Saintier
Saintier
Saintier
Mechanics of Materials
journal homepage: www.elsevier.com/locate/mechmat
a r t i c l e i n f o a b s t r a c t
Article history: An analysis of high cycle multiaxial fatigue behaviour is conducted through the numerical
Received 2 March 2012 simulation of polycrystalline aggregates using the finite element method. The metallic
Received in revised form 26 July 2012 material chosen for investigation is pure copper, which has a Face Centred Cubic (FCC)
Available online 11 September 2012
crystalline microstructure. The elementary volumes are modelled in 2D using an hypoth-
esis of generalised plane strain and consist of 300 equi-probability, randomly oriented
Keywords: grains with equiaxed geometry. The aggregates are loaded at levels equivalent to the aver-
Multiaxial high cycle fatigue
age macroscopic fatigue strength at 107 cycles. The goal is to compute the mechanical
Finite element simulation
Polycrystalline aggregates
quantities at the mesoscopic scale (i.e., average within the grain) after stabilization of
Cubic elasticity the local cyclic behaviour. The results show that the mesoscopic mechanical variables
Crystal plasticity are characterised by high dispersion. A statistical analysis of the response of the aggregates
is undertaken for different loading modes: fully reversed tensile loads, torsion and com-
bined in-phase tension–torsion. Via the calculation of the local mechanical quantities for
a sufficiently large number of different microstructures, a critical analysis of certain mul-
tiaxial endurance criteria (Crossland, Dang Van and Matake) is conducted. In terms of
material behaviour models, it is shown that elastic anisotropy strongly affects the scatter
of the mechanical parameters used in the different criteria and that its role is predominant
compared to that of crystal plasticity. The analysis of multiaxial endurance criteria at both
the macroscopic and mesoscopic scales clearly show that the critical plane type criteria
(Dang Van and Matake) give an adequate estimation of the shear stress but badly reflect
the scatter of the normal stress or the hydrostatic stress.
Ó 2012 Elsevier Ltd. All rights reserved.
0167-6636/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.mechmat.2012.08.006
C. Robert et al. / Mechanics of Materials 55 (2012) 112–129 113
Nomenclature
quantities to reflect certain experimental observations, the terised by a Gumble type distribution, and converge with
authors establish a correlation between the distributions of this distribution, even for a small number of SVE (i.e., 25).
the initiation parameters obtained for alternating tensile Parallel to the development of these approaches, recent
loads and distribution of crack lengths observed on the work by Altus (2006), Morel and Huyen (2008) and Ponc-
surface of a ferrito-martensitic steel subjected to the same elet et al. (2010) have made it possible to take into account
loading type. The results are promising for the Fatemi and the role of microstructural heterogeneities in terms of mul-
Socie type parameter and for the parameter based on the tiaxial fatigue modelling. These models seek to account for
maximum slip amplitude. the effect of the variability of the mechanical behaviour at
Guilhem et al. (2010) continued the work of Bennett the scale of the microstructure, as simply as possible. For
and McDowell (2003) and enriched the crack initiation metallic materials, the fatigue crack initiation mechanisms
analysis by taking into account the effects of a free surface in the HCF regime result in isolated microcracks with no
and the anisotropic elastic material behaviour. The distri- interaction between neighbouring initiated cracks (Flace-
butions of the various parameters were studied, distin- liere et al., 2007). Moreover, when an isolated microcrack
guishing the position of the grains (i.e., in the thickness is not stopped by a microstructural barrier, it may lead to
or on the surface), their direction and the orientation of the final failure. To reflect this behaviour from a probabilis-
neighbouring grains. The authors find that the Dang Van tic point of view, the weakest link hypothesis (Weibull,
equivalent stress and the Fatemi and Socie equivalent slip 1939, 1951) is often used when it comes to deal with high
amplitude do not highlight the well know critical character cycle fatigue. Furthermore, Monchiet et al. (2006, 2008)
of the grains at the free surface when considering high cy- propose to extend the framework of crystal plasticity by
cle fatigue behaviour. computing damage at the grain scale. The proposed crite-
Bertolino et al. (2007) use the concept proposed by rion seems to be promising, especially for high hydrostatic
Dang Van, whereby elastic shakedown of the grains can pressure.
be seen as a limit for crack initiation. They therefore sought This study is undertaken in the context of understand-
to determine, using numerical models, the critical load ing the role of microstructural heterogeneities on the mul-
above which at least one grain does not elastically shake- tiaxial high cycle fatigue response via the numerical
down. This seems to give good predictions for alternating simulation of polycrystalline aggregates. In this work, the
uniaxial tensile loads but not for shear loads. Indeed, in this material investigated is pure copper.
case, the predicted fatigue limit is under estimated. The principal components of this work are:
Przybyla and McDowell (2010) have studied the inter-
action effects between different microstructures for the ex- To investigate the role of the material behaviour (i.e.,
treme values of the Fatemi and Socie criteria. Several stress cubic elasticity and crystal plasticity) and the grain ori-
levels are studied for uniaxial loads. The authors analysed entation on the mechanical response under cyclic loads
these distributions for 25–100 Statistical Volume Elements at levels close to the average macroscopic fatigue limit
(SVE). They show that these distributions are well charac- at 107 cycles.
To undertake an analysis of the distribution of the mes- The crystal plasticity material constitutive law used in
oscopic mechanical parameters (i.e., the shear stress on this work was introduced by Meric and Cailletaud (1991).
a slip plane, the resolved shear stress in a slip system, It is commonly used for the numerical simulation of poly-
the hydrostatic stress and the stress normal to a slip crystalline aggregates (Musienko et al., 2007; Barbe and
plane) for different aggregates and for different macro- Quey, 2011; Aubert et al., 2012).
scopic, fully reversed cyclic loads (i.e., tension, shear The constitutive relations are defined by Eqs. (1)–(6),
and combined in phase tension–shear). where cs represents the plastic slip within system s and
To investigate the ability of the Crossland, Dang Van ms is the accumulated plastic slip in system s. The isotropic
and Matake multiaxial fatigue criteria to reflect certain hardening is defined in each slip system by the variable rs ,
trends observed in HCF under multiaxial loading condi- and the kinematic hardening by the variable xs . c and d are
tions through the results of the numerical simulations. parameters of the constitutive law.
Particular attention is given to the difference between The orientation tensor ms , which is computed for each
the macroscopic predictions and the mesoscopic results slip system s by the tensorial product of the normal to
of the aggregates for each criterion. the slip plane ns and the direction of slip ls , is used to com-
pute the resolved shear stress ss and the plastic strain rate
2. Numerical procedure tensor, e_ pl . The two material parameters, K and n, represent
the sensitivity to strain rate. Finally, the parameters r 0 , Q
This section discusses the different aspects of the mod- and b (and also the components hrs of the interaction ma-
elling approach. After describing the different material trix) are used to model the isotropic hardening.
behaviour models used, the finite element models are The material parameters used for the crystal plasticity,
presented. identified for a high-purity copper by Gérard et al.
(2009), are presented in Table 1.
2.1. Material behaviour models s n
ks xs k r s
c_ s ¼ signðss Þ ¼ m_ s signðss Þ ð1Þ
In this work, four types of material behaviour, all asso- K
ciated with copper, are studied:
xs ¼ cas with a_ s ¼ c_ s dm_ s as ð2Þ
isotropic elasticity; X rs
cubic elasticity; rs ¼ r0 þ Q h ð1 exp ðbmr ÞÞ ð3Þ
r
isotropic elasticity + crystal plasticity;
cubic elasticity + crystal plasticity.
ms ¼ ðns ls þ ls ns Þ=2 ð4Þ
For the case of isotropic elasticity, the elastic properties of
the material are identical in all directions and are charac-
s s ¼ ms : r ð5Þ
terised by Young’s modulus and Poisson’s ratio. Further- X
more, all grains are assumed to have the same properties. e_ pl ¼ c_ s ms ð6Þ
s
The cubic elasticity material constitutive law is character-
ised by 3 independent elastic constants. The elasticity coef-
ficients are the same for all grains (in their respective 2.2. Loading conditions
coordinate systems), and each grain has a different orienta-
tion, defined by three Euler angles. Crystal plasticity is con- The polycrystalline aggregates considered in this work
sidered in the classical manner, in which plastic slip occurs are subjected to load levels corresponding to the average
in a slip system defined by a plane and a direction in the macroscopic fatigue limit at 107 cycles, taken from the re-
planes of maximum atomic density. The crystal plasticity sults of Lukás and Kunz (1989). The fully reversed tensile
law includes combined non-linear isotropic and kinematic and torsional fatigue limit used are s1 ¼ 56 MPa and
hardening. Each grain has 12 slip systems (i.e., an FCC crys- t1 ¼ 36 MPa respectively. From these data the Crossland
tal structure) and all of the grains have a different orienta- criterion, which is defined below, is arbitrarily used to de-
tion. This plasticity model is described in details in the fine the equivalent multiaxial loading conditions in ten-
following section. sion, shear and combined in-phase tension–shear for all
One of the objectives of this work is to better under- analyses regardless of the criterion being investigated.
stand the effect of the choice of the material constitutive For all load conditions investigated, the load ratio used
law on the local distributions of the aforementioned is Rr ¼ rmin =rmax ¼ 1 and in the case of combined ten-
mechanical quantities. For this, the effect of cubic elasticity sion–shear, the biaxiality ratio is ra =sa ¼ 1. These fatigue
and crystal plasticity on the response at the mesoscopic limits are presented in Table 2.
scale (i.e., the scale of individual grain) are discussed.
For the case of isotropic elasticity, Young’s modulus is 2.3. The numerical model
50 GPa and Poisson’s ratio of 0.3 (C 11 ¼ 67:3 MPa,
C 12 ¼ 28:8 MPa and C 44 ¼ 38:5 MPa in Voigt notation). The generation of synthetic polycrystalline microstruc-
For cubic elasticity, the following constants are used: tures has been the subject of numerous studies (Groeber
C 11 ¼ 159 MPa, C 12 ¼ 121:9 MPa and C 44 ¼ 80:9 MPa (Mer- et al., 2008; St-Pierre et al., 2008; Bhandari et al., 2007;
ic and Cailletaud, 1991). Fritzen et al., 2009). Among the most widely used methods,
C. Robert et al. / Mechanics of Materials 55 (2012) 112–129 115
Table 1
Material parameters used in this work, taken from Gérard et al. (2009).
1
K (MPa sn ) n (–) r 0 (MPa) Q (MPa) b (–) c (MPa) d (–) h0 (–) h1 (–) h2 (–) h3 (–) h4 (–) h5 (–)
8 20 15 4 12 32,000 900 1 1 0.2 90 3 2.5
mesh is generated using Gmsh (Geuzaine and Remacle, cients and can be identified from the fully reversed tensile
2009). All of these steps are shown in Fig. 2. and torsional fatigue limits (s1 and t 1 ).
For all of the loading conditions studied here, three The amplitude of the octahedral shear stress is defined
microstructural topologies containing 300 equiaxed grains by:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
were used. In addition, for each topology three sets of ran-
soct;a ¼ J 2;amp ðrÞ ð8Þ
dom grain orientations are studied. This results in a total of
Fig. 2. The principal steps used to generate the mesh of the finite element models of the polycrystalline aggregates.
116 C. Robert et al. / Mechanics of Materials 55 (2012) 112–129
Fig. 3. (a)–(c) Pole figures (100) showing the three different grain orientations and (d)–(f) the three different topologies investigated.
Fig. 4. The finite element mesh of a periodic aggregate with 300 grains and approximately 120,000 elements.
where J 2;amp ðrÞ is determined from the smallest hyper- 3.2. Dang Van criterion
sphere circumscribing the loading path in the deviatoric
space associated with the tensor r. More precisely The Dang Van multiaxial criterion (Dang Van, 1973) is
J 2;amp ðrÞ is the radius of this hypersphere when the norm valid in the high cycle fatigue domain and is a multi-scale
J 2 is used. J 2 is the second invariant of deviatoric stress: approach based on the concept of elastic shakedown.
The Lin–Taylor hypothesis (Taylor, 1938; Lin, 1957) is
1
J2 ¼ Sij Sij ð9Þ used for the scale transition from mesoscopic to macro-
2 scopic quantities. The criterion can be summarised by the
and S is the deviatoric stress tensor defined by: following statement: ‘‘Microscopic fatigue crack initiation
does not occur if the material achieves an elastic shake-
1 down state at both the macroscopic and mesoscopic
S ¼ r tr ðrÞI ð10Þ
3 scales’’. Mathematically the criterion is represented by
I is the second order unit tensor. Eq. (11)
C. Robert et al. / Mechanics of Materials 55 (2012) 112–129 117
Fig. 5. Evolution of the accumulated equivalent plastic strain at the end of each cycle for the tensile loading condition using the cubic elasticity + crystal
plasticity material constitutive law.
3.3. Matake criterion
max max b
s ðn; tÞ þ adv rb hyd ðtÞ 6 bdv ð11Þ
t n
Fig. 6. Evolution of the accumulated equivalent plastic strain at the end of each cycle for the fully reversed shear loading condition using the cubic
elasticity + crystal plasticity material constitutive law.
the fully reversed tensile and torsional fatigue limits (s1 point of the FE mesh. It can be seen that the most plastified
and t 1 ). grains, in terms of accumulated equivalent plastic strain,
during the first cycles remains the most plastified through-
4. Results out the following cycles. Moreover, a strong variation of
this quantity (or large heterogeneity) is observed between
This section presents the cyclic response for the differ- the different grains. Regarding the influence of the loading
ent material behaviour models and the three loading type, the intensity of the accumulated equivalent plastic
modes. Firstly, Figs. 5–7 highlight the spatial evolution of strain after cycling is greater for shear loading (Fig. 6) com-
the accumulated equivalent plastic strain (ccum ), during pared to the tensile load case (Fig. 5). Also, for the shear
the first 10 cycles, for the cubic elasticity + crystal plastic- load case there are more grains for which this quantity is
ity material constitutive law, and for the three loading high. As for the combined tension–shear load case (Fig. 7)
modes studied. Note that these values are not values aver- the intensity of the maximum accumulated plastic defor-
aged in the grains but are the values at each integration mation is in between the tensile and shear cases. In shear
C. Robert et al. / Mechanics of Materials 55 (2012) 112–129 119
Fig. 7. Evolution of the accumulated equivalent plastic strain at the end of each cycle for the tension/shear loading condition using the cubic
elasticity + crystal plasticity material constitutive law.
17% of grains have ccum > 0:5ccummax compared to the tensile The macroscopic scale represents the average behaviour
loading case for which 10% of grains have ccum > 0:5ccum max of the complete aggregate. The mechanical quantities are
and for the tension/shear loading case 15% of grains have computed from the stress tensor, associated with the im-
ccum > 0:5ccum
max . This is also experimentally observed (Ag- posed load. The macroscopic stress is referred to as R.
bessi et al., 2011). The maximum mesoscopic (or grain The mesoscopic scale represents the behaviour in a
averaged value) of ccum is 0.032 for the shear loading case, grain. The mechanical quantities are the volumetric aver-
0.021 for the tensile loading case and 0.026 for tension/ age in the grain of each component of the stress tensor.
shear. Also in tension/shear the grains for which the plas- Elements having at least one node on the grain boundary
ticity is well developed are identical to the shear load case. are excluded from the averaging process. Hence, stress
concentrations at grain boundaries are not taken into
4.1. Mechanical response of the polycrystalline aggregates account.
Note that, all of the mechanical quantities reported at
The results are presented for both the mesoscopic and the mesoscopic scale are averaged quantities within the
the macroscopic scale. grains so that the linear interpolation of three nodes ele-
120 C. Robert et al. / Mechanics of Materials 55 (2012) 112–129
loop is obtained for each loading mode. The CPU time for
each simulation is approximately 60 h on an Intel Xeon
X5677 processor at 3.47 GHz with 64 Gb of RAM.
Fig. 10. The prediction from the Crossland criterion at the macroscopic scale (in black) and mesoscopic scale (in gray).
Fig. 12. The predictions from the Dang Van criterion at the macroscopic scale (in black) and the mesoscopic scale (in gray).
two scales, is very significant for the simulations using the be seen that for the isotropic elasticity constitutive law,
cubic elasticity material constitutive law. Also, the maxi- without crystal plasticity, the maximum values at the mes-
mum shear stress on all slip planes at the mesoscopic scale oscopic scale for each time step are equal to those at the
is greater than the maximum value of macroscopic shear macroscopic scale. For the case of cubic elasticity, the cloud
stress for this type of behaviour. Regarding the influence of points representing the mesoscopic scale is very dis-
of the crystal plasticity, its role seems limited to a slight in- persed but globally has the same shape as the macroscopic
crease in the dispersion of the averaged mesoscopic hydro- load path. The crystal plasticity has less impact than cubic
static stress for the simulations using isotropic elasticity elasticity on the overall shape of the mesoscopic loading
and a reduction in the maximum shear stress for the case path.
of cubic elasticity (which however remains greater than Note that the black line illustrates the load path on the
that computed at the macroscopic scale). In addition, the Dang Van critical plane, while the gray points correspond
maximum observed shear stress in the calculations with to all the slip planes of the aggregates.
cubic elasticity is higher than that for the case of isotropic
elasticity. 4.3.3. Critical plane analysis
Fatigue crack initiation is highly dependent on the ori-
4.3.2. The effect of the material constitutive law on the loading entation of slip planes. In a 3D coordinate system the ori-
path entation of the normal to the slip planes can be defined
Fig. 13 represents the temporal evolution of the two using a spherical coordinate system. The normal can be de-
variables used in the Dang Van criterion, at the mesoscopic fined by two angles: the angle h between the z axis and the
scale (in gray) and the macroscopic scale (in black). It can xy plane of the macroscopic coordinate system and the an-
C. Robert et al. / Mechanics of Materials 55 (2012) 112–129 123
Fig. 13. Temporal evolution of the two variables used in the Dang Van criterion at the macroscopic scale (in black) and the mesoscopic scale (in gray).
Fig. 15, in terms of the two angles ðh; UÞ, for all simulated
slip planes of the polycrystalline aggregates. (Values lower In terms of orientation, very little variation of the distri-
than 0.6 are not plotted.) bution of the coefficient of danger is seen for the isotropic
124 C. Robert et al. / Mechanics of Materials 55 (2012) 112–129
Fig. 15. Normalised coefficient of danger for the Dang Van criterion as a function of the orientation of the normal to the slip plan (values less than 0.6 are
not plotted).
elasticity constitutive law, with and without crystal plas- For the simulations using the isotropic elastic material
ticity. For the case of cubic elasticity, some orientations constitutive law (see Fig. 16(a)), good correspondence is
have a coefficient of danger that is much higher than oth- observed between the point representing the macroscopic
ers. Also, the critical plane orientation scatter is much prediction and the upper part of the envelope of the mes-
higher compared to the case of isotropic elasticity. oscopic points, for which the shear stress amplitude is a
maximum. This implies that the number of slip system in
4.4. Matake criterion the aggregate is sufficient to be representative of the num-
ber of slip planes in the physical space.
Fig. 16 shows the resolved shear stress amplitude as a
function of the maximum normal stress determined from 4.4.1. Effect of the material behaviour on the mechanical
the totality of the modelled slip systems. The results of variables of the criterion
the different loading modes and the different material con- Regarding the influence of the type of material behav-
stitutive laws are shown. As before, the cloud of gray iour at the mesoscopic scale, Fig. 16(b) shows that cubic
points shows the mesoscopic result for the different slip elasticity strongly influences the resolved shear stress
systems. The black point shows the macroscopic result amplitude as well as the maximum normal stress. The im-
and the solid line is the Matake criterion threshold line. pact of crystal plasticity combined with isotropic elasticity
C. Robert et al. / Mechanics of Materials 55 (2012) 112–129 125
Fig. 16. The predictions from the Matake criterion at the macroscopic scale (in black) and the mesoscopic scale (in gray).
(see Fig. 16(c)) is low when considering the maximum va- From a constitutive point of view at the mesoscopic
lue of the resolved shear stress amplitude. However it in- scale, for the case where the fatigue limit is close to the
creases the maximum normal stress by several percent. macroscopic yield stress, crystal plasticity has relatively
In terms of the effect of crystal plasticity combined with little effect on the mesoscopic response. However, cubic
cubic elasticity (see Fig. 16(d)), a slight decrease is ob- elasticity (or local anisotropic elasticity) has a considerable
served in the maximum resolved shear stress amplitude influence on the dispersion of the mesoscopic values.
and maximum normal stress. In the case of a local stress approach, the two criteria
investigated (i.e., Dang Van and Matake) also highlight
5. Discussion the dispersion of the variables computed at the mesoscopic
scale. In terms of behaviour, crystal plasticity combined
Thanks to the preceding investigation of different high with isotropic elasticity has little effect on the maximum
cycle multiaxial fatigue criteria, the dispersion of the shear stress as well as the amplitude of the maximum re-
mechanical quantities at the mesoscopic scale has been solved shear stress compared to the case of isotropic elas-
highlighted. ticity, without plasticity. In terms of hydrostatic stress and
The use of the Crossland criterion, based on a macro- maximum normal stress, a dispersion of a few percent is
scopic approach, shows that the macroscopic point is lo- observed. However, for these two criteria, the critical val-
cated roughly in the middle of the cloud of gray points ues obtained at the mesoscopic scale are very close to
representing the mesoscopic predictions. The use of this those associated with the macroscopic scale. For this type
criterion therefore results in a macroscopic prediction of criteria, cubic elasticity has a strong influence on the re-
which corresponds to the ‘‘average’’ mesoscopic behaviour. sults. The maximum shear stress and the amplitude of the
126 C. Robert et al. / Mechanics of Materials 55 (2012) 112–129
Fig. 17. Comparison of the maximal shear stress (a) between isotropic elasticity (Iso. E.) and cubic elasticity (Cub. E.). and (b) between cubic elasticity (Cub.
E.) and cubic elasticity + crystal plasticity (Cub. E. + CP).
maximum resolved shear stress increase significantly with to hardening. Lower dispersion is also observed (i.e.,
respect to the isotropic elasticity case (even if crystal plas- ‘‘smoothing’’ effect of the microplasticity).
ticity tends to decrease them). In terms of the hydrostatic The maximum resolved shear stress (or the maximum
stress, the maximum is more than 20 MPa greater than shear stress) increases significantly for a behaviour includ-
the maximum value obtained with an isotropic elastic ing cubic elasticity at the mesoscopic scale. To further
behaviour, for the three studied load cases. Crystal plastic- investigate this particularity, apparent Schmidt factors
ity does not appear to influence these maximum values. are analysed in the following. These factors are defined
For the case of cubic elasticity, it is clear that the meso- using the following equation:
scopic hydrostatic stress is not equal to the macroscopic
hydrostatic stress, as assumed in the Dang Van criterion.
sr ðtÞ
F S ¼ max ð16Þ
These differences have a significant impact on the max-
t RI ðtÞ
imum value of the Dang Van coefficient of danger, but in where RI is the largest principal stress at the macroscopic
terms of critical planes orientations, the maximums are scale.
not strongly modify by cubic elasticity. In terms of the Fig. 18 shows the accumulated apparent Schmidt factor
maximum normal stress, its maximum value also increases density, considering all slip systems computed for the nine
by the same order of magnitude as the hydrostatic stress. microstructures, for the different loading modes and for
The maximum shear stress for each slip plane, deter- the four material constitutive laws studied. It can be ob-
mined for the last cycle, is plot in Fig. 17(a) for the case served that the majority of slip systems have an apparent
of isotropic elasticity and cubic elasticity. The gray points Schmidt factors with a lower intensity for the cubic elastic
are plotted for the same slip plane. The solid line repre- case (with or without crystal plasticity). However, for this
sents perfect agreement between isotropic elasticity and type of behaviour, the maximum values are much higher.
cubic elasticity. Large dispersion is observed, which is Globally, crystal plasticity tends to reduce the maximum
globally centred on the solid line. The largest dispersion value.
is observed for the highest values of the shear stress for To better understand these curves, the maximum
the tensile loading condition. This is not the case for the apparent Schmidt factors are plotted as a function of the
other two loading types. percentage of the total number of systems (see Fig. 19).
The maximum shear stress for each slip plane, deter- The 100th percentile corresponds to all of the simulated
mined for the last cycle, is plot in Fig. 17(b) for the case slip systems and 90th percentile corresponds to the 90%
of cubic elasticity and cubic elasticity + crystal plasticity. of systems with the lowest apparent Schmidt factors. For
The gray points are plotted for the same slip plane, and the 90th percentile the tendencies, with respect the type
the solid line represents perfect agreement between the of elasticity, are inverted compared to 100th percentile.
two behaviour types. It can be seen that crystal plasticity The apparent Schmidt factors are lower for cubic elastic
tends to reduce the highest values of the shear stress due behaviour. Crystal plasticity, has a relatively small effect,
C. Robert et al. / Mechanics of Materials 55 (2012) 112–129 127
Fig. 18. Density of the apparent accumulated Schmidt factors for the different material constitutive laws studied.
but never-the-less tends to decrease the maximum values, crease in the apparent Schmidt factors as a function of
for all the loading modes and the two types of elasticity the coefficient of anisotropy (2C 44 =ðC 11 C 12 Þ). He noted
studied. For the case of uniaxial tension, it goes from that with a coefficient of anisotropy of 3.26 the maximum
0.44 to 0.43 for isotropic elasticity and from 0.40 to 0.39 apparent Schmidt factor is 0.43. In our work, for the cubic
for cubic elasticity. elasticity, without plasticity case (i.e., coefficient of anisot-
For the case of uniaxial tension Sauzay (2007) studied ropy equal to 4.36), the maximum apparent Schmidt factor
the effect of the free surface using cubic elasticity on the is 0.71 taking into account all grains and becomes equal to
maximum values of the apparent Schmidt factors. Poly- 0.40 for the 90th percentile. For the case of isotropic elas-
crystalline aggregates with grains modelled using an ex- ticity, these factors are 0.5 and 0.44 respectively. The major
truded hexagonal geometry, surrounded by an isotropic difference, with respect to Sauzay’s work (Sauzay, 2007) is
homogeneous matrix were considered. He obtained a de- the 3D geometry which takes into account the free surface.
128 C. Robert et al. / Mechanics of Materials 55 (2012) 112–129
Fig. 19. Apparent Schmidt factors as a function of the percentile. (Cub. E. = cubic elasticity, Iso. E. = isotropic elasticity, CP = crystal plasticity.)
facilities. International Journal for Numerical Methods in Engineering Poncelet, M., Doudard, C., Calloch, S., Weber, B., Hild, F., 2010. Probabilistic
79 (11), 1309–1331. multiscale models and measurements of self-heating under
Groeber, M., Ghosh, S., Uchic, M.D., Dimiduk, D.M., 2008. A framework for multiaxial high cycle fatigue. Journal of the Mechanics and Physics
automated analysis and simulation of 3d polycrystalline of Solids 58 (4), 578–593.
microstructures. Part 2: synthetic structure generation. Acta Przybyla, C.P., McDowell, D.L., 2010. Microstructure-sensitive extreme
Materialia 56 (6), 1274–1287. value probabilities for high cycle fatigue of ni-base superalloy in100.
Guilhem, Y., Basseville, S., Curtit, F., Stéphan, J.-M., Cailletaud, G., 2010. International Journal of Plasticity 26 (3), 372–394.
Investigation of the effect of grain clusters on fatigue crack initiation Saintier, N., Robert, C., Hor, A., Morel, F., Palin-Luc, T., 2011. High cycle
in polycrystals. International Journal of Fatigue 32 (11), 1748–1763. fatigue strength criteria and microstructurally related modeling of
Lin, T.H., 1957. Analysis of elastic and plastic strains of a fcc crystal. metals. Advances in Heterogeneous Material Mechanics, In: 3rd
Journal of the Mechanics and Physics of Solids 5, 143. Conference for Heterogeneous Materials Mechanics (ICHMM),
Lukás, P., Kunz, L., 1989. Effect of mean stress on cyclic stress–strain Shanghaï.
response and high cycle fatigue life. International Journal of Fatigue Sauzay, M., 2007. Cubic elasticity and stress distribution at the free
11 (1), 55–58. surface of polycrystals. Acta Materialia 55 (4), 1193–1202.
Luther, T., Könke, C., 2009. Polycrystal models for the analysis of St-Pierre, L., Héripré, E., Dexet, M., Crépin, J., Bertolino, G., Bilger, N., 2008.
intergranular crack growth in metallic materials. Engineering 3D simulations of microstructure and comparison with experimental
Fracture Mechanics 76 (15), 2332–2343. microstructure coming from o.i.m. analysis. International Journal of
Matake, T., 1977. An explanation on fatigue limit under combined stress. Plasticity 24 (9), 1516–1532.
Bulletin of the JSME 141, 257–263. Taylor, G.I., 1938. Plastic strains in metals. Journal Institute of Metals 62,
Meric, L., Cailletaud, G., 1991. Single crystal modeling for structural 307.
calculations: Part 2—finite element implementation. Journal of Weibull, W., 1939. A statistical theory of the strength of materials.
Engineering Materials and Technology 113 (1), 171–182. Proceedings of the Royal Swedish Institute for Engineering Research,
Monchiet, V., Charkaluk, E., Kondo, D., 2006. Plasticity-damage based Report, 151.
micromechanical modelling in high cycle fatigue. Comptes Rendus Weibull, W., 1951. A statistical distribution function of wide applicability.
Mécanique 334 (2), 129–136. ASME – Journal of Applied Mechanics 18, 293–297.
Monchiet, V., Charkaluk, E., Kondo, D., 2008. A micromechanical Weyer, S., Fröhlich, A., Riesch-Oppermann, H., Cizelj, L., Kovac, M., 2002.
explanation of the mean stress effect in high cycle fatigue. Automatic finite element meshing of planar voronoi tessellations.
Mechanics Research Communications 35 (6), 383–391. Engineering Fracture Mechanics 69 (8), 945–958.
Morel, F., Huyen, N., 2008. Plasticity and damage heterogeneity in fatigue. Zhang, K.S., Wu, M.S., Feng, R., 2005. Simulation of microplasticity-
Theoretical and Applied Fracture Mechanics 49 (1), 98–127. induced deformation in uniaxially strained ceramics by 3-d voronoi
Musienko, A., Tatschl, A., Schmidegg, K., Kolednik, O., Pippan, R., polycrystal modeling. International Journal of Plasticity 21 (4), 801–
Cailletaud, G., 2007. Three-dimensional finite element simulation of 834.
a polycrystalline copper specimen. Acta Materialia 55 (12), 4121–
4136.