BARI, MD. SHAFIQUL Constitutive Modeling for Cyclic Plasticity and Ratcheting.
(under the supervision of Dr. Tasnim Hassan)
BY
SHAFIQUL BARI
DOCTOR OF PHILISOPHY
RALEIGH
JANUARY 2001
APPROVED BY:
Shafiqul Bari was born on January 11th, 1969 in Faridpur, Bangladesh. After
graduating from the Bangladesh University of Engineering and Technology in 1993, with
a degree in Civil Engineering, he joined the University as a lecturer in the Department of
Civil Engineering. He obtained the Master of Science degree from the same university in
1996. In the fall of 1996, he enrolled at the North Carolina State University to pursue his
doctoral studies in Engineering Mechanics. After completing his doctoral study, he will
work for a company in the automobile industry at Michigan, USA.
ii
ACKNOWLEDGEMENTS
I sincerely thank Dr. Tasnim Hassan for his guidance and support during my
research. His time and effort in guiding the direction of this research was indispensable in
the completion of this thesis. He was always there to support me in every aspect of my
stay at North Carolina State university.
I would like to express my thanks to Prof. Vernon Matzen for his constant support
and suggestions during the course of this thesis. Working with him in conducting
experiments on nuclear power plant piping components for monotonic and cyclic loading
was an enlightening experience for me.
I would also like to express my gratitude to Prof. Kerry Havner for his invaluable
suggestions and counsel. His comments have been extremely valuable in the preparation
of this thesis.
I would also like to thank Dr. Mohammed Zikry for his time, encouragement and
suggestions.
Special Thanks should go to Dr. James Nau for his support and encouragement
during my stay at NC State University.
I like to acknowledge gratefully the financial support provided by the Center for
Nuclear Power Plant Structures, Equipment and Piping and the Department of Civil
Engineering to conduct this research.
I would like to express my heartfelt gratitude to the professors, staff and students
at the Center for Nuclear Power Plant Structures, Equipment and Piping for helping me in
numerous occasions during the course of this thesis.
A very special note of thanks and appreciation goes to my wife, Tahsina Ahmed
for her constant emotional and moral support to keep me in focus.
iii
TABLE OF CONTENTS
List of Tables v
List of Figures vi
1 Introduction 1
Paper I 9
Paper II 49
Paper III 77
References 131
iv
LIST OF FIGURES
1.1 Loading histories; (a) Uniaxial stress cycles, (b) Axial strain cycle with 4
constant pressure, (c) biaxial bowtie cycle, (d) biaxial reverse bowtie
cycle.
1.2 Axial strain ratcheting response for uniaxial stress controlled cycles with a 5
mean stress (from Hassan and Kyriakedes [1992])
1.3 Circumferential strain ratcheting response for (a) constant pressure and (b) 6
bowtie loading histories of Fig 1.1 (from Hassan et al. [1992] and Corona
et al. [1196])
5.2 Yield surface deformation law from Phillips and Tang [1972] 107
5.3 A schematic diagram of deformation and translation of yield surface from 109
Eisenberg and Yen [1984]
5.6 Equiplasticstrain surface I in the moving stress space i of the Ilyushin 117
space
5.7 Biaxial ratcheting predictions by model I and experiments from Hassan 119
and Kyriakides [1992].
5.9 Biaxial ratcheting predictions by model II and experiments from Hassan 123
and Kyriakides [1992].
v
CHAPTER 1
INTRODUCTION
Ratcheting, one of the low cycle fatigue responses, is defined as the accumulation
of plastic strain with cycles. There are many type of structures that are subjected to cyclic
loading where the stress state exceed the elastic limit of the materials used. For design
and analysis of these type of structures, accurate prediction of ratcheting response is
critical as ratcheting can lead to catastrophic failure of the structures. Even for structures
that are designed to be within the elastic limit, plastic zones may exist at discontinuities
or at the tip of cracks. The fatigue cracks can initiate at these plastic zones. Therefore,
better simulation model for cyclic plasticity response is important for the prediction of
the high cycle fatigue life as well.
1
yield surface in stress space) is attributed to be the primary reason for ratcheting.
Therefore, in order to develop and verify a basic model for ratcheting simulation, it is
essential to study the ratcheting responses of stabilized materials. After achieving a
robust model for ratcheting responses of cyclically stable materials, it can be extended to
cyclically hardening and softening materials.
The modeling schemes for the cyclic plasticity and ratcheting simulations can be
classified into two types: coupled and uncoupled. In the coupled modeling scheme, the
plastic modulus calculation is coupled with the kinematic hardening rule through the
consistency condition of the yield surface (Prager [1956]). Consequently, in uniaxial
loading condition, the kinematic hardening rule is effectively used to calculate the plastic
modulus, as the evolution direction is constant in this case. In multiaxial loading
condition, however, the hardening rule affects both the plastic modulus and the evolution
direction. If hardening rule parameters are calibrated from uniaxial loading responses
only, as has been done in many existing coupled models, the hardening rule is basically
designed for better representation of the plastic modulus, which undermines its role in
describing the direction of the yield surface evolution. Because of this approach in
modeling, many popular coupled models perform well in one type of loading condition
while fail in others.
In the uncoupled modeling scheme, on the other hand, the plastic modulus may be
indirectly influenced by the kinematic hardening rule but its calculation does not depend
2
on it. Therefore, the hardening rule parameters can be calibrated from multiaxial loading
conditions without having any effect on the simulation by the model in uniaxial loading
condition. However, the determination of the kinematic hardening rule parameter can be
somewhat subjective in uncoupled modeling scheme, unless simple hardening rules are
used.
Ratcheting experiments
The loading histories of the experiment set are shown in Fig. 1.1. The history in
Fig. 1.1a, which results in ratcheting of axial strains, involves axial stress controlled
cycles with a mean stress. The ratcheting response from this uniaxial loading history is
3
x
xc
xa
xm
t m
x
(a) (b)
xc xc
m m
x x
(c) (d)
Fig 1.1 Loading histories; (a) Uniaxial stress cycles, (b) Axial strain cycle with constant
pressure, (c) biaxial bowtie cycle, (d) biaxial reverse bowtie cycle.
shown in Fig. 1.2. Two sets of uniaxial ratcheting experiments are considered in this
xm) vary with constant amplitude stresses (
study. In the first set, the mean stresses ( xa).
Whereas in the second set, the amplitude stresses vary while mean stresses are kept
constant. The plots of peak axial strains (xp) at each cycle against the number of cycles,
N are presented in the paper I (Fig. 8a and b).
The loading history in Fig. 1.1b involves axial strain cycles in the presence of a
constant internal pressure. This history results in ratcheting of the circumferential strain
as shown in Fig. 1.3a. Two sets of data from this biaxial loading history are also
considered in this study. The peak circumferential strains (p) at each cycle are plotted
against the number of cycles, N in paper I (Figs. 8c and d). For the set shown in Fig. 8c,
the axial strain amplitudes (xc) are varied with same internal pressure (
m), whereas for
4
the set shown in Fig. 8d, the internal pressure varies while the axial strain amplitudes are
unchanged.
x
xa
x
t
Fig. 1.2 Axial strain ratcheting response for uniaxial stress controlled cycles
with a mean stress (Hassan and Kyriakides [1992])
The socalled bowtie and reverse bowtie loading histories shown in Fig. 1.1c
and d also result in circumferential strain ratcheting. The ratcheting response from the
bowtie loading history is shown in Fig 1.3b. It should be noted that the bowtie and
reverse bowtie loading histories are typically observed in piping components with
internal pressure. The peak circumferential strains (p) at each cycle for these histories
are plotted in the paper I (Fig. 8e and f) against the number of cycles, N.
Coupled models
5
(a) (b)
Fig. 1.3 Circumferential strain ratcheting response for (a) constant pressure and (b)
bowtie loading histories of Fig 1.1 (from Hassan et al. [1992 and Corona et al. [1996])
Wang [1993] are discussed anatomically in this chapter. For some of the models,
elaborate parameter determination schemes have been proposed. A journal paper has
been written from this part of the study, which is enclosed in Chapter 2 as paper I.
Uncoupled models
6
3 against the same set ratcheting responses. The ratcheting simulations from the
kinematic hardening rules by Armstrong and Frederick [1966], Phillips [1972,1979],
Tseng and Lee [1983], Voyiadjis and Sivakumar [1991,1994], Kaneko [1981,1984], Xia
and Ellyin [1994,1997], Chaboche [1991] and OhnoWang [1993] are evaluated in this
chapter. The journal paper written from this part of the study is enclosed Chapter 3 as
paper II.
Chapter 4 evaluates several coupled models that try to improve their multiaxial
ratcheting simulations by introducing additional maultiaxial parameter(s) or term(s) in the
kinematic hardening rule. The models studied are by McDowell [1995], Jiang and
Sehitoglu [1996], Voyiadjis and Basuroychowdhury [1998], and AbdelKarim and Ohno
[2000]. A modified kinematic hardening rule using the idea of Delobelle and his
coworkers [1995] in the framework of the Chaboche [1991] model is proposed in this
chapter. A journal paper has been written from this part of the study and is enclosed in
Chaper 4 as paper III.
In all the models considered in the Chapters 2, 3 and 4, the vonMises yield
surface is considered as the yielding boundary and also, the plastic potential surface. But
in reality, the yield surface translates as well as deforms during plastic loading (Phillips
and Tang [1972], Phillips and Lee [1979]). The incompleteness, thus, introduced in the
modeling bars the models from being successful in the general multiaxial loading
conditions. Therefore, investigations have been conducted in Chapter 5 to incorporate the
anisotropically deformed yield surface in the cyclic plasticity modeling (formative
hardening rule).
7
CHAPTER 2
The cyclic plasticity modeling schemes can be classified into two types: coupled
and uncoupled. There are four basic aspects that are used to define the constitutive
relations in cyclic plasticity models: (i) the yield surface, (ii) the flow rule derived from
the normality condition of plastic strain increments, (iii) the hardening rule and (iv) the
consistency condition. In coupled models, following the classical model by Parger
[1956], the plastic modulus is calculated from the kinematic hardening rule and the
consistency condition. This study critically evaluates the performances of five
constitutive models in predicting ratcheting responses of carbon steels for a broad set of
uniaxial and biaxial loading histories. A journal paper "Anatomy of Coupled Constitutive
Models for Ratcheting Responses" is published in the International Journal of Plasticity
from this part of the study. This paper is enclosed next as paper I. Readers are referred to
the enclosed paper I for details of the study conducted in this chapter.
8
Paper IInternational Journal of Plasticity 16 (2000) 381409
ABSTRACT
This paper critically evaluates the performance of five constitutive models in predicting ratcheting
responses of carbon steel for a broad set of uniaxial and biaxial loading histories. The models proposed
by Prager, ArmstrongFrederick, Chaboche, OhnoWang, and Guionnet are examined. Reasons for
success and failure in simulating ratcheting by these models are elaborated. The bilinear Prager and the
The Chaboche and OhnoWang models perform quite well in predicting uniaxial ratcheting responses;
however, they consistently overpredict the biaxial ratcheting responses. The Guionnet model simulates
one set of biaxial ratcheting responses very well, but fails to simulate uniaxial and other biaxial ratcheting
responses. Similar to many earlier studies, this study also indicates a strong influence of the kinematic
dependent on multiaxial ratcheting responses, while dormant for uniaxial responses, into Chabochetype
kinematic hardening rules may be conducive to improve their multiaxial ratcheting simulations. The
uncoupling of the kinematic hardening rule from the plastic modulus calculation is another potentially
viable alternative. The best option to achieve a robust model for ratcheting simulations seems to be the
incorporation of yield surface shape change (formative hardening) in the cyclic plasticity model.
9
Paper IInternational Journal of Plasticity 16 (2000) 381409
I. INTRODUCTION
As the data base and understanding of ratcheting response (accumulation of strains with cycles) are
growing, the number of efforts in developing constitutive models for ratcheting is also increasing
(Chaboche and his coworkers [1979,1986,1991,1994], Voyiadjis and his coworkers [1991,1998],
Guionnet [1992], Ohno and Wang [1993], Hassan and Kyriakides [1994a,b], Delobelle et al. [1995],
McDowell [1995], Jiang and Sehitoglu [1996a], Ohno [1997], Xia and Ellyin [1997] and others). In many
of these models the plastic modulus calculation is coupled with its kinematic hardening rule through the
yield surface consistency condition as in the classical model proposed by Prager [1956]. These models
are referred to as coupled models in this paper. Models proposed by ArmstrongFrederick [1966],
Chaboche [1986,1991,1994], Guionnet [1992], and Ohno and Wang [1993] belong to this class and are
In another class of models, the plastic modulus might be indirectly influenced by the kinematic
hardening rule but its calculation is not coupled to the kinematic hardening rule through the consistency
condition. The models proposed by Mroz [1967], Dafalias and Popov [1976], Drucker and Palgen [1981],
Tseng and Lee [1983] and many others belong to this class. These models are referred to as uncoupled
models and will be discussed in another paper (Bari and Hassan, [1999a]).
Most of the models proposed so far are developed and verified using data from limited or simple
experiments. These models have not been tested against a wide variety of ratcheting responses to verify
the generality of these models. Consequently, most of these constitutive models might predict a special
class of ratcheting responses quite well, but fail to predict a broad class of ratcheting responses (Hassan
Most metals cyclically harden or soften up to a certain number of cycles and subsequently stabilize
or cease to change the size of the yield surface (Morrow [1965], Jhansale [1975], Tuegel [1987], Ishikawa
10
Paper IInternational Journal of Plasticity 16 (2000) 381409
and Sasaki [1988]). Ratcheting, though, keeps on occurring with cycles even after the material stabilizes
(Hassan and Kyrikides [1992], Hassan et al. [1992], Hassan and Kyriakides [1994a,b]). Hence, the
kinematic hardening (translation of the yield surface in stress space) is attributed to be the primary reason
for ratcheting. Thus, in order to develop and verify a model for ratcheting, it is essential to study the
ratcheting responses of stabilized materials. This, in effect, means that the parameters affecting the
isotropic hardening (i.e, yield surface size change) should not be included during the model development
for ratcheting. Also, all of the kinematic hardening (i.e, yield surface translation) parameters should be
determined using experiments performed on stabilized materials. After achieving a robust model for
ratcheting responses of cyclically stable materials, it can easily be extended to cyclically hardening and
softening materials following the approach demonstrated by Hassan and Kyriakides [1994 a,b].
A broad set of quasistatic ratcheting data which include uniaxial to complex biaxial ratcheting
responses of stabilized carbon steels have been developed by Hassan and Kyriakides [1992], Hassan et al.
[1992] and Corona et al. [1996]. These data are used in this study to evaluate the performance of the
models considered. The cyclic loading histories prescribed in the experiments are shown in Fig. 1. The
history in Fig. 1a, which results in uniaxial ratcheting, involves axial stress cycles with a mean stress. The
readers are referred to Hassan and Kyriakides [1992] for demonstration of a uniaxial ratcheting response
(Fig. 7 in the reference). Two sets of uniaxial experiments are considered in this study. In the first set, the
amplitude stresses (xa) in the experiments are the same while the mean stresses (xm) vary. Whereas in
the second set, the amplitude stresses (xa) vary while the mean stresses (xm) are kept constant. Data
from these experiments are presented in Figs. 8a and 8b, where the peak axial strains (xp) of each cycle
The loading history in Fig. 1b involves axial strain cycles in the presence of a constant internal
pressure. This history results in circumferential ratcheting as demonstrated in Fig.1 of Hassan et al.
11
Paper IInternational Journal of Plasticity 16 (2000) 381409
xc xc
xc
x a a
xa
xm m
m m
t
x x x
(a) (b) (c) (d)
Fig. 1. Loading histories; (a) Uniaxial stress cycle, (b) Axial strain cycle with constant internal
pressure, (c) biaxial bowtie cycle, (d) biaxial reverse bowtie cycle.
[1992]. Two sets of data from this biaxial loading history are also considered in this study as shown in
Figs. 8c and 8d. In these plots, the peak circumferential strains (p) of each cycle are plotted against the
number of cycles, N. For the set shown in Fig. 8c, the axial strain amplitude (xc) of the experiments vary
with same circumferential stress (m) due to internal pressure, whereas in the set of Fig. 8d, the internal
pressures of the experiments are different with same axial strain amplitude.
The bowtie and reverse bowtie loading histories in Figs. 1c and 1d also result in circumferential
ratcheting as demonstrated in Figs. 10 and 11 of Corona et al. [1996]. The data set from these two biaxial
loading histories considered in this study are shown in Figs. 8e and 8f.
The above mentioned ratcheting data are used to evaluate the performance of several coupled cyclic
plasticity models. Also, parameters for each model are determined using experiments on stabilized
carbon steels. The same set of parameters for each model are used to simulate uniaxial and multiaxial
ratcheting responses. The reasons for the success and failure in simulating these responses by the models
are presented through anatomical discussion of the influences of modeling schemes and parameters on
ratcheting simulation. Although the Prager [1956] and Armstrong and Frederick [1966] models are not
capable of simulating ratcheting responses satisfactorily, these models are discussed in this study in order
to demonstrate the gradual development of different features of plasticity models over time.
12
Paper IInternational Journal of Plasticity 16 (2000) 381409
The plasticity models with the assumption of rateindependent material behavior has the following
common features:
12
i. vonMises yield criterion: f ( ) = 3 ( s a ) ( s a ) = 0 , (1)
2
p 1 f f
ii. flow rule: d =   d  , (2)
H
p
Where, is the stress tensor, is the plastic strain tensor, s is the deviatoric stress tensor, is the
current center of the yield surface, a is the current center of the yield surface in the deviatoric space, 0 is
the size of the yield surface (constant for a cyclically stable material), and H is the plastic modulus. Also,
iii. The third and most important feature for ratcheting simulation, the kinematic hardening rule is
given by:
p p
da = g ( , , a, d, d , etc ) . (3)
The Kinematic hardening rule dictates the evolution of the yield surface during a plastic loading
increment by translation in the stress space only. Different models discussed in this study basically differ
In coupled models, the plastic modulus H is evaluated using the consistency condition, f = 0 ,
kinematic hardening rule (Eq. 3), flow rule (Eq. 2), and yield criterion (Eq. 1). The models considered in
this study are discussed below in terms of their features and their influence on ratcheting simulation.
13
Paper IInternational Journal of Plasticity 16 (2000) 381409
Prager [1956] proposed perhaps the most simple kinematic hardening rule,
da = C d p (4)
to simulate plastic response of materials. For uniaxial loading, as the yield surface in this model moves
linearly with plastic strain (see x trace in Fig. 2a), the ensuing hysteresis loop is bilinear as shown in Fig.
2a. It is clear from the figure that this model cannot represent the experimental hysteresis curve during the
initial nonlinear part. Also, for a prescribed uniaxial stress cycle with a mean stress, this model fails to
distinguish between shapes of the loading and reverse loading hysteresis curves and consequently
produces a closed loop with no ratcheting (see Fig. 2b and 2c). For all the biaxial histories of Fig. 1, the
ratcheting by Pragers model stabilizes (shakedown) after some initial overprediction of ratcheting as
Improvement to the linear kinematic hardening model was proposed by Mroz [1967] as a
multisurface model, where each surface represents a constant work hardening modulus in the stress
space. Besseling [1958] introduced a multilayer model without any notion of surfaces. Also, Ohno and
Wang [1993] introduced a piecewise linear kinematic hardening rule. In uniaxial loading, all these
models essentially divide the stressstrain curve into many linear segments. When a sufficient number of
segments are chosen, the hysteresis loop simulations by these models are very good as shown in Fig. 3a.
Unfortunately, like the linear kinematic hardening model, multilinear models also predict a closed loop
when subjected to a uniaxial stress cycle with a mean stress and hence produce no uniaxial ratcheting (see
Fig. 3b).
14
Paper IInternational Journal of Plasticity 16 (2000) 381409
x Experiment x CS 1026
(ksi) 40 (ksi) 40 xa = 38.57 ksi
xm = 2.57 ksi
20 20
x
0 0
20 20
40 40
(a) x (%)
p
(b) x (%)
p
xp p
4 4
CS 1026 CS 1026
(%) xa = 32.0 ksi (%) xc m = 9.65 ksi
xm = 6.52 ksi
x m (1) xc = 0.40 %
3 3
xa x (2) xc = 0.50 %
xm (3) xc = 0.65 %
t
2 2 (3)
(3)
(2)
(1)
(2)
Experiment (1)
1 Prager model 1
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(c) N (d)
(c) N
p p
4 4
xc CS 1026 xc CS 1018
(%) a = 2.36 ksi (%) a = 2.36 ksi
a xc = 0.50 % a xc = 0.50 %
3 m 3 m
(2)
x (1) m = 9.65 ksi x (2) (1) m = 9.65 ksi
(2) m = 14.54 ksi (2) m = 14.54 ksi
2 (2) 2 (2)
(1)
(1) (1)
1 1 (1)
Experiment
Prager model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(e) N (f) N
Fig 2. Predictions from Pragers model for (a) straincontrolled stable hysteresis loop, (b)
stresscontrolled hysteresis loop, (c) axial strain at positive stress peaks of uniaxial cycles,
(d) Circumferential strain peaks of axial strain cycles with constant pressure, (e,f)
Circumferential strain peaks under biaxial bowtie and reverse bowtie cycles.
15
Paper IInternational Journal of Plasticity 16 (2000) 381409
x Experiment x CS 1026
(ksi) 40 (ksi) 40 xa = 32.0 ksi
xm = 9.14 ksi
x
20 20
0 0
20 20
40 40
x (%)
p
x (%)
p
(a) (b)
Fig 3. Predictions from Multilinear model for (a) straincontrolled stable hysteresis loop, (b)
stresscontrolled hysteresis loop.
For multiaxial responses, the original Mroz [1967] model or a modified version of the model can
not properly describe the ratcheting responses from nonproportional loadings (Mroz and Rodzik [1996],
Jiang and Sehitoglu [1996b]). The Besseling model [1958] as used in ANSYS [1998] overpredicts the
ratcheting rates for most of the biaxial loading histories considered (Yuan. et al. [1999]). Since the
multilinear models also fail to simulate any uniaxial ratcheting, they are not considered further in this
study. When a slight nonlinearity is introduced into the multilinear Ohno and Wang [1993] model, it
shows promise in simulating both uniaxial and biaxial ratcheting responses. This model is presented later
in this paper.
The most wellknown nonlinear kinematic hardening model has been proposed by Armstrong and
Frederick [1966]. They introduce a kinematic hardening rule containing a recall term which
incorporates the fading memory effect of the strain path and essentially makes the rule nonlinear in
nature. The kinematic hardening rule in this model is given in the form:
16
Paper IInternational Journal of Plasticity 16 (2000) 381409
2 p
da =  Cd a dp , (5)
3
12
p 2 p p
where dp = d =  d d .
3
For uniaxial loading, this rule basically provides an exponential x trace (see Fig. 4a), which always
H = C
+ x (6)
and stabilizes to a value of C/ after traversing some amount of plastic strains. In Eq. (6), the negative
sign is used for forward loading curve and the positive sign for reverse loading curve. Figure 4a shows
this models simulation of the stable hyteresis loop. It is apparent from the figure that the experimental
stressstrain curve is not necessarily exponential in nature and the attempt to simulate it by a single
exponential equation does not yield a good fit. Increasing the value of C would improve the simulation
during the initial nonlinear part, but the simulation for the rest of the curve would suffer. Another
limitation of this model is its inability to produce constant plastic modulus exhibited by experiments for a
high strain range, for which this model always stabilizes to zero plastic modulus.
For a uniaxial stress cycle with mean stress, the recall term in the ArmstrongFrederick kinematic
hardening rule produces change in shapes between forward and reverse loading paths. Therefore, the loop
does not close and results in ratcheting (see Fig. 4b). But the stressstrain loop produced by this model
deviates significantly from the experiment and the ratcheting strain is also overpredicted, as demonstrated
in Fig. 4b. For continued cycles between two fixed stress levels, this model simulates the same ratcheting
loops for all cycles and thus, produces a constant ratcheting rate (strain accumulation per cycle) which is
If the experimental hysteresis loops in uniaxial ratcheting tests from Hassan and Kyriakides [1992]
are examined, it is observed that the loading curves gradually stiffen while the unloading curves soften
with cycles. This results in the gradual decrease in the rate of ratcheting up to a strain level and
17
Paper IInternational Journal of Plasticity 16 (2000) 381409
x Experiment x CS 1026
(ksi) 40 (ksi) 40 xa = 32.0 ksi
xm = 6.52 ksi
x
20 20
0 0
20 20
AF model
40 40 Experiment
(a) x (%)
p
(b) x (%)
p
xp p
4 4
CS 1026 CS 1026
(%) xa = 32.0 ksi (%) m = 9.65 ksi
xm = 6.52 ksi xc = 0.50 %
3 3
xc
m
2 x 2 x
xa
xm
t
1 1
Experiment
AF model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(c) N (d)
(c) N
p p
4 4
CS 1026 CS 1018
(%) a = 2.36 ksi (%) a = 2.36 ksi
xc = 0.50 % xc = 0.50 %
m = 9.65 ksi m = 9.65 ksi
3 3
xc
a xc
2 2
a
m
m
x
1 1 x
Experiment
AF model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(e) N (f) N
Fig 4. Predictions from Armstrong and Frederick model (AF) for (a) straincontrolled stable
hysteresis loop, (b) stresscontrolled hysteresis loop, (c) axial strain at positive stress peaks of
uniaxial cycles, (d) Circumferential strain peaks of axial strain cycles with constant pressure,
(e,f) Circumferential strain peaks under biaxial bowtie and reverse bowtie cycles.
18
Paper IInternational Journal of Plasticity 16 (2000) 381409
subsequent stabilization to a constant rate of ratcheting. The plastic modulus H expressed in Eq. (6), on
the other hand, is a function of x only. Hence, for stress cycles between two fixed limits, the shapes of
forward and reverse loading curves are repeatedly reproduced by this model. This, in effect, results in a
constant rate of ratcheting simulation for all cycles. It should be pointed out that by increasing the value
of the parameter C and selecting a corresponding , the rate of uniaxial ratcheting can be reduced, but this
impairs the simulation of the straincontrolled stable hysteresis loop. Moreover, this model always
grossly overpredicts the circumferential ratcheting strains for all the biaxial loading histories in Fig. 1
Conceptually, the Armstrong and Frederick model has been a leap in representing cyclic plasticity
responses of materials, but is not robust enough to predict the ratcheting responses of materials. Several
improved models which are based on the ArmstrongFrederick kinematic hardening rule have
II.4.1. Model Formulation and Ratcheting Responses: Chaboche and his coworkers [1979, 1986]
M
2
d a i ,
p p
da = da i =  C d i a i dp where dp = d (7)
3 i
i=1
As can be observed in Eq. (7), the Chaboche kinematic hardening rule is a superposition of several
ArmstrongFrederick hardening rules. Each of these decomposed rules has its specific purpose.
A stable hysteresis curve can be divided into three critical segments where the ArmstrongFrederick
model fails: the initial high modulus at the onset of yielding, the constant modulus segment at a higher
strain range and the transient nonlinear segment (knee of the hysteresis curve, see Fig. 4a). Chaboche
19
Paper IInternational Journal of Plasticity 16 (2000) 381409
[1986] initially proposed to use three decomposed hardening rules (M = 3 in Eq. 7) to improve the
simulation of the hysteresis loops in these three segments. They suggested that the first rule (1) should
start hardening with a very large modulus and stabilizes very quickly. The second rule (2) should
simulate the transient nonlinear portion of the stable hysteresis curve. Finally, the third rule (3) should
be a linear hardening rule (3 = 0) to represent the subsequent linear part of the hysteresis curve at a high
strain range. If this scheme is followed (see section II.4.1.1 for detailed parameter determination method),
the simulation for a stable hysteresis loop improves as shown in Fig. 5a (compare to Fig. 4a). Also note in
Fig. 5a, the traces of the three decomposed rules 1, 2 and 3 and the resulting yield surface center x
x Experiment x CS 1026
(ksi) 40 (ksi) 40 xa = 32.0 ksi
xm = 6.52 ksi
x
20 20
2
1
0 3 0
20 20
CH3 model
40 40 Experiment
1 Lp
p
0.5 0 0.5 1 0.5 0 0.5 1
L
x (%)
p
x (%)
p
(a) (b)
xp
4
CS 1026 x
xa = 32.0 ksi xa
(%) xm
xm = 6.52 ksi
t
3
1 Experiment
3 = 0
3 = 9
0
0 10 20 30 40 50
N
(c)
Fig 5. Predictions from Chaboche model with 3 decomposed rules (CH3) for (a) strain
controlled stable hysteresis loop, (b) stresscontrolled hysteresis loop, (c) axial strain at positive
stress peaks of uniaxial cycles.
20
Paper IInternational Journal of Plasticity 16 (2000) 381409
(=1+2+3). The hysteresis curve simulation still deviates slightly from the experimental curve. This
Figure 5b shows the simulation for a stresscontrolled hysteresis loop by the Chaboche model using
three decomposed rules. The simulated loop is a significant improvement from that of the original
ArmstrongFrederick model and traces the experimental loop very closely. But this model still
overpredicts the ratcheting strain, though, by a relatively smaller amount at the end of the first cycle. The
ratcheting strain simulations (by three decomposed rules with 3 = 0) are plotted against the number of
cycles in Fig. 5c. It is observed in Fig. 5c that this model overpredicts the ratcheting strain for some initial
cycles, but gradually approaches to complete shakedown of ratcheting, which is in contrast to the
experimental trend. This shakedown is mainly caused by the incorporation of the linear kinematic
To understand this shakedown phenomenon, consider two points a and b at the same stress level
as shown in Fig. 6, where the total backstress x and the decomposed backstresses 1, 2 and 3 are
plotted against the plastic strain (note that a becomes x in uniaxial case). For 3 =0 (or linear third
rule), the backstress increments dx, for the same plastic strain increment dxp at these two points are:
dxa = d1a+d2a+d3a = (2/3 C1 dxp  11a dp) + (2/3 C2 dxp  2 2a dp) + (2/3 C3 dxp ) (8)
dxb = d1b+d2b+d3b = (2/3 C1 dxp  11b dp) + (2/3 C2 dxp  2 2b dp) + (2/3 C3 dxp ) (9)
Since a and b represent the same stress level, xa= xb for a cyclically stable material, thus
Since 1a = 1b (as evident from Fig. 6b), by subtracting Eq. (9) from Eq. (8) one obtains,
21
Paper IInternational Journal of Plasticity 16 (2000) 381409
30 30
x x
(ksi) (ksi)
a b
20 20
10 10
0 0
10 10
0 0.5 1 1.5 2 0 0.5 1 1.5 2 2.5 3
30 30
1
(ksi)
x 1
(ksi) x
CS 1026 CS 1026
xa xa = 32.0 ksi xa xa = 32.0 ksi
20 xm xm = 6.52 ksi 20 xm xm = 6.52 ksi
t t
10 10
a b
0 0
10 10
0 0.5 1 1.5 2 0 0.5 1 1.5 2 2.5 3
(b) xp
(%) (b) xp (%)
30 30
2 2
(ksi) (ksi)
20 20
a b
10 10
0 0
10 10
0 0.5 1 1.5 2 0 0.5 1 1.5 2 2.5 3
30 30
3 3
(ksi) 3 = 0 (ksi) 3 = 9
20 20
10 10
a b
0 0
10 10
0 0.5 1 1.5 2 0 0.5 1 1.5 2 2.5 3
(d) xp
(%) (d) xp (%)
Fig 6. Uniaxial backstresses by Chaboche Fig 7. Uniaxial backstresses by Chaboche model with
model with three decomposed hardening three decomposed hardening and 3=9; (a) total
and 3=0; (a) total backstress, x, (b) 1, backstress, x, (b) 1, (c) 2, (d) 3 against axial
(c) 2, (d) 3 against axial plastic strain. plastic strain.
It is clear from Eq. (11) that the difference in incremental hardening between points a and b
comes from the second rule. As 3 increases linearly from point a to b making 3b > 3a (Fig. 6d) and
22
Paper IInternational Journal of Plasticity 16 (2000) 381409
as 1a = 1b, so 2 has to decrease linearly at a rate defined by C3 between points a and b in order to
hold the equality of Eq. (10), which is also validated in Fig. 6c. Since 2a > 2b, Eq. (11) dictates that
dxb> dxa. This essentially results in the gradual stiffening of the loading curves with cycles. Similarly,
the unloading curves relax with increasing cycles, resulting in progressive decrease of ratcheting strain
with cycles. The point at which both the loading and unloading curves assume the same shape, the
ratcheting ceases completely (shakedown) as observed in Fig. 5c (for 3 = 0) and Fig. 6a.
If a slight nonlinearity is introduced in the third rule by assigning a relatively small value to 3 (= 9),
keeping other parameters the same, the ratcheting simulation improves as shown in Figs. 5c and 7a. This
small value does not introduce any noticeable change in the straincontrolled stable hysteresis loop
simulation. Note also that a nonzero 3 does not have any effect on 1 (Fig. 7b), but it changes the course
of 3 (Fig. 7d) and thereby of 2 (Fig. 7c), which improve the uniaxial ratcheting simulation and prevent
shakedown (Fig. 7a). When the third backstress (3) reaches its limiting value (near xp = 1.5% in Fig. 7d),
this model starts predicting a constant ratcheting rate which is also the trend in experimental responses.
The higher the value of 3, the quicker the third rule would reach its limiting state and, consequently, the
earlier the steady rate of ratcheting would start. The inception point of constant ratcheting rate also
depends on the stress limits of a stress cycle. One may consider 3 as a ratcheting parameter whose value
II.4.1.1 Parameter Determination (Three Decomposed Rules): In this study, all parameters, except
3, are determined from a uniaxial straincontrolled stable hysteresis curve, not a monotonic curve as sug
p
gested in the original model. This requires a hysteresis loop of reasonable strain limit ( L = 0.85% in
Fig. 5a) which ensures that all, except the third slightly nonlinear kinematic hardening variable get stabi
lized within the strain limit. Also, a stabilized decomposed backstress (1 or 2) should have the same
23
Paper IInternational Journal of Plasticity 16 (2000) 381409
tensile and compressive levels within the strain range of the stable loop. In other words, for the loading
portion of the hardening curves, they should start from Ci /i at the starting plastic strain Lp and reach
the value Ci/i at or prior to the final plastic strain, Lp as shown in Fig. 5a. In addition, the third linear
The equations used for the loading part of the hardening curves are:
i + 0 = x (12)
i=1
Ci p p
i =  1 2 exp i ( x ( L ) ) , for i = 1 and 2 (13)
i
p
where L is the strain limit of the stable hysteresis loop (Fig. 5a).
C1 should be a very large value to match the plastic modulus at the yielding and corresponding 1
also should be large enough to stabilize the hardening of 1 immediately. C3 is determined from the slope
of the linear segment of hysteresis curve at a high strain range. C2 and 2 are evaluated by trials to pro
duce a good representation of the experimental stable hysteresis curve which also satisfy the relationship
C1 C2 C3 p p
 +  + 0 = x  { x ( L ) } (14)
1 2 2
Finally, 3 is determined from a uniaxial ratcheting experiment (xp vs. N plot) to produce the best
possible fit. It should be noted here that choosing a value of 3 relatively smaller in magnitude than 1 or
2 would produce no tangible effect on the stable hysteresis loop simulation, but would produce a signifi
cant effect on the uniaxial ratcheting rate. The parameters determined following this method and used for
24
Paper IInternational Journal of Plasticity 16 (2000) 381409
II.4.1.2. Ratcheting Simulations (Three Decomposed Rule): The above set of parameters is used
to simulate all uniaxial and biaxial ratcheting responses of stabilized carbon steels considered in this
study. Figure 8 shows the predictions by the Chaboche model (three decomposed rules) for all the
experiments. In these figures, the maximum peak strain in each cycle is plotted as a function of the
number of cycles, N. As observed in Figs. 8a and 8b, the Chaboche model has a tendency to overpredict
uniaxial ratcheting rates during initial cycles. As a result, the overall simulation for a low rate ratcheting
experiment (e.g. xa= 28.29 ksi in Fig. 8b) deviates from the experiment. Otherwise, the ratcheting rate
predictions for other uniaxial cases are quite reasonable (Figs. 8a,b). However, this model overpredicts
the ratcheting for all the biaxial loading cases (Figs. 8c,d,e,f).
One might argue here that the incorporation of more than three decomposed rules might improve
the ratcheting simulations. In order to verify this argument, simulations are obtained for all the
experiments by adding a fourth ArmstrongFrederick type decomposed hardening rule. It is observed that
the incorporation of a fourth rule improves the stable hysteresis loop simulation, especially at the knee
(not shown). But the ratcheting simulations for both uniaxial and biaxial experiments do not improve.
II.4.2. Chaboche Model  Fourth Rule with a Threshold: Realizing the above deficiencies of the
model, Chaboche [1991] added a fourth hardening rule with a concept of threshold in his model. This
kinematic hardening rule grows linearly to a certain threshold stress level and subsequently hardens
ai
da i =  C i d i a i 1  dp ,
2 p
for i = 4 (15)
3 f ( i )
25
Paper IInternational Journal of Plasticity 16 (2000) 381409
xp xp
4 4
CS 1026 CS 1026
xa = 32.0 ksi xm = 9.14 ksi xm = 6.5 ksi
(%) (%) a = 33.28 ksi
x
xa
3 3 xm xa= 32.12 ksi
t
xm = 4.18 ksi
1 1 xa = 28.29 ksi
Experiment
CH3 model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(a) (b)
p p
4 4
CS 1026 CS 1026
(2) (3)
(%) m = 9.65 ksi (%) xc = 0.50 %
(3) (4)
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(c) (d)
p p
4 4
(2) CS 1026 (2) CS 1018
(%) a = 2.36 ksi (%) a = 2.36 ksi
(1)
xc = 0.50 % xc = 0.50 %
3 3
(1)
(1) m = 9.65 ksi (1) m = 9.65 ksi
(2) m = 14.54 ksi (2) m = 14.54 ksi
xc
2 (2) 2 (2)
a xc
(1)
m a
(1) m
1 x 1
Experiment
x
CH3 model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(e) (f)
Fig. 8. Ratcheting experiments (From Hassan and Kyriakides [1992], Hassan et al. [1992] and
Corona et al. [1996]) and predictions by Chaboche model with three decomposed rule (CH3).
(a,b) axial strain at positive stress peaks of uniaxial cycles, (c,d) Circumferential strain peaks of
axial strain cycles with constant pressure, (e,f) Circumferential strain peaks under biaxial bowtie
and reverse bowtie cycles.
26
Paper IInternational Journal of Plasticity 16 (2000) 381409
This fourth rule simulates a constant linear hardening with a slope of C4 within the threshold a 4
and becomes nonlinear outside the threshold (see trace of 4 in Fig. 9a). When this fourth rule is added to
the previous three decomposed rules and the parameters are determined using the methodology described
in section II.4.2.1., the model predicts a reduced rate of ratcheting for initial cycles without jeopardizing
the simulation of the stable hysteresis loop. The improvement becomes evident if simulations in Figs. 5a
and 5b are compared to Figs. 9a and 9b, respectively. The simulation in Fig. 9b conforms better with the
experimental uniaxial ratcheting hysteresis loop compared to that in Fig. 5b. The reason for improvement
in uniaxial ratcheting simulation by the fourth rule with a threshold is the fact that within the threshold
level, this rule does not use its recall term and assumes linear hardening in a similar fashion as shown in
x Experiment x CS 1026
(ksi) 40 (ksi) 40 xa = 32.0 ksi
xm = 6.52 ksi
x
20 20
4 1 3
2
0 0
20 20
CH4T model
40 40 Experiment
1 Lp
p
0.5 0 0.5 1 0.5 0 0.5 1
L
(a) x (%)
p
(b) x (%)
p
xp
4
CS 1026 x
(%) xa = 32.0 ksi xa
xm = 6.52 ksi xm
3 t
1 Experiment
CH4T model
0
0 10 20 30 40 50
(c) N
Fig 9. Predictions from Chaboche model with threshold (CH4T) for (a) straincontrolled
stable hysteresis loop, (b) stresscontrolled hysteresis loop, (c) axial strain at positive stress
peaks of uniaxial cycles.
27
Paper IInternational Journal of Plasticity 16 (2000) 381409
Fig. 6d. As stated before, a linear hardening is instrumental in stiffening the loading curves and relaxing
the unloading curves to reduce overall ratcheting with cycles. When the threshold level is reached, the
hardening becomes nonlinear again and the reduction of ratcheting is attenuated to avoid potential
shakedown. The threshold level, a 4 can also be considered a ratcheting parameter which is to be
II.4.2.1. Parameter Determination (Chaboche Model with Threshold): The parameter C1,1 and
C3 are determined using the same method as discussed in section II.4.1.1 using Eqs. (12) and (13). C2,
C4, 2 and 4 are evaluated by trials to produce a good representation of the experimental stable hysteresis
C1 C2 C4 C p p
 +  +  + a 4 + 0 = x 3 { x ( L ) } (16)
1 2 4 2
at or close to the plastic strain Lp (Fig. 9a). The first trial value of a 4 can be taken close to the value of the
meanstress in the uniaxial ratcheting experiment used for the parameter determination (Chaboche
[1991]). The 3 value is determined using the ratcheting rate from a uniaxial ratcheting experiment (xp
vs. N) as discussed in section II.4.1.1. Finally, the value of a 4 is varied a little to improve the simulation
for the uniaxial ratcheting experiment (xp vs. N) further. Care should be taken so that the value of a 4
selected does not deteriorate the simulation for the straincontrolled stable hysteresis loop. All parameters
determined accordingly and used in this study for simulation by the Chaboche model with threshold
are:
C14 = 60000, 3228, 455, 15000 (ksi); 14 = 20000, 400, 11, 5000; a4 = 5 ksi
28
Paper IInternational Journal of Plasticity 16 (2000) 381409
xp xp
4 4
CS 1026 CS 1026 xa = 33.28 ksi
xa = 32.0 ksi xm = 9.14 ksi xm = 6.5 ksi
(%) (%)
x
3 3 xa xa= 32.12 ksi
xm
t
xm = 6.52 ksi
2 2
xa = 30.35 ksi
xm = 4.18 ksi
1 1 xa = 28.29 ksi
Experiment
CH4T model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(a) (b)
4 4
CS 1026 CS 1026
p (2)
m = 9.65 ksi p (3)
xc = 0.50 %
(%) (%)
(3) (4)
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(c) (d)
p p
4 4
(2) CS 1026 (2) CS 1018
(%) a = 2.36 ksi (%) a = 2.36 ksi
(1)
xc = 0.50 % xc = 0.50 %
3 3
(1)
(1) m = 9.65 ksi (1) m = 9.65 ksi
(2) m = 14.54 ksi (2) m = 14.54 ksi
2 xc 2 (2)
(2) xc
a
(1)
a
m
(1) m
1 x 1
Experiment x
CH4T model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(e) (f)
Fig. 10. Ratcheting experiments (From Hassan and Kyriakides [1992], Hassan et al. [1992] and
Corona et al. [1996]) and predictions by Chaboche model with threshold (CH4T). (a,b) axial
strain at positive stress peaks of uniaxial cycles, (c,d) Circumferential strain peaks of axial
strain cycles with constant pressure, (e,f) Circumferential strain peaks under biaxial bowtie
and reverse bowtie cycles.
29
Paper IInternational Journal of Plasticity 16 (2000) 381409
II.4.2.2. Ratcheting Simulations (Chaboche model with Threshold): Figure 10 shows the
ratcheting simulations by the Chaboche model with threshold along with the experimental responses.
The simulations by this model for most of the uniaxial loading cases are similar to those by the Chaboche
model with three decomposed rules (compare Figs. 10a,b and 8a,b). The effect of the forth rule with a
threshold becomes pronounced for the experiments with lower rates of ratcheting (see Figs. 8b and 10b
for xa= 28.29 ksi). But overpredictions for all biaxial loading cases persist (shown in Figs. 10b,c,d,e)
II.5.1. Model Formulation and Ratcheting Responses: The OhnoWang model [1993] is also a
superposition of several kinematic hardening rules. They initially proposed the form:
M ai
da i =  C i d i ai d 
H { a2 ( Ci i )2 }
2
d a i ,
p p
da = (17)
3 f ( i ) i
i=1
Here, each decomposed hardening rule simulates a linear hardening with a slope Ci until it reaches
the critical value Ci/i. After that it does not evolve at all, as shown qualitatively by solid lines in Fig. 11.
Consequently, the model becomes a multilinear model in a uniaxial case. It is demonstrated in section II.2
that the multilinear models produce no uniaxial ratcheting due to the prediction of closed hysteresis
loops. To avoid this limitation, Ohno and Wang [1993] introduces a slight nonlinearity for each rule at the
mi
M a i f ( i )
d a i , da i = 3 C i d i a i d 

2 p p

da = (18)
f ( i ) C i i
i=1
The slight nonlinearity is introduced by replacing the Heaviside step function with the multiplier
with power of mi in Eq. (18). The slight nonlinearities produced are shown qualitatively by dashed lines
30
Paper IInternational Journal of Plasticity 16 (2000) 381409
x,x
(4p,x4)
(3p,x3) x
(2p,x2)
2
(1p,x1)
3
L p 1
4
Lp Sp
(0p,x0) xp
in Fig. 11. These nonlinearities prevent the stresscontrolled hysteresis loops from closing and allow
uniaxial ratcheting to occur (see Fig. 5 in Ohno and Wang [1993] for details).
As several essentially linear hardening rules are employed to simulate a nonlinear hysteresis curve,
this model should use a large number of decomposed rules in order to produce a good representation of
the stable uniaxial hysteresis curve. In this study, ten hardening rules are found to be sufficient to obtain a
good stable loop simulation as demonstrated in Fig. 12a. In addition, this model uses the term
a
d p 
i in place of dp in the ArmstrongFrederick rule (Eq. 5). Although both these terms in a given
f ( )
i
model yield the same result in uniaxial cases, they produce different directions of kinematic hardening in
multiaxial loading cases. Inclusion of the former term improves the ratcheting simulation in multiaxial
cases as demonstrated by Ohno and Wang [1993] (Figs. 6 and 8 in the reference).
The OhnoWang model simulates the stresscontrolled ratcheting hysteresis loop more closely than
the Chaboche model (compare Fig. 9b and 12b). It is evident from Fig. 12c that the overall ratcheting
simulation by the OhnoWang model is very good. The effect of nonlinearity induced by the parameter mi
31
Paper IInternational Journal of Plasticity 16 (2000) 381409
x Experiment x CS 1026
(ksi) 40 (ksi) 40 xa = 32.0 ksi
xm = 6.52 ksi
x
20 20
0 0
20 20
OW model
40 40 Experiment
1 Lp
p
0.5 0 0.5 1 0.5 0 0.5 1
L
x (%)
p
x (%)
p
(a) (b)
xp
4
CS 1026 x
(%) xa = 32.0 ksi xa
xm = 6.52 ksi xm
t
3
1 Experiment
m i = 0.20
m i = 0.45
m i = 0.70
0
0 10 20 30 40 50
(c) N
Fig 12. Predictions from Ohno and Wang (OW) model (a) straincontrolled stable hysteresis
loop, (b) stresscontrolled hysteresis loop, (c) axial strain at positive stress peaks of uniaxial
cycles.
on the ratcheting simulations is also shown in Fig. 12c. Note that the smaller the value of mi the more
nonlinear the decomposed rules are and, the higher the rate of ratcheting.
II.5.2. Parameter Determination: Parameters for this model are also determined from a uniaxial
stable hysteresis curve and a uniaxial ratcheting experiment. The stable hysteresis loading curve are
divided into several segments as shown in Fig. 11 and the corresponding parameters Ci and i for each
32
Paper IInternational Journal of Plasticity 16 (2000) 381409
i i1 i+1 i
x x x x 2
C i =   for i 1 ,
p p p p
i = 
p
;
p
(19a)
i i 1 i + 1 i i + 0
C
 ii + 0
0
and finally C1 is determined using = x (19b)
where ix and pis are as indicated in Fig. 11. The power mi is assumed to be same for all segments and
In the Chaboche model, the hardening rule (3) which is responsible for simulating the linear
segment of the hysteresis curve at higher strain ranges has a significant effect on the simulation of the rate
of uniaxial ratcheting. Figure 5c shows that an increase in the value of 3 increases the uniaxial ratcheting
rate and hastens the inception of the steady rate of ratcheting. In the Ohno model, the last hardening rule
If the straincontrolled stable hysteresis loop in Fig. 12a is represented by ten segments, the value of
10 is determined by taking 10p = Lp in Eq. (19). This essentially means that the backstress 10 reaches
its plateau at Lp and, 10 calculated from Eq. (19) becomes relatively large. Consequently, the Ohno
Wang model overpredicts ratcheting in some cases as shown in Fig. 13a. This overprediction is more
To overcome this problem, it is stipulated that the last linear segment (10th in this study) of the
stable hysteresis curve should be extended up to a high strain limit sp as shown by the longdashed lines
in Fig. 11. This increase in the strain limit from Lp to sp reduces the value of 10 as determined by Eq.
(19). Figures 13b and 13c show the simulations for the same set of experiments when values of sp used
are 2.5% and 5%, respectively. Uniaxial ratcheting simulations are reasonable with sp= 2.5% (Fig. 13b),
33
Paper IInternational Journal of Plasticity 16 (2000) 381409
xp xp
4 4
CS 1026 ps = 0.85 % CS 1026 ps = 2.5 %
(%) xa = 32.0 ksi 10 segments (%) xa = 32.0 ksi 10 segments
3 3
xm = 6.52 ksi xm = 6.52 ksi
xm = 9.14 ksi
xm = 9.14 ksi
2 2
xm = 4.18 ksi
1 1
xm = 4.18 ksi
Experiment Experiment
OW model OW model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(a) N (b) N
xp xp
4 4
CS 1026 ps = 5.0 % CS 1026 ps = 5.0 %
(%) xa = 32.0 ksi 10 segments (%) xa = 32.0 ksi 12 segments
3 3
xm = 9.14 ksi xm = 6.52 ksi
xm = 9.14 ksi
xm = 6.52 ksi
2 2
Experiment Experiment
OW model OW model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(c) N (d) N
Fig 13. Axial (peak) strain ratcheting simulations by Ohno and Wang (OW) model for
uniaxial loading with (a) sp=0.85%, 10 segments; (b) sp=2.5%, 10 segments; (c) sp=5%, 10
segments and (d) sp=5%, 12 segments.
but overprediction still persists for the higher stress level (xm= 9.14 ksi) as the simulated rate reaches a
steady state quickly. If sp is increased to 5% for the tenth segment, 10 decreases further. Consequently,
ratcheting simulation show prolonged nonlinearity and reduced strain (Fig. 13c).
In reality, the straincontrolled stable curve is not expected to maintain the same slope all the way
from Lp (0.85%) to sp = 5%. It is expected to show slight nonlinearity or a decrease in modulus with
increasing strain. This behavior can be modeled by adding a few more segments beyond Lp (0.85%),
34
Paper IInternational Journal of Plasticity 16 (2000) 381409
with gradually smaller slopes than previous segments. When two more segments are added to the
hysteresis curve beyond Lp (0.85%) and extended up to sp=5%, the simulations improve greatly as
shown in Fig. 13d. Simulations are now tracing the nonlinear ratcheting rate trend of the experiments
more closely.
Hence, the best and recommended way to determine the parameters (Ci and is) is from a hysteresis
curve with reasonably larger strain range. If such a hysteresis curve is not available, one should follow the
method discussed above and use more than one uniaxial ratcheting experiment to validate the parameters.
The parameters used in this study for simulations by the OhnoWang model are (with sp=5% and 12
segments):
C112 = 31940, 36214, 2520, 376, 11021, 4551, 3475, 2196, 857, 247, 98, 200 (ksi);
112 = 45203, 13944, 7728, 4955, 3692, 2135, 1230, 585, 295, 119, 50, 20;
mi = 0.45
II.5.3. Ratcheting Simulations: Figure 14 shows the simulations of the OhnoWang model using
the above set of parameters for all the experiments. As the parameters in this model are determined using
uniaxial experiments, the uniaxial ratcheting predictions by this model are quite good and comparable to
those obtained by the Chaboche model (compare Figs. 10a,b and 14a,b). Although this model performs
better in all biaxial cases compared to the Chaboche model, the trend of overprediction persists (compare
a
Figs. 10c,d,e,f and 14c,d,e,f). Note, however, that the incorporation of d p 
i in Eq. (18) yields
f ( i )
nonlinear biaxial ratcheting rates and the trend is similar to the experimental results.
35
Paper IInternational Journal of Plasticity 16 (2000) 381409
xp xp
4 4
CS 1026 CS 1026
(%) xa = 32.0 ksi (%) xm = 6.5 ksi
x xa = 33.28 ksi
3 3 xa
xm
xm = 9.14 ksi xm = 6.52 ksi xa= 32.12 ksi
t
2 2
xa = 30.35 ksi
xm = 4.18 ksi
1 1
xa = 28.29 ksi
Experiment
OW model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(a) N (b) N
p p
4 4
CS 1026 CS 1026
(%) m = 9.65 ksi (%) xc = 0.50 %
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(c) N (d) N
p p
4 4
CS 1026 CS 1018
(%) a = 2.36 ksi (%) a = 2.36 ksi
(2) xc = 0.50 % xc = 0.50 %
3 3
(2)
(1) (1) m = 9.65 ksi (1) m = 9.65 ksi
(2) m = 14.54 ksi (2) m = 14.54 ksi
2 xc 2 (1)
(2) (2)
a xc
(1)
a
m
1 1 (1) m
x
Experiment
x
OW model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(e) N (f) N
Fig. 14. Ratcheting experiments (From Hassan and Kyriakides [1992], Hassan et al. [1992]
and Corona et al. [1996]) and predictions by Ohno and Wang (OW) model. (a,b) axial strain at
positive stress peaks of uniaxial cycles, (c,d) Circumferential strain peaks of axial strain cycles
with constant pressure, (e,f) Circumferential strain peaks under biaxial bowtie and reverse
bowtie cycles.
36
Paper IInternational Journal of Plasticity 16 (2000) 381409
The Chaboche and OhnoWang models, discussed above, include parameters which are determined
from uniaxial experiments only. These models have been shown to predict the uniaxial ratcheting
responses well, but the predictions are not as good for biaxial ratcheting responses. Guionnet [1992]
proposes a model which uses some parameters that are determined from biaxial ratcheting experiments.
This model is expected to simulate the biaxial ratcheting responses better and therefore, is studied in this
paper.
The Guionnet model basically modifies the original ArmstrongFrederick hardening rule by
incorporating the effect of accumulated plastic strain in it. For cyclically stabilized material, the
p
, n = 2 

2 p d
m1 (19)
da = m p1 3 C 1 ( a n ) d ( 2 )a dp 3 dp
n n
where = p 1 = p 1M , for p 1 = p 1M ,
n p 1M
= p 1M  , for p 1 p 1M ,
p 1M + p 1
Q IK
p1 = dp , and p 1M = dp
IK IK 1
Here, p1 is the accumulated plastic strain between the last reversal (Ik) and the current loading point
(Q), and p1M is the accumulated plastic strain between the last two reversals (Ik1 and Ik). The parameters
C and 1 are similar to those in ArmstrongFrederick model and are determined from a uniaxial stable
hysteresis curve. Two ratcheting parameters, 2 and are determined using a biaxial ratcheting response
(experiment (2) in Fig. 16c). No clear guidelines are provided by Guionnet [1992] to determine m and n.
These parameters do not affect the ratcheting simulations greatly. Therefore, the values of m and n as
37
Paper IInternational Journal of Plasticity 16 (2000) 381409
x
Experiment x
CS 1026
(ksi) 40 (ksi) 40 xa = 32.0 ksi
xm = 6.52 ksi
x
20 20
0 0
20 20
Guionnet model
40 40 Experiment
x (%)
p
x (%)
p
(a) (b)
Fig 15. Predictions from Guionnet model (a) straincontrolled stable hysteresis loop, (b) stress
controlled hysteresis loop
The stable, uniaxial hysteresis loop simulation by this model is similar to that from Armstrong
Frederick model (compare Fig. 15a to Fig. 4a). This model predicts the uniaxial stresscontrolled
hysteresis loop poorly and also, overpredicts the amount of ratcheting strain in the first cycle as shown in
Fig. 15b. Figure 16 shows the simulations by the Guionnet model for all the experiments considered in
this study using the following parameters:
The ratcheting strains for the uniaxial experiments are overpredicted by this model (Fig. 16a,b).
This model also has numerical divergence problems for higher uniaxial ratcheting rate responses.
However, its biaxial ratcheting predictions for the loading history shown in Fig. 1b are very good (Figs.
16c,d) despite the constant rate of ratcheting prediction. For the bowtie loading case (Fig. 16e), this
model overpredicts with constant rate of ratcheting, whereas for the reverse bowtie (Fig. 16f), it
underpredicts the ratcheting. Similar erratic behavior is also observed in the simulations for bowtie and
reverse bowtie loading cycles obtained by the DafaliasPopov uncoupled model with Armstrong
38
Paper IInternational Journal of Plasticity 16 (2000) 381409
xp xp
4 4
CS 1026 CS 1026
(%) xa = 32.0 ksi (%) xm = 6.5 ksi
x
3 xa 3
xm xm = 6.52 ksi xa= 32.12 ksi
t
2 2
xa = 30.35 ksi
xm = 4.18 ksi
1 1
Experiment xa = 28.29 ksi
Guionnet model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(a) (b)
p p
4 4
CS 1026 CS 1026
m = 9.65 ksi xc xc = 0.50 %
(%) (%)
m
3 (1) xc = 0.40 % 3 (1) m = 4.88 ksi
(2) xc = 0.50 % x (2) m = 7.32 ksi
(3) xc = 0.65 % (3) m = 9.65 ksi
(4) m =14.54 ksi
2 2
(3)
(2) (4) (3)
(2)
1 (1) 1
Experiment (1)
Guionnet model
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(c) (d)
p p
4 4
CS 1026 CS 1018
(%) a = 2.36 ksi (%) a = 2.36 ksi
xc = 0.50 % xc = 0.50 %
3 3
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(e) (f)
Fig. 16. Ratcheting experiments (From Hassan and Kyriakides [1992], Hassan et al. [1992]
and Corona et al. [1996]) and predictions by Guionnet model. (a,b) axial strain at positive
stress peaks of uniaxial cycles, (c,d) Circumferential strain peaks of axial strain cycles with
constant pressure, (e,f) Circumferential strain peaks under biaxial bowtie and reverse bowtie
cycles.
39
Paper IInternational Journal of Plasticity 16 (2000) 381409
Five wellknown constitutive models are evaluated in this paper in terms of simulations for
ratcheting responses from a series of uniaxial and biaxial experiments. Six uniaxial and ten biaxial
ratcheting responses on stabilized carbon steels are collected from literature (Hassan and Kyriakides
[1992], Hassan et al. [1992] and Corona et al. 1996]) for the study. The rateindependent and cyclically
stable ratcheting response simulations by Prager [1956]), Mroz [1967], Armstrong and Frederick [1966],
Chaboche [1986, 1991], Guionnet [1992], Ohno and Wang [1993] are evaluated.
All these models are categorized as coupled models due to their plastic modulus calculation being
coupled with the kinematic hardening rule of the models. Sequential development of different modeling
features, such as, linear kinematic hardening (Prager [1956]), multilinear model (Mroz [1967]), nonlinear
kinematic hardening (Armstrong and Frederick [1966], Guionnet [1992]), decomposed nonlinear
kinematic hardening (Chaboche and his coworkers [1979,1986]), and finally decomposed nonlinear
kinematic hardening with threshold (Chaboche [1991], Ohno and Wang [1993]) are presented. The
bearings of each of these features on the simulation of ratcheting responses are elaborated.
For uniaxial stresscontrolled history (Fig. 1a), the linear kinematic hardening and multilinear
models produce closed hysteresis loops and hence, cannot simulate a ratcheting response. On the other
hand, the nonlinear kinematic hardening models (Armstrong and Frederick [1966], Guionnet [1992])
grossly overpredict the uniaxial ratcheting strains. This overprediction problem is remedied by Chaboche
[1986] through decomposing the back stress into a number Armstrong and Frederick type nonlinear
kinematic hardening rules. The original Chaboche [1986] model with three or four decomposed rules,
with at least one rule as linear hardening, improves the ratcheting simulation for the initial cycles, but
always stabilizes to shakedown with persistent cycling. Incorporation of a slight nonlinearity into the
linear rule, through a small 3 value instead of 3 = 0, improves the models capability of simulating
40
Paper IInternational Journal of Plasticity 16 (2000) 381409
steady rate of ratcheting and prevents shakedown. This model simulates all the uniaxial ratcheting
experiments, except the one with small rate of ratcheting, reasonably well (Figs. 8a,b). In this case, the
ratcheting strains are overpredicted due to overpredictions during the initial cycles. Subsequently,
Chaboche [1991] introduces the concept of threshold on backstress into the fourth nonlinear kinematic
hardening rule. This model incorporates a linear segment in the uniaxial hysteresis curve within the
threshold level (see Figs. 9a and 9b), thus makes the hysteresis curve stiffer and reduces the rate of
ratcheting compared to the models without any threshold. This visible linear segment in the hysteresis
loop seems unnatural, but works well in improving the uniaxial ratcheting simulation.
The OhnoWang model [Ohno and Wang, 1993] is also composed of a number of the Armstrong
and Frederick type kinematic hardening rules. This model reproduces uniaxial hysteresis loops by using a
number of almost linear segments and thus needs a large number of kinematic hardening rules to
simulate the hysteresis loop. The hysteresis loop simulations however are very good as shown in Figs.
12a and 12b. The simulations of all the uniaxial ratcheting experiments by the OhnoWang model are also
very good.
None of the coupled models studied performed satisfactorily in simulating ratcheting responses of
biaxial loading histories in Figs. 1b, 1c and 1d. The linear kinematic hardening model [Prager, 1956]
overpredicts ratcheting strains during the initial cycles which are followed by shakedown after a few
more cycles for all biaxial cases. The nonlinear kinematic hardening rule [Armstrong and Frederick,
1966] grossly overpredicts the strains in all biaxial ratcheting experiments. Multisurface Mroz model (not
studied here) also overpredicts the ratcheting responses under multiaxial stresses (Mroz and Rodzik
[1996], Jiang and Sehitoglu [1996b]). The multilinear Besseling [1958] model is found to overpredict the
ratcheting strains in most biaxial cases (Yuan, et al., [1999]). The Chaboche [1991] and the OhnoWang
[1993] models, which use decomposed nonlinear kinematic hardening rules with threshold, also
41
Paper IInternational Journal of Plasticity 16 (2000) 381409
The models proposed by Prager, Mroz, Besseling, Chaboche and OhnoWang determine model
parameters using uniaxial experiments only. Therefore, the kinematic hardening rule in each of these
models with the assumption of stable yield surface fails to simulate the biaxial ratcheting strains
satisfactorily. Hassan et al. [1992] and Corona et al. [1996] explained graphically the reason behind this
weakness of models. Each of these kinematic hardening rules stabilizes the backstress loading paths
differently after a few initial cycles and thereby, each rule dictates different normal directions to the yield
surface. These normal directions determine the amount of ratcheting strain during an increment. In
reality, the yield surface changes shape during plastic deformations [Phillips and Lee, 1979]. Hence, the
normals to the actual yield surface are different from those predicted by models [Corona et al., 1996].
This difference can be compensated, if some model parameters are calibrated using biaxial ratcheting
experiments. This methodology is demonstrated by the Guionnet [1992] model which simulates the
circumferential ratcheting strain in the constant pressure biaxial experiment impressively. Unfortunately,
when the loading changes to the biaxial bowtie histories, which are different from the biaxial experiment
used for parameter calibration, this model fails to simulate the ratcheting responses.
The above observations indicate that the incorporation of parameters, which improve multiaxial
ratcheting simulations while being dormant for uniaxial ratcheting responses, in a models kinematic
hardening rule can lead to a better constitutive modeling scheme (Bari and Hassan [1999b]). Uncoupling
of the plastic modulus calculation from its kinematic hardening rule (uncoupled model) and determining
the parameters in the kinematic rule using multiaxial ratcheting responses can be another viable
alternative in achieving a constitutive model for ratcheting simulation (Bari and Hassan [1999a]). Finally,
towards achieving a robust model for ratcheting simulation, the incorporation of yield surface shape
change (formative hardening rule) into the modeling scheme seems to be a more rewarding approach.
42
Paper IInternational Journal of Plasticity 16 (2000) 381409
Acknowledgment  The financial supports from the Center for Nuclear Power Plant Structures, Equipment
and Piping and the Department of Civil Engineering at North Carolina State University are gratefully
acknowledged.
NOMENCLATURE
43
Paper IInternational Journal of Plasticity 16 (2000) 381409
REFERENCES
ANSYS (1998) ANSYS users Manual, Revision 5.4. ANSYS Inc. Providence, Houston, PA 15342
0065, U.S.A.
Armstrong, P.J. and Frederick, C.O. (1966) A Mathematical Representation of the Multiaxial Bauscinger
Bari, S. and Hassan. T. (1999a) Uncoupled Constitutive Models for Simulating Ratcheting Responses. To
Bari, S. and Hassan. T. (1999b) An Improved Constitutive Model for Multiaxial Ratcheting. To be sub
Besseling, J.F. (1958) A Theory of Elastic, Plastic and Creep Deformations of an Initially Isotropic Mate
Chaboche, J.L., DangVan, K. and Cordier, G. (1979) Modelization of the Strain Memory Effect on the
Cyclic Hardening of 316 Stainless Steel. Proceedings of the 5th International Conference on SMiRT,
Chaboche, J.L. (1986) TimeIndependent Constitutive Theories For Cyclic Plasticity. International Jour
Chaboche, J.L. (1991) On Some Modifications of Kinematic Hardening to Improve the Description of
Chaboche, J.L. (1994) Modeling of ratchetting: evaluation of various approaches. European Journal of
Corona, E., Hassan, T. and Kyriakides, S. (1996) On the Performance of Kinematic Hardening Rules in
Predicting a Class of Biaxial Ratcheting Histories. International Journal of Plasticity, Vol 12, pp. 117
145.
Dafalias, Y.F. and Popov, E.P. (1976) Plastic Internal Variables Formalism of Cyclic Plasticity. Journal of
44
Paper IInternational Journal of Plasticity 16 (2000) 381409
Delobelle, P., Robinet, P. and Bocher, L. (1995) Experimental Study and Phenomenological Modelization
of Ratchet Under Uniaxial and Biaxial Loading on an Austenitic Stainless Steel. International Journal
Drucker, D.C. and Palgen, L. (1981) On stressStrain Relations Suitable for Cyclic and Other Loadings.
Hassan, T. and Kyriakides, S. (1992) Ratcheting in Cyclic Plasticity, Part I: Uniaxial Behavior. Interna
Hassan, T., Corona, E. and Kyriakides, S. (1992) Ratcheting in Cyclic Plasticity, Part II: Multiaxial
Hassan, T. and Kyriakides, S. (1994a) Ratcheting of Cyclically Hardening and Softening Materials, Part
Hassan, T. and Kyriakides, S. (1994b) Ratcheting of Cyclically Hardening and Softening Materials, Part
II: Multiaxial Behavior. International Journal of Plasticity, Vol 10, pp. 185212.
Ishikawa, H. and Sasaki, K. (1988) Yield Surfaces of SUS 304 Under Cyclic Loading. Journal of Engi
Jhansale, H.R. (1975) A New Parameter for the Hysteretic StressStrain Behaviour of Metals. Journal of
Jiang, Y. and Sehitoglu, H. (1996a) Modeling of Cyclic Ratchetting Plasticity, Part I: Development of
Jiang, Y. and Sehitoglu, H. (1996b) Comments on the Mroz Multiple Surface Type Plasticity Models.
45
Paper IInternational Journal of Plasticity 16 (2000) 381409
McDowell, D.L. (1995) Stress State Dependence of Cyclic Ratcheting Behavior of Two Rail Steels.
Morrow, J. (1965) Cyclic Plastic Strain Energy in Fatigue of Metals. Proceedings of Symp. on Internal
Mroz, Z. (1967) On the Description of Anisotropic Work Hardening. Journal of the Mechanics and Phys
Mroz, Z. and Rodzik, P. (1996) On Multisurface and Integral Description of Anisotropic Hardening Evo
lution of Metals. European Journal of Mechanics, A/Solids, Vol 15, pp. 128.
Ohno, N. and Wang, J.D. (1993) Kinematic Hardening Rules with Critical State of Dynamic Recovery,
Part I: Formulations and Basic Features for Ratcheting Behavior. International Journal of Plasticity,
Ohno, N. (1997) Current State of the Art in Constitutive Modeling for Ratcheting. Proceedings of the
Phillips, A. and Lee, C.W. (1979) Yield Surfaces and Loading Surfaces. Experiments and Recommenda
tions. International Journal of Solids and Structures, Vol. 15, pp. 715729.
Prager, W. (1956) A New Method of Analyzing Stresses and Strains in Work Hardening Plastic Solids.
Tseng, N.T. and Lee, G.C. (1983) Simple Plasticity Model of the TwoSurface Type. ASCE Journal of
Tuegel, E.J. (1987) Measurements of Kinematic and Isotropic Hardening in 304 Stainless Steel During
Voyiadjis, G.Z. and Sivakumar, S.M. (1991) A Robust Kinematic, A Robust Kinematic Hardening Rule
for Cyclic Plasticity with Ratcheting Effects, Part I: Theoretical Formulation. Acta Mechanica, Vol 90,
pp. 105123.
46
Paper IInternational Journal of Plasticity 16 (2000) 381409
Voyiadjis, G.Z. and Basuroychowdhury, I.N. (1998) A Plasticity Model for Multiaxial Cyclic Loading
Xia, Z. and Ellyin, F. (1997) A Constitutive Model with Capability to Simulate Complex Multiaxial
Ratcheting Behaviour of Materials. International Journal of Plasticity, Vol 13, pp. 127142.
Yuan, X., Hassan, T. and Bari, S. (1999) Improved Ratcheting Simulations by ANSYS and ABAQUS.
(Under preparation).
47
CHAPTER 3
It has been realized from the study in chapter 2 that the distinction of the roles
played by the kinematic hardening rule and the plastic modulus is central to the
understanding of the performance of a cyclic plasticity model. In coupled models, when
kinematic hardening rule parameters are calibrated from uniaxial loading condition, the
ratcheting simulations under multiaxial loading condition suffer and vice versa. In an
effort to avoid this problem in cyclic plasticity modeling, the uncoupled models are
introduced where the plastic modulus may be indirectly influenced by the kinematic
hardening rule but its calculation does not depend on it. Thus, the simulation of uniaxial
ratcheting is dictated only by the plastic modulus calculation scheme, whereas the
simulation of multiaxial ratcheting is mainly influenced by the kinematic hardening rule
adopted in the model. In this study, the modified DafaliasPopov [1976] model is used as
the basic plastic modulus calculation scheme and eight different kinematic rules are
evaluated in terms of their ratcheting simulations for the same set of uniaxial and
multiaxial loading histories considered in chapter 2. A journal paper "Kinematic
Hardening Rules in Uncoupled Modeling for Multiaxial Ratcheting Simulation" written
from this part of the study is accepted for publication in the International Journal of
Plasticity. This paper is enclosed next as paper II. Readers are referred to this paper for
details of the study conducted in this chapter.
48
Paper IITo be published in the International Journal of Plasticity (in press)
ABSTRACT
An earlier paper by the authors evaluated the performance of several coupled models in simulating a
series of uniaxial and biaxial ratcheting responses. This paper evaluates the performance of various
kinematic hardening rules in an uncoupled model for the same set of ratcheting responses. A modified
version of the DafaliasPopov uncoupled model has been demonstrated to perform well for uniaxial
influenced by the kinematic hardening rules employed in the model. Performances of eight different
kinematic hardening rules, when engaged with the modified DafaliasPopov model, are evaluated against
a series of rateindependent multiaxial ratcheting responses of cyclically stabilized carbon steels. The
Kaneko, XiaEllyin, Chaboche and OhnoWang are examined. The ArmstrongFrederick rule performs
reasonably for one type of the biaxial ratcheting response, but fails in others. The VoyiadjisSivakumar
rule and its constituents, the Phillips and the TsengLee rules, can not simulate the biaxial ratcheting
responses. The Kaneko rule, composed of the Ziegler and the prestress directions, and the XiaEllyin
rule, composed of the Ziegler and Mroz directions, also fail to simulate the biaxial ratcheting responses.
The Chaboche rule, with three decomposed ArmstrongFrederick rules, performs the best for the whole
set of ratcheting responses. The OhnoWang rule performs well for the data set, except for one biaxial
49
Paper IITo be published in the International Journal of Plasticity (in press)
I. INTRODUCTION
Pioneering works of Dafalias and Popov [1976], Yoshida et al. [1978], and Chaboche et al. [1979]
developed new concepts in the modeling of ratcheting responses. Motivated by these works, many
researchers have proposed improved models for simulation of ratcheting (Voyiadjis and Sivakumar,
[1991], Voyiadjis and Basuroychowdhury [1998], Hassan and Kyriakides [1992], Hassan et al. [1992],
Ohno and Wang [1993], Chaboche [1994], Delobelle et al. [1995], McDowell [1995], Jiang and
Shehitoglu [1996], Xia and Ellyin [1997] and others). Despite the considerable efforts made so far,
existing constitutive models fail to simulate ratcheting from a broad class of loading histories (Corona et
Performance studies of many existing models demonstrate that the simulation for uniaxial
ratcheting primarily depends on the plastic modulus calculation scheme, whereas the simulation for
ratcheting in multiaxial loading depends essentially on the kinematic hardening rule employed in the
model (Hassan and Kyriakides [1992], Hassan et al. [1992], and Corona et al. [1996]). In this paper,
ratcheting from uniaxial loading is referred to as uniaxial ratcheting and that from multiaxial loading is
referred to as multiaxial ratcheting. Note that in case of multiaxial loading, ratcheting may occur along
one or more directions depending on the control mode of the prescribed loading.
An anatomical study of several coupled constitutive models for simulating ratcheting responses in
both uniaxial and biaxial loading has recently been performed by Bari and Hassan [2000a]. In coupled
models, the plastic modulus calculations are directly tied to the kinematic hardening rule of the model
through the consistency condition. The models proposed by Chaboche [1991,1994] and Ohno and Wang
[1993] perform well to simulate uniaxial ratcheting responses, but consistently overpredict the biaxial
ratcheting responses. This limitation of the models can be attributed to the determination of the model
50
Paper IITo be published in the International Journal of Plasticity (in press)
parameters from uniaxial responses and the assumption of the vonMises yield surface. The
Guionnet [1992] model, on the other hand, determines two of its parameters using biaxial
responses and thus, improves the simulation for a set of biaxial ratcheting responses. This model,
however, fails to simulate the uniaxial ratcheting and other biaxial ratcheting responses. The
authors have improved the biaxial ratcheting simulations of the coupled models by introducing a
parameter in the hardening rule. This new parameter is determined from a biaxial ratcheting
experiment and thus produces representative plastic strain rate directional properties for
multiaxial ratcheting, but does not influence the uniaxial ratcheting simulation. This improved
coupled model and its simulations will be presented in a forthcoming paper by the authors (Bari
simulation is uncoupling of the plastic modulus calculation from the kinematic hardening rule
(Hassan et al. [1992]). In the uncoupled models, the plastic modulus may be indirectly influenced
by the kinematic hardening rule but its calculation does not depend on it. Thus, the simulation of
uniaxial ratcheting is dictated only by the plastic modulus calculation scheme, whereas the
simulation of multiaxial ratcheting is mainly influenced by the kinematic hardening rule adopted
in the model. Following the pioneering work of Mroz [1967], new concepts of uncoupled models
have been introduced by Dafalias and Popov [1976], Drucker and Palgen [1981], Tseng and Lee
[1983] and others. Hassan and Kyriakides [1992, 1994a, 1994b], Hassan et al. [1992] and Corona
et al. [1996] evaluated several of these uncoupled models and various kinematic hardening rules
for ratcheting simulation. They have developed an improved version of the DafaliasPopov model
that can simulate uniaxial ratcheting responses quite well. This model, however, fails to simulate a
broad class of biaxial ratcheting responses with kinematic hardening rules of Ziegler [1959],
51
Paper IITo be published in the International Journal of Plasticity (in press)
ArmstrongFrederick [1966], Mroz [1967], Phillips and Lee [1979] and Tseng and Lee [1983].
In order to identify and develop a robust constitutive model for ratcheting simulation, this paper
evaluates the performances of several kinematic hardening rules with the improved Dafalias and Popov
model. The set of uniaxial and biaxial ratcheting responses that is adopted by the authors to evaluate the
coupled models (Bari and Hassan [2000a]) is also used in this study. The evaluation in this paper starts
with a wellknown kinematic hardening rulethe Armstrong and Frederick [1966] rule. Following this,
the kinematic hardening rules by Voyiadjis and Sivakumar [1991], Phillips and his coworkers
[1972,1979,1984], Tseng and Lee [1983], Kaneko [1981,1984], and Xia and Ellyin [1994,1997] are
evaluated. Finally, an improved model, composed of the modified DafaliasPopov model (Hassan and
Kyriakides [1992, 1994a]) and the kinematic hardening rule by Chaboche [1991, 1994], is presented
along with its ratcheting simulations. The simulations from the OhnoWang [1993] kinematic rule, when
Model performances are evaluated with respect to a series of uniaxial and biaxial ratcheting
responses of cyclically stable carbon steels acquired from Hassan and Kyriakides [1992], Hassan et al.
[1992] and Corona et al. [1996]. The philosophy behind studying the ratcheting responses of cyclically
stable materials is discussed in Bari and Hassan [2000a]. A brief review of the ratcheting responses
studied is given below (see Hassan and Kyriakides [1992], Hassan et al. [1992] and Corona et al. [1996]
The loading histories of the experiment set are shown in Fig. 1. The history in Fig. 1a, which results
in uniaxial ratcheting, involves axial stress cycles with a mean stress. The readers are referred to Hassan
and Kyriakides [1992] for demonstration of a typical uniaxial ratcheting response (Fig. 7 in the
reference). Data from these experiments are presented in Figs. 3a and 3b, where the peak axial strains
52
Paper IITo be published in the International Journal of Plasticity (in press)
xc xc
xc
x a a
xa
xm m
m m
t
x x x
(a) (b) (c) (d)
Fig. 1. Loading histories; (a) Uniaxial stress cycle, (b) Axial strain cycle with constant internal
pressure, (c) biaxial bowtie cycle, (d) biaxial reverse bowtie cycle.
(xp) of each cycle are plotted against the number of cycles, N. The loading history in Fig. 1b involves
axial strain cycles in the presence of a constant internal pressure. This history results in circumferential
ratcheting as demonstrated in Fig.1 of Hassan et al. [1992]. Two sets of data obtained from this biaxial
loading history are studied (see Figs. 3c and 3d). In Figs. 3c and 3d, the peak circumferential strains (p)
in each cycle are plotted against the number of cycles, N. The bowtie and reverse bowtie loading
histories, shown in Figs. 1c and 1d, also result in circumferential ratcheting as demonstrated in Figs. 10
and 11 of Corona et al. [1996]. The data from the bowtie histories are shown in Figs. 3e and 3f.
p 1 f f
i. the flow rule: d =   d  , (1)
H
3 12
ii. the vonMises yield surface: f ( ) =  ( s a ) ( s a ) = 0 , (2)
2
3 12
iii. the bounding surface: F ( ) =  ( s b ) ( s b ) = b , (3)
2
p p
iv. the kinematic hardening rule: da = g ( , , a, d, d , etc ) ; (4)
53
Paper IITo be published in the International Journal of Plasticity (in press)
p
where, is the stress tensor, is the plastic strain tensor, s is the deviatoric stress tensor, is the
current center of the yield surface, a is the current center of the yield surface in the deviatoric space, is
the current center of the bounding surface, b is the current center of the bounding surface in the
deviatoric space, H is the plastic modulus, 0 and b are the sizes of the yield and bounding surfaces,
respectively. indicates the MacCauley bracket and the inner product a b = a ij b ij . For a cyclically
stable material, the values of 0 and b remain unchanged during plastic deformation. Unlike a coupled
model, the Dafalias and Popov model has the flexibility to adopt any kinematic hardening rule.
12
H = E o + h  , , = 3 ( s s ) ( s s )
P a
in
h =  . (5)
in m 2
1 + b 
2 b
The parameters of the model are shown in Fig. 2, where and in Eq. (5), Eop is the plastic modulus at the
bound or bounding surface, ( s ) is the image of ( s ) on the bounding surface, is the distance in stress
space of the current stress state from the bounding surface image point, in is the distance in stress space
2 Yield
Bound P Surface
Eo Bounding
b Surface
o
o
p 1
in n
n
(a) (b)
54
Paper IITo be published in the International Journal of Plasticity (in press)
of the recent yielding state from the corresponding bounding surface image point, and a, b, and m are
model constants (see Dafalias and Popov [1976] for details of the model).
The original DafaliasPopov model [1976] simulates an eventual shakedown for the uniaxial
ratcheting response. Hassan and Kyriakides [1992,1994a] developed a modified version of this model to
improve its uniaxial ratcheting simulation. The improvement incorporates relaxation of the bounding
p
E ds n
db = da m dM and d M = 1 b 
 (6)
H m n
P P 12 3 (s a) 3 (s s)
where, Eb = Eo + Cr[(b b) b n] , n =  
 , m =  
 . (7)
2 2
0
Ebp defines the modulus for the translation of the bounding surface center, not the bound modulus, which
is always Eop; m is the unit vector along the direction ( s s ) , n is the unit vector along the normal of the
Note that in the Dafalias and Popov model [1976] the plastic modulus calculation (Eq. 5) is not
directly dependent on the yield surface kinematic hardening rule (Eq. 4). In fact, the uniaxial stable
hysteresis loop and ratcheting simulations by the DafaliasPopov model are entirely controlled by the
parameters involved in the plastic modulus calculations (Eqs. 5 and 6). In this model, the kinematic
hardening rule specifies the direction of movement of the yield surface center. During a uniaxial stress
cycle, the yield surface moves along the stress direction only. Hence, the DafaliasPopov model always
produces the same uniaxial stressstrain response irrespective of the kinematic hardening rule employed
with the model. It is the multiaxial ratcheting simulation that gets affected by different kinematic
hardening rules. For a given kinematic hardening rule, the amount of movement along the specified
f ( + d d ) = 0 . (8)
55
Paper IITo be published in the International Journal of Plasticity (in press)
The parameter determination scheme in the DafaliasPopov model for the cyclically stabilized
materials is described in Hassan and Kyriakides (1992,1994a). In this scheme, o, b, Eop, a, b, m, and E
are determined by using a stable hysteresis curve and Cr is determined from the ratcheting rate in an
uniaxial experiment. The parameters determined and used for simulations are:
E = 26300 ksi, = 0.302, 0 = 18.8 ksi, b = 38.66 ksi, E0p = 274 ksi,
a = 71100, b = 27, m = 2, Cr = 8.
The Armstrong and Frederick [1966] kinematic hardening rule is used with the modified Dafalias and
12
da = (  Cd a dp )
2 p p 2 p p
where dp = d =  d d . (9)
3 3
The parameter C (= 140 ksi) is determined from the ratcheting rate of a constant pressure biaxial
experiment. In this study, the experiment (2) in Fig. 3c is used for this purpose. The value of the scalar
is determined using the consistency condition, Eq. (8), for each loading increment. Figure 3 shows the
simulations of the ratcheting responses by the ArmstrongFrederick hardening rule. The good simulations
for the uniaxial ratcheting responses, shown in Figs. 3a and 3b, are not surprising, as these are the
outcome of the improved DafaliasPopov model (Hassan and Kyrikides [1992,1994a]). As stated earlier,
any other kinematic hardening rule would also produce the same uniaxial simulation with this uncoupled
model.
The ratcheting simulations for the biaxial loading history of Fig. 1b with the ArmstrongFrederick
kinematic hardening rule are shown in Figs. 3c and 3d. The simulations exhibit constant rate of ratcheting
and conform reasonably well with the experiments. It should be pointed out here that the Armstrong
56
Paper IITo be published in the International Journal of Plasticity (in press)
xp xp
4 4
CS 1026 CS 1026
xa = 32.0 ksi xm = 9.14 Ksi xm = 6.5 ksi
(%) (%) a = 33.28 Ksi
x
xa
3 3 xm
t
xm = 4.18 Ksi
1 1
Experiment
DP model xa = 28.29 Ksi
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(a) (b)
p p
4 4
CS 1026 CS 1026
(%) m = 9.65 Ksi (%) xc = 0.50 %
xc
2 2
(3) (2) (4) (3)
(2)
1 (1) 1 (1)
Experiment
ArmstrongFrederick
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(c) (d)
p p
4 4
CS 1026 CS 1018
(%) a = 2.36 Ksi (%) a = 2.36 Ksi
xc = 0.50 % xc = 0.50 %
3 3
N N
(e) (f)
Fig. 3. Ratcheting Predictions by the ArmstrongFrederick rule with the modified DafaliasPopov
model (DP). (a,b) Axial strain at positive stress peaks from uniaxial cycles, (c,d) circumferential
strain peaks from axial strain cycles with constant pressure, (e,f) circumferential strain peaks from
bowtie and reverse bowtie cycles. Experiments are from Hassan and Kyriakides [1992], Hassan et
al. [1992] and Corona et al. [1996].
57
Paper IITo be published in the International Journal of Plasticity (in press)
Frederick kinematic hardening rule in a coupled model overpredicts the ratcheting for this set of biaxial
experiments (Bari and Hassan, 2000a). In an uncoupled model, the parameter C for this rule is
determined from a biaxial ratcheting response and hence, the simulations for similar biaxial responses are
improved. However, with the same set of parameters, this rule overpredicts the ratcheting rate in the bow
tie experiments (see Fig. 3e) and underpredicts in the reverse bowtie experiments (see Fig. 3f).
The kinematic hardening rule proposed by Voyiadjis and Sivakumar [1991,1994] is a combination
of the Phillips (Phillips and Tang [1972], Phillips and Lee [1979]) and TsengLee [1983] rules. They
reason that the Phillips rule, which stipulates the evolution of the backstress along the stress rate
direction, follow the experimental trend better, but it does not ensure the tangential nesting of the yield
and other surfaces in a multisurface plasticity model. In contrast, the TsengLee rule invokes the desired
nesting feature for both the proportional and nonproportional loading. Hence, Voyiadjis and Sivakumar
[1991] propose a rule to appropriately blend the deviatoric stress rate and the TsengLee rules, to include
both the requirement of surface nesting and the experimental observation of the yield surface movement.
Note that, in the improved DafaliasPopov model, the surface nesting is inherently ensured with any
kinematic hardening rule. In order to provide a broader perspective of this kinematic hardening rule, the
Phillips [1972,1979] and TsengLee [1983] rules are also evaluated here along with the rule proposed by
Phillips Rule: Phillips and his coworkers [1972,1979,1984] recommend a kinematic hardening
rule which prescribes the yield surface translation in the stress rate direction as follows:
da = d ( v D ) (10)
12 12
 , x for any tensor x means  x x =  x ij x ij
ds 3 3
where, v D = l =  2
, and d is a scalar variable to
ds 2
58
Paper IITo be published in the International Journal of Plasticity (in press)
Tseng and Lee Rule: The TsengLee [1983] rule can be described with the help of Fig. 4 as
follows:
da = d ( v T ) , (11)
a' a ds
 , a' = ( s + 'l )  ( s + 'l b ) , l = 
 , ' is calculated using
0
where, v T = 
a' a b ds
2 2 2
' = ( s b ) l + { ( s b ) l } ( s b ) ( s b ) +  b , (12)
3
2 2
( s + 'l b ) ( s + 'l b ) =  b , (13)
3
When the above two rules are used with the modified DafaliasPopov model, the value of the scalar
d, in Eqs. (10) and (11), is determined using the consistency condition, Eq. (8) during each load
increment. Note that, unlike the ArmstrongFrederick kinematic hardening rule discussed in section
III.2.1, these hardening rules do not include any parameters which can be calculated using ratcheting
responses.
n
'
S2 ds
s a
a'
b
Bounding
Surface
S1
S3
Fig. 4. Yield surface movement in the TsengLee kinematic hardening rule
59
Paper IITo be published in the International Journal of Plasticity (in press)
p p
4 4
CS 1026 CS 1026
(%) m = 9.65 Ksi (%) xc = 0.50 %
(1) m = 4.88 Ksi
3 (1) xc = 0.40 % 3 (2) m = 9.65 Ksi
(2) xc = 0.50 % (3) m =14.54 Ksi
xc xc
2 2
(2) m (3) (2) m
x x
(1)
1 1 (1)
Experiment
Phillips
TsengLee
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(a) (b)
2 4
CS 1026 CS 1018
p (2)
a = 2.36 Ksi p a = 2.36 Ksi
(%) (%)
xc = 0.50 % xc = 0.50 %
1 (1) (1) m = 9.65 Ksi 3
(2) m = 14.54 Ksi
(1) m = 9.65 Ksi
(2) m = 14.54 Ksi
0 2 (2)
xc
xc
a
a
1 1 (1)
m
Experiment m
x Phillips x
TsengLee
2 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(c) (d)
Fig. 5. Ratcheting Predictions by the Phillips and TsengLee rules with the modified Dafalias
Popov model. (a,b) Circumferential strain peaks from axial strain cycles with constant pressure,
(c,d) circumferential strain peaks from bowtie and reverse bowtie cycles. Experiments are from
Hassan et al. [1992] and Corona et al. [1996].
Figure 5 shows the biaxial ratcheting simulations by the Phillips and TsengLee rules along with the
experimental data. Since the uniaxial ratcheting simulations are not influenced by the kinematic
hardening rule employed in an uncoupled model, they are not presented again in Fig 5. This figure
demonstrates that both the Phillips and TsengLee rules fail to simulate biaxial ratcheting for the
experiments studied. The Phillips rule can not even reproduce the effect of different levels of internal
60
Paper IITo be published in the International Journal of Plasticity (in press)
The VoyiadjisSivakumar rule blends the stress rate and TsengLee directions by introducing two
ba ba
= 
, = l 
 . (14)
2 ba
 ( b 0 )
3
Depending on the value of , the v D (Eq. 10) and v T (Eq. 11) directions are combined using
0 vD + 1 vT
= 
1
 , 0 + 1 = 1 (15)
0 vD + 1 vT
Where, 0() and 1() are blending functions satisfying the boundary conditions
0 = 1 , 1 = 0 for = 0; thus, = v D
1
0 = 0 , 1 = 1 , for = 1; thus, = v T
1
The direction is further combined with v D using
1
' 0 + ' 1 v D
1
2
 , ' 0 + ' 1 = 1
= 
1 . (16)
' 0 + ' 1 v D
Where, ' 0 ( ) and ' 1 ( ) are blending functions satisfying the boundary conditions
' 0 = 0 , ' 1 = 1 , for = 1; thus, = v D
2
Voyiadjis and Sivakumar [1991] propose the kinematic rule in the form
da = d ( ).
2
(17)
Where, the scalar d is determined using the consistency condition, Eq. (8) during each loading
increment. They propose to determine the blending functions 0(), 1(), ' 0 ( ) and ' 1 ( ) from
experimental responses, but have not provided any guideline to obtain these functions. To evaluate the
performance of their hardening rule the following functions are assumed in this study:
q
0 = 1 , thus, 1 = q (18)
1q
' 0 = 1 , thus, ' 1 = 1 q . (19)
61
Paper IITo be published in the International Journal of Plasticity (in press)
When, q = 1, both these functions vary linearly with and , respectively. If the value of q is increased
further, the dominance of the Phillips rule increases over the TsengLee rule.
Figure 6 shows the simulations for the biaxial ratcheting responses by the VoyiadjisSivakumar
kinematic hardening rule for q = 1 and q = 10. Since, both the Phillips and TsengLee rules fail to
simulate the biaxial ratcheting responses (Fig. 5), the VoyiadjisSivakumar rule that is a combination of
these two rules, also fails to simulate these responses as shown in Fig 6. Note that, for q = 1, the
simulations fall in between the simulations by the Phillips and TsengLee rules, as can be observed by
p p
4 4
CS 1026 CS 1026
(%) m = 9.65 Ksi (%) xc = 0.50 %
(1) m = 4.88 Ksi
3 (1) xc = 0.40 % 3 (2) m = 9.65 Ksi
(2) xc = 0.50 % (3) m =14.54 Ksi
xc
2 xc 2
(2) (3) (2) m
m
x
(1) x
1 1 (1)
Experiment
VoyiadjisSivakumar (q=1)
VoyiadjisSivakumar (q=10)
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(a) (b)
p p
2 4
CS 1026 CS 1018
(2)
(%) a = 2.36 Ksi (%) a = 2.36 Ksi
xc = 0.50 % xc = 0.50 %
1 (1) (1) m = 9.65 Ksi 3
(2) m = 14.54 Ksi
(1) m = 9.65 Ksi
(2) m = 14.54 Ksi
xc
0 a 2 (2)
xc
m
a
x (1)
1 1
m
Experiment
VoyiadjisSivakumar (q=1) x
VoyiadjisSivakumar (q=10)
2 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(c) (d)
Fig. 6. Ratcheting Predictions by the VoyiadjisSivakumar rule with the modified DafaliasPopov
model. (a,b) Circumferential strain peaks from axial strain cycles with constant pressure, (c,d)
circumferential strain peaks from bowtie and reverse bowtie cycles. Experiments are from
Hassan et al. [1992] and Corona et al. [1996].
62
Paper IITo be published in the International Journal of Plasticity (in press)
comparing Figs. 5 and Fig. 6. It is also observed from these figures that, for q = 10, the simulations reside
very close to those obtained from the Phillips rule. The authors would like to point out here that, in the
experiments by Phillips and his coworkers [1972,1979,1984]), the yield surfaces change shape when they
translate in the stressrate direction. This might be a reason why the vonMises yield surface with
kinematic hardening rules along the stressrate direction can not represent multiaxial ratcheting
responses.
Kaneko [1981,1984] proposes a kinematic hardening rule that combines the prestress path
(k)
da = d C ( s a ) + ( s s ) (20)
k
where the s ( k ) are the prestress points in stress space when the stress history changes directions (see
Kaneko [1981] for descriptions and examples). For the biaxial loading history of Fig. 1b, this rule with
DafaliasPopov model produces negative ratcheting as shown in Fig. 7 (for C= 50). If the value of C is
increased, the Ziegler direction becomes dominant and simulates no ratcheting at all. Thus, the inclusion
of the prestress path direction with the Ziegler rule produces adverse effect on the ratcheting simulation
Xia and Ellyin [1994,1997] propose a kinematic hardening rule in the following form:
When the yield and bounding surfaces are in contact at the current stress point, the subsequent
plastic loading is defined as monotonic loading. In this case, both the surfaces translate together along
63
Paper IITo be published in the International Journal of Plasticity (in press)
p
2
CS 1026
(%) (2) m = 9.65 Ksi
xc
1
m Experiment
Kaneko
x XiaEllyin
2
0 10 20 30 40 50
Fig. 7. Ratcheting Predictions by the Kaneko and XiaEllyin rules with the modified
DafaliasPopov model for circumferential strain peaks from axial strain cycles with
constant pressure. Experiments are from Hassan et al. [1992].
the Ziegler [1959] direction. On the other hand, when the yield surface is not in contact with the bounding
surface, the subsequent plastic loading increment is defined as reloading and in such a case, the yield
surface moves along the Mroz [1967] direction and the bounding surface moves according to Eq. (6).
This kinematic hardening rule with the modified DafaliasPopov model overpredicts the ratcheting in
constant pressure biaxial experiments as shown in Fig. 7. This rule also overpredicts the ratcheting rate
for all other biaxial cases considered and hence, are not presented here. The overprediction by this rule is
mainly caused by the Mroz rule (Hassan et al. [1992], Corona et. al [1996]).
The above results clearly indicate that many existing kinematic hardening rules when implemented
with an uncoupled model cannot simulate a broad set of ratcheting responses. Since the Armstrong
Frederick rule performs reasonably well with the uncoupled models in simulating many ratcheting
responses (discussed in section III.2.1.), the hardening rule proposed by Chaboche and his coworkers
[1979,1991], which is the superposition of several ArmstrongFrederick rules, becomes a natural choice
as a viable hardening rule to be used with the uncoupled models. The OhnoWangs [1993] modification
64
Paper IITo be published in the International Journal of Plasticity (in press)
After evaluating the Chaboche kinematic hardening rule with two, three and four decomposed
terms, it has been found that with the modified DafaliasPopov model three decomposed terms
3
i (  C i d a i dp ) ,
2
d a i ,
p
da = d ai = 3
(22)
i=1
yields the optimum performance in simulation of the biaxial ratcheting responses. In Eq. (22), the scalar
is determined using the consistency condition, Eq. (8), for each loading increment and Ci, i are model
parameters to be determined from multiaxial responses. It is demonstrated by Bari and Hassan [2000a]
that in the Chaboche [1991] coupled model, each decomposed rule has its specific purpose in reproducing
different features of the uniaxial, nonlinear stable hysteresis loop. However, with the DafaliasPopov
uncoupled model, this hardening rule loses the physical bearing for uniaxial loading responses as the
plastic modulus is calculated by the uncoupled model itself. Therefore, with an uncoupled model the
parameters of the Chaboche rule have significance only in multiaxial loading conditions and are
In order to determine the parameters of the Chaboche rule, the qualitative nature of the parameters
observed in uniaxial loading cases (Bari and Hassan [2000a]) is also adopted here. Since C = 140 ksi in
the original ArmstrongFrederick rule yields good results for most of the ratcheting responses (shown in
Fig. 3), the same value is used here for the third rule (C3 = 140 and 3 = 1). The rest of the parameters are
determined by trials to conform with the nonlinear ratcheting rate of a constant pressure biaxial
experiment (test (2) of Fig. 8c). The general guideline followed is that the first decomposed rule (i = 1)
should have a low C1 value with a large 1 to represent the initial high rate of ratcheting region, while C2
and 2 should have relatively moderate values to represent the transitional nonlinear region. The values of
the parameters determined are: C13 = 4, 20, 140; 13 = 1200, 20, 1.
65
Paper IITo be published in the International Journal of Plasticity (in press)
xp xp
4 4
CS 1026 CS 1026
xa = 32.0 ksi xm = 9.14 Ksi xm = 6.5 ksi
(%) (%) a = 33.28 Ksi
x
3 3 xm xa
t
xm = 4.18 Ksi
1 1
Experiment
DP model xa = 28.29 Ksi
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(a) (b)
p p
4 4
CS 1026 CS 1026
(%) xc m = 9.65 Ksi (%) xc = 0.50 %
2 2
(3) (2) (4) (3)
(2)
(1)
1 1 (1)
Experiment
Chaboche
OhnoWang
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(c) (d)
p p
4 4
CS 1026 CS 1018
(%) a = 2.36 Ksi (%) a = 2.36 Ksi
xc = 0.50 % xc = 0.50 %
3 3
N N
(e) (f)
Fig. 8. Ratcheting Predictions by the Chaboche and OhnoWang rules with the modified Dafalias
Popov model (DP). (a,b) Axial strain at positive stress peaks from uniaxial cycles, (c,d)
circumferential strain peaks from axial strain cycles with constant pressure, (e,f) Circumferential
strain peaks from bowtie and reverse bowtie cycles. Experiments are from Hassan and
Kyriakides [1992], Hassan et al. [1992] and Corona et al. [1996].
66
Paper IITo be published in the International Journal of Plasticity (in press)
Figure 8 shows the simulations by this model for all the experiments considered. Note again that,
the kinematic hardening rule with an uncoupled model do not have any effect on the uniaxial ratcheting
simulations (compare Figs. 3a,b to 8a,b). The kinematic hardening rule influences only the biaxial
ratcheting simulations, which is evident from Figs. 8c and 8d. The nonlinear nature of the ratcheting rates
in constant pressure biaxial experiments are reproduced well with the improved model. Moreover, the
ratcheting simulations for both the bowtie loading histories are also improved (see Figs. 8e,f) by the
a
It has been shown by Bari and Hassan [2000a] that introduction of the term d p 
i replacing
f ( i )
reducing the overprediction of the rate of ratcheting. This modification by OhnoWang [1993] is also
evaluated in this paper with the modified DafaliasPopov model in the following form:
i  C i d a i d 
3 ai
da = d a i , = 2 p p
(23)
3 f ( i )
d ai
i=1
The parameter determination scheme followed in this rule is the same as in case of the Chaboche rule.
The parameters obtained are: C13 = 4, 20, 30; 13 = 400, 12, 1.
Figure 8 also shows the simulations by the OhnoWang kinematic hardening rule (Eq. 23). It is
observed in the figure that this rule simulates all the biaxial ratcheting responses moreorless similar to
those as obtained from the Chaboche rule (Eq. 22). Only, the nonlinear nature of ratcheting rates is more
The Chaboche (Eq. 22) and the OhnoWang (Eq. 23) kinematic rules are further evaluated for a
reverse bowtie response with larger number of cycles as shown in Fig. 9. It is observed in this figure that
the OhnoWang rule simulates shakedown followed by reversal of ratcheting, which seemingly defies the
67
Paper IITo be published in the International Journal of Plasticity (in press)
p
4
CS 1018
xc
(%) a = 2.36 Ksi
xc = 0.50 % a
3
m = 14.54 Ksi
m
1
Experiment
Chaboche
OhnoWang
0
0 10 20 30 40 50
Fig. 9. Ratcheting Predictions by the Chaboche and OhnoWang rules with the
modified DafaliasPopov model for circumferential strain peaks from reverse
bowtie cycles. Experiments are from Corona et al. [1996].
experimental trend. On the other hand, the Chaboche rule predicts a constant rate of ratcheting at higher
cycles, which does not match the actual material responses exactly, but is clearly more representative.
In a coupled model, the plastic modulus calculation is directly tied to the kinematic hardening rule.
As a result, if all the parameters in the hardening rule are calibrated from uniaxial responses, the
multiaxial simulations by the coupled model suffer and vice versa (Bari and Hassan [2000a]). The
uncoupling of plastic modulus calculation from the kinematic hardening rule allows the plastic modulus
parameters to be determined from uniaxial responses and the kinematic rule parameters from multiaxial
responses. These two parameters sets are uncoupled, i.e, the uniaxial simulations always remain
This study evaluates eight different kinematic hardening rules in order to identify or develop a
kinematic hardening rule that works well with an uncoupled model for both uniaxial and multiaxial
ratcheting simulation. The modified DafaliasPopov model that has been demonstrated to be one of the
68
Paper IITo be published in the International Journal of Plasticity (in press)
best uncoupled model for uniaxial ratcheting simulation (Hassan and Kyriakides [1992,1994a]) is used in
this study. Six uniaxial and ten biaxial ratcheting responses on stabilized carbon steels, acquired from
Hassan and Kyriakides [1992], Hassan et al. [1992] and Corona et al. 1996], are considered for evaluation
of the kinematic rules. The rateindependent and cyclically stable material ratcheting simulations from
the kinematic hardening rules by Armstrong and Frederick [1966], Phillips [1972,1979], Tseng and Lee
[1983], Voyiadjis and Sivakumar [1991,1994], Kaneko [1981,1984], Xia and Ellyin [1994,1997],
Chaboche [1991] and OhnoWang [1993] are evaluated. Many of these rules has demonstrated success in
simulating some specific type of ratcheting experiment. In this paper, however, their performances are
The original Armstrong and Frederick [1966] rule performs very well for all the constant pressure
biaxial ratcheting responses when the parameter C is determined from one of these experiments. But this
The Voyiadjis and Sivakumar [1991] rule is a combination of the Phillips (Phillips and Tang [1972],
Phillips and Lee [1979]) and TsengLee [1983] rules. None of these three rules perform satisfactorily for
the set of biaxial experiments considered. In reality, when the center of the yield surface translates in the
stressrate direction, the yield surface changes its shape too (Phillips and his coworkers
[1972,1979,1984]). Thus any effort to use a hardening rule that dictates the yield surface translation in the
stressrate direction is expected to be futile for ratcheting simulations, unless, the change of the shape of
The Kaneko [1981,1984] rule is a modification over the Ziegler [1959] rule. It has been shown by
Hassan et al. [1992] that the Ziegler rule ceases to simulate biaxial ratcheting strains after the first cycle
for the loading history in Fig. 1b. The Kaneko rule predicts negative ratcheting for this history due to the
addition of the prestress path dependence. The Xia and Ellyin [1994,1997] rule, on the other hand,
69
Paper IITo be published in the International Journal of Plasticity (in press)
overpredicts all the biaxial ratcheting cases due to the influence of the Mroz [1967] rule.
The Chaboche [1991] kinematic hardening rule, with three decomposed terms, when employed in
the modified DafaliasPopov model (Hassan and Kyriakides [1992,1994a]) performs the best for the
whole set of ratcheting responses considered. This rule also captures the transient ratcheting rates at
initial cycles and the constant ratcheting rates at higher cycles as observed in material ratcheting
The ratcheting simulations by the OhnoWang rule, which is a variation of the Chaboche rule, are
reasonable for small number of cycles, but predicts shakedown and subsequent reversal of ratcheting for
higher number of cycles in the reverse bowtie loading case (Fig. 9).
In an effort to obtain a constitutive model to simulate both uniaxial and multiaxial ratcheting, the
modified DafaliasPopov uncoupled model recognizes the essential difference in the modeling of uniaxial
and multiaxial ratcheting and separates the two cases by uncoupling the plastic modulus calculation from
the kinematic hardening rule. As a result, with a suitable kinematic hardening rule (the Chaboche rule in
this study) whose parameters are determined from a biaxial ratcheting experiment, this uncoupled model
manifest success in simulating all the ratcheting responses considered. The improvement in simulation by
this model is found to be significant when the simulations are compared to those from many well known
Since the yield surface undergoes shape changes during plastic deformation (Phillips and his co
workers [1972,1979,1984], and others), the normals to the actual yield surfaces are different from the
ones predicted by any models that use the vonMises surface. This would result in differences between
the actual and the predicted ratcheting strains, unless it is compensated for by an artificial movement of
the idealized yield surface through a kinematic hardening rule (Corona et al. [1996]). Most of the rules
70
Paper IITo be published in the International Journal of Plasticity (in press)
which are subsets of the Ziegler, Mroz and Phillips directions can not perform this compensation
effectively. The ArmstrongFrederick rule with one parameter, calibrated from a biaxial ratcheting
experiment, performs the compensation reasonably for similar biaxial experiments. However, this rule
fails when the loading history is changed significantly (for instance, the bowtie case) from the calibrated
experiment. On the other hand, the Chaboche rule that combines three ArmstrongFrederick rules
performs the required compensation more effectively in disparate loading cases, when its six parameters
are calibrated from a biaxial ratcheting experiment. However, the lack of physical meaning of the
parameters of the Chaboche rule makes the parameter determination scheme less definitive. Further
validation of this modeling scheme using different nonproportional ratcheting experiments is needed in
order to verify its generality. The authors strongly feel that, in order to achieve a more general and robust
ratcheting simulation model, the incorporation of yield surface shape change into the modeling scheme is
Acknowledgment  The financial supports from the Center for Nuclear Power Plant Structures, Equipment
and Piping and the Department of Civil Engineering at North Carolina State University are gratefully
acknowledged.
NOMENCLATURE
71
Paper IITo be published in the International Journal of Plasticity (in press)
REFERENCES
Armstrong, P.J. and Frederick, C.O. (1966) A Mathematical Representation of the Multiaxial Bausch
Bari, S. and Hassan. T. (2000a) Anatomy of Coupled Constitutive Models for Ratcheting Simulation.
Bari, S. and Hassan. T. (2000b) An Advancement in Cyclic Plasticity Modeling for Multiaxial Ratcheting
Simulation. Accepted for publication in the International Journal of Plasticity, Sept. 2000.
72
Paper IITo be published in the International Journal of Plasticity (in press)
Chaboche, J.L., DangVan, K. and Cordier, G. (1979) Modelization of the Strain Memory Effect on the
Cyclic Hardening of 316 Stainless Steel. Proceedings of the 5th International Conference on SMiRT,
Chaboche, J.L. (1991) On Some Modifications of Kinematic Hardening to Improve the Description of
Chaboche, J.L. (1994) Modeling of Ratchetting: Evaluation of Various Approaches. European Journal of
Corona, E., Hassan, T. and Kyriakides, S. (1996) On the Performance of Kinematic Hardening Rules in
Predicting a Class of Biaxial Ratcheting Histories. International Journal of Plasticity, Vol 12, pp. 117
145.
Dafalias, Y.F. and Popov, E.P. (1976) Plastic Internal Variables Formalism of Cyclic Plasticity. Journal of
Delobelle, P., Robinet, P. and Bocher, L. (1995) Experimental Study and Phenomenological Modelization
of Ratcheting Under Uniaxial and Biaxial Loading On an Austenitic Stainless Steel. Internatinal Jour
Drucker, D.C. and Palgen, L. (1981) On stressStrain Relations Suitable for Cyclic and Other Loadings.
Hassan, T. and Kyriakides, S. (1992) Ratcheting in Cyclic Plasticity, Part I: Uniaxial Behavior. Interna
Hassan, T., Corona, E. and Kyriakides, S. (1992) Ratcheting in Cyclic Plasticity, Part II: Multiaxial
73
Paper IITo be published in the International Journal of Plasticity (in press)
Hassan, T. and Kyriakides, S. (1994a) Ratcheting of Cyclically Hardening and Softening Materials, Part
Hassan, T. and Kyriakides, S. (1994b) Ratcheting of Cyclically Hardening and Softening Materials, Part
II: Multiaxial Behavior. International Journal of Plasticity, Vol 10, pp. 185212.
Jiang, Y. and Sehitoglu, H. (1996) Modeling of Cyclic Ratchetting Plasticity, Part I: Development of Con
Kaneko, K. (1981) Proposition of New Translation Rule in Kinematic Hardening. Bulletin of the JSME,
Kaneko, K. (1984) Development of a New Practical Plastic Constitutive Model for Anisotropic Metals
After Various Preloading. Bulletin of the JSME, Vol 27, pp. 26872693.
McDowell, D.L. (1995) Stress State Dependence of Cyclic Ratcheting Behavior of Two Rail Steels.
Mroz, Z. (1967) On the Description of Anisotropic Work Hardening. Journal of the Mechanics and Phys
Ohno, N. and Wang, J.D. (1993) Kinematic Hardening Rules with Critical State of Dynamic Recovery,
Part I: Formulations and Basic Features for Ratcheting Behavior. International Journal of Plasticity,
Phillips, A. and Tang, J.L. (1972) The Effect of Loading Path on the Yield Surface at Elevated Tempera
Phillips, A. and Lee, C.W. (1979) Yield Surfaces and Loading Surfaces. Experiments and Recommenda
tions. International Journal of Solids and Structures, Vol. 15, pp. 715729.
Phillips, A. and Lu, W.Y. (1984) An Experimental Investigation of Yield Surfaces And Loading Surfaces
of Pure Aluminum With StressControlled And StrainControlled Paths of Loading. Journal of Engi
neering Materials and Technology, Trans. ASME, Vol. 106, pp. 349354.
74
Paper IITo be published in the International Journal of Plasticity (in press)
Tseng, N.T. and Lee, G.C. (1983) Simple Plasticity Model of the TwoSurface Type. ASCE Journal of
Voyiadjis, G.Z. and Sivakumar, S.M. (1991) A Robust Kinematic Hardening Rule for Cyclic Plasticity
with Ratcheting Effects, Part I: Theoretical Formulation. Acta Mechanica, Vol 90, pp. 105123.
Voyiadjis, G.Z. and Sivakumar, S.M. (1994) A Robust Kinematic Hardening Rule for Cyclic Plasticity
with Ratcheting Effects, Part II: Application to Nonproportional Loading Cases. Acta Mechanica, Vol
Voyiadjis, G.Z. and Basuroychowdhury, I.N. (1998) A Plasticity Model for Multiaxial Cyclic Loading
Xia, Z. and Ellyin, F. (1994) Biaxial Ratcheting Under Strain or StressControlled Axial Cycling With
Constant Hoop Stress. Journal of Applied Mechanics, Vol 61, pp. 422428.
Xia, Z. and Ellyin, F. (1997) A Constitutive Model with Capability to Simulate Complex Multiaxial
Ratcheting Behaviour of Materials. International Journal of Plasticity, Vol 13, pp. 127142.
Yoshida, F., Tajima, N., Ikegami, K. and Shiratori, E. (1978) Plastic Theory of Mechanical Ratcheting.
Ziegler, H. (1959) A Modification of Paragers Hardening Rule. Quarterly of Applied Mechanics, Vol 17,
pp. 5565.
75
CHAPTER 4
76
Paper IIITo be published in the International Journal of Plasticity (in press)
ABSTRACT
In a search for a constitutive model for ratcheting simulations, the models by Chaboche, Ohno
against a set of uniaxial and biaxial ratcheting responses. With the assumption of invariant shape of the
yield surface during plastic loading, the ratcheting simulations for uniaxial loading are primarily a
function of the plastic modulus calculation, whereas the simulations for multiaxial loading are sensitive
to the kinematic hardening rule of a model. This characteristic of the above mentioned models is
elaborated in this paper. It is demonstrated that if all parameters of the kinematic hardening rule are
determined from uniaxial responses only, these parameters primarily enable a better plastic modulus
calculation. However, in this case the role of the kinematic hardening rule in representing the ratcheting
responses for multiaxial loading is underappreciated. This realization motivated many researchers to
incorporate multiaxial load dependent terms or parameters into the kinematic hardening rule. This paper
evaluates some of these modified rules and finds that none is general enough to simulate the ratcheting
responses consistently for the experiments considered. A modified kinematic hardening rule is proposed
using the idea of Delobelle and his coworkers in the framework of the Chaboche model. This new rule
introduces only one multiaxial load dependent parameter to the Chaboche model, but performs the best in
77
Paper IIITo be published in the International Journal of Plasticity (in press)
I. INTRODUCTION
An earlier paper by the authors evaluates several coupled cyclic plasticity models in terms of their
ability to simulate a set of uniaxial and multiaxial ratcheting responses (Bari and Hassan, 2000a). In these
models, the plastic modulus calculation is coupled with the kinematic hardening rule through the
consistency condition. Among the coupled models evaluated, the Chaboche (1991) and the OhnoWang
(1993) models show promise. Both of these models demonstrate success in simulating ratcheting for
uniaxial loading cases, but consistently overpredict ratcheting for multiaxial loading cases.
ratcheting responses in uniaxial loading primarily depend on the plastic modulus, whereas the
simulations in multiaxial loading depend on both the plastic modulus and the kinematic hardening rule,
with significant influence from the latter. This feature is elaborated through Fig. 1. For a uniaxial load
increment AB, in Fig. 1a, the normal direction, n , at the stress point B remains parallel to that at A
irrespective of the kinematic hardening rule adopted in a model. This is also true for an increment CD
during reverse loading. As a result, during uniaxial loading cycles OBDBD...., direction and magnitude
f
of the term 
in the flow rule (Eq. 2 in section III) remain unchanged and the plastic strain increments
become a function of the plastic modulus (H) only. Therefore, simulations of uniaxial ratcheting
responses depend entirely on the accuracy of the plastic modulus calculation of a model.
This feature in multiaxial loading is demonstrated for a biaxial loading history OACDCD.....,
obtained by superimposing a constant 2 stress to a stress cycle along 1 (Fig. 1b), which is a small
deviation from the uniaxial loading cycle in Fig. 1a. During a biaxial loading increment BC (Fig. 1b),
normal direction at C changes from that at B. Similar changes occur continuously throughout the loading
history. It is clearly observed in Fig. 1 that normal directions in biaxial loading are significantly different
from those in uniaxial loading. This change in normal direction during a biaxial loading results from the
78
Paper IIITo be published in the International Journal of Plasticity (in press)
1 1
n
n
n C n
B
d d
A B
d d
E n
O 2 O A 2
C
D
n
D
n
(a) (b)
Fig. 1. Yield surface translation and normal directions during (a) uniaxial and (b) biaxial loading
cycles.
shift of the stress point along the yield surface when it translates. Hence, the normal direction in
multiaxial loading is a function of the direction and magnitude of the yield surface translation, which, in
turn, is dictated by the kinematic hardening rule of a model. Different kinematic hardening rule produces
different translation direction and, thereby, greatly varied normal directions for a given stress history. As
a result, the simulation of ratcheting responses under multiaxial loading by a plasticity model depends
In coupled models, the plastic modulus (H) is calculated using the kinematic hardening rule and the
consistency condition. Usually, the parameters of a kinematic hardening rule are determined from
uniaxial loading responses. These parameters are, in effect, calibrated to produce a better representation
of the plastic modulus only, and consequently, fall short in predicting the representative yield surface
translation and subsequent normal directions. As a result, ratcheting simulations by many coupled models
are rarely good in multiaxial loading cases (Bari and Hassan, 2000a).
79
Paper IIITo be published in the International Journal of Plasticity (in press)
One solution to this problem can be achieved by uncoupling the plastic modulus calculation from
the kinematic hardening rule and, thereby, employing the kinematic hardening rule only to dictate the
yield surface translation. In this methodology, parameters for the plastic modulus calculation are
determined from uniaxial responses and parameters for the kinematic hardening rule are determined
using multiaxial responses (Hassan et al., 1992). The authors have demonstrated reasonable success from
such a modeling scheme in simulating ratcheting responses for a set of uniaxial and biaxial experiments
A solution to the problem for the coupled models has been attempted by incorporating multiaxial
terms and parameters into the Chaboche or the OhnoWang model (McDowell, 1995, Jiang and
Sehitoglu, 1996a, Voyiadjis and Basuroychowdhury, 1998, and AbdelKarim and Ohno, 2000). Most of
these modified models have simulated the ratcheting responses from tensionshear biaxial loading cycles
and rolling/sliding contact problems (see also McDowell, 1997, Jiang and Sehitoglu, 1996b, 1999, Ekh et
al., 2000, Ringsberg, 2000). In this study these models are evaluated using a different set of ratcheting
responses. This study also proposes an improved kinematic hardening rule by incorporating only one
Model performance in this study is evaluated with respect to the same set of ratcheting responses
used in the earlier works by the authors (Bari and Hassan, 2000a,b). This data set on cyclically stabilized
carbon steels is acquired from Hassan and Kyriakides (1992), Hassan et al. (1992) and Corona et al.
The loading histories of the experiment set are shown in Fig. 2. The history in Fig. 2a, which results
in axial strain ratcheting, involves axial stress cycles with a mean stress. Readers are referred to Fig. 7 in
Hassan and Kyriakides (1992) for demonstration of a typical uniaxial ratcheting response. Data from
80
Paper IIITo be published in the International Journal of Plasticity (in press)
xc xc
xc
x a a
xa
xm m
m m
t
x x x
(a) (b) (c) (d)
Fig. 2. Loading histories studied; (a) Uniaxial stress cycle, (b) Axial strain cycle with constant
internal pressure, (c) biaxial bowtie cycle, (d) biaxial reverse bowtie cycle.
uniaxial loading experiments are presented in Figs. 3a and 3b, where the peak axial strain (xp) in each
cycle is plotted against the number of cycles, N. The biaxial loading history in Fig. 2b involves axial
strain cycles in the presence of a constant internal pressure. This loading history results in circumferential
strain ratcheting as demonstrated in Fig. 1 of Hassan et al. (1992). Ratcheting responses with this biaxial
loading history are shown in Figs. 3c and 3d, where the peak circumferential strain (p) in each cycle is
plotted against the number of cycles, N. The bowtie and reverse bowtie loading histories, shown in Figs.
2c and 2d, also result in circumferential strain ratcheting as demonstrated in Figs. 10 and 11 of Corona et
al. (1996). Response data from the bowtie histories are shown in Figs. 3e and 3f. An important feature of
the bowtie histories is that they have resemblance to the loading histories observed in piping components
(Corona et al., 1996). Hence, a success in simulating these responses by a model would imply its
suitability for analysis and design of ratcheting in piping components. In this study, the ratcheting
responses from the uniaxial loading experiments are referred to as uniaxial ratcheting, and those from the
81
Paper IIITo be published in the International Journal of Plasticity (in press)
12
vonMises yield criterion: f ( ) = 3 ( s a ) ( s a ) = 0 , (1)
2
1 f f
flow rule: d
p
=   d  , (2)
H
p p
kinematic hardening rule: da = g ( , , a, d, d , etc ) , (3)
p
where, is the stress tensor, is the plastic strain tensor, s is the deviatoric stress tensor, is the
current center of the yield surface, a is the current center of the yield surface in the deviatoric space, 0 is
the size of the yield surface (constant for a cyclically stable material), and H is the plastic modulus. Also,
In coupled models, the plastic modulus H is evaluated using the consistency condition, f = 0 , the
kinematic hardening rule (Eq. 3), the flow rule (Eq. 2), and the yield criterion (Eq. 1) (see Chaboche,
1986). From the point of view of multiaxial ratcheting simulation, the primary distinction between
different models lies in the kinematic hardening rule. All of the models studied are modifications over
either the Chaboche (1991) or the OhnoWang (1993) rules, which are briefly reviewed below. The
kinematic hardening rules proposed by McDowell (1995), Jiang and Sehitoglu (1996a), Voyiadjis and
Basuroychowdhury (1998), and AbdelKarim and Ohno (2000) are then evaluated. Finally, a new
kinematic hardening rule is proposed in the framework of the Chaboche (1991, 1994) model and
82
Paper IIITo be published in the International Journal of Plasticity (in press)
Chaboche (1991, 1994) proposes a decomposed nonlinear kinematic hardening rule in the form:
4
2
d a i ,
p
da = da i =  C d i a i dp , for i = 1,2,3 and
3 i
i=1
ai
da i =  C i d i a i 1  dp ,
2 p
for i = 4 (4)
3 f ( i )
p 2 p p
12
3 12
where dp = d =  d d and f ( i ) =  a i a i .
3 2
As can be observed in Eq. 4, the first three decomposed rules are the same as the ArmstrongFrederick
(1966) hardening rule. The fourth rule contains a threshold level of backstress, a4, that makes the
dynamic recovery term inactive within the threshold (Chaboche, 1991). Outside the threshold, the fourth
rule evolves according to the ArmstrongFrederick rule. The role of each decomposed rule is discussed
elaborately in Bari and Hassan (2000a), where it is also demonstrated that the Chaboche model performs
well in simulating the uniaxial ratcheting responses (see Figs. 3a and 3b). However, this model
overpredicts ratcheting strains for all biaxial loading responses considered in the study (see Figs. 3c to
C14 = 60000, 3228, 455, 15000 (ksi); 14 = 20000, 400, 11, 5000; a4 = 5 ksi.
All the parameters in this model are determined using uniaxial responses (or, in effect, are chosen for
better plastic modulus calculation). These parameters can not produce representative yield surface normal
directions (or plastic strain rate directions) and hence, fail to simulate the biaxial ratcheting responses
considered.
83
Paper IIITo be published in the International Journal of Plasticity (in press)
xp xp
4 4
CS 1026 CS 1026 xa = 33.28 ksi
xa = 32.0 ksi xm = 9.14 ksi xm = 6.5 ksi
(%) (%)
x
xa
3 3 xm xa= 32.12 ksi
t
xm = 6.52 ksi
2 2
xa = 30.35 ksi
xm = 4.18 ksi
1 1 xa = 28.29 ksi
Experiment
Chaboche
OhnoWang
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(a) (b)
p p
4 4
CS 1026 CS 1026
(%) m = 9.65 ksi (%) xc = 0.50 %
N N
(c) (d)
p p
4 4
CS 1026 CS 1018
(%) a = 2.36 ksi (%) a = 2.36 ksi
xc = 0.50 % xc = 0.50 %
3 3
(1) m = 9.65 ksi (1) m = 9.65 ksi
(2) m = 14.54 ksi (2) m = 14.54 ksi
2 xc 2
(2)
a (2) xc
(1)
m a
(1)
1 x 1 m
Experiment
Chaboche x
OhnoWang
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(e) (f)
Fig. 3. Ratcheting predictions by the Chaboche and the OhnoWang models. (a,b) Axial strain at
positive stress peaks from uniaxial cycles, (c,d) circumferential strain peaks from axial strain cycles
with constant pressure, (e,f) circumferential strain peaks from bowtie and reverse bowtie cycles.
Experiments are from Hassan and Kyriakides (1992), Hassan et al. (1992) and Corona et al. (1996).
84
Paper IIITo be published in the International Journal of Plasticity (in press)
Ohno and Wang (1993) also introduce thresholds to control evolution of decomposed kinematic
hardening rules. Each decomposed rule stops evolving outside of its threshold, Ci/i. The OhnoWang
mi
M a i f ( i )
da i =  C i d i a i d 

2
d a i ,
p p
da =  . (5)
3 f ( i ) C i i
i=1
where, f ( i ) is the vonMises function as defined in Eq. 1. All the parameters in this model are
determined using responses under uniaxial loading. Hence, the uniaxial ratcheting simulations from the
model are quite good and comparable to those from the Chaboche model (Figs. 3a,b). The parameters
C112 = 31940, 36214, 2520, 376, 11021, 4551, 3475, 2196, 857, 247, 98, 200 (ksi);
112 = 45203, 13944, 7728, 4955, 3692, 2135, 1230, 585, 295, 119, 50, 20;
mi = 0.45, i = 1, 12
A detailed discussion of the OhnoWang model and its parameter determination scheme can be found in
The simulations by the OhnoWang model for the biaxial loading cases are shown in Figs. 3c to 3f.
Note in these figures that the amount by which ratcheting is overpredicted by this model is generally
reduced compared to the simulations by the Chaboche model. This is mainly caused by the term
a
d p 
i that replaces dp (plastic strain increment magnitude) of the ArmstrongFrederick (1966)
f ( )
i
rule. This new term is instrumental in introducing some effects of multiaxial loading and thus improves
85
Paper IIITo be published in the International Journal of Plasticity (in press)
the models multiaxial ratcheting simulation. Moreover, it does not affect uniaxial ratcheting simulation,
as it reduces to dp for uniaxial loading. Although the nonlinear nature of the biaxial ratcheting responses
is captured well by the OhnoWang model (Figs. 3c to 3f), overpredictions still persist for all biaxial
loading cases presented. Similar to the Chaboche model, as all of the parameters of this model are
determined using only uniaxial responses, these parameters fall short in simulating the multiaxial
In an effort to improve the capability of the OhnoWang (1993) model for multiaxial ratcheting
ai Bi p
m i = A i n' 
,  ( s a ) .
d 3
where n' = 
= (6)
f ( i ) dp 2 0
The exponents mi, a constant in the OhnoWang rule, is proposed to be dependent on the noncoaxiality of
the plastic strain rate and the backstress in this model. Also, note that the expression for the exponents mi
(Eq. 6) include the constants Bi, which can be calibrated using multiaxial ratcheting responses to
influence the multiaxial ratcheting simulations without affecting the uniaxial simulations.
a
For uniaxial loading, the exponents mi reduce to the constants Ai as the terms n' 
i become
f ( i )
unity. As a result, for uniaxial loading, this rule reduces to the OhnoWang rule, which is also validated
by the ratcheting simulations in Fig. 4a. As the parameters of the OhnoWang model are determined from
uniaxial responses, these parameters can also be used with the McDowell model (with Ai = mi = 0.45 for
i = 1, 12). When the multiaxial parameters Bi are zeros, for all 12 decomposed terms, the biaxial
ratcheting simulations by the McDowell model exactly match the simulations by the OhnoWang model.
With increasing Bi, the McDowell model predicts higher ratcheting strains than the OhnoWang model
86
Paper IIITo be published in the International Journal of Plasticity (in press)
xp p
4 4
CS 1026 CS 1026
(%) xa = 32.0 ksi (%) m = 9.65 ksi
xc = 0.50 %
3 3
xm = 9.14 ksi xm = 6.52 ksi
xc
m
2 2
x
xm = 4.18 ksi
1 Experiment 1 Experiment
OhnoWang OhnoWang
McDowell, Bi = 2 McDowell, Bi = 2
JiangSehitoglu JiangSehitoglu
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(a) N (b) N
p p
4 4
CS 1026 CS 1018
(%) a = 2.36 ksi (%) a = 2.36 ksi
xc = 0.50 % xc = 0.50 %
3 m = 9.65 ksi 3 m = 9.65 ksi
xc xc
a a
2 2
m m
x x
1 Experiment 1 Experiment
OhnoWang OhnoWang
McDowell, Bi = 2 McDowell, Bi = 2
JiangSehitoglu JiangSehitoglu
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(c) N (d) N
Fig. 4. Ratcheting predictions by the McDowell and the JiangSehitoglu models. (a) Axial strain at
positive stress peaks from uniaxial cycles, (b) circumferential strain peaks from axial strain cycles
with constant pressure, (c,d) circumferential strain peaks from bowtie and reverse bowtie cycles.
Experiments are from Hassan and Kyriakides (1992), Hassan et al. (1992) and Corona et al. (1996).
for biaxial loading cases, as shown in Figs. 4b to 4d for Bi = 2. These results demonstrate that the
incorporation of a multiaxiality effect into the exponents mi according to the McDowell (1995) model
(Eq. 6) is not conducive to better biaxial ratcheting simulations for the experiments considered in this
study.
87
Paper IIITo be published in the International Journal of Plasticity (in press)
Jiang and Sehitoglu (1996a) proposed another variation of the OhnoWang (1993) model, with
exponents mi dependent on the noncoaxiality of plastic strain rate and the backstress, as follows:
mi
M
2 f ( i )
d a i ,
p
da = da i =  C i d i 
 a i dp . (7)
i=1
3 C i i
a p
m i = A 0i 2 i , =  ( s a ) .
d 3
where, n' 
f ( )
n' = 
dp 2 0
i
Similar to the McDowell (1995) model, the exponents mi in this model assume constant values (mi = A0i)
and the model reduces to the OhnoWang model for uniaxial loading. Hence, the OhnoWang model
parameters can also be used with this model and consequently, its uniaxial ratcheting simulations remain
the same as the OhnoWang model (Fig. 4a). The JiangSehitoglu (1996a) model, however, overpredicts
the ratcheting responses for the biaxial loading cases as shown in Figs. 4b to 4d. As this model replaces
p a
i  in the OhnoWang model by dp (compare Eq. 5 with Eq. 7), it overpredicts the
the term d 
f ( )
i
ratcheting responses in a manner similar to the Chaboche (1991) model for the biaxial loading histories
considered.
the Chaboche (1991) kinematic hardening rule by adding the direction of the stress rate ds . Their rule is
used in a slightly modified form in this study as follows:
M
2
d a i ,
p
da = da i =  C d i a i dp + i  ds , for i = 1,2,3 and
3 i m
i=1
ai
da i =  C i d i a i 1  dp + i  ds ,
2 p
for i = 4. (8)
3 f ( i ) m
88
Paper IIITo be published in the International Journal of Plasticity (in press)
where i is a material parameter, is the distance of the current stress point from the bounding surface
along the direction of ds , m is the chord length of the bounding surface along the same direction. Due to
the dependence of and m on the stress rate, the numerical solution for Eq. (8) becomes computationally
intensive because of the iterations involved when strains are prescribed as loading. Therefore, and m
along the ( s s ) direction (Mroz, 1967) are used in this study as follows:
3 12 2 3
=  ( s s ) ( s s ) and m =   ( s b ) ( s s ) . (9)
2 2
The deviatoric stress point s ( ) is the image of s ( ) on the bounding surface and b is the current center
of the bounding surface in the deviatoric space (see Fig. 5 for different parameters in the stress space
where is the current center of the bounding surface). Intersections of the yield and the bounding
2 Yield
Bound Surface
Eo P Bounding
b Surface
o
o
p 1
m
n
n
(a) (b)
Fig. 5. Definition of parameters and m in the modified VoyiadjisBasuroychowdhury
model; (a) in uniaxial p space, (b) in biaxial 12 stress space.
surfaces are avoided by translating the bounding surface according to Dafalias and Popov (1976):
p
E ds n 3 (s s) 3 (s a)
db = da dM , d M = 1 0 
 , =  
 , and n =  
 , (10)
H n 2 2 0
89
Paper IIITo be published in the International Journal of Plasticity (in press)
where E0p defines the plastic modulus at the bounding surface, is the unit vector along ( s s ) and n is
the unit normal to the yield surface at the current stress point. It is important to note here that the purpose
of incorporation of the bounding surface in this model is not to calculate the plastic modulus using the
concept of Dafalias and Popov (1976), but to improve the multiaxiality effect in the modeling.
All parameters of this model which are to be determined using the uniaxial responses are kept the
same as in the Chaboche model (see Section III.1). As stated by Voyiadjis and Basuroychowdhury
(1998), and also observed in this study, parameters 1 and 2 do not have much influence on ratcheting
simulations. Therefore, the same values (1, 2 = 0.15) as used by Voyiadjis and Basuroychowdhury
(1998) are also used in this study. Ratcheting simulations from Eq. 8, for two sets of 3, 4 values (=
0.025, 0.05), are compared to the simulations from the Chaboche model in Fig. 6. As seen in this figure,
the modified Voyiadjis and Basuroychowdhury rule (Eq. 8) is performing inadequately for one uniaxial
experiment (Fig. 6a). For the biaxial loading cases, the simulations by this rule are not improving
AbdelKarim and Ohno (2000) propose a new kinematic hardening rule combining the initial
(multilinear) version of the OhnoWang (1993) and the ArmstrongFrederick (1966) rules as follows:
M
da = d a i ,
2
da i =  C i d
3
p
i i ai dp i ai H { 32 ai ai ( C i i ) 2 } d i (11)
i=1
p ai
where, d i = d 
 i dp ,
Ci i
90
Paper IIITo be published in the International Journal of Plasticity (in press)
xp p
4 4
CS 1026 x CS 1026
(%) xa = 32.0 ksi xa (%) xc = 0.50 %
xm
xm = 6.52 ksi m = 9.65 Ksi
t
3 3
xc
m
2 2
x
1 Experiment 1 Experiment
VB (3,4 =.025) VB (3,4 =.025)
VB (3,4 =.05) VB (3,4 =.05)
Chaboche Chaboche
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(a) (b)
p p
4 4
CS 1026 CS 1018
(%) a = 2.36 Ksi (%) a = 2.36 Ksi
xc = 0.50 % xc = 0.50 %
3 m = 9.65 Ksi 3 m = 9.65 Ksi
xc xc
a a
2 2
m m
x x
1 Experiment 1 Experiment
VB (3,4 =.025) VB (3,4 =.025)
VB (3,4 =.05) VB (3,4 =.05)
Chaboche Chaboche
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(c) (d)
2 p
da i =  C i d
3
i ai dp , (13)
as d i in the third term in Eq. 11 never becomes active. The multilinear version of the OhnoWang
model (Eq. 12) simulates shakedown (cessation of ratcheting) for uniaxial loading and some ratcheting
91
Paper IIITo be published in the International Journal of Plasticity (in press)
for biaxial loading conditions (see Figs. 7a and 7b for i = 0). On the other hand, superposition of several
ArmstrongFrederick rules (Eq. 13) overpredicts ratcheting for both uniaxial and biaxial loading (see
Figs. 7a and 7b for i = 1). The same set of OhnoWang model parameters (except mi), presented in
AbdelKarim and Ohno (2000) propose to blend the ArmstrongFrederick and multilinear Ohno
Wang rules through a small value of i (between 0 and 1). For i = 0.07, this model produces a better
biaxial ratcheting simulation, especially in reproducing the recorded steady rate of ratcheting (see Fig.
7b). With the same value of i (= 0.07) this model, however, underpredicts the ratcheting in uniaxial
loading cases (Fig. 7a). The simulations by this model for other experiments, with i = 0.07, are shown in
Figs. 7c to 7f. It is observed from these figures that the simulations for all of the biaxial loading cases are
improved compared to the simulations by the OhnoWang model (compare Figs. 3c, 3e and 3f with Figs.
7d, 7e and 7f, respectively). However, the uniaxial ratcheting simulations by this model have deteriorated
(underpredicting) compared to the simulations by the OhnoWang model (compare Figs. 7c and 3a). It
should be noted that AbdelKarim and Ohno (2000) have verified their model using uniaxial experiments
If the AbdelKarimOhno (Eq. 11) and the OhnoWang (Eq. 5) kinematic hardening rules are
scrutinized, it is realized that both these models are quite comparable. For instance, i = 0 in Eq. 11 and
m i = in Eq. 5 yield multilinear version of the OhnoWang model, and thus simulate shakedown for
uniaxial loading and some ratcheting for biaxial loading (see Figs. 7a and 7b). Whereas i = 1 in Eq. 11
and m i = 0 in Eq. 5 yield overpredictions of ratcheting for both uniaxial and multiaxial loading. For the
latter case, the kinematic rules in both Eqs. 5 and 11 behave like the ArmstrongFrederick rule. The
parameters i and mi influence both the uniaxial and biaxial ratcheting responses. If the parameter i is
determined from a biaxial ratcheting response (as is done in this study), Eq. 11 yields a good
92
Paper IIITo be published in the International Journal of Plasticity (in press)
xp p
4 4
CS 1026 CS 1026
(%)
Experiment xa = 32.0 ksi (%) m = 9.65 ksi
AbdelOhno xm = 6.52 ksi xc = 0.50 %
i = 1
3 3
i = 1
i = 0.07
2 x 2
i = 0
xa xc
xm
t m
1 i = 0.07 1
x
Experiment
i = 0 AbdelOhno
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(a) N (b) N
xp p
4 4
CS 1026 CS 1026
xa = 32.0 ksi xm = 9.14 ksi Experiment m = 9.65 Ksi
(%) (%)
i = 0.07 AbdelOhno i = 0.07
x
3 3 (1) xc = 0.40 %
xa xm = 6.52 ksi (2) xc = 0.50 %
xm
(3) xc = 0.65 %
t
xm = 4.18 ksi
2 2 (3)
(2)
xc
1 1 (1)
m
x
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(c) N (d) N
p p
4 4
CS 1026 CS 1018
(%) Experiment a = 2.36 Ksi (%) Experiment a = 2.36 Ksi
AbdelOhno xc = 0.50 % AbdelOhno xc = 0.50 %
i = 0.07 i = 0.07
3 3
(1) m = 9.65 Ksi (1) m = 9.65 Ksi
(2) m = 14.54 Ksi (2) m = 14.54 Ksi
(2) (2)
2 2
(1)
xc xc
a (1)
a
1 1
m m
x
x
0 0
0 10 20 30 40 50 0 10 20 30 40 50
(e) N (f) N
Fig. 7. Ratcheting predictions by the AbdelKarimOhno model. (a,c) Axial strain at positive stress
peaks from uniaxial cycles, (b,d) circumferential strain peaks from axial strain cycles with constant
pressure, (e,f) circumferential strain peaks from bowtie and reverse bowtie cycles. Experiments
are from Hassan and Kyriakides (1992), Hassan et al. (1992) and Corona et al. (1996).
93
Paper IIITo be published in the International Journal of Plasticity (in press)
representation of the yield surface normal directions, but a somewhat poor estimation of the plastic
modulus. As a result, the biaxial ratcheting simulations for all cases are improved, whereas the uniaxial
ratcheting simulations are deteriorated (Fig. 7). On the other hand, if the parameter mi is determined from
a uniaxial ratcheting response, Eq. 5 produces good uniaxial ratcheting simulations but a consistent
The basis of the improved model is derived from the model by Burlet and Cailletaud (1986), who
modifies the radial evanescence term in the ArmstrongFrederick (1966) hardening rule as follows:
2 p 2 f 3 (s a)
da =  Cd ( a n )n dp , where n =   =  
 . (14)
3 3 2 0
The plastic modulus expression obtained with this hardening rule by satisfying the consistency condition
( f = 0 ) is the same as that obtained from the ArmstrongFrederick rule. Therefore, for uniaxial loading,
these two rules produce the same simulation, which is the overprediction of ratcheting under uniaxial
loading (Bari and Hassan, 2000a). For multiaxial loading, the radial evanescence term ( ( a n )n dp ) of the
Burlet and Cailletaud rule essentially yields a tensor along the plastic strainrate direction. As a result,
this rule behaves like the Prager (1956) rule and predicts shakedown for biaxial loading cases (see Fig. 8
for ' = 0). On the other hand, Bari and Hassan (2000a) demonstrated that the ArmstrongFrederick
(1966) rule overpredicts ratcheting for the same biaxial loading cases. To obtain ratcheting simulations
for multiaxial loading in between these two extremes (shakedown and overprediction), the idea of
which yields the BurletCailletaud rule for ' = 0, and the ArmstrongFrederick rule for ' = 1.
94
Paper IIITo be published in the International Journal of Plasticity (in press)
The plastic modulus expression (H) obtained for this rule (Eq. 15) does not include '. This means
that ' can be calibrated using a multiaxial ratcheting response and thus can influence multiaxial
ratcheting simulations without having any bearing on the plastic modulus calculation or the uniaxial
responses. However, the plastic modulus expression obtained from Eq. 15 is also the same as that
obtained from the ArmstrongFrederick rule, which overpredicts the uniaxial ratcheting. This drawback
of the rule (Eq. 15) can be fixed by superimposing several rules in the framework of the Chaboche (1991)
model as follows:
4
da i =  C i d i { 'a i + ( 1 ' ) ( a i n )n } dp
2
d a i ,
p
da = for i = 1,2,3 and
3
i=1
ai
da i =  C i d i { 'a i + ( 1 ' ) ( a i n )n } 1  dp ,
2 p
for i = 4. (16)
3 f ( i )
Equation 16 becomes the Chaboche (1991) rule for ' = 1 and the superimposed version of the Burlet and
Cailletaud (1986) rule for ' = 0. The plastic modulus expression (H) obtained from this rule is
independent of ' and the same as that obtained from the Chaboche rule. Therefore, all of the parameters
of the Chaboche model (which are determined using uniaxial responses) can be used with this modified
rule (Eq. 16). The new parameter ' is completely uncoupled from the plastic modulus calculation, and its
value can be calibrated using a multiaxial ratcheting response. Figure 8 shows the ratcheting simulations
for a constant pressure biaxial experiment by this improved rule for different values of '. It is observed
from this figure that, for ' = 0, this model predicts shakedown like the Prager model and, for ' = 1, this
model overpredicts ratcheting like the Chaboche model. However, for ' = 0.18, the ratcheting simulation
by this model improves compared to the simulations by the Chaboche and the OhnoWang models (Fig.
8).
A similar approach is also adopted by Geyer (1995) using only two decomposed rules following the
methodology of Burlet and Cailletaud (1987). This model, however, fails to reproduce several of the
95
Paper IIITo be published in the International Journal of Plasticity (in press)
p
4
CS 1026
(%) = 9.65 Ksi
xc = 0.50 %
3
xc
m
2 x
Experiment
1 = 0.18
= 0
= 1 (Chaboche)
OhnoWang
0
0 10 20 30 40 50
Fig 8. Ratcheting predictions by the improved rule (Eq. 16) for various ' and the Ohno
Wang model. The experimental data is obtained from Hassan et al. (1992)
ratcheting responses considered by Portier et al. (2000). It has been demonstrated by the authors (Bari
and Hassan, 2000a) that, while superposition of at least three (ArmstrongFrederick) decomposed rules is
necessary, four decomposed rules with a threshold term in the fourth rule proposed by Chaboche (1991)
performs the best in simulating the uniaxial hysteresis loops and ratcheting responses. This framework of
the Chaboche (1991) model is adopted here with the proposed kinematic hardening rule (Eq. 16). The
improved ratcheting simulations by the new rule, with ' = 0.18 and other parameters of the Chaboche
model (see section III.1), are presented in Fig. 9. As expected, the uniaxial ratcheting simulations are the
same as those from the Chaboche model (compare Figs. 3a, 3b to Figs. 9a, 9b, respectively). The
ratcheting simulations by this new rule for all biaxial loading cases conform reasonably well with the
experimental responses (see Figs. 9c to 9f). The simulations exhibit desired nonlinear ratcheting for lower
cycles and constant ratcheting rate for higher cycles. The rates of ratcheting at higher cycles match well
96
Paper IIITo be published in the International Journal of Plasticity (in press)
xp x
4 4
CS 1026 CS 1026 xa = 33.28 Ksi
xa = 32.0 ksi xm = 9.14 Ksi xm = 6.5 Ksi
(%) (%)
x
3 3 xa xa= 32.12 Ksi
xm
t
xm = 6.52 Ksi
2 2
xa = 30.35 Ksi
xm = 4.18 Ksi
1 1 xa = 28.29 Ksi
Experiment
Improved model
= 0.18
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(a) (b)
p p
4 4
CS 1026 CS 1026
(%) xc m = 9.65 Ksi (%) xc = 0.50 %
m
3 (1) xc = 0.40 % 3 (1) m = 4.88 Ksi
x (2) xc = 0.50 % (2) m = 7.32 Ksi
(3) xc = 0.65 % (3) m = 9.65 Ksi
(4) (4) m =14.54 Ksi
(3)
2 (2) 2 (3)
(3) (4) (3)
(1) (2)
(2) (2)
(1) (1)
1 1
Experiment (1)
Improved model
= 0.18
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(c) (d)
p p
4 4
CS 1026 CS 1018
(%) a = 2.36 Ksi (%) a = 2.36 Ksi
xc = 0.50 % xc = 0.50 %
3 3
(2) (1) m = 9.65 Ksi (1) m = 9.65 Ksi
(2) m = 14.54 Ksi (2) m = 14.54 Ksi
(2) (1) xc (2)
2 2
a (2)
(1) (1) xc
m
a
1 x 1 (1)
m
Experiment
Improved model x
= 0.18
0 0
0 10 20 30 40 50 0 10 20 30 40 50
N N
(e) (f)
Fig. 9. Ratcheting predictions by the Improved model (Eq. 16). (a,b) Axial strain at positive stress
peaks from uniaxial cycles, (c,d) circumferential strain peaks from axial strain cycles with constant
pressure, (e,f) circumferential strain peaks from bowtie and reverse bowtie cycles. Experiments
are from Hassan and Kyriakides (1992), Hassan et al. (1992) and Corona et al. (1996).
97
Paper IIITo be published in the International Journal of Plasticity (in press)
Evaluation of several coupled models demonstrates that most of the models are not general enough
to simulate the set of uniaxial and biaxial ratcheting responses considered in this study. This drawback of
the models is believed to be related to the assumption of the yield surface shape remain invariant during
plastic loading. With this simplifying assumption, the evolution of the yield surface normal directions
(plastic strain rate directions) is influenced primarily by the kinematic hardening rule of a model. Since
different kinematic rules produce varied estimates for the yield surface normals, the multiaxial ratcheting
Moreover, the simulation of ratcheting in uniaxial loading depends only on the plastic modulus
calculation. In coupled models, the kinematic hardening rule defines both the plastic modulus and the
yield surface translation in stress space. If all parameters of a kinematic hardening rule are calibrated
using only uniaxial responses, this rule can accurately estimate the plastic modulus, but its role in
dictating the yield surface translation is underestimated. Consequently, multiaxial ratcheting simulations
For example, in the Chaboche (1991) and the OhnoWang (1993) models, all parameters are
determined using uniaxial responses. These models perform reasonably well in simulating uniaxial
ratcheting responses (Figs. 3a and 3b), but overpredict consistently for all biaxial ratcheting responses
considered (Figs. 3c to 3f). The OhnoWang model, however, predicts a reduced rate of ratcheting
compared to the Chaboche model for all biaxial responses. This improvement can be attributed to the
a
term d p 
i , which introduces a multiaxiality effect in the model without affecting its uniaxial
f ( i )
ratcheting simulation.
Therefore, in order to achieve better simulations in both uniaxial and multiaxial cases, it is essential
for a coupled model to introduce special multiaxiality terms or parameters in the kinematic hardening
98
Paper IIITo be published in the International Journal of Plasticity (in press)
rule. These terms and parameters should remain dormant under uniaxial loading (or do not take part in
the plastic modulus calculation), but play an important role in defining representative yield surface
translation and thus plastic strain rate directions. As the actual yield surface shape deforms during plastic
loading (Phillips and Tang, 1972, Phillips and Lee, 1979), the lack of exactness introduced in the
modeling through invariant yield surface shape assumption is compensated to some extent by calibrating
This paper evaluates several coupled models that vie to improve their multiaxial ratcheting
simulations by introducing new parameter(s) or term(s) in the kinematic hardening rule. The models
considered are proposed by McDowell (1995), Jiang and Sehitoglu (1996a), Voyiadjis and
Basuroychowdhury (1998), and AbdelKarim and Ohno (2000). All of these models are modifications of
either the Chaboche (1991) or the OhnoWang (1993) model. In this paper, the performances of these
McDowell (1995) modifies the OhnoWang (1993) model by allowing the exponents mi (compare
Eqs. 5 and 6) to depend on the noncoaxiality of plastic strain rate and backstress. This model also
introduces parameters Bi that influence only the multiaxial responses and have no bearing on the uniaxial
responses. McDowells modification, however, does not manifest improvement over the OhnoWang
model in simulating the biaxial ratcheting responses considered in this study (see Figs. 4b to 4d).
Jiang and Sehitoglu (1996a) modify the exponents mi of the OhnoWang model to depend on the
noncoaxiality of plastic strain rate and backstress in a slightly different manner than the McDowell
a
(1995) model (compare Eqs. 5, 6 and 7). In addition, this model repudiates the term d p 
i of the
f ( i )
OhnoWang (1993) model in favor of dp. As a result, the model overpredicts the ratcheting responses in
biaxial loading (see Figs. 4b to 4d) similar to the Chaboche (1991) model.
Voyiadjis and Basuroychowdhury (1998) attempt to improve the Chaboche (1991) model by
99
Paper IIITo be published in the International Journal of Plasticity (in press)
incorporating the stressrate direction in the kinematic hardening rule. A slightly modified form of the
Voyiadjis and Basuroychowdhury (1998) rule is evaluated in this study. The uniaxial ratcheting
simulations by this rule (Eq. 8) are inadequate, whereas the biaxial ratcheting simulations do not
demonstrate any improvement over the Chaboche (1991, 1994) rule (see Fig. 6).
AbdelKarim and Ohno (2000) proposed a new kinematic hardening rule (Eq. 11) by blending the
ArmstrongFrederick (1966) and the multilinear OhnoWang (1993) rules through a parameter i. As the
parameter i is determined using a biaxial ratcheting response in this study, this rule simulates all the
biaxial responses considered quite well (Figs. 7f to 7f). The new rule, however, underpredicts the uniaxial
ratcheting responses (Fig. 7c). Conversely, if i were determined from a uniaxial ratcheting response,
good simulations would have been obtained for the uniaxial responses (OhnoAbdelKarim, 2000), but
multiaxial ratcheting responses would have been overpredicted. This behavior has been observed with the
OhnoWang (1993) model, where the parameters mi are determined to represent a uniaxial ratcheting
response (Fig. 3; see Bari and Hassan, 2000a for details). In case of the AbdelKarimOhno and the
VoyiadjisBasuroychowdhury models, as the new terms or parameters affect both the uniaxial and
multiaxial responses it is not possible to improve the ratcheting simulations for the whole set of responses
studied.
Using the concept introduced by Delobelle et al. (1995) to appropriately blend the Armstrong
Frederick (1966) and the Burlet and Cailletaud (1986) rules, an improved kinematic hardening rule is
proposed in the framework of the Chaboche (1991) model. This hardening rule includes a parameter '
that does not take part in the plastic modulus calculation, but produces meaningful influence on the yield
surface translation under multiaxial loading. As a result, uniaxial ratcheting simulations by this model are
exactly the same as those by the Chaboche (1991) model, regardless of the value of ' (compare Figs. 9a
and 9b to Figs. 3a and 3b, respectively). When ' is calibrated using a biaxial ratcheting experiment (see
Fig. 8), simulations by this rule manifest improvement for the biaxial ratcheting responses considered
100
Paper IIITo be published in the International Journal of Plasticity (in press)
(see Figs. 9c to 9f). The authors have critically evaluated a number of coupled and uncoupled models
against the same set of uniaxial and biaxial ratcheting responses (see Bari and Hassan, 2000a,b). Among
these models, the proposed kinematic hardening rule performs the best in simulating the whole set of
responses considered. Moreover, the parameter determination scheme of this model is simple and
systematic.
The authors would like to emphasize that the calibration of the new parameter in the modified rule
using a biaxial ratcheting experiment merely tries to compensate for the lack of exactness of the model
introduced by various assumptions. As a result, applicability of the model to multiaxial loading histories
depends strongly on the ratcheting response used for the parameter calibration. Realizing this
characteristic of modeling, Corona et al. (1996) have demonstrated the challenge in simulating multiaxial
ratcheting responses. This study proposes a new kinematic hardening rule to successfully meet this
challenge. Such a success does not, by any means, imply that the new rule is general enough for
simulation of ratcheting in other multiaxial histories or structures. Further validation is needed to explore
its general purpose application. The authors, however, believe that a more generalized ratcheting
simulation model can be achieved through the incorporation of yield surface shape change (formative
Acknowledgment  The financial supports from the Center for Nuclear Power Plant Structures, Equipment
and Piping and the Department of the Army, Army Research Office, Grant No. DAAD199910312 are
gratefully acknowledged.
101
Paper IIITo be published in the International Journal of Plasticity (in press)
NOMENCLATURE
102
Paper IIITo be published in the International Journal of Plasticity (in press)
REFERENCES
AbdelKarim, M., Ohno, N., 2000. Kinematic hardening model suitable for ratcheting with steadystate.
Armstrong, P.J., Frederick, C.O., 1966. A mathematical representation of the multiaxial bauschinger
Bari, S., Hassan. T., 2000a. Anatomy of coupled constitutive models for ratcheting simulation. Int. J.
Bari, S., Hassan. T., 2000b. Kinematic hardening rules in uncoupled modeling for multiaxial ratcheting
Basuroychowdhury, I.N., Voyiadjis, G.Z., 1998. A multiaxial cyclic plasticity model for nonproportional
Burlet, H., Cailletaud, G., 1986. Numerical techniques for cyclic plasticity at variable temperature.
Burlet, H., Cailletaud, G., 1987. Modeling of cyclic plasticity in finite element codes. Proc. of 2nd Int.
Conf. on Constitutive Laws for Engng. Mater.: Theory and Applications, 11571164.
Chaboche, J.L., 1986. Timeindependent constitutive theories for cyclic plasticity. Int. J. Plasticity, 2,
149188.
Chaboche, J.L., 1991. On some modifications of kinematic hardening to improve the description of
Chaboche, J.L., 1994. Modeling of ratchetting: evaluation of various approaches. Eur. J. Mech., A/Solids,
13, 501518.
Corona, E., Hassan, T., Kyriakides, S., 1996. On the performance of kinematic hardening rules in
Dafalias, Y.F., Popov, E.P., 1976. Plastic internal variables formalism of cyclic plasticity. ASME J. Appl.
103
Paper IIITo be published in the International Journal of Plasticity (in press)
Delobelle, P., Robinet, P., Bocher, L,. 1995. Experimental study and phenomenological modelization of
ratcheting under uniaxial and biaxial loading on an austenitic stainless steel. Int. J. Plasticity, 11, 295
330.
Drucker, D.C., 1951. A more fundamental approach to stressstrain relations. Proc. of the First U.S.
Ekh, M., Johansson, A., Thorberntsson, H., Josefson, B.L., 2000. Models for cyclic ratcheting
Geyer, P., 1995. On use of radial evanescence remain term in kinematic hardening. Trans. of 13th Int.
Guionnet, C., 1992. Modeling of ratcheting in biaxial experiments. ASME J. Engng. Mater. Techn., 114,
5662.
Hancell, P.J., Harvey, S.J., 1979. The use of kinematic hardening models in multiaxial cyclic plasticity.
Hassan, T., Kyriakides, S., 1992. Ratcheting in cyclic plasticity, part I: uniaxial behavior. Int. J. Plasticity,
8, 91116.
Hassan, T., Corona, E., Kyriakides, S., 1992. Ratcheting in cyclic plasticity, part II: multiaxial behavior.
Hassan, T., Kyriakides, S., 1994a. Ratcheting of cyclically hardening and softening materials, part I:
Hassan, T., Kyriakides, S., 1994b. Ratcheting of cyclically hardening and softening materials, part II:
Jiang, Y., Sehitoglu, H., 1996a. Modeling of cyclic ratchetting plasticity, part I: development of
104
Paper IIITo be published in the International Journal of Plasticity (in press)
Jiang, Y., Sehitoglu, H., 1996b. Modeling of cyclic ratchetting plasticity, part II: comparison of model
Jiang, Y., Sehitoglu, H., 1999. A model for rolling contact failure. Wear, 224, 3849.
McDowell, D.L., 1995. Stress state dependence of cyclic ratcheting behavior of two rail steels. Int. J.
Mroz, Z., 1967. On the description of anisotropic work hardening. J. Mech. Phys. Solids, 15, 163175.
Ohno, N., AbdelKarim, M., 2000. Uniaxial ratcheting of 316FR steel at room temperaturepart II:
Ohno, N., Wang, J.D., 1993. Kinematic hardening rules with critical state of dynamic recovery, part I:
formulations and basic features for ratcheting behavior. Int. J. Plasticity, 9, 375390.
Phillips, A., Tang, J.L., 1972. The effect of loading path on the yield surface at elevated temperatures.
Phillips, A., Lee, C.W., 1979. Yield surfaces and loading surfaces: experiments and recommendations.
Portier, L., Calloch, S., Marquis, D., Geyer, P., 2000. Ratcheting under tensiontorsion loadings:
Prager, W., 1956. A new method of analyzing stresses and strains in work hardening plastic solids.
Ringsberg, J.W., 2000. Cyclic ratcheting and failure of a pearlitic rail steel, Fatigue Fract. Engng. Mater.
Struct., in press.
Voyiadjis, G.Z., Basuroychowdhury, I.N., 1998. A plasticity model for multiaxial cyclic loading and
105
CHAPTER 5
5.1. INTRODUCTION
It has been argued in the previous chapter that in order to have a generalized
ratcheting simulation model that performs well for various multiaxial loading histories,
the yield surface shape change (formative hardening rule) should be considered in cyclic
plasticity modeling. Phillips and his coworkers [1972,1979,1984] along with many other
researchers (Naghdi et. al [1959], Ivey [1961], Ikegami [1979]) have demonstrated that
the yield surface changes shape with progressive plastic loading. This change in shapes
312 C
III
I II
O
B A 11
IV
D
106
results in yield surfaces that are quite different from the simple vonMises yield surface
as shown in Fig. 5.1. Consequently, the normal directions calculated from the vonMises
yield surface are not always representative of the experimental normal directions
(incremental plastic strain directions) observed in multiaxial loading conditions.
Although it has been demonstrated in this study that an improved representation of
normals to the vonMises yield surface can be achieved by introducing a multiaxial
parameter into the Chaboche [1991] kinematic hardening rule, the general applicability of
this model for various multiaxial loading histories is yet to be validated. It is our
postulation that incorporating the shape change of the yield surface into cyclic plasticity
312 PR
B'
B
A' D'
E'
A D
11
Fig. 5.2. Yield surface deformation law from Phillips and Tang [1972]
modeling would result in a more robust and general ratcheting simulation model.
Development of such a model is attempted in this chapter.
Many researchers like Phillips and Tang [1972], Shiratori et. al [1979], Ortiz and
Popov [1983], Eisenberg and Yen [1981,1984], Ishikawa and Sasaki [1988], Wegener
and Schlegel [1996], Kurtyka and Zyczkowski [1996] have tried to model the deformed
107
shape of the yield surface with progressive plastic loading. Most of these models are
quite complex and numerically extensive and hence are impractical for implementation
with a cyclic plasticity model.
Phillips and Tang [1972] present experimental results of subsequent yield surfaces
of pure aluminum under tensiontorsion loading histories. From these experiments, they
outline a qualitative model for deformed shapes of the yield surface in stress space. They
propose that the subsequent yield surface is the superposition of a rigid body translation
in the prestressing direction and of a deformation in the same direction as illustrated in
Fig. 5.2. The deformation reduces the width of the yield surface in the prestressing
direction in such a way that the motion of the forward part ABD of the yield surface is
less than the rear part AED (see Fig. 5.2). They have not put forward any modeling
scheme to implement this idea.
Eisenberg and Yen [1981] present a theory of multiaxial anisotropic plasticity that
is general enough in its structure to accommodate the hardening mechanism of distorted
yield surface. In their 1984 paper (Eisenberg and Yen [1984]), they present a modeling
scheme to implement the qualitative hardening model proposed by Phillips and Tang
[1972] in an effort to simulate the experimentally observed yield surface deformations.
To describe both deformation and translation of the yield surface in deviatoric stress
space, the subsequent yield state sij (see Fig. 5.3) is described by
s ij a ij + Rij = S ij (5.1)
where Sij is the initial stress state, Rij is a measure of the deformation of the yield surface
and aij is the center of the yield surface. The center of the yield surface is defined as the
intersection point of the loading direction and the hyperplane that divides the yield
surface into its forward and rear subdomains. The initial vonMises surface is
S ij S ij = 0
3 2
2 (5.2)
where 0 is the initial yield surface size. Thus, the yielding condition becomes:
108
( s ij a ij + Rij )( s ij a ij + Rij ) = 0 .
3 2
2 (5.3)
n
a kl ) = S ij , where Tijkl = Tijkl
(n) (n) (m )
Tijkl ( s kl . (5.4)
m =1
It can be observed from Eq. 5.4 that after n loading increments Tijkl assumes 2n different
values for 2n subdomains (forward and rear parts) of the yield surface and the description
s(1)
(0)
0'(aij) s(0)
s
s(1)
b(sij) 0,0'
sijaij ij
0 ij
Rij
a(Sij) Sij
(b)
(a)
Fig. 5.3 A schematic diagram of deformation and translation of yield
surface from Eisenberg and Yen [1984]
of the yield surface becomes quite complex and extensive numerically. Therefore, for
simulation of multiaxial cyclic loading responses, where large numbers of loading
increments are used, this model appears to be less attractive.
Ortiz and Popov [1983] put forward a distorted yield function rule that contains a
state tensor of second order. They propose the diameter J 2 of the elastic region to be a
Fouriercosine series of the angle to the backstress direction. Because it is not possible to
construct an invariant sine of the angle to the backstress direction, the yield function is
109
only applicable to yield conditions that are symmetric to the back stress direction. This
effectively means that this yield function is not suitable for nonproportional loading.
Ishikawa and Sasaki [1988] also makes effort to describe the deformed yield
surface during plastic loading with the use of a fourth order state tensor. For the tension
torsion space, after simplification, their model includes four anisotropic coefficients
(parameters) of the fourth order tensor whose values are to be determined from non
proportional cyclic loading. However, they only demonstrate the suitability of their
model for proportional loading and a very special type of nonproportional loading. Also,
the model lacks guidelines for the selection of effective parameters and their
determination for general multiaxial loading condition.
Wegener and Schlegel [1996] put forward and implement a very extensive
numerical model to describe the distorted yield surface shapes observed by Phillips and
Tang [1972]. After evaluating several yielding functions with second to fourth order state
tensors, they demonstrate success in reproducing the experimentally observed yield
surfaces by using a yielding function that contains a sixth order state tensor. The Ortiz
and Popov [1983] model is a special case of this generalized yield function. Because a
sixth order tensor is used, the number of coefficients to be determined for this rule is
large and their determination requires several yield surface experiments. This model
defines a quality function whose values are to be minimized using the MonteCarlo
simulation or other similar artificial intelligence programs. The variation of different
parameters with plastic loading does not seem to follow any general trend that can be
modeled to use this method in general multiaxial loading condition. Therefore, even
though this model demonstrates success in simulating the experimentally observed yield
surfaces, the numerical scheme is very complex, extensive, and requires many
experimental data for parameter determination. As a result, the model becomes
impractical to implement in a constitutive modeling scheme for simulation of cyclic
responses of materials and structures.
Yoshida et. al [1977] and Shiratori et. al [1979] propose a socalled equiplastic
strain surface as the plastic potential surface in place of the subsequent yield surfaces.
110
The locus of equal plastic strains measured from the unloading point in the stress space is
defined as the equiplasticstrain surface. Figure 5.4 shows a qualitative diagram of the
equiplasticstrain surface. For any loading path OSQiPi, after unloading at S, the loading
2
d p (1)
d p ( 2 )
P1
P2
d p ( n )
Q1 S
Pn Q2
Qn
F0
Fn O 1
path SQi (for i = 1,2,....n) is the elastic loading stage within the subsequent yield surface
where no plastic strains accumulate. The point Qi (for i = 1,2,....n) falls on the initial yield
surface F0. If the accumulated plastic strains
( Pi )
p= d p , i = 1,2,.....n (5.5)
( Qi )
for paths QiPi are equal, then the locus of these equal plastic strain points Pi (for i =
1,2,....n) approximately forms a closed surface f called the equiplasticstrain surface. The
subsequent yield surface Fn touches the equiplasticstrain surface f at loading point Pn
where both the surfaces have the same normal. Therefore, from the normality rule of the
plastic strain increment, the equiplasticstrain surface can be considered as a plastic
111
potential surface and the plastic strain increment can be derived from the principle of
maximum plastic work as follows:
d p = d ( grad f ) . (5.6)
The main advantages of this type of potential surface over its yield surface
counterpart are that (i) the equiplasticstrain surfaces assume much simpler shapes in the
Ilyushin [1954] stress space, and (ii) the numerical modeling of this plastic potential
surface is simpler. It also contains fewer parameters that are to be determined from
experimental responses. Yoshida et. al [1977] and Shiratori et. al [1979] have conducted
several equiplasticstrain surface experiments on Brass, Aluminum alloy, NiCr Mo steel
and Mild steels and show that the equiplasticstrain surfaces match experimentally
observed shapes of the yield surfaces for p 0 ,i.e., a rounded corner in the loading side
and a flattening corner at the rear side. The shape of the subsequent equiplasticstrain
surfaces gradually approach a concentric isotropic hardening hypersphere with increase
of the plastic strains as shown qualitatively in Fig. 5.5. Also, the equiplasticstrain
surfaces generated for loading history OSBC are found to be symmetric with respect to
the preloading direction 1 .
112
ratcheting responses successfully for two types of cyclic loading histories; axial stress
cycles on steady internal pressure similar to Fig. 1.1 and axial stress (or shear stress)
cycles on steady torsion (or tension). However, this model has some limitations,
especially in simulations of the ratcheting responses under general multiaxial loading
conditions as follows.
2
2
d p
1
C B S
O 1
f0
f3 f1
f2
(ii) This model ignores the Bauchinger effect in uniaxial cyclic loading that is
critical for uniaxial ratcheting responses.
(iii) It assumes that line about which the equiplasticstrain surfaces are
symmetric always passes through the origin in Ilyushin stress space. In all the
experiments presented by Shiratori et. al [1979], the unloading occurs after sufficient
113
hardening that takes the unloading point very close to the isotropic surface (see Fig. 4 and
Fig. 18 in the reference). As a result, the symmetry lines appear to pass through the origin
in all these experiments. However, in case of an unloading point where the equiplastic
strain surface is much away from the isotropic surface, this symmetry line may not pass
through the origin. For example, if the unloading occurs at points S or C in Fig. 5.5, the
symmetry line is expected to pass through or in the vicinity of the origin, as the normals
at those points also pass through or in the vicinity of the origin. However, if unloading
starts at some point like B, the assumed symmetry line is much different from the normal
or the incremental plastic strain direction at B. Therefore, at unloading point B, the
normal directions to the yield surface and to the equiplasticstrain surface would be
different and the basic premise of using the equiplasticstrain surface as a plastic
potential surface would be violated.
(iv) The methodology for the determination of plastic modulus and other
parameters presented in Shiratori et. al [1979] can not be applied in combined stress
strain controlled loading conditions.
Regardless of these limitations, the Shiratori et. al [1979] model shows promise
and seems to be simpler in taking into account the effect of the distorted yield surfaces in
cyclic plasticity modeling. Therefore, in this study, efforts have been made to present a
constitutive model that includes the equiplasticstrain surface and try to address the
above mentioned limitations in order to improve the simulations of ratcheting responses
in general multiaxial loading. Two types of equiplasticstrain surface models are studied
within the uncoupled modeling scheme of Dafalias and Popov [1976]. Performances of
these models are evaluated against a set of constant pressure biaxial ratcheting
experiments (Fig 1.1b). These modeling schemes and numerical results are presented
below.
114
5.2. PLASTIC POTENTIAL THEORY BASED ON THE EQUI
PLASTICSTRAIN SURFACE
1 = 32 S11 , 2 = 2
3
( S11 + 2 S 22 ), 3 = 3S12 , 4 = 3S 23 , 5 = 3S 32 ; (5.7)
1 = 11 , 2 = ( 11 + 2 22 ) , 3 = 12 , 4 = 23 , 5 = 31
p p p 1 p p p 2 p p 2 p p 2 p
3 3 3 3
(5.8)
where S ij and ij are deviatoric stress and plastic strain tensor components in the three
p
dimensional Cartesian coordinate system respectively. In the Ilyushin space, the stress
and the plastic strain vectors are given as
5 5
= i e i and = i ei
p p
(5.9)
i =1 i =1
1
d p
= d n G = ( nG d )nG (5.10)
H
f
where nG = , f is the equiplasticstrain surface and H is the plastic modulus.
f
115
5.2.3. Model I
Although both Yoshida et. al [1977] and Shiratori et. al [1979] propose two
similar type of expressions of the equiplasticstrain surface in the moving stress space
i of the fixed space i , they do not provide any basis for their selection. On the other
hand, Kurtyka and Zyczkowski [1996] present a general form for defining any
anisotropically deformed surface in the Ilyushin vector space. In this study, a special case
of this general form is adopted to describe the shape of the equiplasticstrain surface.
With the assumption that the equiplasticstrain surface is symmetric with respect to a
direction PR , the expression for the equiplasticstrain surface is proposed in this study
as follows:
( 1 0 *) 2 5
( k vY ) 2
+ =1 (5.11)
R1 + 2d 1 ( 1 0 *) d1
2 2 2
k =2 Rk
where R1 = ( N M ) / 2 , Rk = B C * , d 1 = N 0 * R1 .
All the parameters and variables in Eq. 5.11 are shown in Fig. 5.6. As adopted by
Shiratori et. al [1979], the expressions for 0 * and C * are assumed to depend on the
after unloading. It should be noted that the equiplasticstrain surface in this model
gradually approaches a bounding surface for very large * .
It has been demonstrated in this study (chapter 3) that the plastic modulus H
calculation scheme by Dafalias and Popov [1976] is successful in simulating the
116
k (k = 2,...,5)
k
C*
vb
nG
0*
b Rk
R1 PR
d1
1
vy b
1
q
M
p
N
Bounding surface
ratcheting responses when used with appropriate hardening rule. Therefore, this scheme
is also adopted in this study to calculate the plastic modulus as follows:
a
H = E0 + h( ) , h=
p
(5.13)
in
1 + b( in ) m
2 B
is the plastic modulus of the bounding surface, B is the image point of on the
bounding surface and b is the bounding surface center. The parameters a, b and m are
determined from a uniaxial ratcheting test as described in the paper II.
117
(c) Hardening rule
It has been observed in experiments from Shiratori et. al [1979] that p and q
decrease with the increase in plastic strain value. With them, 0 * and C * also
decrease according to Eq. 5.12 and the equiplasticstrain surface gradually approaches
the bounding surface. When p and q become zero, the equiplasticstrain surface
converges with the bounding surface. This hardening mechanism of the equiplastic
strain surface is modeled in this study as follows:
It should be observed in Eq. 5.14 that the bounding surface does not grow in size,
but translates along the negative PR direction. The magnitude of the scalar K is
determined from the consistency condition at each loading increment.
The experimental data from Shiratori et. al [1979] shows that the equiplastic
strain surfaces appear to be symmetric with respect to a line that passes through the origin
when the unloading points are very close to the bounding surface. However, there is no
experimental data available showing the shapes of the equiplasticstrain surface when
unloading occurs far away from the bounding surface. In order to hold the basic premise
that the equiplasticstrain surface can be used as a plastic potential surface, the line of
symmetry for the equiplasticstrain surface has to be along the plastic strain increment
(or normal to the equiplasticstrain surface) at the unloading stress point. Therefore, in
this study, it is assumed that
PR = nG . (5.15)
118
carbon steels where cyclic axial strains are applied on straight pipe with steady internal
pressure (see loading history (b) in Fig. 1.1). The material constants C1, C2, m1 and m2
that define the characteristic values for 0 * and C * are supposed to be determined
from equiplasticsurface experiments in multiaxial loading condition according to
Shiratori et. al [1979]. As there is no experimental data available on the shape of the equi
plasticstrain surface in stabilized carbon steels, these parameters are selected from a
constant pressure biaxial ratcheting experiment in this study. Fig. 5.7 shows the
simulation of ratcheting responses by this model along with the experimental responses
under biaxial loading condition. The parameters used are:
4
p
CS 1026
=====m = 9.65 ksi
(2)
1.5
(1)
1
(2)
0.5
(1)
0
0 10 20 30 40
N
Fig. 5.7. Biaxial ratcheting predictions by model I and experiments from Hassan
and Kyriakides [1992].
The simulated accumulation of the hoop strains from this model tends to stabilize after
some ratcheting during a couple of initial cycles. The predicted ratcheting responses for
119
both axial strain amplitudes with the same internal pressure behave in similar manner.
This deviation of ratcheting responses from experimental responses can be attributed to
the assumption that the line of symmetry is along the plastic strain increment at the
unloading point (Eq. 5.15). With this assumption, this symmetry direction gradually
flattens until it becomes parallel to the 1 direction when the hoop strain component of
plastic strain increment becomes zero. As a result, shakedown of ratcheting responses
occurs.
5.2.4. Model II
Although it is argued in model I that the symmetry direction may not necessarily
pass through the origin of the Ilyushin vector stress space in all loading histories, it has
been observed in the experiments by Shiratori et. al [1979] that the equiplasticstrain
surfaces tend be symmetric with respect to a line passing through the vicinity of the
origin. It should be noted that, for the ratcheting experiments considered in this study to
evaluate models I and II, enough plastic strains are accumulated in each cycle to bring the
unloading points sufficiently close to the bounding surface. For these experiments, the
direction of the plastic strain increments (normals to the equiplasticstrain surface) at the
unloading points also tend to pass through the origin. Therefore, in model II, the
symmetry direction is considered to be the direction of the line passing through the origin
and the unloading point (OS direction in Fig. 5.8).
120
k (k= 2,...,5)
DafaliasPopov k
bounding surface C *
nG
Rk
*
0
a PR
R1
F = B + b b d1 1
S
0
1
q
M
p
N
Isotropic bounding
surface
( 1 0 *) 2 5
( k ) 2
+ =1 (5.16)
R1 + 2d1 ( 1 0 *) d 1
2 2 2
k =2 Rk
where R1 = ( N M ) / 2 , Rk = F C * , d 1 = N 0 * R1 , F = B + b .
All the parameters and variables in Eq. 5.16 are illustrated in Fig. 5.8. The new parameter
b is the increase in size of the isotropic bounding surface during plastic loading from its
original size b . The expressions for 0 * and C * are assumed to depend on the
121
where p = F N , q = M + F and * is the accumulated plastic strain after
unloading.
The Dafalias and Popov [1976] plastic modulus calculation scheme (Eq. 5.13) is
also adopted in this model. However, an additional DafaliasPopov type kinematic
bounding surface is introduced in this model as shown in Fig. 5.8 where, b is the center
of the kinematic bounding surface. The image point of on the kinematic bounding
surface is used as a measure of in Eq. 5.13. The same parameter determination scheme
d M = K q + dvb , d N = K p + dvb , d b = 0 dp .
p
(5.18)
The kinematic bounding surface center b grows according to the evolution model
proposed by SeyedRanjbari [1986] (also Hassan and Kyriakides [1994]) where
a = 0 nG in this model. The readers are referred to Eqs. 6 and 7 of the enclosed
paper II for expressions of this evolution. The magnitude of the scalar K in Eq. 5.18 is
determined from the consistency condition at each loading increment.
The performance of the modeling scheme II is also evaluated against the same set
of biaxial ratcheting responses as used in evaluating model I. The same guidelines to
determine material constants C1, C2, m1 and m2 as adopted in model I are applied in this
122
4
p
CS 1026
=====m = 9.65 ksi
3.5 x (1) xc = 0.40 %
(2) xc = 0.50 %
3 m (3) xc = 0.65 %
2.5 x
Experiment
Model II
2
(2)
1.5 (3)
(1)
1
0.5
0
0 10 20 30 40
N
Fig. 5.9. Biaxial ratcheting predictions by model II and experiments from Hassan
and Kyriakides [1992].
model. Fig. 5.9 shows the simulation of ratcheting responses by model II with
parameters:
The simulated ratcheting responses for all axial strain amplitudes show an almost
constant rate of ratcheting. For initial cycles, ratcheting simulations match the
experimental responses reasonably, but fail to represent the ratcheting rates observed in
the experiments for higher cycles (see Fig. 5.9). In addition, the changes in the ratcheting
rates observed with different axial strain amplitudes in the experiments of Fig. 5.9 are not
reproduced by model II.
123
5.3. CONCLUSION AND DISCUSSIONS
From the experimental trends observed for the shapes of the yield surface, Phillips
and Tang [1972] put forward a qualitative model of the translation and distortion of the
yield surfaces during plastic loading. Ortiz and Popov [1983], Eisenberg and Yen
[1981,1984], Ishikawa and Sasaki [1988], and Wegener and Schlegel [1996] propose
analytical and numerical modeling schemes to implement the yield surface shape change
observations of Phillips and Tang [1972]. After careful scrutiny of these modeling
schemes, it has been observed that most of these models use complex and numerically
extensive schemes with higher order state tensors. As a result, these models become less
attractive (in some cases, impractical) for implementation with a cyclic plasticity model.
Shiratori and his coworkers (Yoshida et. al [1977], Shiratori et. al [1979]), on the
other hand, propose a socalled equiplasticstrain surface that can be used as the plastic
potential surface instead of the yield surface generally used in cyclic plasticity modeling.
This equiplasticstrain surface can effectively incorporate the effect of distorted yield
surface in the flow rule as the yield surface and the equiplasticstrain surface have the
124
same normal direction at the current stress point. Experimental results of Yoshida et. al
[1977] and Shiratori et. al [1979] also demonstrate the basis behind the applicability of
using the equiplasticstrain surface as the plastic potential surface for incremental plastic
strains. Moreover, the formulation for the equiplasticstrain surface is simpler compared
to other deformed yield surface formulations.
125
CHAPTER 6
6.1. CONCLUSIONS
All of the enclosed journal papers (papers I, II and III) written from this study
contain conclusions from the corresponding part of the study. Conclusions drawn from
study of constitutive modeling with anisotropic deformation of yield surface are
presented in chapter 5. A summary of all these conclusions is presented in the following.
For uniaxial stresscontrolled history (Fig. 1.1a), the linear kinematic hardening
[Prager, 1956] and multilinear models produce closed hysteresis loops and, hence, cannot
simulate the ratcheting response. The Prager model overpredicts ratcheting strains during
the initial cycles, which are followed by shakedown after a few more cycles for all biaxial
loading cases.
The original Chaboche [1986] model with three or four decomposed rules, with at
least one rule as linear hardening, improves the ratcheting simulation for the initial
cycles, but always stabilizes to shakedown with persistent cycling. Incorporation of a
slight nonlinearity into the linear rule improves the model's capability to simulate steady
rate of ratcheting and prevents shakedown.
Chaboche [1991] model introduces the concept of a threshold backstress into the
fourth nonlinear kinematic hardening rule. This model incorporates a linear segment in
the uniaxial hysteresis curve within the threshold level and, thus, makes the hysteresis
126
curve stiffer and reduces the rate of uniaxial ratcheting compared to the models without
any threshold. However, this model overpredicts all of the biaxial ratcheting responses
considered.
The Guionnet [1992] model grossly overpredicts the ratcheting responses for
uniaxial loading cases. However, this model performs impressively in simulating the
circumferential strain ratcheting for the biaxial experiments with constant pressure. When
the loading changes to the biaxial bowtie histories, this model fails to simulate the
ratcheting responses.
The limitations introduced by the idealized yield surface shape and other
simplified assumptions in the coupled cyclic plasticity modeling can be compensated if
some model parameters are calibrated using multiaxial ratcheting experiments or if the
hardening rule is uncoupled from the plastic modulus.
This study evaluates eight different kinematic hardening rules in order to identify
or develop a kinematic hardening rule that works well with the DafaliasPopov uncoupled
model for both uniaxial and multiaxial ratcheting simulation.
The original Armstrong and Frederick [1966] rule performs very well for all the
constant pressure biaxial ratcheting responses, but fails for the ratcheting responses in
bowtie loading histories.
127
[1972,1979] and TsengLee [1983] rules. None of these three rules perform satisfactorily
for the set of biaxial ratcheting experiments considered.
The Kaneko [1981,1984] rule is a modification over the Ziegler [1959] rule. This
rule predicts negative ratcheting for the biaxial history of Fig. 1.1b due to the addition of
the prestress path dependence.
The Xia and Ellyin [1994,1997] rule overpredicts all of the biaxial ratcheting
responses due to the influence of the Mroz [1967] rule.
The Chaboche [1991] kinematic hardening rule, with three decomposed terms,
performs the best for the whole set of ratcheting responses considered. This rule also
captures the transient ratcheting rates at initial cycles and the constant ratcheting rates at
higher cycles as observed in material ratcheting responses of stabilized carbon steels.
Jiang and Sehitoglu [1996] modify the exponents of the OhnoWang model to
depend on the noncoaxiality of plastic strain rate and backstress in a slightly different
manner than the McDowell model. In addition, this model uses the equivalent plastic
128
strain increments in the hardening rules. As a result, the model overpredicts the ratcheting
responses in biaxial loading similar to the Chaboche [1991] model.
The improved coupled model proposed in this study by appropriately blending the
ArmstrongFrederick [1966] and the Burlet and Cailletaud [1986] rules in the framework
of the Chaboche [1991] model manifest significant improvement in simulating ratcheting
responses for all of the biaxial loading cases considered. The study has critically
evaluated a number of coupled and uncoupled models against the same set of uniaxial
and biaxial ratcheting responses. Among these models, the proposed kinematic hardening
rule performs the best in simulating the whole set of responses considered. Moreover, the
parameter determination scheme of this model is simple and systematic.
Most of the models that try to formulate the anisotropic deformation of the yield
surface use complex and numerically extensive schemes with higher order state tensors.
As a result, these models become less attractive (in some cases, impractical) for
implementation in a cyclic plasticity model.
Shiratori and his coworkers (Yoshida et. al [1977], Shiratori et. al [1979]), on the
other hand, propose a socalled equiplasticstrain surface that can be used as the plastic
potential surface. The formulation for the equiplasticstrain surface is simpler compared
to other deformed yield surface formulations.
129
This study demonstrates the basic methodology and promise in incorporating the
equiplasticstrain surface as the plastic potential surface into the cyclic plasticity
modeling. More experimental results on the shapes of the equiplasticstrain surface,
especially when unloading occurs far away from the bounding surfaces are essential in
order to have a representative modeling scheme.
(i) The single multiaxiality parameter introduced in the proposed coupled model
of this study is assumed to be constant for all loading conditions. However, this
parameter should include the effect of plastic loading history in order to predict
ratcheting response in general multiaxial loading cases. Studies can be conducted to
evaluate the performance of the proposed model in general multiaxial loading conditions
by identifying the effect of loading histories on this parameter.
(ii) In an effort to incorporate the yield surface shape changes in cyclic plasticity
modeling, this study demonstrates the prospect of employing the equiplasticstrain
surface as the plastic potential surface. Extensive experimental studies should be
conducted to observe the shapes of the equiplasticstrain surfaces in general multiaxial
loading conditions.
130
REFERENCE
AbdelKarim, M., Ohno, N., (2000). Kinematic Hardening Model Suitable for Ratcheting
with Steadystate. International Journal of Plasticity, 16, 225240.
Besseling, J.F. (1958) A Theory of Elastic, Plastic and Creep Deformations of an Initially
Isotropic Material. Journal of Applied Mechanics, Vol 25, pp. 529536.
Burlet, H., Cailletaud, G., (1987). Modeling of Cyclic Plasticity in Finite Element Codes.
Proc. of 2nd Int. Conf. on Constitutive Laws for Engineering Materials: Theory and
Applications, pp. 11571164.
Dafalias, Y.F. and Popov, E.P. (1976) Plastic Internal Variables Formalism of Cyclic
Plasticity. Journal of Applied Mechanics, Vol 43, pp. 645650.
Delobelle, P., Robinet, P., Bocher, L,. (1995). Experimental study and phenomenological
modelization of ratcheting under uniaxial and biaxial loading on an austenitic
stainless steel. International Journal of Plasticity, 11, 295330.
Hassan, T., Corona, E. and Kyriakides, S. (1992) Ratcheting in Cyclic Plasticity, Part II:
Multiaxial Behavior. International Journal of Plasticity, Vol 8, p. 117146.
Ilyushin, A.A. (1954) On the Relation Between Stresses and Small Strains in the
Mechanics of Continua (in Russian). Prikl. Mat. Mech., Vol 18, pp. 641666.
Ishikawa, H. and Sasaki, K. (1988) Yield Surfaces of SUS 304 Under Cyclic Loading.
Journal of Engineering Materials and Technology, Vol 110, pp. 365371.
Ivey, H.J. (1961) Plastic Stress Strain Relations and Yield Surfaces for Aluminum
Alloys. Journal of the Mechanical Engineering Science, Vol 3, pp. 1531.
Jiang, Y., Sehitoglu, H., (1996) Modeling of cyclic ratchetting plasticity, part I:
development of constitutive relations. ASME J. App. Mech., 63, 720725.
132
Anisotropic Metals After Various Preloading. Bulletin of the JSME, Vol 27, pp.
26872693.
McDowell, D.L. (1995) Stress State Dependence of Cyclic Ratcheting Behavior of Two
Rail Steels. International Journal of Plasticity, Vol 11, pp. 397 421.
Naghdi, P.M., Essenburg, F. and Koff, W. (1958) An Experimental Study of Initial and
Subsequent Yield Surfaces in Plasticity. Journal of Applied Mechanics, Vol 25, pp.
201209.
Ohno, N. and Wang, J.D. (1993) Kinematic Hardening Rules with Critical State of
Dynamic Recovery, Part I: Formulations and Basic Features for Ratcheting
Behavior. International Journal of Plasticity, Vol 9, pp. 375390.
Ortiz, M. and Popov, E.P. (1983) Distortional Hardening Rules for Metal Plasticity.
ASCE Journal of Engineering Mechanics, Vol 109, pp. 10421057.
Phillips, A. and Lu, W.Y. (1984) An Experimental Investigation of Yield Surfaces And
Loading Surfaces of Pure Aluminum With StressControlled And StrainControlled
Paths of Loading. Journal of Engineering Materials and Technology, Trans. ASME,
Vol 106, pp. 349354.
Phillips, A., Lee, C.W., (1979) Yield surfaces and loading surfaces: experiments and
recommendations. International Journal of Solids and Structures, Vol 15, pp. 715
729.
Phillips, A., Tang, J.L., (1972) The effect of loading path on the yield surface at elevated
temperatures. International Journal of Solids and Structures, Vol 8, pp. 463474.
133
Prager, W. (1956) A New Method of Analyzing Stresses and Strains in Work Hardening
Plastic Solids. Journal of Applied Mechanics, Vol 23, pp. 493496.
Tseng, N.T. and Lee, G.C. (1983) Simple Plasticity Model of the Two Surface Type.
ASCE Journal of Engineering Mechanics, Vol 109, pp. 795810.
Voyiadjis, G.Z. and Sivakumar, S.M. (1991) A Robust Kinematic Hardening Rule for
Cyclic Plasticity with Ratcheting Effects, Part I: Theoretical Formulation. Acta
Mechanica, Vol 90, pp. 105123.
Voyiadjis, G.Z. and Sivakumar, S.M. (1994) A Robust Kinematic Hardening Rule for
Cyclic Plasticity with Ratcheting Effects, Part II: Application to Nonproportional
Loading Cases. Acta Mechanica, Vol 107, pp. 117136.
Voyiadjis, G.Z., Basuroychowdhury, I.N., (1998) A plasticity model for multiaxial cyclic
loading and ratchetting. Acta Mechanica., 126, 1935.
Xia, Z. and Ellyin, F. (1994) Biaxial Ratcheting Under Strain or StressControlled Axial
Cycling With Constant Hoop Stress. Journal of Applied Mechanics, Vol 61, pp. 422
428.
Xia, Z. and Ellyin, F. (1997) A Constitutive Model with Capability to Simulate Complex
Multiaxial Ratcheting Behaviour of Materials. International Journal of Plasticity,
134
Vol 13, pp. 127142.
Yoshida, F., Murakami, T., Kaneko, K., Ikegami, K. and Shiratori, E. (1977) Plastic
Behaviour in the Combined Loading along Straight Stress Paths With a Bend.
Bulletin of the JSME, Vol 20, pp. 15491556.
135
Bien plus que des documents.
Découvrez tout ce que Scribd a à offrir, dont les livres et les livres audio des principaux éditeurs.
Annulez à tout moment.