Vous êtes sur la page 1sur 12

Computational Materials Science 40 (2007) 201212

www.elsevier.com/locate/commatsci

Coupled mesoscopic constitutive modelling


and nite element simulation for plastic ow
and microstructure of two-phase alloys
R. Ding, Z.X. Guo *, M. Qian
Department of Materials, Queen Mary, University of London, Mile End Road, London E1 4NS, UK

Received 24 May 2006; received in revised form 27 November 2006; accepted 5 December 2006
Available online 8 February 2007

Abstract

A mesoscopic dislocation-based model was coupled with macro-scale nite element analysis for concurrent study of local plastic ow
and microstructure of two-phase alloys during thermomechanical deformation. The model was implemented in the ABAQUS code to
simulate the thermomechanical processing of a Ti6Al4V alloy in the (a + b) phase eld, with consideration of the eects of local
dislocation density variation, deformation heating and phase volume fraction. The simulation show that the intergranular interaction
results in non-uniform distribution of dislocations within each grain, particularly in the initial stages of deformation. Phase boundaries
pose stronger inuence on deformation than grain boundaries. The onset of shear localization was strongly inuenced by the strain rate
sensitivity parameter, deformation heating, phase volume fraction, and the die/sample friction coecient. Both deformation heating and
phase transformation in the shear-localized region contributes to the ow-stress variation during processing. The phase volume fraction
largely aects the microstructure, distribution of the equivalent stress, but not the equivalent strain.
 2006 Elsevier B.V. All rights reserved.

Keywords: Microstructural modelling; Finite element simulation; Thermomechanical processing; Two-phase alloys; Ti alloys

1. Introduction pensable tool for such purposes, but has not been widely
applied to two-phase materials.
Many engineering alloys are two-phase materials. Their Deformation of single-phase polycrystalline materials
thermomechanical processing is often carried out in the has been investigated by a variety of classic elasto-plastic
two- (major) phase eld to impart optimum microstructure deformation models that consider the interaction among
and mechanical properties. Here, the major concerns are a neighbouring grains [210]. These models treat dierently
balanced grain/phase structure, ow resistance and defor- the two essential deformation constraints of polycrystalline
mability, whereas the issue with preferred crystallographic materials, i.e. intergranular force equilibrium and strain
orientation, or texturing, is not as important as it is in cold compatibility, but none has strictly satised both con-
deformation [1]. Hence, it is of theoretical and practical sig- straints simultaneously. For instance, the full-constraint
nicance to develop easily assessable models for the under- Taylor model assumes an identical strain in all grains,
standing of concurrent plastic ow and microstructural but ignores the intergranular force equilibrium [2,3]. The
variation at the local level for the processing of such alloys. Sachs model satises the force equilibrium but the inter-
Computational mesoscopic modelling has become an indis- grain compatibility is ignored [1,7]. The self-consistent
models consider the equilibrium and compatibility over
an averaged eld [46]. In the past two decades, nite
*
Corresponding author. Tel.: +44 20 7882 5569; fax: +44 20 7882 5154. element (FE) method has been widely used to study
E-mail address: x.guo@qmul.ac.uk (Z.X. Guo). elasto-plastic deformation of crystalline materials. FE

0927-0256/$ - see front matter  2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.commatsci.2006.12.003
202 R. Ding et al. / Computational Materials Science 40 (2007) 201212

modelling is usually combined with classic crystalline method for the FE modelling of multiphase composites,
elasto-plastic deformation theories, where the constraints such as AgNi ber composites and AgNi particulate
of intergranular force equilibrium and strain compatibility composites. The phase boundaries are put inside the ele-
can be strictly satised and specic deformation around ments instead of the normal element edges, and the phase
grain boundaries and in the grain interior can be readily properties are designated to the integration points. Grujicic
simulated [1124]. and Sankaran [48] developed a constitutive model which
Various constitutive models for FE analysis of elasto- describes transformation plasticity accompanying stress-
plastic deformation of crystalline materials have been assisted martensitic transformation in two-phase materials
developed [16,2529]. Most of the models consider defor- consisting of a stable matrix and a transforming dispersed
mation of single-phase crystalline materials. For two- or phase, and the model was used to analyse the uniaxial ten-
multi-phase materials, the mechanical properties and sile behaviour of the two-phase systems. All these studies
deformation behaviour are strongly inuenced by phase that use FE modelling of the deformation characteristics
volume fraction, phase distribution, properties of the con- of two- or multi-phase materials have considered the inu-
stituent phases and interactions across phase boundaries, ence of phase volume fraction and phase distribution, but
particularly when there is a large dierence in their none of these has incorporated the inuence of mesoscopic
mechanical properties. Although various forms of the law dislocation density variation on the deformation behav-
of mixtures have been proposed to predict the mechanical iour. In practice, microstructural evolution of two- or
properties of two- and multi-phase materials [3032], these multi-phase materials during deformation is closely associ-
can only represent a macroscopic approximation of the ated with dislocation activities as well as the phase ratio
macroscopic mechanical properties from the properties of and distribution. The eect of interfacial friction was only
the constituent phases, and are unable to show detailed studied for either single crystals or at the mesoscopic scale
microstructural deformation and interactions at grain [3436].
and phase boundaries. In this paper, a meso-scale dislocation-based constitu-
Friction at die/sample interfaces plays an important role tive model was integrated with a micro-scale rate-depen-
for stressstrain relationships at medium to high levels of dent plastic ow equation, and then coupled with nite
plastic deformation [38,33], particularly at the macroscopic element analysis for accurate study of both the stressstrain
scale [3436]. In an example, friction-induced rapid hard- relationship and microstructural evolution during thermo-
ening for single crystal deformation can lead to 25% incre- mechanical deformation of two-phase alloys, with due con-
ment of the ow stress [37]. So it is essential to consider sideration of thermal softening as a result of deformation
friction in the analysis of polycrystalline deformation. heating. Die/sample interface friction has also been consid-
FE modelling is a powerful tool for simulation of plastic ered for both sticking and slipping frictions, and the
deformation of two- or multi-phase materials. The inu- coecient of slipping friction varies from 0.1 to 0.01. The
ence of phase volume fraction, phase distribution, size model was implemented into the commercial FEM package
and properties of constituent phases on the deformation ABAQUS through the user subroutine UMAT. The inu-
behaviour of two- or multi-phase crystalline materials can ence of thermomechanical processing parameters, such as
be investigated with great insight if FE modelling is cou- strain rate, temperature, and phase volume fraction on
pled with more elaborate constitutive models for polycrys- the deformation characteristics of an (a + b) Ti6Al4V
talline deformation. alloy was studied as an example. The simulated stress
Several FE investigations have been carried out for two- strain curves and grain structure were compared with
or multi-phase composites and crystalline materials. Karls- experimental results for the Ti6Al4V alloy deformed in
son and Linden used FE modelling to study the yield and the (a + b) phase eld.
work hardening of ferritepearlite aggregates with a con-
tinuous ferrite matrix [39]. Ankem et al. [4043] used FE 2. Constitutive modelling
modelling to investigate the stressstrain relationship of
two-phase materials at dierent particle sizes and phase An isotropic constitutive model based on the dislocation
distributions. According to their results, most of the strain theory was originally developed by Estrin et al. [27,49]. The
is carried by the softer phase and the stress by the harder model uses the dislocation density as a state variable to
phase, and transverse stresses are generated as a result of characterise the microstructural state of a single-phase
the interaction between the phases. The stress and strain material. The one-parameter KM model developed by
distribution and the magnitude of the transverse stresses Kocks and Mecking [5053] is used to describe the disloca-
depend on the phase volume fraction and the strength dif- tion evolution during deformation, and the macro ow
ference between the two phases. Bolmaro et al. [44] used stress is related to the microstructural evolution through
FE analysis to study the inuence of volume fraction, the dislocation density. Here, a model based on Estrins
geometry, phase distribution, strain hardening and yield approach was developed, by further considering the inu-
stress ratio on the deformation behaviour of two-phase ences of grain orientation, phase volume and the tempera-
materials. Steinkop et al. [4547] developed a reasoning ture variation, to simulate the microstructural evolution
algorithm for net-adaptation and the multiphase element and viscoplastic deformation of two-phase materials.
R. Ding et al. / Computational Materials Science 40 (2007) 201212 203

According to the theory of plastic deformation of poly- where h0 is the strain hardening rate, and rs is the saturated
crystalline solids, the total strain rate tensor e_ is determined stress. h0 and rs can be obtained from experimental results
by the sum of the elastic and plastic components, e_ e and e_ vp , at given processing conditions.
respectively: According to previous investigations [5456], most of
the plastic work during deformation is converted to heat,
e_ e_ e e_ vp 1
which causes the temperature of a deforming material to
The elastic strain rate is related to the stress tensor through rise. If the deformation rate is suciently high, and/or
the Hookes law: the deformation is under well-insulated conditions, the
process can be considered to be adiabatic. According to
r_ L : e_ e 2 the energy balance for adiabatic process, the temperature
where L is the tensor of elastic moduli, the colon : de- increase can be calculated by
notes the tensor dyadic product, i.e. L : e_ e Lijkl e_ elk . The qcp T_ b
re_ 14
plastic strain rate is expressed by the Levyvon Mises equa-
tion for isotropic materials [49]: where q is dislocation density, cp the specic heat, and b a
coecient that represents the ratio of the amount of plastic
e_ vp pe_ 3 work to the corresponding heat generated (assumed to be
3 r0 0.9 in the current case). The mechanical properties of con-
p 4 stituent phases, such as shear modulus and strain harden-
2 r

ing strength vary with increasing temperature.
where e_ is the von Mises equivalent plastic strain rate, r
 is
0
the equivalent stress, and r is the deviatoric stress tensor. 3. Model implementation in ABAQUS
2
e_ 2 e_ vp : e_ vp 5 The commercial FEM package ABAQUS was employed
3
3 in this study. The constitutive equations described above
r2 r0 : r0 6 were implemented into the package through the user
2
1 subroutine UMAT [57]. For such implementation, the
r0 r  r : II 7 Jacobian matrix [oDrij/oDeij] and the updated stress state
3
expression need to be derived from the above constitutive
where I is the identity tensor. Combining Eqs. (1)(3) yields equations. Because the constitutive equations is numeri-
the relationship between stress and viscoplastic strain rate: cally sti, which usually causes instability and poor conver-
r_ L : _e  pe_ 8 gence in numerical calculations, a tangent modulus method
for rate-dependent materials proposed by Peirce et al. [58]
The equivalent plastic strain rate can be calculated through was adopted here to increase numerical stability, as shown
a power-law viscoplastic strain rate relationship [11,14]: in the following.
 n The increment of the equivalent viscoplastic strain is cal-
r

_e e_ 0 9 culated using a linear interpolation over a time increment:
g
De Dt1  he_ t he_ tDt  15
where e_ 0 and n are materials constants, g characterizes the
strain hardening strength which is associated with the where h is the interpolation parameter and chosen to be
dislocation density and can be calculated using the follow- 0.5, following Peirce et al. [58]. According to Eq. (9), the
ing equation [49] : equivalent strain rate e_ is a function of the equivalent stress
p and the hardening strength. The term e_ tDt can be approx-
g Malb q 10 imated using a Taylor series expansion:
where M is the Taylor factor of grains, a is a constant, l is oe_ oe_
the shear modulus, b is Burgers vector, and q is the dislo- e_ tDt e_ t r Dg
D 16
o
r og
cation density. Variation of the dislocation density during
deformation can be described using the KM model [5053]: From Eqs. (5)(8), D
r can be derived as
dq p r Dtp : L : e_  Dep : L : p
D 17
k1 q  k2q 11
de
For isotropic materials, the elastic constant L can be
where k1 is the strain hardening parameter and k2 is the expressed in the index notation as
softening parameter which is a function of deformation  
2m
temperature and strain rate. k1 and k2 are determined by Lijkl l dik djl dil djk dij dkl 18
1  2m
k 1 2h0 =Malb 12
where m is the Poissons ratio, dij is the Kroneckers delta.
k 2 2h0 =rs 13 Eq. (17) can be simplied to
204 R. Ding et al. / Computational Materials Science 40 (2007) 201212

3lDt 0 Table 1
r
D r : e_  3lDe 19 Values of material parameters used in the model [65]
r

ba = 2.95 1010 m bb = 2.86 1010 m
From Eqs. (10) and (11), Dg can be expressed as
la = 4.36 1010 N/m2 (at 300 K) lb = 2.05 1010 N/m2 (at 300 K)
Dg A  De 20 qa = qb = 4429.0 kg/m3 C ap C bp 965 J=kg K at 1173 K
p e_ 0 1:0 s1 n = 100 ma = mb = 0.27
A alb  k 1  k 2 q  De=2 21 a = 0.2 b = 0.9
Combining Eqs. (16)(21), Eq. (16) can be expressed as
De e_ t gt materials, the rate of evolution of dislocation density varies
rt Br0 : e_ 
 22
Dt C with the constituent phases due to their dierent mechani-
where cal properties. The saturated stress and hardening rate of
each constituent phase need to be determined in order to
B 3lhmDt= r 23 calculate the dislocation density variation of the phases.
Cr t gt Be_ t 3lgt A
rt  24 In many cases, the property values of the constituent
phases cannot be readily obtained from experiment, but
The subscript t in the above equations refers to the values the properties of two-phase materials and of one constitu-
of the corresponding parameters at the time t. Substituting ent phase can be determined from experiment. In this case,
Eq. (22) into Eq. (8), using a tensor index notation, leads to the mechanical properties of the alpha phase titanium were
the following: from our experiments. The stress of the beta-phase was
2lm estimated using an extended law of mixtures approach
Drij 2lDeij dij Dekk  Dr0ij r0kl Dekl
1  2m proposed by Fan and Miodownik [60].
 
1 In order to consider the eects of dierent crystallo-
 H rij  rkk dij 25 graphic orientations of grains in polycrystalline materials,
3
the Taylor factor is set to a value between 2.0 and 3.5, ran-
where parameters D and H are domly for each grain, and kept constant for each simula-
3lBe_ t gt tion. Grain orientation rotation is not considered in this
D 26 model, and hence relative slip along grain boundary is
C rt
not allowed in the current model. The sticking friction con-
3le_ t gt Dt
H 27 dition is assumed in most of the calculation where there is
C no relative displacement between the die/sample interface.
The Einsteins summation rule [59] is used in the model. For comparison, slipping friction was also considered for
The Jacobian matrix can be obtained from Eq. (25) as cases with friction coecient l = 0.010.2 (l = f/p, where
 2 f is the frictional stress and p is the normal pressure). The
oDrxx 2lm 1
2l  D rxx  rii 28 microstructural evolution, stressstrain relationship and
oDexx 1  2m 3 dislocation density variation were investigated at dierent
  
oDrxx 2lm 1 1 thermomechanical processing conditions and the results
 D rxx  rii ryy  rii 29
oDeyy 1  2m 3 3 are shown in the following section.
oDrxy 2l  Dr2xy
30
oDexy 2 4. Simulated results and discussion
It should be noted that during the derivation of Eq. (30)
The phase volume fraction of the Ti6Al4V alloy varies
ABAQUS uses the engineering shear strain cxy, instead of
according to the processing temperature in the (a + b)
the shear strain exy (cxy = 2exy).
phase eld. The simulation was carried out at dierent
During calculation, the dislocation density, its rate of
temperatures in the range of 850950 C. Fig. 1a shows a
variation, shear modulus, and temperature of each integra-
sample simulated at 900 C, where the volume fraction of
tion point are saved as the solution dependent state vari-
the a-phase is 42%. The sample includes 26 grains. The a
ables in ABAQUS. The strain increment, time increment,
grains are shown in white and the b grains in grey.
stress and solution dependent state variables are passed
Fig. 1b shows the corresponding FE mesh, which includes
to the subroutine UMAT from ABAQUS at the beginning
1714 quadrilateral elements. The simulation was carried
of each increment. UMAT calculates the materials Jaco-
out under the 2-D plane strain condition.
bian matrix, updates the stress and the solution dependent
state variables, and returns the results to ABAQUS at the
end of each increment. 4.1. Local stress and strain distribution
The model derived above was used to simulate polycrys-
talline viscoplastic plane-stain deformation of a two-phase Fig. 2 shows the predicted distribution of the von Mises
Ti6Al4V alloy in the (a + b) phase eld. The parameters equivalent stress and the equivalent strain for two overall
used in the model are listed in Table 1. For two-phase compression strains of 0.06 and 0.5, at the deformation
R. Ding et al. / Computational Materials Science 40 (2007) 201212 205

Fig. 1. Schematic of a model sample (a) and the corresponding FEM mesh (b) used in the simulations: the white area in (a) represents the a-phase and the
grey area the b-phase. The area represents a real space of 8 8 mm2.

temperature of 900 C and the strain rate of 0.05 s1. It is dence is relatively weak when the friction coecient is
noted that the equivalent stress of the a-phase is higher large. This point will be further claried in Fig. 6 in the
than that of the b-phase for both cases because both the study of the inuence of phase volume fraction on the
initial yield stress and the strain hardening strength of deformation behaviour.
the a-phase is greater than that of the b-phase. The equiv-
alent stress decreases in the shear-localized region because 4.2. Variation of the dislocation density
of deformation heating. Fig. 2b and c shows that the equiv-
alent strain is quite uniform in the sample except in the The local dislocation density variation during deforma-
shear-localized region, where the equivalent strain is rela- tion was calculated using the KM model. The rate of
tively high as the increase in temperature facilitates disloca- variation of local dislocation density depends on two
tion slip and climb in the region. competing processes: work hardening, due to dislocation
The gradient of the equivalent stress around the phase accumulation, and softening, due to dynamic recovery.
boundaries is much greater than that around the grain These two processes are related with the local strain of each
boundaries of the same phase (Fig. 2). The phase bound- grain, which is inuenced by intergranular interactions in
aries are more inuential to deformation than the grain polycrystalline deformation. Fig. 4 shows the dislocation
boundaries, because the dierence in the deformation density distribution at dierent overall strains of 0.06, 0.2
behaviours of two phases at the phase boundary comes and 0.5. It is noted that the distribution of the dislocation
from two sources: dierent mechanical properties and density is not uniform at the interior of each grain. The
dierent crystallographic orientations of the two phases. dislocation densities of some grains reached the saturated
For a grain boundary, only the latter eect exists. The value at very early stages of deformation, e.g. grain 9, 11,
non-uniformity of the stress and the strain at a phase 13 at the overall strain of 0.06, but the dislocation densities
boundary is more severe than at a grain boundary. The of some grains have not reached the saturated state even at
deformation behaviour and the stress distribution of the the overall strain of 0.5, e.g. grain 2, 3 and 12. Fig. 4 also
whole sample are much inuenced by the distribution of shows that the average dislocation density of the a-phase is
the phase boundaries in order to maintain force equilib- lower than that of the b-phase because the level of defor-
rium and strain compatibility at the phase boundaries. mation of the a-phase is less than that of the b-phase. In
The larger the dierence between the mechanical properties the region of shear localization, the dislocation density
of the two constituent phases, the more important is the decreases because of the increase of temperature. When it
inuence of the phase boundaries. This is further conrmed reaches the saturated value, the dislocation density of a
by the rate of dislocation density variation in the following grain remains constant during further deformation, if there
section. is no inuence of the deformation heating or occurrence of
The friction between the sample and the die inuences dynamic recrystallization.
the deformation behaviour of the sample. Fig. 3 shows Fig. 5 illustrates the distribution of the rate of disloca-
the distribution of the equivalent strain after an overall tion density variation dq=de at dierent overall strains.
compression strain of 0.06 at three dierent friction coe- It is shown in Fig. 5a that the dislocation density increases
cients of 0.01, 0.1 and 0.2. Compared with Fig. 2c, the rapidly in both the a- and the b-phase when the overall
equivalent strain distribution depends on the phase distri- strain is 0.01, but the rate of increase in the a-phase is
bution when the friction coecient is small, but this depen- greater than in the b-phase, especially in grains 8 and 17.
206 R. Ding et al. / Computational Materials Science 40 (2007) 201212

Fig. 2. Distribution of the von Mises equivalent stress and strain at dierent overall strains: (a) equivalent stress at a strain of 0.06; (b) equivalent stress at
a strain of 0.5; (c) equivalent strain at a strain of 0.06; and (d) equivalent strain at a strain of 0.5.

The dislocation densities of both the a- and the b-phase has not reached the saturated value in these areas even at
grains have not reached the saturated values at this stage. the overall strain of 0.2 because the dislocation accumula-
When the strain reaches 0.2, Fig. 5b, a negative rate is tion rate is relatively slow.
noted in most of the area of the a-phase and the b-phase It is also noted that the phase boundaries pose greater
grains. The reason is that the temperature rise due to defor- inuence on the dislocation density and its variation rate
mation heating enhances dynamic recovery and reduces the than the grain boundaries because of a relatively severe
saturated dislocation densities of both phases. The varia- non-uniformity of deformation at the phase boundaries.
tion rate is still positive in the area of the deformation The gradients of the dislocation density and the variation
dead zone, but the value is much smaller than that at rate at the phase boundaries are higher than those at the
the strain of 0.01. It shows that the dislocation density grain boundaries. For example, the phase boundaries
R. Ding et al. / Computational Materials Science 40 (2007) 201212 207

ular interaction and the crystallographic orientations on


the variation of dislocation density. For further simulation
of the microstructural evolution and the dislocation varia-
tion involving, e.g. static recrystallization and dynamic
recrystallization, the KM type model should be coupled
with the polycrystalline viscoplastic constitutive model
with further consideration of the crystallographic orienta-
tion and intergranular interactions. In the recent years, sev-
eral investigations have been carried out in this eld but
only for single-phase alloys [22,6163].

4.3. Shear localization

It is noted from the simulated results that the onset of


shear localization and shear bands depends on many
factors, such as the friction coecient between the sam-
ple/die interface, the strain rate sensitivity parameter, m
(=1/n), the deformation heating and the phase volume
fraction. Fig. 6a1, a2 and c illustrates the deformed meshes
at a strain rate of 0.05 s1 and an overall strain of 0.5 at
900 C, but with dierent rate sensitivity parameters. It
shows that when the rate sensitivity parameter is 0.2, the
shear bands are not formed during deformation, only a
very weak shear-localized region is observed. When the
rate sensitivity parameter decreases to 0.01, shear localiza-
tion becomes much pronounced. This agrees with the
experimental results showing considerable shear localiza-
tion at the processing temperature of 900 C [1]. The
formation of shear bands is directly related to the von
Mises equivalent strain rate. It is understandable because
the magnitude of the equivalent strain rate at each element
determines its equivalent strain, which is represented mac-
roscopically through the non-uniformity of deformation of
the sample, e.g. shear localization. Fig. 6a2, b2 and c shows
the inuence of the friction coecient on the formation of
the shear localization. It is clear that a large friction facil-
itates slip-banding in the samples.
Fig. 3. Distribution of von Mises equivalent strain at an overall
compression strain of 0.06 under dierent friction coecients: (a) 0.01; Fig. 6d shows the distribution of the von Mises equiva-
(b) 0.1; and (c) 0.2. lent strain rate under the same condition as in Fig. 6c. It
can be seen that the shear localization corresponds to the
area where the von Mises equivalent strain rate is relatively
among grains 9, 10 and 16 are one of the regions with the high.
lowest rate of dislocation density variation, whereas the
phase boundaries among grains 16, 17 and 23 show rela- 4.4. Inuence of deformation heating
tively high rates.
Clearly, the variation of the dislocation density is very Fig. 7 shows the temperature distribution at dierent
complex inside each grain and at the grain boundaries, strains, where the temperature is noted to increase with
and is inuenced not only by the material properties of strain in the area of shear localization. The maximum
each phase, but also by the grain interactions. The disloca- increase of temperature is 48 C at the overall strain of
tion density is always changing before reaching the satu- 0.2 and 92 C at the overall strain of 0.5. The temperature
rated value, but remains steady once the dynamic balance outside the shear-localized region remains the same as the
between strain hardening and dynamic recovery is reached. ambient processing temperature (heat conduction across
The deformation heating reduces the dislocation density the sample is not considered in the model). The increase
and inuences its distribution. The KM type model is in temperature because of the deformation heating in the
commonly used to calculate the mean dislocation density shear-localized region may also lead to phase transforma-
variation in the whole sample during thermomechanical tion, especially when the temperature reaches the b transus
processing, but cannot treat the inuence of the intergran- of 995 C at a high strain. Fig. 8 shows the simulated ow
208 R. Ding et al. / Computational Materials Science 40 (2007) 201212

Fig. 4. Dislocation density variation at dierent overall compression strains: (a) 0.06; (b) 0.2; and (c) 0.5.

curves with and without the consideration of deformation cessing of two-phase materials. The simulated results agree
heating at two dierent strain rates, 0.05 and 0.5 s1, well with existing experimental observations [64,65].
respectively. The simulated results were compared with
experiments [1]. The experimental results show that the 4.5. Inuence of phase volume fraction
ow stress decreases with strain. The simulated ow-stress
curves also decrease with strain at all the simulated strain Phase volume fraction and distribution inuence the
rates when the deformation heating is considered, but the deformation of two-phase alloys. In order to single out
magnitude of decrease is less than that of the experiment. the inuence of phase volume fraction, simulated process-
The discrepancy may be attributed to the fact that the ing was carried out in silico at 900 C under a given strain
a ! b-phase transformation due to local temperature rate of 0.05 s1 to an overall strain of 0.5 for two dierent
increase may cause further reduction of the ow stress, phase volume fractions of the two-phase (a + b) Ti6Al
which has not been considered in the simulation. Fig. 8 also 4V alloy: (a) 62% and (b) 33%. Fig. 9ad illustrates the
indicates that the ow stress increases with strain because corresponding simulated equivalent stress and strain distri-
of strain hardening if the deformation heating is not con- butions for the two cases, respectively. It can be seen that
sidered in the simulation, as is the case in many previous the maximum equivalent stress decreases with decreasing
simulations. From the simulated results, it may be inferred the volume fraction of the a-phase. The average stress
that there are two causes for the decrease of the ow stress of the a-phase is higher than that of the b-phase in both
during thermomechanical deformation, one is the increase cases. The maximum stress is located in the a-phase grains.
of temperature because of deformation heating, and the It is noted from Fig. 9b and d that the distributions of the
other is the a ! b-phase transformation during hot pro- equivalent strain are quite similar: the area of the high
R. Ding et al. / Computational Materials Science 40 (2007) 201212 209

Fig. 5. Distribution of the rate of dislocation density variation at dierent overall compression strains: (a) 0.01 and (b) 0.15.

Fig. 6. Shear localization vs. the strain rate sensitivity parameter and friction at 0.05 s1 and 900 C and an overall strain of 0.5. The eect of strain rate
sensitivity parameters (m = 1/n) under sticking friction are noted in: (a1) 0.2; (b1) 0.03; and (c) 0.01; and the inuence of sample/die friction (under
m = 0.01) are compared in: (a2) 0.1; (b2) 0.2; and (c) sticking friction. The distribution of the von Mises equivalent strain rate of (c) is shown in (d).

equivalent strain corresponds to the region of relatively the initial grain geometries are identical, the deformed
high temperature due to deformation heating. Although microstructures of the two samples are very dierent at
210 R. Ding et al. / Computational Materials Science 40 (2007) 201212

Fig. 7. Temperature distribution at a strain rate of 0.05 s1, 900 C and an overall strain of: (a) 0.2 and (b) 0.5.

350 phases, the intergranular interactions are very dierent,


Simulated without deformation heating leading to large variation in their deformation.
300
Flow stress (Mpa)

250 5. Conclusions
200 Simulated with deformation
heating A simulative model for high-temperature deformation
150 and microstructural evolution of two-phase materials is
established by direct coupling of meso-scale dislocation
100 density variation and micro-scale nite element analysis
50 Experimental using the ABAQUS code. The model was successfully
applied to the thermomechanical processing of an (a + b)
0 Ti6Al4V alloy. The inuence of dislocation density var-
0 0.1 0. 2 0.3 0. 4 0.5 0. 6 0.7 iation on the deformation and microstructure was investi-
True strain gated. Deformation heating and phase volume fraction
are noted to aect the overall plastic ow characteristics
450
of the material. The simulated ow curves were compara-
400 Simulated without deformation heating tively analyzed with experimental results. The following
Flow stress (Mpa)

350 specic conclusions may be drawn:


300
Simulated with deformation
250 heating (1) Intergranular interactions considerably inuence the
dislocation density, leading to non-uniform distribu-
200
tion of dislocations within each grain, especially at
150
the initial stages of deformation.
100 (2) Phase boundaries play a more important role in the
50
Experimental deformation of two-phase alloys than grain bound-
0 aries, and the equivalent stress and the dislocation
0 0.1 0. 2 0.3 0. 4 0.5 0. 6 0.7 density are more inhomogeneous near phase bound-
True strain aries than near grain boundaries.
(3) The onset of shear localisation or shear banding
Fig. 8. The experimental and simulated stressstrain curves at 900 C is inuenced by the strain rate sensitivity parameter,
under dierent strain rates: (a) 0.05 s1 and (b) 0.5 s1.
the deformation heating, the phase volume fraction,
and the friction coecient between the sample
dierent phase volume fractions. For example, both grain a and the die. Shear bands are less likely to form at a
in Fig. 9a and grain b in Fig. 9c have identical initial geom- higher strain rate sensitivity with a lower friction
etry, but because their surrounding grains are of dierent coecient.
R. Ding et al. / Computational Materials Science 40 (2007) 201212 211

Fig. 9. Inuence of phase volume fraction on equivalent stress/equivalent strain at 900 C under a strain rate of 0.05 s1 and an overall strain of 0.5,
respectively for: (a) and (b) 62% a-phase and (c) and (d) 33% a-phase.

(4) The reduction in the ow stress of two-phase alloys [16] S.R. Kalidindi, C.A. Bronkhorst, L. Anand, J. Mech. Phys. Solids 40
comes from two sources: enhanced dynamic recovery (1992) 537.
[17] H. Takahashi, H. Motohashi, M. Tokuda, T. Abe, Int. J. Plast. 10
and phase transformation due to temperature increase (1994) 63.
as a result of deformation heating. [18] R. Becker, S. Panchanadeeswaran, Acta Metall. Mater. 43 (1995) 2701.
[19] S. Quilici, G. Cailletaud, Comput. Mater. Sci. 16 (1999) 383.
[20] H. Aretz, H. Luce, M. Wolske, R. Kopp, M. Goerdeler, V. Marx, G.
Pomana, G. Gottstein, Modell. Simul. Mater. Sci. Eng. 8 (2000) 881.
References [21] F. Delaire, J.L. Raphanel, C. Rey, Acta Mater. 48 (2000) 1075.
[22] D. Raabe, R. Becker, Modell. Simul. Mater. Sci. Eng. 8 (2000) 445.
[1] R. Ding, Z.X. Guo, A. Wilson, Mater. Sci. Eng., A 327 (2002) 233. [23] F.T. Meissonnier, E.P. Busso, N.P. ODowd, Int. J. Plast. 17 (2001) 601.
[2] G.I. Taylor, J. Inst. Met. 62 (1938) 307. [24] D. Raabe, M. Sachtleber, Z. Zhao, F. Roters, S. Zaeerer, Acta
[3] J.F.W. Bishop, R. Hill, Philos. Mag. 42 (1951) 414. Mater. 49 (2001) 3433.
[4] J.D. Eshelby, Proc. R. Soc. London, Ser. A 241 (1957) 376. [25] D. Peirce, R.J. Asaro, A. Needleman, Acta Metall. 30 (1982) 1087.
[5] E. Kroner, Z. Phys. 151 (1958) 504. [26] G.G. Weber, A.M. Lush, A. Zavaliangos, L. Anand, Int. J. Plast. 6
[6] E. Kroner, Acta Metall. 9 (1961) 155. (1990) 701.
[7] T. Leers, Script Metall. 2 (1968) 447. [27] F. Reusch, Y. Estrin, Comput. Mater. Sci. 11 (1998) 294.
[8] J.W. Hutchinson, Proc. R. Soc. London, Ser. A 319 (1970) 247. [28] Y.J. Lee, G. Subhash, G. Ravichandran, Int. J. Plast. 15 (1999) 625.
[9] R. Ding, Z.X. Guo, Acta Mater. 49 (2001) 3163. [29] E. Nakamachi, K. Hiraiwa, H. Morimoto, M. Harimoto, Int. J. Plast.
[10] R. Ding, Z.X. Guo, Comput. Mater. Sci. 23 (2002) 209. 16 (2000) 1419.
[11] D. Peirce, R.J. Asaro, A. Needleman, Acta Metall. 31 (1983) 1951. [30] J.E. Dorn, C.D. Starr, Relations of Properties to Microstructure,
[12] R.J. Asaro, A. Needleman, Acta Metall. 33 (1985) 923. ASM, Metals Par, OH, 1954, p. 71.
[13] S.V. Harren, H.E. Deve, R.J. Asarom, Acta Metall. 36 (1988) 2435. [31] M.H. Poech, H.F. Fischmeister, Acta Metall. Mater. 40 (1992) 487.
[14] S.V. Harren, R.J. Asaro, J. Mech. Phys. Solids 37 (1989) 191. [32] K.S. Ravichandran, Acta Metall. Mater. 42 (1994) 1113.
[15] R. Becker, Acta Metall. Mater. 39 (1991) 1211. [33] M. Cook, E.C. Larke, J. Inst. Met. 71 (1945) 371.
212 R. Ding et al. / Computational Materials Science 40 (2007) 201212

[34] B. Avitzur, in: Metal Forming, Processes and Analysis, McGraw-Hill, [51] U.F. Kocks, H. Mecking, in: P. Haasen, V. Gerold, G. Kostorz
New York, 1968, p. 78. (Eds.), Strength of Metals and Alloys, Pergamon Press, Oxford, 1979.
[35] F. Farhat-Nia, M. Salimi, M.R. Movahhedy, J. Mater. Process. [52] H. Mecking, in: M.F. Ashby, R. Bullough, C.S. Hartley, J.P. Hirth
Technol. 177 (2006) 525. (Eds.), Dislocation Modelling of Physical Systems, Pergamon Press,
[36] B.P.P. Almeida, M.L. Alves, P.A.R. Rosa, A.G. Brito, P.A.F. Oxford, 1981.
Martins, Int. J. Mach. Tools Manuf. 46 (2006) 1643. [53] H. Mecking, U.F. Kocks, Acta Metal. 29 (1981) 1865.
[37] H. Cho, T. Altan, J. Mater. Process. Technol. 170 (2005) 64. [54] G.I. Taylor, H. Quinney, Philos. Trans. R. Soc., Ser. A 230 (1931)
[38] A.J. Schwartz, J.S. Stolken, W.E. King, G.H. Campbell, Mater. Sci. 323.
Eng., A 317 (2001) 77. [55] K.S. Vecchio, J. Phys. IV 4 (1994) 301.
[39] B. Karlsson, G. Linden, Mater. Sci. Eng. 17 (1975) 209. [56] W. Pantleon, D. Francke, P. Klimanek, Comput. Mater. Sci. 7 (1996)
[40] S. Ankem, H. Margolin, Metall. Trans. A 13 (1982) 595. 75.
[41] S. Ankem, H. Margolin, Metall. Trans. A 13 (1982) 603. [57] ABAQUS Standard Users Manual, Hibbitt, Karlsson & Sorensen,
[42] S. Neti, M.N. Vijayshankar, S. Ankem, Mater. Sci. Eng., A 145 Pawtucket, RI, 2001; V6.2.
(1991) 47. [58] D. Peirce, C.F. Shih, A. Needleman, Comput. Struct. 18 (1984) 875.
[43] S. Neti, M.N. Vijayshankar, S. Ankem, Mater. Sci. Eng., A 145 [59] G.E. Dieter, Mechanical Metallurgy, McGraw-Hill, London, 1976, p.
(1991) 55. 33.
[44] R.E. Bolmaro, R.V. Browning, F.M. Guerra, A.D. Rollett, Mater. [60] Z. Fan, A.P. Miodownik, Acta Metall. Mater. 41 (1993) 2403.
Sci. Eng., A 196 (1995) 53. [61] B. Radhakrishnan, G.B. Sarma, T. Zacharia, Acta Mater. 46 (1998)
[45] T. Steinkop, M. Sautter, Comput. Mater. Sci. 4 (1995) 10. 4415.
[46] T. Steinkop, M. Sautter, Comput. Mater. Sci. 4 (1995) 15. [62] G.B. Sarma, B. Radhakrishnan, T. Zacharia, Comput. Mater. Sci. 12
[47] N. Lippmann, T. Steinkop, S. Schmauder, P. Gumbsch, Comput. (1998) 105.
Mater. Sci. 9 (1997) 28. [63] D. Raabe, Comput. Mater. Sci. 19 (2000) 13.
[48] M. Grujicic, N. Sankaran, Int. J. Solids Struct. 34 (1997) 4421. [64] J.J. Jonas, M.J. Luton, Advances in Deformation Processing, Plenum
[49] Y. Estrin, in: A.S. Krausz, K. Krausz (Eds.), Unied Constitutive Press, New York, 1978, p. 215.
Laws of Plastic Deformation, Academic Press, London, 1996. [65] S.L. Semiatin, G.D. Lahoti, Metall. Trans. A12 (1981) 1705.
[50] U.F. Kocks, J. Eng. Mater. Technol. 98 (1976) 76.

Vous aimerez peut-être aussi