You are on page 1of 18

j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497

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

Metal ductility at low stress triaxiality application


to sheet trimming

A. Bacha a,b , D. Daniel b , H. Klocker a,∗


a SMS Centre, Ecole des Mines, 158, Cours Fauriel, 42 023 Saint Etienne Cedex 2, France
b ALCAN, Centre de Recherches de Voreppe, BP27, 38 340 Voreppe, France

a r t i c l e i n f o a b s t r a c t

Article history: The growth and coalescence of voids nucleated by decohesion or cracking of second phase
Received 18 July 2007 particles is a common damage process for many metallic alloys. Classical damage models,
Received in revised form based on void growth and coalescence, predict a ductility increase if the stress triaxiality
3 October 2007 is decreased. But experiments show that the material ductility decreases at very low stress
Accepted 23 October 2007 triaxialities typical of sheet metal forming operations. At very low stress triaxiality no void
growth is observed in metals containing second phase particles. In the present work, a
new damage model for metals containing second phase particles submitted to low stress
Keywords: triaxiality loading is proposed.
Ductility The new model is based on the observed physical damage mechanism, i.e. strain localiza-
Elastic plastic material tion by reducing the inter-particle spacing during large material rotations. A two step mod-
Finite elements elling strategy has been followed to determine the ductility at low stress triaxiality. In the first
Fracture mechanisms step Thomason’s void coalescence model is extended to large material rotations and shear-
ing. In the second step the principles of applying this model to damage nucleated at second
phase particles are described. The large material rotations observed under low stress triaxi-
ality loading lead to large changes in the microstructure. Thus, in the second step, first appro-
priate representative volume and material elements are determined and then the critical
damage parameters. Finally, as an example, the trimming behaviour of two aluminium sheet
alloys is analyzed by the new model and the model predictions shown to be in good agree-
ment with the experimental data, in particular for the blade displacement to crack initiation.
The main outcomes of this work are: (1) a void coalescence model valid at low stress triaxi-
ality, (2) a damage criterion valid at small stress triaxiality and large material rotations, (3) a
damage variable expressed in a simple closed form for materials containing second phase
particles. The damage analysis in a small fixed volume with a representative microstructure
(Eulerian approach) and the damage analysis in all the material elements of the considered
structure are compared in detail. In trimming (or similar processes), the major contribution
to damage the material movement bringing second phase particles closer together and the
void growth may be neglected. This simplifies considerable the analysis.
© 2007 Elsevier B.V. All rights reserved.

1. Introduction they eventually coalescence to form a macroscopic crack. In


the early 1970s a new branch of material science (“the local
In ductile metals, voids are first nucleated by decohesion approach of ductile fracture”), was initiated by the pioneer-
or cracking of second phase particles and then grow until ing contributions of Gurland and Plateau (1963), McClintock


Corresponding author. Tel.: +33 477420078; fax: +33 477420000.
E-mail address: klocker@emse.fr (H. Klocker).
0924-0136/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2007.10.054
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497 481

Nomenclature
y yield stress of the composite (matrix + voids)
Cm dilatation tensor expressed in the material axes  y0 initial yield stress in Voce law
Cmij component of the dilatation tensor in the mate- lig
y yield stress in the intervoid ligament
rial axes ∞ flow stress for large deformations
DLST maximum principal stress/localization stress ˙E mean value of  in the Eulerian volume VE :
Dtension
LST critical value of DLST determined in uniaxial ˙E = (1/VE ) (x ) dx
VE
tension ˙Mi mean value of   in the material element
E Young’s modulus Mi ˙Mi = (1/Mi ) M (x ) dx
i
E11 overall cell deformation in the direction of the ˙ pm maximum principal stress acting on the cell
material ligament between voids ˙pm
loc value of ˙ pm leading to strain localization
E33 overall cell deformation perpendicular to the  angle between the intervoid ligament and max-
direction of the material ligament imum principal stress
EE mean value of ε in the Eulerian volume  VE
EE mean strain tensor in VE : EE = (1/VE ) ε(x ) dx
VE
Em strain tensor expressed in the material axes
(1968a,b), Rice and Tracey (1969), Rice and Johnson (1970),
EMi mean strain tensor in the material element Mi :
Needleman (1972), Brown and Embury (1973), Argon (1976),
EMi = (1/Mi ) ε(x ) dx
Mi Gurson (1977), Chu and Needleman (1980), Beremin (1981a,b),
Em11 component of the strain tensor Em Lautridou and Pineau (1981), Tvergaard (1981, 1982), Budiansky
f void (particle) volume fraction et al. (1982), Browning et al. (1983) and Needleman and
fdecrease describes decrease of void height during overall Tvergaard (1984). One outcome of all these analyses is the
plastic deformation major role played by the stress triaxiality.
fgrowth describes void growth during overall plastic Large values of the stress triaxiality (T > 0.4) lead to a
deformation decrease of the fracture strain (ductility). Thus, during the
h height of deforming ligament last 20 years, most efforts concerned essentially medium to
h0 initial value of void height high stress triaxiality loadings (Benzerga et al., 2004a,b; Ragab,
hcrit critical value of void height 2004; Besson, 2004; Huber et al., 2005; Klöcker and Tvergaard,
h∞ final void height 2003; Gologanu et al., 1995, 2001; Siruguet and Leblond, 2004;
H unit cell height Pardoen and Hutchinson, 2000, 2003; Shabrov and Needleman,
H0 initial value of the unit cell height H 2002; Gammage et al., 2004; Worswick et al., 2001; Thomson et
l particle (void) width al., 2003; Kim et al., 2004; Pardoen et al., 2003; Lee and Mear,
L particle (void) spacing 1999; Christman et al., 1989; Joly, 1992; Koplik and Needleman,
L0 initial value of the particle (void) spacing L 1988; Hom and McMeeking, 1989a,b; Worswick and Pick, 1990;
Mi material element (Lagrangian volume) Brocks et al., 1995; Steglich and Brocks, 1997; Gao et al., 1998;
P rotation tensor Faleskog and Shih, 1997; Kuna and Sun, 1996; Bordreuil et
Ph work rate corresponding to homogeneous al., 2003; Kroon and Faleskog, 2005; Doghri and Ouaar, 2003;
deformation Babout et al., 2004; Leblond et al., 1994, 1995; Scheyvaerts,
Pl work rate corresponding to strain localization 2006; Pineau, 1992; Huang, 1991). But, in many forming opera-
T stress triaxiality tions, the material is submitted to large compressive or small
u blade displacement tensile loads corresponding to negative or small stress tri-
vx , vy velocity components axiality loadings. In this context, Bao and Wierzbicki (2004)
VE Eulerian volume deformed an AA2024 alloy under various stress strain paths
Ẇ velocity at the cell boundary in the direction of covering small and large values of the stress triaxiality; their
maximum principal stress results are summarized in Fig. 1. The high stress triaxil-
x position vector ity results are well described by several classical damage
˛ linear strain hardening coefficient in Voce law models (Gurland and Plateau, 1963; McClintock, 1968a,b; Rice
ˇ coefficient used to describe change in void and Tracey, 1969; Rice and Johnson, 1970; Needleman, 1972;
height during overall deformation Brown and Embury, 1973; Argon, 1976; Gurson, 1977; Chu and
 shear angle Needleman, 1980; Beremin, 1981a,b; Lautridou and Pineau,
 max maximum shear angle for a given microstruc- 1981; Tvergaard, 1981, 1982; Budiansky et al., 1982; Browning
ture et al., 1983; Needleman and Tvergaard, 1984; Benzerga et al.,
ı coefficient of Voce law 2004a,b; Ragab, 2004; Besson, 2004; Huber et al., 2005; Klöcker
ε strain tensor and Tvergaard, 2003; Gologanu et al., 1995, 2001; Siruguet and
ε̄p plastic equivalent von Mises strain Leblond, 2004; Pardoen and Hutchinson, 2000, 2003; Shabrov
ε̄˙ p equivalent plastic strain rate and Needleman, 2002; Gammage et al., 2004; Worswick et al.,
v velocity discontinuity at the interface between 2001; Thomson et al., 2003; Kim et al., 2004; Pardoen et al.,
deforming ligament and rigid matrix 2003; Lee and Mear, 1999; Christman et al., 1989; Joly, 1992;
 Cauchy stress tensor Koplik and Needleman, 1988; Hom and McMeeking, 1989a,b;
Worswick and Pick, 1990; Brocks et al., 1995; Steglich and
482 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497

Fig. 1 – Schematic representation of the strain to failure of a


ductile material as a function of the stress triaxiality.
Experimental results (Lautridou and Pineau, 1981) and
classical damage models (McClintock, 1968b; Rice and
Tracey, 1969; Rice and Johnson, 1970; Needleman, 1972;
Brown and Embury, 1973; Argon, 1976; Gurson, 1977; Chu
and Needleman, 1980; Beremin, 1981a,b). In the hatched Fig. 2 – Schematic representation of present work. (a and b)
area classical damage models over estimate the strain to Initial and deformed microstructure of a heterogeneous
failure and a new criterion has to be developed. material containing second phase particles. The material
elements are submitted to large movements and
deformations prior to failure. (c) Material element at failure
and (d) material element used in the coalescence criterion.

Brocks, 1997; Gao et al., 1998; Faleskog and Shih, 1997; Kuna
and Sun, 1996; Bordreuil et al., 2003; Kroon and Faleskog, 2005;
Doghri and Ouaar, 2003; Babout et al., 2004; Leblond et al., 1994,
1995; Scheyvaerts, 2006; Pineau, 1992; Huang, 1991). However,
and b). First appropriate representative volume and mate-
when the stress triaxiality is small (T ∈ [−0.3,0.4]), these classi-
rials element are determined and then the critical damage
cal models significantly over-estimate the strain to failure. To
parameters. The strain leading to void nucleation is consid-
our knowledge there is no adequate model to predict ductility
ered negligible compared to the failure strain. In Section 4 the
at low stress triaxiality.
new damage criterion is applied to sheet metal cutting. Metal
In the present work, a new damage model for met-
cutting leads to very large deformations and very large rota-
als containing second phase particles submitted to low
tions of the material elements and the microstructure. The
stress triaxiality loading and large material rotations is pro-
new model predictions are compared to the experimental val-
posed. During forming of heterogeneous materials, the second
ues of the blade penetration at crack initiation.
phase particles undergo large movements (Fig. 2). Thus, the
particle–matrix interface is broken, but, since the stress tri-
axiality is low, no significant void growth is observed (as will
2. The new damage criterion
be confirmed here). As the broken matrix particle interface no
longer carries any load, it is proposed that failure occurs due to
the reduction of interparticle spacing and subsequent strain In heterogeneous materials, voids are nucleated by deco-
localization. hesion or cracking of second phase particles. At low stress
A two step modelling strategy is followed in this paper to triaxiality, only limited void growth is observed and the dam-
determine the ductility, i.e. the local strain to failure at low age is due to reduction of the interparticle spacing. This
stress triaxiality (Fig. 2). In the first step (Section 2), strain local- situation is described schematically in Fig. 2b and c. In this sec-
ization between rectangular voids in a sheared microstructure tion, the critical particle arrangement leading to local material
is considered (Fig. 2c and d). This first step, an extension of failure is determined. To determine this critical arrangement,
Thomason’s void coalescence model (Thomason, 1990), deter- only the corresponding voids are considered (Fig. 2c and d).
mines the critical void arrangement as a function of the local
stress. Similar approaches were used previously to describe 2.1. Thomason’s void coalescence model
void coalescence under medium or high triaxiality loading
(Benzerga et al., 2004a,b; Pardoen and Hutchinson, 2003). But Thomason (1990) considered a material containing a periodic
under medium or high stress triaxiality loading the material array of rectangular voids (Fig. 3a). Under small loads the cell
does not undergo large movements before void coalescence. deforms homogeneously. Increasing the applied stress ˙ pm
This point is addressed in the second step. The second step leads to strain localization in the ligament between the voids
(Section 3) explains how to determine the critical volume and the cell walls. The work rate corresponding to homoge-
(material) element leading to crack initiation and final fail- neous deformation Ph and the work rate corresponding to
ure in a material containing second phase particles (Fig. 2a strain localization Pl in the intervoid ligament are compared.
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497 483

where h is the height of the ligament. x and y are the velocity


components along the axes {xlig , ylig }. Finally, the work rate
Pl corresponding to flow localized in the intervoid ligament
depends on the yield stress in the ligament, the velocity at the
cell boundary, the ligament height h and width (L − l):

4 lig 1 L−l
Pl = √ y Ẇ (L − l) sin 1+ + (5)
3 4tan2 4h

Comparing the work rate (2) corresponding to homoge-


neous deformation and the upper-bound estimate (5) of the
work rate corresponding to localized flow, leads to an upper-
bound estimate of the maximum principal stress ˙pm loc at the

onset of strain localization (Thomason, 1990):


Fig. 3 – Thomason’s (Tvergaard, 1981) void coalescence

˙pm
loc
2
l 1 L−l
model. A periodic array of rectangular voids is considered.
= √ 1− 1+ + (6)
(a) The main void axis and the maximum principal stress lig
y 3 L 4tan2 4h
are aligned. (b) An angle between the main void axis and
the principal stress is considered. 2.2. The new coalescence criterion for large material
rotations

The criterion for flow localization is as follows: In metal forming, material rotation and shearing lead to the
microstructure evolution represented in Fig. 4. If the small
Ph < Pl : homogeneous flow in the cell, angle ε is neglected, the preceding approach may be used to
analyze the strain localization. Thus the same velocity field as
Ph ≥ Pl : flow localized in the intervoid ligament
above may be used. The work rate corresponding to homoge-
(outside the intervoid ligament the cell material is “rigid”) neous flow is now given by

(1) sin
Ph = 2˙pm ẆL (7)
cos 
For perfectly plastic materials, the work rate corresponding
to homogeneous straining is given by the following expression where  is the shear angle. As the strain rate is constant in
(Fig. 3b): the localization volume and the velocity varies linearly along
the ligament boundary, the work rate corresponding to flow
Ph = 2 ˙pm Ẇ L sin (2)

where ˙ pm is the maximum principal stress applied to the cell


volume. Ẇ is the velocity at the cell boundary in the direction
of the maximum principal stress. is the angle between the
intervoid ligament and the maximum principal stress.
The work rate corresponding to localized flow is given by
the following expression (Fig. 3b):

 lig 
lig y
Pl = y ε̄˙ p dV + √ v dS (3)
V 3 s

lig
where y and ε̄˙ p are respectively the yield stress in the inter-
void ligament and the equivalent strain rate. V and S are
the volume and surface of the intervoid ligament, i.e. the
localization area. v is the velocity discontinuity at the inter-
face between the deforming intervoid ligament and the rigid
matrix. In order to estimate this work rate corresponding to
localized flow, Thomason assumed plane strain deformation
with the following kinematically admissible velocity field:


⎨ x = 2Ẇ (ylig cos − xlig sin ) Fig. 4 – Void coalescence model for large material rotations
h 2Ẇ and sheared microstructures.  is the shear angle between
 = x sin (4)
⎪ h lig
⎩ y = 2Ẇ ylig sin different layers of second phase particles. Only the
h microvoids are represented.
484 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497

localized in the intervoid ligament has a simple expression 3.1. Critical volume element and representative
depending on the shearing angle . material element


3.1.1. Critical volume element (Eulerian approach)
4 lig sin 1 L−l
Pl = √ y Ẇ(L − l) 1+ + (8) As strain localization occurs in the most critical inter-particle
3 cos  4 tan2 4 h cos2 
ligament, applying the damage criterion of Eq. (12) at the Gauss
points of a finite element mesh would not make any sense.
At the onset of flow localization, both work rates are equal This damage criterion has to be applied to a certain volume
and the maximum principal stress may be estimated as fol- VE containing a representative microstructure. The volume VE
lows: must satisfy two opposing conditions:

˙pm
loc
2
l
 1 L−l

• The volume has to be large enough to contain a represen-


= √ 1− 1+ + (9) tative microstructure.
lig
y 3 L 4 tan2 4 h cos2 
• The volume VE should be small enough to enable one to treat
stress and strain gradients due to the overall deformation.
 = 0 leads to Thomason’s expression for the maximum
principal stress at the onset of flow localization. The flow
Failure will occur in a volume VE at an overall deformation
stress in the intervoid ligament (undamaged material) may
Ef (Fig. 5). Thus for strains smaller than Ef , the damage vari-
be related to the alloy flow stress by a simple law of mixture:
able DLST is smaller than the critical value everywhere in the
structure. At an overall strain E = Ef , the damage variable DLST
hl

lig lig equals DC in the volume VE and is smaller than the critical
y = (1 − f )y = 1− y (10)
HL value everywhere else. But, during macroscopic deformation,
different material elements flow through the volume VE (Fig. 5)
Comparing the real value of the maximum principal stress and Ef corresponds to the first material element satisfying
˙ pm to the estimated value at the onset of strain localization DLST = DC .
in the intervoid ligament, ˙pm
loc , allows to introduce a scalar

damage variable for low stress triaxiality failure DLST : 3.1.2. Critical material element
Fig. 5 shows three different material elements M1 , Mf and M2
˙pm embedded in the volume VE at overall deformations E1 , Ef and
DLST = E2 , respectively. Material element M1 , embedded in the volume
loc
˙pm
VE at E1 < Ef , will never satisfy the damage criterion. Material

3 ˙ / (1 − (hl/HL)) elements Mf and M2 satisfy the damage criterion DLST = Dc at
= pm y
2 overall deformations Ef and E2 , respectively. The overall fail-
(1 − (l/L)) 1 + (1/4tan2 ) + ((L − l)/4h cos2 )
ure strain Ef is thus determined by the first material element
(11) satisfying DLST = Dc . Thus in order to determine the macro-
scopic failure strain Ef , the damage history of all the different
material elements flowing through the volume VE has to be
Before the onset of strain localization DLST should be
determined.
smaller than unity and after strain localization DLST should
Fig. 6 shows a typical material element at failure. Of course
be larger than unity. But, as Eq. (11) gives an upper-bound
the particle shape is not rectangular and the void shape may be
estimate, the value of DLST may exceed unity at the onset of
very complicated locally. But, as will be seen the void growth
strain localization. The critical value of DLST leading to strain
is almost negligible. Thus approximating the void shape by
localization may be determined by any simple test. In the
rectangles does not affect the void distance.
present work, the values determined in simple tension will be
used and the strain localization criterion takes the following
expression: 3.2. Determining the parameters used in the damage
variable

˙pm
DLST = ≥ Dtension
LST (12) In this section, the determination of the damage variable in
loc
˙pm
the Eulerian volume VE and in all the material elements is
discussed in details. The Eulerian volume VE and the different
material elements Mi correspond to the smallest volume con-
3. Applying the new damage criterion to taining a representative microstructure. Thus to determine
heterogeneous materials the value of the damage parameter in VE and Mi , the mean
values of the field variables (stress and strain tensor) in VE and
At small values of the stress triaxiality, the voids due to par- Mi have to be used. The shape of each material element Mi is
ticle rupture or failure of the particle–matrix interface shrink defined by a series of vertexes (V1 , . . ., V5 in Fig. 5) fixed to each
(Section 4, Fig. 12), but the particle–matrix interface does not material element during overall deformation. The mean val-
carry any load. Strain localization thus occurs in the weakest ues of the field variables are obtained by interpolation from the
inter-particle ligament satisfying Eq. (12). Gauss points embedded in the considered material element.
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497 485

Fig. 5 – Deformation of a material containing second phase particles. (a) The material is shown at different overall
deformations E (E1 < Ef < E2 ). Ef is the macroscopic failure strain. VE is the critical Eulerian volume. M1 , Mf and M2 are three
material elements defined by their vertexes (V1 ,. . ., V4 ). (b) Evolution of the damage variable DLST in the Eulerian volume VE
and the different material elements as a function of the overall strain E. In M1 the damage never reaches the critical value.
In Mf the critical damage corresponding to material failure is reached at the overall strain Ef and in M2 the critical damage
leading to material failure is reached for E2 > Ef .

3.2.1. Current value of the maximum principal stress ˙ pm The rotation  of the second phase particles may be
˙ pm is the maximum principal stress corresponding either to assumed equal to the material rotation (Fig. 12). As will be
E or to ˙Mi where E and ˙Mi are respectively the mean values seen in Section 4, for metal alloys containing small second
of the stress tensor ␴(x ) in the volume VE and in the mate- phase particles, the particle rotation is identical to the mate-
rial element Mi for an overall macroscopic deformation E. In rial rotation. For other materials (composites, metal matrix
plane strain deformation, the maximum principal stress ˙ pm composites, . . .), a specific model for particle rotation may be
is related to the stress triaxiality T and the flow stress  y : used. The angle ˚ of the maximum principal stress with the
structural axis is obtained from the current value of the mean
˙pm 1 stress tensor, i.e. either ˙ E or ˙Mi .
=T+ √ (13)
y 3 The distortion  may have different origins. If the initial
microstructure is perfectly periodic (Fig. 8a)  corresponds to
3.2.2. Characteristic angles , , the shearing in the material axis {x1m , x2m }:
The angle between the ligament and the direction of the ⎧
maximum principal stress controls to a large extent the value ⎨  = −arcsin √ Cm12 √
of the damage variable DLST . The angle ( ≤ ␲/2) depends on Cm11 Cm22 (15)
⎩ C = P exp(2Pt E P)Pt
the second phase particle rotation , the shearing angle  and m m

the angle ˚ between the maximum principal stress direction


and axis x1 (Fig. 7): where Cm12 , Cm11 , Cm22 are the components of the dilation ten-
sor Cm in the material axes {x1m , x2m } and P the corresponding
 rotation tensor diagonalizing Cm .
= 0 ∀ 0 ∈ ]0,
/2]

0 = +˚−− (14) If the initial microstructure is not perfectly periodic, strain


=
− 0 ∀ 0 ∈ ]
/2,
[ 2
localization occurs in the weakest material ligament and an
486 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497

Fig. 6 – Application of the coalescence criterion to real


materials. The void shape may be approximated by a
rectangle to estimate the void interdistance.

irregular arrangement of particle layers respecting the mean


values L0 and H0 has to be considered (Fig. 8b). The maximum
range of  leading to “overlapping” successive particle layers
as shown in Fig. 8c is given by the following expression:

⎨  ∈ [−max , max ] Fig. 7 – Microstructure at the onset of cracking. (a) Initial
(16)
⎩ max = arctan h microstructure in the Eulerian volume VE . (b) Rotated
L−l microstructure. The particles are assumed to follow the
material axes. (c) Relation between the characteristic angles
The shear angle  in the weakest inter-particle ligament satis- of the microstructure and the loading of the inter particle
fies (16) and maximizes the damage variable. If the angle  is ligament.  gives the rotation of the material axes. 
determined by either (15) or (16) the angle is known by (14). corresponds to shearing of the inter particle ligament in
these material axes.  is the angle between the structural
3.2.3. Cell dimensions L, H axes x1 and the maximum principal stress ˙ pm . is the
The initial microstructure (l0 , h0 L0 , H0 ) has to be characterized angle between the direction of ˙ pm and the material
by image analysis. If the second phase particles are assumed ligament.
to rotate the same way as the material axes, the reference
frame {x1m , x2m } fixed to the material axes makes an angle 
with the fixed reference frame {x1 , x2 }. The mean strain tensor
in the material axes allows to determine the current particle
spacing: 3.3. Damage generated by pure shear strain

L = L0 exp(−Em11 ) The ratio ˙ pm / y , the new cell dimensions L, H and all the
(17)
H = H0 exp(Em11 ) characteristic angles (, , ˚, ) may be determined in the
Eulerian volume VE . The damage variable takes the follow-
where Em11 is the deformation in x1m direction. ing expression depending on the current stress and the initial
microstructure:

√ √
3 [T(E) + (1/ 3)][1 − (hl/H0 L0 )]
DLST =   (18)
2
(1 − (l/L0 exp(−Em11 ))) 1 + (1/4 tan2 (E)) + (L0 exp(−Em11 ) − l/4h cos2 (E))
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497 487

Fig. 8 – (a and b) Two possible microstructures exhibiting the same mean particle volume fraction, particle size and shape in
the undeformed and the deformed sheet. Microstructure (b) leads to greater damage than microstructure (a). (c) Critical
shear angle  max corresponding to “overlapping” layers of particles.

where E corresponds to the mean strain tensor in the volume 4.1. The experimental determination of blade
VE or in a material element Mi . At this stage of the develop- displacement at crack initiation
ment, only the void height h and width l are still unknown.
If the latter can be determined in the Eulerian volume VE , the 4.1.1. Microstructure of the material
damage analysis may be done in this volume. If the void width The two grades contain two kinds of intermetallic parti-
and height are history dependent, i.e. have to be determined cles (Mg2 Si and iron-rich particles, i.e. Al–Fe–Mn–Si-particles)
in the material elements, the damage analysis has to be done which control to a large extent the material ductility. Their size
in each material element Mi . and shape were therefore characterized by FEG SEM quantita-
The most simple, but very realistic approach to shear dam- tive metallography (Bacha et al., 2006a) on the (ND, TD) and
age consists in constant void volume (no void growth). The the (ND, RD)-planes. Fig. 9 shows typical SEM-micrographs of
damage parameter is controlled by the initial microstructure the two grades. The particle shape corresponds to a platelet
and the strain tensor: with equal dimensions in the RD and TD directions. The mean

√ √
3 [T(E) + (1/ 3)][1 − f0 ]
DLST = (19)
2 (1 − (l /L exp(−E 1 + (1/4 tan2 (E)) + (L0 exp(−Em11 ) − l0 /4h0 cos2 (E))]
0 0 m11 )))[

where f0 is the particle volume fraction. The damage variable particle dimensions are summarized in Table 1. The mean par-
may be determined in the volume VE . In Section 4, the value of ticle size is comparable in both grades, but the particle volume
 maximizing the damage variable and satisfying Eq. (16) will fraction in grade AM1 (1.16%) is much larger than in grade AM2
be chosen. (0.73%).

4.1.2. The trimming parameters


The various elements of the trimming facility are the blade,
4. Example: application of the new damage the sheet and the die (Fig. 10a). The parameters used to
model to metal cutting describe the geometry of the plane cutting process are the

In this section, the new damage criterion will be applied to


metal cutting which is a particularly common form of metal
failure under conditions of low stress triaxiality. Both the Eule- Table 1 – Microstructural characteristics of the two
rian and the Lagrangian approach will be used. Metal cutting AA6016 grades
is a complex three-dimensional process. However, the plane Grade Np (mm−2 ) fs (%) di (␮m) df (␮m) rf
cutting process can be used to analyze the physics of damage
AM1 3221 1.16 7.74 2.90 2.08
generated during localized shearing under conditions of low AM2 2526 0.73 9.14 2.89 2.44
T. The blade displacement at crack initiation is an important
parameter controlling the cutting behaviour. Here it is deter- Np and fs are respectively the mean particle number and the mean
mined experimentally for two aluminium alloy sheet samples particle surface fraction. di and df are the mean value of the max-
imum particle Feret diameter and the mean particle distance. rf is
and also predicted from physical data by the new damage
the mean particle aspect ratio.
criterion.
488 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497

Fig. 9 – FEG-SEM pictures showing the typical microstructure of the two AA6xxx grades: AM1 (a and b) and AM2 (c and d).

sheet thickness t, the cutting angle ˛ (i.e. the angle between displacement at crack initiation:
the sheet normal and the direction of blade displacement),
the clearance C between the blade and the die, the blade Uf sz + ro + Rblade
uf = = (20)
radius Rblade , the die radius Rdie and the blade displacement t t
Uh . These parameters were divided by the sheet thickness t
where ro and sz are respectively the dimension of the roll-
and the corresponding non-dimensional values are indicated
over (elastoplastic shearing) and the dimension of the sheared
by small symbols. The two alloys were cut in different tool
zone (indentation of the material). Fig. 11 shows results for the
configurations (Tables 2 and 8).
AM1 and AM2 grade cut in the reference tool geometry. The
blade radius Rblade was measured prior to any cutting oper-
4.1.3. Blade displacement at crack initiation
ation. The dimension of the roll-over ro was measured on
The presence of a clearance and finite material ductility lead
the cut profile (Figs. 11a and b) and the shear zone height
to a complex cutting shape with various characteristic zones
sz was determined on the fracture surfaces as indicated in
(Fig. 10b). The material ductility is characterized by the blade
Fig. 11b and c. The experimental values of uf are summarized
in Tables 7 and 8.

4.1.4. Physical mechanism of crack initiation


Fig. 12 shows three micrographs of sheared AM1 and AM2
samples. In micrograph Fig. 12a the crack initiation for AM1
grade is shown to occur at the tangent point between the blade

Table 2 – Reference tool geometry for the shear test:


sheet thickness t, clearance c, blade radius Rblade and die
radius Rdie

Fig. 10 – Schematic representation of a plane cutting t (mm) 1.00


C (mm) 0.05
process. (a) Trimming facility. (b) Characteristic dimensions
Rblade (mm) 0.05
of cut specimen: roll-over ro , shear zone sz , fracture zone Rdie (mm) 0.1
fz and burr burr . Uf is the blade displacement at crack ˛ (◦ ) 0
initiation, giving the shear ductility.
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497 489

Fig. 11 – Cut profiles (a and c) and fracture surfaces (b and d) of AA6xxx grades: (a and b) AM1, (c and d) AM2.

and the sheet. Fig. 12b shows an AM2 specimen after crack in the Voce law. ˛ ε̄p is an additional term accounting for the
propagation (u > uf ). The crack has been initiated at the same small positive strain hardening at large plastic strains. The
point and propagated as a mode 1 crack without void growth. parameters of the VOCE law are given in Table 3. The sheet
Fig. 12c shows a micrograph of AM1 before cracking. In the materials AM1 and AM2 have been shown to be transversely
volume close to the blade, which controls crack initiation and isotropic. Brunet and Morestin (2001) determined the Lank-
thus the shear ductility no void growth but void shrinking is ford coefficients applicable to AM1 and AM2: r0 = r90 = r45 = 0.7.
observed. Damage is due to a reduction of the intermetallic These coefficients lead to the Hill stress function given in
particle spacing during overall deformation and subsequent Table 4.
strain localization. The deformation history is described by the Voce law (21).
All the parameters in Eq. (12) are determined taking into
4.2. Mechanics of the trimming process account this real material behaviour. Applying a damage cri-
terion for perfectly plastic materials is reasonable because at
4.2.1. Material model and finite element simulation void coalescence (very large strains), the strain hardening has
The large deformation behaviour of the sheet metal was become very small.
determined previously (Bacha et al., 2006a,b) by compression To simulate cutting, an ABAQUS explicit code with adaptive
tests on layered samples. The large deformation material flow mesh generation and Arbitrary Lagrangian Eulerian formula-
stress,  y , was represented by an extended VOCE law (Voce, tion (A.L.E.) was used. Fig. 13a shows a typical FE-mesh. The
1948):


Y = Yo + (∞ − Yo )(1 − exp(−ı ε̄p )) + ˛ ε̄p
Y (21) Table 3 – Voce law parameters for the two 6xxx grades
ε̄p = ε̄ −
E
Grade E (GPa)  0 (MPa)  ∞ (Mpa) ı ˛ (MPa)
where E is the Young’s modulus, ε̄ and ε̄p are respectively the AM1 70 130 310 16 15
total and the plastic equivalent strains.  Yo is the initial flow AM2 70 100 248 12 14
stress.  ∞ corresponds to the stress at very large plastic strains
490 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497

Fig. 12 – (a) Fracture profiles of AM1 (a and c) and AM2-grades (b). (a and b) correspond to u = uf and u < uf . (c) corresponds to
u = 0.85 uf . uf is the blade displacement at crack initiation.

details of the initial mesh are given in Table 5. The area submit- meshed with triangular elements (reduced integration). The
ted to shearing was meshed with four-node elements (reduced blade, the blank holder and the die were assumed rigid and
integration) with a typical size of t/30 in the rolling direction defined analytically. The blank holder and the die were fixed
and t/64 in the normal direction. The rest of the sheet was and the blade displacement u was imposed. Several values of
the friction coefficient between the sheet, the die, the blade
and the blank holder have been tested. ALE can be used for
blade displacements u of 0.7 without significant mesh dis-
Table 4 – Hill’s stress function for the two 6xxx grades tortion. Fig. 13b shows a deformed sample corresponding to
F 0.59 a blade displacement of u = 0.25. Locally (beneath the blade)
G 0.59 mesh refinement of about 2 was obtained.
H 0.41
L 1.5
M 1.5
4.2.2. Stress distribution in the sample and the critical
N 1.41 volume element VE
Fig. 14a shows the stress triaxiality distribution in the sheet.

2 2
F(TD − ND ) + G(ND − RD ) + H(RD − TD )
2 Beneath the blade the material is submitted to compressive
¯ = .
+ 2(LTD−RD
2
+ MND−RD
2
+ NRD−TD
2
) stresses. Fig. 14b shows the variation of the stress triaxial-
ity along line AB for several values of the blade radius. For
all blade penetrations and all friction coefficients between 0
and 0.1, a compressive zone is observed beneath the blade.
All blade radii analyzed here (0.02–0.15 mm) lead to a com-
pressive zone close to the blade (point A). The size of this
compressive zone corresponds closely to the value of the blade
radius.
Crack initiation occurs thus in the compressive zone. At the
critical blade displacement uf , a mode I crack propagates. The
compressive zone (with T < 0) close to the blade constitutes
a barrier to crack propagation. If this barrier is overcome the
Fig. 13 – Finite element mesh of the shear sample (a) crack propagates through the sheet during a very small incre-
non-deformed sample. (b) Sample geometry for a blade ment of the blade displacement. The compressive zone beneath
displacement u = 0.25. the blade is the volume element representative of the damage mech-
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497 491

Table 5 – Initial finite element mesh (Fig. 13)


Physical dimension Initial finite element mesh

x y Element type Number of elements

Nodes Gauss points x y Total

Deformation channel t/2 t 4 1 15 64 960


Outside deform. channel 4t t 3 1 180

t is the sheet thickness (1 mm). The blade movement corresponds to the y direction

Fig. 14 – Finite element results for the AM1 grade at a blade penetration of 0.25. (a) Stress triaxiality map. (b) Variation of the
stress triaxiality along the vertical line AB for several values of the blade radius.

anism involved (Fig. 15). The mean values of all field variables 4.2.3. Critical material element Mu (uh )
(stress, strain) in this Eulerian volume have been determined During the blade displacement, different material elements
as a function of the blade displacement uh : flow through the Eulerian volume VE . The material element
embedded for a given blade displacement in the Eulerian vol-
  ume VE has been determined by defining a series of tracers (i.e.
1 1
˙E (uh ) = (uh , x ) dx EE (uh ) = ε(uh , x ) dx (22) points fixed to the flowing material). Fig. 15 shows the mate-
VE VE
VE VE

Fig. 15 – (a) Definition of the representative volume element VE with a size proportional to the blade radius Rblade (b–d).
Representation of the material element Mi for different blade penetrations. For blade penetrations uh smaller than 0.5 the
material element is somewhere in the sheet beneath the Eulerian volume VE (b and c). At a blade penetration uh = 0.5 the
material element Mi is completely embedded in the Eulerian volume VE .
492 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497

Fig. 16 – Finite element results for AM1. Mean values of (a) the plastic strain and (b) the stress triaxiality in different
material elements Mu and in the Eulerian volume VE as a function of the blade displacement uh .

rial element M0.5 (uh ) for different blade displacements uh . For cell dimensions (L, H), the particle width (l = l0 ) and all the
blade penetrations uh smaller than 0.5, M0.5 (uh ) occupies a vol- characteristic angles (, ,˚, ) could thus be determined in
ume inside the sheet beneath the Eulerian volume VE (Fig. 15b the Eulerian volume VE and for all material elements VL(u) (uh ).
and c). At a blade penetration uh = 0.5 M0.5 is embedded in VE . Only the void or ligament height h needs particular attention.


Mu (uh < u) ≡ some volume inside the sheet 4.3.1. Void dimensions
Mu (uh = u) = VE (u) During shearing, no transverse void growth (perpendicular to
the particle width) is observed, thus l corresponds to the ini-
The mean values of the stress and strain tensors in the tial particle width l0 . The height h of the ligament depends on
material elements Mu (uh ) have been determined for each the voids formed around the second phase particles. The evo-
blade penetration uh by interpolating the Gauss point values. lution of a void around a hard particle in a matrix submitted
A complete Lagrangian description following all material to plane strain deformation and constant stress triaxiality T
elements is thus possible. Fig. 16 shows the variation of the was analyzed by finite element calculations (Appendix A). At
mean equivalent plastic strain and the mean stress triaxial- small overall strains, the void extends to a critical value hcrit .
ity as a function of the blade displacement uh . In Fig. 16a, Beyond this critical value the void height decreases asymptoti-
the plastic strain of the critical material element is plotted cally to h∞ . Thus the evolution of the void height in a material
as a bold curve. All the material elements Mu on the left of element Mu depends on the stress triaxiality, the strain, the
this curve correspond to blade displacements uh smaller than initial microstructure and the current value of h:
uf . All curves to the right of the critical one (dashed lines)
correspond to blade displacements uh larger than uf . The com-
⎧ ∂h

⎪ = fgrowth (T, h, Ep , h0 ) h ∈ [h0 , hcrit ]
pressive zone will crack before these material elements are ⎨ ∂Ep
embedded in VE . The stress strain history for blade displace- ∂h (23)
⎪ = fdecrease (T, h, Ep , h0 ) h ∈ [hcrit , h∞ ]
ments larger than uf , i.e. implying crack propagation, is not ⎪
⎩ ∂Ep ˇ 1−ˇ
analyzed in present work. ˇ verifies h∞ = hcrit h0

4.3. Critical blade displacement at the onset of The major disadvantage of this formulation is its history
cracking dependency. As the evolution law depends on the current
value of the void height h and the stress strain path, it has
In the following section the new model is applied. The crit- to be integrated for each material element considered. This
ical values of DLST at the onset of strain localization were evolution law may not be applied to the Eulerian volume VE
determined previously (Bacha et al., 2007) in simple tension corresponding to different material elements and integration
(Table 6). The shearing of a regular microstructure being dis- for all the materials elements Mu is necessary. Fig. 17 shows
carded as non-realistic, the shear angle  was determined by a schematic representation of the void height in a material
(16). Experimentally, the particle rotation was assumed equal element as a function of the overall strain in this material
to the material rotation . The stress ratio (˙ pm / y ), the new element. Three different cases may be considered; no void
growth at all (h = h0 ), void growth to the critical value hcrit
(ˇ = 1, h∞ = hcrit ) and void growth followed by shrinkage (ˇ = 0,
h∞ = h0 ).
Table 6 – Critical values of the damage parameter DLST
for AM1- and AM2-grades
4.3.2. Trimming of AM1 and AM2 in the reference tool
Grade Dtension
BKD geometry
AM1 1.30 The model has first been applied to the shearing of AM1 and
AM2 1.32 AM2 in the reference tool geometry. The shearing of a reg-
ular microstructure being discarded as non-realistic, three
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497 493

Table 7 – Experimental and predicted values of the


critical blade displacement uf (%) at crack initiation for
the two aluminium grades AM1 and AM2 sheared in
reference tool geometry
Grade Experiment Model

Friction ␮ No void Void growth


growth
ˇ=0 ˇ=1

AM1 23 0.0 19 19 19
0.1 20 20 20
Fig. 17 – Schematic evolution of the void height in a AM2 38 0.0 40 39 39
material element as a function of the strain ε of this 0.1 31 30 31
material element. For strains smaller than εcrit , the void
AM2–AM1 15 0.0 21 20 20
height increases. Depending on the remote stress
triaxiality, strains larger than εcrit lead to a constant or
decreasing void height.
displacement uf , reducing the particle spacing leads to a very
rapid increase of the variable DLST .
different versions of the same model were tested. These three The critical blade displacements corresponding to DLST =
versions were tested for zero friction between the sheet and Dtension
LST are summarized in Table 7 for the three ver-
the blade and a friction coefficient of 0.1 between the blade sions of the model. The difference in ductility predicted
and the sheet. In the first version zero void growth (h = h0 ) between the two grades is larger than the experimen-
is assumed. In the second version void growth followed by tal data. But, to our knowledge, the values predicted are
void shrinking (for h > hcrit ) was considered (ˇ = 0, h∞ = h0 ) and much closer to reality than the predictions of any other
in the last version void growth without void shrinking (ˇ = 1, model dedicated to shearing. The difference between the
h∞ = hcrit ) was considered. version without void growth and the two versions consid-
Fig. 18 shows the variation of the damage variable as a func- ering void growth is negligible. Thus, the void height h has
tion of the blade displacement for both aluminium grades in no real influence on the critical damage leading to failure
the first version of the model (no void growth). The evolu- in shearing and the value of DLST may then be determined
tion of the damage variable in the Eulerian volume and in without considering void growth in the Eulerian volume
the material elements Mu is shown. The damage variable in VE .
the Eulerian volume is represented for blade displacements
larger than 0.1 only. In fact, for blade displacements smaller 4.3.3. Effect of tool geometry on the shearing of AM1
than 0.1, the Eulerian volume contains no material. The dam- AM1 exhibits a smaller blade displacement uf at crack initi-
age in the Eulerian volume increases with increasing blade ation. Different tool geometries were tested experimentally
displacements. As the variable DLST expresses the possibility and numerically. In each test, one parameter was changed
of failure, there is no reason for its initial value to be zero. with respect to the reference tool geometry (Table 2). The cor-
At the beginning of the test, the value of DLST in the differ- responding experimental and numerical values of the blade
ent material elements decreases. The shearing of the material displacement at crack initiation uf are given in Table 8. Zero
ligament leads to a small decrease of the angle . For larger friction between tool and sheet was assumed. The experimen-
blade displacements, the material rotation  overcomes this tal data, i.e. the effect of changing the tool geometry on the
effect and the angle is close to
/2. Close to the critical blade blade displacement at crack initiation, are predicted closely.

Fig. 18 – FE-model results for void growth to a critical value hcrit . Damage variable DLST as a function of the blade
displacement for AM1 (a) and AM2 (b) alloys sheared in the reference configuration without friction and void growth.
494 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497

parameters, except the void height, used in the damage vari-


Table 8 – Experimental and predicted values of the
critical blade displacement uf (%) at crack initiation for able DLST are given by the f.e. simulations of the forming or
AM1 sheared in different tool geometries shaping process. In most forming processes, the stress triax-
iality is so small that the void height will have no influence
Ref. geom. C 20 % ˛ Rblade Scrap
30% holder on the value of the damage variable. If void growth is signifi-
10◦ 20◦ cant, the present model can be combined with any void growth
model.
Exp. 23 28 26 38 76 21
Model 19 25 30 35 68 24
5.1.3. The new damage model gives reliable and useful
results
The new model predicts closely the blade displacements at
An increased clearance between blade and die or the use of a crack initiation observed in the different trimming tests. Sim-
scrap holder do not affect significantly the blade displacement ulating correctly trimming is a very severe test for any model
at crack initiation. A cutting angle ˛ larger than 15◦ leads to due to the very large strains and material rotations. The
significant increase of uf . This effect is due essentially to a present model predicts correctly the influence of the alloy
reduction of the bending stresses. But of course, the cutting microstructure (AM1 and AM2) and the influence of the tool
angle may not always be changed in an industrial produc- geometry (cutting angle, clearance, blade radius) on the blade
tion line. The blade radius Rblade has the largest influence on displacement at crack initiation. The present damage model
the blade displacement at crack initiation. uf increases rapidly may thus be used to optimise the alloy microstructure and the
with increasing blade radius Rblade . This is a consequence of an tool geometry for shaping or trimming operations on sheet
increased compressive zone, i.e. an increased barrier to crack metal.
propagation. Moreover changing the blade radius is simple on
most industrial production lines. 5.1.4. Additional computational costs due to the new
damage model
The damage model may be applied to stress triaxialities larger
5. Concluding remarks
than 0.4 using any adequate void growth model. The dam-
age variable is estimated by a post processing procedure and
The main outcomes of this work are: (1) a void coalescence the CPU-time of the finite element-solver is not affected. The
model valid at low stress triaxiality, (2) a damage criterion valid calculation of the damage variable in each material element
at small stress triaxiality and large material rotations, (3) a or in the Eulerian volume is straight forward. If the damage
damage variable expressed as simple closed form for materials variable is determined in a fixed Eulerian volume the addi-
containing second phase particles. The damage analysis in a tional computational costs correspond to the calculation of
small fixed volume with a representative microstructure (Eule- the mean stress tensor in this volume. If all possible material
rian approach) and the damage analysis in all the material elements have to be tested (Lagrangian approach), the addi-
elements of the considered structure are compared in detail. tional computational costs correspond to the calculation of
the mean stress tensor in these material elements. But even
5.1. The new damage model this approach leads to a negligible increase of CPU-time.

5.1.1. The new damage model is based on physical 5.2. Physical and mechanical analysis of damage
observations
Most damage models assume implicitly void growth and coa- The new damage variable has to be applied to the mean values
lescence. But the present work has shown that low stress determined in the representative volume and representative
triaxiality damage is due to bringing second phase parti- material elements. In fact, material forming leads to large
cles closer together during deformation. During macroscopic strain and stress gradients. The finite element mesh used to
deformation, under low stress triaxiality, the second phase analyze this stress strain distribution is of the same size as the
particles themselves and the matrix particle interface may underlying microstructure.
break, but, in the volume relevant for crack initiation, a com- The present work shows how to combine the microstruc-
pressive stress state is observed and the voids shrink due to tural information, the mechanical analysis of the shear test,
particle or interface break-up. the physical (microscopic) observation of the real damage
The new damage model is based on these physical obser- mechanism and the new damage model. The definitions of the
vations. Strain localization in the ligament between second representative volume element and the representative mate-
phase particles is considered. Thomason’s upper-bound solu- rial elements have to be based on the physics of the damage
tion for strain localization in a periodic arrangement of mechanism but should also include the real boundary condi-
rectangular voids was extended to large deformations includ- tions (blade radius) of the process.
ing large material rotations. The values used in the damage variable are mean val-
ues in the representative volume element. But the material
5.1.2. The new damage model is easy to use in industrial arranges to find the weakest microstructural feature (e.g. the
shaping or forming processes successive particle layers with the smallest interdistance).
The critical value of the damage parameter DLST may be deter- Thus the mechanical values (stress, strain and material ori-
mined by a tensile test and no data fitting is needed. All entation) have to be determined as the mean values in the
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497 495

representative volume. But the “worst” possible case of the to plane strain deformation and to homogeneous vertical and
microstructure will lead to the closest material behaviour with horizontal displacements u3 and u1 . The macroscopic prin-
respect to damage. cipal strains and the effective strain, respectively, are given
by:

Appendix A L H
E11 = ln , E33 = ln (A1)
L0 H0
A.1. Void growth in a unit cell submitted to plane strain
deformation and constant remote stress triaxiality The corresponding macroscopic stresses ˙ 22 and ˙ 33 are
the forces on the boundaries divided by the current areas. Dur-
During shearing, no transverse void growth (perpendicular to ing the loading history, the overall strain rate Ė33 was imposed
the particle width) is observed, thus the void width l corre- and the triaxiality ratio T was kept constant.
√ √
sponds to the initial particle width l0 (Fig. 19a). The height Two constant stress triaxiality ratios (T = −1/ 3, T = 1/ 3)
h of the ligament depends on the voids formed around the and two initial void shape factors have been considered
second phase particles. The evolution of a void around a (l0 /h0 = 0.47, l0 /h0 = 1). Fig. 19 illustrates the evolution of a void
hard particle in a matrix submitted to plane strain defor- in an AM1 alloy matrix submitted to a constant stress triax-
mation and constant stress triaxiality T was analyzed by iality. At small overall strains, the void extends to a critical
finite element cell calculations for several values of the stress value hcrit . Beyond this critical value the void height decreases
triaxiality T. The unit cell calculations performed in this asymptotically.
study follow a long history of finite element analyses of void
growth in plastically deforming materials, initiated by the 5.2.2. Cell model results
work of Needleman in 1972, and pursued by many others (Chu The cell model calculations show three different steps in the
and Needleman, 1980; Tvergaard, 1981, 1982; Needleman and void height evolution:
Tvergaard, 1984; Thomson et al., 2003; Kim et al., 2004; Pardoen
and Hutchinson, 2000; Koplik and Needleman, 1988; Hom and • Void growth up to a critical value hcrit .
McMeeking, 1989a,b; Worswick and Pick, 1990; Brocks et al., • Void shrinkage from hcrit to h∞ .
1995; Steglich and Brocks, 1997; Gao et al., 1998; Faleskog and • Void remains unchanged.
Shih, 1997; Kuna and Sun, 1996; Bordreuil et al., 2003).
During the growth phase the void has a trapezoidal shape
5.2.1. Problem formulation with constant base angle ˛ and constant base length l0 . The
The matrix material is described by the generalized Voce law void height h increases. If the void height reaches the critical
for AM1 or AM2. A rectangular cell of initial height H0 and values hcrit , corresponding to a triangular shape, void shrink-
width L0 (Fig. 19b) containing a rectangular particle of width age starts. Void growth does depend strongly on the material
l0 and length h0 is considered (Fig. 19b). The cell is subjected behaviour (AM1 or AM2). Void shrinkage (occurs only under

√ √
Fig. 19 – Void growth in a matrix submitted to constant stress triaxiality T ∈ [−1/ 3, 1/ 3]. (a) FEG SEM image of voids in the
compressive zone of an AM1 sample at different deformations. (b) Fe-mesh and evolution of the void shape. (c) Evolution of
the void height h as a function of the macroscopic equivalent strain for AM1 and AM2 material behaviours and two constant
stress triaxiality ratios. (d) Schematic evolution of the void height from h0 to hcrit and h∞ .
496 j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497

negative stress triaxialities) depends strongly on the material Bacha, A., Feuerstein, M., Desrayaud, Ch., Klöcker, H., 2006a.
behaviour. In AM1 complete void closure is observed whereas Measuring stress strain curves to large strain on sheet metal.
in AM2, the void height stabilizes at h∞ > h0 . J. Eval. Test. 35, 1812–1821.
Bacha, A., Maurice, Cl., Klöcker, H., Driver, J.H., 2006b. The large
strain flow stress behaviour of aluminium alloys as measured
5.2.3. Heuristic model for the void growth by channel-die compression (20–500 ◦ C). Proc. Mater. Sci.
The void growth stage is described by a Rice and Tracey model Forum 519–521, 783–788.
Bacha, A., Klöcker, H., Daniel, D., 2007. On the determination of

5 3 ˙
 true stress triaxiality in sheet metal. J. Mater. Process.
√ = Ė33 + T Ē (A2) Technol. 184, 272–287.
hl 3 4
Bao, Y., Wierzbicki, T., 2004. On the fracture locus in the
Assuming unidirectional void growth (l = l0 ) leads to the equivalent strain and stress triaxiality space. Int. J. Mech. Sci.
46, 81–98.
following evolution equation for the void height during the
Benzerga, A.A., Besson, J., Pineau, A., 2004a. Anisotropic ductile
growth stage: fracture. Part I. Experiments. Acta Mater. 52, 4623–4638.

 √  Benzerga, A.A., Besson, J., Pineau, A., 2004b. Anisotropic ductile


dh 5 3 3 fracture. Part II. Theory. Acta Mater. 52, 4639–4650.
√ = a l0 + T dĒ (A3)
h 6 4 Beremin, F.M., 1981a. Experimental and numerical study of the
different stages in ductile rupture: application to crack
a is a parameter. The f.e. analysis shows that growth stops initiation and stable crack growth. In: Nemat-Nasser, S. (Ed.),
Three-Dimensional Constitutive Relations and Ductile
if the void reaches a triangular shape of height hcrit :
Fracture. North-Holland Publishing Company, pp. 185–205.
Beremin, F.M., 1981b. Cavity formation from inclusions in ductile
hcrit l0
= 1 + a tan ˛T (A4) fracture of A508 steel. Metall. Trans. A 12, 723–731.
h0 h0 Besson, J. (Ed.), 2004. Local approach to fracture. Collective book,
Presses de l’Ecole des Mines de Paris.
where ˛T is the critical value of the angle between the particle Bordreuil, C., Boyer, J.-C., Sallé, E., 2003. On modeling the growth
and the triangular void (Fig. 19). a is a numerical parame- and the orientation changes of ellipsoidal voids in a rigid
ter tanking account of non perfectly triangular void shapes. plastic matrix. Mod. Simul. Mater. Sci. Eng. 11, 365–380.
The f.e. analysis shows that ˛T is almost independent of Brocks, W., Sun, D.Z., Hönig, A., 1995. Verification of the
the material behaviour but depends strongly on the remote transferability of micromechanical parameters by cell model
√ √ calculations with visco-plastic materials. Int. J. Plast. 11,
stress triaxiality (T = −1 3 leads to ˛T = 58◦ and T = 1 3 leads to
971–989.
˛T = 90). A linear fit gives a good approximation of the numer- Brown, L.M., Embury, J.D., 1973. The initiation and growth of voids
ical results at second phase particles. In: Proceedings of the Third
International Conference on the Strength of Metals and
˛T = 0.483 T + 1.291 (A5) Alloys, I CSMA 3, Cambridge, England, pp. 164–169.
Browning, A., Spitzig, W.A., Richmond, O., Teirlink, D., Embury,
J.D., 1983. The influence of hydrostatic pressure on the flow
5.2.4. Heuristic model for the void closure
stress and ductility of a spheroidized 1045 steel. Acta Metall.
During void closure, ˛ = ˛T and the base length of the void l̂ is
31, 1141–1150.
related to the void height h by Brunet, M., Morestin, F., 2001. Experimental and analytical
necking studies of anisotropic sheet metals. J. Mater. Process.
h l̂
h
crit Technol. 112, 214–226.
=1+ −1 (A6)
h0 l0 h0 Budiansky, B., Hutchinson, J.W., Slutsky, S., 1982. Void growth and
collapse in viscous solids. In: Hopkins, H.G., Sewell, M.J. (Eds.),
Applying the Rice and Tracey formulation for the void base Mechanics of Solids, The Rodney Hill 60th Anniversary
length leads to an evolution equation for l̂: Volume. Pergamon Press, pp. 13–45.
Christman, T., Needleman, A., Nutt, S., Suresh, S., 1989. On
 √  microstructural evolution and micromechanical modeling of
dl̂ hcrit + h0 5 3 3 deformation of a whisker-reinforced metal–matrix composite.
= − + T dĒ (A7)
2 6 4 Mater. Sci. Eng. A 107, 49–61.

Chu, C.C., Needleman, A., 1980. Void nucleation effects in
biaxially stretched sheets. J. Eng. Mater. Technol. 102,
The void height decreases to h∞ depending on the remote
249–256.
stress triaxiality.
Doghri, I., Ouaar, A., 2003. Homogenization of two-phase
elasto-plastic composite materials and structures—study of
ˇ 1−ˇ
∀ˇ ∈ [0, 1], h∞ = hcrit h0 (A8) tangent operators, cyclic plasticity and numerical algorithms.
Int. J. Solids Struct. 40, 1681–1712.
references Faleskog, J., Shih, C.F., 1997. Micromechanics of coalescence. I.
Synergistic effects of elasticity, plastic yielding and
multi-size-scale voids. J. Mech. Phys. Solids 45, 21–45.
Gammage, J., Wilkinson, D., Brechet, Y., Embury, D., 2004. A model
Argon, A.S., 1976. Formation of cavities from non deformable for damage coalescence in heterogeneous multi-phase
second-phase particles in low temperature ductile fracture. J. materials. Acta Mater. 52, 5255–5263.
Eng. Mater. Technol. 98, 60–68. Gao, X., Faleskog, J., Shih, C.F., 1998. Cell model for nonlinear
Babout, L., Brechet, Y., Maire, E., Fougeres, R., 2004. On the fracture analysis. II. Fracture-process calibration and
competition between particle fracture and particle decohesion verification. Int. J. Fract. 89, 374–386.
in metal matrix composites. Acta Mater. 52, 4517–4525.
j o u r n a l o f m a t e r i a l s p r o c e s s i n g t e c h n o l o g y 2 0 3 ( 2 0 0 8 ) 480–497 497

Gologanu, M., Leblond, J.-B., Perrin, G., Devaux, J., 1995. In: Needleman, A., Tvergaard, V., 1984. An analysis of ductile rupture
Suquet, P. (Ed.), Continuum Micromechanics. Springer-Verlag, in notched bars. J. Mech. Phys. Solids 32, 461–490.
Berlin, p. 61. Pardoen, T., Hutchinson, J.W., 2000. An extended model for void
Gologanu, M., Leblond, J.B., Perrin, G., et al., 2001. Theoretical growth and coalescence. J. Mech. Phys. Solids 48, 2467–2512.
models for void coalescence in porous ductile solids. I. Pardoen, T., Hutchinson, J.W., 2003. Micromechanics-based model
Coalescence “in layers”. Int. J. Solids Struct. 38, 5581–5594. for trends in toughness of ductile metals. Acta Mater. 51,
Gurland, J., Plateau, J., 1963. The mechanism of ductile rupture of 133–148.
metals containing inclusions. Trans. Am. Soc. Metals 56, Pardoen, T., Dumont, D., Deschamps, A., Brechet, Y., 2003. Grain
442–454. boundary versus transgranular ductile failure. J. Mech. Phys.
Gurson, A.L., 1977. Continuum theory of ductile rupture by void Solids 51, 637–665.
nucleation and growth. Part I. Yield criteria and flow rules for Pineau, A., 1992. Global and local approaches of
porous ductile media. J. Eng. Mater. Technol. 99, 2–15. fracture—Transferability of laboratory test results to
Hom, C.L., McMeeking, R.M., 1989a. Void growth in elastic–plastic components. In: Argon, A.S. (Ed.), Topics in Fracture and
materials. J. Appl. Mech. 56, 309–317. Fatigue. Springer-Verlag, pp. 197–234.
Hom, C.L., McMeeking, R.M., 1989b. Three-dimensional void Ragab, A.R., 2004. Application of an extended void growth model
growth before a blunting crack tip. J. Mech. Phys. Solids 37, with strain hardening and void shape evolution to ductile
395–415. fracture under axisymmetric tension. Eng. Fract. Mech. 71,
Huang, Y., 1991. Accurate dilatation rates for spherical voids in 1515–1534.
triaxial stress fields. J. Appl. Mech. 58, 1084–1085. Rice, J.R., Johnson, M.A., 1970. The role of large crack tip geometry
Huber, G., Brechet, Y., Pardoen, T., 2005. Void growth and void changes in plane strain fracture. In: Kanninen, M.F., Adler,
nucleation controlled ductility in quasi eutectic cast W.G., Rosenfield, A.R., Jaffee, R.I. (Eds.), Inelastic Behavior of
aluminium alloys. Acta Mater. 53, 2739–2749. Solids. McGraw-Hill, pp. 641–672.
Joly, P., Meyzaud, Y., Pineau, A., 1992. Micromechanisms of Rice, J.R., Tracey, D.M., 1969. On the ductile enlargement of voids
fracture of an aged duplex stainless steel containing a brittle in triaxial stress fields. J. Mech. Phys. Solids 17, 201–217.
and a ductile phase: development of a local criterion of Scheyvaerts, F., Onck, P.R., Brechet, Y., Pardoen, T., A multiscale
fracture. In: Advances in Local Fracture/Damage Models for model for the cracking resistance of 7000 Al. In: Besson, J.,
the Analysis of Engineering Problems, AMD, vol. 137. The Moinereau, D., Steglich, D. (Eds.), Proceedings of the 9th
American Society of Mechanical Engineers, pp. 151–180. European Mechanics of Materials Conference on Local
Kim, J., Gao, X.S., Joshi, S., 2004. Modeling of void growth in Approach to Fracture, Euromech - Mecamat 2006.
ductile solids: effects of stress triaxiality and initial porosity. Moret-sur-Loing, 9–12 May 2006, France, Les Presses de l’Ecole
Eng. Fract. Mech. 71, 379–400. des Mines de Paris, p. 447–452.
Klöcker, H., Tvergaard, V., 2003. Growth and coalescence of Shabrov, M.N., Needleman, A., 2002. An analysis of inclusion
non-spherical voids in metals deformed at elevated morphology effects on void nucleation. Modell. Simul. Mater.
temperature. Int. J. Mech. Sci. 25, 1283–1308. Sci. Eng. 10, 163–183.
Koplik, J., Needleman, A., 1988. Void growth and coalescence in Siruguet, K., Leblond, J.-B., 2004. Effect of void locking by
porous plastic solids. Int. J. Solids Struct. 24, 835–853. inclusions upon the plastic behavior of porous ductile solids.
Kroon, M., Faleskog, J., 2005. Micromechanics of cleavage fracture I. Theoretical modeling and numerical study of void growth.
initiation in ferritic steels by carbide cracking. J. Mech. Phys. Int. J. Plast. 20, 225–254.
Solids 53, 171–196. Steglich, D., Brocks, W., 1997. Micromechanical modeling of the
Kuna, M., Sun, D.Z., 1996. Three-dimensional cell model analyses behavior of ductile materials including particles. Comput.
of void growth in ductile materials. Int. J. Fract. 81, 235–258. Mater. Sci. 9, 7–17.
Lautridou, J.C., Pineau, A., 1981. Crack initiation and stable crack Thomason, P.F., 1990. Ductile Fracture of Metals. Pergamon Press,
growth resistance in A508 steels in relation to inclusion Oxford.
distribution. Eng. Fract. Mech. 15, 55–71. Thomson, C.I.A., Worswick, M.J., Pilkey, A.K., Lloyd, D.J., 2003.
Leblond, J.-B., Perrin, G., Suquet, P., 1994. Exact results and Void coalescence within periodic clusters of particles. J. Mech.
approximate models for porous viscoplastic solids. Int. J. Phys. Solids 51, 127–146.
Plast. 10, 213–235. Tvergaard, V., 1981. Influence of voids on shear band instabilities
Leblond, J.-B., Perrin, G., Devaux, J., 1995. An improved under plane strain conditions. Int. J. Fract. 17, 389–407.
Gurson-type model for hardenable ductile metals. Eur. J. Tvergaard, V., 1982. On localization in ductile materials
Mech. A/Solids 14, 499–527. containing voids. Int. J. Fract. 18, 237–251.
Lee, B.J., Mear, M.E., 1999. Stress concentration induced by an Voce, E., 1948. True stress–strain curves and their application to
elastic spheroidal particle in a plastically deforming solid. J. cold-working processes. Metal Treatment 15, 53–60.
Mech. Phys. Solids 47, 1301–1336. Worswick, M.J., Pick, R.J., 1990. Void growth and constitutive
McClintock, F.A., 1968a. A criterion for ductile fracture by the softening in a periodically voided solid. J. Mech. Phys. Solids
growth of holes. J. Appl. Mech. 35, 353–371. 38, 601–625.
McClintock, F.A., 1968b. Local criteria for ductile fracture. Int. J. Worswick, M.J., Chen, Z.T., Pilkey, A.K., Lloyd, D., Court, S., 2001.
Fract. 4, 103–130. Damage characterization and damage percolation modeling
Needleman, A., 1972. Void growth in an elastic–plastic medium. J. in aluminum alloy sheet. Acta Mater. 49, 2791–2803.
Appl. Mech. 39, 964–970.

You might also like