Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Mechanics
http://ijd.sagepub.com/
Published by:
http://www.sagepublications.com
Additional services and information for International Journal of Damage Mechanics can be found at:
Subscriptions: http://ijd.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://ijd.sagepub.com/content/20/6/845.refs.html
What is This?
ABSTRACT: FEM results of softening materials are well known to show patholog-
ical mesh dependency. The main goal of this work is to revisit and propose efficient
nonlocal damage gradient enhanced formulations able to avoid mesh dependency in
the context of elastoplastic damage models with destination to industrial applica-
tions. This formulation is presented and studied for simple tension tests, with various
spatial discretizations. Numerical aspects and implementation in ABAQUS-standard
environment are discussed. The structure of the nonlocal element needed for those
formulations is presented. For a given set of meshes, the ability of the proposed
formulation to control the size of the necking zone is studied. In the same time the
independence of the global dissipation to the mesh size is checked. Theoretical and
practical limits of the proposed approach are highlighted.
INTRODUCTION
a certain value of damage is reached. For higher values, the damage distri-
bution localizes inside a single element width row corresponding to the mac-
roscopic crack thickness. Finally, crack propagation defined as the location of
totaly damaged points, seems to remain strongly dependent of the discretiza-
tion aspects. This phenomenon can be assumed considering that nonlocality
acts only during necking. Then cracking is considered as a purely local phe-
nomenon and takes place in the smaller spatial zone defined by the spatial
discretization, i.e., a single element width row. Some numerical results are
presented in this sense with a simple 2D tensile test. Extension of this study to
the 3D case is under progress and should be presented in a coming work.
The constitutive equations of the present work are developed in the frame-
work of the thermodynamics of irreversible processes under the isothermal
condition and assuming the full isotropy. For the sake of simplicity and to
enable the comparison with existing works, the kinematic hardening is not
considered (full developments are given in Sornin, 2007). Accordingly, the
following state variables are used:
. (ee, ) for the elastic-plastic flow, assuming strain rate additivity
("_ "_e "_p )
. (r, R) for the nonlinear isotropic hardening
. (D, Y) for the ductile damage
In this work we are concerned with an elastoplastic constitutive equations
exhibiting a positive hardening followed by a negative hardening exclusively
due to the ductile damage effect. Therefore, we assume that the regulariza-
tion should be applied to the damage variables namely D in the strain space
or Y in the stress space. Different couplings of the model with regularized
damage variables are explored. First, nonlocality is introduced using a reg-
ularized damage driving force Y. In a second time, a regularized damage
variable D is introduced. Effects and limits of these approaches are dis-
cussed. The elastoplastic model is limited to the case of non linear isotropic
hardening fully coupled to damage.
Constitutive Equations
In this work the damage effect is introduced in the state potential accord-
ing to:
1 1
hD "e : : "e Qr2 2
2 2
expressed in the strain space where KId Id 2I , and Q are, respec-
tively, the elastic operator and the isotropic hardening modulus of the
damaged material. K and are the Lames coefficients. Consequently,
state relations are:
hD : "e 3
R hDQr 4
1 e e 1 2 @hD
Y " : : " Qr 5
2 2 @D
Assuming the plastic incompressibility, the following yield function f and
plastic potential F are chosen:
r
kk R 3
f y with kk S:S 6
hD 2
1b R 2 1 S Y Y0 s1
F, R, Y, hD f 7
2 Q hD hD s 1 S
Where the second-order tensor S is the deviatoric part of the Cauchy stress
tensor , b the isotropic hardening non linearity coefficient and (S, s, , Y0)
the four constants of the Lemaitre-like damage constitutive equations
(Saanouni and Chaboche, 2003). Y* will be specified later. Using the nor-
mality rule, the dissipation analysis gives the rate relations:
"_ p N_ a
@F _
r_ _ 1 br b
@R hD 8
@F _ Y Y0
D_ _
5 4s c
@Y hD S
Where N is the outward normal to the yield surface in the stress space
defined by:
@f 3 S
N 9
@ 2hD kk
The equivalent plastic strain rate is also defined by:
r
2 p p _
p_ "_ : "_ 10
3 hD
The plastic multiplier _ is given by the consistency condition, namely f 0
and f_ 0, as discussed later.
The local damage variable D issued from (22) is strictly increasing with the
plastic multiplier and is no more limited by any threshold. Consequently, to
avoid spontaneous creation of energy we also assume that D is strictly
increasing:
dD 0 at any time 24
For some authors like (Geers, 2004) and (Engelen et al., 2003) this condition
is the consequence of a damage criterion tested at each integration point.
This criterion-based approach is also proposed for the Y model by (Liebe
et al., 2001). In the next section the proposed nonlocal damage model is
implemented into ABAQUS/standard framework using both UMAT and
UEL users subroutines and its ability to restore the wellposedness of the
IBVP studied.
NUMERICAL ASPECTS
Variational Formulation
Galerkin method the approximation of the nodal fields for a given element
(e) are:
e
fu g Neu fuek g fue g Neu fue
k g
e e e e 27
Z 5 NZ 4 fZk g Z 5 NeZ 4 fZk g
e
with Neu and NeZ the displacement and Z interpolation functions, respec-
tively, inside the element e with k the number of nodal unknowen. Their
associated derivatives are:
@Nu @NZ
Bu BZ 28
@xj @xj
The vector Zk represents the nonlocal nodal variable. The elementary
expression of the weak forms becomes:
8 " #
>
> ! X
npg
jT j X npg X
npgf
>
> e e
F hu i j Bu f gJ j T ~j
j Nu ff gJ i T ~i
i Nu fT gJs
>
<
j1 j1 i1
" npg #
>
> X X
npg
>
> e e j T j j T j j T j
: H hZ i
> j !BZ BZ NZ hNZ i fZk gJ j NZ Z J
j1 j1
29
After assemblage on the whole structure and linearization at time tn+1, the
system (Equation (30)) is iteratively solved by an implicit NewtonRaphson
scheme.
8
s h@Fu, Zn1 is h is
> @Fu, Zn1
< Fu, Zn1 @u u k @Z
Zk 0
h is h is 30
:
Hu, Z s @Hu, Zn1 u @Hu, Zn1 Z 0
>
n1 @u k @Z k
where s indicates the global iteration number. Considering only the varia-
tion of variables and Z, the components of the elementary tangent matrix
are given by:
( )
@Fu, Zn1 X
npg
d j
j T
Kuu j Bu Buj :J 31a
@u j1
d" j
X
npg
@Fu, Zn1 j T d j
Kuz j Bu NZj :J 31b
@Z j1 dZ
( )
@Hu, Zn1 X
npg
j dZj
Kzu j NZ Buj :J 31c
@u j1
d" j
@Hu, Zn1
Kzz
@Z
X
npg
j T j j dZj j
j !BZ BZ hNZ i 1 fNZ g J 31d
j1 dZ j
As discussed in [36] for nonlocal formulation the new element must ver-
ify the BrezziBabuska condition (Babus ka and Narasimhan, 1997) to
avoid any stress instability and locking during loading. In the general
case, interpolation degree of the gradient term rui must be one degree less
than the displacement one. This choice is discussed in (Simone et al., 2004)
to ensure convergence stability of the numerical scheme. For the sake of
simplicity, the shape functions of the regularized variable (Niz ) are chosen
linear. A smart solution can be founded using a P1+/P1 subparametric
element. The displacement field is quadratically interpolated using an addi-
tive central node. This element technology is largely discussed in (Perchat,
2000) for the case of fluids mechanic elements implying mixed velocity/
pressure formulations. The shape functions for this four nodes triangular
element are:
0,
0, N1u 1
13 N4u N1z 1
1,
0, N2u 13 N4u N2z
32
0,
1, N3u
13 N4u N3z
13 ,
13 , N4u 27
1
Numerical integration of the two residuals (Equation (29)) over the refer-
ence element is performed using the Gauss method with three IP. This
corresponds to an over integration for the nonlocal field and exact integra-
tion for the displacement field (See Figure 1). This element is implemented in
the ABAQUS framework using a User ELement (UEL) subroutine. An
updated lagrangian scheme coupled to the Jaumann rate is used with this
element. A HughesWinget (Hughes and Winget, 1980) assumption is
retained for the computation of the rotation increment over the step time.
Finite transformation aspects of this element are detailed in (Sornin, 2007).
This element is applied to the nonlocal damage gradient model in the
next section.
(3)U1,U2,Zbar
(4)U1,U2
(1)U1,U2,Zbar (2)U1,U2,Zbar
Figure 1. The P1+/P1 non local element.
This section describes the numerical scheme used for integration of the
nonlocal fully coupled constitutive equations of section Regularized
Damage Variable. A standard elastic prediction and plastic correction algo-
rithm is used for the computation of the stress tensor together with the other
state variables of the model.
Elastic prediction: Elastic prediction drives in our case to the further expres-
sion of the trial yield function:
k k Rn
fn1 n1 y 33
1 Dn1
where the subscript () corresponds to the trial quantities. Note that
damage-behavior coupling involves the nonlocal variable Dn1 assumed as
a constant during material integration. Classically trial stress is defined as:
n1 1 Dn1 : "n1 1 Dn1 : "en " 34
where "n1 "en
" "en1 p
" is the known trial strain supposed as
purely elastic. If f * < 0 the solution is effectively elastic and the internal
variables issued from the elastic prediction are true. i.e., :
n1 n1 , Rn1 Rn , "pn1 "pn , Dn1 Dn 35
Otherwise ( f * 0) the elastic prediction solution must be corrected to
ensure the nullity of the yield function at the end of the time step.
Where e
N is the outward normal to the trial yield surface defined by
n1
Equation (33). According to Equation (39) and (38) the yield function
(Equation (36)) at tn+1 is written:
N e
"pn1 "pn n1
41a
1 Dn1
Yn1 Y0
Dn1 Dn 5 4s 41b
1 Dn1 S
!
1 Ne
n1
Yn1 Yn " :
2 n1 1 D n1
!
Ne 1R e2
n1 n1
: "n1 41c
1 D n1 2 Q
This achieves the computation of all the local state variables at the end of
the load step.
e
d n1 @ n1 d"n1 @ n1 dN @ d
n1
n1
d4" @"n1 d4" @N e d4" @ d4"
n1 42
dDn1 @Dn1 @Yn1 d"n1 @Dn1 @Yn1 d @Dn1 d
d4" @Yn1 @"n1 d4" @Yn1 @ d4" @ d4"
This leads to the following terms of the consistent stiffness matrix:
d n1 1 Dn1 3
1 D n1 KId Id 2Iddev 42
d"n1 kSn1 k
" " # #
kSn1 k 3 4
Nn1 Nn1 Iddev 43a
3 Q bR en1 2
d n1 n1
43b
dDn1 1 Dn1
" " ##
dDn1 21 Dn1 1 s Qeb=1Dn1
FD
d"n1 3 Q bR en1 1 Dn1 1 Dn1 5 Yn1 Y0 4
e s d n1 e
Nn1 FD :" 43c
1 Dn1 5 Yn1 Y0 4 d"n1 n1
" #
dDn1 1 s Qeb=1Dn1
FD 43d
d Dn1 1 Dn1 1 Dn1 1 Dn1 5 Yn1 Y0 4
1 5 Yn1 Y0 4 s
with FD
43e
1 Dn1 S
In the following section the ability of the proposed fully coupled nonlocal
damage model to conserve the wellposedness of the IBVP solution is stu-
died. A bi-dimensional simulation of a plane strain tensile test is proposed.
The dimensions of the tested zone are a 1 by 4 rectangular sheet with a unit
thickness. The ABAQUS/Standard code is used coupled to UEL and
UMAT users subroutines. The boundary conditions are presented in
Figure (2). The sheet is clamped at the left side and tied to a moving tool
on the right side. A uniform displacement is imposed to this tool along the
major axis of the sheet with a constant velocity of 1 mm/s. The sheet is
meshed with triangular elements using a near uniform mesh size h. The
equivalent stress (von-Mises) is presented at Figure 3 as a function of the
cumulated plastic strain (p). We call un-coupled a solution where damage
has no effects on the plastic behavior. There is no more softening phenom-
enon and hardening can reache saturation. Two virtual materials defined in
Table (1) are tested. The both materials share the same elastic and hardening
behavior. They only differ by their damage evolution. Accordingly, the local
un-coupled stress/strain curves are merged. However, the local fully coupled
stress/strain curves differ significantly as can be seen in Figure 3. Those
materials are not representative of any real materials, they only represent
4
1 3 5
Figure 2. Meshing and boundary conditions of the tensile specimen and position of five
studied IP.
700
Un-coupled
600
500
Equ. stress (Mpa)
400
300
200
Type-I
100
Type-II
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Equ. plastic strain
Figure 3. Stress/strain curves for both coupled and un-coupled model for the two materials
(Type-I and Type-II).
The internal length parameter (!) effect is now studied for both materials.
The simulation results obtained with four different values from ! 0.0 to
! 10.8 are compared. The global force/displcament curves analysis at
Figures 4 (Type-I) and 5 (Type-II) proves the aptitude of the internal
length parameter to act on the post-peak dissipation. Dissipation is increas-
ing with ! because the size of the yielding zone is directly dependent on the
internal length parameter. Consequently, a high value of ! delays the local-
ization phenomenon and alters directly the load carrying capacity of the
structure. Comparison of global force/displacement curves for both materi-
als show two radically different answer. For material Type-I (Figure 4),
nonlocality seems to have no impact on the global dissipation. The distance
between the two local curves is constant whatever the value of !. The Non
local model do not reduce mesh dependency. For this paradoxical simula-
tion the finer mesh gives the higher dissipation. This phenomenon is only
noticed for P1 and P1+ interpolated elements due to the well known poor
700
600
500 w = 0.0
Force (N)
400 w = 0.3
100
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Displacement (mm)
Figure 4. Global answer for 4 different values of ! and 2 mesh sizes h = 0.08 and h = 0.16.
Type-I.
The influence of the mesh size on the FEM results is now studied. The
solutions of the IBVP are presented for various characteristic mesh sizes h
varying from h 0.05 to h 0.16. Figure 5 presents the D field distribution
inside the specimen for three different mesh sizes and three level of ! evol-
ving from 0.3 to 10.8. From this Figure, the following remarks can be
formulated:
. filled contours of regularized damage show that the necking zone area is
mesh independent. The surface (or volume in 3D) is clearly dependent on
the ! parameter. Larger is !, greater is the average zone of strong damage
values D 4 0:7.
. For a given value of D the necking zone is mesh independent but the
necking mode depends on the mesh size. Those differences are more vis-
ible with small values of !.
. In every case, regularized damage fields are homogeneous and do not
present strong gradients. The plastic flow is not focused in a single ele-
ment width row.
. In every case the maximum damage value is 80%. It is obvious that a
great number of elements are in situation of plastic softening. It can be
concluded that the IBVP bifurcation has been largely delayed.
Those results show a good independence of the regularized damage filled
contours to the mesh size. This independence is really effective only for a
minimal value of the internal length parameter correlated to the mesh size.
When ! tends to zero, for a given mesh size, the solution of a purely local
model where necking is focused in a single element width row is recovered.
2.2073E+01
1.4370E+01
8.0962E+00
1.6555E+01
1.0777E+01
6.0722E+00
1.1036E+01
7.1850E+00
4.0481E+00
5.5182E+00
3.5925E+00
2.0241E+00
0.0000E+00
0.0000E+00
0.0000E+00
(a) (b) (c)
Figure 7. D filled contours for 3 mesh sizes h=0.05 (a), h=0.08 (b), and h=0.16 (c) with
!=1.8.(material Type-II) (Note that D > 1 if D
1 see section Regularized Damage
Variable).
Note that the UEL subroutine facility is only available for ABAQUS/
Standard. Consequently, convergence difficulties are encountered using this
implicit scheme when the tangent stiffness matrix is negative. Due to this
lack of convergence, when damage reaches unity, the propagation of the
macroscopical cracking can not be simulated. The use of ABAQUS/Explicit
together with an adaptive mesh procedure in which the fully damaged ele-
ments are killed is shown to be very helpful to solve this problem (Saanouni,
2006; Borouchaki et al., 2005). Mesh size dependencies of D fields are now
studied in (Figure 5). If the D field is smooth and regular, internal variables
fields define a cloud of high values surrounding a highly damaged single-
element width row (Figure 7). The size of this cloud is directly dependent on
the parameter !. Whatever the value of !, the cumulated plastic strain filled
contours presented at Figure 6 show that yielding is focused in a single
element row at the end of necking. The proposed simulation looks able to
control the size of the yielding region but finally macroscopic cracking
happens in a zone directly dependent on the mesh size.
Figure 8 presents the force/displacement curves of this specimen for
various values of !. For the purely local case the coarse mesh leads to the
700
600
Un-coupled
500
Force (N)
400
w=0
300 XL-16-Fu 1:2
XL-08-Fu 1:2
XL-05-Fu 1:2
200 XNC-16-Fu 1:2
XNC-08-Fu 1:2
w = 0.3
XNC-05-Fu 1:2
100 XNL-16-Fu 2:3 w =1.8
XNL-08-Fu 2:3 w =10.8
XNL-05-Fu 2:3
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Displacement (mm)
Figure 8. Global answer for 4 different values of ! and 3 mesh sizes h=0.05, h=0.08, and
h=0.16. Type-II.
Let us compare now the answer of the purely local and nonlocal models
for differently oriented initial meshes. Results are presented for the Type-II
material. In the purely local case (! 0) a strong dependence of damage and
plastic multiplier filled contours is noticed (Figure 9). The shear band focus
in a single element width row clearly dependent on the initial mesh orien-
tation (Sornin, 2007 for more details). However, the nonlocal model ! 1.8
shows a good independence to the initial mesh orientation. The shear band
orientation is not influenced by the mesh. On Figure 10 the nonlocal model
is compared for two different orientations and the same mesh size. The
oriented mesh presents a 40 angle from the loading direction. The textured
mesh presents a 60 angle from the loading direction. For both cases it seems
that internal variables evolution is dependent on mesh orientation. But the
regularized damage variable D (Figure 10 (b),(f)) and the plastic multiplier
(Figure 10 (d),(h)) proves that the size and the shape of the yielding zone as
SDV11
+3.113e - 01
+2.853e - 01
+2.594e - 01
+2.334e - 01
+2.075e - 01
+1.816e - 01
+1.556e - 01
+1.297e - 01
+1.038e - 01
+7.782e - 02
+5.188e - 02
+2.594e - 02
(a) +0.000e + 00
SDV10
+3.432e - 04
+3.146e - 04
+2.860e - 04
+2.574e - 04
+2.288e - 04
+2.002e - 04
+1.716e - 04
+1.430e - 04
+1.144e - 04
+8.581e - 05
+5.721e - 05
+2.860e - 05
(b) +0.000e + 00
Figure 9. Damage D (a) and (b) filled contours for an initially oriented mesh in the local
case (! = 0.0), (with h = 0.05 and = 40 ).
well as the orientation of the shear band are not correlated to the initial
mesh orientation.
The nonlocal model is now tested for different mesh topologies (he1, he2,
he3, and hh) defined in Table 2. Figure 11 presents filled contours of D and
D for different kind of mesh in the deformed configuration at the end of
simulation. Clearly, the central zone of the specimen is the same (h 0.08)
whatever the case. The remaining sheet side present different mesh size
varying from h 0.05 (he1) to h 0.16 (he3). A biased mesh (hh) varying
from mesh h 0.05 to h 0.16 is also presented. It is noticed that the plastic
flow with damage localize preferentially at the interface between the central
zone and the remaining sheet side for meshes he1 and he2. Numerical errors
due to the distortion of elements in the transition zone between the two mesh
sizes region can be incriminated. However, the localization tacks place cor-
rectly at the center of the specimen for mesh he3 and hh. These results, as
well as the one given by mesh (hh), are to be compared with the homogenous
mesh (h08) of Figure 5. The size of the yielding zone is the same in the three
cases (hh, he3, h08). However, in the side areas the internal variables fields
are evolving differently. For a given maximum value of damage (Figure
11(c) and 11(d)), the damage field is more regular for (hh) mesh than for
(he3) one. Results in term of fields distribution appears to be similar for the
(hh) and (h08). The influence of the mesh size variation around the necking
zone on the structural dissipation is shown in Figure 12. It can be noticed that
8.4159E+02
5.0481E05
6.3120E+02
3.7861E05
4.2080E+02
2.5241E05
2.1040E+02
1.2620E05
0.0000E+00
0.0000E+00
(a) (b) (c)
8.8177E+02
1.2417E04
6.6132E+02
9.3128E05
4.4088E+02
6.2085E05
2.2044E+02
3.1043E05
0.0000E+00
0.0000E+00
Figure 11. Filled contours for D and D for three nonuniform meshes. h=0.08 in the central
region. (!=1.8 material Type-II).
700
600
Un-coupled
500
Force (N)
400
300
200
he3
100 hh
he2 he1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Displacement (mm)
Figure 12. Structural global answer regarding to the mesh type. h = hh, he1, he2, he3 for
! = 1.8.
structural dissipation is not only dependent on the mesh size in the necking
zone but also on the mesh size far away from the crack elaboration zone. This
dependence tends to vanish with a smoother variation of mesh size. For the
other meshes like (he1) and (he2) the shifted necking zone gives largely over-
estimated global dissipations. A lack of spatial discretization as well as the
poor P1+/P1 element efficiency can be incriminated to explain such a beha-
vior. It is shown that the nonlocal formulation can be used with various mesh
topologies. This kind of mesh are largely used for adaptive remeshing oper-
ation which includes mesh size gradients over the structure (Borouchaki et al.,
2005). But meshing must be done with some care:
. For a given internal length scale, mesh size must be small enough to avoid
a local answer. With the proposed model, there is no analytical relation
between the value of the internal length scale parameter and the maxi-
mum mesh size allowed in the necking region. The value is numerically
obtained by a mesh size convergence analysis.
. The finest mesh in the structure must be automatically associated to
highly damage region around the crack zone. Adaptive meshing based
on physical error indicators is necessary to avoid spurious cracking path.
. The size of the mesh must be constant in all the cracking zone. It means
that the size of the fine mesh region is connected to the internal length
parameter (!).
. The gradient of mesh size in the structure must be smooth enough to
ensure a good independence to the mesh topology.
This analysis corroborates the work of (Svedberg and Runesson, 2000). The
results and main conclusions of the above section can be helpfully compared
to the works of (Areias, 2003) and (Cesar-De-Sa et al., 2006).
The local behavior is studied for some IPs dispatched over the sheet as
defined in Figure 2. IP1 submitted to a small plastic strain, IP2 at the limit of
the yielding zone when necking occurs, IP3 just around the macroscopical
crack (elastically unloaded when cracking occurs), IP4 in plastically soft-
ening when cracking occurs and IP5 at the center of the final crack (without
any elastic unloading). Those five IP are supposed representative of various
loading paths applied to the material points of the structure. At Figure 13
some results of a purely local model and the nonlocal model are compared at
the IP level. For the local model, the IP1 is elastically unloaded first for
p & 0.015 due to the plastic flow of the central zone. As the plastic flow
increases and localizes in the central zone, others integration points
(IP2IP4) becomes successively elastically unloaded expecting IP5 where
the plastic flow with damage leads to softening behavior. It is clearly noticed
that all the integrations points follow the same stress/strain curve before
theirs respective elastic unloading occurrence. This is confirmed by the
unique damage/strain curve followed by all the IPs at Figure 13. For the
nonlocal model, the local behaves of all the IP are significantly different.
Before elastic unloading occurs a plastic softening is noticed for IP2IP5.
This plastic softening issued from the nonlocal formulation ensures a
progressive lose of the load carrying capacity. It can be noticed that soft-
ening increase with the distance to the macroscopic crack. For IPs far from
the macroscopic crack, softening is fastened (IP1), whereas the IP5 ductility
is considerably increased compared to the local case. For a better under-
standing of the plastic and elastic softening the damage curves D p are
plotted for the different IPs at Figure 13(b). For the nonlocal model multiple
responses function of the loading history in the structure are noticed.
Thereby, IP4 and IP5 belonging to the crack zone reach the critical
damage for a higher plastic strain compared to the local case. However,
for IPs far from the crack (IP1 and IP2), damage evolution is
700
Local
600
PI-4
PI-2
Nonlocal
Equivalent stress (N.m2)
500
400
PI-3
PI-1
300
PI-3
200
PI-5
PI-5
PI-4
100
0
0 0.1 0.2 0.3 0.4 0.5 0.6
(a) Equivalent plastic strain (m1)
0.9
PI-5 PI-4 PI-5
0.8
0.7 Local
Nonlocal PI-3
0.6
Damage ()
0.5
0.4
0.3
PI-2
0.2
0.1
PI-1
0
0 0.1 0.2 0.3 0.4 0.5 0.6
(b) Equivalent plastic strain (m1)
p (b) for the five IP in local and non local cases.
Figure 13. Mises(p) (a) and D
radically fastened. Compared to the purely local case, IP2 reaches 20% of
damage for only 10% of cumulated plastic strain.
It is clearly shown that the nonlocal model induces serious variations
of the local behavior function of the distance from the macroscopic
cracking zone. For a given value of the equivalent plastic strain p the
internal length scale parameter. The necessity of fine meshes and quadrati-
cally interpolated elements to reduce mesh-dependency is also highlighted.
Without kill-element option it is rather difficult to prove the efficiency of
this formulation during crack propagation. Some works are actually driven
to enable an explicit scheme computation for crack propagation modeling.
Extension of the model for 3D problems and more complex structures and
loading paths is a major prospective. The proposed approach leads to a
convenient regularization of the IBVP until highly damaged element occur-
ence. But crack propagation corresponding to a discrete phenomenon still
takes place in a single elements width row. This constatation implies the
choice of both a correct mesh size regarding to the crack dissipation and an
internal length parameter ! for the local necking description.
REFERENCES
Areias, P. (2003). A gradient Model for Finite Strain Elastoplasticity Coupled with Damage,
Finite Elements in Analysis and Design, 39: 11911235.
Bazant, Z. and Jirasek, M. (2002). Nonlocal Integral Formulations of Plasticity and Damage:
Survey of Progress, Journal of Engineering Mechanics, 128: 11191149.
Babus ka, I. and Narasimhan, R. (1997). The Babuska-Brezzi Condition and the Patch Test: An
Example, Computer Methods in Applied Mechanics Engineering, 140: 183199.
Bazant, Z. and Pijaudier-Cabot, G. (1988). Nonlocal Continuum Damage, Localization
Instability and Convergence, International Journal of Applied Mechanics, 55: 287293.
Badreddine, H. (2006). Elastoplasticite anisotrope endommageable en transformations finies:
Aspect theoriques, numeriques et applications, PhD Thesis, Universite de Technologie de
Troyes.
Belamri, N. (2004). Modelisation numerique de decoupage de toles laminees. application au
cisaillage et au refendage en 3d, PhD Thesis, Universite de Technologie de Troyes.
Borouchaki, H., Laug, P., Cherouat, A. and Saanouni, K. (2005). Adaptive Remeshing in
Large Plastic Strain with Damage, International Journal of Numerical Methods in
Engineering, 63: 136.
Cesar-De-Sa, J. and Areias, P. (2004). Gradient Damage Models in Metal Forming Problems.
In: Ghosh, S., Castro, J.C. and Lee, J.K. (eds), Modeling, Simulation and Applications 8th
International Conference on Numerical Methods in Industrial Forming Processes
(NUMIFORM), Columbus, Ohio, June 2004.
Cesar-De-Sa, J., Areias, P. and Zheng, C. (2006). Damage Modelling in Metal Forming
Problems Using an Implicit Nonlocal Gradient Model, Computer Methods in Applied
Mechanics Engineering, 195: 66466660.
Cherouat, A. and Saanouni, K. (2003). Numerical Simulation of Sheet Metal Blanking Process
Using a Coupled Finite Elastoplastic Damage Modelling, International Journal of Forming
Processes, 6(1): 732.
Cherouat, A., Saanouni, K. and Hammi, Y. (2002a). Improvement of Forging Process of a 3d
Complex Part with Respect to Damage Occurrence, Journal of Materials Processing
Technology, 142/2: 307317.
Cherouat, A., Saanouni, K. and Hammi, Y. (2002b). Numerical Improvement of Thin Tubes
Hydroforming with Respect to Ductile Damage, International Journal of Mechanical
Sciences, 44: 24272446.
Saanouni, K. and Chaboche, J.-L. (2003). Comprehensive Structural Integrity, ISBN: 0-08-
043749-4).
Saanouni, K., Badreddine, H. and Ajmal, M. (2008). Advances in Virtual Metal Forming
Including the Ductile Damage Occurrence. Application to 3d Sheet Metal Deep
Drawing, Journal of Engineering Materials and Technology, 130: 021022-1 021022-11.
Saanouni, K., Forster, C. and Ben-Hatira, F. (1994). On the Anelastic Flow with Damage,
International Journal of Damage Mechanics, 3: 140169.
Saanouni, K., Mariage, J.-F., Cherouat, A. and Lestriez, P. (2004). Numerical Prediction of
Discontinuous Central Bursting in Axisymmetric Forward Extrusion by Continuum
Damage Mechanics, Computers and Structures, 82: 23092332.
Samal, M.K. and Seidenfuss, M. (2008). Finit Element Formulation of a New Nonlocal
Damage Model, Finit Elements in Analysis and Design, 44: 358371.
Simone, A., Askes, H. and Sluys, L. (2004). Incorrect Initiation and Propagation of Failure in
Nonlocal and Gradient-enhanced Media, International Journal of Solids and Structures,
41: 351363.
Sornin, D. (2007). Sur Les Formulations Elsatoplastiques Nonlocales en Gradient dendomm-
agement, PhD Thesis, Universite de Technologie de Troyes.
Sornin, D. and Saanouni, K. (2006). Theoretical and Computational Aspects of an
Elastoplastic Damage Gradient Nonlocal Model, In: European Conference on
Computational Mechanics ECCOMAS2006, Lisbon, Portugal, 05-09 June, CD_Rom
Paper No. 326.
Svedberg, T. and Runesson, K. (2000). An Adaptive Finite Element Alghorithm for Gradient
Theory of Plasticity with Coupling to Damage, International Journal of Solids and
Structures, 37: 74817499.
Tvergaard, V. and Needleman, A. (1995). Effects of Nonlocal Damage in Porous Plastic Solids,
International Journal of Solids and Structures, 32: 10631077.