Вы находитесь на странице: 1из 159

XFEM based Partly Cracked Interface Model for

Reinforced Concrete Analysis

A thesis submitted in partial fulfillment of the


requirements for the degree of

Master of Science in Structural Engineering

By
Iason Iakovidis

Cardiff School of Engineering


Cardiff University
Abstract
Non-linear Fracture mechanics deals with the macroscopic heterogeneity of quasi-
brittle materials such as concrete (Exadaktylos, 2006; Karihaloo, 1995). Quasi-brittle
materials are characterized by “strain softening” phenomenon, which is governed by
the length of fracture process zone (Hillerborg et al., 1976). This post-failure zone
includes dissipation mechanisms consuming the elastic stored strain energy of the
body, leading to progressive collapse (Karihaloo, 1995). In addition, concrete is
subjected to size effect, providing a more brittle-homogeneous concept for large
scale structures (Bazant, 1984).

Three basic theoretical concepts were developed through the years in order to
facilitate concrete`s complex characteristics: damage plasticity model, smeared
analysis and brittle approach (Mohammadi, 2008; Shi, 2009). In addition, other
parametrical or user defined approaches were examined to support and improve the
aforementioned techniques.

Besides the plain and reinforced concrete modeling, fracture and damage evolution
criteria were developed to represent initiation and growth of concrete cracks (RILEM,
1988). Apart from the criteria based on yield theories (Mohr-Coulomb, Tresca &
Guest, Rankine & Lame etc.), alternative ones developed governed by fracture
mechanical concepts, such as stress intensity factors, J-integral, fracture energy and
toughness etc. (Karihaloo, 1995; Shi, 2009) in combination with numerical
procedures, such as finite elements and extended finite elements.

In our case, the damage plasticity model will be examined in order to describe
concrete`s behavior. Reinforcement will be represented as an embedded region
simulating the interaction inside an interface element (Ngo & Scordelis, 1967;
Lundgren, 1996). XFEM cohesive crack concept, in combination with Rankine &
Lame`s criterion (1850), will be used to examine crack generation, while crack
growth will be described based on fracture energy due to previous work of Sluys &
Wells (2001), Zi & Belytschko (2003) and Asferg et al. (2006).
Acknowledgements
To begin with, I would like to thank all my professors and colleagues for one beautiful
postgraduate year in Cardiff. All the tutors were available to talk to, discuss problems
and give directions at any moment during the year.

In particular, I would like to thank Dr. Stephane Bordas, who helped me a lot in the
implementation of my thesis by giving me directions and providing me with very
useful information in order to overcome any difficulty I have faced. He was always
there to support me and discuss any idea for the development of my project.

Furthermore, I would like to thank Dr. Gardner, who helped me a lot during this
academic year. Firstly by allocating Dr. Bordas as my supervisor and supporting me
with any difficulty I had faced during my academic life in Cardiff. She was always
available to discuss any problem and willing to help, giving suggestions and
counseling me on my academic work and especially on the dissertation and case
study.

In addition, I would like to thank all my colleagues and especially Skerdi Muco and
Long Qui, who were both good friends and fellow students and helped me a lot
through the year. Studying together and facing academic difficulties made us better
engineers and professionals.

Finally, I would like to thank my family, which supported my academic dreams and
helped me fulfill my inner scientific aims by providing me with financial and
psychological aid during this year.
Contents
List of Figures .......................................................................................................................................... 7

1. Introduction ......................................................................................................................................... 1

2. Fracture Mechanics ............................................................................................................................. 3

2.1 Generation and Theoretical Background .................................................................................... 3

2.1.1. Stress Analysis and Design ............................................................................................... 4

2.1.2. Material Science ................................................................................................................ 5

2.1.3. Non-Destructive Methods .................................................................................................. 8

2.2. Design Principles, Concepts and Limitations ........................................................................... 10

2.2.1. Concrete Characteristics ................................................................................................. 10

2.2.2. Three Point Bending Test ................................................................................................ 14

2.2.3. Cracking due to Eurocode 2 ............................................................................................ 17

3. Linear and non-linear Fracture Mechanics concepts ........................................................................ 20

3.1. Linear-Elastic and elasto-plastic theories ................................................................................. 20

3.1.1. Griffith`s Fracture Strength .............................................................................................. 20

3.1.2. Stress Intensity Factor ..................................................................................................... 24

3.1.3. Brittle and Quasi-Brittle Material...................................................................................... 26

3.1.4. CTOD .............................................................................................................................. 27

3.2. Non-linear Fracture Mechanics Concepts ................................................................................ 29

3.3. Existing Modeling and Computational Techniques .................................................................. 31

3.4. XFEM for Concrete ................................................................................................................... 40

3.5. Asferg et al. (2006) Model Concept ......................................................................................... 41

4. Testing and Results Analysis ............................................................................................................ 43

4.1. Model parameters and Analysis description (Abaqus) ............................................................. 43

4.1.1. Units And Geometrical Definition .................................................................................... 43

4.1.2. Material`s Parameters ..................................................................................................... 44

4.2 Mesh Refinement Study for Notched Un-Reinforced Beam in TPBT ....................................... 48

4.2.1. Introduction ...................................................................................................................... 48

4.2.2. Results and Discussion ................................................................................................... 49

4.3. Post Failure Tension Softening Curve and Energy Investigation ............................................ 51

4.3.1. Introduction ...................................................................................................................... 51


4.3.2. Linear Relation ................................................................................................................ 51

4.3.3. Bi-Linear Response ......................................................................................................... 52

4.3.4. Exponential Form ............................................................................................................ 54

4.3.5. Power Law ....................................................................................................................... 54

4.3.6. Discussion and Conclusions ........................................................................................... 55

4.4. Reinforced Beam in TPBT........................................................................................................ 61

4.4.1. Introduction ...................................................................................................................... 61

4.4.2. Mesh Definition and Steel Embedded Region ................................................................ 63

4.4.3. Results Discussion and Conclusions .............................................................................. 65

4.5. Reinforced Concrete Beam Crack Definition ........................................................................... 70

4.5.1 Cohesive Zone Model....................................................................................................... 70

4.5.2. Initiation, Propagation and Smooth Closure Criterion ..................................................... 70

4.5.3. Analytical Testing for the Generation of the Approximated Model .................................. 72

4.5.4. Results Discussion .......................................................................................................... 87

4.6. Theoretical Discussion of Multi-cracking and Branching Phenomena in the Developed Model
......................................................................................................................................................... 89

5. Conclusion ........................................................................................................................................ 92

5.1. General Evaluation of the Project ............................................................................................ 92

5.2. Aims Succeeded ...................................................................................................................... 93

5.3. Proposition for Further Analysis ............................................................................................... 96

References .............................................................................................................................................. 1

Appendix A
LIST OF FIGURES
Fig. 1: Boston Molasses disaster, Liberty Ship rapture, Comet jet case, Aloha Airlines Boeing incident
(Ramesh, 2003) ...................................................................................................................................... 3
Fig. 2: Fracture Mechanics theoretical components ............................................................................... 4
Fig. 3: Microcracking due to cyclic loading (Exadaktylos, 2006) ............................................................ 5
Fig. 4: Description of crack modes: opening, sliding, tearing (Shi, 2009). .............................................. 7
Fig. 5: Stress distribution and fringe density in circular, elliptical hole and edge crack respectively
(Ramesh, 2003) ...................................................................................................................................... 9
Fig. 6: Three point bending test representation (Asferg et al., 2006) ................................................... 14
Fig. 7: Load definition and maximum moment point ............................................................................. 15
Fig. 8: Load-deflection plot describing elastic response when load is applied and not applied, where
beam returns to its initial state with a slight deflection (Concrete Society UK) ..................................... 15
Fig. 9: Crack Propagation due to load increase (Concrete Society UK) ............................................... 16
Fig. 10: Load-deflection plot of under-reinforced beam ........................................................................ 16
Fig. 11: Under-reinforced beam failure due to shear ............................................................................ 16
Fig. 12: Lattice Model for theoretical strength calculation (Ramesh, 2003) .......................................... 21
Fig. 13: Properties of calculated data of lattice model (Ramesh, 2003) ............................................... 21
Fig. 14: Actual Test Results of Griffith (Karihaloo, 1995) ...................................................................... 22
Fig. 15: Modes of fracture based on SIFs (Kumar &Barai, 2011) ......................................................... 24
Fig. 16: Effective length concept of Irwin (Mohammadi, 2008) ............................................................. 28
Fig. 17: Barenblatt-Dugdale`s effective crack length assumption (Mohammadi, 2008) ....................... 29
Fig. 18: Model introduced by Dugdale and corresponding σ/σy rate (Mohammadi, 2008) .................. 32
Fig. 19: Fictitious Crack Model and fracture energy estimation method (Karihaloo, 1995; Shi, 2009) 33
Fig. 20: Concepts of crack band model (Mohammadi, 2008) ............................................................... 33
Fig. 21: Multi-scale laws and crack opening representation (Needleman, 1990; Strang et al, 2006b) 34
Fig. 22: Crack band representation in front of crack tip (Bažant & Oh, 1983; Karihaloo; 1995) ........... 34
Fig. 23: Comparison of simulated models` load-deflection plots (Asferg et al., 2007) ......................... 42
Fig. 24: Geometrical Definition of the Model (Asferg et al., 2006) ........................................................ 43
Fig. 25: Dilation angle determination based on biaxial stress plot (Kmiecik & Kaminski, 2011) .......... 45
Fig. 26: Eccentricity determination described as the distance between hyperbolic vortex and
asymptotic intersection in the center of hyperbola (Kmiecik & Kaminski, 2011) .................................. 46
Fig. 27: Plot circle for describing failure under biaxial compression for determination of strength
parameter K (Heirany & Ghaemian, 2011) ........................................................................................... 46
Fig. 28: Strain-strain curve for estimation of elasto-plastic properties of concrete damage plasticity
model (Abaqus Manual) ........................................................................................................................ 47
Fig. 29: Mesh schemes used for mesh refinement study ..................................................................... 49
Fig. 30: Load-Load Point displacement plot for refinement study......................................................... 49
Fig. 31: Load-load point deflection plot describing the sudden load drop (Malvar & Warren, 1988) .... 50
Fig. 32: Refinement performed for crack region ................................................................................... 51
Fig. 33: Linear Constitutive Law representation (Karihaloo, 1995) ....................................................... 51
Fig. 4 Stress-applied displacement relation plot and stress-cracking displacement plot based on linear
constitutive law in Abaqus ..................................................................................................................... 52
Fig. 35: Bi-linear constitutive law based on Malvar & Warren (1988) calculated using Gf=160N/m .... 52
Fig. 36: Results Calculated in Abaqus due to bi-linear constitutive law ............................................... 53
Fig. 37: Results Calculated in Abaqus due to exponential constitutive law .......................................... 54
Fig. 38: Power Law Described by Malvar & Warren (1988); RILEM (1988) ......................................... 55
Fig. 39: Calculated results using Abaqus 6.12 and Abaqus manual results for lower properties C40
(3.2.11-11) ............................................................................................................................................. 55
Fig. 40: By introducing also the power law in our results and comparing with Hillerborg (1981) ......... 56
Fig. 41: Rectangular, trapezoid and linear constitutive laws described by Hillerborg et al. (1979) ...... 57
Fig. 42: Bi-linear consistutive laws described by Rokugo et al. (1989) and Petersson et al. (1981) (Shi,
2009) ..................................................................................................................................................... 57
Fig. 43: Bi-linear constitutive law based on experimental, Petersson et al (1981) and Planas et al.
GBF (1992b; 1993) describing the tension softening phenomenon in concrete ................................... 58
Fig. 44: Fracture Energy Idealisation based on bi-linear curve for theoretical outcomes ..................... 59
Fig. 45: Suggested graphical representation of constitutive law........................................................... 61
Fig. 46: Plain concrete subjected to three point bending (Fanella, 2010) ............................................ 62
Fig. 47: Reinforced Concrete and crack initiation (Fanella, 2010) ........................................................ 62
Fig. 48: Reinforced Concrete Beam meshing model ............................................................................ 63
Fig. 49: Typical Stress-Strain plot for steel class 500C and the corresponding bi-linear simplification
(Panagopoulos & Kirtas, 2005) ............................................................................................................. 64
Fig. 50: Load-Deflection response for TPBT described by Asferg et al. (2006) and the one calculated
using Abaqus ........................................................................................................................................ 65
Fig. 51: Projection of deflection calculated point in both diagrammatic plots ....................................... 68
Fig. 52: Represents the areas where the two simulations vary ............................................................ 68
Fig. 53: Representation of fracture process zone between grains of concrete (Karihaloo, 1995) ........ 70
Fig. 54: Crack propagation due to Boundary Conditions/ Uy=U2=0/ 2-point CF ................................... 73
Fig. 55: Crack propagation due to BCs / Uy=U2=0/ Load as Displacement .......................................... 73
Fig. 56: Crack propagation where BCs introduced as loads ................................................................. 73
Fig. 57: Crack Propagation with BC U2=0 and Load defined as pressure ............................................ 74
Fig. 58: Crack Propagation with BC as load (CF) and Load defined as pressure ................................ 74
Fig. 59: Crack Propagation if load is applied in a mid-datum point....................................................... 75
Fig. 60: Cohesive Behavior via Interaction Property 1 .......................................................................... 75
Fig. 61: Cohesive Behavior via Interaction Property 2 .......................................................................... 76
Fig. 62: Cohesive Behavior via Interaction Property 3 .......................................................................... 76
Fig. 63: Cohesive Behavior via Interaction Property 4 .......................................................................... 76
Fig. 64: Interaction as Normal-Hard Contact property .......................................................................... 77
Fig. 65: Interaction as Normal-Penalty Linear Contact Property .......................................................... 77
Fig. 66: Using Only XFEM crack Propagation with no Interaction Property ......................................... 77
Fig. 67: Pull-out Pressure in both Sides ................................................................................................ 78
Fig. 68: Pull-out Pressure in both Sides and restrain of mid-datum point ............................................ 78
Fig. 69: Maxps= 1, 100 Increments....................................................................................................... 80
Fig. 70: Maxps= 10, 100 increments ..................................................................................................... 80
Fig. 71: Maxps equal to 100, 100 increments ....................................................................................... 80
Fig. 72: Maxps= 1000, 100 increments ................................................................................................. 81
4
Fig. 73: Maxps equal to 10 , 100 increments ........................................................................................ 81
5
Fig. 74: Maxps equal to 10 , 100 increments ........................................................................................ 81
6
Fig. 75: Maxps equal to 10 , 100 increments ........................................................................................ 82
7
Fig. 76: Maxps equal to 10 , 100 increments ........................................................................................ 82
Fig. 77: Maxps= 1 and 1000 increments ............................................................................................... 82
Fig. 78: Maxps= 10 and 1000 increments ............................................................................................. 83
2
Fig. 79: Maxps= 10 and 1000 increments ........................................................................................... 83
3
Fig. 80: Maxps equal to 10 and 1000 increments ................................................................................ 83
4
Fig. 81: Maxps equal to 10 and 1000 increments................................................................................ 84
5
Fig. 82: Maxps equal to 10 and 1000 increments................................................................................ 84
6
Fig. 83: Maxps equal to 10 and 1000 increments................................................................................ 85
7
Fig. 84: Maxps equal to 10 and 1000 increments................................................................................ 85
Fig. 85: Simple notched plain Concrete beam cracking (Bordelon, 2008) ........................................... 86
Fig. 86: Maps=0.1 and 1000 increments .............................................................................................. 86
Fig. 87: Maxps=0.001 and 1000 Increments ........................................................................................ 86
Fig. 88: Red Stress concentration for reinforcement yielding first 100 increments .............................. 88
Fig. 89: Shear tensile zone failure (Panagopoulos & Kirtas, 2006; Karagiannis & Chalioris, 1999) .... 88
Fig. 90: Representing equal stress distribution in both sides of the crack ............................................ 89
Fig. 91: Experimental testing representation of under-reinforced concrete beam (Concrete Society
UK) ........................................................................................................................................................ 90
Fig. 92: Load application scheme due to breath and length ................................................................. 90
List of Tables
Table 1: Requirements for Chloride content due to type of concrete (BS EN1008:2002) .................... 13
Table 2: Requirements for harmful contaminations due BS EN1008:2002 .......................................... 13
Table 3: Recommended values of wmax (mm) (Eurocode 2– Table 7.1Ν) ............................................ 20
Table 4: FPZ length for some materials (Karihaloo, 1995) ................................................................... 30
Table 5: Unit`s definition based on analytical systems of measuring ................................................... 43
Table 6: Plasticity options determination based on calculated data ..................................................... 47
Table 7: Compressive and Tensile behavior determination based on tabular data.............................. 48
Table 8: Definition of concrete and steel properties in Abaqus ............................................................ 72
Table 9: Definition of field Output for crack initiation and propagation ................................................. 72
Nomenclature
NLFM Non Linear Fracture Mechanics
LFM Linear Fracture Mechanics
XFEA Extended Finite Elements Analysis
CST Constant Strain Triangulars
LST Linear Strain Triangulars
LSQ Linear Strain Quadrilaterals
MAXPS Maximum Principal Stress
TPBT Three-Point Bending Test
CCM Cohesive Crack Model
CTOD Crack- Tip Opening Displacement
CZM Cohesive Zone Model
NDM Non-Destructive Methods
NDT Non-Destructive Testing
NDE Non-Destructive Evaluation
W/C Water to Cement Ratio
ASG Aggregate Specific Gravity
BSG Bulk Specific Gravity
E Modulus of Elasticity
v Poisson Ratio
FPZ Fracture Process Zone
Wc Critical Crack Displacement
FCM Fictitious Crack Model
EFGM Element Free Galerkin Method
FEA Finite Element Analysis
PoU Partition of Unity
1D One Dimensional
2D Two Dimensional
3D Three Dimensional
DOF Degrees of Freedom
DPM Damage Plasticity Model
SCM Smeared Crack Model
VCCT Virtual Crack Closing Techniques
SAR Static Stress Recovery
MLS Moving Least Squares
LPD Load Point Deflection
GBF General Bi-linear Fitting
D Scalar Variable
Tn Normal Stress Component
Ts Shear Stress Component
Tt Tension Stress Component
CMOD Crack Mouth Opening Displacement
VD Virtual Displacement
RC Reinforced Concrete
MC Moisture Content
EPFM Elasto Plastic Fracture Mechanics
CTODc Critical Crack Tip Opening Displacement
Gf Fracture Energy
KIC Fracture Toughness
SIF Stress Intensity Factor
A Size of the Structure
1. I NTRODUCTION
The key idea of our model is to describe cracking behavior of plain and reinforced
concrete inside the frame of NLFM. In particular, XFEA will be used in combination
with cohesive crack simulation, in order to develop a model able to describe: a) plain
and reinforced concrete, based on damage plasticity approach (Hansen, 2001;
Taqieddin, 2008; Xu, 2013) b) the bond slip zone, using embedded region interface
model (Ngo & Scordelis, 1967; Lundgren, 1999; Asferg et al. 2007) and c) crack
initiation, propagation and smooth closure phenomena, due to the previous works of
Zi & Belytschko (2003), Asferg et al. (2005;2006;2007) and Mougaard et al. (2011).

Most of the aforementioned works used CST, in order to accommodate the


longitudinal crack propagation. Zi & Belytschko`s (2003) main problem was equal
stress representation in both sides of crack surface. Asferg et al. (2006) and
Mougaard et al. (2011) improved that model using second order LST adaptive
method.

In our case, LSQ will be used in order to form the meshed regions, while the crack
initiation, propagation and closure will be simulated using maximum principal stress
criterion as described by Wells & Sluys (2001). The latter requires the maximum
value of stress in the crack tip to exceed the ultimate tension limit strength of the
material (Rankine & Lame; 1850), in order for the crack to grow perpendicular to
principal stress orientation (Zi & Belytschko, 2003).

The aforementioned will be investigated in respect to Eurocode 2, providing results


analysis based on previous experimental data, such as in the case of three point
bending test. In general we can say that there are four basic aims for this project.

To begin with, we will try to create a model combining the concepts of Asferg et al.
(2006) and some of the improvements suggested by him, such as the MAXPS
criterion, which was also examined by Zi & Belytschko (2003).

Secondly, we will try to examine how mesh refinement affects the results of load-
displacement curve and consequently describe tension softening. In this case there

1
will be a comparison with model described in Abaqus manual 6.11. The
representation of plain concrete will be implemented using half the beam, due to
symmetry.

Thirdly, to investigate how the formula used to calculate fracture energy affects the
tension softening. In order to examine this, we will use a linear, bi-linear, exponential
and power constitutive law as described by Malvar & Warren (1988), RILEM (1988),
Karihaloo (1995), and Hillerborg & Petersson (1983). After comparing the three
methods, a theoretical graphical tension softening curve will be suggested.

Last but not least, there will be an in-depth parametric analysis of cracking behavior.
The computational results will be compared with the experimental three point test of
the Concrete Society UK, and the Eurocode 2 deflection procedure, in order to
represent a real time benchmark test and verify the obtained values.

Finally, we will try to theoretically investigate the multi-cracking phenomenon


performing a research summarization, mostly based on previous studies of Geniaut
(2011), Cahill et al. (2010), Paulszny & Matthai (2009) and Sima & Wang (2012).
The purpose of this investigation is to provide us with suggestions for further
research.

2
2. F RACTURE M ECHANICS
2.1 G E NE R A TI ON AND T H E OR E T I CA L B A C K G RO U ND
Structural problem analysis and measurement begins when a structure fails without
an obvious reason. Some of these structural failures triggered the engineering
community to focus and evaluate further the design procedures (Ramesh, 2003).

In particular, Boston Molasses Tank Failure (1919), which caused a characteristic


smell in parts of Boston for many decades, Liberty ship failure (1946), in which the
ship cracked and cut in the middle during the World War II, Comet Jet (1954), which
was the first commercial jet put into service, and Aloha Airlines Boeing fuselage
failure (1988), which happened due to corrosion leading to severe cracking.

Fig. 1: Boston Molasses disaster, Liberty Ship rapture, Comet jet case, Aloha Airlines Boeing incident
(Ramesh, 2003)

In the aforementioned cases, the structure got separated without warning or visible
plastic deformation. Probable, the presence of triaxial stress state, inherent
material`s flaws, low temperature or ageing due to corrosion, critical crack growth
due to fatigue and rapid loading such, as decompression or thermal shock (Ramesh,
2003), led to brittle failure.

Theoretical Background
Fracture Mechanics is a broad area covering several disciplines of structural
engineering (Ramesh, 2003). As a matter of fact, it could be expressed as an

3
theoretical field between stress analysis and design, material science and non-
destructive testing (Fig. 2).

Fig. 2: Fracture Mechanics theoretical components

2.1.1. S T R E S S A N A L YS I S AND D E S I GN
Fracture Mechanics discuss in depth the “phenomenon” of embrittlement in structural
systems. This theory is referred to as “Damage Tolerant Design Approach” (Bansal,
2001), and aims to predict the failure state of a structure and the estimate time until
the fracture growth reaches a critical level causing irreversible damage (Rains,
2013).

Strength of materials examines the behavior of a material with three test


methodologies: split-cylinder tensile test, TPBT and twisting of shaft under torsion
test. The first is able to characterize the material as “isotropic”, in respect to modulus
of elasticity (E) and Poisson ratio (v). The flexural method investigates the stress
distribution under uniaxial concentrated load, while torsion method describes the
variation of shear to cross-sectional dimensions.

The aforementioned can help us estimate the structural sufficiency evaluating the
corresponding stresses due to geometrical section definition (Ramesh, 2003). In this
respect, yielding theories and failure criteria such as Rankine & Lamé stress
criterion, Tresca-Guest`s shear stress and Von-Mises` Principle Strain were
developed in order to accommodate load combination analysis (axial, torsional and
flexural) and in general exploit the mechanical properties of materials preventing
structural failure (Coates et al, 1988).

4
The necessity to investigate materials` behaviour in greater depth, generate a new
test method called “fatigue test”, which describes the failure of an element under the
cyclic loading and known environmental conditions (Williams & Todd, 2000). Fatigue
is the exhausting of strength or power of recovery for a material, deterioration, under
repeated loading and in structural sense “formation of a crack or cracks, as a result
of a repeated application of loads, each of which is insufficient to cause normal static
failure by itself (Heywood, 1962).

The aforementioned test methods, discussed also in Standards, identify the material
as a “linear discontinuity” under constant thermal and environmental conditions,
taking into account the existence of a crack or a disorder by using a safety factor
(Bond et al., 2006).

2.1.2. M A TE R I A L S C I E N C E
According to material science, fracture is the generation of new material surfaces
leading progressively to failure. In particular, in microscopic level fracture is the
bonding destruction between material`s atoms, while macroscopically it is the crack
branching and multiple structural splitting, due to crack growth. Finally, in the
mesoscopic level, fracture occurs due to crack generation, growth and coalescence
i.e. voids and flaws between the material`s grains (Exadaktylos, 2006) (Fig. 3).

Fig. 3: Microcracking due to cyclic loading (Exadaktylos, 2006)

Karmarsch`s (1858) experimental formula estimated load capacity of steel wires,


while Griffith (1921) provides a practical approximation for fiberglass. Wolhers (1860)
and Kommers (1920) observed the strength dependency to surface flaws based on
fatigue test, mentioning a 20-50% increase of component`s strength when subjected
to grinding or polishing. Kirsch (1898) investigated the distribution of stress fields in

5
tension strip with embedded elliptical holes, which was further studied by Inglis
(1913) for specifying the severity of cracks (Exadaktylos, 2006; Ramesh, 2003).

Wieghardt & Leon (1908) studied the stress concentration at the apex of a wedge
opening and the orientation of fracture subjected to load combinations providing
information about the asymptotic behavior of near crack-tip stress fields. Wieghardt
also observed that Rankine criterion (1850) was based on the assumption of infinite
arbitrary distributed stress, over a minute region including the crack-tip. In addition,
Tresca & Guest (1868) improved Rankine, by adding crack angle and direction

prediction based on surface energy γ and net parameter c:  c  ( E / c) 0.5 or

 c  0.25  ( E / 13) .

Smekal (1922; 1935) assumed that unevenness in the material`s bulk and pre-
existing micro-cracks lead to strength degradation. Dislocation theory, which was
based on this concept and described a statistical procedure in which micro-flaws
initiate, and propagate in a field slightly plastically deformed, was introduced in1934.
Weibull (1939) developed a statistical fracture theory widely applied to ceramic
materials assuming, that brittleness and corresponding strength exhaustion are
affected by size, distribution and state of stress (Petrovic, 1987).

Zener-Holloman (1944) studied brittle fracture of inelastic bodies, in respect to


Griffith`s theory (1921), and observed that ballistic rates of thermal softening lead to
shear band formation. Orowan`s study (1945) examined the plastic deformation of
materials failing in a brittle manner, which was later used by Irwin (1955) for
improving Griffith`s brittle fracture energy dissipation law. This law states that the
ultimate cracking dissipated energy has two forms: stored elastic strain energy and
plastic consumption and surface energy. In other words G=2γ+Gp, which was
verified by Orowan mentioning the necessity of Griffith`s principal to investigate
brittle fracture. In addition, Irwin (1957) determined strain energy release rate, G=-
dΠ/dA, where G is the crack driving force or crack propagation force characterized
by Karihaloo (1995) as the requirement for crack extension of one unit cracking area.

6
Previous work of Wieghardt & Leon (1908) was used by Williams (1952, 1957) to
describe the constant singularity in linear elastic materials determined by r-1/2.
Williams assumed that the elemental strain energy and the entire strain energy of a
finite element can be calculated by:

dW   2rr  r 2n2rr
R
W   r 2n1r
0 (Shi, 2009)

Williams` (1957) theory led to the calculation of the stress field in the vicinity of the
crack tip based on three crack propagation modes introduced by Irwin (1957) (Fig.4).

Fig. 4: Description of crack modes: opening, sliding, tearing (Shi, 2009).

Irwin`s potential energy concept (ΔΠ) is based on a crack infinitesimal growth (Δα) in
an infinite body subjected to mix-mode fracture. The opening and sliding
displacements calculated introduce stress intensity factors in respect to surface
density energy γ calculated by:

 K 2 K 2 K 2
  G  I  II  III
 E` E` 2 (Karihaloo, 1995)

In 1959, Barenblatt proposed the CCM describing an infinitesimal growth of constant


length ρ (<<α) inside a plastic region including the crack tips. In order to secure
smooth closure of crack-tips cohesive forces are applied to them (Karihaloo, 1995).

There are three main crack classes due to Exadaktylos (2006): the first in which
cracks generate and intersect having a length up to 1mm, the second characterized

7
by a longitudinal growth when subjected to cyclic loading and finally those extending
without any limits after reaching a critical length W>W IC.

Dugdale et al. (1959; 1963) further improved Barenblatt`s model, assuming a slight
variation of plastic zone`s length (ρ≤α), which increases until it reaches the half crack
length when subjected to critical yield stress σy (Karihaloo, 1995). Furthermore,
CTOD factor, which is dependent on the length of CZM and stress localization, was

introduced. Finally, an infinitesimal σ/σy ratio leading to  ( )0.5  (' y  Wc )0.5 can

emphasize the relation between Irwin`s and Griffith`s criterion for Elasto-plastic
materials with limited plasticity (Karihaloo, 1995; Shi, 2009).

2.1.3. N O N -D E S TR U C T I V E M E TH O DS
NDT, NDE and NDM are economic methods used in order to evaluate the health and
lifetime expectancy of structural systems, in terms of serviceability, while the
structure remains intact and available for further investigation. These methods were
officially introduced after the pressure vessel accident of Hartford Connecticut (1854)
which led to the ratification of law including annual mandatory tests for structures.
Rontgen (1895) invented the X-rays, and „Penetrating Flow Liquid‟ was produced in
1880. Metal radiography was introduced by Lester in 1920 and in 1927 the first
magnetic testing system was used in railways. Firestone (1940, 1944) developed the
ultrasonic method and Kaiser (1950) introduced the method of acoustic emissions.

The tests mostly used in relation with fracture mechanics are: 1) visual Inspection, 2)
radiographic methods, 3) Ultrasonic Inspection 4) Acoustic Emissions (AE) and
Stress-Wave Testing (SWE), 5) Electric methods such as Potential Drop, Eddy-
Current, Remote Field Testing, 6) Magnetic methods, such as Magnetic Particles,
Magnetography, Flux Leakage, Barkhausen Noise Test and 7) Penetrating Liquid
Flow in combination with Radiography and leak detection.

Jeng et al. (2003) used echo-test evaluation and impulse response test for various
concrete structures. Maierhofer et al. (2010) described the NDM economical repair
aspect; Lew et al. (2005) noted the difficulty of applying NDT in concrete in contrast
to steel, due to microscopic heterogeneity and eclectically non-conductivity of bond

8
zone. Krstulovic-Opara et al. (1996) introduced some new methods of testing such
as Rayleigh Waive Dispersion and Zhu & Popovics (2006) described a non-conduct
air coupled sensor system for effective flaw detection. Finally, Shah & Chandra
Kischen (2011) found significant increase of acoustic emissions due to progression
of damage. The increase of AE is proportional to the water cement ratio variation for
damage estimation. Kaiser effect did not occur due to microcracking, thus ultrasonic
test seems able to facilitate the evaluation.

Last but not least concept of photo-elasticity allowed the non- destructive stress
analysis of materials based on optomechanical birefringence. According to Ramesh
(2003) from the density of the fringes, an edged crack exhibits higher deformation,
and stress distribution leading to more severe cracking than central cracks (Fig. 5).

Fig. 5: Stress distribution and fringe density in circular, elliptical hole and edge crack respectively
(Ramesh, 2003)

9
2.2. D E S I GN P R I N C I P L E S , C ON CE P TS A ND L I MI TA T I ONS
2.2.1. C O N C R E T E C H A R A C TE R I S T I CS
Concrete is a composite material, which consists of coarse granular material, fine or
coarse aggregates or filler (sand, gravel etc.), hardened paste (cement or binder),
water, other admixtures and steel (Neville, 2011). Strong and durable concrete
production is based on careful proportioning and mixing of concrete compounds
(Lamond & Pielert, 2006). The voids between aggregates must be filled and the
strength of the mixture increases with the elimination of porosity of the matrix
(Karihaloo, 1995). A properly designed concrete mixture typically consists of 10-15%
cement, 60-75% aggregates and 15-20% water. It is possible that a 5-8% entrained
air is present in the mixture (Lamond & Pielert, 2006).

Advantages
First of all, concrete is inexpensive, because its components are found in plethora in
nature (Panagopoulos & Kirtas, 2006). Moreover, it hardens in ambient temperature,
which means the absence of human intervention (Neville, 2011). It can be casted
and modified at any size and shape, embedding low energy during production in
contrast with steel. In particular plain concrete has 450-750 KWh/ton, reinforced has
800-3200 KWh/ton, while steel reaches 8000 kWh/ton (Zongjin Li, 2011).

Reinforced concrete is used in pipelining, due to its high water resistance, mostly
deteriorating due to chlorides, sulfates, and carbon dioxide (Neville, 2011).
Furthermore, concrete exhibits a high temperature resistance, due to heat
conductivity and heat mass storing capacity enabling it to withstand fire up to 8 hours
(Lamond & Pielert, 2006).

Another important aspect of concrete is its compatibility with steel reinforcement,


because both possess quite the same thermal coefficient (Panagopoulos & Kirtas,
2006). Thus, steel provides tensile strength, while concrete provide protection as
cover for steel corrosion (Neville, 2011). Finally, concrete does not need coating or
painting, thus it has lower levels of maintenance (Lamond & Pielert, 2006).

10
Disadvantages
Concrete has low tensile strength requiring reinforcement installation which
consequently increases its cost and energy consumption (Lamond & Pielert, 2006).
In addition, concrete exhibits low toughness, which can be described as the ability to
absorb energy (Karihaloo, 1995). Also, its specific strength, ratio between strength
and bulk specific gravity, is low, 35-60/ 2.3 for normal concrete (Neville, 2011).
Finally, concrete structures have inherent bulk flaws, which contribute to the
difference between the theoretical and experimental strength estimation (Karihaloo,
1995; Lamond & Pielert, 2006).

Due to Mehta & Monteiro (1993) and Kosmatka et al. (2002) concrete can be
classified based on its unit weight as: ultra-light weight <1200 kg/m3, light weight
1200-1800 kg/m3, normal weight 2400 kg/m3, and heavy-weight >3200 kg/m3. Also,
based on cylindrical specimen compressive strength: low <20MPa, moderate 20-50
MPa, high 50-200 MPa, ultra-high >200 MPa. Finally, due to additives can be
classified as: normal, fiber reinforced, shrinkage-compensating and polymer.

Portland cement
Cement is a binding material connecting aggregates together forming a solid mass. It
consists of raw materials, mainly lime and argillaceous impurities. Cementitious
binders such as blast furnace slag, natural pozzolana (volcanic ash) or fly ash added
to the clinker give further properties to concrete mix (Lamond & Pielert, 2006).

Cement reacts chemically in contact with water (hydration), in which cement paste
exhibits hardening and strengthening. Paste quality influences concrete`s properties,
by changing the water to cement ratio. By lowering W/C and no affecting the
workability, we can produce high-quality concrete (Lamond & Pielert, 2006).
Generally, elimination of porosity means greater microscopic homogeneity and
improved mechanical properties (Karihaloo, 1995). In addition, the interface between
cement and aggregates called weakest link plays a key role in the mechanical
properties of concrete (Exadaktylos, 2006).

11
Aggregates
There are two types of aggregates: Fine aggregates having dmax≤ 4.75 mm (No. 4
Sieve) and coarse aggregates dmax≥4.75 mm. Aggregates have high compressive
strength, which affect the mean mixture compressive strength in light-weight
concretes (Lamond & Pielert, 2006). Cohesion, workability, durability and density of
concrete can be defined based on the physic and mineralogy of aggregates (Bond et
al., 2006).

Smooth and rounded aggregates improve workability, while the angular and rough
improve the bonding, requiring more cement paste to fill the porous (Bond et al.,
2006). The key factor that influences concrete`s strength and durability is
aggregates` grading, which minimizes porosity, paste requirement, and W/C ratio by
filling the voids between larger aggregates (Karihaloo, 1995; Mindess & Young,
1981; Mehta & Monteiro, 1993).

It is important not confusing W/C with moisture content. The latter describes
aggregate`s state of reference: over-dry, air-dry, saturated surface-dry and wet
(Lamond & Pielert, 2006), which defines the water quantity possessed by the
aggregate, able to be transferred or subtracted to the paste. Three basic factors
influence this percentage: absorption capacity, effective absorption and surface
moisture (Mehta & Monteiro, 1993).

The specific gravity is influenced by density and consequently mixture proportioning


for weight – volume relationship. There are two types of gravity, specific and bulk
specific. The former ASG, exclude the porosity in the calculation, while the latter
includes it. BSG is important for mixture proportioning, but it is not directly related to
concrete performance. On the other hand, bulk unit weight influences both mixture
proportioning and concrete performance. Bulk density measures the volume of
graded aggregates in concrete including the voids and particles between them (Bond
et al., 2006; Lamond & Pielert, 2006).

Water
The water classes are: potable water, recycled via industrial concrete procedures,
underground water, natural surface and industrial waste water, sea water or brackish

12
water and sewage water. Portable does not need testing, underground natural
surface and industrial waste water should be tested, water from concrete industry
should meet the requirements, sea water can be used only for un-reinforced
concrete based on chloride content levels, and finally sewage water is not suitable
for concrete. Water origin influences its chemical properties, which are based on
chloride contents, sulphates, alkali, and other harmful chemical contaminations. In
particular, the maximum chloride content of mixing water can be estimated by BS EN
1008:2002 (Table 1).

Table 1: Requirements for Chloride content due to type of concrete (BS EN1008:2002)

The sulphate content is expressed as SO42 and cannot exceed the limit of 2000 mg/l.
Furthermore, in the presence of alkaline reactive aggregates in the mixture, water
should be tested for its alkaline content and in particular, the sodium dioxide content,
which should be less than 1500 mg/l. In addition, a harmful contamination test
should be carried out, in order to test if sugars, phosphates, nitrates, lead and Zinc
meet the requirements (table 2).

Table 2: Requirements for harmful contaminations due BS EN1008:2002

Admixtures
Admixtures are materials added during or after mixing. There are many types of
admixtures, such as chemical affecting cement paste`s hydration rate during setting
and hardening, plasticizers are able to reduce the surface water of the mixture, air-
entrain reducers extending the concrete`s durability and mineral ones e.g.
Pozzolans, which contain silica fume able thermal cracking reduction.

13
Strength of Concrete
The behavior of concrete in construction is quite different from that defined by
theoretical calculations, due to the fact that the loads are not being applied in the
same fashion as in the standard tests, and also the size, shapes, dimensions, age
and service of the structure parts are completely different.

The classification of concrete is based on its compressive strength, which is


estimated by experimental cylindrical specimen of diameter 150 by 300 mm and
cube 150x 150 mm in 28 days. The characteristic compressive strength f ck is found
by examining the 95 percent of specimens with a value near the expecting (EKOS,
2000).

2.2.2. T HR E E P OI N T B E N D I N G T E S T

Fig. 6: Three point bending test representation (Asferg et al., 2006)

TPBT is used in order to examine the behavior of structural elements in bending,


when subjected to external load (Eurocode 2).

The calculation of such model is based on the procedure below:


My
1. stress estimation:  
I

PL3
2. deflection estimation:  
48EI
3.  Fx  0  V1  V2  P

4.  M1  0

14
PL h

bh 3 2  12 PL and young
5. M  Pl / 4 , y  h / 2 , and I  (rectangular). Thus   4
12 bh 3 8bh 2
12

FL3
modulus is given by: E  . The load is applied in the mid-span of the beam in
48I
the compression zone. The maximum moment is concentrated in the midspan tensile
zone (fig. 7).

Fig. 7: Load definition and maximum moment point

Due to our problem only the experimental results in an under-reinforcement concrete


beam based on Concrete Society will be taken. The beam tested is designed to
behave in a ductile manner, because steel yields before concrete crashes and the
main tension reinforcement consists of 3T12 high yield steel bars, while no steel is
provided in the compression zone and no stirrups are available to withstand shear.

Fig. 8: Load-deflection plot describing elastic response when load is applied and not applied, where
beam returns to its initial state with a slight deflection (Concrete Society UK)

15
Initially the load applied approximates the serviceability limit and some visible cracks
are observed. Due to removal of the load, the beam returns to its original position
with a slight residual deflection around 2mm (Fig. 8), something that describes the
linear elastic behavior of the beam.

Fig. 9: Crack Propagation due to load increase (Concrete Society UK)

When the load is increased beyond its service limit, the load-deflection plot defines a
level out response (Fig. 10), with little increase of load for increasing deflection. This
indicates that steel has yielded providing ductility to the section. Finally, as the beam
is loaded further, failure occurs due to tensile shear (fig. 11).

Fig. 10: Load-deflection plot of under-reinforced beam

Fig. 11: Under-reinforced beam failure due to shear

16
2.2.3. C R A C K I N G DUE TO E U R O C ODE 2
With a first look on Eurocode 2, concrete cracks can be categorized in three basic
classes: non-structural, thermal and structural. Non-structural cracking stems from
effects with no significant impact in structural integrity. Thermal cracks occur due to
variation of temperature in response to dynamic movement or during curing process.
Structural cracks are small failure state flaws or defects. Eurocode considers
cracking as an effect generated by low tensile strength of concrete, which has been
exceeded after the application of induced action.

Non-structural cracking is of the same nature as structural. However, it is considered


as surface flaws, unable to provide us with signs of collapse. As a matter of fact,
their effect in concrete can be prevented by improved reinforcement detailing and
careful concrete handling during pouring, curing, placing and setting.
There are two main perspectives for concrete cracking, both based on tension
strength-strain characteristics. The first is tensile strength-restrain extend, which
claims that if the material is unrestrained, it expands unboundedly and without any
hindrance from any agent. However, in most cases the material is restrained and the
tensile stress develops in the vicinity of a certain restraint point, leading to cracking.
In addition, concrete behavior varies in respect to time leading to modulus of
elasticity decrease and affects its compatibility with environmental conditions during
its lifespan.

The second perspective is linked to the tensile strain capacity of concrete. Concrete
is characterized by three states: a) liquid observed during pouring and setting,
exhibiting very high tensile strain capacity and deforming easily, b) solid, which
occurs during hardening followed by a drop in tensile strain, and c) over-time state in
which the concrete`s state is affected by its moisture content due to environmental
conditions leading to amenable deformation in combination with moisture absorption.
Thus, this perspective supports the idea that the higher the tensile strain capacity the
less the possibility for crack formation.

Either way, this type cannot be considered as structural cracking, due to the fact that
it affects some of the materials` properties, due to environmental influence,

17
something that can lead to water ingress, caused by concrete spalling and steel
corrosion.

The basic idea of BS EN 1992-1-1 is that concrete is made for cracking in order to
design reinforcement efficiently. Cracking is an aspect of aesthetics and durability in
structural systems, according to British Standards. For insitu concrete exposed to
environmental conditions, the maximum crack development permitted is 0.3mm in
width. BS EN 1992-3 for marine structures locates the limits for cracking width to
range between 0.05-0.2 mm. It is very important to mention here that only the visible
cracks are taken into consideration, which cannot provide an explanation for non-
visible crack formation, which influences concrete`s mechanical properties. Thus,
crack limits should be further researched.

Structural Cracks
The most known cracking concepts are water ingress and steel corrosion, cracking
due to settlement and shrinkage cracking. The former is connected to moisture
absorption of concrete through the cracks leading the moisture to steel, which
corrodes, degrades and finally swells causing concrete spalling. These phenomena
can be prevented by restraining the level of chloride to certain limits, good detailing,
placing of reinforcement and covering with good quality concrete and absence of
electric stray currents. The limit of crack width of 0.3mm is described by EN 1992-1-
1, for preventing moisture and decreasing the risk of steel corrosion.

Cracking due to settlement is called plastic cracking and describes settlement of


concrete during placing and curing. It initiates at concrete setting vibration stage, in
which vibration water rises, while heavier components (aggregates and cement)
settle towards the bottom. This phenomenon is also called “bleeding” and triggers
plastic cracking in the element`s surface. The introduction of admixtures in the mix,
reduce bleeding, improve detailing and designing for restraints.

Another type of plastic cracking is shrinkage cracking, which mostly appears in large
slabs or walls` pouring, two to three hours after setting. Unprotected setting leads to
quicker drying triggered by wind. This phenomenon can lead to a state called
“crazing”, which is described as a multi-cracking and brunching phenomenon of

18
newly formed cracks, leading to a polygonal shape surface cracking. By floating and
providing finish to concrete, this type of cracks can be graded out. In order to prevent
crazing, technicians control the time of concrete setting leading to higher tensile
strain capacity. The key aspect here is the reduction of evaporation rate to prevent
shrinkage. One idea is to add a concrete compound within a few hours of it being
placed and another is to cover the surface with a sheeting of polythene and hessian.

Finally, the last category is thermal cracking, which occurs due to thermal
movement. Concrete should be able to move regardless the temperature fluctuations
throughout its lifespan. Otherwise, tension cracks can be generated. Mostly large
structures are subjected to this type of cracking. Additional reinforced and movement
joints could be foreseen splitting the structure. This type of cracking can be
prevented by applying a straight crack, where reinforcement is not present and
consequently forming a construction joint.

The critical loading at which cracking occurs first time corresponds to a minimum
value of K, given by K  f ctm /(W 0.5 ) , where W is the serviceability loading and fctm

is concrete`s tensile strength. At frequent combination with load stages, cracking


degree ζ is calculated as a quasi-permanent combination.

Cracking according to Eurocode 2 is examined as a deflection and emphasizing in


functionality and aesthetics. The point at which the cracking occurs is defined by
moment‟s induction and tensile strength of concrete. Once a slab cracked stiffness is
permanently reduced cracking should be limited to a range in which functionality or
durability of structure is aesthetically acceptable. In addition Eurocode present
cracks as a normal phenomenon, in which propagation is permitted without
attempting to manage its width, when it is not impairing structural functionality. The
calculated crack growth limit W max (table 3) is emphasizing only in the nature, class
of structure and cost based on traditional crack arrest and repair methods.

19
Table 3: Recommended values of wmax (mm) (Eurocode 2– Table 7.1Ν)

3. L I NE A R AND NON-LINEAR F R A C TU RE M E CHA NI CS C ON CE P TS

3.1. L I NE A R -E L A S TI C A N D E L A S T O - P L A S T I C THE O RI E S

3.1.1. G R I F FI TH ` S F R A C TU R E S T RE N G TH
Griffith (1921) described brittle fracture based on energy failure criterion (Kumar &
Barai, 2011). According to this theory, the concrete matrix can be characterized as
an “energy reservoir” (Karihaloo, 1995), in which the body possesses stored energy
(potential energy). Either the potential energy must be reduced to the material`s
surface energy (Kumar & Barai, 2011) or the crack itself acts as an dissipation
mechanism propagating until the entire stored energy is flown out (Ramesh, 2003).

Another important factor introduced and examined by Inglis (1913) was fracture
strength. The main drawback of Inglis was the high material experimental strength in
respect to theoretical strength, assuming that the concentrated stress in the vicinity
of a small crack can cause development of fracture (Ramesh, 2003). This was
solved by Griffith (1921) assumption called size to plate strength dependency.
Griffith energy approach identified two new surfaces requiring energy absorption for
system strain energy to develop. Small cracks exhibit greater strength than longer
ones.

To validate his approach, he carried out a set of experiments on pre-cracked and


annealed vessels, to eliminate residual stresses, glass tubes and spherical vessels
subjected to internal pressure. He estimated  a  0.25  0.28MPA m . He also
considered that surface tension of glass is a linear function of temperature, the
extrapolated surface tension values of glass fibers are between 1110 oC and 745 oC

20
to room temperature, E of glass is 62GPa and γ is 0.54 N/m. The calculated fracture

2E
strength based on the aforementioned parameters was  a   0.15MPa m .

Inglis Calculation
Inglis used the atomic lattice, crystalline solid, in order to calculate the theoretical
strength of a material (Fig. 12).

Fig. 12: Lattice Model for theoretical strength calculation (Ramesh, 2003)

The fracture strength here is calculated using the force separation law. The
interatomic force-separation law can be approximated by a function exhibiting three
properties: an initial slope that corresponds to the elastic modulus E, a total work of
separation (i.e. area under the curve) that corresponds to the fracture energy, γ s and
a maximum value that represents the inter-atomic cohesive force, amplitude (fig. 13).

Fig. 13: Properties of calculated data of lattice model (Ramesh, 2003)

Due to the aforementioned, the force-separation law can be expressed:

21
1/ 2  1/ 2 
E s    Eb   x 
 ( x)    sin     , where b represents the equilibrium of inter-
 b    s   b 
 
atomic spacing and x denotes the displacement from the equilibrium separation
distance.

  1 / 2 
Eb  x 
Based on that concept, and because sin       1 , Inglis calculated the
  s   b 
 
1/ 2
 E  s 
maximum theoretical strength using  th    and assume from his
 b 
experiments that  s  Eb / 40 , that for many materials  th  E / 6  33GPa.

From the aforementioned calculation of Inglis, it is obvious that something is going


wrong with the theoretical value obtained.

Griffith Calculation for Fracture Energy (1921)


Griffith made several tests on glass fibers and observed that “the critical breaking
stress of a cracked plate is inversely proportional to the square root of the crack (fig.
 2  E  s 
14), expressing in plane stress:  f   .
   

Fig. 14: Actual Test Results of Griffith (Karihaloo, 1995)

We can see that even Griffith found experimental fracture strength for glass 11000
MPa, which is much greater than the bulk glass strength of 175 MPa. This paradox
was solved using size-effect, which was previously discussed by Leonardo da Vinci

22
(1452-1519) studying wires of constant diameter, who observed an inversely
proportional relationship between strength and length, which can be summarized as
“if the length is shorten the strength increases”. Also Le Blanc (1839) observed that
based on the same diameter, long iron wires exhibit lower strength than short wires.
Therefore, the “size effect”, was expressed as an increase in strength as the
dimensions of the test specimen decreases.

Griffith developed the idea of size-effect further in order to accommodate it in his


experimental data, expressing it as “crack-size effect”. He observed the relationship
between the theoretical strength and the fracture stress in the presence of a crack
f b
0.5
and formulated it as:   .
 th  a 

Using this formula he found that the experimental fracture strength of 11000 MPa
was corresponding to a crack size of 0.025 mm. Thus, the basic idea was that when
the specimen size decreases, the strength increases, and if the length is very small
high strength values are observed.

The critical breaking stresses for different conditions are described:

 2  E   s  1/ 2
Plane Stress:  f   .
   
 2  E   s  1/ 2
Plane Strain:  f   .
  (1  v 2 )   

   E   s  1/ 2
Pennywise Crack from Irwin idealization:  f   .
 2(1  v 2 )   

23
3.1.2. S T R E S S I N TE N S I T Y F A C T OR

Fig. 15: Modes of fracture based on SIFs (Kumar &Barai, 2011)

Irwin (1957) described SIFs (Fig. 15) able to cover the majority of the NLFM
concepts (Kumar & Barai, 2011) by defining the amplitude of the stress singularity
and elastic stress fields in the region of the crack-tip, which are assumed to be
equally distributed (Karihaloo, 1995). As a matter of fact, the stress field increases
proportionally with SIF (Kumar & Barai, 2011).

SIF differs from stress concentration factor (SCF), which is the ratio of actual over
the mean nominal stress in a subjected geometrically defined discontinuity (Kumar &
Barai, 2011). Structural parts subjected to tension or flexure, exhibit a stress field in
front of the crack tip, in which a singularity is observed and governed by SIF, based
on an inverse square root relationship characterized by the distance from the crack-
tip (Mohammadi, 2008). In this direction, the critical stress or displacement condition
is described by the SIF critical value for every failure mode (Kumar & Barai, 2011).

In addition, local yielding is observed in many materials manifesting the role of the
singularity and consequently describing the size of the plastic zone, which is related
to SIF (Mohammadi, 2008; Kumar & Barai, 2011). Thus, the local stress quantity is
controlled by K     r .

The elastic state of stress in the vicinity of the crack can be described by:

 
 ij  r 0.5 K I fijI ( )  K II fijII ( )  K III fijIII ( )  higher  order  term. Therefore, KI, KII,

KIII are equal to:

24
K I  lim  yy  2r K II  lim  xy  2r K III  lim  yz  2r
r 0 r 0 r 0
 0  0  0
KI can be simplified further:

a   3
K I  lim  yy  2r  lim 2r 0 cos (1  sin sin )   0 
2r 2 2 2
r 0 r 0
 0  0

Thus, for failure mode I: Failure Mode 2:


K II   3
KI  3  xx  cos (2  cos sin )
 xx  cos (1  sin sin ) 2r 2 2 2
2r 2 2 2
K   3
K   3  yy  II sin  cos  cos
 yy  I cos (1  sin sin ) 2r 2 2 2
2r 2 2 2
K   3
K   3  xy  I  cos (1  sin sin )
 xy  I  sin  cos  cos 2r 2 2 2
2r 2 2 2
 zz  v( xx   yy ) 
v( xx   yy )  Plane  Strain
 zz    xz   yz  0
 0  Plane  Stress
For  the  displaceme nt  field
For  the  displaceme nt  field K II r  
ux  sin  k  1  2 cos 2 
KI r   2 2 2 2
ux  cos  k  1  2 sin 2 
2 2 2 2 K II r  
uy   cos  k  1  2 sin 2 
KI r   2 2 2 2
uy  sin  k  1  2 cos 2  uz  0
2 2 2 2

Failure Mode 3-Tearing: In Polar Coordinates for mode I:

25
KI  
 rr  cos (1  sin 2 )
 xx   yy   zz  0 2r 2 2
K  
K     I cos (1  sin 2 )
 xz   III sin 2r 2 2
2r 2
K  
K   r  I  sin  cos 2
 yz  III cos 2r 2 2
2r 2

For  Shear  Mode  II :


For  the  displaceme nt  field
K 5  3 3
ux  u y  0  rr  II ( sin  sin )
2r 4 2 4 2
K III r  K 3  3 3
uz  sin    II ( sin  sin )
2 2 2 2r 4 2 4 2
K 1  3 3
 r  II ( sin  sin )
2r 4 2 4 2
Mohammadi (2008) and Karihaloo (1995).

3.1.3. B RI T TL E AND Q U A S I -B R I T T L E M A T E RI A L
Due to Griffith assumption the whole strain energy is equal to surface energy
released during crack growth development and rate of potential strain energy.

 U se W U se U s U  U sP U 
U s  U se  U sP . Thus,         .
 a a a a a a a

 U se W U 
For ideally brittle material U SP is not participating:      2 s .
 a a a
Where γs is the surface energy and number 2 shows the existence of two surfaces
upon fracture.


Griffith crack growth energy is given by G    2 s and thus crack growth


criterion can be expressed as  2 s  or    2 s  . (Karihaloo, 1995;

Mohammadi, 2008).

The strain energy required for a central crack of length 2a, in order to be introduced
in an infinite plane under plane stress conditions is given by:

26

  4  0.5u y ( x)dx
0

Due to Mushkelisvili and Westergaard, displacements can be expressed as:

2 x 2 a 2  x 2  2 2
u x ( x)   u y ( x)   , thus   .
E' E' 2'

  2
Therefore G    and the energy consumption for crack growth of value
a '
   2  2
da is given by:    Gx   x   . The critical stress (Mohammadi,
0 0  '  '

2 E ' s
2011) is:  cr  and the stress criterion for the stress intensity factor Kc is
a
given by: K c   cr a and unbounded crack extension occurs when K=Kc.

On the other hand, for quasi brittle materials


U 
da
 
 2  se   sP , where  se and  sP are

the elastic and plastic work related to crack development respectively. Thus,
2 E ' a p s
 cr  , where ap is a correction factor and UΓ=4aγs.
a

3.1.4. CTOD
CTOD is a time dependent non-linear EPFM parameter applying to very ductile
materials. According to Wells (1963) CTODc should be greater of CTOD for stable
crack growth (Kumar & Barai, 2011). Wells (1963) mentioned that CTOD should
include the calculation of the plastic region developed in front of a crack tip, which is
based on maximum load and the effective critical length of the crack. This direction
was also followed by Irwin and Dugdale.

In particular, Irwin described an effective length aeff, which includes both the length of
2
1 K 
the crack and plastic crack-tip zone given by: aeff  a  ry  a   I  (Fig. 16)
2  y 
 

27
Fig. 16: Effective length concept of Irwin (Mohammadi, 2008)

Therefore, Irwin found that:


2
4 KI 4 G
t    Plane  Stress
 E y   y
2
 1  4(1  v 2 ) K I
t     Plane  Strain
 3  E y (Karihaloo, 1995)

Furthermore, it is important to mention that Irwin includes a correction, in which a


greater in length plastic zone rp is assumed (rp>ry, Fig. 16).

Barenblatt-Dugdale introduced a CTOD factor assuming an effective crack length of


2
 K 
aeff  a  rp  a   I  (Fig. 17). This model requires zeroing of stress singularity
8   y 

at the crack tip, therefore CTOD≠0 (Karihaloo, 1995).

K I2 G
t    Plane  Stress
E y y

1 2 K I2
 t    (1  v )  Plane  Strain
2 E y (Karihaloo, 1995)

28
Fig. 17: Barenblatt-Dugdale`s effective crack length assumption (Mohammadi, 2008)

It is quite interesting to compare both estimations in order to find their difference.


rp 8
Thus:   0.81 . Although the theoretical background of these models is
 2
different, the calculated 19% difference found is not so high. The reason is that Irwin
is based in equilibrium force equations of elastic stress fields, while Dugdale adopt a
more fracture mechanical concept. Thus, for small-scale yielding, solutions are valid,
except for the cases including plastic zones.

Furthermore, CTOD can be related both to Gf, energy release rate and SIF or KIC. In
NLFM  t   tc (, T ) defines the dependency of the critical value on strain and

temperature.

3.2. N ON - L I N E A R F R A C T U R E M E C HA N I CS C O NC E P T S
NLFM is governed by crack propagation laws describing the material as inelastic and
assuming that the FPZ is located in a frontal to crack tip region as it was previously
described by LEFM (Shi, 2009). According to Irwin (1958) the length of FPZ in
combination with tensile strength and fracture energy can define the fracture
E  GF
toughness of concrete lch  . In general, one can say that LEFM does not
f t2
apply to concrete due to four main reasons: a) brittleness concept`s incapability to
describe fracture, b) fracture toughness concept, characteristic FPZ length and size-
effect and c) tension softening based on cohesive material properties such as type of
aggregates and grading.

29
There were numerous scientists such as Kaplan (1961), Glucklich (1963), Romauldi
and Batson (1963), Zaitsev (1971), Swamy (1972), who tried without success to
introduce LEFM concepts to concrete. Kesler et al. (1971) described the lack of
brittleness concept to provide sufficient solution first, while Walsh (1972)
experimentally defined a slope greater than that defined by LEFM (Karihaloo, 1995).
Size-effect appears in LEFM at the crack-tip, with the nominal strength to be in
inversely proportional to A1/2 and varies based on crack size, because the procedure
is punctual (Karihaloo, 1995). The ratio of lengths between crack and structure can
either be determined by LEFM for homogeneous material assumption or LEFM for
heterogeneous assumption (Bažant, 1997).

Fracture toughness (KIC) exhibited a variation from the previous results (Brown,
1972; Shah and McGarry, 1971; Walsh, 1972; Swamy & Rao, 1973; Higgins and
Bailey, 1976; Mindess & Nadeau, 1976; Walsh, 1976; Gjorn et al., 1977; Rossi et al.,
1984). The latter was also supported by Leicester (1969) who introduced a power
size-effect and assumed that stress singularity is greater in sharp cracks than the
sharp notches, but according to Bažant (1997) he should have also included lch in his
investigation.

Another interesting observation was the variation of brittle fracture speed and
consequently of softening curve due to the type of aggregates, which influences the
homogeneity of the mixture (Nallathambi, 1986; Karihaloo, 1995). Also, size effect is
tested based on tension softening plot, supporting the idea that LEFM laws govern
fracture after yield limit surpass in large scale structures (Karihaloo, 1995). In
response to softening curve shape and fracture speed, Bache (1986), Hillerborg
(1983), Hillsdorf & Brameshuber (1991), and Brulwiler et al. (1991) summarized their
experimental data providing a table for FPZ length of some materials (Table 4).

Table 4: FPZ length for some materials (Karihaloo, 1995)

30
Another aspect discussed by Bažant (1983; 1984) was energetic size effect, defined
by a characteristic length lch , observed mostly on large scale-structures. There is no

relation to Weibull`s theory (Karihaloo, 1995), due to the absence of lch from the

latter and the introduction of an exponential power law physical function


exp( D)  nd / m , defined by Weibull`s modulus m and geometrical similarity nd
(Bažant, 1984). In addition, Bažant (1984) described scaling effect as a sequence of
complex response variation of crack growth direction, which was insufficient to
provide acceptable solution due to incapability of energetic effect capture in concrete
(Bažant, 1997).

3.3. E XI S TI N G M O D E L I N G AND C O M P U TA T I O NA L T E CH NI QU E S
There are many theoretical concepts developed in order to accommodate concrete`s
behavior and crack definition. Here we will categorize the concepts based on the
area of application to: a) theoretical fracture concept b) numerical procedure, c)
concrete simulation and c) reinforcement idealization.

Based on concrete theoretical fracture concept we have:


1) Cohesive zone models including, model of Barenblatt (1959) and Dugdale (1963),
fictitious model (Hillerborg et al., 1976), Crack Band Model (Hillerborg and Bažant,
1976; Bažant and Cedolin, 1983; Bažant and Oh, 1983), Bridging stress models
(Bao and Suo, 1992; Li and Maalej, 1996; Needlemann, 1990; Ferro, 2002; Stang et
al., 2006)
2) Microplanes (Bažant & Ozbolt, 1990)
3) Particles (Bažant et al., 1990)
4) Lattice (Van Mier, 1997)

Cohesive Zone Models (CZM)


The first CZM was developed by Barenblatt (1959) and was further improved by
Dugdale et al. (1960; 1963) by extending the cohesive zone length (Fig. 18) allowing
the increase of plastic zone up to the point where stress reaches the ultimate yield
strength σy. The most important fact of cohesive cracks is the absence of SIF, which
happens due to the finite value of stress and the smooth face closure criterion
(Karihaloo, 1995).

31
Fig. 18: Model introduced by Dugdale and corresponding σ/σy rate (Mohammadi, 2008)

c 
Smooth closure criterion  1  cos( ) . (Mohammadi, 2008)
a 2 y

Crack Tip opening displacement is given by:

 y (1  ) ln(1  ) 1 (Karihaloo, 1995)


8 c c
Dimensional: CTOD 
' a a
CTOD  (  ' )
Non-dimensional: CTOD*  (Karihaloo, 1995)
8   y

If the CTOD reach CTODc the fracture is unbounded. In the case of infinitesimal
CTOD and FPZ length, tensile force will be infinitesimal and Wc can be found using

   0.5     y  Wc (Karihaloo, 1995).

The second CZM is the fictitious crack, which was developed by Hillerborg (1976)
based on Barenblatt (1959) and Dugdale et al. (1963) model with an introduction of
FPZ characteristic length concept. The important characteristics of this model are: a)
the elimination of SIF in the crack-tip something that facilitates the analysis, b)
fracture energy estimation based on geometrically defined shape of softening curve
(Hillerborg, 1980; Petersson, 1983; Malvar & Warren, 1988; RILEM, 1985, 1988)
using finite element methods, the results of which are included in the RILEM (1985),
and c) Wc based on fracture energy, which is difficult to calculate (Nallathambi &
Karihaloo, 1991a) (Fig. 19)

32
Fig. 19: Fictitious Crack Model and fracture energy estimation method (Karihaloo, 1995; Shi, 2009)

The last CZM is the band model (Hillerborg et al., 1976; Bažant & Cedolin, 1983a;
Bažant & Oh, 1976) in which Γ is defined by FPZ morphology, based on aggregates`
size variation (Fig. 20).

Fig. 20: Concepts of crack band model (Mohammadi, 2008)

E ft 2
COD: GF  hc (1  )  , where hc  3  d max (maximum aggregate)
E1 2 E

2  GF
Strain:  u  . (Karihaloo, 1995; Shi, 2009; Mohammadi, 2008)
f t'  hc

The third CZM is the stress bridging model used mostly to examine multi-scale
phenomena and especially fiber and aggregate intersection via bridging, based on
cohesive multi-scale laws for bonding and de-bonding including disorders or pre-
existing flaws. Cohesive laws have high dependency on scale and crack opening
(Needlemann, 1990) (Fig. 21), presented by atomic interaction breaks, grain`s
dissociation and exchange of stress between the bonds of the micro-crack. The
numerical implementation is the same as that of FCM and is characterized as
smooth closure model governed by the rate between tip and cohesive law fracture
energy (Cox & Marshall, 1991).

33
Fig. 21: Multi-scale laws and crack opening representation (Needleman, 1990; Strang et al, 2006b)

The fourth CZM is the crack band model, in which the whole FPZ and its inelastic
deformations are gathered in a certain band width in the front of the crack
(Mohammadi, 2008) (Fig. 22). This is the basic difference in respect to the
aforementioned three models, which are characterized by the concept of a lumped
FPZ in an infinitesimal parallel and longitudinal crack region based on a stress
deflection law governed by a softening phenomenon (Mohammadi, 2008). Crack
band theory concept was developed by Hillerborg & Bažant (1976), Bažant &
Cedolin (1983), Bažant & Oh (1983) for pure mode I and assumed an
heterogeneous material subjected to strain softening, due to micro-cracks located
between the aggregates.

Fig. 22: Crack band representation in front of crack tip (Bažant & Oh, 1983; Karihaloo; 1995)

Numerical Procedure
The most known numerical methods used are FEA and XFEA. However, other
existing methods are boundary integral technique, boundary elements, Galerkin
method, partition of unity etc.

Partition of Unity
PoU (Melenk & Babushka, 1996) in combination with elemental enrichment of
discontinuous field (Belytschko & Black, 1999), can format a minimal re-meshing

34
style for simulating arbitrary aligned crack geometry in respect to mesh options.
Moës et al. (1999) proposed an improved version of this method, aiming to
geometrically intersect crack from mesh using a specific independent mesh style.
Dolbow (1999) introduced a model able to provide local enrichment of displacement
field based on fundamental partition of unity concepts.

EFGM
Galerkin method is a clear mathematical procedure based on meshless schemes,
which is quite distant to finite element concepts (Zi & Belytschko, 2003). Galerkin
method is mostly used in combination with smeared crack and discrete approach.
The latter combination simplifies the analysis but lacks in accuracy (Sagaresan,
2011). Galerkin element-free method does not have a requirement for crack
propagation criterion and can be used for representing mode I and mix-mode
situations under dynamic loading (Belytschko et al. 2000).

FEM
Finite Element method is used in the majority of the models for concrete
representation. Petersson (1983) developed an FE model, which provides good
approximation and is included in the RILEM (1985). Carpinteri (1980) described the
concrete`s brittleness number and energy number (Carpinteri, 1986), while Bažant
(1984) introduced energetic size effect and Carpinteri (1986) introduced energy
number. Bosco et al. (1988, 1989a, 1989b) provided experimental support to
previous models of Bazant and Carpinteri, by testing the behavior of sixty un-cracked
low-reinforced concrete beams in bending (TPBT), noticing the ability to predict
minimum reinforcement requirement for all concrete classes due to the brittleness
number.

Common factor of FEM models was fracture energy and the representation of
softening phenomenon, which contributes to gradual strain energy dissipation.
Karihaloo (1995) stated that the linear softening constitutive law (RILEM, 1985) is
able to provide us with good approximation for the majority of practical examples.

Nallathambi & Karihaloo (1991a) defined many limitations to the aforementioned


models, in respect to CTOD estimation. In general, estimation considered

35
cumbersome, due to lack of accurate stress localization in the crack-tip, which is
connected to meshing capabilities. This constitutes the major drawback of FE
procedures and the major advantage of XFEM.

XFEM
The fundamental concept of XFEM is based on PoU (Melenik & Babuska, 1996;
Duarte & Oden, 1996) and generalized finite elements. Initially, this numerical
procedure was used for LEFM (Belytschko & Black, 1999), based on the concepts of
minimum meshing and enrichment of the discontinuous field for the crack definition
(Mohammadi, 2008). The method describes two displacement fields, one continuous
and one discontinuous:

u ( x)   Ni ( x)ui   N j ( x) ( x)e j


i1 j1
( Mohammadi, 2008)

Where I is the node set, Νi(x) is the nodal shape function, ui is the displacements of
nodes, ε is the enriched nodes, ej is the additional DOFs and ψ(x) is the enriched
parameter (Zi & Belytschko, 2003).

Moës et al. (1992) and Dolbow (1999; 2000) called this method as XFEA due to its
geometrical elemental independency on meshing options. Daux et al (2000)
introduced a model in which cracking initiates in two holes and propagates due to
branching. Sukumar et al. (2000) simulated crack propagation in 3D plane based on
elemental enrichment, while Areias & Belytschko (2005a) extended the method for
simulating multi-cracking and branching.

Some of the most important properties of XFEM were the level set method, which
was able to simulate localized crack growth (Stolarska et al, 2001; Belytschko et al.,
2002; Ventura et al. (2003); Zi et al, 2004; Budyn et al., 2004; Bordas & Moran
(2006), Stolarska & Chopp , 2003) , the fast merching method (Sukumar et al., 2003;
Chopp & Sukumar, 2003) and signed distance function for arbitrary discontinuities,
which is a simplification of level set method (Pais, 2012).

36
An alternative method for arbitrary distributed crack simulation was introduced by
Song et al. (2006) including shear band propagation. The scheme based on XFEM
elemental rearrangement of XFEM and introduction of additional DOFs is called
phantom nodes. This method is one of the most used, and numerous models were
developed based on it, such as those of Jirásek & Belytschko (2002), Rabczuk et al
(2000), Olsen (2012) and Asferg et al. (2007).

Concrete Simulation
Based on computational simulation, three are the main models for concrete: damage
model (DPM), smeared crack model (SCM) and embedded crack model (ECM).

DPM
This model took its name based on a scalar damage variable, able to influence
microcracking. There are two classes of scalar models: isotropic used for concrete
classes C30 to C40 and anisotropic models based on geometrically orientated
shapes due to loading history Kmiecik & Kaminski (2011).

Frichant et al. (1999) suggested a simplified model in respect to older anisotropic


models, which does not include crack closure. Baker & de Borst (2005) introduced
another anisotropic model based on thermodynamic concrete`s failure when
subjected to high stresses and periodic thermal variation, based on heat flow and
entropy evolution for surface failure. Kondelka & Krejci (2008) assumed a thermo-
hydro-mechanical concept for concrete`s lift bridge deck layer, which was subjected
to shrinkage due to weather conditions` variation and sun radiation triggering
damage evolution.

SCM
These models can be used for both plain and reinforced concrete, when
reinforcement is simulated as a rebar. In general, these models are able to examine
low confining pressures due to monotonic strain, while including isotropic yield
hardening for cases of compressive stress dominance and crack detection
mechanism for surface cracking failure. The post-failure response is defined due to
elastic damage concepts requiring elastic definition of materials and absence of
localized orientation.

37
Nguyen & Chan (2005) proposed a multi-smeared model to describe reinforced
concrete failure of a bridge under earthquake and cyclic loading. The results show a
significant overestimation of damping and unloaded stiffness. Vecchio & DeRoo
(1995) also proposed a smeared crack model, which was however unable to
simulate concrete cylinder tension splitting and was highly influenced by general and
lateral dilation.

ECM or Brittle Model


Due to Jirasek (2000), embedded crack models are a combination of discrete and
smeared models with initial representation of strain and displacement discontinuities
based on FE procedures. This contributes to strain localization improvement and
elimination of stress locking (Mohammadi, 2008).

Many SCMs are capable of modeling a displacement jump either as a strong or


weak discontinuity. Weak discontinuity examined by Ortiz et al (1987) to simulate the
shear band region and consequently improve strain field localization. The model,
however, was unable to model the band zone in high accurancy. In addition,
Belytschko et al. (1998) and Sluys & Berends (1998) examined SCMs finding them
capable of width dependency reduction in contrast to the size. For strong
discontinuity, Dvorkin et al. (1990) described an improved model, which was later
adopted by Lotfi & Shing (1995) adding a cohesive traction law.

As Jirasek (2000) mentions, there are three embedded crack classes: kinematically
and statically defined (Jirasek, 2000), statically optimized symmetric (SOS),
satisfying the requirements of kinematics and traction law, but leads to static
spurious stresses (Belytschko, 1998; Sluys & Berends, 1988; Jirasek, 2000).
Furthermore, there are two more classes: kinematically optimized symmetrical
elements (KOS), expressing an insufficient traction approximation, and statically and
kinematically optimized asymmetric elements (SKON), leading to elimination of
spurious stress and size dependency of the shear band.

38
Due to reinforcement
In most of the cases, reinforcement is defined as an interaction property between
concrete and steel. In one of them, the interaction is defined as an interface model
described by Ngo & Scordelis (1967), Lundgren (1999) and Østergaard (2003).

Zubelewicz & Bažant (1987) studied the interface of composites with brittle
aggregates, noticing that the force-displacement plot of a rigid particle arbitrary
defined system under normal tensile strength exhibits a sudden force drop to zero.
Overlapping and connection between aggregates is absent and interaction is applied
when the distance between them reaches a limit based on largest aggregate size,
3dmax.

Crack growth is a node by node procedure based on maximum yielding stress


criterion. Lundgren (1999) and Østergaard (2003) examined orthotropic bonding of
bond slip zone, while Dirk Nielsen et al (2006) using the pull out procedure in mode
combination assumed a value for bonding and de-bonding stress. Karihaloo (1995)
examined a pull-out technique for axisymmetric headed anchors.

Two are the basic drawbacks of interface elements. The first is the complexity of the
re-meshing procedure and many studies such as those of Carol et al (2001), Prasad
& Krishmoorthy (2002) and Yang & Chen (2004, 2005) examined coarse meshing
models, or models with simple re-meshing requirement. In addition, Xie & Biggers
(2006) examined the mix-mode situation and simple meshing using VCCT method,
virtual crack closure method (Shi, 2009). The compatibility of interface elements with
FE was quite low. Nevertheless, the same does not happen with XFE, which
provides higher compatibility (Mohammadi, 2008).

The second drawback of interface elements is the interaction in the bond slip zone.
Interface models assume an excellent cohesion connection between concrete
aggregates and rebar. This is a drawback, which should be examined further or
represented as an additional interaction property in the interface model. It is also
important to define a small interface region, because the interaction is localized in a
small vicinity of the rebar (Abaqus manual).

39
3.4. XFEM F OR C ON C R E TE
As mentioned before, CCM was introduced by Barenblatt (1959) and was further
improved by Dugdale et al. (1963) and Hillerborg`s et al. (1976) FCM including FPZ
and lch . According to Zihai (2009), Hillerborg et al. (1976) also described a

numerical procedure for crack growth prediction and mix-mode failure for multi-
cracking phenomena, which was nevertheless questioned by Bažant & Gambarova
based on lack of accuracy in geometrical definition of direction and shear. In
addition, they proposed the shear-lag phenomenon as an alternative to crack
representation during initial cracking development (Bažant & Gambarova, 1980),
also examined by Kristek & Bažant (1987). Other related models were those of van
Mier (1989) and Maji & Shah (RILEM, 1989) providing explanation for mix-mode
phenomena.

In 2003, Zi & Belytschko proposed a partly cracked XFEM scheme providing


elimination of blending. Mergheim et al. (2005) described a model based on the
constitutive law for inelastic crack behavior and simulation of both straight and
curved orientation.

Moës & Belytschko (2001) studied randomly distributed cracks in order to


understand the concept of SIF zeroing and absence of remeshing. For crack
propagation and damage evolution J-Integral factor was used to improve simulation
in combination with FEM. The latter and Zi & Belytschko`s (2003) model for static
cohesive crack-tip element were adopted by the majority of computational software.
The Zi & Belytschko (2003) model used LST and quadratic six-node elements in
combination with MAXPS for damage evolution.

Mergheim et al. (2005) proposed an XFEM CZM able to represent unbounded crack
growth, due to localized enrichment of cut elements and extrapolation of
displacement DOF. The model describes the material as inelastic by applying the
constitutive law for the interface and it is able to define both straight and curved
discontinuities. Xiao & Karihaloo (2006) proposed a model based on SAR and MLS
method, to facilitate the crack tip field enrichment. This study also examined the

40
interaction between enrichment field modification and size of domain with the
meshing based on second order LSQ.

Asferg et al. (2005) described a cohesive XFEM model based on pseudo-node


enrichment, and an interface surface for bond zone. The major drawback of his
method was the absence of damage evolution criterion, which restrained his analysis
to simulate a seam crack. Mougaard et al. (2011) improved the previous model of
Asferg et al. (2006) applying second order meshing based on LST and introducing
Rankine`s failure criterion (1850).

Cox (2008) examined enrichment dependency, accomplishing the smooth


development of a cohesive crack by equal stress side representation. The model
was based on a dynamic step able to provide cyclic loading representation affecting
the characteristic crack orientation.

3.5. A S FE R G ET AL. (2006) M O DE L C O NC E P T


Asferg et al. (2007) mentioned some drawbacks in the model of Zi & Belytschko
(2003). Firstly, there was not an equal stress representation in both sides of the
discontinuity. Secondly, that the enrichment should also be applied to the sharing
nodes of the cut elements and not only on those being cut due to shifted sign
function. In particular, Asferg et al. (2007) described an additional enrichment
scheme using the superposition method of cracked elements and shape functions
based on pseudo-node method. This method helped him in the representation of
sub-triangles in which the discontinuous displacement field vanishes inside the CST
element (Adaptive Sub-triangulation). In addition, he stated that CSTs were
incapable of modeling equal stresses in both ends and that a propagation criterion
for crack growth should be used, something that Zi & Belytschko had already
introduced.

Based on TPBT, Asferg et al. (2006) used the linear constitutive law for the
estimation of the area of strain softening curve as described by RILEM (1985) and
Karihaloo (1995) and in contrast with bi-linear law adopted by Zi & Belytschko
(2003). Asferg et al. (2007) compared the different simulations: one based on a

41
partly cracked scheme and one based on a fully cracked scheme, concluding that
the latter produces a smoother load-deflection plot (Fig. 23).

Fig. 23: Comparison of simulated models` load-deflection plots (Asferg et al., 2007)

The first issue is based on post-failure response of fully cracked scheme,


characterized by a zig-zag variation due to loading, which can be justified due to
localized elemental stresses taken in consideration in respect to tensile material`s
strength. As a matter of fact, Asferg et al. (2006) mentioned that this post-failure
behavior of the tail can be avoided using the mean overall stress values found.

Summarizing, Asferg et al. (2006) further analysis suggestion includes: multi-


cracking phenomenon, equal stress representation in both sides of the crack,
branching of crack and cracking control and bond slip accuracy between concrete
and reinforcement.

Mougaard et al. (2011) improved Asferg`s et al. (2006) model by: a) changing
meshing settings using second order LST, b) verification of equal stress
representation in both sides of crack tip, c) introduced a fracture criterion for damage
evolution based on MAXPS, d) described three surfaces in the model, one for
interface element, one for crack propagation definition (fully cracked domain) and
one for the upper part (partly cracked). The last two changes were: e) the bi-linear
constitutive law for fracture energy estimation and f) is the plotting of load-deflection
diagram based on mean stress value to avoid zig-zag curve`s tail orientation.

42
4. T E S TI N G AND R E S U L TS A N A L YS I S
4.1. M O DE L P A R A M E TE R S A N D ANALYSIS DE S C RI P TI O N (A B A Q US )
4.1.1. U N I TS A N D G E O ME T R I C A L D E FI NI T I O N
Abaqus is unit less therefore we need to follow a standard unit definition procedure
(Table 5).

Table 5: Unit`s definition based on analytical systems of measuring

The system chosen for our implementation is SI. One should be very careful in
introduction of units, because otherwise the model will be unable to provide realistic
results. The most difficult unit to define is that of the energy, which needs to be in
160
Joules. Thus, G f  160 N / m  160 J / m 2   2133.3J .
0.15  0.5

Fig. 24: Geometrical Definition of the Model (Asferg et al., 2006)

For the geometrical definition (Fig. 24) of the model a 2D planar deformable shell is
used. The concrete will described initially as a bi-material by dividing its area in two
rectangulars. The first describing the plain concrete bulk and the second an interface
including a rebar element as an embedded region for bond zone definition. The
crack will be introduced as a wire, while the load and restraints as concentrated
forces, following the description of Karihaloo (1995) for crack band model.

43
4.1.2. M A TE R I A L ` S P A R A ME TE R S
Concrete is a quasi-brittle material (Karihaloo, 1995), which behavior is based on
inelastic mechanical properties and can be represented in three ways: smeared
crack model, damaged plasticity model and cracking model.

In order to have compatibility with the results given by Dr. Asferg the most suitable
model is the DPM. This is due to the fact that both brittle cracking model and
smeared crack model have several limitations. Kmiecik & Kaminski (2011)
mentioned that DPM is based on explicitly physical parameters and mathematical
concepts, which can be described in depth, while the bond slip and dowel action are
introduced in the model via tension stiffening definition corresponding to load transfer
across cracks through rebars.

The major drawback of brittle cracking model is that crack initiation and growth
should be simulated as an interaction between the two faces of the crack (Shi,
2009). This means that the crack is based on higher level of assumption that in
plasticity models. In addition, brittle cracking is a more convenient method for ideally
brittle materials, because is dominated by tensile cracking and compressive behavior
is assumed as linear elastic (Abaqus Manual).

Damage Plasticity Model


This model is a modification of Drucker-Prager (1952) strength hypothesis. Here we
need to define parameters connected with plasticity such as dilation angle,
eccentricity, fb0/ fc0, K, viscosity parameter, and also compressive and tensile
behavior of concrete.

Plasticity Parameters
Dilation Angle: this angle determines the inclination of failure surface towards the
hydrostatic axis and defines the performance of concrete under compound stress
(Fig. 25) (Kmiecik & Kaminski, 2011). It is always defined by the plot of concrete
strength under biaxial stress, and in most of the cases is taken as y= 36 to 40 o. Here
it is taken equal to 38o.

44
Fig. 25: Dilation angle determination based on biaxial stress plot (Kmiecik & Kaminski, 2011)

Eccentricity
Inside the stress plane a variation is observed in the plane meridians. Typically,
these meridians are described as curves, while in DPM are described as hyperbolic
forms (Kmiecik & Kaminski, 2011). This shape variation is adjusted via a potential
plastic eccentricity, which describes the distance of segment between the hyperbolic
vortex and asymptotic intersection in the center of hyperbola (Abaqus Manual).
Therefore, eccentricity can be estimated as the ratio between tensile and
compressive strength. Practical examples and Abaqus default assumes that
f t  f c / 10 and thus e=0.1 (Fig. 26).

When e=0 the approximation is described as the classical Drucker-Prager


hypothesis applied when the meridional plane takes an approximately straight line
2/3
0.3 f c
form. Eurocode 2 assumes that f t  0.3 f c
2/3
and thus e   0.3 f c1 / 3 . It is
fc
obvious here that with default assumption a tensile strength of 3.5 MPa correspond
to C35/ 45, while using Eurocode formula corresponds to C40/ 50, which is a class
higher than the default assumption. This is very important due to the fact that
eccentricity is always less than 0.1, in particular e  0.3 f c1 / 3  0.3  40 1 / 3  0.0877 ,

while for default e  0.3 f c1 / 3  0.3  35 1 / 3  0.0917 . Maybe this difference seems

45
small, but in the reality gives you an error of 14% and a completely different class of
concrete increasing the cost of productivity.

Fig. 26: Eccentricity determination described as the distance between hyperbolic vortex and
asymptotic intersection in the center of hyperbola (Kmiecik & Kaminski, 2011)

Failure under biaxial compression f b0 / f c0

Generally, defines the point where concrete is subjected to failure under biaxial
compression, which is also the ratio of the biaxial to uniaxial strength. Due to Kupler
(1969) the uniaxial compressive strength f cc  1.16248  f c approximated via elliptic

equation. Abaqus default is  b0 /  c0  1.16 .

Strength Parameter K
The modifications of Lubliter, Lee and Fenves observed the dependence of failure
surface on K. K is always higher than 0.5 and when K=1, failure surface is described
as a circle in the deviatoric cross-section (Heirany & Ghaemian, 2011) (Fig. 27).

Fig. 27: Plot circle for describing failure under biaxial compression for determination of strength
parameter K (Heirany & Ghaemian, 2011)

46
Abaqus recommends K to be taken as 2/3= 0.667, which defined by the strength
criterion of William & Warnke (1975).

Viscosity Parameter of Concrete


Due to Hocevar et al. (2013), the viscosity parameter is defined by the viscosity
agent used as a concrete chemical admixture. The “plastic viscosity” is determined
by rheological tests and is commonly taken as zero or 0.002 (Fig. 28)

Therefore, the values of plasticity options are:

Table 6: Plasticity options determination based on calculated data

The compressive behavior of concrete is determined using the yield stress and
inelastic strain. In order to find these, an analytical procedure should be developed:

Fig. 28: Strain-strain curve for estimation of elasto-plastic properties of concrete damage plasticity
model (Abaqus Manual)

1. elim  1  exp(  c )  1  e 0.5  0.3934 , f cm  f ck  SD  40  8  48MPa Thus,


f
80
 c  elim  f cm  18.8832MPa , determination of elastic compressive stress
determined by Majewski`s formula.

47
2. Ec  (1  dc ) E0  0.75  37400  28050MPa , degradation factor taken as 0.75, which

is between the highest and the lowest degradation limits 0.65  dc  0.85 .

3. Ultimate strain is given by:  cu  0.004  0.0011 [1  exp  (0.0215  f cm )  3.2919 103 .

4. Inelastic strain:  cin  (3.2919  0.549)  103  2.787e  3

pl  
 0.549 10 3  1.14e  3
0.75
5. plastic strain:  c  2.787 
 0.25 
6. Uniaxial compressive stress:

 c  (1  d c )  E0  ( c   cpl )  0.25  37400  (2.787  1.14) 103  15.39945MPa


7. Effective stress:
c
 E0  ( c   c )  37400  (2.787  1.14) 10 3  61.5978MPa
pl
c 
(1  d c )

Furthermore, compression and tension behavior of concrete should be determined.


Therefore, for compressive behavior two yield stresses and two inelastic strains will
be introduced as tabular data, while for tensile behavior yield stress and fracture
energy will be introduced (Table 7).

Table 7: Compressive and Tensile behavior determination based on tabular data

4.2 M E S H R E FI N E M E N T S T U D Y FOR N OT C HE D U N -R E I NF O RC E D B E A M IN

TPBT
4.2.1. I N TR O D U C TI O N
The plot used for mesh refinement study is developed based on load to point load
displacement as described by Malvar &Warren (1988). Firstly, we will use a coarse
mesh and we will find the resultant force (RF) given by Abaqus. In order to pick the
right value from the diagram, we need to cross out the points experiencing stress
singularity.

48
As we mentioned before, we develop our mesh refinement study using three
meshing styles: coarse, medium, and fine (Fig. 29).

Fig. 29: Mesh schemes used for mesh refinement study

The refinement method is a method developed to understand and examine the


characteristic strength of a material and plot the load-LPD diagram in order to
compare the accuracy of different meshing styles.

4.2.2. R E S U L TS AND D I S C U S S I ON

Fig. 30: Load-Load Point displacement plot for refinement study

Linear Elastic Region


The first region of the analysis is the elastic region. All three models are able to
represent the linear elastic response of the material with the same accuracy despite
the variation of the defined properties. This region limit is the concrete elastic yield
limit, which is estimated quite the same for three models.

49
Plastic Region
In fine mesh the plastic region is readily observed leading to a peak point, where
concrete exhibits its maximum strength. Furthermore, this point can be characterized
as the point of maximum stored energy (Karihaloo, 1995), due to the fact that
concrete behaves as a reservoir of energy (RILEM, 1988; Ramesh, 2003), which will
be released through cracking procedure.

Post-Peak Response
In this region, the stored energy of body is dissipated due macroscopic mechanisms.
Due to these mechanisms a tension-softening response is observed in concrete
corresponding to a load decrease with parallel increase of displacement (Karihaloo,
1995). With that in mind, the response estimated is distributed smoothly during the
load reduction. Both medium and fine mesh can be characterized as accurate.

Finally, in coarse mesh there is a region in which a sudden load loss of load and
displacement is observed. This phenomenon is called snap-back phenomenon
(Malvar& Warren, 1988; Karihaloo, 1995) (Fig. 31).

Fig. 31: Load-load point deflection plot describing the sudden load drop (Malvar & Warren, 1988)

Generally, refinement leads to higher results accuracy, but further experimental trials
will need to be carried out in order to understand carefully what really happens in
concrete.

50
4.3. P O S T F A I L U R E T E N S I O N S O F TE N I N G C U RV E AND E NE R G Y I NV E S TI GA TI O N
4.3.1. I N TR O D U C TI O N
As we said before the fracture energy limit is given to the system (160 N/m) and
defined as a straight linear relationship between tensile stress (3.5 MPa) and crack
displacement (W c). Here we will compare four equilibriums defining the constitutive
law using a fine meshed symmetrical beam model (Fig. 32). The equivalent stress
resultant is given by S22 output contour, and it is calculated by the 75% of the
maximum value of every colour contour region.

Fig. 32: Refinement performed for crack region

The investigation will be performed through the point load displacement to tensile
stress correspondence. The four approximations for fracture energy will be: a
straight linear, bi-linear, exponential, and power law. Verification of the results will be
discussed independently and in the last step, there will be an analytical plot
combining the calculated plots followed by extensive discussion of the modeling.

4.3.2. L I N E A R R E L A T I O N

Fig. 33: Linear Constitutive Law representation (Karihaloo, 1995)

The linear plot describes the fracture energy (Fig. 33) as a relationship given by:

51
G f  0.5  f t Wc .Where it is obvious that there is a linear correlation between crack

 W 
displacement and fracture energy. In addition,   f t  1   is able to provide us
 Wc 
with information for the variation of the ratio between stress and crack displacement.
 f  
By re-arranging to stress equation to crack displacement, we take: W  Wc  t .
 f t 

Fig. 4 Stress-applied displacement relation plot and stress-cracking displacement plot based on linear
constitutive law in Abaqus

4.3.3. B I -L I N E A R R E S P O N S E

Fig. 35: Bi-linear constitutive law based on Malvar & Warren (1988) calculated using Gf=160N/m

The bi-linear response is a different approximation method for fracture energy


calculation. According to Malvar & Warren (1988) there are two stress classes
defined by different energy equations. In particular:
15
Gf   f t Wc , which contributes to a critical crack displacement given by:
54

52
54 G f 54  160
Wc     16.546  10 5 m . It is obvious that critical crack
15 ft 15  3.5  10 6

displacement is greater than that calculated using linear approximation. ft/3=1.1667


and 2x16.457/9= 3.655 (Fig. 35)

The two stress classes are:

W  1   
  f t  (1  3 )  1 f   f  W   Wc  1   
Wc  t t  3  ft  
3
   and thus: 
3  W  1   7   

   f t  1  
   f t  W  Wc  1  
7  Wc  3
 3  ft  
From these, the diagram calculated in Abaqus gives:

Fig. 36: Results Calculated in Abaqus due to bi-linear constitutive law

The tail of the diagram leads to a high cracking displacement due to the fact that,
while the “kink” point is defined, the end of second linearity is not defined, something
that happens to tri-linear approximation (Spring, 2011). Even with the use of the
formula provided by Petersson`s bi-linear assumption (1981) the fracture energy is:

  ft  wi  1  Wc  keeping the predefined kink coordinates 1  f t and


1 1
Gf 
2 3
2
W1  Wc tested by Li Tongchun et al. (2012) and the corresponding cracking
9
18G f
 16.457 10  2 m.
2 1
displacement is: 2G f  ( f t  f t )Wc  Wc 
9 3 5 ft

53
4.3.4. E XP ON E N TI A L F O R M
w f t

Gf
The exponential form is given by   f t  e , which by solving the equation and

Gf  
re-arranging the equation for W, we have: W   ln   . From values of the
ft  ft 
stresses the plot of tension softening is the following:

Fig. 37: Results Calculated in Abaqus due to exponential constitutive law

4.3.5. P OW E R L AW
The last method used to determine the tensile softening behavior is the power law.
This law has mathematical and physical importance due to the fact that it is able to
provide an exponential type of magnitude in a power of a significant testing factor for
concrete (Easley & Kleinberg, 2010).

For example let`s take y  x cons tan t and y  (cons tan t ) x . Both are positive and turn
to zero asymptotically, however when y=0.5x, the value of y cuts in the middle due to
the increase from x to 1. On the other hand, when y=x-2, x increases from one million
to one million+1. This means that y will be multiplied with a value less than 1 eg.
Million/ million+1= 0.989. In addition, by increasing x to trillions then y is multiplied
with a value even closer to 1, something that does not happen with exponential due
to the fact that it is always multiplied by ½ no matter the of x is (Niwa, 2003;
Tsoukatos, 2008).

54
Fig. 38: Power Law Described by Malvar & Warren (1988); RILEM (1988)

However, the most important fact in this formula is the dependency of a concrete
factor called m exponent, which is the direct measure of the concrete intrinsic
brittleness referring to the initial angle, first linearity, of post-peak curve. The smaller
the m the steeper the initial drop and more brittle the material (Karihaloo, 1995). We
(m  1)  G f 1.11  160
assume here m to be equal to 0.11. Wc    4.613  10 5 m and
m  ft 0.11  3.5
W variation is given by:

log w ( x)  0.11, where x   Wc m
ft

4.3.6. D I S C U S S I O N AND C ON C L US I O NS

Fig. 39: Calculated results using Abaqus 6.12 and Abaqus manual results for lower properties C40
(3.2.11-11)

55
Fig. 40: By introducing also the power law in our results and comparing with Hillerborg (1981)

Discussion
The four representations of constitutive law give clear evidence of the error
estimation of fracture energy. Fracture energy is already determined for the current
concrete model (Asferg, 2006), and in order to find the critical crack displacement
equations of RILEM (1988; Malvar & Warren) used.

Tension softening is described by Hillerborg et al (1976) as the procedure occurring


immediately after the surpass of tensile strength leading to crack propagation. The
propagation follows a path characterized by a combination of strength decrease and
corresponded crack width increase, in contrast with traditional mechanics
emphasizing to unexpected failure occurring followed by a sudden drop of tensile
stress to zero (Roesler et al, 2006). Due to theory of stiffness, the stress-deflection is
quite difficult to be obtained, because during the material testing it is important to
have enough high machine stiffness, which otherwise affects the fracture
phenomenon (Cabral et al., 2012).

Petersson et al. (1981) mentioned that the choice of the tension softening shape is
clearly based on the individual analyst, due to the fact that even a rough estimation
can be expressed based on rectangular area curve (Rossmanith, 1993). Generally
the choice of the shape will not affect much the non-linear analysis.

Linear Curve
Hillerborg et al (1979) suggested linear curve as the best choice between
rectangular and trapezoid shapes, due to the fact that it has high compatibility with
tension tests and FEA. Concrete is subjected to linear elastic behavior until ft is

56
reached. In addition Evans & Marathe (1968) mentioned that a linear shape is able
to interpret both the growth of existing cracks and the initiation of new crack profiles.

Fig. 41: Rectangular, trapezoid and linear constitutive laws described by Hillerborg et al. (1979)

Bi-linear
As we can see, bi-linear curve exhibits a fixed point, which is called “kink” and is
estimated to be at a location with coordinates (ft/3, 2W c/9). Furthermore, another
point is the length of bi-linear shape`s tail. Petersson et al (1981) described a small
tail in contrast with Rokugo`s et al (1989) experimentally obtained greater in length
tail (Rossmanith, 1993) (Fig. 42).

Fig. 42: Bi-linear consistutive laws described by Rokugo et al. (1989) and Petersson et al. (1981) (Shi,
2009)

In particular, Rokugo et al (1989) described the bi-linear shape based on fracture


energy in mode I, given as an assemblage of two critical crack opening
displacements.

57
Fig. 43: Bi-linear constitutive law based on experimental, Petersson et al (1981) and Planas et al.
GBF (1992b; 1993) describing the tension softening phenomenon in concrete

Due to Planas et al. (1992b; 1993) GBF method can be used in comparison with
Petersson`s experimental procedure (Fig. 43). This method states that four important
parameters are defining the bi-linear softening curve: a) tensile strength estimation
based on concrete cylinder specimen`s splitting, b) fracture energy estimation based
on RILEM recommendations for fracture tests in notched specimens, c) abscissa w
parameter for defining center of gravity based on far end load deflection method and
d) initial horizontal segment intercept W I for peak load determination in small
specimen.

Conclusions
Conrelissen et al. (1986) and Hordijk et al. (1987) assumed three different shapes of
softening diagram (Rots & De Borst, 1988). Rots (1986) emphasized the need to
examine carefully fracture energy concept based on softening diagram and
Cornelissen et al (1986) and Hordijk et al (1987) proposed an additional non-linear
exponential curve assumed to be the best approximation. However, it was commonly
observed that 3D effects can exist during testing introducing further structural
drawbacks in the analysis.

In order compare the aforementioned four different tension softening models we can
divide the process in five steps. This will be done by assuming that the bi-linear
model will be modified in order to evaluate our results (Fig. 44).

58
Fig. 44: Fracture Energy Idealisation based on bi-linear curve for theoretical outcomes

The first step is called first linearity, and describes the crack width values between
the maximum tensile strength and the kink. The second step describes the
coordinates of the kink point; the third step is the second linearity describing the
values from the kink point until the critical width point, the fourth step is the critical
width and the final step is the fracture energy estimation due to area.

First Linearity
Due to Abaqus results the linear softening law provides an overestimation in contrast
with the three others. Power describes under-estimation, while bi-linear and
exponential describes a moderate response. This can be expressed as steeper
angle (Karihaloo, 1995), which contributes to the assumption that power law
describes a smaller angle than the other three close to exponential form. This
method of steep angle is described also by Kumar & Barai (2009).

Kink Point
The second step is the kink point. Due to Park et al. (2008) the kink point is assumed
to be in a stress ratio between 0.15 to 0.33. Roesler et al. (2007) describes the value
of 0.34 for C32 with Gf=56.6 N/m. Tenq & Shah (1985) obtained a value of 0.267 for
C40 with Gf= 30.1, while Karihaloo & Nallathambi (1985) a value 0.342 for G f=87.8
N/m. According to Malvar & Warren (1988) formula, the kink point is assumed to be
in coordinates (ft/3 and 2W c/9), which contribute to 1.667 MPa and 3.665 10-2 m,
producing a ratio of 1.667/3.665=0.45. This ratio based on the aforementioned kink
point tests can be characterized as a good assumption.

59
Second Linearity
It is important to mention that the preferred model of C40 with such high fracture
energy (Asferg et al., 2006) is quite rare example. Fracture energy of 160 N/m
corresponds to C40 with maximum aggregate size g= 32 mm (Karihaloo, 1995).
There is plenty of literature reference examining how aggregate size varies load-
deflection diagram (Karihaloo. 1995). Thus, the estimation of fracture energy is size

dependent and G f  a F ( f c' )0.7 (Hilsdorf & Brameshuber, 1991) where aF is an

empirical factor associated with maximum aggregate size. Also due to Gettu & Shah
(1992) the concept of characteristic length can provide with further information about
E ' GF 37.4 160
the type of concrete. l p    488.49mm which corresponds due to
f 't2 3.52

normal concrete (Table 4). This also justifies the low intrinsic brittleness ratio,
m=0.11, which corresponds to characteristic length and describes a slightly more
brittle character of concrete. This is due to the fact that maximum aggregate size is,
32mm, (Karihaloo, 1995). As Karihaloo (1995) states “brittleness is being reduced by
increasing maximum aggregate size but increasing with relative increase of
concrete`s strength.

The second linearity is described as the curve with the kink as the initial point and
the critical width as the terminal. Thus, it is dependent to the critical width based on
the previous calculations. The theoretical assumed curve should be between power
and exponential form, due to three reasons. First the linear model increase the area
under the curve a lot leading to overestimation and very steep angle, bi-linear
describes an under-estimation fracture energy for the first half of the line and a
overestimation for the other half, and third the power is steep for a fracture energy of
160 N/m, which means that the area under the curve should be greater.

Critical Width
Both critical width described by the bi-linear constitutive law (Rokugo et al., 1989;
Malvar & Warren, 1988) describes an extensive tail length. The same does not
happen with the model of Petersson et al. (1981), which describes a more moderate
critical point. Malvar & Warren (1988) linear law overestimates the coordinates of the
critical point, due to the fact that second linearity is overestimated.

60
On the other hand power and exponential forms (Malvar & Warren, 1988) describes
quite the same critical width, which seems to be able to give a good approximation of
the model.

Fracture Energy
This is connected to the area under tensile softening curve (Karihaloo, 1995; RILEM,
1988). This is helpful in order to make an approximation for critical width. As we
mentioned before the bi-linear law of Rokugo et al. (1989) and Malvar & Warren
(1988) gives an extended length in the tail, which something that is not right (Roesler
et al. 2006).

Also, one can consider the results of Karihaloo & Nallathambi (1989) and see that for
C40 with fracture energy of 87.5 gives a critical crack width of 4.25 10-2 m. Results
found here estimates the critical width between 4.645 and 4.565, therefore an
average of 4.605 10-2 m. Therefore, we can assume that a constitutive law
representation between power and exponential plot will be the most suitable choice
(Fig. 45)

Fig. 45: Suggested graphical representation of constitutive law

4.4. R E I N F OR C E D B E A M IN TPBT
4.4.1. I N TR O D U C TI O N
Concrete as a material exhibits high compressive strength, while tensile strength is
about ten times less (Eurocode 2; BS). In TPBT, this drawback of concrete is more

61
clearly observed due to maximum moment developed in the tension midspan
(Fanella, 2009). Figure 46 provides a theoretical description of crack initiation and
propagation in plain concrete.

Fig. 46: Plain concrete subjected to three point bending (Fanella, 2010)

Due to externally applied loads the tensile stress in the member increase. When the
value of this stress reaches the tensile strength of the element cracking occurs
(Exadaktylos, 2006). Unreinforced members are subjected to sudden failing once the
first tension cracks form, because there is nothing to prevent its propagation
(Kosmatka et al., 2002). Many literature works support the idea that by introducing
steel bars will overcome this problem (BS EN 1-1-1992). However, due to
microscopic study of concrete this is not the solution for cracking (Nemati, 1997).

In particular the introduction of steel bars inside the bulk increase the elastic
properties of concrete, while crack formation is not prevented (Fig. 47). As a matter
of fact, by indirect improvement of concrete the material is not becoming
microscopically homogeneous (Singh et al, 2009). In fact only the maximum tensile
strength of concrete is improved, due to reinforcement introduction.

Fig. 47: Reinforced Concrete and crack initiation (Fanella, 2010)

According to Karihaloo (1991) the cracking occurs even before the load is being
applied to the element, in particular during and curing. The bulk includes flaws,
micro-cracks and higher level of porosity due to open and closed pores (Karihaloo,
1991), which are able to decrease further the theoretical properties of concrete

62
(Exadaktylos, 2006). According to Mehta & Monteiro (1993) correct aggregate
grading is one of the most important factors for decreasing the level of porosity.

From the aforementioned it is clear that microscopic parameters of concrete can be


improved highly during the production. Fiber reinforced material such as FRC,
CardiFRC (Karihaloo, 1991) constitute a good idea on how to develop highly
improved materials. The basic ideas behind these improved materials are: a)
elimination of opened and closed porosity, b) improved grading and elimination of
entrained air in the mix, c) high resistance fibers introduction for crack opening
bridging and d) improvement of production and application methods.

All in all, the point of the analysis here is to model the bond slip or interface between
the concrete and reinforcement as one element able to provide higher mechanical
elastic properties to concrete (Lundgren, 1999). The procedure will be discussed
thoroughly and the results will be compared with the load-displacement plot obtained
by Asferg et al. (2006) and Mougaard et al. (2011). Further modification in meshing
and boundary conditions will be introduced for tension softening examination due to
Rots (1986) Rots & De Borst (1988).

4.4.2. M E S H D E FI N I T I O N AND S TE E L E M B E DDE D R E GI O N


Due to refinement study and its previous conclusions we can assume that the
development of highly refined mesh (Fig. 48) will provide us with more information
about stress distribution and prevents the singularities in the asymptotic field in the
crack region.

Fig. 48: Reinforced Concrete Beam meshing model

The method for reinforcement introduction is based on embedded region beam


modeling in Abaqus. This model is described as a constraint to the system able to

63
simulate the interaction in the bond zone. The model will be generated using
symmetry and boundary conditions are introduced as in the refinement study.

For steel determination, mechanical properties will be described such as modulus of


elasticity, Poisson ratio, density, and thermal expansion, 200 GPa, 0.3, 8050 kg/m 3
and 13 x10-6 m/m K respectively (engineeringtoolbox.com). The minimum
reinforcement provided due to Eurocode 2 is 16 , which can be determined using
section, module property and profile in Abaqus CAE. E is similar to all steels until
their yield limit is reached, while due to Eurocode 2 class 500C will be used (Fig. 49).

Fig. 49: Typical Stress-Strain plot for steel class 500C and the corresponding bi-linear simplification
(Panagopoulos & Kirtas, 2005)

Beside the elastic properties of steel, also plastic stress and strain were determined
based on the theoretical conventional yield limit f0.2= 500 MPa, where strain is 0.2%,
provided by the plot of stress-strain (Fig. 49).

As for the geometrical definition, the minimum cover calculation is based on general
equation without fire resistance Cnom  Cmin   cdev , which is 10mm due to

permitted quality assurance reduction allowance of  cdev equal to 5mm (Table 4.3

(N) (BS 8500)). The minimum tension reinforcement is given by: Requirement for
M sd 11.2  0.25
compressive: K    0.05677  K ,  0.31(500C )
bd2  f ck 0.1 0.136 2 
40000
1.5
Thus only tensile reinforcement is needed given by:
f 40000 1.15
As    b  d ck  0.0588 10 13.6   0.4906cm 2 , while the minimum
f yk 500000 1.5

reinforcement is given by As min  0.0027 10 13.6  0.3672cm 2 and the maximum by:

64
As max  0.0115 10 13.6  1.564cm 2 . Therefore, 112  (1.13cm 2 ) is the required

reinforcement.

Theoretically the reinforcement provided can withstand a moment found:


As  f yk
pr 1.13  500000 1.5
   0.1355 , which is corresponding to K=0.125 and
b  d  f ck 10 13.6  40000 1.15

40000
thus M sd  K  b  d 2  f ck  0.125  0.1 0.136 2   6.165 KNm, which corresponds
1.5
to a concentrated load in the midspan of F=24.661 KN

4.4.3. R E S U L TS D I S C U S S I ON AND C O NC L US I ON S
In this case we provided the minimum reinforcement and not the required in order to
simulate the case of under-reinforced beam. It is possible that the introduction of the
concentrated force can display high load singularity, thus an assumption should be
made for the exact value. Besides the load introduction, for the values of
displacement greater than 0.05 mm the load will be introduced by applying
displacement.

Fig. 50: Load-Deflection response for TPBT described by Asferg et al. (2006) and the one calculated
using Abaqus

As we can see from the above the load-deflection plot shows adequate results for
the simulation that we used. The model used was refined, thus the results can
constitute a good approximation of the load-deflection curve.

65
Due to Eurocode 2 there are several parameters affecting concrete deflection.
Generally, a strength increase in the provided reinforcement can lead to decrease of
reinforced area in the beam corresponding to the ultimate limit state followed by an
increase in the internal stresses localized in the reinforcement (Bond et. al., 2006).
Furthermore in order to increase the strength of concrete we need to build a more
durable, stiff, cohesive and workable concrete (Bond et al., 2006).

The most important factors influencing concrete behavior are tensile strength, creep,
elastic modulus, degree of restraint, magnitude of loading, time of loading, duration
of loading, cracking in concrete, shrinkage, ambient conditions, secondary load
paths and stiffening provided by other external or internal elements (BS EN 1992-1-
1). The summarized influence of the aforementioned is given by the deflections
simplified method, used in order to find an existing displacement to a corresponding
load and in a way to verify the existing results of our computational simulation. In
particular:

The moment MQP is 2.8 KN/m found previously in the midspan (load= 11.2 x 0.25).
To obtaining the properties of concrete we have f ctm  3.5MPa and

Ec28  1.05  Ecm  1.05  37.4  39.27GPa , creep coefficient φ taken for t0=10, thus

φ=2, for indoor conditions of relative humidity equal to 50%. Long term elastic
modulus is given: Ee ff  Ec28 /(1   )  13.09GPa . The effective modulus ratio is:

ae  Es / E ff  15.278 , where Es is elasticity of steel reinforcement 200 GPa. Next

we find the depth of the neutral axis for un-cracked condition xu:

bh 2 0.1 0.152
 (ae  1)( As d )  (14.278)(0.000113  0.136)
xu  2  2  0.0809 .
bh  (ae  1) As 0.1 0.15  14.278 1.13
After that, the second moment of area for un-cracked condition lu is given by

lu 
bh3
12
h
 
 bh(  xu ) 2  (ae  1) As (d  xu ) 2 
2
0.1 0.153
12

 
2
 0.15 
0.1 0.15    0.0809   14.278  0.000113  0.15  0.08092  3.75110 5
 2 

66
Furthermore, we need to check if the component is cracked due to the moment. In
order to do that we calculate the cracking moment given by:

0.9 f ctm  lu 0.9  3500  3.75 10 5


M cr    1.70948KN / m , 3500 N/m2= 3.5 MPa, and
h  xu 0.15  0.0809
multiplier 0.9 corresponds to exclusion of loading sequence. Thus, checking
M cr  1.70948  M QP  2.8KN / m for this section, therefore the domain is cracked

defining a zeta   1  0.5  ( M cr  M QP ) 2  0.8136 .

We calculate the depths of the neutral axis xc and lc:




xc   0.000113 15.2782  2  0.1  (0.000113  0.136 15.278)


0.5 
 0.000113 15.278 / 0.1  0.010197

b  xc3 0.1  (0.010197)3


lc   ae  As  (d  xc ) 2   15.278  0.000113  (0.136  0.010197) 2  3.9 10 5
3 3

Calculating the flexural curvature:


1 M QP 2.8 2.8
   0.8136  (1  0.8136)   5525.605
rn Eeff  lu 13.09  3.9 10  5 13.09  3.751 10 5

Calculating the total shrinkage strain:


 cs   cd   ca  395.9  75  470.9 , where drying shrinkage strain is given by:

 cd  K h   cd  0.925  428  395.9 , and nominal un-restrained drying shrinkage


0
 cd for relative humidity 50% for compressive strength of 40 MPa is 428.
0
For long term deflection  ca   as (t )   ca ()   ca ()  75 .

Curve due to shrinkage strain 1 / rs :

1 S S
    cs  ae u  (1   )   cs  ae  c  1539.6536 , where:
rs lu lc

Su  As  (d  xu )  0.000113  (0.136  0.0809)  6.2263 106

Sc  As  (d  xc )  0.000113  (0.136  0.010197)  1.652 105

1 1 1
Total curvature    7065.2586
rt ,QP rn rs

67
Quasi-permanent deflection is given by  2 1 0.5 2
QP  KL r   0.0295mm for a
t ,QP 12  7065.2586

load of 1.12 104 due to diagram, where K=1/12 and a=1/2.

Fig. 51: Projection of deflection calculated point in both diagrammatic plots

Conclusions
In both cases, the method of representing a central notched simple beam under
three-point flexural tests can provide us with a quite good estimation of the load–
deflection plot. As the symmetric representation of the beam is idealized and the
notch is define as a discontinuity in the initial geometric definition of the beam, we
can support the idea of previous literature such as Wang (2004) and Wagoner et al.
(2005), who describe that this method can provide quite good results for the Elasto-
plastic modeling of concrete, however are not so well for the representation of
tension softening even including an refined mesh.

Fig. 52: Represents the areas where the two simulations vary

68
Therefore, the entire concrete beam representation seems to provide a more
convenient tension-softening approximation. Initially the first difference is located
after the elastic limit, where the plastic behavior starts to govern. However, we can
see that the fracture point is quite the same, which contribute to underestimation of
plastic zone.

The tension softening zone is described in Abaqus results with quite different
orientation of curvature. However, this can be refined taking more trial values and re-
meshing the model. In the last part of tension-softening curve, the values seem to be
close enough and improve the approximation. Finally, the load in the last
displacement seems to have the same value in both cases.

Last but not least, the maximum deflection found for the load point, where the
serviceability load is taken as 0.7 times the ultimate, verify both Asferg et al. (2006)
and ours calculation.

69
4.5. R E I N F OR C E D C ON C R E TE B E A M C RA CK D E F I N I T I O N
4.5.1 C OH E S I V E Z ON E M O D E L
In the CZM (Fig. 53), the FPZ is lumped in the crack and is described as a stress-
displacement softening law (Mohammadi, 2011). XFEM cohesive cracks are
developed through literature extensively, thus the theoretical background is able to
provide with various experimental reviews (Moes & Belytschko, 2002; Zi &
Belytschko, 2003; Mariani & Perego, 2003; Mergheim et al., 2005; de Borst et al.,
2004; Asferg et al., 2005, 2006; Mougaard et al. 2009, 2011).

Fig. 53: Representation of fracture process zone between grains of concrete (Karihaloo, 1995)

The main property of cohesive cracks is the vanishing of SIF in the crack tip, which
contributes to the smooth crack closure. In addition, because KII<<KI only KI is taken
into account. Therefore, the conditions highlighting the behavior of cohesive cracks

are: KI=0 and due to the superposition of stress field K Icrack  K Icohesive  0 , while

 
1  KI K I2 
the corresponding crack orientation is given by   2  tan 1     8 .
4  K II 2
K II 
 
(Mohammadi, 2011)

4.5.2. I N I T I A T I ON , P R O P A GA T I ON AND S M O O TH C L OS UR E C RI T E RI ON
Due to Zi & Belytschko (2003) cracking occurs when the projection of stress in the
normal direction n of the crack is considered to be equal to material`s tensile
strength,  tip  f t . The model used in this dissertation can be categorized to

enriched model, which with the introduction of XFEM does not require re-meshing
through crack propagation phenomenon.

70
The criterion used for the initiation and crack propagation MAXPS can be described
as:   max11,  22 ,  33. This can be simulated in Abaqus and as a material

property. In particular, a damage traction separation law called „MAXPS‟ standing for
maximum principal stress criterion is defined. In this case, the degradation and
accidental failure can be modeled in an enriched element by describing a damage
initiation criterion and evolution law.

The crack direction in MAXPS is always orthogonal to the maximum principal stress
in order to satisfy the damage criterion as Wells & Sluys (2001), Dumstorff &
Meschke (2003) and Asferg et al. (2006) mentioned. The criterion can be described
  
by f   max  , where  max
0 is the maximum allowable principal stress,  max is
0
  max 

the maximum stress in the developed model inside Macaulay brackets, which intend
to interpret an small algorithm. In this case  max  0 if  max  0 and

 max   max if  max  0 . In other words the damage does not occur by
compressive stress state and is determined as a ratio able to reach a critical value.

The last definition has to do with evolution of damage, which is described as


degraded stiffness enabled by damage criterion. In order to accommodate this
damage property, a scalar variable D is introduced describing the total damage
taking place between crack surface and edges of cracked elements. D value varies
from 0 to 1 and can be displayed in output using statusxfem in contacts folder. The
scalar affects both normal and shear strength components due to:
(1  D)Tn  Tn  0
tn  
Tn  no  damage  to  compressive  stiffness
ts=(1-D) Ts and tt= (1-D) Tt, where Tn, Ts, Tt are the normal, shear, tension stress
components used in the traction separation behavior. Finally, in order to define the
damage occurred by normal and shear separation in the interface an parameter

called „effective separation‟ used given by  m    m  2  s2   t2 .

71
4.5.3. A NA L YT I C A L T E S TI N G F OR T HE G E NE RA TI O N O F THE A P P R O XI MA TE D
M ODE L
In this case the model is generated using its total geometrical definition. The
concrete`s material properties and field output definition (Table 8 and 9) are:

Table 8: Definition of concrete and steel properties in Abaqus

Table 9: Definition of field Output for crack initiation and propagation

The Rankine (1850) criterion defined previously states that yield occurs if the
maximum principal stress exceeds the uniaxial tensile yield strength (σ≤ σy).
Concrete was simulated using elasticity, plasticity, density and concrete damage
plasticity model. In the latter, the compressive and tensile behavior parameters were
0 , while 
introduced. Tensile stress is  max max is the stress defined by the user in
0 , which means that tensile
MAXPS. For crack propagation simulation  max   max

72
strength is equal to maximum tensile stress, when MAXPS defined stress is 1 based
  
on f   max .
0
  max 

Crack Propagation Geometry Tests (T1)


T11

Fig. 54: Crack propagation due to Boundary Conditions/ Uy=U2=0/ 2-point CF

Both ends are restraint to vertical movement, while the load defined in two points as
a concentrated force. We can see that cracking orientation is affected and turns,
leading to a bad approximation. This model is not acceptable.

T12

Fig. 55: Crack propagation due to BCs / Uy=U2=0/ Load as Displacement

This model describes a completely wrong direction of crack orientation due to


erroneous stress distribution corresponding to boundary conditions. This model is
not acceptable.

T13

Fig. 56: Crack propagation where BCs introduced as loads

73
This model seems to work perfectly and describe the propagation of crack reaching
the mid-height. This model is described by Karihaloo (1995) for finite element
procedures for concrete crack band model, which is a member of CZM family. This
model is acceptable.

T14

Fig. 57: Crack Propagation with BC U2=0 and Load defined as pressure

The pressure of the same load value, 27KN, is applied in the enriched area.
Although stress is concentrated in the vicinity of the crack, the latter does not
propagate at all. The second figure show the propagation of crack based in
statusxfem deformed contour, when pressure is applied and boundary u2=0. This
model is not acceptable.

T15

Fig. 58: Crack Propagation with BC as load (CF) and Load defined as pressure

Based on pressure and equivalent concentrated force for supports, crack


propagates, while stress representation in both sides is not equally distributed
leading to termination of growth. Principal stress direction changes the propagation
orientation leading to erroneous representation. The model is unacceptable.

74
T16

Fig. 59: Crack Propagation if load is applied in a mid-datum point

Here we created a partition of the initial geometry by introducing a mid-datum point.


In that point concentrated force is applied, which led to force singularity and the
crack was unable to propagate. This model it can be used for symmetric model, but
not for model with the entire geometrical definition. The model here is not
acceptable.

Conclusion
The best representation of reinforced concrete behavior is achieved using the
assumption of Karihaloo (1995) and Karagiannis & Chalioris (1999), which introduce
both constraints and loads as concentrated forces of the similar total value (action-
reaction). The model represents a cohesive crack development, therefore is a
member of CZM family.

Cohesive Behavior Tests (T2)


T21

Fig. 60: Cohesive Behavior via Interaction Property 1

Cohesive behavior specified due to the bonding mode set in surface to surface std
interaction. Crack propagates due to initial assumption.

75
T22

Fig. 61: Cohesive Behavior via Interaction Property 2

This model introduces contact property with enrichment for only slave nodes initially
in bonding. Crack propagates as the initial assumption, while this modification is not
influencing the geometry.

T23

Fig. 62: Cohesive Behavior via Interaction Property 3

The property defined for any slave nodes experiencing contact. Crack propagation
was not being affected by this modification.

T24

Fig. 63: Cohesive Behavior via Interaction Property 4

Same as before only toggling off the allowance for cohesive behavior during
repeated post-failure-contacts

76
T25

Fig. 64: Interaction as Normal-Hard Contact property

The interaction property here is introduced as a hard contact in default mode. The
crack propagation geometry is not affected at all.

T26

Fig. 65: Interaction as Normal-Penalty Linear Contact Property

Using a normal contact with penalty linear assumption for interaction property, no
significant change observed in the model in respect to T21-T25 tests.
T27

Fig. 66: Using Only XFEM crack Propagation with no Interaction Property

No interaction property is defined and the crack grows due to XFEM definition. The
cracking geometry is displayed as previously in T21-T26.

77
Conclusion
From T21-T27 models we can assume that the crack growth is not influenced by the
introduction of an interaction property. We achieved the propagation of the crack
through the height of the beam reaching the mid-height. Two things should be
determined further: crack propagation reaching the full height of the beam leading to
complete collapse and secondly the vertical displacement of the beam.

Crack Propagation for fracture to full-height of the beam (T3)


T31

Fig. 67: Pull-out Pressure in both Sides

For achieving the propagation, pull-out pressure applied in both sides of the beam,
leading to the crack propagation through the entire height of the beam. This is both
for plain concrete and reinforced concrete, due to the fact that for plain concrete,
rebar item was suppressed.

T31

Fig. 68: Pull-out Pressure in both Sides and restrain of mid-datum point

78
In this test we restrained the point in compression zone (mid-datum) and introduced
lateral pressures in both sides of the beam. As in T31 the model is not affected and
crack propagates reaching the full-height of the beam.

Conclusion
From test T3 the crack can propagate reaching the top end of the beam only if lateral
pressure is applied in both sides of beam forcing the beam to open. Due to
experimental results, this behavior is acceptable for reinforced concrete beams,
because the crack propagates until the mid-height of the beam.

The basic drawback of the model is that for reinforced and un-reinforced concrete,
crack cannot open the beam, providing displacement. As a matter of fact, this is due
to the reason that the embedded region and the intersection between concrete and
steel-rebar idealized as a new geometrical defined face. For crack propagation,
opening of the crack, and vertical displacement of the beam, the model should be
divided only in two interfaces. The one including the crack and is the enriched region
and the other that represents the concrete. Bond-slip zone is model as a constraint
therefore there is no need to define an interface.

Vertical Displacement in combination with Cohesive crack Propagation and


CMOD Tests (T4)
These tests are made in order to simulate reinforced concrete behavior near
experimental real time situation. The most important facts are that the beam
experience cracking, vertical displacement and crack mouth displacement. In
particular, this can be done by deleting the geometrical face including the steel bar
and just introduce the rebar as a constraint interaction embedded region. The model
now will be divided in two interfaces. The first on the left and on the right of the beam
using the same meshing options and

79
T401

Fig. 69: Maxps= 1, 100 Increments

Slight Propagation and no crack mouth opening displacement (CMOD) and vertical
displacement (VD).

T402

Fig. 70: Maxps= 10, 100 increments


The crack propagates in two elements and no VD and CMOD is observed.

T403

Fig. 71: Maxps equal to 100, 100 increments


The crack propagates in two elements and no VD and CMOD is observed.

80
T404

Fig. 72: Maxps= 1000, 100 increments


The crack is propagating while CMOD and VD are observed in the beam. However,
this model cannot be characterized as representative.

T405

4
Fig. 73: Maxps equal to 10 , 100 increments

VD and CMOD are greater here, while crack is not propagating. This contribute to a
bad approximation

T406

5
Fig. 74: Maxps equal to 10 , 100 increments

VD and CMOD are greater than before, while crack is unable to propagate.

81
T407

6
Fig. 75: Maxps equal to 10 , 100 increments

VD and CMOD are keeping increasing, while crack stopped completely propagating.

T408

7
Fig. 76: Maxps equal to 10 , 100 increments

The crack is not propagating at all, while CMOD and VD are slightly greater than
before.

T409

Fig. 77: Maxps= 1 and 1000 increments

The propagation is the governing phenomenon followed by a small vertical


displacement and CMOD. This is a good approximation.

82
T410

Fig. 78: Maxps= 10 and 1000 increments

This test provide us with a good approximation of our main aim, due to the fact that
vertical displacement, CMOD and cohesive crack propagation is observed. This is
also a good approximation however the CMOD is quite big due to experimental
testing.

T411

2
Fig. 79: Maxps= 10 and 1000 increments
Here propagation, vertical displacement and CMOD are exhibit quite the same type
of increase. This model cannot represent the reality.

T412

3
Fig. 80: Maxps equal to 10 and 1000 increments

83
Vertical displacement is observed, however the crack does not propagate in 1000
increments. This is not representative and also show that the increase of maximum
principal stress influence the model by making vertical displacement and CMOD the
governing outcomes.

T413

4
Fig. 81: Maxps equal to 10 and 1000 increments

This model provides a very high vertical displacement, while the propagation is
slightly less. Thus, this model cannot represent reality.

T414

5
Fig. 82: Maxps equal to 10 and 1000 increments

Definitely the CMOD is the principal influenced parameter and crack does not
propagate. This model is not representative.

84
T415

6
Fig. 83: Maxps equal to 10 and 1000 increments

The crack does not propagate and CMOD is keeping increasing. Here the maximum
principal stress is near the tensile strength of the material. This means that the
maximum principal stress criterion is working. Also in the second figure another
important thing is that stress is not equally distributed in both sides of the crack. This
model is not representing the reality.

T416

7
Fig. 84: Maxps equal to 10 and 1000 increments

Here the crack does not propagate at all. The CMOD and vertical displacements are
very high, however the representation of similar stresses in both sides of crack are
quite reasonable.

Conclusion
From the tests it is obvious that maxps criterion influence a lot the behavior of
cracking. When principal stress is near 1 the crack propagates, while the CMOD and
vertical displacement is low. When the stress is near 3.5 MPa, then the crack does
not propagate and the beam exhibits very high vertical displacement and CMOD.
Due to experimental results the crack propagation should be the governing

85
phenomenon, followed by vertical displacement contributing to CMOD. Referring to
the experimental results of Bordelon (2008), the cracking should be the governing
phenomenon and CMOD and vertical displacement are quite small. Thus, the best
approximation is given by maxps=1 for 1000 increments (T409). Further investigation
based on T409 will be performed in order to see if the crack can reach the top of the
beam, something that leads to error due to the specific load of 27 KN used.

Fig. 85: Simple notched plain Concrete beam cracking (Bordelon, 2008)

T409A

Fig. 86: Maps=0.1 and 1000 increments

The propagation is greater in length than before, while CMOD and VD are observed.
This stress corresponds to a stress 10 times greater than tensile stress.

T409B

Fig. 87: Maxps=0.001 and 1000 Increments

86
The principal stress here is 1000 times greater than tensile strength. The
propagation of the crack occurs immediately and reaches the mid-height of the
beam, just in the first 123 increments. However, after 1000 increments no significant
CMOD and VD are observed. This is not a suitable approximation.

Conclusion
When MAXPS is lower than 1 and reaching the limit of 0, the beam is governed by
crack propagation, while CMOD and vertical displacement are lower. During the
calculation, and when the beam is subjected to a 27 KN load there are some time
increments 200-850, where the crack, CMOD and VD develops slightly. This based
on the fact that in that point we have the yielding of reinforcement, where peak
strength is reached and the reinforcement breaks.

This region is called post-plastic behavior and the concrete experiencing tension
softening. The crack propagates because maximum principal stress is higher than
the tensile strength of the material, while the reinforcement breaks due to maximum
moment in compression. In other words if maxps= 0.1 means that principal stress is
10 times greater than tensile strength.

4.5.4. R E S U L TS D I S C U S S I ON
From the aforementioned tests and conclusions the most suitable model for our
approximation is the one of MAXPS equal to 1 and one thousand increments,
because it provides us with crack propagation governing the model, while also
CMOD and VD are observed. The crack propagates until reaching the mid-height of
the beam as the experimental results show (Fanella, 2008), while vertical
displacement and crack mouth opening displacement of low value are observed.

Due to calculations for the provided reinforcement, the maximum moment the beam
can withstand is 6.165 KNm, corresponding to a concentrated load in the midspan of
F=24.661 KN. Therefore, if we apply a load of 27 KN the crack in Abaqus will start
propagating before the stresses reach their highest values. This happens, because
the crack is propagating between smaller values of applied load. This load is the limit
of the reinforcement provided; Abaqus should represent the destruction of
reinforcement as an over stressed area, red colour stress concentration (Fig. 88).

87
Fig. 88: Red Stress concentration for reinforcement yielding first 100 increments

The stress concentration will be concentrated in the region near the embedded crack
and on the left and right part of the beam, where the component is subjected to
shear tensile zone failure (fig. 89) when the shear reinforcement is not present or
failed (Panagopoulos & Kirtas, 2006).

Fig. 89: Shear tensile zone failure (Panagopoulos & Kirtas, 2006; Karagiannis & Chalioris, 1999)

The tests about the maximum principal stress verified the criterion of Rankine and
Abaqus functionality. When 0< maxps <1, the principal stress is greater than tensile
strength of the beam, while when maxps ≥1, tensile strength of the beam is greater
than principal tensile stress. In the first case, cracking is the governing state in the
model while in the former; CMOD and VD are the governing states. When maxps= ft,
no cracking is observed, while CMOD and VD exhibits their maximum.

Another very important conclusion of the tests is that crack reach the mid-height of
the beam and does propagate further, even if we introduce a maximum principal
stress 1000 times greater than tensile strength, for 1000 increments. This represents
both the experimental data from Concrete Society and Fanella (2008) assumption.

One of our main aims due to Asferg et al. (2006) and Mougaard et al. (2011) is the
representation of equal stresses in both sides of crack (Fig. 90)

88
Fig. 90: Representing equal stress distribution in both sides of the crack

4.6. T HE O R E TI C A L D I S C U S S I O N OF M UL TI - C RA CK I N G AND B RA NC HI N G
P HE N O M E N A IN THE D E V E L OP E D M OD E L
One very interesting idea was presented by Geniaut (2001) describing crack
intersection and branching phenomena using XFEM. The model represented an
elastic strip including the geometry of two cracks, one vertical initiating in the mid-
span of compressive zone and one horizontal, which geometry and cracking mouth
opening was already defined. The former propagates due to stress concentration
increase due to fatigue resulted by side pressure application in both sides of the
strip, while dividing the structure. When it reaches the horizontal crack, the crack
intersects following a mix-mode fracture leading to complete destruction of the part.

Cahill et al. (2010) described an XFEM scheme for multi-cracking orthotropic


materials mostly used in aerospace engineering. The model included the
representation of five cracks with different orientation, but of quite the same general
length. Local extrinsic enrichment of FE used based on Heaviside, while crack
propagation is governed by maximum circumferential stress criterion (Wells & Sluys,
2001).

Furthermore, Paluszny & Matthai (2009) introduced a multi-crack growth model for
brittle geological media. The cracks were represented as quasi-static introduced in a
homogeneous matrix behaving in isotropic manner and governed by linear-elastic
laws. For fracture SIF approach (Atkinson, 1984) used simulating a fixed propagation
of crack every time crack-tip stress failed (Olson, 1993; Segall, 1984a). However, the
most important aspect is the representation of the matrix introducing initial flaws. The
same crack propagation criterion used for any flaw, which is not nucleated.

89
For our model, it is important to introduce multi-cracking phenomena in order to
improve the simulation and approximation of under-reinforced concrete. In particular,
it is clear that cracks initiate in the tensile zone, while failure occurs in the
compressive zone leading due to shear failure (Eurocode 2) (Fig. 91)

Fig. 91: Experimental testing representation of under-reinforced concrete beam (Concrete Society
UK)

As our model is in two-dimensions we can assume that the cross-section is unit and
compressive stress is applied uniformly due to its breath and punctual due its length
(Fig. 92). Besides the simple notch in tension zone, another crack in compressive
zone in the vicinity of load application is generated. This is due to high stress
concentration on this point trying pressing the cross-section downwards. Due to the
absence of vertical shear reinforcement and loading increase the concrete cracks in
the compression zone. Overcoming the strength limit, crack propagates without
bound and rapidly, forming cracking-growth geometry and dividing the material.

Fig. 92: Load application scheme due to breath and length

According to the aforementioned, Eurocode 2 explanation for cracks seems to be


verified. The latter states that a crack is unable for material splitting due to lack of
energy, but able to reduce its strength (Bond et al., 2006). Therefore, multi-cracking
phenomenon can provide us with an answer on why fracture energy consumption
and corresponding length is greater in parallel and slightly near to initial notch
cracks.

90
Crack propagation criterion should be the same, while the major change has to do
with enrichment of crack path. In the model of simple notched beam the enriched
field includes only the notch. For multiple cracking a greater area needs to be
enriched in order to simulate multiple cracks. In addition, mix-mode failure can be
defined to the model in order for Abaqus of simulate the growth. There is a
probability that Abaqus experiencing problems for crack branching. However, a
model introducing crack branching in TPBT is not accurate according to
experimental testing.

91
5. C ON CL U S I O N
5.1. G E NE R A L E V A L U A T I ON O F T HE P R OJ E CT
The evaluation of this dissertation is based on the initial aims described both in the
case study report submitted and in the introduction synopsis. The main idea of our
project was the examination of RC in a different frame. The static and mechanic
analysis concepts used in concrete design of Eurocode 2 and commercial FE
software do not include specific concepts of fracture mechanics.

The basic suggestion arising from this dissertation is the requirement of concrete
analysis based both on traditional structural analysis and on fracture mechanics.
XFEM numerical analysis in combination with theoretical concepts, such as damage
plasticity model, can be powerful tools for a more convenient and scholastic analysis.
The development of concrete technology and the continuous growth of laboratories`
space capacity facilitate the experimental analysis of concrete structures.

As structural engineers our basic task is to assure that our models, drawings,
calculations and analysis are representing as much as possible the real time
structural systems. In order to achieve that, we need to focus in three important
fields: a) into in-depth analysis, which can be provided in the case of concrete and
reinforced concrete in non-linear fracture mechanics, b) previous literature and
failure description of existent and non-existent structures, and c) in the continuous
structural evaluation and testing methods, which are able to provide us with data that
should be introduced to existent fractural mechanic concepts.

This dissertation used non-linear fracture mechanical concepts to examine plain and
reinforced concrete. A new model based on damage plasticity model for concrete
was used in combination with XFEM crack definition and maximum principal stress
criterion for crack evolution. The bond slip zone introduced as a dependent interface,
in which rebar is an embedded region in cohesive interaction with concrete. The
basic model idea was taken from the previous works of Asferg et al. (2006) and
Mougaard et al. (2011). However, the model proposed is based on geometric growth
of crack due to cohesive zone model, which means that it is not developed as a
seam crack able to open and progress linearly, but as a cohesive model with a zig-

92
zag orientation representing the Concrete Society testing data. This is done due to
the fact that our main idea is to represent as far as possible the real time situation.

Practically, the basic outcome of the dissertation is that for reinforced concrete the
fracture occurs due to shear. In other words, shear reinforcement plays a very
important role in the concrete fracture. Although the crack can propagate reaching
about the mid-height of the beam, this does not mean that fracture occurs due to
cracking. This idea can be both explained by Eurocode 2 and experimental
reinforced concrete testing. Cracking influences concrete`s strength, due to
aggregate debonding contributing to variation of its microstructure.

Therefore, cracking initiation and propagation can play an important role in the
microscopic characterization of concrete and the development of its strength. By
improving the microscopic characteristics of concrete we can reduce cracking
phenomena and produce improved concrete materials. In addition, based on fracture
energy assumption we can calculate the minimum reinforcement ratios and reduce
both the cost of concrete and the energy consumption during production.

5.2. A I MS S U C C E E D E D
The first aim was to simulate cracking in both plain and reinforced concrete. This
was achieved by establishing the concrete damage model, while parametric
definition was based on calculated magnitudes. The second step was to perform the
mesh refinement study based on a symmetric model of plain concrete beam under
TPBT. The latter provided us with understanding of how load-deflection plot varies
due to mesh modification. Three different meshing styles were used: a coarse, a
medium and a fined mesh. As expected, highly refined meshing styles provide
improved analysis in respect to models of lower refinement.

The second aim was the study of strain softening curve and how this is influenced by
different equations representing the constitutive law. In order to examine that, we
used again the symmetric representation of the beam subjected to four different
constitutive laws: linear approximation (RILEM, 1985; Karihaloo, 1995), bi-linear
based on Malvar & Warren (1988), Petersson (1981), and Rokugo et al. (1988),
exponential form (Malvar & Warren, 1988) and Power law (Malvar & Warren, 1988;

93
Easley & Kleinberg, 2010). Our analysis is based on bi-linear model, which was
divided in five parts: first linearity, kink point, second linearity, critical width, fracture
energy estimation. The basic outcome is the assumption that the constitutive law
should display a tension curve with values lying between power and exponential law.

The third aim was the representation of the bond slip zone according to previous
work of Ngo & Scordelis (1967), Lundgren (1999), Asferg et al. (2007). To achieve
that, the composite was divided into three interfaces: rebar element, plain concrete
and enriched region. The reinforcement was calculated based on Eurocode 2 and
added to the model as a constraint. The load-deflection plot calculated was
compared to the results of Asferg et al. (2006). The basic difference was located in
strain softening curve, where some points had different values. Nevertheless, in
respect to fracture energy, approximately the same area was covered by the plot.
The symmetric representation of the beam, supports the idea of both Wagover et al.
(2005) and Wang (2004), that the model provides good results for the elasto-plastic
modeling, but not as good for tension softening even for refined meshing schemes.
Finally, to verify our results further, we used simple deflection method of Eurocode 2,
in order to calculate a certain point on load deflection diagram, which was found in
both models.

Crack initiation, propagation and smooth closure phenomena simulated in line with
what is stated by Zi & Belytschko (2003). For XFEM crack propagation enriched
method was used, by generating an interface in the midspan region from the bottom
to top of the beam. The meshing options were based on un-structured free LSQ with
reduced integration and second order accuracy. This meshing is compatible with the
XFEM crack definition, something that is not happening with triangular elements in
Abaqus.

Meshing compatibility is clearly based on the crack definition. In particular, crack


definition in Asferg et al. (2006) and Mougaard et al. (2011) is based on seam crack.
Fracture criterion is not defined, thus the seam crack opens straightly due to the
escalation of loading. This concept is not representing real time experimental
situations, in which the crack initiates by the notch, propagating with undefined
geometry, reaching about half the length of the beam and intersect with the crack

94
initiating due to stress singularity in the point of load application. It is obvious that the
simulation should be based on requirements defined by Eurocode 2 and
experimental test data.

The XFEM crack was represented as a wire describing a cohesive interaction


property, allowing crack propagation during repeated post-failure contacts and
enabling any slave nodes experiencing contact. MAXPS fracture criterion was used
based on the previous work of Wells & Sluys (2001), which requires, the equality of
the maximum stress in the crack tip to the ultimate tensile strength of the material
and exhibits a perpendicular to principal stress orientation growth (Zi & Belytschko,
2003). However, the basic drawback was that concrete`s beam does not experience
vertical displacement and CMOD.

The introduction of a geometrical interface subjected to a different meshing style for


the bond zone, covers a part of enriched region and forbids the propagation of the
crack. Therefore, only two geometrical interfaces were designed, while bond zone
was introduced as an embedded region. This alteration allowed us to simulate crack
propagation followed by vertical beam displacement and CMOD.

Finally, from the multi-cracking literature review based on the previous work of
Geniaut (2011), Cahill et al. (2010), and Paluszny & Matthai (2009), the basic idea is
that further enrichment of field is needed for covering a greater area able to include
all the cracks that need to be simulated, while MAXPS should be defined for mix-
mode fracture.

All in all, the simulation of concrete was based on CDP model and elastic, plastic
parameters, density and thermal expansion factors definition. Fracture energy
calculation can be described as a curve between power and exponential law, both
describing a good approximation for critical crack displacement. The model used
simulate crack propagation as governing phenomenon followed by vertical
displacement of the beam and crack mouth opening displacement due to bending.
The load limit due to tensile reinforcement was 26,667 KN, and when it was reached,
the tensile reinforcement was destroyed. The analysis of the model in Abaqus tried
to simulate the real time situation through experimental results of Fanella (2008) and

95
Concrete Society. Eurocode 2 idea describing cracks affecting material`s strength,
but does not lead to complete collapse is verified. The crack propagates due to
maximum principal stress criterion always reaching the mid-height of the beam. The
beam is subjected to crack propagation during the whole loading step, while vertical
displacement and CMOD are observed more clearly in the last 200 time increments,
which basically represents the steel yielding instance. In that point, destruction
occurs, due to load point stress concentration, which leads to shear failure. In other
words, the reinforced beams always fail due to shear. Complete propagation of the
crack in reinforced concrete was achieved only by the side pulling test of the beam.

5.3. P R OP OS I T I O N F OR F U R TH E R A NA L Y S I S
The model proposed can be further investigated in order to accommodate multi-
cracking phenomena using mix-mode fracture parameters. In addition, the
investigation of both over-reinforced and typically reinforced (tensile, compression
and shear reinforcement) beams, contributes to a more convenient study of multi-
cracking phenomena and verifies previous results.

Another important aspect will be the representation of splitting concrete test in


Abaqus, in order to simulate the behavior of concrete component in contrast with the
requirements given by experimental data, which are included in Eurocode 2. It is
already known that the compressive and tensile strength of concrete is measured in
respect to cylindrical specimen. This test is able to give further verification to the
model introduced previously and also provide information on improved concrete
simulation in Abaqus.

Furthermore, experimental concrete data based on both flexural and splitting tests
should be used to improve the damage plasticity model. Elastic and plastic definition
of the material should be further analyzed by introducing tabular experimental data in
Abaqus, such as elastic and plastic stress and strains.

Last but not least, the behavior between reinforcement and concrete should be
further modified due to the fact that embedded element approach assumes perfect
bonding between rebars and concrete. Therefore, bond slip modeling needs a more
detailed interaction approach.

96
R EFERENCES
ABAQUS. 2009. Abaqus analysis user's manual Version 6.9, Dassault Systèmes [Online].
http://abaqus.ethz.ch:2080/v6.11/books/gsa/default.htm [Accessed: 07 August 2013]

Asferg, J. L. 2006. Modeling of Concrete Fracture Applying the eXtended Finite Element Method. PhD
Thesis, Technical University of Denmark

Asferg, J.L. et al. 2005. Cohesive Crack-Tip Element for XFEM. Proceedings of the 11th International
Conference on Fracture [Online]. Available at: http://www.scribd.com/doc/101476999/Cohesive-
Crack-Tip-Element-Xfem [Accessed: 19 April 2013].

Asferg, J.L. et al. 2007. A direct XFEM formulation for modeling of cohesive crack growth in concrete.
Computers and Concrete Journal 4(2), pp. 83-100

Asferg, J.L. et al. 2007. A consistent partly cracked XFEM element for cohesive crack growth.
International Journal for Numerical Methods in Engineering 72, pp. 464-485

Bažant, Z. P. 1984. Size effect in blunt fracture: Concrete, rock, metal. Journal of Engineering
Mechanics (ASME) 110 (4), pp. 518-535

Bažant, Z.P. 2000. Size effect. International Journal of Solids and Structures 37,pp. 69-80.

Bažant, Z.P. 1997. Scaling of quasi-brittle fracture: asymptotic analysis. International Journal of
Fracture 83 (1), pp. 19-40.

Bordelon, A. 2007. Fracture Behavior of Concrete Materials for Rigid Pavement Systems. MS Thesis,
University of Illinois at Urbana Campaign

Bond, A.J. et al. 2006. How to design Concrete Structures using Eurocode 2. Surrey: The Concrete
Center

Bansal, R. K. 2001. A textbook of strength of Materials. New Delhi: Laxmi Publications (P) Ltd

British Standard Institution. 2002. BS8500-2:2002. Specification for constituent materials and
concrete. London: BSI

Barenblatt, G.I. 1962. The mathematical theory of equilibrium of cracks in brittle fracture. Advances in
Applied Mechanics 7, pp. 55–129.
Bažant, .Z.P. and Cedolin, L. 1983a. Finite element modeling of crack band propagation. Journal of
Structural Engineering Mechanics (ASCE) 109 (2), pp. 69-92.

Bažant, .Z.P. and Gambarova, P. 1980. Rough cracks in reinforced concrete. Journal of Structural
Engineering Mechanics (ASCE) 106, pp. 819-842

Bažant, .Z.P. and Gambarova, P. 1984. Crack shear in concrete: crack band microplane model.
Journal of Structural Engineering (ASCE) 110, pp. 2015-2035

Belytschko, T. et al. 1996. Meshless methods: an overview and recent developments. Computer
Methods in Applied Mechanics and Engineering 139, pp. 3–47.

Belytschko, T. and Black, T. 1999. Elastic crack growth in finite elements with minimal remeshing.
International Journal for Numerical Methods in Engineering 45(5), pp. 601–620.

Bažant, Z.P., and Oh, B.H. 1983. Crack band theory for fracture of concrete. Materials and Structures
(RILEM) 16, pp. 155-177.

Cahill, L. et al. 2010. Enriched finite elements (XFEM) for multi-crack growth simulations in orthotropic
materials, IV European Conference on Computational Mechanics (ECCM) 2010, Paris, France, May
16–21, 2010.

Coates, R. C. et al. 1988. Structural Analysis. London: Chapman & Hall

Cox, J.V. 2009. An extended finite element method with analytical enrichment for cohesive crack
modeling. International Journal for Numerical Methods in Engineering 78, pp. 48-83

Dugdale, D.S. 1960. Yielding of steel sheets containing slits. Journal of the Mechanics and Physics of
Solids 8, pp. 100–104.

Daux ,C. et al. 2000. Arbitrary branched and intersecting cracks with the
extended finite element method. International Journal for Numerical Methods in Engineering 48, pp.
1741–1760.

Dvorkin, E.N. et al. 1990. Finite elements with displacement interpolated embedded
localization lines insensitive to mesh size and distortions. Computer Methods in Applied Mechanics
and Engineering 90, pp. 829–844.

EKOS. 2000. Hellenic Code for Reinforced Concrete. Athens: Ektypotiki Attikis A.E
Elfgren, L. and Shah, S.P. 1991. RILEM: Analysis of Concrete Structures by Fracture
Mechanics. UK: Chapman and Hall.

Exadaktylos, G.E., 2006, Introduction in the theory of elasticity and Fracture Mechanics, Civil
Engineering Module lecture notes. University of Crete, Greece.

European Committee for Standardization. 2005. EN 1992-2: 2005(E) Eurocode 2: Design of Concrete
Structures Part 2 Design and Detailing Rules. Brussels: CEN

Evans, R.H. and Marathe, B.E. 1968. Stress Distribution around Holes on Concrete. Materiaux et
Construction Janvier-Fevrier 1(1). pp. 57-60

Frichant, S. et al. 1999. Isotropic and anisotropic descriptions of damage in concrete structures.
International Journal of Mechanical Cohesive Frictional Materials 4, pp. 339-359

Feenstra, P. H. and de Borst, R. D. 1996. A composite plasticity model for concrete.


International Journal of Solids and structures 33 (5), pp. 707-730

Geniaut, P. et al. 2007. A stable 3d contact formulation for cracks using XFEM. Revue Européenne
de Mécanique Numérique, Calculs avec méthodes sans maillage, 16: pp. 259-275

Gettu, R. and Shah, S.P. 1994. Fracture Mechanics and high strength concrete. In Shah, S.P. and
Ahmad, S.H. (eds) High Performance concretes and application. London, Uk: Edward Arnold. pp.
161-212

Hočevar, A. et al. 2013. Rheological Parameters of fresh Concrete-comparison of rheometers.


Gradevinor 65(2). pp. 99-109

Hillerborg, A. et al. 1976. Analysis of crack formation and crack growth in concrete by means of
fracture mechanics and finite elements. Cement and Concrete Research 6, pp. 773–782.

Haywood, R. B. 1962. Design against Fatigue of Metals. London: Chapman and Hall

Heirany, Z. and Ghaemian, M. 2011. Effect of Concrete Nonlinear Parameters on the Seismic
Response of Concrete Gravity Dams. World Academy of Science, Engineering and Technology. pp.
637-640

Jeng, Y.S. and Shah, S.P. 1985. A Fracture thoughness Criterion for Concrete. Engineering Fracture
Mechanics 21(5). pp. 1055-1069
Karihaloo, B. L. and Xiao, Q. Z. 2006. Improving the accuracy of XFEM crack tip fields using higher
order Quadrature and statically admissible stress recovery. International Journal for Numerical
Methods in Engineering 66, pp. 1378-1410

Kundu, T. 2008. Fundamentals of Fracture Mechanics. London: CRC Press

Karihaloo, B. L. 1995. Fracture mechanics and structural Concrete. Essex: Longman Scientific and
Technical.

Kosmatka, S.H. et al. 2002. Design and Control of Concrete Mixtures. Portland Cement Association

Kumar, S. and Barai, V. S. 2011. Concrete-Fracture Models and Applications. Berlin: Springer-Verlag
Berlin and Heidelberg GmbH & Co. KG

Karihaloo, B. L. and Xiao, Q. Z. 2011. Accurate simulation of mixed-mode cohesive crack propagation
in quasi-brittle structures using exact asymptotic fields in XFEM: an overview. Journal of Mechanics of
Materials and Structures 6 (1-4), pp. 267-276

Kmiecik, P and Kaminski, M. 2011. Modelling of reinforced concrete structures and composite
structures with concrete strength degradation taken into consideration. Archieves of Civil and
Mechanical Engineering Vol. XI, pp. 623-636

Lamond, J. F. and Pielert, J. H. 2006. Significance of Tests and Properties of Concrete and Concrete-
making Materials. New Jersey. ASTM International

Lundgren, K. 1999. Three-dimensional Modeling of Bond in Reinforced Concrete: Theoretical Model,


Experiments and Applications. PhD Thesis, Chalmers University of Technology

Lew, H. S. et al. 2005. Design, Construction and Maintenance of Structural and Life Safety Systems.
Federal Building and Fire Safety Investigation of the World Trade Center. NIST NCSTAR 1(1). pp.
270

Mohammadi, S. 2008. Extended Finite Element Method for Fracture Analysis of Structures. Iran:
Blackwell Publishing.

Mehta, P. K. and Monteiro, P. J. M. 1993. Concrete: Structure, properties, and materials. N.J.
Canada. Prentice Hall

Malvar, J. and Warren, G. 1988. NCEL Technical Report on Fracture Energy for Three-Point Bend
Tests on Single-Edge Notched Beams. California: Naval Civil Engineering Laboratory Port Hueneme
Mousas. V. Ch. 2008. NDT: Control and Evaluation of Structures. TEI of Athens.
[Online]. http://users.teiath.gr/vmouss/ebooks/fmndt/30ndt.html [Accessed: 07 September 2013]

Mergheim, J. et al. 2005. A finite element method for the computational modelling of
cohesive cracks. International Journal for Numerical Methods in Engineering 63, pp. 276-289

Mougaard, J. F. et al .2011. A Consistent Partly and Fully Cracked XFEM Element for Modelling.
International Journal for Numerical Methods in Engineering 85 (13), pp. 1667-1681

Mellenk, J.M. and Babŭska, I. 1996. The partition of unity finite element method: basic theory and
application. Computer Methods in Applied Mechanics and Engineering 139, pp. 289–314.

Möes, N. et al. 1999. A finite element method for crack growth without remeshing. International
Journal for Numerical Methods in Engineering 46:131–150.

Möes, N. and Belytschko, T. 2002. Extended finite element method for cohesive crack growth.
Engineering Fracture Mechanics 69, pp. 813–833.

Maierhofer, C et al. 2010. Non-destructive Evaluation of Reinforced Concrete Structures, Volume 1:


Deterioration Processes and Standard Test Methods. Ed. Taylor and Francis, 2010. p. 1-23.

Maji, A.K. et al. 1989. A Study of Mix-Mode crack Propagation in Mortar Using Holographic
Interferometry. Proceedings of the SEM Spring Conference. Boston, May, 1989.

Mojiri, S. 2010. Numerical Analysis of Cohesive Crack Growth Using Extended Finite Element
Methods (XFEM). MSc Thesis, Ecole Centrale de Nantes (ECN)

Nallathambi, P. and Karihaloo, B.L. 1986. Determination of specimen size independent fracture
toughness of plain concrete. Magazine of Concrete Research 38(135), pp. 67-76

Niwa, H.S. 2003. Power-law versus exponential distribution of animal group sizes. Journal of
Theorical Biology 224(4). pp. 451-457

Neville, A. M. 2011. Properties of Concrete Fifth Edition. Essex: Pearson Education Ltd

Ozbolt, J. and Bažant, Z.P. 1992. Microplane Model for cyclic triaxial behavior of Concrete. Journal of
Engineering Mechanics (ASCE) 118 (7), pp. 1365-1386
Paulszny, A. and Matthai, S.K. 2009. Numerical modeling of discrete multi-crack growth applied to
pattern formation in geological brittle media. International Journal of Solids and Structures 46 (18-19).
pp. 3383–3397

Patzák, B & Jirásek, M. 2002. Process zone resolution by extended finite elements. Research gate
[Online]. Available at: http://www.researchgate.net/publication
/222026640_Process_zone_resolution_by_extended_finite_elements?ev=pub_cit
[Accessed: 19 April 2013]

Petrovic, J. J. 1987. Weibull Statistical Fracture Theory for the Fracture of Ceramics. Metallurgical and
Materials Transaction A (18). pp.1829-1834

Panagopoulos, G. and Kirtas, E., 2005, Reinforced Concrete Structures I lecture notes. Technological
Institute of Serres, Greece.

Park, K. et al. 2007. Determination of kink point in the bilinear softening model for concrete.
Engineering Fracture Mechanics 75. pp. 3806-3818

Ramesh, K. 2003. Engineering Fracture Mechanics. Chennai: IIT Madras

Rabczuk, T. et al. 2005. A numerical model for reinforced concrete structures. International Journal of
Solids and Structures 42 (506), pp. 1327-1354

Rots, J.G. and De Borst, R. 1988. Analysis of Concrete Fracture in Direct Tension. International
Journal of Solid Structures Vol. 25(12). pp. 1381-1394

Rots, J.G. 1986. Strain-softening analysis of concrete fracture specimens. Fracture Toughness and
Fracture Energy of Concrete (Edited by Wittmann F.H.). Elsevier. pp. 137-148

Rossmanith, H.P. 1993. Fracture and Damage of Concrete and Rock FDCR-2. London: Chapman &
Hall

Roesler, J. 2012. Accelerated Performance testing in Concrete Pavement with short slabs.
International Journal of Pavement Engineering 13(6). pp. 494-507

Singh, M.K. et al. 2009. Analytical Solution for Conservative Solute Transport in One-Dimensional
Homogeneous Porous Formation with Time-Dependent velocity. Journal of Engineering Mechanics
135(9). pp. 1015-1021
Shah, S. G. and Chandra Kishen, J. M. 2011. Determination of Fracture Parameters of Concrete
Interfaces using Digital Image Correlation. Experimental Mechanics Volume 55 (3). pp. 303 – 313

Song, S. et al. 2006. Simulation of crack propagation in asphalt concrete using an intrinsic cohesive
zone model. Journal of Engineering Mechanics (ASCE) 132, pp. 1215-1223

Stolarska, M. et al. 2001. Modeling crack growth by level sets in the extended finite element method.
International Journal for Numerical Methods in Engineering 51, pp. 943–960.

Shi, Z. 2009. Crack Analysis in Structural Concrete. London: Elsevier Ltd.


Williams, M. S & Todd, J. D. 2000. Structures theory and Analysis. London: Macmillan Press Ltd

Tsoukatos, K.P and Markowski, A.M. 2008. Power-law is exponential queuing in a network traffic
model. Performance Evaluation 65. pp. 32-50

Unger, J. L. et al. 2006. Modeling of cohesive crack growth in Concrete Structures with the extended
finite element method. Computational Methods and Applications in Mechanical Engineering 196, pp.
4087-4100

Vecchio, F.J and Collins, M. P. 1986. The modified compression field theory for reinforced concrete
elements subjected to shear. Journal of Engineering Mechanics (ACI) 83 (2), pp. 219-231

Vecchio, F.J. 1992. Finite Element Modelling of concrete expansion and confinement. Journal of
Engineering Mechanics (ASCE) 118 (9), pp. 2390-2406

Vecchio, F.J. and DeRoo, A. 1995. Smeared-Crack Modelling of Concrete Tension Splitting. Journal
of Engineering Mechanics (ASCE), pp. 702-708

Van Mier, J. et al. 1993. An experimental and numerical study of mode I (tensile) and mode II (shear)
fracture in concrete. Journal of Mechanical Behavior of Materials 4, pp. 179-190

Van Mier, J.G.M. 1997. Fracture Process of Concrete: Assessment of Material Parameters for
Fracture Models. New York: CRC Press.

Wagoner, M.P. et al. 2005. Development of Single-Edge Notched Beam Test for Asphalt Concrete
Mixtures. ASTM Journal of Testing and Evaluation 33, pp. 452-460

Wang, Z. et al. 2010. A simplified Numerical Elastic-Plastic Contact Model for Rough In surfaces.
Tribology Transactions 53. pp. 224-238
Wells, G.N. and Sluys, L. J. 2001. A new method modelling cohesive cracks using finite elements.
International Journal for Numerical Methods in Engineering 50, pp. 2667-2682

Weibull, W. 1951. A statistical distribution function of wide applicability. Journal of Applied Mechanics
18 (3), pp. 293-297

Wells, G. N and Sluys, L. J. A new method for modelling of cohesive cracks using finite elements.
International Journal for Numerical Methods in Engineering 50(12), pp. 2667–2682.

Zongjin Li. 2011. Advanced Concrete Technology. New Jersey: John Wiley & Sons

Zubelewicz, A. and Bažant, Z.P. 1987. Interface element modeling of fracture in aggreagate
composites. Journal of Engineering Mechanics (ASCE) 113, pp. 1619-1630

Zi, G. and Belytschko, T. 2003. New crack-tip Elements for XFEM and applications to Cohesive
Cracks. International Journal for Numerical Methods in Engineering 57, pp. 2221–2240

Zheng, J. H. 2003. Evaluation of concrete structures by advanced non-destructive test methods –


impact echo test, impulse response test and radar survey. Proceedings of the International
Symposium on Non-Destructive Testing in Civil Engineering (NDT-CE 2003). Berlin, September,
2003. Germany.

Zhu, J. and Popovics, J. S. 2008. Report No. NSEL-010 Non-contact NDT of Concrete
StructuresUsing Air Coupled Sensors. Illinois: Department of Civil and Environmental Engineering
University of Illinois at Urbana-Champaign
Apendix A: Engineering Case Study
XFEM based Partly Cracked Interface Model
for Reinforced Concrete Analysis

Student: Iason Iakovidis


Supervisor: Dr. S. Bordas
Duration: February 2013- September 2013

M.S.c Structural Engineering- Cardiff School of Engineering


Contents
A.1. Introduction
A.1.1. Synopsis 1
A.1.1. Aims of the Project 1-2
A.1.2. Theoretical Background and Developments 2-3
A.1.3. Linear Fracture Mechanics and Concrete 3-9

A.2.Literature
A.2.1. Review 9-15
A.2.2. Computational Techniquess 15-18
A.2.3. Extended Finite Elements 18-19

A.3. Case for Support


A.3.1. Introduction 16
A.3.2. Summary of Previous Work 16-22
A.3.3. Description and Context
A.3.3.1 Background 22-25
A.3.3.2 Program and Methodology 25-28
A.3.3.3 Explanation of Relevance 28-29

References
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

A.1. Introduction
A.1.1. Synopsis
Fracture mechanics is very important for the development of theories related to changes in
material`s behavior affecting the life expectancy and serviceability of structures. In general,
Fracture Mechanics is divided in two main theoretical categories: Linear Fracture Mechanics
(LEFM) and Non-linear fracture Mechanics (NLFM) related to ‘macroscopic homogeneous’
materials, exhibiting linear variation of stress intensity factor (KIC) and ‘macroscopic
heterogeneous’ materials, exhibiting non-linear variation of stress intensity factor (KIC)
respectively.

The macroscopic heterogeneous materials exhibit a larger fracture process zone and
subsequently a phenomenon called “strain softening”, which is capable to characterize
them as quasi-brittle. Concrete exhibits strain softening due to energy dissipation
mechanisms developed through its fracture process zone. In addition, is subjected to a
phenomenon called size effect, which is capable to interpret the ‘macroscopic
homogeneous’ behavior for large scale structures.

All in all, it is of great importance to examine the behavior of concrete and validate the
existing models, in order to understand the fundamental theoretical background and
improve the accuracy of the results leading to optimization of structural representation.

A.1.2. Aims of the Project


The aim of this thesis is the investigation and evaluation of the existing models and
emulations that concern the macroscopic concrete behavior. Firstly, we aim at finding and
implement the most suitable theory for reinforced concrete cracking estimation, improve it,
and secondly indicate a new direction for further analysis and future scientific research.

In order to achieve this aim, we shall determine the essential parameters that will be
investigated. In particular, progression of non-linear fracture mechanics theory, theoretical
background for concrete, computational models and their theoretical validity, proposed
crack analysis model and finally, processions of implementation.

1
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

In the first section there would be an introduction in concrete characteristics and general
fracture mechanics. We will define the importance and effectiveness of fracture mechanics
in the design of concrete structures, covering the most important points in legislation
described by Eurocode, British Standards and ACI.

In the second section, there will be an attempt to analyze the available theory for non-linear
elastic materials, define the quasi-brittle characteristics, present the available non-linear
elastic models and introduce the model chosen for implementation.

Furthermore, there will be an analytic simulation of the proposed model introducing further
improvements, based on previous scientific projects and publications.

A.1.3. Theoretical Background and Developments


The theory related to fracture mechanics is referred as ‘Damage Tolerant Design Approach’.
This approach defines the role cracks existing within the structure and how these develop in
service, reaching critical values capable of affecting the performance and life expectancy of
a structure (Bansal, 2001).

The fundamental idea is related to strength of materials, introducing three test


methodologies for concrete`s strength estimation: tensile, four-point bending and twisting
of shaft under torsion test. The former is capable to characterize the form of the material
i.e. ‘isotropic’, based on two fundamental parameters measured from tensile test, modulus
of elasticity (E) and Poisson ratio (v). The bending four-point test investigates the normal
stresses due to bending, which vary in linear fashion over the depth of the beam. Finally, the
torsion test presents the shear stress, which varies linearly over the cross-sectional
dimensions.

Although the aforementioned procedures are fundamental, scientists had to investigate


further the complex phenomena observed in structures. Various ‘Yield theories’ were
developed, such as ‘Maximum Principal Stress Theory’ (Rankine, Lamé), which
accommodated the analysis of load combination. In addition, theoretical criteria, such as
Tresca- Guest`s ‘maximum shear stress theory’ and Von-Misses ‘Maximum Principle Strain

2
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

Theory’ for axial, torsion and bending loading combinations were introduced. These ‘yield
theories’ consist an attempt to exploit the fundamental properties defining materials’
behaviour, estimated by load combination tests, in order to prevented structural failure due
to ‘yielding’ (Coates et al, 1988). Also ‘buckling’ found in slender structures, affecting mostly
structural systems and components subjected to bending, torsion or even in pure shear,
leads to progressive corruption, fracture and complete collapse of the structural mechanism
(Williams & Todd, 2000).

Although, all the previous approaches allowed us to investigate materials` behaviour, they
can be characterized as limited (Exadaktylos, 2006). Thus, the necessity for a new test
method emerged. This test is called ‘fatigue test’ and its results indicate that a specimen will
fail under the effect of cyclic loading and pre-defined laboratory conditions (Eurocode 2).
The fatigue test is presenting an improved, in service, approximation of material`s behavior
in contrast to tensile test.

The Standards are based on two fundamental approximations for strength and stiffness. As
for the strength, the procedure is affected by the determination of the cross-sectional
dimension and choice of tested material, in order to ensure that ‘ultimate yielding stress’ at
any point is not over the pre-defined limits (Panagopoulos & Kirtas, 2005). For the stiffness,
the determination of the deformation limits is achieved by the definition of relative
geometrical and cross-sectional dimensions for the material chosen, restrained within
optimized material`s strength boundaries called ‘endurance limits’ (Eurocode 2). The
material`s characteristics are presented using theoretically defined safety functions,
therefore fracture mechanics is not participating in the Standards, due to insufficient
theoretical and practical background (Karihaloo, 1995).

A.1.4. Linear Fracture Mechanics and Concrete


By the term fracture, we mean the generation of new surfaces within a material, which
constitute a way of failure for a structure. In microscopic level, the destruction of the
bonding between the atoms of the material is being observed. In macroscopic level, fracture
is the crack intersection and splitting of the structure in multiple parts due to crack
propagation. In the mesoscopic level, fracture occurs by the initiation, propagation and
3
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

coalescence of cracks, i.e. voids and microcracks within the grains of the material (fig. 1)
(Exadaktylos, 2006).

Fig. 1: Microcracks in marble specimen under uniaxial cyclic loading at 75x and 1500x respectively
(Exadaktylos, 2006)

In 1858, Karmarsch developed a practical relationship for estimating the carrying capacity of
steel wires, which was quite similar to Griffith`s (1921) equation for fiberglass. Wolhers
(1860) and Kommers (1920) noted that the strength depends on the surface quality
(disorders, flaws) due to fatigue tests and identified a 20-50% increase at the strength
subjected at grinding or polishing. Wieghardt & Leon (1908) examined the behavior of the
stress distribution at the apex of a wedge type opening and the direction of the crack
initiation under combined point loading. The solution represents the asymptotic behavior of
the stressed field in the tip of the apex and also the special problem governing the crack.

Wieghardt mentioned that although ‘Rankine theory’ assumes an infinite stress for arbitrary
infinitesimal force P, the experiments show a finite value of P and assumed that fracture
does not initiate in a point, but in a minute region. In addition, the orientation and angle of
propagation based on ‘Tresca & Guest Criterion’ led to the fracture criterion σc= (Eγ/c)1/2,
where c is the net perimeter and improved criterion σc= (1/4)-(E/13). Smekal (1922; 1935),
assumed that besides the pre-existing flaws, there was unevenness within the material. This
assumption led to the dislocation theory (1934), which is described as a statistical and
stochastic process by which microcracks form, grow and propagate under very small plastic
deformation. After that, Weibull (1939) developed the statistical theory of fracture.

In 1944, Zener- Holloman relate the crack propagation theory of Griffith with the brittle
fracture of inelastic bodies and observed that at ballistic rates thermal softening can result
in the formation of shear bands (Exadaktylos, 2006).

4
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

In 1945, Orowan observed plastic deformations in surfaces of materials failing in a brittle


manner. This observation was very important due to the fact that, Irwin (1955) used it in
order to suggest an improvement of Griffith`s brittle fracture energy dissipation law, in
which the ultimate energy during a crack propagation it is dissipated as: 1) stored elastic
strain energy and 2) dissipated energy including plastic consumption and surface energy. In
other words G=2γ+Gp (Karihaloo, 1995). The latter recognized by Orowan, who stated that
the modification of Griffith`s principal was necessary, in order to investigate brittle fracture
(Exadaktylos, 2006).

In 1957, Irwin defined the strain energy release rate, G=-dΠ/dA, where G is the crack driving
force or crack propagation force. This rate of release energy is the released energy needed
for a crack to extend, while there is a unit crack area formation (Karihaloo, 1995).

Williams (1952, 1957), used the work of Wieghardt and Leon (1908), which described a
constant singularity for linear elastic materials defined by r-1/2, solving it using Airy stress
function near the crack tip. The latter arises from the analysis of a unit depth element in
polar coordinate system (Shi, 2009). Williams reach the conclusion that the strain energy of
an element with unit thickness and the total strain energy of the finite element are given by:

Williams (1957) theory was very important due to the fact that, led him to calculate the
crack tip stress field for the modes introduced by Irwin (1957). Irwin`s theory of brittle
fracture, recognized three different types of crack propagation related to the state stress
developed in the vicinity of the sharp crack tips (fig. 2) (Shi, 2009).

Fig. 2: Opening, sliding, tearing and sharp crack of length 2α in an infinite elastic body (Shi, 2009).

5
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

Stress and displacement non-zero components introduced for all three modes and mix-
mode I & III. Irwin`s criterion described that a sharp crack will grow when stress intensity
factor K reach a critical value. This critical value of K, KIC, described as a material constant,
called fracture toughness (Karihaloo, 1995).

Energy methods such as Irwin`s potential energy (ΔΠ) concept due to a crack infinitesimal
growth length (Δα), introduced for an infinite body under mix-mode I & II. From that, we
could calculate the opening and sliding displacements letting first the crack to advance and
introduce the stress intensity factors relation with surface density energy γ in three loading
modes. Under three-dimensional loading, the relation between potential energy and
loading modes for a brittle material becomes:

Barenblatt proposed the ‘cohesive crack Model’ (1959), in order to simulate an infinitesimal
propagation of constant length ρ (<<α), in a so called plastic zone located in crack tips,
which is subjected to cohesive forces (q(x)) in order to close smoothly (Karihaloo, 1995).

In general there are three main categories of crack propagation: initiation, where
microcracks intersect until reaching a length of 1mm, the crack longitudinal development,
due to cyclic loading and unbounded propagation, when the crack`s length reach a critical
value of , leading to failure (Exadaktylos, 2006).

Paris (1960) mentioned that the increase ratio of a crack under fatigue is related to

rate variation of stress intensity factor relative to . This

leads to a linear relationship corresponding to the constant state of crack propagation,

which is given by ( ) , 2<n<4 (Kundu, 2008).

An improvement of Barenblatt`s model proposed by Dugdale et al. (1959; 1963), in which


variation in the length ρ≤α of plastic zone assumed, reaching its maximum value equal to a

6
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

(half crack length) corresponding to the yield stress σy characterized as critical for
unbounded propagation (Karihaloo, 1995).

In addition, the factor CTOD (crack tip opening displacement) introduced, which varies due
to cohesive zone length and localized stress. Dugdale et al model, for an infinitesimal σ/σy
ratio reach the value , which is similar to Irwin and Griffith
criterion. This describes an elasto-plastic material with limited plasticity (Karihaloo, 1995;
Shi, 2009).

Many scientific attempts such as Kaplan`s (1961), Glucklich`s (1963), Romauldi`s and
Batson`s (1963), Zaitsev`s (1971), Swamy`s (1972), to apply directly linear elastic fracture
mechanics to concrete did not succeed. Kesler et al. (1971) mentioned that the applied
theory of brittleness was insufficient to give a clear interpretation of concrete behavior and
Walsh (1972) plot the logarithm of the nominal strength to the logarithm of the size for
similar specimens, which described an deviation of concrete`s slope plot, from the value of -
1/2 required for linear elastic fracture mechanics (Karihaloo, 1995).

This phenomenon is based on fracture toughness (KIC) and corresponding toughness (GC),
which vary over the notch depth of beams in contrast with the Irwin`s theory request. Many
scientists such as Brown (1972), Shah and McGarry (1971), Walsh (1972), Swamy and Rao
(1973), Higgins and Bailey (1976), Mindess and Nadeau (1976), Walsh (1976), Gjorn et al.
(1977), Rossi et al. (1984) observed this phenomenon (Karihaloo, 1995).

Nallathambi (1986) in tests using concrete containing coarser aggregated and hardened
cement paste, found a significant decrease of strain hardening zone led to rapid brittle
fracture, lead to the conclusion that microscopic heterogeneity is the main reason of
concrete`s quasi-brittle behavior (Karihaloo, 1995). However, there was a significant
specimen depth, after which linear elastic fracture mechanics can apply to concrete. This
phenomenon was described by a tension softening behavior of concrete, which prevent the
direct failure of specimen after reaching the ultimate yield stress (Karihaloo, 1995).

7
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

The fracture process zone (FPZ) length for various materials found by Bache (1986),
Hillerborg (1983), Hillsdorf and Brameshuber (1991) and Brulwiler et al. (1991) (fig. 3).

Fig. 3: Typical length of fracture process zone for various materials (Karihaloo, 1995)

Griffith`s theory basic premise is that all the energy stored in a body is released for the
growth of the crack. However, in concrete, the microstructure prevents the direct
propagation of the crack (Exadaktylos, 2006). As a matter of fact, energy dissipation
processes observed within the FPZ reducing the available stored energy for crack
development (Karihaloo, 1995). Consequently, microscopic heterogeneity modifies the
length of fracture process zone (Asferg et al, 2005).

Bažant (1983; 1984) referred to “energetic size effect”, which in any point should not be
considered the same as Weibull’s size-effect statistical theory (Karihaloo, 1995). The
energetic size-effect is determined by a characteristic length which is observed only for
large scale-structures. Weibull`s theory does not include characteristic length and this
contribute to the introduction of a power law function D (physical function), which
exponent is given by , where m is the Weibull`s modulus and is single
or multiple geometrical similarity (Bažant, 1984).

Scaling is the changes in the response, when there are variations in the direction of crack
propagation, which contribute to complex shape of growth of all physical materials. Without
scaling there is no applicable solution in the problem of crack propagation (Bažant, 1984).
However, Bažant (1997) described a phenomenon not capable to capture the complex
“energetic size-effect” of concrete.

The crack development depends on fracture energy , tensile strength and

characteristic length of FPZ, Irwin (1958), given by . Leicester (1969) examined

8
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

geometrically similar notched concrete beams and observed that the exponent was greater
than -1/2, required for LEFM. Due to this reason, he introduced a power size-effect, in order
to interpret the phenomenon via the assumption that the strength of singularity stress in
sharp notches of finite angle was smaller than that of sharp cracks. However, propagation
for sharp notched of finite angle, zero flux of energy in notch tip and the power-law, which
does not include characteristic length for size-effect, needed further investigation (Bažant,
1997).

From the aforementioned, Walsh (1972) work was of great importance, due to the fact the
results observed comparing similar to Leicester`s specimen, lead him to the log of nominal
strength to size plot, which supports the assumption that there is a theoretical field
between plasticity and linear elastic fracture mechanics.

All in all, in plasticity theory, there is no size-effect and the application includes only limited
small-scale fracture, while in LEFM maximum size-effect appears in the tip of the crack,
where the nominal strength is inversely proportional to A1/2 (A= size of structure), due to the
reason that the procedure is punctual. The length of crack is large acting as a material
property. If the crack is small then the size-effect is macroscopic and there is no reason to
be determined (Karihaloo, 1995).

A.2. Literature

A.2.1. Review
In general, there are four main theoretical crack models: cohesive crack model, CCM,
(Barenblatt, 1959; Dugdale, 1963), fictitious crack model, FCM, (Hillerborg et al., 1976),
Crack Band Model, CBM, (Hillerborg and Bažant, 1976; Bažant and Cedolin, 1983; Bažant
and Oh, 1983), bridged crack model, BCM, (Bao and Suo, 1992; Li and Maalej, 1996;
Needlemann, 1990; Ferro, 2002; Stang et al., 2006), Microplanes (Bažant & Ozbolt, 1990),
Particle model (Bažant et al., 1990), Lattice Model (Van Mier, 1997) and Interface model
(Ngo & Scordelis, 1967; Lundgren, 1999; Østergaard, 2003).

9
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

The theoretical models can be simulated with finite elements analysis (FEA), extended finite
elements analysis (XFEA) and mathematical approximations such as boundary integral
methods, boundary element methods, which are not able to extend using extended finite
element procedures. For FEA the four main methods are: smeared crack approach, SCA,
(Rachid, 1968) and discrete crack approach, DCA, (Ngo & Scordelis, 1967; Nilson, 1967;
Nilson, 1968; Hillerborg et al., 1976).

Partition of unity (Melenk & Babushka, 1996) and discontinues enrichment of the elements
(Belytschko & Black, 1999), described a minimal remeshing model capable to simulate the
arbitrary alignment of the crack in relation to the mesh. An improved method proposed by
Moës et al. (1999), which aimed at the intersection between geometry of crack and
meshing, including enriched approximation, in a way that the whole crack is independent of
meshing. In addition, Dolbow (1999) introduced a model with arbitrary discontinuities with
local enrichment of displacement, based on unity partition concept. Galerkin method is also
a meshless method, however it has low compatibility with finite element procedures (Zi &
Belytschko, 2003).

Sukumar et al (2000) represented the crack in three-dimensional plane with enrichment of


finite elements. Daux et al. (2000) proposed a model including branched and intersected
cracks originate from holes introduced in the model. Two very important theories for the
XFE analysis, introduced by Stolarska et al (2001) and Belytschko et al (2001): the set level
method and arbitrary discontinuities in signed distance function and components of
elements respectively. Signed distance is a simplification of set level method (Pais, 2012).

Furthermore, a worth mentioned method is ‘scalar damage method’. Frichant et al. (1999)
proposed a model for scalar failure with simplified limits in comparison with anisotropic
models. The results gave a good approximation for uniaxial failure, but crack closure was
not participated in the method. Another anisotropic model introduced by Baker & de Borst
(2005), in which the thermodynamic anisotropic failure of concrete studied under high
stresses and periodic thermal variation. This includes the specification of heat flow and
evolution of entropy in order to define the thermodynamic surface failure. In addition,
Kondelka & Krejci (2008) proposed a thermo-hydro-mechanical model for the lift layer of

10
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

concrete of an examined bridge deck. This thin layer was subjected to shrinkage, weather
conditions and sun radiation, which found to be of great importance to damage evolution.

Cohesive crack model (Barenblatt, 1959; Dugdale et al, 1963)


Barenblatt`s Cohesive crack (1959) model improved by Dugdale et al. (1960-1963) for
simulation of concrete. The basic difference was that Dugdale`s model included a known
distribution of cohesive stresses, cohesive zone length was not constant and of very small
length (ρ≤α). As a matter of fact, the plastic zone can increase up to the point where σ
reaches the ultimate yield stress σy. This finite value of σ leads to smooth tip closure and
elimination of stress intensity factor in the tips of the coplanar plastic zone (Karihaloo,
1995).

Fig. 4: Dugdale et al. crack model and rate of non-dimensional σ/σy relation (Mohammadi, 2008)

c 
Smooth closure criterion  1  cos( ) .
a 2 y

The total displacement of the traction free crack is |x|=α-ρ ≥0 and equal to 2u. The crack tip
opening displacement is given by:

8 c c 1 C T O D (   ')
CTOD    y (1  ) ln (1  ) and for non- dimensional CTOD* 
 ' a a 8  y

The non-dimensional rate of σ/σy is given (fig. 4). There is a value Wc, which is characterized
as critical value of CTOD and describes the fracture criterion. If CTOD reaches this value then
the propagation of crack is unbounded and leading to extreme uncontrolled failure.
However, it is quite difficult to measure the exact Wc (Karihaloo, 1995).

In addition, if the external tensile force applied is very small, subsequently both the CTOD
and cohesive zone are infinitesimal. The introduction of power-series σ(πα)1/2= (Ε’ σyWc),
defining similar to Irwin and Griffith`s law, leading to Gc= σyWc. Therefore, Barenblatt`s

11
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

cohesive model described, which it is not capable to simulate the crack development of
concrete (Karihaloo, 1995).

Fictitious Crack Model (Hillerborg, 1976)


Hillerborg (1976) introduced the characteristic length of fracture process zone and proposed
the concept of fictitious crack model. This model followed the concepts of cohesive crack
model (Barenblatt, 1959; Dugdale et al., 1963), however it did not include any assumptions
about the size of FPZ. In other words, the stresses in the cohesive zone are known.

Fig. 5: FCM model, where a0 is the traction zone, assumed elastic and lp is the FPZ, which includes dissipation
mechanisms contributing to fracture energy GF (Karihaloo, 1995; Shi, 2009)

A traction free elastic zone with a0 length assumed in front of the FPZ (fig. 5) and the
characteristic length of FPZ, lp, is simulated by a fictitious crack model, in which energy is
being absorbed due to crack propagation. This energy is called fracture energy and
designated by GF, which is geometrically presented as the area under the strain softening
curve, from the diagram of tensile strength, ft, over critical displacement, Wc. Nallathambi &
Karihaloo (1991a) stated that Wc is very difficult to be found accurately due to the fact that
the tip of the pre-existing macro-crack is never sharp in order to premise accurate
localization leading to variation in the resulting forms of the strain softening curve affecting
GF. Karihaloo continues saying that the best approximation of GF is made by Hillerborg
(1980) and Petersson (1983) using finite element methods, the results of which included in
the RILEM (1985).

Crack Band Model


Based on the model of Hillerborg et al. (1976), Bažant & Cedolin (1983a) and Bažant & Oh
(1976) concept of a crack band introduced. In this method the crack is simulated as a region

12
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

of densely distributed ‘small lines with obtuse face’ (Bažant & Cedolin, 1983a). As a matter
of fact, this shape of the crack path is due to the morphology of FPZ illustrated by the
variation of aggregate size (Fig. 6).

Fig. 6: Crack path between aggregates, corresponding actual and smoothed path, band crack model and
corresponding stress-strain diagram (Mohammadi, 2008)

The COD (crack opening displacement) is equal to the product of width, hc, and fracture
2
E ft
energy, GF. The latter is given by: G F  h c (1  ) and the corresponding strain by:
E1 2  E

2 GF
u  . This method simulation using finite element analysis, in relevance with element
f t  hc
'

size approximation, in which the value of hc is equal to 3∙dmax, where dmax is the maximum
aggregate size.

The development of finite element procedure helped a lot in the research for the best
approximation and a two-parameter fracture model for concrete proposed by Jeng & Shah
(1985), which is also based in the model of CTOD of Dugdale et al. (1963).

Bridging Crack Models


This is crack model used to simulate multi-scale phenomena such as, fiber and aggregate
bridging, which require cohesive multi-scale laws. The cohesive laws describe the bonding
and include disorders of the material. Due to variation of scale there is a variation in the
cohesive law, which can present atomic dissociation of bonding, dissociation of grain
interfaces and stress exchange between the bonds of the micro-crack. It is obvious that the
cohesive laws introduce scales, which are based to the characteristic crack opening Wc as
Needlemann (1990) noted.

13
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

Fig. 7: Multi-scale concept and illustration of crack behaviour (Needleman, 1990; Strang et al, 2006b)

The equations governing the model are quite the same as those of fictitious crack model.
Cox and Marshall (1991) mentioned that the cohesive model with smooth closing tips can
be characterized as bridged models, due to the elimination of stress intensity factor. This
depends on the ratio of fracture energy of the tip over the fracture energy of cohesive law.

Ferro (2002) used the bridged model for concrete and other cementitious materials.
Ballarini & Muju (1991) used non-linear finite analysis, in order to simulate the propagation
of bridged crack, presented as an orthotropic continuum, which prevent the opening of the
crack. Carpinteri & Massabo (1997) mentioned that there are two available forms of bridge
cracks using continuum or discontinuum and described the ductile to brittle transition of
concrete using fibers. Sanborn & Prevost (2008) described a discontinuity with controlled
interface and noticed that fibers within a composite can transfer the stresses from crack tips
to other points inside the crack.

Microplane Model
This model is based in the theory of plasticity for polycrystalline metals proposed by Taylor
(1938) and ‘slip plasticity theory’ (Batdorf & Budianski, 1949), in order to define the
constitutive law of plastic hardening metals. The basic problems of these two theories,
discussed by (Kroner, 1961; Budianski & Wu, 1962; Hill, 1965, 1966; Lin & Ito, 1965-66; Rice,
1970), were strain hardening origination and vertex effects for the anisotropic loading of
radial path in stress field. Solutions on the problems were given from anisotropic modeling
of rocks (Zienkiewicz & Pande, 1977; Pande & Sharman, 1981-1982; Pande & Xiong, 1982),
who described the phenomenon of inelastic shear strain called ‘slips’, which participate

14
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

statically in stress tensor σij in inelastic vectors of plane continuum, macrolevel (Bažant &
Ozbolt, 1990; Bažant et al, 2000).

As a matter of fact, Bažant (1984) noticed that the fundamental principle of this model is
that the constitutive law is based on vectors, in order to define the relationship between
stresses and strains, which is determined respectively by alterations of the direction of
microplanes within the material. Carol et al. (1999) assumed that these alterations of
directions based on the concept of total density energy as a parameter of strain vector.

A.2.2. Computational Techniques


Interface Elements
Zubelewicz & Bažant (1987) introduced the concept of interface elements by testing brittle
aggregates composites and noticing that the force-displacement relationship of a rigid
particle system, arbitrary defined, under normal tensile strength limit, was characterized by
a sudden drop of force to zero. In this model the aggregates cannot overlap and are not
connected between them, however when the distance between them reaches a limit a
deformed layer introduced in the model. As a matter of fact, the results given by this model
showed a band of cracked inter-particles, which are not connected between them, the
width of which is approximately three times the dimension of the largest aggregate.

In interface elements the cracks propagates from node to node and the crack depth and
yielding stresses are equal to tensile forces in the crack tips. The thickness of the elements is
zero and big stiffness introduced in the model. Works of Lundgren (1999) and Østergaard
(2003) described the orthotropic bonding between reinforcement and concrete, Dirk-
Nielsen et al (2006) noted that the model is subjected to combination loading of mode I and
II and methods such as fiber reinforcement pull-out can be used to simulate more
extensively the bonding. The pull-out method was prior examined by Karihaloo (1995) for
axisymmetric headed anchors.

One of the characteristics of interface elements is the meshing update, which in some cases
is very complex computationally. That’s the reason that works of Carol et al (2001), Prasad &
Krishmoorthy (2002) and Yang & Chen (2004, 2005) introduced models for known meshing,
15
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

or simple remeshing. In addition in these works the crack path is not known and the crack
propagation is characterized by variation in direction. Other works related to meshing and
mix-mode was: Xie & Biggers (2006), Alvarez et al (2012), in which interface elements using
virtual crack closure method and discrete crack method used respectively.

Smeared Crack Models


Smeared crack models presents the material as a continuum and the discontinuity of
displacement field as change of constitutive law, due to the location and propagation. The
difficult in the modeling is the crack location conditions and the variation of meshing.
Smeared model provide better results than discrete model due to the fact that crack
simulation due to changes of constitutive law preferred from meshing variation. In the
beginning the method used was fragile failure, in which stiffness assumed zero orientated to
maximum tensile stress direction and then a method of gradual stress reduction introduced
for results improvement for the post peak phase. Cedolin & Dei Poli (1977), Bažant &
Gambarova (1980), Gupta & Maestrini (1989) and Vecchio (2000), proposed the used of
experimental value functions in order to present the results of stress-strain diagram.

Although the results for reinforced concrete were acceptable, due to smooth distribution of
crack surface, the same did not happen with the behavior of concrete structures were the
crack was governing such as in plain concrete and deep beams. The problems occurred due
to meshing incompatibility. Fracture mechanics and work of Bažant & Cedolin (1983), Bažant
& Oh (1983), Feenstra & de Borst (1995) και Rots & Blaauwendraad (1989), which
subsequently introduce fracture mechanics in modeling, provide us with improved models
and comparison studies between smeared and crack models. In this direction, Borst et al.
(2004) noticed that both models were capable to transmit well distributed micro-cracks to a
governing crack and support the idea of fracture mechanics introduction in modeling. The
comparison of the methods led to new categories of smeared cracks, such as fixed, rotating,
multi-axial fixed (Fig. 8). Ohmenhäuser et al. (1998), referred to one more category
microplanes, which were introduced analytically by Bažant & Ozbolt (1990) and Bažant et al
(2000). Also Ohmenhäuser, stated that all these cracks are fictitious models and proposed
an new method called ‘adaptive crack model’, based on the Mohr`s hypothesis and
transition between extremes.

16
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

Fig. 8: Variation of smeared crack models (Ohmenhäuser et al., 1998)

Nguyen & Chan (2005) analyzed a multi-smeared model for an reinforced concrete bridge
under earthquake and cyclic loading, noticing an overestimation for damping ration and
unloaded stiffness. Of great importance is also the work of Vecchio & DeRoo (1995), who
stated that smeared crack models were not capable to cover the concrete tension splitting
mechanism contributing to general and lateral dilation affecting the response of the
structure.

Embedded Crack Model


These models are a combination between discrete and smeared crack models (Jirasek,
2000). Its characteristic is that the standard finite element includes the strain and the
displacement discontinuities from the beginning. This leads to improvement of strain
localization and elimination of stress locking.

Some of these models are capable of describing a jump in displacement field and
characterized as strong displacement models and there are others not capable, which are
referred as weak displacements model. Ortiz et al (1987) used a one element weak
displacement model, in order to analyze the shear band region, which improve the
localization of the strain field but seemed to be insufficient to model accurate the shear
band zone. Belytschko et al. (1998) and Sluys & Berends (1998) introduced models capable
to decrease the dependence of the width to the size. Dvorkin et al. (1990) proposed an
improve model based on strong displacements, which was adopted later by Lotfi & Shing
(1995) introduction the cohesive traction condition.

Due to Jirasek (2000), there are three main categories of embedded cracks, based on
kinematics and statics (Jirasek, 2000) statically optimized symmetric (SOS), which satisfy the
requirements of kinematics and traction condition, but cannot satisfy the static leading to

17
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

spurious stresses. Such models are introduced by Belytschko (1998), Sluys & Berends (1988)
and Jirasek (2000). Kinematic optimized symmetric elements (KOS), giving an insufficient
approximation of traction condition and statically and kinematically optimized asymmetric
elements (SKON), which leads to elimination of spurious stress and independency of size of
the shear band.

A.2.3. Extended Finite Elements


The basis of this new method is the theory of partition of unity, PU, (Melenik & Babuska,
1996; Duarte & Oden, 1996) and generalization of finite elements, GFEM. The first models
used for linear elastic fracture mechanics (Belytschko & Black, 1999), in which two new
parameters added, minimum meshing and discontinuous enrichment for the crack
description (Mohammadi, 2008). Both these parameters can be described by:

u(x)     ( x )u i    j ( x ) ( x ) e j
  j 

Where I is the set of nodes, Νi(x) is the shape function of node ι, ui is the nodal
displacement, ε is the set of the enriched nodes, ej are the additional degrees of freedom
and ψ(x) is the enriched parameter (Zi & Belytschko, 2003).

This new method took its name by Moës et al. (1992) και Dolbow (1999) and allowed the
simulation of the whole crack independently of meshing including the relation between the
geometry of the crack to meshing supported by the enriched parameter. This method
extension in three-dimensional space included arbitrary branching and intersection of
micro-cracks (Dolbow et al, 2000a, b, c; Daux et al, 2000; Sukumar et al., 2000; Areias &
Belytschko, 2005a).

Of great importance was also the introduction of set level method, which provide the
localized simulation of the crack (Stolarska et al, 2001; Belytschko et al., 2002; Ventura et al.
(2003); Zi et al, 2004; Budyn et al., 2004; Bordas & Moran (2006), Stolarska & Chopp , 2003)
and fast merching method discussed by Sukumar et al. (2003) and Chopp & Sukumar (2003)
combined with XFEM implementation. Also the localization, mentioned by Jirasek &

18
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

Zimmermann (2010a, b) who combined XFEM with damage method and introduced a new
method including a gradual transform model (from smear to discrete).

Moës & Belytschko (2003) modeled the propagation of arbitrary cracks and Zi & Belytschko
(2003), Mariani & Pergo (2003), and Mergheim et al (2005) improved it. Song et al. (2006)
introduced an alternative method for arbitrary crack simulation including shear band
propagation followed by a rearrangement of XFEM on their basis and introduction of
additional degrees of freedom called phantom nodes. The same method also used by Jirásek
& Belytschko (2002), Rabczuk et al (2000), Olsen (2012) and Asferg et al. (2007).

A.3. Case for Support

A3.1. Summary of Previous Work


The fundamental cohesive crack model developed by Barenblatt (1959) and was further
improved by Dugdale et al. (1963). Hillerborg (1976) and his students, Modeer and
Petersson (1976) introduced the concept fictitious crack based on the cohesive model
(Barenblatt, 1959; Dugdale et al., 1963) including the concept of the characteristic length of
FPZ without making any assumptions about its length. In other words, stresses in the plastic
or cohesive zone are pre-defined. Zihai (2009) noted that Hillerborg et al. (1976), besides
the concepts of the FPZ and the tension-softening phenomenon introduced: a numerical
procedure capable of providing growth prediction and multi-cracking representation under
mix-mode conditions. Bažant & Gambarova questioned Hillerborg additional concepts, due
to the fact that multiple cracking leads to various orientations and shear type fracture. To
deal with multi-cracking phenomenon, Bažant & Gambarova (1980), proposed the shear-lag
phenomenon based on the concept of the shear transfer delay in the initial stage of crack
growth. Shear lag improved further by Kristek & Bažant (1987). The works of van Mier
(1989) and Maji & Shah (RILEM, 1989) are capable of providing more information for the
mix-mode phenomenon.

With the introduction of FEM, Petersson (1983) developed a more accurate model, included
in the RILEM (1985). Carpinteri (1980) proposed brittleness number, Bažant (1984)
introduce energetic size effect and Carpinteri (1986) introduced energy number. In the

19
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

RILEM (1989), Bosco et al. (1988, 1989a, 1989b) tested sixty un-cracked low-reinforced
concrete beams in bending (TPBT), mentioning that the brittleness number was capable of
providing minimum reinforcement for all concrete grades.

In these first methods, the fracture energy assumed constant and the softening curve a
material property. This had as a consequence, the elimination of energy dissipation
observed outside cohesive zone. These limitations defined by Nallathambi & Karihaloo
(1991a) mentioning that critical opening displacement is very difficult to be found
accurately, due to the fact that the tip of the pre-existing macro-crack is never sharp in
order to premise accurate localization. Karihaloo (1995) mentioned that the linear
approximation of softening curve, requirement of RILEM (1985), was able to give adequate
results for the majority of practical examples.

Wells & Sluys (2001) introduced string singularity approach able to model a displacement
jump, using finite element approximation. Moës & Belytschko (2001), using XFEM, eliminate
stress intensity factor and using J-integral allowed the smooth development of arbitrary
cohesive cracks. Zi & Belytschko (2003) introduce the partly cracked-tip element model,
which was able to provide meshing and remeshing independency and elimination of
blending. Mergheim et al. (2005) introduce a model applying constitutive law to the
interface to describe the inelastic behavior of the material and provide solution for both
straight and curved discontinuities.

Moës & Belytschko (2001) tried to verify the theoretical concept of zero stress intensity
factors in the tip of random cohesive cracks. The simulations made, introduce a new model
allowing smooth development of cohesive cracks without remeshing. Zeroing KI and by
combing XFEM with J-Integral, Moës & Belytschko achieved improved results in contrast
with FEM, in respect to the lower number of elements used for the simulation of material`s
characteristic length (lch).

Zi & Belytschko (2003) proposed a static cohesive crack-tip element with no requirement for
meshing, remeshing. The implementation of the model based on linear three node
triangular and quadratic six node triangular elements, where the stress projection normal to

20
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

the crack tip is equal to the material`s strength. The results obtained using Newton-
Raphson method. The model includes: 1) remeshing elimination due to XFEM meshless
settings, 2) two displacement field parts assumed: continuous (regular) and discontinuous
(partition of unity and enrichment), 3) formulation of new enrichment technique for static
cohesive crack model with high order enrichment and crack-tip random position in the
element based on Chen`s (2003) constant strain triangle. 4) Stress singularity at the crack-tip
cancelled by K due to cohesive stress on crack surfaces of fracture process zone, which
depends on crack opening displacement and softening curve.

Mergheim et al. (2005) introduced a cohesive crack model, in which the discontinuity
allowed to grow through the element without limitations, enriching the elements cut by the
propagation adding extra displacement degrees of freedom. The model is relevant to partly
cracked scheme, because is divided to two regions using zero and standard functions. In
addition, constitutive law is applied to interface element, in order to describe the inelasticity
of the material. This model provides solutions for both straight and curved discontinuities
independent of mesh.

Xiao & Karihaloo (2006) introduced the concepts of statically stress recovery (SAR) and
moving least squares (MLS), in order to improve the results for the crack tip fields obtained
by enrichment functions. These concepts led to a more specified choice of the enriched
elements and the size of the influenced domain, due to improvement introduced by the
Quadrature order method.

Asferg et al. (2007) mentioned that the model of Zi & Belytschko (2003) is limited, due to: a)
incapability of equal stress representation in both sides of the discontinuity, b) the
enrichment introduced only to the cracked elements, by using shifted sign function, which
influences both the displacement fields of the cut elements and that of the sharing
elements. To deal with the aforementioned, Asferg et al. (2007) proposed an additional
enrichment through the superposition of cracked elements and shape functions. In addition,
he included the independent of degrees of freedom ‘pseudo-node P’ method, in order to
represent a sub-triangle in which the discontinuous displacement field vanishes inside the
CST element (Adaptive Sub-triangulation). Constant strain triangles (CST) were not capable

21
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

of modeling equal stresses in the element and a formulation-criterion for smooth transition
between un-cracked, partially cracked and full cracked representation should be introduced.
For this, a stress interpolation generated by the nodal stresses combined with the weighting
factors for every element used. As for the smooth closure, the crack begin to propagate
when the tensile stress of the crack-tip overcome the tensile strength of the material,
followed by a normal orientation of crack path (Zi & Belytschko, 2003). Asferg et al. (2007)
investigated also the role of crack length in design of the load- deformation plot.

Furthermore, in TPBT, a linear law for the representation strain-softening diagram was
assumed, in contrast to the bilinear one used by Zi & Belytschko (2003). Asferg et al. (2007)
mentions that the partly cracked element gives out smother response in contrast with the
fully cracked scheme. In particular, the maximum load capacity and overall shape of the
diagram load-deformation are found quite same. Asferg et al. (2007) proposed a structured
mesh 25 x 24 elements and an unstructured of 709 elements (fig. 9).

Fig. 9: Structured and unstructured mesh schemes and corresponding load-deflection plots (Asferg et al., 2007)

However, the post peak response for fully cracked scheme shows a zig-zag orientation for
the load, because the stresses are taken equal to the tensile strength of the material and
the criterion for crack propagation is based on local stresses of the element and not of the
average value of the overall stresses.

All in all, due to Asferg et al. (2007) five basic points need further investigation a) multi-
cracking, b) stress representation, c) intersecting crack management and d) relationship
between concrete and reinforcement In particular, the model used depends on the analysis
of a single crack initiate and propagate within reinforced concrete bulk. In real concrete
structures the bulk is subjected to multi cracking. Secondly, stress representation needs

22
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

improvement. In addition, the interface should be capable to manage intersecting cracks.


The effects of the confining pressure needs identification prior the simulation of the crack
model in order to represent better the relationship between concrete and reinforcement.

In 2008, Cox, analyzed extensively the general concepts of cohesive cracks and in particular
the enrichment functions, in order to improve the representation of the displacement
gradients around the cohesive model providing a smooth development of crack path across
each element. In his work, Cox, includes meshing refinement and direction, and propagation
of crack within the FPZ varied over time. The model described mode I and mix-mode
situation.

Karihaloo & Xiao (2011) combined the cohesive crack model with the XFEM and Williams’s
expansion for asymptotic fields. They mentioned that the XFEM is capable to provide
improved results in contrast with FEM, only when the asymptotic field of the crack tip is
given and used as a function of enrichment. This concept allows the use of coarser meshing
around the crack-tip.

Mougaard et al. (2011) used the exact model of Asferg et al. (2007) replacing the constant
strain triangles to linear strain triangular, in order to achieve high order elements. In
particular, the work of Mougaard et al. (2011) involves side local shape functions, variational
formulation, representation of equal stresses in both sides of crack tip and the crack
propagation when the maximum tensile stress in the crack-tip overcome the tensile
strength, leading to normal orientation of propagation perpendicular to principal stress.
However, the differences occur in the method of enrichment of the displacement field, the
restrictions in the crack-tip for the transition from partly cracked to full cracked elements,
introduction of higher order elements based on linear strain triangles (fig. 10) and bilinear
softening curve representation. These modifications led to improvement of stress accuracy
within the crack-tip element, more acceptable transition from partly cracked to full cracked
elements and smoother crack closure avoiding the zig-zag orientation of the post-peak part
of the load-deformation plot for three-point bending test (TPBT).

23
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

Fig.10: Adaptive Method and additional enrichment scheme (Mougaard et al., 2011)

A3.2. Description and Context


A.3.2.1 Introduction of Research Topic
To begin with the main scientific problem is the representation of the complex phenomenon
of cracking in reinforced concrete structures. As a matter of fact, the ultimate strength of
the concrete depends on the macroscopic heterogeneity of the mixture (Fig. 11). There is a
great amount of work done for concrete bulk, but the same does not happen for the
reinforced concrete.

Fig. 11: Macroscopic analysis of cylinders and illustration of bond zone ( Asferg et al., 2007; Lundgren, 1999)

In particular, the basic difference is the requirement for representation of the bond zone,
intersection between steel and concrete, in the later (fig. 11). In order to achieve that, we
proposed a model based on the previous work of Asferg et al. (2007), using a rectangular
six-node interface element.

A.3.2.2 Summarization of Past and Current Work


For the representation of interface element, the previous work of Ngo & Scordelis (1967)
and Lundgren (1999) will be used. The former is the fundamental work done for
representing reinforced concrete by analyzing a rectangular beam under TPBT. Lundgren
(1999) investigated the interaction of steel and concrete, and proposed an interface
element simulated numerically using finite element procedures. It is very important to
mention, that the basic drawback of interface models is the application of the method only

24
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

for already pre-defined geometries (fig 12a). In this case, the concepts are not completely
applicable, due to the fact that our main is to model the initiation and propagation of
randomly generated cracks.

Fig. 12: Limitations of interface elements, cohesive elements and equal stress representation (Exadaktylos,
2006; Sluys & Wells, 2001; Asferg et al, 2007)

As we mentioned before, there is a great amount of work done for the analysis of concrete
bulk. The most relative to our model works are those of Zi & Belytschko (2003), Asferg et al.
(2005;2006;2007) and Mougaard et al. (2011). All these projects used a cohesive CST
scheme able to transform from partly to fully cracked, in order to accommodate the
longitudinal propagation of the crack. The main drawback of Zi & Belytschko (2003) is the
incapability to represent equal stresses in both sides of the crack (fig 12c). Asferg et al.
include equal stresses and Mougaard et al. (2011) improved further the model by using
high-order adaptive method in a LST, using extra enrichment of the nodes and
discontinuous degrees of freedom. By adaptive method, we mean the sub-triangulation of
the triangulate element for improved stress representation.

A.3.2.3 Description
The model we propose is based on two elements: a cohesive nine-node rectangular element
for the propagation of a straight crack within the bulk of concrete, able to transform from
partly to fully cracked and a six-node interface able to represent the intersection between
steel and concrete (Fig. 13). With further improvement the interface element will be able to
represent also crack intersection within the bond zone.

Fig. 13: Combination of cohesive and interface elements and interface modeling procedure

25
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

The basic properties of this model are: firstly the allowance for longitudinal crack growth in
the cohesive element, representing the macroscopic heterogeneity of the concrete bulk and
transform from a partly to a fully cracked element (Fig. 13). The crack propagates reaching
the bond zone and intersect with the crack caused by the tensioned region of concrete
beam (under reinforced bar). The latter is simulated using the interface element. Secondly,
there is the ability of interface element to hold the crack-tip inside the element. To achieve
that the displacement field is divided into two parts, a continuous and discontinuous. The
continuous is normal having standard nodal and degrees of freedom enrichment, while the
discontinuous is divided in two topologies. The first topology includes the crack and is
enriched only in the degrees of freedom of the cut points, while in the remaining part all the
nodes are set to zero (Fig. 13c). The most important fact is that even the crack-tips in both
sides of the crack are set to zero (Asferg et al., 2007)

Fig. 14: Crack propagation criterion and approximation for fracture energy (Asferg et al, 2007; Karihaloo, 1995)

However, there is that smooth crack closure criterion is available (fig. 14). Asferg et al (2007)
mentioned that the model is based on the previous assumptions of Wells & Sluys (2001) and
Asferg et al. (2006), in which the crack propagates element by element in a straight
orientation without any restriction. However, he also mentions that a criterion should be
added in order to check that always the maximum stress on the crack-tip is equal to the
tensile strength of the material for the crack to propagate.

A.3.2.4 Aims and Objectives


As we mention before, the introduction of a smooth closure criterion arise. Therefore, we
propose the use of the concept introduce by Zi & Belytschko (2003), which mentions that
for crack propagation and smooth closure the stress projection in the crack tip should be
equal to tensile strength of the material and perpendicular to the orientation of principal
stresses.

26
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

Secondly, we will change the shape of the tension softening diagram, in order to verify the
aspect mentioned by Karihaloo (1994), in which in some applications the shape of curve is
unimportant. RILEM, adopted the Hillerborg & Petersson (1983) simple approximation for
linear shape but now we will use bilinear, polynomial and exponential approximation for
σ(w). This procedure can be implemented by a simple adjustment of fracture energy
0 wc

governing equation: GF   w ( )d     (w )d w (see also fig.)


ft ' 0

Fig. 15: Intersection of cracks leads to bifurcation and complete failure (Géniaut, 2011)

In addition, there will be an extensive investigation for the crack intersection taking place
inside the interface element. This investigation will lead to multi-cracking phenomena under
mix-mode conditions. As a matter of fact, works of Asferg et al (2005;2006;2007) and
Mougaard et al. (2011) mentioned the requirement for multi-cracking and mix-mode
analysis, because is a phenomenon highly observed in reinforced concrete structures. Our
investigation will focus mostly in previous works of Daux et al. (2000), Mohammadi (2008)
and shear-lag effect proposed by Kristek & Bazant (1987). However, it is important to
understand that this investigation will be able to provide some direction for further study.

Finally, TPBT will be implemented using Matlab. The test will ask for dimensions of beam
and fracture energy as inputs which mean that the fracture energy will be estimated prior
and then introduced in the test. Furthermore, TPBT will be used for the comparison of the
results between the proposed model and previous projects of Asfeg et al., in order to verify
our model and be able to improve it further.

A.3.2.5 Programme and Methodology


Basically, the project can be divided in three main phases: 1) analysis, coding and numerical
implementation, 2) further improvement, evaluation of results and further study
proposition and finally 3) final correction, preparation and submission of project.

27
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

The first phase includes the theoretical examination- investigation of Asferg et al. (2007)
model and the numerical implementation using Matlab, which will be based on the
computational analysis of the cohesive and interface XFEM elements. As the numerical and
computational implementation is in reference to real structures, TPBT will be used for a
crack under mode I fracture.

The second phase includes the collection of analysis data and the extensive evaluation of
the parameters, in order to compare the results with those of Asferg et al. (2007). Finally, in
the last part there will be a synopsis about the finding of the researched issues plus
proposition for further study. Finally, the third includes writing, final correction, printing,
and submission.

There is only one milestone in the progression of the project. At 27/07/2013 the analysis of
the model must be completed, in order to collect the data of the analysis, evaluate them
and progress to introduction of improvements and investigation of intersecting and multi-
cracking phenomena. This is millstone of great importance and need to be achieved, in
order to have at least one month to focus on the investigation of intersecting phenomena.
For further information see the analytic Gantt plan below.

28
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

29
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

A.3.3. Context
A.3.3.1 Justification of Choice
To begin with FEM, for discrete inter-element crack models the analysis is based on a pre-
existing crack geometry incorporated in the model (Mohammadi, 2003). Unger et al (2007),
mentions that there are two alternatives for the use of finite elements in concrete. Firstly,
by introducing interface zero-thickness elements and secondly, by introducing interface
elements between the edges of the elements. The first contribute to remeshing dependency
and the second to mesh dependency. Thus, he points out that XFEM is capable of providing
a solution independent of mesh and remeshing settings called as ‘elegant solution’. The
same is also pointed out by Xiao & Karihaloo (2011) using available asymmetric fields.

Discrete crack element model does not include remeshing of the global model and scaling
local remeshing for the propagation of the crack. This have as a consequence the use of a
global remeshing matrix and meshing becomes very distorted, because it includes elements
with bad aspect ratios due to multiple cracking (Mohammadi, 2003)

Smeared is based on the concept of traction to opening curve law and there is no need of
remeshing, which is governing the crack propagation (Mojiri, 2010). Borst et al. (2004),
Ohmenhäuser et al. (1998), Bažant & Ozbolt (1990) and Bažant et al (2000) introduced four
smeared cracks: fixed single, rotating, multiple fixed and micro planes. The first three
models, apply only for pre-defined geometries and the last overestimated the damping ratio
and unloaded stiffness (Nguyen & Chan, 2005), leading to mesh incompatibility. Vecchio &
DeRoo (1995), mention that smeared crack was not capable to cover concrete tension
splitting mechanism contributing to global and lateral dilation. Both discrete and smeared
models include adaptive methods introduced by Ohmenhäuser et al. (1998) for mesh
regeneration, which found to be quite insufficient to provide exact results (Nguyen & Chan,
2005).

Cohesive crack model using XFEM provides generalization of crack orientation and
development by zeroing KIC. In contrast with FEM, X-FEM includes local enrichment of
functions, providing classical displacement settings and meshing concepts (Mojiri, 2010).

30
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

The main reasons for selecting XFEM is the elimination of meshing and remeshing. This can
also be achieved using Galerkin (Belytschko et al, 1996), however the method is
incompatible with finite element procedures (Asferg, 2007). The cohesive scheme is able to
transform partly to fully cracked elements, due to the fact that allow the longitudinal
propagation of crack (Asferg et al, 2007; Mougaard et al. 2011). In addition, the interface
element is a discrete method able to represent the interaction between steel and concrete.
The previous works of Ngo & Scordelis (1967) and Lundgren (1999) using finite element
procedures are highly compatible with XFEM. In addition, the ability for crack-intersecting
within the interface element provides us with opportunities for further investigation and
progression of this model.

The existence of previous theoretical and computational models relative to the cohesive
partly-cracked model XFEM element and interface element will provide us with a practical
and theoretical background for the implementation of this project. For the numerical
simulation Matlab will be used establishing the procedure step by step.

A.3.3.2 Explanation of Academic and Industrial Relevance


The need for further study and increase of available funds for researching, are two of the
main issues a university should focus on. In this direction the education functions in the
frame of demand and supply. Education is an investment for every student and a step for its
future carrier and professional recognition. The research is a part of this investment and
constitutes a presumption of knowledge and specialization.

With regard to industry, it is necessary to attempt in the direction of education, because it is


the only solution for further development and prosperity. Every business needs to provide
financial support to young scientists and professionals to develop their careers and improve
their level of knowledge. Therefore, the students will find an ally in making their future
professional goals.

As we mentioned before, the main scientific problem is the complex phenomenon of


cracking in reinforced concrete. In this direction, the proposed work will try to validate an
existing theoretical model and improve it, up to the extent of possible. In particular, the
31
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

validation of results of a previous study will help both the creator of the model, and future
scientists in the implementation of their scientific work.

In addition, we shall mention the benefits that will be gained by the student doing the
thesis. The use of analytical numerical procedures and computational software combined
with bibliographic reference and extensive study, will contribute to in-depth knowledge of
the theoretical background and understanding of such a complex phenomenon as crack
analysis of concrete. The implementation of this master thesis will be a further step of his
future professional carrier.

All in all, the proposed research is an attempt for further specification in a wide scientific
field such as concrete behavior, which will be helpful for both the student and the scientific
community.

A.3.3.3 Further Research Propositions


As we mentioned before there is of great importance to investigate the possibility of multi-
cracking phenomena, mix-mode conditions and crack intersection inside the interface
element, which are able to provide us with a model representing a more real time situation.

In addition, Rabinovich et al. (2007) introduced a method for crack detection and
identification in conjunction with non-destructive methods and more specifically response
of the structure to vibration in chosen frequency or combination of frequencies. Used to
estimate if the structure contains a crack and define the parameters as location, size
orientation and shape based on response data. For the implementation of the method a
genetic algorithm is used describing an inverse problem. In other words, a database of pre-
existing crack geometries and characteristics scanned in order to find the most suitable to
the structural response. The geometries scanned are implemented through XFEM analysis of
Zi & Belytschko (2003).

Interface elements as described by Lundgren (1999) are able to represent the intersection
between concrete and reinforcement. As a matter of fact, the interface model of Lundgren
(1999) it was able to analyze already pre-defined geometries. Thus, a combination between
32
Case Study: XFEM based Partly Cracked Interface Model for Reinforced Concrete Analysis

the model of Rabinovich et al. (2007), Lundgren (1999) and Asferg et al. (2007) including
further improvements will be a powerful tool of the representation of cracking in reinforced
concrete structures. In particular, we can add our work it to Rabinovich`s et al. (2007)
generating a database of existing crack geometries that will be easily update by repetition of
crack settings. The interface element can give us much information to enrich the database
of Rabinovich et al. (2007).

All in all, the result is the generation of a procedure able to estimate the strength of existing
reinforced structures including intersecting cracks, using the dynamic analysis of frequencies
proposed by Rabinovich et al. (2007) and also can predict the initiation and propagation of
reinforced concrete cracks for newly build structures.

33
List of Figures

Fig. 1 Microcracks in marble specimen under unixial cyclic loading at 75x and 1500x
respectively. Figure reproduced from Exadaktylos. (2006)
Fig. 2 Opening, sliding, tearing and sharp crack of length 2a in an infinite elastic
body. Figure reproduced from Shi. (1995)
Fig. 3 Typical length of fracture process zone for various materials. Figure
reproduced from Karihaloo (1995).
Fig. 4 Dugdale et al. crack model and rate of non-dimensional σ/σγ relation. Figure
reproduced from Mohammadi. (2009)
Fig. 5 FCM model, where a0 is the traction zone, assumed elastic and lp is the FPZ,
which includes dissipation mechanisms contributing to fracture energy G F.
Figure reproduced from Karihaloo; Shi. (1995; 2009)
Fig. 6: Crack path between aggregates, corresponding actual and smoothed path,
band crack model and corresponding stress-strain diagram. Figure
reproduced from Mohammadi. (2008)
Fig. 7 Multi-scale concept and illustration of crack behaviour. Figure reproduced
from Needleman; Strange et al. (1990; 2006b)
Fig. 8 Variation of smeared crack models. Figure reproduced from Ohmenhauser et
al. (1998)
Fig. 9 Structured and unstructured mesh schemes and corresponding load-
deflection plots. Figure reproduced from Asferg et al. (2007)
Fig. 10 Adaptive Method and additional enrichment scheme. Figure reproduced
from Mougaard et al. (2011)
Fig. 11 Macroscopic Analysis of cylinders and illustration of bond zone. Figure
reproduced from Asferg et al. (2007) and Mougaard et al. (2011)
Fig. 12 Limitations of Interface elements, cohesive elements, and equal stress
representation. Figure reproduced from Exadaktylos (2006), Sluys & Wells
(2001) and Asferg et al. (2007
Fig. 13 Combination of cohesive and interface elements and interface modelling
procedure. Figure reproduced from Asferg et al. (2007)
Fig. 14 Crack propagation and approximation for fracture energy. Figure reproduced
from Asferg et al. (2007) and Karihaloo (1995)
Fig. 15 Intersection of cracks leads to bifurcation and complete failure. Figure
reproduced from Géniaut (2007).
References

Van Mier, J.G.M. 1997. Fracture Process of Concrete: Assessment of Material Parameters for
Fracture Models. New York: CRC Press.

Karihaloo, B. L. 1995. Fracture mechanics and structural Concrete. Essex: Longman Scientific
and Technical.

Mohammadi, S. 2008. Extended Finite Element Method for Fracture Analysis of Structures.
Iran: Blackwell Publishing.

Shi, Z. 2009. Crack Analysis in Structural Concrete. London: Elsevier Ltd.

Williams, M. S & Todd, J. D. 2000. Structures theory and Analysis. London: Macmillan Press
Ltd

Bansal, R. K. 2001. A textbook of strength of Materials. New Delhi: Laxmi Publications (P) Ltd

Elfgren, L. and Shah, S.P. 1991. RILEM: Analysis of Concrete Structures by Fracture
Mechanics. UK: Chapman and Hall.

Exadaktylos, G.E., 2006, Introduction in the theory of elasticity and Fracture Mechanics, Civil
Engineering Module lecture notes. University of Crete, Greece.

Panagopoulos, G. and Kirtas, E., 2005, Reinforced Concrete Structures I lecture notes.
Technological Institute of Serres, Greece.

Coates, R. C. et al. 1988. Structural Analysis. London: Chapman & Hall

Kundu, T. 2008. Fundamentals of Fracture Mechanics. London: CRC Press


European Committee for Standardization. 2005. EN 1992-2: 2005(E) Eurocode 2: Design of
Concrete Structures Part 2 Design and Detailing Rules. Brussels: CEN

Dugdale, D.S. 1960. Yielding of steel sheets containing slits. Journal of the Mechanics and
Physics of Solids 8, pp. 100–104.

Barenblatt, G.I. 1962. The mathematical theory of equilibrium of cracks in brittle fracture.
Advances in Applied Mechanics 7, pp. 55–129.

Hillerborg, A. et al. 1976. Analysis of crack formation and crack growth in concrete by means
of fracture mechanics and finite elements. Cement and Concrete Research 6, pp. 773–782.

Bažant, Z. P. 1984. Size effect in blunt fracture: Concrete, rock, metal. Journal of Engineering
Mechanics (ASME) 110 (4), pp. 518-535

Bažant, Z.P and Li, Y.N. 1995. Stability of cohesive crack model: part II—Eigenvalue analysis
of size effect on strength and ductility of structures. Journal of Applied Mechanics (ASME)
62, pp. 965–969

Bažant, Z.P. 2000. Size effect. International Journal of Solids and Structures 37,pp. 69-80.

Bažant, Z.P. 1997. Scaling of quasi-brittle fracture: asymptotic analysis. International Journal
of Fracture 83 (1), pp. 19-40.

Bažant, Z.P. and Ozbolt, J. 1990. Non-local Microplane model for fracture, damage and size
effect in structures. Journal of Engineering Mechanics (ASCE)116(11), pp. 2485- 2504

Zubelewicz, A. and Bažant, Z.P. 1987. Interface element modeling of fracture in aggreagate
composites. Journal of Engineering Mechanics (ASCE) 113, pp. 1619-1630

Van Mier, J. et al. 1993. An experimental and numerical study of mode I (tensile) and mode
II (shear) fracture in concrete. Journal of Mechanical Behavior of Materials 4, pp. 179-190
Ozbolt, J. and Bažant, Z.P. 1992. Microplane Model for cyclic triaxial behavior of Concrete.
Journal of Engineering Mechanics (ASCE) 118 (7), pp. 1365-1386
Wells, G.N. and Sluys, L. J. 2001. A new method modelling cohesive cracks using finite
elements. International Journal for Numerical Methods in Engineering 50, pp. 2667-2682

Weibull, W. 1951. A statistical distribution function of wide applicability. Journal of Applied


Mechanics 18 (3), pp. 293-297

Vecchio, F.J and Collins, M. P. 1986. The modified compression field theory for reinforced
concrete elements subjected to shear. Journal of Engineering Mechanics (ACI) 83 (2), pp.
219-231

Vecchio, F.J. 1992. Finite Element Modelling of concrete expansion and confinement.
Journal of Engineering Mechanics (ASCE) 118 (9), pp. 2390-2406

Vecchio, F.J. and DeRoo, A. 1995. Smeared- Crack Modelling of Concrete Tension Splitting.
Journal of Engineering Mechanics (ASCE), pp. 702-708

Patzák, B & Jirásek, M. 2002. Process zone resolution by extended finite elements. Research
gate [Online]. Available at: http://www.researchgate.net/publication
/222026640_Process_zone_resolution_by_extended_finite_elements?ev=pub_cit
[Accessed: 19 April 2013]

Bažant, .Z.P. and Cedolin, L. 1983a. Finite element modeling of crack band propagation,
Journal of Structural Engineering Mechanics (ASCE) 109 (2), pp. 69-92.

Bažant, .Z.P. and Gambarova, P. 1980. Rough cracks in reinforced concrete. Journal of
Structural Engineering Mechanics (ASCE) 106, pp. 819-842

Bažant, .Z.P. and Gambarova, P. 1984. Crack shear in concrete: crack band microplane
model. Journal of Structural Engineering (ASCE) 110, pp. 2015-2035
Zi, G. and Belytschko, T. 2003. New crack-tip Elements for XFEM and applications to
Cohesive Cracks. International Journal for Numerical Methods in Engineering 57, pp. 2221–
2240

Mergheim, J. et al. 2005. A finite element method for the computational modelling of
cohesive cracks. International Journal for Numerical Methods in Engineering 63, pp. 276-289

Asferg, J.L. et al. 2005. Cohesive Crack-Tip Element for XFEM. Proceedings of the 11th
International Conference on Fracture [Online]. Available at:
http://www.scribd.com/doc/101476999/Cohesive-Crack-Tip-Element-Xfem [Accessed: 19
April 2013].

Unger, J. L. et al. 2006. Modeling of cohesive crack growth in Concrete Structures with the
extended finite element method. Computational Methods and Applications in Mechanical
Engineering 196, pp. 4087-4100

Karihaloo, B. L. and Xiao, Q. Z. 2006. Improving the accuracy of XFEM crack tip fields using
higher order Quadrature and statically admissible stress recovery. International Journal for
Numerical Methods in Engineering 66, pp. 1378-1410

Asferg, J.L. et al. 2007. A direct XFEM formulation for modeling of cohesive crack growth in
concrete. Computers and Concrete Journal 4(2), pp. 83-100

Bažant, Z.P., and Oh, B.H. 1983. Crack band theory for fracture of concrete. Materials and
Structures (RILEM) 16, pp. 155-177.

Asferg, J.L. et al. 2007. A consistent partly cracked XFEM element for cohesive crack growth.
International Journal for Numerical Methods in Engineering 72, pp. 464-485

Cox, J.V. 2009. An extended finite element method with analytical enrichement for cohesive
crack modeling. International Journal for Numerical Methods in Engineering 78, pp. 48-83
Daux ,C. et al. 2000. Arbitrary branched and intersecting cracks with the
extended finite element method. International Journal for Numerical Methods in
Engineering 48, pp. 1741–1760.

Karihaloo, B. L. and Xiao, Q. Z. 2011. Accurate simulation of mixed-mode cohesive crack


propagation in quasi-brittle structures using exact asymptotic fields in XFEM: an overview.
Journal of Mechanics of Materials and Structures 6 (1-4), pp. 267-276

Mougaard, J. F. et al .2011. A Consistent Partly and Fully Cracked XFEM Element for
Modelling. International Journal for Numerical Methods in Engineering 85 (13), pp. 1667-
1681

Belytschko, T. et al. 1996. Meshless methods: an overview and recent developments.


Computer Methods in Applied Mechanics and Engineering 139, pp. 3–47.

Dvorkin, E.N. et al. 1990. Finite elements with displacement interpolated embedded
localization lines insensitive to mesh size and distortions. Computer Methods in Applied
Mechanics and Engineering 90, pp. 829–844.

Mellenk, J.M. and Babŭska, I. 1996. The partition of unity finite element method: basic
theory and application. Computer Methods in Applied Mechanics and Engineering 139, pp.
289–314.

Belytschko, T. and Black, T. 1999. Elastic crack growth in finite elements with minimal
remeshing. International Journal for Numerical Methods in Engineering 45(5), pp. 601–620.

Möes, N. et al. 1999. A finite element method for crack growth without remeshing.
International Journal for Numerical Methods in Engineering 46:131–150.

Stolarska, M. et al. 2001. Modeling crack growth by level sets in the extended finite element
method. International Journal for Numerical Methods in Engineering 51, pp. 943–960.
Wells, G. N and Sluys, L. J. A new method for modelling of cohesive cracks using finite
elements. International Journal for Numerical Methods in Engineering 50(12), pp. 2667–
2682.

Möes, N. and Belytschko, T. 2002. Extended finite element method for cohesive crack
growth. Engineering Fracture Mechanics 69, pp. 813–833.

Asferg, J. L. 2006. Modeling of Concrete Fracture Applying the eXtended Finite Element
Method. PhD Thesis, Technical University of Denmark

Lundgren, K. 1999. Three-dimensional Modeling of Bond in Reinforced Concrete: Theoretical


Model, Experiments and Applications. PhD Thesis, Chalmers University of Technology

Rabinovich, D. 2007. XFEM-based crack detection scheme using a genetic algorithm.


International Journal for Numerical Methods in Engineering 71, pp. 1051-1080

Mojiri, S. 2010. Numerical Analysis of Cohesive Crack Growth Using Extended Finite Element
Methos (X-FEM). MSc Thesis, Ecole Centrale de Nantes (ECN)

Song, S. et al. 2006. Simulation of crack propagation in asphalt concrete using an intrinsic
cohesive zone model. Journal of Engineering Mechanics (ASCE) 132, pp. 1215-1223

Rabczuk, T. et al. 2005. A numerical model for reinforced concrete structures. International
Journal of Solids and Structures 42 (506), pp. 1327-1354

Feenstra, P. H. and de Borst, R. D. 1996. A composite plasticity model for concrete.


International Journal of Solids and structures 33 (5), pp. 707-730

Frichant, S. et al. 1999. Isotropic and anisotropic descriptions of damage in concrete


structures. International Journal of Mechanical Cohesive Frictional Materials 4, pp. 339-359

Вам также может понравиться