Sunteți pe pagina 1din 52

International Journal of Plasticity 24 (2008) 16421693

Contents lists available at ScienceDirect

International Journal of Plasticity


journal homepage: www.elsevier.com/locate/ijplas

A review of some plasticity and viscoplasticity


constitutive theories
J.L. Chaboche
ONERA DMSM, 29 Avenue de la Division Leclerc, BP72 F-92322 Chtillon Cedex, France
University of Technology of Troyes, LASMIS, 12 rue Marie Curie, 10010 Troyes, France

a r t i c l e

i n f o

Article history:
Received 1 October 2007
Received in nal revised form
14 March 2008
Available online 14 April 2008

Keywords:
Continuum mechanics
Plasticity
Viscoplasticity
Strain hardening
Ratchetting

a b s t r a c t
The purpose of the present review article is twofold:

 recall elementary notions as well as the main ingredients and


assumptions of developing macroscopic inelastic constitutive
equations, mainly for metals and low strain cyclic conditions.
The explicit models considered have been essentially developed
by the author and co-workers, along the past 30 years;
 summarize and discuss a certain number of alternative theoretical frameworks, with some comparisons made with the previous
ones, including more recent developments that offer potential
new capabilities.

2008 Elsevier Ltd. All rights reserved.

1. Introduction
The constitutive equation of the material is an essential ingredient of any structural calculation. It
provides the indispensable relation between the strains and the stresses, which is a linear relation in
the case of elastic analyses (Hookes law) and a much more complex nonlinear relation in inelastic
analyses, involving time and additional internal variables.
In this paper we limit ourselves to considering the conventional Continuum approach, i.e. that
the Representative Volume Element (RVE) of material is considered as subject to a near-uniform macroscopic stress. This Continuum assumption is equivalent to neglecting the local heterogeneity of the
stresses and strains within the RVE, working with averaged quantities, as the effects of the heterogeneities act only indirectly through a certain number of internal variables. Moreover, in the framework of the local state assumption of Continuum Thermomechanics, it is considered that the

E-mail address: Jean-Louis.Chaboche@onera.fr


0749-6419/$ - see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijplas.2008.03.009

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1643

state of a material point (and of its immediate vicinity in the sense of the RVE) is independent of that
of the neighboring material point. Therefore the stress strain gradients do not enter into the
constitutive equations. This assumption is obviously questioned in recent theories of Generalized Continuum Mechanics, that are not addressed here.
The entire presentation will be limited to quasi-static movements considered to be slow enough, in
the framework of small perturbations (small strains of less than 10%, for example). Also, the equations
indicated will be formulated without explicitly stating the effect of temperature (although this may be
very large in certain cases). In other words, in accordance with the common practice for determining
the constitutive equations of solid materials, we will assume the temperature is constant (and uniform
over the RVE). The effect of the temperature will come into play only by the change of the material
parameters dening the constitutive equations. Moreover, the above mentioned Continuum Thermodynamic framework will not be considered in detail. Only a few remarks are made as consequences of
such a theoretical framework for the temperature rate effect in the hardening rules.
The presentation is more directly oriented toward metallic type materials with elasto-plastic or
elasto-viscoplastic properties even though, in a way, viscoelasticity, i.e. the effect of viscosity on elasticity, could be modeled from a viscoplastic model. Among the effects considered, we will thus have:
irreversible strain, or plastic strain, the associated phenomena of strain hardening, the time effects,
whether they enter by the effect of the loading velocity or through slow time variations of the various
variables (static recovery, for example). Aging phenomena (associated with possible changes in the
metallurgical structure) and damage effects will be mentioned only briey. The anelasticity of the
metals (very low viscous hysteresis in the elastic range), which corresponds to reversible motions
of the dislocations, will not be discussed either. Only initially isotropic materials are considered, in
which anisotropy is the result of plastic ow and associated hardening processes.
In the present paper, the presentation of constitutive equations is made by following an increasing
order of complexity. It can essentially be considered in two parts:
 half the paper addresses to readers who are not too much informed about the plasticity/viscoplasticity framework. It is more or less an introduction to unied viscoplastic constitutive models,
mainly based on the works made around the author;
 the second part considers more elaborated aspects, reviewing some other unied viscoplastic constitutive theories, pointing out some similarities and differences. Other constitutive frameworks are
also discussed. The present capabilities of the various kinematic hardening models are compared in
the context of predicting ratchetting effects, including modied ArmstrongFrederick based rules as
well as multi-surface and two-surface theories.
A special mention here about the ArmstrongFrederick Report (Armstrong and Frederick, 1966)
that serves of common basis for many kinematic hardening rules. This work was never published, only
available as a Technical Report from CEGB (Central Electricity Generating Board). By using this rule in
the context of unied viscoplasticity and generalising it continuously, the author contributed to the
knowledge and citation of this report. In 2007, it has been published in Materials at High Temperature, accompanied with a Preface retracing this story (Frederick and Armstrong, 2007).
Let us point out that the review of existing modelling methodologies in the context of cyclic plasticity and viscoplasticity cannot be at all exhaustive. We hope only to provide the indispensable general elements, as well as the main types of modelling. The interested reader should refer to more
complete specialized works (Lematre and Chaboche, 1985; Khan and Huang, 1995; Francois et al.,
1991; Miller, 1987b; Krauss and Krauss, 1996; Besson et al., 2001).
2. Basic notions
The general context of modelling the inelastic behaviour in rate-independent plasticity or in viscoplasticity is supposed to be known, as being sufciently standard. Many more details and interesting
exercises on this current and standard framework can be found in textbooks, like in Khan and Huang
(1995). Only the main assumptions and equations are indicated and briey commented here, as they
could be necessary for understanding further developments in the paper.

1644

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

We assume the small strain framework. This is justied by the domain of application to cyclic loading conditions. The main equations are given below, considering also isothermal conditions. The rst
equation denes the partition of total strain tensor into an elastic strain and a plastic strain, though
the second one gives corresponding Hookes Law of linear elasticity.

e e e e p

r L : e ep

X
k k60
f kr

 H

e_ p


of
k_
k_ n

or


An aside on the notations: the symbol . between two tensors designates the product contracted once
((rik rkj r2ij with Einsteins summation, represents the square of the tensor r
); the symbol : desig
2
nates the product contracted twice (for example the scalar rij rji Tr r
).

In the framework of rate-independent plasticity, we need the use of an elasticity domain, f 6 0, as
given by (3). The yield surface f 0 is dened in (3) with Hills criterion, using a fourth rank tensor H

within a quadratic norm denition as

kr
k
 H

r
r : H
:r



More sophisticated yield surface or loading surface denitions could be used. Examples can be found
in recent works by Cazacu and Barlat (2004) or Barlat et al. (2007), but will not be considered in the
present review.
In Eq. (3) parameter k is the initial yield surface size. Moreover, hardening induced by plastic ow is
assumed to be described by a combination of kinematic hardening and isotropic hardening. We use
the back-stress X
for kinematic hardening and the increase of yield surface size R for the isotropic

hardening.
Figs. 1 and 2 illustrate, in the deviatoric stress plane and in the uniaxial tensioncompression particular case, the transformation of the elastic domain and yield surface by the two particular cases of
pure isotropic hardening and pure linear kinematic hardening.
In what follows we also assume the associated plasticity framework (the ow potential is identical
with the yield surface) and the normality law (4) expresses the consequence of the maximum dissipation principle. In the rate-independent framework, the plastic multiplier k_ is determined by the consistency condition f f_ 0.
In case of a viscoplastic behaviour (or rate dependency), the above plasticity framework is generalized by using a viscoplastic potential Xf . The stress state goes beyond the elasticity domain with a

Fig. 1. Schematics of the isotropic hardening. Left: in the deviatoric plane; right: the stress vs plastic strain response.

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1645

Fig. 2. Schematics of the linear kinematic hardening. Left: in the deviatoric plane; right: the stress vs plastic strain response.

positive value of rv f > 0, that can be called the viscous stress, or the overstress. In that case, normality rule reads:

e_ p


oXf oX of
of

p_
p_ n

or
o
o
of
r
r



_ the norm of the viscoplastic strain rate, as dened by


k_ is replaced by p,

p_ ke_ p kH1

Therefore, p is the length of the plastic strain path in the plastic strain space.
Let us conclude this brief introduction of the general framework by indicating the particular case
where orthotropic Hills criterion is restricted to von Mises one, with



3 d 3
1
I
I  1

1
2
2
3 

where I and I d are respectively the fourth rank unit tensor and deviatoric projector. In such case, von
Mises elastic domain is given by

s
3 0
0 : r0  X0  R  k 6 0
f kr
X
kRk
r  X





2 

0
0
0
where r
and X
are deviatoric parts, like r
r
 13 Tr r 1
. Correspondingly, the direction of the plastic





strain rate is

e_ p


r0  X 0
oX
3 
_

p
p_ n

2 kr
or
X
k




10

The accumulated plastic strain rate then writes

s
2 p p
e_ : e_
p_
3 

11

Let us note that, the yield surface being independent on the rst stress invariant, plastic ow does not
induce a volume change (Tr e_ p 0, n
:n
3=2). Moreover, any stress state can be broken down into


_ is deduced by inversion of the relation p_ oX=of .
the following form, in which the function rv p

_ n
r X R k rv p


12

3. Unied theory of viscoplasticity


To simplify the discussion, we adopt the viscoplasticity scheme directly. The case of rate-independent plasticity will be deduced from this as a limiting case. We begin by giving a rather general form to

1646

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

the constitutive equations, and then we examine the most common particular options, for the viscosity function and for the isotropic and kinematic hardening. The main ingredients in the theory are taken from the unied constitutive model of the author (Chaboche, 1977b). Various other versions will
be discussed in Section 5. We then examine the case of rate-independent plasticity and nish with a
few indications on determining the parameters of the equations from experiments.
3.1. General form of the constitutive equation
Let us point out right away that this can be established in the general formal framework of continuum thermodynamics. This subject will not be addressed here. The interested reader can refer to Germain (1973), Halphen and Nguyen (1975), Chaboche (1996), for example.
The expression for the viscoplastic constitutive equation essentially includes two aspects:
 the choice of the viscosity function (see Section 3.2), or choice of the viscoplastic potential X, which
will act in the expression for the viscoplastic strain rate (its dependency on the viscous stress)
through the normality Eq. (6) stated above;
 the choice of the hardening equations for all of the internal variables. These are provisionally denoted
aj j 1; 2; . . . ; N, which can be scalar or tensorial. The general form includes a strain hardening
term, a dynamic recovery term, and a static recovery term:

a_ j hj   _ep  r Dj   aj e_ p  r Sj . . .aj

13

The rst term gives an (increasing) evolution of aj with the plastic strain. The second, on the other
hand, gives a recall, or evanescent memory effect; but this acts again (instantaneously) with the plastic strain, whence the dynamic recovery term. The third term is called static recovery or time recovery,
or thermal recovery, since it can act independently of any plastic strain. This is very clear in an incremental statement such that da h dep  r D a dep  r S a dt. The functions hj ; rDj ; r Sj are to be dened (see
below). Let us note right away that the static recovery mechanism is thermally activated and that
the effect of the temperature in the function r Sj plays an essential role. Roughly speaking, this terms
is used to express the effects of the thermal agitation, inducing dislocation climbing mechanisms
and the corresponding annihilation possibility, or even recrystallization effects in certain cases. Let
us also indicate a strong analogy with equations of physical origin in Garofalo (1965), Kocks (1976),
Estrin and Mecking (1984), concerning the dislocation density q, for example according to Estrin
(1996) in uniaxial loading:

p
p
dq Mko k1 q  k2 q dep  r S q; T dt

14

3.2. Choice of the viscosity function


This relation between the viscous stress and the plastic strain rate norm is usually highly nonlinear.
Thus, through a large range of velocities, it can be approximated by a power function:

p_

 n D En
f
rv

D
D

15

The McCauley brackets hi are used here to ensure that when f < 0, i.e. inside the elastic domain, p_ cancels out continuously. This expression corresponds to Nortons equation (or Odqvists law in threedimensional context) for the secondary creep, when the hardening is neglected. Exponent n depends
on the material, on the strain rate domain considered, and on the temperature, ranging from a theoretical value of n 1 for the diffusional creep of a perfect alloy to sometimes very high values when
we approach the materials low viscosity range (at low temperatures). In practice, it is usually observed that 3 6 n 6 30 for current engineering materials.
The advantage of expression (15) is that it easily derives from the viscoplastic potential:

D Drv En1
n1 D

16

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1647

For certain materials, an effect of saturation of the rate effect can be felt in the high rate regime. Fig. 3
shows the example of 316L stainless steel at 550 C. The intermediate velocity range, where the relation between log10 rv and log10 e_ p appears to be approximately linear with a slope of n 24, extends to
low rates by a rapid drop in the stress (due to static recovery phenomena that will be studied further
on) and by a stress saturation at high velocities between 103 and 101 s1 . Various expressions may
be proposed to express such a saturation effect in the viscosity function. They are studied and compared in Section 5.7.
3.3. Isotropic hardening equations
When we consider the expression for the norm of the strain rate (15), by replacing f with (9), we
nd:

p_

*k r  X k  R  k+n


D

17

We nd three possibilities for introducing a hardening of the isotropic type:


(i) through the variable R, by an increase in the size of the elasticity domain,
(ii) by increase of the drag stress D,
(iii) by coupling with the evolution law of the kinematic hardening variable X
.

In the rst two cases, the only ones considered here, we just have to dene the one-to-one relationship between R (or D) and the state variable of the isotropic hardening, which is the accumulated plastic strain p (or possibly the accumulated plastic work W p ).

R Rp

D Dp

18

One possibility among others is to let the two evolutions be proportional. We can then dene only
the function Rp and deduce from it

Dp K fRp

19

where K is the initial value of the drag stress and f is a weighting parameter. One special case, corresponding to the Perzyna (1964) approach, is the one obtained with K k and f 1.
By decomposition of the equivalent von Mises stress (in the case without kinematic hardening,
X
0), we can note the different roles of the two types of isotropic hardening


_ p k Rp Dpp_ 1=n
req k Rp rv p;

20

Fig. 3. Overstress vs plastic strain rate on 316 L Stainless Steel at 600 C, and its interpretation with the double slope and
exponential functions.

1648

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

In the rst case, with R, the elastic domain will be increased in the same way whatever the strain rate.
In the second, the increase in D will cause an increase in req that will be greater with greater strain
rate. The simplest and most used form of viscoplasticity equation with isotropic hardening is the
one that is deduced from the combination of the secondary creep law (Nortons law with a power
function between the secondary creep rate and the applied stress) and the primary creep law (power
relation between strain and time). Such approaches may be found in Rabotnov (1969), Lematre
(1971). It is equivalent to neglecting, in (17) any elasticity domain k 0, the corresponding hardening Rp, and adopting a power function for the drag stress D. This would be expressed:

req Kp1=m p_ 1=n

21

This multiplicative form of the work hardening is very practical to determine (Lematre, 1971) and
yields good results in a rather large domain, at least for quasi-proportional monotonic loadings.
3.4. Kinematic hardening equations
As kinematic type of hardening is a nearly general occurrence, at least in the range of moderate
strains, the corresponding models will have to be used when we want to correctly express either
non-proportional monotonic loadings (variation of the loading direction, thermomechanical loadings,
etc.), or cyclic loadings.
The most widespread kinematic hardening models are indicated here in increasing order of complexity. A few more advanced models for expressing special effects can be found in Sections 7 and
8. For the time being, we are discussing only strain hardening, while the time recovery effects are considered in Section 3.7.
The simplest model is Pragers linear kinematic hardening (Prager, 1949), in which the evolution of
the kinematic variable X
(called back-stress) is collinear with the evolution of the plastic strain. Thus


_ 2 C e_ p
X

3 

and

2 p
Ce
3 

22

The linearity associated with the stressstrain response (Fig. 2-b) is rarely observed (except perhaps in
the regime of signicant strains). A better description is given by the model proposed initially by Armstrong and Frederick (1966)1 introducing a recall term, called dynamic recovery:

_ 2 C e_ p  c X p_
X


3 

23

The recall term is collinear with X


(as in the general Eq. (13)) and is proportional to the norm of the

plastic strain rate. The evolution of X
, instead of being linear, is then exponential for a monotonic uni
axial loading, with a saturation for a value C=c. That is, the integration of (23) with respect to ep , for a
uniaxial loading, yields:

Xm



C
expmcep  epo
Xo  m

24

in which m 1 gives the ow direction and where X 0 and ep0 are the values of X and ep at the beginning of the loading branch considered.
For strain-controlled cyclic loading, it is shown that the stabilization occurs when X max X min 0:



DX
C
De p
jX o j tanh c
2
c
2

25

Fig. 4 gives the example of a few materials, treated in the rate-independent case, in which the cyclic
curve is described with (25) and D2r D2X k.
1
Interesting to note: this work was never published, only available as a Technical Report from CEGB (Central Electricity
Generating Board). By using this rule in the context of unied viscoplasticity and generalising it continuously, the author
contributed to the knowledge and citation of this report. In 2007, it has been published in Materials at High Temperature,
accompanied with a Preface retracing this story (Frederick and Armstrong, 2007).

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1649

A better approximation, given in Chaboche et al. (1979), Chaboche and Rousselier (1983), consists
in adding several models such as (23), with signicantly different recall constants ci (factors from 5 to
20 between each of them):

M
X

X
i

i1

_ 2 C i e_ p  c Xi p_
X
i

3 
i

26

allowing the expression of a more extensive strain domain and a better description of the soft transition between elasticity and the onset of plastic ow. Fig. 4 shows, for 35NCD16 hard steel, the significant improvement in the case where only two variables are superposed, one being linear, with c2 0.
Let us note here that the number of parameters introduced by such a superposition of back-stresses
(set of fci ; C i g) should not be considered as material parameters but as a series decomposition of a simpler expression of the tensile curve (or cyclic curve), for instance by a power law. This has been proved
later by Watanabe and Atluri (1986), based on the endochronic theory of Valanis (1980).
Other more complex combinations can be used (Cailletaud and Sa, 1995) instead of (26), but they
do not allow analytical closed form solutions in uniaxial loading. In Section 7 we will also indicate various modications of the basic AF rule used above, especially in order to improve plastic ratchetting
predictions by the constitutive models.
3.5. Cyclic hardeningsoftening
In the framework of kinematic hardening models, isotropic hardening is generally used to express
the cyclic evolution of the materials mechanical strength with respect to the plastic ow. This cyclic
hardening phenomenon (increase of strength) or cyclic softening (decrease) is relatively slow, typically taking between ten and a thousand cycles of ep 0:2%, for example, before stabilizing.
We can control the dimension of the elasticity domain with a law of the type:

R_ bQ  Rp_

27

which is the direct transposition of (23) to isotropic hardening, with b and Q being two coefcients
depending on the material and on the temperature (b will be included between 50 and 0.5 to ensure
the typical saturation mentioned above in 10 and 1000 cycles, respectively). The integration of (27)
leads to an expression Rp Q 1  expbp that can also be used in the context of monotonic loadings (but a much higher value is then needed for b).

Fig. 4. Cyclic curves on various materials and their interpretation by the AF rule or the multikinematic model.

1650

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

Fig. 5, reproduced from Goodall et al. (1980), shows the example of 316 Stainless Steel, using a normalised plot of the maximum stress evolution as a function of the accumulated plastic strain:

rM  rM0
1  expbp
rMS  rM0

28

where rM is the current maximum stress, rM0 and rMS being the corresponding initial (1st cycle) and
stabilised values. The gure shows the validity of the choice (27) because the normalised experimental
responses are approximately independent on the plastic strain range, as the model assumes.
Let us note that, in the case of cyclic softening, we can set Q < 0, so that the stabilised yield surface
size k Q will be lower than the initial one (R is assumed to be the change in the size, usually with
R0 0. Also note that the drag stress D can be used in place of the yield stress R, or the two can
be combined, or a coupling can be introduced between the kinematic hardening and isotropic hardening (Marquis, 1979) with a function /p to be dened

_ 2 C i e_ p  c /pXi p_
X
i


3 
i

29

Fig. 6 illustrates the case of 316 SS with the function / dened as / /1 1  /1 expbp. It
shows a slight dependency on the plastic strain range but not in contradiction with experimental
results.
Another possible choice of /p consists in using the variable R with a dependency deduced from an
endochronic type theory (Valanis, 1980; Watanabe and Atluri, 1986):

/p 1=1 xRp

30

Remark. Let us recall here, without more details, that endochronic theory of plasticity developed by
Valanis (1980) is one based on the hereditary form of thermodynamics of irreversible processes,
though the present formulations are developed in the context of thermodynamics with internal
variables (Germain, 1973). See a few more details in Section 4.1.2. Such hereditary theories, like in
viscoelasticity, uses the complete history of observable variables (strain and temperature), without
using the notion of internal variables. This is done by integral equation to relate stress and strain
tensors histories, which kernel contains most phenomenological information. This is the case for
instance with the theory developed in France by Gulin and co-workers (Gulin and Stutz, 1977;
Boisserie et al., 1983).
It is interesting to underline here the following fact: as demonstrated rst by Watanabe and Atluri
(1986), when using the Valanis theory, for computational purpose, a decomposition of the kernel into

Fig. 5. Modelling of isotropic hardening with the yield stress evolution for 316 Stainless Steel at room temperature (from
Goodall et al., 1980).

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1651

Fig. 6. Modelling of isotropic hardening with the Marquis modication of the dynamic recovery term, for 316 Stainless Steel at
room temperature.

a series of decaying exponentials, the model recovers kinematic hardening/isotropic hardening


separation, and, surprisingly, the back-stress obeys exactly to the multikinematic rule (26) (a
superposition of as many AF type back-stresses than terms in the series). The only difference is that
the accumulated plastic strain dependency of the yield stress automatically appears in the forms (29),
(30) above of coupling effect in the back-stress evolution equations.
3.6. Strain range memorisation and out-of-phase effects
Under cyclic conditions, for some polycrystalline materials, like OFHC copper or Stainless Steels, in
fact materials with a low stacking fault energy, special cyclic hardening effects can be observed, which
we classify here as
(1) Plastic strain range memorisation effects: after applying a large cyclic strain range, the subsequent
materials behaviour has been hardened. For lower strain ranges the stabilised cyclic strength is
higher than under normal cyclic conditions without a prior hardening at a larger strain range.
On the other hand, as shown on Fig. 7, for the increasing level cyclic test on 316 Stainless Steel,
after stabilisation of cyclic hardening at a low strain range, a subsequent cyclic hardening is still
possible when applying a larger strain range (Chaboche et al., 1979). Such a behaviour is clearly
not reproducible by the isotropic hardening law (27), in which R saturates only once to a xed
value Q. For such materials the cyclic curve (relation between stress range and plastic strain
range under stabilised conditions) is no more a unique relationship and clearly depends on
the previous loading histories.
(2) Out-of-phase effects: For materials that harden cyclically, if non-proportional multiaxial loadings
are applied (under strain control for instance), the cyclic hardening effect can be drastically
increased and the stabilised cyclic response (in terms of von Mises invariants of the stress
amplitude and plastic strain amplitude) is much more resistant than under equivalent proportional conditions. This fact was observed rst time by Lamba and Sidebottom (1978) for OFHC
copper, and has been reproduced later on several other materials, especially Stainless Steels
(Kanazawa et al., 1979; Cailletaud et al., 1984; Tanaka et al., 1985; McDowell, 1985; Benallal
and Marquis, 1987). Such an effect can be understood from crystal plasticity and dislocation
behaviour: under a non-proportional multiaxial cyclic loading, many more slip systems are activated, which increases the number of obstacles for subsequent slip to take place.

1652

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

Fig. 7. Experimental response for the 5 levels increasing strain-controlled loading on 316 Stainless Steel at room temperature
(from Chaboche et al., 1979).

We will not be able to give detailed models of such situations. Let us only summarises the existing
possibilities, in terms of macroscopic phenomenological models, as follows:
(1) For the plastic strain range memorisation, a simple way was proposed in Chaboche et al. (1979)
that introduces a new internal state variable, called q. Its evolution rule, not given here, memorises progressively the current plastic strain range (under any multiaxial conditions) provided
it is larger than previously encountered ones. Such a memory variable is taken into account in
the plastic ow rule by its inuence on the asymptotic value of isotropic hardening Q, which
now becomes a varying quantity Q q. The initially proposed rule has been generalised by Ohno
(1982), Ohno and Kachi (1986), as the cyclic non-hardening range. Moreover, in Nouailhas et al.
(1983a), a more sophisticated model, in which some part of the memory was slowly evanescent,
was used in order to describe both monotonic and cyclic hardening of (annealed) 316 SS, but
also, with the same model parameters, many different cold worked initial conditions of the
same stainless steel.
(2) For the description of out-of-phase effects, still using macroscopic models, several attempts
have been made in the eighties (McDowell, 1983; Benallal et al., 1985; Krempl et al., 1986;
Tanaka et al., 1987). One of the simplest and best rule was proposed by Benallal et al. (1989),
using a scalar parameter A based on the current tensorial product of the back-stress and the
back-stress rate. The effect interacts with the ow rule by increasing the limit of isotropic hardening Q A in a way similar to the above method of strain range memorisation. Such a model
was working quite well and, to some extent, was able to describe also the strain range memory:
for a proportional cyclic loading A is more or less related with the current amplitude of the backstress. Another interesting approach was given by Tanaka et al. (1987), introducing a structural
tensor (or polarisation tensor) as well as a non-proportionality parameter. A recent presentation
of this approach was given by Tanaka (2001).
Among the additional possibilities, we can indicate the model developed by Teodosiu and co-workers (Hu et al., 1992; Teodosiu and Hu, 1995), in which the limits of the kinematic hardening variables

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1653

and the coupling effects with isotropic work hardening are introduced specically. This kind of kinematic hardening model, differs from the above considered ArmstrongFrederic type, by the use of specic polarisation tensors. This is a little more complex but is justied by physical considerations
involving the dislocation substructures.
A slightly different form of strain range memorization has been used recently by Yoshida et al.
(2002), Yoshida and Uemori (2002), in order to better describe the work hardening stagnation effect
appearing in nite strain reverse plasticity. Modelling of such effects is important in the context of
sheet metal forming. The memorization model is written in the stress space, instead of plastic strain,
and is coupled differently with the isotropic and kinematic hardening. This model also describes a
mean-strain dependency.
3.7. Static recovery
Hardening recovery with time, whether kinematic or isotropic, generally occurs at high temperature. These thermally activated mechanisms are described macroscopically by relations such as
(13). Thus, for kinematic hardening, we will use for example a power function in the recall term acting
as a function of time (Chaboche, 1977a):

0
1m 1
kX
k i
i
_X 2 C i e_ p  c Xi p_  ci @
A
X
i

i
3 
si T Mi
i

31

where mi ; si ; M i depend on the material and temperature. In practice, we let M i C i =ci and the time
constant si will be strongly dependent on the temperature.
For the static recovery of isotropic hardening, we use any function, for example in Nouailhas et al.
(1983b), Chaboche and Nouailhas (1989):

R_ bQ  Rp_  cr jR  Q r jm1 R  Q r

32

which yields correct results for 316 L stainless steel. Fig. 8 illustrates this with cyclic relaxation test
results in controlled strain De 1:2%, incorporating a more or less long tensile hold time. The longer
the hold time, the less the maximum stabilized stress, which is the result of a reduction of the cyclic
hardening effect obtained by the compromise of the relation (32), between hardening by strain (the
rst two terms) and recovery by time (the last term). Moreover, the relaxed stress (difference between
the maximum value and the value rrel after relaxation) increases greatly, which also requires the
inclusion of the static recovery of the kinematic variables with (31), the parameters of which have
been identied by long-duration creep tests (Chaboche and Nouailhas, 1989).

Fig. 8. Cyclic relaxation on 316 L Stainless Steel at 600 C and its modelling by a unied viscoplastic model with strain range
memorisation effect and static recovery effects.

1654

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

3.8. Limiting rate-independent case


In all of the above we have dealt with the case of viscoplasticity, with a part of the stress that is
dependent on the strain rate (relations of Section 3.2). When the temperature is low enough, the viscosity effect can be neglected. For certain applications, even at high temperature, we may also want to
use the rate-independent plasticity scheme. To do this in a relation like (15) or (20) we have two options: reducing the drag stress D to a zero value having exponent n that tends toward innity.
In the rst case, it necessary follows that rv ! 0 and that the criterion f 6 0 will be automatically
met. Of course, in an expression like (15), we end up with an indetermination (0/0), but this is determined by a consistency condition f f_ 0 in the case of plastic ow. In the second case, we have
rv ! 0 and f  D 6 0. The formal treatment of rate-independent plasticity is somewhat more complex
than that of viscoplasticity, as it further brings in a loading/unloading condition and additional difculties when the material is of the negative hardening type (plastic softening). These aspects will not be
discussed here.
The monotonic or cyclic viscoplasticity equations, with the associated hardening models, simply
degenerate in the rate-independent case, with no other change than the dimension of the pure elasticity domain (see Section 3.9). Let us mention the particular case of isotropic hardening, for which
relation (20) becomes:

req k Rp

33

Quite often in applications, the relation Rp can be considered as dened point by point from the
expression r k Rep , equivalent in the uniaxial case. This function is then directly drawn from
the experimental tensile curve. Quite often it can be likened to a power function, as in the

req k Kp1=m

34

3.9. Determination methods


The determination of unied viscoplasticity models combining isotropic hardening, kinematic
hardening, recovery and viscosity effects, can be fairly strenuous work. Here, we propose a determination process by close approximations that has often proved its worth.
3.9.1. Determination of hardening equations in the rate-independent scheme
Suppose we have uniaxial tests in monotonic and cyclic loading, such as low-cycle fatigue tests up
to the stabilized cycle, with r  ep recorded. Let us also suppose that these are performed for velocities
that are fairly constant _e  Const and relatively high (_e 104 or 103 s1 , for example). From the
cyclic curve, considering that e_ p  e_  Const. at the cycle maxima, we will identify the following relation, which is valid after stabilization of the cyclic hardening or softening effects:



M
X
Dr
Ci
De p

tanh ci
k R s
2
ci
2
i1

35

in which k is the sum k K p_ 1=n assumed to be about constant. R s is the stabilized value of R but may
also include the hardening effect that is present in the drag stress. ci accounts for any coupling with
the isotropic work hardening (ci replaced by ci /sat ). In practice, if the number of back-stresses is suf
cient (three, for example), we will try to adjust k R s to get the lowest possible value. The third
variable can be assumed linear and the slope of the cyclic curve in the region of the high amplitudes
(23%) will provide the value of C 3 .
We then complete the determination of the (rate-independent) equations with the available data in
monotonic tensile loading and possible subsequent compression, with the corresponding experimental curve being expressed by

r k R p

M
X
Ci
i1

ci

1  expci ep

36

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1655

The rapidity coefcient of the isotropic hardening, b, will be provided by the number of cycles needed
to saturate the cyclic hardening or softening with a set amplitude Dep =2: A value near 2bN Dep  5 is a
good saturation criterion of the exponential. A more precise way is to plot the successions of normalized maxima rmax N  rmax 0=rmax N sat  rmax 0 as a function of p  2N Dep for a few low-cycle
fatigue tests, as shown in Fig. 5 above. An iterative processing of all of this data, with a few readjust
ments, provides k ; C i ; ci ; Q ; b (and the function /p).
In case of strain range memorisation effects, when they are evidenced by multilevel or incremental
cyclic straining, it is also during the present rate-independent step that the corresponding parameters
can be determined.
3.9.2. Determination of the viscosity equation
We now use the available data in the variable e_ p velocity domains between, let us say, 108 s1 and
4 1
10 s , to determine the viscosity equation, for example the exponent n, the constant K and the nal
value k of the true elasticity domain. We note the need for the following readjustment:

37

38

_ 1=n
k ! k K p
_ 1=n Rp
R p ! 1 fp

between the version already determined in the rate-independent approximation (with a rate about
_ and the complete version, considering the choice (19) for the isotropic hardening associequal to p )
ated with the drag variation. If we have monotonic or cyclic relaxation tests, the determination of n
and K will be greatly facilitated, with the possible use of a graphic determination method (see Lematre and Chaboche, 1985). A few iterations are needed to reach a satisfactory solution (in all these
analyses, we use the parameters determined in step 1).
3.9.3. Determination of static recovery effects
We use the data available in a regime of very low velocities _ep < 108 s1 , in long-duration creep
or relaxation tests. As Fig. 3 illustrates for 316L, the effect of the recovery mechanism appears directly
visible by the great reduction of the stress supported for a given strain rate. By successive approximations, all the other parameters remaining xed, it is relatively easy to get the static recovery parameters of the models considered mi ; si ; Q r ; mr ; cr .
If we have specic recovery test, these effects and the corresponding parameters can be measured
more directly. Such tests are, for example, a normal cycling up to stabilization, then a partial discharge
and a hold, at temperature, of signicant duration (100 h, for example), then cycling again. The recovery should be carried out at a sufciently low, but non-zero, strain or stress level, chosen such that the
partial restoration of the plastic strain cannot occur. Comparison and identication of the responses
before and after recovery then provides the parameter values sought very directly.
3.10. Generalisation to initially anisotropic materials
Such a set of constitutive equations is quite easy to generalise in the context of an initially anisotropic material. In case of orthotropy, we may use Hills criterion in place of von Mises and fourth rank
tensors in the evolution equations for the back-stresses. Such generalisations were used for example by
Nouailhas (1990), in the context of a single crystal constitutive modelling. In that case Hills criterion is
not sufcient and should incorporate higher order invariants, as shown in Culi and Nouailhas (1993).
Several unied constitutive models have also been developed in order to include such possibilities
of initial anisotropy, for instance in VBO theory (Lee and Krempl, 1991) and in Delobelles model
(Delobelle et al., 1995), among those approaches discussed in Section 5.
4. Temperature effects and microstructural evolutions
4.1. Inuence of temperature under stable conditions
In the previous sections, the constitutive equations were presented in an isothermal context. Inuence of temperature was only underlined for those phenomena, like viscoplasticity or static recovery,

1656

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

that are thermally activated. In fact, there are many parameters in the constitutive equations that may
be considered as depending on temperature.
4.1.1. Parametric dependency on temperature
By stable conditions we indicate the normal ones, without microstructural evolutions, at least
without changes in the mechanical properties that could be induced, at a given temperature, by fac_
tors related independently to the history of temperature (or to both T and T).
In those cases, the inuence of temperature on the material parameters in the constitutive equations may be introduced simply by interpolation techniques, linear, parabolic, spline functions, etc.
Each parameter can be temperature dependant, like with parabolic expressions:

CT ai T  T i 2 bi T  T i ci

for T i < T < T i1

39

The choice of interpolating functions should expect the parameters values determined independently
at the various temperatures where experimental data were available. Most often, doing so, there is a
need for some iterations.
At this stage, it is necessary to check on one or two normalising quantities, like 0.2% proof stress,
stress to 1% creep in 100 h, or others, the correct tness and monotonicity of the quantity even at
intermediate temperatures where full experimental data were not available for a complete determination of the material constitutive parameters.
In some case (Cailletaud et al., 2000), with modern optimisation capabilities, the identication procedure could be done all temperatures together, indentifying with all test together the chosen material
functions of temperature. However, such an automatic process should be applied carefully, depending
on the availability of sufcient experimental informations.
Let us note the interest for some normalisation of parameters. This point can be exemplied with
the simple power law for the viscosity function, which can be written simply as

p_

nT

rv

DT

or p_ e_

rv

nT

D T

40

The temperature dependency in exponent n is the cause of rapid variations in the drag stress DT
when using the rst expression (in that case DT is the viscous stress value when p_ 1). Due to
the usual rate domain at which the constitutive equations are used, most often below 103 s1 , it
may be much better to use the second expression, choosing arbitrarily the normalisation parameter
e_ 104 s1 for instance, giving to D T the character of a normalised drag stress for this typical
strain rate. Assuming the exponent as given, we have the following obvious relation:

D T DT_e 1=nT

41

A different, but not incompatible, way of dening the viscosity function is given by the ZenerHollomon type formulation (Zener and Hollomon, 1944), which combines the effect of the temperature and
the effect of the strain rate into a single master curve. This approach consists in saying:

p_ hTZ

rv
Dro T

42

where Z is a unique monotonic function and where hT and ro T are two functions of the temperature
to be dened. The advantage of this formulation, illustrated in Fig. 9 reproduced from Freed and Walker
(1993), is that it avoids the strong nonlinearity of a power function in which the exponent is strongly
dependent on the temperature. As the function Z is dened on a large number of decades in strain rate
(23, for example), the role of the function hT is then to make the useful rate domain slide by normalization (in practice limited to 68 decades in strain rate). The equivalent exponent (the slope of the
function Z in the bi-logarithmic diagram) thus goes from a very low value in a certain region of the
_
_
curve (low values of p=hT)
to a very high value in the opposite region (high values of p=hT).
4.1.2. Discussion on the temperature rate term in the back-stress evolution equation
The need for such an additional term, proportional to the temperature rate in the evolution equation
for the back-stress, was already considered by Prager (1949) in the context of linear kinematic

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1657

Fig. 9. Stationary creep behaviour on Aluminium and Copper and its interpretation by a Zener-Hollomon function (from Freed
and Walker, 1993).

hardening. Introduced also by the author in the unied viscoplastic constitutive equations using the
nonlinear ArmstrongFrederick format (Chaboche, 1977b), it was a subject of discussion all along
the past twenty years, for instance by Walker (1981), Moreno and Jordan (1986), Hartmann (1990),
Ohno et al. (1989), Ohno (1990), Lee and Krempl (1991). For kinematic hardening this rate term, considered as necessary for obtaining stable conditions, is as follows (only one back-stress is considered
here):

_ 2 CT e_ p  c X p_ 1 oC X T_
X



3
CT oT 

43

Compare to Eq. (23) to see the role of the temperature rate term, directly induced by the variation of
C parameter, more or less the hardening modulus in the model. There are several arguments for such
an additional term. The discussion is made here for the kinematic hardening but some arguments are
valid for other hardening rules:
(1) On the physical level the true state is dened by the dislocation arrangements, and the plastic
strain incompatibilities (from grain to grain). Those quantities are all directly associated with
the plastic strain. For the same microplasticity state, if we rapidly change the temperature,
we do change Youngs modulus, which immediately changes the internal stress elds associated
with the various strain incompatibilities. This is the reason why, in Millers unied model (see
Section 5.1), the back-stress is normalised by Youngs modulus;
(2) From the thermodynamic point of view, and consistently with the rst remark, we usually consider a state potential (Helmholz free energy), that is depending on strain like hardening state
variables:

w we ee ; T wp a
; T


44

from which Hookes law (2) derives, by

r

ow
oee

45

The truly independent state variable is then a back strain tensor a


. If the part wp of the free

energy (the energy stored in the material by kinematic hardening) is expressed as quadratic
in this back strain tensor:

1658

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

wp a
; T


1
:a
CT a
 
3

46

we obtain the corresponding back-stress (the thermodynamic associated force) as

ow 2
CT a

oa
3


47

If considering a
as the true independent state variable, we obtain


_ 2 CTa_ 2 oC a T_ X_ _ 1 oC X T_
X


 T0
3
3 oT 
CT oT 

48

_ _ is the rate expression for the back-stress under constant temperature conditions.
where X
 T0
(3) From the phenomenological point of view, we know that a great temperature dependency is
possible for the 0.2% proof stress, so that, for example, the proof stress at low temperature T 1
may be higher than the rupture stress at a high temperature T 2 . If we imagine now a 0.2% monotonic tensile plastic strain at T T 1 , then a rapid unloading and a rapid temperature change to
T T 2 (without any new plastic ow), upon reloading it is impossible to accept that the plastic
ow will not begin before r0:2 T 1 . Clearly the correct behaviour will be to begin plastic ow for
a stress around r0:2 T 2 . Such experimental evidences were for example given by Chan et al.
(1990), when doing tensile tests interrupted by rapid temperature changes.
(4) The last argument is the fact that, in the absence of this temperature rate term, and with a linear
or nearly linear kinematic hardening, the hysteresis loops may shift unreasonably in stress
(Wang and Ohno, 1991). A simple exercise may show such a situation (Chaboche, 1993), not
reproduced here in detail. Let us consider a reversed strain cycle with temperature changes taking place at the maximum mechanical strain (from low to high temperature) and at the minimum strain (from high to low temperature). Due to higher hardening slope CT at a low
temperature, there will be, cycle-by-cycle, an unlimited increase of the maximum stress (if linear kinematic hardening is used in the model).
4.2. Metallurgical instabilities and aging effects
In the above, we have considered only stable materials for which microstructural transformations are negligible or mechanically unperceivable. The effect of the temperature is in the constitutive
equations, but in one-to-one fashion, for example through a dependency of the material parameters
as a function of the temperature. Under certain temperature conditions, on the other hand, metallurgical changes may occur, like phase changes, dissolution, precipitations, coarsening of precipitates,
etc., that signicantly modify the mechanical properties.
The generic terms aging covers all of the unstable situations, of which there are:
 dynamic aging, due to the dragging of the dislocations by the atoms in solution, leads to an inverse
relation in velocity (the viscosity exponent that would be negative in a certain strain rate regime).
This non-monotonicity of the relation between rv and p_ is a source of instabilities (succession of
localized bands) associated with the Portevin-Le Chatelier effect.
To globally model such phenomena Miller (1987a) is using a strain rate dependency (or plastic
strain rate) of the drag stress in the viscosity function. It leads to an implicit, non-unique, dependency of the viscous stress on the strain rate. Section 5.1 gives more details on this approach. Other
solutions are possible, but we should not forget that modelling those situations in the framework of
classical continuum mechanics becomes debatable, due to the strain localisation phenomena taking
place in the tensile specimen.
 static aging, a growth in material strength with time (from a mechanical response viewpoint, this is
the reverse of a static recovery), that can be expressed by an equation of the type dR h dt. This
phenomenon will occur for example in certain aluminium alloys at ambient temperature, for which
destabilisation effects (metallurgical changes) are effective at high temperature.
More or less sophisticated mechanical models have been proposed for this purpose (Marquis, 1989;
El Mayas, 1994). One of the difculties in these models is to meet (in an a priori way) the

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1659

thermodynamic requirements of a positive dissipation. Some related modelling aspects have been
discussed for example in Chaboche (1993), Chaboche (1996).The temperature dependency of some
parameter for isotropic hardening, together with the memory of maximum strain range (Section 3.6)
has been used by Ohno et al. (1989) to describe some temperature history effects in 304 Stainless
Steel.
 phase changes during heat treatment or sometimes during use. In terms of models attempting to
express the mechanical consequences of these phenomena, we will mention that of Cailletaud
(1979), to express the dissolutions, precipitations, growths of the c0 precipitates in superalloys for
turbine blades, phenomena occurring under certain temperature cycles. This model uses two additional state variables, one related to the volume fraction of the precipitates, the other to their size. It
is obviously impossible here to go any further in the explanation of these phenomena and of the
various modelling possibilities.

5. Other unied viscoplastic constitutive equations


Many unied elasto-viscoplastic constitutive theories have been developed in the literature, since
the middle seventies, especially for modelling the small strain cyclic conditions. Clearly, it is not possible to describe such modelling theories in complete details. We will summarise here their main
properties, underlining the important differences compared to the author constitutive equations.
The notations will generally follow the ones already used in the previous sections, except when mentioned. Only the isothermal conditions will be discussed below.
5.1. Millers MATMOD equations
This unied viscoplastic model (Miller, 1976) uses one back-stress for kinematic hardening and a
drag stress for isotropic hardening. There is no yield stress in the model (the elastic domain is reduced
to one point). The viscosity function is a combination of an hyperbolic sine and a power function, such as

"
p_ hT sinh

3 #n
kr=E  ak 2
D

49

where k:k denotes the von Mises invariant. The back-stress X


Ea
is normalised by Youngs modulus.


The main specicity of this model is the drag stress evolution equation, that contains several terms, as
in Schmidt and Miller (1981):

q
F sol;1 F def 1 F sol;2

50

where F def is the classical isotropic hardening variable. F sol;1 and F sol;2 are factors depending on the
norm of plastic strain rate (as a parameter), in order to include an explicit representation of solute drag
effects and dynamic strain aging, respectively without and with interactions to deformation mechanisms. a
and F def obey a hardening/dynamic recovery/static recovery format in the form of (13).

The static recovery terms use also an hyperbolic sine function.
In more recent versions (Miller, 1987a, 1996), there is a coupling with the back-stress by which the
asymptotic value Q of isotropic hardening is enhanced as Q qkak2=3 . The advantage of this term is to
induce a strain range dependant cyclic hardening of the material, but with an erasing memory instead
of a complete memory like in the model mentioned in Section 3.6.
The expressions for the whole evolution equations are not given in detail. Compared to authors
model presented in Section 3, we may point out some differences:
 no yield stress and corresponding hardening, but a drag stress that includes a complex coupling
with the current total strain rate;
 the viscosity function, as well as the static recovery terms in the F def and a
evolution equations, uses

an hyperbolic sine function;

1660

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

 the back-stress temperature dependency is normalised by Youngs modulus (a


is in fact a back

strain);
 the back-stress evolution was linear in the rst versions (Miller, 1976), no dynamic recovery term
but nonlinearity due to the static recovery term. More recently, in Lowe and Miller (1986), the
model was using three back-stresses of the Armstrong Frederick type;
 another specicity of the model is to have a limited number of functions of temperature, all
expressed as Arrhnius functions with different activation energies.
5.2. Bodners theory
This unied viscoplastic constitutive theory began with the Bodner and Partom (1975) article. First
versions were using only isotropic hardening, as a drag stress. The version briey summarised here is
one of the most advanced ones, taken from Bodner (1987), that uses also a directional hardening.
The equations are presented, showing the main differences with the author constitutive model:
 the viscosity function combines an exponential and a power function:

"   #
2n
Z
p_ p_ 0 exp 

req

p_ 0 104

51

As it will be seen in Section 5.7 such an expression gives a tendency in opposition with most other
kinetic equations;
 the direction of the viscoplastic strain rate is given by the stress deviator, without any translation by
a back-stress as in most other models:

e_ p p_


3
2

r 0

52

req

 the hardening effect is entirely taken as a drag effect Z, including an isotropic part K and a directional one D:

Z K D

Db:u



r

r
:r
1=2



53

 the isotropic hardening variable follows the general hardening/dynamic recovery/static recovery
format (13), with


r 1
_ p  A1 K 1 K  K 2
K_ m1 K 1  KW
K1

54

_ p in place of the accuthe identication with (32) is quite easy except the use of the plastic power W
mulated plastic strain rate as the driving factor;
 the directional hardening variable, introduced in the middle eighties (Bodner, 1987), follows also
the general format (13), with


r2 b

_ p  A2 K 1 kbk
b_ m2 D1 u

b

W



K1
kbk

where kbk

55

r
b : b. The main difference with (31) is the use of plastic power as a driving factor. Let


given by the stress direction in place of the stress


us note also the direction of the driving term u

deviator. Bodners model has not the temperature rate term in the evolution equation for directional hardening, but it was introduced by Chan et al. (1990).
The most important difference among other unied constitutive models is the introduction of
directional hardening without using a back-stress. It has consequences both on the hardening effect
(multiplicative instead of additive in stress) and on directionality effects. The direction of viscoplastic
0
0
ow is always given by the stress deviator r
instead of the difference r
X
.




J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1661

This may have a signicant impact on multiaxial non-proportional loading conditions. Such a situation may be illustrated bypconsidering
the out-of-phase (90) tensiontorsion loading, a circle in the

equivalent stress axes r; s 3, and assuming


no isotropic hardening. It produces also a circular rep
sponse for plastic strain and total strain e; c= 3, visualised in the same axes after multiplying them
by 3l (l = elastic shear modulus). After some transient response, Fig. 10 shows that, under such conditions, the stress and plastic strain response delivered by Bodners equations will automatically have
a phase difference of 90. This is not the case with theories using the back-stress and the direction of

Fig. 10. Simulation of the stabilised out-of-phase stress controlled cycle with a model without back-stress. Strain responses are
indicated, with relative positions and directions.

Fig. 11. Simulation of the stabilised out-of-phase stress controlled cycle with a single AFrule. Responses in strains and backstress are indicated, with relative positions and directions.

1662

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

0
plastic ow given by the difference r
X
. Fig. 11 illustrates schematically the case with one back

stress obeying AF rule, in which we observe the phase order by r
, X, 3l e , 3le p . Many experiments
 
under out-of-phase conditions have shown that a phase difference of 90 between stress and plastic
strain is not realistic at all (Benallal et al., 1989).

5.3. Robinsons constitutive model


This constitutive equation was proposed rst in Robinson (1978). More recent and advanced versions have been developed by Arnold and co-workers (Arnold and Saleeb, 1994; Saleeb et al., 2001).
The main specicities are as follows:
 the model uses a back-stress, a drag stress, a yield stress and a power function for the viscoplastic
ow:

p_ p_ 0

*
+n
kr  Xk2  Y 2

56

D2

 the evolution equations for the drag stress D and the yield stress R are not specied here (Y is given
below);
 the back-stress evolution equation uses an hardening and a static recovery terms. However there is
no introduction of a dynamic recovery effect as in other models;
 the non-linearity of the kinematic hardening is reproduced by a power function of the back-stress
invariant kXk, called G:

e_ p

_X n
 RkXk X
with


Gm
8
0
X
:X
> 0 and kXk P G0
< kXk if r



G
0
: G0
if r
X
:X
< 0 or kXk < G0




57
58

where G0 is a small valued quantity. Two advantages when doing so are: the need of only one backstress and the possibility to have an easy smooth elasticplastic transition, due to the quasi-innite
slope when kXk 0.
 However, to deal with cyclic conditions, this model imposes a very special denition for the elastic
domain and for the rate of hardening:

8
R
>
<

if

0 X 
Y
r
 
>
: Max R;  X : krXk
if


r 0 : r 0  X P 0
r 0 : r 0  X < 0

59

The advantage mentioned above is reduced due to the additional complexity of the modication of the
elastic domain. Though continuity is enforced, there is a possible non-convexity of the effective elastic
domain f kr  Xk  Y 6 0.
Another drawback is the indifferent character of the kinematic hardening. After a tensile plastic
ow for example, and a short stress excursion in compression, the previously positive back-stress is
erased (it vanishes rapidly) and the subsequent tension results in exactly the same response than
the initial tensile curve (at least when isotropic hardening is not considered). This is in contradiction
with most of the experimental results.
Several modications, generalisations and improvements of the original Robinson approach have
been developed by Arnold and Saleeb (1994), Saleeb et al. (2001), in the context of an extended thermodynamic framework. The introduction of a supplementary dynamic recovery term, and complex
couplings between isotropic hardening (yield and drag stresses) and kinematic hardening was solving
the above mentioned difculties, but also reducing the impact of the specicities offered by (57), (58).

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1663

5.4. The Walker model


When compared with the constitutive equations of Section 3, Walkers version has many common
points, and only a few differences, as follows:
 The viscoplastic function is a power law, without a yield stress, and isotropic hardening is introduced in the drag stress:

p_ p_ 0


n
kr  Xk
Dp

60

 The back-stress evolution equation is taken as



_ 2 n1 n2 e_ p  X X0  2 n1 e p n3 /pp_ n6 kXkm1

X




3
3 

61

 There is a special asymmetry in the back-stress, given by the constant tensor X


, that may describe
0
an initial non-recoverable asymmetry of the viscoplastic behaviour.
 When X
0 it is easy to check (Chaboche, 1989) the equivalence with superposing one linear
0
back-stress and one nonlinear with the AF rule:

_ 2 n1 e_ p
X

3 
1
_ 2 n2 e_ p  n3 /pX2 p_  n6 kXkm1 X2
X



3 
2

62
63

 The isotropic hardening is introduced in the second back-stress evolution equation, with the Marquis expression:

/p 1

n4
expn5 p
n3

64

 Except the fact that there is only two evolving back-stresses, the main difference is related with the
static recovery effect: it takes place only for the nonlinear back-stress, and its amount is given by
the norm of the total back-stress.
 In Walkers equations there are temperature rate terms for the two back-stresses, not indicated in
(61) above. As shown in Chaboche (1989) they are exactly in conformation with the thermodynamic
framework discussed in Section 4.1.2.

5.5. The VBO theory of Krempl


The theory of viscoplasticity based on overstress, developed by Krempl and co-workers (Cernocky
and Krempl, 1980; Yao and Krempl, 1985; Krempl et al., 1986; Ho and Krempl, 2002), has also many
common features with the constitutive model of Section 3. One of the main differences is to formulate
the back-stress evolution in terms of total strain rate instead of viscoplastic strain rate:
 The equilibrium stress is a second rank tensor, called g, more or less equivalent with the stress state

projected on the current elastic domain.
 The overstress is the difference r
 g. The viscoplastic function is directly depending on its von


Mises invariant rv k r

g
k:



rv
p_
E/rv

r0  g0

3 
e_ p_

2
p

rv

65

where / may have various forms, for example: /rv k1 1 rv =k2 k3 .
 The growth law for the equilibrium stress is driven by the total strain rate, but the dynamic recovery term is proportional to the norm of the viscoplastic strain rate:

1664

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

g_ wrv e_  g  f


wrv  Et
p_
A

66

 The hardening function wrv introduces rate effects in the evolution equation for the equilibrium
stress, which differs from other theories and induces various possibilities.
 The f tensorial variable obeys a linear kinematic hardening law, where Et is the asymptotic tangent
modulus:

f_ Et e_

67

 Parameter A may be a material constant or a function of the accumulated plastic strain, as in the
Marquis model. In some applications it has also been taken as depending on the plastic strain range
memory and out-of-phase index (Colak, 2004).
 The standard VBO model does not take into consideration the static recovery of the equilibrium
stress, but this is easy to incorporate as in other theories.
It is easy to show that the VBO theory reduces to the superposition of two independent kinematic
variables, g f x
, with (67)

! for f and

x

_ w  Et e_  p_
x


A

68

In Chaboche (1989) it was shown that, under multiaxial proportional conditions and the limiting case
of rate-independent plasticity, the VBO theory does coincide with the theory of Section 3 with two
back-stresses (one linear and one nonlinear). The difference induced by using the concept of equilibrium stress and the total strain rate in its evolution equation will be active only for the viscoplastic
case. However, as discussed for example by Freed and Walker (1990), there could have some advantages concerning the modelling of ratchetting effects, provided the quasi-linear evolution of g during

what is usually considered as a purely elastic loading.
Let us note a specic difculty with this VBO theory for its incorporation into a standard thermodynamic framework. As discussed in Chaboche (1996), it will lead to an unconventional denition for
the elastic strain, due to the use of a reversible term in the evolution equation of the internal state variable. Such discussion was also given, in other terms in Lubliner (1973), Freed et al. (1991), Malmberg
(1990). Recent efforts have been made by Hall et al. (2005) to interpret the stress rate dependent term
as a dissipationless contribution.
5.6. Delobelles approach
The unied constitutive model of Delobelle was developed initially by including two back-stresses
playing role successively instead of simultaneously (Delobelle, 1988), with some complicated coupling
criteria. A more recent but enhanced version is summarised here, due to the works done with Robinet
(1995) and Schfer (1997). Though existing for an orthotropic material Delobelle et al. (1995), it is
written here for the isotropic particular case:
 the viscoplastic function is given by an hyperbolic sine like in Millers model:

p_ p_ 0 T sinh

rv

n

Dp; T

69

where the viscous stress is rv kr  Xk (no yield stress);


 the back-stress evolution is given using a secondary and a tertiary back-stresses as




m X

_ C 2 Yp e_ p  X X1 p_  rm T sinh kXk
X




3
X 0 T
kXk

_ C 1 2 Yp e_ p  X1  X2 p_
X




1
3

2
_ C 2 Yp e_ p  X2 p_
X



3
2

70
71
72

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1665

 for rapid loading conditions, or when the static recovery is negligible, the 3 evolution equations of
the back-stresses are equivalent to the superposition of 3 independent back-stresses, each obeying
the rule (26), as shown in Chaboche (1986);
 the common asymptotic value for the back-stresses Yp is a given function of the accumulated
plastic strain, like for example: Yp Y sat Y 0  Y sat expbp. For saturation under a rapid plastic straining, we have: kX 2 k ! Y, kX 1 k ! 2Y, kXk ! 3Y;
 the drag stress Dp; T is also a given function of temperature and accumulated plastic strain;
 the static recovery appears only on the kinematic hardening. It is implemented directly on the main
back-stress evolution equation, with an hyperbolic sine dependency, like in Millers approach. As in
Walkers model, it plays role as a function of the norm of the total back-stress.
Concerning further modications of static recovery of the back-stresses, we could also mention
more recent ones by Yaguchi et al. (2002) and Zhan and Tong (2006). In this model there are two
back-stresses. The rst one obeys the classical ArmstrongFrederick expression (23), without static
recovery in it. The second one introduces an additional tensorial variable Y
, as


_ 2 C 2 e_ p  c X2  Yp_
X
2 


3 
2
2
3
X
2
_ a4Y sat
Y
Y5kX2 km

kX
k  
2

73
74

Such a modication was shown to greatly improve the modelling of relaxation, under monotonic and
cyclic conditions, for new nickel-based superalloys (Zhan and Tong, 2006).
5.7. Comparison of the viscosity functions
In any unied viscoplastic constitutive equation, there is a function to describe the relation between the stress, or viscous stress, or overstress, and the norm of the viscoplastic strain rate. The reference function is often the power law (Nortons equation for creep) but, very often the observed
exponent is varying with the stress (or strain rate) domain. In many experimental results there is
the appearance of a saturation effect of the viscosity for high rates, which justify the use of an hyperbolic sine function for instance.
In the present section, we systematically compare the viscosity functions (or kinetic equations)
used in the various constitutive models considered in the present Section 5.
We consider also two specic expressions used in applications of Onera constitutive model, which
general framework has been presented in Section 3


n1 
rv
exp a
D
D
 a 
r
n 
r
v
v
p_
1
D
q

p_

r
n
v

75
76

In all these expressions, the exponent n is strongly dependent on the temperature, while the viscosity
phenomenon is thermally activated (n becomes small at high temperatures).
We also add others, like Johnson-Cook equation used in the context of dynamic plasticity, in the
high rate regime, which expresses:

p_ p_ 0 exp

 

1 krk
1
C D

77

A special mention can be made here for the Kocks et al. (1975) expression, also used by many others
(Busso and McClintock, 1996; Cheong et al., 2005). Let us note that it generalises Johnson-Cook ones,
which is recovered when choosing p q 1:

"



p q #
1
krk
1
p_ p_ 0 exp 
C
D

78

1666

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

In all these expressions, the exponent n is strongly dependent on the temperature, while the viscosity
phenomenon is thermally activated (for instance, exponent n becomes small at high temperatures).
However, temperature will not be considered in what follows.
For comparison purpose, all these viscosity functions are adjusted each other by the following way:
 when possible, using one point r1 ; p_ 1 with reference Nortons exponent N 0 , in the low rate regime
(for instance p_ 1 108 s1 ), and one point r2 ; p_ 2 in the high rate regime (for instance p_ 2 1 s1 )
 in other cases (Bodner, Johnson-Cook), we adjust only at an intermediate point r 1 ; p_ 1 like
p_ 1 106 s1 , at which we adjust also the equivalent exponent N 0 .
These comparisons are made for a given state of hardening. However, the viscous stress having not
exactly the same role in each theory, the comparisons are showing only qualitative tendencies. Let us
note that, in most viscoplastic models, the time recovery effects that may take place under low stresses, low strain rates (long durations), are not taken into account in the viscosity functions studied here.
They are considered as playing role directly in the evolution equations of the hardening variables, with
the last term in the general format (13). Therefore, the plot made in Fig. 12, in terms of the viscous
stress (at a xed hardening state), should not contain these effects, like in Fig. 3.
_ ln r are indicated
The considered functions and the corresponding effective exponent N d ln p=d
in Table 1, in the uniaxial format, together with the adjusted parameters, in the case N 0 20,
p_ 1 108 s1 , r1 150 MPa, p_ 2 1 s1 , r2 300 MPa, p_ 1 106 s1 , r 1 190 MPa.
Fig. 12 presents a loglog plot of the viscous stress (or stress) as a function of the viscoplastic strain
rate. The following trends are clearly evidenced:
 three models perform quite similarly, having the same exponent N 0 20, in the low rate regime:
the Delobelle hyperbolic sine and the two Onera versions;

Fig. 12. Comparison of viscoplastic ow functions, overstress vs plastic strain rate. Nortons exponent N 20 in the intermediate regime 108 < e_ p < 104 s1 .

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1667

Table 1
Comparison of various viscosity functions and corresponding effective exponents
Model

Viscoplastic strain rate p_

Equivalent exponent N

Values of parameters

Onera exponential

n an 1Dr n1

n N 0 ; D 376:8; a 546:5

Onera double slope

Dr n expaDr n1

a
Dr n 1 rq

n 1ra=qa

n N 0 ; D 376:8; a 50; q 274

Krempl

r 1 r k3
Ek
k

1 k3 k2rr
r 1:5 1 r 1:5
3
Th D
2n D
r n 1 r n
n D Th
D
2n
n rZ

Ek1 1:1023 ; k2 93:3; k3 31

C 1=N 0 ; D 190:; p_ 0 106

r2 1:5

Miller

Bsinh

Delobelle

Johnson-Cook

A sinh D
2n
p_ 0 exp  rZ


p_ 0 expC1 rD  1

Kocks

p_ 1 exp C1 h1  Dr p iq

Bodner

r n

CD

pq r p
C D h1

 Dr p iq1

n 23 N 0 ; D 311:2; B 0:017
n N 0 ; D 271:8; A 0:00145
n 0:395; Z 11095; p_ 0 105
p 0:1; q 0:44; D 328:5; C 0:0107; p_ 1 105

 the double slope version is interesting in the sense it limits the equivalent exponent value for high
rates, which may have positive consequences at the numerical stage when using explicit time integration schemes;
 Krempl and Miller expressions give much less saturation effect. If enforced to pass the point r2 , p_ 2 ,
in the intermediate regime 105 < p_ < 101 s1 they do not perform as the previous ones. Another
adjustment is possible, as shown on the gure, but it does not change the general tendencies;
 Bodners model leads to a totally inverse tendency, with an increasing viscosity effect (or a decreasing effective exponent N). It may be adjusted to the other models only in the regime
109 < p_ < 104 s1 . Let us recall that the generally observed trends are opposite to this model;
 Johnson-Cook expression, usually applied in the high rate regime (dynamic plasticity), when
adjusted here for 108 < p_ < 101 s1 , shows a slower saturation effect (like Krempl and Miller
expressions). Moreover it leads to a signicant underestimation of the stress in the low rate regime
compared to Nortons like expression, which may lead to a questionable trend;

Fig. 13. Effective exponent of the power law as a function of overstress for various viscoplastic ow functions.

1668

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

 the Kocks expression having more degrees of freedom (exponents p and q) allows a better representation than Johnson-Cook. It leads to a saturation effect in the high rate regime (complete saturation
when r ! D), but less pronounced than Onera exponential or Delobelles hyperbolic sine. Moreover, in the low rate regime, this choice leads to lower stresses than in the power law. Though it
needs to modify consequently the recovery effects (compared to the previous models), it could
be a good alternative.
Fig. 13 shows the evolution with the strain rate of the effective exponent N for all the models. It
conrms the trends observed above: the Delobelle and Onera models have a constant exponent for
low rates, then increasing, which leads to saturation of the viscosity effect. Krempl, Miller and Johnson-Cook shows a continuously increasing exponent, but no saturation effect; Kocks expression is
somewhat intermediary. Bodners model gives a decreasing exponent, contrary to all other models.
The same exercise, repeated for lower reference exponents, N 0 10, N 0 5, N 0 3, shows exactly
the same qualitative trends. However, the domains in which every models can be adjusted each other
is further reduced: 108 to 104 for N 0 10 and 106 to 104 for N 0 3.
6. Other types of modelling
6.1. Plasticitycreep partition
This is the oldest way of describing plasticity and creep phenomena simultaneously, by adding two
independent inelastic strains. Eq. (1) is then replaced by

e e e e p e c

79

Let us note immediately that the plastic strain e p of the previous Sections included both the plasticity
and creep effects in unied fashion. Here, on the other hand, we consider that they are separate and
generally independent. The variations of the two inelastic strains will then be described:
 by rate-independent plasticity theory for e p with a normality rule such as (4), and in association
with the hardening equations that are appropriate for the type of application considered, isotropic
hardening for applications under quasi-monotonic loading, kinematic hardening or a combination
of the two for applications under cyclic loading or when non-proportional multi-axial effects may
arise. Without explaining, these can be written formally:
p
a_ pj hj e_ p  r pj apj ; . . .apj e_ p

80

The advantage of decoupling between plasticity and creep is that it makes it easy to determine
materials parameters either from the monotonic tensile curve or from the cyclic curve (step 1 in
Section 3.9).
 by a creep type of law for e c , incorporating primary creep and secondary creep, in an integrated
form such as

eceq req ; t A1 req t 1=p A2 req t

81

in which req is the von Mises equivalent stress, as dened from (5) and (8), and
is the equivalent
creep strain, dened in the same way as p in (11). In place of (81) a differential form is more correct,
because it brings in the strain hardening:
eceq

e_ ceq g c req ; eceq

82

for which we can also take up an hardening equation of the multiplicative type such as (21), by
replacing p by eceq . Here it must be underlined that such a creep equation with isotropic strain hardening cannot correctly describe cyclic creep conditions, that show evidence of successive primary
creep periods after each reversal. It is also possible to adopt a form such as (13), with additional
hardening variables (combining isotropic and kinematic hardening):
c
a_ cj hj e_ c  r cj acj ; . . .acj e_ c

83

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1669

This form of evolution law combines strain hardening and static recovery (effect of time, important
in long term creep). Examples of such modellings may be found in Murakami and Ohno (1982). Here
again, whatever the form of hardening equation chosen, the decoupling with the plasticity allows
very easy determination either from pure creep tests or from relaxation tests.
This method by partition of the inelastic strain was used routinely up until the experimental observation, reported on numerous occasions, of an obvious coupling between plastic strain and creep
strain, by way of the associated hardening effects. An example of experimental observations can be
found in Ikegami and Niitsu (1985), among many others. Fig. 14 schematically shows the type of
observation made on a quick tensile test, interrupted by a long-duration period of creep (constant
stress). Clearly, there is a nearly immediate forgetting of the creep period and the experimental evidence that hardening is correlated with the sum ep ec , and not dened independently with ep or
ec , which assumption should have given the horizontally translated curve on the gure.
On the other hand, unied viscoplastic constitutive equations take into account the hardening induced by creep strain on the subsequent tensile loading, approaching better the experimental observation. However, this may be quantitatively insufcient, provided strain hardening develops less
signicantly in creep, especially in steady-state creep, due to the effect of time recovery.
This type of observation, and many others going in the same direction, have led to the development
of more sophisticated non-unied approaches but with coupled hardening (Kawai and Ohashi, 1987;
Contesti and Cailletaud, 1989) by writing, for example:
p
pc
a_ pj hj e_ p hj e_ c  r pj apj ; . . .apj

84

cp
c
a_ cj hj e_ p hj e_ c  r cj acj ; . . .acj

85

with all sorts of possible variations. These approaches are followed relatively little, because there are
determination complications or difculties analogous to those of unied theories.
6.2. Multiple mechanisms multiple criteria
6.2.1. A general formulation
The unied viscoplastic constitutive equations consider only one inelastic mechanism, that includes both plastic and viscoplastic (or creep) effects. Most of the rate-independent plasticity frameworks (limit case of the unied viscoplastic ones or others like in Section 7 and 8) are also considering
a unique plastic strain variable.

tension at constant strain rate

tension creep tension

creep
independent creep
& plasticity

p + c

Fig. 14. Schematic results of the tensioncreeptension test and comparison with the test at constant strain rate.

1670

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

However, there are other approaches possible, like in slip plasticity (Mandel, 1965; Hill, 1966; Rice,
1970). A quite general presentation, used in Besson et al. (2001), can be given as follows.
In viscoplasticity, we assume a collection of potentials Xs , s 1; S, dependent on r
and hardening

variables AJ through a yield function f s r
;
A
.
The
plastic
strain
rate
is
a
summation
of
the associated
J

mechanisms:

e_ p


X oXs X of s

v_ s
o
r
or
s
s



86

and, in the Generalized Standard Material framework (Halphen and Nguyen, 1975), the strain-like hardening variables evolve as

a_ I 

X oXs
X of s

v_ s
oAI
oAI
s
s

87

where AI are the thermodynamic forces associated with internal state variables aI , deriving from the
free energy potential by

AI

ow
oaI

88

Let us remark that AI gives generally hardening effect. Therefore, the partial derivative of s =oAI is
negative.
In the rate-independent plasticity framework the v_ s are replaced by the plastic multipliers k_ s to be
determined by a linear system of consistency conditions:
r
_
f_ r n
:r



X of r
A_ I 0
oAI
I

89

From (88) and (87), we get

A_ I

X oAI
X X oAI of s
k_ s
a_ K 
oaK
oaK oAK
s
K
K

90

and (89) rewrites:


r
_
f_ r n
:r



Hrs k_ s 0

91

with

Hrs

X of r oAI of s
oAI oaK oAK
I;K

92

Hardening manifests itself through an interaction matrix, whose components Hrs express the hardening induced by mechanism s on the mechanism r. Using the time derivative of Hookes law (2) and its
projection as
r
_ nr : L :
n
:r





e_  e_ p


93

and using (86), we obtain the system:

X
s

r
Q sr k_ s n
: L : e_


94

r
with Q sr n
:L
: ns Hrs . Its solution may be written as

 

k_ s

X
r

r
Q 1
: L : e_
sr n


95

and the tangent stiffness tensor writes:

L
t L



XX
s

s
r
Q 1
: n
n
: L
sr L




96

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1671

Due to the associated plasticity framework used here (elastic domains and yield surfaces are both described by f s ) the matrices Hrs and Q rs are symmetric and the tangent stiffness tensor also has the principal symmetry. Obviously, the linear system (94) only concerns those mechanisms that are actually
active.
Classical crystal plasticity models (or their viscoplastic counterpart) are clearly described in this
general framework (using for instance a Schmid criterion for f s ). The model used in Section 6.3 below
is exactly under these lines, both at the level of single crystal viscoplasticity and for the polycrystalline
formulation with a micromacro transition rule.
6.2.2. Models with two mechanisms and two criteria (2M2C)
This is an interesting particular case that can be an alternative to extend the capabilities of the
macroscopic models considered in this paper. The potential is written as a sum:

X X1 f 1 Xf 2

97

with two kinematic/isotropic hardening sets and two yield criteria:

f I k r  X I k  RI  k I

98

that leads to the plastic strain rate as

e_ p


oX1 1 oX2 2
n 2n
of 1 
of 

I
with n

of I
or


99

The above formulation can be applied directly but it is necessary to introduce coupled kinematic hardening effects in the free energy, using for instance:

wp

1XX
1X
I
J
C IJ a
:a

bI Q I rI 2


3 I
2 I
J

100

I
Each back-strain a
obeys an ArmstrongFrederick rule like:


a_ I n I 


3cI I oXI
X
2C II  of I

101

and a similar rule for the isotropic hardening variable rI , not specied here. However their effect is
coupled through the corresponding forces that play role in the potential:
I
X

owp 2 X
J

C IJ a
I

3
oa

J

RI

owp
bI Q I r I
or I

102

The above coupled approach of a non-unied plasticitycreep theory by Contesti and Cailletaud
(1989) exactly meets this framework. It sufces to chose I p and J v, and to consider (99) as a summation of a plastic strain rate and a viscoplastic (or creep) strain rate:

e_ in


e_ p e_ v k_

of p
of v _ p
_ v
v_
kn
vn


or
or



103

where k_ is an unknown plastic multiplier and v_ is a given function of the overstress f v , like a power
p
function v_ hf v =Kin . Contesti and Cailletaud (1989) still uses AF rule for both plasticity variables a

p
v
v
and X
and
viscoplastic
ones
a
and
X
.
However,
due
to
the
coupling
effect
in
(102)
the
plastic
con


sistency condition leads to a non-classical term in the plastic multiplier, like:

1
k_
Hp



r_  C vp a_ v : n p


104

where Hp depends on the kinematic and isotropic hardening variables of the plastic mechanism. More
elaborated forms of this approach were developed recently by Taleb et al. (2006).

1672

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

6.2.3. Models with two mechanisms and one criterion (2M1C)


It is also possible to combine several mechanisms (here only two) within only one single criterion.
An example is given, proposed initially by Zarka and Casier (1979). We have two mechanisms e 1 and
e 2 , and two sets of associated kinematic hardening variables, still obeying coupling relations like (102).

They are combined through the single criterion:

q
J 21 J 22  R  k J eq  R  k

105

with J I k r
X
k and with a single isotropic variable. The normality rule gives:


I

e_ p


of
k_
k_
or


1
2
J1 n
J2 n



J eq

k_ n


106

I
with n
oJ I =o r
.


In order to meet thermodynamic requirements, the rst term in the evolution equation of back
strains contains a factor J I =J eq as

a_ I



JI
3cI I _
I
k
n

X
J eq 
2C II 

107

In the rate-independent framework, applying the consistency condition f f_ 0, leads to



1
k_
n
:
r
h  

108

as usual, with h hX 1 hX 2 hR , where



J1
3c1 1
1
1
2
: C 11 J 1 n
n

X
C 12 J 2 n



J eq 
2C 11 


J
3c2 2
2
1
2
2 n

X : C 12 J 1 n
C 22 J 2 n



J eq 
2C 22 

hX 1

109

hX 2

110

and where the isotropic hardening contribution hR is not expressed here (free choice). In such a model
the plastic multiplier k_ does not correspond with the equivalent plastic strain rate as usual. An interesting property of such a model is that, for a non-zero determinant C 11 C 22  C 212 , the ratchetting stops
for a non-symmetric stress control, after a transient evolution, as exemplied by Fig. 15 (Besson et al.,

Fig. 15. Ratchetting response of the 2M1C model with a non-singular interaction matrix (from Besson et al., 2001).

1673

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

2001), obtained by using only linear kinematic hardening in (107), with c1 c2 0. The fact that a
linear hardening is here able to describe quite correctly the nonlinear kinematic hardening has to
be underlined.
Such an approach was proposed and exploited in the context of simplied inelastic analyses of
shot-peening or similar processes, like in Zarka et al. (1980), or Inglebert and Frelat (1989). Some generalizations are still under development. See for instance Taleb et al. (2006), Sai and Cailletaud (2007).
6.3. Micromacro transition approaches
These consist in making use of the basic crystalline plasticity equations, by writing directly in the
model the various slip systems that can be activated, for the various grain orientations considered in a
polycrystal RVE. Fig. 16 gives the operating scheme of such an approach, limiting itself (formally) to an
imposed macroscopic stress situation r
(scheme set up on a time increment), the output being the

macroscopic plastic strain e . The method uses two localization steps and two averaging steps:
p

macro $ grain \g" $ slip system \s"


Among numerous similar formulations, we follow here the one of Cailletaud (1992) and Pilvin (1994),
that is rather simple to use and is, however, sufciently precise (even if it retains a pronounced phenomenological character).
The transition from the macro level to the (average) stress level in each grain is done by the following localization rule, of the Krner (1961) type, but corrected with the so-called beta rule, which is valid for a polycrystal with grains of the same type, and with a macroscopically isotropic elasticity:

r g  r 2labg  B

111

where l is the shear elastic modulus, a an Eshelby based adjustment parameter (near a 0:5), bg a

state variable for each grain, analogous to the average plastic strain in the grain e pg , and B
the corre
sponding average:

r

X
g

g
cg r


ep

cg e pg

X
g

cg bg

112

in which cg is the volume fraction of each orientation considered. Krners elastic localisation rule,
which is known as too stiff (Zaoui and Raphanel, 1993), would be equivalent to replacing bg by

e pg and B
by e . The originality of Cailletaud and Pilvins approach is to continuously adapt this rule.


p

Therefore it becomes quasi-elastic in the regime of low plastic strains, and tends toward a tangent
type rule for higher strains, with a corresponding plastic accommodation effect. This is given by the
following evolution law for bg , which is very similar to a nonlinear kinematic hardening (combined


with a linear kinematic hardening):

macroscopic

constitutive equation

gp

slip constitutive
equation

integration

Fig. 16. Flow chart of a polycrystal material constitutive model based on crystal viscoplasticity.

1674

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

b_ g e_ pg  Dbg  depg ke_ pg k




113

D and d are global adjustment parameters, used to render the model similar to a self-consistent type
model, from numerical nite element analyses (Cailletaud and Pilvin, 1994; Pilvin, 1997).
The transition to the slip system level is done by using the resolved shear stress:

ss r g : ns  ls nsi rsij lsj

114

where ns and ls are, respectively, the normal to the slip plane and the slip direction. Although it is formally present, the index g of the grain is omitted for quantities associated with the slip system. In
slip plasticity Schmids criterion writes:

fs jss  xs j  rs  k 6 0

115

In slip viscoplasticity the slip rate is given here by a power law:

c_ s

 n
fs
Signss  xs
K

116

Let us note that it represents the summation on the grain of all the slips of the same direction. Here, xs
is a scalar kinematic hardening variable associated with each system and expressing the presence or
development of intragranular inhomogeneities (precipitates, inclusions, walls, dislocation cells, etc.).
Its evolution law obeys the nonlinear kinematic format already mentioned several times:
s
x_ s C c_ s  dx jc_ s j

117

r is a scalar variable expressing the size variation of elastic domain for each system. It obeys a nonlinear law that brings in the interactions between the various systems, a law of the type:

r_ s

~ r c
~_ r
hsr expqc

118

~r is the cumulative slip on the system r (c


~_ r jc_ r j). Also, hsr is the
or others of similar versions, where c
interaction matrix constructed from crystallographic informations and on studies made at lower
scales (by Dislocation Dynamics for instance). It has the dimension of the number of directions of systems that can be activated per grain, 12 for the octahedrals, 6 for the cubics, 18 in all in a CFC grain, etc.
We note that the material coefcients k (initial threshold of the Schmid criterion), n, K, C, d, q may
depend on the type of system to which the corresponding relation is applied (octahedral, cubic, or others for non-CFC crystals).
Once the cs quantities are dened by integration over a time increment, we then have to work back
to the average plastic strain of the grain:

e pg


cs ns  ls

119

s2g

and then to the macroscopic plastic strain by (112b). The scheme for a controlled macroscopic strain is
the same, although it requires an iterative solution with the macroscopic elasticity equation
r L : e  e .
p

Note: Although it is a matter of a micromechanical approach guided by the physical mechanisms,


the model thus developed could just as well be called macroscopic, as it is part of the multi-criterion
type approaches (Mandel, 1965) mentioned in Section 6.2. That is, the two localization rules applied
remain very close and refer only to average quantities on each grain, without precise geographical
localization. In effect, all of the above equations can be reduced to the use of a criterion f gs :

e_ p


X  fgs n
cs
K gs
g;s

gs
gs
f gs m
: r
X





gs
gs
X
xgs m
2labg 




X
g0

cg0 bg


gs
where m
ns  ls , for the slip systems of the grain g.


120
121

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1675

7. Some modications of AF type hardening to improve ratchetting modelling


7.1. Position of the problem
As was already mentioned in the original AF Report (Frederick and Armstrong, 2007), the simple
combination of kinematic hardening + dynamic recovery (delay term) was leading to ratchetting under
a non-symmetrical stress cycle.
Such a property can be considered as a good one, compared with linear Pragers rule or with multilinear/multi-surface models, provided the later always leads to cyclic elastic shakedown or to cyclic
plastic shakedown, never to ratchetting. On the other side, most mechanical experiments on polycrystallline metals (or even on single crystals) show signicant ratchetting effects, at least for sufciently
high applied mean-stresses.
As pointed out in Chaboche and Nouailhas (1989), ratchetting corresponds with a progressive accumulation of plastic strain, cycle-by-cycle (a positive strain increment dep under a positive meanstress), that resembles to a creep effect. For example, maximum strain evolution shows a primary
stage (decreasing ratchetting rate dep ) followed by a stationary stage (constant ratchetting rate), like
in the creep test. However, such an accumulation must not be confused with creep (induced by time
under a constant stress), because due to plasticity mechanisms during the unloading-reloading (not
specically inuenced by time).
Another experimental fact can be associated with ratchetting. It is the mean-stress cyclic relaxation
observed under a cyclic strain control, under more or less repeated cycles (non-zero mean-strain). For
purely elastic cycles there is elastic shakedown. When plasticity takes place cyclically (at each reversal), the initial mean-stress relaxes progressively, cycle by cycle. For low plastic strain amplitudes, the
relaxation is incomplete, stops asymptotically with some permanent mean-stress. For larger strain
amplitudes, the mean-stress relaxes completely to a zero or a very low value. As pointed out for instance in Jhansale and Topper (1973), mean-stress relaxation and ratchetting are different manifestations of the same process, resulting from different loading control conditions used.
Important to note, for practical applications, is the low amount of such a progressive strain accumulation when ratchetting takes place under stationary conditions. Each cycle we may have a strain
increase by an amount lower than dep 104 but it is its accumulation, as a stationary process, under
thousands of cycles, that may lead to the material and structural failure.
Another fact is that ratchetting can be observed at the global component scale, which obviously is
not recommended, due to the combination of two typical service loading conditions, primary one,
more or less a permanent (eventually constant) load, like internal pressure, and a secondary one (a
cyclic load), like service cycle, temperature cycles, vibratory loads, etc. Such structural conditions
are generally analysed by engineers, using limit analysis methods (like in Melans theorem) and shakedown envelopes, like Brees diagram. Such a diagram separates, in the loading space (secondary vs primary) the regimes that lead to elasticity, monotonic plasticity followed by elastic shakedown, cyclic
plastic shakedown, and cyclic ratchetting.
Most of the engineer shakedown analyses of components are based on using perfect plasticity
modelling, without hardening, often a too schematic material modelling. However, searching
improvements by introducing hardening is questionable for the following reasons:
 isotropic hardening plays role as a transient effect. After some accumulation of plastic strain,
increasing (or decreasing) the yield stress, materials generally show stabilisation and we come back
to the same question: ratchetting or cyclic plastic shakedown?
 using linear kinematic hardening can be not conservative because, at the material level, under cyclic
stress applied conditions, ratchetting is always prevented. Pragers rule, as well as multilinear (or
multi-surface) models always lead to shakedown (at the RVE level);
 on the other side, nonlinear kinematic hardening based on ArmstrongFrederick rule always lead to
ratchetting under a non-zero mean stress (or to elastic shakedown). Moreover, predicted ratchetting is generally much too high, at the material level, when compared with experimental results,
under uniaxial and multiaxial conditions;

1676

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

 combining one linear kinematic hardening and several nonlinear ones, like using a multi-kinematic
hardening (26), improves the results but not sufciently, and the ratchetting always stops after a
more or less extended period;
Some complex combinations, using multi-mechanisms/multi-criteria models (see Section 6.2)
could offer improved modelling capabilities. Also the polycrystalline aggregate models, with a scale
change and crystal plasticity at the level of each grain, may be a useful way. The model presented
in Section 6.3 for instance, separates kinematic hardening in two parts:
 the localisation rule, in which plays role an accommodation internal variable b obeying an AF rule.

This part could lead to ratchetting;
 the multiple slip mechanisms and associated hardening (at the grain level), which may be or not a
source of ratchetting (depending if it uses or not a dynamic recovery term like in (117)).
Since the work of Cailletaud (1987) such a solution has not been too much studied as a possible
compromise for modelling ratchetting in components, most probably in view of its numerical cost
(thousands of internal variables).
In the sections below, we focus on simple macroscopic kinematic hardening models, trying to improve their capabilities in terms of ratchetting modelling, without increasing too much the difculties
for their practical application to components.
7.2. Use of a threshold in the dynamic recovery term
Many modied forms of the ArmstrongFrederick rule have been studied in Chaboche and Nouailhas (1989) and Chaboche (1991), in the context of using a superposition like (26), with two normal
back-stresses (AF rule) and one (or two) obeying special rules. These additional models include:
(1) the linear kinematic hardening (LK) or a slightly nonlinear kinematic hardening (NLK), both
with a large tangent modulus;
(2) a Nonlinear Prager rule, with two variables, that introduces an asymptotically attained nonlinear relation between the mean-stress and the mean-strain;
(3) the model of Rousselier et al. (1985) (with two variables) in which the additional back-stress
saturates at a position depending on the mean-stress,
(4) a nonlinear factor of the additional back-stress (power function) in the dynamic recovery term,
as done in Henshall et al. (1987).
The main conclusions were as follows, compared with the experimental trends as observed on 316
Stainless Steel in uniaxial conditions:
 in order to quantitatively predict the order of magnitude of ratchetting, all the additional models
except Rousseliers must incorporate a stiff kinematic hardening effect, such that the monotonic
tensile curve is much too stiff, at a level not acceptable, for instance a tangent slope of 7500 MPa
at large strains;
 the relationship between the accumulated strain (for the same number of successive cycles) and the
applied mean stress is more or less linear, though existing experiments clearly show a nonlinear
dependency;
 except for NLK and the model of type 4, the negative ratchetting is complete: when the mean-stress
is suppressed, the previously accumulated strain is progressively recovering the very low value that
would have been attained without a mean-stress. This is in total contradiction with the experimental facts that show some negative ratchetting but much more limited;
 the Rousselier model delivers approximately acceptable accumulated strain, except a total stabilisation and the above mentioned linear relation. Moreover, it has been shown in Chaboche et al.
(1989) that the model predicts totally abnormal responses for combined cycles (low strain range
inside a larger one).

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1677

A modied model was also proposed by Bower (Bower, 1987; Bower, 1989). It is also based on the
use of two additional variables and the rule:

_ 2 C 1 e_ p  c X  Yp_ Y_ C 2 X  Yp_
X
1 





3 

122

Such a system is similar to the model of Rousselier et al. (1985). As indicated by Jiang and Sehitoglu
(1994), it is easy to show that stationary conditions lead to predict ratchetting arrest for any loading
condition, like with Rousseliers model, which is not really consistent with existing experiments.
Instead of using a nonlinear dependency (of the back-stress norm) in the dynamic recovery term, it
was proposed by Chaboche (1991) to introduce a threshold. This was justied by the commonly observed existence (Plnard and Fromont, 1987; Pelissier-Tanon et al., 1980), of a limit of accommodation,
both in terms of the mean-stress and of stress amplitude. The structure of the additional back-stress
evolution equation is then:



_ 2 C e_ p  n 1  X l X p_
X

3 
kXk 

123

Below the threshold, when kXk < X l , the back-stress evolves linearly (null dynamic recovery term).
Above, it evolves like in the standard ArmstrongFrederick rule, and attains the same asymptotic value
C=c, provided we choose n C=C=c  X l .
An advantage of this form is to preserve the existence of a closed form solution for any proportional
cyclic loading. Moreover, for low amplitudes of X
, the properties of the linear kinematic hardening are

recovered (that stops ratchetting after some transient), and the ones of the NLK model (AF rule) are
recovered at larger amplitudes.
By using the superposition of 2 classical NLK, one LK (low stiffness) and a fourth model with a
threshold, it was possible to qualitatively and quantitatively reproduce most of the observed facts
on 316 Stainless Steel (Chaboche, 1991)2:
 a correct tangent slope for the monotonic tensile curve, Fig. 17, as low as 1800 MPa at signicant
strain;
 acceptable shapes for the stabilised cyclic stressstrain loops under reversed or repeated strain control. Fig. 18 shows also the strain range memory modelling, to be compared with experiment on
Fig. 7;
 no ratchetting at all (except during transient conditions) for the lower mean-stresses (Fig. 19, in
which the experiment was taken from Goodman (1983));
 permanent ratchetting (under stationary conditions) for larger mean-stresses, with a nonlinear
dependency as observed experimentally (all under a constant applied stress amplitude);
 a small amount of negative ratchetting when the mean-stress is reduced to its lower value, 20 MPa
(Fig. 19).
Correct results were also obtained for INCO 718 superalloy, in Chaboche et al. (1991) (with 3
thresholds in that case). It must also be mentioned the correct prediction of the mean-stress relaxation
under a repeated cyclic strain control, condition that plays an important role in many design situations. In that case mean-stress does not relax completely for low plastic strain amplitudes, which cannot be captured by AF type models.
For 316 SS, the results above were obtained with only one threshold in one of the 4 back-stresses. It
is more common now to use the model in its generalised form, given in Chaboche (1994), with
X
RX
:

i


mi Xi

_ 2 C i e_ p  C i kXk  xi C i =ci
X
p_

3 
1  xi C i =ci
kX i k
i

124

2
In order to describe cyclic hardening effects it was also taken into account the model with strain range memorisation effect
indicated in Section 3.6, but that plays no role concerning the ratchetting effects.

1678

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

Fig. 17. Calculated response for the tensile test on 316 Stainless Steel at room temperature. 4 back-stresses model with 1
threshold (from Chaboche, 1991).

Exponent mi gives some more degrees of freedom. Parameter xi can be chosen independently for each
back-stress or can be a single valued quantity. It represents the ratio between the threshold and the
asymptotic value C i =ci for the back-stress.
7.3. Other models with similar properties
After the model with a threshold in the dynamic recovery term, it was proposed by Ohno and Wang
(1993) a slightly different approach, introducing a critical state of dynamic recovery. It takes two forms:
 the extreme case, corresponding to some multilinear model. The critical state is introduced by a
surface fi kX i k  C i =ci 0. The recovery term operates only when the back-stress attains this surface. In that case, under proportional loading, the back-stress rate vanishes. It writes:



_ i 2 C i e_ p  c Hfi n : ki Xi p_
X
i

 

3 

125

H is the Heaviside step function, n


the plastic strain rate direction, and k
X
=kX i k the unit direc
i
i
tion of the back-stress. The MacCauley bracket introduces the consistency condition for the critical
state fi f_ i 0. This rst version, at least in uniaxial conditions, recovers the classical multilinear
model of Besseling (1958). This version will never predict ratchetting under stationary conditions.

1679

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

Fig. 18. Simulated response for the 5 levels increasing strain-controlled loading on 316 Stainless Steel at room temperature. 4
back-stresses model with 1 threshold (from Chaboche, 1991).

 the smooth version, more realistic, replaces the Heaviside step function by a power function:


mi 

_ 2 C i e_ p  c kX i k
X p_
X
n
:
k
i

 i  i
3 
C i =ci
i

126

Though not necessary, the McCauley bracket hn


:k
i continues to operate in this smooth version,

i
and will play a signicant role in reducing ratchetting.
A systematic comparison study has been made in Chaboche (1994). The two models (124) and
(126) degenerate to the model (125) when mi ! 1 in OhnoWang or xi ! 1 in the model with a
threshold. The only difference is given by the special factor hn
:k
i in OhnoWang model, which van
i
ishes the dynamic recovery term during the beginning of a reverse plastic ow (when the direction of
plastic ow is more or less opposite with the previously stored back-stress). It therefore decreases further the ratchetting effect than in the version with a threshold.
In the case of smooth versions, it was shown that the responses of the two models can be adjusted
to become very similar, for monotonic, cyclic and ratchetting responses, both for uniaxial and multiaxial loadings, including non-proportional ones. It can be concluded that the two models, when using
the same number of back-stresses, offer very similar capabilities.
In fact, in the common practice with the OhnoWang model (see Bari and Hassan (2000), mi m is
taken as a constant for all the back-stresses and is adjusted from a set of ratchetting experiments (it is
the essential parameter that controls ratchetting). In correspondence, for the Chaboche model with a
threshold, (124), parameters mi m and xi x may also be taken as constants.
During the ninetees, there have been many work done in the literature in order to improve further
the modelling capabilities for both uniaxial ratchetting and multiaxial ratchetting conditions. These
are based on important and very complete experimental studies performed on various materials like
Carbon Steels (1018, 1020, 1026, 1045, 1070), 304 and 316 Stainless Steels, Copper (Bower, 1989;
Hassan and Kyriakides, 1992; Hassan et al., 1992; Xia and Ellyin, 1994; Jiang and Sehitoglu, 1994;

1680

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

Fig. 19. Calculated response for the ratchetting test with four mean-stresses (from Goodman, 1983) on 316 Stainless Steel at
room temperature. 4 back-stresses model with 1 threshold (from Chaboche, 1991).

McDowell, 1995; Corona et al., 1996; Portier et al., 2000). Those experiments include normal cyclic
conditions under stress control or strain control, uniaxial ratchetting, and multiaxial ratchetting, under tensiontorsion or tension-internal pressure.
Further modications of the OhnoWang or Chaboche models were proposed. For instance,
McDowell (1995) as well as Jiang and Sehitoglu (1996) introduced variations of the exponent mi of
OhnoWang model with the non-proportionality index hn
:k
i. Ohno and Abdel-Karim (2000) used

i
combinations of the OhnoWang model and AF rule. Voyiadjis and Basuroychowdhury (1998) used
a combination of Chaboche model (124) and of a bounding surface model. As discussed by Bari and
Hassan (2002), none of these modications deliver decisive improvements. Some multiaxial ratchetting experiments could be better reproduced, but this was generally detrimental for the uniaxial
ratchetting.
As indicated earlier in Chaboche and Nouailhas (1989), multiaxial ratchetting experiments generally
show lower ratchetting rates than uniaxial ones, for equivalent conditions on stress or strain amplitudes. From the modied model proposed by Burlet and Cailletaud (1987), it was then expected that
uniaxial ratchetting was the truly difcult condition to model. The Burlet and Cailletaud modication
of AF rule was consisting of a combination with a so-called radial return in the dynamic recovery term:

_ 2 C e_ p  c X p_  c  c X : n e_ p
X
1 
1 

 
3 

127

Using c1 c recovers exactly AF rule. Using c1 0 delivers the radial return model:

_ 2 C e_ p  cX : n e_ p
X


 
3 

128

In Burlet and Cailletaud (1987) it was shown that the model with radial return always stops multiaxial
ratchetting. Moreover, it can easily be checked that (127) always reduces to the AF rule in uniaxial

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1681

(proportional) conditions, independently of the c1 value. Therefore, the BurletCailletaud modication


can easily be used to adjust the (lower) multiaxial ratchetting, without changing at all the uniaxial
responses.

Fig. 20. Modelling of uniaxial and multiaxial ratchetting responses for 1026 and 1018 Carbon Steels at room temperature.
Classical multikinematic rules of Chaboche and OhnoWang (from Bari and Hassan, 2002).

1682

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

Such a combination was effectively performed with success by Bari and Hassan (2002), following a
way used by Delobelle et al. (1995). They used the 4 back-stresses of Chaboche model (3 AF rules and
one rule with a threshold) combined with the radial return model of Burlet and Cailletaud (the same
combination, with only 1 additional parameter d for the 4 back-stresses).

Fig. 21. Modelling of uniaxial and multiaxial ratchetting responses for 1026 and 1018 Carbon Steels at room temperature.
Simple combination of BurletCailletaud modication and the 4 back-stress Chaboche model, with 1 additional parameter
(from Bari and Hassan, 2002).

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1683

Fig. 20 shows the best results obtained (for the most signicant experiments on 1026 and 1018
Carbon Steels, from Hassan and co-workers) with the Chaboche model (9 parameters for back-stresses) and the OhnoWang model (21 parameters for back-stresses). The OhnoWang model was performing a little better (partly due to the use of 10 back-stresses instead of 4) but there was still
signicant overpredictions of ratchetting for multiaxial conditions, especially for the bow-tie or reverse bow-tie type of cycles.
In contrast, the above combination of the model with a threshold and of BurletCailletaud model
(only 10 parameters for the back-stresses), performed signicantly better (see Fig. 21). Using more
back-stresses and a threshold in each of them would certainly improve further uniaxial as well as multiaxial ratchetting predictions.
8. Multi-surface models
8.1. General
A signicantly different approach that can be used to describe monotonic and cyclic plasticity, is
the one initiated by Mroz (1967), with his multi-surface theory. Many other models have been developed following the same lines, including the two surface models. Such two-surface plasticity models
were used for instance in Krieg (1975), Dafalias and Popov (1976), Ohno and Kachi (1986), McDowell
(1985), Ellyin (1989), Moosbrugger and McDowell (1990). The main characteristics of this theoretical
framework are as follows:
(1) there is a yield surface f 0 and a bounding surface f 0:

f kr
X
k  ra 0



129

f k r  Y k  r 0
b



130

Both surface may move by kinematic hardening as well as isotropic expansion/contraction. The
yield surface is supposed to stay inside the bounding surface. In the original Mroz model, the
two-surface concept is extended by using a family of nested surfaces, but it will be explained
later (Section 8.3);
(2) the translation rule of the yield surface is given by the back-stress evolution equation, that generally obeys Mrozs assumption of collinearity:

_ l_ r
  r
X




131

 is the image point on the bounding surface and l_ a multiplier that will be determined by
where r

the consistency condition f f_ 0 (we limit ourselves to the case of rate-independent
plasticity);
 with the same
(3) the image point on the bounding surface is generally dened by the stress state r

outward normal than the current normal at r
on
the
yield
surface
(the
ow
direction
n
);


(4) the current hardening rate (or hardening modulus) is depending on the distance between the
current stress and the image stress on the bounding surface;
(5) the consistency condition applies for dening the multiplier l_ in the back-stress evolution
equation. Doing so, the current hardening modulus (or tangent plastic modulus) in the ow
equation is written freely as

e_ p




1
_ n
n
:
r
Kd   

132

it is not necessarily related with the kinematic hardening rule as in the other models considered
in this paper. Such a specic methodology has been recently qualied of uncoupled procedure by
Bari and Hassan (2001).
Compared with the theories presented previously in this paper, the two-surface or multi-surface approaches offer two important differences that enlarge their degrees of freedom and the corresponding

1684

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

capabilities: (i) the use of the distance d to dene the current hardening modulus, and (ii) the so-called
uncoupled procedure.
The other ingredients are not so signicantly different. For instance it has been demonstrated by
Marquis (1979) that the nonlinear kinematic hardening models implicitly dene a bounding surface
and the ArmstrongFrederick rule exactly obeys Mrozs translation rule. The generalisation (26) with
multiple elementary back-stresses, each of them obeying AF rule, was shown by Ohno and Wang
(1991) to correspond to a family of nested surfaces which simultaneous movement is obtained by
an image point concept.
On the other hand, the denition of the image point in the two-surface theories may be different
than in Mrozs model, but it may be with the drawback of possible intersections between the yield and
bounding surfaces, as shown for instance by McDowell (1989) for the TsengLee model (Tseng and
Lee, 1983). This model, which belongs to the bounding surface hypoplasticity models according to Dafalias (1986), allows the yield surface to translate in the direction of the deviatoric stress rate.
8.2. The DafaliasPopov model
The two-surface model of Dafalias and Popov (1976) benets of all the above characteristics,
including:
 the possible expansion of both the bounding and yield surfaces (isotropic hardening);
 the back-stress evolution with a direction x
that can be dened freely (Prager or Ziegler or Philipps

or Mroz rule);

_ l_ x
X



133

where l_ is to be determined by the consistency condition;


 the centre of the bounding surface evolves coupled with back-stress evolution by

_ X_  g_ m
Y




134

  r=k r
  r k is the direction of Mrozs rule;
where m
r





 there are two consistency conditions, f_ 0, which yields l_ in (133), and f_ that determines g_ in
(134);
 the generalised plastic modulus Kd in (132) is a decreasing function of distance d. The choice made
by Dafalias and Popov was

Kd K 0 hdin

d
din  d

135
m

where hdin is a given function like h a=1 bdin =2rb


;
 the quantity din represents the value of d at the initiation of a new loading process and measures
how far the material is from the state represented by the bounds at the beginning of plastic ow.
din changes at each reversal, but is constant during plastic ow; thus it is associated with the most
recent event of unloading-reloading;
 when the two surfaces are eventually approaching each other, this is at the current loading stress,
with the same outward normal (when using Mrozs rule (131)) and with a vanishing value for
K  K 0 , which, in the DafaliasPopov model, automatically ensures the non-intersection of the
two surfaces.
The advantage of using freely the distance d is clear in Eq. (135) because it allows to describe an
innite plastic tangent modulus at the onset of each new plastic ow. This property is not available
in the other coupled modelling frameworks. However there is also a difculty associated to it: the fact
that din is updated at each new loading may induce mathematical shortcomings (non-continuity of the
response for a vanishing reverse plastic ow) and possible overshootings of the new tensile response
after a very small compressive plastic incursion (see Chaboche, 1986).
Another inconvenience has been pointed out by Hassan and Kyriakides (1992), Hassan and Kyriakides (1994), concerning the ratchetting prediction with the DafaliasPopov model: ratchetting will always stop asymptotically. Modications of the rule have been proposed that incorporate an additional
translation of the bounding surface in the direction of ratchetting at the rate of ratchetting.

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1685

The additional degrees of freedom offered by the uncoupled procedure in Mrozs like multi-surface
and two-surface models has been illustrated by Bari and Hassan (2001), in the context of multiaxial
ratchetting modelling. They used the framework of DafaliasPopov model to dene the shape of the
stressstrain loops through the current tangent plastic modulus, Eq. (133) for the translation rule of
the bounding surface and the modication by Hassan and Kyriakides (1992) to improve uniaxial ratchetting simulations.
In Bari and Hassan (2001) they simply modied the translation rule of the current back-stress in
the DafaliasPopov formulation, using either ArmstrongFrederick rule, or the models by Voyiadjis
and Sivakumar (1994) (that combines Phillips and Tang, 1972 and Tseng and Lee, 1983 rules), Kaneko
(1981), Xia and Ellyin (1997). They also used the multikinematic model (26) with three normal AF
back-stresses (without threshold) and the Ohno version, with the factor hn
:k
i in the dynamic recov

ery term, but not the factor for critical state of dynamic recovery.
Surprisingly, even with the single AF rule (but the DafaliasPopov uncoupled procedure), the
ratchetting results are quite good, for uniaxial as well as multiaxial conditions. Using (26) with three
back-stresses with or without the Ohno factor delivers even better results, especially for the bow-tie
multiaxial cycles, when compared to existing data on Carbon Steels. Fig. 22, reproduced from Bari and
Hassan (2001) illustrates these results. Contrarily, all the other rules studied perform incorrectly when
used in this context.
Incidentally, an other concept of two-surface plasticity model, incorporating an elastic nucleus, was
proposed by Dafalias (1980), and a similar but different approach was followed by Hashiguchi with a
two-surface formulation and a sub-loading surface (Hashiguchi, 1980; Hashiguchi, 1989).
8.3. The Mroz model
The original Mroz model (Mroz, 1967) was built up in the context of a predened family of
nested surfaces. The inside smallest one is the yield surface. When it meets the subsequent one,
which now becomes the active surface, they will stay attached to each other (and tangent) at
the current loading stress. At each stage, the translation of the active surface is dened by Mrozs
rule (Mroz, 1967), with the image point positioned on the subsequent surface (actually inactive
and not translating). Doing that way, the formulation enforces the non-intersection of the successively active surfaces.
The surfaces do not meet asymptotically like the yield and bounding surfaces in the two-surface
versions. In Mrozs model, it is the active surface that denes the (constant) tangent plastic modulus.
Therefore, the model induces multilinear stressstrain responses.
It must be underlined that the translation movement of the active surface and the tangent plastic
modulus are not related each other (the property 5 mentioned above in Section 8.1). Such a property
offers additional degrees of freedom, not available in the nonlinear kinematic hardening rules (of AF
type or others). Qualied of uncoupled formulation by Bari Hassan, it was the point of departure of
the two-surface models and of many other generalisations of multi-surface models.
8.4. Models with discrete memory surfaces
There have been several modications of multi-surface Mrozs model. For instance Garud (1982)
proposed to dene the image point on the next surface by its intersection with the extension of the
stress increment. Mroz himself has exploited the formulation under several different ways (Mroz,
1981; Mroz, 1983), introducing for instance the memory of maximum pre-stress. Multiple memory
surfaces were also used in Trampczynski and Mroz (1992).
The author of the present paper also proposed a kinematic hardening rule based on the storage of
discrete memory surfaces. Partly based on a rule followed by Chu (1987), the objective of the formulation was to be able to reproduce a perfect (or quasi-perfect) closure of any small cycle performed
inside a larger one. The approach is based on the following main ingredients:
 the discrete memory surfaces are dened in the a space, a
being the back-strain (from which the

back-stress can be deduced by X

2=3C
a
,
see
Section
4.1.2);



1686

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

Fig. 22. Modelling of uniaxial and multiaxial ratchetting responses for 1026 and 1018 Carbon Steels at room temperature. BariHassan modication of DafaliasPopov uncoupled model and Chaboche or OhnoWang kinematic hardening models (from
Bari and Hassan, 2001).

 memorisation takes place under a specic loadingunloading condition (in a space, i.e. when
some reversal of plastic strain is taking place). Consequently, the drawbacks associated with the
updating rules, like in Chu model, can be avoided;

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1687

 there is an active surface and several memorised ones (each related to a previous reversal). Those
are nested surfaces, staying coincident each other at the stored point of unloading;
 the movement of the active surface, i.e. the current increment of the back-strain, is directed by
Mrozs rule, applied with the next stored memory surface, as being the current bounding surface;
 there is an erasing condition when the active surface merges (without intersection) the next
memory surface;
 the model makes a distinction between the rst memory surface (monotonic maximum pre-strain)
and the subsequent ones, to which are associated different parameters (1 and 2), like in the Masing
rule;
 the tangent plastic modulus at the unloading event is stored with the corresponding memory surface. The subsequent translation of the internal active surface takes that value into account, which
allows to obtain the perfect closure of the small cycles when the active and memory surface merge
together;
 in order to describe ratchetting effects, a small (and independently modiable) dynamic recovery
term is added to the translation rule of the surface.
A very similar discrete memory surface model was developed further by Ristinmaa (1995), without invoking the storage of the memory surfaces, but memorising the various segments (in the
stress space) corresponding with the previous evolution of the back-stress between the subsequent
unloading events, with the same loadingunloading condition as above. In fact a careful
examination shows that the rule of storage allows exactly the same set of results, except it is
written directly in the back-stress space and that ratchetting is introduced in a slightly different
way.
McDowell and Lamar (1989) also introduced a memory surface, associated with the long-range
back-stress, in order to model properly ratchetting behaviour.
8.5. A drawback of some multi-surface models
Concerning some multi-surface or two-surface models, related with the additional degrees of freedom offered by the so-called uncoupled procedure, an important drawback was demonstrated by
Jiang and Kurath (1996), for cyclic non-proportional loadings (circle paths in the deviatoric plane or
90 out-of-phase tension-shear experiment).
Under such loading conditions, the Mroz and Garud rules predict that plastic strain magnitude is
not a monotonic function of the stress magnitude as it should be. Moreover the plastic strain magnitude vanishes at the level where changing the active surface to the next one in Mrozs like multi-surface models. The response is then qualitatively and quantitatively very much depending on the
number of surfaces taken into account in the family.
The proof is true for any kind of description of the uniaxial tensile curve. In particular, two-surface
versions of Mrozs and Garuds models are giving similar results, with a non-monotonic evolution of
the plastic strain magnitude, and its vanishing value when the yield and bounding surfaces are
approaching each other.
By contrast, the AF rule and its generalisation as a multi-kinematic model (Chaboche et al., 1979)
always give a monotonically increasing plastic strain magnitude in out-of-phase loadings, which is a
property much desirable, and easy to observe in the available corresponding experiments.
This remark indicates that, associated with the additional degrees of freedom in the multi-surface
and two-surface approaches, there is also a signicant drawback for complex non-proportional conditions like out-of-phase loadings.

9. Concluding remarks
Difcult to conclude after such an already long overview of 30 years of constitutive equations
development. It was mainly an authors oriented point of view, supported by modelling works and
corresponding results obtained around the ONERA group.

1688

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

However some effort has been made, in reviewing other approaches, trying to point similarities and
differences. This include multi-criteria/multi-mechanisms concepts, multi-surface and two-surface
theories, as well as other unied viscoplastic constitutive models.
Several points connected with viscoplasticity were intentionally not presented or discussed in the
present paper, like:
 the thermodynamic framework for the constitutive modelling of plasticity and viscoplasticity, only
very much summarized, without explicit developments;
 temperature effects, aging effects, metallurgical and chemical evolutions (displacive and diffusive
transformations) only evocated very briey;
 the constitutive equations with back-stress evolutions for initially anisotropic materials (see some
words and references in Section 3.10);
 the modelling of yield surface distorsion effects, that needs to couple kinematic hardening with
Hills fourth rank tensor or to use higher order theories. Some attempt have been done in Helling
and Miller (1987), Imatani and Inoue (1987), Francois (2001), Vincent et al. (2002);
 the asymmetry of the initial and subsequent yield surfaces (strength differential), as observed for
instance in HCP polycrystals (a recent and interesting attempt was made in this domain by Cazacu
and Barlat (2004));
 soils and rocks mechanics context and specicities were not considered (pressure sensitive effects,
Cap Models like multi-mechanism models, . . .);
 porous plasticity models, like Gurson (1977) or Rousselier (1980) models, as well as Continuum
Damage Mechanics models (Lematre, 1992), coupled with plasticity should be a signicantly different subject;
 modern multiscale methodologies, for metals as well as for composites, are clearly out of the scope
of the present paper.
There are certainly many aspects that were incompletely developed in this paper, probably several
controversial points to be further discussed, with also many good references and other original models
forgotten. An impression at the end is the already extremely rich accomplishment of the eld, with
still large potentialities for further developments.
References
Armstrong, P.J., Frederick, C.O., 1966. A mathematical representation of the multiaxial Bauschinger effect. Report RD/B/N731,
CEGB, Central Electricity Generating Board, Berkeley, UK.
Arnold, S., Saleeb, A., 1994. On the thermodynamic framework of generalized coupled thermoelasticviscoplastic-damage
modeling. Int. J. Plasticity 10 (3), 263278.
Bari, S., Hassan, T., 2000. Anatomy of coupled constitutive models for ratcheting simulation. Int. J. Plasticity 16, 381409.
Bari, S., Hassan, T., 2001. Kinematic hardening rules in uncoupled modeling for multiaxial ratcheting simulation. Int. J. Plasticity
17, 885905.
Bari, S., Hassan, T., 2002. An advancement in cyclic plasticity modeling for multiaxial ratcheting simulation. Int. J. Plasticity 18
(7), 873894.
Barlat, F., Yoon, J.W., Cazacu, O., 2007. On linear transformations of stress tensors for the description of plastic anisotropy. Int. J.
Plasticity 23, 876896.
Benallal, A., Marquis, D., 1987. Constitutive equations for non-proportional cyclic elasto-viscoplasticity. J. Engng. Mater.
Technol. 109, 326336.
Benallal, A., Cailletaud, G., Chaboche, J.L., Marquis, D., Nouailhas, D., Rousset, M., 1985. Description and modelling of
nonproportional effects in cyclic plasticity. In: Proc. Multiaxial Fatigue Conf. Shefeld.
Benallal, A., Le Gallo, P., Marquis, D., 1989. An experimental investigation of cyclic hardening of 316 stainless steel and 2024
aluminium alloy under multiaxial loadings. Nucl. Engng. Design 114, 345353.
Besseling, J.F., 1958. A theory of elastic, plastic and creep deformations of an initially isotropic material showing anisotropic
strain-hardening, creep recovery and secondary creep. J. Appl. Mech. 25, 529536.
Besson, J., Cailletaud, G., Chaboche, J.L., Forest, S., 2001. Mcanique Non Linaire des Matriaux. Herms, Paris.
Bodner, S.R., 1987. Unied constitutive equations for plastic deformation and creep of engineering alloys. In: Miller, A.K. (Ed.),
Unied Constitutive Equations for Creep and Plasticity. Elsevier Applied Science, pp. 273301.
Bodner, S., Partom, Y., 1975. Constitutive equations for elastic-viscoplastic strain-hardening materials. J. Appl. Mech. 42, 385
389.
Boisserie, J.M., Gulin, P., Terriez, J.M., Wack, B., 1983. Applications of a hereditary constitutive law of discrete memory type. J.
Engng. Mater. Technol. 105, 155.
Bower, A., 1987. Some aspects of plastic ow, residual stress and fatigue due to rolling and sliding contact. Ph.D. dissertation,
Emmanuel College, Department of Engineering, University of Cambridge.

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1689

Bower, A.F., 1989. Cyclic hardening properties of hard-drawn copper and rail steel. J. Mech. Phys. Solids 37, 455470.
Burlet, H., Cailletaud, C., 1987. Modeling of cyclic plasticity in nite element codes. In: Desai, C.S. (Ed.), 2nd Int. Conf. on
Constitutive Laws for Engineering Materials: Theory and Applications. Elsevier, Tucson, pp. 11571164.
Busso, E.P., McClintock, F.A., 1996. A dislocation mechanics-based crystallographic model of a B2-type intermetallic alloy. Int. J.
Plasticity 12, 128.
Cailletaud, G., 1979. Modlisation mcanique dinstabilits microstructurales en viscoplasticit cyclique temprature variable.
Thse de Docteur-Ingnieur, Universit Pierre et Marie Curie, Paris 6.
Cailletaud, G., 1987. Une approche micromcanique phnomnologique du comportement inlastique des mtaux. Thse de
Doctorat detat, Universit Pierre et Marie Curie, Paris 6.
Cailletaud, G., 1992. A micromechanical approach to inelastic behaviour of metals. Int. J. Plasticity 8, 5573.
Cailletaud, G., Pilvin, P., 1994. Utilisation de modles polycristallins pour le calcul par lments nis. Rev. Eur. lments Finis 3
(4), 515541.
Cailletaud, G., Sa, K., 1995. Study of plastic/viscoplastic models with various inelastic mechanisms. Int. J. Plasticity 11, 991
1005.
Cailletaud, G., Kaczmarek, H., Policella, H., 1984. Some elements on the multiaxial behaviour of 316 l stainless steel at room
temperature. Mech. Mater. 3 (4), 333347.
Cailletaud, G., Depoid, C., Massinon, D., Nicouleau-Bourles, E., 2000. Elastoviscoplasticity with aging in aluminium alloys. In:
Continuum Thermomechanics: The Art and Science of Modelling Material Behaviour (Paul Germains Anniversary Volume).
Solid Mechanics and Its Applications. Kluwer Academic Publishers, pp. 7586.
Cazacu, O., Barlat, F., 2004. A criterion for description of anisotropy and yield differential effects in pressure-insensitive metals.
Int. J. Plasticity 20, 20272045.
Cernocky, E.P., Krempl, E., 1980. A theory of thermoviscoplasticity based on innitesimal total strain. Int. J. Solids Struct. 16,
723741.
Chaboche, J.L., 1977a. Sur lutilisation des variables dtat interne pour la description de la viscoplasticit cyclique avec
endommagement. In: Problmes Non Linaires de Mcanique. Symposium Franco-Polonais de Rhologie et Mcanique,
Cracovie, pp. 137159.
Chaboche, J.L., 1977b. Viscoplastic constitutive equations for the description of cyclic and anisotropic behaviour of metal s. Bull.
Acad. Polon. Sci., Srie Sci. Tech. XXV (1), 3339.
Chaboche, J.L., 1986. Time-independent constitutive theories for cyclic plasticity. Int. J. Plasticity 2, 149188.
Chaboche, J.L., 1989. Constitutive equations for cyclic plasticity and cyclic viscoplasticity. Int. J. Plasticity 5, 247302.
Chaboche, J.L., 1991. On some modications of kinematic hardening to improve the description of ratchetting effects. Int. J.
Plasticity 7 (7), 661678.
Chaboche, J.L., 1993. Development of continuum damage mechanics for elastic solids sustaining anisotropic and unilateral
damage. Int. J. Damage Mech. 2, 311329.
Chaboche, J.L., 1994. Modelling of ratchetting: evaluation of various approaches. Eur. J. Mech. A/Solids 13 (4), 501518.
Chaboche, J.L., 1996. Unied cyclic viscoplastic constitutive equations: development, capabilities and thermodynamic
framework. In: Krauss, A.S., Krauss, K. (Eds.), Unied Constitutive Laws of Plastic Deformation. Academic Press Inc., pp.
168.
Chaboche, J.L., Nouailhas, D., 1989. A unied constitutive model for cyclic viscoplasticity and its applications to various stainless
steels. J. Engng. Mater. Technol. 111, 424430.
Chaboche, J.L., Rousselier, G., 1983. On the plastic and viscoplastic constitutive equations, Parts I and II. Int. J. Pressure Vessel
Piping 105, 153158. 159164.
Chaboche, J.L., Dang-Van, K., Cordier, G., 1979. Modelization of the strain memory effect on the cyclic hardening of 316 stainless
steel. In: SMIRT 5, Berlin.
Chaboche, J., Nouailhas, D., Paulmier, P., Policella, H., 1989. Sur les problmes poss par la description des effets de rochet en
plasticit et viscoplasticit cycliques. La Rech. Arospatialei, French and English editions 1989-1, pp. 6379.
Chaboche, J.L., Nouailhas, D., Pacou, D., Paulmier, P., 1991. Modeling of the cyclic response and ratchetting effects on Inconel 718
alloy. Eur. J. Mech. A/Solids 10 (1), 101121.
Chan, K.S., Lindholm, U.S., Bodner, S.R., Nagy, A., 1990. High temperature inelastic deformation of the b1900+hf alloy under
multiaxial loading: theory and experiment. J. Engng. Mater. Technol. 112, 714.
Cheong, K.S., Busso, E.P., Arsenlis, A., 2005. A study of microstructural length scale effects on the behaviour of FCC polycrystals
using strain gradient concepts. Int. J. Plasticity 21 (9), 17971814.
Chu, C.C., 1987. The analysis of multiaxial cyclic problems with an anisotropic hardening model. Int. J. Solids Struct. 23 (5), 569
579.
Colak, O.U., 2004. A viscoplasticity theory applied to proportional and non-proportional cyclic loading at small strains. Int. J.
Plasticity 20, 13871401.
Contesti, E., Cailletaud, G., 1989. Description of creep-plasticity interaction with non-unied constitutive equations, application
to an austenitic stainless steel. Nucl. Engng. Design 116, 265280.
Corona, E., Hassan, T., Kyriakides, S., 1996. On the performance of kinematic hardening rules in predicting a class of biaxial
ratcheting histories. Int. J. Plasticity 12, 117145.
Culi, J.P., Nouailhas, D., 1993. New constitutive equations for single-crystal alloys: implementation in the SAMCEF code, and
applications. La Recherche Arospatiale 1993 (5), 1727.
Dafalias, Y.F., 1980. The concept and application of the bounding surface in plasticity theory. In: IUTAM Symp. on Physical NonLinearities in Structural Analyses, Senlis, France. Springer-Verlag.
Dafalias, Y.F., 1986. Bounding surface plasticity I: Mathematical foundation and hypoplasticity. ASCE J. Engng. Mech. 112, 966
987.
Dafalias, Y.F., Popov, E.P., 1976. Plastic internal variables formalism of cyclic plasticity. J. Appl. Mech. 98, 645.
Delobelle, P., 1988. Sur les lois de comportement viscoplastique variables internes. Rev. Phys. Appl. 23, 161.
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, 295330.

1690

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

Ellyin, F., 1989. An anisotropic hardening rule for elastoplastic solids based on experimental observations. J. Appl. Mech. 56,
499507.
El Mayas, N., 1994. Modlisation microscopique et macroscopique du comportement dun composite matrice mtallique.
Thse de doctorat, ENPC.
Estrin, Y., 1996. Dislocation-density-related constitutive modeling. In: Krauss, A.S., Krauss, K. (Eds.), Unied Constitutive Laws of
Plastic Deformation. Academic Press Inc., pp. 69106.
Estrin, Y., Mecking, H., 1984. A unied phenomenological description of work hardening and creep based on one-parameter
models. Acta Metall. 32, 5770.
Francois, M., 2001. A plasticity model with yield surface distorsion for non-proportional loading. Int. J. Plasticity 17, 703717.
Francois, D., Pineau, A., Zaoui, A., 1991. Comportement mcanique des matriaux: lasticit et plasticit. Herms, Paris.
Frederick, C.O., Armstrong, P.J., 2007. A mathematical representation of the multiaxial Bauschinger effect. Mater. High Temp. 24
(1), 126.
Freed, A. D., Walker, K., 1990. Model development in viscoplastic ratchetting. NASA Tech. Memo. NASA TM-102509, NASA.
Freed, A.D., Walker, K., 1993. Viscoplasticity with creep and plasticity bounds. Int. J. Plasticity 9, 213242.
Freed, A.D., Chaboche, J.L., Walker, K., 1991. A viscoplastic theory with thermodynamic considerations. Acta Mech. 90, 155174.
Garofalo, F., 1965. Fundamentals of Creep and Creep Rupture in Metals. Macmillan, New York.
Garud, Y.S., 1982. Prediction of stressstrain response under general multiaxial loading. Mechanical Testing for Deformation
Model Development. Vol. ASTM STP 765, pp. 223238.
Germain, P., 1973. Cours de Mcanique des Milieux Continus, vol. I. Masson, Paris.
Goodall, R.W., Hale, R., Walters, D.J., 1980. On constitutive relations and failure criteria of an austenitic steel under cyclic loading
at elevated temperature. IUTAM Symp. Creep in Structures, Leicester, UK. Springer-Verlag.
Goodman, A.M., 1983. Development of constitutive equations for computer analysis of stainless steel components. In: 4th Int.
Seminar on Inelastic Analysis and Life-Prediction in High Temperature Environment, Chicago.
Gulin, P., Stutz, P., 1977. Une nouvelle classe de lois de comportement dcrivant les grandes dformations viscolastoplastiques. Arch. Mech. Stosow 29, 1.
Gurson, A.L., 1977. Continuum theory of ductile rupture by void nucleation and growth: Part I yield criteria and ow rules for
porous ductile media. J. Engng. Mater. Technol. 44, 115.
Hall, R. et al, 2005. A thermodynamic framework for viscoplasticity based on overstress (VBO). J. Engng. Mater. Technol. 127,
369.
Halphen, B., Nguyen, Q.S., 1975. Sur les matriaux standards gnraliss. J. Mcanique 14 (1), 3963.
Hartmann, G., 1990. Comparison of the uniaxial behavior of the inelastic constitutive models of Miller and Walker by numerical
experiments. Int. J. Plasticity 6, 189206.
Hashiguchi, K., 1980. Constitutive equations of elastoplastic materials with elasticplastic transition. J. Appl. Mech. 102, 266
272.
Hashiguchi, K., 1989. Subloading surface model in unconventional plasticity. Int. J. Solids Struct. 25, 917945.
Hassan, S., Kyriakides, S., 1992. Ratcheting in cyclic plasticity. Part I: Uniaxial behavior. Int. J. Plasticity 8, 91116.
Hassan, S., Kyriakides, S., 1994. Ratcheting of cyclically hardening and softening materials. Parts I and II. Int. J. Plasticity 10, 149
212.
Hassan, S., Corona, E., Kyriakides, S., 1992. Ratcheting in cyclic plasticity. Part II: Multiaxial behavior. Int. J. Plasticity 8, 117146.
Helling, D.E., Miller, A.K., 1987. The incorporation of yield surface into a unied constitutive model. Acta Mech. 69, 923.
Henshall, G.A., Miller, A.K., Tanaka, J.G., 1987. Modeling cyclic deformation with the MATMOD-BSSOL unied constitutive
equations. In: Desai, C.S. (Ed.), 2nd Int. Conf. on Constitutive Laws for Engineering Materials: Theory and Applications.
Elsevier, Tucson.
Hill, R., 1966. Generalized constitutive relations for incremental deformation of metal crystals by multislip. J. Mech. Phys. Solids
14, 95102.
Ho, K., Krempl, E., 2002. Extension of the viscoplasticity theory based on overstress (VBO) to capture non-standard rate
dependence in solids. Int. J. Plasticity 18 (7), 851872.
Hu, Z.Q., Rausch, E.F., Teodosiu, C., 1992. Work-hardening behavior of mild steel under stress reversal at large strains. Int. J.
Plasticity 8, 839856.
Ikegami, K., Niitsu, Y., 1985. Effect of creep prestrain on subsequent plastic deformation. Int. J. Plasticity 1, 331345.
Imatani, S., Inoue, T., 1987. Material anisotropy induced by prior plastic deformation. In: Desai, C.S. (Ed.), 2nd Int. Conf. on
Constitutive Laws for Engineering Materials: Theory and Applications. Elsevier, Tucson, pp. 573580.
Inglebert, G., Frelat, J., 1989. Quick analysis of inelastic structures using a simplied method. Nucl. Engng. Design 116, 281291.
Jhansale, H.R., Topper, T.H., 1973. Engineering Analysis of the Inelastic Stress Response of a Structural Metal under Variable
Cyclic Strains, Vol. ASTM STP 519, p. 246.
Jiang, Y., Kurath, P., 1996. A theoretical evaluation of plasticity hardening algorithms for nonproportional loadings. Acta Mech.
118, 213234.
Jiang, Y., Sehitoglu, H., 1994. Multiaxial cyclic ratchetting under multiple step loading. Int. J. Plasticity 10, 849870.
Jiang, Y., Sehitoglu, H., 1996. Modeling of cyclic ratchetting plasticity. Part I: Development of constitutive relations. J. Appl.
Mech. 63, 720725.
Kanazawa, K., Miller, K.J., Brown, M.W., 1979. Cyclic deformation of 1% CrMoV steel under out of phase loads. Fat. Engng.
Mater. Struct. 2, 217228.
Kaneko, K., 1981. Plastic internal variables formalism of cyclic plasticity. Bull. JSME 24, 914.
Kawai, M., Ohashi, Y., 1987. Couple effect between creep and plasticity of type 316 stainless steel at elevated temperature. In:
Desai, C.S. (Ed.), 2nd Int. Conf. on Constitutive Laws for Engineering Materials: Theory and Applications. Elsevier, Tucson, pp.
967974.
Khan, A.S., Huang, S., 1995. Continuum Theory of Plasticity. A Wiley-Interscience Publication. John Wiley & Sons, New York.
Kocks, U.F., 1976. Laws for workhardening and low-temperature creep. J. Engng. Mater. Technol. 98, 7685.
Kocks, U.F., Argon, A.S., Ashby, M.F., 1975. Thermodynamics and kinetics of slip. Progress in Materials Science, vol. 19. Pergamon
Press, Oxford.

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1691

Krauss, A.S., Krauss, K. (Eds.), 1996. Unied Constitutive Laws of Plastic Deformation. Academic Press Inc..
Krempl, E., McMahon, J.J., Yao, D., 1986. Viscoplasticity based on overstress with a differential growth law for the equilibrium
stress. Mech. Mater. 5, 3548.
Krieg, R.D., 1975. A practical two-surface plasticity theory. J. Appl. Mech. 42, 641646.
Krner, E., 1961. Zur plastischen verformung des vielkristalls. Acta Metall. Mater. 9, 155161.
Lamba, H.S., Sidebottom, O.M., 1978. Cyclic plasticity for non-proportional paths. J. Engng. Mater. Technol. 100, 96111.
Lee, K.D., Krempl, E., 1991. An orthotropic theory of viscoplasticity based on overstress for thermomechanical deformation. Int. J.
Solids Struct. 27, 14451459.
Lematre, J., 1971. Sur la dtermination des lois de comportement des matriaux lasto-viscoplastiques. Thse de doctorat
detat, Univ. Paris VI.
Lematre, J., 1992. A Course on Damage Mechanics. Springer.
Lematre, J., Chaboche, J.L., 1985. Mcanique des Matriaux Solides. Dunod, Paris.
Lowe, T.C., Miller, A.K., 1986. Modeling internal stresses in the nonelastic deformation of metals. J. Engng. Mater. Technol.
(Trans. ASME) 108 (4), 365373.
Lubliner, J., 1973. On the structure of the rate equations of materials with internal variables. Acta Mech. 17, 109119.
Malmberg, T., 1990. Thermodynamics of some viscoplastic material models with internal variables. In: Proc. Int. Conf. Mech.
Phys. Struct. Mater., Thessaloniki.
Mandel, J., 1965. Une gnralisation de la thorie de la plasticit de W.T. Koiter. Int. J. Solids Struct. 1, 273295.
Marquis, D., 1979. Etude thorique et vrication exprimentale dun modle de plasticit cyclique. Thse de 3me Cycle,
Universit Pierre et Marie Curie, Paris 6.
Marquis, D., 1989. Phnomnologie et thermodynamique: couplage entre thermolasticit, plasticit, vieillissement et
endommagement. Doctorat dEtat, Universit Pierre et Marie Curie, Paris 6.
McDowell, D.L., 1983. On the path dependence of transfer hardening and softening to stable states under complex biaxial cyclic
loading. In: Desai, Gallagher (Eds.), Int. Conf. on Constitutive Laws for Engineering Materials. Elsevier, Tucson.
McDowell, D.L., 1985. A two surface model for transient nonproportional cyclic plasticity. J. Appl. Mech. 52, 298
308.
McDowell, D.L., 1989. Evaluation of intersection conditions for two-surface plasticity theory. Int. J. Plasticity 5, 2550.
McDowell, D.L., 1995. Stress state dependence of cyclic ratcheting behavior of two rail steels. Int. J. Plasticity 11, 397421.
McDowell, D.L., Lamar, A.B., 1989. Modeling ratchetting and anisotropic deformation with hardening dynamic recovery format
models. In: Khan, A.S., Tokuda, M. (Eds.), Proc. of Plasticity89, 2nd Int. Symp. on Plasticity and its Current Applications, Tsu,
Japan. Pergamon, Oxford, pp. 247251.
Miller, A., 1976. An inelastic constitutive model for monotonic, cyclic, and creep deformation: Part I. Equations development
and analytical procedures. J. Engng. Mater. Technol. 98 (2), 97105.
Miller, A.K., 1987a. The MATMOD Equations. Elsevier Applied Science Publishers. pp. 139219.
Miller, A.K. (Ed.), 1987b. Unied Constitutive Equations for Plastic Deformation and Creep of Engineering Alloys. Elsevier
Applied Science Publishers.
Miller, A.K., 1996. Improvements in the MATMOD equations for modeling solute effects and yield-surface distorsion. In: Krauss,
A.S., Krauss, K. (Eds.), Unied Constitutive Laws of Plastic Deformation. Academic Press Inc., pp. 153227.
Moosbrugger, J.C., McDowell, D.L., 1990. A rate-dependent bounding surface model with a generalized image point for
nonproportional cyclic plasticity. J. Mech. Phys. Solids 38, 627.
Moreno, V., Jordan, E.H., 1986. Prediction of material thermomechanical response with a unied viscoplastic constitutive model.
Int. J. Plasticity 2, 223245.
Mroz, Z., 1967. On the description of the work-hardening. J. Mech. Phys. Solids 15, 163175.
Mroz, Z., 1981. On generalized kinematic hardening rule with memory of maximal prestress. J. Mcanique Appl. 5, 241260.
Mroz, Z., 1983. Hardening and degradation rules for metals under monotonic and cyclic loading. J. Engng. Mater. Technol. 105,
113118.
Murakami, S., Ohno, N., 1982. A constitutive equation of creep based on the concept of a creep-hardening surface. Int. J. Solids
Struct. 18, 597609.
Nouailhas, D., 1990. Un modle de viscoplasticit cyclique pour matriaux anisotropes symtrie cubique. C. R. Acad. Sci. Paris t.
310 (Ser. 2), 887890.
Nouailhas, D., Cailletaud, G., Marquis, D., Dufailly, J., Bollinger, E., Lieurade, H.P., Ribes, A., 1983a. On the description of cyclic
hardening and of initial cold-working. Engng. Fract. Mech. 21 (4), 887895.
Nouailhas, D., Policella, H., Kaczmarek, H., 1983b. On the description of cyclic hardening under complex loading histories. In:
Desai, C.S., Gallagher, R.H. (Eds.), Constitutive Laws for Engineering Materials. Theory and Applications. Elsevier.
Ohno, N., 1982. A constitutive model of cyclic plasticity with a non-hardening strain region. J. Appl. Mech. 49, 721.
Ohno, N., 1990. Recent topics in constitutive modeling for cyclic plasticity and viscoplasticity. Appl. Mech. Rev. 43 (11), 283
295.
Ohno, N., Abdel-Karim, M., 2000. Uniaxial ratcheting of 316 fr steel at room temperature. Part II: constitutive modeling and
simulation. J. Engng. Mater. Technol. 122, 3541.
Ohno, N., Kachi, Y., 1986. A constitutive model of cyclic plasticity for nonlinear hardening materials. J. Appl. Mech. 53, 395403.
Ohno, N., Wang, J.D., 1991. Transformation of a nonlinear kinematic hardening rule to a multisurface form under isothermal and
nonisothermal conditions. Int. J. Plasticity 7, 879891.
Ohno, N., Wang, J.D., 1993. Kinematic hardening rules with critical state of dynamic recovery, Parts I and II. Int. J. Plasticity 9,
375403.
Ohno, N., Takahashi, Y., Kubawara, K., 1989. Constitutive modeling of anisothermal cyclic plasticity of 304 stainless steel. J.
Engng. Mater. Technol. 111, 106.
Pelissier-Tanon, A., Bernard, J. L., Amzallag, C., Rabbe, P., 1980. Evaluation of the resistance of type 316 stainless steel against
progressive deformation. In: Low-Cycle Fatigue and Life Prediction. ASTM STP, vol. 770, rminy, France.
Perzyna, P., 1964. On the constitutive equations for work-hardening and rate sensitive plastic materials. Bull. Acad. Polon. Sci.,
Srie Sci. Technol. 12 (4), 199206.

1692

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

Phillips, A., Tang, J.L., 1972. The effect of loading path on the yield surface at elevated temperatures. Int. J. Solids Struct. 8, 463
474.
Pilvin, P., 1994. The contribution of micromechanical approaches to the modelling of inelastic behaviour. In: Pineau, A.,
Cailletaud, G., Lindley, T. (Eds.), Fourth Int. Conf. on Biaxial/multiaxial Fatigue, vol. 1, ESIS, SaintGermain, France, pp. 3146.
Pilvin, P., 1997. Une approche inverse pour lidentication dun modle polycristallin evp. In: Actes du 3me Colloque National
en Calcul de Structures, Giens, pp. 207212.
Plnard, E., Fromont, C., 1987. Some mechanical properties relevant to non-linear elasticity materials. In: Yan, M.G., Zhang, S.H.,
Zheng, Z.M. (Eds.), ICM, vol. 5. Pergamon, Beijing, pp. 765771.
Portier, L., Calloch, S., Marquis, D., Geyer, P., 2000. Ratchetting under tensiontorsion loadings: experiments and modelling. Int.
J. Plasticity 16, 303335.
Prager, W., 1949. Recent developments in the mathematical theory of plasticity. J. Appl. Phys. 20 (3), 235241.
Rabotnov, Y.N., 1969. Creep problems in structural members. North-Holland.
Rice, J.R., 1970. On the structure of stressstrain relations for time-dependent plastic deformation in metals. J. Appl. Mech. 37,
728.
Ristinmaa, M., 1995. Cyclic plasticity model using one yield surface only. Int. J. Plasticity 11 (2), 163181.
Robinet, P., 1995. Etude exprimentale et modlisation du comportement viscoplastique anisotrope du zircaloy-4 dans deux
tats mtallurgiques. Thse de doctorat, Universit de Franche-Comt.
Robinson, D.N., 1978. A unied creep-plasticity model for structural metals at high temperature. Report ORNL-TM-5969, Oak
Ridge Nat. Lab.
Rousselier, G., 1980. Finite deformation constitutive relations including ductile fracture damage. In: Nemat-Nasser (Ed.), ThreeDimensional Constitutive Relations and Ductile Fracture. North-Holland Publ. Comp., Dourdan, pp. 331355.
Rousselier, G., Engel, J.J., Masson, J.C., 1985. Etude comparative de modles de comportement pour la simulation dessais en
traction-compresion sur tubes en acier inoxydable. Report Technical Report N 8, EDF-DER/GIS Rupture Chaud.
Sai, K., Cailletaud, G., 2007. Multi-mechanism models for the description of ratchetting: Effect of the scale transition rule and of
the coupling between hardening variables. Int. J. Plasticity 23 (9), 15891617.
Saleeb, A., Arnold, S., Castelli, M., Wilt, T., Graf, W., 2001. A general hereditary multimechanism-based deformation model with
application to the viscoelastoplastic response of titanium alloys. Int. J. Plasticity 17 (10), 13051350.
Schfer, I., 1997. Modr~elisation du comportement lasto-viscoplastique anisotrope des tubes de gaine du crayon combustible
entre zro et quatre cycles de fonctionnement en racteur eau pressurise. Thse de doctorat, Universit de FrancheComt.
Schmidt, C.G., Miller, A., 1981. A unied phenomenological model for non elastic deformation of type 316 stainless steel. Parts I
and II. Res. Mech. 3, 109129. 175193.
Taleb, L., Cailletaud, G., Blaj, L., 2006. Numerical simulation of complex ratcheting tests with a multi-mechanism model type. Int.
J. Plasticity 22 (4), 724753.
Tanaka, E., 2001. A model of nonproportional cyclic viscoplasticity. In: Lematre, J. (Ed.), Handbook of Materials Behavior
Models, vol. 2. Academic Press, pp. 368376 (Chapter 5.9).
Tanaka, E., Murakami, S., Ooka, M., 1985. Effects of plastic strain amplitudes on non-proportional cyclic plasticity. Acta Mech. 57,
167182.
Tanaka, E., Murakami, S., Ooka, M., 1987. Effects of strain path shapes on non-proportional cyclic plasticity. J. Mech. Phys. Solids
33, 559575.
Teodosiu, C., Hu, Z., 1995. In: Shen, Shan-Fu, Dawson, P. (Eds.), Simulation of Materials Processing: Theory Methods and
Applications. Balkema, Swansea, pp. 173182.
Trampczynski, W.A., Mroz, Z., 1992. Anisotropic hardening model and its application to cyclic loading. Int. J. Plasticity 8, 925
946.
Tseng, N.T., Lee, G.C., 1983. Simple plasticity model of the two-surface type. ASCE J. Engng. Mech. 109, 785810.
Valanis, K.C., 1980. Fundamental consequences of a new intrinsic time measure: plasticity as a limit of the endochronic theory.
Arch. Mech. Stosow 32, 171191.
Vincent, L., Calloch, S., Kurtyka, T., Marquis, D., 2002. An improvement of multiaxial ratchetting modeling via yield surface
distorsion. J. Engng. Mater. Technol. 124, 402411.
Voyiadjis, G.Z., Basuroychowdhury, I.N., 1998. A plasticity model for multiaxial cyclic loading and ratchetting. Acta Mech. 126,
1935.
Voyiadjis, G.Z., Sivakumar, S.M., 1994. A robust kinematic hardening rule for cyclic plasticity with ratcheting effects, Part II:
Application to non-proportional loading cases. Acta Mech. 107, 117136.
Walker, K.P., 1981. Research and development program for non-linear structural modeling with advanced time-temperature
dependent constitutive relationships. Report PWA-5700-50, NASA CR-165533.
Wang, J.D., Ohno, N., 1991. Two equivalent forms of nonlinear kinematic hardening. Int. J. Plasticity 7, 637650.
Watanabe, O., Atluri, S.N., 1986. Constitutive modeling of cyclic plasticity and creep, using an internal time concept. Int. J.
Plasticity 2, 107134.
Xia, Z., Ellyin, F., 1994. Biaxial ratcheting under strain or stress-controlled axial cycling with constant hoop stress. J. Appl. Mech.
61, 422428.
Xia, Z., Ellyin, F., 1997. A constitutive model with capability to simulate complex multiaxial ratchetin behaviour of materials. Int.
J. Plasticity 13, 127142.
Yaguchi, M., Yamamoto, M., Ogata, T., 2002. A viscoplastic constitutive model for nickel-base superalloy. PartI: kinematic
hardening rule of anisotropic dynamic recovery. Int. J. Plasticity 18, 10831109.
Yao, D., Krempl, E., 1985. Viscoplasticity theory based on overstress. the prediction of monotonic and cyclic proportional and
nonproportional loading paths of an aluminium alloy. Int. J. Plasticity 1, 259274.
Yoshida, F., Uemori, T., 2002. A model of large strain cyclic plasticity describing the Bauschinger effect and workhardening
stagnation. Int. J. Plasticity 18, 661686.
Yoshida, F., Uemori, T., Fujiwara, K., 2002. Elasticplastic behavior of steel sheets under in-plane cyclic tension-compression at
large strain. Int. J. Plasticity 18, 633659.

J.L. Chaboche / International Journal of Plasticity 24 (2008) 16421693

1693

Zaoui, A., Raphanel, J.L., 1993. On the nature of intergranular accommodation in the modeling of elastoviscoplastic behavior of
polycrystalline aggregates. In: Raphanel, J.L., Teodosiu, C., Sidoroff, F. (Eds.), Large Plastic Deformations, Mecamat, vol. 91.
Balkema, Rotterdam, pp. 2738.
Zarka, J., Casier, J., 1979. ElasticPlastic Response of Structure to Cyclic Loading. Pergamon Press, New York.
Zarka, J., Engel, J., Inglebert, G., 1980. On a simplied inelastic analysis of structures. Nucl. Engng. Design 57, 333368.
Zener, C., Hollomon, J.H., 1944. Effect of strain rate upon plastic ow of steel. J. Appl. Phys. 15, 22.
Zhan, Z.L., Tong, J., 2006. A study of cyclic plasticity and viscoplasticity in a new nickel-based superalloy using unied
constitutive equations. Parts I and II. Mech. Mater. 39, 6480.

S-ar putea să vă placă și