similar to tying type 23 but no
axisymmetric shell transformation is required. Tied
(Element Type 1 or 89) node node also second retained node.
(see Figure 9-36)
T1
y’
R
T2 x’
Figure 9-35 Standard Tying Type 23, Tie Solid-to-Shell (Element Type 1 or 89)
T1
T2
R
ε T3
B T4
A α
Figure 9-36 Standard Tying Type 25, Tie Solid-to-Beam (Element Type 16)
Standard Tying Type 25, Tie Solid-to-Axisymmetric Shell (Element Type 15)
Standard Tying Type 26, Tie Solid-to-Axisymmetric Shell (Element Type 1 or 89)
CHAPTER 9 643
Boundary Conditions
Table 9-20 Tying Constraints for Beams to Element Edge or Face or Axisymmetric Shell to
Element Edge
Number of
Tying Retained
44 2 is 2-D or Rigidly tie a node with The number of retained nodes
axisymmetric displacements and rotations to is required. This tying type
lower-order a surface patch. This is fully supports large
element internally used for CWELD deformation/rotations. No
edge and CFAST option. transformations are required.
3 is 2-D or See Figure 9-37.
axisymmetric
higher-order
element
edge
3 if 3-D
lower-order
triangular
face
4 if 3-D
lower-order
quadrilateral
face
6 if 3-D
higher-order
triangular
face
8 if 3-D
higher-order
quadrilateral
face
R
T T
R
R
2-D Application Lower-order 2-D Application Higher-order
R R
R R R R
T
T R
R
R
3-D Application Lower-order Quadrilateral Face 3-D Application Higher-order Triangular Face
Figure 9-37 Beam to Edge Moment Carrying Constraints

CHAPTER 9 645
Boundary Conditions
Table 9-21 Tying Constraints for Pipe Bend Element (Elements 14 and 17)
Tying Code Number of Retained Nodes Purpose
15 One less than the number of shell Special tying types for pipe bend Element 17 to
nodes in the z-r plane of the section remove rigid body modes
16 Number of shell nodes in the z-r Special tying types for pipe bend Element 17 to
plane of the section remove rigid body modes
17 2 Special tying types for pipe bend Element 17 to
couple bend section into pipe bend
Table 9-22 Tying Constraints for Shell Stiffener (Element 13 as a Stiffener on Shell Elements 4 or 8)
Number of
Tying Retained
19 2 Use beam Element 13 as a stiffener on Tied node also second
shell Elements 4 or 8. Tied node is retained node
beam node; first retained node is shell
node, second is beam node again.
Beam node should be on or close to the
normal to the shell at the shell node
20 3 Create an extra node in a shell Type 8 Always used after tying
Element tied to the interpolated shell Type 21
displacements with tying Type 21 to tie
a beam Element 13 or a stiffener across
a shell element
21 2 Same as Type 19, but tying beam to an Must be followed by tying
interpolated shell node not a vertex of Type 20
an element (Element Type 8 only) must
be followed by Type 20 to tie
interpolated shell node into shell mesh
Table 9-23 Tying Constraints for Nodal Degrees of Freedom

Number of
Tying Retained
I≤NDEG 1 Tie the Ith degree of freedom at NDEG = number of degrees of
the tied node to the Ith degree of freedom per node
freedom at the retained node
100 1 Tie all degrees of freedom at the
tied node to the corresponding
degrees of freedom at the
retained node
>100 1 Generate several tyings of type Tying code is first degree of
≤NDEG freedom multiplied by 100 added
to last degree of freedom (209
means tie second through ninth
degree of freedom at tied node to
respective second through ninth
degree of freedom at retained
node)
<0 User-defined User-generated tying type through
(negative the UFORMSN user subroutine
integer)
Rigid Link Constraint

Tying type 80 can be used to define a rigid link between nodes. This capability can be used for both small
deformation and large deformation, large rotation problems. In small deformations, a linear constraint
equation is used. In addition to the end points of the link, a second retained node must be given. This node
is used to calculate and store the rigid body rotations. For two-dimensional problems, this represents a
rotation about the Z-axis and this is stored as the first degree of freedom for the node. In 3-D, it should
be noted that the rotations/moments about the x, y, and z axes for the second retained node are stored and
output as its first, second, and third degrees of freedom, respectively.
A complete rigid region can be modeled by using multiple tying type 80’s. In this case, the same two
nodes are used for the retained nodes in all of the constraints. Transformations can be applied to the
retained nodes so that any kinematic or force boundary conditions can be defined in a local coordinate
system. It is not necessary that the same transformations should be applied to both retained nodes. It is
also optional to transform the tied nodes to the same or a different local coordinate system.
The rigid links can be used with all elements except types 4, 8, 24, 15, 16, and 17. In addition, it should
not be used in rigid plastic analysis.
CHAPTER 9 647
Boundary Conditions
To reduce linearization errors in large deformation, large rotation problems using tying type 80, an
optional recycling feature is offered through the CONTROL model and history definition options.
Recycling occurs if the maximum change over the retained nodes of rigid link 80 exceed a rotation
tolerance (default is 0.001 radians). This check can be circumvented by setting the rigid link rotation
tolerance to zero (in conjunction with the FEATURE,5701 parameter) or a negative number.
Table 9-24 Rigid Link Constraint

Number of
Tying Retained
80 2 Define a rigid link between nodes The second retained node is an
unattached node which contains
the rotation
Shell-to-Solid Tying
In many problems, a region exists that is modeled with both brick elements and shell elements. A
particular case of this is shown is Figure 9-38 and Figure 9-39.
Figure 9-38 4-Node Shell-to-Solid Automatic Constraint
Figure 9-39 8-Node Shell-to-Solid Automatic Constraint
In the first case, an 8-node brick which has been reduced to a triangular prism is connected to a 4-node
shell. In the second case, a 20-node brick is connected to an 8-node shell. An automatic constraint
equation is developed between the elements. Note that the thickness of the shell must be entered as the
brick thickness.
Rigid Tying to a Surface Patch

Tying type 44 rigidly ties a node with its translation and rotation degrees of freedom to a surface patch.
There are four different tying types for 3-D analyses tying the displacements and rotations of a node
(assuming it has 6 degrees of freedom) to the motion of a 3- or 6-node triangular patch or a 4- or 8-node
quadrilateral patch. There are two different tying types for 2-D analyses, tying the displacements and
rotation of a node (assuming it has three degrees of freedom) to the motion of a 2- or 3-node line patch.
The tied node can be a node of a beam or a shell element and will generally have translation and rotation
degrees of freedom. The retained nodes are the nodes of the patch and the only retained degrees of
freedom involved in the tying are the patch node translations. The exact tying type is determined from
the tying type number (44) and the number of retained nodes and, therefore, the latter is required in the
input. The location where the tied node pierces the patch is computed in terms of the local parametric
coordinates that define locations inside a patch using the usual shape functions for such a patch. The
translations of this piercing point are computed from the nodal displacements of the patch, again using
the usual shape functions of the patch and these define the translations of the tied node. The rotations can
be computed by constructing a local triad to the patch at the piercing point. The translations of the patch
nodes define the change in orientation of the triad. The orientation change of the triad defines the
rotations of the tied node. The tied node must have a proper projection on the patch inside its boundary.
When it does not lie exactly on the patch, the distance will be treated as an offset. The patch does not
necessarily have to coincide with a shell or brick element face in a 3-D model or a beam or quad element
edge in a 2-D model, but the nodes constituting a patch must adhere to the conventions that are generally
required for the particular patch. The patch also does not have to be a flat surface in 3-D or a straight line
in 2-D, but it is recommended to avoid overly distorted patches. A patch is considered overly distorted
when the patch normals at two of its nodes make an angle larger than 10 degrees.
In case of a heat transfer or a thermo-mechanically coupled analysis the temperatures may be tied as well.
The temperature constraints are similar to the mechanical displacement constraints, but only one degree
of freedom is involved. When the retained nodes are shell nodes, the temperature is always tied to the
first degree of freedom of the shell nodes (i.e., the top surface of the shell), regardless of the orientation
in space. When the tied node is a shell node, its first degree of freedom is tied. In other analysis
procedures, tying type 44 is not supported.
When no geometric nonlinearities are flagged in the model (i.e., neither the LARGE DISP nor the
LARGE STRAIN parameter is activated), Marc uses the small displacement method to apply the
constraints. In all other cases, the large displacement method is used fully accounting for finite rotations.
The small displacement method linearizes the constraint equations w.r.t. the reference configuration. The
large displacement method takes account of all geometric nonlinearities caused by finite displacements
and rotations.
CHAPTER 9 649
Boundary Conditions
e3 e2
η e1
ξ
GA
Patch at end
of increment
Patch at start
of increment
Figure 9-40 Motion of a Patch From Start to End of an Increment
The large displacement method (Figure 9-40) can be summarized as follows. The point GA is the node
of, for example, a beam and it pierces the patch at some point for which the parametric coordinates ξ G A ,
η G A have been computed. A local triad is fixed to the beam cross section and another triad to the
piercing point on the patch. For simplicity, it is assumed that both triads have the same orientation in the
initial configuration. The tying constraints are that the displacements of the beam node and the
displacements of the piercing point on the patch are equal and that the rotations of the triad connected to
the beam cross-section and the triad connected to the patch are equal. The rotations of the triad fixed to
the beam cross-section are expressed in terms of the rotational degrees of freedom of the beam node. The
rotations of the triad fixed to the patch are expressed in terms of the displacement degrees of freedom of
the patch nodes. In this way, the nonlinear relations are established between the beam rotations and the
patch displacements. The relations between the displacements of the beam node and the displacements
of the patch nodes are linear when the beam node lies exactly on the patch. In case of an offset, nonlinear
contributions due to the rotations are added to the constraints.
The tied node generally has translation and rotation degrees of freedom, but this is not a requirement. In
case it has only translation and rotation degrees of freedom, the displacements are tied to the patch,
possibly accounting for an offset. No constraints are applied for rotations; hence, adequate constraints
must be applied to prevent a singular system.
The connectivity conventions for the patches are shown in Figure 9-41 for 2-D and 3-D models:
Line2
1 2
2-D Model
Line3
1 2 3
(a) Line Patch in 2-D Analysis
4 3
Quad4
1 2
3-D Model
4 7 3
Quad8 8 6
1 5 2
(b) Quadrilateral Patch in 3-D Analysis
3
Tria3
1 2
3-D Model
3
Tria6 6 5
1 4 2
(c) Triangular Patch in 3-D Analysis
Figure 9-41 Patch Connectivity Convention
Overclosure Tying
In various engineering applications, it is necessary to define a pre-stress in, for example, bolts or rivets
before applying any other structural loading. Although such a pre-stressed state is often simulated using
a temperature loading, it is rather difficult to arrive at a desired net force in the bolt or rivet. An easier
way is to use overclosure tyings (type 69). These tyings create overlaps or gaps between two parts of a
model. If the motion of these parts is somehow constrained in the direction in which the gap or overlap
is being created, then an overlap will introduce an tensile (pre-) stress in each of the parts and a gap will
result in a compressive stress.
Overclosure tyings have one tied node and two retained nodes. The tied node and the first retained node
are usually nodes on the boundaries of the respective parts (see Figure 9-42). The second retained node
is very often a free node and is usually shared by all overclosure tyings which connect the parts. This
node is also called the control node of the tying, since it can be used to apply load or control the size of
the gap or overlap between the parts.
CHAPTER 9 651
Boundary Conditions
top part top part Fcontrol

top nodes mesh split F1,bot F2,bot
(first retained) overclosure tyings
u1,bot u2,bot ucontrol
(overlap) ucontrol
control node
(second retained) u1,top u2,top
bottom nodes F1,top F2,top
(tied)
bottom part bottom part
undeformed deformed
Figure 9-42 Pre-stressing a Structure by Creating an Overlap Between the top and the Bottom Part
Using Overclosure Tyings
An overclosure tying imposes the following constraint on the model:
u tied = u retained + u control , (9-41)
in which u tied , u retained and u control are the displacement and rotation degrees of freedom (if any)
of, respectively, the tied node, the first retained node, and the control node of the tying. It immediately
follows from this equation that u control is the displacement difference of the tied and the retained node
of the tying and is equal to the size of the overlap or gap between the parts. Hence, by prescribing this
displacement using the FIXED DISP and DISP CHANGE options, gaps or overlaps of a particular size can
be created.
Instead of prescribing the size, gaps or overlaps can also be created by prescribing the total force in the
model. This follows immediately from the fact that the work done by a constraint equation is zero. To
demonstrate this, consider the model displayed in Figure 9-42. The model is split into two disjoint parts:
the top part and the bottom part. At the split, corresponding nodes of the respective parts are connected
by overclosure tyings, in which the nodes of the bottom part are the tied nodes and the nodes of the top
part are the first retained nodes of the tyings. Both tyings share a common (free) control node. If u 1, bot
and u 2, bot are the displacements of the bottom nodes and u 1, top and u 2, top are the displacements of
the top nodes, then the two overclosure tyings impose the following two constraints on this model:
u 1, bot = u 1, top + u control , (9-42)
u 2, bot = u 2, top + u control . (9-43)
Introducing the vectors,

T T
u = u 1, bot u 2, bot u 1, top u 2, top and û = u 1, top u 2, top u control ,
and the constraint matrix

T
I 0 I 0
T = 0 I 0 I ,
I I 0 0
in which I and 0 are the unit and null matrices, Equations (9-42) and (9-43) can be summarized into the
matrix equation:
u = Tû . (9-44)
Let
T
F = and T
F 1, bot F 2, bot F 1, top F 2, top Fˆ = Fˆ 1, top Fˆ 2, top Fˆ control
be the force vectors which are work conjugate to the displacements u and û , respectively. Then the zero-
work principle states that
T T
F u = Fˆ û . (9-45)
Substitution of Equation (9-44) into Equation (9-45) and requiring that the result is valid for all û , it
T
follows that T F = Fˆ , or
F 1, bot + F 1, top = Fˆ 1, top , (9-46)
F 2, bot + F 2, top = Fˆ 2, top , (9-47)
F 1, bot + F 2, bot = Fˆ control . (9-48)
Equations (9-46) and (9-47) express the equilibrium of forces across the split, in which F̂ 1, top and
F̂ 2, top can be viewed as externally applied forces. While Equation (9-48) enables overclosure tyings to
pre-stress a model with a certain total force. It states that if the constraints described by Equations (9-42)
and (9-43) are applied, the force on the control node is the sum of the forces on the tied nodes of all
overclosure constraints which share that control node. Hence, the total force on the bottom part of the
model is prescribed by applying a force to the control node using the POINT LOAD option. Obviously (by
the equilibrium Equations (9-46) and (9-47)) if no external force is applied to the nodes on the split, the
total force on the top part is equal but opposite is sign to the force on the control node.
CHAPTER 9 653
Boundary Conditions
Notes: Both methods to control the size of overlaps or gaps can be combined in one analysis. For
example, in the first part of an analysis, a pre-stress can be defined by prescribing the net force
in the structure using the POINT LOAD option on the control node of the overclosure tyings.
Then in the second part, the size of the gap or the amount of overlap which is the result of this
load, can be fixed by suppressing the displacement changes of the control node using the
DISP CHANGE option.
If similar arguments are applied to tying 100 (which yields Equation (9-41) without the
dependence on the control node), then only the equilibrium Equations (9-46) and (9-47)
are obtained. Equation (9-48) is solely due to the dependence of Equation (9-41) on the
control node.
The control node of an overclosure tying has the same displacement and rotation degrees of freedom as
the tied and the first retained node as given by Equation (9-41). This allows creation of overlaps or gaps
in any particular direction or combination of directions. Local coordinate systems can be defined at any
of the nodes of an overclosure tying using the TRANSFORMATION option. Since the constraint Equation
(9-42) is always applied in the global coordinate system, a local coordinate system at the control node
can be used to pre-stress the model in directions other than the global coordinate directions.
Sufficient boundary conditions must be applied on the control node to suppress any rigid body modes, if
the two parts of the structure are not constrained otherwise. Total forces on the split are available for post-
processing as reaction forces on the control node for all suppressed or prescribed displacements and
rotations.
Overclosure tyings can be used in combination with the CONTACT option, that is, the tied and the first
retained node may be nodes on the boundary of a contact body.
In non-mechanical passes of a coupled analysis, the overclosure tying reduces to tying type 100 between
the tied node and the first retained node. There is no dependence on the control node in that case. This
guarantees continuity of the primary field variable (for example, temperature) across the split.
Table 9-25 Overclosure Tying
Number of
Tying Retained
69 2 Create gaps or overlaps between The second retained node is the
two parts of a model either by control node of the tying. The
prescribing the total force on the force on this node is equal to the
nodes on either side of the gap or total force on the tied nodes of all
overlap or by prescribing the size tyings that share this control node.
of the gap or overlap. The displacement of the node is
equal to size of the gap or overlap
between the parts. In non-
mechanical passes, the tying
reduces to tying type 100 between
the tied and the first retained node.
9 Insert
Boundar Marc provides an INSERT model definition option which allows the definition of host bodies and lists of
y elements or nodes to be inserted in the host bodies. The degrees of freedom of the nodes in the inserted
Conditio node list or element list are automatically tied using the corresponding degrees of freedom of the nodes
ns in host body elements based on their isoparametric location in the elements.
The INSERT model definition option can be used to place reinforcing cords or rods, such as 2-D rebar
membrane elements, into solid elements.
The INSERT model definition option can be used to apply point loads in some specific locations other
than element nodes. It also can be used to link two different meshes.
If a node to be inserted is also a node of a host body element, no tying is applied to the node.
Transformation must not be used at any nodes of host body elements and at inserted nodes, unless the
same set of local coordinate system is used for all nodes involved.
AUTOMSET
This parameter automatically re-writes the constraint equations such that a tied degree of freedom is not
used as a retained degree of freedom in another constraint equation. It also modifies constraint equations
where a constrained degree of freedom is used in a prescribed boundary condition. Because of chaining
of constraint equations, this is not always possible, in which case the analysis will terminate. The
constraint equations that are considered are those obtained from user-specified options:
TYING
SERVO LINK
RBE2
RBE3
RROD
CONRAD GAP
Constraint equations internally generated are not considered from the following options:
CONTACT
INERTIA RELIEF
INSERT
If both the MPC-CHECK and AUTOMSET parameters are included in the model, the MPC-CHECK
parameter is ignored.
Support Conditions
Marc provides linear and nonlinear springs and foundations for the modeling of support conditions. For
dynamic analysis, a dashpot can also be included. Table 9-26 lists input options for linear and nonlinear
springs and elastic foundations.
CHAPTER 9 655
Boundary Conditions
Table 9-26 Input Options for Springs and Elastic Foundations

Input Options
Springs SPRINGS USPRNG
Elastic Foundation FOUNDATION FOUNDATION USPRNG
The force in the spring is
F = K ( u2 – u1 ) + D ( v2 – v1 ) (9-49)
where K is the spring stiffness, u 2 is the displacement of the degree of freedom at the second end of the
spring, and u 1 is the displacement of the degree of freedom at the first end of the spring. In a dynamic
analysis, D is the damping factor and v 2 and v 1 are the velocities of the nodes.
You can specify the elements in Marc to be supported on a frictionless, elastic foundation. The
foundation supports the structure with a force per unit area (force per unit length, for beams) given by
p n = Ku n (9-50)
where K is the equivalent spring stiffness of the foundation, and u n is the normal displacement of the
surface at a point in the same direction as p n .
The same conventions apply to the elastic foundation specification as for pressure specification, in terms
of the face of the element that is used. The force is applied whether the displacement is tensile or
compressive.
Nonlinear spring stiffness for mechanical, thermal, and electrical analyses can be specified through
tables using the parameter and model definition option. For more details, refer to Chapters 2 and 3 in
Marc Volume C: Program Input.
The USPRNG user subroutine allows you to supply or further modify a nonlinear spring stiffness or
specify a nonlinear foundation stiffness as a function of prior displacement and force.
Bushings
A generalized spring-damper capability is provided in Marc to model bushings between various
components. They are modeled using element types 194 and 195 for 2-D and 3-D applications,
respectively. The connectivity of these elements is generally comprised of two nodes which are specified
through the CONNECTIVITY model definition option. Grounded bushing elements can be defined by
providing only the first end-node for connectivity. The properties of these bushing elements are specified
through the PBUSH or PFAST model definition option. These elements may be automatically generated
using the CFAST option.
Bushing Coordinate System

A unique coordinate system is used to specify the bushing properties. A number of options are available
to define this coordinate system:
• local element coordinate system along the bushing element
• global cartesian coordinate system, or
• a user-specified coordinate system specified through the COORD SYSTEM option.
The bushing coordinate system defined through global cartesian coordinates or through the
user-specified COORD SYSTEM option are maintained constant and are not updated for large
displacement analysis.
The local element coordinate system can only be used for 2-node bushing elements. In this case, the local
X axis is along the element length and forms a perpendicular triad with the local Y and Z axes. In 2-D,
the local Z axis is always taken as (0,0,1) and the local Y axis is then perpendicular to the local X axis in
the global x-y plane. In 3-D, the local Y-Z axes are located by using an extra node or an orientation vector.
If both are specified, the extra node takes precedence. The local Y axis is first estimated by using the
orientation vector or using the vector from the first bushing node to the extra node. The local Z axis is
then formed by taking the cross product of the local X and estimated local Y axis. The final local Y axis
is obtained by taking the cross product of the local Z and local X axes. For large displacement analysis,
the local element coordinate system is updated continuously. Mid-increment direction cosines are used
to calculate the strains while end-increment direction cosines are used to calculate the stiffnesses and
internal bushing forces.
Nodal Offsets for Bushing Elements

The exact position of the bushing element can be modified by specifying nodal offsets. This offset point
is generally defined from the first end-node of the bushing element. It serves two purposes: It allows
convenient modeling of the bushing at a location away from the actual physical location. Also, it allows
coupling between translational and rotational stiffnesses. The offset point can be located along the
bushing element axis, in the global cartesian coordinate system, or in a user-defined coordinate system.
In the first case, the offset point is located along the line joining the two end nodes of the bushing element.
The distance of the offset point from the first node S is specified by the user. In the second case, the offset
vector to the bushing location from the first end-node is specified.
The treatment of offsets for bushing elements closely follows the treatment of offsets for beam elements
(described later in Beam - Shell Offsets). Once the offset point location is specified, the offset vector from
each node is determined. The equations governing the treatment and update of the offsets are given by
Equations (9-62) to (9-65). Bushing properties are used to define the stiffness/damping/mass matrix at
the offset position and are then transformed to the geometrical position.
CHAPTER 9 657
Boundary Conditions
zelem
yelem
zelem Bush
Location CID
xelem
(S1,S2,S3) OCID
1 v
S.l yelem
2
(1-S) . l
Bush Location
1
2
xelem
Figure 9-43 Offset location Specification for Bushing Element
Bushing Properties
Five sets of properties can be provided for bushing elements depending on the analysis type. All
properties are provided through the PBUSH or PFAST model definition option:
• mechanical stiffness - for static/dynamic/harmonic analysis
• mechanical dashpot - for dynamic/harmonic analysis
• mass - for dynamic/harmonic analysis
• thermal conduction coefficient - for heat transfer/coupled analysis involving a thermal pass
• electrical resistance coefficient - for electrical/coupled analysis involving a joule pass
Mechanical stiffness for bushing elements can be specified either directly in coefficient form or through
force-displacement curves. Note that 2-D bushing elements have two translational and one rotational
degree of freedom and separate stiffnesses can be specified for the X, Y and RZ degrees of freedom.
Three-dimensional bushing elements have three translational and three rotational degrees of freedom and
separate stiffnesses can be specified for X, Y, Z and RX, RY, RZ degrees of freedom. For nonlinear
bushing elements, these stiffnesses can be varied through tables. These tables can be a function of time,
increment number, position coordinates, frequency (only for harmonic analysis), temperature,
displacement (this variable is mandatory when stiffness is specified via forces). Up to four independent
variables can be used in the tables. Based on the current stiffness value and the change in displacement
between the end-nodes, the bushing forces are updated in the specified bushing coordinate system.
Two forms of damping can be specified for the bushing elements: Nominal damping B i for the ith degree
of freedom of the bushing elements where the damping is specified either directly in coefficient form or
through force-velocity curves. For nonlinear dampers, the nominal damping can be varied through tables
as a function of time, increment number, position coordinates, frequency (only for harmonic analysis),
temperature or velocity (this variable is mandatory when nominal damping is specified via forces). The
second form of damping is structural damping Si for the ith degree of freedom of the bushing elements.
The damping is specified in coefficient form only in this case and can be varied through tables with
similar independent variables as the nominal damping case. A composite damping coefficient is formed
for the bushing element as follows:
G 1
C i = B i + -------- K i + -------- S i K i
W3 W4
where C i is the total damping coefficient; G, W3, and W4 are structural coefficients that are specified
through the COEFFICIENT model definition option; B i is the nominal damping coefficient; S i is the
structural damping coefficient; and K i is the current stiffness coefficient.
For a linear bushing element, the total force is given by Equation (9-49). Note that for bushing elements
defined in the local element coordinate system, the force is positive if the bushing element is in tension
and negative if it is in compression. For bushing elements defined via the global coordinate or
user-coordinate systems, the force along a particular degree of freedom is positive if the displacements
of end-node 2 along the specified degree of freedom is greater than the displacement of end-node 1
along the specified degree of freedom. Note that if tables are used to define the bushing properties,
care should be taken to include both positive and negative values of independent variables like
displacements/velocities since these would influence the calculated stiffnesses.
For dynamic and harmonic analyses, lumped translational and rotational masses can be provided at the
end-nodes of the cbush elements. Uniform masses apply to translational and rotational degrees of
freedom at the end-nodes and are maintained constant through the analysis.
Stresses in the bushing elements can be related to the associated forces through recovery coefficients.
Similarly, strains in the bushing elements can be related to the associated deformations through recovery
coefficients. Separate coefficients are allowed for the translational components and the rotational
components through the PBUSH option. If set to 1.0, the stresses are equal to the forces and the strains
are equal to the difference in displacements between the end-nodes.
The bushing elements can also be used for thermal or electrical field analyses. The coefficients of thermal
conduction and electrical resistance are specified via the PBUSH option. The bushing element behaves
like a general link in this case and dashpot properties are not considered. Tables can be used to vary the
properties as a function of temperature (for thermal conduction), voltage (for electrical resistance), time,
increment number, coordinates.
Cyclic Symmetry
A special set of tying constraints for continuum elements can be automatically generated by the Marc
program to effectively analyze structures with a geometry and a loading varying periodically about a
symmetry axis. Figure 9-44 shows an example where, on the left-hand side, the complete structure is
given and, on the right-hand side, a sector to be modeled.
CHAPTER 9 659
Boundary Conditions
Y Y
y’ x’
B
α
X X
A
Figure 9-44 Cyclic Symmetric Structure: Complete Model (Left) and Modeled Sector (Right)
Looking at points A and B on this segment, the displacement vectors should fulfil:
u' B = u A (9-51)
which can also be written as:

u B = Ru A (9-52)
where the transformation matrix R depends on the symmetry axis (which, in the example above,
coincides with the global Z-axis) and the sector angle α (see Figure 9-44). In Marc, the input for the
CYCLIC SYMMETRY option consists of the direction vector of the symmetry axis, a point on the
symmetry axis and the sector angle α . The following items should be noted:
1. The meshes do not need to line up on both sides of a sector (for example, see Figure 9-45).
Figure 9-45 Finite Element Mesh for Cyclic Symmetric Structure with Different Mesh Densities
on the Sector Sides
2. Any shape of the sector sides is allowed provided that upon rotating the sector 360 ⁄ α times
about the symmetry axis over the sector angle α will result in the complete model.
3. The CYCLIC SYMMETRY option can be combined with the CONTACT option. In this case, both
sides of the cyclic symmetry sector must belong to a contact body and corresponding parts of the
sector sides must belong to the same body (see Figure 9-46). Note that the contact bodies do not
need to span between the sector sides.
Figure 9-46 Cyclic Symmetry and Contact Bodies; upper left: one body (correct); upper right:
two bodies (correct); lower left: two bodies (right); lower right: two bodies (wrong;
corresponding parts of the sector sides are not in the same body).
4. The CYCLIC SYMMETRY option can be combined with global remeshing.

5. In a coupled thermo-mechanical analysis, the temperature is forced to be cyclic symmetric
( T A = T B as in Figure 9-44).
6. A nodal point on the symmetry axis is automatically constrained in the plane perpendicular to the
symmetry axis.
7. The possible rigid body motion about the symmetry axis can be automatically suppressed.
8. Cyclic Symmetry is valid for:
a. only the continuum elements. However, the presence of beams and shells is allowed, but there
is no connection of shells to shells, so the shell part can, for example, be a turbine blade and
the volume part is a turbine rotor. The blade is connected to the rotor and if there are 20 blades,
1/20 of the rotor is modeled and one complete blade.
b. nonlinear static analysis including remeshing as well as coupled analysis.
c. valid for all analysis involving contact.
d. valid also for: eigenvalue analysis such as buckling or modal analysis, harmonic analysis, and
transient dynamic analysis. However, there are restrictions in the case of modal analysis which
are described in more detail in the following paragraphs.
9. Cyclic Symmetry is invalid for pure heat transfer.
10. In order to check which nodes get tying constraints due to cyclic symmetry, nodal post code 38
(contact status) can be used; its value will be 2.
CHAPTER 9 661
Boundary Conditions
To prevent confusion, it must be emphasized that the cyclic symmetry feature described above is
different than linear cyclic symmetry commonly used in modal analysis where physical quantities such
as, x n , displacements, forces, stresses and temperature in the n-th segment are expanded in a Fourier
k
series with terms of the cyclic components, u , in the fundamental region, like:
K N
n – 1 ----
1 0 2 k, c k, s ( –1 ) 2
xn = ---- u +
N
----
N  [u cos ( n – 1 )kα + u sin ( n – 1 ) kα ] + ---------------------- u
N
(9-53)
k = 1
where k is the harmonic order; N is the total number of sectors; α is the fundamental inter-sector phase
shift defined as 2π ⁄ N ; and K is defined as:
N –1
-------------- if N is odd
 2
K =  (9-54)
 N –2
-------------- if N is even
 2
There are considerable savings in both computing time and data storage associated with the use of the
linear cyclic symmetry concept. Assuming a finite element model with a sector size of m degrees of
freedom, a real-valued cyclic symmetry approach leads, in the worst case, to one eigenvalue problem of
size m and ( N – 1 ) ⁄ 2 eigenvalue problems of size 2m ; a complex approach leads to N eigenvalue
problems of size m; while the full analysis leads to a single, but very costly, eigenvalue problem of
size Nm .
Although linear cyclic symmetry can reduce the problem size greatly, it is restricted to linear analysis,
and the sector must have its surface mesh on the symmetry planes to be identical on each side of the
sector. The nonlinear cyclic symmetry implemented in Marc can be used for nonlinear problems, such
as contact, and the nodes do not need to line up on both symmetry planes of the sector (for example, see
Figure 9-45).
Nastran RBE2 and RBE3

The basic formulation of the Nastran rigid body elements RBE2 and RBE3 is similar to tying type 80,
which is also known as a rigid link. However, they offer additional flexibility because the degrees
of freedom may be partly unconstrained and a list of tied/retained nodes can be entered with RBE2
and RBE3.
RBE2
The nonlinear relation of an RBE2 can be expressed as follows:
xt = R ( θr ) ( xt – xr ) (9-55)
where x t is the coordinate of the tied node, x r is the coordinate of the reference node, and R ( θ r ) is the
rotation matrix of the retained node. The linearized relation between the incremental displacement of a
tied node to a retained node and its correction term is expressed as follows:
i i
Δu t = SΔu r + C n o n (9-56)
where the tying matrix S is derived from the linearized rigid body relation. For linear analysis, C n o n is
i
zero. For large displacement analysis, SΔu r may not match the nonlinear constraint exactly. Therefore,
an error vector, C n o n , is required to meet the constraint. It is defined as the difference between the
expected coordinates ( x t e ), using the rigid body kinematics, and the current coordinates ( x t ) of the
tied node.
n+i n n+i 0 0
xt e = xr + R ( θ r ) ( xt – xr )
(9-57)
n+i n+i
Cn o n = xt e – xt
where n is increment number, i is iteration number, and zero is the original value. In order to reduce
linearization errors, an optional recycling feature is offered through the CONTROL model and history
definition options. Recycling occurs if the maximum iterative rotation change over the retained nodes of
the RBE2 exceeds a rotation tolerance (default is 0.001 radians). This check can be bypassed by
setting the rigid link rotation to zero (in conjunction with the FEATURE,5701 parameter) or a
negative number.
When the rotation is finite, then the coordinate system attached to the tied node will be co-rotated
according to the rotation of the retained node. The degrees of freedom of the tied node will be assigned
with this co-rotated system. As an illustration, consider a slider link shown in Figure 9-47. In this case,
a local transformation must be defined for the tied node in which the x-direction is coincident with the
line connecting the tied and the retained node. Then, an RBE2 is defined where y-translational degree of
freedom is tied but not the x-translational and any of the rotation degrees of freedom.
.
y x
y
E 2
Y x
RB 2
RBE
X (a) Increment i (b) Increment i + 1

Figure 9-47 Co-rotated Coordinate System at the Tied Node
As another example of the use of RBE2 options, consider Figure 9-48 where a cylindrical tube is shown.
The tube is clamped at one end and loaded by a torque at the other end.
CHAPTER 9 663
Boundary Conditions
If it is assumed that the loaded end remains circular with a constant radius, then a single RBE2 option
can be easily used instead of a number of tyings of type 80. With the RBE2 option, node 1 is defined as
the retained or reference node and nodes 2 to 13 are defined as the tied nodes, with all the degrees of
freedom constrained.
However, if the tube is allowed to expand in radial direction, then tying type 80 cannot be used. However,
RBE2 option can still be used, but now the radial displacement is not included in the list of degrees of
freedom to be constrained.
2
13
3
12
11 θ
4
Mz Z 1
10 5
R
Clamped end
6
9
7
8
Figure 9-48 RBE2 Example: Cylindrical Tube loaded in Torsion
RBE3
The RBE3 option is a powerful tool to distribute applied loads and massed in a finite element model.
Forces and moments applied to reference or tied node are distributed to a set of independent degrees of
freedom based on the RBE3 geometry and local weight factors.
The way in which the forces are distributed is analogous to the classical bolt pattern analysis. Consider
the bolt pattern shown in Figure 9-49 with a force F A and moment M A acting at reference point A. The
force and moment can be transferred directly to the weighted center of gravity along with the moment
produced by the force offset e .
FA FA
Reference point A
CG MA
MCG = MA + FAe
e
Figure 9-49 Transfer of Force and Moment on a Reference Point to the Weighted Center of Gravity (CG)
The force is distributed on the bolts proportional to their weighting factors. The moment is distributed
as forces proportional to the product of their distance from the center of gravity with corresponding
weighting factors, as shown in Figure 9-50. The total force acting on the bolts is equal to the sum of the
two forces. These results apply to both in-plane and out-of-plane loadings.
Fi
Force Distribution:
 ωi 
F i = F A  -------------
  ω k
Fi Moment Distribution:
 M C G ω i r i
F i =  -----------------------
ri   ω k r 2k 
where F i = force at DOF i

ω i = weighting factor for DOF i
r i = radius from the weighted
center of gravity to point i
Figure 9-50 RBE3 Force and Moment Distribution
As an example, consider the cantilever plate model with a single quadrilateral shell element shown in
Figure 9-51. The plate is subjected to nonuniform pressure represented by a resultant force acting at a
distance of 10 mm from the center of gravity. The easiest way to apply the pressure is to use an RBE3
option, with a reference node 5, to distribute the resultant load to each of the four corner points. Setting
the weighting factors to be equal for all nodes, results in a force distribution based exclusively on the
spatial location of the connected nodes. In this way, nodes 3 and 4 will get 3N each and nodes 1 and 2
will get 2N each.
10N
2
3
m
5
100 m
1
50 mm
4
40 mm
100 m m
Figure 9-51 Using the RBE3 Option to Represent a Nonuniform Pressure Load
The formulation of the RBE3 is based on the idea that the motion at a reference node is the weighted
average of the so-called error terms of the generalized displacements at a set of connected nodes. The
error terms x e l , and θ e l are defined as follows:
CHAPTER 9 665
Boundary Conditions
xl = R ( θt ) ( xl – xt ) + xe l
(9-58)
θl = θt + θe l
where x l and θ l are the coordinates and rotations of the l-th retained node, respectively; x t and θ t are
the coordinates and rotations of the reference node, respectively; R is a rotation matrix of the reference
node. The error measure for translational and rotation degrees of freedom are weighted with weighting
factors, W l .
N
T
et r a n =  xe l Wl xe l
l = 1
(9-59)
N
T
e ro t =  θe l Wl θe l
l = 1
The final nonlinear constraint relation is derived by minimizing the error terms with regard to the
reference node coordinates and rotations using least square fit.
The linearized relation between the reference node and the connected nodes and its correction term can
be expressed as:
i i
Δu t = SΔu r + C n o n (9-60)
with C n o n being zero for linear analyses. For large rotation analysis, the error vector, C n o n , is required
rot
to meet the constraint. The rotational error, C n o n , is assumed to be zero. In other words, the expected
incremental rotation of the reference node is the same with the current value.
t ra n
The translational error, C n o n , is derived by evaluating the difference between the expected coordinate
( x t e ) of the reference node with the current one ( x t ). The expected coordinates are calculated by
introducing the coordinates of the connected nodes at the previous increment and the incremental
rotation of the reference node into the nonlinear constraint equations.
–1 N
 N 
xt e =   W l  W l ( x l + R l ( Δθ t ) )
 
l = 1  l = 1
(9-61)
tran
Cn o n = xt e – xt
Beam - Shell Offsets

In many problems, it is convenient and sometimes necessary to model beam and shell elements at a
geometric location that does not match the actual physical location. Such cases are common when flanges
or surfaces for shell and beam elements with varying cross-section properties have to be aligned. Another
common instance is when beams are used as stiffeners for shells. It is most convenient to model the beam
elements by sharing the shell nodes at the midsurface of the shell, as shown in part (a) of Figure 9-52.
The fact that the beam is actually offset by half the plate thickness and half the height of the beam section
will have to be achieved by providing a suitable beam offset as shown in part (b) of Figure 9-52.
(a) Geometric model with the (b) Physical model with the beam
beam element sharing the offset by half-shell thickness
nodes of the shell elements. + half-beam height.
Figure 9-52 Beam-stiffened Shell Structure
There are two methods by which beam/shell offsets can be modeled:

The first method is to place the beams and shells at the actual offset position and then tie the
nodes of these elements back to the original position through manually defined RBE2 links.
While this method is accurate, it is difficult to used for large models. Furthermore, it is not
possible for the offset elements to contact other bodies since the nodes of the offset elements are
already tied.
The second method involves the beam/shell offset capability in Marc. The elements are specified
at the original position in the model and the offset vectors at the nodes of the element are
specified in a variety of ways through the GEOMETRY option. The current nodal offset is applied
to each node of the offset element to calculate the offset element position.
n+i 0 n+i
x o f f se t = x + voff (9-62)
n+i
where n is the increment number, i is the iteration number, x o f f se t is the current coordinate
0
vector of the node at the offset position, x is the original user-specified coordinate vector of the
n+i
node at the original position, and voff is the current offset vector. For small displacement
0
analysis, the initial user-specified offset vector, voff , is maintained through the analysis. For
large displacement analysis, the offset vector is rotated based upon the rotations at the node.
n+i n+i 0
voff = R(θ )voff (9-63)
n+i n+i
where R ( θ ) is the current rotation matrix at the node and θ is the current rotation vector
CHAPTER 9 667
Boundary Conditions
at the node.
The tying relationship between the offset nodal position and the original nodal position is then
established. The basic formulation for this tying relationship is similar to the RBE2 link
described in the previous section.The relation between the incremental displacement of the node
in the offset position to the corresponding displacement at the original position is expressed as
follows:
Δu oi f f se t = SΔu oi r i g i n a l + C n o n (9-64)
where the tying matrix S is based on a linearized rigid body relation and its terms are derived
from offset vector components. For linear analysis, C n o n is zero. For large displacement
i
analysis, SΔu r may not match the nonlinear constraint exactly. Therefore, an error vector,
C n o n , is required to meet the constraint. It is defined in a manner similar to Equation (9-57).
In order to reduce linearization errors, an optional recycling feature is offered through the
CONTROL model and history definition options. Recycling occurs if the maximum iterative
rotation change over the offsets nodes exceeds a rotation tolerance (default is 0.0001 radians).
This check can be by passed by setting the rigid link rotation tolerance to zero (in conjunction
with the FEATURE,5701 parameter) or a negative number.
The element stiffness matrix is formed on the offset element geometry. It is then converted to the
original element geometry through a suitable tying relationship.
T
Ko r i g i n a l = S Ko f f s e t S (9-65)
All distributed loads on the offset element are also treated in the same way. Note that point load
vectors are not offset and are always calculated at the original nodal positions. The nodal
displacements at the original positions are calculated during the solution phase. Finally, during
the recovery phase, the displacements at the offset position are calculated from Equation (9-64)
and element strains, stresses are calculated on the offset element geometry.
Additional details for beam/shell offsets in Marc are provided below:
Offset Vector Specification

For beams, offset nodal vectors can be specified in four different ways:
1. Offset nodal vector specified in the global coordinate system.
2. Offset nodal vector specified in the local beam coordinate system:
For 2-D beams (element types 5, 45), the x-axis is along the beam axis, the z-axis is (0,0,1) and
the y axis is perpendicular to both.
For 3-D beams, the z-axis is along the beam, the x-axis is user-specified and the y-axis is
perpendicular to both. This option is not available for heat transfer links.
3. Offset nodal vector along the associated shell normal:

This option is particularly useful when beams are used as stiffeners for shells. In the case, only
the offset magnitude is provided by the user. The associated shell normal vector at the node is
automatically calculated and is used for defining the offset vector. This option is not available for
2-D beams.
4. Offset nodal vector specified in the local coordinate system at the node:
The local coordinates at the node can be specified using the TRANSFORMATION, CYLINDRICAL,
or similar options.
For shells, the offset magnitude is provided by the user and the element normal calculated at the centroid
of the shell element is used to define the nodal offset vector.
Supported Elements
Beam/Shell offset support is provided for all standard beam and shell elements with global translational
and rotational degrees of freedom at the corner nodes.
Supported mechanical beam elements include element types 5, 14, 25, 45, 52, 76, 77, 78, 79, 98. Heat
transfer links, element type 36, 65, 99, and 100 are also supported (see section on field analysis below).
Non-standard beam elements that are not supported include element types 13, 16, 51. Supported
mechanical shell elements include element types 1, 22, 75, 89, 138, 139, 140. Heat transfer shell
elements, element types 50, 85, 86, 87, 88 are also supported (see section on field analysis below). Non-
standard shell elements that are not supported include elements types 4, 8, 15, 24, 49, 72.
For higher order beam/shell elements (element types 45, 22, 89), there is a choice to use interpolated
values of the corner node offset vectors at the midside nodes. If this is not chosen, the offset is taken as
0 at the mid-side nodes. For beam elements 76, 77, where the midside node degree of freedom is not
compatible with the corner node degree of freedom, only the nodal position is updated for the offset and
no other tying is undertaken at the midside node.
Post File
All nodal vectors on the post file are calculated and depicted at the original nodal positions. This allows
visual compatibility when elements with offsets and elements without offsets are in the model. Element
quantities like strains and stresses on the post file are calculated based on the offset element geometries.
Adaptive Meshing
There is no support for global remeshing with beam/shell offsets. Local adaptive meshing is supported.
Nodal offsets are automatically interpolated and transferred over to the new nodes created during the
adaptive meshing process.
Field Analysis
Beam/shell offsets are predominantly supported for mechanical analysis. Since offset elements can also
be used in thermo-mechanical analysis and other field analyses, offset support is extended for field
analyses other than purely mechanical analysis. In this case, only the coordinates of the element are
adjusted for the offsets, as per Equation (9-62). No other tying of the field variables or element matrices
is undertaken.
CHAPTER 9 669
Boundary Conditions
Contact
The average nodal offset vector is calculated at every node of all beam/shell contact bodies. All touching
nodes and touched patches are then updated using the average nodal offset vectors.
Dynamics
The stiffness matrix, mass matrix and damping matrix for the offset element are all calculated in a
manner similar to Equation (9-65). Implicit in this calculation is the assumption that the displacements
as well as velocities and accelerations between the original and offset positions are tied in a manner
similar to Equation (9-64).
Pin Code for Beam Elements

Pin code is used to remove connections between the node and selected degrees of freedom of the beam.
It can only be applied on the nodes at the ends of the beam. It is not allowed to use it for the middle node
of 3-noded beams. The degrees of freedom are defined in the element’s coordinate system (see Marc
Volume B: Element Library for more detail description how the local coordinate system of beam element
is set up), and the pin code is applied at the offset ends of the beam.
Marc manipulates the pin code in two different ways. The first one, by default, is done by internally
generating a new node which is tied to the flagged node. The tied degrees of freedom are the negation of
the pinned ones. The tying process is carried out at element level. The element stiffness matrix and
internal forces of the element are calculated using a standard procedure. During the assembly process
they are “expanded” to include the degrees of freedom of the new internally generated node. This
procedure works fine both for quasi static and dynamics analysis.
The second one, using FEATURE, 6901, is done by using the so-called static condensation technique.
In this way, no internal node is generated. The pinned degrees of freedoms are condensed out from the
element stiffness and internal forces before they are assembled into the global system. This feature is
added to allow Marc pin code fully compatible with Nastran implementation. For quasi-static analysis,
both ways give exactly the same behavior. For dynamics analysis, the coefficients used to condensed out
the element stiffness matrix is reused for mass matrix. This is an approximation to the correct procedure
where the static condensation has to be done on the element operator matrix before being assembled into
global system.
The pin code is activated using PIN CODE or the GEOMETRY option. For postprocessing purposes, the
element connectivity is always presented with the original nodes. For example, in a stress-free large
displacement analysis for a beam element with pin code that simulate slider, the beam length will become
shorter or longer although the element is stress free (see Figure 9-53; node 4 is generated internally. If
FEATURE, 6901 is activated, node 4 is not generated). The pin code is defined on element 10 at node
2. The axial-displacement and out-of-plane rotation degrees of freedom are flagged.
2
10 20 4
10
20
1 1
3 3
Figure 9-53 Slider Connection using Pin Code
Mesh Independent Connection Methods

In Marc, there are various ways to model connections between structural components in a finite element
model. These connections may represent spot welds, seam welds, bolts, screws, rivets, and so on.
Depending on the modeling goals, they can be modeled using line elements (like trusses and beams),
linear or nonlinear springs, rigid body connections (RBE2, RBE3), and tyings or servo links (MPC
constraints). In this way, general connections can be made between parts, but they often require a lot of
modeling effort in defining the connections or in assuring that meshes of the parts involved are congruent
so that they match in those points where connections are desired. Data and input preparation for these
connections can be a formidable task in real-world applications. Often, it is desired to model certain
stiffness characteristics in the connections. An easy way to establish this is by making point-to-point
connections between the parts and using beam elements or springs as the connector elements. Apart from
the high meshing effort, this modeling technique may also produce poor results. Increasing mesh
refinement can introduce stiffness errors; point-to-point connections in which the effective cross-
sectional areas are larger than 20% of the characteristic element face areas can often lead to significant
underestimation of connector stiffness. Point-to-point connections can also cause numerical problems in
finite element shell models. The stiffness of the rotations normal to the shell plane may not exist or may
be represented by a penalty stiffness to suppress the associated singularity, but in the point-to-point
connections, such rotations may be excited. If the surface of one of the parts is defined by the faces of
continuum elements, there are no rotational degrees of freedom for the nodes of these elements and point-
to-point connections may introduce undesired singularities.
The CWELD and CFAST connection methods that are available in Nastran have also been implemented
in Marc to address and solve the aforementioned modeling issues. Connections can now be established
with ease between points, elements, patches, or any of their combinations and load transfer between the
parts can be established over multiple nodes on each side of a connection providing a more accurate
stiffness representation of the connection. The following connections between two parts can be made:
patch-to-patch, point-to-patch and point-to-point as illustrated in Figure 9-54.
CHAPTER 9 671
Boundary Conditions
Patch-to-patch Connection Point-to-patch Connection Point-to-point Connection

For noncongruent meshes, For noncongruent meshes, For nearly congruent meshes,
area wise connection point to area wise connection point wise connection
Figure 9-54 Supported Connections Between Different Parts
Each connection involves the two end nodes of a connector element and two surfaces that are to be
connected at a certain location. The connector element can be any Marc beam element, a truss element
or a cbush element. The surfaces can be defined by shell elements, the faces of continuum elements or
by nodal points. These surface element faces are referred to as patches, can be triangular or quadrilateral,
and may be linear or quadratic. On each side of the connection, either a patch connection or a point
connection is made thus giving the three connections shown in Figure 9-54. A patch-to-patch connection
can be shell-to-shell, shell-to-solid, or solid-to-solid. A point-to-patch connection can be point-to-shell
or point-to-solid. A point connection is the simplest connection since the end node of the connector
element is directly connected to a node of the surface; the connector element and the surface share a
common node and no additional constraints are required to establish the connection. A patch connection
connects the end node of a connector element to one or more element faces and is more general as it
eliminates the need of congruent meshes. The location where the end node pierces the surface is
automatically determined. The element face being pierced is called the master patch. Internal constraints
are generated to assure a rigid (i.e., a force and moment carrying) connection of the connector end node
to the surface. Two types of patch connections can be made: Direct Patch Connection which connects
the connector end node directly to the nodes of the master patch with tying type 44; the constraints
involved in this tying are also known as Kirchhoff constraints, (see Rigid Tying to a Surface Patch in this
chapter) or the Indirect Patch Connection which connects the connector end node indirectly to the nodes
of the surface by constructing a four-node auxiliary patch at the end node normal to the connector. The
size of this auxiliary patch is determined from the diameter or area of the connector. The end node of the
connector makes a rigid (tying type 44) connection to this auxiliary patch. The four nodes of this
auxiliary patch (the auxiliary nodes) are connected to the surface by additional constraints. The location
where each auxiliary node pierces the surface is automatically determined. The patch being pierced is
called the secondary patch. By default, RBE3 constraints are generated between each auxiliary node and
the nodes of the secondary patch it pierces, but other constraints may be chosen as well. Note that,
depending on the mesh density, two or more auxiliary nodes may connect to the same secondary patch.
All piercing points are determined by normal projections to the surface to which the connection is made.
The two patch connection types are illustrated in Figure 9-55 which displays a top view of one side of
the connection involving regular four-node quadrilateral element faces. Other element face types and
element faces with collapsed edges may be present as well. All constraints are applied internally as tying
constraints and they are automatically generated when processing the input.
master patch
auxiliary patch
master patch
secondary patch
Direct Patch Connection Indirect Patch Connection

Figure 9-55 Patch Connection Types
The direct patch connection is the simpler type, but may lead to asymmetrical load transfer between the
surfaces when the projections are near the element face boundaries and the element faces are not nicely
facing each other. It may also lead to a very localized load transfer in case of fine meshes, resulting again
in a poor stiffness representation of the connection. The indirect patch connection improves the load
transfer by incorporating more element faces, and thus more nodes, in the connection, but this will also
lead to more constraint equations.
In general, a connection between two surfaces can be defined by entering a point that identifies the
approximate location of the connection and the two surfaces that are to be connected. This point can be
entered by referencing a node or by entering its coordinates when defining the connection and it is
referred to as the GS-node or GS-point. These GS reference points determine the spacing of the
connections and they can be defined beforehand, independently of the surface meshes. The surfaces can
be identified by entering the element faces to which the connection will be made or by entering element
or face sets from which the element faces to connect to will be sought. For indirect patch connections,
only the master patches need to be entered. The secondary patches are automatically found in the vicinity
of the master patches. If the surfaces are entered as sets, the master patches and the secondary patches
are searched from these sets. The connection between the two surfaces is established as follows. The
GS-node is projected to the first (side A) and second (side B) surface by a normal projection. These two
projections determine the end nodes of the connector and therefore its final position and are referred to
as the GA-node for side A and the GB-node for side B. The nodes on side A and B of the master patches
are referred to as GAi and GBi respectively, where i ranges from one to the number of nodes in the patch.
The auxiliary nodes are referred to as GAHi and GBHi respectively, where i always ranges from one to
four. The nodes of a secondary patch are referred to as GAij and GBij respectively, where j ranges from
one to the number of nodes in the patch and i refers to the ith auxiliary node. The definitions of the
connection are made on the CWELD or CFAST model definition options. The properties of the connector
elements are defined on the PWELD or PFAST model definition options. These properties are material
and geometric properties in case of CWELD connections and stiffness properties in case of CFAST
connections. A number of control parameters can be entered on the SWLDPRM option to enhance the
search and projection process and to control its output.
CHAPTER 9 673
Boundary Conditions
CWELD patch-to-patch connections

There are several methods to define CWELD patch-to-patch connections. The methods available are
ELEMID, GRIDID, ELPAT, and PARTPAT. In each of these methods, the surfaces on both sides have
to be defined together with the GS reference point of the connection. The ELEMID and GRIDID
methods make the simpler direct patch connections and the ELPAT and PARTPAT methods make the
more advanced indirect patch connections. All methods can be used to connect noncongruent meshes of
any type and they are recommended when the cross-sectional area of the connector is larger than 20% of
the characteristic element face area. The connector element can be any Marc beam element. The default
type used is 98 which is a straight beam in space that includes transverse shear effects. By default, it
assumes elastic behavior and a solid circular cross-section. It is always required to include the
ELEMENTS parameter for each element type used as a connector, even when it does not appear in the
CONNECTIVITY input. The data for each connection are defined on the CWELD model definition input.
Some data inputs are always required, because no defaults exist for them. Other data inputs are optional
and if not defined, Marc assumes appropriate default values for them. These optional data inputs concern
the connector element type, the way the projections are carried out, the type of constraints and the
orientations of specific CWELD connections.
Direct Patch-to-patch Connections

The ELEMID and GRIDID methods define the connection directly between two master patches. For
ELEMID the patches are entered as shell elements or the faces of continuum elements (SHIDA and
SHIDB). For GRIDID, the patches are entered as nodes (GAi and GBi) following the same connectivity
conventions as for corresponding shell patches. The two methods are illustrated in Figure 9-56. The
GS-node is projected to both patches and, in case it is already positioned between the two patches, these
projections determine the connector end node locations. But it is not a requirement to position the
GS-node between the two patches, and if it is not, the projection is done in two steps as illustrated in
Figure 9-57. First, a normal projection to patch A and patch B is made resulting in the points GA’ and
GB’. Then the midpoint GC’ between these two points is projected to both patches and this determines
the connector end point locations GA and GB. The nodal coordinates of the connector element are
updated after this projection. The GS-node must have a proper projection to the patch within its
boundaries. The PROJTOL parameter can be defined to accept projections outside but near patch
boundaries.
GS GS
SHIDA GA4 SHIDA

GB3
GA3
GA1 GA2
GB2
SHIDB
GB1
ELEMID GRIDID
Figure 9-56 ELEMID and GRIDID Connection Methods
GS GS
PIDB Shel
lB
GB′ GB
GB
SHIDB
L PIDA
GA GC′ GC
SHIDA GA′
D A GA
Shell
Figure 9-57 Determination of the Connector End Node Locations GA and GB
The data needed to define a connection with ELEMID are:

1. The reference node GS or its coordinates (XS,YS,ZS).
2. The patches SHIDA and SHIDB on both sides, where for shell elements these are the element
numbers and for continuum elements these are element and face number pairs.
3. A PWID, which is the identification number of a corresponding PWELD property entry.
The data needed to define a connection with GRIDID are:
2. The patch nodes GAi and GBi on both sides. The number of nodes determines the patch type,
which can be quad4, quad8, tria3 or tria6. No mixed patches with partly linear and partly quadratic
edges are supported.
Indirect Patch-to-patch Connections

The ELPAT and PARTPAT methods make the connection of the connector end nodes to the surfaces
involved in two phases, as illustrated in Figure 9-58 and Figure 9-59. First the reference GS-node is
projected to a master patch. For the ELPAT method, the master patches are specified in the input. For the
PARTPAT method, the master patches are searched from sets that have been specified in the input.
Around each projection point, a square four-node auxiliary patch is constructed. This auxiliary patch is
normal to the line connecting the two projection points. The dimensions of this auxiliary patch follow
from the diameter or the area of the connector as displayed in the right part of Figure 9-58. For the four
auxiliary nodes, new projections are computed on the master patches or on patches in their vicinity. The
patches to which these new projections are found are called secondary patches and each auxiliary node
is connected by default with RBE3 constraints to the nodes of its respective secondary patch.
CHAPTER 9 675
Boundary Conditions
GBH4 GBi
GBH3
GB D/2
Shell B D/2
GAH4 GAH3
GBH1
GBH2
a
D
GA a = π ----
GAH4 4
a
GAH1 GAi
GAH3
GA GAH1 GAH2
a a
GAH2
Shell A
Figure 9-58 ELPAT or PARTPAT Connections
The master patches and the secondary patches do not have to be distinct as can be seen from Figure 9-58
and Figure 9-59. For a coarse mesh, they may even be one and the same patch but for fine meshes they
may all be distinct. The requirement is that the secondary patches do not span a larger area of the mesh
than a 3 x 3 area as shown in the right part of Figure 9-59 otherwise a number of nodes not involved in
the constraints defining the connection is present under the area covered by the auxiliary patch. If the
NREDIA parameter is set and the secondary patches violate the 3 x 3 requirement, the auxiliary patch size
is reduced a number of times in an attempt to fulfill this requirement.
GAij GAij
GAH4 GAH3
GAH4 GAH3
GA
GAH1 GAH2
GAH1 GAH2
Coarse Mesh Fine Mesh

Figure 9-59 Connections in Coarse and Fine Meshes for the ELPAT and PARTPAT Methods
The main purpose of the ELPAT and PARTPAT methods is to improve the load transfer between the two
surfaces. These methods allow for a more symmetrical load transfer than the simpler ELEMID or
GRIDID methods as illustrated in Figure 9-60. On side A in the left part of the figure, the shaded patch
is the master patch and the constraints are only applied between the connector node and the nodes of this
patch. On side A in the right part of the figure the neighboring patches also become involved in the
connection and the load at the connector node is transferred through the four auxiliary nodes to all the
nodes of the secondary patches in the surface, leading to a better load transfer. It also avoids a too
localized load transfer as may be clear from the right part in Figure 9-59. If this connection were of the
ELEMID type, the only patch involved in the load transfer would be the one entirely covered by the
shaded area leading to a much more localized load transfer.
The PARTPAT method defines a connection between two sets. In case the part is meshed with shell
elements, the set can be an element set identifying the elements that make up the surface. In case the part
is meshed with continuum elements, the set must be a face set identifying the faces that make up the
surface. A shell surface can also be identified by a face set, but the face information is not essential and
therefore element sets can be used as well.
The ELPAT method is similar to the PARTPAT method. The only difference is that the ELPAT method
defines the master patches, whereas the PARTPAT method searches them from a set of elements. A
further difference is that for shell meshes the ELPAT method searches its secondary patches from the
whole mesh, whereas the PARTPAT method searches them from the same sets from which the master
patches were found, but sets may be used to limit the search regions. Sets are only required for the ELPAT
method if the surfaces contain faces of continuum elements.
SHIDB SHIDB
SHIDA
SHIDA
Connected Elements for ELEMID Format Connected Elements for ELPAT Format
Figure 9-60 Comparison of Load Transfer in the ELEMID and the ELPAT Methods
The data needed to define a connection with PARTPAT are:

2. The sets SetA and SetB for both sides, where, for shell elements, these sets may be element sets,
and for continuum elements, they must be face sets containing element and face number pairs. The
two sets SetA and SetB do not have to be disjoint, but the search procedure is facilitated if they are.
CHAPTER 9 677
Boundary Conditions

The data needed to define a connection with ELPAT are:
2. The patches SHIDA and SHIDB on both sides, where for shell elements these are the element
numbers and for continuum elements these are element and face number pairs.
3. The sets SetA and SetB for both sides, where, for shell elements, these sets may be element sets
and for continuum elements they must be face sets containing element and face number pairs. The
two sets SetA and SetB do not have to be disjoint. For surfaces defined by shell elements, these
sets are optional, but for surfaces defined by faces of continuum elements, they must be specified.
Parameters for the Projection Process

When searching master or secondary patches, it may happen that no projection is found at all to any of
the candidates because the surface is not flat and the GS-node or an auxiliary node only has a projection
outside two neighboring patches as illustrated in the left part of Figure 9-61 or because one or more of
the auxiliary nodes lie outside a free edge as illustrated in the right part of Figure 9-61.
GS Free Edge
Figure 9-61 Exceptional Cases needing Special Consideration
In the first case, a projection is accepted to the common edge of the two patches and the nearest patch is
selected as the master or secondary patch. In the second case, two parameters may be used to improve
the situation. The GSMOVE parameter allows the GS-node to be moved in the direction of the auxiliary
nodes that have found successful projections. The NREDIA parameter allows the area of the auxiliary
patch to be reduced without moving the GS-node. Both parameters may be used simultaneously. In that
case, it is first tried to move the GS-node a number of times and if that is not successful the area is
reduced a number of times. If that still fails, it is once more tried to move the GS-node a number of times,
now using the reduced area for the auxiliary patch. Successful projections of the auxiliary nodes are
counted in pairs. A successful pair means that the two corresponding auxiliary nodes on both sides of the
connection have found a successful projection. So in the situation in the right part of Figure 9-61, two
pairs of auxiliary nodes are successful and two pairs have failed, because their projections fall over a free
edge of one or both of the surfaces. In addition, the PROJTOL parameter may be used to accept
projections outside but near patch boundaries. If GSMOVE is nonzero, the new GS location is determined
from the successfully projected GAHi and GBHi pairs as follows:
• If one pair of GAHi and GBHi finds a successful projection, GS is moved to the middle position
between that pair and the original GS.

• If two pairs of GAHi and GBHi find a successful projection, GS is moved to the middle position
between those two pairs and the original GS.
• If three pairs of GAHi and GBHi find a successful projection, GS is moved closer to the second
pair found. The move is one quarter of the distance from GS to the second pair.
• If no pairs are found at all, moving GS is abandoned.
If the number of moves has been exhausted and NREDIA is nonzero, the characteristic diameter of the
CWELD connector is halved a number of times reducing the auxiliary patch size, thus bringing its nodes
closer to each other. If one or more pairs still fail after this diameter reduction, a last attempt is made to
move the GS node while maintaining the reduced sized of the auxiliary patch. The size of this movement
is twice that of the movement before the size reduction which, in general, is still a small movement,
because of the reduced auxiliary patch size. This second attempt to move the GS-node is only carried out
when NREDIA is nonzero.
In some occasions, it may happen that a proper projection is found far away from the GS point (e.g.,
when the surface is curved and the situation displayed in the left part of Figure 9-61 occurs for patches
near to the GS point. In that case, the GSTOL parameter may be used to reject projections that are further
away than specified by the parameter, so projections will only be accepted that are relatively near to the
GS point.
It is assumed that the two surfaces that are being connected are reasonably parallel. The GSPROJ
parameter may be used to check this explicitly as illustrated in Figure 9-62. If the master patch normal
vectors on both sides make an angle larger than specified by the parameter, the projection is rejected.
It is also assumed that master patches and associated secondary patches are reasonably parallel. It is not
desirable that elements almost normal to the master patch are selected as secondary patches. The GSCURV
parameter may be used to reject secondary patches that make a large angle with the master patches as
illustrated in Figure 9-62. This may occur in sharp corners of the surface which in practice may represent
edges of the structure.
The parameters discussed in this section are defined on the SWLDPRM option.
GSCURV
GSPROJ
nB
GSCURV
nA nA H 1
nA
GAH1 GA GAH2
Figure 9-62 Angle Parameter GSPROJ and GSCURV to Control the Patch Selection Process
CHAPTER 9 679
Boundary Conditions
Internal Constraints to Connect Two Surfaces

Each connector end node, that is not making a point connection, makes a rigid connection to its master
or auxiliary patch by means of a Kirchhoff constraint (tying type 44). This constraint rigidly constrains
the displacements and rotations of the end node to the motion of the patch it connects to. The possible
patches to which such a connection can be made are linear and quadratic quadrilateral and linear and
quadratic triangular patches. With the ELPAT and PARTPAT methods, the auxiliary patch nodes are
connected to their partnering secondary patches by additional constraints. By default, these will be RBE3
constraints, but alternatively Kirchhoff constraints may be used or a combination of RBE2 and RBE3
constraints. The RBE2/RBE3 combination makes a RBE2 constraint between the auxiliary patch node
and its projection point on the secondary patch and RBE3 constraints between this projection point and
the nodes of this patch.
When a surface is curved, the auxiliary patch nodes will in general not lie on the surface. A special
parameter may be set in the CWELD input to force a repositioning of these nodes onto the surface. This
slightly alters the orientation of the auxiliary patch and in some cases this might allow the connector to
follow the curvature of the surface somewhat better. By default, this repositioning is not done.
For the RBE3 constraints involved in the CWELD connection, a weighting other than uniform weighting
can be used by specifying a distance weighting exponent. The weight factor for each retained node in the
RBE3 is:
1
f i = -----n-
di
where f i is the weighting factor for retained node i, d i is the distance from the tied node to retained node
i and n is the weighting exponent. Negative values for n are not recommended, since they will result in
heavier weighting for nodes further away. The default results in uniform weighting ( f i = 1 ) .
The RBE2/RBE3 combination is only useful when the auxiliary nodes are not lying on the secondary
patches and when a nonuniform distance weighting is used. It also comes with a cost penalty because it
involves more constraint equations in each connection.
If the CONNECTIVITY input contains no elements with nodes having displacements and rotations as
degrees of freedom, it is necessary to set the RBE parameter.
Multiple Surface Connections

The patch-to-patch connection can be used to connect more than two layers of shell elements by using
the same GS-node in multiple CWELD definitions. For example, if three layers need to be connected, a
second CWELD definition is made, that refers to the same GS-node as the first CWELD definition.
Surface B on the first CWELD definition is repeated as surface A in the second CWELD definition. If the
surfaces are not curved, the locations of the connector end nodes connecting from the top and bottom
side to the surface in the middle automatically coincide, but for curved surfaces, this is not guaranteed.
For all connection types except PARTPAT, there is no real difference between repeating the GS-node or
defining it a multiple number of times.
When the PARTPAT method is used and a set pair is repeated in a number of connections that use the
same GS-node (e.g., the same set is used for side A and side B for each of these connections), this
repeated usage of the GS-node results in searching through a stack of patches that can be found from the
two sets, in order to connect multiple plate surfaces. However, when the set pairs are different, i.e., at
least one side in a new connection refers to a set with a name different from the name referred to by this
side in a previous connection using the same GS-node, the effect is as in all other connection types.
The definition for a search through a stack of surfaces can only be made by repeated usage of a GS-node
(i.e., a repetition of its node ID, not just a repetition of its coordinates XS, YS, and ZS) in combination
with the same sets for side A and side B for all the connections that must be present in the stack. It only
applies to the PARTPAT connection type and in that case the sets necessarily have an overlap, i.e., they
are not disjoint. The stack is limited to 10 connections, i.e., 11 plate surfaces.
Coupled Thermo-mechanical Analysis

Cweld connections are be used in a coupled thermo-mechanical analysis. In the heat transfer pass of an
increment, the Kirchhoff constraints will tie the temperature of the tied node to the temperature of the
projection point on its corresponding patch. For RBE3 constraints, the temperature of the tied node will
be the least squared weighted average of the temperatures of its retained nodes. The same weighting
factors apply as in the mechanical pass. If the retained nodes are the nodes of a shell element, all
temperature degrees of freedom of the shell node are used in the weighting process. For RBE2
constraints, the temperature of the tied node is set equal to the temperature of its corresponding retained
node, so the constraint in the heat pass is identical to typing type 1.
Rules for usage of the GS-node or the GA-, GB-node Pair in

the Projection Process
In most cases, the GS-node will be used to find the projections determining the end nodes of the
connector element. The GS-node and its XS,YS,ZS-coordinates are ignored on a CWELD definition if
both the GA and the GB end nodes of the connector are specified. In that case, GA and GB should, at
least, be approximately in the right locations and GA is projected to side A and GB is projected to side
B of the connection. If GS and GA are not specified, but GB is, then GB is used as GS and if GS and GB
are not specified, but GA is, then GA is used as GS. If GS, GA, and GB are not specified, then the
XS,YS,ZS coordinates must be specified. If GS is specified and maximally one of the GA or GB, then
GS is used for the projections on both sides and the GA or GB is ignored for this purpose. If one side of
the connector makes a point connection and the connector node for that side has been left blank and the
GS-node has been specified, the point connection will be made to the GS-node and the GS-node will be
projected onto the opposite side to determine the location of the other connector node. If the opposite side
also makes a point connection, the node for that side must have been specified.
CWELD Point-to-patch Connections

It is possible to connect one end of the connector element to the surface by a point connection as is shown
in Figure 9-63. This may even be a free node to which certain boundary conditions may be applied. In
such cases the surface information that identifies the master patch for the particular side is omitted. This
means that for the ELEMID and ELPAT method, the input for SHIDA or SHIDB is omitted, for the
GRIDID method, the nodes GAi or GBi are omitted, and for the PARTPAT method, the set SetA or SetB
CHAPTER 9 681
Boundary Conditions
is omitted. The GS-node is projected to the surface on the other side according to the specifications made
in the input.
n
GS
GA4 GA3
GA
GA1
GA2
Figure 9-63 Point-to-Patch Connection on Side A using GRIDID
CWELD Point-to-point Connections

All methods may be used to define point-to-point connections by omitting the surface information that
identifies the master patches and by specifying the connector end nodes on the CWELD input. These must
be existing nodes in the structure, because they can no longer be generated internally. Using any of the
patch connection methods would lead to cumbersome definitions of such point-to-point connections and
the ALIGN method was designed to simplify their definition. It is shown in Figure 9-64. The GS-node
is no longer required, because the connector end nodes in general will be nodes belonging to the two
surfaces. Point-to-point connections are only recommended for nearly congruent mesh, so the connector
element will be approximately normal to the surfaces. However, they are not recommended when the
connector cross sectional area exceeds 20% of the characteristic element face area.
n
GA Upper Shell Midsurface
GB
Lower Shell Midsurface
Figure 9-64 Point-to-point Connection using ALIGN
CWELD Properties
PWELD Property Definition
The properties for the connector element can be entered on the PWELD model definition option. The
material properties are defined by entering the material identification number of a valid material for the
element. The default cross section is a solid circular section for which the diameter can be entered on the
PWELD input. If other geometric properties than those of a circular section are required, a complete set
of geometric input data can be entered here as well. In that case, the diameter is only used to determine
the size of the auxiliary patches. If no diameter is entered but a set of geometric data is, an attempt is
made to estimate the auxiliary patch size from the geometric data. The geometric data input also specifies
the local directions of the cross section. If their corresponding data values are all zero, the local directions
are determined by the procedure outlined in the Connector Orientation section in this chapter, otherwise
the local directions are determined from the geometric data input according to the usual Marc convention.
The default length of the connector is given by the distance between the two end points GA and GB. Two
control values on this length can be entered to prevent overly stiff or overly flexible behavior of the
connector element. Additionally, a spot weld connection type can be selected as an alternative to the
default general weld connection type. In both cases, the effective stiffness of the element is changed by
making certain length modifications. Their details are discussed later in the Connector Length
Modifications and Spot Connections section.
Connector Orientation
The connector element is positioned between the two projections GA and GB and these points determine
the end nodes of the element and thus its direction. The section forces for the default beam type are
displayed in Figure 9-65 and the local section directions are determined as follows. The element z-axis
points from GA to GB.
The first procedure outlined in this section is used when the 4th, 5th, and 6th fields of the appropriate data
blocks in the PWELD or GEOMETRY options are all zero and no coordinate system (MCID) is specified
in the CWELD option. This procedure is as follows:
xB – xA
e 3 = ------------------------- is the local z-axis
xB – xA
In case of zero length, the average normal of the two surfaces is taken. Find the index j of smallest
component of e 3 .
j i
e 3 = min { e 3 } i = 1, 2, 3
CHAPTER 9 683
Boundary Conditions
In case of two equal components, the smaller index i is taken. Define the corresponding base vector
as follows:
δ1 j 0
bj = δ 2 j , where δ k j = 1 when k = j and 0 otherwise, e.g. for j = 3 , b 3 = 0
δ3 j 1
mz
fz
ze my B
fy
GB fz axial force
fx shear force plane 1
fx fy shear force plane 2
mx B mz torque
mx A
plane 2 m x A bending moment end A, plane 2
fx m x B bending moment end B, plane 2
GA ye m y B bending moment end A, plane 1
plane 1 m y A bending moment end B, plane 1
fy
xe
my A
fz
mz
Figure 9-65 Connector Orientation, Local Cross-section Directions and Forces
It provides a good directional choice for e 1 . In addition, the vector e 1 must be orthogonal to e 3 .
ẽ 1
ẽ 1 = b j – ( e 3 ⋅ b j )e 3 and ẽ 1 = ----------- is the local x-axis
ẽ 1
The cross product of local z and local x provides the local y-axis
e2 = e3 × e1
The second procedure outlined in this section is used when the 4th, 5th, and 6th fields of the appropriate
data blocks in the PWELD or GEOMETRY options are all zero, but a coordinate system (MCID) is
specified in the CWELD option. In that case the local directions are determined by a procedure that uses
a coordinate system defined in the COORD SYSTEM option. This procedure is as follows:
xB – xA
e 3 = ------------------------- is again the local z-axis.
xB – xA
In case of zero length, the average normal of the two surfaces is taken. The 2nd direction (T2) defined by
MCID is used to define the orientation vector v of the connector element.
The local y-axis is then defined as

e3 × v
e 2 = ---------------------
e3 × v
The cross product of local y and local z provides the local x-axis
e1 = e2 × e3
If any of the 4th, 5th, or 6th fields of the appropriate data blocks in the PWELD or GEOMETRY options
is nonzero, the conventional Marc procedure is used to determine the local element directions,
irrespective of the presence of a MCID input in the CWELD option. A coordinate system may be associated
with the first beam node (GA), but offsets are not possible and are ignored.
An angle α may be specified for each connector element, which rotates the local axes about the line
connecting the two end points. This rotation takes effect after the orientation has been found by either
one of the procedures outlined above or by the geometric specifications made in the PWELD or
GEOMETRY options.
Connector Length Modifications and Spot Connections

In the default general connection, the distance between the points GA and GB determines the length
of the connector element. To prevent the connector element from overly stiff or overly flexible behavior
two tolerance values, LDMIN and LDMAX can be entered on the PWELD input. LDMIN and LDMAX
are the minimum and maximum length to diameter ratio of the connection. If the length to diameter ratio
is less than LDMIN, the behavior of the connection may become overly stiff and if it is more than LDMAX,
it may become overly flexible. In either case, a length modification of the CWELD is made such that
its behavior remains within the two limits. There are three methods available to account for this
length modification.
The first method is stiffness scaling. For linear elastic material behavior, the length modification could
be accounted for by scaling the Young’s modulus such that the axial, shear, and torsional stiffness of the
beam with its actual length (i.e., the length determined by the distance between its end nodes) is identical
to the axial, shear, and torsional stiffness of a beam with the modified length using the original Young’s
modulus. This scaling can only be done when the beam material behavior is linear elastic and the cross-
section properties are not entered through the BEAM SECT parameter. Internally, the cross-section
geometric properties of the beam are scaled and not the Young’s modulus.
The second method is repositioning of the end nodes of the beam. Each beam node is moved in a direction
normal to its patch such that the beam obtains the desired length. The distance to the patch is treated
CHAPTER 9 685
Boundary Conditions
as an offset in the Kirchhoff constraints. If there is a point connection on one side, then that node is
not repositioned.
The third method is repositioning of the end nodes and the auxiliary nodes. For methods ELPAT and
PARTPAT, the auxiliary nodes can be repositioned together with the beam nodes such that the beam
obtains the desired length. This requires updating the projections of the auxiliary nodes which may fail
if these nodes become located far away from their respective surfaces and these surfaces are curved.
In that case, they may no longer have a projection inside any of the element face boundaries.
If the element and its material behavior allow it, the default method is the first method, stiffness scaling,
otherwise, no length modification is made. The length modification method can be activated by defining
the CWSPOT parameter on the SWLDPRM option. If a CWELD has a point connection on both sides
(like in ALIGN) or the distance between the two end nodes of the beam is zero, no length modification
will be made.
Alternatively, a spot weld connection can be defined. In this case, the length of the connecting beam is
modified to be 0.5∗ ( t A + t B ) where t A is the patch shell thickness on side A and t B is the patch shell
thickness on side B. If the surface on one side is defined by the faces of a solid element, its thickness
contribution is zero. If both sides are faces of solid elements, the connection type is automatically reset
to a general connection.
The type of connection, general connection or spot connection, can be selected on the PWELD input.
Spot connections are only possible when one of the methods ELEMID, ELPAT, or PARTPAT is used.
If the distance between the connector end nodes GA and GB is zero, the two nodes will be connected
rigidly using tying type 100 and no beam element is activate in the connection.
If stiffness scaling is requested as the length modification method and the material is nonlinear or the
cross section properties are entered through the BEAM SECT parameter, the scaling is not done and a
warning message is written to the output.
CWELD Input Styles

For CWELD connections, two different input styles can be used to define the connections and these styles
may be mixed in one model.
The first style is the most natural style, where only the GS-node and the two surfaces are defined in the
CWELD input. This style was used in the foregoing discussion. The beam connector elements and its
nodes are generated internally and are not present in the CONNECTIVITY and COORDINATES inputs.
All necessary geometric properties of the connection are defined in the PWELD input and its material
properties are defined by entering a valid material identification number in the PWELD input.
With the second style, the connector elements are entered explicitly in the CONNECTIVITY option and
their nodes in the COORDINATES option. The connector nodes do not have to be in the right location
as long as the GS-node is in the approximate location. The material properties of the connectors may
then be entered through the usual material options and their cross-section properties through the
GEOMETRY option. All these properties are assigned to the elements directly. The CWELD option still
defines the connection, but only refers to the connector elements and a PWELD input is generally no
longer required. The second style requires that the CWELD input is made before the CONNECTIVITY
input of the element.
However, it is not required to rigidly adhere to the input styles mentioned above. A connector element
with its ID, type, and nodes may be defined entirely on the CWELD option. If no element ID is specified,
Marc generates it internally. If no node IDs are specified, Marc generates them internally. It no type is
specified, type 98 is chosen for a 3-D analysis and type 5 for a 2-D analysis. If the CONNECTIVITY for
an element is entered after its CWELD input, the CONNECTIVITY input determines the element type and
the nodes of the element. If the CWELD input for an element is entered after its CONNECTIVITY input,
the CWELD input determines the element type and nodes of the element. If the CWELD input specifies
the element ID, a CONNECTIVITY input for the element is not required. It is optional and can be used to
redefine the nodes if these were left undefined in the CWELD input.
If the connector element is only defined through the CWELD input, it is necessary to enter an ELEMENTS
parameter for the type of the connector element as it is necessary to enter an ELEMENTS parameter for
each element type that is used in the CONNECTIVITY input.
If a side of a connection makes a point connection, the node to which this connection is made cannot be
an automatically generated node. It must be a node defined by the user in the input. The user has no
control over automatically generated nodes; i.e., how their ID’s are assigned to them. Therefore, it is not
possible to load or constrain such nodes or to connect them to other parts of the model.
CWELD Error and Warning Messages

A number of error and warning messages can be written to the output file for any of the connector end
nodes or the auxiliary nodes. Error messages obtain a positive code and terminate the job with exit
number 13. Warning messages obtain a negative code and allow the job to continue. The possible
messages are listed in Table 9-27.
Table 9-27 Codes for CWELD Error and Warning Messages
Value Description
1 Projection algorithm did not converge within desired tolerance.
2 Projection converged, but not inside the patch boundary.
-2 Projection converged, but near the patch boundary (inside: 0.01, outside: PROJTOL).
3 The patch is severely distorted or inside out (normal vectors in different locations make
larger than 90° angles).
-3 The patch is OK, but still has much distortion (normals in different locations make
larger than 10° angles).
4 The angle of an element/face normal to its auxiliary or master patch normal is too large;
the patches are almost normal to each other.
-4 The angle of an element/face normal to its auxiliary or master patch normal is
rather large.
5 GS projects OK, but is located too far away from the projection.
CHAPTER 9 687
Boundary Conditions
Table 9-27 Codes for CWELD Error and Warning Messages (continued)
Value Description
-5 GS or an auxiliary node projects OK, but is located far away from the projection.
For GS nodes we use GSTOL as the tolerance.
For auxiliary nodes, we use XDIM (characteristic cweld dimension) as the tolerance.
XDIM = 1 ⁄ 4πD , where D is characteristic cweld diameter.
6 The element we want to project to is unknown (i.e., useless, because element type = 0).
-6 A projecting is being made to a deactivated element, but accepted anyway.
7 The beam is almost parallel to the patch. (its angle with the patch is less than GSPROJ)
-7 The beam is not very normal to the patch (it makes an angle larger than 2*GSPROJ to
the patch normal).
8 The patches to which the auxiliary nodes tie are more than 3 x 3.
9 The patch nodes have invalid degrees of freedom.
10 The patch has an invalid number of retained nodes.
11 A search over a group of patches resulted in no success at all; no projection has been
even attempted to any of the patches.
12 A search over a group of patches resulted in no success at all; although projections were
tried, they were all unsuccessful.
13 The projection is outside the patch boundary and there are no neighbors; so, there is no
positioning to a common edge or a common corner.
-14/14 Warning/Error: Position the cweld node to corner 1 of patch.
-18/18 Warning/Error: Position the cweld node to edge 1-2 of patch.
-20/20 Warning/Error: Position the cweld node to edge 3-4 (quad) or 3-1 (tria) of patch.
22 No face id was specified for the continuum element.
24 No viable patches in the set (PARTPAT, ELPAT).
26 No viable elements in the whole model (PARTPAT, ELPAT).
-27 To account for a length modification, the beam node on this side of the patch has been
offset with regard to the patch it is connecting with.
28 The beam node is shared with other elements, while not making a point connection.
Or the GS-node is shared with elements on the side that makes a point connection.
-28 The GS-node is shared with other elements.
Table 9-27 Codes for CWELD Error and Warning Messages (continued)
Value Description
29 The face ID identifying a patch is missing for a continuum element.
30 The connector node cannot connect to this element because it is not a valid
connection type.
32 In 3-D, a connection to a line type of element was attempted; this is not possible.
-33/33 The connector node made a point connection but is not connected to any other element
than the connector element. If this node is an automatically generated node, the result is
an error; otherwise, a warning.
34 The projection is to an element/face sharing the point connection node.
35 An element/face of another connection lies between the current element/face pair.
CWELD Output
A CWELD connection defines a beam element as the connector element and generates internal tyings to
define the connector constraints. Output to the .t16 or .t19 post file can be requested for the beam
elements by using the standard element post codes available for them. Tying forces can be requested by
using the standard nodal post codes available for them. The displacements of all nodes involved in the
connection will by default be on the post file.
Four sets are automatically generated and available on the post file when the CWSETS parameter on the
SWLDPRM option is set to one. The first set contains all the beam elements used in the connections. The
second and third sets contain all patches defining side A and side B of all connections, respectively. The
fourth set contains all elements that have caused warnings. An error set is not available because, upon
encountering an error, no attempt is made to complete a connection. This means that, in case of an error,
the information at best is incomplete, but because of its intermediate character it may often be misleading
and often it is not yet available at all.
The sets are called:
Set 1: fastener_all_beams_inc0000
Set 2: fastener_all_faces_sidea_inc0000
Set 3: fastener_all_faces_sideb_inc0000
Set 4: fastener_all_warnings_inc0000
Set definitions defining sets with any of these names should not be made in the input.
The PRTSW parameter in the SWLDPRM input can be used to control the amount of output of each
CWELD connection. If PRTSW = 0 or is left undefined, no diagnostic messages are printed to the
output file. If PRTSW = 1, only errors are printed. If PRTSW = 2, errors and warnings are printed. if
PRTSW = 3, all projection diagnostics are printed for each CWELD connection. If PRTSW = 4, all
projection diagnostics and all specifics of the tying constraints involved in each connection are printed.
CHAPTER 9 689
Boundary Conditions
Two Dimensional Models

The previous sections described the CWELD connections for three dimensional models. However,
CWELD connections can also be made in two dimensional models. In that case, the patches consist of
beams, trusses, or edges of continuum elements. The same connection methods that are available for
three dimensional models can be used in two dimensional models. The connector element must be a two
dimensional beam or an axisymmetric shell depending on the type of two dimensional analysis. The
number of auxiliary nodes is two instead of four; one to the left and one to right of the projection point
on the master patch. Two patch types are available: linear (with two nodes) and quadratic (with three
nodes) patches. The 3 x 3 requirement only applies in one direction now, so a connection on one side can
not span more than three patches of the surface.
CFAST Connections
CFAST connections can only be defined using indirect patch-to-patch connections. Two methods,
ELEM or PROP, are available to create them. They are equivalent with the CWELD ELPAT and PARTPAT
methods, respectively (see Figures 9-57 and 9-58). In each of these methods, the surfaces on both sides
have to be defined together with the GS reference point of the connection. Both methods can be used to
connect noncongruent meshes of any type. The constraints between the connector nodes and the
auxiliary patches are applied through Kirchhoff constraints (typing type 44) and the constraints between
the auxiliary nodes and the secondary patches are applied through RBE3 constraints.
The connector element is a cbush element. In a two dimensional model, it is element type 194 and in a
three dimensional model, it is element type 195. For more details about the cbush elements, refer to Marc
Volume B: Element Library. It is always required to include the ELEMENTS parameter for each element
type used as a connector even when it does not appear in the CONNECTIVITY input. The data for each
connection is defined on the CFAST model definition input. Some data inputs are always required
because no defaults exist for them. Other data inputs are optional and if not defined, Marc assumes
appropriate defaults values for them. These optional data inputs concern the connector element type, the
way the projections are carried out, the type of constraints, and the orientations of the specific CFAST
connections. For a more detailed description of the CFAST option, refer to Marc Volume C: Program
Input.
CFAST Properties
PFAST Property Definition
The properties of the CFAST connection are defined using the PFAST model definition option. It
contains the diameter of the fastener, the stiffness matrix, damping coefficient, and mass. The diameter
data is only used to compute the locations of the auxiliary nodes involved in the CFAST connection. It
will not affect the element stiffness as the latter has to be explicitly entered. For a more detailed
description of the PFAST option, refer to Marc Volume C: Program Input.
CFAST Element Coordinate System

The stiffness and damping values are given is a specified coordinate system ( e 1 ,e 2 ,e 3 ) . This coordinate
system is used as the element coordinate system. The output of CFAST is given in this system. There are
two options to define it:
1. Using the connector element orientation. The x-axis points from GA to GB.
xB – xA
e 1 = -------------------------
xB – xA
In case of zero length, the normal of the first patch is taken. Find the index of j of smallest
component of e 1
j i
e 1 = min { e 1 }
i = 1 ,2 ,3
In case of two equal components, the smaller index i is taken. Define an auxiliary vector:
σ1 j
b = σ 2 j , where σ k j = 1 when k = j and 0 otherwise (for example, for j = 3 ,
σ3 j
0
b = 0 ) e 3 is created by the normalized cross product of e 1 and b . And, finally, e 2 is the
1
normalized cross product of e 3 and e 1 . For geometrically nonlinear analysis, this coordinate
system co-rotates with the element.
u u u
2. Using a global or user-defined coordinate system ( e 1 ,e 2 ,e 3 ) . For this option, you may choose
between a relative and an absolute orientation.
a. the absolute orientation:
u u u
( e 1 ,e 2 ,e 3 ) = ( e 1 ,e 2 ,e 3 )
For a geometrically non-linear analysis, this coordinate system is fixed in time.

b. The relative orientation: the x-axis points from GA to GB.
xB – xA
e 1 = -------------------------
xB – xA
In case of zero length, the normal of the first patch is taken. e 3 is defined by the normalized
u
cross product of e 1 and e 2 . e 2 is defined by the normalized cross product of e 3 and e 1 . For
a geometrically nonlinear analysis, this coordinate system co-rotates with the element.
CHAPTER 9 691
Boundary Conditions
SWLDPRM, Special CWELD/CFAST Parameters

A number of global parameters that control the behavior of CWELD and CFAST connections and their
output to the jobid.out file can be entered through the SWLDPRM model definition option. These
parameters and their descriptions are summarized in Table 7-28.
Table 7-28 SWLDPRM Parameter Names and Descriptions

Name Type Default Description
CHKRUN Integer > 0 0 This parameter is available in MD Nastran but has no
(0 or 1) meaning in Marc and is ignored.
GSMOVE Integer > 0 0 Maximum number of times GS is moved in case a
complete projection of all points has not been found.
NREDIA 0 < Integer < 4 0 Maximum number of times the characteristic diameter
D is reduced in half in case a complete projection of all
points has not been found.
PRTSW 0 < Integer < 4 0 Parameter to control the CWELD/CFAST diagnostic
output to the Marc output file (jobid.out).
0 = no diagnostic output
1 = print errors only
2 = print errors and warnings only
3 = print projection diagnostics with no tying details
4 = print all diagnostics
GSPROJ -90 < Real < 90 20.0 Maximum angle allowed between the normal vectors of
master patch A and master patch B. The connection will
not be generated if the angle between these two normal
vectors is greater than the value of GSPROJ. If GSPROJ
is negative, the program will always accept the
connection and will only issue a warning if the angle is
larger than |GSPROJ| (see Figure 9-62).
Table 7-28 SWLDPRM Parameter Names and Descriptions (continued)

GSCURV -90 < Real < 90 20.0 Maximum angle allowed between the normal vectors of
a patch to which an auxiliary node projects and its
corresponding auxiliary and master patches. It provides
a measure to monitor the curvature of a surface and to
recognize patches that belong to, for example, stiffeners.
A connection is not generated if the angle between the
normal vectors is greater than 90-GSCURV meaning
that the patches are almost normal to each other. In that
case, the patch is rejected and the search proceeds to the
next patch in the list. If the angle is between zero and
GSCURV, no message is displayed. If the angle is between
GSCURV and 90-GSCURV, a large angle warning is
displayed. The following three tests are performed in the
order given below when GSCURV is positive:
If 0 < angle < GSCURV => OK
If GSCURVE < angle < 90-GSCURV => trigger a warning.
If angle > 90-GSCURV => reject.
Note that the warning condition is never triggered when
GSCURV > 45 as it is overruled by the reject condition.
If GSCURV is negative, the projection is always accepted
and a warning is issued when the angle is larger than
|GSCURV| (see Figure 9-62)
GSTOL Real 0.0 Maximum allowable distance of the node GS to its
projection on a patch. IF GSTOL is positive, the distance
is relative to the characteristic CWELD/CFAST diameter
D, (the tolerance is GSTOL*D). If GSTOL is negative, the
distance is absolute (i.e., the tolerance is -GSTOL). If GS
is used for the projection together with one of the
methods PARTPAT/PROP or ELPAT/ELEM, an error is
issued if the distance is too large. If GA and GB are used
for the projection or if one of the ELEMID or GRIDID
methods is used, the test only issues a warning if the
distance is too large. If GSTOL is zero, any distance
is accepted.
PROJTOL 0.0 < Real < 0.2 0.0 Tolerance to accept the projected point GA or GB if the
computed coordinates of the projection point lie outside
the patch boundary, but are located within
PROJTOL*(dimension of the patch).
CHAPTER 9 693
Boundary Conditions

ACTVTOL Integer > 0 1111 Parameter controlling the behavior of PROJTOL for the
different CWELD/CFAST connection methods. This
Integer < 2211
parameter is entered as an integer and is converted to a
four-character string. If its value is less than 1000, the
string is prepended with zeros. The first character (from
the left) controls the behavior when the PARTPAT/PROP
method is used. The second controls the behavior when
the ELPAT/ELEM method is used. The third controls the
behavior when the ELEMID method is used and the
fourth controls the behavior when the GRIDID method is
used. For ALIGN, the PROJTOL tolerance has no
significance. Each digit ( d i ) in the string can have the
value 0 or 1 or 2, where the value 2 only has significance
for the ELPAT/ELEM or PARTPAT/PROP methods. The
values have the following meaning:
0 = PROJTOL is completely deactivated
1 = PROJTOL is activated for ELEMID and GRIDID,
PROJTOL is activated in initial projections for
ELPAT/ELEM, PROJTOL is only activated over free
edges of the patch in auxiliary projections for
ELPAT/ELEM, and in initial and auxiliary
projections for PARTPAT/PROP. Free edges have
no neighbors within the set that defines the
complete surface.
2 = PROJTOL is always activated

CWSETS Integer > 0 0 Parameter to control the automatic creation of four
element sets with the elements involved in the
(0 or 1)
CWELD/CFAST connections.
0 = the sets are not created
1 = four sets are created automatically:
fastener_all_beams_inc000”, the set
containing all connector beam elements.
fastener_all_faces_sidea_inc0000,
the set containing all elements with patches on side
A of the connection.
fastener_all_faces_sideb_inc0000,
the set containing all elements with patches on side
B of the connection.
fastener_all_warnings_inc0000, the set
containing all elements involved in CWELD/CFAST
warning messages.
Defining sets with any of these names must be avoided
and are considered an error.
MAXEXP Integer > 0 2 Parameter to control the maximum number of
expansions in the search for projections of the auxiliary
nodes. First, the master patch is tried. If no projection is
found on the master patch, a first expansion is made
including all neighboring patches of the master patch. If
no projection is found on any of the new patches, a
second expansion is made including all neighbors of the
patches tried so far. This process continues until the
number of expansions exceeds MAXEXP. Two patches are
neighbors if they share at least one node in their
connectivities.
DLDMIN Real > 0.0 0.2 Default value for LDMIN; the smallest ratio of length to
characteristic diameter.
DLDMAX Real > 0.0 5.0 Default value for LDMAX; the largest ratio of length to
characteristic diameter.
MAXITR Integer > 0 25 The maximum number of iterations allowed in the
iteration process for finding the projection on a patch.
EPSITR Real > 0.0 1.0E-5 Tolerance to terminate the iteration process for finding
the projection on a patch. If the parametric coordinate
change in an iteration is less than EPSITR, the projection
is accepted as converged.
CHAPTER 9 695
Boundary Conditions

DELMAX Real > 0.0 0.1 Maximum allowable parametric coordinate change
during the iteration process for finding the projection on
a patch. At first DELMAX is not activated (i.e., the
parametric coordinate change is not limited during the
iteration process). The parameter is only activated when
the full Newton Raphson iteration process for a
projection did not converge. In that case, the iteration
process is restarted with DELMAX activated.
CWSPOT 0 < Integer < 3 1 Parameter to choose the method for modifying the
beam length.
1 = scale the stiffness of the beam
2 = reposition the end nodes of the beam
3 = reposition the auxiliary patch nodes and the end
nodes of the beam.
RBE3WT Real 0.0 Default RBE3 distance weighting exponent.
The weight factor for each retained node in a RBE3
1
involved in a CWELD/CFAST connection is: f i = -----
n
di
where
fi is the weighting factor for retained node i.
d i is the distance from the tied node to retained node i
n is the weighting exponent RBE3WT.
Negative values for RBE3WT are not recommended,
since they result in heavier weighting for nodes further
away. The default results in uniform weighting
( fi ) = 1 .
BOXING -1 < Integer < 1 0 Parameter to control the boxing algorithm used to speed
up the search for master patches when connection
method PARTPAT/PROP is used.
-1 =The boxing algorithm is always deactivated
0 = The boxing algorithm may or may not be activated
depending on the number of elements in the sets.
1 = The boxing algorithm is always activated
Chapter 10 Element Library
10 Element Library

Truss Elements 700
 Membrane Elements 700

Continuum Elements 700

Beam Elements 701
 Plate Elements 702

Shell Elements 702

Heat Transfer Elements 702
 Electromagnetic Analysis 704

Soil Analysis 704

Fluid Analysis 704
 Piezoelectric Analysis 704

Special Elements 705

Incompressible Elements 706
Marc includes an extensive element library. The element library allows you to model various types of
one-, two-, and three-dimensional structures, such as plane stress and plane strain structures,
axisymmetric structures, full three-dimensional solid structures, and shell-type structures. (See Marc
Volume B: Element Library for a detailed description of each element, a reference to the use of each
element, and recommendations concerning the selection of element types for analysis).
After you select an element type or a combination of several element types for your analysis with the
ELEMENTS or SIZING parameter, you must prepare the necessary input data for the element(s). In
general, the data consist of element connectivity, thickness for two-dimensional beam, plate and shell
elements, cross section for three-dimensional beam elements, coordinates of nodal points, and face
identifications for distributed loadings.
You can choose different element types to represent various parts of the structure in an analysis. If there
is an incompatibility between the nodal degrees of freedom of the elements, you have to provide
appropriate tying constraints to ensure the compatibility of the displacement field in the structure. Marc
assists you by providing many standard tying constraint options, but you are responsible for the
consistency of your analysis.
You can use almost all of the Marc elements for both linear and nonlinear analyses, with the
following exceptions.
• No plasticity or cracking is allowed in the elastic beams, the shear-panel, and
the Fourier elements.
• The updated Lagrange and finite strain plasticity features are not available for all elements.
• Plasticity is available for the Herrmann elements when using the FeFp formulation only.
• Only heat links and two-dimensional heat transfer elements can be used for hydrodynamic
bearing analysis.
• Fourier analysis can be carried out only for a limited number of axisymmetric elements.
Marc defines all continuum elements in the global coordinate system. Truss, beam, plate, and shell
elements are defined in a local coordinate system and the resulting output must be interpreted
accordingly. You should give special attention to the use of these elements if the material properties have
preferred orientations. The ORIENTATION option is available to define the preferred directions.
All Marc elements are numerically integrated. Element quantities, such as stresses, strains, and
temperatures, are calculated at each integration point of the element if you use the ALL POINTS
parameter. This is the default in Marc. These quantities are computed only at the centroid of the element,
if the CENTROID parameter is used.
Distributed loads can be applied along element edges, over element surfaces, or in the volume of the
element. Marc automatically evaluates consistent nodal forces through numerical integration. (See
Marc Volume B: Element Library for details on this process). Concentrated forces must be applied at the
nodal points.
All plate and shell elements can be used in a composite analysis. You can have a variable thickness shell,
and control the thickness and material property and orientation for each layer. For the thick shell elements
(types 22, 45, 75, or 140), the interlaminar shear can also be calculated.
CHAPTER 10 699
Element Library
Five concepts differentiate the various elements types. These concepts are listed invalid for pure heat
transfer. below.
1. The type of geometric domain that the element is modeling. These geometric domains are:
Truss
Membrane
Beam
Plate
Shell
Plane stress
Plane strain
Generalized plane strain
Axisymmetric
Three-dimensional solid
Special
2. The type of interpolation (shape) functions used in the element. These functions are:
Linear
Quadratic
Cubic
Hermitian
Special
The interpolation function is used to describe the displacement at an arbitrary point in the body.
u i ( x ) = N i ( x )u i (10-1)
where u ( x ) is the displacement at x, N are the interpolation (shape) functions, and u are the
generalized nodal displacements.
Engineering strain is
∂u i ( x ) ∂u j ( x )
ε i j ( x ) = 1 ⁄ 2  ---------------- + ---------------- (10-2)
 ∂x j ∂x i 
Therefore, the computational evaluation is
∂N i u i ∂N j u j
ε i j ( x ) = 1 ⁄ 2  -------------- + -------------- = β i j u i (10-3)
 ∂x j ∂x i 
Hence,
∂N i
β i j = --------- (10-4)
∂x j
3. The number of nodes in a particular element.
4. The number of degrees of freedom associated with each node, and the type of degrees of freedom.
5. The integration method used to evaluate the stiffness matrix. Marc contains elements which use
full integration and reduced integration.
Truss Elements
Marc contains 2- and 3-node isoparametric truss elements that can be used in three dimensions. These
elements have only displacement degrees of freedom.
Since truss elements have no shear resistance, you must ensure that there are no rigid body modes in
the system.
Membrane Elements
Marc contains 3-, 4-, 6-, and 8-node isoparametric membrane elements that can be used in three
dimensions. These elements have only displacement degrees of freedom.
Since membrane elements have no bending resistance, you must ensure that there are no rigid body
modes in the system.
Membrane elements are often used in conjunction with beam or truss elements.
Continuum Elements
Marc contains continuum elements that can be used to model plane stress, plane strain, generalized plane
strain, axisymmetric and three-dimensional solids. These elements have only displacement degrees of
freedom. As a result, solid elements are not efficient for modeling thin structures dominated by bending.
Either beam or shell elements should be used in these cases.
The solid elements that are available in Marc have either linear or quadratic interpolation functions.
They include
• 3-, 4-, 6-, and 8-node plane stress elements
• 3-, 4-, 6-, and 8-node plane strain elements
• 6 (4 plus 2)- and 10 (8 plus 2)-node generalized plane strain elements
• 3-, 4-, 6-, and 8-node axisymmetric ring elements
• 8-, 10-, and 20-node brick elements
• 4- and 10-node tetrahedron
• 6-node and 15-node pentahedral
• 12- and 27-node semi-infinite brick elements
In general, the elements in Marc use a full-integration procedure. Some elements use reduced integration.
The lower-order reduced integration elements include an hourglass stabilization procedure to eliminate
the singular modes.
Continuum elements are widely used for thermal stress analysis. For each of these elements, there is a
corresponding element available for heat transfer analysis in Marc. As a result, you can use the same
mesh for the heat transfer and thermal stress analyses.
Marc has no singular element for fracture mechanics analysis. However, the simulation of stress
singularities can be accomplished by moving the midside nodes of 8-node quadrilateral and 20-node
brick elements to quarter-point locations near the crack tip. Many fracture mechanics analyses have used
this quarter-point technique successfully.
CHAPTER 10 701
Element Library
The 4- and 8-node quadrilateral elements can be degenerated into triangles, and the 8-and 20-node solid
brick elements can be degenerated into wedges and tetrahedra by collapsing the appropriate corner and
midside nodes. The number of nodes per element is not reduced for degenerated elements. The same
node number is used repeatedly for collapsed sides or faces. When degenerating incompressible
elements, exercise caution to ensure that a proper number of Lagrange multipliers remain. You are
advised to use the higher-order triangular or tetrahedron elements wherever possible, as opposed to using
collapsed quadrilaterals and hexahedra.
Beam Elements
Marc’s beam elements are 2- and 3-node, two- and three-dimensional elements that can model straight
and curved-beam structures and framed structures and can serve as stiffeners in plate and shell structures.
A straight beam of a circular cross section can be used for modeling the straight portion of piping
systems. Translational and rotational degrees of freedom are included in beam elements. The cross
section of the beam can be solid or thin-walled. The solid sections can be standard elliptical, rectangular,
trapezoidal, or hexagonal sections or arbitrary solid sections. The thin-walled sections can be standard
closed circular sections or arbitrary closed or open sections. The BEAM SECT parameter is used to define
all solid sections and to define arbitrary thin-walled open or closed sections.
The beam elements are numerically integrated along their axial direction using Gaussian integration. The
stress strain law is integrated through thin-walled sections using a Simpson rule and through solid
sections using a Simpson, Newton-Cotes, or Gauss rule depending on the input specifications. Stresses
and strains are evaluated at each integration point through the thickness. This allows an accurate
calculation if nonlinear material behavior is present. In elastic beam elements, only the total axial force
and moments are computed at the integration points.
If element 52 or 98 is used with numerical section integration, six stiffness factors may be used to modify
its stiffness behavior in axial tension, bending, torsion and shear. The modified internal virtual work of
beam element type 98 when using any of the stiffness factors is given as:
L
δW i n t =  ( f 1 N z δε z + f2 M z δκ z + f 3 M y δκ y + f 4 D x δγ x + f 5 D y δγ y + f 6 T z δκ z ) dz
0
For element type 52 the transverse shear terms are absent. In this expression, f 1 is the normal stiffness
factor, f 2 and f 3 are the bending stiffness factors, f 4 and f 5 are the transverse shear stiffness factors
and f 6 is the torsional stiffness factor. All stiffness factors have a default value of 1 and must be positive.
If stiffness factors with other values than 1 are used, the stress output must be interpreted as follows:
Generalized stresses will be scaled with their respective stiffness factors (i.e., f 1 N z , f 2 M x , etc.
will be written to the output or post file)
Layer stresses are not scaled with their respective stiffness factors, meaning that their section
integration will result in the unscaled generalized stresses (i.e., N z , M x , etc.)
If the section is used as a pre-integrated section all stiffness factors can have arbitrary positive values, but
if the section is used with numerical integration throughout the analysis the stiffness contributions
resulting from the direct stresses must use identical positive stiffness factors (i.e., f 1 = f 2 = f 3 ) and
the stiffness contributions resulting from the shear stresses must use identical positive stiffness factors
(i.e. f 4 = f 5 = f 6 ) in order to guarantee symmetry of the stiffness matrix. If these conditions are not
met, f 2 and f 3 will silently be set to f 1 and f 4 and f 5 will silently be set to f 6 .
Plate Elements
The linear shell elements (Types 49, 72, 75, 138, 139, or 140) or the quadratic element (Type 22) can be
used effectively to model plates and have the advantage that tying is unnecessary. For element type 49,
you can indicate on the GEOMETRY option that the flat plate formulation is to be used. This reduces
computational costs. To further reduce computational costs for linear elastic plate analysis, the number
of points through the thickness can be reduced to one by use of the SHELL SECT parameter.
Shell Elements
Marc contains three isoparametric, doubly curved, thin shell elements: 3-, 4-, and 8-node elements (Types
4, 8, and 24, respectively) based on Koiter-Sanders shell theory. These elements are C1 continuous and
exactly represent rigid-body modes. The program defines a mesh of these elements with respect to a
surface curvilinear coordinate system. You can use the FXORD model definition option to generate the
nodal coordinates. Tying constraints must be used at shell intersections.
Thin shell analysis can be performed using either the 3-node discrete Kirchhoff theory (DKT) based
element (Type 138), the 4-node bilinear DKT element (Type 139), the 6-node bilinear semi-loof element
(Type 49), or the 8-node bilinear semi-loof element (Type 72). Thick shell analysis can be performed
using the 4-node bilinear Mindlin element (Type 75), the 4-node reduced integration with hourglass
Mindlin element (Type 140), or the 8-node quadratic Mindlin element (Type 22). The thick shell elements
have been developed so that there is no locking when used for thin shell applications.
The global coordinate system defines the nodal degrees of freedom of these elements. These elements
are convenient for modeling intersecting shell structures since tying constraints at the shell intersections
are not needed.
Marc contains three axisymmetric shell elements: 2-node straight, 2-node curved, and 3-node curved.
You can use these elements to model axisymmetric shells; combined with axisymmetric ring elements,
they can be used to simulate the thin and thick portions of the structure. The program provides standard
tying constraints for the transition between shell and axisymmetric ring elements.
Heat Transfer Elements

The heat transfer elements in Marc consist of the following:
• 2- and 3-node three-dimensional links
• 3-, 4-, 6-, and 8-node planar and axisymmetric elements
• 6- and 9-node planar and axisymmetric semi-infinite elements
CHAPTER 10 703
Element Library
• 8-, 10-, and 20-node solid elements

• 4- and 10-node tetrahedral
• 6-, 15-node pentahedral
• 12- and 27-node semi-infinite brick elements
• 2-, 3-, 4-, 6-, and 8-node shell elements.
• 3-, 4-, 5-, and 8-node 3-D membrane elements
• 4-, 8-node planar interface elements
• 4-, 8-node axisymmetric interface elements
• 6-, 8-, 15-, 20-node solid interface elements.
For each heat transfer element, there is at least one corresponding stress element, enabling you to use the
same mesh for both the heat transfer and thermal stress analyses. Heat transfer elements are also
employed to analyze coupled thermo-electrical (Joule heating) problems.
In heat transfer continuum elements or link elements, temperature is the only nodal degree of freedom.
The heat transfer shell elements can be used in two modes. In the first mode, the number of degrees of
freedom is two or three for linear or quadratic temperature distribution across the complete thickness. In
the second mode, the number of degrees of freedom are M + 1 or 2 * M + 1 for linear or quadratic
temperature distribution for each M layer. This is defined through the HEAT parameter.
The heat transfer membrane elements have only one degree of freedom per node, so there is no thermal
gradient through the thickness.
In Joule heating, the voltage and temperature are the nodal degrees of freedom. Shell elements are not
available for Joule heating.
Acoustic Analysis
Heat transfer elements are also used to model the compressible media in acoustic analysis. In this case,
the pressure is the single nodal degree of freedom.
Heat transfer elements are also used for electrostatic analysis. The scalar potential is the degree
of freedom.
Coupled Electrostatic-Structural
Stress elements are used for a coupled electrostatic-structural analysis. For each of these elements, there
is a corresponding heat transfer element available for the electrostatic pass. When a medium has no
stiffness, like air, heat transfer elements can be chosen for this region. Then this region will be inactive
in the stress pass.
Fluid/Solid Interaction
Heat transfer elements are used to model the inviscid, incompressible fluid/solid interaction problems.
The hydrostatic pressure is the degree of freedom.

The three-dimensional heat transfer links and planar elements are used to model the lubricant film. As no
variation occurs through the thickness of the film, two-dimensional problems are reduced to one-
dimensional, and three-dimensional problems are reduced to two-dimensional. The pressure is the degree
of freedom.
For two-dimensional problems, a scalar potential can be used; hence, the heat transfer planar and
axisymmetric elements are employed. The single degree of freedom is the potential. For three-
dimensional analyses, magnetostatic elements are available. In such cases, there are three degrees of
freedom to represent the vector potential.
In electromagnetic problems, a vector potential, augmented with a scalar potential, is used. There are
lower-order elements available for these analyses.
Soil Analysis
There are three types of soil/pore pressure analysis. If a pore pressure analysis only is performed, “heat
transfer” elements (41, 42, or 44) are used. If an uncoupled soil analysis is performed, the standard
elements (21, 27, or 28) are used. If a coupled analysis is performed, the Herrmann elements (32, 33, or
35) are used. In this case, the last degree of freedom is the pore pressure.
Fluid Analysis
When performing fluid analysis, any planar, axisymmetric or solid continuum element can be used.
These elements either used a mixed formulation or a penalty formulation based upon the value of the
FLUID parameter. For the mixed formulation, each node of lower- or higher-order continuum element has
velocities as well as pressure degrees of freedom. The penalty method yields elements with only the
velocities as the nodal variables.
In a piezoelectric analysis, a strong coupling exist between stress and electric field. There are lower-order
elements available for this analysis. The first two or three degrees of freedom (in resp. 2-D or 3-D) are
available for displacement components, and the last (3rd or 4th) degree of freedom is available for the
electric potential.
CHAPTER 10 705
Element Library
Special Elements
Marc contains unique features in the program.
Gap-and-Friction Elements
The gap-and-friction elements (12, 97) are based on the imposition of gap closure constraint and
frictional-stick or frictional-slip through Lagrange multipliers. These elements provide frictional and
gapping connection between any two nodes of a structure; they can be used in several variations,
depending on the application. In the default formulation, the elements simulate a gap in a fixed direction,
such that a body does not penetrate a given flat surface. Using the optional formulation, you can constrain
the true distance between two end-points of the gap to be greater than some arbitrary distance. This
ability is useful for an analysis in which a body does not penetrate a given two-dimensional circular or
three-dimensional spherical surface. Finally, you can update the gap direction and closure distance
during analysis for the modeling of sliding along a curved surface.
Pipe-bend Element
The pipe-bend (3-node elbow) element 17 is designed for linear and nonlinear analysis of piping
systems. It is a modified axisymmetric shell element for modeling the bends in a piping system. The
element has a beam mode superposed on the axisymmetric shell modes so that ovalization of the cross-
section is admitted. The twisting of a pipe-bend section is ignored because the beam has no flexibility in
torsion. Built-in tying constraints are used extensively for coupling the pipe-bend sections to each other
and to straight beam elements.
Curved-pipe Element
The curved-pipe (2-node) element 31 is designed for linear analysis only. The stiffness matrix is based
upon the analytic elastic solution of a curved pipe.
Shear Panel Element

The shear panel (4-node) element 68 is an elastic element of arbitrary shape. It is an idealized model of
an elastic sheet. This element only provides shearing resistance. It must be used with beam stiffeners to
ensure any normal or bending resistance. The shear-panel element is restricted to linear material and
small displacement analysis.
Cable Element
The cable element (51) is an element which exactly represents the catenary behavior of a cable. It is an
elastic element only.
Rebar Elements
The rebar elements (23, 46, 47, 48, 142, 143, 144, 145, 146, 147, 148, 165, 166, 167, 168, 169, and 170)
are hollow elements in which you can place single-direction strain members (reinforcing cords or rods).
The rebar elements are used in conjunction with other solid elements (filler) to represent a reinforced
material such as reinforced concrete. The reinforcing members and the filler are accurately represented
by embedding rebar elements into solid elements. The rebar elements can be used for small as well as
large strain behavior of the reinforcing cords. Also, any kind of material behavior can be simulated by
the rebar elements in Marc.
Interface Elements
Elements 186 through 193 are special elements which are used to model the delamination in composite
structures where each ply is modeled with solid elements. The material behavior of these elements is
defined with the COHESIVE model definition option.
Elements 220 through 227 are analogous elements for thermal analysis of interfaces. The material of
these elements is defined with the COHESIVE thermal option.
Note that these interface elements do not have a mass matrix or a specific heat matrix and distributed
loads, fluxes, or films are not applied to these elements. One can use either a Guass or Newton-Cotes
intergration scheme. The latter may be advantageous in cases where the interface has significantly
different stiffness compared to the surrounding material.
Incompressible Elements
Incompressible and nearly incompressible materials can be modeled by using a special group of elements
in the program. These elements, based on modified Herrmann variational principles, are capable of
handling large deformation effects as well as creep and thermal strains. The incompressibility constraint
is imposed by using Lagrange multipliers. Generally, the low (linear) order elements have a single
additional node which contains the Lagrange multiplier, while the high-order (quadratic) elements have
Lagrange multipliers at each corner node.
Elements 155 (plane strain triangle), 156 (axisymmetric triangle), and 157 (3-D tetrahedron) are low-
order elements and are exceptions within the incompressible element group. They have an
additional node located at the center of the elements and have Lagrange multipliers at each corner node.
The shape function of the center node is a bubble function. See Figure 10-1 for an element 155 case. The
degrees of freedom of the center node is condensed out on the element level before the assembly of the
global matrix.
3 Nodes with both displacements and pressure as
degrees of freedom
4
x Center node with only displacement degrees of
x freedom. The shape function for this node is a
bubble function.
2
1
CHAPTER 10 707
Element Library
Figure 10-1 Element 155
Large Strain Elasticity

The incompressible elements based on Herrmann formulations can be used for large strain analysis of
rubber-like materials, using either total Lagrangian formulation or updated Lagrangian formulation.
Large Strain Plasticity

Elemen The incompressible elements can be used for large strain analysis of elasto-plastic materials using
t updated Lagrangian formulation, based on multiplicative decomposition of deformation gradient.
Library
Incomp Rigid-Plastic Flow

ressibl In rigid-plastic flow analysis, with effectively no elasticity, incompressible elements must be used. For
e such analyses, you have a choice of using either the elements with Lagrange multipliers discussed above,
Elemen or standard displacement continuum elements. In the latter case, a penalty function is used to satisfy the
ts incompressibility constraint.
Constant Dilatation Elements

You can choose an integration scheme option, which makes the dilatational strain constant throughout
the element. This can be accomplished by setting the second field of the GEOMETRY option to one.
Constant dilatational elements are recommended for use in approximately incompressible, inelastic
analysis, such as large strain plasticity, because conventional elements can produce volumetric locking
due to overconstraints for nearly incompressible behavior. This option is only available for elements of
lower order (Types 7, 10, 11, 19, and 20). For the lower-order reduced integration elements (114 to 117)
with hourglass control, as only one integration point is used, these elements do not lock, and effectively,
a constant dilatation formulation is used. (The constant dilatation approach has also been referred to
mean dilatation and B-bar approach in the literature.)
Reduced Integration Elements

Marc uses a reduced integration scheme to evaluate the stiffness matrix or the thermal conductivity
matrix for a number of isoparametric elements. The mass matrix and the specific heat matrix of the
element are always fully integrated.
For lower-order, 4-node quadrilateral elements, the number of numerical integration points is reduced
from 4 to 1; in 8-node solid elements, the number of numerical integration points is reduced from 8 to 1.
For 8-node quadrilateral elements, the number of numerical integration points is reduced from 9 to 4; for
20-node solid elements, the number of numerical integration points is reduced from 27 to 8.
The energy due to the higher-order deformation mode(s) associated with high-order elements is not
included in the analysis. Reduced integration elements and fully integrated elements can be used together
in an analysis. Reduced integration elements are more economical than fully integrated elements and
they can improve analysis accuracy. However, with near singularities and in regions of high-strain
gradients, the use of reduced integration elements can lead to oscillations in the displacements and
produce inaccurate results. Using reduced integration elements results in zero-energy modes, or
breathing nodes. In the lower-order elements, an additional stabilization stiffness is added which
eliminates these so-called “hourglass” modes.
Continuum Composite Elements

There is a group of isoparametric elements in Marc which can be used to model composite materials.
Different material properties can be used for different layers within these elements. The number of
continuum layers within an element, the thickness of each layer, and the material identification number
for each layer are input via the COMPOSITE option. A maximum number of 1020 layers can be used
in one 2-D continuum composite element. A 3-D continuum composite element only allows a maximum
of 510 layers within the element.
These structural elements are available as both lower- and higher-order and can be used for plane strain
(element types 151 and 153), axisymmetric (element types 152 and 154) or 3-D solid analysis (element
types 149 and 150). Corresponding heat transfer elements are available for lower- and higher-order and
can be used for planar (element types 177 and 179), axisymmetric (element types 178 and 180) or 3-D
solid analysis (element types 175 and 176).
Fourier Elements
A special class of elements exists which allows the analysis of axisymmetric structures with
nonaxisymmetric loads. The geometry and material properties of these structures do not change in the
circumferential direction, and the displacement can be represented by a Fourier series. This
representation allows a three-dimensional problem to be decoupled into a series of two-dimensional
problems. Both solid and shell Fourier elements exist in Marc. These elements can only be used for linear
elastic analyses.
Semi-infinite Elements
This group of elements can be used to model unbounded domains. In the semi-infinite direction, the
interpolation functions are exponential, such that the function (displacement) is zero at the far domain.
The rate of decay of the function is dependent upon the location of the midside nodes. The interpolation
function in the non semi-infinite directions are either linear or quadratic.
These elements can be used for static plane strain, axisymmetric, or 3-D solid analysis. They can also be
used for heat transfer, electrostatic, and magnetostatic analyses.
CHAPTER 10 709
Element Library
Cavity Surface Elements

This group of elements can be used to define the boundaries of cavities where standard finite elements
are not used; for example, along rigid boundaries. These elements are for volume calculation purposes
only and do not contribute to the stiffness equations of the model. They can be used for plane strain, plane
stress, axisymmetric, and 3-D analyses. No material properties are needed for these elements. They do
not undergo any deformation except if they are attached or glued to other elements or surfaces.
Assumed Strain Formulation

Conventional isoparametric four-node plane stress and plane strain, and eight-node brick elements
behave poorly in bending. The reason is that these elements do not capture a linear variation in shear
strain which is present in bending when a single element is used in the bending direction. For the six
elements (3, 7, 11, 160, 161, and 163), the interpolation functions have been modified such that shear
strain variation can be better represented. For the lower-order reduced integration elements (114 to 123),
an assumed strain formulation written with respect to the natural coordinates is used. For elastic isotropic
bending problems, this allows the exact displacements to be obtained with only a single element through
the thickness. This is invoked by using the ASSUMED STRAIN parameter or by setting the third field of
the GEOMETRY option to one.
Follow Force Stiffness Contribution

When activating the FOLLOW FOR parameter, the distributed loads are calculated based upon the current
deformed configuration. It is possible to activate an additional contribution which goes into the stiffness
matrix. This improves the convergence. This capability is available for element types 3, 7, 10, 11, 18, 72,
75, 80, 81, 82, 83, 84, 114, 115, 116, 117, 118, 119, 120, 139, 140, 149, 151, 152, 160, 161, 162, or 163.
Inclusion of the follower force stiffness can result in nonsymmetric stiffness for nonenclosed volumes,
thereby, resulting in increased computational times. You can flag the nonsymmetric solver in the
SOLVER option.
Explicit Dynamics
The explicit dynamics formulation IDYN=5 model is restricted to the following elements:
1, 2, 3, 5, 6, 7, 9, 10, 11, 18, 19, 20, 52, 64, 75, 89, 98, 114, 115, 116, 117, 118, 119, 120, 130,
139, and 140.
When using this formulation, the mass matrix is defined semi-analytically; for example, no numerical
integration is performed. In addition, a quick method is used to calculate the Courant stability limit
associated with each element. For these reasons, this capability is limited to the elements
mentioned above.
Adaptive Mesh Refinement

Marc has a capability to perform local adaptive mesh refinement to improve the accuracy of the solution.
This capability is invoked by using the ADAPTIVE parameter and model definition option. The adaptive
meshing is available for the following 2-D and 3-D elements:
2, 3, 6, 7, 10, 11, 18, 19, 20, 37, 38, 39, 40, 43, 75, 80, 81, 82, 83, 84, 111, 112, 113, 114, 115,
116, 117, 118, 119, 120, 121, 122, 123, 138, 139, 140, 155, 156, 157, 160, 161, 162, 163, 164,
196, 198, and 201.
CHAPTER 11 711
Solution Procedures for Nonlinear Systems
Chapter 11 Solution Procedures for Nonlinear Systems
11 Solution Procedures for

Nonlinear Systems

Considerations for Nonlinear Analysis 712
 Newton-Raphson Method 726

Modified Newton-Raphson Method 727

Direct Substitution 729
 Arc-length Methods 729


 Solution of Linear Equations 740

Flow Diagram 746

Remarks 747
 References 748
This chapter discusses the solution schemes in Marc for nonlinear problems. Issues of convergence
controls, singularity ratio, and available solvers for linearized system of equations are also discussed. In
a nonlinear problem, a set of equations must be solved incrementally. The governing equation of the
linearized system can be expressed, in an incremental form, as
Kδu = r (11-1)
where δu and r are the correction to the incremental displacements and residual force vectors,
respectively.
There are several solution procedures available in Marc for the solution of nonlinear equations:
• Newton-Raphson Method
• Modified Newton-Raphson Method
• Direct Substitution
• Arc-length Methods
Considerations for Nonlinear Analysis

Nonlinear analysis is usually more complex and expensive than linear analysis. Also, a nonlinear
problem can never be formulated as a set of linear equations. In general, the solutions of nonlinear
problems always require incremental solution schemes and sometimes require iterations (or recycles)
within each load/time increment to ensure that equilibrium is satisfied at the end of each step.
Superposition cannot be applied in nonlinear problems.
The iterative procedures supported in Marc are: Newton-Raphson, Modified Newton-Raphson,
Newton-Raphson with strain correction, and direct substitution. If the R-P flow contribution model is
chosen, a direction substitution is used.
A nonlinear problem does not always have a unique solution. Sometimes a nonlinear problem does not
have any solution, although the problem can seem to be defined correctly.
Nonlinear analysis requires good judgment and uses considerable computing time. Several runs are often
required. The first run should extract the maximum information with the minimum amount of computing
time. Some design considerations for a preliminary analysis are:
• Minimize degrees of freedom whenever possible.
• Halve the number of load increments by doubling the size of each load increment.
• Impose a coarse tolerance on convergence to reduce the number of iterations. A coarse run
determines the area of most rapid change where additional load increments might be required.
Plan the increment size in the final run by the following rule of thumb: there should be as many
load increments as required to fit the nonlinear results by the same number of straight lines.
Marc solves nonlinear static problems according to one of the following two methods: tangent modulus
or initial strain. Examples of the tangent modulus method are elastic-plastic analysis, nonlinear springs,
nonlinear foundations, large displacement analysis and gaps. This method requires at least the following
three controls:
• A tolerance on convergence
• A limit to the maximum allowable number of recycles
• Specification of a minimum number of recycles
CHAPTER 11 713
An example of the initial strain method is creep or viscoelastic analysis. Creep analysis requires the
following tolerance controls:
• Maximum relative creep strain increment control
• Maximum relative stress change control
• A limit to the maximum allowable number of recycles
To input control tolerances, use the model definition option CONTROL for all simulations. These values
can be reset upon restart or through the CONTROL history definition option. For creep simulations,
addition controls are provided in the AUTO CREEP or AUTO STEP options. See Convergence Controls
in this chapter for further discussion on tolerance controls.
Behavior of Nonlinear Materials

Nonlinear behavior can be time- (rate-) independent, or time- (rate-) dependent. For example, plasticity
is time-independent and creep is time-dependent. Both viscoelastic and viscoplastic materials are also
time-dependent. Nonlinear constitutive relations must be modeled correctly to analyze nonlinear
material problems. A comprehensive discussion of constitutive relations is given in Chapter 7.
Scaling the Elastic Solution

The SCALE parameter causes scaling of the linear-elastic solution to reach the yield stress in the highest
stressed element. Scaling takes place for small displacement elastic-plastic analysis, where element
properties do not depend on temperature. The SCALE parameter causes all aspects of the initial solution
to be scaled, including displacements, strains, stresses, temperature changes, and loads. Subsequent
incrementation is then based on the scaled solution.
Load Incrementation
Several history definition options are available in Marc to input mechanical and thermal load increments
(see Table 11-1). The choice is between a fixed and an automatic load stepping scheme.
Table 11-1 History Definition Options for Load Incrementation
Load Type Fixed Adaptive
Mechanical AUTO LOAD AUTO STEP
DYNAMIC CHANGE AUTO INCREMENT
CREEP INCREMENT AUTO THERM
AUTO CREEP
AUTO THERM CREEP*
Thermal TRANSIENT NON AUTO AUTO STEP
TRANSIENT*
*The option is obsolete; use AUTO STEP instead.
For a fixed scheme, the load step size remains constant during a load case. The fixed schemes are AUTO
LOAD for static mechanical, CREEP INCREMENT for creep, DYNAMIC CHANGE for dynamic
mechanical, and TRANSIENT NON AUTO for thermal/thermo-mechanically coupled.
In many nonlinear analyses, it is useful to have Marc automatically determine the appropriate load
step size. For an adaptive scheme, the load step size changes from one increment to the other and also
within an increment depending on convergence criteria and/or user-defined physical criteria. The
adaptive schemes are AUTO STEP and AUTO INCREMENT for static mechanical, AUTO STEP and
AUTO CREEP for creep, AUTO STEP, AUTO THERM, and AUTO THERM CREEP for thermally driven
mechanical problems, and AUTO STEP for dynamic mechanical, and AUTO STEP and TRANSIENT for
thermal/thermo-mechanically coupled.
The adaptive stepping scheme of choice is AUTO STEP. AUTO STEP has been designed as a unified load
stepping scheme and many of the capabilities of the other stepping schemes can now be handled by AUTO
STEP. AUTO STEP can be used for mechanical (static, creep, dynamic), thermal, and thermo-
mechanically coupled analysis problems.
• In the AUTO STEP scheme, either a recycle based convergence criterion is used to automatically
determine the time step based upon a comparison of the actual number of recycles needed in an
increment to a user-specified desired number of recycles or a damping energy procedure is used.
In addition to this, user-defined or program-determined physical criteria based upon strain, stress,
displacement, or temperature increments can be used to control the time step. Reductions in the
time step through cut-backs are used to satisfy both convergence criteria and physical criteria.
More details on the AUTO STEP option are provided in the next section.
• Post buckling or snap-through analyses require the so-called arclength method which is available
through the AUTO INCREMENT option. This option can only be used in static mechanical
analyses and the applied load is automatically increased or decreased in order to maintain a
certain arclength.
Note: AUTO INCREMENT can also be used for general situations without instabilities, but, in
general, the AUTO STEP option is preferred for these situations.
• For creep analysis, the available adaptive options are AUTO CREEP and AUTO STEP. The AUTO
CREEP option determines the time step in an explicit creep analysis based on the creep strain
change and the stress change (see Volume C: Program Input, Chapter 3). These checks are not
available by default in AUTO STEP. These may however be incorporated by using an absolute or
relative creep strain increment criterion and an absolute or relative stress increment criterion as
additional user-defined physical criteria in combination with the default convergence criterion.
Addition of these user-defined criteria for AUTO STEP is quite simple. If an appropriate input
flag is set, two physical criteria are automatically added by the program at run-time for explicit
creep problems: creep strain increment/elastic strain = 0.5, and stress increment/stress at
beginning of increment = 0.5. It should also be noted that AUTO STEP is usually more reliable in
cases involving creep and contact.
• For the case of automatic load stepping for a thermally loaded elastic-creep/elastic-plastic-creep
stress analysis, the available adaptive schemes are AUTO THERM CREEP and AUTO STEP.
Allowable increments for normalized creep strain, normalized stress and state variables can be
optionally prescribed for AUTO STEP either through user-defined criteria or program determined
automatic physical criteria.
CHAPTER 11 715
• For thermally driven mechanical problems, the available options are AUTO THERM and AUTO
STEP. The thermal loads derived from a thermal analysis are applied using the CHANGE STATE
option in a mechanical analysis. In the AUTO THERM scheme, the load step of the mechanical
analysis is automatically adjusted based upon user-specified (allowed) changes in temperature
from the thermal analysis per increment. For example, if there is a change of 50° in the thermal
analysis in one increment but only a change of 10° per increment is allowed in the mechanical
analysis, AUTO THERM splits up the thermal increment into five mechanical increments.
Allowable state variable increments can be optionally prescribed for AUTO STEP either through
a user-defined criterion or program determined automatic physical criterion. If these criteria are
violated in an increment, AUTO STEP cuts the time-step back and repeats the increment with a
smaller time step.
Selecting Load Increment Size

Selecting a proper load step increment is an important aspect of a nonlinear solution scheme. Large steps
often lead to many recycles per increment and, if the step is too large, it can lead to inaccuracies and
nonconvergence. On the other hand, using too small steps is inefficient.
Fixed Load Incrementation

When a fixed load stepping scheme is used, it is important to select an appropriate load step size that
captures the loading history and allows for convergence within a reasonable number of recycles. For
complex load histories, it is necessary to prescribe the loading through time tables while setting up
the run.
For fixed stepping, there is an option to have the load step automatically cut back in case of failure to
obtain convergence. When an increment diverges, the intermediate deformations after each recycle can
show large fluctuations and the final cause of program exit can be any of the following: maximum
number of recycles reached (exit 3002), elements going inside out (exit 1005 or 1009) or, in a contact
analysis, nodes sliding off a rigid contact body (exit 2400), and nodes not being projected properly onto
3-D NURBS (exit 2401). These deformations are normally not visible as post results (there is a feature
to allow for the intermediate results to be available on the post file, see the POST option). If the cutback
feature is activated and one of these problems occur, the state of the analysis at the end of the previous
increment is restored and the increment is subdivided into a number of subincrements. The step size is
halved until convergence is obtained or the user-specified number of cutbacks has been performed. Once
a subincrement is converged, the analysis continues to complete the remainder of the original increment.
No results are written to the post file during subincrementation. When the original increment is finished,
the calculation continues to the next increment with the original increment count and time step
maintained. If the global remeshing option is activated in conjunction with the cutback feature, then, for
exit 1005 or 1009, the chosen contact body is remeshed and the analysis is repeated with the original time
step before the first cutback.
Arc Length Method

The arc-length procedures assume that the control of the nonlinear behavior and possible instabilities is
due to mechanical loads, and that the objective is to obtain an equilibrium position at the end of the
loadcase. Hence, while the program may increase or decrease the load, the load can always be considered
to be F = F b + λ ( F e – F b ) , where F b and F e are the loads at the beginning and end of the loadcase.
The scale factor does not necessarily vary linearly from 0 to 1 over the increments, and may, in fact,
become negative. This would result in negative time steps as well; hence, the AUTO INCREMENT history
definition option cannot be used with dynamics and should not be used with table driven input where the
load is a function of time.
Mechanical loads, as shown above, are applied in a proportional manner and thermal loads are
applied instantaneously.
This means that any automatic load incrementation method is limited to mechanical input histories that
only have linear variations in load or displacement and thermal input histories that have immediate
change in temperature. For example, one may not use a rigid body with a linearly changing velocity, since
the resulting displacement of the rigid body would give parabolically changing displacements. In this
case, one would need to use a constant velocity for the arc length method to work properly.
For the arc length method, care must be taken to appropriately define the loading history in each loadcase.
The load case should be defined between appropriate break points in the load history curve. For example,
in Figure 11-1, correct results would be obtained upon defining three distinct loadcases between times
0 – t 1 , t 1 – t 2 , and t 2 – t 3 during the model preparation. However, if only one load case is defined for
the entire load history between 0 – t 3 , the total applied load for the loadcase is zero. This is also true
when the AUTO STEP history definition option is used when the table driven input is not used.
P2 P2
P (Load)
P (Load)
P1 P1
t1 t2 t3 t3
0 0
t (Time) t (Time)
a. Three Defined Loadcases b. One Defined Loadcase
Figure 11-1 Defining Loadcases for Automatic Load Incrementation
When the table driven input method is used with the AUTO INCREMENT history definition option and
behavior as shown in Figure 11-1 is desired, the independent variable should be the load case number
and not time. A quasi-static total mechanical load can, hence, be given which is both a function of
position and the load case number.
CHAPTER 11 717
AUTO STEP
The scheme appropriate for most applications is AUTO STEP. The primary control of the load step is
based upon the number of recycles needed to obtain convergence. There are a number of optional user-
specified physical criteria that can be used to additionally control the load step. The user inputs needed
to define the AUTO STEP scheme are described in Marc Volume C: Program Input.
Recycling Criterion
The default recycle based criterion works as follows: The user specifies a desired number of recycles.
For most problems, it is sufficient to provide a value in the range of three to five. For problems with
severe nonlinearities or for problems with very small convergence tolerances, it may be necessary to
increase this number. This number is used as a target value for the load stepping scheme. If the number
of recycles required in the current increment is less than the desired number, the load step for the next
increment is increased. The time step increase is based on a factor, S u , that can also be specified by the
user. Typical values for S u are in the range of 1.2 to 1.5. While the time step increase is obviously more
aggressive with larger scale factors, it should be noted that there may be excessive recycling and
cutbacks if sudden nonlinearities are encountered. In order to avoid this, Marc uses the following logic
for higher scale factors. If the actual number of recycles in an increment is greater than 60 % of the
desired number of recycles (i.e., the current increment did not converge easily), the increase scale factor
for the next increment is limited to 1.25 for scale factor values between 1.25 and 1.5625, and to 80 % of
the value for scale factors above 1.5625.
See Damping Energy for alternative approach.
Time Step Cutback Scheme

The load step is never increased during an increment. If the number of recycles needed to obtain
convergence exceeds the desired number, the load step size is scaled back, the recycling cutback number
N r is incremented by 1 and the increment is performed again with the new load step. The scaleback
Nr
factor for the N r th cutback is taken as s , where the factor s is calculated from the expression
2 ⁄ ( Nr m ( Nr m + 1 ) )
s = [ Ts ⁄ Tm ] ; where N r m is the maximum number of recycling related cutbacks
for the increment and is calculated from N r m = log 10 ( 10 5 * T s ⁄ T m ) , T s is the time increment before
any recycling related cutbacks occur for the increment and T m is the minimum possible time step for
the increment. T m is equal to the value set by the user ( 10 – 5 by default) if there is no quasi-static
artificial damping and is equal to 10 – 3 times the value set by the user ( 10 – 8 by default) if there is quasi-
Nr
static artificial damping. The scaleback factor for any cutback is the smaller of ( s , 1 ⁄ S u ). This
scheme guarantees that no matter what the starting time step for an increment, the minimum time step is
reached in a reasonable number of cutbacks if the increment consistently fails to converge. For special
cutbacks such as maximum number of recycles reached (exit 3002), elements going inside out (exit 1005
or 1009) or, in a contact analysis, nodes sliding off a rigid contact body (exit 2400) nodes not being
Nr
projected properly onto a 3-D NURB (exit 2401), the scaleback factor is the smaller of ( s , 0.5). If the
minimum time step is reached and the analysis still fails to converge, it is terminated with exit 3015. If
the ‘proceed when not converged’ option is used, then the analysis proceeds to the next increment if and
when the maximum number of recycles are reached.
Exceptions
There are some exceptions to the basic scheme outlined above. If an increment is consistently converging
with the current load step and the number of recycles exceeds the desired number, the number of recycles
is allowed to go beyond the desired number until convergence or up to the user specified maximum
number. The time step is then decreased for the next increment by 1 ⁄ S u . An increment is determined to
be converging if the convergence ratio was decreasing in three previous recycles.
Special rules also apply in a contact analysis. For quasi-static problems, the AUTO STEP option is
designed to only use the automated penetration check option (see CONTACT option, 7th field of 2nd data
block; option 3 is always used). During the recycles, the contact status can keep changing (new nodes
come in contact, nodes slide to new segments, separate etc.). Whenever the contact status changes during
an increment, a new set of contact constraints are incorporated into the equilibrium equations and more
recycles are necessary in order to find equilibrium. These extra recycles, due to contact changes, are not
counted when the recycle number is checked against the desired number for determining if the load step
needs to be decreased within the increment. Thus, only true Newton-Raphson iterations are taken into
account. For the load step of the next increment, the accumulated number of recycles during the previous
increment is used. This ensures that the time step is not increased when there are many changes in contact
during the previous increment.
Thermal Analysis
For the most part, the recycling criterion works in a similar fashion for thermal analysis (heat transfer
analysis, or thermal part of a coupled thermo-mechanical analysis). The recycling criterion is used to
satisfy the tolerance value provided for the temperature error in estimate on the CONTROL option; that
is, the analysis cuts back and starts the increment over if the temperature error in estimate is not
consistently converging within the desired number of recycles. If the temperature estimate consistently
converges in three previous recycles, the analysis continues recycling. Once the temperature estimate
tolerance is satisfied, the actual incremental temperature change ΔT a is calculated and checked against
the corresponding tolerance ΔT m provided on the CONTROL option. It should be noted that if a user
criterion on temperature is also provided, the latter parameters over-ride the one provided on the
CONTROL option. More details on how user criteria are handled are described in the next section. If the
incremental temperature change is not satisfied, the time step is scaled-back using a factor =
0.8 ( ΔT m ⁄ ΔT a ) . In this case, the thermal pass is simply continued with the smaller time step without a
cut-back because if there was a cutback, the analysis would have redo the temperature error in estimate
convergence from scratch. However, if the maximum number of allowed recycles is reached before
Nr
thermal convergence is achieved, a cutback with a scale-back factor of min( s , 0.5) is made and the
increment is repeated.
CHAPTER 11 719
User-defined Physical Criteria

In addition to allowing Marc to use the number of recycles for automatically controlling the step size for
AUTO STEP, user-specified physical criteria can optionally be used for controlling the step size. The
user-specified physical criteria work as follows. The user can specify the maximum allowed incremental
change within certain ranges for specific quantities during an increment. The quantities available are
displacements, rotations, stresses, strains, strain energy, temperature (in thermal or thermomechanically
coupled analyses), and state variables (in mechanical or thermomechanically coupled analysis). These
criteria can be utilized in one of two ways.
By default, they are used as limits, which means that the load step is immediately decreased if a criterion
is violated during any iteration of the current increment, but they do not influence the decision to change
the load step for the next increment; that is, only the actual number of recycles versus desired number of
recycles controls the load step for the next increment.
The criteria can also be used as targets; in which case, they are used to control the time step for the current
and next increments. If the calculated values of the criteria are higher than the user-specified values in
any iteration, the time step is scaled down and the current increment is repeated. If the calculated values
of the criteria for the current increment are consistently smaller than the user-specified values prior to
convergence, the time step for the next increment is scaled up. The scale factor used for reduction or
increase is the ratio between the actual value and the target value and this factor is limited by
user-specified minimum and maximum factors (defaults to 0.1 and 10 respectively). If this type of load
step control is used together with the recycle based control, the time step can be reduced in the current
increment due to whichever criterion that is violated. The decision to increase the step size for the next
increment is solely based upon the physical criteria.
Specification of user-defined physical criteria can be further simplified by setting a special flag in the
AUTO STEP option that allows for physical criteria to be automatically added by the solver at run-time.
These automatic criteria serve as upper-bound controls to prevent run-away Newton-Raphson iterations
that ultimately cause the program to abort. Currently, four mechanical criteria are automatically added
depending on the kind of analysis that is being run: a total strain criterion is added for any large
displacement analysis and the maximum allowable equivalent total strain increment at any point in the
model is set to 50%; a plastic strain criterion is added for any large displacement, finite strain analysis
and the maximum allowable equivalent plastic strain increment at any point in the model is set to10%;
a relative creep strain criterion and a relative stress change criterion are added for any explicit creep
analysis wherein the maximum allowable creep strain change/elastic strain and the maximum allowable
equivalent stress change/equivalent stress are each set to 0.5; a state variable criterion is added for any
large displacement analysis wherein the maximum allowable temperature increment is such that the
equivalent stress increment associated with the change in thermal properties of the materials does not
exceed 50% of the total equivalent stress. These criteria are only added in the analysis if there are no
competing explicitly defined user-criteria found. It should also be noted that these automatic criteria
are only used as limits; i.e. they are used to control the time step within an increment but not for the
next increment.
Failure to satisfy user-defined physical criteria can occur due to two reasons — the maximum number
of cutbacks allowed by the user can be exceeded, or the minimum time step can be reached. In this case,
the analysis terminates with exit 3002 and exit 3015, respectively. These premature terminations can be
avoided by using the option to continue the analysis even if physical criteria are not satisfied. If this flag
is set on the AUTO STEP option, and either the maximum number of user-allowed cutbacks or the
minimum time step is reached, a mechanical analysis moves on to the next increment if it is otherwise
converged (see Convergence Controls in this chapter) or continues to recycle and scales back based
solely on the recycling criterion. Setting this flag for a thermal analysis simply allows it to move on to
the next increment.
User Programmed Time Steps Selection

In addition to controlling the time step through the recycle based criterion and the physical criteria, direct
control of the time step is possible through the use of the UTIMESTEP user subroutine. In this case, the
new time step that is determined by the auto step algorithm enters the program as input and the modified
time step by the user is returned as output. More details are provided in Marc Volume D: User Subroutines
and Special Routines, Chapter 2.
Post Files Output

In many analyses it is convenient to obtain post file results at specified time intervals. This is naturally
obtained with a fixed load stepping scheme but not with an automatic scheme. Traditionally, the post
output frequency is given as every nth increment. With the AUTO STEP history definition option, you can
request post output to be obtained at equally spaced time intervals. In this case, the time step is
temporarily modified to exactly reach the time for output. The time step is then restored in the
following increment.
Table Considerations
When tables are used in the new input format to specify loads with complex time variations, in most
cases, it is important that the exact peaks and valleys of the loading history are not missed due to the
adaptive time stepping. By default, AUTO STEP adjusts the time step temporarily so that the peaks and
valleys of the loading history are reached exactly. The time step is then restored in the following
increment. In analyses where the boundary conditions are obtained from experimental data containing
many points, it is often useful to turn this behavior off, in which case large time steps will be achieved
and the load will be smoothed out.
Quasi-Static Damping Scheme

The AUTO STEP option also has an optional artificial damping feature available for mechanical
statics analyses.
Damping Energy
A comparison between the incremental strain energy and the estimated damping energy is used as a
criteria for time step control when either a 4 or 5 is given on the 10th field of the 2nd data block (idampit
flag).The criteria based upon the number of cycles is bypassed. Decisions to decrease and increase the
time step are based on the damping energy rate of the system. Furthermore, if a 4 is entered, then artificial
damping is added to the system. The details of the time stepping procedure and the damping procedure
are as follows:
CHAPTER 11 721
A damping factor, F d , is introduced, which at the start of the loadcase, is set to 0. The time step for the
first increment is set equal to the user given initial time step. During the first cycle of the loadcase, a
“small” strain energy value ΔE s m a l l is calculated as
ΔE s m a l l =  0.5D ( 1, 1 )ε ( 1 ) 2 dV
V
where D ( 1, 1 ) is the first component of the stress-strain relation at the integration point and E ( 1 ) is
10-6. This small energy value is used later to distinguish between real deformations of the structure and
zero-energy modes.
During the assembly of the stiffness matrix K and the right-hand side vector F , the following
contributions are added to K and F , respectively.
K d a m p = F d M ⁄ timinc
F da m p = F d Mdu ⁄ timinc
where M is the lumped mass matrix, du is the incremental displacement and timinc is the time
increment. Note that when F d = 0 , there is no change to the equations. During the recovery phase, the
incremental damping energy for the nth iteration is calculated as
n
ΔE d a m p = du T Mdu ⁄ timic
when F d = 0 and as
n
ΔE d a m p = F d Mdu ⁄ timinc
when F d is not 0 .
The check to determine if the time-step should be reduced through a cut-back is made as follows: If
F d = 0 , perform a cutback if
n 1
ΔE > ΔE sm a l l and ΔE d m a p > 4ΔE d a m p
where ΔE is the incremental strain energy. The time step reduction factor in this case is given by
1 (n)
fr e d u c = ΔE d a m p ⁄ ΔE d a m p
Such a cutback is triggered in the first increment of the loadcase in the case of severe non-linearities, as
expressed by the fact that the damping energy for the nth iteration deviates considerably from that of the
first iteration.
If F d is not 0 , perform a cutback if

n n
ΔE d a m p > 4ΔE p r e d i c t e d and ΔE d a m p > ΔE m a x and tol > 2
where ΔE p r e d i c t e d is the predicted damping energy for the current increment, ΔE m a x is the maximum
damping energy in any previously converged increment and tol is the global user-defined convergence
tolerance. The reduction factor in this case is set to 0.5. This cut-back is triggered in the case where the
solution is clearly diverging and avoids unnecessary recycles. The same reduction factor (0.5) is also used
for cases when the maximum number of iterations are reached or when user criteria are violated.
For the first increment of the loadcase, the calculation of F d and predicted energy is as follows:
• Estimated strain energy for the loadcase E e s t i m a t e d = ΔE timinc 2
• E m a x = max ( E e st i m a t e d ,E ) , where E = total strain energy in system .
n
• Estimated damping energy for the loadcase E d a m p _ e s t i m a t e d = ΔE d a m p totinc ⁄ timinc ,
where totinc = total time period .
• F d = δE m a x ⁄ E d a m p _ e st i m a t e d where δ is a user-defined factor. Recommended default value
is 10-4.
• Predicted damping energy for next increment
n
ΔE p r e d i c t e d = F d ΔE d a m p timincnew ⁄ timincold
where timincnew is the time step for the next increment and timincold is the time increment
for the previous increment.
n n
• Set ΔE p r e v = F d ΔE d a m p ⁄ timinc and ΔE m a x = F d ΔE d a m p
For subsequent increments of the loadcase, the modification of F d and the time step is as follows:
• If the total strain energy E is larger than twice the estimated strain energy E e s t i m a t e d , then
Fd = Fd E ⁄ Ee s t i m a t e d .
n
• If ΔE d a m p ⁄ timinc < 1.1ΔE p r e v , then increase the time step by a factor of 1.5.
Avoiding Exit 3015

When the minimum time step t m i n is reached, Marc normally exits with 3015. If the quasi-static
damping schemes (10th field, 2nd data block) is a 0 or 1, a quasi-static damping option is triggered in an
attempt to avoid this premature exit: When exit 3015 is encountered for the first time, the increment is
repeated with a new time step given by t n e w = fclrg t m i n , where fclrg is the maximum ratio between
steps (4th field of the 2nd data block of the AUTO STEP option = 10 by default). The increment is further
CHAPTER 11 723
stabilized by added the factored lumped mass matrix to the stiffness matrix and modifying the force
vector consistently. The damping factor associated with the lumped mass matrix is again based on the
ratio of the estimated strain energy to the estimated damping energy in a manner similar to the idampit
= 4 scheme. If exit 3015 is again encountered, a second attempt is made with t n e w = fclrg2 t m i n and
again stabilizing the incremental solution by adding the quasi-static damping in the solution. It should
be noted that this procedure is not available for the first increment of the auto step loadcase since the
damping factor is only calculated after the first increment is completed. It should also be noted that once
exit 3015 is encountered and damping is turned on, it remains on for the rest of the loadcase.
Auto Step for Transient Dynamics

The AUTO STEP algorithm is further modified for transient dynamics problems:
• When penetration is detected in dynamic contact problems, instead of using the default iterative
penetration procedure, a time cutback is made and the increment is repeated with a smaller time
step that avoids the penetration. This scheme allows for momentum conservation without
spurious penetration induced oscillations in the response. If no other cutbacks are made, the time
step is restored in the following increment.
• Additional, optimal checks are available in transient dynamic problems to control the time step
since larger time steps that may have been assessed based on the recycle based criterion or
physical criteria can give rise to unacceptable time integration errors. The time integration error
check is only turned on if the 13th field of the 3rd data block of the AUTO STEP option is set to
1. If this field is not present or set to 0, the time integration error check is skipped and the time
stepping algorithm for dynamic problems is then similar to that of static problems. If the check is
turned on, an additional check is made at the end of each increment to see if the time step needs
to be reduced for the following increment. This check is only made for the Newmark-Beta (NB)
and the Single Step Houbold (SSH) methods. The scheme that is followed is a modified version
of the scheme outlined by Bergan and Mollestad [Ref. 1]. The time step for increment n + 1
should satisfy the inequality
Δu nT MΔu n
( Δt ) n + 1 ≤ 2πα ---------------------------------
Δu nT K T Δu n
where Δu n is the incremental displacement vector at increment n ; M is the mass matrix; K T is

the tangent stiffness matrix, and α is a user-customizable scaling factor. The default value of α
is as follows:
for the SSH operator α = 0.375 * fcsml . and
for the NB operator α = 0.75 * fcsml ,
where fcsml is the smallest ratio between time steps (3rd field of the 2nd data block of the
AUTO STEP option = 0.1 by default). The smaller default value of α = 0.0375 for SSH
problems compared to α = 0.075 for NB problems is to better control the artificial damping
inherent in the SSH method. The user can control the accuracy of the solution by changing
fcsml . The time step is only reduced if the value predicted by the above equation is less than
67% of the current time step. The check is bypassed if t n + 1 is already at t m i n , if the strain
energy is negligible (for example, rigid body motion). When there are multiple loadcases in a
transient dynamic run and the time integration error check is flagged for all of them, the initial
time step of a loadcase that follows a transient dynamic loadcase is also adjusted if it is too high
compared to the time step predicted by the error check. This avoids accuracy problems associated
with using an initial user-prescribed time step that is too large.
The defaults of the AUTO STEP option are carefully chosen to be adequate in a wide variety of
applications. There are cases, however, when the settings may need to be modified. Assume that the
default settings are used, which means that the recycle based control is active with an initial load of one
per cent of the total. If the structure is weakly nonlinear, convergence is obtained in just a few recycles
and the time steps for successive increments get progressively larger. This can lead to problems if the
initially weakly nonlinear structure suddenly exhibits stronger nonlinearities; for instance, occurrence of
plasticity or parts coming into contact. Possible remedies to this problem include:
a. decrease the time step scale factor to a smaller number so the step size does not grow so rapidly;
b. use a physical criterion like maximum increment of displacements to limit the load step;
c. use the maximum time step to limit large steps;
d. decrease the desired and maximum number of recycles to decrease the load step if more
recycles are needed.
Another situation is if the structure is highly nonlinear and convergence is slow. In this case, it may be
necessary to increase the desired number and maximum number of recycles. In general, there is a close
connection between the convergence tolerances used and the desired number and maximum number of
recycles. In many cases, it may be beneficial to use one or more physical criteria; for example, the
increment of plastic strain as targets for controlling the load step. This can easily be achieved by allowing
the program to add automatic physical criteria where appropriate. This is especially a good idea if the
‘proceed if not converged’ option is used or if the ‘non-positive definite flag’ is set since the added
physical criteria then serve as controls to limit the time step and produce a realistic numerical solution in
each increment rather than letting the solution proceed unchecked with unrealistic results.
Residual Load Correction

The residual load is applied as a correcting force to ensure that equilibrium is maintained and, hence, that
an accurate solution is obtained for nonlinear problems.
The residual load correction enforces global equilibrium at the start of each new increment. This prevents
the accumulation of out-of-equilibrium forces from increment to increment and makes the solution less
sensitive to the step size. Figure 11-2 shows how stiffness is based on the state at the start of a step. The
variables are defined below for increments i = 1,2,3:
• F i applied forces for i = 1,2,3
• u i calculated displacements for i = 1,2,3
• R i residual loads for i = 1,2,3
CHAPTER 11 725
F3
F2
R3
F1 R2
R1
U1 U2 U3
Figure 11-2 Stiffness Based on State at Start of Step
The residual load correction is the difference between the internal forces and the externally applied loads.
The residual load correction is expressed as
R = P –  β T σdV (11-2)
where β is the differential operator which transforms displacements to strains, σ is the current
generalized stresses, P is the total applied load vector, and R is the residual load correction.
In order to evaluate the residual load correction accurately, evaluate the integral by summing the
contributions from all integration points. The residual load correction feature requires that stresses be
stored at all the integration points. Data storage at all integration points is the default in Marc, but can be
overridden in linear analysis by use of the CENTROID parameter.
Residual load correction should always be used unless the ELASTIC parameter is invoked; this is
the default.
Restarting the Analysis

The RESTART model definition option creates a restart file for the current analysis which can be used in
subsequent analyses. It can also be used to read in a previously generated file to continue the analysis.
The RESTART option is very important for any multi-increment analysis because it allows you to
continue the analysis at a later time. The default situation writes the restart information to unit 8 and reads
a previously generated file from unit 9. For post processing, the RESTART option can be used to plot or
combine load cases (see CASE COMBIN). Upon restart, you can use the model definition REAUTO
option to redefine parameters associated with an automatic load sequence.
To save storage space, it is not necessary to store each increment of analysis. The frequency can be set
using the RESTART option, and subsequently modified using the RESTART INCREMENT option. It is
also possible to store only the last converged solution by using the RESTART LAST option. The input
data describing the problem is not saved, and therefore must be read in with each restart. This should
include the model geometry including CONNECTIVITY and the COORDINATES material properties
contact data and boundary conditions. Additional tables may be added, or they can have additional data
added. The RESTART option specifies restart parameters; for example, input/output files, restart
increment, and intervals at which restarts are to be written.
Newton-Raphson Method
The Newton-Raphson method can be used to solve the nonlinear equilibrium equations in structural
analysis by considering the following set of equations:
K ( u )δu = F – R ( u ) (11-3)
where u is the nodal-displacement vector, F is the external nodal-load vector, R is the internal nodal-
load vector (following from the internal stresses), and K is the tangent-stiffness matrix. The internal
nodal-load vector is obtained from the internal stresses as
T
R =  β σ dv (11-4)
elem V
In this set of equations, both R and K are functions of u . In many cases, F is also a function of u (for
example, if F follows from pressure loads, the nodal load vector is a function of the orientation of the
structure). The equations suggest that use of the Newton-Raphson method is appropriate. Suppose that
i i
the last obtained approximate solution is termed δu , where indicates the iteration number. Equation
(11-3) can then be written as
i–1 i–1
K ( u n + 1 )δu = F – R ( u n + 1 ) (11-5)
i
This equation is solved for δu and the next appropriate solution is obtained by
i i–1 i i i
Δu = Δu + δu and u n + 1 = u n + Δu (11-6)
Solution of this equation completes one iteration, and the process can be repeated. The subscript n
denotes the increment number representing the state t = n . Unless stated otherwise, the subscript
n + 1 is dropped with all quantities referring to the current state.
The Newton-Raphson method is the default in Marc (see Figure 11-3). The Newton-Raphson method
provides good results for most nonlinear problems, but is expensive for large, three-dimensional
problems, when the direct solver is used. The computational problem is less significant when the iterative
solvers are used. Figure 11-3 illustrates the graphical interpretation of the Newton-Raphson iteration
technique in one dimension to find the roots of the function F ( u ) – 1 = u – 1 = 0 , starting from
increment 1 where F ( u 0 ) = 0.2 to increment 2 where. F ( u l a s t ) = 1.0 .The iteration process stops
when the convergence criteria are satisfied. Similarly, Figure 11-4 graphically demonstrates the modified
Newton-Raphson iteration technique for the same one dimensional problem.
CHAPTER 11 727
Newton-Raphson Method
1.2
F = u1/2 Convergence Criteria:
1 Inc 2
Residual Checking
Residual 1.001
|| Fr esidua l ||
0.8 1
---------------------------- < TOL 1
8
1
0.999
4 || Fr ea ction ||
8
Force
0.998
0.997
0.996
0.6 0.995
3
0.992 0.993 0.994 0.995 0.996 0.997 0.998 0.999 1 1.001
Displacement Checking
0.4 || δu ||
--------------- < TO L1
8
Δu1 δu2 δu3
|| Δu ||
8
Recycle 1 2 3
0.2 Inc 1
Displacement (u)
0
0 0.5 1 1.5
Figure 11-3 Newton-Raphson
Modified Newton-Raphson Method

The modified Newton-Raphson method is similar to the Newton-Raphson method, but does not
reassemble the stiffness matrix in each iteration.
0 i i–1
K ( u )δu = F – R ( u ) (11-7)
Modified Newton-Raphson Method

1.2
1/2
F=u Convergence Criteria:
1 Inc 2
Residual Checking
0.8
Residual
1
|| Fr esidua l ||
---------------------------- < TOL 1
1.02
8
1.01
|| Fr ea ction ||
Force
0.99
0.6 0.98
0.97
0.97 0.98 0.99 1 1.01 1.02 1.03 Displacement Checking
0.4 || δu ||
δu3
--------------- < TO L1
8
Δu1 δu2
|| Δu ||
8
Recycle 1 2 3
0.2 Inc 1
u0
Displacement (u)
0
0 0.5 1 1.5
Figure 11-4 Modified Newton-Raphson
The process is computationally inexpensive because the tangent stiffness matrix is formed and
decomposed once. From then on, each iteration requires only forming the right-hand side and a backward
substitution in the solution process. However, the convergence is only linear, and the potential for a very
large number of iterations, or even nonconvergence, is quite high.
If contact or sudden material nonlinearities occur, reassembly cannot be avoided. The modified Newton-
Raphson method is effective for large-scale, only mildly nonlinear problems. When the iterative solver
is employed, simple back substitution is not possible, making this process ineffective. In such cases, the
Newton-Raphson method should be used instead.
If the load is applied incrementally, Marc recalculates the stiffness matrix at the start of each increment
or at selected increments, as specified.
Strain Correction Method

The strain correction method is a variant of the Newton method. This method uses a linearized strain
calculation, with the nonlinear portion of the strain increment applied as an initial strain increment in
subsequent iterations and recycles. This method is appropriate for shell and beam problems in which
rotations are large, but membrane stresses are small.
In such cases, rotation increments are usually much larger than the strain increments, and, hence, the
i+1
nonlinear terms can dominate the linear terms. After each displacement update, the new strains E α β
i i
are calculated from u and δu ( = δ u ) which yield
i+1 i 1 i
Eα β = E α β + --- ( δ u α, β + δu β, α ) + u κ, α δu κ, β + δu κ, α u κi β + δu κ, α δu κ, α (11-8)
2
This expression is linear except for the last term. Since the iteration procedures start with a fully
linearized calculation of the displacement increments, the nonlinear contributions yield strain increments
inconsistent with the calculated displacement increments in the first iteration. These errors give rise to
either incorrect plasticity calculations (when using small strain plasticity method), or, in the case of
elastic material behavior, yields erroneous stresses. These stresses, in their turn, have a dominant effect
on the stiffness matrix for subsequent iterations or increments, which then causes the relatively poor
performance.
The remedy to this problem is simple and effective. The linear and nonlinear part of the strain increments
are calculated separately and only the linear part of
l i 1 i i
( E αβ ) = E α β + --- ( δu α, β + δu β, α ) + u κ, α δu κ, β + δu κ, α u κ β (11-9)
2
is used for calculation of the stresses. The nonlinear part
nl i+1 1
( E αβ ) = --- δu κ, α δu κ, β (11-10)
2
is used as an “initial strain” in the next iteration or increment, which contributes to the residual load
vector defined by
C αβ γ δ nl
R =  δ κ, β X κ, α L ΔE γ δ dV (11-11)
V0
This “strain correction” term is defined by

i i C
K ( u n + 1 )δu = F – R ( u n + 1 ) – R (11-12)
CHAPTER 11 729
Since the displacement and strain increments are now calculated in a consistent way, the plasticity and/or
equilibrium errors are greatly reduced. The performance of the strain correction method is not as good if
the displacement increments are (almost) completely prescribed, which is not usually the case. Finally,
note that the strain correction method can be considered as a Newton method in which a different
stiffness matrix is used.
Direct Substitution
In the Eulerian formulation (R-P FLOW parameter), the governing equation of the system can be
expressed as
Kv = F (11-13)
where v is a velocity vector, and F is a force vector. This equation is very nonlinear because K is a
i
nonlinear function of v . By default, a direct substitution method is used to solve the problem. If v is
the velocity at iteration i , the result of iteration i + 1 is
i i+1
K( v ) v = F (11-14)
If this method does not converge in 10 iterations, it is possible to switch into a Newton-Raphson method.
Arc-length Methods
The solution methods described above involve an iterative process to achieve equilibrium for a fixed
increment of load. Besides, none of them have the ability to deal with problems involving
snap-through and snap-back behavior. An equilibrium path as shown in Figure 11-5 displays the features
possibly involved.
Snap-through Behavior
Load (F)
um Path
uilibri
r : Eq
6
2
5 Displacement (u)
Figure 11-5 Snap-through Behavior

The issue at hand is the existence of multiple displacement vectors, u , for a given applied force vector,
F . The arc-length methods provide the means to ensure that the correct displacement vector is found by
Marc. If you have a load controlled problem, the solution tends to jump from point 2 to 6 whenever the
load increment after 2 is applied. If you have a displacement controlled problem, the solution tends to
jump from 3 to 5 whenever the displacement increment after 3 is applied. Note that these problems appear
essentially in quasi-static analyses. In dynamic analyses, the inertia forces help determine equilibrium in
a snap-through problem.
Thus, in a quasi-static analysis sometimes it is impossible to find a converged solution for a particular
load (or displacement increment):
λ n + 1 F – λ n F = ΔλF
This is illustrated in Figure 11-5 where both the phenomenon of snap-through (going from point 2 to 3)
and snap-back (going from point 3 to 4) require a solution procedure which can handle these problems
without going back along the same equilibrium curve.
As shown in Figure 11-6, assume that the solution is known at point A for load level λ n F . For arriving
at point B on the equilibrium curve, you either reduce the step size or adapt the load level in the iteration
process. To achieve this end, the equilibrium equations are augmented with a constraint equation
expressed typically as the norm of incremental displacements. Hence, this allows the load level to change
from iteration to iteration until equilibrium is found.
g : Co Arc- Length Procedure
nstr
ain
λn+1 F t Su
r fa
λn+2 F ce r : Equilibri
λn+3 F um
Path
Load (F)
th
rc leng
l: a
λn F
A
Displacement (u)
Figure 11-6 Intersection of Equilibrium Path with Constraining Spherical Surface
The augmented equation, c ( u, λ ) , describes the intersection of the equilibrium curve with an auxiliary
surface g for a particular size of the path parameter η :
r ( u, λ ) = λF – R ( u ) = 0 c ( u, λ ) = g ( u, λ ) – Δη = 0 (11-15)
CHAPTER 11 731
Variations of the parameter η moves the surface whose intersection with the equilibrium curve r
generates a sequence of points along the curve. The distance between two intersection points, denoted
with η 0 and η , denoted by l is the so-called arc-length.
Linearization of Equation (11-15) around point A in Figure 11-6 yields:
K P  δu   –r 
  =  –r  (11-16)
n n 0  δλ 
T
 0
where:
∂r ∂r
K = ------ : P = ------ (11-17)
∂u ∂λ
T ∂c ∂c
n = ------ : n 0 = ------ (11-18)
∂u ∂λ
r = λF – R (11-19)
r 0 = g ( u, λ ) – Δη (11-20)
It can be noted that a standard Newton-Raphson solution procedure is obtained if the constraint condition
is not imposed. The use of the constraint equation causes a loss of the banded system of equations which
would have been obtained if only the K matrix was used. Instead of solving the N + 1 set of equations
iteratively, the block elimination process is applied.
i–1
Consider the residual at iteration i to which the fraction of load level λ corresponds
i i–1 i–1 i i–1

r (λ ) = λ F – R (u ) (11-21)
i
The residual for some variation of load level, δλ , becomes
i i–1 i i i i–1
r (λ + δλ ) = δλ F + r ( λ ) (11-22)
which can be written as:
i i–1 i i i–1 i i
δu ( λ + δλ ) = δu ( λ ) + δλ δu * (11-23)
i i–1 i –1
where δu ( λ ) = (K ) r (11-24)
i i –1
and δu * = ( K ) F (11-25)
i
Notice that δu * does not depend on the load level. The equation above essentially establishes the
i
influence of a change in the load level δλ during one iteration on the change in displacement increment
for that iteration. After one iteration is solved, this equation is used to determine the change in the
load level such that the constraint is followed. There are several arc-length methods corresponding to
different constraints.
11 Among them, the most well-known arc-length method is one proposed by Crisfield, in which the iterative
solution in displacement space follows a spherical path centered around the beginning of the increment.
This requirement is translated in the formula:
2 i i
Solution c = l = Δu Δu (11-26)
Procedu
where l is the arc length. The above equation with the help of Equations (11-25) and (11-26) is applied as:
res for
Nonlinea i T i i 2 i–1 i i–1 T i i
r [ ( δu * ) δu * ] ( δλ ) + [ 2 ( Δu + δu ( λ ) ) δu * ] ( δλ ) +
Systems (11-27)
i–1 i i–1 T i–1 i i–1 2
[ ( Δu + δu ( λ ) ) ( Δu + δu ( λ )) – l ] = 0
1
The equation above is interpreted with i = 1 and δu = 0 in the prediction phase while retaining the
full form of Equation (11-27) in the correction phase. Two solutions for δλ are available. There are
several methods to determine which root to select. By default, we choose the one that maintains a positive
angle of the displacement increment from one iteration to the next.
i i
The two roots of this scalar equation are ( δλ ) 1 and ( δλ ) 2 . To avoid going back on the original load-
i–1 i
deflection curve, the angle between the incremental displacement vectors, Δu and Δu (before and
i
after the current iteration, respectively) should be positive. Two alternative values of Δu (namely,
i i i i
( Δu ) 1 and ( Δu ) 2 corresponding to ( δλ ) 1 and ( δλ ) 2 are obtained and the cosine of two
corresponding angles ( φ 1 and φ 2 ) are given by
i T i–1
[ ( Δu n + 1 ) 1 ] Δu n + 1
cos φ 1 = ----------------------------------------------------
- (11-28)
l
i T i–1
[ ( Δu n + 1 ) 2 ] Δu n + 1
and cos φ 2 = ----------------------------------------------------
- (11-29)
l
0
Once again, the prediction phase is interpreted with i = 1 and Δu n + 1 = Δu n , while Equations
(11-28) and (11-29) retain their full form in the correction phase.
CHAPTER 11 733
i i
As mentioned earlier, the appropriate root, ( δλ ) 1 or ( δλ ) 2 is that which gives a positive cos φ . In case
both the angles are positive, the appropriate root is the one closest to the linear solution given as:
i–1 i i–1 i 2
i ( Δu + δu ) ( Δu + δu ) – l
δλ = -------------------------------------------------------------------------------------- (11-30)
i–1 i i
2 ( Δu + δu )δu *
Crisfield’s solution procedure, generalized to an automatic load incrementation process, has been
implemented in Marc as one of the options under the AUTO INCREMENT history definition option.
Various components of this process are shown in Figure 11-7.
λ1 F
Load (F)
F2 = λ2 F
r1
B
δu2(λ1) = K2-1(λ1F - r1)
Displacement (u)
Figure 11-7 Crisfield’s Constant Arc Length
The constraints in Equations (11-26) and (11-27) are imposed at every iteration. Disadvantage of the
quadratic equation suggested by Crisfield is the introduction of an equation with two roots and thus the
need for an extra equation to solve the system for the calculated roots if two real roots exists. This
1 1
situation arises when the contribution Δu (or δu ) is very large in comparison to the arc-length. This
can be avoided in most cases by setting sufficiently small values of the error tolerance on the residual
force. In case the above situation still persists despite the reduction of error tolerance, Marc has two
options to proceed:
a. To attempt to continue the analysis with the load increment used in the initial step of auto
increment process.
b. Use the increment resulting from the linear constraint for the load.
There are two alternate methods using the Crisfield method for selecting the roots:
• Singularity ratio method; where if the system of equations is positive definite, the largest root is
used; while if the system was negative, the smallest root is used.
• Falzon’s method described in Reference 7 is applicable for problems which have multiple nearly
equal buckling modes.
This is circumvented in Ramm’s procedure due to the linearization.
Another approach to impose the constraint is due to Ramm, who also makes use of a quadratic equation
to impose the constraint giving rise to the Riks-Ramm method. The difference is that while Crisfield
imposes the constraint as a quadratic equation, Ramm linearized the constraint.
Geometrically, the difference between the two methods is that the Crisfield method enforces the
correction on the curve of the augmented equation introducing no residual for the augmented equation.
Ramm takes the intersection between the linearizations of the curves which gives a residual of the
augmented equation for the next step. Both methods converge to the same solution, the intersection of
the two curves, unless approximations are made.
The Riks-Ramm constraint is linear, in that:
2
c = l = Δu n Δu n + 1
which results in a linear equation for δλ :

T i i 2
Δu n ( δu + δλ δu * ) = l
Thus, the load parameter predictor is calculated as:
T i –1 i
1 Δu n l – ( Δu n ) [ ( K ) r ]
δλ n + 1 = ---------------------------------------------------------------------- (11-31)
T 1
Δu n ( δu * )
while during the corrector phase it is:
i T i –1 i
i ( Δu n + 1 ) [ ( K ) r ]
δλ n + 1 = – ----------------------------------------------------
i
(11-32)
i T
( Δu n + 1 ) ( δu * )
It is noted that in the definition of the constraint, the normalized displacement of the previous step is used
∂c
for the normal to the auxiliary surface ------ = n . Thus, problems can arise if the step size is too big. In
∂u
situations with sharp curvatures in the solution path, the normal to the prediction may not find
intersections with the equilibrium curve. Note that the norm of the displacement increment during the
iterations is not constant in Riks-Ramm method.
In contact problems, sudden changes of the stiffness can be present (due to two bodies which are initially
not in contact suddenly make contact). Hence, a potential problem exists in the Riks-Ramm method if
i
the inner-product of the displacement due to the load vector δu * and the displacement increment Δu n
is small. This could result in a very large value of the load increment for which convergence in the
subsequent iterations is difficult to achieve. Therefore, a modified predictor can be used resulting in a
modified Riks-Ramm procedure as:
1 1 T 1
1 l n – 1 δu – [ αδu * ] δu
Δλ *
= -----------------------------------------------------------------
- (11-33)
1 1
[ αδΔu * ]Δu *
CHAPTER 11 735
where
T i
Δu n δu *
α = -------------------------
- (11-34)
T i
Δu n δu *
This method effectively scales the load increment to be applied in the prediction and is found to be
effective for contact problems.
Refinements and Controls

The success of the methods depend on the suitable choice of the arc-length:
2
C = l
The initial value of the arc-length is calculated from the initial fraction β of the load specified by you in
the following fashion:
Kδu = βF – R (11-35)
2
li n i = Δu (11-36)
In subsequent steps the arc-length can be reduced or increased at the start of a new load step depending
on the number of iterations I 0 in the previous step. This number of iterations in compared with the
desired number of iterations I d which is typically set to 3 or 5. The new arc-length is then given by:
rt = I d ⁄ I o (11-37)
2 2
l new = rt ⋅ l p r e v
Often, this results in too large of an arc-length growth, especially initially when the system is nearly
linear and I o is small (often one). In this case, one can gradually grow the value as:
l n e w = sf ⋅ l p re v
sf = .1 if rt ≤ .1
sf = 1 if .1 < rt < 1
sf = 1.25 if 1 < rt ≤ 3
sf = 1.5 if 3 ≤ rt
Two control parameters exist to limit the maximum enlargement or the minimum reduction in the
arc-length.
2
l
min < ------- < max (11-38)
2
l ini
The min and max defined here are the 7th and 5th field of the 2nd data block. It is also possible that the
maximum fraction of the load allowed can vary during the loadcase. This is often useful if one knows
that one is approaching the limit load.
In addition, the maximum fraction of the total load may be defined. In general, control on the limiting
values with respect to the arc-length multiplier is preferred in comparison with the maximum fraction of
the load to be applied in the iteration since a solution is sought for a particular value of the arc-length.
Also, attention must be paid to the following:
1. In order to tract snap-through problems, the method of allowing solution if the stiffness matrix
becomes nonpositive needs to be set.
2. The maximum number of iterations must be set larger than the desired number of iterations.
The default procedure for convergence criterion in Marc is based on the magnitude of the maximum
residual load compared to the maximum reaction force. This method is appropriate since the residuals
measure the out-of-equilibrium force, which should be minimized. This technique is also appropriate for
Newton methods, where zero-load iterations reduce the residual load. The method has the additional
benefit that convergence can be satisfied without iteration.
The basic procedures are outlined below.
1. RESIDUAL CHECKING
F r e si d u a l ∞
------------------------------- < TOL 1 (11-39)
F re a c t i o n ∞
F r e si d u a l ∞ M r e si d u a l ∞
------------------------------- < TOL 1 and --------------------------------- < TOL 2 (11-40)
F re a c t i o n ∞ M re a c t i o n ∞
< TOL 1 (11-41)
< TOL 1 and M r e si d u a l ∞
< TOL 2 (11-42)
Where F is the force vector, and M is the moment vector. TOL 1 and TOL 2 are control
tolerances. F ∞ indicates the component of F with the highest absolute value. Residual
checking has two drawbacks. First, if the CENTROID parameter is used, the residuals and
reactions are not calculated accurately. Second, in some special problems, such as free thermal
expansion, there are no reaction forces. The program uses displacement checking in either of these
cases.
CHAPTER 11 737
11 2. DISPLACEMENT CHECKING
δu ∞
Solution - < TOL 1
--------------- (11-43)
Δu ∞
Procedu
res for
Nonlinea δu ∞ δφ ∞
- < TOL 1 and ---------------
--------------- - < TOL 2 (11-44)
r Δu ∞ Δφ ∞
Systems
δu ∞ < TOL 1 (11-45)
δu ∞ < TOL 1 and δφ ∞ < TOL 2 (11-46)
where Δu is the displacement increment vector, δu is the correction to incremental displacement

vector, Δφ is the correction to incremental rotation vector, and δφ is the rotation iteration vector.
With this method, convergence is satisfied if the maximum displacement of the last iteration is
small compared to the actual displacement change of the increment. A disadvantage of this
approach is that it results in at least one iteration, regardless of the accuracy of the solution.
δi Correction to incremental displacements of ith iteration

F un Displacements at increment n
δ0 i
δ1 δ -
--------------- ≤ Tolerance
δ k i
 δj
0 j = 0
k+1
u u u
n+1 n+1
Figure 11-8 Displacement Control
3. STRAIN ENERGY CHECKING

This is similar to displacement testing where a comparison is made between the strain energy of
the latest iteration and the strain energy of the increment. With this method, the entire model is
checked.
δE
------- < TOL 1 (11-47)
ΔE
where ΔE is the strain energy of the increment and δE is the correction to incremental strain
energy of the iteration. These energies are the total energies, integrated over the whole volume.
A disadvantage of this approach is that it results in at least one iteration, regardless of the accuracy
of the solution. The advantage of this method is that it evaluates the global accuracy as opposed
to the local accuracy associated with a single node.
Different problems require different schemes to detect the convergence efficiently and accurately.
To do this, the following combinations of residual checking and displacement checking are also
available.
4. RESIDUAL OR DISPLACEMENT CHECKING

This procedure does convergence checking on both residuals (Procedure 1) and displacements
(Procedure 2). Convergence is obtained if one converges.
5. RESIDUAL AND DISPLACEMENT CHECKING
This procedure does a convergence check on both residuals and displacements (Procedure 4).
Convergence is achieved if both criteria converge simultaneously.
In several types of analyses, maximum reactions or displacements are extremely small (even close to the
round-off errors of computers). In such circumstances, not all types of relative convergence criteria may
work properly. For example, in a problem with stress-free motion, the convergence check based on
relative displacement increments works correctly but not the convergence check based on relative
residual or strain energy. In this situation, it is necessary to check the convergence with absolute values
of reactions or strain energy; otherwise, the analysis may terminate prematurely. Similarly, this kind of
situation may happen for problems with springback and free thermal expansion or constraint thermal
expansion. The details for the cases where convergence checking with relative values may encounter
difficulties are listed in Table 11-2. The AUTO SWITCH option is designed for Marc to switch to the
proper convergence check scheme automatically if any of the situations mentioned above occur during
the analysis. The AUTO SWITCH option allows Marc to automatically switch the convergence check
scheme to check as required on either residuals or displacements if small reactions or displacements are
detected, or to use the absolute strain energy checking if necessary.
Table 11-2 Effectiveness of various Relative Tolerance Convergence Testing Criterion
Convergence Variable
Displacement/ Residual Strain Energy
Analysis Type Rotation Force/Torque
Stress-free motion Yes No No
Springback No Yes No
Free Thermal Expansion Yes No No
Constraint Thermal Expansion No Yes Yes
Yes – relative tolerance testing works.
No – relative tolerance testing does not work.
If AUTO SWITCH is turned on, it:

1. Switches on the relative residual checking if the relative displacement criterion is used (which
fails when the maximum incremental displacement becomes very small
Max._incremental_displacement/Smallest_element_size<1.0e-6)
2. Switches on the relative displacement checking if the relative residual force (moment) criterion is
used (which fails when the maximum reaction force becomes very small <1.0e-8)
3. To switch on the absolute energy checking if the structure is free of stress and deformation
(strain energy density < 1.0e-15).
CHAPTER 11 739
Notes: 1. If any criterion with absolute value testing is being used as convergence checking,
this feature is deactivated.
2. If the combined convergence check scheme with relative displacements or residuals
is chosen as convergence criterion, it is not necessary to turn the AUTO SWITCH on.
3. Once the extreme cases disappear in the analysis, Marc recovers to use the
convergence criteria given by user.
Singularity Ratio
The singularity ratio, R , is a measure of the conditioning of the system of linear equations. R is related
to the conditioning number, C , which is defined as the ratio between the highest and lowest eigenvalues
in the system. The singularity ratio is an upper bound for the inverse of the matrix conditioning number.
1⁄R≤C (11-48)
C and R establish the growth of errors in the solution process. If the errors on the right-hand side of the
equation are less than E prior to the solution, the errors in the solution will be less than δ , with
δ ≤ CE (11-49)
The singularity ratio is a measure that is computed during the Crout elimination process of Marc using
the direct solver. In this process, a recursive algorithm redefines the diagonal terms
k–1
(k) (k – 1)
Kk k = Kk k –  Km k Km k 1 ≤i≤k–1 (11-50)
m = i
where i is a function of the matrix profile. K k k is a diagonal of the kth degree of freedom. The
singularity ratio is defined as
(k) (k – 1)
R = min K k k ⁄ K k k (11-51)
(k) (k – 1)
If all K k k and K k k are positive, the singularity ratio indicates loss of accuracy during the Crout
elimination process. This loss of accuracy occurs for all positive definite matrices. The number of digits
lost during the elimination process is approximately equal to
n l o s t = – log 10 R (11-52)
The singularity ratio also indicates the presence of rigid body modes in the structure. In that case, the
(k)
elimination process produces zeros on the diagonal K k k ≅ 0 . Exact zeros never appear because of
numerical error; therefore, the singularity ratio is of the order
– nd i g i t
R = O  10  (11-53)
 
where n digit is the accuracy of floating-point numbers used in the calculation. For most versions of Marc,
(k)
n digit > 12 . If rigid body modes are present, K k k is very small or negative. If either a zero or a negative
diagonal is encountered, execution of Marc is terminated because the matrix is diagnosed as
being singular.
You can force the solution of a nonpositive definite or singular matrix. In this case, Marc does not stop
(k)
when it encounters a negative or small term K k k on the diagonal. If you use Lagrangian multiplier
elements, the matrix becomes nonpositive definite and Marc automatically disables the test on the sign
(k)
of K k k . However, it still tests on singular behavior.
Note: The correctness of a solution obtained for a linearized set of equations in a nonpositive
definite system is not guaranteed.
AutoSPC
The autospc capability, which can be activated either on the AUTOSPC parameter or on the SOLVER
option, attempts to eliminate rigid body motion by constraining degrees of freedom. This procedure
occurs if the singularity ratio is smaller than 1.e-11. This option is only available with the direct solution
procedure. This is different than setting the non-positive definite flag, which puts a small number into the
stiffness matrix.
Solution of Linear Equations

The finite element formulation leads to a set of linear equations. The solution is obtained through
numerically inverting the system. Because of the wide range of problems encountered with Marc, there
are several solution procedures available.
Most analyses result in a system which is real, symmetric, and positive definite. While this is true for
linear structural problems, assuming adequate boundary conditions, it is not true for all analyses. Marc
has two main modes of solvers – direct and iterative. Each of these modes has two families of solvers,
based upon the storage procedure. While all of these solvers can be used if there is adequate memory,
only a subset uses spill logic for an out-of-core solution. Finally, there are classifications based upon
nonsymmetric and complex systems. This is summarized in the following table:
CHAPTER 11 741
Table 11-3 Solver Behavior

Vendor
Direct Iterative Direct Provided Multifrontal
Profile Sparse Sparse Sparse* Sparse
Solver Option 0 2 4 6 8
Real Symmetric Yes Yes Yes Yes Yes
Real Nonsymmetric Yes No No No Yes
Complex Symmetric Yes No No No Yes
Complex nonsymmetric No Yes No No Yes
Out-of-core Yes Yes Yes SGI only Yes
Possible problem with poorly No Yes No No No
conditioned systems
Solver Parallelization No No No Yes Yes**
Memory Architecture N/A N/A N/A Shared Shared
Available with DDM Yes Yes Yes Yes Yes
Mixed Pardiso
Direct Direct
CASI Iterative Iterative Sparse** MUMPS Direct**
Solver Option 9 10 11 12
Real Symmetric Yes Yes Yes Yes
Real Nonsymmetric No No Yes Yes
Complex Symmetric No No Yes No
Complex nonsymmetric No No Yes Yes
Out-of-core Yes No No No
Possible problem with poorly No No Yes Yes
conditioned systems
Solver Parallelization Pre-release No Yes Yes
Memory Architecture Shared or Distributed N/A Shared Shared or Distributed
Available with DDM Pre-release No Yes Yes
*Available for SGI platforms only.

**Available for Win32, Win64, and Linux 64 platforms only.
The choice of the solution procedure is made through the SOLVER option.
Direct Methods
Traditionally, the solution of a system of linear equations was accomplished using direct solution
procedures, such as Cholesky decomposition and the Crout reduction method. These methods are usually
reliable, in that they give accurate results for virtually all problems at a predictable cost. For positive
definite systems, there are no computational difficulties. For poorly conditioned systems, however, the
results can degenerate but the cost remains the same. The problem with these direct methods is that a
large amount of memory (or disk space) is required, and the computational costs become very large.
This version of MUMPS is based on public domain software developed during the Esprit IVEuropean
project PARASOL (1996-1999) by CERFACS, ENSEEIHT-IRIT and RAL. Since this first public
domain version in 1999, the developments are supported by the following institutions: CERFACS,
ENSEEIHT-IRIT, and INRIA. [Refs. 8., 9., and 10.]
Solver 8 is the preferred direct solver. If a multiprocessor machine is available, then Solver 11 or Solver
12 is recommended.
Iterative Methods
Marc offers iterative solvers as a viable alternative for the solution of large systems. These iterative
methods are based on preconditioned conjugate gradient methods. The single biggest advantage of these
iterative methods is that they allow the solution of very large systems at a reduced computational cost.
This is true regardless of the hardware configuration. The disadvantage of these methods is that the
solution time is dependent not only upon the size of the problem, but also the numerical conditioning of
the system. A poorly conditioned system leads to slow convergence – hence increased computation costs.
When discussing iterative solvers, two related concepts are introduced: fractal dimension, and
conditioning number. Both are mathematical concepts, although the fractal dimension is a simpler
physical concept. The fractal dimension, the range of which is between 1 and 3, is a measure of the
“chunkiness” of the system. For instance, a beam has a fractal dimension of 1, while a cube has a fractal
dimension of 3.
The conditioning number is related to the ratio of the lowest to the highest eigenvalues of the
system. This number is also related to the singularity ratio, which has been traditionally reported in
the Marc output when using a direct solution procedure. In problems involving beams or shells, the
conditioning number is typically small, because of the large differences between the membrane and
bending stiffnesses.
The CASI iterative solver is preferred.
Mixed Direct-Iterative Solver

The mixed direct iterative solver attempts to combine the advantages of both solution methods. In the
first iteration of an increment, the Multifrontal Sparse Solver (type 8) is used to obtain a solution while
simultaneously creating the Cholesky pre-conditioner. In subsequent iterations, the sparse iterative solver
is used with the previously calculated preconditioner. If solver convergence is not being obtained at a fast
enough rate, the program switches back to the direct solver. This procedure has been shown to be
effective for contact problems where small deformation or inelastic problems occur. This solver is not
supported with DDM.
CHAPTER 11 743
Preconditioners
The choice of preconditioner can substantially improve the conditioning of the system, which in turn
reduces the number of iterations required. While all positive definite systems with N degrees of freedom
converges in N iterations, a well conditioned system typically converges in less than the square root of
N iterations.
The available preconditioners in the sparse iterative solver are diagonal, scaled diagonal, and
incomplete Cholesky.
The available preconditioners in the CASI sparse iterative solver are: CASI Primal and CASI Standard.
The iterative solvers require an error criteria to determine when convergence occurs. The default is to
use an error criteria based upon the ratio between the residuals in the solution and the reaction force.
c
After obtaining the solution of the linear equations u evaluate:
C C
Ku = F (11-54)
The residual from the solution procedure is:
A C A C
Res = F – F = F – Ku (11-55)
If the system is linear ( K does not change) and exact numerics are preformed, then Res = 0 .
Because this is an iterative method the residual is nonzero, but reduces in size with further iterations.
Convergence is obtained when
Res ⁄ Reac < TOL (11-56)

The tolerance is specified through the SOLVER option.
Storage Methods
In general, a system of linear equations with N unknowns is represented by a matrix of size N by N , or
2
N variables. Fortunately, in finite element or finite difference analyses, the system is “banded” and not
all of the entries need to be stored. This substantially reduces the memory (storage) requirements as well
as the computational costs.
In the finite element method, additional zeroes often exist in the system, which results in a partially full
bandwidth. This profile storage method is used in Marc to store the stiffness matrix when solver type 0
is chosen. When many zeroes exist within the bandwidth, the sparse storage methods can be quite
advantageous. Such techniques do not store the zeroes, but require additional memory to store the
locations of the nonzero values. You can determine the “sparsity” of the system (before decomposition)
by examining the statements:
“Number of nodal entries excluding fill in” x
“Number of nodal entries including fill in” y
If the ratio ( x ⁄ y ) is large, then the sparse matrix storage procedure is advantageous.
Nonsymmetric Systems
The following analyses types result in nonsymmetric systems of equations:
Inclusion of convective terms in heat transfer analysis
Coriolis effects in transient dynamic analysis
Fluid mechanics
Steady state rolling
Soil analysis
Follower force stiffness
Frictional contact
The first four always result in a nonsymmetric system. The last three can be solved either fully using the
nonsymmetric solver, or (approximately) using a symmetric solver. The nonsymmetric problem uses
approximately twice as much memory for storing the stiffness matrix.
Complex Systems
Marc utilizes a complex operator matrix for dynamic harmonic analyses when the damping matrix is
present or for harmonic electromagnetic, acoustic, or piezoelectric analyses. The matrix is
always symmetric.
Iterative Solvers
In Marc, an iterative sparse solver is available using a sparse matrix technique. This method is
advantageous for different classes of problems.
There exist certain types of analyses for which the sparse iterative solver, CASI iterative or Mixed
direct-iterative solvers are not appropriate. These types include:
elastic analysis
explicit creep analysis
complex harmonic analysis
superelements/substructures
central difference techniques
eigenvalue analysis
use of gap elements
auto increment loadcases
If the CASI solver is used with an option that is not supported, the program will switch to the direct solver
type 8.
Elastic or explicit creep analysis involves repeated solutions using different load vectors. When a direct
solver is used, this is performed very efficiently using back substitution. However, when an iterative
solver is used, the stiffness matrix is never inverted, and the solution associated with a new load vector
requires a complete re-solution.
The sparse iterative solver can exhibit poor convergence when shell elements or Herrmann
incompressible elements are present.
CHAPTER 11 745
The CASI iterative solver should not be used with higher order Herrmann elements which includes
element types 155, 156, and 157.
If the mixed direct-iterative solver is used and convergence is poor, Marc switches to the multifrontal
sparse direct solver.
Parallel Direct Solvers

The solution of the linear equations may be solved using multi-processors using the hardware provided
solver, the multifrontal solver, the Pardiso solver, or the MUMPS solver depending on the hardware
environment as shown in Table 11-3. For the case of the hardware, multifrontal and Pardiso solver, the
number of processors used is controlled by the -nthread ntx command when submitting the job. Here,
ntx is the number of threads. For a shared memory machine with a single chip but with four cores, one
would use -nthread 4 to reduce the amount of wall time consumed. When using a distributed memory
machine or cluster, one would use the MUMPS solver and the -nsolver nsx control line argument.
Where nsx is the number of processors/threads to be used. If network computing is used, then a host file
is required as well.
Basic Theory
A linear finite element system is expressed as:
Ku = F (11-57)
And a nonlinear system is expressed as:
T
K Δu = F – R = r (11-58)
T
where K is the elastic stiffness matrix, K is the tangent stiffness matrix in a nonlinear system, Δu is
the displacement vector, F is the applied load vector, and r is the residual.
The linearized system is converted to a minimization problem expressed as:
T T
ψ ( u ) = 1 ⁄ 2u Ku – u F (11-59)
For linear structural problems, this process can be considered as the minimization of the potential energy.
The minimum is achieved when
–1
u = K F (11-60)
The function ψ decreases most rapidly in the direction of the negative gradient.
∇ψ ( u ) = F – Ku = r (11-61)
The objective of the iterative techniques is to minimize function, ψ , without inverting the stiffness
matrix. In the simplest methods,
uk + 1 = uk + α k rk (11-62)
where
T T
α k = r k r k ⁄ r k Kr k (11-63)
The problem is that the gradient directions are too close, which results in poor convergence.
An improved method led to the conjugate gradient method, in which
uk + 1 = uk + αk Pk (11-64)
T T
α k = P k r k – 1 ⁄ P k KP k (11-65)
The trick is to choose P k to be K conjugate to P 1, P 2, …, P k – 1 . Hence, the name “conjugate gradient

methods. Note the elegance of these methods is that the solution may be obtained through a series of
matrix multiplications and the stiffness matrix never needs to be inverted.
Certain problems which are ill-conditioned can lead to poor convergence. The introduction of a
preconditioner has been shown to improve convergence. The next key step is to choose an appropriate
preconditioner which is both effective as well as computationally efficient. The easiest is to use the
diagonal of the stiffness matrix. The incomplete Cholesky method has been shown to be very effective
in reducing the number of required iterations.
Flow Diagram
Figure 11-9 is a diagram showing the flow sequence of Marc. This diagram shows the input phase,
equivalent nodal load vector calculation, matrix assembly, matrix solution, stress recovery, and output
phase. It also indicates load incrementation and iteration within a load increment.
CHAPTER 11 747
Input Phase:
Read Input Data
Space Allocation
Data Check
Incremental
Loads
Equivalent Nodal
Load Vector
Matrix Assembly
Iteration Loop
Matrix Solution
Time Step Loop
Stress Recovery
No
Convergence
Yes
Output Phase
Adapt Mesh
Yes Next
Increment
No
Stop
Figure 11-9 Marc Flow Diagram
Remarks
Which solution method to use depends very much on the problem. In some cases, one method can be
advantageous over another; in other cases, the converse might be true.
Whether a solution is obtainable or not with a given method, usually depends on the character of the
system of equations being solved, especially on the kind on nonlinearities that are involved.
As an example in problems which are linear until buckling occurs, due to a sudden development of
nonlinearity, it is necessary for you to guide the arc-length algorithm by making sure that the arc length
remains sufficiently small prior to the occurrence of buckling.
Even if a solution is obtainable, there is always the issue of efficiency. The pros and cons of each solution
procedure, in terms of matrix operations and storage requirements have been discussed in the previous
sections. A very important variable regarding overall efficiency is the size of the problem. The time
required to assemble a stiffness matrix, as well as the time required to recover stresses after a solution,
vary roughly linearly with the number of degrees of freedom of the problem. On the other hand, the time
required to go through the direct solver varies roughly quadratically with the bandwidth, as well as
linearly with the number of degrees of freedom.
In small problems, where the time spent in the solver is negligible, you can easily wipe out any solver
gains, or even of assembly gains, with solution procedures such as a line search which requires a double
stress recovery. Also, for problems with strong material or contact nonlinearities, gains obtained in
assembly in modified Newton-Raphson can be nullified by increased number of iterations or
nonconvergence.
The development of new solution procedures is still an active field of research in the
academic community.
References
1. Bergan, P. G. and E. Mollestad, “An Automatic Time-Stepping Algorithm for Dynamic
Problems”, Computer Methods in Applied Mechanics and Engineering, 49, (1985), pp. 299 - 318
2. Bathe, K. J. Finite Element Procedures, Prentice Hall, Englewood Cliffs, NJ, 1996.
3. Riks, E. “An incremental approach to the solution of solution and buckling problems”, Int. J. of
Solids and Structures, V. 15, 1979.
4. Riks, E. “Some Computational Aspects of the Stability Analysis of Nonlinear Structures”, Comp.
Methods in Appl. Mech. and Eng., 47, 1984.
5. Crisfield, M. A. “A fast incremental iterative procedure that handles snapthrough”, Comput. &
Structures, V. 13, 1981.
6. Ramm, E. “Strategies for tracing the nonlinear response near limit points,” in K. J. Bathe et al
(eds), Europe-US Workshop on Nonlinear Finite Element Analysis in Structural Mechanics, Ruhr
University Bochum, Germany, Springer-Verlag, Berlin, pp/ 63-89. Berlin, 1985.
7. Cerini, M. and Falzon, B. G., “Use of Arc-Length Method for Capturing Mode Jumping in
Postbuckling Aerostructures”, AIAA Journal, V43,3,(2005), pp 681-689.
8. P. R. Amestoy, I. S. Duff and J.-Y. L'Excellent, Multifrontal parallel distributed symmetric and
unsymmetric solvers, in Comput. Methods in Appl. Mech. Eng., 184, 501-520 (2000).
9. P. R. Amestoy, I. S. Duff, J. Koster and J.-Y. L'Excellent, A fully asynchronous multifrontal
solver using distributed dynamic scheduling, SIAM Journal of Matrix Analysis and Applications,
ol 23, No 1, pp 15-41 (2001).
10. P. R. Amestoy and A. Guermouche and J.-Y. L'Excellent and S. Pralet, Hybrid scheduling for the
parallel solution of linear systems. Parallel Computing Vol 32 (2), pp 136-156 (2006).
Chapter 12 Output Results
12 Output Results

Workspace Information 750
 Increment Information 754

Selective Printout 755

Restart 757
 Element Information 758

Nodal Information 762

Post File 765
 Forming Limit Parameter (FLP) 765

Program Messages 768

Marc HyperMesh Results Interface 769
 Marc SDRC I-DEAS Results Interface 769

Marc - ADAMS Results Interface 770

Status File 772
 Element Group Information 772

References 773
This chapter summarizes the information that Marc provides in the output. In addition to reviewing
your input, Marc provides information about the procedures the program uses and the workspace
allocation. All calculated results are automatically written to the output file unless the user specifically
requests otherwise.
Workspace Information
Marc reports several aspects of workspace information, specifically, the allocation of memory workspace
and the size of the work files.
For the general memory, Marc first gives the workspace needed for the input and stiffness assembly. This
number tells how much memory is needed to store the user-supplied data, the program-calculated data,
and two-element stiffness matrices. Each set of data comprises three parts: overhead, element
information, and nodal information. Element information consists of properties, geometries, strains, and
stresses, and nodal information consists of coordinates, displacements, and applied forces.
Next, Marc specifies the internal core allocation parameters. These values are useful in user subroutines
and are often provided to the user subroutines. The values provided here are:
• number of degrees of freedom per node (ndeg)
• maximum number of coordinates per node
• maximum number of invariants per integration point (neqst)
• maximum number of nodes per element (nnodmx)
• maximum number of stress components per integration point (nstrmx)
• number of strain components per integration point (ngens)
Note the following:
• The number of degrees of freedom per node indicates the number of boundary conditions
necessary to eliminate rigid body modes.
• The number of stresses per integration point is the number of stresses per layer multiplied by the
number of layers.
• The number of invariants per integration point is the number of layers for either shell or
beam elements.
The ELSTO parameter stores element information on an auxiliary storage device, rather than in main
memory. You can invoke this option, or Marc invokes it automatically. In this option, Marc:
• sets a flag for element storage (IELSTO)
• specifies how many words are used per element (NELSTO)
• specifies the number of elements per buffer (MXELS)
• specifies the total amount of space needed to store this information
The number of elements per buffer should be at least two, and you can change the buffer size in the
ELSTO parameter option to increase the number of elements which will be in the buffer.
Marc reports the memory for different parts as it is allocated.
CHAPTER 12 751
Output Results
For contact bodies:

allocated workspace of 4567 words of memory for body 1
For incremental backup:
space needed for incremental backup: 555313
For boundary conditions:
allocated 600 words of memory due to kinematic boundary
conditions
We also get printouts like
total workspace needed with in-core matrix storage = 893281
This printout refers to the memory in general memory (see Marc Workspace Requirements in Chapter 2
in this volume) and is not the total memory used.
The solution of the equations of equilibrium result in the formation of the operator matrix and the
decomposition/inversion of the matrix. In an mechanical analysis, this is the global stiffness matrix. For
some solvers, there is an intermediate stage where the matrix is restructured from a sparse format to a
more optimal format for numerical calculations. The allocation of this memory is reported in the memory
summary (see below) either under solver:first part or under the allocated separately:
category. This is summarized as follows:
Operator Decomposition or
Solver Matrix Restructured Matrix Inverted Matrix
0 -Profile direct first part first part first part
2 - Sparse Iterative first part N/A N/A
4 - Sparse Direct first part first part first part
6 - Hardware Provided first part allocated separately allocated separately
8 - Multi-front Direct Sparse first part N/A allocated separately
10 - Mixed Direct Iterative first part N/A allocated separately
11 - Pardiso Direct first part N/A allocated separately
12 - MUMPS direct first part N/A allocated separately
For solvers 0, 2, and 4, all the memory for the matrix solution is allocated within the general memory.
For solver 6 and 8, the workspace for the restructuring stage (solver 6 only) and decomposition part is
allocated separately outside of the general memory.
Solvers 0, 4, and 8 have an out-of-core option for both storing the operator matrix and for inverting the
matrix. Solver 6 (SGI) has an out-of-core option for inverting the matrix only. For this solver, there must
be enough memory to assemble the operator matrix in-core.
When using solver 0, if the out-of-core solver is invoked, one observes:
total workspace needed with in-core matrix storage = 89686633
out-of-core matrix storage will be used
core allocation based on 50498 nodal entries per assembly buffer.
...file 14-- maximum record length= 40960
approximate no. of words on file= 3686400

If solver 4 were used, one would observe:
total workspace needed with sparse in-core matrix storage =
34288801
out-of-core matrix storage will be used
Note that until 14 contains the true sparse matrix, and, as the same problem is used, the I/O is the same
for both solvers 0 and 4. For solver 0, this matrix gets expanded on unit 11 using a row blocking system,
and decomposed to unit 12. For solver 4, it gets expanded and decomposed using unit 2. Solver 4 uses
substantially less I/O than solver 0.
For solver 8, we observe
solver workspace needed for out-of-core matrix storage =
6389794
solver workspace needed for in-core matrix storage =
31339042
matrix solution will be out-of-core
approximate disk space for out-of-core matrix storage =
62678084
At the end of the run, Marc prints out a summary of the memory used. This typically looks something like
memory usage: mbyte words %
of total
within general memory:

element stiffness matrices: 0 64982
0.2
solver: first part 4 963934
3.6
overallocation initial allocation 49 12957302
48.1
CHAPTER 12 753
Output Results
other: 0 15713
0.1
allocated separately:
incremental backup: 5 1253317
4.7
solver 8 10 2581002
9.6
nodal vectors: 1 349427
1.3
contact: 1 261625
1.0
tyings: 0 13362
0.0
transformations: 0 25182
0.1
kinematic boundary conditions: 0 4200
0.0
element storage: 5 1324844
4.9
executable and common blocks: 27 7000000
26.0
miscellaneous 0 121763
0.5
---------------------------------------------------------------
total: 103 26936653
general memory allocated: 53 14001931

general memory used: 4 1044629
peak memory usage: 110 28936653
In a parallel analysis, Marc also prints out the memory usage summed over the domains. The memory
printout, during the execution, can also be obtained by means of the control file. Suppose the job defined
in the input file jobname.dat is running. The memory summary printout is obtained by creating a file
called jobname.cnt with a single line containing the word “memory”. The information is output at the
beginning of the next increment.
If the ALLOCATE parameter is to be used for initial memory allocation, the value to look for is the one
for the line
general memory used: 4 1044629
In this example, one would choose a value of 5 for ALLOCATE to insure that no dynamic reallocation of
memory is performed.
The top part defines the memory in general memory. The “overallocation initial memory” specifies the
amount of general memory specified with the ALLOCATE option that is not used. The above example is
very small and there is a large overallocation due to the default used (although the total amount is small).
For larger problems, this part should be small. The line “other” specifies the part of the general memory
that is not categorized.
Further down are the parts that are allocated separately from the general memory, listed for each category.
Similar to “other” above, “miscellaneous” specifies memory not categorized.
The two lines for general memory is for highlighting the memory allocated and the memory actually
used. The difference between the two is reflected in the “overallocation” line in the table. “executable
and common blocks” is an estimate of the memory needed for loading the executable and storing values
in common blocks. This is added so the total better agrees with the amount of memory the operative
system reports for the program. Finally, the peak memory denotes the maximum amount of memory
allocated at any instant during the job. In some cases, memory is allocated and then released. The “total:”
in the summary table may then show a smaller value than the maximum during the run.
Increment Information
A variety of information is given for each increment, and often for each iteration, of the analysis. This
information is useful in determining the accuracy and the stability of the analysis.
Summary of Loads
The printout entitled “Load Increments Associated with Each Degree of Freedom” allows you to check
the load input quickly. It represents the sum, over all nodes in the mesh, of the point loads and the
equivalent forces obtained after distributed loads (pressures, body forces) are applied. For example,
you can easily check a pressure because the total force in a global coordinate direction would be the
projected area normal to that direction of the surface upon which the pressure is applied, multiplied by
the pressure magnitude.
Timing Information
The amount of CPU necessary to reach the given location in the analysis is indicated by the
following output:
• start of increment
• start of assembly
• start of matrix solution
• end of matrix solution
• end of increment
As an example, subtract the time associated with the start of the matrix solution from the time associated
with the end of the matrix solution to determine how much time was spent in the equation solver.
CHAPTER 12 755
Output Results
Singularity Ratio
The singularity ratio is a measure of the conditioning of the matrix (see Chapter 11 Solution Procedures
for Nonlinear Systems for further details). This ratio is printed each time there is a solution of the matrix
equations. You can measure the influence of the nonlinearities in the structure by examining the change
in singularity ratios between increments. The singularity ratio is not available when using the iterative
solvers, and when using the Pardiso and MUMPS direct solver.
Convergence
Several messages are printed that concern the convergence of the solution. These messages indicate the
displacement, velocity, or residual error and are very important because they provide information
concerning the accuracy of the solution procedure. These messages also indicate the ratio of the error
and its relative quantity. This ratio must be less than that given in the CONTROL option for convergence
to occur. See Convergence Controls in Chapter 11for details on convergence testing.
When the iterative solver is used, additional messages are printed regarding the convergence of the
solution procedures.
Selective Printout
Marc gives you several options and user subroutines for the control of the Marc output.
Options
PRINT CHOICE allows you to select how much of the element and nodal information is to be printed.
The possible selections are
• group of elements
• group of nodes
• which layers (form beam and shell elements)
• which integration points
• increment frequency between printouts
The data entered through this option remain in control until you insert a subsequent PRINT CHOICE set.
You can include such a set with either the model definition or with the history definition set. To obtain
the default printout after a previous PRINT CHOICE, invoke PRINT CHOICE using blank entries.
Note: The PRINT CHOICE option has no effect on the restart or post file.
You can use the PRINT ELEMENT capability as a replacement for, or in conjunction with, the PRINT
CHOICE option. The enhancements this option offers over the PRINT CHOICE option are the following:
1. You have a choice of integration and layer points for each element to be printed, which is
especially useful when several different element types are used in one analysis.
2. You have a choice of the type of quantity to be printed, for instance, stresses could be printed for
all elements, but strains for only a few.
3. The types of output quantities that can be selected are stresses, strains, creep strain, thermal strain,
cracking strain, Cauchy stress, state variables, strain energy, or all nonzero quantities.
4. The output can be placed in a file other than the standard output file.
5. The PRINT NODE option can be used to obtain element quantities such as stresses and strains at
the nodal points of each element. These are obtained by extrapolating the integration point values
to the nodes.
The PRINT NODE option is an alternative to the PRINT CHOICE option for controlling the output
of nodal quantities. The additional capabilities of this option compared to the PRINT CHOICE are
as follows:
1. You can choose which of the following nodal quantities are to be printed:
incremental displacements
total displacements velocities
accelerations
reaction forces
generalized stresses
2. Different quantities can be printed for different nodal points.
3. The output can be placed in a file other than the standard output file.
The PRINT CONTACT and the NO PRINT CONTACT options control the output of the summary of the
loads on the contact bodies and their positions.
Note: All quantities are saved on the restart file, and you can obtain them at a later time by using
the RESTART option. Quantities that were not printed out are available for later use.
The SUMMARY option allows you to get a quick summary of the results obtained in the analysis. This
option prints the maximum and minimum quantities in tabular form. The is designed for direct placement
into reports. You control the increment frequency of summary information and the file unit to which the
information is written.
The SUMMARY option reports on the physical components and the Tresca, von Mises, and mean values
of stress, plastic strain, and creep strain. It also reports on such nodal quantities as displacements,
velocities, accelerations, and reaction forces.
The sort options (ELEM SORT and NODE SORT) allow you to sort calculated quantities in either
ascending or descending order. These quantities may either be sorted by their real magnitude or their
absolute magnitude. The sort options also allow you to control the type of quantity to be sorted, for
example, equivalent stress and the number of items to be sorted. The sort options print the sorted values
in tabular form. The is designed for direct placement into reports. You may control the increment
frequency of sorted values and the file unit to which these values are written.
CHAPTER 12 757
Output Results
The PRINT VMASS option allows you to selectively choose which elements and associated volumes,
masses, first and second moment of inertia, and strain energy are to be printed. In order to have correct
mass computations, mass density for each element must be given. The volumes, masses, and energies
can be written on either standard output file unit 6, or user specified unit. The center of mass is also
calculated and output.
Grid Force Balance

The GRID FORCE option allows the user to output the different contributions to the total force applied
on an element or a node.
On an element level, the grid force balance is based upon the
Internal forces
Distributed Loads
Foundation Forces
Reaction Force
On a nodal basis it is much more complete and includes
Internal Forces Distributed + Point Forces
Foundation Forces Spring Forces
Contact Normal Forces Contact Friction Forces
Tying/MPC Forces Inertia Forces
Damping Forces DMIG Forces
Reaction Force
User Subroutines
User subroutines can be used to obtain additional output, which can be accessed at each time/load
increment. The available subroutines are the following:
• IMPD – obtains nodal quantities, such as displacements, coordinates, reaction forces, velocities,
and accelerations
• ELEVAR – obtains element quantities such as strains, stresses, state variables, and
cracking information
• ELEVEC – outputs element quantities during harmonic subincrements
• INTCRD – obtains the integration point coordinates used for forming the stiffness matrix
Restart
One capability of the RESTART option allows you to recover the output at increments where printout
was suppressed in previous runs. This option can be used to print time/load increments for a number of
consecutive increments. Under this option, Marc does not do any analysis.
In conjunction with the RESTART option, you can use either the PRINT CHOICE or PRINT ELEMENT,
or PRINT NODE option to select a region of the model for which you want to obtain results.
Element Information
The main output from an increment includes element information followed by nodal information. The
system provides the element data at each integration point. If you use the CENTROID parameter, the
system provides the element data at the centroidal point. You can control the amount of printed output by
using either the PRINT CHOICE or PRINT ELEMENT option. All quantities are total values at the current
state (at the end of the current increment), and the physical components are printed for each tensor
quantity (stress, strain, and generalized stress and strain). The orientation of these physical components
is generally in the global coordinate system; however, their orientation depends on the element type.
(Marc Volume B: Element Library explains the output for each particular element type.) Also, the
physical components may be printed with respect to a user defined preferred system.
In addition to the physical components, certain invariants are given, as follows:
1. Tresca intensity - the maximum difference of the principal values, and the measure of intensity
usually required for ASME Code standards.
von Mises intensity - defined for stress as
3--- 1
σ = S S S i j = σ i j – --- δ i j σ kk (12-1)
2 ij ij 3
where σ is a stress tensor and S is the deviatoric stress tensor. For beam and truss elements,
σ 22 = σ 33 = 0 , and additional shears are zero; for plane stress elements, including plates and
shells, σ 33 = 0 .
2. von Mises intensity - calculated for strain type quantities as
2
ε = --- ε i j ε i j (12-2)
3
Marc uses these measures in the plasticity and creep constitutive theories. For example,
incompressible metal creep and plasticity are based on the equivalent von Mises stress. For beam,
truss, and plane stress elements, an incompressibility assumption is made regarding the
noncalculated strain components.
a. For beam and truss elements, this results in
1
ε 22 = ε 33 = – --- ε 11 (12-3)
2
b. For plane stress elements
ε 33 = – ( ε 11 + ε 22 ) (12-4)
c. For beam and plane stress elements taken as a whole

CHAPTER 12 759
Output Results
εk k = 0 (12-5)
3. Mean normal intensity - calculated as
( σ 11 + σ 22 + σ 33 ) ⁄ 3
or (12-6)
( ε 11 + ε 22 + ε 33 ) ⁄ 3
Equation (12-6) represents the negative hydrostatic pressure for stress quantities. For strain
quantities, the equation gives the dilatational magnitude. This measurement is important in
hydrostatically dependent theories (Mohr-Coulomb or extended von Mises materials), and for
materials susceptible to void growth.
4. The principal values are calculated from the physical components. Marc solves the eigenvalue
problem for the principal values using the Jacobi transformation method. Note that this is an
iterative procedure and may give slightly different results from those obtained by solving the
cubic equation exactly.
5. State variables are given at any point where they are nonzero.
Mentat Computed Element Results

These element results are computed based on the Marc output. By default, Marc always outputs the
elemental results at the integration points. In Marc Mentat, by default, these data are extrapolated linearly
to the nodal positions. The data are then averaged with the contributions from the neighboring elements.
Principle values and equivalent values are computed based on these nodal data in Marc Mentat. If a stress
or strain tensor is output in the post file, Marc Mentat will compute their principal values and the
equivalent items (such as, equivalent stress and strain). Note that the equivalent items may be different
from ones directly output by Marc because of the way they are computed in Marc Mentat. For example,
the equivalent stress computed in Marc Mentat may show a slightly different result compared to the
equivalent von Mises stress from Marc output.
It is also possible to turn off the extrapolation procedure in Marc Mentat. This is recommended in
problems of high gradients. It can be done through the RESULTS>scalar plot
SETTING>EXTRAPOLATION menu by picking TRANSLATE.
Solid (Continuum) Elements

For solid (continuum) elements, stress, strain, and state variables are the only element quantities given.
Marc prints out stresses and total strain for each integration point. In addition, it prints out thermal,
plastic, creep, and cracking strains, if they are applicable. Note that the total strains include the
thermal contribution.
Shell Elements
Marc prints generalized stresses and generalized total strains for each integration point. For thick shell
(Type 22, 72, 75) elements interlaminar shear stresses are printed at the interface of two laminated layers
if the TSHEAR parameter is included.
The generalized stresses printed out for shell elements are

+t⁄2
1
---
t  σ i j dy
(12-7)
–t ⁄ 2
( Section Force per Unit Thickness )
+t⁄2
1
---
t  yσ i j dy
(12-8)
–t ⁄ 2
( Section moment per Unit Thickness )
The generalized strains printed are
Eα β ; α, β = 1, 2
(12-9)
( Stretch )
κα β ; α, β = 1, 2 (12-10)
( Curvature )
Physical stress values are output only for the extreme layers unless you invoke either PRINT CHOICE or
PRINT ELEMENT. In addition, thermal, plastic, creep, and cracking strains are printed for values at the
layers, if applicable. Although the total strains are not output for the layers, you can calculate them using
the following equations.
ε 11 = E 11 + hκ 11 (12-11)
ε 22 = E 22 + hκ 22 (12-12)
γ 12 = E 12 + hκ 12 (12-13)
where h is the directed distance from the midsurface to the layer; E i j are the stretches; and κ i j are the
curvatures as printed. You can obtain these quantities by using the ELEVAR subroutine.
Beam Elements
The printout for beam elements is similar to shell elements, except that the section values are force,
bending and torsion moment, and bimoment for open section beams. For beam element type 45,
interlaminar shear stresses are printed at the interface of laminated layers if the TSHEAR parameter
is used.
These values are given relative to the section axes (X, Y, Z) which are defined in the COORDINATES or
the GEOMETRY option.
CHAPTER 12 761
Output Results
Before a beam member can be designed, it is necessary to understand the section forces distribution
along the axial direction of the beam. For example, if variations of shear force and moment along axial
direction are plotted, the graphs are termed shear diagram and moment diagram, respectively. To
postprocess these diagrams, post code 261 needs to be specified along with the post codes (shown in
Table 12-1) which define the corresponding section forces of the beam element.
Table 12-1 Post Codes for Beam Section Forces
Post Code
265 266 267 268
264 (Moment xx (Moment yy (Shear (Shear yz
Element (Axial or Moment or Moment xz or or Shear 269 270
Type 2-D/3-D Force) ip) op) Shear ip) op) (Torque) (Bimoment)
5 2-D X X
16 2-D X X
45 2-D X X X
13 3-D X X X X X
14 3-D X X X X
25 3-D X X X X
31 3-D X X X X X X
52 3-D X X X X
76 3-D X X X X
77 3-D X X X X
78 3-D X X X X
79 3-D X X X X
98 3-D X X X X X X
The stresses at the section points in the cross section are only output if the PRINT CHOICE or PRINT
ELEMENT option are included. If any nonlinearity exists at the section points, they will be provided in
the output file.
Gap Elements
Marc prints the gap contact force, friction forces, and the amount of slip.
Linear and Nonlinear Springs

Marc prints the spring forces.

The element temperatures are provided at the integration points. When requested, the temperature
gradient and the flux is also provided.
Joule Heating Analysis

In addition to the temperatures, the voltage, current, and the heat generated at the integration points
is output.

The lubricant thickness, pressure, and pressure gradient are provided.
The electrical potential ( V ), electric field intensity vector ( E ), and electrical displacement vector/flux
( D ) are provided.
Piezoelectric or Electrostatic-Structural Analysis

In addition to the mechanical stresses and strains, the electrical field intensity vector ( E ) and electrical
displacement/flux ( D ) are provided.
The magnetic potential ( A ), the magnetic field density ( B ), and the magnetic flux density ( H ) are
provided. For two-dimensional problems, only the A z component is given.
Electromagnetics Analysis
The electric field intensity vector ( E ), the electrical displacement vector/flux ( D ), the magnetic flux
density ( H ), and the current density ( J ) are provided.
Acoustic Analysis
The pressure and the gradient of the pressure is provided. Note in an Acoustic-Structural analysis, the
pressure is only provided as a nodal variable.
Nodal Information
Marc also prints out the following quantities at each nodal point.
CHAPTER 12 763
Output Results
Stress Analysis
• Incremental displacements - the amount of deformation that occurred in the last increment
• Total displacements - the summation of the incremental displacements
• Total equivalent nodal forces (distributed plus point loads) - the total force applied to the model
through distributed loads (pressures) and point loads
• Reaction forces at fixed boundary conditions
• Residual loads at nodal points that are not fixed by boundary conditions
Note: Each value listed above is given at each nodal point, unless you invoke the PRINT
CHOICE or PRINT NODE option. If you invoke the TRANSFORMATION option, Marc
prints the nodal information relative to the user-defined system, rather than the global
coordinate system.
Reaction Forces
Marc computes reaction forces based on the integration of element stresses. This is the only
way to compute total reaction forces in a nonlinear analysis. Since such integration is only exact if the
stresses are known at each integration point, the reaction forces are not printed if you use the
CENTROID parameter.
In a nonlinear analysis, you should check that the reaction forces are in equilibrium with the external
forces. If they are not in equilibrium, the analysis will be inaccurate, usually due to excessively large
incremental steps. In most cases, the equilibrium is automatically ensured due to the convergence testing
in Marc.
Residual Loads
The residual loads are a measure of the accuracy of the equilibrium in the system during analysis. This
measure is very important in a nonlinear analysis and should be several orders of magnitude smaller than
the reaction forces.
Dynamic Analysis
In a dynamic analysis, Marc prints out the total displacement, velocity, and acceleration at each
time increment.

In a heat transfer analysis, Marc prints out nodal temperatures, externally applied fluxes, and calculated
nodal heat fluxes.
Joule Heating Analysis

In Joule heating analysis, in addition to the thermal vectors, the nodal voltage and the applied and
calculated currents are provided.
Rigid-Plastic Analysis
In steady-state rigid-plastic analysis, Marc prints nodal velocities.

In a hydrodynamic bearing analysis, Marc gives the mass flux at the nodal points.
In an electrostatic analysis, Marc prints the scalar potential and the applied and calculated charge.
In a magnetostatic analysis, Marc prints the vector potential and the applied and calculated current.
In an electromagnetic analysis, Marc prints out the vector and scalar potential, the applied and calculated
current and charge.
Piezoelectric or Electrostatic-Structural Analysis

In a piezoelectric or electrostatic-structural analysis, in addition to the structural quantities, such as
displacement and reaction force, the electric potential, and reaction charges.
Acoustic Analysis
In an acoustic analysis, Marc prints the pressure and the source.
Supplementary Information
Contact Analysis
The standard output includes, at the end of each increment, a summary of information regarding each
body. This information reports the increment’s rigid body velocity, the position of the center of rotation,
and the total loads on the body. These last values are obtained by adding the contact forces of all nodes
in contact with the rigid body. Deformable bodies being in equilibrium have no load reported.
Additional information can be obtained by means of the PRINT,5 parameter. In such cases, all the contact
activity is reported. Namely, every time a new node touches a surface, or separates from a surface, a
corresponding message is issued.
CHAPTER 12 765
Output Results
Contact with rigid surfaces entails an automatic transformation. Displacement increments and reactions
in the transformed coordinates - tangent and normal to the contact interface - are also reported for every
node that is in contact in the usual Marc manner.
If the EMRESIS or EMCAPAC options are invoked, a summary is output providing the electrical
resistance or capacitance for the conducting body.
Post File
The post file contains results of the analysis performed. This information can be viewed using Marc
Mentat or MD Patran to observe the displaced mesh, contours of element quantities, principal quantities,
time history behavior, response gradients, design variables, etc. The post file contains the finite element
mesh and any changes to the finite element mesh due to either rezoning or adaptive mesh refinement.
Basic nodal quantities (such as displacements, velocities, acceleration, applied loads, and reaction
forces) are automatically placed on the post file. It is also possible to include user-defined nodal vectors
by use of the UPOSTV user subroutine. Element quantities, such as stress, strain, plastic strain) must be
explicitly requested by using the POST model definition option. It is possible to have user-defined
element quantities placed on the post file for subsequent display by using the PLOTV user subroutine.
Forming Limit Parameter (FLP)

For continuum or shell elements, the Principal Engineering Strains can be calculated based on the true
strain values. Correspondingly, the Forming Limit Parameter can be obtained for shell/membranes
according to the data of the Forming Limit Diagram (FLD).
The principal engineering strains are calculated based on the true strain according to following relation:
e i = exp ( ε i ) – 1 (12-14)
where e i are the principal engineering strain components and ε i are the principal true strain components.
The principal engineering strain values can be selected for postprocessing. In addition, if
shell/membrane elements are used, the Forming Limit Parameters (FLP) can also be selected. The FLP
is defined as the ratio of the major principal engineering strain to the maximum allowable major principal
engineering strain given by the Forming Limit Diagram. Based on this definition, it is calculated by
FLP = e 1 ⁄ FLD ( e 2 ) (12-15)
where e 1 means the major principal engineering strain, e 2 is the minor principal engineering strain.
FLD ( e 2 ) is the forming limit corresponding to e 2 . FLD is defined by input data in model definition
through material properties. By plotting the major principal engineering strain of each integration point
of an element as point ( e 2 , e 1 ) on the chart, as shown in Figure 12-1, it is easy to find that for points
below the FLD line is safe. Strain rates above the FLD line imply failure. To consider the FLP value of
these fields, one can find that in the field which is above FLD line, the value of FLP is larger than 1. On
the FLD line, FLP equals 1, and FLP is less than 1 if a point located in the field below the FLD line.
Therefore, by plotting the FLP value in Marc Mentat, it can be seen if the sheet metal forming process is
successful.
Major Principal Engineering Strain (e1)
FLD(e2)
Fail
Fail
FLP>1 FLD(e2)
Safe
Safe
FLP<1
Minor Principal Engineering Strain (e2)

Figure 12-1 The Definition of Forming Limit Parameter (FLP)
The methods employed in Marc to define the Forming Limit Diagram are:
1. Fitted function definition
2. Predicted function definition
3. Table definition (accepts piecewise linear FLD curves)
The details of the above methods are described below:
1. Fitted function:
By this method, the polynomial functions are utilized to fit the FLD curve. The functions are
shown in Equations 12-16a and 12-17b:
FLD ( e 2 ) = C 0 + D 1 e 2 + D 2 e 22 + D 3 e 23 + D 4 e 24 ( e2 ≤ 0 ) (12-16)
FLD ( e 2 ) = C 0 + C 1 e 2 + C 2 e 22 + C 3 e 23 + C 4 e 24 ( e2 ≥ 0 ) (12-17)
2. Predicted function:
The predicted functions are generated based on the theories of local necking and diffuse necking.
Both theories assume that the material obeys the power-law strain hardening, σ = Kε n .
Local necking: The critical strain for the onset of local necking is influenced by strain ratio ρ ,
where ρ = ε 2 ⁄ ε 1 , ρ ≤ 0 . Equation shows the relationship of critical strain versus strain ratio
(by Hill).
CHAPTER 12 767
Output Results
n
ε* = ------------------ ρ≤0
(1 + ρ) (12-18)
For uniaxial tension, ρ = – 0.5 . For plane strain, ρ = 0 .

Diffuse necking: Diffuse necking is the phenomenon that while necking happens, deformation
continues. The critical strain for the diffuse necking to happen is determined by Equation 12-19
(by Swift):
2n ( 1 + ρ + ρ 2 )
ε* = ------------------------------------------------------ ρ≤0 (12-19)
( 1 + ρ ) ( 2ρ 2 – ρ + 2 )
Note: For the case of ρ < 0 , the predicted critical strain by Equation 12-19 is well below the
experiment data. Therefore, for the purpose of prediction, the diffuse necking theory
and the local necking theory are chosen for cases of ρ > 0 and ρ ≤ 0 , respectively.
Experiments (mainly with low carbon steels) showed that the predicted FLD usually is at a level
below the experimental FLD curve. The experimental forming limit diagram is characterized by
the value of ε* corresponding to plane strain. This value is referred at FLD 0 as shown in
Figure 12-2. This value increases with the strain-hardening exponent, n , and the strain-rate
exponent, m . The value of ε* is observed to rise as the thickness increases. This phenomenon is
referred as thickness effect and it is characterized as thickness coefficient t c . Experiments by
Keeler tended to express this relationship as shown in Equation (12-20).
FLD 0 = Q ⋅ ( 0.233 + t c ⋅ t ) (12-20)
where Q = n ⁄ 0.21 if n is less than 0.21. Otherwise, Q = 1.0 . T is the thickness of the sheet
metal. The thickness coefficient t c is set as 3.59 if the unit used to define the thickness is “Inch”.
If the unit used is “mm”, t c is set as 0.141. Similarly, if the unit “cm” is used, t c is then set as
1.41, etc.
It should be emphasized that, in most cases, friction is also a very important parameter that will
affect the occurrence of necking. On the other hand, Equation (12-20) is mainly from data on low-
carbon steels that have relatively high strain-rate dependence. In materials with low strain-rate
dependence (for example, aluminum alloys), the thickness effect should be much less.
Finally, it is suggested that for safety purposes, a safety zone, as shown as the shadow area in
Figure 12-2, must be set. For points above the safety zone, material is considered as failure. For
points below the safety zone, material is considered as safe. If a point locates in the safety zone,
the material is in the marginal status. Usually, the thickness of the safety zone is set as 0.1 or 10%.
3. Table definition:
The table function in Marc allows users to define any curves through the TABLE model definition
option. For example, if the user has FLD point of material, it is possible to define the FLD as
piecewise linear curve. For details, refer the Marc Volume C: Program Input, Chapter 3 Model
Definition Options, TABLE.
Major Principal Engineering Strain e1
Experimental
forming limit
FLD0
Local necking Diffuse necking
0
Minor Principal Engineering Strain e2
Figure 12-2 Predicted FLD and Experimental FLD Curves
Program Messages
The messages provided by Marc at various points in the output show the current status of the problem
solution. Several of these messages are listed below.
• START OF INCREMENT x indicates the start of increment number x.
• START OF ASSEMBLY indicates Marc is about to enter the stiffness matrix assembly.
• START OF MATRIX SOLUTION indicates the start of the solution of the linear system.
• SINGULARITY RATIO prints yyy, where yyy is an indication of the conditioning of the matrix.
This value is typically in the range 10-6 to 1. If yyy is of the order of machine accuracy (10-6 for
most machines), the equations might be considered singular and the solution unreliable.
• END OF MATRIX SOLUTION indicates the end of matrix decomposition.
• END OF INCREMENT x indicates the completion of increment number x.
• RESTART DATA at INCREMENT x ON UNIT 8 indicates that the restart data for increment
number x has been written (saved) on Unit 8.
• POST DATA at INCREMENT x on UNIT y indicates that the post data for increment number x
has been written (saved) on Unit y.
CHAPTER 12 769
Output Results
In addition to these Marc messages, exit messages indicate normal and abnormal exists from Marc.
Table 12-2 shows the most common exit messages.
Table 12-2 Marc Exit Messages
Message Meaning
MARC EXIT 3001 Normal exit.
MARC EXIT 3004 Normal exit.
MARC EXIT 13 Input data errors were detected by the program.
MARC EXIT 2004 Operator matrix (for example, stiffness matrix in stress analysis) has
become non-positive definite and the analysis terminated.
MARC EXIT 3002 Convergence has not occurred within the allowable number of recycles.
Marc HyperMesh Results Interface

Marc now outputs results for postprocessing with HyperMesh. The writing of the HyperMesh results
file is invoked by the HYPERMESH model definition option described in detail in Marc Volume C:
Program Input. Marc Mentat can also be used to write the necessary commands into the Marc data file.
The related button can be found one level below each of the ANALYSIS CLASS buttons within the JOBS
menu.The binary file created as a result has the title jobid.hmr where jobid is the name of the data
file for the job.
The geometry data may be imported into HyperMesh from a Marc data file. Alternatively, it can be
created by HyperMesh or read in from other formats via the HyperMesh import capability.
The types of data that may be selected for writing into the HyperMesh results file are listed in Marc
Volume C: Program Input under the HYPERMESH model definition option. It is important to note that,
in case eigenvectors for buckling or eigenfrequency analysis are to be written into the results file, the
corresponding Marc file should have the BUCKLE INCREMENT or MODAL INCREMENT model
definition option, as appropriate. The BUCKLE, MODAL SHAPE, and RECOVER history definition
options are not to be used in these cases.
Since HyperMesh allows only one deformed shape plot per simulation, each eigenvector of an
eigenvalue analysis is saved as a separate simulation. Thus, when using HyperMesh, these eigenvectors
can be plotted by skipping to the next simulation rather than skipping to the next data type of a
simulation. The contour plots can be obtained for all data types including eigenvectors.
In case the number of requested eigenvalues is more than the number extracted, the data type in the
HyperMesh “deformed” screen will inform you for those modes that are not extracted. The “next” button
may need to be clicked to see the data type in the “deformed” mode of plotting.
Marc SDRC I-DEAS Results Interface

A facility exists in Marc to write out an SDRC I-DEAS universal file of the analysis results, in order for
postprocessing by the I-DEAS program. The writing of the universal file is invoked by the SDRC model
definition option described in detail in Marc Volume C: Program Input. Marc Mentat can also be used to
write the necessary commands into the Marc data file. The related button can be found one level below
each of the ANALYSIS CLASS buttons within the JOBS menu. The formatted ASCII file created as a
result has the title jobid.unv, where jobid is the name of the data file for the job.
The model data is imported into I-DEAS by way of the universal file as well.
The types of data that may be selected for writing into the I-DEAS universal file are listed in Marc
Volume C: Program Input under the SDRC model definition option.
Marc - ADAMS Results Interface

ADAMS/Flex allows flexible components to be included into ADAMS models to achieve more realistic
simulation results. Marc is capable of generating a Modal Neutral File (MNF) representing the flexible
component to be integrated into the ADAMS model. The flexible component can undergo linear or
nonlinear loading in Marc up to the point where the MNF is requested to be generated. No more
analysis takes place in Marc after generating the MNF. Generating an MNF from Marc is based on
performing the most general method of Component Mode Synthesis (CMS) techniques, namely the
Craig-Bampton method.
Using the Craig-Bampton method, the degrees of freedom, u , of the flexible component are partitioned
into two sets of DOFs:
• Boundary or Interface DOFs, u B : These DOFs can be used to apply loads in ADAMS or to
connect the flexible component to other rigid bodies.
• Interior DOFs, u I : These internal DOFs of the flexible component can be condensed out using
the superelement technique.
Two sets of modes shapes arranged in matrices, Φ I C and Φ I N , and their corresponding modal
coordinates vectors, q C and q N , are also defined:
• Constraint Modes, Φ I C : These modes are the static shapes obtained by giving each boundary
degree of freedom a unit displacement while holding all other boundary DOFs fixed.
Corresponding to the constraint modes is the set of modal coordinates q C . The basis of
constraint modes completely spans all possible motions of the boundary DOFs, with a one-to-one
correspondence between the modal coordinates of the constraint modes and the displacement in
the corresponding boundary DOFs, u B = q C .
• Fixed-Boundary Normal Modes, Φ I N : These modes are obtained by fixing the boundary DOFs
and computing an eigensolution. There are as many fixed-boundary normal modes as the user
desires. These modes define the modal expansion of the interior DOFs. The quality of this modal
expansion is proportional to the number of modes retained by the user. Corresponding to the
fixed-boundary normal modes is the set of modal coordinates q N .
The Craig-Bampton modes consist of both the constraint modes and the fixed-boundary normal modes.
The relationship between the physical DOFs and the Craig-Bampton modes and their modal coordinates
is given by
CHAPTER 12 771
Output Results
   
 uB  IB C OB N  qC 
  =   (12-21)
 uI  ΦI C ΦI N  qN 
   
where the subscripts B, I, C, and N denote Boundary DOF, Interior DOF, Constraint Mode and Normal
Mode, respectively. The above equation indicates that the Craig-Bampton analysis can be considered as
a transformation from the physical DOFs to the Craig-Bampton modal basis q in the form
u = Φq
where Φ is the modal matrix or transformation matrix. The finite element equation of motion is given by
Mu·· + Ku = f
in which M is the mass matrix, K is the stiffness matrix and f is the load vector. Substituting by the
T
transformation relation into the equation of motion and premultiplying by Φ gives
T ·· T T
Φ MΦq + Φ KΦq = Φ f
The above equation defines the generalized mass and stiffness matrices as:
ˆ : Projection of the mass matrix onto the modal space given by
• Generalized Mass Matrix, M
ˆ = Φ T MΦ
M
ˆ : Projection of the stiffness matrix onto the modal space given by

• Generalized Stiffness Matrix, K
ˆ = Φ T KΦ
K
Note that the ADAMS MNF interface expects lumped mass matrices to be used. Thus, the LUMP
parameter must be included in the Marc input. The MNF interface distinguishes between two types
of loads:
• Modal Preload, ˆf P L : A projection onto the modal space of the internal load vector
corresponding to the loads that have already been applied in the finite element analysis up to the
point when the MNF is generated.
• Modal Loads, ˆf L : A projection onto the modal space of the loads that may be defined in the Marc
input using the ASSEM LOAD history definition option but not applied in the finite element
analysis. They may be scaled and introduced in the ADAMS simulation. ASSEM LOAD
loadcases, if exist, should precede any actual preloading to be applied to the component in Marc.
Both types of loads can be computed using the relation
ˆf = Φ T f
Corresponding to each Craig-Bampton mode, a stress and a strain mode can be computed. These are the
incremental nodal stresses and strains found by considering each mode at a time to be the incremental
displacement solution and performing a stress recovery. To request the computation and export of stress
and/or strain modes, the MNF parameter must be used.
The Craig-Bampton analysis is triggered when the SUPERELEM model or history definition option is
reached in the input. The SUPERELEM option allows direct definition of the boundary or interface
DOFs, u B . The option also allows automatic definition of interface DOFs of the nodes that get in contact
with selected rigid contact bodies. This is very useful for some nonlinear analyses such as tire footprint
analysis in which the interface DOFs are not known a priori. It also allows the specification of interface
DOFs of the control nodes of selected load-controlled rigid contact bodies. The two control nodes for
load-controlled rigid bodies are consolidated into one node with six degrees of freedom before exporting
to the MNF.
A MODAL SHAPE loadcase using the Lanczos method must be present in the input immediately
following the SUPERELEM loadcase. No more analysis can be performed afterwards.
The units used in the Marc model are specified using the MNF UNITS model definition option. Successful
MNF generation produces Marc Exit 3018. The MNF generated will have the same filename as the input
file and the extension .mnf. Assuming the input file is jobname.dat, the MNF will have the name
jobname.mnf.
Status File
The status file contains summarized information of the analysis. Each line within this file includes the
load case number, the associated increment number, and the number of cycles, contact separations, and
cutbacks associated with that increment; the accumulated total cycles, increment splittings, separations,
cutbacks, and remeshing times numbers in the analysis as well as the time step size of each increment
and the overall time achieved by the analysis. Marc reports all this information on one line upon
completion of each increment. If the increment is partially completed (for example, the completion of
one part of a split increment or one part of the whole increment due to cutback), Marc also reports one
line for each part of the increment.
Element Group Information

The finite element model is internally subdivided into element groups. The groups are based upon the
element type and material ID. The division into groups is internal to the program and entirely automatic.
The memory allocation is done separately for each element group, allowing the memory usage to be
optimized for cases when different element types and materials are used. Information about the element
groups is printed to the output file. This printout occurs in two places. The program first prints the amount
of memory used for each group. Here is a sample printout for a job with two element groups
memory usage per element group
group # elements nelsto MByte words
1 1368 1480 8 2024640
2 456 1284 2 585504
----------------------------------------------------------
total 1824 10 2610144
CHAPTER 12 773
Output Results
For each groups, it lists the number of elements in the group; nelsto, which is the amount of memory in
words used for each element and the total amount of memory used for all elements in the group, given
in Megabytes and words.
A second printout is written further down in the output file after reading the model definition input data:
number of element groups used: 2
group # elements element type material formulation
1 1368 7 1 S
2 456 140 1 S
formulation:
S: small displacement
This listing shows the number of elements in the group and also the element type, the material ID, and
the formulation used in the group. If composites or rebars are used, it lists the composite or rebar group
instead of the material ID. The formulation column shows a number of characters indicating the type of
formulation used. The table below shows the characters that are used.
Character Description
S small displacement small strain analysis
T total Lagrange formulation
U updated Lagrange formulation
M multiplicative plasticity or hyperelastic formulation
A additive plasticity formulation
F finite strain plasticity for additive formulation
E enhanced/assumed strain formulation
C constant dilatation formulation
References
1. R. R. Craig and M. C. C. Bampton. Coupling of substructures for dynamics analyses. AIAA
Journal, 6(7): 1313-1319. 1968.
2. R.R. Craig. Structural Dynamics: An Introduction to Computer Methods. John Wiley &
Sons. 1981.
3. ADAMS User Guide: Using ADAMS/Flex. 2002.
Chapter 13 Parallel Processing
13 Parallel Processing

Different Types of Machines 776
 Supported and Unsupported Features 776

Matrix Solvers 778

Contact 779
 Domain Decomposition 779
Marc can make use of multiple processors when performing an analysis. Most, but not all, features of
Marc are supported in parallel mode. The type of parallelism used is based upon domain decomposition.
A commonly used name for this is the Domain Decomposition Method (DDM). The model is
decomposed into domains of elements, where each element is part of one and only one domain. The
nodes which are located on domain boundaries are duplicated in all domains at the boundary. These nodes
are referred to as inter-domain nodes below. The total number of elements is thus the same as in a serial
(nonparallel) run but the total number of nodes can be larger. The computations in each domain are done
by separate processes on the machine used. At various stages of the analysis, the processes need to
communicate data between each other. This is handled by means of a communication protocol called MPI
(Message Passing Interface). MPI is a standard for how this communication is to be done and Marc
makes use of different implementations of MPI on different platforms. Marc uses MPI regardless of the
type of machine used. The types of machines supported are shared memory machines, which are single
machines with multiple processors and a memory which is shared between the processors, and cluster of
separate workstations connected with some network. Each machine (node) of a cluster can also be a
multiprocessor machine.
Marc also has matrix solvers which can perform the matrix solution in parallel while the rest of the
analysis is serial. The MUMPS solver supports both shared memory machines and clusters. It uses MPI
for the communication. The SGI and Pardiso solvers support shared memory machines. These solvers
can also be used with DDM.
Different Types of Machines

As mentioned above, Marc can run on a shared memory machine and on a cluster of workstations. The
main reason for running a job in parallel on a shared memory machine is speed. Since all processes run
on the same machine sharing the same memory, the processes all compete for the same memory. There
is an overhead in memory usage so some parts of the analysis need more memory for a parallel run than
a serial job. The matrix solver, on the other hand, needs less memory in a parallel job. Less memory is
usually needed to store and solve several smaller systems than a single large one.
In the case of a cluster, the picture is somewhat different. Suppose a number of workstations are used in
a run and one process is running on each workstation. The process then has full access to the memory of
the workstation. If a job does not fit into the memory of one workstation, the job could be run on, say,
two workstations and the combined memory of the machines may be sufficient.
The amount of speed-up that can be achieved depends on a number of factors including the type of
analysis, the type of machine used, the size of the problem, and the performance of communications. For
instance, a shared memory machine usually has faster communication than a cluster (for example,
communicating over a standard Ethernet). On the other hand, a shared memory machine may run slower
if it is used near its memory capacity due to memory access conflicts and cache misses etc.
Supported and Unsupported Features

Most of the main features of Marc are supported in parallel mode. For instance, a fully coupled
thermomechanical contact analysis can be performed in parallel. Local adaptivity is fully supported. New
elements and nodes that are created remain in the same domain as the parent element. Global adaptive
meshing is supported with DDM, but each body to be remeshed cannot be in multiple domains.
CHAPTER 13 777
Parallel Processing
The list of unsupported features is:

• Acoustics
• Auto therm creep
• Beam-to-beam contact
• Bearing
• Buckling
• Convective terms in heat transfer
• Design sensitivity and optimization
• Eigenvalue extraction
• Electromagnetics
• Explicit dynamics
• Fluids and its coupled analysis
• Gap elements (but the normal contact is supported)
• Harmonics
• Hydrodynamics
• Insert
• Radiation
• Response spectrum
• Steady state rolling analysis
• Superelements
• VCCT with crack propagation
Global remeshing is supported by DDM but with the restriction that each remeshing body must be fully
contained in a domain.
The DDM single input file capability will not support any analyses types and input options that are not
currently supported by the traditional multiple input files DDM.
Moreover, some additional model and history definition options and command line arguments are not
supported in the current release. These are:
• ACTIVATE and DEACTIVATE
• ACTUATOR
• BEGIN SEQUENCE and END SEQUENCE
• CHANNEL
• CFAST
• CONRAD GAP
• CWELD
• DAMPING
• DISP CHANGE if not FIXED DISP format. The DISP CHANGE option is supported only if the 1st
entry of the data block following DISP CHANGE is equal to 0.
• EXCLUDE
• FXORD
• GLOBALLOCAL
• GRID FORCE
• INCLUDE
• INSERT
• NEW
• POINT LOADS with follower forces
• SUPERPLASTIC
• QVECT
• TIME-TEMP
• TYING CHANGE
• UDUMP
• def command line option
• Auxiliary input files
The following notes apply to the use of the single post file option:
• Although general DDM jobs support post revisions 7 and higher, the ability to generate a single
post file from a DDM job is restricted to post revisions 9 and higher.
• Reading a single post file back into a Marc DDM job using the -pid command line option is
only available for the AXITO3D and PRE STATE options in this release.
• In the case of single input file jobs with continuous post files where the POST option follows the
RESTART option, both jobs should either use multiple post files or single post files. If multiple
post files are used, then the number of domains should be the same in both jobs.
Matrix Solvers
All matrix solvers in Marc are supported in conjunction with DDM parallel mode with the exception of
the unsymmetric solver and the mixed direct-iterative. Only the multifrontal solver (solver type 8)
supports out-of-core solution in parallel. The matrix solvers work somewhat different in parallel mode
as compared to serial mode.
For some of the direct solvers (type 0, 4, and 8), a two-stage approach is used. In a mechanical analysis,
all inter-domain nodes are first given fixed displacements and each domain solves the resulting equation
system independently. The second stage operates on the interdomain nodes to relax the extra imposed
fixed displacements. This stage is not necessarily if there are no connections between the meshes in the
different domains. An iterative procedure based upon the conjugate gradient method is performed for this
stage. This procedure operates as mentioned on the inter-domain nodes. The number of conjugate
gradient iterations needed to converge to the final solution is reported in the Marc output file as “nn inter-
domain iterations”.
The conjugate gradient iterative solver (solver type 2) operates simultaneously on the whole model. It
works to a large extent like in a serial run. For each iteration cycle, there is a need to synchronize the
residuals from the different domains.
The third type of solver supported is the hardware solver (SGI), Pardiso, and MUMPS. For this case, the
global stiffness matrix (or the corresponding matrix for nonstructural analysis) is gathered to the parent
process (where the job was started). The original child processes are put to sleep and the solver creates
new processes (multiple threads) to solve the equation system in parallel. When the solution is finished,
the child processes are awakened and the data related to each domain are distributed. The hardware
solver, Pardiso solver, or MUMPS may be used in conjunction with DDM on a shared memory
architecture. For the hardware or Pardiso solver, multithreading is used to obtain the performance
improvement. For distributed memory or cluster environments, the MUMPS solver may be used in
conjunction with DDM.
CHAPTER 13 779
Parallel Processing
Contact
Contact analysis poses a special challenge for parallelization due to its global nature. The current version
of Marc fully supports contact analysis including thermal and electrical contact but excluding beam-to-
beam contact.
In order to enable a fast contact search procedure, some contact data are duplicated across all domains.
A node in one domain must be able to find contact with a segment in a different domain. The duplicated
data includes nodes on the exterior boundary of deformable contact bodies. This data duplication is not
necessary in a pure rigid contact job, so it is important to define a contact table to indicate that no
deformable contact can take place in this case. The default, if no contact table is used, is that all
deformable bodies check for contact with all other bodies including itself.
If a node touches a segment which is located in a different domain, the nodes making up the segment are
imported into the domain of the touching node. If the node separates, the nodes are deleted from the
domain. This means that the number of nodes in a domain can dynamically change. This procedure
allows the creation of tyings spanning domains. This also has implications on how the domains should
be created. One should avoid having contact bodies with a large number of nodes coming into contact
across different domains. An extreme case would be two sheets of equal number of shell elements
coming into contact on top of each other. If two domains are used and each sheet is in a separate domain,
the number of nodes in one domain would double.
Domain Decomposition
The way the model is decomposed into domains, in general, affects the performance of the parallel run.
Traditionally, domain decomposition had to be performed in the preprocessor stage using Mentat or
MD Patran. Both Mentat and MD Patran use Metis to perform the domain decomposition using the
multiple file option. For a single input file case, the domain decomposition can also be performed in
Marc.
The general goal for a good decomposition is to minimize the number of interdomain nodes. Keeping
this number low minimizes the overhead in having duplicate nodes and also minimizes the size of the
system solved in the interdomain part of the direct solver. It is important to get a good balance between
the workload in the different domains if, for instance, different types of elements are used or if the nodes
of a cluster have different performance.
Another aspect of deformable contact as mentioned above is to avoid having a large number of nodes
touching segments of a different domain.
Several different methods for creating the decomposition are available in Marc (single input), Mentat,
and MD Patran. One should note that some of them are based on element connectivity so they tend to
put separate contact bodies in separate domains which is not always optimal.
The optimal number of domains to use for a given job is pretty hard to estimate. It obviously depends on
the number of processors available. There is, in general, an optimal number of processors to use for a
given job and a given machine or cluster of machines. A general rule is that the larger the model is, the
more it pays off to use more processors.
Running a Parallel Job

Running a Marc parallel job is not much different from running a serial job. The model can be
decomposed into domains by the preprocessor (Mentat or MD Patran) and a separate input file is created
for each domain. From a user’s perspective, the extra steps consist of activating the procedure for
decomposing the model and activating the parallel run.
Marc can also be used to create the domains in case a single input file containing the whole model is used.
Two methods can be used to trigger the single input file option; the first is through the command line
option as explained below and the second is through the use of the PROCESSOR parameter.
If the command line option is used and the PROCESSOR parameter is not defined, the default domain
decomposition method, Metis Best, is used. Currently, the single input file option does not support any
additional features that are not supported by the traditional multiple input file option as described by the
list of unsupported features above.
For the case of a run on a cluster, one would additionally have to define which machines make up the
cluster (unless an automatic cluster tool is used). The run is started as usual and, by default, separate
results files are created for each domain. The postprocessor automatically picks up the results from the
different domains and combines them into a single model which looks the same as in a serial run. One can
also choose to look at the results of each domain separately, which is particularly useful for large jobs.
As opposed to the default multiple post file option, the user also has the option to request a single post
file for the whole model. For details on how to use this functionality, refer to Marc Volume C: Program
Input for the POST model definition option.
When running a parallel job from Marc Mentat or MD Patran, one has to activate the button for parallel
and start the job as for a serial job. From the command line, one needs to use an extra argument to specify
the number of domains:
run_marc -v n -b n -j jobname -np 5
for running a job with five domains using multiple input files generated by a preprocessor. To indicate
that a single input file will be used, the -np option is replaced with -nps and the command line in this
case becomes:
run_marc -v n -b n -j jobname -nps 5
User subroutines are handled in the same way as in a serial run. See Marc Volume D: User Subroutines
and Special Routines for a description on special considerations for user subroutines in a parallel run.
Running on a cluster (a set of connected computers) requires that the machines communicate properly.
The actual connection could be anything from a telephone modem up to a high speed direct connected
cluster. A standard 100 Mbit Ethernet network is usually sufficient for reasonable performance, although
it can become bottlenecked with fast processors. Faster connections give better performance on the
communication part.
In order to perform an analysis over a network, one needs to specify which machines are to be part of the
analysis. This is done through a so-called host file, which is generic for a Marc run. The Marc startup
script converts it into a format that the current MPI implementation expects. The host file is a simple text
file that lists the computers that are to be used in the run and some info about the run. The host file
typically looks something like (using a Unix system)
CHAPTER 13 781
Parallel Processing
host1 2
host2 1 /users/joe/run /programs/marc
host3 2 /usr/people/joe /programs/marc/marc
Here host1, host2, and host3 are the host names of the respective machines. It specifies a job with
five domains where two processes are run on host1 and host3 and one on host2. /users/joe/run
is the directory where the job is run on host2. This is where the input files are copied and where the
results files are created. This directory could be a local directory on a hard drive on host2 or it could be
a shared directory (via NFS for example). This directory must be accessible from host2 (but not
necessarily from the other hosts). The fourth entry of the host file specifies where the Marc installation
directory is located. This can also be a local directory (if Marc is installed on the local machine), or it
can point to a shared directory. It is recommended that the run directory be a local directory to avoid I/O
traffic over the network. Not all MPI implementations currently supported uses the fourth entry.
Specifically, the commonly used HP-MPI and Intel MPI (used on Linux, HP-UX and Windows) do not
support the specification of different installation directories. They need to use the same path on all hosts.
They can be shared or local but the path name must be the same.
The input files for the remote hosts are automatically copied to each host in case the run directory is local.
When the run is finished, the post files are copied back. Please note that only the input files and the
post files (not output files, etc.) are copied. This copying can be suppressed by buttons in Mentat and
MD Patran or by a command line option if Marc is manually started. If a user subroutine is used together
with local run directories, the new executable is automatically transferred to all hosts before the Marc
job starts.
One can also set up Marc to run via external cluster tools. A useful run time option in this context is the
command line option -dir. It specifies the run directory for the job and is where the scratch files and
results files are created. Suppose Marc is set up to run on a cluster using some kind of cluster tool to
distribute the processes and each machine in the cluster has a local disk called /scratch. By using the
-dir /scratch option from the command line, the run uses the local directory during the run. The
input files are still read from where the job is started (which could be a shared directory on a control
workstation which is not even part of the run).
The following command line options related to parallel processing are available for the Marc run_marc
start-up script:
-np nn specify nn number of processes.
-nps nn specify nn number of processes if using a single input file.
-ci y|n if y, copy input files to local run directories; if n, do not copy.
-cr y|n if y, copy post files from local run directories; if n, do not copy.
-ho hostfile specify the host file that describes the hosts to use in a network run.
-dir rundir specify that the run directory of the run is rundir.
-mpi specifies the MPI version to use if other than the default.
Domain Decomposition Methods

Metis Best - method performs Metis element based, Metis node based, and Vector decomposition and
picks the best decomposition with respect to the number of interdomain nodes as well as the domains
balance. (Figure 13-1)
Metis_element_based - performs decomposition of graph constructed from element connectivity.
Metis_node_based - performs decomposition of graph constructed from nodal connectivity.
Figure 13-1 Metis Best Decomposition
Vector - performs decomposition based on IDs of elements. The element ids are sorted by coordinates of
the element centroids. The default direction of sorting is in the maximal dimension of the model
bounding box in x-, y-, and z-directions respectively. User may also specify direction of sorting by
supplying the direction vector. See Marc Volume C: Program Input, Chapter 2, Parameters –
PROCESSOR. (Figure 13-2)
Figure 13-2 Vector Decomposition with Specified User Direction (1,1,0)

CHAPTER 13 783
Parallel Processing
Radial - performs decomposition based on IDs of elements. The element ids are sorted by distance
between the element centroids and axis of rotation. The default (axis of rotation) is specified by default
direction and default point on the axis of rotation.
Default point for axis of rotation is defined as the centroid of the model bounding box. User may also
specify the axis of rotation by point and direction. See Marc Volume C: Program Input, Chapter 2,
Parameters – PROCESSOR. (Figure 13-3)
Figure 13-3 Radial Decomposition with User Direction(.5,.5,.707) and

Point on Axis (2,2,2.141)
Angular - performs decomposition based on ids of elements. The element ids are sorted by angle
between the element centroids, axis of rotation and selected start vector. The default (axis of rotation) is
specified by default direction and default point on the axis of rotation. User may also specify the axis of
rotation by point and direction. See Marc Volume C: Program Input, Chapter 2, Parameters –
PROCESSOR. (Figure 13-4).
Figure 13-4 Angular Decomposition with User Direction(.5,.5,.707) and Point on Axis (2,2,2.141)
Recursive Coordinate Bisection - performs recursively vector decomposition into two domains. Each
domain is subsequently subdivided into two or more domains until all domains are created (Figure 13-5).
Figure 13-5 Recursive Coordinate Bisection
User - performs decomposition based on element sets in the input file. The element sets must be named
domaini where i is the domain number. The domain sets are usually created by the user in the
preprocessor program.
CHAPTER 13 785
Parallel Processing
The Fine Graph flag in the 2nd data block on the PROCESSOR parameter tells Metis to use a fine graph
for decomposition. This results in higher memory requirements and lower speed for the decomposition.
In some cases, it produces better decomposition than the default (Coarse Graph). The Coarse Graph
choice might produce more islands. It is recommended to use the default unless there is a need for better
quality of domain decomposition.
Sometimes the domain decomposition produces disconnected domains resembling islands (Figure 13-6).
The best performance usually is obtained when the domains are contiguous. Islands can be added to
neighboring domains by turning on the Island Remove flag in the 2nd data block on the PROCESSOR
parameter. This might result in worse balance of domains as shown in Figure 13-7.
Figure 13-6 Metis Node Based Decomposition without Island Removal
Figure 13-7 Metis Node Based Decomposition with Island Removal

Chapter 14 Code Coupling Interfaces
14 Code Coupling Interfaces

Code Coupling 788
 Coupling Regions 790

References 793
This chapter describes the coupling interface to external solvers available through the user subroutine
programming. The interface allows code coupling libraries such as MpCCI (see [Ref. 1]) to couple Marc
with commercial CFD codes, such as STAR-CD and Fluent. The interface may also be used for
developing dedicated interfaces to external solvers, for applying complex boundaries to certain regions
of the model, or for dedicated postprocessing.
Code Coupling
Analyzing fluid-structure interaction problems has become increasingly important in industrial
applications. Effects of the deforming structure on the fluid flow and, vice versa, of the fluid pressure on
the structure, as well as heat transfer between fluid and structure can be of critical importance for a
design, especially if the deformations of the structure are large.
The fluid-structure capabilities of Marc (see Chapter 6 Nonstructural Procedure Library in this manual)
are limited to nonreactive, incompressible, single phase, laminar flows. This implies that interaction
problems such as combustion reactions in airbag gas generators, involving complex turbulent flows with
large deformations of the structure cannot be solved. An additional drawback is that, at the fluid-structure
interface, the fluid and structural meshes must share the same nodes. This may be inconvenient, as the
fluid analysis tends to require a denser mesh than the structural analysis.
Code coupling software, such as the MpCCI (Mesh-based parallel Code Coupling Interface) software
from Fraunhofer SCAI (see [Ref. 1]), allows structural codes such as Marc to be coupled with
commercial CFD codes such as STAR-CD or Fluent. Fluid-structure interaction problems can then be
solved using the capabilities of both solvers in their respective domains. The two solvers run
simultaneously and exchange data regularly during the analysis on the common boundary between the
fluid and the structure.
Figure 14-1 shows a flow chart of such a coupled analysis, in which the CFD solver and the structural
solver alternately solve their respective parts of the coupled problem. The CFD solver solves the fluid
flow problem for a given period of time (the coupling time step) to provide the pressure that the fluid
exerts on the structure, the temperature of the fluid at the common boundary with the structure and the
film coefficient to the structural solver. The structural solver applies the fluid pressure and the film
boundary condition to the structure and computes the deformations and temperatures of the structure for
the same period of time. Then it sends the current coordinates of the nodes and the temperature at the
fluid-structure interface to the CFD code. The CFD code uses these coordinates to update the flow
domain and continues with the next coupling time step. Note that the coupling time step may consist of
multiple time steps in each of the solvers.
CHAPTER 14 789
Code Coupling Interfaces
Code Coupling Software

CFD Analysis Structural Analysis
CFD Structural
Mesh Mesh
Initialization Initialization
CFD
Time Step(s)
Fluid Pressure, Fluid
Temperature, Film
Coefficient Coupling
Time Step
Structural Time
Step(s)
Current Coordinates,
Structural Temperature
CFD
Time Step(s)
Fluid Pressure, Fluid
Temperature, Film
Coefficient
Coupling
Time Step
Structural Time
Step(s)
Current Coordinates,
Structural Temperature
Figure 14-1 Flow Chart of a Coupled Fluid-Structure Analysis Using Code Coupling Software
The code coupling software serves as a communication layer between the two codes. It handles the data
transfer between the solvers and interpolates the data from the CFD mesh to the structural mesh and vice
versa, in case the meshes do not match at the common boundary (Figure 14-2). To be able to interpolate
the data from one mesh to the other, the code coupling software determines the geometrical relationship
between the meshes, using information (connectivity and coordinates) provided by the two solvers at the
start of the analysis.
CFD Mesh
Structural Mesh
Figure 14-2 Non-Matching Meshes at the Common Boundary Between Fluid and Structure
The basic stages, from a code coupling point of view, in a coupled analysis are:
1. Initialization. The relationship between the CFD mesh and the structural mesh is established. Both
the CFD solver and the structural solver must provide the code coupling software with the
connectivity and the coordinates of the patches on the common surface between the fluid and the
structure.
2. Data exchange. Physical quantities such as the fluid pressure and the temperature of the fluid at
the common surface with the structure, are extracted from the CFD code, interpolated to the
structural mesh, transferred to the structural code and applied through appropriate boundary
conditions to the structural mesh. Likewise, current coordinates and temperature of the structure
at the common surface with the fluid are extracted from the structural code, interpolated to the
CFD mesh and transferred to the CFD code. The CFD code uses the new coordinates to update
the flow domain.
3. Finalization.
Coupling Regions
Coupling regions are the basic concept to couple Marc with external solvers via code coupling software
such as MpCCI. A coupling region is that part of the surface or volume of the structural model where the
interaction with the external solver takes place. A surface region consists of a list of edges or geometric
curves in 2-D and a list of faces or geometric surfaces in 3-D. A volumetric region consists of a list of
elements or contact bodies. Coupling regions are defined by the COUPLING REGION model definition
option and are accompanied by an extensive application programming interface (API). The basic
CHAPTER 14 791
mechanical and thermal quantities (current coordinates, displacement, force, pressure, temperature,
heat flux, film coefficient) can be exchanged with an external solver on coupling regions via API
calls. Quantities that are received from the external solver are applied to the coupling region via
The coupling region API consists of three user subroutines and a set of utility routines that can be called
from these user subroutines (see Chapter 9 Special Routines - Marc Post File Processor, Marc Volume
D: User Subroutines and Special Routines). The three user subroutines correspond to the basic three
stages of a coupled analysis, as outlined in the previous section:
1. Initialization: CPLREG_INIT. This subroutine is called once at the start of the analysis and can be
used to extract the connectivity and coordinates of the coupling region and pass it to the code
coupling software to determine the geometric relationship between the coupling region mesh and
the mesh of the external solver.
2. Data exchange: CPLREG_EXCHANGE. This subroutine is called twice per coupling time step,
once at the start and once at the end. The call at the start can be used to set the new values of
physical quantities that have been received from the external solver for the next coupling time
step. The call at the end can be used to extract the new values of physical quantities that must be
sent to the external solver.
3. Finalization: CPLREG_FINALIZE. This subroutine is called once at the end of the analysis and
can be used to inform the external solver that the Marc analysis has ended.
The utility routines can be called from the user subroutines to obtain the connectivity and coordinates of
the coupling region, to obtain the current values of physical quantities on coupling regions, to set the new
values of physical quantities for the next coupling time step and even to specify the next coupling time
step. The available utility routines are listed in Table 14-1.
Table 14-1 Available Utility Subroutines for Coupling Regions
Available in User
Utility Subroutine Subroutines Purpose
CPLREG_FIND_NAME CPLREG_INIT Find coupling regions
CPLREG_EXCHANGE by name.
CPLREG_GET_INFO CPLREG_INIT Get general information about
CPLREG_EXCHANGE a coupling region.
CPLREG_GET_QUANTS CPLREG_INIT Get the prescribed physical
CPLREG_EXCHANGE quantities on a coupling
region.
CPLREG_GET_MESH CPLREG_INIT Get the mesh of a coupling
CPLREG_EXCHANGE region.
CPLREG_GET_GLOBAL_VALUES CPLREG_EXCHANGE Get the values of a global
quantity.
CPLREG_GET_NODE_VALUES CPLREG_EXCHANGE Get the values of a
CPLREG_GET_ALL_NODE_VALUES node-based quantity at a
coupling region.
Table 14-1 Available Utility Subroutines for Coupling Regions (continued)

Available in User
Utility Subroutine Subroutines Purpose
CPLREG_PUT_GLOBAL_VALUES CPLREG_EXCHANGE Put the values of a global
quantity.
CPLREG_PUT_NODE_VALUES CPLREG_EXCHANGE Put the values of a node-based
CPLREG_PUT_ALL_NODE_VALUES quantity at a coupling region.
CPLREG_PUT_EDGE_VALUES CPLREG_EXCHANGE Put the values of an
CPLREG_PUT_ALL_EDGE_VALUES edge-based quantity at a
coupling region.
CPLREG_PUT_FACE_VALUES CPLREG_EXCHANGE Put the values of a face-based
CPLREG_PUT_ALL_FACE_VALUES quantity at a coupling region.
CPLREG_PUT_ELEM_VALUES CPLREG_EXCHANGE Put the values of an
CPLREG_PUT_ALL_ELEM_VALUES element-based quantity at a
coupling region.
Time Step Control

The default coupling time step is equal to the Marc time step, so that the CPLREG_EXCHANGE user
subroutine is called at the start and at the end of each increment. If the AUTO STEP time stepping scheme
is used, the coupling time step may be prescribed by calling the CPLREG_PUT_GLOBAL_VALUES
utility routine from the CPLREG_EXCHANGE user subroutine. In that case, the AUTO STEP algorithm
matches the next coupling time, but multiple increments may be needed to reach this coupling time. If
multiple increments are performed within a given coupling time step, then the CPLREG_EXCHANGE
user subroutine is called at the start of the first increment and at the end of the last increment of the
coupling step.
Shell Elements
The top and bottom faces of a shell element are independent surface entities, as far as coupling regions
are concerned. They can both be part of the same coupling region or they can be part of two different
coupling regions. The faces share the same nodes, but the connectivity of the bottom face, as returned by
the CPLREG_GET_MESH utility routine, is exactly the opposite of that of the top face (which is the
connectivity of the element).
In heat transfer and coupled thermo-mechanical analyses, the temperatures (and nodal heat fluxes) at the
top and bottom of a shell element are distinct. If a coupling region contains only top faces of shell
elements, the CPLREG_GET_NODE_VALUES utility routine returns the top temperatures (and heat
fluxes) at the nodes of the region. Similarly, the CPLREG_PUT_NODE_VALUES utility routine
prescribes the top temperatures (and heat fluxes) at the nodes of the region. If the coupling region
contains only bottom faces of shell elements, these routines return and prescribe the bottom temperatures
(and heat fluxes) at the nodes. If both the top and the bottom faces of shell elements interact with the mesh
CHAPTER 14 793
of an external solver, two coupling regions must be defined to properly access the temperatures and nodal
heat fluxes on both sides of the shell elements: one containing the top faces and the other the bottom faces
of the shells.
Parallel Processing
In a parallel run with Marc, the finite element mesh is subdivided into domains where each element is
part of one domain. Nodes at the boundary between domains are present in all domains sharing that
boundary. Coupling regions are allowed to span multiple domains. If a physical quantity is prescribed on
a such a coupling region, using the CPLREG_PUT_NODE_VALUES utility routine, the value of the
quantity for a given node must be the same in all domains in which the node appears. Similarly, if the
coupling time step is prescribed via CPLREG_PUT_GLOBAL_VALUES, the value must be the same in
all domains.
References
1. Fraunhofer Institute for Algorithms and Scientific Computing SCAI. MpCCI 3.0.4: Manuals and
Tutorials. April 25, 2005, http://www.scai.fraunhofer.de/mpcci.
Appendix A Finite Element Technology in Marc
A Finite Element Technology in

Marc
 Governing Equations of Various Structural Procedures 796


System and Element Stiffness Matrices 798

Load Vectors 799
 References 800
Governing Equations of Various Structural Procedures

This section describes the basic concepts of finite element technology. For more information, you are
urged to refer to the finite element text [Ref. 5]. Marc was developed on the basis of the displacement
method. The stiffness methodology used in Marc addresses force-displacement relations through the
stiffness of the system. The force-displacement relation for a linear static problem can be expressed as
Ku = f (A-1)
where K is the system stiffness matrix, u is the nodal displacement, and f is the force vector.
Assuming that the structure has prescribed boundary conditions both in displacements and forces, the
governing Equation (A-1) can be written as
K 11 K 12  u 1   f1 
  =   (A-2)
K 21 K 22  u 2   f2 
u 1 is the unknown displacement vector, f 1 is the prescribed force vector, u 2 is the prescribed
displacement vector, and f 2 is the reaction force. After solving for the displacement vector u , the strains
in each element can be calculated from the strain-displacement relation in terms of element nodal
displacement as
ε e l = βu e l (A-3)
The stresses in the element are obtained from the stress-strain relations as
σ e l = Lε e l (A-4)
where σ e l and ε e l are stresses and strains in the elements, and u e l is the displacement vector associated
with the element nodal points; β and L are strain-displacement and stress-strain relations, respectively.
In a dynamic problem, the effects of mass and damping must be included in the system. The equation
governing a linear dynamic system is
Mu·· + Du· + Ku = f (A-5)

where M is the system mass matrix, D is the damping matrix, Equation (A-6) is the acceleration vector,
and u is the velocity vector. The equation governing an undamped dynamic system is
Mu·· + Ku = f (A-6)
The equation governing undamped free vibration is
Mu·· + Ku = 0 (A-7)
APPENDIX A 797
Finite Element Technology in Marc
Natural frequencies and modal shapes of the structural system are calculated using this equation.
2
Kφ – ω Mφ = 0 (A-8)
The equations governing some other procedures are similar. For example, the governing equation of
transient heat transfer analysis is
·
CT T + KT T = Q (A-9)
where C T is the heat capacity matrix, K T is the thermal conductivity matrix, Q is the thermal load
vector (flux), T is the nodal temperature vector, and t is the time derivative of the temperature. Equation
(A-9) reduces to
KT T = Q (A-10)
for the steady-state problem. Note that the equation governing steady-state heat transfer (Equation
(A-10)) and the equation of static stress analysis (Equation (A-1)) take the same form. Similarly, the
hydrodynamic bearing problem is analogous to a steady-state heat transfer problem. This problem is
governed by an equation similar to Equation (A-10).
The matrix equation of the electrical problem in coupled thermo-electrical analysis is
K E ( T )V = I (A-11)
The equation governing the thermal problem is:
· E
C T ( T )T + K T ( T )T = Q + Q (A-12)
in both Equation (A-11) and Equation (A-12), v is the voltage, K E ( T ) is the temperature-dependent
electrical-conductivity matrix, I is the nodal-current vector, C T ( T ) is the temperature-dependent, heat-
capacity matrix, and K T ( T ) is the thermal-conductivity matrix. T is the nodal temperature vector, Q
E
is the heat-flux vector, and Q is the internal heat-generation vector that results from the electrical
E
current. Electrical and thermal problems are coupled through K E ( T ) and Q .
The matrix equations for the thermal-mechanical problem are as follows:
Mu·· + Du· + K ( T ,u ,t )u = F (A-13)

· I
C T ( T )T + K T ( T )T = Q + Q + Q F (A-14)
In Equations (A-13) and (A-14) the damping matrix D , stiffness matrix K , heat-capacity matrix C T and
I
thermal-conductivity matrix K T are all dependent on temperature. Q is the internal heat generated due
to inelastic deformation. The coupling between the heat transfer problem and the mechanical problem is
due to the temperature-dependent mechanical properties and the internal heat generated. If an updated
Lagrangian analysis is performed, K and K T are dependent upon prior displacement.
The governing equations described above are either sets of algebraic equations (Equations (A-1), (A-10), and
(A-11)), or sets of ordinary differential equations (Equations (A-5) through (A-8) and all equations
through Equation (A-4)). The time variable is a continuous variable for the ordinary differential
equations. Select an integration operator (for example, Newmark-beta, Houbolt, or central difference for
dynamic problems, and backward difference for heat transfer) to reduce the set of differential equations
to a set of algebraic equations. The final form of governing equations of all analysis procedures is,
therefore, a set of algebraic equations.
System and Element Stiffness Matrices

The previous section presented system matrices assembled from element matrices in the system. For
el
example, the system stiffness matrix K is expressed in terms of the element stiffness matrix K i as
N
el
K =  Ki (A-15)
i = 1
where N is the number of elements in the system. The system stiffness matrix is a symmetric-banded
matrix. See Figure A-1 for a schematic of the assemblage of element matrices. When the element-by-
element iterative solver is used, the total stiffness matrix is never assembled.
The element stiffness matrix can be expressed as
el T el
K =  β Lβdv (A-16)
el
v
el
where v is the volume of the element, β is the strain-displacement relation, and L is the stress-strain
relation.
ε = βu (A-17)
σ = Lε (A-18)
The mass matrix can be expressed as
el T
M =  N ρNdv (A-19)
el
v
APPENDIX A 799
Finite Element Technology in Marc
1 2
2 3
1
(a) 3 4 5
4 5
6 8
7
1 2 3 4 5
(b) δ
[K] {F}
(c)
δ
Band
Figure A-1 Schematic of Matrix Assemblage
The integration in (A-16) is carried out numerically in Marc, and is dependent on the selection of
integration points. The element stiffness matrix can be fully or under integrated. The mass matrix is
always fully integrated.
Note: The β matrix is directly associated with the shape functions and geometry of each element.
Shape functions associated with different element types are explained in Marc Volume B:
Element Library. Stress-strain relations L are discussed in Chapter 7 of this volume.
Load Vectors
The nodal force vector f in (A-1) includes the contributions of various types of loading.
f = f p o i n t + f s u r f a c e + f b o d y + f∗ (A-20)
where f point is the point load vector, f surface is the surface load vector, f body is the body (volumetric)
load vector, and f* represents all other types of load vectors (for example, thermal and creep strains, and
initial stress).
The point load is associated with nodal degrees-of-freedom and can be added to the nodal force vector
directly. Equivalent nodal force vectors f surface , f body y must be calculated from the distributed
(surface/volumetric) load first and then added to the nodal force vector. In Marc, the computation of
equivalent nodal forces is carried out through numerical integration of the distributed load over the
surface area of volume to which the load is applied. This can be expressed as
T
f s u rf a c e = N pdA (A-21)
A
T
fb o d y = N pdV (A-22)
V
where p is the pressure.

Figure A-1 shows the assemblage of the nodal force vector.
References
1. Zienkiewicz, O. C. and R. L. Taylor. The Finite Element Method (4th ed.) Vol. 1. Basic
Formulation and Linear Problems (1989),) Vol. 2. Solid and Fluid Mechanics, Dynamics, and
Nonlinearity (1991) McGraw-Hill Book Co., London, U. K.
2. Bathe, K. J. Finite Element Procedures, Prentice-Hall, Englewood Cliffs, NJ, 1995.
3. Hughes, T. J. R. The Finite Element Method–Linear Static and Dynamic Finite Element Analysis,
Prentice-Hall, Englewood Cliffs, NJ. 1987.
4. Ogden, R. W. “Large Deformation Isotropic Elasticity: On The Correlation of Theory and
Experiment for Incompressible Rubberlike Solids,” Proceedings of the Royal Society, Vol. A
(326), pp. 565-584, 1972.
5. Cook, R. D., D. S. Malkus, and M. E. Plesha, Concepts and Applications of Finite Element
Analysis (3rd ed.), John Wiley & Sons, New York, NY, 1989.
Appendix B Finite Element Analysis of NC Machining Processes
B Finite Element Analysis of NC

Machining Processes
 General Description 802


NC Files (Cutter Shape and Cutter Path Definition) 802

Intersection Between Finite Element Mesh and Cutter 804
 Deactivation of Elements 804

Adaptive Remeshing 804

Input of Initial Stresses 805
General Description
Machining (or Metal Cutting) is a type of material removal processes widely used to produce a part with
the desired geometry. After removal of the machined material, re-establishment of equilibrium within the
remaining part of the structure causes some distortion due to the relief of the residual stresses in the
machined materials.
Depending on the stress level and geometry of the part, the severe distortion due to machining can result
in high scrap rates and increased manufacturing costs, especially for structures with thin wall or large
thin plates.
A numerical analysis procedure has been developed in Marc to predict the distortion during the 3-D bulk
machining so that engineers can optimize their structure design and machining processes. This procedure
is based on the assumption that effects introduced by the cutting process are local and small compared to
those due to pre-existing residual stresses.
The simulation procedure uses the cutter shape and path information obtained from NC machining data.
Cutter motion information is used to predict the material to be removed by automatically deleting finite
elements that are located within the cutter path. The distortion of the remaining part can be predicted by
re-calculating the equilibrium. An interface has been developed in Marc to convert cutter path data into
the finite elements to be removed. The cutter path data is stored in either the Automatically Programmed
Tools (APT) source or the Cutter Location (CL) data format. The APT source is the NC data output by
CAD software CATIA. The CL data is the cutter location data provided by APT compilers.
For details on how to use this functionality, refer to Marc Volume C: Program Input for the MACHINING
parameter and the DEACTIVATE model definition option and the DEACTIVATE history definition option.
NC Files (Cutter Shape and Cutter Path Definition)

The cutter shape and the cutter path are the key parameters to determining the volume of the material to
be cut. The cutter shape is defined by the CUTTER statement in the APT source or CL files. The expanded
form of the CUTTER statement is as follows:
CUTTER ⁄ d ,r ,e ,f ,α ,β ,h
where the parameters

d The tool diameter.
r The radius of the corner circle; it can be zero or larger than d ⁄ 2 .
e The radial distance from the tool axis to the center of the corner circle.
f The distance from the tool endpoint to the center of the corner circle measured parallel with the
tool axis.
α The angle from a radial line through the tool endpoint to the lower line segment; it is in range of
zero to 90°, and positive.
β The angle between the upper segment and the tool axis; it is in the ranges of -90° to 90°.
h The cutter height measured from the tool endpoint along the tool axis.
APPENDIX B 803
Finite Element Analysis of NC Machining Processes
have the following meaning as shown in Figure B-1.

Tool axis
Upper Line
segment
Envelope h
e
Lower line
r segment
f
α
Endpoint
d
Figure B-1 Definition of Cutter Geometry
The cutter path is calculated based on the motion of the cutter and cutter shape. The motion is defined
by cutter motion statements like GOTO/, GODLTA/, CYCLE/, and DRILL/, etc. In Marc, these motion
types are translated into point-to-point, circular, or drilling motion modes, respectively. For details, refer
to references by Irvin H. Kral and Dassault Systems.
Notes: 1. Marc accepts APT data files written by CATIA V4 and the corresponding CL data files
by APT compilers.
2. The cutter can be of either multi-axis or fixed axis.
3. Two cutter motion types: Point-To-Point (for example, GOTO/ and GODLTA/ statements)
and Cycle (DRILL/…) are directly supported. Circular motions (either in-plane or out-of-
plane) can be interpolated into Point-To-Point motions. The interpolation can be done
inside CATIA by choosing the proper option when writing out APT source file.
4. In addition, some statements in the APT/CL files are skipped. If the user chooses
statements like TRACUT/, COPY/, /MACRO, or /CALL to define cutter motions, they must
be converted into explicit motion statements like Point-to-Point or Cycle (DRILL/)
motions before writing out the APT data file. The cutting process is stopped if statements,
such as /FINI, /STOP, or /END, or the end of APT/CL file is reached.
5. It is assumed that the part and cutter are defined in the same coordinate system. So, if a
part needs to be flipped over between two machining stages, the user can rotate the cutter
but keep the part unchanged in space.
6. If the user wants to change the boundary conditions for some nodes during the cutting
process defined by one APT/CL file, time synchronization should be turned on when
defining the machining load case.
7. For visualization purpose, the user can specify a contact surface as cutter surface. During
postprocessing, user will be able to visualize the cutter motion along the cutting process.
In order to do this, you MUST exclude this cutter surface ID from all kinds of contact
checking during analysis. The user can do this by creating CONTACT TABLE and turn off
the contact body ID of the cutter surface. In addition, it is also necessary to make sure that
the cutter surface is initially set at the location of the origin of the coordinate system and
the orientation is initially set to be in the direction of Z-axis.
Intersection Between Finite Element Mesh and Cutter

When the cutter moves, its path forms the volume of the material to be cut. This volume is calculated
based on the cutter geometry and its path in space. The finite element mesh is checked against this volume
to decide the intersection between the volume and the finite element nodes and elements.
Deactivation of Elements
The automatic deactivation of finite elements depends on the intersection status between the cutter path
and the mesh. The approach is designed for elements that are fully or partially located inside the cut
volume. Elements that are fully located inside the cut volume are automatically deactivated and the stress
that existed before deactivation is released. For elements that are partially intersected by the cut volume,
integration points are checked. In the current version, the check of integration points is limited to the
element centroid only. If element centroid is found inside the cut volume, the associated element is
deactivated. For deactivated elements, the stresses/strains are set to zero permanently.
After the deactivation of elements along the motion of cutter, the part will distort progressively. The
updated geometry of the part is used for the next cutting step. Using this method, the whole machining
process is simulated with Marc. After the machining is finished, the final distortion of the work piece can
be calculated by spring back analysis. The spring back analysis is conducted by removing all fixturing
restraints except for those needed to remove rigid body motion.
Adaptive Remeshing
Generally speaking, the accuracy of the finite element analysis of machining processes depends on, plays
a key role to get the accuracy of initial (residual) stresses. Accurate initial stress the correct deformation
of the part. From the view point of FE analysis, the initial FE mesh may affect the accuracy significantly.
Finer mesh is needed for accuracy but it results in higher analysis cost due to the increased number of
finite elements and increased total number of analysis increments. Also, for most machining processes,
the cutter path is complicated enough that it is not possible to properly define the best suited fine mesh
at the beginning of the analysis. The adaptive remeshing technology has been enhanced in Marc to obtain
accurate results for the FE mesh located inside the cutter path.
APPENDIX B 805
Finite Element Analysis of NC Machining Processes
An element is subdivided if the cutter only cuts part of the element in the current cutting step. For
example, one 3-D brick element generates 8 new elements after first level subdivision and 64 new
elements after the second level subdivision. Therefore, users need to be aware of the potentially large
number of new elements with increased levels of subdivision.
Typical Features for Machining

1. Multiple level subdivision of element in one incremental step:
Typically, Marc subdivides only once per load step. If the 17th criterion of ADAPTIVE model
definition option is used (for NC machining), Marc automatically subdivides the element until the
maximum specified level has been reached. This process may generate a huge number of new
elements and nodes. Consequently, these newly created elements and nodes may result in
excessive memory usage and CPU time, though most of them are immediately cut off. For some
problems, it is possible to cut the part with a few large elements but without sacrifice of accuracy
if only allowing remeshing to be conducted at some specific locations of the part where finer
meshing is required. In such circumstances, it might be reasonable to conduct one round of rough
cutting followed by fine cutting based on finer mesh after adaptive remeshing. Details are
described in Item 3.
2. ADAPTIVE REMSHING at EACH increment:
At the load case of machining, Marc does adaptive remeshing at each increment based on the
intersection of the cuter path and the FE mesh.
3. ADAPTIVE REMSHING at LAST increment:
A cutting analysis in Marc is not done without adaptive remeshing until the cutting process is
completed. However, the cutting process may not be completely finished yet because some
elements that are partially cut by cutter are still retained. In order to cut those elements off, the
cutter path is re-calculated and those elements that are partially cut are subdivided in order to
improve the accuracy of the cutting. In this case, extra incremental steps are added into this
machining load case. The added number of extra load steps equal the maximum subdivision level
that is defined by the ADAPTIVE option.
Input of Initial Stresses

There are various approaches to input the initial stresses into the FE model for machining analysis.
Generally speaking, Marc allows user to input the initial stress through the INIT STRESS model
definition option. It also allows user to import the data through the PRE STATE option. With the PRE
STATE option, the user can transfer information, including strain, stress etc., from previous analysis into
the model for new analysis.
By using option of INIT STRESS, user may define the stress based on the input values or tables (if the
stresses are defined as function of, say, coordinates). Marc also allows the user to define the initial
stresses with an ASCII format text file. The text file contains the raw initial stress data of points randomly
located in a space field. In Marc, the given initial stress data is processed and interpolated into the finite
element mesh according to the location of each element. For details about this data structure and format,
see INIT STRESS model definition option in Marc Volume C: Program Input.
References
1. Irvin H. Kral, Numerical Control Programming in APT, Prentice-Hall, 1986.
2. Dassault Systems, CATIA Training Guide, Version 4, Release 1.6, April, 1996.
3. Houtzeel Manufacturing systems, Inc., APT Language Reference Manual, March, 1997.
MSC.Fatigue Quick Start Guide
Index
Marc Volume A: Theory and User Information
ABCDEFGHIJKLMNOPQRSTUVWXYZ
electristatic-structural 539
Index
I N D E X
A electromagnetic 316, 538, 632, 704, 764

ablation 259 electromagnetics 762
ABLATION (parameter) 259, 260 electrostatic 538, 631, 703, 762, 764
acoustic electrostatic-structural 764
analysis 324, 539, 631, 703, 762, 764 electro-structural 762
element types 324, 342 error 79
input options 631 fluid 325, 539, 704
ACOUSTIC (parameter) 325 fluid/solid interaction 342, 538, 703
ACTIVATE (history definition option) 59 Fourier 102
ADAMS 770 harmonic 319
ADAPT GLOBAL (model definition option) 260 heat transfer 230, 537, 628, 762, 763
ADAPTIVE (model definition option) 82, 101, 102, hydrodynamic bearing 296, 538, 630, 704, 764
710 Joule heating 764
ADAPTIVE (parameter) 710 large strain 707
adaptive mesh linear 100
contact 578 magnetostatic 304, 538, 704, 762, 764
global 85 nonlinear 105
local 82 perturbation 123
refinement 710 piezoelectric 320, 537, 631, 704, 762, 764
adaptive time control 141 post buckling 714
ALL POINTS (parameter) 291, 698 rigid cavity acoustic 324
ALLOCATE (parameter) 32 rigid-plastic 190, 707, 764
analysis 352, 538 snap-through 714
acoustic 324, 539, 631, 703, 762, 764 soil 195, 512, 704
contact 764 steady state rolling 213
coupled 335, 569 stress 763
coupled acoustic-structural 339 structural zooming 216
coupled contact 569 thermal contact 571
coupled electromagnetic-thermal 355 thermo-electrical 538
coupled electrostatic-structural 703 analytical contact 584
coupled thermo-electrical (Joule heating) 348 ANELAS (user subroutine) 101, 361, 364
coupled thermo-mechanical 337 ANEXP (user subroutine) 101, 364, 507
crack 147 ANISOTROPIC (model definition option) 101, 293,
creep 714 321, 361, 363, 367, 389, 443, 467, 469, 505
design sensitivity 202 anisotropic viscoelastic material 494
dynamic 763 ANISTROPIC (model definition option) 468
eigenvalue 168 ANKOND (user subroutine) 231, 537
electrical 629 ANPLAS (user subroutine) 364, 443
application phase 164 post file 668
APPROACH (history definition option) 550 supported elements 668
Arbitrary Eularian-Lagrangian (AEL) formulation beam-to-beam intersection tying constraints 641
119 BEARING (parameter) 296
arc-length methods 729 BEGIN SEQUENCE (history definition option) 627
ASSEM LOAD (model definition option) 771 behavior
ASSUMED STRAIN (parameter) 709 incompressible 125
assumed strain formulation 709 nearly incompressible 127
ATTACH EDGE (model definition option) 81, 618 nonlinear 713
ATTACH FACE (model definition option) 618 Bergan and Mollestad 723
AUTO CREEP (history definition option) 141, 143, B-H RELATION (electromagnetic, model definition
146, 147, 490, 713, 714 option) 320, 538
AUTO INCREMENT (history definition option) 72, B-H RELATION (magnetostatic, model definition
122, 512, 621, 713, 714, 716, 733 option) 305, 306, 538
AUTO LOAD (history definition option) 72, 346, Bingham fluid 327
348, 499, 621, 713 block definition 59
AUTO STEP (history definition option) 146, 177, body motion 548
232, 236, 346, 348, 355, 356, 494, 499, 512, 584, boundary conditions
590, 627, 713, 714, 715, 717, 718, 719, 720, 722, input options 634
723, 724 loading types 605
AUTO THERM (history definition option) 627, 713, time-dependent 184
714, 715 treatment of 128
AUTO THERM CREEP (history definition option) breaking glue 567
145, 713, 714 BUCKLE (history definition option) 122, 123
automatic global remeshing 85 BUCKLE (parameter) 105, 122
Automatic Thermally Loaded Elastic-Creep/Elastic- BUCKLE INCREMENT (model definition option)
Plastic-Creep Stress Analysis (AUTO THERM 105, 122, 123
CREEP) 145 buckling analysis 122
Automatically Programmed Tools (APT) 802 bushing
AUTOMSET (parameter) 654 coordinate system 656
AUTOSPC (parameter) 568 bushings 655
AXITO3D (model definition option) 212 nodal offsets for elements 656
properties 657
B C
bandwidth optimization 70 Cam-Clay model 513
Barlat’s (1991) yield function 444 capacitor 302
beam elements 701, 760 Carreau model 328
pin code 669 CASE COMBIN (model definition option) 101, 103,
BEAM SECT (parameter) 74, 684, 685, 701 725
beam-shell offsets 666 CASI iterative solver 190, 741, 743, 744, 745
adaptive meshing 668 cavity
contact 669 pressure loading 622
dynamics 669 radiating 283
field analysis 668 rigid, acoustic analysis 324
offset vector specification 667 surface elements 626, 709
Index 809
CAVITY DEFINITION (model definition option) 276, CONRAD GAP (model definition option) 231
286, 288, 630 constant dilatation elements 707
central difference operator 178, 180, 181, 182 constitutive relations 470
CENTROID (parameter) 112, 115, 121, 291, 627, constraint
698, 725, 736, 758, 763 rigid link 646
CFAST (model definition option) 655, 672, 689, CONSTRAINT (model definition option) 63
691, 692, 693, 694, 695 contact
Chaboche 473, 488 adaptive meshing 578
CHANGE PORE (model definition option) 199 analysis 764
CHANGE STATE (history definition option) 145, analytical 584
715 beams 574
CHANGE STATE (model definition option) 34, 231, bodies 542
291, 627 constraint implementation 554
CHANNEL (model definition option) 231, 336 coupled 569
classical lamination theory, multi-Layered shells deformable 587
366 detection 550
closest point projection procedure 136 dynamic 577
code coupling friction modeling 556
basic stages 790 mathematical aspects 580
software 788 neighbor relations 553
COEFFICIENT (model definition option) 658 optimize ordering 545
COHESIVE (model definition option) 706 quadratic 574
COHESIVE (with TABLE Input - Thermal) 706 release 568
cohesive zone 529 segment-to-segment 592
coil current 310 separation 568
COIL CURRENT (magnetostatic, model definition shell 552, 576
option) 311, 313, 354 thermal 571
coking 253 tolerance values 579
linear model 255 CONTACT (model definition option) 119, 120, 128,
water drying model 255 178, 231, 275, 346, 348, 542, 544, 575, 576, 581,
combined hardening 468 605, 653, 659, 718
common material characteristics 123 CONTACT CHANGE (rezoning option) 577
COMPOSITE (model definition option) 210, 362, CONTACT NODE (model definition option) 543
363 contact penetration criteria, remeshing 90
composite continuum elements 363 CONTACT TABLE (history definition option) 550,
composite material 362 551, 569
conditioning number 742 CONTACT TABLE (model definition option) 556,
CONN FILL (model definition option) 70 565, 566, 567, 571, 576, 589
CONN GENER (model definition option) 70 CONTINUE (history definition option) 44, 50, 51
CONNECT (model definition option) 64 continuum composite elements 708
connection methods 670 interlaminar stresses 386
CONNECTIVITY (model definition option) 59, 69, continuum elements 700
337, 353, 354, 355, 655, 673, 679, 685, 686, 689, CONTROL (electromagnetic, model definition
725 option) 316
Conrad Gap 289
CONTROL (heat transfer, model definition option) cracking
337 closure 512
CONTROL (history definition option) 145, 233, 662, low tension 511
667, 713 uniaxial 510
CONTROL (magnetostatic, model definition option) Craig-Bampton 770
305 creep 140
CONTROL (model definition option) 112, 123, 232, buckling 144
283, 580, 662, 667, 713, 718, 755 control tolerances 142
control algorithm 72 implicit formulation 486
convergence controls 170 Maxwell model 480
CONVERT (model definition option) 338, 354, 355, CREEP (model definition option) 146, 480, 481,
356 482, 487, 490, 506
COORD SYSTEM (model definition option) 371, CREEP (parameter) 144, 146, 147, 479, 486, 487,
549, 656 489
COORDINATES (model definition option) 59, 69, creep analysis 714
685, 725, 760 CREEP INCREMENT (history definition option) 713
Coulomb friction model 556 Crisfield 732, 734
arctangent model 558 criteria
bilinear model 562 equivalent values 84
stick-slip (modified step function) model 560 mean strain energy 82
COUPLE (parameter) 337, 353 node within a box 84
coupled acoustic-structural analysis 339 nodes in contact 84
acoustic medium 342 previously refined mesh 85
divergence theorem of Green 340 remeshing 90
fluid pressure 341 temperature gradient 84
harmonic analyses 339 user-defined 85
coupled analysis 335, 569 Zienkiewicz-Zhu 82
coupled contact analysis 569 Zienkiewicz-Zhu creep strain 83
coupled electrical-thermal-mechanical 352 CRPLAW (user subroutine) 141, 142, 143, 147, 480,
coupled electrical-thermal-mechanical analysis 352 482, 485, 487, 490, 507
coupled electrostatic-structural CRPVIS (user subroutine) 146, 479, 489, 490
analysis 539 crushing 512
coupled emectromagnetic-thermal analysis 355 cure kinetics 221
coupled thermo-electrical analysis (Joule heating) CURE SHRINKAGE (model definition option) 221
348 CURVES (model definition option) 81, 371
coupled thermo-mechanical analysis 337 Cutter Location (CL) 802
coupling interface 788 CWELD (model definition option) 672, 673, 678,
coupling region 679, 680, 681, 682, 684, 685, 686, 688, 689, 691,
parallel processing 793 692, 693, 694, 695
shell elements 792 CWELD patch-to-patch connections 673
time step 792 direct 673
coupling regions 790 indirect 674
crack analysis 147 cyclic loading 627
CRACK DATA (model definition option) 510 cyclic symmetry
tying constraints 658
Index 811
CYCLIC SYMMETRY (model definition option) 659, direct
660 constraints 581
CYLINDRICAL (model definition option) 58, 668 integration 176
mesh definition input 54
methods 742
D DISP CHANGE (history definition option) 145, 338,
DAMAGE (model definition option) 391, 529 353, 651
damage models DIST CHARGE (electromangetic, model definition
cohesive zone 529 option) 316
ductile metals 524 DIST CHARGE (model definition option) 321, 346
elastomers 526 DIST CHARGES (electrostatic, model definition
damping 185 option) 300, 346, 356
DAMPING (model definition option) 172, 175, 186 DIST CURRENT (electromagnetic, model definition
DAMPING COMPONENTS (history definition option) 316
option) 297 DIST CURRENT (history definition option) 353
deact glue 567 DIST CURRENT (Joule, history definition option)
DEACT GLUE (model definition option) 158, 162, 349, 356
567 DIST CURRENT (magnetostatic, model definition
deactivate 59 option) 305, 354
element method 273 DIST CURRENT (model definition option) 353
DEFINE (model definition option) 43, 44, 45 DIST FLUXES (history definition option) 349, 353,
deformable contact, solution strategy 589 356
deformable-deformable contact 587 DIST FLUXES (model definition option) 195, 230,
degree of freedom transformation 634 236, 297, 349, 353, 354
DELAMINATION (model definition option) 166, 167 DIST LOADS (history definition option) 145, 353
DENSITY EFFECTS (model definition option) 450 DIST LOADS (model definition option) 196, 321,
DESIGN DISPLACEMENT CONSTRAINTS (model 346, 348, 353, 619, 621
definition option) 210 DIST LOADS (parameter) 209
DESIGN FREQUENCY CONSTRAINTS (model DIST MASS (model definition option) 293
definition option) 210 DIST SOURCE (model definition option) 325
DESIGN OBJECTIVE (model definition option) 210 Domain Decomposition Method (DDM) 776, 782
design optimization 205 ductile metals 524
DESIGN OPTIMIZATION (parameter) 209, 361 dynamic
DESIGN SENSITIVITY (parameter) 209, 361 analysis 763
design sensitivity analysis 202 crack propagation 160, 164
design space 206 fracture methodology 163
DESIGN STRAIN CONSTRAINTS (model definition impact 577
option) 210 DYNAMIC (parameter) 105, 165, 168, 175, 177
DESIGN STRESS CONSTRAINTS (model DYNAMIC CHANGE (electromagnetic, history
definition option) 210 definition option) 316, 319
DESIGN VARIABLES (model definition option) 210 DYNAMIC CHANGE (history definition option) 175,
diffusion 293 177, 183, 325, 343, 713
DIFFUSION (parameter) 293
dilatational creep 486
E electromagnetic 316
electrostatic 300
eddy 314 fluid 333
Edge I-J orientation 368 Fourier analysis 104
effects of temperature and strain rates on yield stress gap-and-friction 705, 761
462 groups 772
eigenvalue analysis 168 heat transfer 702
ELASTIC (parameter) 79, 100, 101 heat transfer convection 292
elastic models 513 hydrodynamic bearing 296
ELASTICITY (parameter) 101 incompressible 706
elastomeric material 429 interface 167, 529, 581, 706
elastomers 526 magnetostatic 304
electrical membrane 700
analysis 629 piezoelectric 320
input options 631 plate 702
ELECTRO (parameter) 300, 345 quarter point 164
electromagnetic rebar 706
analysis 316, 538, 632, 704, 764 reduced integration 707
element types 316 semi-infinite 708
input options 317, 632 shell 576, 702
electromagnetics special 705
analysis 762 truss 700
electrostatic ELEMENTS (parameter) 168, 673, 686, 689, 698
analysis 538, 631, 703, 762, 764 ELEVAR (user subroutine) 757, 760
Coulomb force 345 ELEVEC (user subroutine) 757
coupled electrostatic-structural 344, 539, 632, EL-MA (parameter) 316, 355
703 ELSTO (parameter) 31, 32, 33, 34, 750
element types 300 EMCAPAC (history definition option) 302, 573
input options 631 EMINDUC (history definition option) 313
electrostatic-structural EMISSIVITY (model definition option) 286
analysis 762, 764 EMLAMIN (history definition option) 316, 354
ELEM SORT (model definition option) 756 EMRESIS (history definition option) 350, 573
element distortion criteria, remeshing 90 EMWINDING (magnetostatic, model definition
elements option) 311, 313, 354
acoustic analysis 324, 342 END (parameter) 50
beam 701, 760 END OPTION (model definition option) 42, 50, 51,
beams, contact 574 209, 325
cavity surface 709 END REZONE (rezoning option) 51
classes of stress-strain relations 364 END SEQUENCE (history definition option) 627
composite continuum 363 Equivalent Values Criterion 84
connectivity data 54 error analysis 79
constant dilatation 707 ERROR ESTIMATE (model definition option) 79,
contact 574 102, 128
continuum 700 Eulerian formulation 118
continuum composite 708 examples of lists input 45
deactivate 273 EXCLUDE (history definition option) 552
Index 813
exponential cap 449, 450, 457, 461 fluid mechanics 325
EXTENDED (parameter) 43 Bingham fluid 327
Carreau model 328
element types 333
F piecewise non-Newtonian flow 327
FAIL DATA (model definition option) 371, 386, 387, power law fluid 328
389, 511 FLUID SOLID (model definition option) 343
FEATURE (parameter) 662, 667 fluid/solid interaction analysis 342, 538, 703
file units 33 FLUX (user subroutine) 231, 236, 238, 297, 300,
FILM (user subroutine) 231, 236, 237, 238, 349, 353, 305, 321, 325, 346, 349, 353, 354, 356, 629, 630
354, 356, 629, 630 FOAM (model definition option) 499
FILMS (model definition option) 230, 236, 247, 275, FOLLOW FOR (parameter) 112, 121, 619, 620, 637,
349, 353, 354, 356, 629 709
FILMS (with TABLE input, model definition option) follow force stiffness contribution 709
630 FORCDF (user subroutine) 356
finite element technology FORCDT (user subroutine) 184, 235, 236, 238, 297,
basic concepts 796 300, 305, 321, 334, 346, 348, 354
FIXED DISP (model definition option) 59, 63, 196, force components 606
321, 338, 343, 346, 348, 353, 651 FORCEM (user subroutine) 81, 184, 316, 321, 346,
FIXED POTENTIAL (electromagnetic, model 348, 354, 356
definition option) 316 Forming Limit Diagram (FLD) 765
FIXED POTENTIAL (electrostatic, model definition Forming Limit Parameter (FLP) 765
option) 300, 356 FOUNDATION (model definition option) 101, 121
FIXED POTENTIAL (magnetostatic, model FOURIER (model definition option) 101, 102, 105
definition option) 305, 354 FOURIER (parameter) 102, 105
FIXED PRESSURE (model definition option) 195, Fourier analysis 102
293, 296, 325 coefficients 103
FIXED TEMPERATURE (model definition option) elements used 104
230, 235, 286, 334, 338, 349, 353, 354, 356 fractal dimension 742
FIXED VELOCITY (model definition option) 334 fracture mechanics 147
FIXED VOLTAGE (model definition option) 349, 353 friction modeling 556
FLD (Forming Limit Diagram) 765 FXORD (model definition option) 65, 69, 702
flow
diagram 746
rule 470 G
FLP (Forming Limit Parameter) 765 gap-and-friction element 705, 761
fluid gasket 393
analysis 325, 539, 704 GASKET (model definition option) 396
elements 333 generalized alpha operator 177, 180
FLUID (parameter) 704 HHT 181
FLUID DRAG (model definition option) 621 spectral radius 181
fluid drag and wave generation phase 164
input options 621 GENT (parameter) 115
loads 621 geometric domains 699
GEOMETRY (model definition option) 74, 210, 296,
666, 669, 682, 683, 684, 685, 702, 707, 709, 760
GLK (model definition option) 217 hydrodynamic bearing
global analysis 538, 630, 704, 762, 764
meshing 577 elements 296
ZX plane orientation 369 mass fluxes 630
GLOBALLOCAL (model definition option) 34, 217 pump pressures 630
glue model, friction 567 restrictors 630
gradient HYPELA2 (user subroutine) 112, 398, 399, 513
objective or response function 202 HYPOELASTIC (model definition option) 398
pressure 196 hysteresis 314, 423, 501, 527
Gurson model, modified 524
I
H IBOOC (parameter) 31, 32, 34
harmonic IMPD (user subroutine) 757
analysis 319 incompressible
response 171 Herrmann elastic formulation 126
HARMONIC (electromagnetic, history definition material behavior 125
option) 320 material modeled 126
HARMONIC (electromagnetic, history definition) incompressible elements 706
356 rigid-plastic flow 707
HARMONIC (parameter) 172, 173 incremental mesh generators 69, 70
Hashin induction
fabric failure criterion 379 heating 355
failure criterion 378 magnetic 47, 304
tape failure criterion 380 motors 314
HEAT (parameter) 291, 292, 354, 355, 703 inertia relief 187
heat fluxes 569, 628 loads evaluation 188
(special) conditions 629 Support method 187
input options 629 initial conditions 183
heat transfer previous analysis result data 212
analysis 230, 537, 628, 762, 763 INITIAL DISP (model definition option) 183
convection 291 INITIAL PC (model definition option) 199
convection elements 292 initial plastic strain 628
elements 702 INITIAL PORE (model definition option) 199
temperature effects 234 INITIAL POROSITY (model definition option) 195,
hereditary integral model 490 196, 293
Hill initial stress and initial plastic strain input options
failure criterion 374 628
yield function 442 INITIAL TEMP (model definition option) 230, 235,
history definition options 50 349, 353, 354, 356
Hoffman failure criterion 375 INITIAL VEL (model definition option) 183
HOOKLW (user subroutine) 101, 361, 364 INITIAL VOID RATIO (model definition option) 195,
HOOKVI (user subroutine) 494 196
host file 37, 745, 780 input conventions 42
Houbolt operator 177
hydrodoynamic bearinh 762
Index 815
input options
acoustic analysis sources 631
J
boundary conditions 634 J-INTEGRAL (model definition option) 101
electrical 631 Joule
electromagnetic 317, 632 analysis 764
electrostatic 631 heating 348
fluid drag and wave loads 621 structural 352
heat fluxes 629 JOULE (model definition option) 349, 353, 354
initial stress and initial plastic strain 628 JOULE (parameter) 348, 353
magnetostatic currents 632
mass fluxes and restrictors 630 K
mechanical loads 619
Kachanov factor 528
thermal loads 627
Kelvin-Voigt model 489
INSERT (model definition option) 654
kinematic
INTCRD (user subroutine) 757
constraints 632
interface
hardening 467
ADAMS 770
hyperMesh results 769
SDRC I-DEAS results 769 L
interference fit 650 Lagrange multipliers 580
interpolation (shape) functions, type of 699 Lagrangian formulation 110
inverse power sweep 169 Lanczos method 169
ISOTROPIC (acoustic, model definition option) 325, large deformations 125
539 LARGE DISP (parameter) 111, 112, 116, 121, 122,
ISOTROPIC (electromagnetic, model definition 127, 398, 463, 648
option) 316, 538 large strain
ISOTROPIC (electrostatic, model definition option) analysis 707
300, 538 elasticity 124
ISOTROPIC (fluid, model definition option) 343, plasticity 127
538, 539 LARGE STRAIN (parameter) 111, 112, 115, 138,
ISOTROPIC (heat transfer, model definition option) 178, 287, 338, 398, 463, 515, 637, 648
348, 355, 537, 538, 628 linear analysis 100
ISOTROPIC (hydrodynamic, model definition linear elastic material 360
option) 538 linear fracture mechanics 148
ISOTROPIC (magnetostatic, model definition linear springs and elastic foundations input options
option) 305, 306, 538 655
ISOTROPIC (model definition option) 43, 210, 230, linked/unlinked 210
234, 252, 286, 287, 293, 296, 321, 334, 354, 361, list of items, input 43
363, 389, 396, 442, 443, 447, 448, 462, 463, 465, load
467, 468, 469, 482, 485, 487, 488, 492, 494, 505, AUTO STEP scheme 717
510, 619 automatic incrementation 716
ISOTROPIC (with TABLE input - Stress) 463 cavity pressure 622
isotropic hardening 466 fixed 715
fluid drag and wave 621
history definition options 713
increment size 715 Hashin Tape 380
mass fuxes 630 Hill 374
mechanical 618 Hoffman 375
mechanical and thermal increments 713 maximum strain 373
piezoelectric 631 maximum stress 372
pump pressures 630 Puck 381
recycling criterion 717 strain failure 373
residual 763 Tsai-Wu 377
residual load correction 724 user-defined 378
restrictors 630 material model
surface/volumetric 606 composite 362
thermal 627 elastomers 429
types 605 gasket 393
user-defined physical criteria 719 linear elastic 360
vectors 799 Mohr-Coulomb (hydrostatic yield dependence)
LOAD COR (parameter) 619 447
load incrementation nonlinear hypoelastic 398
adaptive scheme 714 time-independent inelastic 438
fixed scheme 713 Mean Strain Energy Criterion 82
Lorentz 346 mean-normal method 137
LORENZI (model definition option) 151, 153, 165 mechanical
losses 314, 354 erosion 248
erosion by other actions 249
loads 618
M loads input options 619
machining (or metal cutting) 802 mechanical shape memory model
MAGNETO (parameter) 305, 348, 354 conversation thermo-mechanical to mechanical
magnetostatic SMA 427
analysis 538, 632, 704, 762, 764 experimental data fitting 426
current input options 632 mechanical wear 199
elements types 304 membrane elements 700
MANY TYPES (model definition option) 64 MERGE (model definition option) 60
Marc mesh
host systems 30 definition 54
overview 24 refinement tying constraints 639
mass fluxes and restrictors 630 severe distortion 128
MASSES (model definition option) 172 mesh generation, remeshing techniques 92
material advancing front meshing 92
instabilities 128 Delaunay Triangulation 93
low tension 510 overlay meshing 92
mechanical shape memory 423 mesh splitting 165
powder 449 MESH2D 59
preferred direction 367 MESH2D (model definition option) 64
material dependent failure criteria 371 mixture model 389
Hashin 378 MNF (parameter) 772
Hashin Fabric 379 MNF UNITS (model definition option) 772
Index 817
modal NLELAST 398
stresses and reactions 171 (bimodulus) elasticity model 411
superposition 171 (bimodulus) elasticity model with either no
MODAL INCREMENT (model definition option) 105 tension, limited tension, no compression, or
MODAL SHAPE (history definition option) 169, 170, limited compression enhancements 411
325, 343 basic assumptions and definitions 399
mode separation 153 converting engineering strain/stress to true
model connections strain/stress 400
CWELD patch-to-patch 673 input of uniaxial stress-strain data 399
model definition options 50 Nastran like nonlinear elasticity model 401
Mohr-Coulomb nonlinear orthotropic elasticity model 413
linear material 447 principal strain space model 409
parabolic material 448 strain invariant model 405
MOONEY (model definition option) 115, 389, 432, NLELAST (model definition option) 112, 363, 398,
436, 505 401, 402
MOONEY (parameter) 115 nodal degrees of freedom tying constraints 646
Mooney-Rivlin 432, 437 NODE CIRCLE (model definition option) 70
MOTION (2-D, user subroutine) 548 NODE FILL (model definition option) 70
MOTION CHANGE (history definition option) 550, NODE GENER (model definition option) 70
569 NODE MERGE (model definition option) 70
MPC-CHECK (parameter) 654 NODE SORT (model definition option) 756
MPI (Message Passing Interface) 776 Node Within A Box Criterion 84
Mullin’s effect 526 Nodes In Contact Criterion 84
multiplicative decomposition 132, 438 nodes, merging of 60
multistage return mapping 138 nonlinear
MUMPS 37, 741, 776 analysis 105
fracture mechanics 150
nonlinear hypoelastic material 398
N HYPELA2 user subroutine 398
Narayanaswamy model 146, 500 NLELAST model definition 398
NASSOC (user subroutine) 147, 487 nonlinearities, three types 109
Nastran RBE2 and RBE3 661 non-Newtonian 327
Navier-Stokes 325, 328, 334, 336 nonstructural materials 537
NC machining data 802 NURBS 543, 544, 555, 584, 585
cutter path 803 trimmed 543, 579
cutter shape 802
element deactivation 804
intersection 804 O
nearly incompressible behavior 127 Oak Ridge National Laboratory (ORNL) 141, 462,
neighbor relations 553 485
Neo-Hookean 432 objective function 202, 203, 208, 209, 211, 212
Newmark-beta operator 177, 178 objective stress rates 130
Newton Cooling 509 OGDEN (model definition option) 433, 499, 505,
Newton-Raphson method 529
modified 727 OGDEN (parameter) 115
NiTi alloys 413
operator CAVITY DEFINITION (model definition) 276,
central difference 178 286, 288, 630
generalized alpha operator 177 CFAST (model definition) 655, 672, 689, 691,
Houbolt 177 692, 693, 694, 695
Newmark-beta 177 CHANGE PORE (model definition) 199
single step Houbolt 177 CHANGE STATE (history definition) 145, 715
OPTHOTROPIC (model definition option) 468 CHANGE STATE (model definition) 34, 231,
optimization algorithm 208 291, 627
OPTIMIZE (model definition option) 70 CHANNEL (model definition) 231, 336
Option COEFFICIENT (model definition) 658
ACTIVATE (history definition) 59 COHESIVE (model definition) 706
ADAPT GLOBAL (model definition) 260 COIL CURRENT (magnetostatic, model
ADAPTIVE (model definition) 82, 101, 102, 710 definition) 311, 313, 354
ANISOTROPIC (model definition) 101, 293, COMPOSITE (model definition) 210, 362, 363
321, 361, 363, 367, 389, 443, 467, 468, 469, CONN FILL (model definition) 70
505 CONN GENER (model definition) 70
APPROACH (history definition) 550 CONNECT (model definition) 64
ASSEM LOAD (model definition) 771 CONNECTIVITY (model definition) 59, 69, 337,
ATTACH EDGE (model definition) 81, 618 353, 354, 355, 655, 673, 679, 685, 686, 689,
ATTACH FACE (model definition) 618 725
AUTO CREEP (history definition) 141, 143, CONRAD GAP (model definition) 231
146, 147, 490, 713, 714 CONSTRAINT (model definition) 63
AUTO INCREMENT (history definition) 72, CONTACT (model definition) 119, 120, 128,
122, 512, 621, 713, 714, 716, 733 178, 231, 275, 346, 348, 542, 544, 575, 576,
AUTO LOAD (history definition) 72, 346, 348, 581, 605, 653, 659, 718
499, 621, 713 CONTACT CHANGE (rezoning) 577
AUTO STEP (history definition) 146, 177, 232, CONTACT NODE (model definition) 543
236, 346, 348, 355, 356, 494, 499, 512, 584, CONTACT TABLE (history definition) 550, 551,
590, 627, 713, 714, 715, 717, 718, 719, 720, 569
722, 723, 724 CONTACT TABLE (model definition) 556, 565,
AUTO THERM (history definition) 627, 713, 566, 567, 571, 576, 589
714, 715 CONTINUE (history definition) 44, 50, 51
AUTO THERM CREEP (history definition) 145, CONTROL (electromagnetic, model definition)
713, 714 316
AXITO3D (model definition) 212 CONTROL (heat transfer, model definition) 337
BEGIN SEQUENCE (history definition) 627 CONTROL (history definition) 145, 233, 662,
B-H RELATION (electromagnetic, model 667, 713
definition) 320, 538 CONTROL (magnetostatic, model definition)
B-H RELATION (magnetostatic, model 305
definition) 305, 306, 538 CONTROL (model definition) 112, 123, 232,
BUCKLE (history definition) 122, 123 283, 580, 662, 667, 713, 718, 755
BUCKLE INCREMENT (model definition) 105, CONVERT (model definition) 338, 354, 355,
122, 123 356
CASE COMBIN (model definition) 101, 103, COORD SYSTEM (model definition) 371, 549,
725 656
Index 819
COORDINATES (model definition) 59, 69, 685, DIST CURRENT (magnetostatic, model
725, 760 definition) 305, 354
CRACK DATA (model definition) 510 DIST CURRENT (model definition) 353
CREEP (model definition) 146, 480, 481, 482, DIST FLUXES (history definition) 349, 353, 356
487, 490, 506 DIST FLUXES (model definition) 195, 230, 236,
CREEP INCREMENT (history definition) 713 297, 349, 353, 354
CURE SHRINKAGE (model definition) 221 DIST LOADS (history definition) 145, 353
CURVES (model definition) 81, 371 DIST LOADS (model definition) 196, 321, 346,
CWELD (model definition) 672, 673, 678, 679, 348, 353, 619, 621
680, 681, 682, 684, 685, 686, 688, 689, 691, DIST MASS (model definition) 293
692, 693, 694, 695 DIST SOURCE (model definition) 325
CYCLIC SYMMETRY (model definition) 659, DYNAMIC CHANGE (history definition,
660 electromagnetic) 316, 319
CYLINDRICAL (model definition) 58, 668 DYNAMIC CHANGE (history definition) 175,
DAMAGE (model definition) 391, 529 177, 183, 325, 343, 713
DAMPING (model definition) 172, 175, 186 ELEM SORT (model definition) 756
DAMPING COMPONENTS (history definition) EMCAPAC (history definition) 302, 573
297 EMINDUC (history definition) 313
DEACT GLUE (model definition) 158, 162 EMISSIVITY (model definition) 286
DEFINE (model definition) 43, 44, 45 EMLAMIN (history definition) 316, 354
DELAMINATION (model definition) 166, 167 EMRESIS (history definition) 350, 573
DENSITY EFFECTS (model definition) 450 EMWINDING (magnetostatic, model definition)
DESIGN DISPLACEMENT CONSTRAINTS 311, 313, 354
(model definition) 210 END OPTION (model definition) 42, 50, 51,
DESIGN FREQUENCY CONSTRAINTS (model 209, 325
definition) 210 END REZONE (rezoning) 51
DESIGN OBJECTIVE (model definition) 210 END SEQUENCE (history definition) 627
DESIGN STRAIN CONSTRAINTS (model ERROR ESTIMATE (model definition) 79, 102,
definition) 210 128
DESIGN STRESS CONSTRAINTS (model EXCLUDE (history definition) 552
definition) 210 FAIL DATA (model definition) 371, 386, 387,
DESIGN VARIABLES (model definition) 210 389, 511
DISP CHANGE (history definition) 145, 338, FILMS (model definition) 230, 236, 247, 275,
353, 651, 653 349, 353, 354, 356, 629
DIST CHARGE (electromangetic, model FILMS (with TABLE input, model definition)
definition) 316 630
DIST CHARGE (model definition) 321, 346 FIXED DISP (model definition) 59, 63, 196,
DIST CHARGES (electrostatic, model 321, 338, 343, 346, 348, 353, 651
definition) 300, 346, 356 FIXED POTENTIAL (electromagnetic, model
DIST CURRENT (electromagnetic, model definition) 316
definition) 316 FIXED POTENTIAL (electrostatic, model
DIST CURRENT (history definition) 353 definition) 300, 356
DIST CURRENT (Joule, history definition) 349, FIXED POTENTIAL (magnetostatic, model
356 definition) 305, 354
FIXED PRESSURE (model definition) 195, ISOTROPIC (model definition) 43, 210, 230,
293, 296, 325 234, 252, 286, 287, 293, 296, 321, 334, 354,
FIXED TEMPERATURE (model definition) 230, 361, 363, 389, 396, 442, 443, 447, 448, 462,
235, 286, 334, 338, 349, 353, 354, 356 463, 465, 467, 468, 469, 482, 485, 487, 488,
FIXED VELOCITY (model definition) 334 492, 494, 505, 510, 619
FIXED VOLTAGE (model definition) 349, 353 ISOTROPIC (with TABLE input - Stress) 463
FLUID DRAG (model definition) 621 ISOTROPIC (with TABLE input - thermal) 354
FLUID SOLID (model definition) 343 J-INTEGRAL (model definition) 101
FOAM (model definition) 499 JOULE (model definition) 349, 353, 354
FOUNDATION (model definition) 101, 121 LORENZI (model definition) 151, 153, 165
FOURIER (model definition) 101, 102, 105 MANY TYPES (model definition) 64
FXORD (model definition) 65, 69, 702 MASSES (model definition) 172
GASKET (model definition) 396 MERGE (model definition) 60
GEOMETRY (model definition) 74, 210, 296, MESH2D (model definition) 64
666, 669, 682, 683, 684, 685, 702, 707, 709, MNF UNITS (model definition) 772
760 MODAL INCREMENT (model definition) 105
GLK (model definition) 217 MODAL SHAPE (history definition) 169, 170,
GLOBALLOCAL (model definition) 34, 217 325, 343
HARMONIC (electromagnetic, history MOONEY (model definition option) 115
definition) 320, 356 MOONEY (model definition) 389, 432, 436, 505
HYPOELASTIC (model definition) 398 MOTION CHANGE (history definition) 550, 569
INITIAL DISP (model definition) 183 NLELAST (model definition) 112, 363, 398,
INITIAL PC (model definition) 199 401, 402
INITIAL PORE (model definition) 199 NODE CIRCLE (model definition) 70
INITIAL POROSITY (model definition) 195, NODE FILL (model definition) 70
196, 293 NODE GENER (model definition) 70
INITIAL TEMP (model definition) 230, 235, 349, NODE MERGE (model definition) 70
353, 354, 356 NODE SORT (model definition) 756
INITIAL VEL (model definition) 183 OGDEN (model definition) 433, 499, 505, 529
INITIAL VOID RATIO (model definition) 195, OPTIMIZE (model definition) 70
196 ORIENTATION (model definition) 362, 363,
INSERT (model definition) 654 367, 443, 494, 698
ISOTROPIC (acoustic, model definition) 325, ORTHO TEMP (model definition) 230, 348,
539 363, 462, 505
ISOTROPIC (electromagnetic, model ORTHOTROPIC (electrical, model definition)
definition) 316, 538 538
ISOTROPIC (electrostatic, model definition) ORTHOTROPIC (electromagnetic, model
300, 538 definition) 316, 538
ISOTROPIC (fluid, model definition) 343, 538, ORTHOTROPIC (electrostatic, model
539 definition) 300, 346, 348
ISOTROPIC (heat transfer, model definition) ORTHOTROPIC (magnetostatic, model
348, 355, 537, 538, 628 definition) 305, 306, 538
ISOTROPIC (hydrodynamic, model definition) ORTHOTROPIC (model definition) 101, 210,
538 230, 252, 286, 287, 293, 321, 361, 363, 389,
ISOTROPIC (magnetostatic, model definition) 442, 443, 462, 465, 467, 468, 469, 487, 488,
305, 306, 538 492, 494, 505, 507, 619
Index 821
ORTHOTROPIC (thermal, model definition) POTENTIAL CHANGE (history definition) 356
348, 355, 537, 538, 628 POWDER (model definition) 450
ORTHOTROPIC (with TABLE input - Stress) PRE STATE (model definition) 34, 212, 213
463 PRINT CHOICE (model definition) 173, 384,
ORTHOTROPIC (with TABLE input - thermal) 755, 756, 758, 760, 763
354 PRINT ELEMENT (model definition) 384, 755,
PARAMETERS (history definition) 179 758, 760
PARAMETERS (model definition) 191, 192, PRINT NODE (model definition) 756, 758, 763
287, 334, 556, 570, 596 PRINT VMASS (model definition) 757
PBUSH (model definition) 655, 657, 658 PROPORTIONAL INCREMENT (history
PERMANENT (magnetostatic, model definition) 72, 145, 147, 621
definition) 305, 306 PSHELL (model definition) 367, 392
PFAST (model definition) 655, 657, 672, 689 PWELD (model definition) 672, 674, 677, 682,
PIEZOELECTRIC (model definition) 321 683, 684, 685
PIN CODE (model definition) 669 RAD-CAVITY (model definition) 247, 286, 287,
POINT CHARGE (electrostatic, model 630
definition) 317, 346 RADIATING CAVITY (model definition) 276
POINT CHARGE (model definition) 300, 321 RBE2 (model definition) 158, 162, 662, 663,
POINT CHARGE (with TABLE Input, model 670, 679
definition option) 356 RBE3 (model definition) 663, 664, 670, 671,
POINT CURRENT (electromagnetic, history 674, 679, 689, 695
definition) 316 REAUTO (model definition) 141, 725
POINT CURRENT (history definition) 353 REBAR (model definition) 510
POINT CURRENT (Joule, history definition) RECEDING SURFACE (model definition) 259
349 RECOVER (history definition) 171
POINT CURRENT (Joule, model definition) 349 RELATIVE DENSITY (model definition) 450
POINT CURRENT (magnetostatic, model RELEASE (history definition) 568
definition) 305, 354 RESTART (model definition) 33, 101, 141, 725,
POINT CURRENT (model definition) 353 756, 757, 758
POINT CURRENT (with TABLE Input - RESTART INCREMENT (history definition) 725
Magnetostatic, history definition) 356 RESTART LAST (model definition) 725
POINT FLUX (history definition) 349, 353, 354 RESTRICTOR (model definition) 296
POINT FLUX (model definition) 195, 230, 236, REZONE (rezoning) 51
297, 349, 353, 356 ROTATION A (model definition) 619
POINT LOAD (electrostatic, model definition) RROD (model definition) 158, 162
346, 348 SERVO LINK (model definition) 605, 639
POINT LOAD (history definition) 145, 353 SHELL TRANSFORMATION (model definition)
POINT LOAD (model definition) 43, 196, 321, 634, 637
353, 548 SHIFT FUNCTION (model definition) 146, 497
POINT MASS (model definition) 293 SOIL (model definition) 195, 196, 199, 513
POINT SOURCE (model definition) 325 SOLVER (model definition) 709, 740, 741, 743
POINTS (model definition) 81 SPECIFIED NODES (model definition) 64
POROSITY CHANGE (model definition) 392 SPLINE (model definition) 261, 543, 588, 595,
POST (model definition) 145, 231, 291, 386, 599, 600
715, 765 SPRINGS (model definition) 101, 121, 172
START NUMBER (model definition) 64 TRANSIENT (history definition) 231, 232, 233,
STEADY STATE (electrostatic, history 236, 349, 355, 356, 713, 714
definition) 301 TYING (model definition) 231, 343, 605, 638,
STEADY STATE (history definition) 232, 331, 639
349, 355, 356 TYING CHANGE (history definition) 639
STEADY STATE (magnetostatic, history UFCONN (model definition) 59
definition) 305 UFXORD (model definition) 59
STIFFNS COMPONENTS (history definition) UTRANFORM (model definition) 634
297 VCCT (model definition) 147, 154, 160
STIFFSCALE (model definition) 74 VELOCITY (model definition) 231, 292, 296
STRAIN RATE (model definition, fluid) 539 VELOCITY CHANGE (history definition) 231,
STRAIN RATE (model definition) 192, 334, 292, 334
363, 463, 464 VIEW FACTOR (model definition) 280, 630
SUMMARY (model definition) 756 VISCEL EXP (model definition) 146, 503
SUPERELEM (history definition) 772 VISCELFOAM (model definition) 146, 499
SURFACE ENERGY (model definition) 240, VISCELMOON (model definition) 146, 432, 499
259 VISCELOGDEN (model definition) 146, 433,
SURFACES (model definition) 81 499, 529
SWLDPRM (model definition) 672, 685, 688, VISCELORTH (model definition) 494
691 VISCELPROP (model definition) 494
SYNCHRONIZED (history definition) 550 VOID CHANGE (model definition) 392
TABLE (model definition) 286, 287, 315, 395, VOLTAGE CHANGE (history definition) 349,
467, 468, 492, 505, 507 353
TEMP CHANGE (history definition) 230, 235, WELD FILL (model definition) 268, 272, 273
334, 338, 349, 353, 354, 356 WELD FLUX (model definition) 268, 273, 274
TEMPERATURE EFFECTS (coupled thermal- WELD PATH (model definition) 268, 270
stress, model definition) 538 WORK HARD (model definition) 192, 363, 464,
TEMPERATURE EFFECTS (heat transfer, 467, 468
model definition) 348, 537 ORIENT (user subroutine) 101, 362, 443, 494
TEMPERATURE EFFECTS (hydrodynamic, ORIENT2 (user subroutine) 367, 368, 371
model definition) 538 ORIENTATION (model definition option) 362, 363,
TEMPERATURE EFFECTS (model definition) 367, 443, 494, 698
230, 234, 287, 296, 363, 432, 450, 462, 464, ORTHO TEMP (model definition option) 230, 348,
492, 497, 505, 507, 539 363, 462, 505
THERMAL CONTACT (2-D) (model definition) ORTHOTROPIC (electrical, model definition
313, 315, 573 option) 538
THERMAL CONTACT (model definition) 231, ORTHOTROPIC (electromagnetic, model definition
275, 302, 310, 350 option) 316, 538
THERMO-PORE (model definition) 252 ORTHOTROPIC (electrostatic, model definition
THICKNESS (model definition) 296 option) 300, 346, 348
THICKNS CHANGE (history definition) 297 ORTHOTROPIC (magnetostatic, model definition
TIME STEP (history definition) 463, 494, 499 option) 305, 306, 538
TIME-TEMP (model definition) 508, 509 ORTHOTROPIC (model definition option) 101, 210,
TRANSFORMATION (model definition) 549, 230, 252, 286, 287, 293, 321, 361, 363, 389, 442,
634, 642, 653, 668, 763 443, 462, 465, 467, 469, 487, 488, 492, 494, 505,
507, 619
Index 823
ORTHOTROPIC (thermal, model definition option) EXTENDED 43
348, 355, 537, 538, 628 FEATURE 662, 667
ORTHOTROPIC (with TABLE input - Stress) 463 FLUID 704
ORTHOTROPIC (with TABLE input - thermal) 354 FOLLOW FOR 112, 121, 619, 620, 637, 709
out-of-core storage options 32 FOURIER 102, 105
output results GENT 115
workspace information 750, 788 HARMONIC 172, 173
overclosure 650 HEAT 291, 292, 354, 355, 703
tying 650 IBOOC 31, 32, 34
JOULE 348, 353
LARGE DISP 111, 112, 116, 121, 122, 127, 398,
P 463, 648
parallel processing LARGE STRAIN 111, 112, 115, 138, 178, 287,
contact analysis 779 338, 398, 463, 515, 637, 648
domain decomposition 779 LOAD COR 619
solvers 778 MAGNETO 305, 348, 354
types of machines 776 MNF 772
unsupported features 777 MOONEY 115
Parameter MPC-CHECK 654
ABLATION 259, 260 OGDEN 115
ACOUSTIC 325 PIEZO 321
ADAPTIVE 710 PORE 293
ALL POINTS 291, 698 PRINT 123, 240, 286, 562, 764
ALLOCATE 32 RADIATION 276, 280, 537, 630
ASSUMED STRAIN 709 RBE 679
AUTOMSET 654 RESTRICTOR 296
AUTOSPC 568 REZONING 71
BEAM SECT 74, 684, 685, 701 R-P FLOW 191, 729
BEARING 296 SCALE 713
BUCKLE 105, 122 SHELL SECT 702
CENTROID 112, 115, 121, 291, 627, 698, 725, SIZING 698, 753, 754
736, 758, 763 STATE VARS 296
COUPLE 337, 353 STOP 32
CREEP 144, 146, 147, 479, 486, 487, 489 STRUCTURAL 346, 348
DESIGN OPTIMIZATION 209, 361 TSHEAR 384, 760
DESIGN SENSITIVITY 209, 361 T-T-T 508
DIFFUSION 293 UFLUXMEC 240
DIST LOADS 209 UPDATE 178
DYNAMIC 105, 165, 168, 175, 177 VISCO ELAS 146
ELASTIC 79, 100, 101 parameters 50
ELASTICITY 101 PARAMETERS (history definition option) 179
ELECTRO 300, 345 PARAMETERS (model definition option) 191, 192,
ELEMENTS 168, 673, 686, 689, 698 287, 334, 556, 570, 596
EL-MA 316, 355 Pardiso 741, 776
ELSTO 31, 32, 33, 34, 750 particle tracking 34
END 50
PBUSH (model definition option) 655, 657, 658 POINT MASS (model definition option) 293
penalty 191, 331 POINT SOURCE (model definition option) 325
methods 581 POINTS (model definition option) 81
PERMANENT (magnetostatic, model definition Poisson’s 100, 360, 361, 362, 363, 450, 505, 508,
option) 305, 306 510, 513, 515
perturbation analysis 123 PORE (parameter) 293
PFAST (model definition option) 655, 657, 672, 689 POROSITY CHANGE (model definition option) 392
piecewise linear representation 505 POST (model definition option) 145, 231, 291, 386,
PIEZO (parameter) 321 715, 765
piezoelectric post buckling analysis 714
analysis 320, 537, 631, 704, 762, 764 post file 765
coupling 323 POTENTIAL CHANGE (history definition) 356
element types 320 powder 106
loads 631 POWDER (model definition option) 450
PIEZOELECTRIC (model definition option) 321 powder material 449
pin code 669 Power Law Fluid 328
PIN CODE (model definition option) 669 PRE STATE (model definition option) 34, 212, 213
pipe bend element tying constraints 645 preconditioners, iterative solvers 743
plate elements 702 presentation of the energy equation 257
PLOTV (user subroutine) 765 Previously Refined Mesh Criterion 85
POINT CHARGE (electrostatic, model definition Principal Engineering Strains 765
option) 317, 346 PRINT (parameter) 123, 240, 286, 562, 764
POINT CHARGE (model definition option) 300, 321 PRINT CHOICE (model definition option) 173, 384,
POINT CHARGE (with TABLE Input - Electrostatic, 755, 756, 758, 760, 763
model definition option) 356 PRINT ELEMENT (model definition option) 384,
POINT CURRENT (electromagnetic, history 755, 758, 760
definition option) 316 PRINT NODE (model definition option) 756, 758,
POINT CURRENT (history definition option) 353 763
POINT CURRENT (Joule, history definition option) PRINT VMASS (model definition option) 757
349 program
POINT CURRENT (Joule, model definition option) initiation 36
349 messages 768
POINT CURRENT (magnetostatic, model definition progressive composite failure 386
option) 305, 354 PROPORTIONAL INCREMENT (history definition
POINT CURRENT (model definition option) 353 option) 72, 145, 147, 621
POINT CURRENT (with TABLE Input - PSHELL (model definition option) 367, 392
Magnetostatic, history definition option) 356 PSPG 328
POINT FLUX (history definition option) 349, 353, Puck failure criterion 381
354 PWELD (model definition option) 672, 674, 677,
POINT FLUX (model definition option) 195, 230, 682, 683, 684, 685
236, 297, 349, 353, 356 PWL (piece-wise linear) 543
POINT LOAD (electrostatic, model definition option) pyrolysis 249
346, 348 alternative formulation 253
POINT LOAD (history definition option) 145, 353 presentation of the mass equation 250
POINT LOAD (model definition option) 43, 196, principle of 251
321, 353, 548
Index 825
Q resistor 350
RESTART (model definition option) 33, 101, 141,
quadratic contact 574 725, 756, 757, 758
quarter point elements 164 restart analysis 725
RESTART INCREMENT (history definition option)
R 725
RESTART LAST (model definition option) 725
RAD-CAVITY (model definition option) 247, 286,
RESTRICTOR (model definition option) 296
287, 630
RESTRICTOR (parameter) 296
radial return 135
result evaluation, contact 578
RADIATING CAVITY (model definition option) 276
REZONE (rezoning option) 51
radiation 275
rezoning 71, 577
computational approaches 276
REZONING (parameter) 71
direct adaptive integration 282
rigid contact, solution strategy 585
Monte Carlo method 279
rigid link constraint 647, 653
pixel based modified Hemi-Cube method 277
rigid-plastic analysis 190, 707, 764
viewfactors 276
Riks-Ramm 733, 734
RADIATION (parameter) 276, 280, 537, 630
ROTATION A (model definition option) 619
radiation viewfactors 276
R-P FLOW (parameter) 191, 729
RBE (parameter) 679
RROD (model definition option) 158, 162
RBE2 (model definition option) 158, 162, 662, 663,
rubber foam model 436
670, 679
RVCNST 558, 559
RBE2 and RBE3 661
RBE3 (model definition option) 663, 664, 670, 671,
674, 679, 689, 695 S
reaction forces 763 SCALE (parameter) 713
REAUTO (model definition option) 141, 725 scaling, linear-elastic solution 713
REBAR (model definition option) 510 segment-to-segment 592
REBAR (user subroutine) 510 selective printout 755
rebar elements 706 grid force balance 757
RECEDING SURFACE (model definition option) options 755
259 semi-infinite elements 708
RECOVER (history definition option) 171 sensitivity 202
reduced integration elements 707 separation, contact 568
RELATIVE DENSITY (model definition option) 450 SEPFOR (user subroutine) 586
RELEASE (history definition option) 568 SEPSTR (user subroutine) 586
remeshing Sequential Quadratic Programming 208
automatic global 85 SERVO LINK (model definition option) 605, 639
criteria 90 shape memory
remeshing techniques 91 mechanical 423
adaptive 804 thermo-mechanical material 413
mesh generation 92 shell
outline extraction and repair 93 contact 552
target number of elements 95 elements 576, 702
residual load correction 724 stiffener tying constraints 645
resistance 348, 350, 573, 657, 765 transformation 634
SHELL SECT (parameter) 702 steady state rolling analysis 213
shell stiffener tying constraints 645 Stefan-Boltzmann 630
SHELL TRANSFORMATION (model definition stick-slip model, friction 562
option) 634, 637 STIFFNS COMPONENTS (history definition option)
shell-to-shell intersection tying constraints 641 297
shell-to-solid intersections tying constraints 642 STIFFSCALE (model definition option) 74
SHIFT FUNCTION (model definition option) 146, STOP (parameter) 32
497 storage methods 743
single input strain correction method 728
DDM 37, 777, 779, 780, 781 STRAIN RATE (model definition option) 192, 334,
single step Houbolt operator 177 363, 463, 464
singularity ratio 739 STRAIN RATE (model definition, fluid option) 539
SIZING (parameter) 698, 753, 754 strain rate effects on elastic-plastic constitutive
snap-through analysis 714 relation 472
SOIL (model definition option) 195, 196, 199, 513 stress
soil analysis 195, 512, 704 analysis 763
solver intensity factor defined 148
conjugate gradient iterative 778 intensity factors 149
direct 778 STRUCTURAL (parameter) 346, 348
hardware 778 structural zooming analysis 216
matrix 778 substructuring 71
multifrontal 778 SUMMARY (model definition option) 756
SOLVER (model definition option) 709, 740, 741, SUPERELEM (history definition option) 772
743 SUPG 329
special elements 705 surface energy 240
cable 705 SURFACE ENERGY (model definition option) 240,
curved-pipe 705 259
Gap-and-Friction 705 SURFACES (model definition option) 81
pipe-bend 705 SWLDPRM (model definition option) 672, 685, 688,
rebar 706 691
shear panel 705 symmetries 63
SPECIFIED NODES (model definition option) 64 SYNCHRONIZED (history definition option) 550
spectrum response 173 system and element stiffness matrices 798
SPLINE (model definition option) 261, 543, 588,
595, 599, 600
SPRINGS (model definition option) 101, 121, 172
T
springs, linear and nonlinear 761 TABLE (model definition option) 286, 287, 315, 395,
START NUMBER (model definition option) 64 467, 468, 492, 505, 507
STATE VARS (parameter) 296 TEMP CHANGE (history definition option) 230,
status file 772 235, 334, 338, 349, 353, 354, 356
STEADY STATE (electrostatic, history definition TEMPERATURE EFFECTS (coupled thermal-stress,
option) 301 model definition option) 538
STEADY STATE (history definition option) 232, TEMPERATURE EFFECTS (heat transfer, model
331, 349, 355, 356 definition option) 348, 537
STEADY STATE (magnetostatic, history definition TEMPERATURE EFFECTS (hydrodynamic, model
option) 305 definition option) 538
Index 827
TEMPERATURE EFFECTS (model definition Tsai-Wu failure criterion 377
option) 230, 234, 287, 296, 363, 432, 450, 462, TSHEAR (parameter) 384, 760
464, 492, 497, 505, 507, 539 T-T-T (parameter) 508
Temperature Gradient Criterion 84 tying
temperature-dependent overclosure 650
creep 506 shell-to-solid 647
material properties 505 TYING (model definition option) 231, 343, 605, 638,
tension softening 511 639
thermal TYING CHANGE (history definition option) 639
loads 627 tying constraints
loads input options 627 beam-to-beam intersection 641
THERMAL CONTACT (2-D) (model definition cyclic symmetry 658
option) 313, 315, 573 mesh refinement 639
THERMAL CONTACT (model definition option) nodal degrees of freedom 646
231, 275, 302, 310, 350 pipe bend element 645
thermal contact analysis 571 rigid link 647
thermal expansion coefficient 507 shell stiffener 645
thermochemical ablation and surface wnergy shell-to-shell intersection 641
balance 241 shell-to-solid intersections 642
thermo-electrical 538
thermo-mechanical shape memory model 413
constitutive theory 417
U
phase transformation strains 417 UADAP (user subroutine) 85
transformation induced deformation 415 UADAPBOX (user subroutine) 274
THERMO-PORE (model definition option) 252 UBEAR (user subroutine) 297
Thermo-Rheologically Simple material 495 UBREAKGLUE (user subroutine) 568
THICKNESS (model definition option) 296 UCOHESIVE (user subroutine) 533
THICKNS CHANGE (history definition option) 297 UCOHESIVET (user subroutine) 533
TIME STEP (history definition option) 463, 494, 499 UCOMPL (user subroutine) 172
time step definition 183 UCOORD (user subroutine) 81
time-dependent UCRACK (user subroutine) 510
boundary conditions 184 UCRACKGROW (user subroutine) 159, 160
inelastic behavior 476 UCRPLW (user subroutine) 141, 144, 487, 507
time-independent, inelastic behavior 438 UDELAM (user subroutine 167
TIME-TEMP (model definition option) 508, 509 UELASTOMER (user subroutine) 436
Time-Temperature-Transformation (T-T-T) 507 UELDAM (user subroutine) 529
tolerance values, contact 579 UENERG (user subroutine) 436
total Lagrangian procedure 112 UEPS (user subroutine) 300, 305, 316, 538
TRANSFORMATION (history definition option) 653 UFAH (user subroutine) 243
TRANSFORMATION (model definition option) 549, UFAIL (user subroutine) 371, 378
634, 642, 653, 668, 763 UFCONN (model definition option) 59
TRANSIENT (history definition option) 231, 232, UFCONN (user subroutine) 59
233, 236, 349, 355, 356, 713, 714 UFINITE (user subroutine) 442
TRSFAC (user subroutine) 497 UFLUXMEC (parameter 240
truss elements 700 UFLUXMEC (user subroutine) 249
UFMEC (user subroutine) 249
UFORMSN (user subroutine) 639, 646 HYPELA2 112, 398, 399, 513
UFOUR (user subroutine) 102 IMPD 757
UFRIC (user subroutine) 565, 566 INTCRD 757
UFTHP (user subroutine) 246 MOTION (2-D) 548
UFXORD (model definition optiion) 59 NASSOC 147, 487
UFXORD (user subroutine) 59, 65 ORIENT 101, 362, 443, 494
UGLAW (user subroutine) 246 ORIENT2 367, 368, 371
UGMEC (user subroutine) 249 PLOTV 765
UGROOV (user subroutine) 296 REBAR 510
UHTCOE (user subroutine) 571, 573 SEPFOR 586
UHTCON (user subroutine) 571, 573 SEPSTR 586
UHTNRC (user subroutine) 571, 573 TRSFAC 497
UMOONY (user subroutine) 432 UADAP 85
UMU (user subroutine) 316, 538 UADAPTBOX 274
UNEWTN (user subroutine) 192 UBEAR 297
uniaxial cracking data 510 UBREAKGLUE 568
UOGDEN (user subroutine) 433 UCOHESIVE 533
UPDATE (parameter) 178 UCOHESIVET 533
updated Lagrange Formulation, nonlinear elasticity UCOMPL 172
438 UCOORD 81
updated Lagrangian Procedure 115 UCRACK 510
UPNOD (user subroutine) 191 UCRACKGROW 159, 160
UPOSTV (user subroutine) 765 UCRPLW 141, 144, 487, 507
URESTR (user subroutine) 296 UDELAM 167
URPFLO (user subroutine) 192 UELASTOMER 436
User subroutine UELDAM 529
ANELAS 101, 361, 364 UENERG 436
ANEXP 101, 364, 507 UEPS 300, 305, 316, 538
ANKOND 231, 537 UFAH 243
ANPLAS 364, 443 UFAIL 371, 378
CRPLAW 141, 142, 143, 147, 480, 482, 485, UFCONN 59
487, 490, 507 UFINITE 442
CRPVIS 146, 479, 489, 490 UFLUXMEC 249
ELEVAR 757, 760 UFMEC 249
ELEVEC 757 UFORMSN 639, 646
FFORCEM 354 UFOUR 102
FILM 231, 236, 237, 238, 349, 353, 354, 356, UFRIC 565, 566
629, 630 UFTHP 246
FLUX 231, 236, 238, 297, 300, 305, 321, 325, UFXORD 59, 65
346, 349, 353, 354, 356, 629, 630 UGLAW 246
FORCDF 356 UGMEC 249
FORCDT 184, 235, 236, 238, 297, 300, 305, UGROOV 296
321, 334, 346, 348, 354 UHTCOE 571, 573
FORCEM 81, 184, 316, 321, 346, 348, 356 UHTCON 571, 573
HOOKLW 101, 361, 364 UHTNRC 571, 573
HOOKVI 494 UMOONY 432
Index 829
UMU 316, 538 VELOCITY CHANGE (history definition option)
UNEWTN 192 231, 292, 334
UOGDEN 433 VIEW FACTOR (model definition option) 280, 630
UPNOD 191 virtual crack closure technique (VCCT) 147
UPOSTV 765 VISCEL EXP (model definition option) 146, 503
URESTR 296 VISCELFOAM (model definition option) 146, 499
URPFLO 192 VISCELMOON (model definition option) 146, 432,
USIGMA 316, 538 499
USINC 183, 235 VISCELOGDEN (model definition option) 146, 433,
USPLIT_MESH 166 499, 529
USPRNG 121, 655 VISCELORTH (model definition option) 494
USSD 173 VISCELPROP (model definition option) 494
UTHICK 296, 297 VISCO ELAS (parameter) 146
UTIMESTEP 720 viscoelastic material
UTIMP 249 incompressible isotropic 494
UTRANS 634 isotropic 492
UVELOC 231, 292, 296 Thermo-Rheologically Simple 495
UVSCPL 142, 143, 147, 488 viscoelasticity 145
UWATERSL 256 viscoplasticity
UWELDFLUX 268, 270 explicit formulation 487, 488
VSWELL 142, 143, 486, 490 explicit method 146
WKSLP 467, 468 implicit method 147
YIEL 147, 487 VOID CHANGE (model definition option) 392
ZERO 147, 487 VOLTAGE CHANGE (history definition option) 349,
user subroutine input 59 353
user-defined Criterion 85 von Mises 135, 136, 143, 206, 441, 444, 450, 466,
user-defined failure criteria 378 467, 470, 471, 472, 482, 756, 758, 759
USIGMA (user subroutine) 316, 538 VSWELL (user subroutine) 142, 143, 486, 490
USINC (user subroutine) 183, 235
USPLIT_MESH (user subroutine) 166
USPRNG (user subroutine) 121, 655
W
USSD (user subroutine) 173 weighted mesh 63
UTHICK (user subroutine) 296, 297 WELD FILL (model definition option) 268, 272, 273
UTIMESTEP (user subroutine) 720 WELD FLUX (model definition option) 268, 273, 274
UTIMP (user subroutine) 249 WELD PATH (model definition option) 268, 270
UTRANFORM (model definition option) 634 welding 268
UTRANS (user subroutine) 634 weld flux 268
UVELOC (user subroutine) 231, 292, 296 weld path 270
UVSCPL (user subroutine) 142, 143, 147, 488 Williams-Landel-Ferry (WLF) 146, 496
UWATERSL (user subroutine) 256 windings 310
UWELDFLUX (user subroutine) 268, 270 WKSLP (user subroutine) 467, 468
WORK HARD (model definition option) 192, 363,
464, 467, 468
V
VCCT (model definition option) 147, 154, 160
VELOCITY (model definition option) 231, 292, 296
workhardening
isotropic 466
rules 466
slope 440
workspace allocation 31
Y
YIEL (user subroutine) 147, 487
Young’s 363
Young’s modulus 360, 363, 437, 440, 466, 505, 508,
510, 511, 513
Z
ZERO (user subroutine) 147, 487
Zienkiewicz-Zhu Creep Strain Criterion 83
Zienkiewicz-Zhu Criterion 82
zooming 216

Msc-Marc-manual Volome A

Загружено:

Сведения о документе

Исходное описание:

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Msc-Marc-manual Volome A

Загружено:

Авторское право:

Доступные форматы

Marc 2010

Volume A: Theory and User Information

How to Use This Manual 21

1 The Marc System

Features and Benefits of Marc 26

Examples of Running Marc Jobs 39

4 Introduction to Mesh Definition

BEAM SECT Parameter 74

Automatic Global Remeshing 85

5 Structural Procedure Library

Nonlinear Analysis 105

Fracture Mechanics 147

Dynamic Fracture Methodology 163

Mesh Splitting 165

Inertia Relief 187

Rigid-Plastic Flow 190

Soil Analysis 195

Mechanical Wear 199

Design Optimization 205

Steady State Rolling Analysis 213

Structural Zooming Analysis 216

Cure-Thermal-Mechanically Coupled Analysis 220

6 Nonstructural Procedure Library

Hydrodynamic Bearing 295

Electrostatic Analysis 300

Magnetostatic Analysis 304

Piezoelectric Analysis 320

Acoustic Analysis 324

Fluid Mechanics 325

Coupled Analyses 335

Composite Material 362

Material Dependent Failure Criteria 371

Nonlinear Hypoelastic Material 398

Thermo-Mechanical Shape Memory Model 413

Mechanical Shape Memory Model 423

Time-independent Inelastic Behavior 438

Time-dependent Inelastic Behavior 476

Temperature Effects and Coefficient of Thermal Expansion 503

Low Tension Material 510

Damage Models 524

Nonstructural Materials 537

Motion of Bodies 548

Detection of Contact 550

Implementation of Constraints 554

Friction Modeling 556

Coupled Analysis 569

Thermal Contact 571

Joule Heating 572

Contact in an Electrostatic or Piezoelectric Analysis 573

Contact in Magnetostatic Analysis 573

Contact in Electromagnetic Analysis 573

Contact in Soil Analysis 573

Contact in an Acoustic Analysis 573

Element Considerations 574

Dynamic Impact 577

Global Adaptive Meshing and Rezoning 577

Adaptive Meshing 578

Result Evaluation 578

Tolerance Values 579

Segment-to-Segment Contact 592

Piezoelectric Loads 631

Kinematic Constraints 632