Engineering Fracture Mechanics 68 (2001) 365±378

www.elsevier.com/locate/engfracmech

Finite element model fracture prediction during sheet-metal

blanking processes
Ridha Hambli a,b,*
a
ISTIA-LASQUO, 62, Avenue Notre Dame du Lac, 49000 Angers, France
b
ENSAM ± LPMI, 2, Bld du Ronceray, B.P. 3525, 49035 Angers Cedex, France
Received 9 November 1998; received in revised form 10 December 1999; accepted 22 September 2000

Abstract
In order to accurately simulate sheet-metal cutting processes by shearing mechanisms, such as blanking and punching,
we have developed a ®nite element model (FEM) valid for the numerical description of such processes. Damage and crack
propagation have been taken into account by means of an elastoplastic constitutive law. To study the e€ects of varying the
process parameters on the geometry of the sheared edges, and the evolution of the force-punch penetration, we have
implemented a calculation algorithm by means of the user routine (UMAT) of ABAQUS/standard ®nite element code. Final
results given by the FEM were compared with the experimental ones. Ó 2001 Elsevier Science Ltd. All rights reserved.

Keywords: Fracture; Damage; Blanking process; Finite element; Experiment

1. Introduction

Sheet-metal forming is, after machining, the most widely used process in industry. Certain processes, such as
blanking, lead to the total rupture of the sheet as opposed to other processes, such as stamping and folding
which are aimed at deforming the sheet plastically. As the process is being performed the material whose
behavior is non-linear, is submitted to complex stress and strain states. Before complete rupture, the material is
subjected to ductile damage and crack propagation phenomena. To describe the sheetÕs behavior while the
process is being carried out, various parameters may be used such as the material state of hardening and
damage, the metallurgical morphology, the crystallographical texture and the intragranular structures as they
evolve during the procedure [4,8,11,13,19].
The main diculty encountered in simulating this operation is describing the behavior of the sheet
continuously from the beginning of the operation up to the total rupture [9,10,17]. In this case, choosing

*
Address: ISTIA-LASQUO, 62, Avenue Notre Dame du Lac, 49000 Angers Cedex, France. Tel.: +33-2-41-36-57-57; fax: +33-2-4136-57-53.
E-mail address: ridha.hambli@istia.univ-angers.fr (R. Hambli).
366 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378

0013-7944/01/$ - see front matter Ó 2001 Elsevier Science Ltd. All rights reserved.
PII: S0013-7944(00)00106-5

Fig. 1. Schematic diagram of punching forces vs. punch penetration from Ref. [8]. 1: Elastic stage, 2: elastoplastic stage, 3: elastoplastic
stage in which damage occurs, 4: initiation and propagation of cracks leading to ®nal rupture.

a behavior law is crucial, since each successive step in the whole process, as illustrated in Fig. 1, has to be
described accurately.
From a numerical point of view, a good description of the above stages requires the development of reliable
algorithms. Thus, the simulation of crack initiation and propagation can be modeled, without diverging
computation. In previous studies that have dealt with this topic, simpli®ed hypothesis were formulated
sometimes in a rather coarse manner.
The aim of this work was to provide a general FEM capable of numerically simulating the whole blanking
process.
The numerical results obtained by the simulation were compared with the experimental ones in order to
verify the validity of the proposed FEM.

2. Modeling the ductile fracture

Accurate knowledge of the failure process is essential to the selection of a suitable damage model. In the
case of sheet blanking by shearing processes, numerous authors have studied the di€erent physical
mechanisms leading to the ®nal rupture, and proposed their own models.
Dos Santos et al. [4] carried out a viscoplastic study of a rectangular-bar cropping operation. They analyzed
the deformation of a pattern engraved on the surface of the sheared area of the testing bar. They
demonstrated, by means of observation that the rupture is of a ductile nature and that the cracks generated by
the cropping process follow an ÔSÕ, or double ÔSÕ pro®le, parallel to the direction of maximum intensity of
the shear stress.
These results correspond to those obtained by Wong and Das [19] with regards to crack initiation. Examining
the faces of the sheared edge, Jana and Ong [11], con®rmed that the blanked edge exhibits cupules which
characterize the ductile fracture.
Recently, an analytical study and experimental tests [7,8], showed that the physical mechanisms leading to
the complete failure of the sheet material can be described as follows:
First, due to the penetration of the punch into the sheet, cracks initiate at the cutting edges ÔAÕ and ÔBÕ of
the tools (Fig. 2). Secondly, the cracks propagate when and where the equivalent strain exceeds a critical value,
and thus progressively cut the material ®bres one after the other.
A conclusion drawn from the above papers, is that the description of the damage evolution leading to cracks
initiations and propagations must take into account the equivalent plastic strain and the hydrostatic pressure
value rH. Such models [3,5,6,14,16] would be suitable to describe the damage process of the material before
®nal fracture.
In the Le Ma^õtreÕs [14] model for isotropic damage evolution, which has been retained in the following,
the damage law is written in the incremental form:
 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378 367

8<
: 0R; c D 23 eqH 2 p for epp 6eDD

dD 1
ˆ D
e ÿe 1 ‡ m† ‡ 3 1 ÿ 2m† rr de ; for e > e †

ep the equivalent plastic strain:

2
ˆ
ep r3 eijpl eijpl 2†

eijpl is the plastic part of the strain tensor.

eD is the threshold logarithmic strain at which damage initiates, e R is the logarithmic strain value at fracture.
Dc is the critical damage value at fracture. This model requires the knowledge of three parameters D c, eR and eD
which are easily identi®ed by means of a tensile test [14].

3. Finite element approach

The law describing the material behavior should describe the di€erent stages of the process observed
experimentally from the initial elastic state to the ®nal rupture. For this, a behavioral law including damage and
failure phenomena must be chosen. In this paper, the behavior law coupled with Le Ma^õtreÕs [14] damage
law has been employed.

3.1. Local integration

From a numerical point of view the problem consists in resolving the constitutive equations describing the
behavior of the sheet.
The set of equations de®ning the problem are derived from:

1. The decomposition of the deformation into an elastic part and a plastic part:
etot ˆ eel ‡ epl 3†
2. The elastic law coupled with Le Ma^õtreÕs damage law:

rij ˆ 1 ÿ D† ke dijekkel ‡ 2le eijel† 4†

368 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378

or in Jaumann (corotational) rate form:

r_ijJ ˆ 1 ÿ D† ke dij e_kkel ‡ 2le e_ijel† 5†

where ke and le are LameÕs constants.

3. The plastic ¯ow law (implicit scheme): of
Dep ˆ Dk or; k being the plastic multiplier: 6†

4. The integration of the damage law:

D ˆ D rij;eij;ak† 7†

where ak indicates the intrinsic coecients of the material.

5. The integration of the yield function coupled with damage:

f ˆ req ÿ 1 ÿ D† rel ‡ r0† 8†

where rel is the initial yield stress, and r0 is the isotropic non-linear hardening law.

The integration methods of the non-linear constitutive equations are based on the use of a special algorithm
which solves the equations in incremental form. For this purpose, during a small time interval ‰tn;tn‡1Š, we
suppose that the whole increment is purely elastic, then an elastic prediction is de®ned as:

rnT‡1 ˆ rn ‡ Dr ˆ 1 ÿ Dn†Cel etot†n‡1 ÿ epl†n 9†

The superscript † T refer to Trial test. f < 0, the prediction is true and the local procedure
If this elastic predictor satis®es the yield condition: is
completed. Then it can be stated that:

rn‡1 ˆ rnT‡1 10†

Otherwise, this state must be corrected by means of a plastic correction de®ned as developed as follow:
The variables at tn‡1 must satisfy the yield condition, i.e. the evolution laws written in the incremental form
and elasticity law must satisfy the system:
fˆ0 11†
rn‡1 ÿ ÿ1 Dn‡1†Celÿen ‡ De ÿ enpl ÿ Depl ˆ 0
12†

13†
DHa ˆ haDeij; rij†n‡1;Hija;Dn‡1
In this case, we must project the stresses rnT‡1 on the surface of the yield function df, as follows:

rn‡1 ˆ PÿrnT‡1 T 14†

 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378 369

whereThe Newton method is used to solve the non-linear global equilibrium equations as well as the non-
linearP: R5 ! df: denotes the orthogonal projection on df of the trial stresses r n‡1. local equations obtained by full
implicit integration of the constitutive equations. The consistent local tangent modulus is obtained by the exact
linearization of the algorithm. In comparison with the use of the analytical ÔstandardÕ modulus, the consistent
modulus leads to a faster convergence rate when the Newton algorithm is used to perform iterations in the
balance equations. In this way, we have developed an implicit algorithm which allows for the integration of the
constitutive equations. The scheme obtained seems well suited to the non-linearity of the behavioral law
[2,12,15,18].
The described procedure, was implemented using the general purpose code ABAQUS with the help of the user
subroutine UMAT [1].

Fig. 3. Modeling of crack propagation.

Within the framework of the displacement formulation of FEM, the global equilibrium equations to be
satis®ed at each instant tn‡1 can be written in the general form [1]:

FU n‡1† ˆ0 15†

Un‡1 is the displacement ®eld at step (n+1).

If this non-linear problem is solved iteratively by a Newton method, at each global iteration r the following
equation can be written:

‡
F Uÿ r
n‡1 ÿKrn‡1ÿUrn‡‡11 ÿ Urn‡1 ˆ 0 16†

where

r oF r 1 Z T r 1

Kn‡1 ˆ oU n‡ ˆ X B Jn‡ BdX 17†

B is the strain-displacement matrix and Jrn‡1 is the Jacobian tensor obtained by:

ˆ o r
Jrn‡1 o e nr‡1 18†
370 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378

3.2. Modeling crack initiation and propagation

During the analysis, the initiation of crack is supposed to occur at any point in the structure where the
damage reaches its critical value Dc. The crack propagation is simulated by the propagation of a completely
damaged area. From a numerical point of view, the damage value is abruptly set to the value D R (near the value
1). This method leads to the decrease in the sti€ness of the elements concerned. These concepts are illustrated
by Fig. 3 in which the propagation of cracks is illustrated.

4. Damage model calibration for the mesh

In the conventional approach of fracture mechanics, material failure occurring in crack tip process zone is
assumed to be governed by some stress ®eld and the crack growth can be correlated with these parameters. In
this paper, the continuum damage mechanics approach has been used to describe the crack growth into the
sheet during the blanking process. In this approach, a crack propagation is modeled by a completely damaged
zone.
Mesh-dependence in ®nite element analysis of crack growth by continuum damage mechanics approach and
its regularization has been investigated by di€erent authors [20±23]. The prediction of the crack

Fig. 4. Illustration of a bar under uniaxial loading.

initiation time, the crack growth and the ®nal rupture time have been observed to be sensitive to the size of
®nite element discretization. The mesh dependence phenomenon has been mainly attributed to a consequence
of the strain softening induced by the presence of damage. The regularization method used in this paper
consists in calculating the damage law coecients by taking into account the mesh-size of the elements [1].
The strain energy just prior rupture in the case of a straight bar subjected to uniform traction applied over its
ends (Fig. 4) is equals to:

WˆWV 19†

W indicates the strain energy density and V indicates the material volume accumulating this energy. The
strain energy of the material is given by the area under the stress±strain curve by:
W ˆ Z0 rde 20† eR
Eq. (20) can be written in following form:

W ˆ W Sl 21†
 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378 371

S indicates the cross sectional area of the bar and l (Fig. 4), the length of the zone which undergoes the
accumulation of the strain energy density.
In fracturing, the energy required to create a crack of length l and width w can be determined using the
energy release rate of the material Ws de®ned by:

Ws ˆ W l 22†

In the following, we consider the same tensile bar (Fig. 4) discretized in a certain number of ®nite elements
as shown in Fig. 5.
In order to simplify calculation, we will consider that the material is elastic perfectly with zero o€set plastic
strain, the strain hardening law takes the form:
r < ry if e ˆ 0 23a†

r ˆ ry if e ˆ ep 6ˆ 0 ep 23b†

The strain energy of the material just prior rupture of a ®nite element is given by:
ˆ 24†
U Z0 rde e
R

Substituting Eq. (23b) into Eq. (24), the strain energy becomes:

U ˆ ry e R 25†

Ue ˆ UVe 26†

According to Fig. 5, it can be written that:

Se ˆ bc and Ve ˆ DeSe 27†

Using Eq. (27), Eq. (26) becomes:

Ue ˆ U DeSe 28†

Then the energy release can be obtained as:

Therefore for one ®nite element of Ve volume, the strain energy Ue can be written in the following form:
372 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378

Us ˆ U De 29†

The FEM should allow the description of the experimental behavior of the tensile bar, therefore we obtain
the following relation:
Ws ˆ U s 30†

Using Eq. (25), Eq. (30) becomes:

Ws ˆ Dery eR 31†

Eq. (31) leads to the regularized equivalent plastic strain at rupture depending on the mesh size as:

Ws
eR ˆ D ery 32†
This value can be used in the ®nite element calculation.
The characteristic mesh-size De, can computed as follows:

1. In the case of 1D problems, De is equal to the length of the ®nite element.

2. In the case of 2D problems, the value of De can be taken equal to the square root of the area of the ®nite
element.
3. In the case of 3D problems, the value of De can be taken equal to the cube root of the volume of the ®nite
element.

To obtain the regularized strain value e D, we suppose that the ratio between the experimental value e R and
the regularized one (Eq. (32)) is equal to the ratio between the experimental value of the strain e D and the
ˆ
regularized one: eRR†simexp eeDD††simexp 33†

e†

This equation leads to: D sim D exp eRR†simexp 34†

e† ˆ e† e†

Finally, to calculate the calibrated value D c, we assume that the term (D c=eR ÿ eD) of the damage law (1) is a
constant for both experimental and FEM evolution: Dc

e Rÿ eD ˆ a ˆ Cste: 35†

Therefore, the regularized value Dc can be obtained as:

Dc ˆ a e Rÿ eD† 36†

5. Numerical test

A ¯at plate subjected to uniaxial tensile loading was studied as a testing model for the numerical algorithm
which was developed. The FEM model of this includes just one rectangular 8-node plane strain element.
 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378 373

Clamped on one side, it is subjected to a prescribed displacement U on the other side. Geometry and boundary
conditions are depicted in Fig. 6. The materialÕs elastic parameters are taken to be Young

Fig. 6. Tensile test of a quadrangular ®nite element with eight nodes.

Fig. 7. Equivalent stress and damage evolution versus equivalent strain.

modulusel ÿ pEnˆwith the value of210,000 MPa and Poisson ratiorel ˆ 200 MPa, Km ˆˆ 0480:3. For the isotropic
hardening, the equation isMPa, n ˆ 0:406. eD:ˆr 0ˆ, r ‡ K e ^õtre damage law (1), the material parameters take
the following values: Accounting for Le Ma eR ˆ 1, Dc ˆ 0:35, DR ˆ 0:98.
Di€erent results are obtained depending on whether or not damage of the element is taken into account.
This is illustrated by the two curves in Fig. 7a. These curves correspond to the equivalent stress evolution versus
equivalent total strain. If the damage is taken into account, the failure of the structure can be predicted.
Otherwise, the structure would never fail.
As it is shown in Fig. 7b, when the equivalent strain exceeds the value e R, the damage increases rapidly from
Dc to DR value inducing a decreasing stress due to the reduced sti€ness of the material, as can be seen in Fig. 7a.
Taking these results into account, it can be concluded that the algorithm implemented accurately predicts the
material behavior and damage evolution, until the complete rupture of the structure, without computing
divergence.
In the following, the simulation of a blanking process for 1060 sheet-metal is described by employing the
numerical model outlined above.
374 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378

6. Numerical simulation of a blanking operation

The problem studied here consists of an axisymmetric blanking operation of a metal-sheet with 3 mm
thickness. The geometrical data are shown in Fig. 8.

Fig. 8. Axisymmetric model of blanking operation.

Fig. 9. Finite element meshing (1400 FE).

The mechanical characteristics of the material obtained by a tensile test are: E ˆ 210,000 MPa, m ˆ 0:29. The

corresponding strain hardening law takes the non-linear form: r ˆ r el ‡ Kenpl with the values of rel ˆ 250 MPa, K ˆ

1048 MPa, n ˆ 0:196. The rheological constants for the damage law associated with the Le Ma^õtre model were

found to be: eD ˆ 0, eR ˆ 0:8, Dc ˆ 0:37, DR ˆ 0:94.

6.1. Selection of the mesh

In this particular FEM simulation, the selection of the mesh is of great importance because it must be
suitable to describe the phenomenon of crack initiation and propagation. Therefore re®ning the mesh in the
clearance between the punch and the die is essential throughout the thickness of the sheet. In the regions
where the material is not subjected to high stress levels, a coarse mesh is sucient. The meshing of the model
was carried out by means of 1400 quadrangular four node axisymmetric elements. Fig. 9 shows how the mesh
was constructed.
To simulate the crack propagation in a realistic way, the mesh-size must have a physical characteristic length.
In the case of blanking processes, the ®nite elements sizes in the punch-die clearance must be in same order of
 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378 375

the material grains sizes. In this paper, the average mesh-size De is chosen equal to 0.05 mm. This value
correspond to average grain size of the material used for the blanking experiments.
The tool is modeled by adopting the rigid body hypothesis. The contact surfaces law, is de®ned by a Coulomb
friction model with a friction coecient of 0.15 at the sheet-punch and the die-sheet interfaces.

7. Results and discussion

Experiments using devices equipped with electrical gauges and a force transducer, were performed by a
4000 kN hydraulic press in order to verify the validity of the proposed FEM.
In a simpli®ed model, the blanking process was simulated without damage in¯uence. Using this in tandem
with the developed algorithm, it is possible to simulate punch penetration up to 100% of the sheetÕs thickness,
in spite of a large mesh distortion. When the material damage is taken into account, the same penetration may
be reached, but with a computing time about 1.7 times greater. This increase in computing time is the time
required to obtain numerical convergence. E€ectively, convergence will be reached in an increasingly dicult
manner as the damage reaches the critical value DR, at which damage the ®nite element will break.

Fig. 10. Crack propagation in the mesh (relative clearance ˆ 10%).

The blanked pro®le of the sheet obtained by experiment is presented in Fig. 10d, it is noted that the ®nal
rupture is obtained at a punch displacement of about 70% of the sheet thickness. The computation results
corresponding to di€erent displacement steps of the punch penetration and the corresponding graphical issues
376 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378

are presented in Fig. 10a±c. The crack propagation can be followed in the mesh until rupture occurs. Referring
to Fig. 10c and d shows that the predicted blanked pro®le is in good agreement with the experimental one.
As can be observed, the distortion of the mesh is restricted to a small area near the die-punch clearance.
The di€erent numerical computations, indicated that due to the punch penetration into the sheet, cracks
initiate at the cutting edges of the tools (Fig. 10a). Secondly, the cracks propagate in the same direction of the
punch penetration. This is illustrated by the two curves in Fig. 11. These curves correspond to the relative crack
propagation (expressed in percent of the sheet thickness) vs. the relative punch displacement obtained by the
experiment and the simulation.
The crack initiation and propagation can accurately be predicted without computational divergence. From
the moment of crack initiation to the complete rupture of the sheet, experimental and numerical results are
always in good agreement.
The two curves are suciently close to conclude that the procedure that we propose to simulate crack
initiation and propagation, gives good results.

Fig. 11. Crack propagation vs. punch penetration.

Fig. 12. Punching forces against punch displacement (clearance ˆ 10%).

As the optimal choice of the press and corresponding tools is always an industrial goal, we have computed
the punch force vs. the punch displacement as occurs throughout the blanking process. In Fig. 12, for the case
of an optimal relative clearance of 10% of the sheet thickness, three curves are drawn. These correspond to the
numerical predictions, and the experimental results.
It can be seen that the more realistic description of the process corresponds to the model that includes
damage.
 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378 377

7.1. E€ect of the geometry of cutting edge of the punch

The design of the tool is one of the main features in the industrial process. Therefore, it is necessary to
predict as accurately as possible, the geometry of the sheared edge of the blank for di€erent geometries of the
tool cutting edges. For this purpose, several calculations were performed. The in¯uence of the tool-wear on the
evolution of the sheared pro®le was accounted for by changing the values of the edge radii R p and Rd (Fig. 13).
Two di€erent states of punch wear, corresponding to a new punch with a cutting edge radius R p ˆ 0:01 mm and
a used punch with Rp ˆ 0:2 mm were chosen, the radius of the cutting edge of the die was chosen to be R d ˆ 0:01
mm.

Fig. 13. Geometry of the cutting edges of the tool.

Fig. 14. Predicted pro®les corresponding to 2 states of the punch wear (clearance ˆ 10%).

The two corresponding blanked pro®les are shown in Fig. 14a and b. It was veri®ed in Ref. [8] that the
numerical prediction was in very good agreement with the experimental results.
From a competitive industrial point of view, the best quality of mechanical parts must be obtained with
minimum costs. Therefore attention was also focused here on the geometrical quality of the blanked pro®le. As
it can be expected, in the case of a used punch, the pro®le of the part boundary presents a bad quality due to
the presence of a burr.
The numerical results, when compared with the experimental ones, show the reliability of the FEM in
predicting the results of punch wear in the blanking process.
378 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378

8. Conclusion

As has been shown, the simulation of sheet blanking by a press is a dicult problem due to the choice of
realistic hypotheses, especially in accounting for damage, and to the complexity of numerical calculations
involved in large displacements.
The crack initiation and propagation can accurately be predicted without computational divergence. From
the moment of the crack initiation to the complete rupture of the sheet-part, experimental and numerical
results are all in good agreement.
The computational results obtained by means of ABAQUS/standard routines which, account for the contact law
between tool and sheet, incremental computations strategy, give good results compared to experimental ones
for punch penetrations up to 70% of the sheet thickness, the distortion of the ®nite elements are still
satisfactory and calculations can be done without problems.

Acknowledgements

The author wishes to thank Professor Alain POTIRON for his assistance during this investigation.

References

[1] ABAQUS-HKS, Theory manual, Version 5.2-1.

[2] Cries®eld MA. Non linear ®nite element analysis of solids and structures, vol. 1. New York: Wiley; 1991.
[3] Clift SE, Hartley P, Sturgess CEN, Rowe GW. Fracture prediction in plastic deformation process. Int J Mech Sci 1990;32(1):1±17.
[4] Dos santos W, Organ AJ. Deformation in the ductile fracture processes examined by the visioplasticity method. Int J Mach Tool Des Res
1973;13:217±32.
[5] Grabacki JK. Constitutive equations for some damaged materials. Eur J Mech A/Solids 1994;13(1):51±71.
[6] Gurson AL. Continuum theory of ductile rupture by void nucleation and growth. J Engng Mater Tech Trans ASME 1977;99: 2±15.
[7] Ghosh E, Popat PB, Raghu V. A new approach to the mechanics of the blanking operation: theoretical model and experimental
veri®cation. J Mech Work Tech 1985;11:215±28.
[8] Hambli R. Etude experimentale, numerique et theorique du decoupage des toles en vue de l'optimisation du proc^ ede. These de
Doctorat ± ENSAM d'Angers, 1996 [in French].
[9] Homsi M, Wronski M, Roelandt JM. Modelisation numerique de la coupe. Strucome 1994;94:677±90.
[10] Hartley P, Pillinger I, Sturgess C. Numerical modeling of material deformation processes research, development and applications.
Berlin: Springer; 1992.
[11] Jana S, Ong NS. E€ect of punch clearance in the high-speed blanking of the thick metal using an accelerator for mechanical press. J
Mech Work Tech 1989;19:55±72.
[12] Lee CH, Kobayashi S. Elastoplastic analysis of plane-strain and axisymmetric ¯at punch indentation by the ®nite element method. Int J
Mech Sci 1970;12:349±70.
[13] Kasuga Y, Tsutsumi S, Mori T. Investigation into shearing process of ductile sheet metals. Mem Fac Engng NAYAGOYA Univ, Japan,
1979. p. 1±46.
[14] Le Ma^õtre J, Chaboche JL. Mecanique des materiaux solides. 2eme ed. Dunod, 1988.
[15] Marques JMMC. Stress computation in elastoplasticity. Engng Comput 1984;1:42±51.
[16] Murakami S. Mechanical modeling of material damage. Trans ASME 1998;55:280±6.
[17] Popat PB, Ghosh A, Kishore NN. Finite element analysis of the blanking process. J Mech Work Tech 1989;18:269±82.
[18] Simo JC, Taylor RL. A return mapping algorithm for plane stress elastoplasticity. Int J Num Meth Engng 1991;22:649±6710.
[19] Wong VG, Das MK. Analysis of stresses in bar cropping. 15th Int Mach Tool Res Conf, Birmingham: Macmillan; 1975. p. 617±24.
[20] Bensussan P. A critical review of the application of global and local approaches to creep crack initiation and growth. Proceedings of
MECAMAT, Dourdan, France, 1987. p. 1±17.
[21] Chaboche JL. Phenomenological aspects of continuum damage mechanics. First IUTAM Int Congr Theor Appl Mech Grenoble, 1988. p.
41±56.
[22] Chaboche JL, Saanouni K, Lesne PM. On the creep crack-growth prediction by non local damage formulation. Eur J Mech A/ Solids
1989;8(6):437±59.
 R. Hambli / Engineering Fracture Mechanics 68 (2001) 365±378 379

[23] Liu Y, Murakami S, Kanagawa Y. Mesh-dependence and stress singularity in ®nite element analysis of creep crack growth by continuum
damage mechanics approach. Eur J Mech A/Solids 1994;13(3):395±417.