Elasto-Plastic and Damage Modeling of Reinforced Concrete
Elasto-Plastic and Damage Modeling of Reinforced Concrete
Elasto-Plastic and Damage Modeling of Reinforced Concrete
CONCRETE
A Dissertation
in
by
Ziad N. Taqieddin
B.S., Applied Science University, Amman, Jordan, 2001
M.S., Louisiana State University, Baton Rouge, Louisiana, 2005
August 2008
TO THOSE WHO TAUGHT ME EVERYTHING THAT
MATTERS
ii
ACKNOWLEDGEMENTS
I would like to gratefully and sincerely thank Dr. George Z. Voyiadjis for his
impeccable guidance, knowledge, understanding, patience, and most importantly, his
friendship during my graduate studies at LSU. His mentorship was paramount in
providing me with a well rounded experience throughout the years. Prof. Voyiadjis
encouraged me not only to grow as a scientist but also as an instructor and an
independent thinker. His enthusiasm and unlimited zeal have been major driving forces
throughout my graduate career. For everything youve done to help me, Dr. Voyiadjis, I
cant thank you enough.
I would also like to thank all of the members of the advanced Computational Solid
Mechanics (CSM) Laboratory. Through five years of Masters and PhD, I developed a
very strong bond of friendship with a dozen of the CSM Tigers (as we like to call
ourselves). These friends were from all over the world yet they were my family in Baton
Rouge. I attended the graduation of many of them, just as many of them will attend the
graduation of mine. My experience with them is one of the memories that I will hold dear
to my heart forever. To all of you CSM Tigers, thank you. For you, my dear friends, have
provided much needed humor and entertainment in what could have otherwise been a
stressful and boring concentrated research environment.
My parents: Dr. Noureddin and Mona. Thank you for being the infinite source of
love and support that enlightened my path from six thousand miles away. Thank
you for always believing in me and for encouraging me to be strong and focused
in the times I needed encouragement most. Thank you for welcoming my dreams
and making sure they become true throughout the years.
My brother and sisters. Thank you for all the joy that you bring into my life.
Thank you for sending me pictures of your new born babies and for allowing me
to be a part of their daily life as an uncle even though Im distantly far away.
Thank you for being the wonderful people that you are.
My wife and best friend, Rawan. Your support, encouragement, quiet patience
and unwavering love were undeniably the bedrock upon which the past two years
of my life have been built. Your sprit lifting and cheering smile conquered all my
fears. Your tolerance of my occasional stress-outbursts is a testament in itself of
your unyielding love and devotion. You are the reason why my glass is always
half full, and your happiness and welfare will be the primary goal in my life.
iii
TABLE OF CONTENTS
DEDICATION . . . . . . . . . ii
ACKNOWLEDGMENTS . . . . . . . . iii
ABSTRACT . . . . . . . . . . vii
CHAPTER 1: INTRODUCTION . . . . . . . 1
1.1 Scope and Objectives . . . . . . . . 3
1.2 Outline of the Dissertation . . . . . . . 4
iv
CHAPTER 4: CONCRETE MATERIAL MODEL . . . . 76
4.1 Introduction . . . . . . . . . 76
4.2 Framework for the Elastic-Plastic-Damage Model . . . . 78
4.3 Consistent Thermodynamic Formulation . . . . . . 80
4.4 The Helmholtz Free Energy Function . . . . . . 84
4.4.1 The Elastic/Damage Part of the Helmholtz Free Energy Function . . 84
4.4.2 The Plastic Part of the Helmholtz Free Energy Function . . . 89
4.5 Plasticity Formulation . . . . . . . . 91
4.5.1 Plasticity Yield Surface and Hardening Functions . . . . 91
4.5.2 Plasticity Non-associative Flow Rule and Consistency Condition . . 93
4.6 Damage Formulation . . . . . . . . 94
4.6.1 Tensile and Compressive Damage Surfaces and Hardening Functions . 94
4.6.2 Damage Consistency Conditions . . . . . . 96
4.7 Numerical Integration of the Constitutive Model . . . . . 96
4.7.1 The Effective (Undamaged) Elastic-Plastic Steps . . . . 97
4.7.1.1 The Elastic Predictor . . . . . . . 97
4.7.1.2 The Plastic Corrector . . . . . . . 98
4.7.1.3 The Effective Configuration Integration Algorithm . . . 103
4.7.2 The Degradation (Damage) Step . . . . . . 107
4.7.2.1 The Damage - Corrector Step . . . . . . 107
4.8 The Consistent Elastic Plastic Damage Tangent Operator . . . 109
4.9 Implementation and Verification of the Integration Scheme . . . 114
4.9.1 Identification of the Proposed Models Parameters . . . . 114
4.9.2 Monotonic Uniaxial Tensile Test . . . . . . 115
4.9.3 Monotonic Uniaxial Compressive Test . . . . . 117
4.9.4 Monotonic Biaxial Tests . . . . . . . 119
4.9.5 Three Point Bending Test of a Notched Beam . . . . 126
BIBLIOGRAPHY . . . . . . . . . 148
v
APPENDIX B: UPDATING THE EFFECTIVE STRESS USING THE
RETURN MAPPING EQUATION (CHAPTER 4) . . 165
VITA . . . . . . . . . . . 166
vi
ABSTRACT
Modeling the mechanical behavior of Reinforced Concrete (RC) is still one of the
most difficult challenges in the field of structural engineering. The Nonlinear Finite
Element Analysis (NFEA) and modeling of the behavior of RC members are the primary
goals of this study. The macroscopic components of RC, Concrete material and
reinforcing steel, are represented herein by separate material models. These material
models are combined together using a model that describes the global effect of
interaction between reinforcing steel and concrete in order to simulate the behavior of the
composite RC material.
The Numerical implementation and application are important parts of this study. A
suitable elastoplasticity-implicit/damage-explicit scheme is adapted here for the
integration of the incremental constitutive equations. The elastic-predictor, plastic-
corrector and damage-corrector steps are used to facilitate the integration procedure. The
constitutive approach is implemented, through numerical algorithms; in the advanced FE
software ABAQUS via user defined material subroutine UMAT to analyze and better
describe the overall behavior of such a composite material. Concrete and RC beams
subjected to static-short-term-monotonic loading are analyzed in an assumed isothermal
environment. The simulated results are compared to experimental studies conducted by
other researchers.
vii
CHAPTER 1
INTRODUCTION
Reinforced Concrete (RC) is one of the most commonly used building materials
nowadays. It is a composite material made of plain concrete, which possesses relatively
high compressive strength but low tensile strength, and steel bars embedded in the
concrete, which can provide the needed strength in tension. The economy, efficiency,
strength and stiffness of RC make it an attractive material for a wide range of structural
engineering applications, such as nuclear power-plants, bridges, cooling towers and
offshore platforms. For RC to be used as a structural material, it should satisfy special
criteria including:
Strength and Stiffness
Safety and Appearance
Economy
By applying the principles of structural analysis, the laws of equilibrium and the
consideration of the mechanical properties of the components studied; RC design
procedure should yield a sufficient margin of safety against collapse under ultimate loads.
Serviceability analysis is conducted to control the deflections under service loads and to
limit the crack width to an acceptable level for the structural component to perform and
appear safe and inhabitable for the human eye. Economical considerations are satisfied
by optimizing the usage of steel/concrete quantities to account for the difference in unit
costs of steel and concrete.
The ultimate objective of design is the safety and economy of the RC structural
member. The design process is usually based on a linear elastic analysis to calculate the
internal forces in the member which are then used to design the reinforcement and the
details of the member using some code provision. Codes are usually based on empirical
approaches that utilize experimental data and provide design rules to satisfy safety and
serviceability requirements. Although the design of RC structures based on linear-elastic
stress analysis is adequate and reliable in many cases, the extent and impact of a disaster
in terms of human and economical losses in the event of structural failure of large scale
modern structures necessitate more careful and detailed structural safety analysis. Thus,
Nonlinear Finite Element Analysis (NFEA) is often required to obtain detailed
information regarding the ultimate loading capacity and the post-failure behavior of RC
structures. The importance and interaction of different nonlinear effects on the response
of RC structures can be studied analytically using NFEA.
The complex behavior of concrete, which arises from the composite nature of the
material, is characterized by a reduction of the load carrying capacity with increasing
deformations after reaching a certain limit load. This global behavior is usually caused by
a material behavior which is described as strain softening and occurs in tension and in
compression. This necessitates the development of appropriate constitutive models to
describe such behavior.
1
In RC, the response of the structure is even more complicated. In general a number of
cracks will develop in the structure due to the bond action between concrete and
reinforcement. This results in a redistribution of the tensile loads from concrete to the
reinforcement. This phenomenon is called tension-stiffening, because the response is
stiffer than the response with a brittle fracture approach.
A large variety of models have been proposed to characterize the stress-strain relation
and failure behavior of RC materials. All these models have certain inherent advantages
and disadvantages which depend to a large degree on their particular application and
complexity. Macroscopic constitutive studies have been conducted with different levels
of complexity and applicability in order to address the different aspects of the concrete
material behavior. On the other hand, microscopic modeling and multi-scale modeling
offer useful ways to model the material behavior, but their applicability to full-scale
structural problems is still problematic, due to their requirement for huge amounts of
computer resources. Therefore, further development in the macroscopic constitutive
modeling of concrete is justified and needed, with the motivation of incorporating
contemporary experimentally observed features of the material behavior in the modeling.
In this study, constitutive models representing the macro constituents of RC (Concrete
and Steel Reinforcement) are developed with emphasis on the rigor and consistency in
formulation and implementation into the Finite Element Analysis (FEA) commercial
software ABAQUS.
Concrete and reinforcing steel are represented herein by separate material models
which are combined together using a model that describes the interaction between
reinforcing steel and concrete to simulate the overall behavior of the composite RC
material. An elasto-plastic-damage constitutive model is used to describe the behavior of
concrete, while steel reinforcement is modeled as an elastoplastic material with strain
hardening using the classical von Mises plasticity. Bond considerations are accounted for
within the steel reinforcement model. Coupling between damage and plasticity in the
constitutive model is employed to capture the observed phenomenological behavior of
concrete. In this combined approach, damage theory is used to model the material
deterioration, while the permanent deformation and some other behavioral features of
2
concrete can be captured using plasticity theory. All features of the two theories can be
incorporated in this combined approach, making it very promising for use in constitutive
modeling or RC structures.
The model will be translated into algorithms that will simulate the nonlinear material
behavior of concrete, steel reinforcement, and their interaction. They will also facilitate
the reproduction of experimentally observed load carrying capacity curves of RC beams.
These algorithms will be incorporated into the FEA software ABAQUS via a user-
defined material subroutine (UMAT). While ABAQUS performs the standard FE
procedure using standard types of finite elements, the UMAT will govern the behavior of
these materials during different loading stages, i.e., elastic, inelastic, failure, post-failure
loads. The model is intended to be robust, efficient and reliable but by no means vague
and complicated.
The complexity of the behavior of RC, and the scatter of the experimental data
associated with machine precision, variations in testing techniques, and statistical
distributions of material properties from one sample to another is one of the main factors
enforcing the notion that the primary goal of any constitutive model should be set in the
prediction of essential features of experimentally observed behavior, rather than in
exactly replicating the entire history of stress-strain curves. Along this line, it should be
emphasized that numerical implementation of a proposed constitutive model into a
computer code is almost as important an issue to consider as the model itself. A literature
survey can easily reveal models that are mathematically very elegant, but pose
overwhelming computational difficulties. It is thus important that a constitutive model,
although rigorous in theory, should also be suitable for use in computation and should
lend itself well to an efficient implementation in computer codes.
The objective of this study can be cast into the following points:
Capturing the elasto-plastic-damage behavior of concrete under monotonic loads
in tension, compression, tension-tension, compression-compression, and
compression-tension stress states. This incorporates defining the continuous
damage mechanism in concrete that will represent the strain softening in the post-
peak regions and the degradation of elastic stiffness.
Modeling the behavior of structural steel - embedded in the concrete - as an
elastoplastic material with isotropic hardening.
3
Accounting for the effect of bond between steel and concrete and their interaction
on the overall short term material (stress-strain) behavior of RC.
Verifying the applicability of the proposed model by comparing the predicted
behaviors with those observed in experimental results obtained by other
researchers.
This study starts with a literature review on the physical behaviors of RC and its
macroscopic constituents, i.e., steel and concrete, as well as the constitutive models
applied to describe such behaviors, all of which are presented in Chapter 2. Emphasis
here is placed on the experimental work done by various researchers to study concrete
and RC in structural members and on the constitutive models applied to capture the
important features of the experimentally observed material behaviors. This emphasis
leads to the application of combined approaches employing both damage mechanics and
plasticity theory to describe the concrete behavior and classical hardening plasticity
theory to describe that of steel reinforcement. The bond effect and material interactions as
well as modeling of these phenomena are also addressed.
Chapter 3 addresses the von Mises plasticity-based material constitutive models used
to describe the physical behavior of steel reinforcement. This includes perfectly plastic
and (linear, bilinear and nonlinear) hardening models. The chapter starts with an
introduction discussing the application of the FE method to nonlinear continuum
mechanics. The fundamentals of FE analysis procedures for elastic-plastic problems are
considered, with algorithms required for convergence at the local (Gauss point) and
global (equilibrium) levels. This is followed by a discussion of the incremental
constitutive theory of metal plasticity and the numerical techniques employed to integrate
the constitutive equations. Examples are provided to verify the implemented algorithms.
This chapter is intended to serve as an introduction to the chapters that follow in terms of
theoretical formulation and computational implementation methods.
In Chapter 4, the material model for plain concrete is discussed. Rigorous consistent
thermodynamic formulation is employed to derive the framework of the elastic-plastic-
damage constitutive model. The Helmholtz free energy function is discussed and the
dissipation potentials for plastic and damage processes are postulated. To capture the
different responses in tension and compression, the approach makes use of the separation
of tensile and compressive behaviors, obtained through the decomposition of stress tensor
and the introduction of the damage variables, all of which are integrated into the
thermodynamic framework. The dissipation process therefore consists of three separate
dissipation mechanisms: tensile and compressive damage coupled with plasticity.
Detailed description of the theoretical formulation of the constitutive model in terms of
the theories of plasticity with multiple hardening rules and continuum damage mechanics
is then provided, followed by a step-by-step detailing of the numerical scheme applied to
integrate the incremental constitutive equations. Non associative plasticity - with multiple
hardening rules - is combined with continuum damage mechanics, where the accumulated
damage in the concrete material is represented by two internal damage parameters, one in
4
tension and one in compression. The constitutive model is formulated as a relation
between the undamaged stress and an internal accumulated damage parameter defined
using the tensile and compressive damage parameters. The integration scheme of elastic
predictor followed by plastic and damage correctors is adopted in order to overcome the
difficulties that arise from the combined plastic-damage approach. Several verification
examples are provided in order to test the models predictions under uniaxial and biaxial
stress states, as well as under three point bending test of a single-edge-notched concrete
beam. The numerical results are compared to the experimental ones in order to evaluate
the performance of the proposed concrete model and to asses the models ability to
capture the experimentally observed behaviors of concrete materials.
Throughout this work, (+) and ( ) superscripts are used to indicate tension or
compression, respectively. All symbols with an over bar, e.g. , are considered to be in
the effective fictitious undamaged configuration. On the other hand, the absence of the
over bar indicates the actual damaged configuration.
5
CHAPTER 2
LITERATURE REVIEW
The nonlinear response is caused by three major effects, namely, cracking of concrete
in tension, yielding of the reinforcement or crushing of concrete in compression, and the
interaction of the constituents of RC. Interaction includes bond-slip between reinforcing
steel and surrounding concrete, aggregate interlock at a crack and dowel action of the
reinforcing steel crossing a crack. The time-dependent effects of creep, shrinkage and
temperature variation also contribute to the nonlinear behavior. Furthermore, the stress-
strain relation of concrete is not only nonlinear, but is different in tension than in
compression and the mechanical properties are dependent on concrete age at loading and
6
on environmental conditions, such as ambient temperature and humidity. The material
properties of concrete and steel are also strain-rate dependent to some extent.
The earliest publication on the application of the finite element method to the analysis
of RC structures was presented by Ngo and Scordelis (1967). In their study, simple
beams were analyzed with a model in which concrete and reinforcing steel were
represented by constant strain triangular elements, and a special bond link element was
used to connect the steel to the concrete and describe the bond-slip effect. A linear elastic
analysis was performed on beams to determine principal stresses in concrete, stresses in
steel reinforcement and bond stresses. Ngo and Scordelis (1967) reported that one of the
main difficulties in constructing an analytical model for RC member is due to the
composite action of steel and concrete. Prefect bonding between steel and concrete can
only exist at an early stage under low load intensity. As the load is increased, cracking as
well as breaking of bond inevitably occurs, and a certain amount of bond slip will take
place in the beam, all of which will in turn affect the stress distributions in concrete and
steel.
Since Ngo and Scordelis published their landmark paper in 1967, the analysis of RC
structures has enjoyed a growing interest and many publications have appeared.
Important progress has also been made in the finite-element-based numerical analysis of
plain and RC structures (recent examples: Fantilli et. al., 2002; Sumarac et. al., 2003;
Marfia et. al., 2004; Phuvoravan and Sotelino, 2005; Junior and Venturini, 2007; just to
mention a few). However, despite this progress, modeling the mechanical behavior of
concrete is still one of the most difficult challenges in the field of structural engineering.
This is due to the inherent complexity and uncertainty concerning the properties of
concrete, which make it excessively difficult to develop accurate constitutive models or
algorithms that are sufficiently robust to obtain reliable and converged solutions in
numerical analyses.
7
RC is a composite material made up of concrete and steel, two materials with very
different physical and mechanical behavior.
Concrete exhibits nonlinearities even under low level of loading due to nonlinear
material behavior, environmental effects, cracking, biaxial stiffening and strain
softening.
Reinforcing steel and concrete interact in a complex way through bond-slip and
aggregate interlock.
Because of these factors and the differences in short- and long-term behaviors of the
constituent materials, it is common practice among researches to model the short- and
long-term response of RC members and structures based on separate material models for
reinforcing steel and concrete, which are then combined along with models of interaction
between the two constituents to describe the behavior of the composite RC material. This
will be the approach adopted in this study of short-term behavior of RC beams. In the
following, literature review of the research done to describe the behavior of concrete,
reinforcing steel, and their bond-interaction analysis is presented.
8
of the material. Concrete may be treated as a composite material and it may contain
porosity in its matrix. The porosity in the matrix is not homogenous and a strong porosity
gradient is observed around the inclusions formed by the aggregates (Panoskaltsis and
Lubliner, 1994; Ollivier et. al., 1995). This area in the matrix affected by the surface of
the aggregate is known as transition zone. This transition zone has a damaging effect
on the mechanical behavior of concrete. It is clear from the above discussion that the size
and texture properties of the aggregates will also have a significant effect on the
mechanical behavior of concrete under various types of loading (Chen, 1982).
The nonlinear material behavior of concrete can be attributed to two distinct material
mechanical processes; plasticity and damage mechanisms. The cracking process in
concrete is distinguished from cracking of other materials in that it is not a sudden onset
of new free surfaces but a continuous forming and connecting of micro-cracks. The
formation of micro-cracks is presented macroscopically as softening behavior of the
material, which causes localization and redistribution of strains in the structure. This
phenomenological behavior at the macroscopic level can be modeled by classical
plasticity (Pramono and Willam, 1989). On the other hand, the micro-processes such as,
micro-cracking, micro-cavities, and their nucleation and coalescence, also cause stiffness
degradation, which is difficult to represent with classical plasticity (Lee and Fenves,
1998). This introduces the need for continuum damage mechanics, where the stiffness
degradation can be modeled by making use of the effective stress concept (Kachanov,
1958) as will be shown later. Damage mechanics can also be used to represent the post-
peak softening behavior of concrete materials; a behavior that cannot be addressed by
classical plasticity theory.
9
(shown as 45% in Fig. 2.3). In this initial stage, one can observe the highest tensile strain
concentration at particular points where new micro-cracks are initiated as shown in Fig.
2.2b. At this load state, localized cracks are initiated, but they do not propagate
(stationary cracks). Hence, the stress-strain behavior is linearly elastic. Therefore, 0.3 f c'
is usually proposed as the limit of elasticity. Beyond this limit, the stress-strain curve
begins to deviate from a straight line. Stresses up to 70-90% of the ultimate strength
(shown as 85% in Fig. 2.3) characterize the second stage. In this stage, as the applied load
is progressively increased, the crack system multiplies and propagates as shown in Fig.
2.2c. The increase of the internal damage in concrete, revealed by deviation from the
linear elastic behavior, reduces the material stiffness and causes irrecoverable
deformation in unloading. Although the relief of strain concentration continues during
this stage, void formation causes the rate of increase of the tensile strain in the direction
normal to that of branching to increase with respect to the rate of increase of the strain in
the direction of branching (Kotsovos and Newman, 1977). The start of such deformation
behavior is called onset of stable fracture propagation (OSFP). In this load stage, the
mortar cracks tend to bridge the aggregate-matrix bond cracks.
A third stage shown in Fig. 2.3 extends up to the ultimate strength. Interface micro-
cracks are linked to each other by mortar cracks as shown in Fig. 2.2c, and void
formation (dilation) begins to have its effect on deformation at this stage. The start of this
stage is called onset of unstable fracture propagation (OUFP). This level is easily
10
defined since it coincides with the level at which the overall volume of the material
becomes a minimum. In this stage, the progressive failure of concrete is primarily caused
by cracks through the mortar. These cracks merge with bond cracks at the surface of
nearby aggregates and form crack zones of internal damage. Following that, a smoothly
varying deformation pattern may change and further deformation may be localized.
A fourth stage defines the region beyond the ultimate strength. In this region, the
energy released by the propagation of a crack is greater than the energy needed for
propagation. Therefore, the cracks become unstable and self-propagating until complete
disruption and failure occurs. In this stage, the major cracks form parallel to the direction
of the applied load, causing failure of the concrete. The volume of voids increases
dramatically causing a rapid dilation of the overall volume of concrete as shown in Fig.
2.2d.
All the above mentioned stages are for the uniaxial compression case. Stages I, (II and
III), and IV could be categorized into the linear elastic, inelastic, and the localized stages
respectively. Understanding these stages is crucial for the development of any concrete
model.
Figure 2.4 shows a typical uniaxial tension stress-elongation curve. In general the
limit of elasticity is observed to be about 60-80% of the ultimate tensile strength. Above
this level, the aggregate-matrix interface micro-cracks start to grow. As the uniaxial
tension state of stress tends to arrest the cracks much less frequently than the compressive
stage of stress, one can expect the interval of stable crack propagation to be quite short,
and the unstable crack propagation to occur much earlier. That is why the deformation
behavior of concrete in tension is quite brittle in nature. In addition, the aggregate-matrix
11
interface has a significantly lower tensile strength than the matrix, which is the primary
reason for the low tensile strength of concrete.
Nonlinearities in concrete behavior are well documented and arise from two distinct
micro-structural changes that take place in the material: one is the plastic flow; the other
is the development of micro-cracks and micro-voids. From a plasticity point of view, the
number of bonds between atoms during the plastic flow process is hardly altered;
therefore, the elastic compliances remain insensitive to this mode of micro-structural
change. On the other hand, micro-cracking destroys the bond between material grains,
affects the elastic properties, and may also result in permanent deformations, which can
be modeled by damage mechanics.
The idea of combined plasticity and damage mechanics theories through the
description of plasticity and damage surfaces has been explored and used in the past
(Oritz, 1985; Lubliner et. al. 1989). Many researches attempted to expand the application
of plasticity and damage theories to concrete (Chen and Chen, 1975; Lubliner et. al.,
1989; Yazdani and Schreyer, 1990; Abu-Lebdeh and Voyiadjis, 1993; Voyiadjis and
Abu-Lebdeh, 1994; Luccioni et. al., 1996; Lee and Fenves, 1998, 2001; Faria et. al.,
1998; Hansen et. al., 2001; Nechnech et. al., 2002; Gatuingt and Pijaudier-Cabot, 2002;
Kratzig and Polling, 2004; Salari et. al., 2004; Shen et. al., 2004; Jankowiak and
Lodygowski, 2004; Luccioni and Rougier, 2005, Rabczuk et. al., 2005; Wu et. al., 2006;
Grassl and Jirasek, 2006; Nguyen and Korsunsky, 2006; Shao et. al., 2006; Jason et. al.,
2006; Contrafatto and Cuomo, 2006; Nguyen and Houlsby, 2007; Mohamad-Hussein and
Shao, 2007; Ananiev and Obolt, 2007; Voyiadjis et. al., 2008a,b; Yu et. al., 2008; and
others).
Plasticity by itself fails to address the softening behavior of concrete under tension and
compression caused by damage propagation due to micro-cracking in the strained
12
material. On the other hand, damage mechanics is only concerned with the description of
this progressive weakening of solids due to the development of micro-cracks and micro-
voids (Loland, 1980; Oritz and Popov, 1982; Krajcinovic, 1985; Simo and Ju, 1987a,
1987b; Ju et. al., 1989; Voyiadjis and Kattan, 1989, 1999, 2006). Therefore, a
constitutive model should equally address these two distinct physical modes of
irreversible changes and should satisfy the basic postulates of mechanics and
thermodynamics governing these phenomena. Concrete plasticity and damage will be
further discussed in what follows.
The uniaxial behaviors of plain concrete under tension and compression up to tensile
and compressive failure are shown in Fig. 2.5. For tensile failure, the behavior is
essentially linear elastic up to the failure load, followed by a strain softening response
which was generally neglected or idealized in the past. For compression failure, the
material initially exhibits almost linear behavior up to the proportional limit at point A,
after which the material is progressively weakened by internal micro-cracking up to the
end of the perfectly plastic flow region CD at point D. The nonlinear deformations are
basically plastic, since upon unloading only the elastic portion e can be recovered from
the total deformation . It is clear that the phenomenon in regions AC and CD
corresponds exactly to the behavior of a work-hardening elastoplastic and elastic
perfectly plastic solid, respectively. As can be seen from Fig. 2.5, the total strain in a
plastic material can be considered as the sum of the reversible elastic strain e and the
permanent plastic strain p . A material is called perfectly plastic or work-hardening
according to whether it does or does not admit changes of permanent stain under constant
stress (Chen, 1982).
The plastic response of concrete exhibits some characteristics that the classical theory
of plasticity can not describe. It was shown experimentally that there is a lack in
simulating the normality rule (Adenaes et. al., 1977). Furthermore, the descending branch
of the uniaxial stress-strain diagram of concrete resembles a violation of the Druckers
stability postulate. On the other hand, and from a macroscopic point of view; classical
plasticity can simulate the behavior of concrete particularly in the pre-peak regime such
as the nonlinearity of the stress-strain curve and the significant irreversible strain upon
loading. Therefore, the plasticity theory can be used in the modeling of strain hardening
behavior of concrete. Many works were presented by researchers to modify the classical
theory of plasticity in order to make it more suitable for concrete materials (Feenstra and
de Borst, 1996; Bicanic and Pearce, 1996; Grassl et. al., 2002; Park and Kim, 2005; and
others).
13
Figure 2.5 Uniaxial stress-strain curve, pre- and post failure regimes, (Chen, 1982)
The main characteristics of the plasticity models (Chen and Schnobrich, 1981; Han
and Chen, 1985; Oritz, 1985; Lubliner et. al., 1989; Voyiadjis and Abu-Lebdeh, 1994;
Lee and Fenves, 1998, 2001; Grassl and Jirsek, 2006; and others) used to describe the
behavior of concrete include: pressure and path sensitivity, non-associative flow rules,
work or strain hardening, and limited tensile strength. Many of those models have been
developed for the finite element analysis of structural elements. Some of theses models
are associated with high mathematical complexity which renders them undesirable for
many engineering applications, especially the analysis and design of simple structural
elements, such as beams and columns.
Concrete contains numerous micro-cracks even before the application of any external
loads. These inherent micro-cracks are mainly present at the aggregate-cement interface
as a result of shrinkage, and thermal expansion in the cement paste or aggregates.
Damage in concrete is primarily caused by the propagation and coalescence of these
micro-cracks. The growth of these micro-cracks during loading causes reduction in
strength and deterioration in the mechanical properties of the concrete material.
14
Therefore, modeling of crack initiation and propagation is very important in the failure
analysis of concrete structures.
To model damage in concrete, various types of constitutive laws have been presented
including different approaches such as the endochronic theory (Bazant, 1978) the plastic
fracturing theory (Dragon and Mroz, 1979) the total strain models (Kotsovos 1980),
plasticity with decreasing yield limit (Wastiels 1980), microplane models (Bazant and
Ozbolt, 1990), and the bounding surface concept (Voyiadjis and Abu-Lebdeh, 1993).
Continuum damage mechanics was first applied to metals and later modified to model
different materials. The term damage mechanics has been conventionally used to refer to
models that are characterized by a loss of stiffness or a reduction of the secant
constitutive modulus. It was first introduced by Kachanov (1958) for creep-related
problems and then later applied to the description of progressive failure of metals and
composites and to represent the material behavior under fatigue (Kachanov, 1986). The
use of continuum damage mechanics in concrete began in 1980s. Damage models were
used to describe the strain-softening behavior of concrete. Since then, researchers
developed different damage models to represent concrete (Lemaitre and Mazars, 1982;
Krajcinovic, 1986; Mazars and Pijaudier-Cabot, 1989; Voyiadjis and Abu Al-Lebdeh,
1992; and others).
Damage theory in concrete materials can represent the post-peak region (Krajcinovic,
1983b; Mazars and Pijaudier-Cabot, 1989; and others.). The use of fracture mechanics in
concrete materials, on the other hand, was debated extensively (Mindess, 1983;
Krajcinovic and Fanella, 1986). The debate of the use of fracture mechanics for concrete
materials was supported by some experimental evidence (Pak and Trapeznikov, 1981),
which showed that energy dissipation and planar crack idealization in concrete make the
application of fracture mechanics in concrete materials very arguable. Therefore,
continuum damage mechanics is applied for the determination of macroscopic damaging
variables and material properties (Krajcinovic, 1979; Dragon and Mroz, 1979; among
others).
There are several approaches of how the stiffness degradation is included in a model.
Some researchers coupled damage with elastic analysis only (Mazars, 1984; Willam et.
al., 2001; Tao and Phillips, 2005; Labadi and Hannachi, 2005; Junior and Venturini,
2007; Khan et. al., 2007), while others coupled damage with plasticity. In the plastic-
damage model (Simo and Ju, 1987a,b; Ju, 1989; Lubliner et. al., 1989; Voyiadjis and
Kattan, 1989; Luccioni et. al., 1996; Armero and Oller, 2000; Salari et. al., 2004;
Voyiadjis et. al., 2008a) stiffness degradation is embedded in a plasticity model. These
models introduce the use of elastic and plastic damage variables to represent the stiffness
degradation. The damage variables are coupled with the plastic deformation in the
constitutive formulations which provide help for calibrating the parameters with the
experimental results. Yet, the coupled relations are complex and result in an unstable
numerical algorithm. This kind of algorithm may cause unrealistic representation of the
plastic behavior of the concrete during numerical implementation and iteration
procedures (Lee and Fenves, 1994).
15
In the effective-plasticity and damage model, plasticity is formulated in the effective
stress space to model the plastic irreversible phenomena while continuum damage
mechanics is adopted to represent stiffness degradation (Simo and Ju, 1987a,b; Ju, 1989;
Hansen and Schreyer, 1992; Yazdani and Schreyer, 1990; Fichant et. al., 1999; Voyiadjis
and Kattan, 1999; Mazars and Pijaudier-Cabot, 1989; Salari et. al., 2004; Shen et. al.,
2004; Lee and Fenves, 1998; Faria et. al., 1998; Jefferson, 2003; Voyiadjis and Kattan,
2006; Jason et. al., 2006; Voyiadjis et. al., 2008b). It has the advantage to decouple the
stiffness degradation from the plastic deformation by linearizing the evolution equations
(e.g., Jason et. al. 2006).
There are models that present thermodynamic theories of anisotropic damage mainly
by the use of tensor damage variables (Krajcinovic and Lemaitre, 1986; Krajcinovic and
Fonseka, 1981; Krajcinovic, 1983b; Voyiadjis and Kattan, 1989, 1999, 2006). Chow and
his co-workers (Chow and Wang, 1988; Chow and Lu, 1989) proposed an energy-based
elastic-plastic damage model in order to describe the difference in the observed failure
16
modes of geological materials under compression and tension, by using a damage tensor
identified by the fourth rank order and fourth rank projection tensors. The coupling
between elastic-plastic deformation and anisotropic damage using symmetric second
order damage tensors is widely presented by several authors (Simo and Ju, 1987a,b; Ju,
1989; Voyiadjis and Kattan, 1989, 1990, 1992a,b; Hansen et. al., 2001; Tikhomirov and
Stein, 2001; Voyiadjis et. al., 2008a,b; and others).
Normalized stress fnom= f / fc0
The variety in all of these anisotropic models is somewhat puzzling because a) the
relation between these models is difficult to establish (except maybe in the case of the
isotropic damage) and b) the comparison of damage-induced anisotropy with
experimental data is difficult and therefore the characterization of the damage-induced
anisotropy of the material requires three-dimensional experimental facilities. Because of
this difficulty in the experimental simulation of concrete specimens, together with the
inconvenience related to computational aspects, many researchers avoided these
complexities by resorting to simplified (single or multiple) isotropic damage models.
17
compressive loading and therefore, more damage occurs in tension than in compression.
Thus, the characterization of material damage in tension and in compression should be
considered when modeling the response of plain concrete, justifying the use of multiple
damage variables.
Axial deformation, m
Figure 2.7 Stress-deformation history for concrete subjected to cyclic tensile
loading (data from Reinhardt, 1984)
18
2.2 Steel Reinforcement
Reinforcement comes in different types and shapes. Those most commonly used are
the deformed circular cross-sectional bars. The spiral deformation pattern on the bars
strengthens the mechanical bond between the bars and concrete. The properties of
reinforcing steel, unlike concrete, are generally not dependent on environmental
conditions or time. Thus, the specification of a single stress-strain relation is sufficient to
define the material properties needed in the analysis of RC structures.
Typical stress-strain curves for reinforcing steel bars used in concrete construction are
obtained from coupon tests of bars loaded monotonically in tension. For all practical
purposes steel exhibits the same stress-strain curve in compression as in tension. The
steel stress-strain relation exhibits an initial linear elastic portion, a yield plateau, a strain
hardening range in which stress again increases with strain and, finally, a range in which
the stress drops off until fracture occurs. The extent of the yield plateau is a function of
the tensile strength of steel. High-strength, high-carbon steels, generally, have a much
shorter yield plateau than relatively low-strength, low-carbon steels (Wang and Salmon,
1998).
Two different idealizations, shown in Fig. 2.8, are commonly used depending on the
desired level of accuracy (ASCE 1982). The first idealization neglects the strength
increase due to strain hardening and the reinforcing steel is modeled as a linear elastic,
perfectly plastic material, as shown in 3.8a (Mazars and Pijaudier-Cabot, 1989; Sumarac
et. al., 2003; Kratzig and Polling, 2004; Luccioni and Rougier, 2005; and Wu et. al.,
2006). This assumption underlies the design equations of the ACI code. If the strain at the
onset of strain hardening is much larger than the yield strain, this approximation yields
very satisfactory results. This is the case for low-carbon steels with low yield strength.
If the steel hardens soon after the onset of yielding, this approximation underestimates
the steel stress at high strains. In several instances it is necessary to evaluate the steel
stress at strains higher than yield to more accurately assess the strength of members at
19
large deformations. This is, particularly, true in seismic design, where assessing the
available ductility of a member requires that the behavior be investigated under strains
many times the yield strain. In this case more accurate idealizations which account for the
strain hardening effect are required, as shown in Fig. 2.8b for the case of bilinear stress-
strain models (Salari and Spacone, 2001; Limkatanyu and Spacone, 2003). The
parameters of these models are the stress and strain at the onset of yielding, the strain at
the onset of strain hardening and the stress and strain at ultimate (Fig. 2.9). These
parameters can be derived from experimentally obtained stress-strain relations.
Figure 2.9 Linear elastic, linear strain hardening steel stress-strain relation
In this study the reinforcing steel is modeled as a linear elastic, linear strain hardening
material with yield stress y , as shown in Fig. 2.9. The reasons for this approximation
are:
1- The computational convenience of the model.
2- The behavior of RC members is greatly affected by the yielding of reinforcing
steel when the structure is subjected to monotonic bending moments. Yielding is
accompanied by a sudden increase in the deformation of the member. In this case
the use of the elastic-perfectly plastic model in Fig. 2.8a leads to numerical
convergence problems near the ultimate member strength. It is, therefore,
advisable to take advantage of the strain-hardening behavior of steel to improve
the numerical stability of the solution. The assumption of a linear strain hardening
behavior immediately after yielding of the reinforcement does not adversely affect
the accuracy of the results, as long as the slope of the strain hardening branch is
determined so that the strain energy of the model is equal to the strain energy of
the experimental steel stress-strain relation (Fig. 2.9). Such a model has been
successfully used for the analyses of RC structures (Ngo and Scordelis, 1967;
Feenstra, 1993; Kwak and Fillippou, 1997; Lowes, 1999; Tikhomirov and Stein,
20
2001; Rabczuk et. al., 2005; Phuvoravan and Sotelino, 2005; He et. al., 2006;
Junior and Venturini, 2007, and others).
Different approaches can be used for modeling reinforcing steel in RC structure using
the finite element method. Some researchers use two dimensional elements to represent
steel reinforcement in their finite element meshes (Ngo and Scordelis, 1967; Mazars and
Pijaudier-Cabot, 1989; Pijaudier-Cabot et. al., 1991; Luccioni and Rougier, 2005; and
Wu et. al., 2006; and others). On the other hand, one dimensional representation of steel
reinforcement is more widely used since it is unnecessary to introduce the complexities
of multiaxial constitutive relationships for structural steel (Chen, 1982; Feenstra, 1993;
Kwak and Fillippou, 1997; Faria et. al., 1998; Lowes, 1999; Soh et. al., 2003; Sumarac et.
al., 2003; Kratzig and Polling, 2004; Rabczuk et. al., 2005; Phuvoravan and Sotelino,
2005; He et. al., 2006; and others). The discrete model, the distributed or smeared model,
and the embedded model are examples of one dimensional representation of steel
reinforcement in a RC finite element mesh. The most widely used approach is the
discrete one dimensional truss element, which is assumed to be pin connected to the
concrete elements and posses two degrees of freedom at each node (Feenstra, 1993; Faria
et. al., 1998; Kratzig and Polling, 2004; Phuvoravan and Sotelino, 2005; He et. al., 2006;
and others). Alternatively, beam elements can also be used where three degrees of
freedom are allowed at each end of the bar element (Rabczuk et. al., 2005). In the
distributed model, the reinforcing steel is assumed to be distributed over the concrete
elements at a certain orientation angle, and in the embedded steel model the reinforcing
steel is considered as an axial member built into the isoparametric elements representing
the concrete. A significant advantage of the discrete representation, in addition to the
simplicity, is that it can include the slip of reinforcing steel with respect to the
surrounding concrete in contrast with the other 1D models where perfect bond between
concrete and steel is usually assumed (e.g. de Borst et. al., 2004).
21
The main stress transfer mechanisms between concrete and steel in RC elements are
represented by adhesion, mechanical interaction, and friction (Luccioni et. al., 2005).
Adhesion is constituted by chemical bonds and stresses developed during the curing
process of concrete. This transfer mechanism, schematically represented in Fig. 2.10a, is
prevailing in the case of bars of smooth surface and its failure is characterized by the
initiation and propagation of cracks in the concrete/steel interface. In the case of
corrugated bars, the described mechanism is secondary and the stress transfer is mainly
due to the interaction between ribs and the surrounding concrete. Adhesion is relatively
soon exhausted in the global response and consequently, the transfer force is transmitted
by friction and mechanical interaction between the ribs and the adjacent concrete.
As the force in the reinforcing bar is increased, the transfer forces are dominated by
the mechanical interaction concentrating at the faces of the ribs. In this state, the term
adhesion stress refers to the mean force per unit of surface. At increased loading, the
concrete begins to fail near the ribs with two different modes of failure; by failure of the
concrete adjacent to the contact area, as illustrated in Fig. 2.10b and by transverse
cracking.
Figure 2.10 a) Schematic presentation of adhesion between concrete and steel and its
failure, b) Stress concentration: Transverse cracking
Transverse cracking initiates at the ribs presenting a characteristic cone shape, also
called secondary cracking or adhesion cracking. The bond zone, constituted by a
damaged zone with finite depth surrounding the steel bar, is defined by the extension of
the transverse cracking (Fig. 2.11a). With the stable propagation of transverse cracking,
the concrete next to the bar seems to form inclined struts that are known as compression
cones (Fig. 2.11b). The adhesion stiffness is usually characterized by the stiffness of
these struts (Luccioni et. al., 2005).
When the load is further increased, radial splitting forces can be developed. This
phenomenon is due to the rotation of the inclined struts (Fig. 2.11b) that produces a
considerable radial component of the contact force, the increase of the radial force that is
produced by the increase of the effective contact angle between ribs and concrete due to
the deposition of the crushed concrete at the faces of the ribs, the wedge action of the
ribs, and the radial component of the contact forces. Without an accurate confinement, a
splitting failure can occur, spreading the effect of bond outside the bond zone.
22
Figure 2.11 a) Failure of concrete in contact with face of rib, b) Change in direction of
struts: Longitudinal cracking
The bond zone dilation, due to the longitudinal cracking, takes place when the
adhesion stress reaches values next to the limit one. Following this moment, a stress
softening is observed in the response. This stress softening, characteristic of the bond
slip behavior, is frequently interpreted as a progressive shear failure of concrete between
the ribs. When the confinement stresses are low, the geometric variation in the contact
occurring between the ribs and concrete can also contribute to the stress softening. Both
the progressive shear failure of concrete and the contact zone geometric variation are
stimulated by the reduction of the confinement stresses produced by the propagation of
longitudinal cracking. In both cases, the stress softening reveals a strong discontinuity
between the reinforcing bar and the surrounding concrete.
Based on the bond zone mentioned above, some researchers combined 2D interface
elements with regular 2D elements representing steel and concrete (Rots, 1988; Clement,
1987). The nonlinear response of concrete near the steel bar is lumped into a fictitious
interface that has special constitutive equations. In addition some mechanisms such as the
friction that depends on the shape of the steel bars (deformed or not) have been included
into the behavior of interface elements. The implementation of such relations raises the
problem of the identification of additional material properties. Others lumped the damage
zone characteristics into an interface of zero thickness. In the work of (Dragosavic and
Groneveld, 1984), it is assumed that the thickness of the damaged layer around the
reinforcement is equal to the radius of the bar. This technique is justified using
homogenization in (Clement et. al., 1985). Thus omitting interface elements and
considering that concrete is progressively damaged around the steel bar is strictly
23
equivalent to using interface elements if the bar presents surface deformations, i.e., if
displacements are continuous at the interface. In finite element applications, omitting the
interface elements should certainly reduce the computational cost. However, it requires
the implementation of accurate constitutive equations for concrete which can be a very
sensitive subject.
2.4 Methodology
It was mentioned earlier that because of the many factors affecting the nonlinear
behavior of RC and the differences in short- and long-term behavior of the constituent
materials, it is a common practice among researches to model the short- and long-term
response of RC members and structures based on separate material models for reinforcing
steel and concrete, which are then combined along with models of the interaction
between the two constituents to describe the behavior of the composite RC material. This
will be the adopted approach in this study of short-term behavior of RC beams under
monotonic loads. Time dependent effects such as creep, shrinkage, and temperature
change will not be considered here. Thermodynamically consistent derivation of the
formulations will be shown throughout this work. The RC behavior predicted by the
proposed model will be eventually compared to results obtained through experiments
conducted by other researchers. In the following, the methodologies used to describe the
behavior of concrete, reinforcing steel, and their bond-interaction analysis is presented.
The plasticity concrete model selected for this study is the model presented originally
by Lubliner et. al., (1989) (also known as the Barcelona model) and later modified by Lee
and Fenves, (1998) and Wu et. al., (2006). The model is based on an internal variable-
formulation of plasticity theory for the nonlinear analysis of concrete. The model makes
use of the fact that concrete eventually exhibits strain-softening in tension and
compression, leading to complete loss of strength. In this regard concrete resembles such
24
materials as cohesive soils, and may be classified with them as frictional materials with
cohesion, where the eventual loss of strength may be thought of as the vanishing of the
cohesion (Lubliner et. al., 1989). For such a model to be capable of representing the
behavior of concrete materials, the yield criterion must be of the form in which the
concept of cohesion is clear, and the hardening rule should be such that it will eventually
lead to vanishing of the cohesion (total damage).
Any plasticity model must include three basic components: an initial yield surface that
defines the stress level at which plastic deformation begins, a hardening rule that defines
the change of loading surface as well as the change of the hardening properties of the
material during the course of plastic flow, and a flow rule which gives an incremental
plastic stress-strain relation to allow for the plastic strain evolution during the course of
loading.
The adopted yield surface accounts for plasticity in tension and compression, and is
observed as a successful yield criterion in simulating the behavior of concrete under
uniaxial, biaxial, multiaxial, and cyclic loadings (Lee and Fenves, 1998, 2001; We et. al.,
2006). Multiple hardening variables are introduced through this yield surface to account
for different hardening. The uniaxial strength functions are factored into two parts,
corresponding to the effective stress and the degradation of elastic stiffness. The
constitutive relations for elastoplastic responses are decoupled from the degradation
damage response (discussed in the next section), which provides advantages in the
numerical implementation. This criterion is given in the effective (undamaged)
configuration as follows:
( f b0 / f 0 ) 1
= (2.2)
2( fb0 / f 0 ) 1
c ( )
( ) = (1 ) (1 + ) (2.3)
c + ( + )
where fb0 and f 0 are the initial biaxial and uniaxial compressive yield stresses,
respectively, c and c + are the tensile and compressive cohesions, respectively.
25
Figure 2.12 Concrete behavior under uniaxial loading in, a) tension, b) compression
(ABAQUS, 2004)
The tensile and compressive cohesions c in Eq. (2.3), which are functions of the
tensile and compressive equivalent plastic strains , should be scaled so that their initial
26
values are f 0 , the initial yield strength in uniaxial tension and compression, respectively,
and their final values vanishes when the damage parameters reach their maximum.
However, unlike the usual plasticity models with isotropic hardening, the cohesions c
themselves are assumed to be internal variables governed by rate equations that simulate
their evolution.
It should be noted here that for tensile loading, damage and plasticity are initiated
when the equivalent applied stress reaches the uniaxial tensile strength f o+ as shown in
Fig. 2.12a. However, under compressive loading, damage is initiated at a different stage
than plasticity. Once the equivalent applied stress reaches f o (i.e. when nonlinear
behavior starts) damage is initiated, whereas plasticity occurs once f u is reached (Fig.
2.12b). Therefore, generally f o+ = f u+ for tensile loading, but this is not true for
compressive loading (i.e. f o f u ).
The flow rule gives the relation between the plastic flow direction and the plastic
strain rate. In contrast with metals, the nonassociative flow rule is necessary to control the
dilatancy in modeling frictional materials (Chen and Han, 1988). Because the yield
surface in Eq. (2.1) is a combined geometric shape from two different Drucker-Prager
type functions, a Drucker-Prager type function is used as the plastic potential function
(Lee and Fenves, 1998, 2001; Wu et. al., 2006).
F p = 3 J 2 + p I1 (2.4)
F p 3 sij
= + p ij (2.5)
ij 2 3J 2
The flow rule also connects the loading surface/function and the stress-strain relation.
When the current yield surface f is reached, the material is considered to be in plastic
flow state upon increase of the loading. The flow rule is presented as follows:
F p
ijp = (2.6)
ij
Damage mechanics can be illustrated using the effective stress concept proposed first
by Kachanov (1958) as explained below: Consider a uniform bar subjected to a uniaxial
uniform tensile stress, , as shown in Fig. 2.13a. The cross-sectional area of the bar in
the stressed configuration is A and it is assumed that both voids and cracks appear as
damage in the bar. The uniaxial tensile force T , acting on the bar is expressed using the
relation T = A . In order to use the principles of continuum damage mechanics, one
27
considers a fictitious undamaged configuration (effective configuration) of the bar as
shown in Fig. 2.13b. In this configuration all types of damage, including voids and
cracks, are removed from the bar. The effective stressed cross-sectional area of the bar in
this configuration is denoted by A and the effective uniaxial stress is ij . The bars in
both the damaged configuration and the effective undamaged configuration are subjected
to the same tensile force T . Therefore, considering the effective undamaged
configuration, one obtains the relation T = A . Equating the two expressions of T
obtained from both configurations, the following expression is derived:
A
= (2.7)
A
T T
Remove Both
Voids and Cracks
A A
T
T
Damaged Effective Undamaged
Configuration Configuration
(a) (b)
Figure 2.13 A cylindrical bar subjected to uniaxial tension: both voids and cracks
are removed (Voyiadjis and Kattan, 1999, 2006)
Moreover, as it is seen from Fig. 2.13, the effective area A is obtained from A by
removing the surface intersections of the micro-cracks and cavities (Voyiadjis and
Kattan, 1999, 2006) and correcting for the micro-stress concentrations in the vicinity of
discontinuities and for the interactions between these effects. Therefore, the damage
parameter in case of uniaxial loading can be defined as follows:
A
= 1 (2.8)
A
In the above equation, in the case of no damage (effective state) in the material the
damage parameter is equal to zero (i.e. = 0 for A = A ). The critical damage = cr
corresponds to the rupture of the element. Lemaitre (1984) showed that the damage
parameter value ranges between 0.2 and 0.8 ( 0.2 0.8 ) for metals. The theoretical
value of damage parameter, , for the general case lies in the range 0 1 .
28
The effective area A can be obtained through mathematical homogenization
techniques (Voyiadjis and Kattan, 1999, 2006). Homogenization techniques can be used
when the shape and the size of the defects are known which is somewhat difficult to
obtain even with electron microscopes.
Making use of Eqs. (2.7) and (2.8), one obtains the following expression for the
effective uniaxial stress (Kachanov, 1958):
= (2.9)
1
For a three-dimensional state of stress, Eq. (2.9) can be generalized for isotropic
damage as follows:
ij
ij = (2.10)
1
where ij and ij are the stress tensors in the damaged and effective configurations,
respectively.
The transformation equations from the nominal configuration to the effective one can
be obtained using either the strain energy equivalence hypothesis or the strain
equivalence hypothesis (Voyiadjis and Kattan, 1990, 1999, 2006). The strain equivalence
hypothesis states that the strain in the effective configuration is equal to the strain in the
nominal configuration such that in the constitutive equations one can simply replace the
nominal strain by the corresponding effective strain. On the other hand, the strain energy
equivalence hypothesis states that the elastic strain energy density in the damaged
configuration is equal to the elastic strain energy density in the effective (undamaged)
configuration.
The constitutive equation for the damaged material is written in terms of that of the
virgin material; that is, the damaged material is modeled using the constitutive laws of
the effective undamaged material in which the Cauchy stress tensor ij is replaced by the
effective stress tensor, ij (Murakami and Ohno, 1981):
ij = M ijkl kl (2.11)
where M ijkl is the fourth-order damage effect tensor which has many definitions in
literature (isotropic or anisotropic). One of these definitions is given by Voyiadjis and
29
Venson (1995) and Voyiadjis et. al. (2008a,b) for anisotropic damage as follows (in an
effort to symmetrize the stress tensor):
M ijkl = 2 ( ij ij ) kl + ij ( kl kl )
1
(2.12)
In this study, two isotropic damage variables are used to model the tensile and
compressive damage phenomena in concrete material. The constitutive equation is
written in terms of an accumulated damage variable that is function of the tensile and
compressive damage variables and the stress state using the spectral decomposition of the
stress tensor. Detailed formulation is presented in Chapter 4.
In this study the reinforcing steel is modeled as a linear elastic, linear strain hardening
material with yield stress y , as shown in Fig. 2.9. A J 2 elasto-plasticity model with
linear hardening will be adopted to describe the behavior or steel reinforcement. The von
Mises yield criterion, associative flow rule and isotropic hardening are suitable for
modeling structural steel. The von Mises yield criterion can be written as:
F ( , R) = eq y R( p) 0 (2.13)
where y is the yield stress, R( p ) is the isotropic hardening stress (linear function of the
accumulated plastic strain R = kp ), and eq is the von Mises equivalent stress defined
by:
3
eq = Sij Sij (2.14)
2
1
Sij = ij mm ij (2.15)
3
The plastic flow rule that governs the evolution of the plastic strain is given as follows:
f
ijp = (2.16)
ij
The scalar is the plastic multiplier which is equal in this case to the rate of the
accumulated plastic strain:
30
2 p p
= p = ij ij (2.17)
3
Figure 2.14 Monotonic stress-strain response of reinforcing bar fiber with different
degrees of bond (Monti and Spacone, 2000)
The effect of bond between steel and concrete will be incorporated in this study
through the description of the steel constitutive model. The global effect of bond and
interaction on the stress-strain diagram will be accounted for by reducing the yield stress
and the elastic and hardening moduli of the steel bars according to the parametric studies
conducted by Belarbi and Hsu (1994), Kwak and Kim (2006), and the references therein.
Further discussion is provided in Chapter 5.
Once the material constitutive models for concrete, steel reinforcement, and bond-
interaction are developed, they will be all implemented into the advanced finite element
analysis software ABAQUS via a user defined material subroutine (UMAT). While
ABAQUS performs the standard finite element procedure using standard types of finite
elements, the UMAT will govern the behavior of these materials during different loading
31
stages, i.e., elastic, inelastic, failure, post-failure loads. Combining a UMAT subroutine
with standard finite element procedure will manifest the simplicity and applicability of
the proposed model into any engineering problem where the average user will not be
exposed to the complexities associated with introducing non standard finite elements.
32
CHAPTER 3
3.1 Introduction
Typical stress-strain curves for reinforcing steel bars used in RC construction are
obtained from coupon tests of bars loaded monotonically in tension. For all practical
purposes, reinforcing steel exhibits the same stress-strain curve in compression as in
tension. Thus, the specification of a single stress-strain relation is sufficient to define the
material properties needed in the analysis of RC structures. The steel stress-strain relation
exhibits an initial linear elastic portion, a yield plateau, a strain hardening range in which
stress again increases with strain and finally, a range in which the stress drops off
(softens) until fracture occurs (Fig. 3.1).
The extent of the yield plateau is a function of the tensile strength of steel. High-
strength, high-carbon steels, generally, have a much shorter yield plateau than relatively
low-strength, low-carbon steels (Wang and Salmon, 1998). The experimentally obtained
33
stress-strain diagram for the reinforcing steel is usually replaced in research by an
idealized characteristic diagram (Fig. 3.2).
Two different idealizations, shown in Fig. 3.2, are commonly used in literature
depending on the desired level of accuracy (ASCE 1982). The first idealization neglects
the strength increase due to strain hardening and the reinforcing steel is modeled as a
linear elastic, perfectly plastic material, as shown in Fig. 3.2a. This assumption underlies
the design equations of the ACI code. If the strain at the onset of strain hardening is much
larger than the yield strain, this approximation yields very satisfactory results. This is the
case for low-carbon steels with low yield strength.
If the steel hardens soon after the onset of yielding, this approximation underestimates
the steel stress at high strains. At several instances it is necessary to evaluate the steel
stress at strains much higher than yield to more accurately assess the strength of members
at large deformations. This is, particularly, true in seismic design, where assessing the
available ductility of a member requires that the behavior be investigated under strains
many times the yield strain. In this case, more accurate idealizations which account for
the strain hardening effect are required. The second idealization - the case of the bilinear
stress strain models - would be more appropriate to model the steel behavior. As shown
in Fig. 3.2b, the reinforcing steel is modeled as a linear elastic, linear-plastic-hardening
material. The parameters of these models are the stress and strain at the onset of yielding
and the stress and strain at the ultimate load (Fig. 3.3). These parameters can be derived
from experimentally obtained stress-strain relations.
The first idealization (Fig. 3.2a) is frequently used by researches (Mazars and
Pijaudier-Cabot, 1989; Sumarac et. al., 2003; Kratzig and Polling, 2004; Luccioni and
Rougier, 2005; and Wu et. al., 2006) because of its simplicity, especially in the cases
where most of the interesting activities in the RC member occur before the strain-
hardening of reinforcing steel starts (Vecchio and Collins, 1982). The second idealization
(Fig. 3.2b) is also used in many studies (Ngo and Scordelis, 1967; Feenstra, 1993; Kwak
and Fillippou, 1997; Lowes, 1999; Tikhomirov and Stein, 2001; Rabczuk et. al., 2005;
34
Phuvoravan and Sotelino, 2005; He et. al., 2006; and Junior and Venturini, 2007). It is
claimed by some researchers (e.g. Kwak and Fillippou, 1997 and the references therein)
that adopted the second idealization that the use of the elastic-perfectly plastic model
shown in Fig. 3.2a leads to numerical convergence problems near the ultimate member
strength in situations where yielding is accompanied by a sudden increase in the
deformation of the member under monotonic bending moments. In these situations, the
RC member is greatly affected by the hardening of reinforcing steel, the fact that
encourages the use of bilinear stress-strain models to improve the numerical stability of
the solution. They also claimed that the assumption of a linear strain hardening behavior
immediately after yielding of the reinforcement does not adversely affect the accuracy of
the results, as long as the slope of the strain hardening branch is determined so that the
strain energy of the model is equal to the strain energy of the experimental steel stress-
strain relation (Fig. 3.3). Despite the discussion, both idealizations have been reported to
be successfully implemented into FE codes used to model RC structures.
Figure 3.3 Linear elastic, linear strain hardening steel stress-strain relation
35
idealization requires the development of a new criterion to determine the strains at which
hardening starts, a procedure that increases the complexity of this idealization technique.
Different approaches can be used to model reinforcing steel in RC structure using the
FE method. Some researchers chose to use two dimensional elements to represent steel
reinforcement in their FE meshes (Ngo and Scordelis, 1967; Mazars and Pijaudier-Cabot,
1989; Pijaudier-Cabot et. al., 1991; Luccioni and Rougier, 2005; and Wu et. al., 2006).
On the other hand, one dimensional FE representation of steel reinforcement is more
widely used to avoid introducing the complexities associated with multiaxial constitutive
relationships for structural steel (Chen, 1982; Feenstra, 1993; Kwak and Fillippou, 1997;
Faria et. al., 1998; Lowes, 1999; Soh et. al. 2003; Sumarac et. al., 2003; Kratzig and
Polling, 2004; Rabczuk et. al., 2005; Phuvoravan and Sotelino, 2005; and He et. al.,
2006).
Figure 3.4 Linear elastic, yield plateau, plastic hardening steel stress-strain relation
The introduction of the steel reinforcing elements into RC FE analysis can be done in
many different ways. The discrete element model, the distributed or smeared model, and
the embedded model are examples of representing steel reinforcement in an RC FE mesh.
The most widely used approach is the discrete element (e.g., Wu et. al., 2006), where
steel and concrete are modeled as two distinguished elements with different material
properties and behaviors. Whereas, in the distributed model, the reinforcing steel is
assumed to be distributed over the concrete elements at a certain orientation angle, and in
the embedded steel model the reinforcing steel is considered as an axial member built
into the isoparametric elements representing the concrete. A significant advantage of the
discrete representation, in addition to its simplicity, is that it eliminates the need to
develop more sophisticated types of finite elements needed by the other models (e.g.,
Soh et. al., 2003), and therefore facilitates the use of classical finite elements available to
a wider spectrum of engineers and researchers.
In this study, the steel reinforcing bars are modeled by two-dimensional constitutive
relations to comply with the primary objective of this work which is the two-dimensional
FE analysis of RC structures. The elastic-plastic with linear strain hardening model (i.e.,
36
the second idealization, Fig. 3.2b) will be eventually used in the RC beam analysis in a
subsequent chapter. Nevertheless, numerical algorithms facilitating the use of different
idealizations will be discussed and presented in this chapter.
The following sections describe how metal plasticity may be applied to solve practical
boundary-value problems. Emphasis is given to the development of numerical procedures
used in solving elastic-plastic boundary-value problems by the nonlinear FE method and
the theory of plasticity. The application of the FE method to nonlinear mechanics will be
presented first, followed by the discussion of the aspects related to the constitutive
equations. Implementation of incremental elastic-plastic constitutive relations in a
numerical process where strain and stress increments are no longer infinitesimal but with
a finite size, will also be discussed and presented.
The basic formulations and procedures of the FE method for elastic analysis are
described briefly as an introduction, followed by a description of the formulations and
procedures for elastic-plastic analysis, where elastic-plastic FE analysis procedure results
in a set of simultaneous nonlinear equations, the solution of which will also be discussed.
V
ij ij dV = Ti ui dA + qi ui dV
A V
(3.1)
where ij and ij are the stress and strain tensors, respectively, ui and ij are virtual
displacement and virtual strain increments, respectively, where these increments form a
compatible set of deformations, and ij with Ti and qi form an equilibrium set. In matrix
form, which is adopted more frequently for FE formulations than tensorial notations, Eq.
(3.1) can be represented as:
{ } { } dV = { u} {T } dA + { u} {q} dV
T T T
(3.2)
V A V
where the vectors for displacement {u} , strain { } (similarly { } and { u} ) and stress
{ } are defined in general (3D) terms as follows:
37
{u} = {u x , u y , u z }
T
(3.3)a,b
{ u} = { ux , u y , uz }
T
{ } = { x , y , z , yz , zx , xy }
T
(3.4)a,b
{ } = { x , y , z , yz , zx , xy }
T
{ } = { x , y , z , yz , zx , xy }
T
(3.5)
{ } = [ L]{u}
(3.6)a,b
{ } = [ L ]{ u}
x 0 0 0
z y
[ L]
T
= 0 0 0 (3.7)
y z x
0 0 0
z y x
{u}m = [ N ]m {U } (3.9)
38
four nodal points, say points i, j, k, l, the shape function matrix will have non-zero sub
matrices of 3 by 3 only at columns i, j, k, l, and may be expressed as:
{ }m = [ B ]m {U }
(3.11)a,b
{ }m = [ B ]m { U }
[ B ]m = [ L ][ N ]m (3.12)
Substituting Eqs. (3.11)a,b and (3.9) into Eq. (3.2), one obtains the governing equation
for small-deformation FE analysis as:
[ B ] { } dV = [ N ] {T } dA + [ N ] {q} dV
T T T
m m m
(3.13)
m Vm m Am m Vm
where m = 1,2,.,M, or summation is over every element. This equation may also be
written in a simplified form as:
[ B ] { } dV = [ N ] {T } dA + [ N ] {q} dV
T T T
(3.14)
V A V
{R} = [ N ] {T } dA + [ N ] {q} dV
T T
(3.15)
A V
denote the equivalent external force acting on the nodal points, then accordingly Eq.
(3.14) can be reduced to:
[ B ] { } dV = {R}
T
(3.16)
V
For an elastic analysis, the stress-strain relationship may be generally written as:
{ } = [C ]{ } (3.17)
39
where [C ] is the elasticity matrix. The governing Eq. (3.16) may be further written as:
[ K ]{U } = {R}
(3.18)a,b
[ K ] = [ B ] [C ][ B ] dV
T
where [ K ] is termed the stiffness matrix of the FE system. Equation (3.18)a represents a
set of linear simultaneous equations. The displacement vector {U } may be determined by
solving this set of equations, and the strain and stress of each element can then be
determined by Eqs. (3.11)a and (3.17).
It should be noted here that in Eqs. (3.9), (3.11), and (3.12), the element shape
function matrix [ N ]m and the strain-displacement matrix [ B ]m are expressed in terms of
the global displacement vector {U } . More compact form of these two matrices may be
obtained if they are expressed in terms of the element displacement vector {U }m . For a
four-node element with nodal points i, j, k, and l, one can write:
where each sub-vector in the equation represents the displacement vector at a nodal point.
In terms of {U }m , Eqs. (3.9) and (3.10) may be rewritten as:
{u}m = [ N ]m {U }m (3.20)
[ N ]m = Ni I , N j I , N k I , Nl I (3.21)
For most elastic-plastic problems, closed-form solutions are not possible and
numerical solutions are sought via the FE method. Due to the nonlinear relationship
between the stress { } and the strain { } , the governing equation, Eq. (3.16), is a
nonlinear equation of strains, and thus, a nonlinear equation of nodal displacements {U } .
Iterative methods are therefore necessary to solve this equation for a given set of external
forces. Moreover, because of the deformation history dependence of an elastic-plastic
constitutive relation, an incremental analysis following the actual variation of the external
40
forces must be used to trace the history of the displacements, strains and stresses along
with the applied external forces.
{R} = { R} + {R}
n +1 n
(3.22)
where the right superscript (n) is used to refer to the n-th incremental step, and {R} is
the force increment from step (n) to step (n+1). Assuming that the solution at the n-th
step, {U } , { } and { } are known, and at the (n+1)th step, corresponding to the load
n n n
{U } = {U } + {U }
n +1 n
(3.23)a-b
{ } = { } + { }
n +1 n
{F } = {R}
n +1 n +1
(3.24)
{F } = [ B ] { }
n +1 T n +1
dV (3.25)
V
where { F } is the stress equivalent force acting on the nodal points. Substituting Eq.
n +1
[ B ] { } dV = { R} [ B ] { } dV
T n +1 T n
(3.26)
V V
which is another way of noting that the stiffness matrix, at any (n+1) step, is a function of
the displacement. Therefore, the structural equation, Eq. (3.18)a, which now can be
written as:
K (U ) {U } = { R} (3.27)
41
Equation (3.24) in fact represents the equilibrium of the external force { R}
n +1
with the
internal force { F } . Equation (3.26) is the governing equation of the incremental FE
n +1
formulations. Two types of numerical algorithms are involved in solving this equation for
the displacement increment {U } and the stress increment { } . One is the algorithm
used for solving the nonlinear simultaneous equations generated from Eqs. (3.24) or
(3.26) for the displacement increment {U } (global/equilibrium iterations). Another is
the algorithm used to determine the stress increment { } corresponding to a strain
increment { } , which is computed from {U } , at a given stress state and a given
deformation history (local iterations) . These two algorithms constitute the nonlinear FE
procedure for elastic-plastic analysis. Many algorithms exist for solving the nonlinear
equations shown above. What follows is a description of two equation-solving techniques
applicable to the time-independent nonlinear equations.
( ) { ( )}
n +1
{U n +1} = F {U n +1} { R}
n +1
(3.28)
( )
{U n +1} using the Taylor series expansion at {U }i 1 and neglecting all higher-order
n +1
42
(
{U }i 1 +
n +1
)
{U }
(
{U } {U }i 1 = 0
n +1 n +1
) (3.29)
is evaluated at {U }i 1 , which can be written in terms of the external and
n +1
where
{U }
internal forces as:
F
{F }i 1 +
n +1
{U }
( {U }i ) {R} = 0
n +1
(3.30)
F
is evaluated at {U }i 1 ,
n +1
where
{U }
{U }i = {U } {U }i 1
n +1 n +1
(3.31)
and
{ ( )}
n +1
{F }i 1 = F {U }i 1 = [ B ] { }i 1 dV
n +1 n +1 T n +1
(3.32)
V
F
evaluated at {U }i 1 is equal to:
n +1
Recalling that
{U }
[ K ]i 1 = [ B ] C ep [ B ] dV
n +1 T
(3.33)
V
tangential stiffness matrix of the structure. Therefore, the iteration scheme of a Newton-
Raphson algorithm is obtained as follows:
[ K ]i 1 {U }i = {R} { F }i 1
n +1 n +1 n +1
(3.34)a,b
{U }i = {U }i 1 + {U }i
n +1 n +1
43
iteration step. This can be prohibitively expensive for a large scale system. Moreover for
a perfectly-plastic or a strain-softening material, the tangential matrix may become
singular or ill-conditioned during the iteration. This may cause difficulty in finding the
solution of the nonlinear equation system. Modification of the Newton-Raphson
algorithm might therefore become necessary.
evaluated at a load step m , where m < n + 1 . If the matrix is evaluated only at the
beginning of the first lead step, the initial elastic matrix [ K ]0 is used throughout all load
steps. Such a method is referred to as the initial stress method. Most commonly, when the
stiffness matrix is evaluated at the beginning of each load step, or for the ( n + 1 )th step,
the following stiffness matrix is used:
[K ] = [ K ]0 = [ K ]
m n +1 n
(3.35)
The iteration scheme for the Modified Newton Raphson algorithm is expressed as:
[ K ] {U }i = {R} { F }i 1
n n +1 n +1
(3.36)a,b
{U }i = {U }i 1 + {U }i
n +1 n +1
44
Another problem associated with the Modified Newton-Raphson algorithm is that if a
change in the external load causes an unloading in the structure analyzed, this algorithm
may not result in a convergent iteration, unless the stiffness matrix is re-evaluated once
an unloading is detected. This problem increases the programming complexity of the
implementation of the Modified Newton-Raphson.
[ N ] =
1 1
(1 r ), (1 + r ) (3.37)
2 2
where r is the natural coordinate with origin at the center of the bar element. The strain-
displacement matrix of the bar element is then obtained as:
1
[ B ] = [ 1, 1] (3.38)
L
45
The element stiffness matrices of the two-bar elements are:
E1 (u ) A 1 1 E2 (u ) A 1 1
[ K1 ] = [ K2 ] = (3.39)a,b
L 1 1 L 1 1
and the global stiffness matrix, using Eq. (3.18)b, is given as:
E1 (u ) E1 (u ) 0
A
[ K ] = [ B ] [C ][ B ] Adx = E1 (u ) E1 (u ) + E2 (u ) E2 (u )
T
(3.40)
L
L
0 E2 (u ) E2 (u )
Using the governing equation, Eq. (3.25), and the boundary condition u1 = u3 = 0 , the
governing equation at (n+1)th step becomes:
f (u ) = F (u ) R = 0 (3.41)
where u = u2 is the displacement of node 2 (see Fig. 3.5), and F (u ) is the stress
equivalent force given by:
F (u ) = A ( 1 (u ) 2 (u ) ) (3.42)
E1 (u ) E1 (u ) 0 u1 R1
A
E (u ) E (u ) + E (u ) E (u ) u = R (3.43)
L
1 1 2 2 2
0 E2 (u ) E2 (u ) u3 R3
After applying the loads and boundary conditions and removing the subscripts of the load
and displacement (only one load and one displacement remains), Eq. (3.43) becomes:
A
( E1 (u ) + E2 (u ) )( u ) = R (3.44)
L
The tangential stiffness matrix of the two-bar structure may be expresses as:
46
F A
(evaluated at ui ) = K i = ( E1 (ui ) + E2 (ui ) ) (3.45)
u L
where
E , 1 < y E, 2 < y
E1 (u ) = E and E2 (u ) = E (3.46)
, 1 > y ,2 > y
4 2
K i u = R F (ui ) = R , ui = ui 1 + u , u0 = 0 (3.47)
The iteration converges in three steps ( i = 3 ). These steps are shown in details below:
Step 1
Start with u0 = 0 , F (u0 ) = A ( 1 (u0 ) 2 (u0 ) ) = 0 , R = R F (u0 ) = 3 A y 0 = 3 A y
when u0 = 0 , 1 and 2 are both equal to zero, i.e., < y which means that
A
E1 (u0 ) = E2 (u0 ) = E , into K 0 u = R F (ui ) = R gives ( E + E ) u = R = 3 A y
L
3 y L
which yields u = . The next step starts
2E
Step 2
3 y L u1 3 y 3
u1 = u0 + u = , by rearranging 1 = 2 = = = y which means that both
2E L 2E 2
materials have yielded, and the tangential matrix must be used for the two bar elements
E E
after yielding. E1t (u1 ) = , E2t (u1 ) = , therefore,
4 2
E 1 E 1 1 1
F (u1 ) = A y + ( y ) ( y ( y ) = A y + y + y + y = 2.375 A y
4 2 2 2 8 4
A E E
K1 u = R F (ui ) = R or + u = 3 A y 2.375 A y = 0.625 A y which gives
L 4 2
L
u = 0.8333 y , the third and final step starts
E
Step 3
yL u1
u2 = u1 + u = 2.333 , by rearranging 1 = 2 =
= 2.333 y = 2.333 y which
E L E
means that both materials have yielded, and the tangential matrix must be used for the
47
E E
two bar elements after yielding. E1t (u2 ) = , E2t (u2 ) = into F (u2 ) gives
4 2
E 4 E 4 1 2
F (u2 ) = A y + ( y ) ( y ( y ) = A y + y + y + y = 3 A y
4 3 2 3 3 3
R = R F (ui ) = 3 A y 3 A y = 0 , which means that K 2 u = R F (ui ) = R =0
u = 0 end of solution procedure.
The solution procedure using Newton-Raphson scheme was shown here in full details
because it consists of only three steps and it forms the basis for all the successive
discussions related to solving the nonlinear equilibrium equations. Note that the solution
procedure terminates when u = 0 , which originates from R = 0 when the material
stiffness in not equal to zero. The term R is defined as the out-of-balance, or the
difference between the external force R and the stress equivalent force F . This term,
R , can be used as a convergence criterion to end the iterative procedure.
F A
K0 = (evaluated at u0 ) = ( E1 (u0 ) + E2 (u0 ) ) (3.48)
u L
Only the first three iterations will be shown here to describe the procedure. The results
will be shown afterwards:
Step 1
Start with u0 = 0 , F (u0 ) = A ( 1 (u0 ) 2 (u0 ) ) = 0 , R = R F (u0 ) = 3 A y 0 = 3 A y
when u0 = 0 , 1 and 2 are both equal to zero, i.e., < y which means that:
A
E1 (u0 ) = E2 (u0 ) = E , into K 0 u = R F (ui ) = R gives ( E + E ) u = R = 3 A y
L
3 y L 2EA
which yields u = , where K 0 = . The next step starts
2E L
Step 2
3 y L u1 3 y 3
u1 = u0 + u = , by rearranging 1 = 2 = = = y which means that both
2E L 2E 2
48
materials have yielded, and the tangential matrix must be used for the two bar elements
E E
after yield. E1t (u1 ) = , E2t (u1 ) = , therefore,
4 2
E 1 E 1 1 1
F (u1 ) = A y + ( y ) ( y ( y ) = A y + y + y + y = 2.375 A y , the
4 2 2 2 8 4
2EA
tangent stiffness matrix K 0 is not updated and remains constant K 0 =
L
2 EA
K 0 u = R F (ui ) = R or u = 3 A y 2.375 A y = 0.625 A y , which gives
L
L
u = 0.3125 y , the third step starts
E
Step 3
3 L L L
u2 = u1 + u = y + 0.3125 y = 1.8125 y , by rearranging:
2E E E
u
1 = 2 = 1 = 1.8125 y = 1.8125 y which means that both materials have yielded, and
L E
the tangential matrix must be used for the two bar elements after yielding.
E E
E1t (u2 ) = , E2t (u2 ) = into F (u2 ) gives:
4 2
E E
F (u2 ) = A y + (0.8125 y ) ( y (0.8125 y ) = 2.6094 A y , again the tangent
4 2
2EA
stiffness matrix K 0 is not updated and remains constant K 0 =
L
R = R F (ui ) = 3 A y 2.6094 A y = 0.3906 A y , which means that:
2 EA yL
K 0 u = R F (ui ) = R , or u = 0.3906 A y , which gives u = 0.1953 , the
L E
fourth step then starts.
By observing the value of R through the first three steps, we realized that as long as
the solution is converging, the value for R is decreasing. This iterative procedure is
repeated until the convergence criterion R reaches a practically small value (in this
case, when R 0.0001 ) . Twenty one iterations are performed to complete the analysis.
The results of the two iterative procedures used to solve the one-degree-of-freedom
nonlinear problem are shown in Fig. 3.6, where the stress equivalent force is normalized
in terms of A y , i.e., the ordinate is F / A y .
The results for the Modified Newton-Raphson method showing the variation of the
element normalized stresses in bar (1) and bar (2) along with the iteration steps are
demonstrated in Fig. 3.7. In the case of the Newton-Raphson method; only three points
would be plotted for each stress.
49
Newton-Raphson .vs. Modified Newton-Raphson
1.5
1 Modified NR
NR
0.5
0
0 5 10 15 20
Iteration number
Modified Newton-Raphson
1.5
1
0.5
Sigma bar 1
Stress
0
-0.5 0 5 10 Sigma
15 bar 2 20
-1
-1.5
-2
Iteration number
50
(strain hardening material with Et 2 = E / 2 ), the following results are obtained for the
Newton-Raphson and the Modified Newton-Raphson methods. Again, only three steps
were required for the Newton-Raphson method to converge. On the other hand, the
iterative procedure converges after 70 iterations ( R 0.0001 ) for the Modified Newton-
Raphson method (see Fig. 3.9).
2.5
2
Modified NR
1.5
NR
1
0.5
0
0 20 40 60
Iteration number
Figure 3.10 shows the normalized stresses in bar (1) and bar (2) along with iterations
for the Modified Newton-Raphson method. Note that as bar (1) fails to carry anymore
load, the entire load will be carried out by bar (2) in compression.
As can be seen from the previous two examples, the numerical algorithms/procedures
discussed above are applied to satisfy the weak form of the equilibrium equation. In each
step, the stresses are found and used to calculate the residual R which in turn is used to
calculate a new u until the equilibrium equation, K u = 0 , is satisfied. If the
equilibrium equation is not satisfied, a new step is applied where a new approximation to
the nodal displacement is proposed by solving the linear system. The procedure is
repeated until the equilibrium equation is satisfied.
Up to this point, only the global/equilibrium iterations - required to solve the nonlinear
governing equation at a specific time step for the entire structure - were discussed. The
stresses were calculated using very simple elastic E and elastic-plastic (tangent) Et
moduli. This is straight forward for a one-degree-of-freedom nonlinear problem, but
when discussing more complicated problems, the elastic-plastic tangential operator
depends on the incremental properties of the material as does the stresses. The
displacement increment u will be used to calculate the strain increment using an
incremental format of Eq. (3.11)a, and the latter, , will then be used to update the
stress. This introduces the need for the incremental theory of plasticity and the integration
of the constitutive equations to yield the required solution parameters at the local level.
51
Modified Newton-Raphson
1
0.5
0
-0.5 0 20 40 60
Stress -1
-1.5 Sigma 1
-2 Sigma 2
-2.5
-3
Iteration number
What follows is a discussion of the local or Gaussian iterations, i.e., how to integrate
the plasticity constitutive model to update the quantities at the point level. The
constitutive equations and computational algorithms for rate-independent elastic-plastic
analysis of metals are presented. Everything here will be demonstrated for isotropic metal
plasticity, which will be used to model the steel reinforcement in a RC beam. Different
forms of metal strain-hardening plasticity will be presented.
When the stress tensor, ij , and the strain tensor, ij , are related by a strain-dependent
matrix rather than a matrix of constants, the material studied is said to be nonlinear, i.e.,
material nonlinearity is observed. Computational challenges arise from the fact that the
equilibrium equations (discussed earlier) must be written using material properties that
depend on strains, but strains are not known in advance. The loading history and
geometry, support conditions, and material properties are assumed to be known (as
discussed in the examples of section 3.3.3). The deformations and stresses in the body as
a function of the applied load are being sought.
52
number of sampling points must be small to reduce computational expense. Accordingly,
simple elements should be used. A contrary argument is that many sampling points are
needed to accurately capture the spread of yielding in individual elements. In practice, the
choice should be made between many simple elements and a small number of more
sophisticated elements.
The total strain tensor ij is well accepted to be composed of two parts: a small
recoverable (reversible) elastic strain ije and a large irreversible plastic strain ijp such
that:
where Cijkl is the fourth order Hookes elastic operator given as:
The constants G and K are the shear and bulk moduli, respectively given by:
53
2E E
G= and K= (3.53)a,b
(1 + v) 3(1 2v)
dev
where E and v are the Youngs modulus and Poissons ratio, respectively, and I ijkl is
the deviatoric part of the fourth-rank identity tensor I ijkl given as:
1 1
dev
I ijkl = ( ik jl + il jk ) ij kl (3.54)
2 3
In order to develop a description of the relation between the plastic stresses and strains,
one needs to establish a yield criterion that will predict the onset of inelastic deformation.
This is done by adopting a yield function f ( , R ) in terms of the stress ( ) and the
hardening equivalent stress R (linear strain-hardening is assumed).
By studying the general nature of polycrystalline solids, these solids are assumed to be
isotropic and their yield criteria can be assumed to be independent of the hydrostatic
pressure. Therefore, the yield criterion here will only depend on the deviatoric
components of stresses and furthermore, because of isotropy, the yield criterion will only
depend on the magnitudes (not the directions) of the deviatoric stresses ( J 2
elastoplasticity or J 2 flow theory). This leads to the von Mises yield criterion (maximum
distortion-energy criterion) where the yield criterion is a function of the invariants of the
deviatoric stresses, given here as:
f ( , R) = eq y R ( p) 0 (3.55)
54
where y is the yield stress, R( p ) is the isotropic hardening stress (linear function of the
accumulated plastic strain p ), and eq is the von Mises equivalent stress defined by:
3
eq = Sij Sij (3.56)
2
where Sij is the deviatoric part of the Cauchy stress tensor, given as:
1
Sij = ij mm ij (3.57)
3
1
and mm is the mean stress and ij is the Kronecker delta.
3
Isotropic strain hardening can be modeled by relating the size of the yield surface to
plastic strain in some appropriate way. One of the most commonly used forms of
hardening stress R( p ) for steel reinforcement is (Ngo and Scordelis, 1967; Feenstra,
1993; Lowes, 1999; Tikhomirov and Stein, 2001; Phuvoravan and Sotelino, 2005;
Rabczuk et. al., 2005; He et. al., 2006; and Junior and Venturini, 2007):
R( p) = kp (3.58)
which is known as linear isotropic hardening law, k is the isotropic hardening constant,
and y + R ( p) represents the radius of the yield surface (a cylinder) in the space of
principal stresses. As the accumulated plastic strain ( p ) increases, the radius of the yield
surface increases.
In order to calculate the plastic strains induced by loading beyond yield, we note that
the magnitude of the plastic strain can be determined from the hardening behavior of the
material. Since the stress must be on the yield surface at all times (the consistency
condition), and the radius of the yield surface is related to the magnitude of the
accumulated plastic strain, the magnitude of the increment of the plastic strain must be
related to the stress increment. This leads to the definition of a plastic flow rule. A plastic
flow rule derived from the yield surface (associative plastic flow rule) governs the
evolution of the plastic strain, and is given as follows:
ijp = N ij (3.59)
f
where the scalar is called the plastic multiplier and N ij = is the gradient of the
ij
yield function with respect to the stress tensor. The geometric interpretation of the flow
rule is very simple. The gradient N ij is perpendicular to the yield surface, and therefore,
55
the flow rule is also referred to as the normality rule. It should be noted here that the
plastic flow rule is incremental in nature.
Using the equation for the yield criterion given in Eq. (3.55), the gradient N ij can be
evaluated for a von Mises material as:
f 3 Sij
N ij = = (3.60)
ij 2 eq
3 Sij
ijp = (3.61)
2 eq
If the scalar ( p) , which is the accumulated plastic strain, is defined as the integration
of the rate of the accumulated plastic strain during an iterative procedure, i.e.:
t
p(t ) = pd
(3.62)
o
2 p p
p = ij ij (3.63)
3
where ijp is the plastic strain rate, the plastic multiplier can then be easily proven to be
equal to the rate of the accumulated plastic strain, using Eqs. (3.56), (3.61) and (3.63):
56
leading to the following possible situations in the evaluation of plastic multiplier :
f ( , R) < 0 = 0 (elastic)
f < 0 = 0 (elastic unloading)
= 0 (neutral loading)
(3.66)
f ( , R ) = 0 f = 0
f = 0 > 0 (plastic loading)
If the updated yield function f n +1 (a scalar) at the end of a step is written in terms of
its initial value at the beginning of that step f n plus the rate of change df (which can be
considered equal to f in a numerical procedure), one can write:
f n +1 = f n + df = 0 (3.67)
but f n = 0 , which means that df = 0 is the only possible outcome. Taking the rate of the
yield function, Eq. (3.55), results in the following:
f f f f R
df = ij + p = ij + p = 0 (3.68)
ij p ij R p
f
and making use of Eqs. (3.50) and (3.51) along with substitutions as = 1 and
R
R
= k , we obtain the following expression for the plastic multiplier ( = p ):
p
N ij Cijkl kl
= (3.69)
N ij Cijkl N kl + k
Beyond the elastic limit, plastic flow rules are incremental in nature and stresses are
related to strains by means of an incremental constitutive relation obtained by considering
the rate of Hooks law (Eq. (3.51)):
as well as the rate of the strain additive decomposition equation, Eq. (3.50):
57
and substituting for the increment of the plastic strain, klp , using the expression obtained
from the normality rule, Eq. (3.59), along with the expression for the plastic multiplier,
Eq. (3.69), the following equation can be obtained:
N mn Cmnpq pq
ij = Cijkl (kl N kl ) (3.73)
N ab Cabcd N cd + k
In order to find an operator relating ij and kl , a number of tensorial manipulations and
rearrangements (shown below) leads to the following incremental constitutive relation:
N pq C pqrsrs
ij = Cijmn (mn N mn )
N ab Cabcd N cd + k
N pq C pqrs kr lskl
ij = Cijmn ( km lnkl N mn ) (3.74)
N ab Cabcd N cd + k
ij = Cijkl
ep
kl
where
Cijmn N mn N pq C pqkl
ep
Cijkl = Cijkl (3.75)
N abCabcd N cd + k
ep
Cijkl is the fourth-order tensor known as the elastoplastic tangent operator. It has the same
ep
symmetries as Cijkl , but unlike Cijkl , Cijkl is not constant; it depends on the deviatoric
stress and the hardening parameter (Doghri, 2000).
If the expression in Eq. (3.75) is simplified to compare with the previously discussed
ep
examples of the one-degree-of-freedom FE elastoplastic problem, Cijkl simplifies to
Et (section 3.3.3) which is given by:
E Ek
Et = E 1 = (3.76)
E+k E+k
58
Ek1
Et1 = E / 4 , but Et1 = which gives k1 = E / 5 (softening occurs)
E + k1
Ek2
Et 2 = E / 2 but Et 2 = which gives k2 = E (hardening occurs)
E + k2
where k for a given material is a constant property known as the strain-hardening
(softening) parameter or the plastic modulus.
It should be noted here that in ABAQUS nonlinear FE analysis associated with the
UMAT subroutine, ABAQUS carries out the global/equilibrium iterations while UMAT
integrates the material incremental constitutive model (local iterations). Yet, satisfying
the equilibrium equations (i.e., convergence) depends on the tangent operator passed back
to ABAQUS from the UMAT subroutine. This was illustrated previously in the one-
degree-of-freedom examples (3.2.2.2). Recalling that the value of tangential stiffness
matrix K i was calculated based on the tangential modulus Et , for both bar elements,
obtained in each step. In a multi-degree-of-freedom problem, Et will become the elastic-
plastic tangent operator C ep calculated at each integration point and passed back to
ABAQUS where it will be used to calculate the stiffness matrix [ K ] shown in Eq. (3.18).
The purpose here is to devise accurate and efficient algorithms for the integration of
the constitutive relations governing the material behavior. In the context of the FE
analysis, the integration of constitutive equations is carried out at the Gauss points for a
given deformation increment. The unknowns to be found are the updated stresses and
plastic variables.
Efficient algorithms should satisfy three basic requirements: consistency with the
constitutive equations to be integrated, numerical stability, and incremental plastic
consistency. The soundness and efficiency of the algorithm ensure the effectiveness of
the numerical procedure. An improper algorithm may lead not only to an inaccurate stress
solution, but may also delay the convergence of the equilibrium iterations or even lead to
divergence of the iteration.
59
In the scope of time-independent plasticity, it is common to use fictitious time
increments (pseudo-time) merely as a way of counting successive increments of loads,
stresses and strains. Therefore, we denote all quantities at time tn with a subscript ( n )
while all quantities at time tn +1 are denoted by a subscript ( n + 1 ). Increments within the
time step [ tn , tn +1 ] ( tn +1 tn = t ) are left with no subscripts. Tensorial notation will be
dropped whenever an algorithm is introduced to reduce the complication of indices and
subscripts. Each term will be written to correspond to its original format shown in the
previous sections. The reader can refer to the original format of each equation in the
algorithms to figure out what each symbol represents (scalar, vector, or tensor). The
procedure starts at time tn with knowledge of the stresses n and the total strain
increment passed to the UMAT file from the previous equilibrium iteration carried
out by ABAQUS. The goal now is to find the updated stresses n +1 , plastic strains np+1 ,
and the equivalent plastic strain pn +1 .
Two schemes for the integration of the constitutive equations will be presented next:
The Explicit (Forward-Euler) scheme and a Radial Return form of the Implicit
(Backward-Euler) scheme. Other integration methods exist and are numerous with
different levels of complexity. But due to the advantage of the presence of high level
processors, complicating the integration procedure will not result in any better outputs
than those obtained using these two schemes with a reasonable number of time
increments.
In both integration schemes, the first step is to use an elastic procedure to update the
stresses. If these updated stresses are found to lie within the yield surface, the material at
the Gauss point is assumed to have remained elastic. In this case, there is no need to
integrate the rate equation; elastic analysis can resume. However, if the elastic stresses
are outside the yield surface, an integration scheme is adopted to bring the stresses back
to the yield surface.
This integration scheme is famous for its simplicity and being straightforward to
implement. By calculating the gradient of the yield function, Eq. (3.60), and thus the
plastic multiplier, Eq. (3.69) at the beginning of the increment, the Explicit integration
scheme uses the parameters evaluated at the previous time step tn to calculate the
updated parameters at the end of the current time step tn +1 . The following is a block
representation of the Explicit (Forward-Euler) integration scheme:
Given n , np , pn and
If f ( n , Rn ) < 0 then exit the integration scheme-End If
If f ( n , Rn ) > 0 then
60
N nC
Calculate the plastic multiplier =
N nCN n + k
Calculate the increment of plastic strain p = N n
Calculate the increment of elastic strain using e = p
Calculate the increment of stress = C e
Update all the quantities
Stresses n +1 = n +
Strains np+1 = np + p
Equivalent plastic strain pn +1 = pn +
CN n N nC
Calculate the elastic-plastic tangent operator C ep = C
N nCN n + k
End If
Associated with its simplicity, however, are three of its disadvantages that need to be
discussed: a) Because it is an Explicit procedure, it is conditionally stable. b) The
accuracy of the integration depends on the increment size chosen. c) The plastic
multiplier was obtained in such a way that the yield condition is satisfied at time tn .
Satisfying the yield condition at time tn +1 was not checked. Therefore, the solution at
time tn +1 may drift away from the yield surface over many time steps. This drift can be
reduced by decreasing the size of the increment. Furthermore, if the total strain increment
1 (see Fig. 3.11) at time tn causes the stress to cross the yield surface, then
f ( n , Rn ) < 0 , and the increment will be treated elastically, which will over predict the
location of the yield stress from y to n +1 . Then the new total strain increment 2 and
all the consequent increments will result in a shifted stress strain diagram as shown in
Fig. 3.11. This leads to accumulation of errors and overestimation of the computed
ultimate (collapse) load.
61
In order to minimize the drift from the yield surface, the total strain increment 1
causing the stress to cross the yield surface for the first time needs to be resolved into two
parts (Fig. 3.11). One part of 1 is elastic ( 1e ) taking the stress to the first yield y
while the other part is plastic ( 1p ). Only the plastic part needs to be passed into the
integration algorithm shown above. The elastic part needs to be considered such that:
n + C 1e = y and ne+1 = ne + 1e . The remainder plastic part, 1p = 1 1e ,
becomes the new total strain increment passed to the integration algorithm with the
known value of the stress being equal to the yield stress, i.e., = 1p and n = y .
Different methods are suggested in literature for finding a scalar scaling factor such
that the elastic and plastic part of the strain increment 1 can be found: 1e = 1 and
1p = (1 )1 . In terms of stresses, the yield function should be satisfied at
n + = n + C 1e = y , i.e., f ( y ) = 0 . Note that f ( n ) < 0 as shown in Fig. 3.11.
When reaches its maximum value ( = 1 ), the elastically predicted stress would be
n + = n + C 1 , which is equal to n +1 . At a given value of , say i the stress
would be equal to i = n + i and the yield function becomes
f ( i ) = f ( n + i ) . By using a truncated Taylor series, the value of the yield function
f i f
f ( i ) can be given as: f ( i ) = f ( n ) + = f ( n ) + ( ) = 0 . Such a
i i i
f ( n )
scheme might start with an initial estimate o = to calculate
f ( n +1 ) f ( n )
f
o = n + o , evaluate and then calculate o using the truncated Taylor series,
o
f
f ( n ) + ( ) o = 0 . The next iteration then starts with 1 = o + o to obtain
o
f
1 = o + 1 . By calculating , and using the truncated series again,
i
f
f ( o ) + ( )1 = 0 , the next estimate of i can be found. The iterations continue
1
until the value of i practically approaches zero.
Once has been evaluated, the new strain increment = 1p (see Fig. 3.11) will
be passed into the Forward-Euler Explicit integration algorithm shown above to complete
the analysis of this yield point defining increment. A block representation of the
correction to the Explicit (Forward-Euler) integration algorithm is shown below:
Given n , np = 0 , pn = 0 and
If f ( n ) < 0 then
62
Calculate a trial elastic stress ntrial
+1 = n + C
If f ( ntrial
+1 ) < 0 then exit integration scheme- End If
This method is popular because, for the von Mises yield criterion, it takes a
particularly simple form. At a certain step in time, tn , applying the strain increment
takes the elastically updated stress n + C = ntrial +1 outside of the yield surface (see Fig.
3.12). This trial stress is known as the elastic predictor. The stress is then updated with a
plastic corrector C p to bring it back onto the yield surface at the end of the current
time step [ tn , tn +1 ] as follows: n +1 = n + C e = n + C ( p ) = n + C C p
= ntrial
+1 C . This plastic correction involves calculating the plastic multiplier
p
strain p .
63
In contrast to the Explicit scheme, all quantities are written here at the end of the time
increment, ensuring that the yield condition is satisfied at the end of the time increment,
and therefore, avoiding the drift from the yield surface which was observed in the
Explicit scheme. This scheme also leads to a faster (larger time increments) solution.
In the deviatoric stress space, the plane stress von Mises ellipse becomes a circle, and
the plastic correction term is always directed towards the center of the yield surface
(because of the normality of the flow rule). This fact gives the technique its name, i.e.,
the radial return method, which can be derived as follows:
as ijtrial ):
Expanding the plastic strain increment, using Eq. (3.59), one obtains:
Only the deviatoric part Sij of the stress tensor ij will affect the plastic analysis merely
due to the fact that the hydrostatic pressure remains constant. By obtaining the deviatoric
parts of each term in Eq. (3.78), the following expression is obtained:
However, using Eqs. (3.52) and (3.60):, the following can be derived in details for the last
term in Eq. (3.79):
3 S kln +1
Cijkl N kln +1 = ( 2GI ijkl
dev
+ K ij kl ) N kln +1 = ( 2GI ijkl
dev
+ K ij kl )
2 n +1
eq
n +1 n +1
dev 3 S kl 3 S kl
Cijkl N kln +1 = 2GI ijkl + K ij kl (3.80)
2 eq n +1
2 eqn +1
3 Sijn +1
Cijkl N kln +1 = 2G = 2GN ijn +1
2 eqn +1
Note that S kln +1 is deviatoric, therefore, I ijkl S kl = Sijn +1 and S kkn +1 = 0 (incompressible
dev n +1
plasticity). Substituting Eq. (3.80) back into Eq. (3.79), one obtains the following:
Sijn +1
S n +1
=S trial
2GN n +1
=S trial
3 G (3.81)
ij ij ij ij
eqn +1
64
Multiplying Sijn +1 (Eq. (3.81)) by itself gives:
Sijn +1Sitrial
+ 6 ( G )
2
S n +1 n +1
S =S trial
S trial
6 G j
ij ij ij ij
n +1
eq
S n +1 S n +1S n +1
Sijn +1Sijtrial = Sijn +1 Sijn +1 + 3 G ijn +1 = Sijn +1Sijn +1 + 3 G ij n +1ij
eq eq (3.83)
Sijn +1Sijtrial = Sijn +1Sijn +1 + 2 G eqn +1
S n +1 n +1
S =S trial
S trial
6 G
(S n +1 n +1
ij S ij + 2G eqn +1 )
+ 6 ( G )
2
ij ij ij ij
n +1
eq
3
Multiplying all the terms in Eq. (3.85) by and taking the square root of both sides
2
gives:
3 trial trial 3 n +1 n +1
Sij Sij + 6 G eqn +1 + 9 ( G )
2
Sij Sij =
2 2
eqtrial = ( ) n +1 2
eq + 2 ( 3G ) ( eqn +1 ) + ( 3 G )
2
(3.86)
eqtrial = eqn +1 + 3G
and by multiplying all terms of Eq. (3.81) by eqn +1 , and rearranging on obtains:
65
eqn +1Sijn +1 = eqn +1Sijtrial 3 GSijn +1
(3.87)
( n +1
eq + 3 G ) Sijn +1 = eqn +1Sijtrial
Note that the term between brackets in Eq. (3.87) is equal to eqtrial . This is shown in Eq.
(3.86). Therefore, Eq.(3.87) can now be written as:
3
and recalling that eqtrial , eqn +1 are scalars, multiplying both sides by and rearranging
2
terms, the above relation reduces to the following:
n +1 trial
3 Sij 3 Sij
= or N ijn +1 = N ijtrial (3.89)
2 eqn +1 2 eqtrial
which is the proof of the radial return using the von Mises yield criterion.
When the yield criterion, Eq. (3.55), is satisfied, i.e., eqn +1 = y + R( p n +1 ) , Eq. (3.86)
can be rewritten as:
eqtrial = eqn +1 + 3G = y + R( p n +1 ) + 3G
(3.90)
y + R( p n +1 ) + 3G eqtrial = 0
This equation can be used to solve for the plastic multiplier . For linear hardening, the
hardening equivalent stress R( p n +1 ) , which is a function of the equivalent plastic strain
p n +1 , can be linearly decomposed into the following:
R( p n +1 ) = kp n +1 = k ( p n + p ) = kp n + k p (3.91)
y + kp n + k p + 3 G eqtrial = 0 (3.92)
Rearranging the previous equation, and substituting for p (Eqs. (3.62) to (3.64)),
the following expression for the plastic multiplier is derived:
eqtrial y + kp n f trial
= = (3.93)
k + 3G k + 3G
66
Given n , np , pn and
Calculate a trial elastic stress ntrial
+1 = n + C
If f ( ntrial
+1 , Rn ) < 0 then exit integration scheme-End If
If f ( ntrial
+1 , Rn ) > 0 then
f trial
Calculate the plastic multiplier =
3G + k
Calculate the increment of plastic strain = N trial
p
N CN + k
End If
In FE nonlinear material analysis, and when the return mapping algorithm has
converged but the weak form of equilibrium, Eq. (3.1), is not satisfied, a new global
iteration is needed in order to propose new approximation to the nodal displacements
used to calculate the strain field at time tn +1 during the same time increment [tn , tn +1 ] .
When the Newton-Raphson method, Eq. (3.29), is used to iterate on the global level, the
so called consistent tangent operator has been reported to preserve a quadratic rate of
convergence, i.e., significantly faster convergence rate than that of the classical tangent
operator C ep . This results in the substitution of C ep calculated above by a type of
material consistent tangent operator or one of its approximations, in order to achieve
higher convergence rates. Simo and Taylor (1985) demonstrated the importance of the
consistent linearization in preserving a quadratic rate of convergence when using the
Newton-Raphsons method.
3.4.2.3 Other Forms of Strain Hardening: Nonlinear Strain Hardening and Linear
Strain Hardening Following Perfectly-Plastic Behavior
For nonlinear hardening, the breakdown of R( p n +1 ) , Eq. (3.91), is no longer valid, and
the above algorithms (Forward-Euler or Radial Return) need to be modified. Equations
(3.68) for Explicit and (3.90) for Radial Return, are now nonlinear functions of the
plastic multiplier . These equations are solved using a local Newton-Raphson
iteration scheme. The iteration scheme is termed local to distinguish it from the
gobal/equilibrium Newton-Raphson iterations.
67
The algorithms can be easily altered to account for a nonlinear isotropic hardening
rule, such as a power hardening law, R( p ) = kp n , where ( n ) is a material constant that
can be calibrated to match experimental results. The expressions given by the algorithms
for , i.e. Eq. (3.69) for Explicit and Eq. (3.93) for Radial Return, are replaced by the
following expressions given as:
2GN ij ij f trial
= and = (3.94)a,b
R R
3G + 3G +
p p
R
where = n k p n 1 is no longer a constant. This means that the above equations will
p
have to be solved iteratively for using a nonlinear equation solver scheme. The
scheme selected here is the Newton-Raphson iterative procedure. Once the plastic
multipliers (for both algorithms) are obtained through nonlinear iterations, the rest of
the algorithms proceed as shown above for the case of linear hardening except for the
elastoplastic tangent operators C ep , where the linear hardening parameter k in the
R
expressions for C ep needs to be replaced with .
p
Beyond the yield point, perfect plasticity dominates the analysis until the onset of
strain hardening. At that point and beyond, the strain increment is divided into elastic and
plastic parts, where the elastic part is used to update the stress. The issue here is to
determine the point at which the strain hardening starts. In a one dimensional analysis,
the onset of strain hardening is the experimentally obtained strain value. However, in a
two dimensional analysis, a more general term has to be developed to determine the onset
of hardening. In this work, the maximum strain criterion is used to define a value for the
maximum principal strain in steel, where the latter is used as an indicator of whether the
onset of hardening has been reached or not through comparing it to the experimental
value of strain at the onset of hardening. This is accomplished through adding IF
statements to the UMAT FORTRAN file after the yield condition to determine when the
stress is updated. The additional IF statements acquire the value of the maximum
68
principal strain and allow hardening and stress update to take place if the maximum strain
criterion is satisfied.
Given n , np , pn and
If f ( n , Rn ) < 0 then exit the integration scheme-End If
If f ( n , Rn ) > 0 then
Update the strain tensor n +1 = n +
Calculate the maximum principal strain max
p
If max
p
uniaxial strain hardening then allow hardening
Else use perfectly plastic analysis
End If
N nC
Calculate the plastic multiplier =
N nCN n + k
Calculate the increment of plastic strain p = N n
Calculate the increment of elastic strain using e = p
Calculate the increment of stress = C e
Update all the quantities
Stresses n +1 = n +
Strains np+1 = np + p
Equivalent plastic strain pn +1 = pn +
CN n N nC
Calculate the elastic-plastic tangent operator C ep = C
N nCN n + k
End If
The two integration schemes discussed above were implemented into the FE
commercial code ABAQUS where their applicability, with different post yield behaviors,
was verified. The Implicit code (ABAQUS Standard) will be used. In such an Implicit
code, provision of both the integration scheme of the plasticity constitutive equations
(whether Implicit or Explicit) along with the material tangent operator/matrix is
necessary. These will be written using FORTRAN into a material behavior subroutine
called UMAT. This UMAT is then linked to ABAQUS during the execution of the
command that runs the analysis using the data stored in the input file. While ABAQUS
performs the standard FE procedure along with the nonlinear global/equilibrium
iterations, the UMAT file will govern the behavior of the materials during different
loading stages; elastic and plastic.
69
Plane stress
Steel plate 0.2x0.2x0.02m3
Left edge is pinned
Right edge subjected to uniaxial
displacement control of 0.01m applied
through different time increments.
y = 414 MPa, u = 720 MPa
E = 200 GPa , v = 0.3
k = 5.3 GPa
Figure 3.13 Schematic diagram of a steel plate modeled using plane stress elements
A wide range of information and parameters are passed in and out the subroutine
UMAT during FE analysis at the beginning and end of each time increment. Data passed
into UMAT at the beginning of the time increment would include, but is not limited to,
the values of stresses, strains, state variables as well as the strain increments. Data passed
back to ABAQUS would include the updated stresses, state variables and material
tangent operator (the Jacobian). The Jacobian is required for the global iterative
procedure used to minimize the force residual. If convergence occurs after a given
number of iterations, the type of the Jacobian does not influence the accuracy of the
solution as much as it influences the rate of convergence (Dunne and Petrinic, 2005).
The verification of the UMAT file is carried out under the plane stress condition
shown in Fig. 3.13. The plate is subjected to uniaxial displacement control at one of the
vertical ends while pinned at the other. Four ABAQUS CPS4R (4 nodded Continuum
Plane Stress Reduced Integration) elements are used. The analytical solution to this
problem provides guidelines that help in assessing the applicability of the UMAT
subroutines. Using the parameters given in Fig. 3.13, the strain at yield will be equal to
y = 0.00207.
At the beginning of each time increment, ABAQUS passes the stress, strain and strain
increment into UMAT. These are all passed in vector form, which can then be transferred
into tensors using proper functions/subroutines in UMAT file. The material model is then
used to update the stresses and return them back to ABAQUS in vector format. The
tangent operator is returned in the form of a matrix.
The results obtained by running the problem shown above with different UMAT files
are shown below for the case of linear hardening first, followed by other hardening laws.
The number of increments used is indicated on each figure.
Decreasing the size of the total strain increment 1 (Fig. 3.11) - causing the stress to
cross the yield surface for the first time - will improve the results of the Explicit
integration scheme (Fig. 3.14), eliminating the need for a correction procedure. The
correction applied to the Explicit integration will have a pronounced effect at a lesser
70
number of time increments (Fig. 3.15) rather than at high number of increments where
the correction would be rendered unnecessary.
800
700
600
Stress , M Pa
500
400 Explicit 100 Inc.
300 Explicit 1000 Inc.
200
100
0
0 0.01 0.02 0.03 0.04 0.05 0.06
Total Strain
Figure 3.14 Explicit integration scheme using 100 and 1000 increments
800
700
600
Stress , M Pa
500
400
Corrected Explicit 100 Inc.
300
200
Standard Explicit 100 Inc.
100
0
0 0.01 0.02 0.03 0.04 0.05 0.06
Total Strain
As was discussed earlier, the Implicit integration scheme requires lesser amount of
increments to give good results than the Explicit integration scheme. The differences
between the 100 and 1000 increments curves shown below, Fig. 3.16, for the Implicit
integration scheme are the smoothness of each line, and the smoothness of the region of
transition from elastic to plastic behavior.
Fig. 3.17 shows the difference between Explicit and Implicit integrations at different
time increments. It shows the difference between the Implicit and Explicit integrations at
100 increments. It also shows that Implicit and Explicit integrations will eventually give
the same result at increased numbers of time increments.
71
Implicit Integration Scheme
800
700
600
Stress , M Pa
500
400 Implicit 100 Inc.
300 Implicit 1000 Inc.
200
100
0
0 0.01 0.02 0.03 0.04 0.05 0.06
Total Strain
Figure 3.16 Explicit integration scheme using 100 and 1000 increments
800
700
600
Stress , M Pa
500
Explicit 100 Inc.
400
Implicit 100 Inc.
300
200 Implicit 1000 Inc.
If the linear hardening rule is replaced by a nonlinear hardening rule, such as a power
hardening law, R( p) = kp n , the following figures (Fig. 3.19 and Fig. 3.20) can be
obtained. One should note the difference in the values of the yield stress between the
Explicit and Implicit integrations at 100 increments. Reducing the size of these
increments results in better agreement between the two integration schemes. The Explicit
integration results can be further enhanced by incorporating the correction procedure
discussed above.
72
Numerical .vs. Theoretical Stress-Strain Curves
800
700
600
Stress, M Pa
500
Theoretical
400 Implicit 1000 inc.
300 Explicit 1000 inc.
200 Implicit 100 inc.
100 Explicit 100 inc.
0
0 0.05 0.1 0.15
Strain
Figure 3.18 Numerical .vs. theoretical stress strain curve for Reinforcing steel
1500
n = 0.9
Stress , M Pa
1000 n = 0.8
n = 0.6
500 n = 0.5
0
0 0.01 0.02 0.03 0.04 0.05 0.06
Total Strain
It is worthy to note that as the power (n) increases from 0.5 to 0.9, the solution
approaches that of a linear hardening law. Substituting unity for the value of (n) is not
admissible. In addition, increasing the number of increments to 1000 will give identical
results for both integration schemes, as shown in Fig. 3.21.
A UMAT file was also used to produce an elastic perfectly plastic behavior where the
plastic slip in the reinforcing steel equals the applied strain rate which renders the
increment of the stress to be equal to zero, resulting with a plateau beyond yielding, see
Fig. 3.22. As discussed earlier, this algorithm neglects the strain hardening effect of the
steel and therefore, might not be suitable for large deformation analyses such as seismic
analysis, etc.
73
Implicit Integration 100 inc.
1600
1400
1200 n = 0.9
Stress , M Pa 1000 n = 0.8
800
n = 0.6
600
400 n = 0.5
200
0
0 0.01 0.02 0.03 0.04 0.05 0.06
Total Strain
1600
1400
Ex n = 0.9
1200
Stress , M Pa
Ex n = 0.8
1000
800 Ex n = 0.6
600 Ex n = 0.5
400 Im n = 0.9
200 Im n = 0.8
0 Im n = 0.6
0 0.01 0.02 0.03 0.04 0.05 0.06
Im n = 0.5
Total Strain
500
400
Stess, MPa
100
0
0 0.01 0.02 0.03 0.04 0.05 0.06
Total Strain
74
Elastic - Perfectly Plastic - Linear Hardening
700
600
Stress, MPa
500
400
300
200 Corrected Explicit Full Range 100 inc.
100 Implicit 100 Full Range 100 inc.
0
0 0.01 0.02 0.03 0.04 0.05 0.06
Total Strain
Another modification was applied to the UMAT file in order to obtain a refined
description of the stress-strain behavior of reinforcing steel (Fig. 3.23). Here the analysis
beyond the yield point presumes under the perfectly plastic condition, until the point
where the strain reaches the strain hardening value of the reinforcing steel, where
hardening behavior starts. The analysis changes from a perfectly plastic to a plastic
hardening analysis leading to the stress-strain curve shown in Fig. 3.23. It should be
mentioned here that this algorithm is sensitive to increment size and sometimes it leads to
divergence of the solution. Further work is needed in order to enhance the performance of
such an algorithm and to substitute the linear hardening stage with nonlinear hardening to
better describe the steel behavior.
75
CHAPTER 4
4.1 Introduction
The analysis and design of a concrete structure requires prior knowledge of its
mechanical properties. When continuum mechanics is considered, elastic damage models
or elastic plastic constitutive laws are generally the standard approaches to describe the
behavior of concrete. In the first case, the mechanical effect of the progressive
microcracking and strain softening are represented by a set of internal state variables
which act on the elastic behavior (decrease of the stiffness) at the macroscopic level (e.g.
Mazars, 1984; Simu and Ju, 1987a,b; Mazars and Pijaudier-Cabot, 1989; Labadi and
Hannachi, 2005; Tao and Phillips, 2005; Junior and Venturini, 2007; Khan et. al., 2007).
In plasticity models, softening is directly included in the expression of a plastic yield
surface by means of a hardeningsoftening function (e.g. Feenstra and de Borst, 1996;
Bicanic and Pearce, 1996; Grassl et. al., 2002; Park and Kim, 2005).
There are several possibilities for coupling plasticity and damage effects in a single
constitutive relation. Historically, damage has first been coupled to plasticity (Lemaitre
and Chaboche, 1984) in the so-called ductile failure approaches for metal alloys. The
underlying assumption was that void nucleation is triggered by plastic strains.
Applications to concrete were proposed among others, (e.g. Oller et. al., 1990; Voyiadjis
and Abu-Lebdeh, 1993, 1994; Abu-Lebdeh and Voyiadjis, 1993; Kratzig and Polling,
2004). In these models, damage growth is a function of the plastic strains. There is a
difficulty, however. In uniaxial tension there is little plasticity and quite a lot of damage
while in uniaxial compression, the picture is reversed with little damage and important
plastic strains. Furthermore, it can be hardly explained how plastic strain may develop in
concrete prior to microcracking. A common assumption is that irreversible strains are due
to microcrack sliding and internal friction. Such a process requires the prior formation of
internal surfaces (microcracks).
The second approach, that is more suited to both tension and compression responses,
uses the effective stress. The plastic yield function is written in the effective
configuration pertaining to the stresses in the undamaged material. Many authors (Simo
76
and Ju, 1987a,b; Ju, 1989; Mazars and Pijaudier-Cabot, 1989; Yazdani and Schreyer,
1990; Hansen and Schreyer, 1992; Lee and Fenves, 1998; Faria et. al., 1998; Fichant et.
al., 1999; Voyiadjis and Kattan, 1999, 2006; Jefferson, 2003; Salari et. al., 2004; Shen et.
al., 2004; Jason et. al., 2006, Voyiadjis et. al., 2008b) applied this approach to isotropic
and anisotropic damage coupled to elasto-plasticity. It has been extended to other sources
of damage, for instance to thermal damage by Nechnech et. al. (2002) and Willam et. al.
(2002).
A last possibility is what could be called the strong coupled approach. As opposed to
the above where the plastic yield function is written in term of the effective stress, the
applied stress appears in the plastic process, which becomes coupled to damage. Luccioni
et. al. (1996), Armero and Oller (2000) and Voyiadjis et. al. (2008a) provided the
thermodynamic consistent backgrounds of such a model.
The effective stress approach has been chosen because it provides a simple way to
separate the damage and plastic processes. Plastic effects, driven by the effective stresses,
can be described independently from damage ones and vice versa. One of the main
interests is to ease the numerical implementation which is Implicit/Explicit. The plastic
part is Implicit and the damage part is Explicit, same as in classical continuum damage
computations. As a consequence, existing robust algorithms for integrating the
constitutive relations can be implemented. The calibration of the material parameters is
also easier to handle as a consequence of the separation of damage and plasticity
processes.
In this contribution, the damage process is (elastic) strain controlled. The isotropic
damage model proposed by Tao and Phillips (2005) will be adopted here to describe the
damage behavior of concrete. While the Tao and Phillips (2005) model incorporated
strain-softening in an elastic-damage framework, it is used in this work simultaneously
with the effective stress space plasticity in order to describe the damage and irreversible
phenomena in concrete materials under tension and compression. The plastic process
shall be described using a yield function inspired from Lubliner et. al. (1989) and later
modified by Lee and Fenves (1998) and Wu et. al. (2006). It is an isotropic hardening
process in the present model. Softening is controlled by damage, while plasticity controls
hardening, in tension and compression. Fracture energy related coefficients have been
defined and incorporated in order to achieve a reasonable degree of discretization
insensitivity in numerical calculations (Feenstra and de Borst, 1996; Lee and Fenves,
1998; Wu et. al., 2006).
77
In the following sections, the general framework of the elastic plastic damage model is
discussed first, followed by a consistent thermodynamic formulation. The Helmholtz free
energy function is then introduced with specific forms to be used in the FE code. The
effective stress space plasticity is then introduced followed by the discussion of the
damage yield criteria and evolution laws for concrete under tension and compression.
Elements of validation and comparisons between the model and existing damage and
damage plasticity approaches are also presented. The behavior of the model is then tested
using four elementary loading cases: simple tension, simple compression, biaxial tension,
and biaxial compression. The model is also used to reproduce the load capacity of a
notched beam subjected to three point bending test.
The model presented in this work is thermodynamically consistent and comes from a
generalization of the effective space plasticity theory and isotropic damage theory applied
simultaneously under the assumptions of small strains and isothermal conditions.
The transformation from the effective (undamaged) configuration to the damaged one
can be done by utilizing the strain equivalence hypothesis, which basically states that the
strains in the undamaged (effective) configuration are equal to the strains in the damaged
configuration. This hypothesis is commonly applied to the coupling of plasticity and
continuum damage mechancis (Lemaitre and Chaboche, 1990; Steinmann et. al., 1994;
Lammer and Tsakmakis, 2000; Menzel et. al., 2005; Voyiadjis and Kattan, 2006). Using
the additive decomposition of the total strain tensor ij (= ij ) into elastic (reversible) ije
(= ije ) and plastic (irreversible) ijp (= ijp ) strain tensors, along with the strain equivalence
hypothesis, the following arrangements can be assumed:
ij = ije + ijp
p
ij = ij (4.1)
ij = ij + ij
e
The equivalence of the elastic strains will be used to obtain an expression for the
elasticity tensor Eijkl ( ) in the damaged configuration as well as the damage
thermodynamic conjugate forces Y used in the damage yield criteria g . The
equivalence of the plastic strains, on the other hand, will be justified through the use of
the effective stress space plasticity (Ju, 1989) and the definition of the plastic Helmholtz
free energy function. Both equivalences will be discussed later in this chapter.
By taking the time derivative of the arrangements in Eq. (4.1), the following strain rate
equations necessary for the plastic-damage incremental procedure are obtained:
78
The effective stress tensor (stresses in the undamaged configuration) can now be
written in terms of the strain equivalence hypothesis and using Hooks law as:
dev
where I ijkl = I ijkl 13 ij kl is the deviatoric part of the fourth-order identity tensor
I ijkl = 12 ( ik jl + il jk ) , and G and K are the effective shear and bulk moduli,
respectively. The tensor ij is the Kronecker delta, and is equal to one, ij = 1 when i = j
or zero, ij = 0 when i j .
Taking the time derivative of Eq. (4.3), the rate of the constitutive equation in the
effective configuration can be obtained as:
The damage configuration counterpart of Eq. (4.3), i.e. the stress tensor for the
damaged material, can be written as follows:
The stress-strain behavior is affected by the development of micro and macro cracks in
the material body. Concrete contains a large number of micro cracks, especially at
interfaces between coarse aggregates and mortar, even before the application of external
loads. These initial microcracks are caused by segregation, shrinkage, or thermal
expansion in the cement paste. Under applied loading, further micro-cracking may occur
at the aggregate-cement paste interface, which is the weakest link in the composite
system. These microcracks which are initially small (invisible), will eventually lead to
visible cracks that extend as the applied external loads are increased. These cracks
contribute to the generally obtained nonlinear stress-strain behavior. Since a
phenomenological continuum approach is followed in this work, these effects are
smeared out (i.e. averaged) throughout the body where the material is considered as a
mechanical continuum with degraded (damaged) properties.
In this work, the stress-strain relation involves a scalar (isotropic) damage variable
which is a weighted average of the tensile and compressive damage scalar variables, +
79
and . It was shown in Chapter 2 that the Cauchy stress tensor ij is related to the
effective stress tensor ij by:
ij = (1 ) ij (4.7)
By substituting Eqs. (4.3) and (4.6) into Eq. (4.7) one obtains the following relations:
The expression for the fourth order elasticity tensor in the damaged configuration Eijkl in
terms of its effective counterpart Eijkl will also be derived from the elastic dissipation
potential later on. The damage variable has values from zero to one. The value = 0
corresponds to the undamaged (effective) material and the value = 1 corresponds to the
fully damaged material. Damage associated with the failure mechanisms of the concrete
(cracking and crushing) results in a reduction in the elastic stiffness (Eq. (4.8)). Within
the context of the scalar-damage theory, the stiffness degradation is isotropic (i.e. the
same damage evolution is assumed in different directions) and represented by a single
degradation value . The time derivative can now be applied to Eq. (4.9) to obtain the
following:
E e
ij = (1 ) Eijkl kle (4.10)
ijkl kl
which represents the constitutive relation for the elastic-plastic-damage model used in
this work.
80
on to derive the evolution equations for the proposed model. The constitutive equations
are derived from the second law of thermodynamics, the expression of Helmholtz free
energy, the additive decomposition of the total strain rate into elastic and plastic
components, the Clausius-Duhem inequality, and the maximum dissipation principle.
The Helmholtz free energy can be expressed in terms of a suitable set of internal state
variables that characterizes the elastic, plastic, and damage behaviors of concrete. In this
work, the following internal state variables are assumed to satisfactorily characterize the
behavior of concrete both in tension and compression: the elastic strain tensor ije , a set
of plastic hardening variables ( + , ) defined as the equivalent plastic strains in tension
and compression, respectively, and the scalar damage variables ( + and ) representing
the damage densities in the material under tension or compression, respectively, such
that:
= ( ije , + , , + , ) (4.11)
= e ( ije , + , ) + p ( + , ) (4.12)
It can be noted from the above decomposition that damage affects only the elastic
properties and not the plastic ones. This can be justified by the following: once micro-
cracks are initiated during loading of a concrete material, local stresses are redistributed
to undamaged material micro-bonds over the effective area ( A shown in Fig. 2.13,
Chapter 2). Thus, effective stresses of undamaged material points are higher than nominal
stresses. Accordingly, it appears reasonable to state that the plastic flow occurs only in
the undamaged material micro-bounds by means of effective quantities (Ju, 1989). The
plastic response is therefore characterized in the effective stress space and the yield
function is no longer written in term of the applied stress, rather, it is a function of the
effective stress, i.e., the stress in the undamaged material in between the microcracks.
This approach, which is more suited for brittle materials like concrete, has been
extensively used by researchers (Simo and Ju, 1987a,b; Ju, 1989; Mazars and Pijaudier-
Cabot, 1989; Yazdani and Schreyer, 1990; Hansen and Schreyer, 1992; Lee and Fenves,
1998; Faria et. al., 1998; Fichant et. al., 1999; Salari et. al., 2004; Jefferson, 2003; Jason
et. al., 2006; and others).
In the following, the thermodynamic conjugate forces associated with the internal
state variables in Eq. (4.12) are derived based on the second law of thermodynamics. For
81
isothermal behavior, the second-law of thermodynamics states that the rate of change in
the internal energy is less than or equal to the external expenditure of power such that:
dv P
v
ext (4.13)
where Pext is the external power which according to the principle of virtual power should
be equal to the internal power such that:
Substituting Eq. (4.14) into Eq. (4.13), one obtains the following:
dv dv 0
v v
ij ij ( ) dv 0
v
ij ij (4.15)
In a stepwise sense, the Clausius-Duhem inequality can be inferred from Eq. (4.15) as
follows:
ij ij 0 (4.16)
Taking the time derivative of Eq. (4.12), the following expression can be written:
e e e + e p + p
= + = e ij + + + +
e p
+ (4.17)
ij +
By substituting the rate of the Helmholtz free energy density, Eq. (4.17), into the
Clausius-Duhem inequality, Eq. (4.16), one can write the following relation:
e e e + e p + p
ijijp + ij
ij
0 (4.18)
ije + +
The above equation is valid for any admissible internal state variable such that the
Cauchy stress tensor can be define as:
e
ij = e (4.19)
ij
ijijp + Y + + + Y c + + c 0 (4.20)
82
where the damage and plasticity conjugate forces that appear in the above expression are
defined as follows:
e
Y + = (4.21)
+
e
Y = (4.22)
p
c =
+
(4.23)
+
p
c =
(4.24)
= p + d 0 (4.25)
The plasticity and damage dissipation potentials are given, respectively, as follows:
p = ijijp c + + c 0 (4.26)
d = Y + + + Y 0 (4.27)
The rate of the internal variables associated with plastic and damage deformations are
obtained by utilizing the calculus of functions of several variables with the plasticity and
damage Lagrange multipliers p and d , respectively. Thus the following general
objective function can be defined:
= p F d+ g + d g 0 (4.28)
where F and g are the plastic potential function and the tensile and compressive
damage potential functions, respectively, to be defined later.
Use is now made of the well known maximum dissipation principle (Simo and
Honein, 1990; Simo and Hughes, 1998), which describes the actual state of the
thermodynamic forces ( ij , Y , c ) as the state that maximizes the dissipation function
over all other possible admissible states. Hence, one can maximize the objective function
by using the following necessary conditions:
83
= 0, = 0, =0 (4.29)
ij Y
c
Substituting Eq. (4.28) into Eqs. (4.29) along with Eqs. (4.26) and (4.27) yields the
following corresponding thermodynamic laws:
F p
ijp = p (4.30)
ij
+
+ g
= d
+
(4.31)
Y +
g
= d (4.32)
Y
F
+ = p + (4.33)
c
F
= p (4.34)
c
Note that Eq. (4.30) is defined in terms of a plastic potential F p different from F to
indicate the use of a non-associative flow rule. It is also worthy to note that in this work,
the damage criteria are characterized with scalar quantities scalar thermodynamic
conjugate forces and scalar damage parameter - therefore, the above general
thermodynamic evolution laws for damage, Eqs. (4.31) and (4.32), will be greatly
simplified in the implementation procedure.
Based on the additive decomposition of the Helmholtz free energy function into
elastic-damage and plastic parts discussed earlier, Eq. (4.12), this section introduces
specific forms for the elastic-damage and plastic parts of the Helmholtz free energy
function adopted in this work. The elastic-damage part of the Helmholtz free energy
function will be defined first, followed by a definition for the plastic part.
In order to define the elastic-damage part of the Helmholtz free energy, the
spectral decomposition procedure for obtaining the tensile and compressive parts of the
Cauchy stress tensor, as well as the combined scalar damage variable, , are to be
defined first.
84
(e.g. Ortiz, 1985; Ju, 1989; Lubliner et. al., 1989; Faria et. al., 1998; Lee and Fenves,
1998; Wu et. al., 2006) such that:
ij = ij+ + ij (4.35)
The total effective stress tensor ij can be written in terms of its principal values ( k )
and their corresponding principal directions ni( k ) ( k =1, 2, 3) as follows:
3
ij = ( k ) ni( k ) n(jk ) = (1) ni(1) n (1)
j +
(2) ni(2) n(2)
j +
(3) ni(3) n(3)
j (4.36)
k =1
The positive part ij+ can be obtained by considering only the tensile principal
values as follows:
3
ij+ = H ( ( k ) ) ( k ) ni( k ) n (jk ) (4.37)
k =1
where H is the Heaviside step function (H = 1 for max > 0 and H = 0 for max < 0 ).
The principal stresses ( k ) in Eqs. (4.36) and (4.37) are defined in the following form:
By substituting Eq. (4.38) into Eq. (4.37), the tensile stress can be written as:
3
ij+ = H ( ( k ) )n(pk ) pq nq( k ) ni( k ) n(jk ) (4.39)
k =1
ij+ = Pijpq
+
pq (4.40)
where
3
P +
ijpq = H ( ( k ) )ni( k ) n (jk ) n (pk ) nq( k ) (4.41)
k =1
and substituting Eq. (4.40) into Eq. (4.35), the following expressions are obtained:
85
ij = Pijpq
+
pq + ij
ij = ij Pijpq
+
pq = I ijpq Pijpq
+
pq = Pijpq
pq (4.42)
+
I ijpq = Pijpq + Pijpq
+
where Pijpq and Pijpq are the tensile and compressive fourth-order projection tensors,
respectively. A numerical procedure is provided in Appendix A to illustrate the spectral
decomposition concept using the software Maple.
Next, the combined scalar damage variable is defined here similar to that defined by
(Tao and Phillips, 2005) as follows:
c + f + + c f
de ij hg ed ij gh
= (4.43)
c f
ed ij gh
where + and are the tensile and compressive damage crack densities, respectively,
ij+ and ij are the positive and negative spectral decomposition parts of the Cauchy
stress tensor, ij , and cde X ij fgh represents the scalar contraction of the second order tensor,
i.e., cde X ij fgh = X ij X ij . This definition implies that damage under uniaxial loading is
governed by the corresponding damage parameter, while under bi-axial loading two
damage parameters, + and , both contribute to the induced damage. The effective
contribution is in proportion to the ratio of positive and negative part contractions to the
total stress contraction. This definition is slightly different than that given by Tao and
Phillips (2005) where they avoided the decomposition of the stress tensor into positive
and negative parts by separating the principal values of the stress tensor into positive and
negative parts to simplify the implementation in an FE code.
1 1
e = ije Eijkl kle = ij ije (4.44)
2 2
In order to account for the stiffness degradation induced by the concrete material
damage, the elastic free energy in the damaged configuration can be written in terms of
the elastic strain equivalence hypothesis as follows:
1 1 1
e = (1 ) e = ije Eijkl ( ) kle = (1 ) ije Eijkl kle = ij ije (4.45)
2 2 2
86
The previous equation can be substituted into Eq. (4.19) to give the constitutive stress
strain relation in the damaged configuration, Eq. (4.9). It also shows that the elasticity
tensor in the damaged configuration Eijkl () is given in terms of the elasticity tensor in
the effective configuration Eijkl as shown in Eq. (4.8).
1 1 1
e = (1 )(eije + mm
e
ij ) Eijkl (ekle + nne kl ) (4.46)
2 3 3
In the above equation, the elastic strain tensor, uve , has been additively decomposed
1
into deviatoric, euve , and hydrostatic, epp , parts such that:
3
1
uve = euve + pp
e
uv (4.47)
3
1 1 1
e = (1 )(eije Eijkl ekle + nne eije Eijkl kl + mm
e
ij Eijkl ekle
2 3 3
(4.48)
1 e 2
+ ( mm ) ij Eijkl kl )
9
The term involving pure hydrostatic strains can be isolated from the rest of the terms
as follows:
1 1 1
e = (1 )(eije Eijkl ekle + nne eije Eijkl kl + mm
e
ij Eijkl ekle )
2 3 3
(4.49)
1 1 e 2
+ (1 )( ( mm ) ij Eijkl kl )
2 9
To reduce the susceptibility of the hydrostatic part to damage, Tao and Phillips (2005)
used a damage multiplier in the term involving pure hydrostatic strain,
1
(1 )( mm
e
) 2 ij Eijkl kl , as follows:
18
87
1 1 1
e = (1 )(eije Eijkl ekle + nne eije Eijkl kl + mm
e
ij Eijkl ekle )
2 3 3
(4.50)
1 1 e 2
+ (1 )( ( mm ) ij Eijkl kl )
2 9
Considering the last term in the previous equation, the following manipulation can be
performed in order to reach an objective result:
1 1 e 2 1
Zero
1 e 2
(1 )( ( mm ) ij Eijkl kl ) = (1 + )( ( mm ) ij Eijkl kl )
2 9 2 9
1 1 1 e 2
= (1 ) + ( ) ( ( mm ) ij Eijkl kl ) (4.51)
2 2 9
1 1 1 e 2
= (1 ) + (1 ) ( ( mm ) ij Eijkl kl )
2 2 9
Substituting the final term in the above equation to Eq. (4.50), one obtains the
following relation:
1 1 1 1
e = (1 )(eije Eijkl ekle + nne eije Eijkl kl + mm
e
ij Eijkl ekle + ( mm
e
) 2 ij Eijkl kl )
2 3 3 9
(4.52)
1 1 e 2
+ (1 ) ( ( mm ) ij Eijkl kl )
2 9
Comparing the above equation with Eq. (4.48), it can be seen that the first term on the
right hand side of Eq. (4.52) is identical to the right hand side of Eq. (4.48), therefore, the
elastic free energy function can be now written in terms of the total strain tensor and the
hydrostatic strain tensor with the effect reduction factor, , as follows:
1 1 1
e = (1 ) ije Eijkl kle + (1 ) ( ( mm
e
) 2 ij Eijkl kl ) (4.53)
2 2 9
Based on the failure characteristics of concrete and the experimental fact that the
effect of the hydrostatic strain component on damage is less than that of the deviatoric
component, Tao and Phillips (2005) designed the damage multiplier (or damage
reduction factor) to provide this effect reduction. Clearly, is less than or equal to
one ( 0 1 ). For the uniaxial version, they defined the damage multiplier as the
ratio of the average stress mm / 3 to the maximum principal stress 1 , i.e., ( mm / 3) / 1 .
Whereas, under bi-axial loadings, the response of concrete is dependent on the stress
ratio, and since a relationship between damage and stress ratios is not straightforward to
establish and due to the fact that whatever the stress ratios are, it is with no doubt that
material damage is the consequence of energy dissipation from the damage mechanics
point of view, consequently, the bi-axial version of the damage multiplier is proposed to
88
be a damage energy release rate, Y , dependent parameter (Tao and Phillips, 2005).
Different mathematical forms of the damage multiplier can be assumed as long as they
can match the corresponding experimental data. Tao and Phillips (2005) adopted the
following form of the damage reduction factor :
1
= 1 (4.54)
1 + cY exp(dY )
where exp is the base of natural logarithms, and c and d can be regarded as two
material constants to make dimensionless and be so determined as to match the
experimental data. It is worth mentioning here that this form of the damage reduction
factor introduces nonlinearity to the definition of the damage energy release rate (damage
conjugate force) Y , and requires local iterations when solving for that damage release
rate during a given strain increment.
Based on the thermodynamic framework, one can obtain expressions for the damage
thermodynamic conjugate forces Y + and Y from Eqs. (4.53), (4.21), and (4.22) in the
following form:
e 1 ij e
+
1
Y =
+
= ij Eijkl kl (1 )( mm ) ij Eijkl kl
e e 2
(4.55)
+
2 ij 9
e 1 ij e
1
Y =
= ij Eijkl kl (1 )( mm ) ij Eijkl kl
e e 2
(4.56)
2 ij 9
Since the magnitudes of the damage energy release rates Y + and Y are measures of
how susceptible the material is to damage, the damage energy release rates Y are
therefore used to define the damage criteria g in tension or compression.
The plastic part of the Helmholtz free energy is postulated to be a function of the
plastic variables + and in the following form:
1 1
p = f 0+ + + h( + ) 2 + f 0 + Q + exp( ) (4.57)
2
where f 0+ and f 0 are the uniaxial tensile and compressive yield stresses, respectively.
The hardening parameters + and are introduced as the equivalent plastic strains
under tension and compression, respectively, defined as:
t
+ = + dt (4.58)
0
89
t
= dt (4.59)
0
where + and are the tensile and compressive equivalent plastic strain rates,
respectively, which are assumed to be evaluated according to the following expressions
(Lee and Fenves, 1998):
+ = r ( i )max
p
(4.60)
= (1 r ( i ))
p
min (4.61)
where max
p
and min
p
are the maximum and minimum eigenvalues of the plastic strain rate
tensor p such that p > p > p and p = p and p = p . It should be mentioned
ij 1 2 3 max 1 min 3
here that the procedure for obtaining the eigenvalues of a second-order tensor (e.g. stress
or strain) is a built-in function in ABAQUS readily available for the UMAT subroutine.
Under uniaxial loading, these eigenvalues reduce to max
p
= 11p in tension and min
p
= 33p in
compression. The dimensionless parameter r ( i ) is a weight factor 0 r ( i ) 1
depending on the effective principal stresses i ( i = 1, 2,3 ) and is defined as follows (Lee
and Fenves, 1998):
i
r ( i ) = i =1
3
(4.62)
i
i =1
Substituting Eq. (4.57) into Eqs. (4.23) and (4.24) yields the following expressions for
the plasticity conjugate forces c + and c (the tensile and compressive hardening
functions):
p
c+ = = f 0+ + h + (4.63)
+
and
p
c = = f 0 + Q 1 exp( ) (4.64)
90
such that by taking the time derivative of the above two expressions, one can easily
obtain the following evolution equations of the hardening functions c + and c in terms
of the plastic internal state variables + and :
c + = h + (4.65)
c = (Q ci ) (4.66)
Equation (4.66) can be obtained by realizing that the exponential term, exp( ) , in
the rate equation, c = Q ( exp( ) ) , can be replaced by the following term,
c f 0 c
exp( ) = 1 = 1 i , obtained from rearranging Eq. (4.64).
Q Q
In this section, the effective stress space plasticity and its components will be
discussed. Owing to the coupling between the damage evolutions and the plastic flow in
the elastic plastic damage models, the so-called (effective stress space plasticity) was
introduced by Ju (1989). In this approach the effective (undamaged) configuration is
resorted in order to establish the evolution laws for the plastic strains governing the
plastic irreversible behavior in the material (Wu et. al., 2006). To determine the required
effective stress tensor ij , the evolution law for the irreversible plastic strains tensor ijp
has to be established first. The additive decomposition of the effective total strain tensor
into elastic and plastic parts is assumed, Eq. (4.1). An effective stress plasticity yield
criterion with multiple hardening rules is used along with a non-associative flow rule.
Both take into account the dilatation effect of concrete materials. A Kuhn-Tucker
consistency condition is applied to obtain the evolution of the magnitude of plastic
strains.
A crucial component of any material model that involves plasticity theory is the yield
surface/criterion. This criterion should address and model the experimentally observed
non-symmetrical behavior of concrete under tensile and compressive loadings. Assuming
the same yield behavior for both tension and compression in concrete materials leads to
over/under estimation of plastic deformations (Lubliner et. al., 1989). The yield criterion
adopted in this work was first introduced in the Barcelona model by Lubliner et. al.
(1989), and later modified by Lee and Fenves (1998, 2001) and Wu et. al. (2006). These
works reported that the yield criterion is successful in simulating the concrete behavior
under uniaxial, biaxial, multiaxial, and cyclic loading. This criterion is given in the
effective stress space and expressed using the undamaged configuration parameters as
follows:
91
where J 2 = sij sij / 2 is the second-invariant of the effective deviatoric stress
sij = ij kk ij / 3 , I1 = kk is the first-invariant of the effective stress ij , denote a
suitable set of plastic variables (Wu et. al., 2006) given as the equivalent plastic strains
defined in Eqs. (4.60) and (4.61), H( max ) is the Heaviside step function defined in
Eq.(4.37), and max is the maximum principal stress.
( f b0 / f 0 ) 1
= (4.68)
2( fb0 / f 0 ) 1
and the parameter , defined as a constant in the Barcelona model, was later modified
by Lee and Fenves (1998), and given as a dimensionless function of the tensile and
compressive cohesions c (hardening internal state variables) in the following form:
c ( )
( ) = (1 ) + + (1 + )
(4.69)
c ( )
where f b0 and f 0 are the initial equibiaxial and uniaxial compressive yield stresses,
respectively. Experimental values of the ratio fb0 / f 0 lie between 1.10 1.20 (Wu et. al.,
2006); yielding to be between 0.08 0.14. For further details about the derivation of
both parameters, and , the reader is referred to Lubliner et. al. (1989).
c ( ) = f 0 + Q 1 exp( ) (4.70)
where Q and are material constants characterizing the saturated stress and the rate of
saturation, respectively. On the other hand, a linear expression is assumed for the tensile
hardening function c + such that:
c + ( + ) = f 0+ + h + (4.71)
92
where h is a material constant obtained from the uniaxial tensile stress-strain diagram.
The evolution equations of the hardening parameters were shown in Section 4.4.2, Eqs.
(4.65) and (4.66).
The flow rule gives the relation between the plastic flow direction and the plastic
strain rate. A non-associated flow rule means that the yield function f and the plastic
potential F p do not coincide, and therefore, the direction of the plastic flow is not
normal to the yield surface. This is important for realistic modeling of the volumetric
expansion (dilatancy) under compression for frictional materials such as concrete. Using
an associated flow rule for the type of yield surface shown in Eq. (4.67) gives an
unrealistically high volumetric expansion in compression, which leads in some cases to
an overestimated strength - peak stress (Chen and Han, 1988). Therefore, the shape of the
concrete loading surface at any given point in a given loading state should be obtained by
using non-associative plasticity. The plastic strain rate can be written in terms of the
effective stress ij as:
F p
ijp = (4.72)
ij
where is the plastic flow parameter (consistency factor) known as the Lagrangian
multiplier, which can be obtained using the plasticity consistency condition, and the
plastic potential function F p takes the following Drucker-Prager format as given in Lee
and Fenves (1998):
F p = 3 J 2 + p I1 (4.73)
such that:
F p 3 sij
= + p ij (4.74)
ij 2 3J 2
where p is a parameter chosen to provide proper dilatancy with common range between
0.2 and 0.3 for concrete (Lee and Fenves, 1998; Wu et. al., 2006).
The plasticity consistency condition can be obtained by taking the time derivative of
the plasticity yield function, f = 0 , and satisfying the following Kuhn-Tucker
loading/unloading conditions:
93
f 0, p 0, p f = 0, p f = 0
If f <0 then p = 0;
(4.75)
If f = 0 and f 0 then p = 0;
If f = 0 and f > 0 then p > 0
This concludes the plastic formulation for the present model. The damage formulation
is discussed next where the tensile and compressive damage surfaces are defined.
The isotropic damage in this work is responsible for the softening response and the
degradation in the elastic stiffness. The tensile and compressive damage surfaces and
their hardening functions will be presented first, followed by a brief discussion of the
damage consistency conditions.
To determine the stress states during a damaging process from the thermodynamic
constitutive relations, Eqs. (4.28), (4.31) and (4.32), tensile and compressive damage
surfaces and their evolution laws have to be specified. Referring to the definitions of a
yield function and the plastic flow rule in plasticity theory, Tao and Phillps (2005)
defined for an isothermal process the following two damage surfaces g as functions of
the damage thermodynamic conjugate forces Yij and the scalar damage parameters ,
with a similar form to that of La Borderie et. al. (1992):
where Y0 are initial damage thresholds (tension and compression) which govern the
onset of tensile or compressive damage, respectively. As damage progresses, initial
damage surfaces change by means of evolution laws defined by hardening/softening
parameters Z . These paramters Z can be expressed mathematically in different forms,
such as polynomials, power and exponential functions, etc. Amongst them power and
exponential functions have the best match for the shapes of loading curves of concrete
(Lubliner et. al., 1989; Lee and Fenves, 1998; Nechnech et. al., 2002; Tao and Phillips,
2005; Wu et. al., 2006 and others). Tao and Phillips (2005) assumed that the softening of
damage surfaces follow a power law in the form of:
1
1 b
Z =
(4.77)
a 1
94
effects of a and b on Z as damage progress. They showed that the shape of damage
surface varies with b , whilst a determine the magnitude of Z . In other words,
parameters a mainly dominate the magnitude of damage surfaces with units of MPa 1 ,
whilst b , being dimensionless parameters, influence generally the characteristics of
softening/hardening (see Fig. 4.1). The same trend was observed in this work as will be
shown in the verification section. Tao and Phillips (2005) claimed that a proper selection
of parameters a and b tailors Z to the demands of different types of concrete and
their corresponding tensile and compressive strengths. Other researchers that used Eq.
(4.76) to model their isotropic damage criteria include Salari et. al. (2004), Shao et. al.
(2006), and Grassl and Jirasek (2006). All three used single isotropic damage variable
and thus, a single damage criterion.
f 0
(b)
f +
(a)
0 G f
G +f
p p
Fig.4.2 ( ) relation for uniaxial test, a) tension, b) compression
p
In this work, and in an effort to reduce the sensitivity of the FE analysis of concrete to
the refinement of the FE meshes, the damage magnitude parameters a are adjusted to
include dimensionless embedded coefficients that are related to the fracture energies
in tension and compression, G f (see Fig. 4.2), and to the (geometrical) characteristic
length, A , of the applied FE mesh obtained from ABAQUS. These coefficients are
given as (Oliver et. al., 1990; Labadi and Hannachi, 2005; Wu. et. al., 2006):
95
1
G f E 1
= 2
(4.78)
Af o 2
A stress point in principal stress space can be either within or on the current damage
surface. When within the damage surface, being tensile or compressive, the stress point
may be loading, but it has not violated the current damage criterion yet. Once it is on the
damage surface, two damage states are possible. One may be unloading or neutral
loading, having = 0 . The other is loading, accompanied by the evolution of damage
and defined as > 0 . Mathematically, the above description is expressed as:
If g < 0 then = 0;
If g = 0 and g 0 then = 0; (4.79)
If g = 0 and g > 0 then > 0
The above conditions are damage extension of the classical plasticity KuhnTucker
conditions (Voyiadjis and Kattan, 1992).
ij = (1 ) ij ij
ELASTIC PLASTIC
PREDICTOR CORRECTOR
In the first step, the elastic-predictor problem is solved with the initial conditions
being the converged values of the previous iteration ( t = tn ) along with the new increment
while keeping the irreversible variables frozen. This produces a trial stress state,
ijtrial = ijn + ijtrial , which, if outside the plastic surface f and the damage surfaces g
is taken as the initial conditions for the solution of the plastic-corrector and damage-
corrector problems. The scope of the second and third steps is to restore the generalized
96
plasticity and damage consistency conditions by returning back the trial stress to the
plastic surface f and the damage surfaces g .
During the elastic-predictor ijtrial and the plastic-corrector Eijkl klp steps the
damage variables are fixed, so that the elastic-plastic behavior is decoupled from damage,
constituting a standard elastic-plastic problem in the effective stress space. Regarding the
adopted plastic yield function, Eq. (4.67), the spectral decomposition form (Lee and
Fenves, 2002) of return-mapping algorithm (Simo and Hughes, 1998) is applied to update
the effective stress tensor ijn +1 . Once the effective stress tensor ijn +1 is updated in the
elastic-predictor and plastic-corrector steps, the damage variables ( ) n +1 (and therefore
n +1 ) and the Cauchy stress tensor ijn +1 can then be updated correspondingly in the
damage-corrector step (Wu et. al., 2006).
A fully Implicit (Backward-Euler) scheme is used for the stress computation problem
in the effective space, followed by an explicit integration scheme for the updated damage
variables and Cauchy stress tensor. The integration procedure start at the beginning of the
( n, n + 1 ) step where ijn , ( ijp ) n , ( ) n (adopted here for simplicity as n ) and ij are
all known from the previous step ( n 1, n ).
ijn +1 = Eijkl ( kle ) n +1 = Eijkl ( kln +1 ( klp ) n +1 ) = Eijkl ( kln + kl ( klp ) n klp )
= Eijkl ( kln ( klp ) n ) + Eijkl ( kl klp ) = Eijkl ( kle ) n + Eijkl kle (4.81)
= ijn + ij
The elastic predictor problem defines the stress in the trial state assuming an entirely
elastic strain increment ( ij = ije ) as:
97
ijtrial = Eijkl ( kln +1 ( klp ) n ) = Eijkl ( kln + kl ( klp ) n )
= Eijkl ( kln ( klp ) n ) + Eijkl kl = Eijkl ( kle ) n + Eijkl kl (4.82)
= ijn + Eijkl kl = ijn + ijtrial
In order to verify the correctness of this elastic prediction, the trial stress is applied
into the yield function f ( ijtrial , n ) . If f ( ijtrial , n ) < 0 , the process is elastic and the
trial state is admissible and accepted as the final state since there is no change in the
plastic strain, such that:
If the current step is not an elastic state, f ( ijtrial , n ) > 0 , the plastic strain tensor will
change. To compute the effective stress along with the plastic strain tensor at the current
time increment requires iterations within that time increments for the effective stress in
the plastic-corrector step. During these iterations, the discrete version of the plastic
consistency condition is imposed as a constraint, f ( ijn +1 , n +1 ) = 0 , at the end of each
iteration. Therefore, the analysis is transformed into a linear set of equations that depend
on the material parameters and on the current coordinates of the integration points within
each iteration. The outcomes of the plastic-corrector step are the updated effective stress
tensor ijn +1 , plastic strain tensor ( ijp ) n +1 , and plastic variables ( ) n +1 . The plastic-
corrector step can be derived as follows:
Within the plastic-corrector step, and in order to compute klp in Eq. (4.84), the
radial-return method is used. The flow rule in Eq. (4.72) can be written implicitly for the
(n + 1) step as follows:
F p
klp = p (4.85)
kln +1
98
F p
where (Eq. (4.74)) is given here as the following:
kln +1
F p 3 Skln +1
= + p kl (4.86)
kln +1 n +1
2 Smn
n +1 n +1 n +1
where Smn = S mn Smn is the norm of the updated effective deviatoric stresses.
By substituting the elasticity tensor in the undamaged configuration, Eijkl (Eq. (4.4)),
into Eq. (4.84), the effective stress can be updated using the return-mapping equation
given as follows (derivation shown in Appendix B):
2
ijn +1 = ijtrial 2G ijp + ( K G ) kkp ij (4.87)
3
where kkp can now be obtained from Eqs. (4.85) and (4.86) as follows:
kkp = 3 p p (4.88)
By substituting Eqs. (4.85), (4.86), and (4.88) into Eq. (4.87), one obtains:
3 Sijn +1 2
ijn +1 = ijtrial 2G p n +1
+
p ij + ( K G )(3 p p
) ij
(4.89)
2 S
3
mn
Expanding and then simplifying the above equation, one can obtain the following form:
Sijn +1
n +1
ij = trial
ij 6G p
n +1
+ 2G p p
ij + 3 K p p
ij 2G p p
ij
S mn
(4.90)
S n +1
ijn +1 = ijtrial p 6G ijn +1 + 3K p ij
Smn
Separating Eq. (4.90) into deviatoric and volumetric parts gives the following:
99
S n +1
kl = I ijkl
dev n +1
I ijkl kl p I ijkl
dev trial dev
6G kln +1 + 3K p kl
S mn
1 1 S n +1
Sijn +1 = Sijtrial p ( ik jl + il jk ) ij kl 6G kln +1 + 3K p kl (4.91)
2 3 Smn
Sijn +1
Sij = Sij 6 G n +1
n +1 trial p
Sij
and
P 0
S n +1 P3
iin +1 = iitrial p 6G iin +1 + 3K p ii
Smn (4.92)
I1 = I1 9 K p
n +1 trial p
To solve for the plastic multiplier p from Eq. (4.91), the radial-return method is
applied. The proof of the radial-return applied to Eq. (4.91) was provided in Chapter 3,
Eq. (3.88) and is stated again here as:
Sijn +1 Sijtrial
= (4.93)
Sijn +1 Sijtrial
Therefore, substituting Eq. (4.93) into Eq. (4.91), the radial-return form of Eq. (4.91) can
be provided as follows:
Sijtrial
Sijn +1 = Sijtrial 6 p G (4.94)
Sijtrial
and rewriting Eq. (3.86) from Chapter 3 using the current notation, one obtains the
following:
Sijtrial
n +1
ij = trial
ij 6G trial + 3K p ij
p
(4.96)
S mn
100
Next, in order to account for the effective principal stress term in the yield function,
Eq. (4.67), the spectral return-mapping algorithm of (Lee and Fenves, 2001) is used.
This algorithm has the advantage for yield function which includes principal stress terms
in addition to the stress tensor invariants (Wu et. al., 2006). A decoupled version of the
return-mapping algorithm is derived (using the spectral decomposition concept) and
shown below.
where lmn is the principal direction tensor corresponding to the effective principle stress
tensor. Lee and Fenves, (2001), showed that the plastic potential function F p can be
written in terms of the effective principal stresses such that the increment of the effective
plastic strain ijp can be written as follows:
F p
ijp = p lir l js (4.98)
rsn +1
By substituting Eqs. (4.88), (4.97), and (4.98) into Eq. (4.87), one can write the return-
mapping equation in the following form:
F p 2
lir rsn +1l js = ijtrial 2G p lir l js + 3( K G ) p p ij (4.99)
rs
( n +1)
3
Using the following relation for the Kroneckar delta, which holds for any direction tensor
lmn :
ij = lir rs l js (4.100)
one can write Eq. (4.99) in the following form to obtain the effective trial stress directly
as:
F p 2
ijtrial = lir rsn +1l js + p lir 2G + 3( K G ) p rs l js (4.101)
rs
n +1
3
Lee and Fenves (2001) proved that any principal direction tensor for ijn +1 is also a
principal direction tensor for ijtrial . This proof is necessary to show that the eigenvector
spaces of ijn +1 and ijtrial are exactly identical despite the fact that symmetric tensors do
not have a unique spectral decomposition form if they have repeated eigenvalues.
Accordingly, the spectral decomposition of the trial stress tensor is given as:
101
ijtrial = lir rstrial l js (4.102)
From Eq. (4.101) along with Eq. (4.102) one can write the decoupled form of the
return-mapping equation as follows:
F p 2
trial
lir rs l js = lir rs l js + lir 2G
n +1 p
+ 3( K G ) l js
n +1 p rs
rs 3
(4.103)
F p 2
ijn +1 = ijtrial p 2G + 3( K G )
p ij
ij
n +1
3
and from Eq. (4.98), the eigenvalue tensor of the plastic strain increment becomes:
F
ijp = p (4.104)
ijn +1
where the derivative of the potential function with respect to the principal stress tensor is
given as follows:
n +1
F p 3 Sij
= + p ij (4.105)
ijn +1 2 S n +1
mn
However, using Eq. (4.93), one can obtain the following expression (Lee and Fenves,
2001):
trial 1 trial
F p
S trial
3 ij
I1 ij
3 ij
+ p ij = 3
= + p ij
ijn +1 2 S trial 2 S trial
mn mn
(4.106)
3 ij
trial 1 I1trial
= + ij
trial
2 S mn p 6 Smn trial
where I1tr = I1tr and Sijtr = Sijtr are invariants ( Sijtrial = 2( J 2 )trial , I1trial = mm
trial
).
Now, by substituting the final form of Eq. (4.106) into Eq. (4.103) one can obtain the
following form:
102
3 trial
1 I1trial 2
ijn +1 = ijtrial p 2G ij
+ ij + 3( K G ) p ij
2 Smn tr
p tr
6 S mn 3
ijtrial 2 I trial
ijn +1 = ijtrial p 6 G trial
+ 2G p ij G 1trial ij + 3K p ij 2G p ij (4.107)
Smn 3 Smn
ijtrial 2 I trial
n +1
ij = ij 6 G trial + 3K p
trial p
G 1trial ij
Smn 3 Smn
In order to obtain the expression sought out of the spectral radial-return algorithm for the
maximum principal stress max
n +1
used in the yield function, Eq. (4.67), one can contract
the above relation such that:
max
trial 2 I trial
max
n +1
= max
trial
p 6 G + 3K p G 1trial (4.108)
trial
S mn 3 Smn
It can be seen from Eqs. (4.103) and their final result, Eq. (4.108), that the eigenvectors
are preserved throughout the corrector-steps which basically means that the effective
stress eigenvectors are calculated at the predictor step and only the principal stress needs
to be determined during the iterations of the plastic-corrector step (Lee and Fenves,
2001). It should also be noted that, if the eigenvalues of the plastic strain increment
tensor, ijp , are obtained by a linear combination of ijn +1 and ij , such as for the
Drucker-Prager model, Eq. (4.73), the algebraic order of the effective (undamaged)
principal stresses is preserved throughout the corrector-steps. This was shown by Lee and
Fenves (2001) through checking Eq. (4.107) and realizing that the updated effective
principal stress tensor ijn +1 is obtained only by a scalar multiplication and constant-
tensor addition/subtraction on the trial stress. Therefore, the order of the diagonal entries
in the trial stress tensor cannot be changed. This argument, however, is not valid for the
case where the yield criterion f given in Eq. (4.67) is used as a plastic potential
function. This means that if one takes the derivative with respect to the maximum
stresses, max
n +1
, the algebraic order in the eigenvalue tensor does not preserve the same
order.
The radial-return mapping algorithm is derived next in order to solve for the plastic
multiplier p using the yield function discussed in Eq. (4.67), which can be written
here at the end of the ( n + 1 ) step as:
103
f ( ijn +1 , ( ) n +1 ) = 0
3 n +1 (4.109)
Sij + I1n +1 + ( ) n +1 H( max ) max (1 ) c ( ) n +1 = 0
n +1 n +1
Applying the plasticity consistency condition given in Eq. (4.75) to the previous equation
yields the following:
f f f f
f n +1 = f n + ij + max + + + + = 0 (4.110)
ij max
where ij is defined here in the incremental form using Eqs. (4.90) and (4.96) as:
F p
ij = Eijkl kle = Eijkl ( kl klp ) = Eijkl kl Eijkl p
kl
(4.111)
2 S trial
= trial
ij 2G ijp + ( K G ) kkp ij = ijtrial p 6G ijtrial + 3K p ij
3 Smn
and the principal increment max can be obtained using Eq. (4.108) as follows:
max
trial 2 I1trial
max = max
trial p
6 G trial + 3K p G trial (4.112)
S mn 3 Smn
The increments of the equivalent plastic strains are expressed using Eqs. (4.60),
(4.61) and (4.104) as follows:
F p
+ = r p (4.113)
max
F p
= (1 r ) p (4.114)
min
where r is defined in Eq. (4.62). Substituting Eqs. (4.111), (4.112), (4.113) and (4.114)
into Eq. (4.110), one can obtain the following relation:
104
f Sijtrial
n +1
f = f + n
ij 6G trial + 3K p ij
trial p
ij Smn
f trial trial 2 I1trial
+ max p 6 G max + 3 K G (4.115)
max
p
trial
S mn
3 trial
Smn
f F p f p F
p
+ + r p + (1 r ) =0
max
min
F p F p
The plastic potential function derivatives, and , are given using Eq. (4.105)
max min
along with the Eq. (4.93) as functions of the trial state:
1
( max
trial
I1trial )
F p 3 3
= +p (4.116)
max 2 trial
S mn
1
( min
trial
I1trial )
F p 3 3
= +p (4.117)
min 2 trial
S mn
f f
and the yield function derivatives, and , are obtained using Eqs. (4.67), (4.69),
+
(4.70), and (4.71) as follows:
f (1 )c h n +1
= max (4.118)
+ (c + ) 2
f (1 )
= Q exp( ) max
n +1
(1 )Q exp( )
c+
(4.119)
n +1
= (1 )Q exp( ) max +
1
c
f f
fn+ ijtrial + max
trial
ij
max
p = (4.120)
H
where
105
f Sijtrial
H= 6G trial + 3K p ij
ij Smn
f max
trial 2 I1trial
+
6 G trial + 3K p G trial (4.121)
max Smn 3 Smn
f F p f F p
r + + (1 r )
max min
trial
f 3 Sij
= + ij (4.122)
ij trial
2 Smn
f
= ( ) n +1 H( max
n +1
) (4.123)
max
where ( ) n +1 are the updated plastic variables. Applying Eqs. (4.122) and (4.123) into
Eq. (4.121) yields the following simplified expression for H :
trial 2 I1trial
H = 3G + 9 K p + ( ) n +1 H( max
n +1
) 6 G max + 3 K G
Smntrial
p
3 trial
Smn
(4.124)
f F p f F p
r + + (1 r )
max min
It should be noted here that Eqs. (4.118), (4.119) and (4.123) are functions of the updated
forms of the principal stress in tension max
n +1
and the plastic variables ( ) n +1 . This is
the reason why local iterations are required to obtain the plastic multiplier p . In each
iteration the new evaluated trial stress will be used to update the plastic variables and the
hardening functions in order to obtain an updated plastic multiplier. The process goes on
until a convergence tolerance for the yield function is satisfied.
The numerator of Eq. (4.120) can be shown to be equal to f trial during the iterative
procedure by substituting Eqs. (4.92), (4.95), and (4.108) into Eq. (4.109) and
considering only the terms that don not involve the plastic multiplier p to obtain the
following:
106
3 trial
f trial = Sij + I1trial + ( )trial H( max ) max (1 ) c ( ) n
trial trial
(4.125)
2
f trial
p = (4.126)
H
This concludes the elastic-plastic steps in the effective configuration. Upon convergence,
the updated effective stress ijn +1 along with its updated elastic strain tensor ( ije ) n +1 are
now available to be used in the damage-corrector step in order to update the damage
variables ( ) n +1 (and therefore n +1 ) and the Cauchy stress tensor ijn +1 . The flowchart
shown in Fig. 4.3 demonstrates the effective elastic-plastic integration procedure.
The degradation process is termed explicit because the spectral stress ratios used in
defining the damage parameter are assumed to depend on the initial stress states of
each degradation increment such that Eq. (4.43) can now be written as:
ij+ + + ij
= (4.127)
ij
where ij+ and ij are the positive and negative spectral decomposition parts of the
effective stress tensor, ij ; another modification to the Tao and Phillips (2005) definition
for , since their model (only elastic-damage analysis) did not include the plastic effect.
Nevertheless, satisfying the damage criteria and updating the damage variables are
accomplished by using the projected values (n+1) of the effective elastic strain tensor.
Lee and Fenves (2001) showed that the eigenvectors of the current stress ijn +1 are the
same as those of the effective stress ijn +1 due to the fact that scalar degradation damage is
assumed. Therefore, at the degradation corrector step, the final form of the stress tensor
ijn +1 is obtained by computing the degradation damage variable , such that:
This step starts after the effective elastic-plastic steps pass the updated forms of the
effective stress tensor ijn +1 and its corresponding elastic strain tensor ( ije ) n +1 . The
spectral decomposition of the updated effective stress tensor ijn +1 into positive and
107
negative parts is then performed according to the procedure shown in Section 4.4.1.
These spectral components are then used to calculate the damage release rates (damage
thermodynamic conjugate forces) Y + and Y by substituting Eq. (4.54) into Eqs. (4.55)
and (4.56) to obtain the following two equations:
1 ( ij ) e n +1
+ n +1
1 1 e n +1 2
Y = +
n +1
( ij ) Eijkl ( kle ) n +1 + +
(( mm ) ) ij Eijkl kl (4.129)
2 ( ij ) 9 1 + cY exp(dY )
1 ( ij ) e n +1
n +1
1 1 e n +1 2
Y =
n +1
( ij ) Eijkl ( kle ) n +1
(( mm ) ) ij Eijkl kl (4.130)
2 ( ij ) 9 1 + cY exp( dY )
Equations (4.129) and (4.130) are nonlinear functions of the conjugate forces Y + and Y ,
respectively. Therefore, Newton-Raphson iterative procedure is used in order to solve
each of these two equations independently. The procedure is demonstrated next for the
thermodynamic force Y i for expediency, where i represents (+) or ( ) such that:
1 ( ij ) e n +1
i n +1
1 1 e n +1 2
Y
i
( ij ) Eijkl ( e n +1
kl ) (( mm ) ) ij Eijkl kl = 0 (4.131)
2 ( ij ) n +1 9 1 + cY exp( dY )
i i
Equation (4.131) can be written as K (Y i ) = 0 with function roots Y i obtained using the
Newton-Raphson iterative technique as follows:
K (Y i )
Ymi +1 = Ymi + Y i = Ymi K (Ymi ) (4.132)
Y
i
K (Y i )
where the derivative is given as follows (evaluated at Y i = Ymi ):
Y i
i
1 ( ij )
i n +1
K (Y i ) 1 c exp( dY ) cdY exp( dY ) (( e ) n +1 ) 2 E (4.133)
i i
= 1
9 ( ) mm
ij ijkl kl
Y i 2 ( ij ) n +1 1 + cY i
exp( dY i
)
2
Since the outcome of this iterative process is highly dependent on the guess-value
initially suggested for Ymi = Y0i , the following form of Y0i was observed in this study to
be convenient:
1 ( ij )
i n +1
Y = 0
i
2 ( ij ) n +1
( ( ije )n+1 Eijkl ( kle )n+1 ) (4.134)
108
The iterative procedure is terminated upon convergence of a tolerance criterion (e.g.
Y i 0 ).
Now that the thermodynamic conjugate forces are determined, Y + and Y are
substituted into Eqs. (4.76) in order to check the activation of the tensile and compressive
damage yield criteria g + and g , respectively:
g = (Y ) n +1 Y0 ( Z ) n 0 (4.135)
If g + < 0 and g < 0 , the damage-corrector step ends and the Cauchy stress tensor takes
the form of the updated effective stress tensor. If either of the damage criteria is
activated, g > 0 , damage evolution takes place and the Cauchy stress tensor is
evaluation using Eq. (4.128). An integration algorithm is thus devised as demonstrated
here to obtain the updated damage variables. These updated damage variables, along with
the thermodynamic conjugate forces Y + and Y , should satisfy the damage consistency
conditions shown in Eq. (4.79), such that:
(Y ) n +1 Y0 ( Z ) n +1 = 0 (4.136)
1
1 ( ) n +1 b
( Z ) n +1 = (4.137)
a 1 ( ) n +1
Substituting Eq. (4.137) into Eq. (4.136), explicit expressions for the damage variables
( ) n +1 satisfying the consistency conditions can be obtained as follows:
( a (Y ) )
b
n +1
Y0 1
( ) n +1 = = 1 (4.138)
1 + ( a (Y ) ) ( )
b b
n +1
Y0 1 + a (Y ) n +1 Y0
These updated damage variables can now be used to evaluate the combined scalar
damage variable using Eq. (4.127). The later can then be used to update the Cauchy
stress tensor as shown in Eq. (4.128). This concludes the damage-corrector step. Figure
4.3 shows an entire step of the integration scheme demonstrating the effective elastic-
predictor plastic-corrector steps followed by the damage-corrector step.
109
equation, Eq. (4.7), with respect to the strain tensor as follows (all parameters are at the
(n+1) state):
ij ij
= (1 ) ij (4.139)
kl kl kl
where the elasto-plastic tangent operator appears in the first term on the right hand side of
this equation. Since damage depends on the elastic strain only, can be written as:
kl
mn
e
= e (4.140)
kl mn kl
and the derivative of the elastic strain tensor with respect to the total strain tensor can be
obtained by taking the derivative of the constitutive equation, Eq. (4.3), with respect to
the total strain tensor as follows:
mn
e
pq
= ( E 1 ) (4.141)
kl mnpq
kl
Substituting Eqs. (4.140) and (4.141) into Eq. (4.139), the following arrangement can be
obtained:
ij ij pq
kl
= (1 )
kl
mn
e ( E 1
) mnpq
ij (4.142)
kl
ij pq
= (1 ) pi qj e ( E 1 ) ij (4.143)
kl mn mnpq
kl
Equation (4.143) is equivalent in format to that given by Wu et. al. (2006) as follows:
I
R
ij
= ( pi qj ) ( pi qj ) e ( E 1 ) ij pq (4.144)
kl mn
mnpq
kl
ij
In order to obtain , the derivative of the damage variable with respect to the elastic
kl
strain tensor, , needs to be evaluated. This can be accomplished by considering the
mn
e
110
n , ( e ) n , ( p ) n , ( ) n , ( )n , n known
Given
trial = n + E
p =0
yes
f ( trial ) < 0 n +1 = trial , ( e ) n +1 = ( e ) n +
no
f trial
Update p = p +
H
Calculate p , ( )
Update trial = trial E p
Update hardening functions c + , c
no Convergence
f ( trial ) = 0
yes
Update ( p ) n +1 , ( e ) n +1
Update n +1 = trial
Evaluate ( ij ) n +1 , (Y ) n +1
111
no yes
g ((Y ) n +1 , ( ) n ) < 0
Update ( ) n +1 ( ) n +1 = ( ) n
Using (Y ) n +1
Update
( ) n+1
n +1 = (1 ) n +1
ij + ij
+
= + (4.145)
mn
e
ij mn
e
ij mn
e
Note that the stresses in Eq. (4.127) are used to obtain scalar ratios that are used as
weighing factors; therefore, they are not considered as contributing components to the
foregoing derivative. Using Eq. (4.138), it can be seen that the damage variables are
functions of the thermodynamic conjugate forces Y , respectively, and Y are functions
of the elastic strain tensor mn
e
, Eqs. (4.129) and (4.130), such that the following
expression can be obtained:
Y
= e (4.146)
mn
e
Y mn
112
( )
( b ) 1
a b a Y Y0
= (4.147)
Y 2
( )
b
1 + a Y Y0
and the second term is obtained by applying the linearization technique given by (Simo
and Hugues, 1998) using Eqs. (4.129) and (4.130) as (4.152):
ij 1
Emnkl kl (1 ) pp mn ij Eijkl kl
e e
Y ij 9
= (4.148)
mn
e
1 ij c exp( dY ) cdY exp(dY ) 1 e 2
1 ( qq ) ij Eijkl kl
2 ij (1 + cY + exp(dY ) ) 9
2
where the derivatives of with respect to the elastic strain tensor were expressed as
follows:
Y
= e (4.149)
mn
e
Y mn
ij
The effective consistent tangent operator can be obtained using the linearization
kl
technique given in the references mentioned above and is stated here as given by Wu et.
al. (2006) as:
1
ij F p p 2 F p
= ( E 1 ) + + p (4.150)
kl ijkl ij kl ij kl
p
where is given as:
kl
f
p
kl
= (4.151)
kl f ( / p ) i
+ p
p i p
113
4.9 Implementation and Verification of the Integration Scheme
This section is dedicated to the numerical validation of the concrete model. The
numerical algorithm of the proposed model was implemented in the non-linear FE code
ABAQUS via the user material subroutine UMAT. Several analytical examples are
provided here in order to investigate the capability, applicability, and effectiveness of the
proposed elastic-plastic-damage model in capturing material behavior in both tension and
compression under uniaxial and bi-axial loadings. The results obtained by the proposed
model are compared with corresponding experimental results to evaluate the models
performance. Tensile and compressive verification tests under unaxial loading are
demonstrated first, followed by biaxial tests in tension and compression. Then a three
point bending test of a notched beam is investigated.
The proposed model contains 17 parameters: two elastic constants for the undamaged
material ( E and ), five parameters for the characterization of plasticity ( , p , h , Q
and ), eight parameters for damage characterization ( a , b , Y0 , c and d ), and two
parameters for the fracture energy of concrete under tension and compression ( G f ). All
the parameters can be identified from a series of tensile and compressive experimental
tests (Tao and Phillips, 2005 and Wu et. al., 2006). Since the elastic-plastic-damage
model presented in this work is a combination of the effective elastic-plastic constitutive
relations presented by (Lee and Fenves, 1998 and Wu et. al., 2006) and the damage
model presented by (Tao and Phillips, 2005); the model parameters used in these works
are used here with some adjustment to the damage parameters in order to take into
consideration the plastic effect introduced in this work.
The initial elastic constants are determined from the linear part of stressstrain curves
before the initiation of damage and plastic deformation (Shao et. al., 2006). For concrete
materials, these averaged elastic parameters are documented in literature. The initial
damage thresholds in tension and compression, Y0 , can be determined by locating the
onset points of variation of elastic properties in the unloading paths (Tao and Phillips,
2005). As the damage evolution is coupled with plastic flow, it seems to be reasonable to
consider that the damage initiation occurs at the same time as the plastic initiation under
tensile loading. They are identified as the end of the linear part of the stressstrain curves
in uniaxial tensile tests. Under compression, however damage start at an earlier stage than
plasticity depending on the initial damage threshold Y0 . The plastic hardening
parameters ( Q and ) are related to the saturated stress in the plastic regime and the rate
of saturation.
Lee and Fenves (1998) and Wu et. al. (2006) used the following values for the elastic-
plastic material parameters: Poissons ratio = 0.20; the equibiaxial to uniaxial
compressive strength ratio fb0 / f 0 = 1.16, resulting with = 0.12, and the dilatancy
parameter p was chosen as 0.20. Tao and Phillips, (2005), used the following values for
114
the material parameters used to split the strain tensor into hydrostatic and deviatoric
compoenets: c = 2.0 MPa 1 and d = 0.7 MPa 1 . These values are used throughout this
work.
82.6 mm
a) 82.6 mm b)
c) d)
Figure 4.5 Quadrilateral finite elements under uniaxial (a - tension, c -
compression) and Biaxial (b - tension, d- compression)
In the first example of uniaxial tension test, the following material properties are used
(Lee and Fenves, 1998; Tao and Phillips, 2005; Wu et. al., 2006; Nguyen and Houlsby,
2008a,b) in order to compare the results with the experimental work of Gopalaratnam and
Shah (1985): E = 3.1104 MPa, f 0+ = 3.48 MPa and G +f = 40 N/m. This test is
conducted using a single quadrilateral finite element (82.6 mm x 82.6 mm) shown in Fig.
4.5a to comply with the results of the studies mentioned above. The models plastic
hardening parameter is given here for the tensile case as: h = 2.5 104 MPa. The models
damage parameters are provided by Tao and Phillips, (2005) as: b + = 1.2 , Y0+ = 1.9 104
MPa. Only parameter a + = 14 103 MPa 1 was adjusted to account for the plastic effect
introduced in this work. Note that the uniaxial tests (in tension and compression) were
used to determine the values of the material parameters to be used along with the fracture
energy related factors in further examples. The use of the fracture energy factors
115
a) Stress-Strain curve b) Strain .vs. Damage Variable
Figure 4.6 The behavior of the proposed model under uniaxial tension
a) Effect of h b) Effect of a +
116
The stress-strain response is plotted in Fig. 4.6a while damage evolution is plotted in
Fig 4.6b. It can be observed from Fig. 4.6a that the predictions obtained from the
numerical model agree well with the experimental data (Gopalaratnam and Shah, 1985),
especially for the post-peak nonlinear softening branches.
The effect of the model parameters mentioned above on the stressstrain response and
damage evolution in tension is shown in Fig. 4.7a-d. Each model parameter in turn is
varied, while others are kept fixed, to show the corresponding effect on the stressstrain
curve and damage evolution.
Note in Fig. 4.7a, that as the plastic hardening parameter h decreases, the plastic
behavior becomes more dominant than the elastic one, resulting in smaller elastic strains,
which affect the magnitude of the tensile damage release rate Y + and therefore negatively
affect the damage growth. This shows the coupled effect of damage and plasticity on the
predicted behavior.
The models ability to reproduce the concrete behavior under monotonic uniaxial
compression is verified here and compared to the experimental results of Karsan and Jirsa
(1969). The material properties used here are (Lee and Fenves, 1998, Tao and Phillips,
2005; Wu et. al., 2006; Nguyen and Houlsby, 2008a,b): E = 3.1104 MPa, f 0 =10.2
MPa, f c' = 27.6 MPa and G f = 5690 N/m. The test is conducted using a single
quadrilateral finite element (82.6 mm x 82.6 mm) shown in Fig. 4.5c. The models plastic
hardening parameters are given here for the compressive case as: Q = 2.5 103 MPa and
= 200 . The models damage parameters are provided as: a = 22 MPa 1 , b = 0.98 ,
Y0 = 3.0 104 MPa 1 . The models parameters a and b were adjusted in order to
account for the introduction of the plastic effect in this work. The stress-strain response is
117
a) Effect of b) Effect of Q
c) Effect of a d) Effect of b
e) Effect of Y0
Figure 4.9 Effect of material parameters on the model response in compression
plotted in Fig. 4.8a while damage evolution is plotted in Fig 4.8b. Whether in the
hardening or in the softening regimes, the overall nonlinear numerical performance
118
predicted by the model and the experimentally obtained stressstrain curve are rather
close.
The effect of model parameters on the stressstrain response and damage evolution in
compression is shown in the Fig. 4.9. Each model parameter in turn is varied, while
others are kept fixed, to show the corresponding effect on the stressstrain curve and
damage evolution; Fig. 4.9a-d:
Figures 4.9b shows a trend similar to that observed in Fig. 4.7a. As the magnitudes of the
hardening parameters increase, the damage growth and thus the strain softening is more
pronounced up to the point where the exponential hardening function becomes saturated.
This again shows the coupled effect of damage and plasticity on the response of the
proposed model.
In this section, the performance of the proposed model subjected to combined loading
situations (biaxial tension, biaxial compression, and biaxial tension compression) is
investigated.
In the biaxial tension case, the same material parameters as those for the uniaxial
tension test are used to analyze the Quadrilateral FE setup shown in Fig. 4.5b. The
numerical results are compared to the experimental ones reported by Kupfer et. al.
(1969). In Fig. 4.10a, the ordinate represents the normalized stress 11 in terms of the
compressive strength f c' = 27.6 MPa for the case ( R = 22 / 11 = 1 ). Note that the results
of Kupfer et. al. (1969) cover only the range of ( 0 11 / f c' 0.09 ). The results are in
good agreement. The full range of the stress ratio ( 0 R 1 ) is investigated next.
119
Table 4.1 Biaxial Tension Test
11 22 11 22 11 / f c' 22 / f c'
0.0006 0 3.48573 0 0.126295 0
0.0006 0.0001 3.48279 1.39312 0.126188 0.050475
0.0006 0.0002 3.39821 1.82981 0.123124 0.066297
0.0006 0.0003 3.29641 2.19761 0.119435 0.079624
0.0006 0.0004 3.14892 2.47415 0.114091 0.089643
0.0006 0.0005 2.94521 2.64054 0.106711 0.095672
0.0006 0.0006 2.80792 2.80792 0.101736 0.101736
0.0005 0.0006 2.64054 2.94521 0.095672 0.106711
0.0004 0.0006 2.47415 3.14892 0.089643 0.114091
0.0003 0.0006 2.19761 3.29641 0.079624 0.119435
0.0002 0.0006 1.82981 3.39821 0.066297 0.123124
0.0001 0.0006 1.39312 3.48279 0.050475 0.126188
0 0.0006 0 3.48573 0 0.126295
During the biaxial tension test, the total displacements in the horizontal and vertical
directions of the setup shown in Fig. 4.5b are specified in the input file. In order to
retrieve the full spectrum of stress ratios ( 0 R 1 ), the displacement in one direction is
fixed while the displacement in the other direction is incremented during multiple runs of
the input file. The results are shown in Table 4.1.
Figure 4.11 The tensile quadrant of the biaxial failure surface of concrete
The last two columns of Table 4.1 are plotted against each other to obtain Fig. 4.11
showing the tensile quadrant of the biaxial failure surface of concrete. The experimental
values were obtained by Kupfer et. al. (1969) and used by Lee and Fenves (1998) and Wu
et. al. (2006).
The tensile biaxial results are compared to those of the uniaxial tension test. Fig 4.12a
shows that under the biaxial state ( 11 = 22 ), the model predicts higher damage growth
120
rate than under uniaxial tension. This is also obvious from Fig. 4.12b, where damage
starts at an earlier stage and grows at a higher rate.
In the case of biaxial compression, the proposed model in its given form is not capable
of capturing the trend observed experimentally. A modification that had to be applied to
the proposed model in order to enhance its performance under biaxial compression is
discussed next. This modification is easily incorporated into the UMAT file using proper
IF statements. Experimental results (e.g. Kupfer et. al. (1969)) showed an increase in
concrete compressive strength as the biaxial stress ratio ( R = 22 / 11 ) increases up to the
point where the strength in one direction is 1.3 f c' , followed by a reduction in strength
that reaches 1.16 f c' when R = 1.0 (see Fig. 4.13). This is a result of the consolidation of
concrete under biaxial compressive loading which leads to reduced damage growth (Wu
et. al., 2006). This experimental observation is modeled here through the reduction of the
damage encountered by concrete as the stress ratio R is increased. Since the compressive
damage parameter a is responsible for the magnitude of damage endured by concrete as
121
was shown in Fig. 4.9c, this parameter is related to the biaxial strain ratio ( 22 / 11 ) in an
effort to account for damage reduction during biaxial compressive loading in a
displacement-controlled environment. The same material parameters as those used for the
uniaxial compressive test are used here with f 0 =15.2 MPa (Wu et. al., 2006). The FE
setup shown in Fig. 4.5d was used.
Figure 4.14 The behavior of the proposed model under biaxial compression ( 11 = 22 )
It is worth mentioning here that the relation between damage growth and the damage
parameter a is a proportional relation; the higher the value of a , the higher is the
magnitude of damage . For example, if the damage magnitude is fixed at = 0.3 , and
122
a is increased from 11 to 22 MPa 1 , Fig. 4.15a can be obtained to show the damage
hardening function Z (MPa) plotted against a . Furthermore, if some arbitrary value is
assigned to the damage release rate, say Y = 0.01 MPa, and the value of a is increased
from 11 to 22 MPa 1 again, then damage growth can be potted against a as shown in
Fig. 4.15b.
By checking the biaxial compression stress envelop of concrete, Fig. 4.13, one can
easily see that when the stress ratio R is zero, the point on the envelop is (-1,0) and the
consolidation effect doesnt exist. On the other hand, the consolidation effect starts to
increase as R moves away from (-1,0) towards higher values of R up to a certain point
where the consolidation effect starts to decrease until the stress ratio R =1 is reached. By
plotting the values of the strain ratio 22 / 11 versus different power evolution equations
of the damage parameter a , it was realized that the equation that best describes the
evolution of a with respect to R is the one given as follows:
22 n
a = 22 11( ) (4.152)
11
Figure 4.16a shows different curves for the evolution of a obtained by plotting Eq.
(4.152) over the full range of the strain ratio 22 / 11 using different values of the power
parameter (n). Figure 4.16b, on the other hand, shows a symmetric half of the biaxial
compressive quadrant of the concrete envelop. The results in Fig. 4.16b are obtained by
varying the horizontal and vertical displacements in the input file during multiple runs in
order to obtain different stress ratios and their corresponding values of the damage
variable a . Table 4.2 shows the numerical values for the series of runs used to obtain
the curve representing (n = 1/6), which is the closest fit to experimental results reported
by Kupfer et. al., (1969) as shown in Fig. 4.17.
123
Table 4.2 Biaxial Compression Test (n = 1/6 )
22 / 11 11 22 22 / 11 11 / f c' 22 / f c' a
0 -27.6 0 0 -1 0 22
0.055 -33.9815 -11.0433 0.32498 -1.23121 -0.40012 15.22063
0.14 -35.018 -13.9127 0.397301 -1.26877 -0.50408 14.09824
0.2 -35.3416 -15.9112 0.450212 -1.28049 -0.57649 13.58803
0.4 -35.297 -21.4989 0.609086 -1.27888 -0.77895 12.55788
0.6 -34.4252 -25.9382 0.753465 -1.24729 -0.93979 11.89776
0.8 -33.3097 -29.444 0.883947 -1.20687 -1.06681 11.40158
0.9 -32.7517 -30.9081 0.94371 -1.18666 -1.11986 11.19147
1 -32.1874 -32.1874 1 -1.16621 -1.16621 11
Figure 4.18 The mixed (tension/compressive) quadrant of the biaxial concrete envelop
In the case of biaxial tension and compression, the model did not fit the experimental
results, even with the modification used for the biaxial compression case mentioned
above. The proposed model underestimates the strength of concrete under combined
124
tension and compression (Fig. 4.18). Varying the damage parameters a + and a with the
strain ratio 22 / 11 did not result in any improvements. This region of the concrete
strength envelop still requires further research in order for the proposed model to
adequately describe the biaxial tensile-compressive behavior of concrete. The results
shown in Fig. 4.18 were obtained by varying the magnitude of the damage release rates
Y to account for the different tensile and compressive contributions to damage growth.
Despite the difference between the numerical and experimental results observed here,
it is worth mentioning that the combined damage variable in the mixed quadrants of
the concrete envelope is made out of two parts, + and , unlike the other two
quadrants (pure tensile or compressive quadrants). These two damage variables, + and
125
, are combined using Eq. (4.127) to obtain the total damage variable (Fig. 4.19),
where the latter is used to update the Cauchy stress tensor.
Combining the results from different quadrants discussed above, the biaxial strength
envelop of concrete can be plotted as shown in Figure 4.20. The experimental results
pertain to Kupfer et. al., (1969).
In this section, the monotonic testing of the damaging process for a single-edge-
notched plain concrete beam is simulated using experimental data from Malvar and
Warren (1988). The square-cross-section beam, with an initial notch depth of 51 mm, is
subjected to three-point loading test as illustrated in Fig. 4.21a. The following material
properties are used (Lowes, 1999; Lee and Fenves, 2001): E = 2.17 104 MPa, f c' = 29.0
MPa, f 0+ = 2.4 MPa, G +f = 35 N/m. Two-dimensional FE mesh for the symmetric left
part of the specimen is shown in Fig. 4.21b. Displacement control is used to apply the
loading in the simulation.
126
Applying the material parameters given above along with the characteristic length of
the FE mesh given by ABAQUS as ( A = 12.6e 3 m) into Eq. (4.78), the mesh sensitivity
parameter + can be calculated and the adjusted value for the damage parameter
a + = 1400 MPa 1 can be obtained. Since the elastic modulus, E = 2.17 104 MPa, in this
test is lower than that considered in the uniaxial test, the tensile plastic hardening
parameter h is reduced here to h = 1.9 104 MPa to maintain the same ratio ( h / E ). In
Fig. 4.22, the load versus load-point deflection curve from the simulation is compared
with the experimental result of Malvar and Warren (1988). Good agreement between the
numerical and experimental results is observed, which demonstrates the effectiveness of
the proposed model.
It should be noted here that the non-smoothness of the curve beyond the peak load is a
phenomenon frequently observed in any numerical simulations that does not involve a
regularization technique (Feenstra, 1993; Fichant et. al., 1999; Lowes, 1999; Tikhomirov
and Stein, 2001; Sumarac et. al., 2003; Rabczuk et. al., 2005; He et. al., 2006; Nguyen
and Houlsby, 2008a,b; Yu et. al., 2008; and others). He et. al. (2006) reported that the
load-displacement curve beyond the maximum load and during the softening region, may
reach a valley which is then followed by a new local peak in the next increment. These
local peaks, if not robustly tolerated and accounted for, will cause the algorithm to
diverge.
The evolution of damage is demonstrated next using the two dimensional FE mesh
shown in Fig. 4.21b. By applying a displacement control of -0.5 mm to the point of load
application through 100 time increments, Figs. 4.23, 4.24 and 4.25 show how damage
propagates starting at the tip of the notch working its way toward the top of the cross-
section.
127
Figure 4.23 Evolution of damage ( ) at 50 time increments
128
CHAPTER 5
5.1 Introduction
Since Ngo and Scordelis published their landmark paper on RC beams in 1967,
important progress has been made in the FE-based numerical analysis of RC structures.
However, despite this progress, modeling the mechanical behavior of RC is still one of
the most difficult challenges in the field of structural engineering. This is due to the
inherent complexity and uncertainty concerning the properties of concrete, which make it
excessively difficult to develop accurate constitutive models or algorithms that are
sufficiently robust to obtain reliable and converged solutions in numerical analyses.
In order to ensure that the serviceability requirements are met in any RC structure, it is
necessary to predict the cracking and the deflection of the structure under service loads.
Therefore, estimation of the ultimate load is indeed essential in assessing the margin of
safety of RC structures against failure. Furthermore, it is necessary to predict the load
deformation behavior of the structure for responses ranging from elastic to inelastic as
well as under all possible loading conditions. Therefore, tracing the entire response of an
RC structure is an essential step in the process of understanding the performance of RC.
In this study, the two dimensional RC beam analysis will consist of three major
components: steel reinforcing bars, concrete material, and bond effect. Steel reinforcing
bars will be modeled using the discrete representation in the FE mesh. The reinforcing
129
steel elastoplastic material model, described in Chapter 3, will be used here with linear
strain hardening. The concrete behavior will be described using the elastic-plastic-
damage model discussed in Chapter 4. The effect of bond deterioration and the transfer
of the stresses from concrete to the reinforcing bars will be accounted for using the work
of Belarbi and Hsu (1994) which was modified later on by Kwak and Kim (2006). The
bond-slip effect along the reinforcing bars is quantified with the force equilibrium and
compatibility condition at the post-cracking stage and its contribution is indirectly
implemented into the stressstrain relation of reinforcing steel. The advantage of the
analytical procedures proposed by Belarbi and Hsu (1994) and Kwak and Kim (2006) is
taking into account the incorporation of the bond slip effect while using the conventional
discrete representation of steel, without the need for additional considerations such as
using double nodes or interface elements.
5.2 Bond-Slip Effect on the Stress Strain Relation of the Steel Reinforcement
Reinforcing steel is usually modeled as a linear elastic, linear strain hardening material
with a yield stress y (see Chapter 3). However, when reinforcing bars are surrounded
by concrete, the average behavior of the stressstrain relation is quite different, as shown
in Fig. 5.1 (Belarbi and Hsu, 1994). The most strikingly different feature is the lowering
of the yield stress below the value of y . Yielding of an RC member occurs when the
steel stress at a cracked section reaches the yield strength of the bare bar. However, the
average steel stress at a cracked element still maintains an elastic stress that is less than
the yield strength, because the concrete matrix located between cracks is still partially
capable of resisting tensile forces, owing to the bond between the concrete and the
reinforcement.
Figure 5.1 Stress Strain Relation for Steel Reinforcement (Belarbi and Hsu, 1994)
130
overestimation of the structural response at the post-yielding range. Since this
phenomenon is accelerated with increased deformation, an analysis of RC members
subjected to loading accompanied by relatively large deformations requires the use of
average stressstrain relations (Stevens et. al., 1991; Belarbi and Hsu, 1994).
Accordingly, the average stressstrain relation of steel needs to be defined so as to trace
the cracking behavior of RC beams up to the ultimate limit state. This can be
accomplished here by using continuum damage mechanics, in which the local
displacement discontinuities at cracks are distributed over the finite element and
wherethe behavior of cracked concrete is represented by the average stressstrain
relations (Stevens et. al., 1991). Considering these factors, Belarbi and Hsu (1994)
proposed the bilinear average stressstrain relation shown in Fig. 5.1, which was
introduced from experimental data.
n = y (0.93 2 B) (5.1)
where B is defined as B = ( ft ' / f y )1.5 / s , f y and y are the yield stress and the
corresponding yield strain of the bare bar, ft ' is the tensile strength of concrete, and s is
the steel reinforcement ratio.
Figure 5.2 Stress Strain Relation for Steel Reinforcement (Kwak and Kim, 2006)
Kwak and Kim (2006) elaborated on the above mentioned works and emphasized that
reinforcing bars transfer tensile stresses to concrete through the bond stresses located
along the surface between reinforcements and surrounding concrete. They extracted part
of an RC member subjected to uniaxial tension and bounded by two adjacent cracks and
used it as a free body diagram to obtain the equilibrium equations for concrete and steel
using a linear bond stress-slip relation. A reinforcing bar equivalent stress-strain relation
131
that incorporates the effect of bond-slip is therefore developed. Their final adopted stress-
strain relation for the reinforcing bars is shown in Fig. 5.2. The equivalent elastic
modulus Eeq is given as:
lt
where s1 is the strain in the steel bar assuming perfect bond, and eqs = s ( x)dx / lt .
0
The parameter lt represents the transfer length which can be determined following the
linear relationship proposed by Somayaji and Shah (1981) on the basis of experimental
data obtained from pull-out tests as follows:
Fc
lt = K p (5.3)
o
Steel reinforcement can now be included in the FE mesh as individual elements with
the equivalent stress-strain relation that accounts for bond effects while assuming perfect
bond between steel and concrete elements. Furthermore, the corresponding equivalent
modulus of elasticity for steel Eeq can be used up to the yielding point in the stressstrain
relation of the discrete reinforcing steel elements, as depicted by the solid line in Fig. 5.2.
The same ratio of Eeq to Es given by Belarbi and Hsu (1994) was assumed by Kwak and
Kim (2006) to be maintained at the post-yielding region.
The proposed model is applied to study the simply supported RC beam tested
experimentally by Burns and Siess (1962) and later modeled by Kwak and Filippou
Figure 5.3 Geometry and cross section of RC beam (Burns and Siess, 1962)
132
Figure 5.4 FE discretization of steel and concrete, loading, and support conditions
133
Figure 5.8 Idealized tensile damage distribution, coarse mesh 78 elements
Figure 5.9 Idealized tensile damage distribution, finer mesh 156 elements
Figure 5.10 Idealized tensile damage distribution, finer mesh 312 elements
78 Elements
156 Elements
312 Elements
134
Figure 5.12 Damage distribution, results obtained by Kwak and Filippou (1997)
(1997). The tested specimen consisted of a simply supported beam with a span of 12 ft
(3.7 m) which was subjected to a concentrated load at midspan. The geometry and the
cross section of the beam are shown in Fig. 5.3 and the material properties are given as
follows: The moduli of elasticity for concrete and steel are Ec = 3800 Ksi and Es =
29500 Ksi, respectively. The tensile and compressive strengths of concrete are f o+ =
0.347 Ksi and f c' = 4.82 Ksi, respectively. The yielding stress for the steel bars is f y =
44.9 Ksi. The reinforcement consists of #6 (0.75 in diameter) rebars with a reinforcement
ratio of s = 0.99%. The reinforcement is modeled using 2 dimensional elements with a
thickness. The constitutive model of these elements accounts for the bond effect as shown
in section (5.2). The Poissons ratios for concrete and steel are c = 0.167 and s =
0.333, respectively. The fracture energy is given as G f = 0.5 lb/in.
Only half of the beam is modeled in this FE simulation by taking advantage of the
symmetry in the geometry and loading. The FE discretization, the arrangement of the
steel reinforcement, and the loading and support conditions are shown in the Fig. 5.4.
Three analyses were carried out using four-nodded quadrilateral elements and 100
deflection/time increments. The same model parameters used for the concrete beam
analysis in chapter 4 are used here. The fracture energy along with the characteristic
lengths for different meshes provided by ABAQUS are used to reduce the effect of mesh
sensitivity. Figures 5.5, 5.6, and 5.7 show the results of the tensile damage variable + at
the end of the FE incremental procedures involving three different mesh sizes, 78, 156,
and 312 elements. Figures 5.8, 5.9, and 5.10 show idealized tensile damage distributions -
plotted over undeformed meshes - that are used to compare results to each other (Fig.
5.11) as well as to the damage distribution obtained by Kwak and Filippou (1997) as
shown in Fig. 5.12.
The load-deflection curves obtained using the three simulations discussed above are
shown in Figs. 5.13 through 5.16. The results are compared to the experimental output of
the work of Burns and Siess (1962). It can be seen that the numerical results exhibit
fluctuations that are related to convergence issues, especially in the zone where concrete
is softening under compression (crushing of concrete). The compressive damage material
parameters are the ones that govern the shape of the load-deflection curves beyond the 30
135
Ksi stress level. The mesh sensitivity is reduced by using the fracture energy related
coefficients .
Figure 5.13 Load-Deflection relations for RC beam (Burns and Siess, 1962)
Figure 5.14 Load-Deflection relations for RC beam (Burns and Siess, 1962)
136
Figure 5.15 Load-Deflection relations for RC beam (Burns and Siess, 1962)
Figure 5.16 Load-Deflection relations for RC beam (Burns and Siess, 1962)
Figures 5.17, 5.18, and 5.19 show the total damage ( ) distributions for the three
simulations discussed above. The total damage variable (scalar) is obtained using Eq.
4.129. It should be noted here that the total damage variable is a weighted average
function of the stress tensor as well as its spectral decomposition parts, and the tensile +
137
and compressive scalar damage variables. This averaging technique results in the
differences between the tensile damage distributions and the total damage distributions
plotted in the previously indicated figures.
The proposed model is also applied to study the simply supported RC beams tested
experimentally by Karihaloo (1992) and later modeled by Tikhomirov and Stein (2001).
Three point bending of the experimental RC specimens was carried out. Two beams with
one and two 12 mm diameter bars are analyzed until failure occurred in a displacement
controlled environment. The geometry and the cross sections of the beams are shown in
Figs. 5.20 and 5.26. The material properties used are as follows: The moduli of elasticity
138
for concrete and steel are Ec = 30 GPa and Es = 200 GPa, respectively. The tensile
strength of concrete is f o+ = 2.26 MPa. The yielding stress for the steel bars is f y = 463
MPa. The reinforcement is again modeled using 2 dimensional elements with a thickness.
The constitutive model of these elements did not account for the bond effect, since the
results underestimate the RC beam strength, see Figs. 5.25 and 5.27. The Poissons ratios
for concrete and steel are c = 0.2 and s = 0.3, respectively. Only half of each beam is
modeled in these FE simulations by taking advantage of the symmetry in the geometry
and loading conditions.
The first analysis was performed using a 360 element mesh with 45 elements
representing the steel reinforcement as can be seen in Fig. 5.21. The damage distributions
are similar in pattern to those of the previous example. The averaging of the properties
used in obtaining the total damage distribution is again the reason for the differences
between the tensile + and total damage distributions (see Figs 5.21 and 5.22).
In Fig. 5.23, damage is plotted over the undeformed mesh in order to compare the
proposed models damage distribution to that of Tikhomirov and Stein (2001), see Fig.
5.24. The load-deflection curves obtained using the FE simulation discussed above is
shown in Fig. 5.25. The results are compared to the experimental output of the work of
Karihaloo (1992). The numerical results exhibit fluctuations similar to those in Fig. 5.16,
139
Figure 5.22 Total Damage distribution, 360 elements mesh
140
and are again believed to be related to convergence issues. The models behavior
underestimates the strength of the RC beam in the intermediate stage where concrete is
deteriorating. This was not observed in the previous example, where all three simulations
showed overestimation of the strength in the same region. Nevertheless, the trend of the
results is close to the experimental ones demonstrating the ability of the proposed
models to capture the physical behavior of RC.
Figure 5.27 shows the load-deflection curves obtained using the FE simulation of the
second beam with two 12mm reinforcing bars. The same number of elements used in the
previous example is adopted here. The simulated results are again compared to the
experimental output of the work of Karihaloo (1992). The experimental results show a
more brittle failure of the RC beam when compared to the previous one (one 12mm bar)
141
where the beam fails through yielding of steel. This is a direct result of increasing the
amount of reinforcement (2 bars). The numerical behavior, on the other hand,
underestimates the strength of the RC beam in the intermediate stage where concrete is
deteriorating. The trend of the results is close to the experimental ones, demonstrating
again the ability of the proposed models to capture the physical behavior of RC.
142
CHAPTER 6
This study introduces a continuum FE approach that is appropriate for predicting the
physical behavior of RC members subjected to short term monotonic loading assuming
isothermal conditions. The macroscopic components of RC members are modeled using
separate constitutive models that are brought together through the consideration of the
steel-concrete bond and interaction analysis. Elastoplastic constitutive models with strain
hardening are introduced to model the behavior of steel reinforcement, and a continuum
approach based on damage mechanics and plasticity theory is adopted to describe the
complex behavior of concrete material in structural elements. The concept of energy
dissipation (fraction energy) is used in order to reduce the effect of mesh sensitivity on
the FE numerical results. Bond and interaction between steel and concrete is accounted
for by modifying the global stress-strain curve of steel reinforcement to account for the
reduction in the strength (elastic and plastic moduli and yield stress) observed
experimentally. All these components are integrated into the proposed approach in an
attempt to properly describe the complex physical behavior of RC composite materials.
The proposed material model for concrete is derived using rigorous and consistent
thermodynamic formulation. The additive decomposition of the Helmholtz free energy
concept is used to define the thermodynamic conjugate forces associated with the internal
state variables, including the damage thermodynamic conjugate forces (damage energy
release rates). The energy dissipation mechanisms are formulated to satisfy the first
inequality of thermodynamics, and to postulate the plastic and damage dissipation
functions. Three dissipation mechanisms (plasticity, tensile and compressive damage) are
present to control the dissipation process of the material model.
143
to the damaged configurations. A concrete plasticity yield function with multiple
isotropic hardening criteria and a non-associative plasticity flow rule is adopted to
represent the irreversible plastic behavior of concrete. The non-associative flow rule
includes a hydrostatic term to account for the dilatation effect of concrete materials. On
the other hand, two damage growth criteria that are based on the hydrostatic-deviatoric
sensitive thermodynamic-conjugate forces are used to model stiffness degradation and
material deterioration in concrete. Two damage variables, tensile and compressive, are
used to model the different damaging behaviors in concrete. These two damage variables
are combined using a relation that incorporates the stress tensor and its spectral
decomposition into tensile and compressive components. The combined damage variable
is consequently used to map the stress from the effective to the damaged configuration.
Fracture energy related coefficients have been defined and incorporated in the damage
model to achieve a reasonable degree of discretization insensitivity in numerical
calculations.
The computational algorithm for the concrete model is based on the operator split
concept, where the incremental constitutive equation is decomposed into elastic, plastic
and damage parts, leading to the corresponding numerical elastic-predictor, plastic-
corrector and damage-corrector steps. During the first two steps, the damage variables are
fixed, so that the elastic-plastic behavior is decoupled from damage, constituting a
standard elastic-plastic problem in the effective stress space. The damage variables and
the Cauchy stress tensor can then be updated correspondingly in the damage-corrector
step. An elastic-plastic implicit, damage explicit integration scheme is adopted. The
consistent elastic-plastic-damage tangent operator is derived and a flow chart of the
integration technique is provided.
The proposed concrete model is implemented and tested to verify its capability,
applicability, and effectiveness in capturing the material behavior in both tension and
compression under uniaxial and bi-axial loadings. The results obtained by the proposed
model are compared with corresponding experimental results to evaluate the models
performance. Tensile and compressive verification tests under unaxial loading are
demonstrated first, followed by biaxial tests in tension and compression. Then a three
point bending test of a notched beam is investigated.
The numerical results of the verification examples under tensile and compressive
unaxial loads demonstrate the effectiveness of the model in capturing the uniaxial
behavior of concrete. During all stages of loading, the simulated results match the
experimental ones in terms of physical behavior. The experimentally observed softening
branch of the stress-strain diagram under uniaxial loading (tension or compression) is
reproduced efficiently.
Under biaxial tension, the model easily depicts the experimentally observed
phenomenon; concrete strength suffers a reduction as the biaxial stress ratio is increased.
Whereas, under biaxial compression, an additional equation governing the damage
variables is introduced in order to account for the strengthening of concrete due to
144
consolidation under biaxial compression. This improvement leads to a more efficient
numerical representation of the experimental results under biaxial compression.
The proposed concrete model in its current form is not capable of capturing the
experimentally observed behavior under combined tensile and compressive loadings. The
reduction in compressive strength was not observed numerically as the tensile stress is
increased and vise versa. Further development of the model is therefore required in these
regions of loading to enhance the numerically simulated results.
When a three point bending test of a single notched concrete beam is investigated, the
computational algorithm demonstrated robustness in simulating the entire stress-strain
diagram of concrete and the evolution of damage. Nevertheless, some difficulties are
encountered: the non-smoothness in the numerical results beyond the peak point is
believed to be related to the global equilibrium iterations. Another possible source is the
use of non-associative plastic flow rule, where the direction of the plastic flow is not
normal to the yield surface. It is well documented in literature that local damage
approaches exhibit mesh sensitivity and non-smooth results in the softening region.
Regularization methods should be incorporated in order to overcome such drawbacks.
In order to analyze RC beams, the bond-slip effect along the reinforcing bars is
quantified with the force equilibrium and compatibility condition at the post-cracking
stage. This effect is indirectly implemented into the stressstrain relation of reinforcing
steel. The advantage of such procedure is taking into account the incorporation of the
bond slip effect while using the conventional discrete representation of steel, without the
need for additional considerations such as using double nodes, interface elements, or
modified finite element formulation to produce more complicated element library. This
approach is more suited to the material modeling using the UMAT subroutine, where all
the numerical procedures are written in terms of stresses and strains.
The implemented algorithms are then used to analyze two simply supported RC beams
subjected to concentrated loads applied at mid spans. The proposed model is able to
capture the three stages of loading discussed in Chapter 2. An attempt to reduce the
variation in the numerical results is carried out by means of adjusting the fracture energy
related parameters. As the mesh size is reduced, the results varied yet remained close to
one another, a behavior attributed to a non regularized constitutive approach.
145
6.2 Future Work
The use of the proposed model to predict the response of plain and RC systems
indicates a number of characteristics of the current model that could be enhanced to
improve modeling accuracy. It also identifies a number of areas in which additional
research focused on the model development and verification, can greatly enhance
simulation of RC structural behavior.
The most significant deficiency of the current model is the failure to represent the
biaxial behavior under combined tension and compression. Next is the dependence of the
numerical results on mesh refinement. Although this subject has been partially accounted
for using fracture energy induced parameters, regularization methods should be employed
in order to significantly trivialize the effect of mesh dependency. At the global level,
results of this study indicate that additional research is required to improve solution
algorithms for systems that represent material softening (non-smooth material response).
Although the calibration of the model parameters governing the initial yield and the
damage surfaces has been carried out, a proper and unified procedure for the
identification and determination of the model parameters and their effect under different
loading conditions is an important aspect yet to be addressed. A parametric study may
reveal important information regarding the parameters that are most susceptible to mesh
sensitivity effect, and therefore, help in calibrating these parameters using the fracture
energy concept.
146
In order to further investigate the effect of bond deterioration on the global behavior
of RC composite material, the traction forces at the interfaces between steel and concrete
finite elements can be studied and modeled using ABAQUS. This will result in the
introduction of the slip concept into the analysis as a result of the decay of the traction
forces. The slip was not accounted for in the current study since all the analysis was
based on modeling of stresses and strains in the RC composite material using the user
defined material subroutine UMAT.
147
BIBLIOGRAPHY
1. Taqieddin, Z. N. and Voyiadjis, G. Z., 2008. Elastic Plastic and Damage Model
for Concrete Materials. Part II: Implementation and Numerical Aspects. Sent for
publication.
2. Taqieddin, Z. N. and Voyiadjis, G. Z., 2008. Elastic Plastic and Damage Model
for Reinforced Concrete Materials. Sent for publication.
5. Voyiadjis, G. Z., Kattan, P.I., and Taqieddin, Z. N., 2005. Analysis of Damage in
Composite Materials using Fabric Tensors and the Continuum Approach.
Proceedings of McMat2005: Joint ASME/ASCE/SES Conference on Mechanics
and Materials, June 1 - 3, 2005, Baton Rouge, Louisiana, USA.
6. Voyiadjis, G. Z., Kattan, P. I., and Taqieddin, Z. N., 2007. Continuum Approach
to Damage Mechanics of Composite Materials with Fabric Tensors Int. J. of
Damage Mech., 16(7), 301-329.
7. Voyiadjis, G. Z. and Taqieddin, Z. N., 2008. Elastic Plastic and Damage Model
for Concrete Materials. Part I: Theoretical Formulation. Sent for publication.
10. Voyiadjis, G. Z., Taqieddin, Z. N., and Kattan, P. I., 2008. Anisotropic Damage-
Plasticity Model for Concrete, Int. J. of Plasticity, available online, in press.
148
Part II: By Other Authors
1. ABAQUS (2004) User Manual, version 6.4, ABAQUS Inc., Pawtucket, Rhode
Island, 2004.
3. Adenaes, E., Gerstle, K.H. and Ko, H.Y., 1977. Response of mortar and concrete
to biaxial compression. J. Eng. Mech. Div. ASCE, 103(4), 515-526.
4. Ananiev, S. and Obolt, J., 2007. Plastic-damage model for concrete in principal
directions. Condensed Matter-Materials Science, Li V. et. al. (eds.). FraMCoS-5
Fracture Mechanics of Concrete and Concrete Structures. 8 pages.
5. Armero, F. and Oller, S., 2000. A general framework for continuum damage
models. I. Infinitesimal plastic damage models in stress space. Int. J. Solids
Struct., 37(48-50), 74097436.
8. Bahn, B. Y., and Hsu, C. T., 1998. Stress-strain behavior of concrete under
cyclic loading, ACI Materials Journal, 95(2), 178-193.
10. Bazant, Z.P., Ozbolt, J. 1990. Nonlocal microplane model for fracture, damage
and size effect in structures. J. Eng. Mech. ASCE, 116(11), 2485-2505.
11. Bazant, Z.P., Pan, J., Pijaudier-Cabot, G., 1987. Softening in reinforced concrete
beams and frames. J. Struct. Eng., ASCE, 133(12), 2333-2347.
12. Belarbi, A., Hsu, TTC., 1994. Constitutive laws of concrete in tension and
reinforcing bars stiffened by concrete. ACI Structural Journal, 91(4), 46574.
13. Ben Romdhane, M. R. and Ulm, F. J., 2002. Computational mechanics of the
steel-concrete interface. Int. J. Numer. Anal. Meth. Geomech., 26:99-120.
149
14. Bicanic, N. and Pearce, C. J., 1996. Computational aspects of a softening
plasticity model for plain concrete. Mech. of Cohesive-frictional Mat., 1(1), 75-
94.
15. Burns, N.H. and Siess, C.P., 1962. Load-Deformation Characteristics of Beam-
Column Connections in Reinforced Concrete. Civil Engineering Studies, SRS
No. 234, University of Illinois, Urbana.
16. Chaboche, J. L., 1993. Development of continuum damage mechanics for elastic
solids sustaining anisotropic and unilateral damage. Int. J. Damage Mech., 2,
311-329.
17. Chaboche, J. L., Lesne, P. M. and Maire, J. F., 1995 "Continuum Damage
Mechanics, Anisotropy and Damage Deactivation for Brittle Materials Like
Concrete and Ceramic Composites", Int. J. Damage Mech., Vol. 4, pp. 5-22.
18. Chen, F.W., 1982. Plasticity in reinforced concrete. McGraw-Hill, New York.
19. Chen, A. C. T. and Chen, F. W., 1975. Constitutive relations for concrete. J.
Eng. Mech. Div. ASCE, 101(4), 465-481.
20. Chen, F. W. and Han, D. J., 1988. Plasticity for structural engineers. Springer-
Verlag, New york.
21. Chen, E. Y. and Schnobrich, W. C., 1981. Material modeling of plain concrete.
Advanced Mechanics of Reinforced Concrete, IABSE Colloquium, Delft, 33-51.
22. Chow, C. L. and Lu, T. J., 1989. On evolution laws of anisotropic damage. Eng.
Fract. Mech. 34(3), 679-701.
23. Chow, C. L., and Wang, J., 1988. Ductile fracture characterization with an
anisotropic continuum damage theory. Eng. Fract. Mech., 30(5), 547-563.
25. Clement, J. L., Mazars, J. and Zaborski, A., 1985. A damage model for concrete
reinforcement bond in composite concrete structures. Struct. and Crack
Propagation in Brittle Matrix Composites. A. M. Brandt, ed., Elsevier, New York,
N. Y., 443-454.
26. Coleman, B. D. and Gurtin, M. E., 1967. Thermodynamics with internal state
variables. J. Chem. Phys., 47, 597613.
27. Comi C. and Perego U., 2001. Fracture energy based bi-dissipative damage
model for concrete. Int. J. of Solids and Struct., 38(36), 6427-6454.
150
28. Contrafatto, L. and Cuomo, M., 2006. A framework of elasticplastic damaging
model for concrete under multiaxial stress states. Int. J. of Plasticity, 22(12),
2272-2300.
29. Cox, J. V. and Herrmann, L. R., 1998. Development of a plasticity bond model
for steel reinforcement. Mech. of Cohes. Frict. Mater. 3, 155-180.
30. de Borst, R., Remmers, J., Needleman, A., and Abellan, M., 2004. Discrete vs
smeared crack models for concrete fracture: bridging the gap. Int. J. Numer.
Anal. Meth. Geomech., 28, 583-607.
31. Doghri, I., 2000. Mechanics of Deformable Solids Linear, Nonlinear, Analytical
and Computational Aspects. Springer Verlag, Berlin, Germany.
32. Dragon, A. and Mroz, Z., 1979. A continuum theory for plastic-brittle behavior
of rock and concrete. Int. J. Eng. Sci., 17(2), 121-137.
33. Dragosavic, M. and Groeneveld, H., 1984. Bond model for concrete structures.
Int. Conf. on Computer Aided Analysis and Design of concrete Structures, F.
Damjanic et. al., ed., Pineridge Press, 203-214.
34. Dunne, F.P.D., and Petrinic, N., 2005. Introduction to Computational Plasticity.
Oxford University Press.
35. Fantilli, A. P. , Ferretti, D., Iori, I., and Vallini, P., 2002. Mechanical Model for
Failure of Compressed Concrete in Reinforced Concrete Beams. J. Struct.
Engrg., 128( 5), 637-645.
36. Faria, R. Oliver, J. and Cervera, M., 1998. A strain-based plastic viscous-damage
model for massive concrete structures. Int. J. Solids and Struct., 35(14), 1533-
1558.
37. Feenstra P.H., 1993. Computational aspects of biaxial stress in plain and
reinforced concrete Dissertation, Delft University, the Netherlands.
38. Feenstra, P.H. and Borst, R.D., 1996. A composite plasticity model for
concrete. Int. J. of Solids and Struct., 33(5), 707-730.
39. Ferrara, L. and di Prisco, M., 2001. Mode I Fracture Behavior in Concrete:
Nonlocal Damage Modeling. J. Engrg. Mech., 127(7), 678-692.
40. Fichant, S., La Borderie, C., Pijaudier-Cabot, G., 1999. Isotropic and anisotropic
descriptions of damage in concrete structures. Int. J. Mech. Cohesive Frictional
Mater., 4, 339359.
151
41. Gatuingt, F., and Pijaudier-Cabot G., 2002. Coupled Gurson Plasticity and
Damage Model for Concrete Structures in Dynamics.15th ASCE Engineering
Mechanics Conference, June 2-5, Columbia University, New York, NY.
42. Ghandehari, M., Krishnaswamy, S. and Shah, S., 1999. Technique for
Evaluating Kinematics between Rebar and Concrete. J. Eng. Mech., 125(2), 234-
241.
43. Gopalaratnam, V.S. and Shah, S.P., 1985. Softening response of plain concrete
in direct tension. ACI Journal, 82, 311-323.
44. Grassl, P., Lundgren, K.; Gylltoft, K., 2002. Concrete in compression: a
plasticity theory with a novel hardening law. Int. J. of Solids and Struct., 39(20),
5205-5223.
45. Grassl, P. Jirsek, M., 2006. Damage-plastic model for concrete failure. Int J. of
Solids and Struct., 43(22-23), 7166-7196.
46. Han, D. J. and Chen, W. F., 1985. A nonuniform hardening plasticity model for
concrete materials. J. Mech. Mat., 4(4), 283-302.
48. Hansen, E., Willam, K., and Carol, I., 2001. A two-surface anisotropic
damage/plasticity model for plain concrete. Proceedings of Framcos-4
Conference Paris, May 28-31, 2001, Fracture Mechanics of Concrete Materials,
R. de Borst (ed.), A.A. Balkema, Rotterdam, 549-556.
49. He, W., Wu, Y. F., Liewand, K.M., Wu, Z., 2006. A 2D total strain based
constitutive model for predicting the behaviors of concrete structures. Int. J. of
Eng. Science, 44(18-19), 1280-1303.
50. Hoehler, M. S. and Stanton, J. F., 2006. Simple Phenomenological Model for
Reinforcing Steel under Arbitrary Load. Journal of Structural Engineering,
132(7), 1061-1069.
51. Hordijk, D. A., 1991. Local approach to fatigue of concrete. Dissertation, Delft
University of Technology.
152
53. Jason, L., Huerta, A., Pijaudier-Cabot, G. and Ghavamian, S., 2006. An elastic
plastic damage formulation for concrete: Application to elementary tests and
comparison with an isotropic damage model. Comp. Meth. in Applied Mech. and
Eng., 195(52), 7077-7092.
54. Jefferson, A.D., 2003. Crafta plastic damage contact model for concrete - I.
Model theory and thermodynamic considerations. Int. J. Solids Struct., 40, 5973-
5999.
55. Jirsek, M., 2004. Non-local damage mechanics with application to concrete.
RFGC 8/2004. Failure, Degradation and Instabilities, 683-707.
57. Ju, J. W., Monterio, J. M., and Rashed, A. I., 1989. Continuum damage of
cement paste and mortar as affected by porosity and sand concentration. J. Eng.
Fract. ASCE, 115(1), 105-130.
59. Kachanov, L. M., 1958. On the creep fracture time. Izv. Akad. Nauk USSR
Otd. Tech 8, 26-31 (in Russian).
62. Karsan, I.D., Jirsa, J.O., 1969. Behaviour of concrete under compressive
loadings. Journal of Structure Division. ASCE, 95(12), 25352563.
63. Khan, A.R., Al-Gadhib, A.H., Baluch, M.H., 2007. Elasto-damage Model for
High Strength Concrete Subjected to Multiaxial Loading. Int. J. of Damage
Mech., 16(3), 361-398.
65. Kotsovos, M. D., and Newman, J. B., 1977. Behavior of concrete under
multiaxial stress. ACI, 74(9), 443-444.
153
66. Krajcinovic, D., 1979. A distributed damage theory for beams in pure bending.
J. Appl. Mech., 46(3), 592-596.
67. Krajcinovic, D., 1983a. Constitutive equations for damaging materials. J. Appl.
Mech., 50, 355-360.
68. Krajcinovic, D., 1983b. Continuum damage mechanics. Appl. Mech. Rev., 37,
1-6.
69. Krajcinovic, D., 1985. Continuous damage mechanics revisited: Basic concepts
and definitions. J. Appl. Mech., 52(4), 829-834.
70. Krajcinovic, D. and Fanella, D., 1986. A micromechanical damage for concrete.
Eng. Fract. Mech., 25(6), 585-596.
71. Krajcinovic, D. and Fonseka, G. U., 1981. The continuous damage theory of
brittle materials I and II. J. Appl. Mech., 48(4), 809.
72. Krajcinovic, D. and Lemaitre, J., 1986. Continuum damage mechanics theory
and applications. Springer-Verlag, Berlin.
73. Krtzig, W. B. and Plling, R., 2004. An elasto-plastic damage model for
reinforced oncrete with minimum number of material parameters. Computers &
Structures, 82(15-16), 1201-1215.
74. Kukla, K., Biani, N., Pearce, C.J., Pamin, J., 2005. Numerical analysis of
fracture in concrete. 13th ACME Conference: University of Sheffield, March 21-
22 2005.
75. Kupfer, H.B., Hildorf, H.K., and Rusch, H., 1969. Behavior of concrete under
biaxial stresses. ACI Journal 66 (8), 656666.
76. Kwak, H. G. and Filippou F. C., 1997. Nonlinear FE Analysis of R/C Structures
Under Monotonic Loads. Computers & Structures, 65(1), l-16.
77. Kwak, H.G. and Kim, J. K., 2006. Implementation of bond-slip effect in analyses
of RC frames under cyclic loads using layered section method. J. Eng. and
Struct., 28, 1715-1727.
78. Labadi, Y. and Hannachi, N. E., 2005. Numerical Simulation of Brittle Damage
in Concrete. Strength of Materials, 268-281.
79. La Borderie, C., Mazars, J., Pijaudier-Cabot, G., 1992. Response of plain and
reinforced concrete structures under cyclic loading. ACI SIP, 134, 14772.
154
80. Lammer, H., Tsakmakis, C., 2000. Discussion of coupled elastoplasticity and
damage constitutive equations for small and finite deformations. Int. J. Plasticity
16, 495523.
82. Lee, J. and Fenves, G. L., 1998. A plastic-damage model for cyclic loading of
concrete structures. J. Eng. Mech., ASCE, 124, 892-900.
83. Lee, J. and Fenves, G. L., 2001. A return-mapping algorithm for plastic-damage
models: 3-D and plane stress formulation. Int. J. Num. Meth. Eng., 50, 487-506.
84. Lemaitre, J., 1984. How to Use Damage Mechanics. Nuclear Engineering
Design, 80(2), 233245.
87. Liang, Q. Q., Uy, B., Bradford, M. A. and Ronagh, H. R., 2005. Strength
Analysis of SteelConcrete Composite Beams in Combined Bending and Shear.
J. of Struc. Eng., 131(10), 1593-1600.
88. Limkatanyu, S. and Spacone, E., 2003. Effects of reinforcement slippage on the
non-linear response under cyclic loading of RC frame structures. Earthquake
Engng. Struct. Dyn., 32, 2407-2424.
89. Loland, K. E., 1980. Continuous Damage model for load-response estimation of
concrete. Cem. Concr. Res., 10(3), 395-402.
90. Lowes, L. N., 1999. Finite Element Modeling of Reinforced Concrete Beam-
Column Bridge Connections. Dissertation, University of California, Berkeley,
California, USA.
92. Lubliner, J., Oliver, J., Oller, S. and Onate, E., 1989. A plastic-damage model for
concrete. Int. J. Solids Struct., 25(3),299-326.
155
93. Luccioni, B. M., Lopez, D. E. and Danesi, R. F., 2005. Bond-slip in reinforced
concrete element. J. Struct. Eng., 131(11), 1690-1698.
94. Luccioni, B., Oller, S. and Danesi R., 1996. Coupled plastic-damaged model.
Comp. Meth. in Applied Mech. and Eng., 129(1-2), 81-89.
95. Luccioni, B.M. and Rougier, V.C., 2005. A plastic damage approach for
confined concrete. Computers & Structures, 83(27), 2238-2256.
96. Marfia, S., Rinaldi, Z., and Sacco, E., 2004. Softening behavior of reinforced
concrete beams under cyclic loading. Int. J. of Solids and Struct., 41(11-12),
3293-3316.
97. Malvar, L. J., Warren, G., E., 1988. Fracture energy for three-point-bend tests on
single-edge-notched beams. Exp. Mech., 45, 266-272.
99. Mazars, J., 1986. A model of unilateral elastic damageable material and its
application to concrete. Fracture toughness and fracture energy of concrete, F. H.
Wittmann, ed., Elsevier Science Publishers, Amsterdam, Netherlands, 61-71.
101. Menzel, A., Ekh, M., Runesson, K., Steinmann, P., 2005. A framework for
multiplicative elastoplasticity with kinematic hardening coupled to anisotropic
damage. Int. J. of Plasticity, 21, 397434.
102. Mindess, S., 1983. The application of fracture mechanics of cement and
concrete: A historical review, Fracture mechanics of concrete, F. H. Wittmann,
ed., Elsevier, New York, 1-30.
103. Mirza, S. M., Houde, J., 1979. Study of bond stressslip relationships in
reinforced concrete. ACI J. 76(1), 1945.
105. Monti, G. and Spacone, E., 2000. Reinforced concrete fiber beam element with
bond-slip. J. Struct. Eng., 126(6), 654-661.
156
106. Murakami, S., 1983. Notion of continuum damage mechanics and its
application to anisotropic creep damage theory. J. Eng. Tech., 105, 99-105.
108. Murakami, S. and Ohno, N., 1981. A continuum theory of creep and creep
damage. Proc. Of the third IUTAM Symposium on creep structures. Springer,
Berlin, 422-444.
110. Ngo, D. and Scordelis, A.C., 1967. Finite element analysis of reinforced
concrete beams. Journal of ACI, 64(3), 152-163.
115. Oliver, J., Cervera,M., Oller, S., Lubliner, J., 1990. Isotropic damage models
and smeared crack analysis of concrete. In: Proc. 2nd Int. Conf. Comput. Aided
Analysis Design Conc. Structures, Zell am See, 945957.
116. Oller, S.; Onate, E.; Oliver, J. and Lubliner, J., 1990. Finite element nonlinear
analysis of concrete structures using a plastic damage model. Eng. Fracture
Mech., 35(13), 219231.
117. Ollivier, J.P., Maso, J.C., Bourdette, B., 1995. Interfacial Transition Zone in
Concrete. Advanced Cement Based Materials, 2(1), 30-38.
157
118. Oritz, M., 1985. A constitutive theory for the inelastic behavior of concrete.
Mech. Mat., 4(1), 67-96.
119. Oritz, M. and Popov, E. P., 1982. Plain concrete as a composite material,
Mech. Mat., 1(2), 139-150.
121. Panoskaltsis, V. P. and Lubliner, J., 1994. Rate dependent plasticity and
damage for concrete. Proc. of the Eng. Foundation Conference, Cement
Manufacture and Use, 27-40.
122. Park, H. and Kim, J. Y., 2005. Plasticity model using multiple failure criteria
for concrete in compression. Int. J.of Solids and Struct., 42(8), 2303-2322.
123. Peterson, P. E., 1981. Crack growth and development of fracture zones in plain
concrete and similar materials. Rep. No. TVBM-11-6, Division of Building
Materials, University of Lund, Lund, Sweden.
124. Phuvoravan, K. and Sotelino E. D., 2005. Nonlinear Finite Element for
Reinforced Concrete Slabs. J. Struct. Engrg., 131(4), 643-649.
125. Pijaudier-Cabot, G., Mazars, J. and Pulikowski, J., 1991. Steel-concrete bond
analysis with nonlocal continuous damage. J. Struct. Eng., 117(3), 862-882.
127. Rabczuk, T., Akkermann, J. and Eibl, J., 2005. A numerical model for
reinforced concrete structures. Int. J. of Solids and Struct., 42(5-6), 1327-1354.
130. Resende L., 1987. Damage mechanics constitutive theory for the inelastic
behaviour of concrete. Comput. Meth. Appl. Mech. Eng., 60, 5793.
131. Rots, J. G., 1988. Computational modeling of concrete failure. Thesis, Delft
Uni. Tech., Delft, the Netherlands.
158
132. Sain T. and Kishen, J.M.C., 2007. Energy-based equivalence between damage
and fracture in concrete under fatigue. Eng. Fracture Mech., 74(15), 2320-2333.
133. Salari, M. R., and Spacone, E., 2001. Finite element formulation of one-
dimensional elements with bond slip. Engineering Structures, 23, 815-826.
134. Salari, M. R., Saeb, S., Willam, K. J., Patchet, S. J., and Carrasco, R. C., 2004.
A coupled elastoplastic damage model for geomaterials. Comp. Meth. in
Applied Mech. and Eng, 193(27-29), 2625-2643.
135. Selby, R.G., and Vecchio, F.J., 1997. A Constitutive Model for Analysis of
Reinforced Concrete Solids. Canadian Journal of Civil Engineering, 24(3), 460-
470.
136. Shao, J.F., Jia, Y., Kondo, D. and Chiarelli, A.S., 2006. A coupled
elastoplastic damage model for semi-brittle materials and extension to unsaturated
conditions. Mech. of Mat., 38(3), 218-232.
137. Shen, X., Yang, L., and Zhu, F., 2004, A Plasticity-based Damage Model for
Concrete. Adv. in Struct. Eng., 7(5), 461-467.
140. Simo, J. C. and Ju, J. W., 1987a. Strain and stress-based continuum damage
model. Part I: Formulation. Int. J. Solids Struct., 23(7), 821-840.
141. Simo, J. C. and Ju, J. W., 1987b. Strain and stress-based continuum damage
model. Part II: Computational aspects. Int. J. Solids Struct., 25(7), 841-869.
142. Simo, J.C, and Taylor, R.L, 1985. "Consistent tangent operators for rate-
independent elastoplasticity". Comp. Meth. in App. Mech. and Eng., 48, 101-18.
143. Soh, C. K., Chiew, S. P., Dong, Y. X., 1999. Damage Model for Concrete-
Steel Interface. Journal of Eng.Mech, 125(8), 979-983.
144. Soh, C. K., Liu, Y., Dong, Y. X., and Lu X. Z., 2003. Damage Model Based
Reinforced-Concrete Element. J. Mat. in Civ. Engrg., 15(4), 371-380.
145. Somayaji, S., Shah, S. P., 1981. Bond stress versus slip relationship and
cracking response of tension members. ACI J., 78(3), 21725.
159
146. Spacone, E. and El-Tawil, S., 2004. Nonlinear Analysis of Steel-Concrete
Composite Structures: State of the Art. J. of Struct. Eng., 130(2), 159-168.
147. Steinmann, P., Miehe, C., Stein, E., 1994. Comparison of different finite
deformation inelastic damage models within multiplicative elastoplasticity for
ductile materials. Comput. Mech. 13, 458474.
148. Stevens, N.J., Uzumeri, S. M., Collins, M. P., Will, G. T., 1991. Constitutive
model for reinforced concrete finite element analysis. ACI Struct. J., 88(1), 49-
59.
149. Sumarac, D., Sekulovic, M., and Krajcinovic, D., 2003. Fracture of Reinforced
Concrete Beams Subjected to Three Point Bending. Int. J. of Damage Mech.,
12(1), 31-44.
150. Tao, X. and Phillips, D. V., 2005. A simplified isotropic damage model for
concrete under bi-axial stress states. Cement and Concrete Composites, 27(6),
716-726.
151. Tikhomirov, D. and Stein, E., 2001. Finite element computations of anisotropic
continuum damage in reinforced concrete. Computers & Structures, 79(22-25),
2249-2260.
153. Voyiadjis, G. Z. and Abu Lebdeh, T. M., 1992. Biaxial testing of repaired
concrete. ACI J., 564-573.
154. Voyiadjis, G. Z. and Abu Lebdeh, T. M., 1993. Damage model for concrete
using bounding surface concept. J. Eng. Mech., 119(9), 1865-1885.
155. Voyiadjis, G. Z. and Abu Lebdeh, T. M., 1994. Plasticity model for concrete
using the bounding surface concept. Int. J. Plast., 10(1), 1-21.
156. Voyiadjis, G. Z. and Deliktas, B., 2000. A coupled anisotropic damage model
for the inelastic response of composite materials. Comp. Meth. Appl. Mech.
Eng., 183, 159-199.
157. Voyiadjis, G. Z., Deliktas, B. and Aifantis, E. C., 2001. Multiscale analysis of
multiple damage mechanics coupled with inelastic behavior of composite
materials. J. Eng., 127(7), 636-645.
158. Voyiadjis, G. Z. and Kattan, P. I., 1989. Local Approach to Damage in Elasto-
Plastic Metal Matrix Composites. Int. J. Damage Mech., 2(1), 92-114.
160
159. Voyiadjis, G. Z. and Kattan, P. I., 1990. A coupled theory of damage
mechanics and finite strain elasto-plasticity. I: Damage and elastic deformations.
Int. J. Eng. Sci., 28(6), 505-524.
160. Voyiadjis, G. Z. and Kattan, P. I., 1992a. A plasticity-damage theory for large
deformations of solids. Part I: Theoretical formulation. Int. J. Eng. Sci., 30,
1089-1108.
161. Voyiadjis, G. Z. and Kattan, P. I., 1992b. Finite strain plasticity and damage in
constitutive modeling of metals with spin tensors. Appl. Mech. Rev., 45, S95-
S109.
163. Voyiadjis, G. Z., and Kattan, P. I., 2006, Advances in Damage Mechanics:
Metals and Metal Matrix Composites, with an Introduction to Fabric Tensors,
Second Edition, Elsevier, Oxford.
164. Voyiadjis, G. Z. and Park, T., 1997. Local and interfacial damage analysis of
metal matrix composites using finite element method. Eng. Fract. Mech., 56(4),
483-511.
165. Voyiadjis, G.Z., Park, T., 1999. Kinematics description of damage for finite
strain plasticity. Int. J. Eng. Sci., 37, 803830.
167. Wang, C.K., Salmon, C.G., 1998. Reinforced Concrete Design, Sixth Edition.
Addison Wesley Education Publisher.
168. Wastiels, J., 1980. A softening plasticity model for concrete. Proc.
International conference of the numerical method of nonlinear Problems,
Pinebridge Press, Swansea, U.K., 481-493.
169. Willam, K. J., Hurlbut, B. and Sture, S., 1986. Experimental and constitutive
aspects of concrete failure. In Finite Element Analysis of Reinforced Concrete
Structures. Edited by C. Meyer and H. Okamura, ASCE, NY, 226-245.
170. Willam, K., Rhee, I., Beylkin, G., 2001. Multiresolution Analysis of Elastic
Degradation in Heterogeneous Materials. Meccanica, 36(1), 131-150.
161
171. Willam, K., Rhee, I., and Xi, Y., 2003. Thermal degradation of heterogeneous
concrete materials. Special issue on Durability, J. Mat. Civil Eng., ASCE, 17(3),
276-285.
172. Wu, J. Y., Li, J., and Faria, R., 2006. An energy release rate-based plastic-
damage model for concrete. Int. J. of Solids and Struct., 43(3-4), 583612.
173. Yazdani, S. and Schreyer, H. L., 1990. Combined plasticity and damage
mechanics model for plain concrete. J. Eng. Mech., ASCE, 116(7), 1435-1450.
174. Yu, R. C., Ruiz, G., and Chaves, E. W.V., 2008. A comparative study between
discrete and continuum models to simulate concrete fracture. Eng. Fracture
Mech., 75, 117127.
162
APPENDIX A
> restart;
> with(linalg):
Warning, the protected names norm and trace have been redefined and
unprotected
163
0.07005077413 0.1600873628 0.1987855614
Egmat3 := 0.1600873628 0.3658484015 0.4542855763
0.1987855614 0.4542855763 0.5641008243
>Egval1:=multiply(multiply(transpose(Egvec1),Stress),Egvec1);
Egval1 := -0.02414811403
>
Egval2:=multiply(multiply(transpose(Egvec2),Stress),Egvec2);
Egval2 := 2.940332529
>
Egval3:=multiply(multiply(transpose(Egvec3),Stress),Egvec3);
Egval3 := 14.08381559
>
Stress:=matadd(matadd(Egmat1,Egmat2,Egvals[1],Egvals[2]),Egma
t3,1,Egvals[3]);
1.000000001 2.000000003 3.000000000
Stress := 2.000000003 7.000000006 5.000000000
3.000000000 5.000000000 8.999999996
> sigma:=matrix(3,3,[0,0,0,0,0,0,0,0,0]);
0 0 0
:= 0 0 0
0 0 0
> for k from 1 to 3 do
for i from 1 to 3 do
for j from 1 to 3 do
sigma[i,j]:=sigma[i,j]+Egvals[k]*vecs[i,k]*vecs[j,k]
end do:
end do:
end do:
> print(sigma);
1.000000001 2.000000003 3.000000000
2.000000003 7.000000006 5.000000000
3.000000000 5.000000000 8.999999996
164
APPENDIX B
165
VITA
Ziad N. Taqieddin was born in Amman, Jordan. Ziad graduated from the Educational
College and received a certificate of excellence for his distinguished accumulated
average (92.7%) in the Jordanian national high-school examination, 1996. He
subsequently joined the Civil and Environmental Engineering Program at the Applied
Science University (ASU), Amman, Jordan, to pursue his bachelor degree. Five years
later, 2001, he graduated with a bachelor and first degree honors for rating excellent and
ranking first among his class. He received the prestigious ASU Golden Watch for his
distinguished academic performance. Engineer Ziad was immediately hired by the Civil
and Environmental Engineering Department, ASU, as a teaching assistant. He helped in
teaching several courses and laboratory practices.
Ziad continued his path of passion for science towards his doctoral degree with his
mentor, Boyd Professor George Z. Voyiadjis. Ziad got married to the love of his life
Rawan in 2006. Two years later, he completed the research presented in this dissertation
and maintained a 4.0 GPA. Ziad graduated with a structures/mechanics Doctor of
Philosophy degree in civil engineering is August, 2008.
Throughout his stay at LSU, Ziad was always involved in research in addition to his
teaching responsibilities. He indulged himself in the mechanics of solids and structures,
mechanics of composites, computational mechanics, plasticity, fracture and damage
mechanics. He has several publications in prestigious journals in the field of engineering
mechanics. He has also participated in a number of internationally recognized
conferences.
166