DAMAGE MECHANISMS AND LOCAL APPROACH TO FRACTURE

Part I: Ductile fracture

C. BERDIN1, P. HAUŠILD1,2
1
Ecole Centrale Paris, LMSS-Mat, Grande Voie des Vignes,
92 295 Châtenay-Malabry, France
2
Czech Technical University, Faculty of Nucl. Sci. & Phys. Eng.,
Dept. of Materials, Trojanova 13, 120 00 Praha 2, Czech Republic

Abstract: This work aims at the application of local approach to ductile fracture. GTN
model is used in order to model damage for different materials. The problem of parameter
identification is pointed out. Mesh size dependency of the results is also tackled. GTN
model is found to be able to predict ductile fracture for different materials with different
damage mechanisms, provided that damage mechanisms are correctly determined: cast iron
with high initial void volume fraction, duplex stainless steel with void nucleation as a major
part of damage, low alloy steel, with a combination of small initial void volume fraction and
void nucleation.

1. Introduction

Most of the structural components are designed in order to remain in the elastic,
reversible part of their mechanical behaviour. For structures submitted to high cycle
fatigue, the design is performed in order to keep the structure in such a stress state that
the endurance limit is not reached. However, fail safe design needs to predict the
fracture of components. This is important for components such as automotive engine
supports, which act as breaking pieces in accidental case, but also when the damage
tolerance concept has to be applied. This paper concerns the prediction of ductile
fracture of materials used for such components.
For cracked structures, the global approach is currently applied. Linear or non-linear
fracture mechanics theory (Irwin, Rice) is used in order to predict the propagation of a
crack under the “in-service” loading. In this approach, a global parameter (G or J) which
is related to the local stress field, is compared to a critical value which depends only on
the material properties (fracture toughness). Material mechanical behaviour
(characterised by e.g., Young’s modulus, Poisson’s ratio, yield stress and plastic
hardening exponent) and the specimen geometry (with the crack) are sufficient to obtain
167
I. Dlouhý (ed.) Transferability of Fracture Mechanical Characteristics,, 167-180.
© Kluwer Academic Publisher. Printed in the Netherlands.
168

the global parameter which represents the damage level in the structure. Nevertheless,
various difficulties arise from this approach:
(a) the global parameter is only well-known for simple geometry,
(b) the critical value has to be measured with specific experiment,
(c) the critical value is dependent on specimen geometry,
(d) the complex mechanical behaviour of the material (kinematic hardening, loading
rate dependence) is not really taken into account.
For these reasons, the transferability of the global approach from laboratory specimens
to complex structural components is not ensured.
An alternative approach is the local approach to fracture [1,2] which has been more
recently developed. It is based on the following steps:
a) identification of the physical damage mechanisms and damage variables,
b) establishment of the relations between the macroscopic local stress-strain fields and
the damage variables,
c) computation of the relevant field variables.
The last step can be achieved using the widespread finite element method. The local
approach needs a very good knowledge of the local mechanical behaviour and the
physical damage mechanisms of the material. However, it leads to a geometry
independent approach, and allows damage assessment for any structural component.
This paper is dedicated to various applications of the local approach to ductile
fracture. Comparisons are made with experimental results for different materials
presenting very different damage mechanisms. The application of local approach to
brittle fracture is proposed in the second part of this paper.
Ductile damage is classically considered to initiate by void nucleation followed by
void growth and void coalescence leading to a macroscopic crack. Hence, the damage
variable can be simply represented by the void volume fraction. The first model for void
growth was proposed by Mac Clintock [3] who computed the exact solution for
cylindrical void growth into an infinite rigid plastic matrix under generalised plane
strain and axisymmetric loading. An approximate solution was found for spherical void
growth into an infinite rigid plastic matrix by Rice and Tracey [4]. It was improved by
Budiansky et al. [5] and later by Huang [6]. These models pointed out that stress
triaxiality and (plastic) strain are the relevant mechanical variables controlling damage.
Ductile fracture is then assumed to occur at a critical void volume fraction, which is
supposed to be only material dependent. The applications of this model to experimental
results [7,8] showed a correct prediction of the ductile fracture strain evolution versus
stress triaxiality.
However, these models are not relevant neither for intermediate and high void
volume fraction nor to model crack propagation. In these cases the damage controlling
mechanical variables are dependent on the damage level; this needs the so-called
“coupled local approach”, in which damage variables directly appear in the constitutive
equations modelling the mechanical behaviour of the material. The most famous model
for ductile fracture modelling is the Gurson’s model [9]. Following the same procedure
as Rice and Tracey did, this model is based on a micro-mechanical approach for porous
material. It was subsequently modified in a more or less heuristic manner by Tvergaard
and Needleman [10,11] in order to correctly predict void growth and fracture for
hardening material.
 169

This constitutive model requires to identify quite a few parameters: this is one of the
main difficulties for the applications of the model [12]. Different ways can be followed
for this identification: fully phenomenological method, i.e. determination of the damage
parameters from macroscopic mechanical response [13], unit cell numerical
computation [14], or direct damage kinetics assessment [15]. Any combination of these
three methods can be used.
In the following, we propose to review different applications of this model, in order
to enlighten the relation between damage mechanisms and model parameter’s
identification. The size of the characteristic volume for the physical damage
mechanisms is an additional parameter for any local approach to fracture [16,17]. This
parameter becomes very important when large mechanical gradients are present in the
structure to be designed as in the case of pre-cracked component. In a finite element
analysis, it is directly related to the element size ahead of the crack tip. The choice of
the mesh size at the crack tip will be discussed.
After a short review of the GTN model, three different cases are presented:
prediction of fracture toughness of cast iron, strength prediction of pre-cracked bend
specimens made of cast duplex stainless steel, and finally prediction of the ductile
fracture preceding cleavage in the ductile to brittle transition temperature (DBTT) range
for a low alloy steel.

2. Gurson Tvergaard Needleman model (GTN model)

The yield function of the GTN model is [10,11] :

Σ eq 2  3Σ m  
φ= + 2q1 f * cosh q2  − 1 + q12 f * 2  (1)
σ y2  2σ y   
 

q1 and q2 are parameters, Σeq the macroscopic von Mises equivalent stress, Σm the
macroscopic hydrostatic stress, σy the flow stress of the matrix, which depends at least
on the equivalent plastic strain of the matrix, εy, and f* is the parameter related to the
void volume fraction f, by :

f = f0 before any loading (2.a)

f* = f f ≤ fc (2.b)
f = f + δ ( f − fc )
*
f ≥ fc (2.c)
1/q − f
δ = 1 c (2.d)
f f − fc

f0, fc, ff, are respectively the initial void volume fraction, the critical void volume
fraction at void coalescence and the void volume fraction at fracture. Fracture occurs
when f=ff or f*=1/q1. The stress carrying capacity of the material is then lost.
170

Both nucleation of new voids and growth of existing voids contribute to the increase
of the void volume fraction, so its evolution can be expressed as :

f = fgrowth + fnucleation (3)

.
 
fgrowth = (1 − f ) tr  E p  (4)
 
 
fnucleation =An(εy) ε y (5)

E is the macroscopic plastic strain rate, and An(εy) is a function which depends on the
p

mode of void nucleation.

The parameters q1 and q2 are related to the hardening of the matrix and they are
difficult to identify because of their strong interaction with the stress-strain behaviour of
the matrix which is generally unknown. In this study the classical values (q1=1.5 and
q2=1)[11] where chosen.
The Abaqus/standard code [18] was used for numerical modelling. The GTN model,
not readily available in the software package, was implemented via a ‘user subroutine’
provided by GKSS [19]. The stress carrying capacity loss of the material at fracture was
modelled by a drastic drop of the elastic rigidity of the finite element.

3. Applications

3.1. FRACTURE TOUGHNESS OF NODULAR CAST IRON

In nodular cast iron, since specimen thickness exerts a large influence on toughness
[20,21], global approach is difficult to apply. So, deformation and damage mechanisms
were investigated, and the local approach was applied in order to predict fracture
toughness.
In this material, damage occurs mainly by debonding of the interface at the pole cap
of the nodules [22]. Debonding propagates along the interface between matrix and
graphite nodules thus inducing void nucleation (Figure 1). This step is followed by void
growth and void coalescence. In fact, depending on the material, damage kinetics can be
different [23]. Two cast irons were studied, GS52 with 10% of graphite nodules and
GGG40 cast iron with 7.7% of graphite nodules. The chemical composition of ferritic
matrix in the both materials differed mainly by the Si and Ni contents, which play a
major role in plastic hardening.
The damage evolution during tension was observed thanks to in-situ tensile tests i.e.
tensile tests performed into a scanning electron microscope [22]. For the GGG40 cast
iron, most of the graphite nodules were debonded from the matrix in the very early
stages of macroscopic yielding. For the GS52 one the decohesion of the graphite/matrix
interface seemed to start later, at about 2% of macroscopic strain with a more
heterogeneous damage spatial distribution.
 171

As the presence of voids modifies the stiffness of the material, the evolution of the
Young's modulus was measured as a function of plastic strain (Fig. 2). This has been
achieved by tensile loading followed by unloading after various strain increments
according to the procedure described in reference [22].

0,95 GGG40
GS52
0,9

Young's modulus ratio

0,85
0,8
0,75
0,7
0,65

0,6
0 2 4 6 8 10 12
Plastic strain (*100)
Figure 1. Damage in nodular cast iron: Figure 2. Evolution of the Young’s modulus ratio versus
debonding of graphite nodule during tensile plastic strain during a tensile test.
test (the direction of tension is horizontal).

Damage kinetics can consequently be precisely characterised by Young’s modulus

measurements associated with in situ tensile test observations. In GGG40 from the
results presented in Figure 2, one can consider that all the graphite nodules are
debonded when the macroscopic yield strength of the material is reached; so the initial
void volume fraction can be the graphite nodule volume fraction, and only void growth
can be considered for damage mechanism.
Once damage kinetic is established, it is then possible to identify the hardening law
of the matrix σy(εy) using an inverse method: the parameters of a Ludwik’s law were
adjusted in order to give a good fit of the hardening of the damaged material, i.e. the
mechanical behaviour in tension.
The void volume fraction at coalescence, fc, was identified in order to predict the
fracture diameter of notched specimen (fc=0.12). Fracture elongation measured by
tensile test cannot be used to determine the fracture parameters [24] because it is
strongly dependent on the viscoplasticity of the material, so that the strain rate
dependence of the mechanical behaviour has to be precisely established. The diameter
at fracture for notched specimen appears to be less sensitive to strain rate. The
measurement of the area ratio occupied by dimples over total fracture surface observed
with a scanning electron microscope allowed the determination of void volume fraction
at fracture (ff = 0.2 or δ=6.8)
The mesh size at the crack tip of the CT specimen used for fracture toughness
assessment was chosen in order to account for the experimental crack opening
displacement at crack initiation measured by the multiple specimen method [25]. It was
found to be 100*100 (microns)2. This is about 3 times the mean nearest neighbour
distance between graphite nodules, whereas Steglich et al. [26] found about 6 times of
172

this distance for very similar materials. In fact, it should be noted that in this work, the
distance between integration points of finite elements is about the mean nearest
neighbour distance between graphite nodules: the volume associated to the integration
point is therefore corresponding to the elementary representative volume with respect to
the damage process.
For the GS52 cast iron, direct (in-situ tensile test) and non-direct (Young’s modulus
measurements) damage observations showed a more progressive damage kinetics than
in GGG40 (see for example Fig. 2): a void nucleation process with constant void
nucleation rate can be assumed: An(εy)=An. the initial void volume fraction was obtained
from Young’s modulus measurements via a relation between Young’s modulus ratio
and void volume fraction [27] (f0=0.04). The void nucleation rate and the matrix
hardening law were numerically adjusted in order to correctly predict the tensile and CT
tests. Since the mean nearest neighbour distance was not so different between both cast
iron, the same values of fc, ff and mesh size at the crack tip as for GGG40 cast iron were
used.
Prediction of PQ [28] is correct as it can be seen in Table 1; for 2D computation,
experimental value lies between the results from the two extreme assumptions, that is
plane stress assumption for the lower bound, and plane strain assumption for the upper
bound [23].
The difference in the fracture toughness of these two cast irons seems to be due to
the existence of different damaged zone sizes, related to different damage kinetics : the
spatial distribution of the graphite nodules is pointed out as the main origin of the
difference in damage kinetics for the two materials as it was modelled. In GS52, the
spatial distribution of voids is less homogeneous than in GGG40. This is emphasised by
the higher graphite content in GS52. Aggregates of graphite nodules lead to a more
localised damage. In the mechanical modelling, this is taken into account in a
phenomenological way, thanks to the void nucleation law.

TABLE I. Comparison between experimental values of PQ [28]

and computed values

PQ (kN) GGG40 GS52

Experiment 12 17.5
Computation 11.3-13* 17.8**
* 2D-plane stress and 2D-plane strain computation
** 3-D computation

3.2. FRACTURE TOUGHNESS OF DUPLEX STAINLESS STEEL

CF8M duplex stainless steels are widely used for temperature applications because of
their good mechanical properties and corrosion resistance. The fracture toughness of
this material was investigated in order to compare the influence of shrinkage cavities
aggregate and an equivalent crack on the mechanical behaviour of a 3-point bend
specimen [29].
 173

The microstructure of CF8M is composed of interconnected ferritic lathes and

austenitic islands (Fig. 3). The material under study contained 25% volume fraction of
ferrite. During ageing below 400°C, spinodal decomposition of ferrite occurs and
increases ferrite brittleness. Fracture of the thermally aged material under monotonic
loading is induced by crack nucleation in embrittled ferrite (Fig. 4), followed by ductile
tearing of the austenite ligaments between micro-cracks (e.g. [30]).

Figure 3. Microstructure of cast duplex stainless steel Figure 4. Damage mechanisms: cracking into a ferrite
(dark grey: ferrite – white grey: austenite) lath under the main fracture surface

The physical mechanisms of deformation and damage being correctly identified, a

local approach can be applied. Recently Besson et al. [30] following [15] have proposed
a model of duplex stainless steel mechanical behaviour based on the local approach to
ductile fracture, which was used in [29]: void nucleation (i.e. cleavage microcracking)
was considered to be driven by plastic strain, as suggested by experimental observations
[15] with a constant rate (as for cast iron GS52). No initial void volume fraction was
considered.
For that material, matrix flow stress was first determined thanks to compression
loading experiments assuming that no damage occurs during this test. The damage
parameters, specially An, were then adjusted in order to predict tensile test results. No
void coalescence law (Eq. 2) was needed, so fc=ff=1/q1.
Modelling a pre-cracked three points bend specimen [29], a relationship was found
between the void nucleation rate and the mesh size in front of the crack (Fig. 5). For the
damage mechanisms under consideration, the elementary representative volume should
contain a sufficient number of ferrite lathes with favourable crystallographic
orientations for cleavage. Since the nucleation rate is an average coefficient over this
representative volume, there is a strong correlation between this parameter and the mesh
size. Therefore, delayed crack initiation due to the use of a coarse mesh can be
compensated by an increase of the nucleation rate taking into account the higher actual
plastic deformation at the crack tip: for a given nucleation rate, the predicted strength of
the pre-cracked bend specimen increases with the mesh size at the crack tip (Fig. 5.a),
whereas for a given mesh size, increasing the nucleation rate induces an earlier damage
and a lower predicted strength of the modelled specimen (Fig. 5.b).
174

For 3D modelling, the element size at the crack tip was imposed as 0.5x0.5x1.25
mm3 in order to reduce the computation time. A nucleation rate was identified in order
to account for the strength of the pre-cracked specimen. The optimised value is An=1,
compared to 0.6 which was the value determined in [30] for tensile test prediction; the
associated mesh size was 0.25x0.25mm2 in section, which is approximately two or three
times the mean distance between the ferrite laths. Since it has been identified for a given
mesh size, the nucleation rate becomes a fully phenomenological parameter and is no
more available for tensile tests prediction. However, this procedure can be interesting
for the modelling of large structural cracked component, even though the initial physical
basis of the void nucleation rate is partly lost.
The results of 3D numerical modelling of the pre-cracked specimen is shown in
Figure 6 and compared with experimental results (load drop corresponds to an
unloading imposed by the operator as soon as crack initiation occurred). This
comparison shows that the load versus deflection is correctly predicted.

(a) (b)

Figure 5. Influence of mesh size and void nucleation rate on the prediction of the mechanical behaviour of
precracked specimens (2D plane strains). (a) constant void nucleation rate: An=1; (b) constant mesh size
constant: d=500x500 µm2.

Figure 6. Comparison between experiment and 3-D modelling

for a precracked specimen (An=1, d=500x500 µm2 in section).
 175

3.3. DUCTILE FRACTURE PREDICTION IN THE DBTT RANGE

The examples presented above concerned only crack initiation and fracture toughness
prediction. The following work focuses on ductile crack propagation modelling in the
DBTT range in a low alloy bainitic steel (16MND5). In this temperature range, ductile
fracture precedes cleavage. The second part of this paper deals with the cleavage
fracture prediction. Ductile crack advance is shown to modify significantly the
mechanical field thus inducing a significant increase of the maximum principal stress
[31,32,33] in Charpy V-notch (CVN) specimen, as well as an increase of the loaded
volume susceptible to contain potential cleavage initiation sites. So, for cleavage
fracture prediction, ductile fracture has to be well represented.
Three types of inclusions playing a role in ductile fracture were distinguished in this
material:
a) large elongated Manganese sulfide (MnS) inclusions (10µm in diameter of short
section, 100µm in length) usually formed clusters,
b) smaller spheroïdal MnS inclusions and oxydes (1µm in diameter),
c) carbides (0.1µm)
According to the chemical composition (see part II) the volume fraction of carbides was
about 6.10-3, and the volume fraction of MnS together with part of oxydes was
estimated at 5.10-4 [34]. For the mechanical test considered, the elongated inclusions
were mainly loaded perpendicular to their length which induced early debonding. This
can be observed in the figure 7, which is a cross section of a CT specimen loaded at
0°C: large cavities coming from MnS initiating void growth are linked by smaller voids
nucleated on finer particles.

(a) (b)

Figure 7. Ductile damage in the low alloy bainitic steel. (a) Void growth on large MnS inclusions linked by
void nucleation on smaller particles. Cross section of a CT specimen loaded at 0°C (nital etching); (b)
nucleation and void growth from carbides near the notch root in a CVN specimen tested at –30°C.

Following schematically these observations the initial void volume fraction was
chosen as the volume fraction of MnS with part of oxydes. A gaussian void nucleation
176

law (Eq. 6) was considered [35], in order to account for the second population of
cavities. Three parameters are included in this function: the volume fraction of
nucleated void fn, which is related to the particle volume fraction inducing void, the
mean plastic strain εn of the matrix for which the nucleation rate is maximum and the
standard deviation sn, indicating the range of plastic strain inducing void nucleation; the
total volume fraction of nucleated voids, fn, was taken as 25% of the carbides content
[36]. For the other nucleation parameters, classical values were used : εn=0.3 and sn=0.1
[35].

 1  ε y − ε n  
2
( )
An ε y =
fn
exp −


2  s n  
 (6)
sn 2π
 

The mechanical behaviour of the matrix was modelled by a Cowper-Symonds law to

account for the high strain rates applied on Charpy impact specimens and identified on a
large experimental data base [31,32]. A Linear temperature dependence was also
included. So σy was defined by the following expression:

  ε 1/ p 

(
σ y ε y ,ε y ,T ) = σ (y1 ) (ε y ) 1 +  y
 D


 (α − βT )

(7)
   

where D, p, α, β were parameters, and σ (y1 ) the static hardening law. Parameters values
can be found in [31,32].
The parameters for void coalescence and fracture prediction (fc and δ) were
identified in a phenomenological way on smooth and notched tensile test results. They
were adjusted, in order to correctly predict the diameter at fracture of specimens tested
at 0°C (fc=0.04, δ=4 which is still a common value for obtaining a high void growth
acceleration with a good numerical stability). The parameters associated with ductile
damage were supposed to be independent on temperature and strain rate for the ranges
concerned here (10-4-1000 s-1 and –90°C-0°C). It should be noted that matrix hardening
was found to be independent on these variables, so that q1, q2 which depend on matrix
hardening can be considered as constant too.
A mesh size of 100*100µm2 in section provides a correct ductile crack growth rate
for CT specimens tested at 0°C (Fig. 8.a). In this study the mesh size identification for
ductile crack extension modelling in CT specimen is validated by the ductile crack
prediction for Charpy specimens (Fig. 8.b). Furthermore, the ductile crack growth is
correctly predicted all along the loading. Quantitative microstructure analysis [37]
revealed a 50µm mean nearest neighbour distance between large MnS inclusions, which
is about half of the mesh size.
Figure 9 shows the comparison between macroscopic fractography of a Charpy
specimen and iso-contour of the maximum principal stress which is the mechanical
variable associated with brittle damage (part. II). It can be seen that the ductile crack
 177

advance is very well predicted (region where maximum principal stress value is about
zero) with crack tunnelling propagation. However, shear lips are not predicted, which is
a classical result with that model, since no void growth can occur under pure shearing.
Consequently, lateral necking is slightly overestimated. The same figure shows that the
cleavage initiation point lies in the region of higher maximum stress value; this is in
agreement with the classical theory of brittle fracture (Part. II).

(a) (b)

Figure 8. Ductile crack advance versus loading (deflection for KCV specimen and COD, crack opening
displacement for CT specimen). Comparison between experiment and modelling. (a) side-grooved CT
specimen tested at 0°C (2-D modelling); (b) KCV specimen tested at –30°C (3-D modelling).

Figure 9. Comparison between fractography of a Charpy specimen tested

at -30°C, and iso-contour of the maximum principal stress.
178

4. Conclusions

Different materials with different damage kinetics were modelled with the GTN model:
a) cast iron with high initial void volume fraction,
b) duplex stainless steel with void nucleation as a major part of damage,
c) low alloy steel, with a combination of small initial void volume fraction and void
nucleation.
The identification of the damage parameters and hardening laws of matrix were
performed thanks to a combination of direct damage measurements (in situ tensile tests,
fractography analysis), non-direct damage measurements (Young’s modulus
measurements, comparison between compressive and tensile tests results), and in a
phenomenological way, comparing macroscopic mechanical test results and the model
prediction. The last procedure was essentially used for the determination of the void
volume fraction at void coalescence, fc, the void growth acceleration slope δ, as well as
for fitting the mesh size. This last parameter can be defined only thanks to cracked
specimen mechanical results. Mesh size is closely related to the elementary
representative volume associated with the damage mechanisms. It shall be larger at least
than this characteristic volume, what is effectively found in these studies.
However, the relation between mesh size and microstructure is not so clear, and such
a mesh size dependence of the modelling is a real problem for fracture prediction of
cracked components, since mesh size shall be chosen in order to resolve the mechanical
fields, but may be still dependent on the problem size and on a reasonable computation
time. These criterion can be in contradiction with a mesh size allowing a correct
prediction of ductile crack initiation and growth by the GTN model. So, the best way to
resolve that sort of conflict is certainly to use non-local approach of ductile fracture [38]
which includes a length scale directly into the constitutive equations of the model,
associated with adaptive meshing. However, waiting for these developments, a relation
can be found between damage parameters and mesh size as it was demonstrated in the
case of duplex stainless steel.
Finally, The GTN model allows good fracture prediction in very different cases if
the parameter’s identification is performed on a sufficiently large database. Damage
parameters were found to be strain rate and temperature independent (at least in the
ranges of identical plasticity mechanisms, i.e. outside the adiabatic shear and creep
domains). Nevertheless, this result should be confirmed on other materials.

9. References
1. Lemaitre, J. (1986) Local approach of fracture, Engineering Fracture Mechanics 25, 523-537.
2. Mudry, F. (1987) A local approach to cleavage fracture, Nuclear Engineering and Design 105, 65-76
3. Mc Clintock (1968) On the ductile enlargement of voids in triaxial stress fields, J. of Mechanics Physics
and Solids 17, 201-217.
4. Rice, J.R. and Tracey, D.M. (1969) On the ductile enlargement of voids in triaxial stress fields, J. of
Mechanics and Physics of Solids 17, 201-217.
5. Budiansky, B., Hutchinson, J.W. and Slutsky, S. (1982) Void growth and collapse in viscous solids, in
Mechanics of Solids, eds Hopkins H.G., Sewell M.J., 13-45.
6. Huang, Y. (1991) Accurate dilatation rates for spherical voids in triaxial stress fields, J. of Applied
 179

Mechanics 58, 1084-1086.

7. Hancock, J.W. and Mackenzie, A.C. (1976) On the mechanisms of ductile failure in high strength steels
subjected to multi-axial stress-states, J. of Mechanics and Physics of Solids 24, 147-169.
8. Marini, B., Mudry, F. and Pineau, A. (1985) Experimental study of cavity growth in ductile rupture,
Engineering Fracture Mechanics 22, n°6, 989-996.
9. Gurson, L.A. (1977) Continuum theory of ductile rupture by void nucleation and growth, J. of
Engineering Materials and Technology, Jan., 2-15.
10. Tvergaard, V. (1981) Influence of voids on shear band instabilities under plane strain conditions,
International J. of Fracture 17, 389-407.
11. Tvergaard, V. (1982) On localization in ductile material containing spherical voids, International J. of
Fracture 18, n°4, 337-252.
12. Zhang, Z.L. and Hauge, M. (1998) On the micro-mechanical parameters, ASTM-STP 1321.
13. He, R., Steglich, D., Heerens, J., Wang, G.W., Brocks, W. and Dahms, M. (1998) Influence of particle
size and volume fraction on damage and fracture in Al-Al3Ti composites and micromechanical modelling
using the GTN model, Fatigue and Fracture of Engineering Materials 21, 1189-1201.
14. Brocks, W., Sun, D.Z. and Hönig, A. (1996) Verification of micromechanical models for ductile fracture
by cell model calculations, Computational Material Science 7, 235-241.
15. Joly, P., Pineau, A. and Meyzaud, Y. (1993) Fracture micromechanisms of an aged duplex stainless steel;
application to the simulation of fracture of notched tensile and compact specimen, in Proceedings
Mecamat 93: International seminar on micromechanisms of materials, 210-221.
16. Lemaitre, J. and Dufailly, J. (1987) Damage measurements, Engineering Fracture Mechanics 28, n°5-6,
643-661.
17. Rousselier, G. (1987) Ductile fracture models and their potential in local approach to fracture, Nuclear
Engineering and Design 105, 97-111.
18. HKS ABAQUS/Standard, Version 5.8 (1998), Theory Manual, Hibbit, Karlsson and Sorensen Inc., USA.
19. Siegmund, T. and Brocks, W. (1997) A user-material subroutine incorporating the modified Gurson-
Tvergaard-Needleman model of porous metal plasticity into ABAQUS finite element program, Technical
report GKSS/WMG/97/2, GKSS, Geesthacht, Germany.
20. Bradley, W.L. and Srinivasan, M.N. (1990) Fracture and fracture toughness of ductile iron and cast steel,
International Materials Reviews 35, n°3, 129-161.
21. Kobayashi, T. and Yamada, S. (1994) Evaluation of static and dynamic fracture toughness in ductile cast
iron, Metallurgical and. and Materiala Transactions 25A, 2427-2437.
22. Dong, M.J., Prioul, C. and François, D. (1997) Damage effect on the fracture toughness of nodular cast
iron: Part I and Part.II, Metallurgica and Materiala Transactions 28A, 2245-2262.
23. Berdin, C., Dong, M.J. and Prioul, C. (2001) Local approach of damage and fracture toughness for
nodular cast iron, Engineering Fracture Mechanics 68, 1107-1117.
24. Bernauer, G. and Brocks, W. (2000) Micromechanical modelling of ductile damage and tearing – Results
of a european numerical round robin, Report GKSS 2000/15.
25. Dong , M.J., Berdin, C., Béranger, A.S. and Prioul, C. (1996) Damage effect in the fracture toughness of
nodular cast iron, J. de physique IV, 65-74.
26. Steglich, D. and Brocks, W. (1998) Micromechanical modelling of damage and fracture of ductile
materials, Fatigue and Fracture of Engineering Materials 21, 1175-1188.
27. Bompard, P. and François, D. (1984) Damaging effects of porosity on fracture of sintered Nickel
Proceedings of ICF6, New Dehli, 1279.
28. ASTM, Standard method for measurement of fracture toughness, E1820-99.
29. Haušild, P., Berdin, C., Bompard, P. and Verdière, N. (2001) Ductile fracture of duplex stainless steel
with casting defects, International J. of Pressure Vessels and Piping 78, 607-616.
30. Besson, J., Devillers-Guerville, L. and Pineau, A. (2000) Modeling of scatter and size effect in ductile
fracture: application to thermal embrittlement of duplex stainless steels, Engineering Fracture Mechanics
67, 169-190.
31. Rossoll, A. (1998) Détermination de la ténacité d’un acier faiblement allié à partir de l’essai Charpy
instrumenté, Ph.D thesis, Ecole Centrale Paris, France.
32. Rossoll, A., Berdin, C. and Prioul, C. (2002) Determination of the fracture toughness of a low alloy steel
by the instrumented Charpy impact test, International J of Fracture, in press.
33. Haušild, P., Nedbal, I., Berdin, C., and Prioul, C. (2002) The influence of ductile tearing on fracture
energy in the ductile-to-brittle transition temperature range, Material Science and Engineering A, in press.
34. Mäntylä, M., Rossoll, A., Nedbal, I., Prioul, C. and Marini, B. (1999) Fractographic observation of
180

cleavage fracture initiation in a bainitic A508 steel, Journal of Nuclear Materials 264, 257-262.
35. Chu, C.C. and Needleman, A. (1980) Void nucleation effects in biaxially stretched sheets, Journal of
Engineering Materials and Technology 102, 249-256.
36. Fisher, J.R. and Gurland, J. (1981) Void nucleation in spheroidized carbon steels- part1. Experiment,
Metals Science 15, 185-192.
37. Haušild, P. (2001) Transition ductile-fragile d’un acier faiblement allié, Internal Report, Ecole Centrale
Paris, L.MSS-Mat, France.
38. Tvergaard, V. and Needleman, A. (1995) Effects of nonlocal damage in porous plastic solids,
International J. of Solids and Structures 32, n°8-9, 1063-1077.