Influence of stress state dependency on ductile

fracture modeling in standard test specimens using

LS-dyna explicit analysis

A thesis submitted to the

Graduate School of the University of Cincinnati

in partial fulfillment of the requirements for the degree of

Master of Science

in the Department of Mechanical, Industrial and Nuclear Engineering

of the College of Engineering and Applied Sciences

November 2014

by

Jonny M. Shelke

Bachelor of Engineering (B.E)

University of Pune, India, 2008

Thesis Advisor and Committee Chair: Dr. Ala Tabiei

 Abstract

Damage and fracture in engineering materials are a complex phenomenon. Accurate modeling &
simulation requires close approximation of standard test conditions and definition of parameters
influencing the phenomenon with optimization. An implementation of a strain-based, stress state
dependent failure criterion for a ductile material is validated through simulations of standard
experiments.

For a strain-based failure criterion, the strain required to reach fracture depends on the material,
rate of loading, temperature and the state of stress. The rate of loading or strain rate and
temperature dependence has been extensively studied by many researchers. Here the effect of
stress state on ductile fracture is given importance to achieve accurate results for quasi-static
loading without temperature influence. Studying the stress state dependency of the strain to failure
is the main objective of this thesis. The state of stress is defined by triaxiality (hydrostatic pressure)
and the lode angle parameter (deviatoric stress third invariant). The stress state not only varies
with different kinds of loading but also with the location of the stress concentration, geometry of
specimen and deformation. It is for this reason that exact original shape of a specimen before
plastic deformation cannot be achieved by simply reversing the same forces causing the
deformation. The difference lies in the stress state defined by pressure and lode angle.

Fracture surface in the principal stress space based on the triaxiality and lode angle parameter is
implemented in the damage model GISSMO (Generalized Incremental Stress State dependent
MOdel). Experimental data from the tests on Aluminum alloy 2024-T351, a material used widely
in aerospace applications is validated through different specimen tests covering combinations of
stress states. GISSMO is also implemented & tested for 1045 steel and the already validated Dual
Phase 600 steel to evaluate correctness of the present approach. Apart from simulating the correct
mode of fracture, a satisfactory match between experimental and simulation results is achieved.

ii
iii
Acknowledgments

First and foremost I would like to express my deeply felt gratitude towards my thesis advisor Dr.
Ala Tabiei for his constant guidance throughout this entire project with his technical expertise,
immense experience and timely feedback. I am thankful to him for believing in me and helping
me improve the implementation patiently even in his busy schedule. Without his ideas, initiative,
motivation and support I could not have completed this kind of work.

I would like to thank Dr. David Thompson and Dr. Yijun Liu for taking the time to review my
work and serve on the thesis committee. Their valuable comments have proved to be constructive
in applying my best effort to achieve the objectives of this research.

The support received from Ohio Supercomputer Center staff has been commendable in resolving
the issues faced while carrying out the simulations. I am thankful to them for extending the
efficient computing resources. A special mention for the professional curricular practical training
on LS dyna and Hypermesh software tools provided by Tech Americas Inc. I am also indebted to
the University of Cincinnati for awarding the University Graduate Scholarship for my masters
education. The UC libraries and office of college computing were always of great assistance.

I am thankful to my colleagues Priyanjali Chatla, Sandeep Medikonda, Pushkaraj Bhagwat, Nikhil

Noronha, Abhijit Dani, Bhargava Sista, Sudarshan Kasera and Himneet Singh for the fruitful
discussions and help received even during casual conversations as the work progressed.

Finally, I am very thankful to my family for being extremely patient and standing by me throughout
my masters program.

iv
Contents
1 Introduction…………………………………………………………………………….……1
1.1 Background and Motivation………………………………………………………..…1
1.2 Research Scope & Objectives………………………………….……………………..3
1.3 Outline of Thesis……………………………………………………………………...4

2 Literature Review…………………………………………………………………..……….6
2.1 Existing Failure/Damage Modeling Approaches……………………………………..7
2.1.1 Stress dependent failure criteria……………………………………….…..8
a) Maximum principal stress criterion…………………………………...….8
b) Tresca’s failure criteria…………………………………………….…….9
c) Mohr’s failure criterion…………………………………………...……..10
d) Von Mises’s failure criteria……………………………………………...11
e) Maximum shear stress criterion (MS)……………………………..…….13
2.1.2 Strain dependent failure criteria………………………………………….14
a) Maximum principal strain criterion…………………………………..…14
b) Constant equivalent strain criterion……………………………….…….14
c) Forming Limit Diagram (FLD)……………………………………...…..14
2.1.3 Cockcroft-Latham…………………………………………………….….16
2.1.4 CrachFEM………………………………………………………………..17
2.1.5 Damage models…………………………………………………………..18
a) Gurson…………………………………………………………...………19
b) Johnson-Cook (JC)………………………………………………..……..22
c) Wilkins…………………………………………………………………..23
d) Xue-Wierzbicki……………………………………………...…………..24
e) Continuum damage mechanics (CDM)……………………….………....27
2.2 Failure/Damage models from LS dyna materials library……………………………28

3 Stress state, failure criterion and implementation………………………………….….…30

3.1 Stress state characterization……………………………………………….…..……..31
3.2 Stress state influence on equivalent plastic strain to failure………………………….36
3.2.1 Pressure or triaxiality dependence η………………………………..……36
3.2.2 Lode angle parameter dependence ξ……………………………………..38
3.3 Typical specimen geometries for approximate deviatorically proportional
loading……………………………………………………………………………….39
3.3.1 Axisymmetric tension……………………………………………………39
3.3.2 Uniaxial compression…………………………………………………….41
3.3.3 Pure shear…………………………………………………………..…….43
3.4 GISSMO model…………………………………………………………………..….46
3.5 Plasticity, damage & failure ……………………………………………..………….47

v
 3.5.1 Plasticity modeling…………………………….…………………………47
3.5.2 Damage accumulation………………………………………………...….48
3.5.3 Failure prediction…………………………………………………..…….50
3.6 LS dyna implementation and material keyword input………………………………50

4 Modeling, simulation of experiments and results…………………………….…….......…53

4.1 2024-T351 aluminum alloy……………………………………..…………….……..54
4.1.1 Tensile specimens………………………………………….……….……….56
a) Un-notched round 9 mm diameter specimen…………………………..58
b) Notched round 18 mm notch radius specimen………………….……...60
c) Notched round 9 mm notch radius specimen…………………………..61
d) Notched round 4.5 mm notch radius specimen………………..……….63
4.1.2 Upsetting specimens………………………………………………………...64
a) 11.25 mm long upsetting specimen………….……………………...….66
b) 8 mm long upsetting specimen…………………………….…….……..69
4.1.3 Grooved flat plate specimen………………………………………...………70
4.1.4 3-point bending specimens………………………………………………….75
a) 10 mm thick 3-point bending specimen………………………………..75
b) 30 mm thick 3-point bending specimen………………………………..78
c) 60 mm thick 3-point bending specimen………………………………..79
4.1.5 Projectile impact and ballistic limit………………………….………..……..81
a) 0.063” thick 2024-T351 plate…………………………….……….……83
b) 0.125” thick 2024-T351 plate…………………………….…………….85
c) 0.25” thick 2024-T351 plate…………………………….……...………86
4.2 1045 steel……………………………………………………….………………..…..90
4.2.1 Tensile specimens………………………………………………………..92
a) Un-notched round tensile specimen………………………………...….93
b) 10.5mm radius notched tensile specimen……………………………...94
4.2.2 Grooved flat plate specimen……………………………………………..96
4.2.3 Torsion specimen…………………………………………………...……98
a) Pure torsion test…………………………………………………………..99
b) Tension test of torsion specimen…………………………………..……101

5 DP600 steel experiments, damage exponent effect and regularization ...……..…...….104

5.1 DP600 steel specimen experiments……………………….……………………..…105
5.1.1 Tensile specimens………………………………………………………….107
a) Un-notched flat specimen…………………………………………….108
b) Notched flat 2 mm notch radius specimen…………………………....110

vi
 c) Notched round 0.5 mm notch radius specimen……………….………112
d) Notched round 2 mm notch radius specimen…………………………114
e) Notched round 4 mm notch radius specimen………………………....115
5.1.2 Grooved flat specimens……………………………………………………117
a) 0.5 mm radius grooved flat plate specimen…………………………..117
b) 1 mm radius grooved flat plate specimen…………………………….119
5.2 Effect of damage exponent variation………………………………………………124
5.3 Mesh size based regularization…………………………………………...………..126
5.3.1 2024-T351 un-notched round tensile test………………………………….127
5.3.2 1045 steel un-notched round tensile test…………………………..………130
5.3.3 DP600 steel un-notched flat tensile test…………………………………...134

6 Conclusions and future work…………………………………………………………….138

6.1 Advantages of presented J2 plasticity and GISSMO implementation……………..140
6.2 Limitations of presented J2 plasticity and GISSMO implementation……………..141
6.3 Future scope………………………………………………………………………..142

Bibliography…………………………………………………………………………………...143

vii
List of Tables
2.1: Failure models from the LS dyna material library…………………..................….........…...29
3.1: Triaxiality and Lode parameter values for typical loading…………………….……..…..…44
3.2: Ten classical specimens for plasticity and fracture calibration…………………….…..……45
4.1: 2024-T351 material properties……………………………………………………….……...54

4.2: Specimens used to calibrate the fracture locus using a six parameter analytical formulation by
Xue……………………………………………………………………………….…..….……….54
4.3: Calibrated parameters for symmetric 2024-T351 fracture locus……………………...…….55
4.4: Six calibrated parameters for 2024-T351 fracture surface………………………………….55
4.5: Material properties for 1045 steel………………………………………………….……..…91
4.6: Six calibrated parameters for 1045 steel fracture locus by Bai……………………….…….91
4.7: Initial values of stress state parameters for the 1045 steel experiments…………………....92
4.8: Dimensions of 1045 steel grooved flat plate with 3.97 mm radius notch…………………..98
5.1: Material properties used for DP600 steel…………………………………………….……105
5.2: Parameters for the DP600 analytical fracture surface……………………………….…….106
5.3: DP600 experiments selected and their typical stress state parameters…………………….107

5.4 Mesh sizes for regularization of tensile tests using different materials…………….………127

viii
List of Figures
1.1: Comparison of effective stress versus equivalent strain curves for tension, compression and
torsion experiments on 2024-T351……………………………………..…………………………2

2.1: Classification of phenomenological failure/damage models…………………………………8

2.2: Maximum principal stress criterion…………………………………………………………..8

2.3 Tresca’s and Von mises failure criterion in plain stress case………………………………....9

2.4: Mohr-Coulomb failure criterion………………………………………………………….…10

2.5: Tresca & Mohr-Coulomb fracture criteria in principal stress space……………….………..11

2.6: Von mises & Drucker-Prager failure criterion in 3D principal stress space………………..12

2.7: Yield surface in principal stress space for Tresca & Von mises…………………………....12

2.8: Π-plane or deviatoric plane representation of classical yield functions…………….……....13

2.9: Forming limit Diagram…………………………………………………….…….………….15

2.10: Transformation of the FFLD fracture locus into the space of (ԑf ; η)………………….…..16

2.11: Cavity nucleation rate………………………………………………………………….…..21

2.12: Xue’s symmetric fracture surface in the space of ԑ and triaxiality……………………….26

2.13: Xue’s 3-dimensional fracture surface in the principal stress space…………………….….26

3.1: Representation of vector in Haigh-Westergaard principal stress space ( , , ) and

cylindrical coordinates (r, θ, z) ………………………………………………………………….32

3.2: Geometrical representation of the principal stresses ( , , ) and deviatoric principal

stresses ( , , ) on the deviatoric plane…………………………………………………..…..33

ix
3.3: General trend of reduction of fracture strain with pressure………………………………....37

3.4: Axisymmetric tension and principal stress tensor………………………………………..…40

3.5: Uniaxial compression and principal stress tensor…………………………………………...41

3.6: Pure shear and principal stress matrix………………………………………………………43

4.1: Xue’s symmetric fracture surface in the form of ԑ vs η curves for each ξ value……..……56

4.2: Bai’s asymmetric fracture surface in the form of ԑ vs η curves for each ξ value……….…56

4.3: 2024-T351 Tensile specimen dimensions tested by Xue…………………………………...57

4.4: Yield curve used for 2024-T351 tensile experiments…………………………………….....57

4.5: 2024-T351 Un-notched tensile specimen mesh…………………………………….…….…58

4.6: Force-Displacement plot for un-notched tensile specimen for 2024-T351…………..……..59

4.7: Stress state in 2024-T351 un-notched tensile specimen…………………………….………59

4.8: Notched round specimen mesh with 18 mm notch radius………………………..…………60

4.9: Force-Displacement plot for 18 mm radius notched tensile specimen for 2024-T351….….60

4.10: Stress state in 2024-T351 18 mm notch radius tensile specimen at neck………….……....61

4.11: Notched round specimen mesh with 9 mm notch radius……………………………….….61

4.12: Force-Displacement characteristics for the 9 mm radius notched specimen test for 2024-
T351……………………………………………………………………………………..……….62

4.13: Stress state in 2024-T351 9mm notch radius tensile specimen at neck……………………62

4.14: Notched round specimen mesh with 9 mm notch radius………………………………..…63

4.15: Force-Displacement characteristics for the 4.5 mm radius notched tensile specimen test for
2024-T351………………………………………………………………………………………..63

4.16: Stress state in 2024-T351 4.5 mm notch radius tensile specimen at neck……………....…64

x
4.17: Yield curve used for 2024-T351 simulations other than tensile tests……………….…..…65

4.18: 8 mm diameter upsetting specimen cylindrical meshes and their corresponding lengths for
2024-T351…………………………………………………………………………….………….65

4.19: Stress state in 2024-T351 upsetting specimen before crack failure……………………..…66

4.20: Force-Displacement plot for 11.25mm length 2024-T351 upsetting test…………….……66

4.21: Von mises stress plot for fracture initiation in 11.25 mm long 2024-T351 upsetting specimen
obtained with GISSMO………………………………………………………….……….………67

4.22: Von mises stress plot for fracture initiation in 11.25 mm long 2024-T351 upsetting specimen
obtained without GISSMO………………………………………………………….…………....67

4.23: Actual fracture initiation in 11.25 mm long 2024-T351 upsetting specimen experiment (a)
cut of A-A plane (b) cut off B-B plane……………………………………………………….….68

4.24: Force-Displacement plot for 8 mm long 2024-T351 upsetting test…………………..……69

4.25: Von mises stress plot for fracture initiation in 8 mm long 2024-T351 upsetting specimen
obtained with GISSMO…………………………………………………………………….…….69

4.26: Von mises stress plot for fracture initiation in 8 mm long 2024-T351 upsetting specimen
obtained without GISSMO………………………………………………………………………70

4.27: Grooved flat plate dimensions for 2024-T351……………………………………………..71

4.28: Slant crack observed during the groove flat plate experiment (a) entire crack surface of one
side of specimen (b) side view of interested section………………………………………..……71

4.29: Crack prediction in 2024-T351 grooved flat plate specimen mid-section (a) With GISSMO
(b) Without GISSMO…………………………………………………………………………....72

4.30: Stress state in 2024-T351 grooved flat specimen before fracture…………………………73

4.31: Force-displacement plot for 2054-T351 grooved flat plate specimen…………….….……73

xi
4.32 (a): 2024-T351 specimen experiments with unique stress states for fracture surface
testing………………………………………………………………………………………….....74

4.32 (b): 2024-T351 experiments represented on the fracture strain – triaxiality plot of the fracture
surface upper bound (ξ = 1 or -1) and lower bound (ξ = 0)……………………………..……….74

4.33: Schematic arrangement for 2024-T351 3-point bending…………………….……….……75

4.34: Mesh used for 2024-T351 10 mm wide 3-point bending test………………………….…..76

4.35: Force-Displacement plot for 2024-T351 10 mm wide 3-point bending specimen test…....76

4.36: Fracture obtained in the 10 mm wide 2024-T351 3-point bending simulation (a) without
GISSMO (b) with GISSMO………………………………………………………….……….…77

4.37: Mesh used for 2024-T351 30 mm wide 3-point bending test………………………….….78

4.38: Force-displacement plot for 30 mm wide 2024-T351 3-point bending specimen test…....78

4.39: Fracture obtained in the 30 mm wide 2024-T351 3-point bending simulation (a) without
GISSMO (b) with GISSMO…………………………………….…………………………….…79

4.40: Mesh used for 2024-T351 60 mm wide 3-point bending test………………………..……80

4.41: Force-displacement plot for 60 mm wide 2024-T351 3-point bending specimen test…….80

4.42: Fracture obtained in the 60 mm wide 2024-T351 3-point bending simulation (a) without
GISSMO (b) with GISSMO…………………………………………………………………..…81

4.43: Quarter model mesh used for 0.063” thick plate and 0.5” diameter spherical projectile….83

4.44: Velocity (mm/ms) VS time (ms) for 0.063” plate with initial velocity of -140.208
mm/ms………………………………………………………………………..…………….……84

4.45: Initial VS residual velocity chart for 0.063” thick plate target………………………….….84

4.46: Quarter model mesh used for 0.125” thick plate and 0.5” diameter spherical projectile….85

xii
4.47: Velocity (mm/ms) VS time (ms) for 0.125” plate less with initial velocity of -219.456
mm/ms…………………………………………………………………………………………...86

4.48: Initial VS residual velocity chart for 0.125” thick plate target…………………..……….…86

4.49: Quarter model mesh used for 0.25” thick plate and 0.5” diameter spherical projectile…...87

4.50: Velocity (mm/ms) VS time (ms) for 0.125” plate less with initial velocity of -304.8
mm/ms…………………………………………………………………………………………...88

4.51: Initial VS residual velocity chart for 0.25” thick plate target……………………………….88

4.52: Ballistic limit predictions for different plate thickness……………………….…………….89

4.53: Failure mode transition with plate thickness: (a) Petaling: Bending & Necking (b) Mixed-
mode: Bending & Spalling (c) Plugging: Shearing & Spalling…………………………….……89
4.54: 1045 steel yield curve for LCSS input……………………………………………………..90
4.55: 1045 steel fracture surface consisting of ԑ vs η curves for each ξ value…….…………...91
4.56: Necked cross-section in a round bar specimen………………………………………….....92
4.57: Stress state parameters for 1045 steel un-notched tensile specimen………………..….….94
4.58: Force-displacement results comparison for 1045 steel un-notched tensile specimen.….…94
4.59: Stress state parameters for 1045 steel 10.5 mm radius notched tensile specimen…….…...95
4.60: Force-displacement plot for 1045 steel 10.5mm notch radius tensile specimen test………95
4.61: Principal stresses in an element 91458 at the critical section of 1045 steel grooved flat plate
specimen…………………………………………………………………………….……..…….96
4.62: Stress state parameters in 1045 steel grooved flat plate…………………………..……….97
4.63: Force-displacement plots for 1045 steel grooved flat specimen…………………………...98
4.64: 1045 steel torsion specimen dimensions…………………………………………………...99
4.65: Stress state parameters fringe plot for 1045 steel torsion specimen……………………….99
4.66: Principal stresses (GPa) with time (quasi-static) plotted at an element in critical section of
1045 steel torsion specimen…………………………………………………………………….100
4.67: Torque (Nm) vs rotation (degrees) for 1045 steel torsion specimen………………..……100
4.68: Stress state parameters in 1045 steel tube tension loading……………………………….101
4.69: Force vs displacement for 1045 steel tube tension……………………………….………102

xiii
4.70: Different experiments with unique stress states for 1045 steel………………………..…103
4.71: 1045 steel experiments represented on the fracture strain – triaxiality plot of the fracture
surface upper bound (ξ = 1 or -1) and lower bound (ξ = 0)…………………………….………103
5.1: DP600 steel yield curve for LCSS input………………………………………………..….105
5.2: DP600 fracture surface plotted in the (η-ξ) space…………………….…………..……….106
5.3: DP600 steel un-notched flat specimen (a) Dimensions [5] (b) Mesh…………….………..108
5.4: The plastic strain, triaxiality and lode parameter before fracture at the critical section of DP600
un-notched flat tensile specimen…………………………………………………………..……109
5.5: Normalized force-displacement plots for DP600 un-notched tensile flat specimen experiment
and simulation…………………………………………………………………………………..109
5.6: DP600 2 mm notch tensile flat specimen (a) Dimensions [5] (b) Mesh…………….….….110
5.7: The plastic strain, triaxiality and lode parameter distribution at the critical section of DP600 2
mm notch tensile specimen before failure………………………………………………….…...111
5.8: Normalized force vs displacement for DP600 2 mm notch flat tensile specimen……..…...111
5.9: 2 mm notch DP600 round tensile specimen (a) Dimensions[5] (b) Mesh………..………..112
5.10: Plastic strain, triaxiality and lode parameter distribution for 0.5 mm notch DP600 tensile
specimen…………………………………………………………………………………..……113
5.11: Force-Displacement plot for DP600 0.5 mm notch radius tensile specimen test…..…….113
5.12: Mesh for DP600 2 mm radius notch round tensile specimen……………….……..……..114
5.13: The plastic strain, triaxiality and lode parameter distribution at the critical section for DP600
2 mm notch round tensile specimen before failure……………………………………………..114
5.14: Force displacement plot for DP600 2 mm notch round tensile specimen test…………....115
5.15: Mesh for the DP600 4 mm notch round tensile specimen………………………………..115
5.16: Plastic strain, triaxiality and lode parameter distribution in DP600 4 mm notch round tensile
specimen critical section just before failure…………………………………………………….116
5.17: Normalized force displacement plot for DP600 4 mm notch round tensile specimen
test………………………………………………………………………………………………116
5.18: DP600 0.5 mm notch radius grooved flat plate specimen…………………………..……118
5.19: The plastic strain, triaxiality and lode parameter distribution at the critical section before
failure of DP600 0.5 mm notch radius grooved flat plate………………………………………118

xiv
5.20: Normalized force-displacement plot for DP600 0.5 mm notch radius grooved flat plate
specimen test……………………………………………………………………….…………...119
5.21: DP600 1 mm notch radius grooved flat plate specimen…………………….……………120
5.22: The plastic strain, triaxiality and lode parameter distribution at the critical section before
failure of DP600 1 mm notch radius grooved flat plate…………………………….….……….121
5.23: Normalized force-displacement plot for DP600 1 mm notch radius grooved flat plate
specimen test…………………………………………………………………………………....121
5.24: DP600 experiments simulated characterizing different stress states on the fracture
locus…………………………………………………………………………………………….123
5.25: DP600 experiments simulated characterizing different stress states on the fracture strain-
triaxiality plot for the fracture locus……………………………………………………………123

5.26: Force-displacement plots for different damage exponents in the case of 2024-T351 grooved
flat plate specimen loading………………………………………………………………….….124
5.27: Fracture pattern for damage exponents ranging from 1 to 5 for 2024-T351 aluminum alloy
grooved flat plate specimen loading………………………………………………………...….125
5.28: Differently discretized models for 2024-T351 un-notched tensile test specimen………..128
5.29: Force-displacement plots for 2024-T351 un-notched tensile specimen simulated with
different mesh sizes without regularization………………………………………………...…..129
5.30: Fracture strain vs mesh size plot for 2024-T351 un-notched tensile specimen…………..129
5.31: Force-displacement plots for different mesh sizes of 2024-T351 un-notched tensile
specimens after regularization……………………………………………………………….…130
5.32: Differently discretized meshes for 1045 steel un-notched tensile test specimen……..….131
5.33: Force displacement plots for 1045 steel un-notched tensile specimens without
regularization……………………………………………………………….…………………..132
5.34: Fracture strain vs mesh size for 1045 steel un-notched tensile specimen………………..133
5.35: Force-displacement plots for different element size 1045 steel un-notched tensile specimen
test simulations………………………………………………………………………..………..133
5.36: Differently discretized DP600 un-notched flat tensile specimen………………………...134
5.37: Normalized force vs displacement plots for different mesh size models for DP600 un-notched
tensile flat specimens………………………………………………………..………..135

xv
5.38: Fracture strain variation with mesh size for DP600 un-notched tensile flat specimen
test……………………………………………………………………………………………...136
5.39: Normalized force-displacement plots superimposed for different mesh sizes of DP600 un-
notched tensile flat specimen tests after regularization……………………………………..…136

xvi
Chapter 1

Introduction

From the Stone Age times man has been in search for better and long lasting materials to use them
for the advancement of technology. Aluminum alloys & steels have revolutionized the automobile,
aerospace and marine industry to an extent that transportation has become a necessity more than a
luxury. In an effort to better understand the behavior of different materials, computational tools
such as Finite Element Analysis, Numerical methods and simulation have been frequently
employed.

Ductile fracture is the kind of fracture where the material undergoes considerable amount of plastic
deformation before it fails. It’s important to model the accurate behavior of ductile fracture because
numerous factors affect the phenomena significantly. Among the many developments that took
place in the field of fracture mechanics, this thesis presents an implementation of a stress-state
dependent strain-based fracture criterion.

1.1 Background and motivation

Damage and fracture are a common phenomenon in the materials used in engineering field.
Scientists and Engineers have been trying to better understand the microscopic as well as

1
macroscopic characteristic of the material failure. Since the pre-industrial revolution times,
material behavior in different structures and loading conditions have been studied, although
significant breakthrough has only been achieved in the past few decades.

The widely used failure models like Von-mises, Drucker-Prager, Johnson-Cook, Wilkins, etc.
offer different depths of insights into the initiation & propagation of failure. Recent research at
MIT in Impact & Crashworthiness Laboratory suggested that apart from other factors influencing
failure of a material, the pressure & principal stress proportions play a major role. The need for
this thesis was realized when the exploration of these stress state parameters was thought to be
constructive for accurate failure modeling. For aluminum alloy 2024-T351, figure 1.1 shows three
different stress-strain curves observed for three different kinds of loadings – tension, compression
& torsion. This difference is caused by the difference of stress states during loading.

Figure: 1.1 Comparison of effective stress versus equivalent strain curves for tension,
compression and torsion experiments on 2024-T351 [29]

Finite element modeling is the most widely used tool for any mechanical design validation using
the aid of computers. As of today, every automobile, aerospace, construction, heavy machinery

2
and recently even bio-mechanics industry highly benefits from FE modeling. As the failure of
materials in service can be catastrophic, analysis of engineering materials is of utmost importance
in order to decide its life, durability & reliability. Computer simulations greatly reduce the cost
and time of conducting actual full-scale experiments especially in crash situations. The important
responsibility to achieve close prediction of results makes concentration of efforts on accurate
constitutive mathematical material failure modeling and simulation an urgent need. New ideas for
improving ductile material failure behavior & simplification of its characterization will contribute
a great deal to the research community.

1.2 Research Scope, approach & objectives

Most material models don’t have any control over the post-failure behavior of the material. The
commonly used approach of element deletion upon reaching a failure criterion does not serve the
purpose of producing realistic simulation very well due to lack of deleted element participation
post failure. That is when the significance of accounting for damage comes into picture. An attempt
has been made to have some control over post-failure behavior by including damage, critical strain
and stress state dependent equivalent plastic strain to failure. A relatively simple plasticity model
based on von-mises flow rule in combination with erosion criteria is chosen to replicate the
experiments. The strain rate, anisotropy and temperature effects are not considered in this thesis.
All the examples validated assume room temperature and quasi-static loading conditions. The
material model used is isotropic although the materials show slight anisotropy in the experimental
results due to strain effects inherent to the manufacturing process. 2024-T351 aluminum alloy,
DP600 steel & 1045 steel are the three materials investigated for failure in different cases of stress
state.

The stress-strain response obtained from experimental results is captured by the plasticity model.
Stress state parameters can be obtained from parallel trial simulations. A fracture locus in the form

3
of equivalent plastic strain to failure in the triaxiality and lode angle parameter space was obtained
by Xue [3], Bai [4] and Basaran [5]. This fracture locus is directly fed to the material model in the
form of load curves. Based on the stress state, the code selects the appropriate value of equivalent
plastic strain to fracture or simply ‘fracture strain’. This advantage of stress state dependency is
not present in simple fixed failure strain based approach. Results with fixed failure strain are
compared with the ones with stress state dependent failure strain to determine the feasibility of this
additional computational work.

For 2024-T351 aluminum alloy and 1045 steel materials, the tensile & grooved flat plate
specimens are simulated with both the approaches – 1) With fixed failure strain 2) Stress state
dependent failure strain. Some special specimens like torsion, compact-tension, upsetting & 3-
point bending specimens are also simulated to produce fracture transition from normal to shear
mode. The fracture characteristics, pattern, location and accuracy are studied. The tensile, grooved
flat plate, torsion & upsetting specimens are unique as they have a fairly constant stress state for
major duration of loading (proportional loading). Although an optimum fine mesh size is selected
for the simulations, a study of regularization is carried out to study the influence of mesh size on
the results in the post-critical damage region. Finally, the ballistic impact simulations are carried
out for aluminum alloy 2024-T351 using the implemented plasticity & erosion model to evaluate
the feasibility of this method over the previously employed highly sophisticated tabulated Johnson-
Cook [12] material model by DuBois et al. [9]. All the simulation results using both approaches
are compared to experimental results and the pros & cons of each approach are discussed to
determine suitability to different applications.

1.3 Outline of Thesis

This thesis consists of total 6 chapters. All the chapters follow a logical sequence leading to more
elaborate information about the work done in step-by-step fashion.

4
Chapter 1 gives a general background information and motivation behind this work. The research
scope, approach, objectives and the basic outline of thesis is given at the beginning.
A brief introduction to the existing failure modeling approaches in engineering materials is given
in chapter 2, along with references to some of the important work done in this field. Since the
macro-mechanical models are more widely used and also available in material libraries of most
commercial codes than the micro-mechanical ones, they are discussed in more detail. The models
that account for damage are learnt to give closer approximation to reality.
Chapter 3 describes the effect of different ratios and the average of principal stresses in the form
of lode angle parameter dependence and triaxiality dependence. These effects are realized by
definition of stress state parameters lode angle & triaxiality. Some simple specimen geometries
having nearly proportional ratio of principal stresses are shown to be helpful in determining stress
states analytically. These specimen geometries are important to determine the ‘fracture locus’ in
the space of lode angle parameter & triaxiality. Finally, an introduction to GISSMO damage model
and its implementation is presented.
Chapter 4 deals with the approach adopted in the modeling, meshing, and simulation of
experiments along with the results that were achieved from the previous steps. Simulations for
experiments on 2024-T351 aluminum alloy and 1045 steel are used to validate the previously done
experiments.
In chapter 5, the already validated Dual Phase 600 steel experiments are simulated to evaluate
correctness of the present approach. Methods to minimize errors and avoid mesh-size influence on
results by regularization are described as well.
Chapter 6 concludes this work by commenting on the quality of results approached. The
advantages, limitations and computational effort required in following the present modeling
method are elaborately discussed. The future scope for improving this kind of implementation is
briefly laid out.

5
Chapter 2

Literature Review

While most failure criteria are able to satisfactorily control the occurrence of failure, the recent
efforts led to the development of models, which control the post-failure behavior. Oscar [1]
presented a review of commonly used criteria, and compared the results of Cockroft-Latham [20],
Bressan-Williams [25] and Critical thickness strain criteria. While these are simple failure criteria,
new modeling approaches with breakthrough research sought to address complex applications. For
example, simple Mohr-Coulomb criterion [11] acknowledges the pressure effect by considering
different values of fracture stress for tension and compression. Further, Bridgeman’s pressure
chamber experiments (1952) studied in [3] point out variation in occurrence of fracture at different
pressures. Based on experiments on 2024-T351 aluminum alloy, Bao [19] studied that stress
triaxiality and equivalent strain are the two most important parameters governing ductile crack
formation while other parameters have secondary importance. Xue [3] studied the dependence of
hydrostatic and deviatoric components of stresses on the fracture strain and came up with a
symmetric ‘fracture surface’ or fracture locus for 2024-T351. After validating experiments for the
same material, Wierzbicki et.al [2] presented a comparison of seven different fracture criteria and
confirmed that the fracture surface has strong stress state dependency. According to this fracture
surface, there is a unique equivalent plastic strain to fracture or ‘fracture strain’ at a given triaxiality
(pressure or mean stress) and lode angle parameter (deviatoric stress third invariant). A 3D fracture

6
locus for 1045 steel, A710 steel, DH36 steel and 2024-T351 aluminum alloy was proposed and
validated through experiments by Bai [4]. Basaran [5] implemented the GISSMO model proposed
by Neukamm et.al. [26, 27] in LS dyna for DP600 steel. This model approach is versatile to include
different failure criteria simultaneously, account for damage and also control the post-failure
behavior. We choose a similar approach to validate the experimental results by directly feeding
the ‘fracture surface’ based on the stress state in the form of a table but make no attempt to compare
the different failure models. Carney, et.al. [28] proposed a new parameter denoted product
triaxiality based on the third stress invariant. Additional research on the simultaneous inclusion of
anisotropy, strain rate and temperature effects on ductile fracture for aluminum alloy 2024-T351
can be found in [29] and is beyond the scope of this thesis.

2.1. Existing Failure/Damage modeling approaches:

The concept of fracture attained more and more complexity as a number of factors influencing it
were discovered. The range varies from micro-scale modeling where breaking of atomic bonds is
considered to macro scale modeling where crack growth developed at meso-scale level is studied.
Due to the high complexity and the limited computational capability offered by modern processors,
micro-scale modeling approaches are still not having substantial application outside research
organizations. In this thesis, we only discuss macro-scale or phenomenological modeling
approaches used widely in the industry. The literature review is brief and makes no attempt to be
exhaustive. The following is the general classification of some of the phenomenological failure
models:

7
 Failure/Damage
models

Stress Strain Cockroft Damage CrachFEM

dependent Dependent Latham Models

Max. principal
stress Max. principal strain
Gurson

Tresca's failure Constant

criterion Johnson-
equivalent strain Cook
Max. shear
stress Forming limit diagram Wilkins
(FFLD)
Mohr's failure Xue-
criterion Wierzbicki

Von-Mises CDM

Figure 2.1: Classification of common phenomenological failure/damage models [1]

2.1.1 Stress dependent failure criteria

a) Maximum principal stress criterion: Also called the Coulomb or Rankine failure
criterion, it predicts failure when any of the principal stresses exceeds the maximum tensile
stress σt allowed or maximum compressive stress σc allowed.

Figure 2.2: Maximum principal stress criterion. No failure when stress lies inside box

8
 -σc ≤ (σ1, σ2, σ3) ≤ σt (2.1)
Where σ1, σ2 and σ3 are the principal stresses.
This criteria can be implemented in LS dyna using the SIGP1 input of the
*MAT_ADD_EROSION keyword for the maximum principal stress for failure.

b) Tresca’s failure criteria: It predicts failure when the maximum difference between the
principal stresses is lower than the fracture stress σf.

Max. (|σ1-σ2| |σ2-σ3| |σ3-σ1|) ≤ σf (2.2)

For plain stress case (σ3 = 0) the Tresca’s failure criteria resembles a hexagon in the stress
plane as seen in figure 2.3. For a general 3 dimensional problem, Tresca’s failure criterion
becomes a hexagonal prism as seen in figure 2.5.

Von-mises
Tresca

Figure 2.3 Tresca’s and Von mises failure criterion in plain stress case

Due to the development of other simpler criteria without significant compromise in

accuracy, this criterion is not widely preferred over Von-mises.

9
c) Mohr’s failure criterion: Also known as Mohr-Coulomb fracture criteria, [11] it is based
on Mohr’s circle. In addition to σt & σc from maximum principal stress criterion, this also
considers the shear stresses to predict failure. In the figure 2.4, two Mohr’s

Figure 2.4: Mohr-Coulomb failure criterion

circles are shown in bold with the shear stress as ordinate and normal stress as abscissa.
The first circle has radius σc/2 and center ( , 0). The second circle has radius of σt/2 and

center (− , 0). The circles are connected with dotted lines as shown in figure 2.4. Failure

is not expected as long as the Mohr’s circles lie within the shaded region. This material
model has some application in Geomaterial modeling [14]. In LS dyna library,
*MAT_MOHR_COULOMB (*Mat_173) for sandy soils and other granular materials is
based on the normal & shear stress for its yield surface.

10
 Figure 2.5: Tresca & Mohr-Coulomb fracture criteria in principal stress space.

d) Von Mises’ failure criterion: Based on principal stresses, the Von-Mises failure criteria
is given by

= 3 = [( − ) +( − ) +( − ) ] ≤ σf (2.3)

Where σf is the fracture strain or strain at which failure occurs and is the second
deviatoric principal stress invariant.

For plain stress, Von-mises criteria resemble an ellipse surrounding Tresca’s criterion as
seen in figure 2.3. For a general 3 dimensional problem, Von-mises failure criterion
becomes an elliptical cylinder as can be seen in figure 2.6. Including the hydrostatic effect
makes up another criterion known as the Drucker-Prager failure criterion [16]. It should be
noted that Von mises criteria only depends on the second deviatoric stress invariant ,
while Tresca’s criteria depends on the second and third deviatoric stress invariants and
respectively. This can also be concluded as Lode dependence. Due to its simplicity, Von
Mises is still a widely used material model. Simple LS dyna implementation could be
through FAIL input in *MAT_PIECEWISE_LINEAR_PLASTICITY or SIGVM input in
*MAT_ADD_EROSION.

11
Figure 2.6: Von mises & Drucker-Prager failure criterion in 3D principal stress space

Figure 2.7: Yield surface in principal stress space for Tresca & Von mises [17]

12
 Figure 2.8: Π-plane or deviatoric plane representation of classical yield functions [14]

e) Maximum shear stress criterion (MS): It has been observed for any material that the
shear strain is smaller than the tensile or compressive strain required causing failure. Hence
the material is more likely to fail in the maximum shear stress zone. This shear fracture
near the equatorial region is quite evident in the upsetting tests discussed later.
The maximum shear stress criterion postulates failure to occur when maximum shear stress
reaches the failure shear strain. Thus, failure is not predicted as long as
τmax ≤ (τmax)f. (2.4)
Where τmax = max. ( , !
, !
) (2.5)

Where, , and are the principal stresses in decreasing order of magnitude.

This criterion can be implemented in LS dyna by using the EPSSH input of
*MAT_ADD_EROSION

b) Strain dependent failure criterion

13
a. Maximum principal strain criterion: This is similar to the maximum principal stress
criterion except that it predicts failure if any of the principal strains exceed the failure
strain ԑf instead of principal stresses
i.e. failure occurs when {ԑ1, ԑ2, ԑ3} ≥ ԑf (2.6)

Where ԑ1, ԑ2, ԑ3 are the principal strains in the decreasing order of magnitude.
This criterion can be implemented in LS dyna by MNEPS input of
*MAT_ADD_EROSION keyword.

b. Constant equivalent strain criterion: This criteria is based on the equivalent plastic
strain ԑp. Failure is expected to occur when:
ԑp = ԑf (2.7)
Where, ԑf is the equivalent plastic strain to produce failure.

And ԑp = (ԑ + ԑ + ԑ ) (2.8)

This criteria can be implemented by using fixed value for the SIGVM input of
*MAT_ADD_EROSION

c. Forming Limit Diagram (FLD): This criterion was developed to represent the forming
strain up to fracture or formability in the plane of principal strains. As can be seen in figure
2.9, this criterion is mostly useful for sheet-metal plain stress analyses. The curve
represents the limit up to which the metal can be formed. Since its difficult to measure the
principal stresses, three unique stress states of proportional loading are generally used to
find the fracture strain ԑ namely, 1) Uni-axial tension, 2) Plain strain and 3) Equi-biaxial
tension. Proportional loading means the principal stresses have a constant ratio throughout
the loading. Specially shaped specimens are used to achieve these stress states. These three
states are also useful for determining the ‘fracture locus’ in the triaxiality – lode angle
space.
"
From [5], Triaxiality, η = (2.9)
#$

14
 Figure 2.9: Forming limit Diagram

% % !
Where mean stress, σm = (2.10)

, & are principal stresses in descending order

ԑeq is equivalent or von-mises stress from eqn (2.3) for equivalent stress. Figure 2.10 shows
corresponding values of fracture strain ԑf for the three stress states on the FLD as well as
ԑf - η space.

For most isotropic materials including those investigated by Lee [15], the FLD
approximates a straight line which can be represented in the simple form:

ԑ +ԑ =-ԑ =C (2.11)
The subscript ‘f’ denotes strain magnitude at the point of fracture.

FLD can be implemented in LS dyna as load curves. In the LCFLD option in

*MAT_ADD_EROSION, minor strains in percent are defined as abscissa values and
major strains in percent are defined as ordinate values [18]. However, this criterion is only
for shell elements starting with release 961 R6. Additionally, the NLP option in *MAT_3-

15
 PARAMETER_BARLAT (MAT_36) or ICFLD option in
*MAT_TRANSVERSELY_ANISOTROPIC_ELASTIC_PLASTIC (*MAT_37) can be
used to implement the forming limit diagram in LS dyna.
*MAT_FLD_TRANSVERSELY_ANISOTROPIC (MAT_39) uses the LCFLD option to
include FLD using a curve.

ԑf-η space

FFLD

Figure 2.10: Transformation of the FFLD fracture locus into the space of (ԑf, η) [2]

2.1.3 Cockcroft-Latham

Cockcroft and Latham [20] suggested a ductile failure criterion based on plastic strain ԑp,
maximum principal stress , equivalent stress σ and fracture strain ԑp. It was proposed that
fracture occurs in a ductile material at a given temperature and strain rate when critical value
of constant C is reached in the below formulation:

16
 ( ) *ԑ+ = C
ԑ(
') (2.12)

The change in peak stress , is also taken into account caused by necking and stress
concentration. No fracture will occur if all the principal stresses are smaller or equal to zero.
Thus, the effect of both stress and strain on ductile fracture is considered. Further modifications
to this criterion are published in Heung and Keun-Hwan [21]. A user defined implementation
in LS dyna is discussed in [22] and *MAT_135 can be used to implement this criteria from
LS-dyna material library. This criteria can also be implemented through FLAG2 = 1 in
*Mat_107 [18].

2.1.4 CrachFEM

This criterion postulates two competing fracture modes – ductile and shear. One is based on
the void growth and coalescence and the other is based on shear failure. Accordingly, there are
different expressions for the equivalent fracture strain: ԑ,-./01 and ԑ23 45
. The one for which

the calculated fracture strain (ԑ,-./01 or ԑ23 45

) is the lowest, determines the fracture limit.
In the case of ductile void growth model, the equivalent strain to fracture is assumed to depend
only on the stress triaxiality (from equation 2.9):
ԑ,-./01 = *) 7 ( .8)
+ * 7( .8)
(2.13)
Where c, *) and * are the three material constants that can be found from tests.
In case of shear failure model, the equivalent strain to fracture is postulated to depend both on
hydrostatic and deviatoric state:
ԑ23 45
= * 7( 9)
+ * 7( 9)
(2.14)
Where the shear fracture parameter θ accounts for both hydrostatic and deviatoric state given
by:
θ=: (1-3=2 η) (2.15)
";<

=2 , f, * and * are the four material constants that can again be found from tests.
Principal stress based tensor representation for every stress state has a symmetric (hydrostatic)
part and an asymmetric part (deviatoric). The shear fracture model acknowledges the joint

17
 effect of stress triaxiality and the deviatoric state through the definition of the parameter θ.
Hence, this model is somewhat similar to the Xue-Wierzbicki model which will be later
reviewed. This model was developed for industrial application and consists of seven free
material parameters. It was proposed by a group of researchers of the BMW R&D center and
MATFEM Co. in Munich. It is now available in PAM-CRASH (Mat #52, 128). CrachFEM
was implemented as user defined material model by Gese H. et. al. [23], M. Chauffray et. al.
[24] and integrated with LS dyna.

2.1.5 Damage models

Material fracture is characterized by complete loss of load carrying capability and

deformability [3]. At an intermediate damaged state, the damage parameter accounts for the
remaining ductility or load carrying capacity up to fracture. It is argued that this damage
parameter is influenced by factors like stress state, anisotropy, strain rate etc., and hence needs
to be represented as a tensor. However, most failure models apply isotropy for simplification
and have shown satisfactory results with a scalar damage parameter. Microscopically, scalar
damage assumes micro-voids and micro-cracks to be evenly distributed in all directions.
Studies have shown that the exclusion of damage over-estimates the material strength and
produces a stiffer material behavior [40]. Further works that describe the benefits and selection
of material parameters for models concerning damage variables can be found in Feucht
et.al.[39], Poizat et.al., [42], etc.

One simple approach to consider damage proposed by Lemaitre and Chaboche [41] is by
relating initial area and damaged area of the specimen in a particular plane perpendicular to
loading direction (cross-sectional areas).
>?
>
D= (2.16)

Where D is the damage parameter, Ad is the damaged area and A is the initial area. According
to this approach, complete failure is reached when damage parameter D = 1 or the entire cross-
sectional area is damaged with no material left to resist deformation.

18
Out of the many approaches to model damage, the most commonly used damage models are
discussed in the next few sections. The GISSMO damage model accounts damage similar to
proposition made in Johnson-Cook [12] with a power function as presented in the coming
sections. The Gurson model, discussed next uses a damage parameter called void volume
fraction.

a) Gurson model: This model can also be categorized as micro-mechanical failure model.
However, due to its inability to capture shear-dominated failure it is often implemented in
extended form to take into account the shear failure and lode angle dependence using
phenomenological approach. From micromechanical point of view, ductile fracture is
defined as material separation due to void nucleation, evolution of existing micro voids
and cracks, followed by progressive void coalescence. Voids usually preexist, but can also
be initiated at material defects such as inclusions.

In microscopic models, material is handled as inhomogeneous cells with considerations of

voids and micro-cracks. The global response of the void containing cell is used to
determine the macroscopic material behavior. The three stages for micromechanical
fracture can be categorized as: 1) Void nucleation 2) Void growth 3) Void coalescence.
Void nucleation requires the modeling of inclusions, interface between inclusions and
material matrix. The inclusions are assumed as elastic and brittle particles. Void nucleation
occurs by fracture or de-bonding of inclusions. Due to difficulty in understanding void
nucleation, it has not been studied as much as void growth or void coalescence. In the
original Gurson model [30] a damage parameter ‘f’ called void volume fraction measures
the porosity of the material. Void volume fraction is the ratio of volume occupied by the
void to the total volume occupied by the material matrix along with the void in a
representative volume element (RVE). It can vary from 0 for an undamaged material to 1
for a completely damaged material. Because of the inability of predicting instability caused
by coalescence of micro-voids, the Gurson model was extended by Tvergaard and
Needleman [31] by using cohesive model between the inclusions and the material matrix,
where the interfacial strengths are relatively weak. The void nucleation stage is followed

19
by void growth. McClintock [32] studied a material containing a cylindrical void and Rice
and Tracy [33] studied a material containing a spherical void. In both the studies, stress
triaxiality had an effect on void growth. Rice and Tracy found that void growth is
exponentially dependent on stress triaxiality. Analytical expressions defining void growth
assume the void surrounding material as perfectly plastic. Gurson defined a pressure
defined yield function from an isolated spherical void in a continuum that is dilatant and
elasto-plastic. The most widely used Gurson-Tvergaard-Needleman or GTN model uses
the yield condition:

) - 1 - (B D ∗ )2 = 0
@" /5C
Φ= + 2B f* cosh ( (2.17)
A A

Where σ is macroscopic stress tensor, FG is the Von mises equivalent stress, H is the
actual yield stress, B is a material constant that amplifies the hydrostatic stress effect for
all strain levels and f* is the new void volume fraction given by,

D, JD D ≤ DL
f*(f) = I
DL + (D − DL ) JD D > DL
$ M (2.18)
( M

Where DL is the critical void volume fraction at coalescence and D is the void volume
fraction at rupture. At small deformations, the total void volume fraction evolution is equal
to the sum of growth of existing voids and the nucleation of new voids:
DO = DP5QR/3
O + DS-.1 O 4/TQS (2.19)
Since the material is plastically incompressible, the growth of voids can be expressed as:
O
DP5QR/3 = (1 − D)VWԑ+O (2.20)
Where, ԑ+O is the plastic strain rate. The phenomenon of nucleation depends on the stress
level, strain level, or a combination of both depending on the material. One of the most
studied cases corresponds to the situation when nucleation is produced by plastic strain
which is given by:
DS-.1 O 4/TQS = A ԑ+O (2.21)

ԑ+O is the equivalent plastic strain rate and A is the cavity nucleation rate given by:

20
 ]
ԑ#$^ ԑZ
(
X= Y
7 _Z
2Z √ \
(2.22)

Where D` is total void volume fraction that

can be nucleated,
` is the standard deviation,
ԑ` is the mean strain for nucleation,
ԑ+ is effective equivalent plastic strain

Figure 2.11: Cavity nucleation rate [34]

This model is implemented in LS dyna as *MAT_GURSON (or Mat 120). Although the
GTN model is designed to describe void nucleation, growth and coalescence, it has some
inherent limitations:
1. It is incapable of modeling the localization and fracture for low stress triaxiality,
shear dominated deformations, since it does not predict void growth and damage
under shear loading when fN < ff .
2. It does not include the lode dependence.
3. The geometrical change of shape and size are over-simplified. For example, it is
assumed that spherical voids retain their spherical shape. This gives incorrect
results especially at high stress triaxialities. Also, the shapes of the voids tend to be
elliptical at low triaxialities after deformation.
4. Only void volume fraction is used as a measure of damage. Volumes less defects
such as cracks are not considered.
5. The calibration of micromechanical parameters is very cumbersome and done
usually in a stochastic way with reverse engineering, since experimentally the
determination of parameters is very difficult.

21
 To overcome these limitations, the GTN model was modified by many researchers. Xue
[3] further extended the GTN model to incorporate void shearing mechanism in a
phenomenological way. Nahshon and Hutchinson [35] accounted for shear deformation
and lode dependence. Gologanu et. al. [36] enhanced the model with consideration of
axisymmetric elliptical voids. Pardoen and Hutchinson [37] further enhanced model of
Gologanu with void spacing effects. A comprehensive review of GTN model and its
extension was published by Lassance et. al. [38]. The optional addition of Johnson-Cook
failure criteria for void growth in shear dominated states is presented by Feucht et.al. [39]
along with validation for tensile & losipescu shear specimen fracture. This approach is
included in LS dyna material liabrary as *MAT_GURSON_JC (or *MAT_120_JC, an
extension of *MAT_120). The GTN model was also extended by the Wilkins model and
is included in LS dyna material library as *MAT_GURSON_RCDC (or
*MAT_120_RCDC after *MAT_120).

b) Johnson-Cook
Johnson-Cook [12], [43] is a phenomenological model which uses damage parameter D
that varies from 0 at no damage to 1 at complete or critical damage Dc.
,ԑ]
D='
ԑ(
(2.23)

Where, ԑ+ is the equivalent plastic strain and ԑ is the equivalent plastic strain at fracture.
The fracture strain ԑ further depends on the stress triaxiality, strain rate and temperature
as follows:
ԑ = (D1 + D27 a! ) (1 + D4ln(ԑO ∗ )) (1 + D5b ∗ )
∗
(2.24)
where, Di, i = 1 to 5 are material constants,
G/
∗
= is the stress triaxiality (ratio of mean and equivalent stress)
ԑO ∗ = ԑO /ԑ) is dimensionless ratio of plastic strain rate and user defined reference
strain rate
b∗ =
d deff"
d"#g deff"
is the homologous temperature with T as current temperature,

b5QQG as ambient temperature and bG 1/ as the material melting temperature

22
 Similarly, the Von mises equivalent flow stress σy is the product of three factors
representing strain hardening, strain rate and temperature given as:
σy = (A + Bԑ+ n) (1 + c ln(ԑO ∗ )) (1 - (b ∗ )m (2.25)
where, A, B, C, n & m are calibrated material constants.
Damage accumulates in the material element during plastic straining and affects the stress
field as:
σy = (1 - D)[(A + Bԑ+ n) (1 + c ln(ԑO ∗ )) (1 - (b ∗ )m] (2.26)
Few shortcomings of this model also presented in [44] are corrected in the later
developments. This model is unable to predict variation in flow stress with strain rate.
Another difficulty in calibrating the JC parameters is that triaxiality changes during loading
[2]. Since this model does not consider Lode dependence, it was modified by Wilkins et.al.
[45] in a cumulative damage formulation.
In LS-dyna material library, *Mat_15 is the implementation of JC model. The advanced
*Mat_224 allows complete tabular input for this model. *Mat_98 simplifies the
implementation by ignoring the temperature effects and damage but also has an option to
include orthotropic damage through the extension *Mat_99 for aluminum panels using
shell elements [18]. *Mat_107 can be used to implement modified Johnson Cook criteria
with different constitutive relations and fracture criteria developed prior to Johnson Cook.

c) Wilkins
The Wilkins failure model although based on the ductile fracture theory of McClintok
(2004) has been extended and modified by many researchers. The simplest expression for
damage D which accounts for both the hydrostatic and asymmetric stress is,

D = ' h h dԑ+ (2.27)

Where ԑ+ is effective plastic strain

The hydrostatic stress accounts for growth of holes by spalling and asymmetric stress
accounts for the effect of lode parameter like decrease of fracture strain in shear-dominated
regions.

23
 Also,
j
h =
( i "
(2.28)

is the triaxial stress weighting term, G is the mean stress,

h = (2 − Xa l
(2.29)
is the asymmetric strain weighting term,
α, γ and β are material parameters,

2 2
Xa = min mn n , n no
2 2!
(2.30)

is the ‘stress eccentricity’ with , and the ordered principal stress deviators
AD ranges from 0 to 1 and when AD = 1 stress field is symmetric (and asymmetric when
AD = 0).
Fracture is initiated when accumulation of damage is
a
a
>1 (2.31)

Where critical damage Dc = Do (1 + b|qr|s ) (2.32)

Do, b and λ are again material constants and qr is damage gradient [18]
A fracture fraction
(a a
a_
F= (2.33)

defines the degradation of the material by the Rc-Dc model.

Due to its non-local form, equation (2.31) is less mesh-dependent than other criteria like
Johnson-Cook model.

A modification of this criterion has been implemented in the PAM CRASH FE code as the
EWK model [46]. In LS dyna, Wilkins criteria can be implemented by the *MAT_82 RcDc
option (*MAT_PLASTICITY_WITH_DAMAGE) [18].

d) Xue-Wierzbicki

24
Xue [3] implemented symmetric fracture surface in the space of pressure and lode angle
parameter as a user defined subroutine that accounts for incremental damage influence over
matrix stress.
The equivalent yield stress σeq is affected by the damage as:
σeq = (1 - Dβ) t (2.34)
Where, (1 - Dβ) is the weakening function, β is material parameter and t is the
matrix stress

Damage rate, rO is
(G )
ԑ] ԑ]O
rO = m vԑ w ԑ(
(2.35)
(

Where m is the damage exponent, a material parameter and ԑ+ is equivalent plastic strain
The fracture strain ԑ depends on the pressure and lode dependence functions x+ & x9
ԑ = ԑ ) x+ x9 (2.36)
Where, ԑ ) is the reference fracture strain, a material parameter
Pressure dependence function with pressure p,
x+ = 1 – q log (1 - +
+
) (2.37)
gy"

Where, shape parameter q and limiting pressure z10G are material parameters
Lode dependence function,
|9{ | }
x9 = 1 – (1 – γ) m o
\/|
(2.38)

Where, ~• is the lode angle parameter

k is a material constant called lode dependence exponent
ԑ( 4/ +140S 2/540S .QS,0/0QS
ԑ( 4/ 4€02HGG /50. .QS,0/0QS
fracture strain ratio γ =

Xue’s lode angle parameter,

~• = V•‚ ƒ (2„ − 1)…

√
(2.39)

2 2! !
2 2!
x= = (2.40)
!

(x is another form of lode angle parameter calculated from the ratio of principal
stress deviators , and in descending order)

25
Thus Xue’s damage theory has two internal variables (ԑ+ and D) and six material
parameters (ԑ ) , z10G , q, γ, m and β) which define the fracture locus. Xue’ fracture locus
is made up of fracture strain values (ԑ ) for different lode angle parameters (x) and stress
triaxilities. One limitation of Xue’s fracture locus is that it is symmetric about the lode
angle parameter values, which is not the case for many materials. It is bounded by the plain
strain state and the axisymmetric state of loading as shown in the fracture strain-triaxiality
space below:

(a) (b)
Figure 2.12: Xue’s symmetric fracture surface in the space of ԑ and triaxiality [3]

26
 Figure 2.13: Xue’s 3-dimensional fracture surface in the principal stress space [3]

e) Continuum Damage Mechanics (CDM)

Continuum damage mechanics model is a phenomenological damage model extended by

Lemaitre [47] from Kachanov’s [48] creep damage model. Due to voids and micro-cracks,
the load-carrying area reduces, thereby reducing the effective stress:
Effective stress, † =
a
[18] (2.41)

Where, D is damage variable and σ is Cauchy stress.

The evolution equation for damage is defined as:
ˆ
WO DŠW W N Wa •‚* N0
rO = ‡‰( a
0 ŠVŒ7W•J 7
(2.42)

Where Y is a positive material constant, S is the strain energy release rate Wa is the damage
threshold and is the maximum principal stress.
The damage strain energy release rate (conceptually similar to stress intensity factor SIF)
is given by
•#$ ‘@
Y=ԑ ∶•∶ ԑ = ’( “
(2.43)

27
 Where ԑ is elastic strain tensor, C is the material stiffness matrix tensor, E is young’s
modulus and σ is the Von mises stress given by:

σ = ∶ (2.44)

R F is the function that accounts for stress triaxiality as:

R F = (1 + ν) + 3(1 - 2ν) v "

w (2.45)
#$

Where G is hydrostatic or mean stress and ν is Poisson’s ratio

The plastic strain rate ԑ+O given by

ԑ–O = ԑ+O : ԑ+O (2.46)

This plastic strain rate is affected due to damage accumulation. The damage accumulated
plastic strain rate W+O is given by:
W+O = ԑ–O (1 – D) (2.47)

This model has been used with modifications for composite materials for example [49, 50,
51] and ductile fracture in metals such as 2024-T351 [52] and Weldox 460E [53]. From LS
dyna material library, CDM can be implemented for anisotropic visco-plastic material
using the *MAT_104 (*MAT_DAMAGE_1) by setting FLAG ≥ 1. *MAT_105 or
*MAT_DAMAGE_2 can be used to apply CDM model elastic visco-plastic material.

2.2 Failure/Damage models from LS dyna materials library

Some of the failure criteria can have better predictions than others for a particular case. Some
are simple with a few parameters while others require more data, experiments and even trial
simulations for the implementation. This work does not intend to explore all different options
available but attempts to validate a simple but efficient method as far as possible to include
most of the advantages, especially the stress state effect.

28
Below are some material models in LS dyna library capable of modeling failure for ductile
fracture in metals [18]:

No. Material Model

1 Mat_013 (Isotropic elastic failure)
2 Mat_015 (Johnson_Cook)
3 Mat_017 (Oriented crack) (Brittle damage)
4 Mat_019 (Strain rate dependent plasticity)
5 Mat_24 (Piecewise linear plasticity) (Von Mises)
6 Mat_36 (3_Parameter_Barlat)
7 Mat_40 (Non-linear orthotropic)
8 Mat_52 (Bamman damage)
9 Mat_081/82 (Plasticity_with_damage) (Wilkins
damage model implemented in 82)
10 Mat_096 (Brittle damage)
11 Mat_098/99_Simplified_Johnson_Cook
12 Mat_104_Continuum_damage (anisotropic &
viscoplastic CDM)
13 Mat_105_Damage (elastic viscoplastic CDM)
14 Mat_107_Modified_Johnson_Cook
15 Mat_120_Gurson_JC
16 Mat_120_Wilkins_RcDc
17 Mat_123 (Modified piecewise linear plasticity)
18 Mat_124 (Plasticity compression tension)
19 Mat_153_Kinematic/isotropic_hardening
20 Mat_155 (Plasticity compression tension EOS,
extension of Mat_124)
21 Mat_190 (FLD 3-parameter Barlat)
22 Mat_221_Orthotropic_simplified_damage
23 Mat_224_Tabulated_Johnson_Cook
24 Mat_238 (Perturbation piecewise linear plasticity
based on Mat_24)
25 Mat_251 (Tailored properties, similar to Mat_24)
26 Mat_255 (Piecewise linear plastic thermal)
Table 2.1: Failure models from the LS dyna material library

29
Chapter 3

Stress State, failure criterion and

implementation

FE methods although originally developed for structural analysis, also extend to thermal, magnetic,
biomechanics, fluid mechanics and electrical analysis for numerical simulation [10]

Due to the advanced computational capabilities offered by modern processors, many commercial
efficient FE codes have been introduced in the market and are continuously under development.
A good knowledge of analysis requirements is necessary for evaluating feasibility in choosing
particular software offering different multitude of capabilities. LS DYNA, ABAQUS, ANSYS,
MSC NASTRAN, MSC DYTRAN, COSMOS, MARC, RHINO, I-DEAS, ADINA, COMSOL,
etc. are some of the popularly used software in the industry. ANSA, LS Prepost, Abaqus CAE,
Hypermesh, MSC Patran are some of the preprocessors used to generate finite element models for
the codes mentioned above. While most of these have post processing and time history processing
features, some special post processors such as TecPlot, HyperGraph, Hyper View, LS Tarus, LS
Prepost, etc. are also employed based on the analysis requirement.

In our case, we have used LS Prepost and Hypermesh for the preprocessors, mpp971_d_R7.0.0 for
solver and LS Prepost for the post processor. Before post processing, most solvers provide a log
file or batch output file, which should be searched for warnings or errors, and which will also

30
provide a quantitative measure of how well-behaved the numerical procedures were during
solution. The other important output files are plots, ASCII files, binary files and message files.
The results from these will be discussed in chapter 5.

The below paragraphs provide details of basics for a stress state dependent failure criterion:

3.1 Stress state characterization

The stress tensor 0˜ has six components and representing it in a six dimensional space makes
it too complicated.

C=™ š (3.1)

This component matrix can be simplified in the form of a diagonal matrix with the eigenvalues
or principal stress components ( , , , also called as principal stress invariants. For many
applications, it is easier to work with this diagonal principal stress matrix than fully populated
stress matrix.

0 0
C=™0 0š (3.2)
0 0

A vector notation made up of these principal stress directions in the Cartesian space is called
the Haigh-Westergaard space as shown in figure 3.1.

31
 Figure 3.1: Representation of vector in Haigh-Westergaard principal stress space
( , , ) and cylindrical coordinates (r, θ, z) [5]

The z-axis is the hydrostatic axis, where all the principal stresses are equal. The plane passing
through the origin and perpendicular to the hydrostatic axis is called the Π-plane. In general,
any plane perpendicular to the hydrostatic z-axis is called the octahedral or deviatoric plane,
like the one passing through the point N. In the deviatoric plane, the lode angle θ can be
measured along the positive direction from axis as shown in fig (3.2). In some research
works, alternative measurement of lode angle along positive direction of =- axis is also
used shown as ~4 in fig. (3.2). However, this thesis uses the former definition of lode angle
measurement.

32
 Figure 3.2: Geometrical representation of the principal stresses ( , , and deviatoric
principal stresses ( , , on the deviatoric plane [5]

In cylindrical coordinate system, the vector can be decomposed into two components:
1) Hydrostatic component › along the hydrostatic axis
Magnitude of vector › = | ›| = √3 G (3.3)
œ % % !
Where mean stress, G = = (3.4)

2) Deviatoric component › lying in the deviatoric plane

Magnitude of vector › = |› | = (3.5)

Where, is equivalent or Von mises stress from eqn. (2.3)

From Haigh-Westergaard space, the three principal stresses ( , , ) constitute the

Principal stress invariant triplet and the three cylindrical space coordinates (r, θ, z) constitute
the Lode coordinate invariant triplet [67]. Other invariant triplets are Principal invariant triplet
(• , • , • ), Mechanics invariant triplet (• , , ) and Fracture invariant triplet ( G, η, ξ)
[54]. Where the principal invariants are given by:

• = tr (σ) (3.6)

33
 • = tr [(tr (C – tr (σ2)] (3.7)

• = det (σ) (3.8)

If principal stresses are known, principal invariant triplets can be found by substituting eqn.
(3.2) into above eqns. (3.6), (3.7) and (3.8). Conversely, if values of principal invariants are
known, then values of principal stresses or eigenvalues can be found by solving characteristic
equation:

-• +• -• =0 (3.9)

Further the Lode coordinate invariants or cylindrical coordinates (r, θ, z) can be given by:

œ
√
z= (3.10)

r= 2 (3.11)

!
¡!
θ= J‚ ƒ¡ … ¢ (3.12)

œ
Coordinate z is constant multiple of pressure p = - . The lode angle θ varies monotonically

with the position of the middle principal stress relative to the high and low principal stresses.
Specifically, as middle eigenvalue moves from left side to the right side of outer Mohr’s circle
\ \
|
(fig. 2.4), the lode angle will vary from to - | . When r = 0, lode angle is mathematically

undefined, but is generally assumed as 0.

The principal invariants of deviatoric stress tensor S are ( , − , ) where:

= tr (S) = 0 (3.13)

= - tr [(tr (£ – tr (S2)] = tr (S2) (3.14)

= det (S) = tr (S3) (3.15)

34
 Negative sign is introduced in the definition of because it should always be positive. In
= 0, the set ( , − ,
√
other words, a mod value should be taken for as = ||S||. Since

) cannot be regarded as independent triplet. Specifically ( , − , ) are not sufficient to

determine • . Hence, the mechanics invariants are (• , , ) and can be considered as
independent.

And finally, the fracture invariants ( G, η, ξ) can be given as:

G is the mean or hydrostatic stress given by eqn. (3.4)

η is stress triaxiality, a ratio of mean (or hydrostatic) stress and equivalent (or Von mises)
stress given by eqns. (3.4) and (2.3) respectively.

"
i.e. η = (3.16)
#$

!
¡!
ξ= m o = sin 3θ
¡
(3.17)

Where ξ is alternative measure of Lode angle θ from fig. (3.1).

The fracture invariant triplet ( G, η, ξ) and lode coordinate (r, θ, z) invariant triplet are
homogenous of degree zero with respect to stress, meaning knowing the ratio of the stresses
is sufficient to calculate their values. Hence, values for η and ξ can be computed without stress
magnitude. The values for common stress states will be discussed in section 3.2 and can be
found in table 3.1. It can be shown that ( G, η, ξ) constitute an admissible independent
invariant triplet because having values for this triplet is sufficient to find values of any other
invariant triplet set. Thus, having one invariant triplet is sufficient to obtain any other invariant
triplet. Also, just one set of three invariants is sufficient to uniquely define the stress state at
a material point.

3.2 Stress state influence on equivalent plastic strain to failure

35
Almost all of the literatures on the stress state influence indicate that the two stress state
parameters triaxiality and lode parameter can be assumed to act independent of each other.
However, it is very difficult to design experiments that can control only one parameter at a
time. Due to this reason, the triaxiality and lode dependence has been studied separately. The
effect of hydrostatic pressure & lode angle can be uncoupled and determined from a separate
series of experiments. There are some standard experiments which give a particular
combination of average stress state parameters. These experiments can be used to study the
dependence of failure strain on a particular stress state. Xue [3], Bai [4], Bao [19], Basaran
[5] etc., used the axisymmetric tension specimens and grooved flat plate specimens to study
the effect of change of triaxiality with introduction of different notch sizes at constant lode
parameter for a particular type of specimen. Further, they also studied the effect of change in
lode angle parameter when they changed the specimen from axisymmetric tension (ξ = 1) to
grooved flat plate (ξ = 0). These two experiments are sufficient to establish the upper and
lower bounds depicted in figure (2.12) for the fracture locus. By controlling the notch sizes in
the two kinds of specimens, the lateral pressure acting on the specimen can be controlled,
thereby controlling the stress triaxiality. The next couple of sections discuss the effect of each
of the stress state parameters individually.

3.2.1 Pressure or triaxiality dependence η

It is generally observed that the pressure or triaxiality does not have a significant
influence on the plasticity [59-61]. Flow strength is considered independent of
pressure in this thesis. A conventional plasticity theory with the von mises yield
criteria is found to give satisfactory results without much error. From eqn. 3.4, the
mean or hydrostatic stress G accounts for the pressure or triaxiality effect.

Bridgeman [59] observed that subjecting the specimen to higher pressure in a pressure
chamber increases the strain at fracture. Pugh et.al [63] found that zinc exhibits a
sudden transition from brittle to ductile at a critical pressure of about 70 MPa. It was
concluded that continuous increase in pressure increases the fracture strain till the

36
value of pressure where no failure will occur at all, called the limiting pressure
designated as plim. Conversely, too low pressures may decrease the failure strain to
zero, called as the cutoff pressure shown as pcutoff. However, these extreme values of
pressures or triaxialities are never reached in practical situations and are only
theoretical extrapolations. Figure 3.3 shows the representation of observed change of
fracture strain with the pressure. It should be noted that compressive pressure is
considered positive whereas compressive stress is considered negative. For this
reason, pressure and mean stress represented on the same axis increase in the opposite
directions and vice versa.

Figure 3.3: General trend of reduction of fracture strain with pressure [3]

For some applications, yielding is also simply considered failure. In such a case,
simple relation between pressure and triaxiality can be established as follows:

From Von-Mises yield criterion, second deviatoric stress invariant,

J2 = k2 (3.18)

Where ‘k’ is yield stress in pure shear

= ( )
A
√
i.e. k = (3.19)

Pressure from eqn. 3.16 can be given as,

p = ƞ * σv = ƞ * σeq … as triaxiality ƞ = σm/ σeq = p/ σv (3.20)

37
 p = ƞ * (3 … considering σeq = σv = σy = (3 at fracture

p = ƞ * (3=

( ∗ A)
p=ƞ*
√

p=ƞ* H (3.21)

Using this relation, pressure values can be converted to triaxiality values and vice
versa.

Rice and Tracy [33] proposed a fracture strain dependence function on the triaxiality
as:
ԑ (η) = ¤ 7 ¥ 8
(3.22)
The Johnson-Cook triaxiality dependence has a similar form discussed in 2.1.5 (b).

3.2.2 Lode angle parameter dependence ξ

Clausing [66] observed that fracture strain for plain strain specimen (grooved flat
specimen) is significantly lesser than tensile specimen for seven structural steels.
Thus, for the same material, the fracture strain decreases when lode parameter changes
from axisymmetric (ξ =1) to plain strain ξ = 0). The reduction of fracture strain favors
the shear mode failure before the tensile or compressive failure mode. This explains
to some extent why in some cases of loading, the failure mode transitions to shear
mode from say, simple tensile.

It is generally observed that some materials have more lode dependency than others.
For example, aluminum alloy like 2024-T351 has more lode parameter dependency
than 5083-H116 [5]. On similar lines 1045 steel has more lode dependency than DH36
steel [4]. These works also indicate, aluminum alloys show more lode dependency
than steels.

38
 The dimensionless variable ξ from eqn. (3.17) was defined by Xue [3] and Wierzbicki
to characterize stress states based on lode angle parameter.

¦ ¡!
ξ= ! -1 ≤ ξ ≤ 1 (3.23)
#$

3.3. Typical specimen geometries for deviatorically proportional stress states

After having known the analytical formulation of stress state parameters, it is possible to derive
the values of triaxiality and lode angle parameter for simple loading cases. These loading cases
are constructive in determining the approximate fracture locus for a given material. However,
since the calculation of principal stresses becomes difficult with complex loading, trial
simulations can be performed where the FEA software calculates the stress state parameters
from the principal stresses.

Below are few loading cases where it is easy to calculate stress state parameters analytically:

3.3.1 Axisymmetric tension (calculation of ξ and η)

This stress state can be found in a loaded un-notched tensile bar. Figure 3.24 shows
the single element stress tensor. When un-notched bar is axially loaded with tensile
force, σ1 = σ1, σ2 = σ3 = 0

The lode angle parameter from Xue’s thesis [3] from eqn. 2.39 is,

2 2! ! )
2 2!
x= = = =0 (3.24)
!

39
 0 0
™0 0 0š
0 0 0

Figure 3.4: Axisymmetric tension and principal stress tensor

Lode angle θ can be calculated from lode angle parameter x as

θ = cot v m − ow
√ €
(3.25)

= cot m o
)
(3.26)

Therefore, cot(~ = ) and tan(~ = 0

Hence θ = 0o

Therefore, lode angle parameter ξ will be,

ξ = cos(3θ) = 1.00 (3.27)

Thus, lode angle parameter ξ = 1 corresponds to lode angle parameter x = 0

From eqn. 2.3, the Von mises or equivalent stress is given as:

= [( − +( − +( − ]

When un-notched bar is axially loaded with tensile force, σ1 = σ, σ2 = σ3 = 0

Therefore, = (3.28)

The mean stress, G from eqn. 3.4 is given by

40
 ( % %
G = = (3.29)

From eqn. 3.27 and 3.28, the stress triaxiality can be calculated as,

©
" !
η= = = = 0.33 (3.30)
#$

3.3.2 Uniaxial compression (calculation of ξ and η)

An example for uniaxial compression could be reduced friction upsetting tests. The
ideal single element principal stress matrix is shown in figure 3.5.

0 0 0
™0 0 0 š
0 0 −

Figure 3.5: Uniaxial compression and principal stress tensor

For frictionless upsetting test or uniaxial compression, σ1 = σ2 = 0; σ3 = -σ3

The lode angle parameter x [3] can be given by

2 2! ! !
2 2!
x= = = =1 (3.31)
! !

Therefore, lode angle parameter θ will be,

θ = cot v m − ow = cot-1 (1/ √3)

√ €

Therefore, cotθ = m o and tanθ = √3.

√

41
 Hence, θ = 60o (3.32)

Therefore lode angle parameter ξ will be,

ξ = cos(3θ) = -1.00

Thus, lode angle parameter ξ = -1 corresponds to lode angle parameter x = 1

From eqn. 2.3, the Von mises or equivalent stress is given as:

= [( − +( − +( − ]

For σ1 = σ2 = 0; σ3 = -σ3

= F = (3.33)

The mean stress, G from eqn. 3.4 is given by

( % % !
G = = (3.34)

From eqn. 3.33 and 3.34, the stress triaxiality for uniaxial compression is,

^©!
" !
η= = = -0.33 (3.35)
#$ !

3.3.3 Pure shear (calculation of ξ and η)

In most cases, shear failure occurs in combination with other modes. For pure shear
failure, the single element principal stress matrix is shown in figure 3.6

42
 0 0
™0 0 0 š
0 0 −
Figure 3.6: Pure shear and principal stress matrix

For pure shear, principal stresses σ1 = σ1, σ2 = 0, and σ3 = -σ1

The lode angle parameter x can be given by

2 2! !
2 2!
x= = = = 0.5 (3.36)
!

Therefore, lode angle parameter θ will be,

θ = cot v m − ow = cot-1 (3/ √3)

√ €

√
Therefore, cotθ =
√
and tanθ =

Hence, θ = 30o (3.37)

Lode angle parameter ξ will be,

ξ = cos(3θ) = cos (90) = 0.

Thus, for pure shear, lode angle parameter x = 0.5 corresponds to lode angle
parameter ξ = 0.

From eqn. 2.3, the Von mises or equivalent stress is given as:

= [( − +( − +( − ]

For σ1 = σ1, σ2 = 0, and σ3 = -σ1

43
 = 3 (3.38)

The mean stress, G from eqn. 3.4 is given by

( % % )
G = = = 0. (3.39)

From eqn. 3.38 and 3.39, stress triaxiality η will be,

" )
η= = = 0. (3.40)
#$

The table 3.1 gives the stress state parameter values for typical loading conditions:

Table 3.1: Triaxiality and Lode parameter values for typical loading [3, 4, 5, 29, 67]

Further, Bai [4] presented analytical formulation and typical values for the fracture strain,
triaxiality and lode parameter using 10 classical specimens as tabulated in table 3.2:

The notations used in the table are as follows:

R is the radius of a notch or groove, a is a radius of round bar at notch or the outer radius of
a torsional specimen, •) is initial radius, • is the radius at fracture, t is the thickness of the

44
 flat grooved plate at the groove, V) is the initial thickness, V is the thickness at fracture, «) is
the initial gauge length of the torsional specimen, ∆- is angular displacement to fracture of
a torsional specimen, *) is the initial grid spacing of an equi-biaxial testing specimen, and *
is the corresponding spacing to fracture.

No. Specimen type and Analytical value or The lode Analytical

loading formula for stress angle formula for
triaxiality η parameter ξ fracture strain ԑ®
1 4
2 ln v4¯ w
1 Smooth round bars, 1
tension 3 (

4
2 Notched round bars, + √2 lnm1 + °
o 1 4
2 ln v ¯ w
4(
tension (Bridgeman,
1952)
3 Plastic plain strain, √3 0 /
ln v ¯ w
√ /(
tension 3

√3 V /
ln v ¯ w
±1 + 2 ln v1 + w¶
4 Flat grooved plates, 0
√ /(
tension 3 4µ
5 Torsion or shear 0 0 ∆- •)
3«)
1 4(
2 ln m o
−
6 Cylinders, compression -1
3 4¯

7 Equi-biaxial tension 2 -1 ,
2 ln m ( o
3 ,¯

2 ,
8 Equi-biaxial compression
−
1 2 ln v ¯ w
3 ,(

/(
√3 ln m o
−
9 Plastic plain strain, 0
√ /¯
compression 3
4 4(
10 Notched roundbars, -ƒ + √2 ln m1 + o… -1 2 ln m4 o
° ¯
compression Bridgeman
(1952)
Table 3.2: Ten classical specimens for plasticity and fracture calibration [4]

45
3.4. The GISSMO model

The GISSMO (Generalized incremental stress state dependent damage model) is a

phenomenological failure/damage model proposed by Neukamm et.al. [68, 26, 27, 18]
originally for sheet-metal forming and crash modeling. This model can be used with other
material models and is invoked by the *MAT_ADD_EROSION keyword. This keyword
has many options for failure, damage initiation and damage evolution which can also be
applied simultaneously. Prime advantages in using this approach are the flexibility to
couple it with almost any material model and direct use of test data as input.

The parameter IDAM = 1 is the flag for GISSMO damage model. The damage can be
accumulated along the loading path and coupled to flow stress in different proportions
depending on the parameters.

An incremental damage accumulation similar to the one proposed by Johnson-Cook [12]

is available in GISSMO with a power function to address non-linearity.
m ^ o
at·¸¹º » a
∆r = ∆ԑº
¼½¾¿ÀÁ

ԑ(
(3.41)

Where, D is damage parameter ranging from 0 for no damage to 1 for complete damage
DMGEXP is the damage exponent defined in the input card
ԑ is the fracture strain which in turn depends on the lode angle parameter and
triaxiality. ԑ is obtained from the fracture locus defined using load curves of
fracture strain vs triaxiality in the LCSDG input. For different lode angles,
different load curves of fracture strain vs triaxiality can be tabulated using table
input for LCSDG.
For constant values of failure strain, this damage rate can be integrated to get a simple
relation of damage and actual equivalent plastic strain:
at·¸¹º
ԑ]
D=v w with ԑ as constant
ԑ(
(3.42)

46
 The coupling of damage is similar to Lemaitre’s [47] effective stress concept. In the post-
critical damage region, the inherent mesh-size dependency of results can be corrected using
the fading exponent ‘m’ (FADE input defined in the next few sections), damage exponent
(DMGEXP input) and the regularization curve (LCREGD input). A regularization curve is
nothing but a locus of scaling factors for the fracture criteria (failure strain in case of
GISSMO) which provides a correction to the over estimation of material strength for
coarser meshes.

3.5. Plasticity, damage & failure modeling

The main criteria for selection of a ductile material model are:

a) It should be able to give right results in different types of loading
b) It should be easy to calibrate the material parameters from experiments
c) It should have the correct combined plasticity and failure correlation
d) It should be able to account the effect of material weakening and deterioration

The approach chosen in this thesis is presented in next few sections.

3.5.1 Plasticity modeling:

There are numerous studies about complex plastic hardening behavior of ductile
materials under reversal loading. These studies include the kinematic models [55,
56] and kinematic-isotropic hardening models [57, 58]. However, in this thesis
reversal loading is not studied. Since it was found that neglecting hardening does
not induce significant errors under monotonous loadings, this topic is not given
importance.

Moreover, researchers like Bridgeman [59] and Brownrigg et.al [62] observed a
weak hydrostatic pressure dependence on material plasticity for some steels and
aluminum alloys. As far as lode dependence is concerned, the ductile fracture
47
 behavior strongly depends on the material. For example, Teng, Wierzbicki et.al [64,
65] showed that cast aluminum alloy ductile fracture behavior has considerable lode
parameter dependence. On the other hand, Bai [4] showed that DH36 steel, 1045
steel have weak lode dependence on the plasticity and fracture behavior. Further,
Basaran [5] studied lode dependence of different materials and established that
aluminum alloys like 2024-t351, 5083, etc. show somewhat more sensitivity than
steels like DP600, 1045 or DH36 for the plasticity behavior.

Experiments on 1045 steel, A710 steel, DH36 steel DP600 steel & 2024-T351
aluminum alloy were carried by Xue & Wierzbicki [3], Bao [19], Bai [4], and
Basaran [5]. They achieved good co-relation using J2 plasticity for simulation
results. Hence, we have used the Von mises yield criteria (J2 plasticity) based
*MAT_PIECEWISE_LINEAR_PLASTICITY material model for the plasticity
behavior in this thesis.

3.5.2 Damage accumulation:

The damage is initialized to a value of 0.01 for numerical reasons as soon as the
critical damage DCRIT is reached coupling of damage begins. Localization [6] of
stresses occurs and elements will be deleted if D ≥ 1. GISSMO offers three ways
in which damage can be coupled to material strength:

1) Input of fixed value of critical plastic strain (ECRIT ≥ 0): As soon as the
magnitude of plastic strain reaches the value defined by ECRIT, current damage
parameter D is stored as critical damage and the damage coupling flag is set to
unity, in order to facilitate the identification of critical elements in post
processing. From this point onward, damage is coupled to the stress tensor using
the relation:
a a¥°œd Ã>a¸¹º
 ±1 − m
σ=σ a¥°œd
o ¶ (3.43)

48
 This causes continuous reduction of stress, up to the load-bearing capacity
completely vanishing as D reaches unity. The fading exponent can be defined
element size dependent, to allow for the consideration of an element size
dependent amount of energy to be dissipated during element fadeout.

2) Input of a load curve defining critical plastic strain vs triaxiality (ECRIT ≤ 0):
This kind of input points to a load curve id. |ECRIT|. A triaxiality-dependent
material instability can be defined which controls the occurrence of instability
and localization depending on the actual load case. It is possible to use a
transformed forming limit diagram as an input for the expected onset of
softening and localization. Using this load curve, instability measure F is
accumulated using the following relation similar to the accumulation of
damage:

at·¸¹º
∆Ä = Äm ¼½¾¿ÀÁ
o
∆ԑ+
ԑ],gf
(3.44)

Where, F is instability measure (0 ≤ F ≤ 1)

ԑ+,1Q. is equivalent plastic strain to instability, determined from ECRIT

∆ԑ+ is the equivalent plastic strain increment

As soon as instability measure F reaches unity, the current value of damage D

in the respective element is stored. From this point onward, damage will be
coupled to the flow stress using the relation given above.

3) DCRIT input: If no input of ECRIT is made, parameter DCRIT will be

considered. Coupling of damage to the stress tensor starts if this value (damage
threshold) is exceeded (0 ≤ DCRIT ≤ 1) using the relation described above. This
input allows for the input of extreme values also. For example, DCRIT = 0
would lead to no coupling at all, and element will be deleted under full load
causing brittle fracture.

49
 We have used the first approach of damage accumulation as we don’t have the data
for instability curve (critical plastic strain vs triaxiality) or the transformed forming
limit curve (figure 2.10). A fixed value of 0.2 is used for ECRIT input in most cases,
depending on the type of loading and material for better results.

3.5.3 Failure prediction:

After the accumulated damage D exceeds unity, the material point is considered to
be failed and the element is deleted once a defined number of its material points
fail. The effective plastic strain to reach failure is selected from the input fracture
locus in the form of load curves depending on the triaxiality and lode angle
parameter. Depending on the choice of material model, state of stress, location of
stress concentration, rate of loading and temperature effects, failure can be
formulated to occur in the most favorable combination of parameters.

3.6. LS-dyna implementation and material keyword input

While GISSMO uses a strain-based failure criterion, the damage weighing function is
stress-based. This framework validates the transition from ductile to brittle or
tensile/compressive to shear fracture with stress state evolution. This section describes the
general approach used in this work for modeling the plasticity-failure behavior of the
materials used in experiments. The limited resources, infrastructure and time did not allow
carrying out actual experiments for validation. Hence, scope of this work is only to study
prior experiments done and simulate them using the GISSMO model implementation.
Some of them have been already validated using older material modeling approaches in the
literature.

Chapters 4 and 5 deal with different experiments from literature on ductile fracture. For a
particular experiment, we have attempted to use the same material data presented in the

50
original literature containing the experiment. However, in a few cases, because of lack of
sufficient documentation of material properties, we have used other sources for the material
properties of the same material. Although every data source is well documented in this
work, slight variation in the results is unavoidable in case the data and experiments are
from different literature for a particular material.

Following are the general steps for material model keyword definition:

a) Define the digitally obtained yield stress-equivalent plastic strain curve from the
references in the LCSS input of *Mat_24 card. This curve may be obtained from tensile
test experiments on axisymmetric specimens for the plasticity modeling.
b) Define the material properties like density, young’s modulus, poisson’s ratio and yield
strength in the *Mat_24 card
c) For fixed failure strain, input the value of FAIL in the *Mat_24 card. For stress state
dependent failure, define the *Mat_add_erosion card to apply GISSMO model.
d) For the GISSMO model, derive the fracture locus from the analytical fracture surface
expression in matlab and input it in the form of load curves in the LCSDG input of
*Mat_add_erosion. Define the pressure (triaxiality) dependence using equivalent
plastic strain to failure VS triaxiality curves. Define the lode angle dependence using
separate strain to failure VS triaxiality curves for each lode angle parameter in a tabular
form using *Define_table.
e) Assume trial values for critical plasitic strain (ECRIT), fading exponent (FADEXP),
damage exponent (DMGEXP). If necessary, eliminate the mesh-size effect of lack of
convergence after critical damage is reached by using suitable regularization curve in
LCREGD. For most cases the values of 0.2, 2 and 2 for the ECRIT, FADEXP and
DMGEXP give satisfactory results to start with depending on the material.

The input material data in steps a) through d) are unique for a given material, except for
the input of FAIL in *Mat_24 card. This is because the failure strain varies depending on
the stress state for a given temperature and strain rate. However, the input of FAIL is not
required when GISSMO model is used, as the failure strain is defined by the load curves

51
in LCSDG input instead. Apart from this, there is slight variation in the material properties
calibrated even for the same material by different researchers. Hence, to get better results
as far as possible, we have used the same material properties given in the literature
containing the experimental results if they are reported. The influence of the additional
effort to define LCSDG instead of FAIL on the ductile fracture prediction is studied in this
thesis. To do this, simulations are run using both the approaches for the same mesh
(experiments). In the next couple of chapters the simulation of experiments and their results
are discussed.

52
Chapter 4

Modeling, Simulation of Experiments and

Results

Apart from the difficulty in controlling any one stress state parameter in an experiment, the design
of experiments with negative stress state parameters can be further challenging. Although, the
possibility to conduct experiments on standard specimens which give a particular combination of
stress state parameters is often exploited to get few points of the fracture surface, the rest of the
points of the fracture locus needs assumptions, interpolation and extrapolation.

Few such experiments for the materials 2024-T351 and 1045 steel are presented in the next few
sections. Then the experiments on DP600 which were already validated by GISSMO previously
are simulated again to validate our GISSMO implementation approach. The mesh used is the
default Lagrangian with explicit time integration. Depending on the smallest element, LS dyna
automatically calculates the time step. Since these are temperature and strain rate independent
experiments, mass scaling [69] is exploited in some cases for faster processing of the quasi-static
simulation. For the low Poisson’s ratio, faster processing and stable simulation, single point
integration solid elements are used.

4.1. 2024-T351 aluminum alloy

53
Even though different literatures show very little difference in properties for the same
material, we have made best attempt to use the same properties as far as possible for the
experiments presented in that particular literature. The following properties for 2024-T351
aluminum alloy simulations were used by default unless they are presented in the reference
material to be different than this [70]:

Material property Value and consistent unit

Density 2.780e-006 kg/mm3
Young’s modulus 73.1 GPa
Poisson’s ratio 0.33
Yield strength 0.324 GPa
Table 4.1: 2024-T351 material properties

Xue [3] calibrated a symmetric fracture surface using four standard specimens namely un-
notched tensile, notched tensile, doubly grooved and upsetting specimen. These four
specimens are found to be sufficient for the positive load angle and triaxiality range, but
extrapolation is needed for extreme negative values of triaxiality and lode angle parameters.
The range of triaxiality and lode angle parameters covered by these four specimens is given
in table 4.2. Trial simulations can be performed to investigate the stress state from the
principal stresses in different elements.

Specimen Lode angle parameter Triaxiality

Un-notched tensile 1 0.33
Notched tensile 1 0.66
Grooved flat (plain strain) 0 0.5
Upsetting (compression) -1 -0.33
Table 4.2: Specimens used to calibrate the fracture locus using a six parameter analytical
formulation by Xue

54
 However, this fracture surface assumes symmetry and possibly causes over-estimation of
the fracture strain in the negative lode angle parameter range. Hence, we have also
implemented the non-symmetric six parameter fracture surface proposed by Bai [4] and
compared it with Xue’s symmetric fracture surface for 2024-T351. After implementing both
the analytical fracture surfaces proposed by Bai and Xue using the same parameters calibrated
by them in the literature for GISSMO, the fracture surface is tested for above four experiments
in addition to the 3-point bending and the projectile impact experiments for ductile fracture.
Table 4.3 shows the fracture locus parameters calibrated by Xue in his formulation which is
also briefly presented in section 2.1.5 d):

Fracture surface parameters ԑ ) z10G q γ m β

Xue’s calibrated value 0.8 800 MPa 1.5 0.40 2 2

Table 4.3: Calibrated parameters for symmetric 2024-T351 fracture locus [3]

Figure 4.1 shows Xue’s symmetric fracture locus in the space of triaxiality and lode angle
parameter obtained by substituting these parameters into (eqn. 2.34 through 2.37). Figure 4.2
shows Bai’s asymmetric fracture surface obtained by substituting parameters r through r|
from table 4.4 into the equation 4.1 for fracture surface. Note that θ here is alternative
definition of lode angle presented in eqns. 3.12 and 3.17:

Fracture surface parameters r r r rÅ rÆ r|

Bai’s calibrated value 0.5862 1.3576 0.2170 0.0040 0.4859 0.700

Table 4.4: Six calibrated parameters for 2024-T351 fracture surface [4]

Bai’s asymmetric fracture surface expression:

ԑ (η,θ) = ƒ (r 7 a 8
+ rÆ 7 aÇ 8
− r7 aÈ 8
… ~ + (r 7 a 8
− rÆ 7 aÇ 8
~

+r 7 aÈ 8
(4.1)

55
Figure 4.1: Xue’s symmetric fracture surface in the form of ԑ vs η curves for each ξ value

Figure 4.2: Bai’s asymmetric fracture surface in the form of ԑ vs η curves for each ξ value

4.1.1 Tensile specimens

56
For all the tensile specimens used by Xue the dimensions are shown in figure 4.3. A
constant diameter of 9 mm was maintained for all the specimens at the weakest section
to have the initial cross-sectional area at fracture constant.

Figure 4.3: 2024-T351 Tensile specimen dimensions tested by Xue [3]

The stress strain curve used in the LCSS field for all tensile tests for 2024-T351
aluminum alloy is shown in Figure. 4.4:

2024-T351 Tensile yield curve for LCSS

0.6

0.5

0.4
Yield stress

0.3
Tensile…
0.2

0.1

0
0 0.05 0.1 0.15 0.2
Equivalent plastic strain

Figure 4.4: Yield curve used for 2024-T351 tensile experiments [3]

57
 Single point integration solid elements of size 0.5 mm were used for general
discretization and stable behavior. The end surfaces of the specimen were meshed with
the butterfly mesh for proper discretization. However, to eliminate inaccuracies due to
mesh size dependent post critical damage behavior, regularization was also carried out
which will be dealt with in the next chapter. While one end of the gauge length was
fixed, an axial displacement was defined at the other end so as to cause the bar to
extend. Different notch sizes give different triaxialities but the lode parameter remains
constant at ξ = 1. After the un-notched tensile specimen, notched tensile specimens are
validated with the notch radius reduced by half with every next notched specimen to
reduce the triaxiality at the notch. For all the specimens, simulations were run with and
without GISSMO. In case GISSMO was not used, a constant failure strain was defined
in the FAIL input.

a. Un-notched round 9 mm diameter tensile specimen:

Figure 4.5: 2024-T351 Un-notched tensile specimen mesh

The un-notched tensile specimen is usually tested to calibrate the stress-strain curve
which captures the plasticity behavior of the material shown in figure 4.4. The mesh

58
used in the simulation is shown in figure 4.5. From the experiments, a cup-cone fracture
is expected for all the simulations. There is not a significant difference in the
predictions with and without GISSMO for the fracture in this test, except for the Force-
Displacement characteristics shown in figure 4.6. The stress state in the critical section
just before fracture is shown in figure 4.7. *MAT_24 is able to model the elastic and
the plastic flow prior to failure satisfactorily. But the GISSMO plot shows a smoother
drop in the post critical damage region thus giving closer approximation to the
experimental plot after the neck formation.

Force-Displacement plot for 2024-T351 Un-

notched tensile specimen
35
30
25
Experiment
20
Force (kN)

Mat24
15
GISSMO
10
5
0
-5 0 2 4 6
Displacement (mm)

Figure 4.6: Force-Displacement plot for un-notched tensile specimen for 2024-T351

Figure 4.7: Stress state in 2024-T351 un-notched tensile specimen

59
b. Notched round 18 mm notch radius specimen:

The notched tensile specimens change the triaxiality value at the location of fracture
by changing the pressure at the notch. The mesh for this experiment is shown in figure
4.8. The force-displacement plot for this specimen test is given in figure 4.9 and the
stress state in figure 4.10. It can be noticed that in the post-failure region, GISSMO
gives a better match than just *MAT_24.

Figure 4.8: Notched round specimen mesh with 18 mm notch radius

Force-Displacement for 18 mm radius

2024-T351 notched tensile specimen
40

30 Experiment
Force (kN)

Mat24
20 GISSMO

10

0
0 2 4 6
Displacement (mm)

Figure 4.9: Force-Displacement plot for 18 mm radius notched tensile specimen for
2024-T351
60
 Figure 4.10: Stress state in 2024-T351 18 mm notch radius tensile specimen at neck

c. Notched round specimen with 9 mm radius notch:

This specimen further reduces the pressure as compared to the 18 mm notch radius
specimen. This causes increase in mean stress and hence triaxiality at the critical
section. The specimen mesh is shown in figure 4.11 and the corresponding force-
displacement characteristics for the experiment in figure 4.12.

Figure 4.11: Notched round specimen mesh with 9 mm notch radius

61
 Force-Displacement for 9 mm radius 2024-
T351 notched specimen

40

30 Experiment
Force (kN)

Mat24
20 GISSMO

10

0
0 2 4 6
Displacement (mm)

Figure 4.12: Force-Displacement characteristics for the 9 mm radius notched

specimen test for 2024-T351

There is no significant difference between failure with GISSMO or *MAT_24, except

in the latter, the failure strain is provided by the fracture locus instead of fixed failure
strain. For constant temperature and strain rate, judgment of the fixed failure strain
should be based on the stress state which is not provided by *MAT_24. The stress state
at the critical section before fracture is shown in figure 4.13.

Figure 4.13: Stress state in 2024-T351 9mm notch radius tensile specimen at neck

62
d. Notched round specimen with 4.5 mm radius notch:

This is the specimen with the least notch radius and also the least pressure at the critical
section for 2024-T351. The mesh is shown in figure 4.14 and corresponding Force-
Displacement plot is shown in figure 4.15. Again, this specimen shows no difference
between the predictions with and without GISSMO, except for the post critical damage
region where there is a smooth drop in the axial force rather than abrupt one giving
closer approximation to the experimental plot.

Figure 4.14: Notched round specimen mesh with 9 mm notch radius

Force-Displacement for 4.5 mm

50
radius notched tensile specimen
40
Force (kN)

30
Experiment
20 Mat24
10 GISSMO

0
0 2 4 6
Displacement (mm)

Figure 4.15: Force-Displacement characteristics for the 4.5 mm radius notched tensile
specimen test for 2024-T351

63
 The stress state in the critical section before failure is shown in figure 4.16.

Figure 4.16: Stress state in 2024-T351 4.5 mm notch radius tensile specimen at neck

4.1.2 Upsetting specimens

These specimens are in the form of a cylinder of constant diameter with varying
lengths. For the compression of the cylinder, it is placed on fixed platen at the bottom
and another platen presses the specimen from the top. The platens were modeled as
rigid bodies to reduce the simulation time. No calculation is done for the rigid elements
by the solver. A surface to surface contact is defined between the platens and the
specimen and an eroding single surface contact is defined for the elements of the
specimen for avoiding internal penetration of the cracked surface. A very small gap of
0.05 mm is maintained between the specimen and the platens to maintain the shell
contact thickness between the solid elements of the specimen and the shell elements of
the platen. A co-efficient of friction of 0.4 is assumed between the contacting surfaces
between the specimen and the platen so as to avoid the specimen slipping out of the
platens. The yield curve from upsetting tests was used by Xue [3] for all his simulations
except the tensile tests mainly because: 1) Plastic deformation induced material
weakening is less significant in upsetting tests. 2) Fracture strain in upsetting test is
significantly higher than tensile test which allows more accurate yield curve fitting for
higher strains. Hence, we have also used the yield curve in compression for simulating

64
the 2024-T351 experiments presented in [3] i.e. Upsetting, 3-point bending and
grooved flat plate specimen tests. This curve is shown in figure 4.17. Two cylindrical
specimens of 8 mm constant diameter and 11.25 mm and 8 mm lengths are compressed
between platens. The force-displacement plot was compared between the simulations
and the experiments. The mesh and corresponding lengths are shown in figure 4.18.

Compressive yield curve for LCSS

0.7
0.6
0.5
0.4
Yield stress (GPa)

0.3
0.2
0.1
0
0 0.1 0.2 0.3
Plastic strain

Figure 4.17: Yield curve used for 2024-T351 simulations other than tensile tests [3]

Figure 4.18: 8 mm diameter upsetting specimen cylindrical meshes and their

corresponding lengths for 2024-T351 [3]

65
 The stress state just before crack occurrence is shown in figure 4.19.

Figure 4.19: Stress state in 2024-T351 upsetting specimen before crack failure

a. 11.25 mm long Upsetting specimen:

This is the longest upsetting specimen and hence the highest strained specimen before
fracture. The superimposed force-displacement plots are shown in figure 4.20.

11.25 mm long 2024-T351 Upsetting

specimen
50
45
40
35 Experiment
Force (kN)

30
25 Mat24
20
15 GISSMO
10
5
0
0 1 2 3 4
Displacement (mm)

Figure 4.20: Force-Displacement plot for 11.25mm length 2024-T351 upsetting test

It can be seen that there is no significant advantage with GISSMO for obtaining
accurate Force-Displacement plot in upsetting tests. However, there is an advantage of
using GISSMO when it comes to crack formation location and propagation direction.
66
Figures 4.21 and 4.22 show the fracture obtained with and without GISSMO. Figure
4.23 shows fracture in actual experiment.

Figure 4.21: Von mises stress plot for fracture initiation in 11.25 mm long 2024-T351
upsetting specimen obtained with GISSMO

Figure 4.22: Von mises stress plot for fracture initiation in 11.25 mm long 2024-T351
upsetting specimen obtained without GISSMO

It is observed that without GISSMO, the fracture is initiated at the center of the
specimen and then it propagates towards the outer surface. While with GISSMO, the

67
fracture is initiated at the outer surface corners of the specimen and then propagates
towards the center. This prediction of GISSMO is found to be a closer approximation
of the actual fracture in the experiment as shown in the figure 4.18. For upsetting tests,
it was also observed that the symmetric fracture surface proposed by Xue gave better
results than the asymmetric surface proposed by Bai as the latter gives low fracture
strain values in the negative stress state parameter range. Hence, for the upsetting tests
we have used the fracture surface proposed by Xue depicted in figure 4.1 and
formulated by eqn. 2.34 through 2.37. For all the other tests for 2024-T351 aluminum
alloy, Bai’s asymmetric fracture surface presented in figure 4.2 and equation 4.1 is
found to give a better result.

Figure 4.23: Actual fracture initiation in 11.25 mm long 2024-T351 upsetting

specimen experiment (a) cut of A-A plane (b) cut off B-B plane [3]

68
b. 8 mm long Upsetting specimen:

Figure 4.24 shows the force-displacement plot and figures 4.25 and 4.26 the crack
formation in 8 mm upsetting tests with and without the application of GISSMO
respectively. GISSMO does not give any edge over just *MAT_24 for the prediction
of the force-displacement characteristics, however, again, for fracture prediction, it
does.

8 mm long 2024-T351 Upsetting

specimen
80
Force (kN)

60
Experiment
40
Mat24
20
GISSMO
0
0 1 2 3 4
Displacement (mm)

Figure 4.24: Force-Displacement plot for 8 mm long 2024-T351 upsetting test

Figure 4.25: Von mises stress plot for fracture initiation in 8 mm long 2024-T351
upsetting specimen obtained with GISSMO

69
 Similar to the fracture in the previous 11.25 mm specimen, GISSMO predicts fracture
to initiate at the diagonally opposite outer surface corners where stress concentration
bands are formed and then propagate towards the center. The fixed failure strain criteria
of *Mat_24 gives the opposite result, where the fracture initiates at the center and then
propagates outwards. The primary reason for this mode change of fracture is that the
stress state favors slant shear fracture before the failure strain in compression is
reached.

Figure 4.26: Von mises stress plot for fracture initiation in 8 mm long 2024-T351
upsetting specimen obtained without GISSMO

4.1.3 Grooved flat plate

This specimen is characterized by two opposite grooves cut along the longer dimension
or length of the flat plate. The dimensions are shown in figure 4.27. Adams [71] had
previously conducted tests on such doubly grooved plates and suggested that the length
to height ratio need to be greater than 10 to ensure a plain strain condition for the
grooved plate. Trial simulation shows the stress state is plain strain or generalized shear
superimposed by hydrostatic tension when tensile force is applied perpendicular to
groove direction. The groove length direction is also the direction of intermediate
principal stress. Due to this, the crack is found to be developed at an angle of 45o to the
major principal stress direction as seen in figure 4.28. GISSMO is successfully able to

70
predict this slant crack as seen in Figure 4.29. Because of symmetry of the specimen,
two competing shear stress bands are seen perpendicular to each other at the same time
also called a slip planes. The fracture propagates in one of these slip planes as this mode
is favored for its low fracture strain before tensile failure. GISSMO also gives a closer
match to the force-displacement plots as the damage and weakening are considered in
the post critical damage area as shown in figure 4.31.

Figure 4.27: Grooved flat plate dimensions for 2024-T351 [3]

Figure 4.28: Slant crack observed during the groove flat plate experiment (a) entire
crack surface of one side of specimen (b) side view of interested section [3]

71
The same yield curve shown in figure 4.17 was used along with the properties from
table 4.1 by default for 2024-T351 aluminum alloy.

(a)

(b)
Figure 4.29: Crack prediction in 2024-T351 grooved flat plate specimen mid-section
(a) With GISSMO (b) Without GISSMO

The stress state in the critical section just before the slant fracture is shown in figure
4.30. The simulation results with GISSMO are found to be significantly better for 2024-
T351 grooved flat plate specimen.

72
 Figure 4.30: Stress state in 2024-T351 grooved flat specimen before fracture

Force-displacement plot for 2024-

T351 grooved flat plate tension
140
120
100
Force (kN)

80 Experiment
60 Experiment
GISSMO
40
Mat24
20
0
0 0.1 0.2 0.3 0.4 0.5 0.6
Displacement (mm)

Figure 4.31: Force-displacement plot for 2054-T351 grooved flat plate specimen

Overall, the classical experiments covering different stress states for 2024-T351
fracture surface can be represented as shown in figures 4.32 (a) and (b).

73
Figure 4.32 (a): 2024-T351 specimen experiments with unique stress states for
fracture surface testing [3]

Stress state location for different specimens

1.6
1.4
Upper bound
1.2 Lower bound
Fracture strain

1 Un-notched tensile
0.8 18 notch tensile

0.6 9 notch tensile

4.5 notch tensile
0.4
Grooved Flat
0.2
Upsetting
0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Triaxiality

Figure 4.32 (b): 2024-T351 experiments represented on the fracture strain –

triaxiality plot of the fracture surface upper bound (ξ = 1 or -1) and lower
bound (ξ = 0)

74
4.1.4 3-Point bending

Three-point bending experiments are often used to test crack propagation in ductile
metals. A rectangular cross-section beam is placed on two stationary pins and pressed
by a third central pin moving in the direction of the beam. All the pins are modeled as
rigid bodies as they are not analyzed structurally and also to save computational time
by excluding them from all calculations. Figure 4.33 shows the schematic arrangement
and the dimensions [3]. Since the simulation is quasi-static, mass-scaling is used to
save simulation time. The center 50 mm diameter pin is given a motion in the
downward direction so as to bend the beam. An automatic single surface contact was
used to deal with all kinds of contact situations encountered during deformation. Three
beam widths namely 10 mm, 30 mm and 60 mm are used in the direction perpendicular
to the plane of paper in the figure 4.34.

Figure 4.33: Schematic arrangement for 2024-T351 3-point bending

a. 10 mm wide 3-point bending specimen:

75
 Figure 4.34: Mesh used for 2024-T351 10 mm wide 3-point bending test

An isometric view of the mesh used in the 10 mm wide 3-point bending specimen
simulation is shown in figure 4.34. Figures 4.35 shows the force-displacement plots
obtained. The slight noise in the results is due to mass-scaling effect, although it does
not change the results.

Force-displacement plot for 10 mm

wide 3-point bending specimen
30
25
20
Force (kN)

15 Experiment
10 GISSMO
Mat24
5
0
0 20 40 60
Displacement (mm)

Figure 4.35: Force-Displacement plot for 2024-T351 10 mm wide 3-point bending

specimen test

76
 A crack in the tensile zone is attained in the experiment. Without GISSMO, this is not
correctly predicted in simulation and a crack in the compression zone is developed
instead as shown in figure 4.36 (a). Figure 4.36 (b) is able to show the correct fracture
prediction in the tensile zone using GISSMO.

Figure 4.36 (a)

Figure 4.36 (b)

Figure 4.36: Fracture obtained in the 10 mm wide 2024-T351 3-point bending
simulation (a) without GISSMO (b) with GISSMO

b. 30 mm wide 3-point bending specimen:

77
 Figure 4.37: Mesh used for 2024-T351 30 mm wide 3-point bending test

Figure 4.37 shows an isometric view of the 30 mm wide 3-point bending specimen and
the figure 4.38 shows the force-displacement plot for the experiment as well as
simulations.

Force-displacement plot for 30 mm

wide 3-point bending specimen
50

40

30
Force (kN)

Experiment
20 Mat24
GISSMO
10

0
0 20 40 60
Displacement (mm)

Figure 4.38: Force-displacement plot for 30 mm wide 2024-T351 3-point bending

specimen test

Similar to the 10 mm wide specimen, a fracture in the tensile zone is achieved in the
30 mm wide specimen experiment as well. Figure 4.39 (a) and (b) shows the crack

78
 prediction without and with GISSMO respectively. GISSMO predicts crack in the
tensile zone prior to its propagation/initiation in the compressive zone.

Figure 4.39 (a)

Figure 4.39 (b)

Figure 4.39: Fracture obtained in the 30 mm wide 2024-T351 3-point bending
simulation (a) without GISSMO (b) with GISSMO

As for the force-displacement plots, both GISSMO and *MAT_024 simulations give
satisfactory results.

c. 60 mm wide 3-point bending specimen:

79
 Figure 4.40: Mesh used for 2024-T351 60 mm wide 3-point bending test

Figure 4.40 above shows an isometric view of the mesh and figure 4.41 shows the
force-displacement plot obtained for 60 mm wide 3-point bending test specimen
simulation.

Force-displacement plot for 60 mm

wide 3-point bending test
100

80
Force (kN)

60
Experiment
40 GISSMO
Mat24
20

0
0 20 40 60
Displacement (mm)

Figure 4.41: Force-displacement plot for 60 mm wide 2024-T351 3-point bending

specimen test

The crack predictions without and with GISSMO are shown in figure 4.42 (a) and (b)
respectively. It is clearly seen that GISSMO correctly predicts crack initiation in only
the tension zone, as was established by many researchers like Tsuchida et. al. [73].
They validated the *MAT_81_ORTHO implementation which eliminated unwanted
failure in the compression zone.

80
 Figure 4.42 (a)

Figure 4.42 (b)

Figure 4.42: Fracture obtained in the 60 mm wide 2024-T351 3-point bending
simulation (a) without GISSMO (b) with GISSMO

Once again, there is no significant advantage in using GISSMO over the fixed failure
strain based *MAT_024 as far as force-displacement plots are concerned.

4.1.5 Projectile impact and ballistic limit

These simulations are often employed by civil aviation firms for turbine blade
containment studies [9, 13], by spacecraft designers for space debris effect studies
and survivability analysts for protecting aircraft or other vehicles from man-made
threats. An important measure of vulnerability is the probability of kill given a hit
(Pk/h), evaluated on a system or aircraft level and probability of component

81
dysfunction given a hit (PCD/H) [74]. If the threat is taken to be some kind of
projectile, PCD/H is among other things, a function of mass and velocity of the
projectile. The ballistic limit or limit velocity is the velocity required for a particular
projectile to reliably (at least 50% of the time) penetrate a particular piece of
material. In other words, a given projectile will not pierce a given target when the
projectile velocity is lower than the ballistic limit [75].

There are different methods for ballistic experiment analysis like empirical
equations, FE material models, witness plate, etc. While analytical models are
physics-based, they use simplifying assumptions to reduce the governing equations
to one and two dimensional equations. The projectile impact simulations performed
by *MAT_224 or MAT_MODIFIED_JOHNSON_COOK have been found to
produce very good results for ductile metal like 2024-T351 aluminum alloy [9].
However, the *MAT_224 constitutive model has a very complex implementation
and requires a lot of data for input. In comparison, GISSMO is a relatively simpler
implementation that can be combined with any suitable constitutive model. This
section provides a comparison of the approximations provided by the GISSMO -
*MAT_24 combination with the experiments similar to the previous results achieved
by *MAT_224 and penetration equations.

The *CONTACT_ERODING_SURFACE_TO_SURFACE keyword was used for

the impact with the rigid Ti6Al4V spherical projectile and subsequent failure of the
2024-T351 plate elements. The projectile was given an initial velocity, which was
varied until accurate ballistic limit was found. The residual velocity was noted for
each initial velocity including the cases with no perforation. The initial velocity VS
residual velocity plots were plotted to study the trend of velocity change. For
completion of simulation in reasonable time, only a quarter of the mesh for the plate
and projectile was modeled. Symmetry was exploited by applying symmetry
boundary conditions to the quarter model. With all other dimensions and mesh
discretization constant, three different thicknesses were used for the plate, viz.:
0.063”, 0.125” and 0.25”. To maintain consistency of units, all the quantities in the

82
input file were converted to the kilogram-millimeter-millisecond system. The
dimensions and velocities in the inch and ft/s units presented in the literature [9] are
also converted to mm and mm/ms for comparison in the plots respectively.

a) 0.063” thick 2024-T351 plate

This is the thinnest plate tested for ballistic limit. The mesh as shown in figure
4.43 is biased to increase in size away from the projectile monotonously so as to
discretize critical region of impact with finest element size and accommodate
larger elements at the non-critical region for cost-effectiveness.

Figure 4.43: Quarter model mesh used for 0.063” thick plate and 0.5” diameter
spherical projectile.

The velocity VS time plot at an initial velocity of -140.208 mm/ms which is just
marginally less than the ballistic limit for this plate is presented in figure 4.44.

83
 Figure 4.44: Velocity (mm/ms) VS time (ms) for 0.063” plate with initial
velocity of -140.208 mm/ms

Note that the initial velocity is negative and the residual velocity of 4.91 mm/ms
is positive indicating that the projectile has bounced back after impacting the
plate. Any increase in initial velocity beyond this would result in perforation of
the plate and a negative residual velocity. After impact/penetration, the velocity
reaches a constant value which is the residual velocity. Figure 4.45 shows the
chart marking the initial VS residual velocities for the 0.063” thick plate.

0.063" Plate Initial vs Residual Velocity

160.40
151.92
Initial velocity (mm/ms)

142.67
131.09
126.45
122.59 GISSMO
120.27 Experiment
119.47
119.46
100.26

-25 25 75 125
Residual velocity (mm/ms)

Figure 4.45: Initial VS residual velocity plot for 0.063” thick plate target

It is observed that GISSMO produces satisfactory results in combination with

*MAT_24 comparable to the complex tabulated Johnson-Cook (*MAT_224)
material model.

84
b) 0.125” thick 2024-T351 plate

The mesh for this plate is shown in figure 4.46. The only difference this mesh has
from the previous 0.063” thick plate mesh is the thickness dimension. The same
discretization is maintained, except the increased number of elements in the plate
thickness direction. This would eliminate any mesh size dependent results. For
this plate, the velocity VS time plot for velocity just below the ballistic limit i.e.
-219.456 mm/ms is shown in figure 4.47. Note that the residual velocity of 17.1
mm/ms is positive indicating the projectile bounces back after impact with the
plate. This residual velocity remains constant. Figure 4.48 shows the initial VS
residual velocity plot for 0.125” thick plate.

Figure 4.46: Quarter model mesh used for 0.125” thick plate and 0.5” diameter
spherical projectile.

85
 Figure 4.47: Velocity (mm/ms) VS time (ms) for 0.125” plate less with initial
velocity of -219.456 mm/ms

The bar graphs are categorized as per the initial velocities to compare residual
velocities. The GISSMO simulation results are comparable to the experiments
and a similar close match like the previous *MAT_224 simulations is attained.

0.125" Plate Initial vs Residual Velocity

257.99

247.97
Initial velocity (mm/ms)

237.17

227.13

220.15 GISSMO

218.53 Experiment

202.40

160.16
-20 0 20 40 60 80 100 120 140 160
Residual velocity (mm/ms)

Figure 4.48: Initial VS residual velocity chart for 0.125” thick plate target

c) 0.25” thick 2024-T351 plate

86
This is the thickest plate for which the ballistic limit is validated through
simulations. Once again, the same mesh discretization was used as in the previous
two cases except for the higher number of elements in the thickness direction
because of the double plate thickness. Because of fine discretization for all the
cases of ballistic limit simulations in the region of impact, no regularization study
was necessary. The mesh for this case is shown in figure 4.49. The force-
displacement plot for an initial velocity of -304.8 mm/ms just below the ballistic
limit is shown in figure 4.50. The constant residual velocity of 2.8 mm/ms
obtained is positive and indicates the projectile bounce back after impact with the
plate. Initial velocity increased above this velocity causes perforation of the plate.
Figure 4.51 shows the initial VS residual velocities graph for the 0.25” thick
2024-T351 aluminum alloy plate target. It is found that simulations with
GISSMO in combination with *MAT_24 are able to give better predictions than
the previous tabulated Johnson-Cook simulations presented in the literature for
the case of 0.25” thick target plate.

Figure 4.49: Quarter model mesh used for 0.25” thick plate and 0.5” diameter
spherical projectile.

87
 Figure 4.50: Velocity (mm/ms) VS time (ms) for 0.25” plate less with initial
velocity of -304.8 mm/ms

0.25" Plate Initial vs Residual Velocity

483.27
470.92
460.91
452.40
Initial velocity (mm/ms)

443.93
435.48
437.77
426.97 GISSMO
420.03
412.35 Experiment
409.24
406.08
409.89
412.83
402.85
318.37
311.46
-15 25 65 105 145 185 225 265 305 345 385
Residual velocity (mm/ms)

Figure 4.51: Initial VS residual velocity chart for 0.25” thick plate target

Overall, the comparison of ballistic limit predictive capability with experiments and
GISSMO simulations is presented in figure 4.52 [9]. For the results achieved by the
GISSMO approach, additional computational effort required is completely justified as
advantageous and comparable to the results achieved by previous *MAT_224.

88
 Ballistic limit for different plate thickness

Ballistic Limit Projectile Velocity (m/s)

400
350
300
250 Experiment

200
GISSMO
150
100
50
1 2 3 4 5 6 7
Plate thickness (mm)

Figure 4.52: Ballistic limit predictions for different plate thickness.

Apart from the ballistic limit velocities, we also compare the plate failure
characteristics in the below figures 4.53 (a) through (c).

(a) (b) (c)

Figure 4.53: Failure mode transition with plate thickness:
(a) Petaling: Bending & Necking (b) Mixed-mode: Bending & Spalling (c) Plugging:
Shearing & Spalling

89
 It is observed that the bending deformation of the plate before failure is reduced as the
plate thickness is increased. The thickest plate shows failure by shearing and spalling
while the thinnest plate shows failure by bending and necking. The GISSMO
simulations are able to predict this change in the failure mechanism as shown in the
figures. The pictures from the experiment are also shown for comparison.

Overall, the ballistic limit experiments were well simulated by a simple von-mises
plasticity model combined with GISSMO. The ballistic limit, residual velocities,
failure mechanism and other quantities which would otherwise be difficult to measure
were also predicted properly by the method applied.

4.2. 1045 steel

The 1045 steel material exhibits moderate plastic hardening and stress state dependency
which makes it suitable to be modeled with the ductile failure approach. The yield curve,
material properties, fracture locus parameters and the fracture locus are shown in figure
4.54, table 4.5, table 4.6 and figure 4.55 respectively. Once again, the same 6-parameter
asymmetric analytical fracture locus proposed by Bai [4] as presented in equation 4.1 was
used for 1045 steel. Notched tensile specimen, un-notched tensile specimen, grooved flat
specimen and torsion specimen were used to achieve different states of stress for the
fracture locus.
1045 steel Yield stress VS equivalent
1.5 plastic strain
Yield stress (GPa)

0.5

0
0 0.2 0.4 0.6 0.8 1
Equivalent plastic strain

Figure 4.54: 1045 steel yield curve for LCSS input [4]

90
 Material property Value and consistent unit
Density 7.8e-006 kg/mm3
Young’s modulus 220 GPa
Poisson’s ratio 0.3
Yield strength 0.5 GPa
Table 4.5: Material properties for 1045 steel [4]

Fracture surface parameters r r r rÅ rÆ r|

Bai’s calibrated value 0.7121 1.6968 0.5187 1.9454 0.7121 1.6968

Table 4.6: Six calibrated parameters for 1045 steel fracture locus by Bai [4]

Figure 4.55: 1045 steel fracture surface consisting of ԑ vs η curves for each ξ value [4]

The table 4.7 presents the different initial triaxiality and lode parameter values that the
selected 1045 steel specimens characterize. These values tend to change by some extent
with deformation, which should be taken into account for accurate value of fracture strain.

91
Also, these experiments are good to cover the positive region (0 to 1) of the fracture locus
stress state parameters.

Specimen Lode angle Triaxiality

parameter

Unnotched tensile 1 0.33

Notched tensile 1 0.5

Doubly grooved 0 0.68

(Plain strain)
Tube torsion 0 0

Tube tension 1 0.3

Table 4.7: Initial values of stress state parameters for the 1045 steel experiments

4.2.1 Tensile specimens

As is already known, these specimen have characteristic lode angle parameter of 1

and the triaxiality value can be controlled by introducing a notch of different sizes.
As per Bridgeman [59], the stress triaxiality at the center of the specimen is controlled
by a ratio of a/R empirically.

Figure 4.56: Necked cross-section in a round bar specimen

92
The Von mises or equivalent stress as in eqn (2.3) is given as:

= [( − ) +( − ) +( − ) ]

When un-notched bar is axially loaded with tensile force, typically σ1 = σ, σ2 = σ3 ~

0

Substituting, principal stresses, we get equivalent stress =

The mean stress, G is increased due to notch depending on notch radius ‘R’

and cross-sectional area ‘a’ as per empirical formula [4]

4 4
G =É 1 + 2 + 3Ê v + ln m1 + ow = v + ln m1 + ow
° °

From above equations, the stress triaxiality can be calculated as,

;
v %lnm % ow 4
= + ln m1 + o = 0.33 to 0.7 depending on notch radius.
" ! Ë
η=
°
=
#$

Two values of a/R were assigned to tensile specimens, a/R = 0 (smooth round bars)
and a/R = 1/3 (notched round bars). The gauge length chosen was 20.6 mm for both
specimens as per Bai’s experiments [4].

a) Un-notched tensile specimen

This specimen had a uniform gauge cross-sectional diameter of 9 mm with typical

stress state values as shown in below figure 4.57. The figure shows the mesh at
the neck cross-section just before the failure.

93
 Figure 4.57: Stress state parameters for 1045 steel un-notched tensile specimen

The force displacement plot shown in figure 4.58 shows satisfactory match of
simulation with experimental results for methods both with and without using
GISSMO.

Force-Displacement for 1045 steel

unnotched tensile specimen
60
50
40
Force (kN)

Experiment
30 Mat_24
20 GISSMO
10
0
0 1 2 3
Displacement (mm)

Figure 4.58: Force-displacement results comparison for 1045 steel un-notched

tensile specimen

b) 10.5 mm radius notched tensile specimen

94
This specimen had a minimum initial cross-sectional diameter of 7 mm at the
notch of radius 10.5 mm. The mesh at the neck cross-section just before failure
along with the stress state parameters is shown in figure 4.59.

Figure 4.59: Stress state parameters for 1045 steel 10.5 mm radius notched
tensile specimen

Similar to the un-notched tensile specimen, there is no significant difference in

the force-displacement results with and without using GISSMO for the notched
tensile specimen as well. Comparison is shown by superimposed force-
displacement plots in figure 4.60. The contribution of stress state dependent
modeling here is only to predict correct fracture strain.

Force-Displacement for 1045 steel 10.5

mm radius notched tensile specimen
50

40

30
Force (kN)

Experiment
Mat_24
20
GISSMO
10

0
0 1 2 3
Displacement (mm)

Figure 4.60: Force-displacement plot for 1045 steel 10.5mm notch radius tensile
specimen test

95
4.2.2 Grooved flat plate specimen

For most elements in the critical section, the typical principal stresses in a loaded
grooved flat plate before failure are found to be σ1 = σ, σ2 = σ/2, and σ3 = 0 as is seen
in the principal stresses plotted in figure 4.61.

Figure 4.61: Principal stresses in an element 91458 at the critical section of 1045
steel grooved flat plate specimen

σ 0 0
Typical stress tensor using principal stresses would look like ™0 σ/2 0š
0 0 0

The lode angle parameter from Xue’s thesis [3] as given in eqn. 2.39 is,
2 2! ! σ/2
2 2!
x= = = = 0.5
! σ

Lode angle θ can be calculated from lode angle parameter x as

θ = cot v m − ow = cot v m − ow
√ € √ ).Æ

= cot (1.732)
Therefore, cot(~ ) = 1.732 and tan(~ ) = 0.57736
Hence θ = 30o

96
Therefore, lode angle parameter ξ will be,
ξ = cos(3θ) = cos(90) = 0

Substituting the same values for principal stress i.e. σ1 = σ, σ2 = σ/2, and σ3 = 0 the,
von mises or equivalent stress from eqn. 2.3 would be:

= [( − ) +( − ) +( − ) ] = 0.866σ

The mean stress from eqn. 3.4, G is given by

É % % Ê
=m o =m o = 0.5σ
σ%).Æσ%))
G

From above equations, the stress triaxiality can be calculated as,

" ().Æ)
η=
).Í||
= = 0.577
#$

Figure 4.62 shows the stress states found in the 1045 steel 3.97mm notch radius
grooved flat plate specimen.

Figure 4.62: Stress state parameters in 1045 steel grooved flat plate

The used dimensions for the 1045 steel grooved flat plate [4] were as in table 4.8:

97
 Description Dimension
Thickness at the groove at the weakest cross section 1.6 mm
Radius of the groove 3.97 mm
Plate thickness at the specimen shoulder 5 mm
Gauge length 25.9mm
Table 4.8: Dimensions of 1045 steel grooved flat plate with 3.97 mm radius notch

Figure 4.63 shows the force-displacement results obtained for this specimen using
methods with and without GISSMO superimposed with the force-displacement plots
from the experiment for comparison.

Force-Displacement for 1045 steel

grooved flat plate specimen
80
70
60
Force (kN)

50 Experiment
40 Mat_24
30
20 GISSMO
10
0
0 0.1 0.2 0.3
Displacement (mm)

Figure 4.63: Force-displacement plots for 1045 steel grooved flat specimen

4.2.3 Torsion specimen

This specimen is in the form of hollow circular tube with a central notch as seen
in the figure 4.64 below which also shows the dimensions:

98
 Figure 4.64: 1045 steel torsion specimen dimensions [4]

The tubular type specimens were documented in Lindholm et al., 1980; White et
al., 1990. This specimen is important to represent the unique stress state
parameters of zero at the critical section when loaded.

a. Pure torsion test

Two plug-in aluminum cylinders were tightly placed inside the two ends of the
tube specimen and turning motion was applied to cause torsion quasi-statically
at 0.028o/s. The torque and the rotation in degrees was measured until failure.
Because torsion fracture is a pure shear condition, the two stress state
parameters are zero i.e η = ξ = 0 as derived in section 3.3.3. The triaxiality and
lode parameters at the critical section before failure are plotted in figure 4.65.

Figure 4.65: Stress state parameters plot for 1045 steel torsion specimen

99
For pure shear, the principal stress would be σ1 = σ, σ2 = 0.0 and σ3 = −σ as
can be seen from the principal stresses plotted at an element 809651 from the
critical section in figure 4.66.

Figure 4.66: Principal stresses (GPa) with time (quasi-static) plotted at an

element in critical section of 1045 steel torsion specimen.

The torque vs rotation plot comparision for the experiment A and simulations
with and without GISSMO are superimposed and presented in figure 4.67:

Torque vs rotation plot for 1045

steel torsion specimen
500
450
400
350
Torque (Nm)

300 Experiment
250 Mat24
200
150 GISSMO
100
50
0
0 5 10 15 20 25
Rotation (degrees)

Figure 4.67: Torque (Nm) vs rotation (degrees) for 1045 steel torsion
specimen

100
 There is no significant advantage in using GISSMO in this case as far as
fracture prediction is considered, except for the stress state dependency that
affects the fracture strain.

b. Tension test of torsion specimen

This test basically uses the same specimen from the previously discussed
torsion test but instead of applying torque or rotation, an axial load tensile load
is applied quasi-statically until fracture. The force and displacement over the
gauge length of 12.7 mm is noted. The resulting stress state at the critical section
before fracture is shown in figure 4.68.

Figure 4.68: Stress state parameters in 1045 steel tube tension loading

The corresponding force-displacement plot is shown in figure 4.69 for the

experiment and simulations with and without using GISSMO.

101
 Force vs displacement in 1045
steel tube tension
100

80

60 Experiment
Force (kN) 40 Mat24

20 GISSMO

0
0 0.1 0.2 0.3
Displacment (mm)

Figure 4.69: Force vs displacement for 1045 steel tube tension

There is no significant difference in fracture shape prediction with or without

GISSMO, except for the slow deterioration of material in the post-critical
region in the case of the former. Also, because of stress state consideration, the
fracture strain is predicted correctly. Thus the simulations capture the stress
state well and the corresponding ductility predicted is also studied to be close
to the experimental results.

Overall, the experiments for 1045 steel cover different states of stress and can
be represented on the fracture surface as shown in figure 4.70 and 4.71

102
Figure 4.70: Different experiments with unique stress states for 1045 steel

Stress state location for different 1045 steel

specimens
4
3.5
Upper bound
3
Fracture strain

Lower bound
2.5
Tube tension
2
Un-notched tension
1.5
Notched tension
1
Grooved flat
0.5
Tube torsion
0
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Triaxiality

Figure 4.71: 1045 steel experiments represented on the fracture strain –

triaxiality plot of the fracture surface upper bound (ξ = 1 or -1) and lower
bound (ξ = 0)

103
Chapter 5

DP600 experiments, Damage exponent effect

and Regularization

It is a known fact that a fracture surface fitted using optimized parameters is bound to have slight
error due to interpolation, extrapolation and simplification tolerances. As GISSMO only targets
the damage and failure prediction, it does not work separately but has to be used in conjunction
with a constitutive model that provides the underlying plasticity formulation. The yield curve and
material parameters used affect the results before the critical damage is reached. After the
implementation of GISSMO failure criteria, this chapter aims to analyze the correctness of the
plasticity-failure approach used. For this purpose, experiments previously validated with GISSMO
are again simulated to find the variation. Also, the accuracy is affected by the selection of the mesh
sizes. Optimum selection of mesh size would be the size that gives reasonably low variation in
results on reducing the mesh size further, without significant computing time increase. One of the
methods to eliminate or rather reduce the effect of mesh size variation is defining the regularization
factors to scale the failure criteria, in our case, the equivalent plastic strain to fracture. This chapter
aims to highlight few of the simplifications used in our approach and its implications for adopting
effective guidelines to achieve better results as a pre-requisite for future scope of this kind of work.

104
5.1 DP600 specimen experiments

The DP600 experiments were simulated solely to validate the approach in this thesis,
especially the constitutive model used. Unlike the 2024-T351 aluminum alloy and the 1045
steel experiments, the objective is not to study the influence of stress state dependency on the
ductile fracture behavior but to study the correctness of the definition of *MAT_24 and
*MAT_ADD_EROSION cards. For this reason we have not done comparison of separate
simulations run with and without using GISSMO failure model. Instead, we directly compare
simulation results with experiments. Since these DP600 experiments were already simulated
using GISSMO, it is easier to know what should be expected from this present
implementation. The yield curve and material properties used are shown in figure 5.1 and
table 5.1 respectively.

DP600 Yield stress VS equivalent

plastic strain
0.8
Yield stress (GPa)

0.6

0.4

0.2

0
0 0.05 0.1 0.15 0.2
Plastic strain
Figure 5.1: DP600 steel yield curve for LCSS input

These general DP600 parameters were used from [76] as the original source [5] did not
specify the exact parameters that were used:
Material property Value and consistent unit
Density 7.830e-006 kg/mm3
Young’s modulus 200 GPa
Poisson’s ratio 0.30
Yield strength 0.362 GPa
Table 5.1: Material properties used for DP600 steel

105
For the fracture surface, we have used the 9 parameter analytical surface presented by Basaran
[5] with the following formulation.

(r + r% − r) + (r exp(−r Ò + r% exp(−r % Ò −
ԑ (η, ξ) = Î Óξ
r) exp(−r) Ò

+ ƒ (r% − r + (r% exp(−r% Ò − r exp(−r Ò … Õ

+ r) + r) exp(-r) η) (5.1)

Table 5.2 shows the already fitted parameters used for DP600 analytical fracture surface:

Fracture surface r % r% r% r) r) r) r r r
parameters
Basaran’s 0.783 0.391 0.953 0.404 0.534 0.706 0.018 1.298 0.725
calibrated value
Table 5.2: Parameters for the DP600 analytical fracture surface [3]

When plotted in Matlab as separate fracture strain-triaxiality curves for each lode angle
parameter, the fracture surface appears in the (η-ξ) space as shown in figure 5.2.

Figure 5.2: DP600 fracture surface plotted in the (η-ξ) space

106
Table 5.3 lists the different stress states obtained with the different DP600 specimen
experiments as reported by Basaran [5]. Since we are not interested in validating the DP600
stress state effect, we only choose the following few simple tests covering the positive range
(0 to 1) of stress state parameters.

Lode parameter → 1 0.36 0.73 0

Triaxiality ↓

1 0.5 mm radius
grooved flat

0.95 0.5 mm
notch round
tensile

0.74 2 mm notch
round tensile

0.7 4 mm notch
round tensile

0.88 1 mm radius
grooved flat

0.61 Un-Notched
tensile flat

0.67 2mm notch

flat tensile

Table 5.3: DP600 experiments selected and their typical stress state parameters [5]

The following sections describe the DP600 specimens simulated for evaluation of the adopted
von-mises and GISSMO approach:

107
5.1.1 Tensile specimens

a. Un-notched flat specimen

Figure 5.3 (a) shows the dimensions used for the un-notched flat tensile specimen.
One end of the specimen is held fixed while axial displacement is defined at the
other end away from the specimen.

(a) (b)

Figure 5.3: DP600 steel un-notched flat specimen (a) Dimensions [5] (b) Mesh

To minimize mesh size effects, fine mesh of mesh size of 0.2 mm was used for
sufficient resolution without the need for regularization. To complete the simulation
in reasonable times, non-critical regions of the specimen having low deformation
are modeled with coarse mesh. The mesh is shown in figure 5.3 (b). Cross-sectional
force is extracted using the SECFORC option in *DATABASE_ASCII_OPTION.
This force is normalized by dividing with the original cross-sectional area A in the

108
gauge length. The displacement is also normalized by dividing with the gauge length
of the specimen L0. Figure 5.4 shows the plastic strain, triaxiality and lode parameter
in the critical section of the specimen before fracture. Figure 5.5 shows the
normalized force-displacement plots for experiment as well as simulation.

Plastic strain Triaxiality Lode angle parameter

Figure 5.4: The plastic strain, triaxiality and lode parameter before fracture at the
critical section of DP600 un-notched flat tensile specimen

DP600 Un-notched tensile flat

specimen
Normalized force (F/A0)

0.8

0.6
kN/mm2

0.4 Experiment
GISSMO
0.2

0
0 0.1 0.2 0.3 0.4 0.5 0.6
Normalized displacement (L/L0)

Figure 5.5: Normalized force-displacement plots for DP600 un-notched tensile flat
specimen experiment and simulation

It is seen that there is deviation between experiment and simulation results in the
elastic-plastic as well as post-critical damage region. This deviation is not present in
Basaran’s GISSMO implementation [5] which means he might have used slightly
different material parameters and yield curve. Nonetheless, our aim is only to study

109
 the reproducibility of the results using similar approach. Further tests will confirm
if the method we followed was close to the previous one or not.

b. Notched flat 2 mm notch radius specimen

The dimensions and the mesh for the specimen are shown in figures 5.6 (a) and (b)
respectively. Once again the optimum mesh size of 0.2 mm was selected in the
critical region and coarse mesh elsewhere for conservative model size and
computing time.

(a) (b)

Figure 5.6: DP600 2 mm notch tensile flat specimen (a) Dimensions [5] (b) Mesh

The plastic strain, triaxiality and lode parameter distribution at the critical section
before failure is shown in figure 5.7.

110
 Plastic strain Triaxiality Lode angle parameter

Figure 5.7: The plastic strain, triaxiality and lode parameter distribution at the
critical section of DP600 2 mm notch tensile specimen before failure

The normalized force vs displacement plot is shown in figure 5.8.

DP600 2mm notch flat tensile

specimen
Normalized force (F/Ao) kN/mm2

0.8

0.6

0.4 Experiment
GISSMO
0.2

0
0 0.02 0.04 0.06
Normalized displacement (L/Lo)

Figure 5.8: Normalized force vs displacement for DP600 2 mm notch flat tensile
specimen

It is seen that the elasticity and plasticity region result matches well with the
experiment but the damage and ductile fracture behavior is captured with less
accuracy by the applied plasticity and failure model. This indicates that the
approximated failure locus we used from the analytical formulation and fitted

111
 parameters does not correspond well to the actual fracture locus calibrated from the
experiments. Hence, the original paper [5] has better simulation results than this
implementation.

c. Notched round 0.5 mm notch radius specimen

The figure 5.9 (a) shows the specimen dimensions and figure 5.9 (b) shows the mesh
for this specimen. For round tensile specimens, the lode parameter is constant at
unity while the triaxiality can be varied by changing the notch radius. This is the
least notch radius specimen hence has the highest triaxiality of 0.95. The higher
notch sizes will also be dealt with in the next sections to check effect of reduced
triaxiality on fracture strain.

(a) (b)

Figure 5.9: 2 mm notch DP600 round tensile specimen (a) Dimensions[5] (b) Mesh

112
Because the small size of the specimen, a fine mesh of 0.1 mm size was used. For
the round end surfaces, a butterfly type of mesh was used. The plastic strain,
triaxiality and lode parameter distribution at the critical section just before fracture
is shown in figure 5.10.

Plastic strain Triaxiality Lode angle parameter

Figure 5.10: Plastic strain, triaxiality and lode parameter distribution for 0.5 mm
notch DP600 tensile specimen

Figure 5.11 shows the force-displacement plot comparison between simulation and
experiment which shows the simulation results to match satisfactorily.

Normalized Force-Displacement for

DP600 0.5 mm notch round tensile
specimen
Normalized force F/Ao (kN/mm2

1
0.8
0.6
Experiment
0.4
GISSMO
0.2
0
0 0.01 0.02 0.03 0.04
Normalized displacement L/Lo

Figure 5.11: Force-Displacement plot for DP600 0.5 mm notch radius tensile
specimen test

113
 The initial deviation in the plastic region can be attributed to the inaccurate von
mises plasticity model definition due to lack of data.

d. Notched round 2 mm notch radius specimen

The dimensions for this specimen remains the same as the previous 0.5 mm radius
notch round tensile specimen except for the notch radius in this case has increased
to 2 mm. The mesh for the specimen with 0.1 mm size elements is shown in figure
5.12 below:

Figure 5.12: Mesh for DP600 2 mm radius notch round tensile specimen

While the lode parameter remains constant at 1, the triaxiality has reduced due to
increased notch radius. The plastic strain, triaxiality and lode parameter distribution
at the critical section just before fracture is shown in figure 5.13.

Plastic strain Triaxiality Lode angle parameter

Figure 5.13: The plastic strain, triaxiality and lode parameter distribution at the
critical section for DP600 2 mm notch round tensile specimen before failure

114
 The force-displacement plot for this specimen simulation gets better than the
previous specimen and is presented in the figure 5.14 below:

Normalized force-displacement plot for

Normalized force F/Ao (kN/mm2
DP600 2 mm notch round tensile specimen
0.8

0.6

0.4 Experiment
GISSMO
0.2

0
0 0.02 0.04 0.06
Normalized displacement L/Lo

Figure 5.14: Force displacement plot for DP600 2 mm notch round tensile
specimen test

There is still some deviation in the simulation results in the plastic flow region which
is again due to lack of proper material data presented in the source for the
constitutive model.

e. Notched round 4 mm notch radius specimen

Again for this specimen, the dimensions remain the same as the previous specimen
except for the notch radius of 4 mm to have reduced triaxiality. The mesh for the
specimen is shown in figure 5.15. An optimum mesh size of 0.1 mm was used for
this specimen.

Figure 5.15: Mesh for the DP600 4 mm notch round tensile specimen

115
The figure 5.16 shows the plastic strain, triaxiality and lode parameter distribution
at the critical section just before fracture.

Plastic strain Triaxiality Lode angle parameter

Figure 5.16: Plastic strain, triaxiality and lode parameter distribution in DP600 4
mm notch round tensile specimen critical section just before failure

For this specimen, the normalized force-displacement plot gets even better with the
simulation plastic flow region actually showing lower values than the experiments.
The plot is shown in figure 5.17.

Normalized force-displacement plot for

DP600 4mm notch round tensile
specimen
Normalized force F/Ao (kN/mm2

0.7
0.6
0.5
Experiment
0.4
0.3 GISSMO
0.2
0.1
0
0 0.02 0.04 0.06
Normalized displacement L/Lo

Figure 5.17: Normalized force displacement plot for DP600 4 mm notch round
tensile specimen test

116
 However, the simulation results are still not as close to the ones presented in the
source [5]. These results indicate the inefficiency in this modeling effort and its
efforts on other simulations performed, which will be discussed in the concluding
section. Suggestions to improve these results will also be proposed.

5.1.2 Grooved flat plate specimens

These specimens are important from the standpoint of producing stress state of zero
lode parameter. The same triaxialities (ranging 0.5 to 1) as the tensile specimens can
be achieved with the variation of notch radius. Here we will choose two notch radius
sizes to study effect of reduced triaxiality on fracture strain as was presented by Basaran
[5]:

a. 0.5 mm notch radius grooved flat plate specimen

The stress state in the case of grooved flat plates in usually classified into the plain
strain category, since the specimen has flat geometry but also a sufficient thickness.
If it were not for the thickness, very thin specimens could be classified into plane
stress condition. Because of more thickness, more resistance is applied to the
stretching of the elements along the plane, hence the stresses normal to the plane of

117
the grooved flat plate are not zero like in plane stress condition.

(a) Dimensions (b) Mesh

Figure 5.18: DP600 0.5 mm notch radius grooved flat plate specimen

Figure 5.18 shows the dimensions of the specimen and the mesh. Because of the
plain strain loading condition, the fracture strain is found to be less than the previous
axisymmetric loading condition. Figure 5.19 shows the plastic strain, triaxiality and
lode parameter distribution at the critical section of the specimen before fracture.

Plastic strain Triaxiality Lode angle parameter

Figure 5.19: The plastic strain, triaxiality and lode parameter distribution at the
critical section before failure of DP600 0.5 mm notch radius grooved flat plate

118
 The normalized force-displacement plots for the experiment and gissmo simulation
are superimposed for comparison as shown in figure 5.20.

DP600 0.5mm radius notch grooved flat plate

Normalized force F/Ao (kN/mm2 0.9
0.8
0.7
0.6
0.5
Experiment
0.4
0.3 GISSMO
0.2
0.1
0
0 0.005 0.01 0.015 0.02

Normalized displacement L/Lo

Figure 5.20: Normalized force-displacement plot for DP600 0.5 mm notch radius
grooved flat plate specimen test.

The above simulation results match better with the experiment in the plastic flow
region than the tensile specimens discussed before for the same yield curve and
material properties. However, there is little deviation in the post-critical damage
region which can be attributed to the different size of the specimen and hence the
element size that was selected for economical computing resource utilization. Mesh
size based regularization discussed in the next part of the chapter could fix this
deviation in the damage region. However, our aim is only to evaluate the present
method of implementing the ductile fracture method and attempt to replicate
previous similar GISSMO - *MAT_024 implementation for DP600 specimens with
approximate data.

b. 1 mm notch radius grooved flat plate specimen

119
The dimensions for this specimen remain the same as of the previous 0.5 mm notch
radius grooved flat plate except for the increased notch radius of 1mm. Figure 5.21
shows the dimensions and mesh for this specimen.

(a) Dimensions (b) Mesh

Figure 5.21: DP600 1 mm notch radius grooved flat plate specimen

Because of the increased notch radius which is twice that of the previous specimen,
the triaxiality is reduced to 0.88 from about 1. The lode angle parameter remains the
same at 0. This would make the fracture strain to be reached somewhat earlier than
the previous case. Figure 5.22 shows the plastic strain, triaxiality and lode parameter
distribution at the critical section of the specimen before fracture.

120
 Plastic strain Triaxiality Lode angle parameter
Figure 5.22: The plastic strain, triaxiality and lode parameter distribution at the
critical section before failure of DP600 1 mm notch radius grooved flat plate

Figure 5.23 shows the normalized force-displacement plot comparison for

experiment and simulation of this specimen. While a good match is achieved in the
initial elastic-plastic flow region, the post-critical damage region shows more
deviation. However, the original simulation performed in the reference [5] achieves
closer results because there actual variety of experiments performed supported the
fracture locus points directly. Although the same constitutive material model and
fracture model as adopted in this thesis was used, better data was used to produce
more realistic and accurate simulation rather than analytical approximation.

DP600 1 mm notch radius grooved flat

plate
1
Normalized force (kN/mm2

0.8
0.6
Experiment
0.4
GISSMO
0.2
0
0 0.01 0.02

Normalized displacement L/Lo

Figure 5.23: Normalized force-displacement plot for DP600 1 mm notch radius

grooved flat plate specimen test

121
 With this specimen test one more point in the zero lode parameter and nearly unity
triaxiality region of fracture surface is established. These experiments are sufficient
to cover the positive side of the fracture surface stress state parameters. Since we are
not establishing the complete fracture surface but only testing the implementation of
an analytical approximation of previously established fracture surface, the number
of different kinds of experiments simulated are limited. With approximate data, we
get an idea of what accuracy of fracture surface a complete set of actual experiments
can accomplish for a given material. The results indicate what measures can be taken
to have a better validation which will be discussed in the next chapter.

Graphically, the DP600 experiments simulated in this thesis can be represented on the fracture
surface by the following points as shown in figures 5.24 and 5.25. We have used these
simulations just to evaluate the GISSMO and constitutive model implementation and not to
study the stress state effect as mentioned before. Hence, we have only selected few specimen
experiments to be simulated not covering all the cases from the source as they were previously
already validated using GISSMO and *MAT_024.

122
Figure 5.24: DP600 experiments simulated characterizing different stress states on the
fracture locus

Stress state location for different DP600 specimens

3
Lode parameter -1
2.5
Lode parameter 0
2 Lode parameter 1
Fracture strain

0.5 mm R grooved flat

1.5 0.5mm notch round tensile
2mm notch round tensile
1 4mm notch round tensile
1mm R grooved flat
0.5
Unnotched tensile flat
0 2mm notch tensile flat
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Triaxiality

Figure 5.25: DP600 experiments simulated characterizing different stress states on the
fracture strain-triaxiality plot for the fracture locus

123
5.2 Effect of damage exponent variation

Here damage exponents 1, 2, 3, 4 & 5 were studied to have the following effect on the force-
displacement plots of 2024-T351 aluminum alloy grooved flat plate discussed in section 4.1.3
of the thesis. In all the cases a slant fracture was observed except the fracture pattern started
becoming slightly irregular as the damage exponent increased. The plots in figure 4 show the
force-displacement results.

2024-T351 Grooved flat plalte with DMGEXP from 1 to 5

160

140

120

100 DMGEXP 1
DMGEXP2
Force kN

80
DMGEXP 3
60
DMGEXP 4
40 DMGEXP 5

20 Experimental

0
0 0.1 0.2 0.3 0.4 0.5 0.6
-20
Displacement mm

Figure 5.26: Force-displacement plots for different damage exponents in the case of 2024-T351
grooved flat plate specimen loading

124
 (DMGEXP = 1) (DMGEXP = 2)

(DMGEXP = 3) (DMGEXP = 4)

(DMGEXP = 5)

Figure 5.27: Fracture pattern for damage exponents ranging from 1 to 5 for 2024-T351
aluminum alloy grooved flat plate specimen loading

125
 The fracture pattern for each of the simulations for different damage exponents with 2024-
T351 aluminum alloy grooved flat specimen are presented in figure 5. It can be seen that
although the pattern remains of slant nature, it tends to become more irregular as the damage
exponent increases. Also, increasing the damage exponent increases the predicted strength of
the ductile material in the simulation results. In our case, the damage exponent = 2 matches
well with the experimental results. For all the 2024-T351 simulations we used damage
exponent of 2 except for the 8 mm long upsetting specimen where it was 4.

5.3 Mesh size based regularization

Since the direct determination of the yield curves from the specimen tests is not possible for
the post-critical range of deformation, stress extrapolation based on engineering assumptions
(or models) is frequently used. Also, because of the inherent mesh-dependency of results in
the post critical range, the used parameters of extrapolation would determine the material
properties in the post-critical range, and lead to mesh dependent results.

The critical damage represents the point of material instability and therefore end of mesh size
convergence of results. For practical application of the model to finite element simulations
with limited mesh sizes, this marks the beginning of the need for regularization of different
mesh sizes. For GISSMO, the regularization treatment is combined with the damage model.
Basically, this controls the amount of energy that is dissipated in the process of crack
development and propagation [72]. For a finite element model this results in variation of rate
of stress reduction through the element fadeout and Lemaitre’s effective stress concept
involving damage. This strategy allows for regularizing not only the fracture strains but also
the energy consumed during the post-critical deformation. A reasonably good regularization
of the resulting force-displacement curves in tensile tests with different mesh sizes can be
achieved by proper selection of mesh-size based scaling factors for the fracture criteria, which
in our case is the fracture strain. This regularization method is different than
*MAT_NONLOCAL [77]. The progression of each force-displacement curves is compared
to the experimental results aiming at a perfect correlation between both and the difference

126
between engineering fracture strains is minimized. Damage based regularization for the post-
critical range is presented with the following mesh sizes for materials shown in table 5.4:
Material Mesh sizes
For 2024-T351 2mm, 1mm, 0.5mm & 0.25mm
For 1045 steel 2mm, 1mm, 0.5mm & 0.25mm
For DP600 steel 1mm, 0.5mm, 0.2mm, 0.1mm

Table 5.4 Mesh sizes for regularization of tensile tests using different materials

For cost-effectiveness, big simulations like the full-scale car crash simulations have to be done
using coarse elements. Also, to save the simulation time coarse elements are used in locations
where there is not failure or stress concentration and higher resolution of nodal quantities is
not necessary. This variation in mesh size in the component might affect the results if the
discretization is not efficient. For the range of mesh size variation, regularization can correct
this problem with proper selection of mesh size dependent regularization factors for
equivalent plastic strain to failure defined in the fracture locus. However, for all the
simulations an optimum mesh size is selected which would minimize the effect of mesh size
dependency. This also eliminates the need for convergence study and regularization for
different mesh sizes. However, the regularization presented in the next few sections aims to
demonstrate the effect mesh size can have on the post critical damage behavior of the
deformation until fracture.

5.3.1 2024-T351 aluminum alloy un-notched round specimen tensile test:

The uniaxial tensile test with un-notched specimen is chosen for regularization study as
they are simple yet large enough to accommodate the mesh size variation from 0.25 mm
to 2 mm. For the un-notched tensile specimen presented in section 4.1.1 a) figure 5.26
shows differently discretized models:

127
 0.25 mm

Figure 5.28: Differently discretized models for 2024-T351 un-notched tensile test
specimen

Apart from the mesh, exactly same keywords and parameters were defined for each of
these simulations. Without regularization, the results for each of these simulations were
compared for difference in the results. Figure 5.27 shows the force-displacement plots
for each of these simulations without regularization. It is evident that mesh size has a
huge influence on the results in the post-critical damage region.

128
 2024-T351 Force-displacement without
regularization
35
30
25 Experiment
Force (kN)

20 2 mm mesh
15
1 mm mesh
10
0.5 mm mesh
5
0 0.25 mm mesh
0 2 4 6
Displacement (mm)

Figure 5.29: Force-displacement plots for 2024-T351 un-notched tensile specimen

simulated with different mesh sizes without regularization.

The fracture strain is different in each of the simulations due to different mesh size,
although the rest of the parameters remain the same. Figure 5.28 shows the plot of mesh
size vs fracture strain or equivalent plastic strain to fracture. The fracture strain
monotonously increases with the mesh size, except for the mesh sizes 1 mm and 2 mm,
which fall in the asymptotic region. This means, coarse mesh over-predicts the ductility
of the material in general and refinement is required.

2024-T351 Fracture strain vs mesh size

Equivalent plastic strain to fracture

0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0 0.5 1 1.5 2 2.5

Element size (mm)

Figure 5.30: Fracture strain vs mesh size plot for 2024-T351 un-notched tensile
specimen
129
 To counter this effect and to get uniform results as far as possible, suitable ad hoc mesh-
size dependent regularization factors were defined which scale the equivalent plastic
strain to fracture suitably. To come up with these factors might need some prior
simulation experience and some trial and error. Figure 5.29 shows the force-
displacement plots that were obtained with regularization.

2024-T351 Post Regularization Force-Displacement

35
Experiment
30
25 2 mm mesh
20
Force (kN)

1 mm mesh
15
0.5 mm mesh
10
0.25 mm
5
mesh
0
0 1 2 3 4 5 6
Displacement (mm)

Figure 5.31: Force-displacement plots for different mesh sizes of 2024-T351 un-
notched tensile specimens after regularization

It is observed that selection of 0.5 mm mesh size produces satisfactory results for a
specimen of this size even without regularization. Choosing the appropriate mesh size
is also important from the point of view of reasonable computing time and justified use
of disk space. Thus, regularization helps in minimizing the effect of mesh size in the
post-critical damage region but it should be used after at least two different mesh sizes
have been tested for difference in results. Sometimes, defining inappropriate
regularization factors might completely change the results which is not desirable.

5.3.2 1045 steel un-notched round specimen tensile test:

130
For 1045 steel we again choose the un-notched tensile specimen for regularization study
because of the same reasons mentioned in the previous section. Because of the size of
this specimen is almost similar to the previous specimen discussed, the same mesh sizes
are chosen to carry out regularization study. Figure 5.30 shows the differently
discretized meshes for the same un-notched tensile specimen.

Figure 5.32: Differently discretized meshes for 1045 steel un-notched tensile test
specimen

131
Same set of keywords and parameters of the input file were defined for each of the
simulations with the above meshes. Without regularization the force-displacement
results for each of the runs were plotted and compared by super-imposing the plots.
Figure 5.31 shows the superimposed plots without regularization.

1045 Steel Force-Displacement without

regularization
70
60
50 Experiment
40 2 mm mesh
Force (kN)

30
1 mm mesh
20
0.5 mm mesh
10
0.25 mm mesh
0
0 1 2 3
Displacement (mm)

Figure 5.33: Force displacement plots for 1045 steel un-notched tensile specimens
without regularization

The higher mesh size causes delayed fracture or increases ductility. In the case of 1045
steel un-notched tensile specimen, the simulation having highest mesh size of 2 mm
doesn’t fail at all in the defined displacement range. This means that the failure strain is
even higher than the maximum failure strain reached. As the mesh size reduces, the
failure strain also reduces thereby causing early failure as seen for the 0.25 and 0.5 mm
meshes. The fracture strain vs mesh size plot is presented in figure 5.32. As previously
noted, fracture stain tends to increase with the mesh size.

132
 Fracture strain vs mesh size for 1045 steel
0.4

Effective plastic strain to fracture

0.35
0.3
EpsilonP (GPa) 0.25
0.2
0.15
0.1
0.05
0
0 0.5 1 1.5 2 2.5
Element size (mm)

Figure 5.34: Fracture strain vs mesh size for 1045 steel un-notched tensile specimen

After defining mesh-size dependent regularization factors, the simulations were run
again and the force-displacement results obtained are shown in figure 5.33

1045 Steel Force-Displacement post

regularization
60
Experiment
50
40 2 mm mesh
Force (kN)

30 1 mm mesh
20
0.5 mm mesh
10
0 0.25 mm mesh
0 1 2 3
Displacement (mm)

Figure 5.35: Force-displacement plots for different element size 1045 steel un-notched
tensile specimen test simulations

The quality of regularization can be improved by better selection of regularization

factors, but for that more trial simulations with different mesh sizes are necessary. It is

133
 noted that simulations with 0.5 mm mesh size produces optimum results with reasonable
computer time.

5.3.3 DP600 steel un-notched flat specimen tensile test:

The un-notched flat specimen is chosen for mesh size based convergence and
regularization study because out of the experiments presented in the literature, it has the
maximum fracture strain. This is helpful in studying the variation of fracture strain with
element size. Since this specimen is smaller than the previous two specimens discussed,
the range of mesh size for regularization study is selected to be from 1 mm to 0.1 mm.
Figure 5.34 shows the differently discretized un-notched flat tensile specimen meshes:

1 mm 0.5 mm 0.2 mm 0.1 mm

Figure 5.36: Differently discretized DP600 un-notched flat tensile specimen

134
The normalized force vs normalized displacement plot obtained from simulations
without regularization for each of these meshes with the same input keywords and
parameters is shown in figure 5.35. As is previously established, the ductility of the
material is found to increase with the mesh size. The biggest mesh size model with 2 mm
element size is found to have higher fracture strain than is reached with the applied
displacement boundary condition so it does not fail when simulated. As the mesh size is
reduced, the failure is reached earlier.

DP600 Force-displacement without

regularization
Normalized Force F/Ao (kN/mm2

0.7
0.6
0.5 Experiment
0.4 1 mm mesh
0.3 0.5 mm mesh
0.2
0.2 mm mesh
0.1
0 0.1 mm mesh
0 0.2 0.4 0.6
Normalized Displacement L/Lo

Figure 5.37: Normalized force vs displacement plots for different mesh size models for
DP600 un-notched tensile flat specimens

Further, figure 5.36 shows the fracture strain vs mesh size plot for the DP600 un-notched
tensile specimen.

135
 DP600 fracture strain vs mesh size
1.2
1
Fracture strain 0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2
Mesh size

Figure 5.38: Fracture strain variation with mesh size for DP600 un-notched tensile flat
specimen test

After defining suitable regularization factors, the super-imposed force-displacement

plates for each of the element sizes is plotted as shown to study the effect of regularization
as shown in figure 5.37.

Normalized force-displacement for DP600 un-

notched tensile flat specimen after
regularization
0.7
0.6
Normalized force F/Ao

0.5
Experiment
0.4 1 mm mesh
0.3 0.5 mm mesh
0.2 0.2 mm mesh
0.1 0.1 mm mesh
0
0 0.2 0.4 0.6
Normalized displacement L/Lo

Figure 5.39: Normalized force-displacement plots superimposed for different mesh

sizes of DP600 un-notched tensile flat specimen tests after regularization

136
Overall, the DP600 specimens in this chapter show inferior match when compared to 2024-T351
aluminum alloy and 1045 steel specimen results from previous chapter. The unavailability of exact
material properties and use of approximated fracture locus may be the prime reasons for this
difference. Still, the DP600 specimen simulations highlight the importance of having the exact
data to reproduce the experimental results closely. The conclusions from all the previous
simulations up till now will be presented in the next chapter along with the comments on the
advantages, limitations and future scope for improved modeling of ductile fracture.

137
Chapter 6

Conclusions and future work

In the present thesis, the influence of stress state dependent modeling on the ductile fracture is
studied. The previous damage modeling approaches are briefly described and the concept of stress
state in cylindrical coordinate system and principal stress space is analyzed. The experiments
previously validated with stress state dependent constitutive models were re-validated using
*MAT_024 and GISSMO (*MAT_ADD_EROSION) from the LS dyna library. The input data for
the material like its properties, yield curve and fracture locus was obtained from the documented
references. Since we did not perform any experiments, the fracture locus was directly reproduced
using the fitted parameters from past experiments and analytical expression from the literature.
The accuracy of this reproduced fracture locus depends on the quality of approximation of
analytical expression and fitted parameters and is bound to small deviation from the actual fracture
locus obtained by experiments. This implementation is tested for different specimen fracture and
the plots of force versus displacement are compared to a satisfactory match.

The stress state achieved in the simulations also correspond to the experiments and simple cases
like uniaxial axisymmetric tension, pure shear and axisymmetric compression were supported by

138
analytical calculations directly from the stress tensor. The transition of crack from flat to slant due
to damage induced material deterioration and lode parameter dependence is also captured well like
the case of 2021-T351 grooved flat plate. The resulting formation of shear bands with zero lode
parameter and faster damage accumulation in shear bands having localized deformation is studied
to be in conformation to the actual phenomenon. In addition, the initiation of fracture is correctly
predicted as in the case of 3 point bending specimens where the fracture in tensile zone occurs
before the compression. The fracture mode transition in projectile impact experiments from
petaling to plugging with increased plate thickness is well captured by the simulations. The
increased fracture strain in upsetting specimens and the diagonally opposite fracture initiation is
also simulated by the present implementation. The reduced ductility in the case of pure shear
experiments like torsion compared to the tensile ones in 1045 steel further establishes the effect of
stress state.

The materials investigated are found not to have any significant stress state dependency as far as
plasticity modeling is concerned. Satisfactory results are obtained with just failure strain
dependency on the stress state. Both the lode parameter and triaxiality play an almost equally
important role in predicting the strain required to fracture although steels show lower lode
parameter dependence than aluminum alloys in general. For 2024-T351 experiments, Xue’s [3]
symmetric fracture locus over-predicted the fracture strains for most experiments compared to
Bai’s [4] asymmetric fracture surface except for the upsetting tests. The damage exponent
parameter, critical plastic strain and fading exponent are optimized by trial and error with prior
simulation experience.

The results obtained with GISSMO have been found to have some key advantages over the other
complex material models even for the same degree of accuracy. The following are the advantages
of the ductile fracture modeling approach adopted in this thesis:

139
6.1. Advantages of presented J2 plasticity and GISSMO implementation:

1. Experimentally calibrated data can be directly input in the form of lode curves for the
yielding behavior as well as fracture locus. This makes it simple to apply this modeling
method to a variety of different kinds of materials.
2. GISSMO simulations give better control over the results beyond the critical damage
region. The control offered by parameters like fading exponent FADE, critical damage
DCRIT, critical plastic strain ECRIT, stress state dependent failure strain LCSDG,
damage exponent DMGEXP in combination with regularization factors can achieve
fairly accurate post failure results.
3. Correct mode of failure is predicted. For example, the transition of crack from straight to
slant or from normal to shear strain failure can be modeled as in the case of 2024-T351
grooved flat plate specimen discussed in section 4.1.3. Or the higher probability of failure
occurring in tension rather than in compression is well captured in the 3-point bending
specimens discussed in section 4.1.4. Another example would be the predicted
bending/spalling failure in thin plates and shear/plugging failure in thick plates on
projectile impact in section 4.1.5. Also, the prediction of initiation of crack at the
diagonally opposite corners rather than center of the upsetting specimens in section 4.1.2
corresponds to the experimental results. Finally, the correct prediction of early failure
due to reduced fracture strain for 1045 steel torsion specimen in section 4.2.3 a
4. Inclusion of GISSMO gives a smooth realistic drop in stresses when material damage is
encountered rather than an uncontrolled abrupt failure predicted by a fixed failure strain
input.
5. The failure surface is stress state dependent. This gives ability to handle different failure
conditions affected by the stress state and make sense of different failure strains
associated with informed selection of appropriate failure strain. For the simple fixed
failure strain model, it becomes difficult to fix the exact equivalent plastic strain at which
the specimen would fail. Also, this fixed failure strain would need separate definition for
each case of stress state.
6. GISSMO is has the flexibility to be coupled with any suitable material model.

140
Consideration of the stress state effect and damage certainly helps improve the fracture results.
However, apart from the additional experimental data and computational effort required, there are
also some limitations to this approach such as:

6.2. Limitations of presented J2 plasticity and GISSMO implementation:

1. The J2 plasticity model does not consider stress state effect on plastic flow although it
is found to be insignificant.
2. The stress state changes with the deformation from the initial loading condition. This
also causes change in ductility of the material which needs to be accounted for.
3. This modeling approach causes localized stresses. Which means, it causes the stress
concentrated elements to affect the behavior of neighboring elements which might fail
immediately in a chain reaction, leading to complete failure of the specimen. Hence,
regularization methods like the one using LCREGD keyword presented in section 5.2
and non-local methods [77] should be adopted.
4. The von-mises plasticity model used is a simplification of the plastic flow. For
example, in the documentation of the experiments, slight anisotropy for the 2024-T351
and DP600 materials was presented which might lead to slight variation in results.
5. The fracture locus used in this thesis was not obtained from experimental results, but
from previously proposed analytical formulation and fitted parameters. Since no actual
experiments were performed, fracture locus is approximate.
6. Some DP600 experiment simulations show slight deviation in results even in the plastic
flow region which means the input data for constitutive models was not very accurate.

141
6.3. Future scope

1. Complex material models addressing anisotropy, strain history, strain rate, hardening,
stress state and material weakening can be used in place of von-mises although care
must be taken to maintain feasibility in the additional computational and experimental
effort.
2. The weighted stress state parameters take the average values over the range of
deformation and stress concentration. To have realistic values of fracture strain, actual
stress state parameters could be analyzed by complex experiments like the digital image
correlation (DIC).
3. Local material models which inherently calculate effect of mesh size on the fracture
strain can be used instead of separate regularization treatment.
4. Accurate input parameters like the material properties, yield curve and fracture locus
could be employed from actual experimental tests rather than reference literature and
approximate analytical formulations.
5. User defined constitutive and failure model can be used instead of library material
models for better more flexible analysis.
6. In this thesis the effect of stress state on the material plasticity is assumed negligible
since the results do not show any significant error. This is especially true for the steels
than aluminum alloys. The plasticity can also be explored for stress state dependency.
7. A variety of additional experiments to cover different complex stress states like
butterfly specimen, combined torsion-tension experiment, biaxial-stress experiment,
etc. could help in better fracture locus fitting, especially in the negative range of
triaxiality and lode parameter.

142
Bibliography:

[1] Björklund O. (2008): Modelling of failure. Master’s thesis, Division of Solid Mechanics
Linköpings University LIU-IEI-TEK-A--08/00381—SE

[2] Wierzbicki T., Bao Y., Lee Y.-W., Bai Y., (2005), Calibration and evaluation of seven fracture
models. International Journal of Mechanical Sciences 47 pp. 719-743.

[3] Xue L. (2007): Ductile Fracture Modeling - Theory, Experimental Investigation and
Numerical Verification. Phd. Dissertation, Massachusetts Institute of Technology

[4] Bai Y. (2008): Effect of loading history on necking and fracture. Phd. Dissertation,
Massachusetts Institute of Technology

[5] Basaran M. (2011): Stress state dependent damage modeling with a focus on lode angle
dependence. PhD. Dissertation, Aachen University, Germany ISBN 978-3-8440-044-5

[6] De Borst R., Sluys L. J., Mühlhaus H. –B., Pamin J. (1993): Fundamental issues in finite
element analysis of localization and deformation. Engineering Computations, Vol. 10 lss 2 pp
99-21.

[7] Chen G., Ming X., Ming F. Shi, Kalpundi G., Wehner T., Yarlagadda R. (2004): Material
and processing modeling of dual phase steel front rails for crash. Proceedings of International
Conference on Advanced High Strength Sheet Steels for Automotive Applications, pp 161-170

[8] Rios, P.R., Guimaräes, J.R.C., and Chawla, K.K., (1981): Modelling the stress-strain curves
of dual phase steels, Scripta Mat., 15(8), pp 899-904

143
[9] Du Bois P., Buyuk M., He J., Kan S. (2010): Development, implementation and validation of
3D failure model for aluminum 2024 for high speed impact applications. 9th LS Dyna forum,
October 12-13, Bamberg, Germany.

[10] Bathe K. J. (1996): Finite Element Procedures New Jersey Prentice Hall

[11] Mohr O. (1906): Abhandlungen aus dem Gebiete der technischen Mechanik; mit zahlreichen
Textabbildungen. Wilhelm Ernst & Sohn

[12] Johnson G. R. and Cook W. H. (1985): Fracture characteristics of three metals subjected to
various strains, strain rates, temperatures and pressures. Eng. Fracture Mechanics 21, pp 31-
48

[13] Du Bois P., Bayuk M., He J., Kan S. (2010): Development, Implementation and Validation
of 3-D failure model for Aluminum 2024 for high speed impact applications. 11th
International LS dyna users conference, June 6-8, Bamberg, Germany

[14] Schwer L. E. and Murray Y. D. (2002): Continuous surface cap model for Geomaterial
modeling: A New LS-dyna material type. 7th International LS dyna user’s conference

[15] Lee Y.W. (2004): Fracture prediction in metal sheets. Ph.D. thesis, Cambridge, MA:
Department of Ocean Engineering, Massachusetts Institute of Technology.

[16] Drucker D.C. and Prager W. (1952): Soil mechanics and plastic analysis or limit design. Q.
Appl. Math., 10(2) pp 157–165.

[17] http://en.wikipedia.org/wiki/Image:Yield surfaces.png (2008-02-28)

[18] LS-Dyna Keyword User's Manual Vol II: Material Models (May 2014): Livermore Software
Technology Corporation (LSTC) revision 5442

144
[19] Bao Y. (2003): Prediction of ductile crack formation in uncracked bodies. Ph.D. thesis.
Cambridge, MA: Department of Ocean Engineering, Massachusetts Institute of Technology

[20] Cockcroft M. G., Latham D. J., (1968): Ductility and the workability of metals. Journal of the
institute of metals 96 pp 33-39

[21] Heung N.H., Keun-Hwan K., (2003): A ductile fracture criterion in sheet metal forming
process. Journal of Materials Processing Technology 142 pp 231-238

[22] Takada K, Sato K and Ma N.(2012): Fracture prediction of high strength steels with ductile
fracture criterion and strain dependent model of anisotropy. 12th International LS dyna user’s
conference, Detriot USA

[23] Gese H., Oberhofer G., Dell H. (2007): Consistent modelling of plasticity and failure in the
process chain of deep drawing and crash with user material model MF-GenYld + CrachFEM
for LS dyna. 6. LS dyna Anwenderforum, Frankenthal

[24] Chauffray M., Delattre G. and Guerin L. (2011): Prediction of failure on high strength steel
in seat mechanisms simulation. 8th European LS dyna user’s conference, Strasbourg.

[25] Bresan J. D and Williams J. A., (1983): The use of shear instability criterion to predict local
necking in sheet metal deformation. International Journal of Mechanical Sciences, 25 pp 155-
168.

[26] Neukamm F., Feucht M., and Haufe A. (2009): Considering damage history in
crashworthiness simulations. 7th European LS-dyna conference, Salzburg, Austria.

[27] Neukamm F., Feucht M., and Haufe (2008) : Consistent damage modelling in the process
chain of forming to crashworthiness simulations. 7th German LS dyna conference, Bamberg,
Germany, volume 30.

145
[28] Carney K.S., DuBois P.A., Buyuk M. and Kan S. (2009): Generalized, Three-Dimensional
Definition, Description, and Derived Limits of the Triaxial Failure of Metals. Journal of
Aerospace Engineering, Vol. 22.

[29] Seidt J. D. (2010): Plastic deformation and ductile fracture of 2024-T351 aluminum under
various loading conditions. PhD dissertation , Ohio State University

[30] Gurson A. L., (1977): Continuum theory of ductile rupture by void nucleation and growth:
Part I – Yield criteria and flow rules for porus ductile media. Journal of Engineering Materials
and Technology 99(1) pp 2-15.

[31] Needleman A., Tvergaard V., (1984): An analysis of ductile rupture in notched bars. Journal
of Mechanical Physics of Solids 32 pp 461-490.

[32] McClintock. F. A., (1968): A criterion for ductile fracture by the growth of holes. Journal of
applied mechanics, 35(2) pg 363-371.

[33] Rice J. R. and Tracey. D.M. (1969): On the ductile enlargement of voids in triaxial stress
fields. Journal of mechanics of physical solids, 17(3) pp 201-217.

[34] Haufe A., Neukamm F., Feucht M., DuBois P. and Borvall T., (2010): Recent developments
in damage and failure modeling with LS-Dyna. Nordic LS dyna user’s forum.

[35] Nahshon K., Hutchinson J.W. (2008): Modification of the Gurson model for shear failure.
Eur. J. Mech. A. Solids, 27(1) pp1-17.

[36] Gologanu M.,.Leblond J.-B and Devaux J. (1993): Approximate models for ductile metals
containing non-spherical voids – case of axisymmetric prolate ellipsoidal cavities. J. Mech.
Phys. Solids. 41(11) pp 1723-1754

[37] Pardoen T. and Hutchinson J.W. (2000): An extended model for void growth and coalescence.
J. Mech. Phys. Solids, 48(12) pp2467-2512.

146
[38] Lassance D., Fabr´egue D., Delannay F. and Pardoen T. (2007): Micromechanics of room and
high temperature fracture in 6xxx alloys. Prog. Mater Sci., 52(1) pp 62-129.

[39] Feucht M., Dong-Zhi S., Erhart T. and Frank T. (2006): Recent development and applications
of the Gurson model. 5. Ls dyna Anwenderforum, Ulm.

[40] Ockewitz A., Dong-Zhi S., (2006): Damage modelling of automobile components of
aluminium materials under crash loading. 5. LS-Dyna Anwenderforum, Ulm.

[41] Lemaitre J., Chaboche J.-L., (1990): Mechanics of solid materials. Cambridge: Cambridge
university press.

[42] Poizat C., Campagne L., Daridon L., Ahzi S., Husson C. & Merle L. (2005): Modeling and
simulation of thin sheet blanking using damage and rupture criteria. International Journal of
Forming Prosses 8 pp. 29-47.

[43] Johnson G.R. and Cook W.H. (1983): A constitutive model and date for metals subjected to
large strains, high strain rates and high temperatures. Seventh International Symposium on
Ballistics, Hague

[44] Lesuer D.R., Kay G.J. and LeBlanc M.M. (2001): Modeling large-strain, high rate
deformation in metals. Lawrence Livermore National Laboratory, L-342, Livermore, CA
94551.

[45] Wilkins M.L., Streit R.D. and Reaugh J.E. (1980): Cumulative-strain-damage model of
ductile fracture: Simulation and prediction of engineering fracture tests. Lawrence Livermore
National Lab., CA (USA); Scinece Applications Inc., San Leandro, CA (USA) Technical
report, UCRL-53058.

[46] Nguyen N.T., Kim D.Y., Song J.H., Kim K.H., Lee I.H. and. Kim H.Y (2012): Numerical
prediction of various failure modes in spot welded metals. International Journal of
Automotive Technology, Vol. 13, No. 3, pp 459-467

147
[47] Lemaitre J. (1985): A continuous damage mechanics model for ductile fracture. Journal of
engineering material technology. 107:83-9.

[48] Kachanov M. (1999): Rupture time under creep conditions. International Journal of
Fracture. 97 pp 11-18.

[49] Williams K. V., Vaziri R. and Poursartip A. (2002): A physically based continuum damage
mechanics model for thin laminated composite structures. International journal of solids and
structures. 40 pp 2267-2230.

[50] Xiao X. (2008): Simulation of composite tubes axial impact with damage mechanics based
composite material model. 10th Intenrational LS-dyna user’s conference.

[51] Williams K. V. and Vaziri R. (2000): Application of a damage mechanics model for predicting
the impact response of composite materials. Computers and structures. 79 pp 997-1011.

[52] Teng X. (2008): Numerical prediction of slant fracture with continuum damage mechanics.
Engineering fracture mechanics 75 pp 2020-2041.

[53] Børvik T., Hopperstad O.S., Berstad T., Langseth M. (2001): A computational model of
viscoplasticity and ductile damage for impact and penetration. Structural impact laboratory
(SIMlab), Department of structural engineering, Norwegian university of science and
technology. N-7491, Trondheim, Norway.

[54] Brannon R., Fossum A. and Strack O. (2009): “Kayenta: theory and user’s guide. Tech. Rep.
SAND2009-2282, USDOE.

[55] Zeigler H. (1959): A modification of Prager’s hardening rule. Quart. Appl. Math. 17(1), pp
55-65.

[56] Chaboche J. L. and Rousselier G. (1983): On the plastic and viscoplastic constitutive
equations – Part I: Rules developed with internal variable concept. Journal of Pressure Vessel
Techonlogy.105(2), pp 153-158.

148
[57] Chung K., Lee M. –G., Kim D., Kim C., Wenner M. L. and Barlat F. (2005): Spring-back
evaluation of automotive sheets based on isotropic-kinematic hardening laws and
nonquadratic anisotropic yield function: Part I: Theory and formulation. International journal
of plastics. 21(5): pp 861-882.

[58] Geng L. and Wagoner R. H. (2002): Role of plastic anisotropy and its evolution on
springback. International journal of mechanical sciences. 44(1), pp 123-148.

[59] Bridgeman P. W. (1952): Studies in large plastic flow and fracture. McGraw-Hill, New York.

[60] Lewandowski J. J. and Lowhaphandu P. (1998): Effects of hydrostatic pressure on mechanical

behavior and deformation processing of materials. International Materials Reviews, 43(4): pp
145-187.

[61] French I. E. and Weinrich P.F. (1975): The influence of hydrostatic pressure on the tensile
deformation and fracture of copper. Metallurgical and materials transactions A. 6(4): pp 785-
790.

[62] Brownrigg A., Spitzig W. A., Richmond O., Teirlinck D. and Embury J. D. (1983): The
influence of hydrostatic pressure on the flow stress and ductility of a spherodized 1045 steel.
Acta Metalluigica 31(8) pp 1141-1150.

[63] Pugh H. Ll. D., Hodgson G. and Gunn D. A. (1963): Tensile strain measurement under high
hydrostatic pressure using an optical method. Journal of scientific instruments, 40 pp 221-
224.

[64] Mae H., Teng X., Bai Y. and Wierzbicki T. (2009): Correlation between tensile/shear fracture
strains and pore sizes in a cast aluminum alloy. The Japanese society for experimental
mechanics, 9(2) pp 129-135.

[65] Teng X., Mae H., Bai Y. and Wierzbicki T. (2008): Statistical analysis of ductile fracture
properties of an aluminum casting. Engineering fracture mechanics, 75(15) pp 4610-4625.

149
[66] Clausing D. P. (1970): Effect of plastic strain state on ductility and toughness. International
journal of fracture mechanics, 6(1) pp 71-85

[67] Swan M. S. (2012): Incorporation of a general strain-to-failure fracture criterion into a stress-
based plasticity model through a time-to-failure softening mechanism. Master’s thesis,
University of Utah. Publishing No. 1506806.

[68] Neukamm F., Feucht M. and Bischoff M. (2009): On the application of continuum damage
models of sheet metal forming simulations. Proceedings, X international conference on
computational plasticity. CIMNE, Barcelona, Spain.

[69] Pan F., Zhu J., Helminen A. O., Vatanparast R. (2006): Three point bending analysis of a
mobile phone using LS dyna explicit integration method. 9th International LS dyna user’s
conference.

[70] ASM material data sheet: http://www.aerospacemetals.com/contact-aerospace-metals.html

[71] Adams C. M. Jr. (1967): Effective ductility in castings and weldments. American society for
metals, Metals park, Ohio. In Ductility, chapter 6, pp 179-197.

[72] Effelsberg J., Haufe A., Feucht M., Neukamm F., Du Bois P. (2012): On parameter
identification for the GISSMO damage model. 12th International LS dyna users conference,
Detriot, Michigan, USA.

[73] Tsuchida T., Yamamoto S. and Isomura K. (2003): The application of damage and fracture
material model to crashworthiness evaluations for aluminum cars. 4th European LS dyna
user’s conference, Ulm, Germany.

[74] Chelluru S. K. (2007): Finite element simulations of ballistic impact on metal and composite
plates. Master’s thesis, Wichita State University.

150
[75] Carlucci D. E. and Jacobson S. S. (2008): Ballistics: Theory and design of guns and
ammunition. CRC Press p. 310, ISBN 978-1-4200-66180.

[76] Padmanabhan R., Sung J., Lim H., Oliviera M. C., Menzes L.F. and Wagoner R.H. (2008):
Influence of draw restraining force on the springback in advanced high strength steels.

151