ARTICLE IN PRESS

International Journal of Mechanical Sciences 51 (2009) 453464

Contents lists available at ScienceDirect

International Journal of Mechanical Sciences

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

A micromechanical model for inelastic ductile damage prediction in

polycrystalline metals for metal forming
M. Boudifa a, K. Saanouni a,, J.-L. Chaboche b,a
a
ICD/LASMIS, University of Technology of Troyes, 10010 Troyes Cedex, France
b
ONERA DMSE/LCME, BP72, 92322 Chatillon Cedex, France

a r t i c l e in fo abstract

Article history: This paper deals with micromechanical modeling of ductile damage and its effects (coupling) on the
Received 8 September 2008 plastic behavior of FCC polycrystalline metallic materials. The fully coupled constitutive equations are
Received in revised form written in the framework of rate-dependent polycrystalline plasticity where a ductile damage variable
10 March 2009
has been introduced at a crystallographic slip system (CSS) scale in order to describe the material
Accepted 30 March 2009
degradation by initiation, growth and coalescence of microdefects inside the aggregate. Both, theoretical
Available online 16 April 2009
and numerical (FEA) aspects of the proposed micromechanical coupled model are presented. The ability
Keywords: of the obtained model to predict the plastic strain localization, due to the ductile damage effect, in the
Crystal plasticity classical tensile test is carefully analyzed. Application is also made to the fracture prediction in deep
Self-consistent models
drawing of a cylindrical cup using a thin sheet. Finally, some concluding remarks and perspectives are
Micromacro approach
pointed out.
Constitutive equations
Continuum damage mechanics & 2009 Elsevier Ltd. All rights reserved.
Finite element analysis

0. Introduction [18,19,20,40,41], this approach uses a scalar or tensorial damage

variables to represent the ductile defects initiation and growth
Sheet or bulk metal forming or machining processes undergo and their consequences on the other thermomechanical elds.
large plastic (or viscoplastic) deformations generating, often, This kind of CDM-based phenomenological approach has been
inelastic ow localization zones at the origin of ductile fracture widely used in the literature and applied to various engineering
by nucleation, growth and coalescence of microvoids and problems under various loading paths including the prediction of
microcracks. The constitutive equations currently used to simu- the ductile damage occurrence in metal forming and machining
late these processes are mostly written on macroscopic scale, processes [51,52].
using macroscopic constitutive equations generally accounting for However, in most of these works dedicated to the damage
the hardening (kinematic and isotropic) and the ductile damage- occurrence in metal forming using the CDM theory, the inelastic
induced softening. behavior (plastic or viscoplastic ow) is supposed to be incom-
Two main approaches are extensively used to describe the pressible despite the presence of the microvoids and microcracks.
ductile damage occurrence and its effect on the thermomechani- This means that the homogeneously distributed microvoids
cal behavior of the deformed metal. The rst one pioneered have a very small size which preserves the initial volume of the
by Gurson [33] and developed later ([32,49,56,57], among many deformed and damaged material. However, in a recent work the
others) is based on the micromechanisms of voids nucleation, authors proposed a modied formalism of CDM in order to allow
growth and coalescence. It uses the voids volume fraction as a the description of the damage-induced plastic volume variation or
scalar damage variable introduced on the plastic potential in compressibility [9,10,21,23]. This leads to using two scalar damage
order to model the effect of these voids on the plastic ow variables: one affects the deviatoric part of the stress and strain
including the associated volume variation. The second approach tensors and is volume preserving, and the second affects the
ignores the micromechanisms of the microdefects initiation hydrostatic stress and strain tensors and governs the induced
and growth and tries to represent damage effects on the overall volume variation or damage-induced compressibility.
elastic and inelastic thermomechanical behavior of the material. In this paper, we are interested in presenting a micromecha-
Based on the Continuum Damage Mechanics (CDM) theory nical model for polycrystalline materials accounting for different
kind of hardenings fully coupled with ductile crystallographic
damage. In fact, the ductile damage has undoubtedly microscopic
 Corresponding author. Tel.: +33 3 25 71 56 54; fax: +33 3 25 71 56 75. origins in the sense that the microvoids initiate around certain
E-mail address: saanouni@utt.fr (K. Saanouni). local heterogeneities as second-phase particles inside the grains

0020-7403/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijmecsci.2009.03.014
 ARTICLE IN PRESS
454 M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464

themselves or at the grain boundaries [45,46]. In this work, the the intergranular defects are neglected and all the transgranular
microscopic origin of the ductile damage is taken into account at inclusions are supposed to give rise to stress concentration
the crystallographic slip systems (CSS) scale inside the grains leading to the activation of CSS passing through these
(or single crystals) neglecting the creep damage which mainly inclusions. Due to this stress concentration, some of these
takes place at the grain boundaries and is governed by some brittle inclusions are broken giving rise to some microcracks or
thermally activated mechanisms. A scalar damage variable is microvoids nucleation. In other situations the matrix-inclusion
introduced at the CSS level together with an evolution equations interface is rst delaminated giving rise to the nucleation of other
derived from thermodynamical considerations. Starting from types of microvoids and microcracks. When the external loads
a micromechanical damage model initially proposed for cyclic continue to increase, the nucleated defects facilitate the slip along
loading paths (low cycle fatigue) by Saanouni and co workers the concerned CSS and generate voids growth under the effect of
[13,50], a new version is adapted for purely ductile damage both resolved shear stress and (probably) stress normal to the slip
under monotonic loading paths. Both, theoretical and numerical direction.
(FEA) aspects of the proposed micromechanical coupled model are To model these possible damage mechanisms in a very simple
presented. This leads to a useful micromechanical model, able way (no experimentally observed results concerning the effect of
to describe the inelastic behavior of FCC polycristalline materials, these microdefects in crystallographic slip are available), a single
including the ductile damage effects and accounting for possible scalar damage variable is introduced on each activated CSS, in
damage-induced plastic compressibility. The ability of the obtained order to represent the mean effects of these microvoids on the
model to predict the plastic strain localization in the classical tensile concerned CSS as proposed by Saanouni et al. [50] based on the
test until nal fracture, due to the ductile damage effect, is carefully original rate-dependent polycrystalline plasticity modeling work
analyzed. Application is also made to the fracture prediction in deep of Cailletaud [15]. Following this approach, some couples of state
drawing of a cylindrical cup using a thin sheet. variables are introduced at different scales in order to describe the
rate-dependent ow with both transgranular and intergranular
hardening, and are affected by the ductile damage effects.
1. Micromechanical modeling of the viscoplastic ow with Accordingly, and after the complete denition of the aggregate,
ductile damage the following state variables are introduced (see Fig. 2):

For metallic materials the Representative Volume Element  At the CSS scale two couples of state variables are introduced
(RVE), in the continuum sense, is considered a highly hetero- (qs, Rs) for the transgranular self-hardening and latent hard-
geneous aggregate of differently oriented grains. Each grain, taken ening, and (ds, Ys) for the ductile damage on this CSS.
as a single crystal, contains some Crystallographic Slip Systems,  At the granular scale only one couple of tensorial variables
potentially activated under the effect of the loading path applied g
b ; X g is introduced to describe the intergranular kinematic
to the overall aggregate at macroscopic level. This CSS activation hardening without damage effect since it was supposed that
under the effect of the dislocation movement denes the origin of no defects can take place along the grain boundaries (no creep
the plastic deformation as can be found in the pioneering book by damage) as assumed above.
Mura [46]. Fig. 1 shows the schematic description of the main  At the macroscopic scale (the overall aggregate) it is assumed
ductile damage micromechanisms as will be modeled in this work that the couple of the macroscopic small elastic strain and
[9]. The ductile fracture surface of a given material is represented Cauchy stress tensors Ee ; S are sufcient to represent the
in Fig. 1a. Fig. 1b denes schematically the representation of some elastic behavior supposed as homogeneous and isotropic at
grains of the aggregate containing some second-phase particles the RVE scale.
(inclusions) as well as some microvoids distributed inside
the grains (transgranular inclusions) as well as along the grain
boundaries (intergranular inclusions). Fig. 1c represents an According to this multiscale approach we have to follow the
isolated grain with its overall intergranular and transgranular following steps:
inclusions where some favourably oriented CSS are activated.
Since this work deals only with the ductile transgranular damage,
RVE Level
p e
Eij = Eij + Eij
ij

Grain Level
HOMOGENIZATION
LOCALIZATION

ijg ijg, p
g g
(ij, Xij)

CSS Level
n b

(qs, Rs)

s (ds, Ys) s, ds

Fig. 2. Schematization of the micromacro polycrystalline self consistent model-

Fig. 1. Schematic representation of the microstructure and failure mechanisms. ing.
 ARTICLE IN PRESS
M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464 455

1. Denition of the aggregate or RVE by giving the number of evaluation of the granular stress was the object of various works
phases and grains composing the RVE. In this simple micro (see the excellent book by Mura [46] and many recent works in
maco approach, monophasic FCC polycrystalline aggregates are the eld). One among these is Sachss model [53], which suggests
considered and only the crystallographic orientation of each that the stress in each grain of the polycrystalline aggregate sg is
grain is taken into account without describing neither the identical to the macroscopic one S. On the other hand, the Taylor
grains shape (supposed as spherical) nor their size and spatial model [55] corresponds to Voigt upper bound, and considers that
localization. Accordingly, the RVE is only dened by the total the granular plastic deformation is equal to the macroscopic one.
number of the FCC grains and their spatial orientations inside The same rule was proposed by Lin [42], which supposes uniform
the macroscopic frame using classical Euler angles. total deformations. The elastic accommodation model of Kroner
2. Localization step which consists of deducing the mesoscopic [38], based on the Eshelby [27] solution of spherical inclusion,
stress sg (strain g ) tensor applied on each grain from the leads to a linear relation between the macroscopic and micro-
macroscopic stress tensor S (or strain rate tensor D) applied scopic stresses (or strains). Meanwhile, there was the tangent
to the overall aggregate. The resolved shear stress ts and the approximation of the elasticplastic deformations proposed by
normal stress ssn for each CSS of each grain is deduced from sg Hill [35], which was simplied to secant estimation by Berveiller
using adequate projections on the normal and shear direction and Zaoui [7] for the case of isotropic elasticity and spherical form
of the slip plane. The superscripts (g) and (s) refer to the grains of the inclusions. For transition rules with viscous effects, we can
and the crystallographic slip systems, respectively. mention in the works [12,26,36,44]. In this work the so-called
3. Modeling step: for each level, the required constitutive parametric (b) scale transition rule introduced by Cailletaud and
equations should be deduced from the appropriate state and Pilvin [17] and Pilvin [47] is adopted. Assuming that the stress
dissipation potentials as will be shown later. tensor S is applied to the RVE, the resulting meso-stress tensor sg
4. The homogenization step which allows deducing the macro- on a given grain (g) of the aggregate is given by
scopic behavior from the local constitutive equations. !
XNg
g g g
sg S C f b b (1)
In the next sub-section, each of the above steps will be described. g1

However, let us note that this micromechanically based approach g

where f is the density of the grains with the same orientation and
as long as it neglects the description of the 3D initial texture of C is a phenomenological parameter characterizing the intergra-
the material (grains morphology: shape, size and spatial location) nular hardening which plays an important role on the plastic
including the second phases, as well as the 3D morphology (shape, accommodation inside the aggregate.
size and spatial location) of the initial microcracks/voids embedded, The resolved shear stress ts on each CSS of each grain (g) of the
cannot pretend to realistically describe the very complex mechan- aggregate results from sg thanks to Schmid projection [54]:
isms of the microcracks/voids initiation, growth and coalescence as
well as the complex interactions between a great number of these ts sg : ms (2)
microdefects. Accordingly, the present micro-damage approach s
where m is the well-known Schmid orientation tensor dened for a
should be seen as a mean eld approach. This means that only given CSS by the symmetrical part of the tensorial product of the
s
average plastic strain and (continuum) damage are considered for s
n times the slip direction on this plane ~
normal to the slip plane ~ b:
each orientation family (average in each grain and over all grains
s s s s
with the same orientation). n ~
ms 12~ n ~
b ~ b (3)
s
Similar to this resolved shear stress t , the normal stress to the CSS
1.1. Denition of the RVE (aggregate) sns dened by

In this work, a mono-phase aggregate of FCC grains is used to ssn sg : Zs (4)

dene the RVE. Only the crystallographic orientation of each family with
of grains is considered, neglecting the spatial location as well as the
s s
morphology (shape and size) of each grain. This means that the Zs ~
n ~
n (5)
grains have neither specic shape nor specic spatial location inside
is used. The introduction of this normal stress will be discussed
the aggregate; only the total number of the families of FCC grains,
and justied in the following section.
their volume fraction and their spatial orientations inside the
macroscopic frame are considered. The other physical entities like
second-phase particles, voids, grain boundaries, etc. that will be 1.3. Constitutive equations at different levels
mentioned later will be accounted for in by a phenomenological
manner and do not refer to a direct description in the modeling. The The model presented here is inspired by the work of Saanouni
possible evolution of the crystallographic orientation of the grains and Abdul-Latif [3] and Abdul-Latif and Saanouni [50] who carried
during the large inelastic strain with damage effect can be taken out modeling low cycle fatigue damage. In what follows, this idea
into account. This problem will be addressed in a forthcoming paper will be developed and applied to the ductile damage occurrence
and is not considered in the present work. So, the number of grains inside a given polycrystalline aggregate under the monotonic
and their spatial orientations are supposed as material parameters loading paths, assuming the small elastoplastic strain hypothesis
and should be dened for each case studied. for the sake of simplicity. The extension to nite polycrystalline
plasticity will be shortly discussed at the end of the present section.
1.2. Localization process
1.3.1. State potential and state relations
This stage of modeling is crucial since it is a question of nding The free energy is taken as the state potential dened at the
how the macroscopic Cauchy stress tensor S localizes inside the RVE scale by the following additive decomposition:
aggregate; i.e. what is the local stress tensor sg is resulting at each   gNg
X
grain (g) of the aggregate. For metallic material, it is common to rcEe ; bg ; qs ; ds rcel Ee ; DT ds rcgan bg ; qs ; ds (6)
use the mean eld approaches for the scale transition. The g1
 ARTICLE IN PRESS
456 M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464

The variable DT is a macroscopic measure of damage at the RVE 1.3.2. Dissipation analysis
scale, deduced from the local damage values for each CSS of each As mentioned above, the dissipative phenomena, i.e. hard-
grain (see later). As a rst approximation, the DT in the elastic ening, rate-dependent plastic yielding and ductile damage, are
potential is considered as a single parameter, which measures the dened at the grain and CSS levels. Using the same idea as in
macroscopic damage effect on the elastic behavior and not as [9,10,50] and taking into account the damage-induced volume
an internal state variable. In order to perform the effect of the variation from Boudifa [9] and Boudifa et al. [21] a yield function
damage on the mechanical behavior, the so-called effective state fsand a plastic potential fs are dened for each CSS (s) of each
e
~ ) and (q~ s , R~ s )) within the energy
variables are introduced ((E~ , S grain (g) of the aggregate by
equivalence, see [50]. Following this idea, the macroscopic s
X s s
effective variables are f jt~ s j ahs
~ sn i d  R~  t0 (15)
s

e
p S
E~ 1  DT Ee and ~ p
S (7) s 3 ag g
1  DT fs f s bqs R~ s s
X : X g fd d ; Y s (16)
4 cg
and similarly, at the CSS scale the following internal state where the damage part in the plastic potential fds is dened by
hardening variables we have
 s b1
1 Y  Y0 1
p s Rs fsd s (17)
s
q~ 1  d qs
s
and R~ ps (8) b1 S 1  d m
1d
~ sn i denotes
In the crystalline plasticity criterion equation (15), hs
The elastic part of the potential is supposed purely macro- the positive part of the crystallographic normal effective stress s ~ sn
scopically dened on the RVE scale by the classical isotropic dened above. This term is introduced into the classical Schmid
elastic free energy in which DT is a simple parameter: criterion in order to describe the plastic compressibility induced
h i by the overall damage occurrence. Let us note that several works
r cel Ee ; DT 1  DT 12ltrEe 2 mtrEe 2 (9) proposed similar approaches of non-Schmid criteria but without
any explicit reference to the damage effects (see for example
This leads to the following macroscopic stress tensor: [11,24,25,39]).
To insure the coherence of the yield function, by similarity
@cel
Sr 1  DT l trEe 1 2mEe  (10) with the effective state variables, the effective resolved and
@Ee
normal stresses are dened by
The inelastic part of the state potential is dened at the granular ts ssn
scale as t~ s ps ; s~ sn p
s
(18)
1d 1d
1 QX Ns XNs
Afterwards, hs ~ sn i plays the same role as the hydrostatic granular
rcgan bg ; qs ; ds C bg : bg s r
Hrs q~ q~ (11) ~ g projected by the normal tensor Zs dened by
stress: 13trs
3 2 s1 r1
Eq. (5). The normal stress s ~ sn is considered as a non-Schmid stress
where C is the kinematic (intergranular) hardening modulus contribution and the parameter a is introduced to balance the
introduced in Eq. (1), Q is the isotropic (transgranular) hardening effects of both the shear stress ts and the normal stress sns
modulus and Hrs is the symmetric square matrix dening the to the critical resolved stress. Let us note that if a 0 the
hardening interaction between the neighboring CSS of the same compressibility effect disappears and the classical Schmid yield
grain [31] and accounts for self-hardening and latent hardening. criterion is recovered. We also remark that these effects are
For the FCC crystals this matrix is of size 12  12 according to enhanced as the cumulated damages inside of the overall CSSs
the symmetry of octahedral CSSs. Note that no damage effect is of the same grain increases.
introduced in the rst term of the state potential (Eq. (11)) From the plastic potential equation (16), the evolution
according to the fact that no intergranular (or creep) damage is equations of the internal variables associated to the CSS level
considered. This leads to dene: can be derived:
s s
s @f l_ s
 the inter-granular kinematic hardening deviatoric stress q_ s l_ r p
s
1  bl_ (19)
@R 1d
tensor:
g s  s b
@can 2 g s
d_ l_
s @f
_s Y  Y0 1
Xg r g Cb (12) s l s (20)
@b 3 @Y S 1  d m
s
 the thermodynamic forces associated with the transgranular where l_ is the so-called viscoplastic multiplier, dened using
hardening and with the ductile damage at the CSS level: the framework of the time dependant plasticity [5,14] by
g p X
Ns
 s n
@can s f
Rs r Q 1d
s
Hrs q~ r (13) l_ (21)
@qs
r
K
where K and n are the viscosity parameters at the microscopic
scale (CSS). The material parameter b represents the non-linearity
g ss
@can 1 R~ q~ 1 Rs qs of the intergranular hardening, while S, Y0, b and m in Eq. (17)
Y s r s s (14)
@d 2 1  d 2 1  ds characterize the crystallographic ductile damage evolution. The
From these local equations, we can note that the size of granular plastic strain rate _ gp and the kinematic intergranular
g
the yield domain (Rs) is directly affected by the hardening hardening rate b_ are dened by
matrix. Also, the damage force (Ys) is only governed by the s !
sX
Ns
s @F
s XNs
l_ XNs
hardening stress (Rs) without any direct effect of the Cauchy stress _ gp l_ p signt s
ms
aZs
d
r
(22)
@sg 1d
s
tensor [50]. s1 s1 r1
 ARTICLE IN PRESS
M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464 457

g
sX
Ns
s @F s XNs
s metallic materials as can be found in the large literature devoted
b_  l_ _g
g p  ab
g
l_ (23) to this aspect. In fact, within this assumption in hand, the
s1
@X s1
kinematic non-linearities are easily separated from the material
where a is the non-linearity parameter and g_ s is the accumulated non-linearities thanks to the use of appropriately rotated inter-
slip rate on the crystallographic systems dened by mediate congurations in order to fulll the objectivity require-
s ment. In this work the kinematical framework of polycrystalline
s @f
s
l_
g_ s l_ p signts (24) plasticity proposed in [30] is used. Therefore, the RVE is
@ts 1d
s
transformed to its objective conguration by a standard kine-
Once again, in the case where a 0, we nd the initial model matic formulation, by using either the rigid body rotation coming
described in [10]. Moreover, a standard polycrystalline plasticity from the polar decomposition of the total transformation gradient
model is recovered if there is no coupling with damage (d 0). or the Jaumann rate of rotation dened by the anti-symmetric
More details of the whole modeling procedure can be found in [9]. part of the velocity gradient. Note that the macroscopic spatial
Note that, if the following modied orientation tensor is velocity gradient is additively decomposed into elastic and
introduced: inelastic parts. Only the inelastic contribution is assumed to arise
from crystallographic slip including both the plastic strain rate
X
Ns
M s signts ms aZs
r
d (25) given by Eqs. (26) and (28) and the plastic spin given by Eqs. (27)
r1 and (29).
the granular plastic strain rate tensor dened by Eq. (22) can be
rewritten as follows:
s 2. Numerical implementation
X
sNs
l_
_ gp ps M s (26)
s1 1d For the numerical simulation of large mechanical structures
Furthermore, the plastic spin which can be used to estimate the behavior including ductile damage occurrence, using the multi-
texture evolution (which is not considered in the present work) is scale or micromacro approach, we follow the classical way. It
given by consists in representing each integration (or Gauss) point of each
macroscopic nite element by an aggregate of a nite number
sX
Ns
of grains (RVE) to which the macroscopic strain tensor obtained
wgp g_ s A ms (27)
from the classical FEA is homogeneously applied. The mechanical
s1
s s s s answer of this RVE in term of macroscopic stress tensor is
n ~
where A ms 12 ~ n ~
b ~ b denotes the anti-symmetric part computed following the self-consistent model presented above
of ms . ([4,6,16,28,29,34,37,43,48] among many others).
In this work, the model developed above has been implemen-
1.4. Fields mean values at the aggregate scale ted in the Z-MAT package connected to the FE code ZeBuLoN [58].
For the global resolution scheme, the ZeBuLoN solver with static
The macroscopic plastic strain rate at the RVE (or aggregate) implicit (SI) standard procedure is used. For a more complex
level is a simple average and is given by calculations as the numerical simulation of metal-forming
processes, the dynamic explicit (DE) resolution strategy is used,
gNg
X g with ABAQUS/explicit FE code connected with the material
Dp f _ gp (28)
g1 models in Z_MAT using the interface Z_ABA that connects
ABAQUS to Z_MAT [8,58]. The DE resolution scheme is preferred
In the same way, the macroscopic plastic rotation rate (plastic since it avoids the iteration process as well as the calculation of
spin) of the RVE is expressed by the consistent tangent matrix. Starting from the principle of
gNg
X virtual power written for the velocity-based FE method leads,
g
Op f wgp (29) after the assembly operation, to the following global non-linear
g1 algebraic system (the matrix notations are used):
Similarly, the macroscopic damage rate is given by fF ext g  fF int g
MfUg (32)
0
gNg
X sX
Ns
T s
d_
_ g where [M] is the global consistent (lumped) mass matrix and Fint
D f (30)
g1 s1 and Fext are the well-known global internal and external forces
vectors, respectively, dened in the natural frame by
where Ng0 pNg is the number of grains where the crystallographic
damage has taken place (i.e. ds40). X Z e nXe o e
Finally let us mention that the macroscopic measure of the fF int g B T Jv dV (33)
e V
damage-induced plastic compressibility can be easily derived
from the above equations by ( )
X Z Z
! e
Ne T ff gJev dV Ne T fF e gJes dV
Ns _ s PrNs r fF ext g (34)
g g
gN
X g g gN
X g sX l a r1 d Ve GeF
trDp tr f _ p f ps a0 (31) e

g1 g1 s1 1d P P
where { e} ije is the macroscopic Cauchy stress tensor inside
Clearly, if a 0, there is no more volume variation and the the element (e) computed from the numerical integration of the
plastic ow is volume preserving. fully coupled micromechanical constitutive equations. [Ne] is the
Before closing this section let us shortly indicate that matrix of the interpolation functions of the element (e) and [Be]
the above-developed polycrystalline inelastic model can be easily is its spatial derivatives. {Fe} and {fe} are the applied boundary
extended to the large inelastic deformation by assuming that the and body forces, respectively. Jve and Jse are the determinants of
elastic strain component is still innitesimal. This widely used the volume and surface Jacobian matrices, respectively, linking
assumption is very sound when dealing with large deformation of the global and the natural frames.
 ARTICLE IN PRESS
458 M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464

Eq. (30) denes a highly non-linear system expressing the of grains, the theta method [22] have been used to integrate
mechanical equilibrium of the work piece at each time step. Over the constitutive equations, and used to validate the two local
each load (or time) increment of size Dt tn+1tn, the explicit integration strategies.
integration procedure can be summarized as follows:
M1 ffF ext g  fF ixt gg
fUg (35)
n n 3. Application of the model

_ _ Dtn1 Dt n Two 3D examples are solved to validate the proposed

fUg n1=2 fUgn1=2 fUgn (36)
2 micromacro model. The rst one concerns the uniaxial tensile
specimen subjected to the monotonic tension. The second deals
_
fUgn1 fUgn Dt n1 fUg (37) with the deep drawing of a thin sheet of a cylindrical cup known
n1=2
as swift test.
where the subscripts (n), (n+1/2) and (n+1) refer to the beginning,
the middle and the end of the time step, respectively. It is worth
noting that the time step size Dt should be efciently and 3.1. Tensile test for polycrystalline materials
automatically controlled to satisfy the accuracy and stability
requirements as discussed deeply in the ABAQUS/Explicit users A tensile test is carried out under 3D stress conditions using
manual. two different polycrystalline aggregates. The rst one called
As mentioned above, the self-consistent micromacro consti- Poly40, is composed of randomly distributed 40 grains (Fig. 3b)
tutive equations have been implemented into the Z_MAT package taken from Pilvin [47] representing a quasi-isotropic distribution
of the FE code ZeBuLoN. For the local integration, due to the large of FCC crystals. The second one called Poly24 (Fig. 3c) contains 24
number of variables, the constitutive equations are integrated grains taken from Bugat [13], corresponding to the 24 possible
using an explicit RungeKutta scheme with automatic time step orientations compared to the reference mark of ferrite.
control. For FCC polycrystals, 12 slip systems are considered for The sample (see Fig. 1a) has dimension 12  36  1 mm3 and is
each grain g, leading to ((12  3+6)  g+6+1 42  g+7) ordinary discretized by 432 hexahedral nite elements (C3D20R from
differential equations to be integrated at each time step and for ABAQUS element library) with 20 nodes and reduced integration.
each gauss point. The corresponding variables are: qs, ds, ls, bijg, The mesh and boundary conditions are also shown in Fig. 1a.
Eije and DT dened by the non-linear ordinary differential For the sake of simplicity, one row of elements in the thickness,
equations (Eqs. (19), (20), (23), (28) and (30)). For a small number 12 elements in the width and 36 elements in the length are used

Fig. 3. Pole gures of the (a) 40 and (b) 24 grain aggregate following (2 0 0) plane. (c) 3D Model for the tensile test sample.

Table 1
Material parameters for the two aggregates Poly40 and Poly24.

Aggregate Macroscopic Intragranular kinematic Intragranular isotropic hardening Damage

elasticity hardening

Poly40 E 200 GPa Cg 23082.3 MPa t0 400 MPa S 121.28 MPa, s 1, m 4.05
n 0.3 ag 30.13 Q 50 MPa a 0.1, Y0 0 MPa
b 6.851
h1 1, h2 1, h3 2, h4 1.5, h5 1,
h6 2.5
K 50 MPa, n 25

Poly24 E 200 GPa, n 0.3 Cg 30067.72 MPa, ag 26.75 s0 145 MPa S 0,846 MPa, s 1.194, m 47.858, a 0.5,
Y0 0 MPa
Q 50 MPa, b 74,78
hi 1, i 1,6
 ARTICLE IN PRESS
M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464 459

giving a constant size element of 1 mm3. Note that each of the

8 integration (or Gauss) points of each element is an aggregate
of Ng (Ng 24 for Poly24 and Ng 40 for poly40) FCC grains.
The materials parameters of both aggregates are given in Table 1,
determined on the basis of the reference curve shown in Fig. 4
with the help of an inverse approach discussed in [9]. This
reference curve is obtained using a fully coupled macroscopic
model including the damage occurrence leading to the plastic
deformation at fracture (or ductility) around 40%. A uniform
displacement is applied at the bottom side of the sample and the
specimen is xed at the opposite side.
The obtained equivalent local macroscopic stressstrain curves
are shown in Fig. 4, where the curves obtained with both
aggregates are compared with the reference curve.
For the two aggregates, the distributions of the macroscopic
equivalent plastic strain and damage are given in Figs. 5 and 6,
respectively, for different values of the imposed displacement.
Clearly, both computations give the diffuse and the localized
necking along the well-known shear bands without using any
geometrical or material initial perturbations. However, the Poly24
Fig. 4. Macroscopic stressstrain curves for the two aggregates compared to the
aggregate shows an early localized necking at U 2.5 mm while
reference stressstrain curve.
the Poly40 aggregate gives the localized neck around U 3 mm.

Fig. 5. Maps of macroscopic accumulated plastic strain for Poly40 and Poly24 at different stages of loading. (a) U 2.0 mm, (b) U 2.5 mm, (c) U 3.0 mm.
 ARTICLE IN PRESS
460 M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464

Fig. 6. Maps of macroscopic damage DT for Poly40 and Poly24 at different stages of loading. (a) U 2.0 mm, (b) U 2.5 mm, (c) U 3.0 mm.

lated crystalline plastic slip gs response inside two RVEs or

aggregates (Gauss Points):

 Gauss Point No. 6 of element 16 (denoted E16-PI6) which is

located inside the shear band
 Gauss Point No. 1 of element 351 (denoted E351-PI2) which is
located outside the localized shear band. In fact this point is
situated inside the diffuse neck but outside the localized neck.

For the RVE (E351-PI2) located inside the diffuse necking area, 7
CSS have been strongly activated until the complete saturation
Fig. 7. Maps of macroscopic compressible plasticity for Poly24.
of the hardening, which happens around t 20 s (Fig. 8a). This
saturated state is kept until t 40 s where the RVE is elasti-
Fig. 7 shows clearly that a macroscopic-induced com- cally unloaded when the localized neck takes place without any
pressibility is well described by the model and this underlines sensitive ductile damage occurrence within these CSSs. However,
the fact that compressibility follows the same localization pattern for the second RVE (E16-PI6) located fully inside the localized
as the accumulated plastic deformation. Also this, reinforces the neck zone (or shear band) the plastic slip increases strongly inside
fact that the microcavities are growing, i.e. expand in volume in some CSSs as can be seen in Fig. 7b. Due to the damage increase
the same direction where the damage and the plastic deformation inside these activated CSSs, the shear stresses decrease and far to
are localizing (Figs. 5c and 6c). zero as the damage increases.
In the case of the Poly24 aggregate, let us examine now the These observations on a microscopic scale reveal a good
local distribution of some mechanical elds in terms of accumu- concordance with the numerical results on a macroscopic scale.
 ARTICLE IN PRESS
M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464 461

Fig. 8. Evolution of the crystalline plastic slip jgsj and the crystalline ductile damage ds inside two RVEs located outside and inside the shear bands. (a) El351-Pil, (b) El16-Pi6.

v = 10 mm/sec constant and equal to 10 mm/s. The contact between the blank
Punch and the tools is modeled with a Coulomb friction model with a
constant friction coefcient Z 0.17.
Blank-holder The Dynamic Explicit scheme is used to solve the equilibrium
problem thanks to ABAQUS/Explicit connected to the Z_MAT
package in which our model has been previously implemented.
The results of this simulation are summarized in Fig. 10, where
the distribution of the macroscopic equivalent plastic strain and
the macroscopic damage are presented for different values of the
punch displacement. As expected, the plastic strain localizes
quasi-axisymmetrically while the damage localizes at the same
zones without any tendency to the axisymmetry as indicated in
Fig. 10a for the punch displacement u 6 mm. This heterogeneous
damage localization breaks the axisymmetry of the plastic ow
localization (Fig. 10b and c). The rst macroscopic crack appears
for u 8 mm and tends to propagate as the punch displacement
Die increases (Fig. 10b and c). The elements where the macroscopic
Blank damage reaches its critical value (DT 1) at all the associated
integration points (aggregates) have been removed from the
structure giving a discontinuous macroscopic cracks propagation
Fig. 9. Schematic illustration of Swift test. as the punch displacement increases.
Fig. 11 shows the evolution of the equivalent granular stresses
Accordingly, this micromacro approach is able to predict the for each of the 24 grains of not completely damaged RVE
macroscopic localization modes observed in the tensile tests (Gauss point 1 of the element no. 76). It illustrates the
starting from the highly heterogeneous state of the damage at the characteristic stress history until the nal fracture of all the
microscopic scale. grains of the aggregate for a macroscopic accumulated plastic
strain around 50%.
In terms of these examples, one can conclude that this
3.2. Sheet metal drawing of polycrystalline material micromacro approach to describe the ductile damage occurrence
inside the deformed parts gives very encouraging results. It is able
The polycrystalline damaged model used above for tensile to predict when and where macroscopic cracks can take place
test is now applied to simulate the damage occurrence in deep- inside the deformed parts as in metal forming by large plastic
drawing of FCC polycrystalline sheet using the non-isotropic deformation for example. In addition, introducing more physical
aggregate Poly24. When drawing cups out of single and aspects with a micromechanical model can lead to good
polycrystalline sheets, the tops of the cups do not show a constant prediction of induced anisotropy with respect to plastic ow in
height, as one would expect in the case of isotropic material the worked sheet. This can lead to an effective methodology for
behavior. Instead a periodic sequence of lower and higher points metal working by large plastic deformations where the ductile
can be observed, the so-called troughs and ears, which are caused damage could occur.
by the plastic heterogeneity of the blank.
In Fig. 9 the schematization of the deep drawing of cylindrical
cup (or Swift test) process is given. The blank is a circular sheet 4. Conclusion
of 73 mm diameter and 1 mm thickness. It is meshed using
hexahedral elements of C3D8R type with reduced integration This paper introduces a micromechanical damage model based
already used for the tensile test simulation. All the tools (the on crystal plasticity theory and continuum damage mechanics.
punch, the blank holder, and the die) are supposed as rigid bodies The modeling scheme focuses on the CSS scale giving a good
and meshed using the rigid shell elements C3D4 from the ABAQUS representation of macroscopic failure with or without damage-
elements library. The rate of displacement of the punch is kept induced volume variation. Finite element simulations were made
 ARTICLE IN PRESS
462 M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464

Macroscopic equivalent plastic strain Macroscopic Damage

Fig. 10. Macroscopic equivalent plastic strain and damage maps for different values of the punch displacement. Punch displacement: (a) u 6 mm, (b) u 8 mm, (c)
u 10 mm.

using polycrystalline materials to show the capacity of the

proposed methodology to predict when and where the ductile
damage can occur during metal forming. The rst results obtained
with this model are very encouraging even if more investigation
should be made in order to encourage this kind of approaches.
Clearly, the proposed approach cannot pretend to realistically
simulate the true and complex mechanisms at the microscale. It is
built up at some intermediate level, having the potential to raise
to the macro-scale some average information from the lower
scale, like the effect of the initial texture of the polycrystalline
aggregate in the context of a damage mechanics approach of
ductile fracture of or the micro-damage coupling effects playing
signicant roles in metal-forming applications.
Moreover, still with the same information from microscale,
the fact that the approach takes into account the plastic
compressibility effects, associated with the cavities growth, like
in Gursons macroscopic models, appears to be useful. Potentially,
it allows additional capabilities in terms of damage localiza-
tion and associated thickness changes in sheet metal-forming
Fig. 11. Evolution of the granular stresses inside a totally damaged RVE. context.
 ARTICLE IN PRESS
M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464 463

Several investigations can be made at different levels with [20] Chaboche JL. Damage mechanics. In: Milne I, Ritchie RO, Karihaloo B, editors.
further nite element calculations, where granular quantities may Comprehensive structural integrity, vol. 2. Amsterdam, Oxford: Elsevier-
Pergamon; 2003. p. 21384.
reveal a range of information about residual stresses, grains [21] Chaboche JL, Boudifa M, Saanouni K. A CDM approach of ductile damage with
deformation heterogeneity, texture evolution (grains rotation). plastic compressibility. International Journal of Fracture 2006;137(1):5175.
Particularly, the effect of the micro-damage on the initial texture [22] Chaboche JL, Cailletaud G. Integration methods for complex plastic consti-
tutive equations. Computer Methods in Applied Mechanics and Engineering
evolution during the large inelastic deformation should be 1996;133(12):12555.
investigated. On the other hand some experimental results [23] Chaboche JL, Saanouni K, Boudifa M. A CDM approach of ductile damage with
at the microscopic scale are required to better determine the plastic volume changes. ICTAM04, Krakov (Invited paper); 2004.
[24] Dao M, Asaro RJ. Non-Schmid effects and localized plastic ow intermetallic
material constants entering this kind of micromacro constitutive
alloys. Material Science and Engineering 1993;A170:14360.
equations. Then, we will be able to simulate working processes [25] Diard O, Leclercq S, Rousselier G, Cailletaud G. Distribution of normal stress at
such as stamping or deep drawing with about texture evolution grain boundaries in multicrystals: application to an intergranular damage
modeling. Computational Materials Science 2002;25(12):7384.
(grains rotation and reorientation) or induced anisotropy with
[26] Dingli JP, Abdul-Latif A, Saanouni K. Predictions of the complex cyclic
which much more accuracy cannot be easily obtained by classical behavior of polycrystals using a self-consistent modeling. International
macroscopic models. Journal of Plasticity 2000;16(34):41137.
[27] Eshelby JD. The determination of the elastic eld of an ellipsoidal inclusion
and related problems. Proceedings of the Royal Society A 1957;241:37696.
[28] Feyel F, Chaboche J-L. FE2 multiscale approach for modelling the elastovisco-
plastic behaviour of long bre SiC/Ti composite materials. Computer Methods
in Applied Mechanics and Engineering 2000;183(34):30930.
Acknowledgement [29] Forest S, Cailletaud G, Jeulin D, Feyel F, Galliet I, Mounoury V, et al.
Introduction au calcul de microstructuresElements of microstructural
mechanics. Mecanique & Industries 2002;3(5):43956.
The nancial support by the French government (CRCA) [30] Forest S, Pilvin P. Modelling nite deformation of polycrystals using local
of Champagne-Ardenne is gratefully acknowledged. Thanks to objectives frames. Zeitschrift fur angewandte Mathematik und Mechanik
1999;79:S199202.
ENSMP and ONERA for allowing the use of Z-ABA and ZeBuLon [31] Franciosi P. The concept of latent hardening and strain hardening in metallic
licencing. single crystals. Acta Metallurgica 1985;33:1601985.
[32] Gologanu M, Leblond JB, Perrin G, Devaux J. Recent extensions of Gursons
model for porous ductile metals. In: Suquet P, editor. Continuum Micro-
References mechanics, CISM Courses and Lectures 377. Berlin: Springer; 1997. p. 61130.
[33] Gurson AL. Porous rigid-plastic materials containing rigid inclusionsyield
[1] Abdul-Latif A, Dingli JP, Saanouni K. Modeling of complex cyclic inelasticity in function, plastic potential and void nucleation. In: Proceedings of the
heterogeneous polycrystalline microstructure. Mechanics of Materials 1998; conference on fracture, vol. 2, 1977. p. 35764.
30(4):287305. [34] Habraken AM, Duchene L. Anisotropic elasto-plastic nite element analysis
[2] Abdul-Latif A, Saanouni K. Damaged anelastic behavior of FCC polycrystalline using a stressstrain interpolation method based on a polycrystalline model.
metals with micromechanical approach. International Journal of Damage International Journal of Plasticity 2004;20(89):152560.
Mechanics (USA) 1994;3(3):23759. [35] Hill R. Continuum micro-mechanics of elastoplastic polycrystals. Journal of
[3] Abdul-Latif A, Saanouni K. Effect of some parameters on the plastic fatigue the Mechanics and Physics of Solids 1965;13:89101.
behaviour with micromechanical approach. International Journal of Damage [36] Hutchinson JW. Elasticplastic behaviour of polycrystalline metals and
Mechanics (USA) 1997;6(4):43352. composites (stressstrain curves of polycrystalline metals and composites
[4] Anand L. Single-crystal elasto-viscoplasticity: application to texture evolution using single crystal elastoplastic model). Royal Society/London/Proceedings,
in polycrystalline metals at large strains. Computer Methods in Applied Series A 1970;319(1537):24772.
Mechanics and Engineering 2004;193(4851):535983. [37] Kalidindi SR, Schoenfeld SE. On the prediction of yield surfaces by the crystal
[5] Asaro RJ. Crystal plasticity. Journal of Applied Mechanics 1983;50:921. plasticity models for fcc polycrystals. Materials Science and Engineering A
[6] Barbe F, Decker L, Jeulin D, Cailletaud G. Intergranular and intragranular 2000;293(12):1209.
behavior of polycrystalline aggregates. Part 1: F.E. model. International [38] Kroner E. Zur plastischen verformung des vielkristalls. Acta Metallurgica
Journal of Plasticity 2001;17(4):51336. 1961;9:15561.
[7] Berveiller M, Zaoui A. An extension of the self-consistent scheme to plasticity [39] Kurouda M, Kwabar T. Shear-band development in polycristalline metal with
owing polycrystals. Journal of the Mechanics and Physics of Solids 1979; strength-differential effect and plastic volume expansion. Proceedings of the
26:32544. Royal Society of London 2002;A458:224359.
[8] Besson J, Foerch R. Large scale object-oriented nite element code design. [40] Lematre J. A continuum damage mechanics model for ductile fracture.
Computer Methods in Applied Mechanics and Engineering 1997;142(12): Journal of Engineering Materials and Technology 1985;107:839.
16587. [41] Lemaitre J, Chaboche JL. Mechanics of solid materials. Cambridge, UK:
[9] Boudifa M. Multi-scale modeling of damaged polycrystalline materials. PhD Cambridge University Press; 1990.
Thesis, University of Technologie of Troyes, France, 2006. 223p [in French]. [42] Lin TH. Analysis of elastic and plastic strains of face centered cubic crystal.
[10] Boudifa M, Saanouni K, Chaboche JL. A Micromechanical Model For Inelastic Journal of the Mechanics and Physics of Solids 1957;5:1439.
Ductile Damage Prediction In Polycrystalline Metals. In: The seventh [43] Miehe C, Schotte J. Anisotropic nite elastoplastic analysis of shells:
international conference on Computation Structures Technology (CST2004), simulation of earing in deep-drawing of single- and polycrystalline sheets
Lisbon, Portugal, 79 September 2004. by Taylor-type micro-to-macro transitions. Computer Methods in Applied
[11] Brunig M, Obrecht H. Finite elasto-plastic deformation behavior of crystalline Mechanics and Engineering 2004;193(12):2557.
solids based on a non-associated macroscopic ow rule. International Journal [44] Molinari A, Canova GR, Ahzi S. A self consistent approach of the large
of Plasticity 1998;14(12):1189208. deformation polycrystal viscoplasticity. Acta Metallurgica 1987;35(12):298394.
[12] Budiansky B, Wu TT. In: Rosenberg RM, editor. Proceedings of the fourth US [45] Montillet F, Moussy F. Physique et mecanique de lendommagement. Les Ulis,
national congress of applied mechanics. Berkeley, CA. 1821 June 1962. New France: Edn. Les editions de physique; 1986.
York: American Society of Mechanical Engineers; 1962. p. 1175. [46] Mura T. Micromechanics of defects in solids. Second, Revised ed. Dordrecht,
[13] Bugat S. Comportement et endommagement des aciers austeno-ferritiques The Netherlands: Martinus Nijhoff Publishers; 1987. 587pp.
vieillis: une approche micromecanique. Doctorat Sciences et genie des [47] Pilvin P. The contribution of micromechanical approaches to the modelling of
materiaux, ENSMP-Centre des Materiaux P.M. Fourt, ENSMP, 2000. p. 290. inelastic behaviour of polycrystals. In: Societe Francaise de Metallurgie et de
[14] Cailletaud G. A micromechanical approach to inelastic behavior of metals. materiaux, vol. I, May 1994. p. 3145.
International Journal of Plasticity 1991;8(1):5574. [48] Raabe D, Roters F. Using texture components in crystal plasticity nite
[15] Cailletaud G. A micromechanical approach to inelastic behaviour of metals. element simulations. International Journal of Plasticity 2004;20(3):33961.
International Journal of Plasticity 1992;8(1):5573. [49] Rousselier G. Ductile fracture models and their potential in local approach of
[16] Cailletaud G, Forest S, Jeulin D, Feyel F, Galliet I, Mounoury V, et al. Some fracture. Nuclear Engineering and Design 1987;105(1):97111.
elements of microstructural mechanics. Computational Materials Science [50] Saanouni K, Abdul-Latif A. Micromechanical modeling of low cycle fatigue
2003;27(3):35174. under complex loadingsPart I. Theoretical formulation. International
[17] Cailletaud G, Pilvin P. Utilisation de modeles polycristallins pour le calcul par Journal of Plasticity 1996;12(9):111121.
elements nis. Revue Europeenne des Elements nis 1994;3(4):51541. [51] Saanouni K, Chaboche JL. Computational damage mechanics. Application to
[18] Chaboche JL. Continuous Damage Mechanics: a tool to describe phenomena Metal Forming; 2003.
before crack initiation. Nuclear Engineering and Design 1981;64:23347. [52] Saanouni K, Cherouat A, Mariage JF, Nesnas K. On the numerical simulation of
[19] Chaboche JL. Continuum damage mechanics. I-General concepts. II-Damage 3D forging of some complex parts considering the damage occurrence. In: 4th
growth, crack initiation, and crack growth. ASME, Transactions, Journal of international ESAFORM conference on material forming; Liege; Belgium,
Applied Mechanics 1988;55:5972. 2325 April 2001. p. 5936.
 ARTICLE IN PRESS
464 M. Boudifa et al. / International Journal of Mechanical Sciences 51 (2009) 453464

[53] Sachs G. Zur Ableitung einer Fliessbedingung. Zeitschrift des Vereins [56] Tvergaard V. On localization in ductile materials containing spherical voids.
deutscher Ingenieure 1928;72:734. International Journal of Fracture 1982;18(4):23752.
[54] Schmid E, Boas W. Kristallplastizitat mit besonderer Berucksichtigung der [57] Tvergaard V, Needleman A. Analysis of the cup-cone fracture in a round
Metalle. Berlin: Springer; 1935. tensile bar. Acta Metallurgica 1984;32(1):15769.
[55] Taylor GI. Plastic strain in metals. Journal of the Institute of Metals [58] ZeBuLoN. Zset/Zebulon: Developer Manual, 2005.
1938;62:307.