MSC Nastran 2008 r1 Release Guide

MSC Nastran 2008 r1
Release Guide
Main Index
Corporate
MSC.Software Corporation
2 MacArthur Place
Santa Ana, CA 92707 USA
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, Shinjyku-Ku
Tokyo 160-0023, JAPAN
Telephone: (03)-6911-1200
Fax: (03)-6911-1201
Worldwide Web
www.mscsoftware.com
Disclaimer
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 text are for illustrative and educational purposes only,
and are not intended to be exhaustive or to apply to any particular engineering problem or design. MSC.Software
Corporation assumes no liability or responsibility to any person or company for direct or indirect damages resulting
from the use of any information contained herein.
User Documentation: Copyright 2008 MSC.Software Corporation. Printed in U.S.A. All Rights Reserved.
This notice shall be marked on any reproduction of this documentation, in whole or in part. Any reproduction or
distribution of this document, in whole or in part, without the prior written consent of MSC.Software Corporation is
prohibited.
This software may contain certain third-party software that is protected by copyright and licensed from
MSC.Software suppliers.
MSC, MD, Dytran, Marc, MSC Nastran, MD Nastran, Patran, MD Patran, the MSC.Software corporate logo, and
Simulating Reality are trademarks or registered trademarks of the MSC.Software Corporation in the United States
and/or other countries.
NASTRAN is a registered trademark of NASA. PAMCRASH is a trademark or registered trademark of ESI Group.
SAMCEF is a trademark or registered trademark of Samtech SA. LS-DYNA is a trademark or registered trademark
of Livermore Software Technology Corporation. ANSYS is a registered trademark of SAS IP, Inc., a wholly owned
subsidiary of ANSYS Inc. ABAQUS is a registered trademark of ABAQUS Inc. All other brand names, product
names or trademarks belong to their respective owners. PCGLSS 6.0, Copyright 1992-2005, Computational
Applications and System Integration Inc. All rights reserved. PCGLSS 6.0 is licensed from Computational
Applications and System Integration Inc.
Revision 0. June 4, 2008
NA:V2008R1:Z:Z:Z:DC-REL
Main Index
Cont ent s
MSC Nastran 2008 r1 Release Guide
A Word About Prerelease Features x
List of Books xi
Technical Support xii
Internet Resources xiv
1 Overview of MSC Nastran 2008 r1
Overview 2
Optimization 2
Aeroelasticity 2
Symbolic Subsitution 2
List of Errors Resolved 3
List of Example Problems for the MSC Nastran 2008 r1 Release 3
List of MSC Nastran Documents Released with MSC Nastran 2008 r1 3
2 Implicit Nonlinear
Implicit Nonlinear - SOL 600 6
Support of Large Grid and Element IDs 6
Multiple RFORCE Entries in the Same Subcase 6
BCONTACT Case Control Command Clarification 6
Generalized Alpha Dynamic Integration Method 12
MATVP Material Property Entry 12
MATSMA Shape Memory Alloy Material Property Entry 12
Nonlinear Elastic Orthotropic Materials 13
Composite Integration Methods to Reduce Computer Time 13
New SOL 600 Bulk Data Entries and Parameters 14
3 NVH and Acoustics
NVH Enhancements 18
ACMS with Acoustic External Superelement Creation 18
Multiple RANDOM Looping 18
Sparse OUTPUT4 Format for External Superelement Creation 18
jp`=k~~=OMMT=
o~=d
Table of Contents
Main Index
==
iv
Binary op2 and op4 Compatibility Robustness 18
Merged Superelement Results 19
4 Numerical Methods and High Performance Computing
Factor Matrix Caching for Lanczos 22
Introduction 22
Benefits 22
Method and Theory 22
Inputs 22
Outputs 22
Guidelines and Limitations 23
Demonstration Examples 23
New TAUCS Indefinite Solver Improves Lanczos Performance 24
Introduction 24
Benefits 24
Inputs 24
Outputs 24
Shared Memory Parallel (SMP) Scalability Improvements for Static
Analysis 26
Introduction 26
Benefits 26
Inputs 26
Outputs 26
Guidance and Limitations 26
New MAXRATIO Information Output 28
Introduction 28
Benefits 28
Inputs 28
Outputs 28
Demonstration Example 29
Example Input Data 29
Example Output 31
New SPARSESOLVER MDTSTATS Information Output 32
Main Index
v Contents
Introduction 32
Benefits 32
Inputs 32
Outputs 32
Demonstration Example 33
Example Input Data 33
Example Output 35
5 Upward Compatibility
TEMPERATURE Case Control Command 38
Improvements in Fluid Eigenvalue Analysis 40
FLUID GRID Points and Partitioning 41
Distributed Memory Parallel (DMP) Diagnostic Messages 43
System Information Message (SIM) 6916 44
6 Optimization
Enhancements in DRESP3 46
Introduction 46
Benefits 46
User Inputs 46
Output 51
Examples 53
Topometry Optimization 56
Introduction 56
Benefits 56
Input 57
Output 59
Example 1 - Three-bar Truss (tomex1.dat) 59
Input 61
Output 63
Example 2 Car Model Topometry Design 63
Topography (Bead or Stamp) Optimization 65
Introduction 65
Main Index
==
vi
Benefits 65
Input 65
Outputs 68
Example 3 A Square (togex1.dat) 69
Input 70
Output 70
Randomization of an Input Data File 71
Introduction 71
Benefits 71
Input 71
Output 72
Random Elimination of Element Types 73
Introduction 73
Benefits 73
Input 73
Output 73
Enhancements in SOL 200 Optimization 74
Introduction 74
Benefits 74
Input 74
Example 75
Output 76
Guidelines 77
Limitations 77
7 Aeroelasticity and Rotor Dynamic Improvements
A New Aerodynamic Interpolation Method 82
Introduction 82
Inputs 82
Outputs 82
Examples 83
External Spline Server 85
Introduction 85
Inputs 85
API Changes 85
Main Index
vii Contents
Sparse Matrix Format 86
Upgrading an Existing Spline Server 86
Blade Vibration Analysis 87
Main Index
==
viii
Main Index
The 2005 New Template
m~
A Word About Prerelease Features
List of Books
Technical Support
Internet Resources
Main Index
x
MSC Nastran 2008 r1 may contain a number of features that have been labeled as prerelease.
A prerelease feature or enhancement is defined as a feature or enhancement that has not yet completed
MSCs exhaustive verification and validation (V and V) testing and qualification process. Therefore,
prerelease features are to be used at the clients own risk.
Main Index
xi
Preface
List of Books
Below is a list of some of the MD Nastran and MSC Nastran documents. You may order any of these
documents from the MSC.Software BooksMart site at http://store.mscsoftware.com/.
Insta||at|on and Re|ease Gu|des
Installation and Operations Guide
Release Guide
Reference Books
Quick Reference Guide
DMAP Programmers Guide
Reference Manual
User's Gu|des
Getting Started
Linear Static Analysis
Basic Dynamic Analysis
Advanced Dynamic Analysis
Design Sensitivity and Optimization
Thermal Analysis
Numerical Methods
Aeroelastic Analysis
Superelement
User Modifiable
Toolkit
Implicit Nonlinear (SOL 600)
Explicit Nonlinear (SOL 700)
MD Users Guide - Application Examples
Topology Optimization
SCA Service Guide
User Defined Services
Main Index
Technical Support
xii
Technical Support
For help with installing or using an MSC.Software product, contact your local technical support services.
Our technical support provides the following services:
Resolution of installation problems
Advice on specific analysis capabilities
Advice on modeling techniques
Resolution of specific analysis problems (e.g., fatal messages)
Verification of code error.
If you have concerns about an analysis, we suggest that you contact us at an early stage.
You can reach technical support services on the web, by telephone, or e-mail.
Web Go to the MSC.Software website at www.mscsoftware.com, and click on Support. Here you can find a
wide variety of support resources including application examples, technical application notes, training
courses, and documentation updates at the MSC.Software Training, Technical Support, and
Documentation web page.
Phone
and
Fax
Ema|| Send a detailed description of the problem to the email address below that corresponds to the product you
are using. You should receive an acknowledgement that your message was received, followed by an
email from one of our Technical Support Engineers.
United States
Telephone: (800) 732-7284
Fax: (714) 784-4343
Frimley, Camberley
Surrey, United Kingdom
Telephone: (44) (1276) 60 19 00
Fax: (44) (1276) 69 11 11
Munich, Germany
Telephone: (49) (89) 43 19 87 0
Fax: (49) (89) 43 61 71 6
Tokyo, Japan
Telephone: (81) (03) 6911 1200
Fax: (81) (03) 6911 1201
Rome, Italy
Telephone: (390) (6) 5 91 64 50
Fax: (390) (6) 5 91 25 05
Paris, France
Telephone: (33) (1) 69 36 69 36
Fax: (33) (1) 69 36 45 17
Moscow, Russia
Telephone: (7) (095) 236 6177
Fax: (7) (095) 236 9762
Gouda, The Netherlands
Telephone: (31) (18) 2543700
Fax: (31) (18) 2543707
Madrid, Spain
Telephone: (34) (91) 5560919
Fax: (34) (91) 5567280
Main Index
xiii
Preface
q~
The MSC Institute of Technology is the world's largest global supplier of CAD/CAM/CAE/PDM
training products and services for the product design, analysis, and manufacturing markets. We offer
over 100 courses through a global network of education centers. The Institute is uniquely positioned to
optimize your investment in design and simulation software tools.
Our industry experienced expert staff is available to customize our course offerings to meet your unique
training requirements. For the most effective training, The Institute also offers many of our courses at
our customer's facilities.
The MSC Institute of Technology is located at:
2 MacArthur Place
Santa Ana, CA 92707
Phone: (800) 732-7211
Fax: (714) 784-4028
The Institute maintains state-of-the-art classroom facilities and individual computer graphics
laboratories at training centers throughout the world. All of our courses emphasize hands-on computer
laboratory work to facility skills development.
We specialize in customized training based on our evaluation of your design and simulation processes,
which yields courses that are geared to your business.
In addition to traditional instructor-led classes, we also offer video and DVD courses, interactive
multimedia training, web-based training, and a specialized instructor's program.
Course Information and Registration. For detailed course descriptions, schedule information,
and registration call the Training Specialist at (800) 732-7211 or visit www.mscsoftware.com.
m~~=p
jp`=k~~=p
a~=p
jp`=c~=p
j~=p
jp`=f=`=f~
~~K]~K
~~K]~K
~K]~K
~K]~K
~K]~K
~K]~K
Main Index
Internet Resources
xiv
Internet Resources
MSC.Software (www.mscsoftware.com)
MSC.Software corporate site with information on the latest events, products, and services for the
CAD/CAE/CAM marketplace.
Main Index
Chapter 1: Overview of MSC Nastran 2008 r1 MSC Nastran 2008 r1 Release Guide
1
Overview of MSC Nastran 2008 r1
Overview
Main Index
Overview
2
Overview
MSC Software is proud to release MSC Nastran 2008 r1. This release of MSC Nastran significantly
advances the multidiscipline capabilities available to you. The following sections briefly describe some
of the major and minor enhancements to MSC Nastran 2008 r1.
Optimization
MSC Nastran has had very powerful optimization routines since it was released in 2006. The
functionality in that release included shape, sizing, and basic topology optimization. MSC Nastran 2007
r1 introduced manufacturing and symmetry constraints for topology optimization. MSC Nastran 2008
r1 extends this functionality in the areas of topography and topometry optimization.
In topography optimization, the nodes on a surface mesh are moved normal to the surface during the
optimization loop to arrive at an optimal shape. In contrast, topometry optimization considers each
element in a design region to have a unique property and it will be modified to achieve an optimal design.
Additional optimization enhancements include:
Automatic randomization of input variables rapid stochastic analysis set-up,
Random element elimination for sensitivity studies of spot welds and connectors,
More information on these optimization enhancements can be found in Optimization (Ch. 6).
Aeroelasticity
MSC Nastran 2007 r1 introduced an external spline evaluation capability. This capability has been
enhanced in R3 to support storage of the spline matrix in sparse format. This change allows larger models
to fit into memory.
MSC Nastran 2008 r1 also introduces new capabilities for aeroelasticity analyses. There is a new
aerodynamic interpolation method that interpolates each term in the generalized aerodynamic matrix
individually. Examples for using this new interpolation method are given in Aeroelasticity and Rotor
Dynamic Improvements (Ch. 7).
Symbolic Subsitution
Using the new Symbolic Substitution feature, you can run multiple analyses on an input file, while
modifying fields automatically. Using Symbolic Substitution you specify a special symbol in the input
file that identifies the location where changes are to be made. When you run your job, you specify a
replacement symbol value that replaces the special symbol in your input file, but only for that job. You
can then make several runs, each with a different value, without having to make any additional
modifications to the input file.
For more information, please see Symbolic Substitution (App. A) in the MSC Nastran Installation and
Operations Guide.
Main Index
3
CHAPTER 1
Overview of MSC Nastran 2008 r1
List of Errors Resolved
The list of errors resolved in this release can be found at:
http://www.mscsoftware.com/support/prod_support/nastran/errorlist/files/error2008.lst
List of Example Problems for the MSC Nastran 2008 r1
Release
The table below is a list of the example problems in this release guide and the associated file name that
can be found in the test problem library, or in the documentation directory in your MSC Nastran 2008 r1
installation.
List of MSC Nastran Documents Released with MSC Nastran
2008 r1
Along with this Release Guide, the following documents are updated for the MSC Nastran 2008 r1
release:
MSC Nastran Installation and Operations Guide
MSC Nastran Quick Reference Guide
Example Problems File Name
Three-bar Truss page 59 tomex1.dat
A Square page 69 togex1.dat
Fluid Modes as Design Constraints page 79 d200fmd1.dat
Main Index
Overview
4
Main Index
Chapter 2: Implicit Nonlinear MSC Nastran 2008 r1 Nastran Release Guide

2
Implicit Nonlinear
Implicit Nonlinear - SOL 600

Main Index
6
The following is a discussion of the new additions and improvements made for MSC Nastran 2008 r1.
Support of Large Grid and Element IDs
The largest addition to SOL 600 for MSC Nastran 2008 r1 is the addition of a capability to support very
large grid and element IDs. Up to 10-digit IDs may now be used for grids and elements when SOL 600
is used, however to be compatible with other solution sequences, IDs should not normally exceed a value
of 99999999. Large IDs may be specified separately for grids, elements, or both items. Large ID
capability is not the default for MSC Nastran 2008 r1 and, if it is needed for a particular model, it must
be activated by placing one of the following items shown in bold on the SOL 600,ID, 155 in the MSC
Nastran Quick Reference Guide (most other items are omitted to prevent confusion):
SOL 600,SID MRENUMBR= MRENUELE= MRENUGRD=
Please see parameters, MRENUMBR, 786, MRENUELE, 784, and MRENUGRD, 785 in the MSC Nastran
Quick Reference Guide. These key words are only required if the number of digits is greater than seven.
Multiple RFORCE Entries in the Same Subcase
SOL 600 now supports multiple RFORCE entries in the same subcase so that different portions of the
structure can rotate with different angular velocities, or even in different directions. To accomplish this,
the two or more RFORCE entries should have the same SID (see below) and field 4 of the each
continuation entry should specify IDRF which points to a SET 3 entry designating which elements apply
to that particular RFORCE entry.
RFORCE (addition to the RFORCE entry for SOL 600)
Format:
BCONTACT Case Control Command Clarification
Normally, only one form of this entry may be used in any given analysis. The exception, for SOL 600
only, is that BCONTACT=NONE may now be used for any subcase desired and/or for increment zero
1 2 3 4 5 6 7 8 9 10
RFORCE SID G CID A R1 R2 R3 METHOD
RACC MB IDRF
IDRF (SOL 600
only)
ID indicating to which portion of the structure this particular RFORCE entry applies.
It is possible to have multiple RFORCE entries in the same subcase for SOL 600 to
represent different portions of the structure with different rotational velocities. IDRF
corresponds to a SET3 entry specifying the elements with this acceleration.
Main Index
7
CHAPTER 2
Implicit Nonlinear
and some other form such as BCONTACT=N used for the other subcases. This allows some subcases to
have contact and others to have no contact. Analysis restarts must use the same form as the original run
BCONTACT=ALLxxx cannot be mixed with
BCONTACT=NONE or BCONTACT=N in the same input file.
BCTABLE Bulk Data Entry Additions
Several new fields have been made in the BCTABLE entry to clarify which shell surfaces may contact
for SOLs 101 and 600. The new fields are shown in bold:
Format:
For detailed descriptions on the new fields and Remarks 22 and 23 see BCTABLE (SOL 600), 1114 in the
MSC Nastran Quick Reference Guide.
Other BCTABLE Clarifications
If the user leaves IDSLAVE and IDMAST blank, then NGROUP is normally required and continuation
entries are usually expected for NGROUP SLAVE/MASTER combinations. Exceptions are (a) for
SOL 700 where self-contact may be designated using a slave IDSLA1 of zero and no MASTER entry
and (b) for SOL 600 if no contact is desired in increment zero or a particular subcase, fields 1 and 2 of
the primary BCTABLE entry for that subcase is entered, all other fields left blank and no continuation
1 2 3 4 5 6 7 8 9 10
BCTABLE ID IDSLAVE IDMAST NGROUP COPTS COPTM
SLAVE IDSLA1 ERROR FNTOL FRIC CINTERF IGLUE
ISEARCH ICOORD JGLUE TOLID DQNEAR DISTID
FBSH FRLIM BIAS SLIDE HARDS COPTS1 COPTM1
BKGL BGST BGSN BGM BGN
HHHB HCT HCV HNC BNC EMISS HBL
FK EXP METHOD ADAPT THICK THICKOF PENV
FACT TSTART TEND MAXPAR PENCHK FSF VSF
EROSOP IADJ SOFT DEPTH BSORT FRCFRQ SNLOG
ISYM I2D3D IGNORE SPR MRP VDC SBOPT
SFS SFM SST MST SFST SFMT AUTO
LCID FCM US PSF FA ED INTTYPE
NFLS SFLS IGNOFF FSLIM PYS TDIC CDIST
NFLF SFLF NEN MES TBLCID TBLAB IGAP
FTBID VC SMOOTH FLANGL PENMAX THKOPT SHLTHK
SLDTHK SLDSTF
DBID TIDRF TIDNF DBDTH DFSCL NUMINT
MASTERS IDMA1 IDMA2 IDMA3 IDMA4 IDMA5 IDMA6 IDMA7
IDMA8 IDMA9 ...
Main Index
8
lines are entered. The SOL 600 no contact condition may be achieved in either of two ways - set Case
Control BCONTACT=ID and enter a matching BCTABLE with that ID in field 2 and all other fields
blank or set BCONTACT=NONE and do not enter BCTABLE for that subcase.
New Triangular Plane Stress Element
The MRALIAS parameter or the ALIASM option may be used to specify that a 3-node plane stress
element is to be used by specifying type 201.
New Solid Shell Element
A new solid shell element (CSSHL) has been added to SOL 600. The solid shell is normally used for
contact problems where contact occurs on both the top and bottom faces. This element may be used with
either homogeneous properties or by referencing a PCOMP or PCOMPG.
CSSHL (SOL 600)
Defines the connection for a Solid Shell with 6 or 8 grid points.
Format:
Examples:
1 2 3 4 5 6 7 8 9 10
CSSHL EID PID G1 G2 G3 G4 G5 G6
G7 G8
CSSHL 44 11 1 2 3 4 5 6 quad
7 8 quad
CSSHL 51 22 11 12 13 21 22 tria
23 tria
CSSHL 51 22 11 12 13 13 21 22 tria
23 23 tria
Note: The second and third examples are equivalent to each other.
Field Contents
EID Element identification number. (1 < Integer < 1E11, Required)
Main Index
9
CHAPTER 2
Implicit Nonlinear
See CSSHL (SOL 600) (p. 1342) in the MSC Nastran Quick Reference Guide for additional details. An
example is tpl model hextqb-sshl2.dat
PSSHL (SOL 600)
Defines the properties for Solid Shell (CSSHL) elements.
Format:
Example:
New Penta 15 Solid Element
Support for Penta 15 elements with mid-side nodes has been available in previous releases but they were
formed using hexa 20 elements with a collapsed side. This formulation is not as accurate as the new true
penta 15 formulation. The old collapsed side formulation is no longer used starting with this release and
the new formulation is used automatically. There are no changes to the input data required.
PID Property identification of a PSHELL, PCOMP, or PCOMPG entry. (Integer > 0,
Required). Note that the MID2 entry on the PSHELL or PCOMP is ignored.
Gi Grid point identification number of connection points. (Integer or blank, for quad
shapes all eight values are required, for triangle shapes only G4 and G8 may be left
blank in which case G4=G3 and G8=G7.)
1 2 3 4 5 6 7 8 9 10
PSSHL PID MID IT SF
PSSHL 11 33 .8333
Field Contents
PID Property identification number. (Integer > 0, Required)
MID Identification of a MAT1xxx entry. All MAT entries available in SOL 600 can be
specified except for hyperelaastic materials. (Integer > 0)
IT Transition thickness - Enter only if a solid shell is attached to a standard shell (such
as CQUAD4), in which case TT is the thickness of the standard shell. (Real, Default
= 0.0)
SF Transverse shear factor - Leave blank if transverse shear is not to be considered.
(Real or blank, if entered SF must range between 0.0 and 1.0)
Field Contents
Main Index
10
MARCOUT t16 Output Results Changes
All output quantities supported by Marc are available in the SOL 600 t16 file and may be specified using
the MARCOUT Bulk Data entry. Additions for this release are as follows:
Please note that for SOL 600, MSC Nastran Case Control commands such as SET ID=, DISP=,
STRESS=, STRAIN= only control the output in the .op2, .xdb, punch, .f06 and/or jid.marc.out file(s).
The Case Control requests do not affect the. t16 output.
Limiting t16 Output to Selected Elements or Grids
For large nonlinear models the output can become very large. Sometimes only a certain portion of the
structure is of concern. The following new entry may be used in SOL 600 to specify which elements or
nodes should be output. The default is all elements and nodes will be output if the entry is not made. This
entry may be used in combination with MARCOUT or the default MARCOUTs may be used.
MT16SEL Limits elements and/or grid results to selected elements or grids for t16
and t19 file results
Format
E-USER 1
st
user-defined element post code(s) are generated by user subroutine plotv.f
E-USER1 2
nd
user-defined element post code(s)
E-USER2 3
rd
E-USER3 4
th
E-USER99 100
th
These outputs are only available in the .t16 file, not in .op2, .xdb, .f06, punch. A
maximum number of 100 user-defined element post codes may be entered for SOL 600.
N-USER 1
st
user-defined nodal post code are generated by user subroutine upstnd.f
N-USER1 2
nd
N-USER2 3
rd
N-USER3 4
th
N-USER4 5
th
N-USER99 100
th
user-defined nodal post code are generated by user subroutine upstnd.f, etc.
User-defined outputs are only available in the .t16 file, not in .op2, .xdb, .f06, punch. A
maximum of 100 user-defined nodal post codes may be entered for SOL 600.
1 2 3 4 5 6 7 8 9 10
MT16SEL TYPE ID1 THRU ID2 BY ID3
Main Index
11
CHAPTER 2
Implicit Nonlinear
Example:
See MT16SEL (SOL 600), 1851 in the MSC Nastran Quick Reference Guide for more details.
Analytical Contact Threshold Angle
Starting with this release it is possible to define analytical contact threshold angles for different subcases.
To do so, include the following entry:
Defines automatic analytical contact threshold angle for multiple subcases - SOL 600 only.
Format:
Example:
Please see SANGLE (SOL 600), 2202 in the MSC Nastran Quick Reference Guide for more details. An
example using SANGLE is TPL model sangle1a.dat.
Additions to NLSTRAT
The following additions have been made to the NLSTRAT entry to support heat transfer analyses which
was introduced in the previous release.
MT16SEL GRID 1 THRU 100 BY 5
ELEM 100 THRU 500 BY 2
1 2 3 4 5 6 7 8 9 10
SANGLE IDC IDB Angle IDC IDB Angle
SANGLE 1 4 50.0 1 6
2 4 -1.0 2 6 55.0
Field Contents
IDC Identification number of a SUBCASE Case Control command. (Integer, no Default) To
enter a value corresponding to Marcs increment zero, set IDC=0.
IDB Identification of a contact body (must be the same as a BCBODY ID) (Integer,
no Default)
Angle Threshold automatic analytical contact angle (SANGLE). (Real, Default = 60.0)
A value of -1.0 turns off analytical
PLANKS Planks second constant (Real, Default=14387.69 microMK) PARAMETERS (4,6)
CLIGHT Speed of light in a vacuum (Real, Default=2.9979E14 micor M/s) PARAMETERS
(4,7)
RAPMAX Maximum change in the incremental displacement in a Newton-Raphson iteration
(Real, Default = 1.0E30) PARAMETERS (4,8)
Main Index
12
Generalized Alpha Dynamic Integration Method
For previous releases, several numerical integration methods were available for dynamic analysis. One
additional method, called the Generalized Alpha or Hilber-Hughes Taylor Method has been added. This
method is sometimes superior to the others for difficult dynamics problems, particularly those involving
contact. The single step Houbolt method is still the default for this release, but the new method may
become the default in subsequence releases. To select any of these methods, enter the following bulk
data parameter:
MHOUBOLT
Integer, Default = 0, MSC Nastran Implicit Nonlinear (SOL 600) only.
If MHOUBOLT=0, SOL 600 transient dynamics will use the single step Houbolt numerical
integration method.
MHOUBOLT=1, SOL 600 transient dynamics will use the Newmark Beta numerical integration method.
MHOUBOLT=2, SOL 600 transient dynamics will use the standard multi-step Houbolt numerical
integration method.
MHOUBOLT=7, SOL 600 transient dynamics will use the generalized alpha (Hilber-Hughes Taylor)
numerical integration method.
MATVP Material Property Entry
The MATVP entry has completely changed so that both SOL 400 and 600 may use the same entry. Please
refer to MATVE (SOL 600), 1775 in the MSC Nastran Quick Reference Guide for the new description and
be sure to update any existing input files that have an older entry and need to be run with this release. An
example is TPL model vcreep.dat.
MATSMA Shape Memory Alloy Material Property Entry
A new shape memory allow material property entry is now available for use both in SOL 400 and 600.
Please see MATSMA (SOL 600), 1742 in the MSC Nastran Quick Reference Guide for a full description
of this entry.
FISTIF Initial friction stiffness for model 6 used in first cycle of an increment to define the
friction stiffness matrix in cases where a touching node has a zero normal force and
the amount of sliding does not exceed the elastic sticking limit (Real, Default = 0.0 in
which case the program calculates it) PARAMETERS (5,1)
SNGMIN Minimum value that indicates a singularity if a direct solver is used (Real, Default =
0.0 in which case the value is set internally by the program) PARAMETERS (5,2)
RTMAX Maximum change in temperature per iteration in radiation simulations (Real, Default
= 10 times the maximum error in temperature estimate or 100.0) PARAMETERS (5,3)
Main Index
13
CHAPTER 2
Implicit Nonlinear
Nonlinear Elastic Orthotropic Materials
An orthotropic material model that allows the user to enter the nine material parameters as a function of
strain and temperature is now available. This is defined through the MATNLE6 and TABLE3Di options.
Composite Integration Methods to Reduce Computer Time
SOL 600 allows composite materials to be fully nonlinear. The properties of each layer may have
plasticity and/or have properties that vary with temperature. Often analyses are conducted where the
material properties are assumed to remain linear and are at a constant temperature. For such cases,
computer time can be reduced by significant amounts by taking these factors into account. A new entry,
PCOMPF, is available to specify which elements can use the faster integration methods. This entry is
shown below and an example is compos1-fast3.dat in the tpl directory.
Format:
Alternate Formats:
1 2 3 4 5 6 7 8 9 10
PCOMPF INT PID1 THRU PID2 BY N
1 2 3 4 5 6 7 8 9 10
PCOMPF INT PID PID PID1 THRU PID2 PID3 THRU
PID4 PID5 TO PID6 PID PID PID PID7
THRU PID8 BY N
1 2 3 4 5 6 7 8 9 10
PCOMPF INT ALL
Field Contents
PID1 Property identification number. (0 < Integer < 10000000) corresponds to a matching
PCOMP or PCOMPG entry.
INT INT=1, (Default), conventional through the thickness integration of each layer, allows
all available material behavior through the thickness.
INT=2, linear elastic material, fast-integrated through the thickness - thermal strains
and temperature dependent material properties are not allowed.
INT=3, linear elastic material, fast integrated through the thickness, temperature
dependent elasticity, and thermal strains are allowed.
Main Index
14
New SOL 600 Bulk Data Entries and Parameters
Table 2-1 contains new Bulk Data entries for SOL 600 in MSC Nastran 2008 r1. More details can be
found in the MSC Nastran Quick Reference Guide.
Table 2-2 contains new Parameters for SOL 600 in MSC Nastran 2008 r1. More details can be found in
the MSC Nastran Quick Reference Guide.
Table 2-1 New Bulk Data Entries for SOL 600
New for MSC Nastran 2008 r1 (SOL 600)
Bulk Data Entries Description
CSSHL Defines the connection for a solid shell with 6 or 8 grid points.
MATSMA Material properties for shape memory alloys (SOLs 400 and 600 only)
MATNLE6 Properties for nonlinear orthotropic elastic material
MT16SEL Limits elements and/or grid results to selected elements or grids for t16
and t19 file results
PSSHL Defines the properties for solid shell (CSSHL) elements.
SANGLE Defines automatic analytical contact threshold angle for multiple
subcases.
Table 2-2 New Parameters for SOL 600
Parameters Description
MARCMATT (Integer) Determines if Marc input file will be created with materials using the
table-driven formats or not (Default = -1 if parameter is not entered)
MARROUTT Determines whether an inconsistent set of outputs between the Marc t16 file
(selected using MARCOUT) and standard Nastran output selected using Case
Control requests (and param,post) is allowed or not (Default = -1 if parameter is
not entered)
MBENDCAP Determines how PBEND internal pressure will be treated for SOL 600, (Default
= 1 if this parameter is not entered).
MDAREAMD Option to modify or not modify all DAREA entries which are not associated with
any other loads (DAREA entries that supply the actual load)
MFORCOR1 Option to correct forces entered twice (at the same node) in multiple subcases.
MINVASHF Inverse power auto sift value.
MINVCITR Inverse power method, number of iterations.
MINVCSHF Inverse power shift frequency in Hz.
Main Index
15
CHAPTER 2
Implicit Nonlinear
MINVCTOL Inverse power convergence tolerance.
MINVFMAX Inverse power max frequency to extract in Hz.
MINVNMOD Inverse power max number of modes to extract.
MRENUELE It is best if MRENUELE is specified in the SOL entry. Some models will not have
memory allocated properly if this parameter is placed in the bulk data.
(Integer) Determines if SOL 600 elements will be renumbered or not (Default = -1
if parameter is not entered and MRENUELE is not entered on the SOL 600 entry)
MRENUGRD It is best if MRENUGRD is specified in the SOL entry. Some models will not
have memory allocated properly if this parameter is placed in the bulk data.
(Integer) Determines if SOL 600 grid ids will be renumbered or not (Default = -1
if parameter is not entered and MRENUGRD is not entered on the SOL 600
entry)
MRENUMBR Determines if both grid and element IDs for SOL 600 will be renumbered or not.
MRPELAST Determines whether PELAST will be skipped or cause the job to abort for
SOL 600, (default = -1 if parameter is not entered). SOL 600 does not support
PELAST. PBUSHT along with CBUSH and PBUSH should be used instead.
MRPREFER Determines to output SOL 600 stresses on the t16 file in the standard Marc
coordinate system for the element or the preferred (layer) coordinate system
when the model contains composite elements.
MSPEEDCB Determines whether CBEAM increased speed options are to be applied. This
option may be necessary for models with a large number of beams whose element
IDs are large.
MTABLD1M Option to modify or not to modify all TABLED1 entries which do not start with
the first point of (0.0, 0.0)
MTABLD1T Specifies the second time value of all TABLED1 entries that do not start with the
first point being (0.0, 0.0) if PARAM,MTABLD1M=1.
MULRFORC Option to activate multiple RFORCE entries for different portions of the model in
the same subcase.
Table 2-2 New Parameters for SOL 600
Parameters Description
Main Index
16
Main Index
Chapter 3: NVH and Acoustics MSC Nastran 2008 r1 Release Guide

3
NVH and Acoustics
NVH Enhancements
Main Index
NVH Enhancements
18
NVH Enhancements
ACMS with Acoustic External Superelement Creation
The ACMS feature (see the DOMAINSOLVER ACMS (PARTOPT=DOF) Executive Control statement
in the MSC Nastran Quick Reference Guide) is now fully integrated with the creation of an external
acoustic superelement which contains both the fluid cavity and the fluid-structural boundary. External
superelement creation is requested with the EXTSEOUT Case Control command. For an acoustic
external superelement, the component modes and their associated reduced stiffness, mass, etc. matrices
are computed separately for the fluid and structure. If QSETi and SPOINT Bulk Data entries are used
to define the generalized coordinates then there must be a sufficient number to accommodate both the
fluid and structure modes. If there are insufficient generalized coordinates then the program will truncate
both the fluid and structural modes proportionally. It is for this reason that PARAM,AUTOQSET,YES is
strongly recommended to avoid potential modal truncation. Fluid points may also be specified on the
boundary of the superelement using the ASETi entry. However, free-fixed or free-free fluid or structure
boundaries are not permitted with ACMS.
Multiple RANDOM Looping
Prior to this release, only one set of RANDPS Bulk Data entries could be selected per run. In other
words, the RANDOM Case Control command could only reference a single RANDPS set identification
number (SID). In this version multiple SIDs may be specified on the SET command if its identification
number is in turn referenced on a RANDOM command. For example;
SET 1000 = 101 103 107 110
RANDOM = 1000
where 101, 103, 107, and 110 refer to multiple RANDPS SIDs. It should be noted for this type of usage
the SET id must be unique with respect to all RANDPS SIDs; e.g., 1000 is not an SID on any RANDPS
entry.
Sparse OUTPUT4 Format for External Superelement Creation
The sparse OUTPUT4 format option is now used for EXTSEOUT (MATRIXOP4=unit) Case Control
command. This will result in significant disk space reduction of the resulting op4 file.
Binary op2 and op4 Compatibility Robustness
Starting in version V2004 r3, binary op2 and op4 files could be read across dissimilar platforms.
However, several errors were encountered since V2004 r3 and are now corrected in MSC Nastran 2008
r1.
Main Index
19
CHAPTER 3
NVH and Acoustics
Merged Superelement Results
PARAM, FULLSEDR, YES may be specified in a superelement analysis to merge several types of
results (displacements, stresses, etc.) across all superelements into a single non-superelement results
format. FULLSEDR is intended for superelement models which contain unique IDs across all element
and grid points. FULLSEDR benefits third party post-processing programs which have difficulty
digesting superelement results in the op2 or .pch files.
Main Index
NVH Enhancements
20
Main Index
Chapter 4: Numerical Methods and High Performance Computing MSC Nastran 2008 r1 Release
Guide
=
4
Numerical Methods and High
Performance Computing
Factor Matrix Caching for Lanczos
New TAUCS Indefinite Solver Improves Lanczos Performance
Shared Memory Parallel (SMP) Scalability Improvements for Static

Analysis
New MAXRATIO Information Output
New SPARSESOLVER MDTSTATS Information Output

Main Index
22
Introduction
In solution sequences where many linear systems are solved using the same coefficient matrix, the FBS
time can be significant when the factor matrix is stored out of core. Examples include dynamic solution
sequences which use the Lanczos method and nonlinear transient response. In MSC Nastran 2008 r1,
new logic has been introduced to cache as much of the factor as possible in memory.
Benefits
The reduced I/O can cut the elapsed time for FBS by five to 30 percent, depending on the size of the factor
matrix, the number of right-hand sides, and the amount of memory available.
Method and Theory
The underlying method has not changed; only the memory usage. Previously, only the minimum amount
of factor data needed to perform the FBS was read from the factor data block each time and FBS was
required. Now, as much of the factor as possible is cached in memory between FBS calls, reducing the
I/O required.
Inputs
For Lanczos, no input is required, except when running on Linux IA64. This option has not been
beneficial on Linux IA64, but it can be turned-on setting SYSTEM cell 146 to -1. For nonlinear transient
analysis, the factor caching logic must be activated by setting SYSTEM(146) to -1. For comparison
purposes, the factor-caching logic can be deactivated by setting SYSTEM(146)=+1.
Outputs
A new System Information Message 4157 will appear in the. f04 file:
In addition, when SYSTEM cell 166 to 2, additional time stamps FBSI BGN and FBSI END appear
in the .f04 file.
MEMORY REQUIREMENTS FOR IN-CORE FACTOR OPTION:
AVAILABLE MEMORY: 229909 KWORDS
NUMBER OF TOTAL FRONTS: 42935
NUMBER OF FRONTS WHICH FIT IN CORE: 14225
EST MEMORY FOR ENTIRE FACTOR TO IN CORE: 586268 KWORDS
Main Index
23
CHAPTER 4
Numerical Methods and High Performance Computing
Guidelines and Limitations
It is difficult to estimate the amount of memory to specify for a given model so that most of the factor
can be cached. The estimate program can give a starting point. A rule of thumb would be to give 3
times the amount of memory recommended in the estimate output, but no more than 75% of the
physical memory available on the machine.
Demonstration Examples
Example 1:
The following example is a powertrain model with 160,000 grids and 940,000 degrees of freedom.
Twenty mode shapes are required using Lanczos. A total of 14 FBS operations are performed. The job
is run on workstation with two dual-core 64-bit Pentium Xeon processors running at 3GHz, 8GB of
physical memory. The job is submitted with mem=6gb. By caching the factor, the total FBS time is
reduced by 30%, resulting in a 13% reduction in the overall READ time.
Lanczos Performance Improvement
0
50
100
150
200
250
300
350
FBS time
[seconds]
READ time
[seconds]
C
P
U

s
e
c
o
n
d
s
Factor out of core
Factor Cached
Main Index
New TAUCS Indefinite Solver Improves Lanczos Performance
24
New TAUCS Indefinite Solver Improves Lanczos
Performance
Introduction
A new symmetric indefinite factorization method from the TAUCS software project
www.tau.ac.il/~stoledo/taucs/ has been integrated into MSC Nastran 2008 r1. It is available in the
DCMP, DECOMP, SOLVE, RMG2, SDR2, and READ modules, with the main focus on the READ
module. The new solver keeps all data in memory.
Benefits
The new method can significantly improve the factorization and FBS time, particularly in the Lanczos
procedure, for problems which fit in memory. In this release, the method is only recommended on the
Linux x86_64 platforms (Intel EM64T and AMD Opteron).
Method and Theory
The new method is based on a hybrid left-looking /supernodal multifrontal technique, with emphasis on
data locality.
Inputs
The new method can be selected by setting SYSTEM cell 166 to 16384.
Outputs
The following information is printed in the .f04 file.
In this release, this method has only been tuned for the Linux x86_64 platform (Intel EM64T/AMD
Opteron) and is only recommended for that platform. This method must keep all data in memory, so it
is recommended that it be submitted with 75% of the physical memory available.
Elimination tree depth is 7043
Symbolic Analysis of LDL^T: 1.53e+08 nonzeros, 9.48e+10 flops, 1.40e+09 bytes in L
Relaxed Analysis of LDL^T: 1.80e+08 nonzeros, 1.14e+11 flops, 1.72e+09 bytes in L
Symbolic Analysis = 3.415 seconds (3.411 cpu)
12:07:41 1:05 15753.0 24.0 61.1 6.5 TAUD END
12:07:47 1:11 16890.0 1137.0 66.7 5.5 TAUD BGN
Using blocked update in dense factorization.
Supernodal Left-Looking LDL^T = 49.121 seconds (48.390 cpu)
Post Analysis of LDL^T: 1.80e+08 nonzeros, 1.14e+11 flops, 1.72e+09 bytes in L
Main Index
25
CHAPTER 4
A normal modes computation has been performed using the default factorization method (SPDC) and
the new TAUCS method on the following models. The jobs are run on a workstation with two dual-core
64-bit Pentium Xeon processors running at 3GHz, 8GB of physical memory. Each job was submitted
with mem=6gb.
Model Grids DOF Modes Factorizations FBSs
Powertrain 160,000 940,000 20 2 14
Rotor 197,000 592,000 8 2 8
Viga 139,000 413,000 5 2 5
Van 103,000 584,000 73 2 43
Lanczos Performance
0
200
400
600
800
1000
1200
1400
Rotor viga Powertrain van
Model
R
E
A
D

t
i
m
e

[
c
p
u

s
e
c
o
n
d
s
]
SPDC
TAUCS
Main Index
Shared Memory Parallel (SMP) Scalability Improvements for Static Analysis
26
Shared Memory Parallel (SMP) Scalability
Improvements for Static Analysis
Introduction
In solution sequences where a linear system must be solved with a large number of right-hand sides,
several passes over the factor matrix may be needed to compute all of the solution vectors. This
performance enhancements keeps as much of the factor matrix as possible in memory to reduce the I/O,
improve overall performance, particularly SMP performance.
Benefits
The performance of any large FBS with many right-hand sides will be improved by as much as 30%. For
example, superelement models with static (Guyan) reduction, statics models with many load cases, and
heat transfer models with radiation will benefit from this enhancement.
Method and Theory
The underlying method has not changed; only the memory usage. Previously, only the minimum amount
of factor data need to perform the FBS was read from the factor data block during each FBS pass. Now,
as much of the factor as possible is cached in memory between FBS passes, reducing the I/O required.
Inputs
The feature is automatically activated on all platforms except Linux IA64 when enough memory is
available to store at least 32 right-hand side vectors, and when the factor and right-hand side have the
same data type (both real or both complex). The minimum number of right-hand sides required to
activate the feature can be overridden with the value of SYSTEM cell 70.
Outputs
A system information message is printed in the .f04 file if any part of the factor is cached:
Guidance and Limitations
1. This feature is not available on the Linux IA64 platform.
*** SYSTEM INFORMATION MESSAGE 4157 (PREFAC1)
A PORTION OF THE SPARSE FACTOR HAS BEEN CACHED IN MEMORY FOR THE FBS.
34435 FRONTAL MATRICES OUT OF A TOTAL OF 37881 ARE STORED IN MEMORY.
MEMORY AVAILABLE: 412 M WORDS
ESTIMATED ADDITIONAL MEMORY NEEDED TO STORE THE ENTIRE FACTOR: 281 M WORDS
Main Index
27
CHAPTER 4
2. This feature requires additional memory. The recommended amount is three times the amount
specified by the estimate program plus enough to hold 32 right-hand side vectors.
The following example is a linear statics model of a car body with 42,000 grid points, 246,000 degrees
of freedom, and 8,300 load cases. The job was run on an IBM pSeries workstation with 8 1.9GHz
power5 processors, and 8GB of physical memory.
Linear Static Performance
0
100
200
300
400
500
600
700
800
900
1000
Serial SMP=2 SMP=4
E
l
a
p
s
e
d

T
i
m
e

[
S
e
c
o
n
d
s
]
R2
R3(Factor Cached)
Main Index
28
Introduction
A new interface is now available for analysts to better control the generation of matrix diagonal term ratio
statistics produced by the sparse symmetric matrix decomposition process in the DCMP module. The
matrix diagonal term ratio statistics are sometimes useful in determining the quality of the matrix
decomposition process. In general, for linear static analysis, high or negative ratios indicate a loss of
accuracy and could be indicative of a modeling error. The MAXRATIO functionality was a pre-release
capability in the MSC Nastran 2007 r1 release. For MSC Nastran 2008 r1 this is now a production
capability.
Benefits
The new interface provides analysts with more control over the process than the existing method of
supplying a value for the MAXRATIO DMAP parameter. In addition, a new output data option is
available in the form of a simple bar chart that provides a more comprehensive view of the ratio data.
Method and Theory
No new theory is involved. The method simply involves the computation of a ratio defined as the original
matrix diagonal term divided by the decomposed matrix diagonal term. These ratios are placed in a table
together with the external identifier associated with the row/column of the term. This table is then
processed according to the options requested by the user.
Inputs
The matrix diagonal term ratio output options are controlled by keywords specified on the
SPARSESOLVER Executive Control statement. See New SPARSESOLVER MDTSTATS Information
Output, 32 for a complete description of this statement.
Outputs
The matrix diagonal term ratios can be presented in two different views. The first view is the table view,
in which each ratio is listed together with the external identifier of the row/column of the matrix, as well
as the original input matrix diagonal term. This format is virtually identical to that produced by the
previous version when any ratio exceeds the value of the MAXRATIO input parameter.
The second view of the ratios is statistical in nature. It is similar to a bar chart. A series of bar segments
are generated. There are two options for specifying the segment widths of the bars. The default option
uses powers of 10 as the widths (e.g., 10.0 to 100.0, and 100.0 to 1000.0). The second option allows the
user to specify how many segments are desired. The program will compute the segment width using the
Main Index
29
CHAPTER 4
maximum and minimum ratios. For each bar in the chart, the total number of terms in the range is
tabulated together with a visual indication of the percentage number of terms in that particular bar.
Note that when negative matrix diagonal term ratios are detected, they will always be output if the
TABLE option is specified.
These new views of the ratios do not replace any existing diagnostics generated by the DCMP module if
a problem is detected. Under these conditions, output from the table view may duplicate previous output
generated by DCMP module error processing.
The matrix diagonal term ratio statistics are sometimes useful in determining the quality of the matrix
decomposition process. In general, high ratios indicate a loss of accuracy. The feature can be used by
taking all of the program defaults for the various control variables. These defaults produce both the table
and bar outputs. The table is limited to 25 ratios that exceed 1.0E+05. The bar chart uses powers of ten
for segment widths. This can be done by adding
SPARSESOLVER DCMP (MDTRATIO)
to the Executive Control Section of the input data file.
The use of this new feature is currently limited to sparse symmetric matrix operations in the DCMP
module.
If there are scalar-type points present in the problem, the degrees of freedom associated with these points
will be grouped into the results for the translational degrees of freedom output.
Demonstration Example
A simple example is presented that demonstrates the use of some of the new features available for output
of the matrix diagonal term ratios. The SPARSESOLVER Executive Control statement is used to specify
the desired features. The example is for demonstration purposes only, and is not representative any
particular modeling situation. The model data consists of a simple plate structure subject to an end load.
Example Input Data
$
$ Example problem to demonstrate matrix diagonal term ratios
$
id test,case
sol 101
SPARSESOLVER DCMP (MDTRATIO)
cend
spc=100
load=1000
disp=all
begin bulk
grdset,,,,,,,6
cquad4,101,101,1,2,52,51
Main Index
30
cquad4,102,101,2,3,53,52
cquad4,103,101,3,4,54,53
cquad4,104,101,4,5,55,54
cquad4,105,101,5,6,56,55
cquad4,106,101,6,7,57,56
cquad4,107,101,7,8,58,57
cquad4,108,101,8,9,59,58
cquad4,109,101,9,10,60,59
cquadr,1101,101,1,2,52,51
cquadr,1102,101,2,3,53,52
cquadr,1103,101,3,4,54,53
cquadr,1104,101,4,5,55,54
cquadr,1105,101,5,6,56,55
cquadr,1106,101,6,7,57,56
cquadr,1107,101,7,8,58,57
cquadr,1108,101,8,9,59,58
cquadr,1109,101,9,10,60,59
grid, 1,, 0.0,0.0,0.0
grid, 2,, 1.0,0.0,0.0
grid, 3,, 2.0,0.0,0.0
grid, 4,, 3.0,0.0,0.0
grid, 5,, 4.0,0.0,0.0
grid, 6,, 5.0,0.0,0.0
grid, 7,, 6.0,0.0,0.0
grid, 8,, 7.0,0.0,0.0
grid, 9,, 8.0,0.0,0.0
grid,10,, 9.0,0.0,0.0
grid,51,, 0.0,1.0,0.0
grid,52,, 2.4,1.0,0.0
grid,53,, 3.5,1.0,0.0
grid,54,, 4.6,1.0,0.0
grid,55,, 5.7,1.0,0.0
grid,56,, 6.8,1.0,0.0
grid,57,, 7.9,1.0,0.0
grid,58,, 9.0,1.0,0.0
grid,59,,10.1,1.0,0.0
grid,60,,11.2,1.0,0.0
$
ctria3,201,101,101,102,151
ctria3,202,101,102,152,151
ctria3,203,101,102,103,152
ctria3,204,101,103,153,152
ctria3,205,101,103,104,153
ctria3,206,101,104,154,153
ctria3,207,101,104,105,154
ctria3,208,101,105,155,154
ctriar,1201,101,101,102,151
ctriar,1202,101,102,152,151
ctriar,1203,101,102,103,152
ctriar,1204,101,103,153,152
ctriar,1205,101,103,104,153
ctriar,1206,101,104,154,153
ctriar,1207,101,104,105,154
ctriar,1208,101,105,155,154
grid,101,, 0.0,0.0,0.0
grid,102,, 1.0,0.0,0.0
grid,103,, 2.0,0.0,0.0
grid,104,, 3.0,0.0,0.0
grid,105,, 4.0,0.0,0.0
grid,151,, 0.0,1.0,0.0
Main Index
31
CHAPTER 4
grid,152,, 3.4,1.0,0.0
grid,153,, 4.5,1.0,0.0
grid,154,, 5.6,1.0,0.0
grid,155,, 6.7,1.0,0.0
$
pshell,101,1,0.05,1
mat1,1,10.+6,,0.33
spc1,100,123,1,101
spc1,100,3,5,55,105,155
spc1,100,1,55,155
spc1,100,2,1,101
force,1000,10,,1000.0,1.0,0.0,0.0
force,1000,60,,1000.0,1.0,0.0,0.0
force,1000,105,,1000.0,1.0,0.0,0.0
force,1000,155,,1000.0,1.0,0.0,0.0
enddata
Example Output
The output generated by the example is shown as follows. Notice that there are two separate sections of
output: one for translational degrees of freedom, and one for rotational degrees of freedom. Within each
section, both a bar chart and table of matrix diagonal term ratios are output.
TRANSLATIONAL DOF DIAGONAL TERM RATIO STATISTICS CHART FOLLOWS FOR THE DECOMPOSITION OF MATRIX KLL
------------------------------------------------|---------------------------------------------------------------------------
DIAGONAL TERM RATIO RANGE #TERMS % TOT |MAXIMUM RATIO = 6.90963E+02 MINIMUM RATIO = 1.00000E+00
------------------------------------------------|---------------------------------------------------------------------------
1.0000E+00 TO 1.0000E+01 62 79.49 |**************************************************************************>
1.0000E+01 TO 1.0000E+02 12 15.38 |***************
1.0000E+02 TO 1.0000E+03 4 5.13 |*****

0
0 MATRIX/FACTOR DIAGONAL TERMS RATIO SUMMARY TABLE FOR TRANSLATIONAL DOF SORTED ON DIAGONAL RATIO
GRID POINT ID DEGREE OF FREEDOM MATRIX/FACTOR DIAGONAL RATIO MATRIX DIAGONAL
(TOP 1 RATIOS>MAXRAT= 6.90963E+02)
58 T3 6.90963E+02 5.65535E+04

ROTATIONAL DOF DIAGONAL TERM RATIO STATISTICS CHART FOLLOWS FOR THE DECOMPOSITION OF MATRIX KLL
------------------------------------------------|---------------------------------------------------------------------------
DIAGONAL TERM RATIO RANGE #TERMS % TOT |MAXIMUM RATIO = 3.35974E+02 MINIMUM RATIO = 1.00000E+00
------------------------------------------------|---------------------------------------------------------------------------
1.0000E+00 TO 1.0000E+01 38 63.33 |***************************************************************
1.0000E+01 TO 1.0000E+02 18 30.00 |******************************
1.0000E+02 TO 1.0000E+03 4 6.67 |*******
0
0 MATRIX/FACTOR DIAGONAL TERMS RATIO SUMMARY TABLE FOR ROTATIONAL DOF SORTED ON DIAGONAL RATIO
GRID POINT ID DEGREE OF FREEDOM MATRIX/FACTOR DIAGONAL RATIO MATRIX DIAGONAL
(TOP 1 RATIOS>MAXRAT= 3.35974E+02)
58 R2 3.35974E+02 2.14135E+04
Main Index
32
Introduction
A new interface is now available to control the generation of matrix diagonal term statistics for the input
matrix to the sparse symmetric matrix decomposition process in the DCMP module. The matrix diagonal
term statistics can be useful in determining the quality of the model in regions that produce unusually
large or small terms. In general, for linear static analysis, model degrees of freedom with small
stiffnesses could indicate areas where loads will produce large displacements. This feature complements
the MDTRATIO option that controls MAXRATIO output described previously. The SPARSESOLVER
MDTSTATS functionality was a pre-release capability in the MSC Nastran 2007 r1 release. For MSC
Nastran 2008 r1 this is now a production capability.
Benefits
The new interface provides another means of identifying potential modeling errors other than monitoring
the MAXRATIO statistics. One of the new output data options is a simple bar chart that provides a more
comprehensive view of the diagonal term data.
Method and Theory
No new theory is involved. The method involves adding the original matrix diagonal term to the ratio
table where the computation of the ratio is defined to be the original matrix diagonal term divided by the
decomposed matrix diagonal term. These terms are placed together in a table with the external identifier
associated with the row/column of the term. This table is then processed according to the options
requested by the user.
Inputs
The matrix diagonal term statistical output options are controlled by keywords specified on the
SPARSESOLVER Executive Control statement. See the MSC Nastran Quick Reference Guide for a
complete description of this statement.
Outputs
The matrix diagonal term statistics can be presented in two different views. The first is the table view in
which each diagonal term is listed together with the external identifier of the row/column of the matrix,
as well as with the Aii/Lii diagonal term ratio. This format is almost identical to that produced now when
any ratio exceeds the value of the MAXRATIO input parameter. The second view of the diagonal terms
is statistical in nature, similar to a bar chart. A series of bar segments is generated. There are two options
for specifying the segment widths of the bars. The default option uses powers of 10 as the widths (e.g.,
10.0 to 100.0, and 100.0 to 1000.0). The second option allows the user to specify how many segments
are desired. The program will compute the segment width using the maximum and minimum diagonal
Main Index
33
CHAPTER 4
terms. For each bar in the chart, the total number of terms in the range is tabulated together with a visual
indication of the percentage number of terms in that particular bar.
These new views of the diagonal terms do not replace any existing diagnostics generated by the DCMP
module if a problem is detected. Under these conditions, output from the table view may duplicate
previous output generated by DCMP module error processing.
The matrix diagonal term statistics are sometimes useful in determining areas of the model that may pose
problems during the decomposition process, or afterwards during the solution of equations that produce
displacements. In general, unusually large or small values could indicate a modeling problem. The
feature can be used by taking all of the program defaults for the various control variables. These defaults
produce both the table and bar outputs. The table is limited to the 25 largest terms that exceed 1.0E+10,
and the 25 smallest terms less than 1.0. The bar chart uses powers of ten for segment widths. This can
be done by adding
SPARSESOLVER DCMP ( MDTSTATS )
to the Executive Control Section of the input data file.
The use of this new feature is currently limited to sparse symmetric matrix operations in the DCMP
module.
If there are scalar-type points present in the problem, the degrees of freedom associated with these points
will be grouped into the results for the translational degrees of freedom output.
Demonstration Example
A simple example is presented that demonstrates the use of some of the new features available for output
of the matrix diagonal term statistics. The SPARSESOLVER Executive Control statement is used to
specify the desired features. The example problem is used for demonstration purposes only, and is not
representative of any particular model. The model data consists of a simple plate structure subject to an
end load. The model properties have been designed to indicate a potential problem in the bending
properties at grid points 4 and/or 54.
Example Input Data
$
$ Example problem to demonstrate matrix diagonal term statistics
$
id test,case
sol 101
$ Note: SPARSOLVER DCMP options must be enclosed in ()
$ Note also that MDTSTATS options must also be enclosed in their own
()
SPARSESOLVER DCMP ( MDTSTATS = ( CHART,
TABLET, NMAXVALT=10, MAXVALT=1.0e+08,
Main Index
34
NMINVALT=20, MINVALT=1.0,
TABLER, NMINVALR=30, MINVALR=100.0 ) )
cend
spc=100
load=1000
disp=all
begin bulk
grdset,,,,,,,6
cquad4,101,101,1,2,52,51
cquad4,102,101,2,3,53,52
cquad4,103,102,3,4,54,53
cquad4,104,102,4,5,55,54
cquad4,105,101,5,6,56,55
cquad4,106,101,6,7,57,56
cquad4,107,101,7,8,58,57
cquad4,108,101,8,9,59,58
cquad4,109,101,9,10,60,59
cquadr,1101,101,1,2,52,51
cquadr,1102,101,2,3,53,52
cquadr,1103,102,3,4,54,53
cquadr,1104,102,4,5,55,54
cquadr,1105,101,5,6,56,55
cquadr,1106,101,6,7,57,56
cquadr,1107,101,7,8,58,57
cquadr,1108,101,8,9,59,58
cquadr,1109,101,9,10,60,59
grid, 1,, 0.0,0.0,0.0
grid, 2,, 1.0,0.0,0.0
grid, 3,, 2.0,0.0,0.0
grid, 4,, 3.0,0.0,0.0
grid, 5,, 4.0,0.0,0.0
grid, 6,, 5.0,0.0,0.0
grid, 7,, 6.0,0.0,0.0
grid, 8,, 7.0,0.0,0.0
grid, 9,, 8.0,0.0,0.0
grid,10,, 9.0,0.0,0.0
grid,51,, 0.0,1.0,0.0
grid,52,, 2.4,1.0,0.0
grid,53,, 3.5,1.0,0.0
grid,54,, 4.6,1.0,0.0
grid,55,, 5.7,1.0,0.0
grid,56,, 6.8,1.0,0.0
grid,57,, 7.9,1.0,0.0
grid,58,, 9.0,1.0,0.0
grid,59,,10.1,1.0,0.0
grid,60,,11.2,1.0,0.0
$
ctria3,201,101,101,102,151
ctria3,202,101,102,152,151
ctria3,203,101,102,103,152
ctria3,204,101,103,153,152
ctria3,205,101,103,104,153
ctria3,206,101,104,154,153
ctria3,207,101,104,105,154
Main Index
35
CHAPTER 4
ctria3,208,101,105,155,154
ctriar,1201,101,101,102,151
ctriar,1202,101,102,152,151
ctriar,1203,101,102,103,152
ctriar,1204,101,103,153,152
ctriar,1205,101,103,104,153
ctriar,1206,101,104,154,153
ctriar,1207,101,104,105,154
ctriar,1208,101,105,155,154
grid,101,, 0.0,0.0,0.0
grid,102,, 1.0,0.0,0.0
grid,103,, 2.0,0.0,0.0
grid,104,, 3.0,0.0,0.0
grid,105,, 4.0,0.0,0.0
grid,151,, 0.0,1.0,0.0
grid,152,, 3.4,1.0,0.0
grid,153,, 4.5,1.0,0.0
grid,154,, 5.6,1.0,0.0
grid,155,, 6.7,1.0,0.0
$
pshell,101,1,0.05,1
pshell,102,1,0.05,2
mat1,1,10.+6,,0.33
mat1,2,10.+1,,0.33
spc1,100,123,1,101
spc1,100,3,5,55,105,155
spc1,100,1,55,155
spc1,100,2,1,101
force,1000,10,,1000.0,1.0,0.0,0.0
force,1000,60,,1000.0,1.0,0.0,0.0
force,1000,105,,1000.0,1.0,0.0,0.0
force,1000,155,,1000.0,1.0,0.0,0.0
enddata
Example Output
The output generated by the example is shown as follows. There are two separate sections of output: one
for translational degrees of freedom and one for rotational. Within each section, both a bar chart and
table of matrix diagonal terms are output.
===============================================================================================================================
TRANSLATIONAL DOF Aii DIAGONAL TERMS STATISTICS CHART FOLLOWS FOR MATRIX KLL Matrix Trace(Aii) = 1.27351E+08
------------------------------------------------|---------------------------------------------------------------------------
| MAXIMUM VALUE = 5.94121E+06 MINIMUM VALUE = 4.07806E-01
MATRIX DIAGONAL TERM RANGE #TERMS % TOT | GRID ID = 104, DOF = T2 GRID ID = 54, DOF = T3
------------------------------------------------|---------------------------------------------------------------------------
1.0000E-01 TO 1.0000E+00 2 2.56 |***
1.0000E+03 TO 1.0000E+04 2 2.56 |***
1.0000E+04 TO 1.0000E+05 18 23.08 |***********************
1.0000E+05 TO 1.0000E+06 6 7.69 |********
1.0000E+06 TO 1.0000E+07 50 64.10 |****************************************************************
===============================================================================================================================
Main Index
36
0 MATRIX/FACTOR DIAGONAL TERMS SUMMARY TABLE FOR TRANSLATIONAL DOF SORTED ON Aii DIAGONAL
GRID POINT ID DEGREE OF FREEDOM Aii TERM Lii TERM Aii/Lii RATIO
(TOP 1 VALUES > 5.94121E+06)
104 T2 5.94121E+06 3.57301E+06 1.66280E+00
0 MATRIX/FACTOR DIAGONAL TERMS SUMMARY TABLE FOR TRANSLATIONAL DOF SORTED ON Aii DIAGONAL
(TOP 2 VALUES < 1.00000E+00)
54 T3 4.07806E-01 4.07806E-01 1.00000E+00
4 T3 4.70350E-01 2.84250E-01 1.65471E+00
===============================================================================================================================
ROTATIONAL DOF Aii DIAGONAL TERMS STATISTICS CHART FOLLOWS FOR MATRIX KLL Matrix Trace(Aii) = 4.52211E+05
------------------------------------------------|---------------------------------------------------------------------------
| MAXIMUM VALUE = 2.34493E+04 MINIMUM VALUE = 4.71107E-02
MATRIX DIAGONAL TERM RANGE #TERMS % TOT | GRID ID = 9, DOF = R2 GRID ID = 54, DOF = R1
------------------------------------------------|---------------------------------------------------------------------------
1.0000E-02 TO 1.0000E-01 2 3.33 |***
1.0000E-01 TO 1.0000E+00 2 3.33 |***
1.0000E+02 TO 1.0000E+03 3 5.00 |*****
1.0000E+03 TO 1.0000E+04 36 60.00 |************************************************************
1.0000E+04 TO 1.0000E+05 17 28.33 |****************************

===============================================================================================================================
0 MATRIX/FACTOR DIAGONAL TERMS SUMMARY TABLE FOR ROTATIONAL DOF SORTED ON Aii DIAGONAL
(TOP 1 VALUES > 2.34493E+04)
9 R2 2.34493E+04 4.87230E+03 4.81277E+00
0 MATRIX/FACTOR DIAGONAL TERMS SUMMARY TABLE FOR ROTATIONAL DOF SORTED ON Aii DIAGONAL
(TOP 4 VALUES < 1.00000E+02)
54 R1 4.71107E-02 3.08701E-02 1.52610E+00
4 R1 4.71334E-02 9.03202E-03 5.21847E+00
54 R2 1.48139E-01 8.62427E-02 1.71770E+00
4 R2 1.48232E-01 7.96377E-02 1.86133E+00
Main Index
Chapter 5: Upward Compatibility
=
5
Upward Compatibility
TEMPERATURE Case Control Command
Improvements in Fluid Eigenvalue Analysis
FLUID GRID Points and Partitioning
Distributed Memory Parallel (DMP) Diagnostic Messages
System Information Message (SIM) 6916

Main Index
38
According to Remark 8 under the Case Control command TEMPERATURE (Ch. 4) in the MSC Nastran
Quick Reference Guide, TEMPERATURE(INITIAL) and TEMPERATURE(MATERIAL) cannot be
specified in the same run and User Fatal Message 633 will be issued. User Fatal Message 633 is also
issued if TEMPERATURE(BOTH) is specified with TEMPERATURE(INIT) and
TEMPERATURE(MATERIAL). However, in MSC Nastran 2007 r1 and prior, this rule was not enforced
when just TEMPERATURE was specified with TEMPERATURE(INITIAL) or
TEMPERATURE(MATERIAL); and, depending on their relative locations in the Case Control Section,
one of them would be ignored and results will be wrong. For example, the following input file (modified
from TPL problem tempload):
sol 101
cend
temp(init) = 10
subcase 1
temp = 20
load = 100
spc = 10
disp = all
begin bulk
force,100,3,0,100.0,1.0,0.0,0.0
grid, 1,, 0.0,0.0,0.0
grid, 2,,10.0,0.0,0.0
grid, 3,,20.0,0.0,0.0
cbar,1,10,1,2,0.0,0.0,1.0
pbar,10,100,1.0,1.0,1.0,1.0
mat1,100,1.+4,,0.3,,1.-3
rbar,2,2,3,,,,,2.0-4
temp,10,1,51.0
temp,10,2,52.0
temp,10,3,53.0
temp,20,1,61.0
temp,20,2,62.0
temp,20,3,63.0
spc1,10,123456,1
enddata
produces the following results in versions MSC Nastran 2007 r1 and prior:
In MSC Nastran 2008 r1, this rule is now enforced with TEMPERATURE and UFM 633 will be issued.
To avoid User Fatal Message 633 in MSC Nastran 2008 r1, simply replace TEMPERATURE with
Case Control T1 Displacement at Grid 2 Comment
TEMP(INIT) and TEMP(LOAD) 0.200 Correct answer
TEMP(LOAD) 0.715 Correct answer
TEMP(INIT) and TEMP 0.715 Wrong answer because
TEMP(INIT) is ignored
Main Index
39
CHAPTER 5
TEMPERATURE (LOAD). In MSC Nastran 2009 or later, the BOTH keyword may be removed from
the documentation and the program as an option of the TEMPERATURE command.
Main Index
40
1. The elemental mass matrix formulation for the 4-noded CTETRA fluid element has been
modified to prevent spurious modes. Changes may be observed in the fluids natural frequencies
especially at the higher frequencies. Use NASTRAN SYSTEM(446)=1 to obtain the previous
versions formulation.
2. Householder method is automatically selected for the fluids system modes if the acoustic cavity
is defined in a superelement and there exist fluid boundary points. The Householder method is
more reliable method when there are fluid points on the boundary of an acoustic superelement.
The switch to Householder occurs if the number of estimated fluid modes is less than or equal to
the value of user PARAMeter FLUIDNE (Default=500).
Main Index
41
CHAPTER 5
The GP4 module processes displacement set definition Bulk Data entries (e.g. ASET/ASET1). During
this process, it performs various integrity tests on the data supplied by users. One of these tests verifies
that the degree-of-freedom (DOF) components exist for the points specified. The allowable components
depend upon the point type. For instance, a GRID point has six degrees of freedom and one may specify
any (or all) of the components one through six. An SPOINT on the other hand has only a single DOF
and one may specify only a blank or zero as the component. One must also remember that for some types
of analysis, a GRID point may have a reduced number of components available. For example, in
acoustics, one can define GRID points attached to fluid that have only a single component DOF. If the
DOF integrity test fails, Nastran issues message 2049 that informs the user of the problem. The severity
of the message depends upon whether one uses the standard input format (e.g. ASET) or the alternative
format (ASET1) for the Bulk Data entry. When one uses the standard format, one defines each point and
DOF component code explicitly and it must exist. Otherwise, GP4 issues a FATAL 2049 message
indicating that the point is missing. When one uses the alternate entry format, GP4 is prepared for the
possibility that one or more points may not exist in THRU ranges defined on the entry. For this case, a
missing point/DOF produces a WARNING 2049 message.
Consider the following Bulk Data entries:
Since point 1130 is a fluid GRID point, it has only a single DOF associated with it. This DOF is
referenced with DOF component 1. Standard format entry #2 and alternate format entry #4 both use the
proper DOF component code and GP4 places the entries in the a-set without generating any messages.
Standard format entry #3 and alternate format entry #5 on the other hand, contain DOF components that
do not exist for the specified point.
Previous versions handle the processing of entries #3 and #5 as follows:
For entry #3 (ASET), GP4 issues a FATAL message 2049 indicating that it could not find the
point and the job stops.
For entry #5 (ASET1), GP4 issues a WARNING message 2049 indicating that it could not find
the point and the job continues, BUT, the point is NOT placed in the a-set as requested.
1. GRID,1130,,0.0,0.0,0.0,-1 $ this is a FLUID GRID point (OCID=-1)
2. ASET,1130,1 $ standard format
3. ASET,1130,123456 $ standard format
4. ASET1,1,1130 $ alternate format
5. ASET1,123456,1130 $ alternate format
Main Index
42
MSC Nastran 2008 r1 contains a modification to this process for FLUID GRID points and SPOINTs as
follows:
For entry #3 (ASET), GP4 issues a WARNING message 2049 indicating that certain DOFs are
not available at the point, places the one DOF available at the point in the a-set and continues the
job.
For entry #5 (ASET1), GP4 issues a WARNING message 2049 indicating that certain DOFs are
not available at the point, places the one DOF available at the point in the a-set and continues the
job
Note the difference between MSC Nastran 2008 r1 and previous versions in this area applies only to
FLUID GRID and SPOINT entries found on displacement set membership (partitioning) definition Bulk
Data entry (ASET, ASET1, OMIT, OMIT1, etc.). Existing bulk data files containing illegal
specifications for DOF component codes for FLUID GRIDs and SPOINTs on the partitioning bulk data
entries that ran successfully on previous versions will continue to run, but may produce different results
when run with MSC Nastran 2008 r1 if a DOF becomes a part of the a-set.
Main Index
43
CHAPTER 5
Distributed Memory Parallel (DMP) Diagnostic
Messages
Several DMP diagnostic messages used to indicate one or more of the following keywords MDMODES,
GDMODES, FDMODES, MDACMS, GDSTAT, MDSTAT, FDFREQ in the .f06 and f04 files. They
have been replaced by their proper DOMAINSOLVER description, for example, MDMODES was
replaced by DOMAINSOLVER MODES (PARTOPT=DOF).
Main Index
44
SIM 6916 which looks similar to the example below is no longer printed in the .f06 file unless you set
system(294) to a value greater than zero.
*** SYSTEM INFORMATION MESSAGE 6916 (DFMSYM)
DECOMP ORDERING METHOD CHOSEN: BEND, ORDERING METHOD USED: BEND
Main Index
Chapter 6: Optimization MSC Nastran 2008 r1 Release Guide
6
Optimization
Enhancements in DRESP3
Topometry Optimization
Topography (Bead or Stamp) Optimization
Randomization of an Input Data File
Random Elimination of Element Types
Enhancements in SOL 200 Optimization

Main Index
46
Introduction
DRESP3 is a feature of SOL 200 in MSC Nastran that allows the user to invoke external software to
calculate design responses that are not available as standard DRESP1 quantities or that cannot be
synthesized using the DRESP2 capability. The DRESP3 is a special purpose capability that requires
some work on the users part to function effectively, but it has its adherents who appreciate its ability to
include design responses that are not available from Nastran. Use of this capability has identified three
enhancements for this capability that have been implemented for MSC Nastran 2008 r1:
1. Provision for a capability to provide analytic gradients for the response
2. The ability to produce multiple response outputs from a single DRESP3 call.
3. Reordering of finite difference sensitivities when the DRESP3 has only DRESP1 flags and there
are more DRESP1 responses in the DRESP3 than there are independent design variables in the
model.
Benefits
Analytical gradients provide a performance benefit as well as more robust results than can be expected
from a finite difference approach to obtaining gradients.
The multiple response requirement arises from a typical scenario where a number of design criteria for a
particular component share a common set of inputs. For example, a panel may have criteria on stress,
buckling and fatigue that share parameters for geometry, properties and internal responses. By evaluating
all of these criteria in a single call, duplicate calculations are avoided and the number of calls to the server
are reduced.
The third enhancement above is for the very special application where there are perhaps thousands of
DRESP1 entries and a few hundred design variables. In this case, it makes sense to do the finite
difference gradient calculation by perturbing all the DRESP1 quantities for a particular design variable
and then calling the DRESP3 evaluator. In this way, the number of call to the evaluator is reduced from
2*NRESP1 to 2*NDVI. When NRESP1 >> NDVI, this can provide a major performance improvement
to the extent it enables performing design tasks that were previously out of reach.
User Inputs
The format of the DRESP3 Bulk Data entry is unchanged. The user is required to modify the two server
subroutines that serve to supply Nastran with the information required to evaluate the DRESP3
responses. These two subroutines are R3SGRT and R3SVALD and have the same names as has been
used in previous releases of this capability. They now have additional inputs and outputs as shown here
by examples.
The R3SGRT now not only checks that the DRESP3 Bulk Data entry is supported by the server, but also
identifies the number of responses that are produced from the server and whether analytic or finite
Main Index
47
CHAPTER 6
Optimization
difference gradient techniques will be used during the sensitivity and optimization evaluations.
Listing 6-1 shows an R3SGRT subroutine that utilizes the new features in solving the DRESP3 example
contained in External Response to Include Alternative Buckling Response (p. 504) in the MSC Nastran
Design Sensitivity and Optimization Users Guide.
Listing 6-1 R3SGRT Subroutine
SUBROUTINE R3SGRT(GRPID,TYPNAM,NRESP, GRDTYP, ERROR)
C ----------------------------------------------------------------------
C
C PURPOSE: VERIFY THE EXTERNAL RESPONSE TYPE
C
C GRPID INPUT INTEGER - GROUP ID
C TYPNAM INPUT CHARACTER*8 - NAME OF EXTERNAL RESPONSE TYPE
C NRESP OUTPT INTEGER - NUMBER OF RESPONSES FOR THIS DRESP3
C GRDTYP OUTPT INTEGER - INTEGER ARRAY OF LENGTH NRESP
C INDICATING HOW GRADIENT ARE TO BE
C COMPUTED
C = 2 THE USER WILL SUPPLY ANALYTIC
C GRADIENTS
C = -2 FINITE DIFFERENCE TECHNIQUES ARE USED
C ERROR INPUT/OUTPUT INTEGER -ERROR CODE FOR THE CALL.
C
C METHOD
C MATCH THE USER INPUT: TYPNAM WITH THE LIST OF AVAILABLE
C EXTERNAL RESPONSE TYPES. IF NO MATCH IS FOUND, SET ERROR CODE.
C SPECIFY THE NUMBER OF RESPONSES AND THE GRADIENT TECHNIQUE TO
C BE USED FOR EACH
C
C CALLED BY
C R3CGRT
C
C NOTE:
C THE WRITER OF THIS ROUTINE IS RESPONSIBLE TO SPECIFY
C NTYPES AND R3TYPE.
C ----------------------------------------------------------------------
C
C VARIABLES PASSED IN
C
INTEGER GRPID, ERROR, NRESP
INTEGER GRDTYP(*)
CHARACTER*8 TYPNAM
C
C LOCAL VARIABLES
C
INTEGER NTYPES, BADTYP
PARAMETER(NTYPES=6)
CHARACTER*8 R3TYPE(NTYPES)
C
DATA BADTYP/7554/
DATA R3TYPE/'USEVAR1 ','USEVAR10','USEALL',
. 'USEMIXVS','FREQMOD ','EULJOH '/

ERROR = 0
DO 100 ITYPE = 1, NTYPES
IF (TYPNAM .EQ. R3TYPE(ITYPE)) THEN
NRESP = 2
GRDTYP(1) = 2
Main Index
48
GRDTYP(2) = 2
GOTO 200
END IF
100 CONTINUE
ERROR = BADTYP
200 CONTINUE
RETURN
END
This is an update of Listing 7-33 in the Design Sensitivity and Optimization Users Guide and items in
bold are highlighted for the following discussion. There are two additional arguments for the subroutine:
NRESP indicates how many responses are to be calculated for this response type
GRDTYP indicates how gradients are to be supplied to an optimization or sensitivity analysis.
GRDTYP is a vector of length NRESP. Setting GRDTYP(iresp)=2 specifies that analytic gradients will
be provided while =-2 indicates that finite difference techniques will be required to compute gradient
information. In the Listing 6-1, the user has specified that there are two responses and that analytical
gradients will be supplied for each.
The corresponding R3SVALD subroutine is an update of Listing 7-34 in the MSC.Nastran Design
Sensitivity and Optimization Users Guide:
Listing 6-2 R3SVALD Subroutine
SUBROUTINE R3SVALD(GRPID,TYPNAM,
. NITEMS,ARGLIS,
. NSIZE, ARGVAL,
. NWRDA8,ARGCHR,
. FORG,NRESP,NARG,
. DR3VAL,SENVAL,
. ERROR)
C ----------------------------------------------------------------------
C
C PURPOSE: COMPUTE THE EXTERNAL RESPONSE
C
C GRPID INPUT INTEGER - GROUP ID
C TYPNAM INPUT CHARACTER*8 - NAME OF EXTERNAL RESPONSE TYPE
C NITEMS INPUT INTEGER - DIMENSION OF ARRAY ARGLIS
C NSIZE INPUT INTEGER - DIMENSION OF ARRAY ARGVAL
C NWRDA8 INPUT INTEGER - DIMENSION OF CHARACTER ARRAY ARGCHR
C ARGLIS INPUT INTEGER - ARRAY OF NO. OF ITEMS FOR EACH
C ARGUMENT TYPE
C ARGVAL INPUT DOUBLE - ARRAY OF ARGUMENT VALUES (EXCEPT
C CHARACTERS)
C ARGCHR INPUT CHARACTER*8 - ARRAY OF CHARACTER VALUES
C NRESP INPUT INTEGER - NUMBER OF RESPONSES
C FORG INPUT INTEGER - TYPE OF CALL
C = 0 FUNCTION EVALUATION
C = 1 SENSITIVITY EVALUATION
C NARG INPUT INTEGER - NUMBER OF ARGUMENTS NEEDING GRADIENTS
C DR3VAL OUTPUT DOUBLE - VALUE OF THE EXTERNAL RESPONSES
C SENVAL OUTPUT DOUBLE - MATRIX OF THE SENSITIVITY OF THE IRTH
C RESPONSE TO THE IARGTH ARGUMENT
C ERROR INPUT/OUTPUT INTEGER -ERROR CODE FOR THE CALL.
C
Main Index
49
CHAPTER 6
Optimization
C METHOD
C A)SET UP VARIOUS PARAMETERS FROM THE ARGUMENT LIST
C B)IF FORG = 0 EVALUATE THE EXTERNAL RESPONSE BASED ON THE
C GIVEN TYPNAM
C C)ELSE IF FORG = 1 EVALUATE THE SENSITIVITIES OF THE EXTERNAL
C RESPONSES TO THE ARGUMENTS THAT CAN VARY FOR
C THE GIVEN TYPNAM
C D)RETURN BADTYP ERROR IF TYPNAM IS NOT MATCHED HERE.
C
C NSIZE - THE NUMBER OF ARGUMENTS OR VALUES IN A DRESP3 ENTRY
C
C NSIZE=NV+NC+NR+NNC+NDVP1+NDVP2+NDVC1+NDVC2+NDVM1+NDVM2+NRR2
C WHERE:
C NV = NUMBER OF DESVARS NR = NUMBER OF DTABLES
C NR = NUMBER OF DRESP1S NNC = NUMBER OF DNODE PAIRS
C NDVP1 = NUMBER OF DVPREL1S NDVP2 = NUMBER DVPREL2S
C NDVC1 = NUMBER OF DVCREL1S NDVC2 = NUMBER DVCREL2S
C NDVM1 = NUMBER OF DVMREL1S NDVM2 = NUMBER DVMREL2S
C NRR2 = NUMBER OF DRESP2S
C NARG = NSIZE - NC
C
C CALLED BY
C VARIOUS
C ----------------------------------------------------------------------
C
C VARIABLES PASSED IN
C
CHARACTER*8 TYPNAM, ARGCHR(NWRDA8)
INTEGER FORG , NRESP
INTEGER GRPID, NITEMS, NSIZE, ARGLIS(NITEMS), ERROR, NWRDA8
DOUBLE PRECISION ARGVAL(NSIZE), DR3VAL(*), SENVAL(NRESP,*)
C
C
C LOCAL VARIABLES
C
INTEGER BADTYP, IDBG
DOUBLE PRECISION PI, FAC, FACT, SLNDER
DOUBLE PRECISION R,L,E,SIGMA,SIGMAC, RGYRA
C
DATA BADTYP /7554/, BADFG /7555/
C
PI = 3.14159
PI2 = PI * PI
C
C THE USER-SUPPLIED EQUATION TO DEFINE THE EXTERNAL RESPONSES
C SIGMA = DRESP1, R=DESVAR, L, E AND SIGMAC = DTABLE CONSTANTS
C
C EULER : EULER= -SIGMA * (L/ RGYRA ) **2 / (PI**2 * E)
C RGYRA = R / 2.0
C
C JOHNSON: JOHNSON = -SIGMA / (SIGMAC * FACTOR )
C FACTOR = 1. - SIGMAC * (L/RGYRA)**2 /(4 * PI**2 * E)
ERROR = 0
C
C SET UP PARAMETERS FOR VARIOUS ARGUMENT ITEMS
C
IF (TYPNAM .EQ. 'EULJOH ') THEN
C FUNCTION EVALUATION
R = ARGVAL(1)
L = ARGVAL(2)
Main Index
50
E = ARGVAL(3)
SIGMAC = ARGVAL(4)
SIGMA = ARGVAL(5)
RGYRA = R / 2.0
SLNDER = L / RGYRA
FACT = PI * SQRT(2.0D0 * E / SIGMAC)
FAC = 1.0D0 - SIGMAC * (SLNDER) ** 2 /(4.0D0 * PI2 * E )
IF ( FORG .EQ. 0 ) THEN
C FUNCTION EVALUATION
C JOHNSON CRITERION
DR3VAL(1) = -SIGMA / (SIGMAC * FAC)
C EULER CRITERION
DR3VAL(2) = -SIGMA * SLNDER**2 / (PI2 * E *10.0D0)
ELSE IF ( FORG .EQ. 1 ) THEN
C GRADIENT EVALUATION
DO 10 IRESP = 1, NRESP
DO 20 IARG = 1, NARG
SENVAL(IRESP,IARG) = 0.0D0
20 CONTINUE
10 CONTINUE
C NOTE THAT ARGVAL(2,3 AND 4) ARE CONSTANT AND THEREFORE HAVE
C ZERO SENSITIVITY
DSLDR = -SLNDER / R
DFACR = -SIGMAC * SLNDER * DSLDR / (2.0D0 *PI2 * E)
C SENSITIVITY OF THE FIRST RESPONSE TO THE FIRST ARGUMENT
SENVAL(1,1 ) = DFACR * SIGMA / ( SIGMAC * FAC ** 2)
C SENSITIVITY OF THE SECOND RESPONSE TO THE FIRST ARGUMENT
SENVAL(2,1) = -2.0D0*SIGMA * SLNDER * DSLDR /
1 (PI2 * E * 10.0D0)
C SENSITIVITY OF THE FIRST RESPONSE TO THE SECOND ARGUMENT
SENVAL(1,2) = - 1.0D0 / (SIGMAC * FAC)
C SENSITIVITY OF THE SECOND RESPONSE TO THE SECOND ARGUMENT
SENVAL(2,2) = -SLNDER**2 / (PI2 * E* 10.0D0)
DO 25 IDBG =1,2
25 CONTINUE
ELSE
ERROR = BADFG
ENDIF
ELSE
ERROR = BADTYP
END IF

RETURN
END
There are three new arguments and one modified argument in the calling statement:
FORG input integer - flag to indicate whether this call is to perform function evaluations or
gradient evaluations. 0-function, 1-gradient
NRESP input integer - indicates how many responses are to be calculated for this response
type
NARG input integer -number of arguments requiring gradients
DR3VAL output real vector of responses
SENVAL, - output real matrix of sensitivities
Main Index
51
CHAPTER 6
Optimization
DR3VAL is the modified argument in that it previously was a scalar and now is a vector. A comparison
with the listing in the Users Guide shows that now two responses are being returned (one for the Euler
criteria and one for the Johnson criteria) rather than a single argument which was the most critical of the
two criteria. NARG is used to supply the number of columns in the SENVAL matrix and it is important
to note that any constant terms (i.e, those input using DTABLE) are not included in the count of NARG
even though they are in the ARGVAL vector. SENVAL has NRESP rows and NARG columns.
DR3VAL is output when FORG=0 while SENVAL is output when FORG=1. Additional discussion of
these arguments is provided in Guidelines and Limitations, 51.
Output
.f06 output associated with the DRESP3 has been altered in one subtle respect: a response number field
has been added to the print as a count of which of the multiple responses is associated with the print. As
an example, the
It is seen that a single DRESP3 entry has generated 10 responses. These are 2 responses in each five
elements that have the buckling criteria imposed on them. It is up to the user to decipher that RESP NO.
1 is the Johnson buckling criterion while RESP NO. 2 is the Euler criterion.
Modifying Existing Server Subroutines
The enhanced capability does not require any changes in the input files that have been developed to
utilize the DRESP3, but it does require changes in the R3SGRT and R3SVALD server subroutines. To
retain the current capability for an existing DRESP3, the changes required in the R3SGRT subroutine are
to:
1. Add arguments NRESP and GRDTYP
2. Type NRESP and GRDTYP as integers.
3. Once the appropriate TYPNAM has been selected, add NRESP = 1 and GRDTYP(1) = -2
---- RETAINED DRESP3 RESPONSES ----
-----------------------------------------------------------------------------------------------------------------------
INTERNAL DRESP3 RESP RESPONSE GROUP TYPE LOWER UPPER
ID ID NO LABEL NAME NAME BOUND VALUE BOUND
------------------------------------------------------------------------------------------------------------------------
1 32 1 JOHNSON TESTGRP EULJOH N/A 1.4018E+00 1.0000E+00
Main Index
52
For the R3SVALD subroutine, the changes are to:
1. Add arguments FORG, NRESP, NARG, and SENVAL
2. Type FORG, NRESP and NARG as integers and DR3VAL(*) and SENVAL(NRESP,*) as
double precision
3. Replace the current DR3VAL = statements with DR3VAL(1) =.
4. Since FORG=1 is not supported, it is not necessary to specify any SENVAL output.
Other Guidelines
In R3SGRT, GRDTYP needs to be defined for all NRESP responses and the values must be either 2 or -
2. It is an error if any other value is used.
As mentioned previously, NARG is an input to R3SVALD and this value is determined from all the
DRESP3 arguments minus the constants and the string inputs. If a particular response is not a function
of one of the arguments, it is necessary to explicitly set the corresponding SENVAL output to zero. It is
a good practice to initialize the entire SENVAL array to 0.0.
It is important to realize that the gradients that are provided are for the responses with respect to the
DRESP3 arguments and not (necessarily) the design variables. This takes the burden of performing the
chain rule calculations from the user and uses existing Nastran operations to compute terms such as:
Instead, the R3SVALD subroutine provides the and terms and the remaining operations
are performed within Nastran.
Limitation
There is a current limitation that all GRDTYPs for a particular TYPE must be the same, either -2 or 2.
The GRDTYPs do not need to all be the same for all the DRESP3s in an input file. That is, one can
specify analytic gradients for one TYPE and finite difference gradients for another type.
Validation and Verification
Checking that the gradients are correct is an important and challenging process. Tips for facilitating this
include:
1. Setting DSAPRT(END=SENS) = n will stop the run after printing the sensitivities of the
responses in set n.
2. Setting DSAPRT(START=1) = n will provide sensitivities for the response in set n on the first
design cycle.
3. One can use two different versions of a DRESP3 to have the program check on itself. One would
use finite difference gradients while the second would use analytic gradients. The results should
agree except for numerical rounding due to the finite difference calculation.
dr
3
dx
JJJJJJJJ
r
3
x
JJJJJJJJ Z
r
3
r
i
JJJJJJJJ
r
i
x
JJJJJJJ H
r
3
x r
3
r
i
Main Index
53
CHAPTER 6
Optimization
Examples
Three test cases are discussed here. The first of these is ds13grad and is a variation of the dsoug13
example in External Response to Include Alternative Buckling Response (Ch. 7) in the MSC Nastran
Design Sensitivity and Optimization Users Guide. The R3SGRT and R3SVALD subroutines listed
above are used in this example.
The second example is dresp3aa is a test example to demonstrate all the types of arguments that can be
included in a DRESP3. The DRESP3 input in this example is:
Listing 6-3 ds13grad
$ F101 = X1
DRESP3 101 EXTERNR3TESTGRP USEVAR1
DESVAR 1
DTABLE CONST
DRESP1 808
DNODE 2 1
DVPREL1 1
DVCREL1 3
DVMREL1 5
DVPREL2 2
DVCREL2 4
DVMREL2 6
DRESP2 909
$ F102 = R2
DRESP3 102 EXTERNR3TESTGRP USEVAR10
DESVAR 1
DTABLE CONST
DRESP1 808
DNODE 2 1
DVPREL1 1
DVCREL1 3
DVMREL1 5
DVPREL2 2
DVCREL2 4
DVMREL2 6
DRESP2 909
$ F103 = F(X1,CONST,R1,G,DVP1,DVC1,DVM1,DVP2,DVC2,DVM2,R2)
$234567
DRESP3 103 EXTERNR3TESTGRP USEALL
DESVAR 1
DTABLE CONST
DRESP1 808
DNODE 2 1
DVPREL1 1
DVCREL1 3
DVMREL1 5
DVPREL2 2
DVCREL2 4
DVMREL2 6
DRESP2 909
$ F104 = F(X,g,P1,C1,M1,p2,R2)
DRESP3 104 EXTERNR3TESTGRP USEMIXVS
DESVAR 1
DTABLE CONST
DRESP1 808
DNODE 2 1
Main Index
54
DVPREL1 1
DVCREL1 3
DVMREL1 5
DVPREL2 2
DVCREL2 4
DVMREL2 6
DRESP2 909
USRDATA thisisa teststring
foraddingxxx
$ F105 = F(X,R1)
DRESP3 105 EXTERNR3TESTGRP FREQMOD
DESVAR 1
DRESP1 505
The relevant part of the R3SGRT subroutine that goes with this input file is:
Listing 6-4 dresp3aa
nresp = 1
IF (TYPNAM .EQ. 'FREQMOD' ) then
grdtyp(1) = 2
Else if (typnam .eq. 'USEVAR1' ) then
grdtyp(1) = -2
ELSE IF (TYPNAM .EQ. 'USEVAR10') THEN
grdtyp(1) = 2
ELSE IF (TYPNAM .EQ. 'USEALL') THEN
grdtyp(1) = -2
ELSE IF (TYPNAM .EQ. 'USEMIXVS') THEN
grdtyp(1) = 2
else
ERROR = BADTYP
endif
There is a single response for each DRESP3 and analytic gradients are to be provided for
TYPNAM=FREQMOD,USEVAR10 and USEMIXVS
The relevant part of the R3SVALD subroutine is:
Listing 6-5 dresp3sig
if ( forg .eq. 1 ) then
c gradient evaluation
do 10 iresp = 1, nresp
do 20 iarg = 1, narg
senval(iresp,iarg) = 0.0d0
20 continue
10 continue
Endif

IF (TYPNAM .NE. 'FREQMOD') THEN
x = argval(1)
const = argval(2)
r1 = argval(3)
g = argval(4)
p1 = argval(5)
c1 = argval(6)
m1 = argval(7)
p2 = argval(8)
c2 = argval(9)
m2 = argval(10)
Main Index
55
CHAPTER 6
Optimization
r2 = argval(11)
END IF
IF (TYPNAM .EQ. 'USEVAR1') THEN
dr3val(1) = x+r1+r2
ELSE IF (TYPNAM .EQ. 'USEVAR10') THEN
dr3val(1) = r2
else
senval(1,9) = 1.0d0
endif
ELSE IF (TYPNAM .EQ. 'USEALL') THEN
dr3val(1) = x+const+r1+g+p1+c1+m1+p2+c2+m2+r2
ELSE IF (TYPNAM .EQ. 'USEMIXVS') THEN
dr3val(1) = x+g+p1+c1+m1+p2+r2
else
senval(1,1) = 1.0d0
senval(1,3) = 1.0d0
senval(1,4) = 1.0d0
senval(1,5) = 1.0d0
senval(1,6) = 1.0d0
senval(1,9) = 1.0d0
endif
ELSE IF (TYPNAM .EQ. 'FREQMOD') THEN
x = argval(1)
r1 = argval(2)
dr3val(1) = x*r1
else
senval(1,1) = r1
senval(1,2) = x
endif
ELSE
ERROR = BADTYP
END IF
It is expedient to zero out all the gradient values, whether they are needed or not. For the TYPNAMs
that dont support analytic gradients, it is only necessary to provide the response value in DR3VAL(1).
For the TYPNAMs that do support analytic gradients, an if test on FORG is provided. For FORG=0, a
function evaluation is made while a gradient evaluation is made for FORG=1. Note that for TYPNAM=
USEVAR10, the input file shows 11 inputs, 1 for each of the available Flags while the actual response
shown in R3SVALD only uses the DRESP2 argument, the eleventh ARGVAL. Furthermore, the gradient
calculation has a single non-zero result: senval(1,9) = 1.0d0, indicating that the sensitivity of the first
response to the ninth argument that can vary is 1.0. The constant term in the ARGVAL list and the
undesigned DNODE do not count as one of the NARG sensitivity arguments, hence the discrepancy
between eleven and nine.
A final example is entitled dresp3sig and demonstrates the feature that reorders the sensitivity
calculations when NRESP1>>NDVI. In this case, there are 181 DRESP1s and 10 DESVARs so the
criteria is satisfied. The job has two DRESP3s that have the same arguments but one has
TYPNAM=RSS and the other has TYPNAM=RSSA. The RSS response has its gradients calculated
using finite difference techniques while the RSSA uses analytical gradients. Since these are the same
response, the test case serves to demonstrate that the same sensitivity information is generated using
analytic or finite difference gradient techniques. The problem is too small to make any assessment of
the performance gains that have resulted from the third enhancement mentioned in the Introduction, 46.
Main Index
56
Introduction
Topometry optimization is an element-by-element sizing optimization. Unlike conventional sizing
optimization where all elements referencing a property entry are grouped as one design variable, each
designable element has an independent design variable in topometry optimization. Since element-by-
element optimization has many design variables, it may find a better design than conventional sizing
optimization. In previous versions of Nastran, the user can use the design variable Bulk Data entry
DESVAR and the relation of model property and design variables Bulk Data entry DVxREL1 to support
element-by-element sizing optimization. However, with this approach the user must generate a unique
property data entry for each element and perhaps prepare thousands of DESVAR and DVPREL1 entries.
With the topometry optimization capability released in MSC Nastran 2008 r1, the user can utilize a new
Bulk Data entry, TOMVAR, to select designable regions (model property or material property
identification number), design parameters (such as thickness of PSHELLs, or Youngs Modulus of
materials), input initial values, lower and upper bounds to perform element-by-element sizing
optimization. The MSC Nastran program internally generates DESVAR and DVPREL1 (and/or
DVMREL1) for each designed element. The implementation provides a very simple user interface to do
element-by-element sizing design optimization. In addition, topometry optimization supports the fully
stressed design algorithm in MSC Nastran. FSD is very efficient for certain problems with many stress
constraints.
Topometry optimization released in MSC Nastran 2008 r1 can be applied to all elements that can be
resized through Bulk Data entries DVPREL1 and DVMREL1. Those element types include not only
volume-based elements like CQUAD4 but also non-volume elements like CWELD, CBUSH, and
CFAST.
Topology optimization is another element-by-element optimization technology. However, topology
optimization and topometry optimization are fundamentally different. Topology optimization is a 0 or
1 discrete element-by-element optimization methodology. Topology optimization can be used to
decide which element should be retained and which element should be discarded from the design space.
One the other hand, topometry optimization aims to get a continuous variation of the designed properties.
Although topometry optimization is not recommended for topology optimization tasks, it is observed
topometry optimization can be used to get similar topological results for some cases. It is particularly
useful for non-structural elements like CELAS, CFAST, and CBUSH that MSC Nastran topology
optimization does not support.
In a single optimization problem, it is allowable to resize (or shape, topology) certain properties while
topometry optimizing other properties.
Benefits
Topometry optimization is easy-to-use. One TOMVAR Bulk Data entry replaces many
thousands of DESVAR and DVxREL1 entries for large element-by-element design optimization
problems.
Main Index
57
CHAPTER 6
Optimization
Topometry optimization is good to identify critical design regions.
Topometry optimization is good to locate where to add/or remove material to improve structural
performance.
Topometry optimization is good for finding the optimal location of spot welds. In particular,
topometry optimization is very useful for some properties that MSC Nastran topology does not
support; for example, PDAMP, PELAS, PMASS, PBUSH, PVISC, PGAP, PACBAR, and
PFAST.
Input
The TOMVAR Bulk Data entry is used to select a topometry designable region and designed property
name. The initial, lower, and upper bound of the designed property value are also specified on the
topometry entry. The program automatically generates one design variable for each element
referencing a property PID. The relationship between design variables and the element property
given by
where is the analysis model property value for the ith element. NE is the total number of elements
referencing to the property PID. The user must input an initial value (such as the analysis model input
property value). The default of lower bound (XLB) on is , and default of upper bound on
(XUB) is .
The topometry Bulk Data entry is:
Format:
Example:
Design all element's thickness referencing PSHELL ID = 5 with initial design = 10.0 ( input
element thickness), lower bound and upper bound .

Example:
Design all element's Young Modulus referred by PSHELL ID = 100 with initial design XINIT = 3.E+5,
XLB=1.0, and XUB= 1.0E+6.
N 2 3 4 5 6 7 8 9 10
TOMVAR
ID TYPE PID PNAME
/FID
XINIT XLB XUB DELXV
TOMVAR
10 PSHELL 5 T 10.0
DV
i
DV
i
P
i
P
i
DVi Z i 1 NE , Z
XLB DV
i
XUB
P
i
DV
i
0.5 DV
i
DV
i
1.5 DV
i
t
0
10.0 Z
0.5 t
0
1.5 t
0
Main Index
58
Remarks:
1. Multiple TOMVARs are allowed in a single file.
2. Property name and FID > 0 can be used for element property values just like a Bulk Data entry
DVPREL1. Only property name can be used for material property values like DVMREL1. If a
property name is shared by both property and material (such as A for PROD and MAT1), this
name is taken as a material name. The user must provide a FID for property name (FID=4 for
PROD). PCOMP, PCOMPG, PBEAML, PBARL, PBMSECT, PBRSECT are not supported. If
material property name is selected, PSHELL (with multiple MID inputs) must reference a unique
material ID.
3. Combined topometry, topography, topology, sizing, and shape optimization is supported in a
single file. However, topometry and topology cannot reference the same property ID. It is possible
to topometry certain elements while sizing others. It is allowed to simultaneously design the same
elements with topometry and desvar (sizing and/or shape) variables but topometry and sizing
cannot reference the same property name.
4. The design response DRESP1=FRMASS (fractional mass) can be used for topometry
optimization. The initial FRMASS is defined as1.0 at the initial design specified on a TOPVAR
entry. For non-volume elements like CELAS, a artificial mass = 1.0 is assumed for each element.
TOMVAR
10 PSHELL 100 E 3.E+5 1.0 1.E+6
Field Contents
ID Unique topometry design region identification number. (Integer > 0)
TYPE Property entry type. Used with PID to identify the elements to be designed.
(Character: PBAR, PSHELL, PSOLID, etc. see Remark 2.)
PID Property entry identifier (Integer > 0). This PID must be unique for PIDs referenced
by other TOPVAR, DVPREL1, DVPREL2, DVMREL1, and DVMREL2 entries.
(Integer > 0). See Remark 2.
PNAME/FID Property name or property material name, such as T, A, E, and GE, or field
position of the property entry or word position in the element property table of the
analysis model. Property names that begin with an integer such as 12I/T**3 may
only be referenced by field position. (Character or Integer > 0. see Remark 2.)
XINIT Initial value. (Real or blank, no default). Typically, XINIT is defined to match the
mass target constraint (so the initial design does not have violated constraints) or the
analysis model input property value.
XLB Lower bound. (Real or blank; Default = blank). The default is XLB=0.5*XINIT.
XUB Upper bound. (Real or blank; Default = blank). The default is XLB=1.5*XINIT.
DELXV Fractional change allowed for the design variable during approximate optimization.
(Real > 0.0; Default = 0.5. See Remark 3.).
Main Index
59
CHAPTER 6
Optimization
Output
A regular SOL 200 summary table is produced. In addition, a Patran element result file jobname.des
contains the optimal design values for each element. This Patran element result file can be imported to
Patran a third party post-processor to display topometry optimization results. Two parameters DESCPH
and DESPCH1 are used to specify in SOL 200 when the optimized topometry results are written to the
jobname.des.
BIGDOT is the default optimizer of topometry optimization since topometry optimization
usually involves many thousands of design variables. BIGDOT requires a Topology
Optimization license. For SOL 200 design optimization clients without access to topology
optimization, optimizer MSCADS, method=4 (SUMT) is recommended through the
optimization control Bulk Data entry DOPTPRM.
Since SOL 200 adjoint design sensitivity analysis method does not support element responses
(such as stress), a direct design sensitivity analysis method is automatically selected for
problems with element response constraints. In this case, topometry optimization with element
response constraints are slow due to many design variables. Fully stressed design (FSD) can be
used for certain problems.
Topology optimization can be used for analysis model properties PDAMP, PELAS, PMASS,
PBUSH, PVISC, PGAP, NSM, NSM1, PACBAR and PFAST. Topology optimization is limited
to analysis properties that can reference material property MAT1.
P2 > 13 on DOPTPRM prints design variables in *.f06.
Example 1 - Three-bar Truss (tomex1.dat)
A simple sizing optimization example three-bar truss (a TPL file DSOUG1.dat) is used here to
demonstrate topometry optimization solved by the fully stressed design algorithm. Figure 6-1 shows the
three-bar truss that must be built to withstand two separate loading conditions. The objective is to
minimize structural weight and subjected to displacement and stress constraints. The sizing design
variables are the cross-sectional areas. The detailed descriptions of analysis model and design
DESPCH DESPCH specifies when the topometry optimized design values are written to the
element result history file jobname.des. The Default = 0 writes the last design cycle
only. DESPCH < 0 never. DEPSCH1 > 0 at every design cycle that is a multiple of
DESPCH and the last design cycle.
DESPCH1 DESPCH1 > 0, write all topometry designed and non-designed element values to the
element result history file jobname.des. 1.0 is assigned to the non-designed element
value. DESPCH1 < 0, write all topometry designed element values to the element
result history file jobname.des.
Main Index
60
optimization model can be seen in Chapter 7 of the MSC Nastran Design Sensitivity and Optimization
User's Guide.
Figure 6-1 Three Bar Truss
The goal of this example is to show an alternate method of setting design variables by a TOMVAR entry.
The objective and constraints are not changed. In conventional sizing optimization, the set of DESVAR
and DVPREL1 entries define the relations Ai=1.0Xi (i=1, 2, 3) where A is the rod element cross-
sectional area and X is the design variable. In DSOUG1.dat, we have:
$...DESIGN VARIABLE DEFINITION
$DESVAR ID LABEL XINIT XLB XUB DELXV(OPTIONAL)
DESVAR 1 A1 1.0 0.1 100.0
DESVAR 2 A2 2.0 0.1 100.0
DESVAR 3 A3 1.0 0.1 100.0
$
$...DEFINITION OF DESIGN VARIABLE TO ANALYSIS MODEL PARAMETER $RELATIONS
$DVPREL1 ID TYPE PID NAME PMIN PMAX C0 +
$+ DVID1 COEF1 DVID2 COEF2 ...
DVPREL1 10 PROD 11 A
1 1.0
2 1.0
3 1.0
In DSOUG1.dat, rod elements 11 and 12 have different property groups. Then, the DLINK entry is used
to explicitly link the design variables 1 and 3 together. In this example, we try to do element-by-element
optimization. Thus, we take three design variables (rod element cross-sectional areas) as independent
variables. The rod elements 1 and 3 have the same property group (PROD=1). TOMVAR entry 1
Main Index
61
CHAPTER 6
Optimization
(Listing 6-6) is used to define two independent design variables with an initial value = 1.0 (and element
cross-sectional area = 1.0) for rod element 11 and 13 respectively. This is equivalent to four entries in
DSOUG1.dat:
DESVAR 1 A1 1.0 0.1 100.0
DESVAR 3 A3 2.0 0.1 100.0
1 1.0
3 1.0
TOMVAR entry 2 (Listing 6-6) is used to define one independent design variable with an initial value =
2.0 (and element cross-sectional area = 2.0) for rod element 12. This is equivalent to two entries in
DSOUG1.dat:
DESVAR 2 A2 2.0 0.1 100.0
2 1.0
Input
The input data for this example is given in Listing 6-6.
Listing 6-6 Input File for Example 1
ID MSC TOMEX1 $
TIME 10 $
SOL 200 $ OPTIMIZATION
CEND
TITLE = THREE BAR TRUSS TOPOMETRY OPTIMIZATION
SUBTITLE = 3 CROSS SECTIONAL AREAS AS DESIGN VARIABLES
ECHO = SORT
SPC = 100
DISP = ALL
STRESS = ALL
DESOBJ(MIN) = 20 $ (DESIGN OBJECTIVE = DRESP ID)
DESSUB = 21 $ DEFINE CONSTRAINT SET FOR BOTH SUBCASES
ANALYSIS = STATICS
SUBCASE 1
LABEL = LOAD CONDITION 1
LOAD = 300
SUBCASE 2
LABEL = LOAD CONDITION 2
LOAD = 310
BEGIN BULK
$
$------------------------------------------------------------------------
$ ANALYSIS MODEL
$------------------------------------------------------------------------
$
$ GRID DATA
$ 2 3 4 5 6 7 8 9 10
GRID 1 -10.0 0.0 0.0
GRID 2 0.0 0.0 0.0
GRID 3 10.0 0.0 0.0
GRID 4 0.0 -10.0 0.0
$ SUPPORT DATA
Main Index
62
SPC1 100 123456 1THRU3
$ ELEMENT DATA
CROD 1 11 1 4
CROD 2 12 2 4
CROD 3 11 3 4
$ PROPERTY DATA
PROD 11 1 1.0
PROD 12 1 2.0
MAT1 1 1.0E+7 0.33 0.1
$ EXTERNAL LOADS DATA
FORCE 300 4 20000. 0.8 -0.6
FORCE 310 4 20000. -0.8 -0.6
$
$------------------------------------------------------------------------
$ DESIGN MODEL
$------------------------------------------------------------------------
$
$...DESIGN TOPOMETRY DESIGN DEFINITION
$TOMVAR, ID, PRYPE, PID, PNAME, XINIT, XLB, XUB, DELXV(OPTIONAL)
TOMVAR, 1 , PROD, 11, 4 , 1., .1 , 100.0
TOMVAR, 2 , PROD, 12, 4 , 2., .1 , 100.0
$
$...STRUCTURAL RESPONSE IDENTIFICATION
$DRESP1 ID LABEL RTYPE PTYPE REGION ATTA ATTB ATT1 +
$+ ATT2 ...
DRESP1 20 W WEIGHT
DRESP1 21 U4 DISP 12 4
DRESP1 23 S1 STRESS PROD 2 11
12
$...CONSTRAINTS
$DCONSTR DCID RID LALLOW UALLOW
DCONSTR 21 21 -0.20 0.20
DCONSTR 21 23 -15000. 20000.
$
$...OPTIMIZATION CONTROL (FULLY STRESSED DESIGN):
$
DOPTPRM FSDMAX 20 DESMAX 0 P1 1 P2 15
$
$.......2.......3.......4.......5.......6.......7.......8.......9.......0
ENDDATA
Main Index
63
CHAPTER 6
Optimization
Output
A regular SOL 200 output can be found as:
Example 2 Car Model Topometry Design
A real complex example car body is used here to demonstrate topometry optimization for graphical post-
processing. This example also shows that SOL 200 is able to deal with very large optimization problems.
The objective is to minimize structural compliance and keep weight unchanged. SOL 200 produces an
element thickness distribution file *.des that can be used by Patran or other post-processors to view
topometry optimization results.
***************************************************************
S U M M A R Y O F D E S I G N C Y C L E H I S T O R Y
***************************************************************
(HARD CONVERGENCE ACHIEVED)
NUMBER OF FINITE ELEMENT ANALYSES COMPLETED 17
NUMBER OF FULLY STRESSED DESIGN CYCLES COMPLETED 16
NUMBER OF OPTIMIZATIONS W.R.T. APPROXIMATE MODELS 0
OBJECTIVE AND MAXIMUM CONSTRAINT HISTORY
--------------------------------------------------------------------------------------------------------------------------------
OBJECTIVE FROM OBJECTIVE FROM FRACTIONAL ERROR MAXIMUM VALUE
CYCLE APPROXIMATE EXACT OF
OF
NUMBER OPTIMIZATION ANALYSIS APPROXIMATION CONSTRAINT
-------------------------------------------------------------------------------------------------------------------------------
INITIAL 4.828427E+00 -
3.234952E-01
1 FSD 3.862742E+00 N/A
-1.543690E-01

2 FSD 3.225798E+00 N/A
-7.883203E-03
.
16 FSD 2.741757E+00 N/A
1.664062E-04
DESIGN VARIABLE HISTORY
-------------------------------------------------------------------------------------------------------------------------------
---------------------------------------
INTERNAL | EXTERNAL | |
DV. ID. | ELEMENT ID | LABEL | INITIAL : 1 : 2 : 3 :
4 : 5
: -------------------------------------------------------------------------------------------------------------------------
-----------------------------------------
1 | 1 | TOMVAR | 1.0000E+00 : 8.0000E-01 : 6.8794E-01 : 6.8306E-01 : 6.9978E-
01 : 7.2284E-01 :
2 | 2 | TOMVAR | 2.0000E+00 : 1.6000E+00 : 1.2800E+00 : 1.0240E+00 : 8.1920E-
01 : 6.5536E-01 :
3 | 3 | TOMVAR | 1.0000E+00 : 8.0000E-01 : 6.8794E-01 : 6.8306E-01 : 6.9978E-
01 : 7.2284E-01 :
-------------------------------------------------------------------------------------------------------------------------------
Main Index
64
Figure 6-2 Optimal Thickness Distribution of Car Model - Note that this figure is meaningful
only when viewed in color.
Main Index
65
CHAPTER 6
Optimization
Introduction
Topography optimization (also called bead or stamp optimization) is used to generate a design proposal
for reinforcement bead patterns. In MSC Nastran 2008 r1, topography optimization is treated as a special
shape optimization and built on SOL 200 shape optimization technology. In topography optimization,
finite element grids are moved in as normal vectors to the shell surface or the user's given direction. New
algorithms were developed to generate shape design variables and shape basis vectors automatically
based on the user's provided bead dimension (minimum bead width, maximum bead height, and draw
angle). Since many design variables are generated in the topography optimization, the adjoint design
sensitivity analysis method and large scale optimizer play key roles in solving topography optimization
problems.
Benefits
Topography optimization is particularly powerful for designing sheet metal parts.
Topography optimization can be used for all SOL 200 analysis types such as statics, normal
modes, buckling, complex eigenvalue, dynamic frequency response, and aeroelastic analyses.
Input
The BEADVAR Bulk Data entry is used to define topography design regions.
N 2 3 4 5 6 7 8 9 10
BEADVAR
ID PTYPE PID MW MH ANG BF SKIP
DESVAR NORM/XD
YD ZD CID XLB XUB DELXV
GRID NGSET DGSET
Field Contents
ID Unique topography design region identification number. (Integer > 0)
PTYPE Property entry type. Used with PID to identify the element nodes to be
designed. (Character: PSHELL, PSHEAR, PCOMP, or
PCOMPG.)
PID Property entry identifier. See Remark 1. (Integer > 0)
MW Minimum bead width. This parameter controls the width of the beads. The
recommended value is between 1.5 and 2.5 times the average element
width. See Remark 2. (Real > 0.0)
MH Maximum bead height (Real > 0.0). This parameter sets the maximum
height of the beads when XUB=1.0 (as Default). See Remark 2.
Main Index
66
ANG Draw angle in degrees (0.0 < Real < 90.0). This parameter controls the
angle of the sides of the beads. The recommended value is between 60 and
75 degrees.
BF Buffer zone ('yes' or 'no'; Default='yes'). This parameter creates a buffer
zone between elements in the topography design region and elements
outside the design region when BF='yes'. See Remark 3.
SKIP Boundary skip (bc, load, both, or none; Default = both). This
parameter indicates which element nodes are excluded from the design
region. bc indicates all nodes referenced by SPC and SPC1 are
omitted from the design region. "load" indicates all nodes referenced by
FORCE, FORCE1, FORCE2, MOMENT, MOMENT1,
MOMENT2, and SPCD are omitted from the design region. both
indicates nodes with either bc or load are omitted from the design
region. none indicates all nodes associated with elements referencing
PID specified in field 4 are in the design region.
DESVAR Indicates that this line defines bead design variables that are automatically
generated.
NORM/XD, YD, ZD Bead vector (draw direction). Norm indicates the shape variables are
created in the normal directions to the elements. If XD, YD, and ZD are
provided, the shape variables are created in the direction specified by the
xyz vector defied by XD/YD/ZD that is given in the basic coordinate
system or CID. See Remark 4. (Character or Real, Default = blank = norm).
CID Coordinate system ID used for specifying draw direction (Blank or Integer
> 0; Default = blank = basic coordinate system)
XLB Lower bound. (Real < XUB or blank; Default = blank = 0.0). This ensures
the lower bound on grid movement equal to XLB*MH. See Remark 5.
XUB Upper bound. (Real > XLB or blank; Default = 1.0). This sets the upper
bound of the beads equal to XUB*MH. See Remark 5.
DELXV Fractional change allowed for the design variable during approximate
optimization. See Remark 3. (Real > 0.0; Default = 0.2)
GRID Indicates this line defines what element nodes can be added and/or removed
from topography design regions.
NGSET All grids listed on Bulk Data entry SET1 = NGSET are removed from
topography design regions.
DGSET All grids listed on Bulk Data entry SET1 = DGSET are added to topography
design regions.
Field Contents
Main Index
67
CHAPTER 6
Optimization
Remarks:
1. Multiple BEADVARs are allowed in a single file. Combined topometry, topology, topography,
sizing, and shape optimization is supported in a single file.
2. The user can provide allowable bead dimensions.
Bead Dimensions
3. It is recommended to set buffer zone = yes to maintain a good quality of mesh during topography
optimization.
4. The grids moves in the normal direction. All element grids referenced by one BEADVAR entry
must follow the right hand rule.
MW
MH
ANG
Design elements
Buffer zone
Nondesign elements
No buffer zone
Nondesign elements
Element Normal
Element normal vectors
Baseline surface
Optimized
surface
Main Index
68
5. To force the grids to move only in the positive bead vector direction (one side of the surface), use
XLB = 0.0. To force the grids to move only in the negative bead vector direction (another side of
the surface), use XUB = 0.0. To allow girds to move in both positive and negative bead vector
directions, use XLB < 0.0 and XUB > 0.0. For example,
6. The jobname.op2 has topography results (shape change) that can be viewed in Patran. The text
file jobname.pch also has updated grid coordinates that can be copied to replace the grids in the
original file, and imported to Patran on other post-processors to view topography optimization
results.
Outputs
A regular SOL 200 design history summary table is produced. The jobname.op2 (with PARAM,POST,-
1) and jobname.pch can be imported to Patran and other post-processors to view topography optimization
results.
User defined draw vector
Baseline surface
Optimized
surface
Users Provided Draw Direction
Bead Vector
Bead Vector
Optimized Surface
Base Surface
(a) XLB = 0.0 and XUB = 1.0 (b) XLB = -1.0 and XUB = 0.0 (c) XLB = -1.0 and XUB = 1.0
Optimized Surface
Main Index
69
CHAPTER 6
Optimization
BIGDOT is the default optimizer of topography optimization since topography optimization
usually involves many design variables. BIGDOT requires a Topology Optimization license. For
SOL 200 design optimization clients without access to topology optimization, the optimizer
MSCADS method=4 (SUMT) is recommended through the optimization Bulk Data entry
DOPTPRM.
Since SOL 200 adjoint design sensitivity analysis method does not support element responses
(such as stress), a direct design sensitivity analysis method is automatically selected for
problems with element response constraints. In this case, topography optimization with element
response constraints are slow.
Since adjoint design sensitivity analysis does not support rigid body elements (RBE1, RBE2,
RBE3, RROD, RBAR, RTRPLT, RSPLINE), all grids connected to rigid body elements must be
fixed in topography optimization for static and dynamic frequency response analyses.
The minimum bead width and maximum bead height have significant effects on optimal
designs. A smaller minimum bead width results in more small beads.
Mesh distortion is a challenge for topography optimization. It is recommended that a relatively
coarse mesh be used for highly curved areas.
P2 > 13 on DOPTPRM prints design variables in *.f06
Example 3 A Square (togex1.dat)
A square model shown in Figure 6-3 is used to demonstrate MSC Nastran 2008 r1 topography
optimization capabilities. The square is modeled with quadrilateral plate elements (CQUAD4) and is
fixed at all four edges. The objective is to maximize the first frequency of the structure with a given bead
dimension (minimum bead width = 10.0, maximum bead height = 20.0, draw angle = 70.0).
Figure 6-3 A Square
Main Index
70
Input
The input data for this example is given in Listing 6-7. The Bulk Data entry 1 defines the topography
designable region. It is noticed that element normals are used for bead vectors (draw direction) and all
grids associated with the boundary condition are fixed during optimization. PARAM, POST, -1 outputs
results for Patran.
Listing 6-7 Input File for Example 2
$Topography opt example one
SOL 200
CEND
TITLE = MSC Nastran job created on 28-Nov-07
ECHO = NONE
$ Direct Text Input for Global Case Control Data
DESOBJ(MAX) = 1
SUBCASE 1
$ Subcase name : Default
SUBTITLE=Default
SPC = 2
METHOD = 1
DISPLACEMENT(SORT1,REAL,PLOT)=ALL
ANALYSIS=MODES
BEGIN BULK
EIGRL,1,,,20
$ Direct Text Input for Bulk Data
$ Elements and Element Properties for region : ps1
$
$ BEADVAR, ID, TYPE, PID, MW, MH, ANG, BF, SKIP.
$
BEADVAR, 1 , PSHELL, 1, 10., 20.0, 70.0, YES, BOTH
DRESP1, 1, MODES, FREQ,,,1
PARAM POST -1
Output
Figure 6-4 shows the topography optimized result by using Patran. The first frequency has increased from
0.568HZ at the initial design to 4.78 HZ.
Figure 6-4 Topography Optimal Design of A Square
Main Index
71
CHAPTER 6
Optimization
Introduction
The stochastic capability in MSC Nastran is the first step toward a complete and automatic self-
randomization of a finite element model. The current capability offers the possibility to automatically
distribute tolerances and uncertainties with minimum effort. This dramatically reduces the complexity
of large-scale stochastic simulations. In fact, once the stochastic option is triggered, the entire Bulk Data
file is automatically randomized without further user intervention. The resulting model, which needs to
be incorporated in a Monte Carlo Simulation loop (there are numerous off-the-shelf products which
support this capability) possesses unprecedented levels of realism.
In order to make full use of this new capability, it is necessary to use a multi-run environment which can
spawn a certain number of independent MSC Nastran executions, collect the results, and perform
statistical postprocessing. With the self-randomization capability in MSC Nastran, the user need only
define the outputs to be monitored, such as stresses, Eigenfrequencies, temperatures, displacements, etc.
There is no need to define inputs, as these are defined automatically by MSC Nastran. The
Randomization of an Input Data File functionality was a pre-release capability in MSC Nastran 2007 r1.
For MSC Nastran 2008 r1 this is now a production capability.
Benefits
It is sometimes assumed that the inputs to an MSC Nastran analysis are known exactly, and thus the
computed responses are exact. This is an invalid assumption since there will always be some uncertainty
in the input values with a corresponding variation in the results. MSC Nastran 2007 r1 provides a way
of introducing this uncertainty into the analysis process by automatically randomizing user input real
numbers based on the input values and statistical quantities that characterize the variation.
Input
The randomization capability is driven by a new STOCHASTICS Case Control command, as described
in the MSC Nastran Quick Reference Guide. If STOCHASTICS=ALL is used, all real quantities on
connectivity (those starting with C), material, and property Bulk Data entries, as well as any loads and
SPCD quantities, are modified based on a covariance factor of 0.05. A Gaussian distribution is used to
randomly select the perturbed quantity with the restriction than the value can be no more that a specified
number of standard deviations from the user input mean value. The default number of maximum
standard deviations is three.
Alternatively, the STOCHASTICS Case Control command can point to a STOCHAS Bulk Data entry
that provides the ability to selectively randomize different types of input quantities by means of user-
specified covariance values and user-prescribed numbers of allowed standard deviations. In this case,
only the types of input specified are randomized so that, for example, it is possible to randomize the load
inputs while leaving the property values unchanged.
Main Index
72
Output
There is no new output produced by this capability.
The randomization algorithm involves using a random number generator, a Gaussian distribution, a
prescribed covariance, and a mean value based on user input to determine a randomized value that is to
be used in the analysis. In order to avoid physically meaningless properties, the random value is
prescribed to be within m standard deviations of the input value, where m is a user input value with a
default value of 3.0.
The product of m * COV should not be greater than 1.0 to eliminate the possibility of the property
changing sign.
Any real value in the Bulk Data file will be randomized unless otherwise specified by the user. To keep
a particular field or fields from being randomized, the user must set them equal to a value of 0.0.
Main Index
73
CHAPTER 6
Optimization
Random Elimination of Element Types
Introduction
There has been a long-standing capability in MSC Nastran that allows the user to specify the random
elimination of a specified percentage of the CWELD elements contained in a bulk data file. This was
done using the PARAM CWRANDEL entry, with an additional CWDIAGP PARAM providing the
option of printing the IDs of the deleted elements. In the current release, this capability has been extended
to the CELASi, CFAST, CSEAM, and 1-D mass (CMASSi, CONM1, and CONM2) elements. In
addition, the user interface has been changed from the NASTRAN statement to the MDLPRM entry. The
Random Elimination of Element Types functionality was a pre-release capability in the MSC Nastran
2007 r1 release. For MSC Nastran 2008 r1 this is now a production capability.
Benefits
The ability to randomly delete various 1-D elements provides the user with some assessment of the
integrity of the design being modeled. For instance, if randomly deleting 20% of, say, of the CWELD
elements from a model caused a negligible change in the first ten natural frequencies, this was taken as
an indication of the robustness of the structure. Extending this approach to other element types provides
more options in this type of analysis. Placing the input on the MDLPRM entry consolidates that input so
that the user does not have to deal with the PARAM entry.
Input
The MDLPRM entry has ten new PARAMi names that support this capability. Five of these names (e.g,
DELELAS) select the element type to which the random elimination applies and the ratio to be deleted,
while an additional five names (e.g., PRTELAS) provide control as to whether the IDs of the deleted
elements are to be printed. The default is that the IDs will not be printed.
Output
There is no new output produced by this capability.
The deletion ratio is input as a real number between 0.0 and 1.0, with 0.0 indicating that no deletion is
to take place, while 1.0 eliminates all elements of the specified type.
It is possible that the elimination of a series of elements will introduce mechanisms in the structure that
will cause the analysis to fail. It is the users responsibility to determine whether this failure has occurred.
A likely scenario for the use of this capability would be to submit the same file multiple times and
determine the variation in the results. MSC does not offer an automated way of doing this at this time.
Main Index
74
Introduction
Capabilities of SOL 200 have been expanded to support:
Using properties on PCOMPG as design variables
Using responses from fluid model
A modified objective function
Benefits
PCOMPG
The implementation provides a simple user interface to design and to track a particular ply over many
PCOMPGs which has the potential to significantly increase the productivity of engineers and designers.
Fluid Modes
Fluid modes can be utilized as design constraints in the optimization.
Objective Function Modification
Frequently, auto and aircraft manufacturers use SOL 200 to design just a tiny portion of the structure.
The mass of design portion can be 3 to 4 orders of magnitude smaller than the full structural mass.
Modifying the objective function provides a quick way avoid premature convergence.
Input
PCOMPG as Design Variables
The following KEYWORDs are added to the TYPE field of DVPREL1 Bulk Data entry.
For type=PCOMPG, PID field should have the ID of the PCOMPG entry and the PNAME/FID field can
have input of property name of fields or field number.
For type=GPLY, PID field should have the GPLYID on the continuation lines of PCOMPG and the
PNAME/FID field can only have T or THETA as input.
It should be noted that:
1 2 3 4 5 6 7 8 9 10
DVPREL1 ID TYPE PID PNAME/
FID
PMIN PMAX C0
DVPREL1 100 PCOMPG
or GPLY
PID or
GPLYID
Main Index
75
CHAPTER 6
Optimization
1. When DVPREL1 has TYPE=GPLY, all PCOMPG entries with GPLYID will participate in the
design. The relationship between design variable and properties are defined by the following
equation
or is the original thickness or THETA angle on PCOMPG which is determined
automatically from the PCOMPG entry and is utilized as multiplier to the design variables. This
formulation allows a ply with same GPLYID on different PCOMPGs to change in tandem
percentage-wise.
2. For with original value equal to 0.0, is taken as 1.0 and it is recommended
to have XINIT of DVID set to 0.0.
Example
$ $ $ $ $ $ $ $ $
DESVAR 1 T100000 1.00 0.01 100.0
$
DVPREL1 1 GPLY 100000 T
1 1.0
Fluid Modes as Design Constraints
With RTYPE=EIGN or FREQ, the default for PTYPE field is STRUC.
New parameter OBJMOD for DOPTPRM is implemented as a flag for objective function modification.
With DOPTPRM,OBJMOD,1, the original objective function value will be reset to 0.0. From the second
cycle onward, the objective function value represents the change of objective function with respect to the
original design.
The default value for OBJMOD is 0, meaning the total objective function value will be used.
1 2 3 4 5 6 7 8 9 10
DRESP1 ID LABEL RTYPE PTYPE REGION ATTA ATTB ATT1
ATT2 -etc.-
RTYPE
New Option for
PTYPE Field
Response
Attributes
ATTA ATTB ATTi
EIGN or FREQ STRUC or FLUID Normal Modes Number Approximation code
P
i
C0 Z T0
i
or THETA0
i
( ) H DVID
j
COEF
j
( )
T0 THETA0
THETA0 THETA0
Main Index
76
Output
Output for the previous new features of SOL 200 is presented in the following paragraph. New features
are highlighted in BOLD characters.
A single set of DESVAR/DVPREL1 with GPLY will cover multiple PCOMPG entries which has a ply
with GPLYID. The output of comparison of analysis and design model, will then look like:
In the previous output, the number printed under the PROPERTY ID column is the ID of PCOMPG.
Fluid Modes as Design Constraints
A sample of sensitivity for fluid mode responses is shown as follows. The output is produced via the
DSAPRT Case Control command.
With DOPTPRM,OBJMOD,1, objective function modification algorithm is activated. A sample of
objective function history is shown as follows. The output is available for all optimization jobs.
----- COMPARISON BETWEEN INPUT PROPERTY VALUES FROM ANALYSIS AND DESIGN MODELS -----
----------------------------------------------------------------------------------------------------------------------------
PROPERTY PROPERTY PROPERTY ANALYSIS DESIGN LOWER UPPER DIFFERENCE SPAWNING
TYPE ID NAME VALUE VALUE BOUND BOUND FLAG FLAG
----------------------------------------------------------------------------------------------------------------------------
GPLY 12 T 5.400000E-03 5.400000E-03 N/A N/A NONE
****************************************************************************
* *
* D E S I G N S E N S I T I V I T Y M A T R I X O U T P U T *
* *
* *
* R E S P O N S E S E N S I T I V I T Y C O E F F I C I E N T S *
* *
****************************************************************************

-------------------------------------------------------------------------------------------------------------------------------
DRESP1 ID= 101 RESPONSE TYPE= FREQ MODE ID= 1 FLUID SEID= 0
SUBCASE RESP VALUE DESIGN VARIABLE COEFFICIENT
-------------------------------------------------------------------------------------------------------------------------------
1 8.6023E+01 1 T 1.4641E-01

-------------------------------------------------------------------------------------------------------------------------------
DRESP1 ID= 102 RESPONSE TYPE= FREQ MODE ID= 2 FLUID SEID= 0
SUBCASE RESP VALUE DESIGN VARIABLE COEFFICIENT
-------------------------------------------------------------------------------------------------------------------------------
1 2.6020E+02 1 T 4.4286E-01
Main Index
77
CHAPTER 6
Optimization
Guidelines
For type=GPLY, the recommended values for fields of DESVAR and DVPREL1 are
The design model for exterior acoustic must be part of main input file which is after the BEGIN
BULK entry. Any design model entries placed after BEGIN BULK AFPM=xxxx are ignored.
The effectiveness of DOPTPRM,OBJMOD,1 is not consistent. Hence, it is recommended only
for optimization problems that design just a tiny portion of the full structure.
Limitations
DVPREL2 must not be used to link design variable and properties of PCOMPG.
Properties associated with MICRO feature of PCOMPG are not supported in SOL 200.
Example
A simple file, d200pcg1, with multiple PCOMPG entries is utilized here to demonstrate the features
implemented for PCOMPG support in SOL 200. Some key bulk data entries are shown as follows:
$
DESVAR 1 T100000 1.00 0.01 100.0
$
DVPREL1 1 GPLY 100000 T
1 1.0
DVPREL1 2 PCOMPG 12 T2
---------------------------------------------------------------------------------------------------------------
CYCLE APPROXIMATE EXACT OF OF
---------------------------------------------------------------------------------------------------------------
>>> OBJECTIVES in COLUMN 2/3 ARE INCREMENTAL TO OBJECTIVE OF ORIGINAL DESIGN = 1.0000E+05 <<<
>>> ADD INCREMENTAL OBJECTIVE TO ORIGINAL TO ARRIVE AT REAL OBJECTIVE OF EACH CYCLE <<<
---------------------------------------------------------------------------------------------------------------
INITIAL 0.000000E+00 1.249923E+00
1 8.947921E-03 7.812500E-03 1.453339E-01 9.368142E-01
2 -1.010694E-02 -7.812500E-03 -2.936888E-01 1.003581E+00
3 -1.317651E-02 -1.562500E-02 1.567034E-01 3.951643E+00
4 -1.562500E-02 -1.562500E-02 0.000000E+00 3.951643E+00
---------------------------------------------------------------------------------------------------------------
Bulk Data Entry Field Name Recommended Value
DESVAR X0 1.0
DVPREL1 C0 0.0
DVPREL1 COEF1 1.0
Main Index
78
1 0.0054
$
pcompg,12,,,5000.,hill,0.0,,,
,100000, 1, .0054, 45., yes,
,400000, 1, .0054, 90., yes,
,500000, 1, .0054, 90., yes,
,600000, 1, .0054, 0.0, yes
,700000, 1, .0054,-45., yes
,800000, 1, .0054, 45., yes
pcompg,22,,,5000.,hill,0.0,,,
,100000, 1, .0054, 45., yes,
,300000, 1, .0054, 0.0, yes,
,400000, 1, .0054, 90., yes,
,500000, 1, .0054, 90., yes
,600000, 1, .0054, 0.0, yes
,800000, 1, .0054, 45., yes
pcompg,33,,,5000.,hill,0.0,,,
,100000, 1, .0054, 45., yes,
,200000, 1, .0054,-45., yes,
,300000, 1, .0054, 0.0, yes,
,400000, 1, .0054, 90., yes
,500000, 1, .0054, 90., yes
,800000, 1, .0054, 45., yes
$
DVPREL1,1 links the thickness of ply 100000 in PCOMPG 12, 22 and 33 to DESVAR,1 and
DVPREL1,2 connects the thickness of PLY 400000 in PCOMPG,12 to DESVAR,1. Note that
DVPREL1,2 uses the existing equation for relation between design variables and properties while
DVPREL1,1 uses the new one shown in INPUT section. PCOMPG entries corresponding to the new
design are shown as follows,
Main Index
79
CHAPTER 6
Optimization
Fluid Modes as Design Constraints TPL test file: d200fmd1.dat
The output section shows results from this file.
$ *************************************************************
$ * *
$ * CONTINUOUS DESIGN CYCLE NUMBER = 4 *
$ * *
$ *************************************************************
$
$
$ UPDATED DESIGN MODEL DATA ENTRIES
$
DESVAR * 1T100000 3.37500000E+00 9.99999978E-03+D 1V
*D 1V 1.00000000E+02
$
$ UPDATED ANALYSIS MODEL DATA ENTRIES
$
PCOMPG* 12 0.00000000E+00 5.00000000E+03*
* HILL 0.00000000E+00 0.00000000E+00 *
* 100000 1 1.82250012E-02 4.50000000E+01*
* YES *
* 400000 1 1.82250012E-02 9.00000000E+01*
* YES *
* 500000 1 5.40000014E-03 9.00000000E+01*
* YES *
* 600000 1 5.40000014E-03 0.00000000E+00*
* YES *
* 700000 1 5.40000014E-03 -4.50000000E+01*
* YES *
* 800000 1 5.40000014E-03 4.50000000E+01*
* YES
PCOMPG* 22 0.00000000E+00 5.00000000E+03*
* HILL 0.00000000E+00 0.00000000E+00 *
* 100000 1 1.82250012E-02 4.50000000E+01*
* YES *
* 300000 1 5.40000014E-03 0.00000000E+00*
* YES *
* 400000 1 5.40000014E-03 9.00000000E+01*
* YES *
* 500000 1 5.40000014E-03 9.00000000E+01*
* YES *
* 600000 1 5.40000014E-03 0.00000000E+00*
* YES *
* 800000 1 5.40000014E-03 4.50000000E+01*
* YES
PCOMPG* 33 0.00000000E+00 5.00000000E+03*
* HILL 0.00000000E+00 0.00000000E+00 *
* 100000 1 1.82250012E-02 4.50000000E+01*
* YES *
* 200000 1 5.40000014E-03 -4.50000000E+01*
* YES *
* 300000 1 5.40000014E-03 0.00000000E+00*
* YES *
* 400000 1 5.40000014E-03 9.00000000E+01*
* YES *
* 500000 1 5.40000014E-03 9.00000000E+01*
* YES *
* 800000 1 5.40000014E-03 4.50000000E+01*
* YES
Main Index
80
Test file, d200zobj, is used. The design model covers just a tiny portion of the structure. The original
file produced following optimization history.
From column 3 of the previous output, the change in objective function is not visible at all. The same
output from d200zobj with DOPTPRM,OBJMOD,1 is shown as follows
Column 3 of the previous table shows the change of objective function. The original objective function
value can be found on the first line bracketed by >>> and <<<. Note that the change of objective
function is 8 orders of magnitude smaller than the original objective function value.
---------------------------------------------------------------------------------------------------------------
---------------------------------------------------------------------------------------------------------------
INITIAL 1.000000E+05 1.249923E+00
1 1.000000E+05 1.000000E+05 0.000000E+00 9.361193E-01
2 1.000000E+05 1.000000E+05 0.000000E+00 1.003858E+00
3 1.000000E+05 1.000000E+05 0.000000E+00 3.077129E+00
4 1.000000E+05 1.000000E+05 0.000000E+00 3.077129E+00
---------------------------------------------------------------------------------------------------------------
---------------------------------------------------------------------------------------------------------------
---------------------------------------------------------------------------------------------------------------
>>> OBJECTIVES in COLUMN 2/3 ARE INCREMENTAL TO OBJECTIVE OF ORIGINAL DESIGN = 1.0000E+05 <<<
>>> ADD INCREMENTAL OBJECTIVE TO ORIGINAL TO ARRIVE AT REAL OBJECTIVE OF EACH CYCLE <<<
---------------------------------------------------------------------------------------------------------------
INITIAL 0.000000E+00 1.249923E+00
1 8.947921E-03 7.812500E-03 1.453339E-01 9.368142E-01
2 -1.010694E-02 -7.812500E-03 -2.936888E-01 1.003581E+00
3 -1.317651E-02 -1.562500E-02 1.567034E-01 3.951643E+00
4 -1.562500E-02 -1.562500E-02 0.000000E+00 3.951643E+00
---------------------------------------------------------------------------------------------------------------
Main Index
Chapter 7: Aeroelasticity and Rotor Dynamic Improvements MSC Nastran 2008 r1 Release Guide
7
Aeroelasticity and Rotor
Dynamic Improvements
A New Aerodynamic Interpolation Method
External Spline Server
Blade Vibration Analysis

Main Index
82
Introduction
Solution of the flutter equations entails interpolation to compute the aerodynamics at the exact reduced
frequency of the flutter solution as a function of the aerodynamics that have been computed explicitly
based on the k (reduced frequency) values input on the MKAEROi entries. For many years, the
available interpolation schemes have been adaptations of the beam and surface spline methods used in
the splining of displacements and forces in aeroelasticity. Chapter 2.6 of the MSC Nastran Aeroelastic
Analysis Users Guide documents these methods. All of these methods perform their interpolation based
only on the k values and each term in the generalized aerodynamics matrix is weighted in the same way.
For MSC Nastran 2008 r1, an alternative interpolation is provided that interpolates each term in the
generalized aerodynamic matrix individually.
Inputs
The existing FLUTTER Bulk Data entry contains an IMETH field that allows the user to select between
L (linear interpolation on k-only) and S (surface interpolation on Mach number and k) methods. Under
this enhancement, an additional option (TCUB) has been provided to invoke a termwise cubic
interpolation technique. For legacy purposes, if the FLUTTER entry has a METH field of PK,
PKNL, PKS or PKNLS and the IMETH field is blank, S or L, the linear beam spline is used to
interpolate the aerodynamics as a function of reduced frequency. If IMETH is TCUB, the termwise cubic
spline technique is employed. Any other value of IMETH results in an error. If the flutter method is K
or KE, IMETH=S selects a surface spline on Mach and reduced frequency and IMETH=L selects a
linear method on reduced frequency and using the Mach number that is closest to the Mach number
specified on the FLFACT entry. It is an error to select METH = K or KE and IMETH=TCUB
Outputs
There are no new outputs as a result of this implementation.
The interpolation scheme involves determining weighting coefficients using cubic spline techniques
based on the k values entered on the MKAEROi entries and the generalized aerodynamics computed at
these ks. During the PK flutter analysis, an estimate of the k value is made. The interpolation is then
performed using:
Q
i j
k
est
( ) Q
i j
k
0
( ) Z C
i j
3
k C
i j
2
H ( ) k C
i j
1
H ( ) H k
Main Index
83
CHAPTER 7
Aeroelasticity and Rotor Dynamic Improvements
where:
If only one value is provided for the MKAEROi input, no interpolation is performed and the
aerodynamics are invariant.
If the value falls outside the range of ks input using the MKAEROi entries, no extrapolation is
performed. Instead, the aerodynamics at the lowest input k value are used if the desired k is lower than
the input ks and the aerodynamics at the highest input k value are used if the desired k is higher than any
input ks.
For sensitivity analysis, it is necessary to provide the sensitivity of the aerodynamics due to a change in
. Differentiation of the equation above gives:
If only one value is provided of if the falls outside the range of MKAEROi values, the sensitivity
is zero.
A convenient way to check interpolation using the TCUB method with the beam spline method
(IMETH=L) is to perform a flutter analysis with two subcases with the only difference being the IMETH
value. The flutter summary results should be close, but not identical. DIAG 39 can be turned on around
the FA1 module to provide debug data for the flutter analysis while DIAG 30 will print even more data.
Turning DIAG 30 and DIAG 39 on around the DSFLTE module in the FLUTSENS dmap will provide
information on the sensitivity analysis. It is cautioned that the output is particularly voluminous for the
sensitivity diagnostics.
IMETH=TCUB is only supported for the PK method of flutter analysis and its variants, i.e., it is not
supported for the K and KE methods.
Examples
Two examples are available with the release demonstrating this new capability. The first is named
csint.dat and is a variation of the simple HA145A example found in the MSC Nastran Aeroelasticity
Users Guide. An extra subcase has been provided that repeats the PK flutter analysis of the example
= A term in the generalized aerodynamic matrix. Real and imaginary terms are splined
separately
= k at which aerodynamics are required
= largest k value from the MKAEROi input that is
=
= Interpolation coefficients determined using a cubic spline (1,2,3 are superscripts, not
exponents.)
Q
i j
k
est
k
0
k
est
<
k k
est
k
o
C
ij
1 2 3 , ,
k
k
est
k
dQ
i j
k
est
( )
dk
JJJJJJJJJJJJJJJJJJJJJJJJJJ 3.0 C
i j
3
k 2.0 C
i j
2
H ( ) k C
i j
1
H Z
k k
est
Main Index
84
while setting IMETH to TCUB. A comparison of the flutter summary results for the two examples show
virtually identical results.
The second example is entitled cintopt.dat and is a variation of the HA200B example from the same
Users Guide. In this case, the flutter subcases have been converted from using the beam spline
interpolation to using the new TCUB interpolation. Again, there is virtually no change in the results.
Main Index
85
CHAPTER 7
Introduction
The external spline evaluation capability that was introduced with MSC Nastran 2007 r1 required that
every term in the server-generated spline matrix be stored in memory. This limited the capability since
very large, but very sparse, spline matrices would not fit into the available memory.
With this release, the API was updated to allow the spline matrix to be stored in a sparse format. The
fully-populated spline matrix format is still supported.
Inputs
No changes were made to the Nastran input file or to how the external spline server is used.
API Changes
The interface between Nastran and an external spline server was modified to support the sparse matrix
format.
Two changes were made to the calling sequence of the main spline server interface routine (sxsevd.c),
they are noted in bold and a slightly larger font:
void sxsevd (
INTEGER group_id, /* Group id */
INTEGER spline_id, /* Spline id */
INTEGER *usage, /* Usage string (stored as hollerith) */
INTEGER n_int_data, /* Number of integer data */
INTEGER *int_data, /* Integer data */
INTEGER n_real_data, /* Number of real data */
MACHINEPRECISION *real_data, /* Real data */
INTEGER n_char_data, /* Number of character data */
INTEGER *char_data, /* Character data (stored as hollerith) */
INTEGER n_dep_grid, /* Number of dependent grids */
INTEGER *dep_grid_id, /* Dependent grid ids */
MACHINEPRECISION *dep_grid_xyz, /* Dependent grid x,y,z locations */
INTEGER n_indep_grid, /* Number of independent grids */
INTEGER *indep_grid_id, /* Independent grid ids */
MACHINEPRECISION *indep_grid_xyz, /* Independent grid x,y,z locations */
INTEGER n_dep_elem, /* Number of dependent elements */
INTEGER *dep_elem, /* Dependent element table */
INTEGER n_indep_elem, /* Number of independent elements */
INTEGER *indep_elem, /* Independent element table */
char *command_line, /* Optional command line argument */
char *connect_data, /* Optional connect data */
INTEGER *ginfo, /* Output information about the spline matrix */
MACHINEPRECISION **gmat, /* The computed spline matrix */
INTEGER *error) { /* Error code */
1. A new integer parameter called ginfo is now output. This variable stores the total number of
nonzero terms in the spline matrix if it is stored in the sparse format. ginfo should have a value
of zero if the spline matrix is stored in the fully-populated format.
Main Index
86
2. Previous versions of Nastran allocated the memory for the spline matrix on the client (Nastran)
side of the problem. With the new version, the client does not know if the server will be storing
the spline matrix in the sparse or full formats. Therefore, it is now the servers responsibility to
allocate the memory to store the spline matrix. As a result of this change, the gmat variable,
which was previously a pointer is now a pointer to a pointer.
Sparse Matrix Format
If the sparse format is used to store the spline matrix, then the data must be stored in gmat in triplets of
(row number, column number, value) for each nonzero term in the spline matrix:
The ginfo variable must store the number of nonzero terms (n) in the spline matrix. The sparse gmat
will store numbers total.
Upgrading an Existing Spline Server
This section provides one method for upgrading an MSC Nastran 2007 r1 spline server to be compatible
with R3. The experienced C programmer may wish to implement the changes differently. It will be
assumed that the spline matrix will be stored in the fully-populated format.
1. Update the calling arguments of sxsevd.c to be exactly as listed above.
2. Declare a local variable to store the spline matrix:
MACHINEPRECISION *server_gmat=NULL;
3. Set the value of ginfo:
*ginfo = 0;
4. Allocate the memory to store the spline matrix.
gmat
Row number for value 1
Collumn number for value 1
Value 1
Row number of value 2
Column number for value 2
Value 2
.
.
.
Row number for value n
Column number for value n
Value n
Z
3 n ( )
Main Index
87
CHAPTER 7
In prior versions of Nastran; MSC Nastran 2007 r1, special options were added to SOL 106 to support
blade vibration analysis but were not documented in the MSC Nastran Release Guide or MD Nastran
Quick Reference Guide. The options are documented here for your convenience.
Frequency (Forced) Response Analysis - in addition to the current normal modes analysis in
SOL 106, the user can request a frequency response analysis. Both the normal modes and
frequency response analysis are requested with a separate subcase followed by the Case Control
commands ANALYSIS=MODES (Ch. 4) and ANALYSIS=DFREQ (Ch. 4) in the MSC Nastran
Quick Reference Guide.
Hot-to-Cold Analysis - allows the user to input the stressed or hot (deformed) geometry
using standard Bulk Data input and then unload the structure to determine the unstressed or
cold shape. See the description of the Case Control command ANALYSIS=HOT2COLD
(Ch. 4) and user parameters HTOCITS (Ch. 5), HTOCPRT (Ch. 5), and HTOCTOL (Ch. 5) in the
MSC Nastran Quick Reference Guide.
Tangential Acceleration and Coriolis Follower Forces in Frequency Response Analysis - Include
the effects of the tangential acceleration and Coriolis follower forces in the nonlinear differential
stiffness matrix to be used frequency (forced) response analysis. See the description of user
parameter HTOCITS (Ch. 5) in the MSC Nastran Quick Reference Guide.
These options enable the analyses of a rotating nonsymmetrical structure connected to a nonsymmetrical
stationary structure. The rotating component will be assumed to be spinning at a constant rate. The
procedure permits dynamic response calculations of the bypass fan, compressor, and turbine blades for
aerojet engines. It also allows analyses of rotating wing aircraft. The methodology can also be used for
dynamic response of the crankshaft/engine block of a reciprocating engine.
Main Index
88
Main Index

MSC Nastran 2008 r1 Release Guide

Загружено:

Сведения о документе

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 Nastran 2008 r1 Release Guide

Загружено:

Авторское право:

Доступные форматы

MSC Nastran 2008 r1

Implicit Nonlinear - SOL 600

Factor Matrix Caching for Lanczos

New TAUCS Indefinite Solver Improves Lanczos Performance

Shared Memory Parallel (SMP) Scalability Improvements for Static

New MAXRATIO Information Output

New SPARSESOLVER MDTSTATS Information Output

TEMPERATURE Case Control Command

Improvements in Fluid Eigenvalue Analysis

FLUID GRID Points and Partitioning

Distributed Memory Parallel (DMP) Diagnostic Messages

System Information Message (SIM) 6916

Topography (Bead or Stamp) Optimization

Randomization of an Input Data File

Random Elimination of Element Types

Enhancements in SOL 200 Optimization

A New Aerodynamic Interpolation Method

External Spline Server

Blade Vibration Analysis

Вам также может понравиться