Takaki 2013

International Journal of Plasticity xxx (2013) xxxxxx
Contents lists available at ScienceDirect
International Journal of Plasticity

journal homepage: www.elsevier.com/locate/ijplas
Multiscale modeling of hot-working with dynamic

recrystallization by coupling microstructure evolution
and macroscopic mechanical behavior
Tomohiro Takaki a,, Chihiro Yoshimoto a, Akinori Yamanaka b, Yoshihiro Tomita c
a
Graduate School of Science and Technology, Kyoto Institute of Technology, Matsugasaki, Sakyo, Kyoto 606-8585, Japan
b
Mechanical Systems Engineering, Tokyo University of Agriculture and Technology, Koganei, Tokyo 184-8588, Japan
c
Department of Mechanical Engineering, Fukui University of Technology, Gakuen, Fukui 910-8505, Japan
a r t i c l e i n f o a b s t r a c t
Article history: Dynamic recrystallization (DRX) occurs during the hot-working of a metallic material with
Received 4 September 2012 low-to-medium stacking-fault energy. The macroscopic mechanical behavior during hot-
Received in nal revised form 1 July 2013 working is largely affected by the microstructure evolution due to DRX. In this study, a
Available online xxxx
novel multiscale hot-working model was developed by coupling the multi-phase-eld
dynamic recrystallization (MPF-DRX) model and large deformation elasticplastic nite
Keywords: element (FE) method using J2 ow theory to evaluate the microstructure evolution and
A. Microstructures
macroscopic mechanical behavior, respectively. We call this model the multi-phase-eld
A. Thermomechanical processes
B. Polycrystalline material
and nite element dynamic recrystallization (MPFFE-DRX) model. Compression simula-
C. Finite differences tions with nonuniform deformation of a cylinder conrmed that the newly developed
C. Finite elements MPFFE-DRX model can be used to evaluate the macroscopic mechanical behavior during
hot-working by considering the DRX microstructure evolution, which differs depending
on the area. We also conrmed that the MPFFE-DRX model can be used to simulate mac-
roscopic mechanical behavior depending on the initial microstructure by varying the initial
grain size.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
Hot-working is a process in which a metallic material is plastically deformed at an elevated temperature higher than the
recrystallization temperature (Humphreys and Hatherly, 1995). Because hot-working is performed under such high temper-
atures, dynamic recovery (DRV) due to a thermally activated process actively occurs in addition to the dislocation accumu-
lation due to plastic deformation. In particular, for a metallic material with low-to-medium stacking-fault energy, dynamic
recrystallization (DRX) also occurs, in which the recrystallized grains having low dislocation density are nucleated when the
dislocation density exceeds a critical value and grow to the deformed materials driven by the stored energy difference be-
tween the plastically deformed grain and the recrystallized grain. For DRX materials, the macroscopic mechanical behavior
during hot-working is largely affected by the microstructure evolution due to DRX (Roberts and Ahlblom, 1978; Sakai and
Jonas, 1984; Humphreys and Hatherly, 1995). Therefore, predicting and evaluating the macroscopic mechanical behavior
during hot-working with high accuracy requires considering the DRX microstructure evolution.
Corresponding author.
E-mail address: takaki@kit.ac.jp (T. Takaki).
0749-6419/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijplas.2013.09.001
Please cite this article in press as: Takaki, T., et al. Multiscale modeling of hot-working with dynamic recrystallization by coupling micro-
structure evolution and macroscopic mechanical behavior. Int. J. Plasticity (2013), http://dx.doi.org/10.1016/j.ijplas.2013.09.001
2 T. Takaki et al. / International Journal of Plasticity xxx (2013) xxxxxx
Finite element (FE) computation is widely used in numerical simulations of hot-working with DRX. The DRX volume frac-
tion, average diameter of DRX grain, and so on that are used to express the DRX microstructure evolution are incorporated
into the constitutive equations as internal state variables (Gbel, 1991; Busso, 1998; Yanagimoto et al., 1998; Manonukul
and Dunne, 1999; Davenport et al., 2000; Cho et al., 2005; Yeom et al., 2005; Fan and Yang, 2011; Brown and Bammann,
2012; Momeni et al., 2012). In these cases, we must model the changes of such internal state variables depending on the
histories of plastic deformation and temperature. However, models that precisely express the complicated histories of defor-
mation and temperature during actual hot-working are very difcult to develop. In addition, we cannot obtain the micro-
structure itself from these simulations.
Ding and Guo (2001) developed a cellular automaton dynamic recrystallization (CA-DRX) model that allows the average
mechanical behavior of the computational domain to be investigated based on changes in the DRX microstructure. In this
model, the DRX grain growth, changes in dislocation density due to plastic deformation and DRV, and average stress of
the entire computational region are computed by the cellular automaton (CA) method, KocksMecking model (Mecking
and Kocks, 1981), and Bailey and Hirschs equation (1960), respectively. This CA-DRX model is widely applicable and has
already been applied to various materials and problems (Ding and Guo, 2004; Qian and Guo, 2004; Kugler and Turk,
2004; Goetz, 2005; Xiao et al., 2008; Zheng et al., 2008). Takaki et al. (2008, 2009) proposed a multi-phase-eld dynamic
recrystallization (MPF-DRX) model in which the multi-phase-eld (MPF) method (Steinbach and Pezzolla, 1999), which
can be used to accurately simulate the grain growth process, is used instead of the CA method in the CA-DRX model. The
MPF-DRX model can also be used to express transient deformation (Sakai et al., 1983; Tanner and McDowell, 1999; Tanner
et al., 1999; Frommert and Gottstein, 2009), where the strain rate and temperature change during deformation (Takaki et al.,
2011). Although the CA-DRX and MPF-DRX models can express the microstructure evolution during DRX and calculate the
mechanical behavior of the domain based on the computed microstructure, these models are limited to uniform deforma-
tion. Because the deformations in actual hot-working are nonuniform, these models cannot be used to evaluate the macro-
scopic mechanical behavior of hot-working processes such as hot-rolling, extrusion, and drawing.
Recently, multiscale models using the FE method for the macroscopic hot-working process and a model for the DRX grain
growth and phase transformation microstructure evolution have been investigated (Qiang and Esche, 2005; Gawad et al.,
2008; Won and Im, 2010; Svyetlichnyy, 2012). However, in these computations, the FE simulations are performed by using
a constitutive equation that depends on the internal state variables, which are independent of the microstructure evolution.
The grain growth and phase transformation of the microstructure are simulated based on the FE simulation results. There-
fore, these multiscale models are not a complete coupling model.
Many multiscale models have been developed for cold-working, as reviewed by McDowell (2010) and Horstemeyer and
Bammann (2010). For example, Groh et al. (2009) developed three different length scale models that use molecular dynam-
ics, discrete dislocation dynamics, and crystal plasticity; Ohashi et al. (2007) expressed scale-dependent mechanical prop-
erties using discrete dislocation dynamics and crystal plasticity nite elements; and Sundararaghavan and Zabaras (2006)
proposed FE homogenization based on microstructure evolution. However, because grain boundary migration is not active
in the cold-working process, no multiscale models are available that consider grain boundary migration.
In this study, a novel multiscale hot-working model was developed by coupling the MPF-DRX model and a large defor-
mation elasticplastic FE method to evaluate the microstructure evolution and macroscopic mechanical behavior, respec-
tively, during hot-working. In this model, the micro and macro elds interfere with each other, and the model acts as a
complete coupling model as a homogenization method (Terada and Kikuchi, 1997; Terada et al., 1997). We call this model
the multi-phase-eld and nite element dynamic recrystallization (MPFFE-DRX) model. To the best of our knowledge, this is
the rst attempt at using multiscale modeling to completely couple microstructure evolution and macroscopic mechanical
behavior. After deriving the MPFFE-DRX model, we conrmed its validity by simulating compression of cylinders with non-
uniform deformation. We conrmed that the MPFFE-DRX model can be used to simulate macroscopic mechanical behavior
depending on the initial microstructure by varying the initial grain size.
2. MPFFE-DRX model
In our developed MPFFE-DRX model, the microstructure evolution during DRX is evaluated by the MPF-DRX model, and
the macroscopic mechanical behavior during hot-working is calculated by the FE method. The DRX microstructure evolution
is computed by a nite difference (FD) method, and the FD computations are performed for every element used in the mac-
roscopic FE simulation. The concept is very similar to that of the homogenization method (Terada et al., 1997; Terada and
Kikuchi, 1997). The tangent modulus of the equivalent stressstrain curve is transferred from the micro eld to the macro
eld, and the equivalent strain rate and temperature are transferred from the macro eld to the micro eld.
2.1. Micro eld model
In the MPF-DRX model (Takaki et al., 2008, 2009), the polycrystalline structure and growth are expressed by the MPF
model proposed by Steinbach and Pezzolla (1999). In the MPF model, the ith grain out of N grains in the polycrystal is indi-
cated by the phase-eld variable /i . /i takes the value of 1 in the ith grain and 0 in the other grains; it smoothly changes at
the grain boundary between the ith grain and other grains. The time evolution equation of /i is derived as the total free
T. Takaki et al. / International Journal of Plasticity xxx (2013) xxxxxx 3
PN
energy of a system decreasing monotonically with time under the constraint condition i1 /i 1 at a lattice point (Stein-
bach and Pezzolla, 1999); it is expressed as
" #
@/i 2X n Xn
1 2 8 q
M/ij W ik W jk /k aik a2jk r2 /k /i /j Dfij ; 1
@t n j1 k1
2 p
where n is the number of phase-eld variables whose value is larger than 0 at a lattice point; aij is the gradient coefcient
between the ith and jth grains; W ij is the amount of energy varier; and M /ij is the phase-eld mobility. aij ; W ij , and M /ij are
related to the grain boundary thickness d, grain boundary energy c, and grain boundary mobility M as follows (Takaki
and Tomita, 2010):
2 p 4c p2
aij 2dc; W ij ; M /ij M: 2
p d 8d
For the grain boundary mobility M, we use M M 0 =T expQ b =RT, where T is the temperature; M 0 is the pre-exponential
factor; Q b is the activation energy; and R is the gas constant. Dfij in Eq. (1) is the stored energy difference between the ith and
jth grains, or
Dfij Ed qi qj : 3
2
Here, Ed is the elastic strain energy caused by one dislocation, and it is simply expressed as Ed 1=2Gb (Humphreys and
Hatherly, 1995), where G is the shear modulus, b is the value of the Burgers vector, and qi and qj are the dislocation densities
in the ith and the jth grains, respectively. Eq. (1) is discretized by the normal nite difference method, where spatial discret-
ization is realized by the second-order central difference and temporal discretization is realized by the forward difference.
The dislocation density qi in the ith grain, which was assumed to be constant in the grain, is calculated by the KocksMec-
king model as follows (Mecking and Kocks, 1981):
dqi p
k1 qi k2 qi ; 4
de
where e is an equivalent strain; k1 is a constant that represents hardening; and k2 is a function that represents the DRV of the
equivalent strain rate e_ and temperature T. k1 and k2 are determined from the uniaxial stressstrain curve without DRX,
which is a monotonic increasing curve converging to a steady-state stress rst . The steady-state stress rst is modeled by (Ding
and Guo, 2001):
1=A2
rst fA1 e_ expQ a =RTg ; 5
where A1 and A2 are constants and Q a is the activation energy.
For a single-phase material, the DRX grains have been reported to be nucleated by bulging of the grain boundary (Ponge
and Gottstein, 1998; Wusatowska-Sarnek et al., 2002; Miura et al., 2004; Wusatowska-Sarnek, 2005). The critical dislocation
density qc required to nucleate the DRX grain by bulging of the grain boundary is calculated by (Roberts and Ahlblom, 1978)
!1=3
20ce_
qc 2
; 6
3blMEd
p
where l is the average mean free path of the mobile dislocation; it is expressed by l 10=0:5 q0 . The nuclei of the DRX
grain are placed on the lattice point in the grain boundary; its dislocation density becomes greater than the critical value
qc based on the nucleation rate

T0
n_ ce_ d exp : 7
T
where c; d, and T 0 are constants (Ding and Guo, 2001). The nucleation lattice point is determined by a random number.
The initial dislocation density was assumed to be constant in all grains, and the dislocation density in each grain increases
with deformation according to Eq. (4). Therefore, the dislocation densities in all grains increase uniformly until nucleation
starts. At nucleation, the dislocation density in the DRX nucleus returns to the initial value. Then, the dislocation density
evolves differently for each grain. The average stress r
of a computational domain is calculated using the average dislocation
density q of the entire computational region by the BaileyHirsch equation as follows (Bailey and Hirsch, 1960):
p
r aGb q ; 8
where a is the dislocation interaction coefcient and has a value of approximately 0.5.
In the present MPF-DRX model, although the macro-FE domain deforms extensively, the FD computational domain keeps
the initial rectangular shape. This approximation model is widely accepted and very convenient for the present coupling
model from the view point of computational cost. In particular, at a relatively low deformation rate, the approximation is
thought to be reasonable because the DRX grains grow in a round shape. Although it is difcult to express the elongated
grain shape in a high deformation rate, the stressstrain curves and variations of the average grain size can be accurately
calculated even at a relatively high deformation rate.
2.2. Macro eld model
The normal FE formulation for hot-working needs the strain rate and temperature-dependent constitutive equation
(Gbel, 1991; Busso, 1998; Yanagimoto et al., 1998; Manonukul and Dunne, 1999; Davenport et al., 2000; Cho et al.,
2005; Yeom et al., 2005; Momeni et al., 2012; Brown and Bammann, 2012). In the present MPFFE-DRX model, we instead
use a simple elasticplastic constitutive equation that is independent of the strain rate and temperature because the effects
of the strain rate and temperature are incorporated into the MPF-DRX model in the micro eld. We use a constitutive equa-
tion where the elastic and plastic strains are derived from
O
Hooks law and J2 ow theory, respectively. In this case, the rela-
tion between the Jaumann rate of the Kirchhoff stress, Sij , and strain rate e_ ij is indicated by (Kitagawa and Tomita, 1972)

O 2G 0 0
Sij Deijkl rij rkl e_ kl ; 9
g
where Deijkl is the elastic coefcient tensor, r0ij is the deviatoric stress, and g is expressed by the equation

2 2 h
g r 1 : 10
3 2G
In Eq. (10), r
is the Mises equivalent stress dened by r
2 3=2r0ij r0ij , and h is calculated by

1 3 1 1
; 11
h 2 Et E
where E is the Youngs modulus and Et is the tangent modulus of the uniaxial stressstrain curve.
In this study, the temperature T was assumed to be constant. When a temperature change is being considered, the heat
conduction equation that considers the heat generation due to working should be solved to obtain the space distribution and
time change of the temperature (Tomita, 1994).
2.3. Coupling of micro and macro elds
In the coupling computation of the micro and macro elds, the time increment Dt is determined from a numerical stable
condition and solving Eq. (1) explicitly. When solving Eq. (1) in two-dimensional space, the time increment Dt becomes

Dt Dx2 = 4a2 M/ . If the nucleation rate calculated by Eq. (8) is too high, the time increment Dt should be adjusted. The
equivalent strain rate e_ and temperature T (constant in this study) of every element are transferred from the macro eld
to the micro eld. In the micro eld computation, the dislocation density in the grain is calculated by substituting the equiv-
alent strain rate e_ , temperature T, and equivalent strain increment De e_ Dt into Eq. (4). The critical dislocation density qc
of Eq. (6) and rate n_ of Eq. (7) for DRX nucleation also change depending on e_ and T. At the end of the micro eld computation,
the equivalent stress r is calculated by Eq. (8), and the equivalent stress increment Dr is obtained as the difference between
r in the present and previous steps. The tangent modulus Et is then calculated as Et Dr =De and is transferred to the macro
eld. In the macro eld computation, the tangent modulus Et is incorporated into the constitutive equation of Eq. (9) through
Eqs. (11) and (10).
3. Computational conditions for cylinder compression simulation
The validity of the MPFFE-DRX model was conrmed by simulating the compression of a cylinder. Fig. 1 shows the com-
putational models for the micro and macro elds.
The top and bottom surfaces of the cylinder were perfectly bonded onto the rigid body plates, and the cylinder was then
compressed. In consideration of the symmetry, the axisymmetric model shown in Fig. 1(b) was used; its size was
DR DZ = 5 mm. The cylinder was compressed to up to half the thickness at a constant displacement rate of
_ = 0.025 mm/s under the displacement constraint of the z-direction for the bottom and r-direction for the left and top.
u
The macro domain was meshed into 5 5 regular elements with a crossed-triangle element (Nagtegaal et al., 1974).
Fig. 1(c) shows the micro eld computational model. To evaluate the effects of the initial grain size dini , FD lattices of
192 166, 192 166, 190 164, and 286 247 were used for dini = 12.5, 25.0, 50.0, and 75.0 lm, respectively. Because
we used a regular constant lattice size of Dx Dy = 0.5 lm, the computational domains were different for every dini model.
The initial grain was set to a regular hexagon, and 64, 16, 4, and 4 grains were prepared for dini = 12.5, 25.0, 50.0, and 75.0 lm
models, respectively, under the periodic boundary conditions for Eq. (1). Here, Fig. 1(c) corresponds to the model of dini = 50.0
and 75.0 lm. These macro computational models were prepared for all 100 triangular elements of the macro computational
model shown in Fig. 1(b).
The following material parameters for copper (Takaki et al., 2008, 2009) were used in the present simulations: c 0:2 J/
m2, d 5Dx, M 0 0:139 m4K/Js, Q b 110 kJ/mol, R = 8.314 J/mol, k1 4:00 108 /m, k2 aGbk1 =rst , G 42:1 GPa,
Fig. 1. Micro and macro computational models for simulation of compression of cylinder.
b 2:56 1010 m, A1 2:0 1044 , A2 7:6,

Qa
13 2 7
275 kJ/mol, q0 = 1:7 10 /m , c = 1.125 10 , d 1, and T 0 = 2400 K.
The time increment was set to Dt Dx2 = 8a2 M / , where a and M/ are the diagonal components of aij and M /ij , respectively.
1
For DRX nucleation in the micro eld, one DRX nucleus was placed every n_ Dtngb Dx2 =d steps, where ngb is the number of
lattice points satisfying 0 < /i < 1 and qi P qc . The initial dislocation density was set to qini 109 /m2. The Youngs modulus
E needed in the macro computation was calculated from Eqs. (4) and (8) in the rst step. The temperature was set to be con-
stant at T = 800 K. The following Sections of 4.1 and 4.2 present the numerical results for the initial grain diameter
dini = 50.0 lm. Although the grain boundary thickness, given by d 5Dx = 2.5 lm, is signicantly larger than an actual grain
boundary thickness, it is smaller than the average grain size shown in Section 4; its accuracy has already been conrmed in
our previous study (Takaki et al., 2008). However, in the computation with smaller grain, we should use smaller grain bound-
ary thickness.
4. Numerical results
4.1. Uniform compression simulation
Before the nonuniform compression simulation using the model shown in Fig. 1, a uniform compression simulation with-
out the displacement constraint in the r-direction of the top surface in Fig. 1(b) was performed to check that the MPFFE-DRX
model can simulate the uniaxial condition properly.
Fig. 2 shows the changes in the true stress and average grain diameter dav e against the true strain. Although the stress and
strain in the compression direction had negative values, the compression direction of the stress and strain is indicated as a
positive value in Fig. 2 and in the following gures. The stressstrain curve of the black line indicates the macroscopic
mechanical behavior calculated by the MPFFE-DRX model, where the true stress and true strain are calculated by
F=A0 1 u =DZ and ln1 u
=DZ, respectively; F; A0 , and u
are the force in the z-direction applied to the macro model, dis-
placement of the top surface in the z-direction, and initial cross-sectional area of the cylinder, respectively, as shown in
Fig. 1(a)(c). The light-blue lines are the r e curves for the hundred triangular elements that is, all of the micro elds.
The stress uctuations of 5 MPa can be observed by the light-blue lines because the nucleation site of the DRX grain dif-
fered in every microscopic computational domain. The macroscopic r e curve took the middle value of the hundred micro-
scopic r e curves. The deformation of the macro computational domain showed a completely uniaxial compression, as
shown in the lower right-hand-side gure in Fig. 3. The red and blue lines indicate the MPF-DRX computation results when
Fig. 2. Changes in true stress and average grain diameter dav e against the true strain in simulation of uniform compression of cylinder (dini = 50.0 lm).
=DZ (dini = 50.0 lm).

Fig. 3. Relation between average macroscopic stress F=A0 and strain u
Fig. 4. Deformation of macroscopic computational model and variations of (a) equivalent stress r , (b) equivalent strain rate e_ , and (c) average grain
=DZ = 0.1, 0.2, 0.3, 0.4, and 0.5 from top to bottom (dini = 50.0 lm).
diameter dav e distributions at u
only the micro eld was used, as shown in Fig. 1(c). The red1 and blue lines indicate the results for true strain rates of e_ = 0.005/
s and 0.01/s, respectively. For uniaxial compression under a constant true strain rate, the stress and average grain diameter are
known to converge to a value with increasing deformation (Blaz et al., 1983; Prasad and Rao, 2005). In the present MPF-DRX
computations, the true stress and average grain diameter converged to 35 MPa and 15 lm for e_ = 0.005/s and 40 MPa and
11 lm for e_ = 0.01/s, respectively. Because the compression displacement rate was constant in the MPFFE-DRX simulation,
the true strain rate e_ changed from 0.005/s to 0.01/s, whereas u =DZ changed from 0 to 0.5. As shown in Fig. 2, the black
line overlaps the red line at the beginning of deformation, begins to diverge from the red line with deformation, and then over-
laps the blue line at the end of deformation. Here, the variation in dav e for the MPFFE-DRX computation was the result for a
triangular element. Although experiments under a constant displacement rate have not been reported so far, the MPFFE-DRX
simulation results are plausible considering the experimental results for temporal deformation, where the true strain rate sud-
denly changes during deformation (Sakai et al., 1983; Frommert and Gottstein, 2009; Takaki et al., 2011).
4.2. Nonuniform compression simulation
=DZ as obtained by the MPFFE-DRX

Fig. 3 shows the relations between the average macroscopic stress F=A0 and strain u
simulations. The solid line indicates the nonuniform compression simulation performed under the conditions shown in
Fig. 1, and the dashed line indicates the uniform compression simulation shown in Section 4.1. The right-hand-side gure
in Fig. 3 shows the deformation sequences for every 0.1 u =DZ. Because the deformation stiffness for the nonuniform
1
For interpretation of color in Figs. 1, 2, 4, 5 and 10, the reader is referred to the web version of this article.
deformation was higher than that for the uniform deformation, the average stress F=A0 for the nonuniform deformation was
higher in all strain regions than that for the uniform deformation.
Fig. 4 shows the distribution variations of the (a) equivalent stress r , (b) equivalent strain rate e_ , and (c) average grain
diameter dav e together with the mesh deformation. As shown in Fig. 4(a) and (b), the color changed from blue to red with
increasing r and e_ ; as shown in Fig. 4(c), the color changed from blue to red with decreasing dav e . As shown in Fig. 4(a),
=DZ = 0.1 the stresses in all regions except for around the center of the top end showed high values. When the deforma-
at u
tion increased up to u =DZ = 0.2, the stress values decreased overall, and the stress values along the diagonal line from the left
lower side to the right upper side were relatively high. With further deformation, the stresses increased again in the inner
part of the cylinder. Around the outer perimeter, there were no large changes after u =DZ = 0.2. In the vicinity of the top end,
increasing deformation produced successively increasing stress from the outer region to the inner region. For e_ and dav e
shown in Fig. 4(b) and (c), variations similar to r were observed. Changes in the microstructure size distribution shown
in Fig. 4(c), or the ne grains, were observed in the inner part of the cylinder; coarse grains were observed in the outer perim-
eter and top end. These observations qualitatively agree with the experiment (Qu et al., 2005) and FE computation results
(Wang et al., 2011).
To investigate these variations in greater detail, the variations of the (a) equivalent stress r , (b) equivalent strain e, (d)
equivalent strain rate e_ , and (e) average grain diameter dav e in every triangular element for the macroscopic average strain
=DZ were examined, as shown in Fig. 5. As shown in Fig. 5(c), the abscissa was e. As shown in Fig. 4, the variations of r
u ; e_ ,
and dav e were roughly classied into variations in three regions: inner part, outer perimeter, and top end of the cylinder. The
triangular elements were colored as shown in Fig. 5(f); the values of the colored elements are shown in Fig. 5(a)(e) using the
corresponding color. The red elements for the inner part, green elements for the outer perimeter, and blue elements for the
top end are indicated by the G1, G2, and G3 element groups, respectively. The white elements were located at the boundary
of the colored elements; these data are not shown in Fig. 5. Fig. 6 shows the variations of (a) r , (b) e_ , and (c) dav e for the ele-
ments g1, g2, and g3 indicated in Fig. 5(f), where the three elements represent the three element groups G1, G2, and G3,
respectively. As shown in Fig. 6(a), the abscissa was e, similar to Fig. 5(c). The dots in Fig. 6 correspond to u =DZ = 0.1,
0.15, 0.2, 0.3, 0.4, and 0.5, and the microstructures are indicated in Fig. 7 by a DRX number and grain boundary. The DRX
numbers for the initial grains, DRX grains nucleated from the initial grains, DRX grains nucleated from the rst DRX grains,
and DRX grains nucleated from the second DRX grains were 0, 1, 2, and 3, respectively (Takaki et al., 2009). The grain bound-
ary was dened as the region satisfying R/2i 6 0:55.
Immediately after deformation began, as shown in Fig. 5(d), there was a variance of the equivalent strain rates e_ of every
element because of the nonuniform deformation, although the macroscopic average strain rate was a constant u _ =DZ =
0.005/s. As the deformation increased, e_ increased in the G1 group, decreased in the G3 group, and remained almost con-
stant in the G2 group. Because the strain rates e_ in the G1 group were high, the amount of equivalent strain e was also
higher than that in the other groups, as shown in Fig. 5(b). Hence, the dislocation densities in the G1 group quickly
achieved the critical value to nucleate the DRX grains, as indicated by Eq. (6). Because the number of DRX grains increased,
the DRX microstructure quickly reduced the size dav e , as shown in Fig. 5(e). Because the steady state stress rst calculated by
Eq. (5) depends on the strain rate, the peak stresses shown in Fig. 5(a)(c) were the highest for the G1 group and lowest for
the G3 group. From the microstructure at u =DZ = 0.1 shown in Fig. 7, some DRX nuclei were observed at the grain boundary
for g1 and g2, but none were observed for g3. At u =DZ = 0.1, r
of g1 already exceeded the peak value, as shown in Fig. 6(a);
, (b) equivalent strain e, (d) equivalent strain rate e_ , and (e) average grain diameter dav e for equivalent strain
Fig. 5. Variations of (a) equivalent stress r e and
(c) r
-e curves for colored elements shown in (f) (dini = 50.0 lm).
, (b) equivalent strain rate e_ , and (c) average grain diameter dav e for elements g1, g2, and g3 shown in Fig. 5(f)
Fig. 6. Variations of (a) equivalent stress r
(dini = 50.0 lm).
Fig. 7. Microstructure evolution for elements (a) g1, (b) g2, and (c) g3 shown in Fig. 5(f) (dini = 50.0 lm).
the softening was caused by growth of the DRX grain. g2 at u =DZ = 0.1 was in the balance condition, as shown in Fig. 6(a),
of the hardening of the initial grains and softening due to DRX grain growth. As shown in Fig. 5(d), after around
=DZ = 0.05, the strain rate e_ of the G1 group accelerated; on the other hand, that of the G3 group decreased further. This
u
is because the deformation was concentrated in the G1 group because of softening by the DRX grain growth; deformation
of the G3 group was difcult to occur. When the deformation increased to around u =DZ = 0.1, the rate of increase of the
strain rate e_ of the G1 group decreased because the changes in the tangent modulus of the r -e curve became small after
exceeding u =DZ = 0.1, as shown in Figs. 5(c) and 6(a). On the other hand, the strain rate e_ of the G3 group began to increase
instead of decreasing. Once it exceeded the deformation u =DZ = 0.15, e_ of the G1 group increased slowly and almost line-
arly, similar to the uniform compression shown in Fig. 2. At u =DZ = 0.15, as shown in Fig. 7(a), the hardening of the rst
DRX grains and softening of the second DRX grains were in equilibrium. Depending on the change, the equivalent stress
and average grain diameter after u =DZ = 0.15 for the G1 group gradually increased and decreased, respectively, as shown
in Fig. 5(a)(e). As shown in Fig. 5(e) and 6(c), the microstructure change shown in Fig. 7 caused the average grain size to
gradually decrease for the G1 group. In the G3 group, after around u =DZ = 0.15, the strain rate e_ increased. The dislocation
density in the G3 group successfully achieved the critical value qc from the outer region to the inner region; therefore, the
average grain size dav e changed as shown in Fig. 5(e). Because the strain rate e_ of element g3 continuously increased from
=DZ = 0.15 to 0.5, as shown in Fig. 6(b), the microstructure also gradually decreased in size, as shown in Figs. 6(c) and 7(c).
u
As a result, the stress r increased with increasing strain e, as shown in Figs. 5(c) and 6(a). For the G2 group, the equivalent
strain rate e_ was almost constant through deformation, and the equivalent stress r and average grain diameter dav e became
almost constant after around u =DZ = 0.15 because the G2 group, which was located in the outer perimeter of the cylinder,
could deform relatively freely.
The results shown in Figs. 47 conrmed that the developed MPFFE-DRX model can be used to evaluate the macroscopic
mechanical behavior depending on the microstructure evolution and the microstructure evolution depending on the mac-
roscopic nonuniform deformation. To conrm the validity of the developed MPFFE-DRX model, we compared the numerical
results obtained in this study with the experimental ndings of other studies. Qu et al. (2005) and Lee and Im (2010) have
performed compression tests on 26Cr2Ni4MoV and pure copper cylinders in DRX conditions, respectively. They measured
Fig. 8. Macroscopic mechanical behavior depending on initial grain size.
, (b) equivalent strain rate e_ , and (c) average grain diameter dav e for element g1 shown in Fig. 5(f).
Fig. 9. Variations of (a) equivalent stress r
the DRX grain-size distributions from observations of cross sections cut in parallel to the compression axis, of the quenched,
barrel-shaped specimens (Fig. 4). From these observations, it has been concluded that the grain size increases in three direc-
tions: (i) from the bottom to the top along the central axis of the cylinder; (ii) from the center to the outside along the center
of thickness direction; and (iii) from the outside to the center along the top surface. Although the used materials, as well as
the compression and friction conditions are not identical in these experiments and our simulations, a qualitative agreement
in the DRX grain-size distribution between the developed MPFFE-DRX model and the experiments can be conrmed.
4.3. Effects of initial grain size
We performed the MPFFE-DRX simulations by changing the initial grain size to dini = 12.5, 25.0, 50.0, and 75.0 lm for the
model shown in Fig. 1. The results for dini = 50.0 lm were identical to those of the simulation presented in Section 4.2. Fig. 8
shows the relation between the macroscopic average stress F=A0 and strain u =Dz. All curves had local maximum and min-
imum values. Although the changes in these curves were similar, the stresses in the region of 0:07 < u =Dz < 0:2 differed
depending on the initial grain size; the stress values were smaller for small initial grain sizes.
Fig. 9 shows the variations of the (a) equivalent stress r , (b) equivalent strain rate e_ , and (d) average grain size dav e of
element g1 shown in Fig. 1. The abscissa was the equivalent stress e for Fig. 9(a) and the macroscopic average strain u =Dz
for Fig. 9(b) and (c). The dots in Fig. 9 correspond to u =Dz = 0.05, 0.10, 0.15, and 0.20, and the microstructure evolutions
are indicated in Fig. 10 by the DRX number.
Because the nucleation rate of the DRX grain increased for a small initial grain size owing to the high area fraction of the
grain boundary, the area fraction of the recrystallized grains became high, as shown by u =Dz = 0.05 and 0.10 in Fig. 10. Be-
cause softening occurred during an early stage when the recrystallization rate was high, the peak stress became small, as
shown in Fig. 9(a). These tendencies were also observed for all elements, although Figs. 9 and 10 show the results for only
element g1. Thus, the macroscopic mechanical behavior changed depending on the initial grain size, as shown in Fig. 8. The
=Dz = 0.1 shown in Fig. 10(a) showed that the nucleation of the second DRX grain (green
microstructures of dini = 12.5 at u
grain) occurred after the computational region was lled with the rst DRX grains (light-blue grain). Thus, hardening and
softening occurred alternately; consequently, the r e curve showed multiple peaks, as shown in Fig. 9(a). As shown by
=Dz = 0.1 and 0.15 in Fig. 10(d), the nucleation of the second DRX grains occurred before
the microstructure of dini = 75.0 at u
the entire computational region was lled out by the rst DRX grains when the initial grain size was large. In this case, hard-
ening and softening occurred simultaneously; as a result, the single peak r e curve occurred as shown in Fig. 9(a). The mac-
roscopic curve in Fig. 8 showed a smooth variation for the larger initial grain. Considering the case of the small grain size,
because hardening and softening never occurred within the same complete cycle, the microstructure of dini = 12.5 lm at
=Dz = 0.2 in Fig. 10(a) became similar to that for a large grain size. Therefore, the uctuation of the curve for
u
dini = 12.5 lm in Fig. 9(a) decreased with the deformation. Finally, r ; e_ , and dav e for all initial grain sizes converged to the
Fig. 10. Microstructure evolution of element g1 for four different initial grain sizes dini =(a) 12.5, (b) 25.0, (c) 50.0, and (d) 75.0 lm.
same value independent of the initial grain size, as shown in Fig. 9. Therefore, the macroscopic stressstrain curves with dif-
=Dz = 0.2.
ferent initial grain sizes in Fig. 9(a) overlapped with each other after u
5. Conclusions
In this study, a novel multiscale model for hot-working with DRX (i.e., MPFFE-DRX model) was developed by coupling the
MPF-DRX model and a large deformation elasticplastic FE method using J2 ow theory for DRX microstructure evolution
and macroscopic mechanical behavior, respectively. Simulations of compression with nonuniform deformation of the cylin-
der conrmed that the developed MPFFE-DRX model can evaluate the macroscopic mechanical behavior during hot-working
while considering the DRX microstructure evolution, which differs depending on the area. We also conrmed that the
MPFFE-DRX model can be used to simulate the macroscopic mechanical behavior depending on the initial microstructure
by varying the initial grain size. By comparing the numerical results with the experimental ndings performed on different
materials at different compression conditions, we could conrm a qualitative agreement in the DRX grain-size distributions.
Therefore, we believe that the developed MPFFE-DRX model is applicable to the computation of realistic material properties.
On the other hand, we need further validation to achieve a model with higher accuracy.
The newly developed MPFFE-DRX model affords the following advantages: (1) it does not require a complicated consti-
tutive equation that depends on the inter-state variables used in macroscopic FE computation, (2) deformation and temper-
ature histories can be expressed through DRX microstructure changes, and (3) microstructures that differ depending on the
space can be obtained. Similar to the homogenization method, a disadvantage is the computational cost: the micro eld
needs to be solved for every macro FE element. Although the computation in this study was limited to simulation of the com-
pression of a cylinder to check the developed model, future work will involve improving the efciency of the numerical com-
putation in order to achieve more realistic simulations of hot-working.
Acknowledgments
This work was supported by MEXT KAKENHI Grant No. 24109504 and JSPS KAKENHI Grant No. 25630011.
References
Bailey, J.E., Hirsch, P.B., 1960. The dislocation distribution, ow stress, and stored energy in cold-worked polycrystalline silver. Philos. Mag. 5 (53), 485497.
Blaz, L., Sakai, T., Jonas, J.J., 1983. Effect of initial grain size on dynamic recrystallization of copper. Met. Sci. 17, 609616.
Brown, A.A., Bammann, D.J., 2012. Validation of a model for static and dynamic recrystallization in metals. Int. J. Plast. 3233, 1735.
Busso, E.P., 1998. A continuum theory for dynamic recrystallization with microstructure-related length scales original research article. Int. J. Plast. 14 (45),
319353.
Cho, J.R., Jeong, H.S., Cha, D.J., Bae, W.B., Lee, J.W., 2005. Prediction of microstructural evolution and recrystallization behaviors of a hot working die steel by
FEM. J. Mater. Process. Technol. 160 (1), 18.
Davenport, S.B., Silk, N.J., Sparks, C.N., Sellars, C.M., 2000. Development of constitutive equations for modelling of hot rolling. Mater. Sci. Technol. 16 (5),
539546.
Ding, R., Guo, Z.X., 2001. Coupled quantitative simulation of microstructural evolution and plastic ow during dynamic recrystallization. Acta Mater. 49,
31633175.
Ding, R., Guo, Z.X., 2004. Microstructural evolution of a Ti6Al4V alloy during-phase processing: experimental and simulative investigations. Mater. Sci.
Eng. A365 (12), 172179.
Fan, X.G., Yang, H., 2011. Internal-state-variable based self-consistent constitutive modeling for hot working of two-phase titanium alloys coupling
microstructure evolution. Int. J. Plast. 27, 18331852.
Frommert, M., Gottstein, G., 2009. Mechanical behavior and microstructure evolution during steady-state dynamic recrystallization in the austenitic steel
800H. Mater. Sci. Eng. A506 (12), 101110.
Gawad, J., Madej, W., Kuziak, R., Pietrzyk, M., 2008. Multiscale model of dynamic recrystallization in hot rolling. Int. J. Mater. Form. 1 (Suppl. 1), 6972.
Goetz, R.L., 2005. Particle stimulated nucleation during dynamic recrystallization using a cellular automata model. Scr. Mater. 52 (9), 851856.
Groh, S., Marin, E.B., Horstemeyer, M.F., Zbib, H.M., 2009. Multiscale modeling of the plasticity in an aluminum single crystal. Int. J. Plast. 25,
14561473.
Gbel, I.R., 1991. A stochastic recrystallization model: the description of recrystallization by a Markov process original research article. Int. J. Plast. 7 (3),
161198.
Horstemeyer, M.F., Bammann, D.J., 2010. Historical review of internal state variable theory for inelasticity. Int. J. Plast. 26, 13101334.
Humphreys, F.J., Hatherly, M., 1995. Recrystallization and Related Annealing Phenomena. Pergamon Press, Oxford.
Kitagawa, H., Tomita, Y., 1972. Note on incremental stressstrain relations of elasto-plastic materials referred to a convected coordinate systems. J. Appl.
Math. Mech. 52 (3), 183186.
Kugler, G., Turk, R., 2004. Modeling the dynamic recrystallization under multi-stage hot deformation. Acta Mater. 52 (15), 46594668.
Lee, H.W., Im, Y.-T., 2010. Numerical modeling of dynamic recrystallization during nonisothermal hot compression by cellular automata and nite element
analysis. Int. J. Mech. Sci. 52, 12771289.
McDowell, D.L., 2010. A perspective on trends in multiscale plasticity. Int. J. Plast. 26, 12801309.
Manonukul, A., Dunne, F.P.E., 1999. A model for the initiation of dynamic recrystallization in two-phase materials. Philos. Mag. 79, 113132.
Mecking, H., Kocks, U.F., 1981. Kinetics of ow and strain-hardening. Acta Metall. 29 (11), 18651875.
Miura, H., Sakai, T., Mogawa, R., Gottstein, G., 2004. Nucleation of dynamic recrystallization at grain boundaries in copper bicrystals. Scr. Mater. 51 (7), 671
675.
Momeni, A., Abbasi, S.M., Badri, H., 2012. Hot deformation behavior and constitutive modeling of VCN200 low alloy steel. Appl. Math. Model. 36 (11), 5624
5632.
Nagtegaal, J.C., Parks, D.M., Rice, J.R., 1974. On numerically accurate nite element solutions in the fully plastic range. Comput. Methods Appl. Mech. Eng. 4
(2), 153177.
Ohashi, T., Kawamukai, M., Zbib, H.M., 2007. A multiscale approach for modeling scale-dependent yield stress in polycrystalline metals. Int. J. Plast. 23, 897
914.
Ponge, D., Gottstein, G., 1998. Necklace formation during dynamic recrystallization: mechanisms and impact on ow behavior. Acta Mater. 46 (1),
6980.
Prasad, Y.V.R.K., Rao, K.P., 2005. Kinetics and dynamics of hot deformation of OFHC copper in extended temperature and strain rate ranges. Z. Metallkd. 96
(1), 7177.
Qian, M., Guo, Z.X., 2004. Cellular Automata Simulation of Microstructural Evolution during Dynamic Recrystallization of an HY-100 Steel. Mater. Sci. Eng.
A365 (12), 180185.
Qiang, Y., Esche, S.K., 2005. A mutlti-scale approach for microstructure prediction in thermo-mechanical processing of metals. J. Mater. Process. Technol.
169, 493502.
Qu, J., Jin, Q.L., Xu, B.Y., 2005. Parameter identication for improved viscoplastic model considering dynamic recrystallization. Int. J. Plast. 21 (7), 12671302.
Roberts, W., Ahlblom, B., 1978. A nucleation criterion for dynamic recrystallization during hot working. Acta Metall. 26, 801813.
Sakai, T., Akben, M.G., Jonas, J.J., 1983. Dynamic recrystallization during the transient deformation of a vanadium microalloyed steel. Acta Metall. 31, 631
641.
Sakai, T., Jonas, J.J., 1984. Overview no. 35 dynamic recrystallization: mechanical and microstructural considerations. Acta Metall. 32, 189209.
Steinbach, I., Pezzolla, F., 1999. A generalized eld method for multiphase transformations using interface elds. Physica D 134 (4), 385393.
Sundararaghavan, V., Zabaras, N., 2006. Design of microstructure-sensitive properties in elasto-viscoplastic polycrystals using multi-scale homogenization.
Int. J. Plast. 22, 17991824.
Svyetlichnyy, D.S., 2012. Simulation of microstructure evolution during shape rolling with the use of frontal cellular automata. ISIJ Int. 52 (4), 559568.
Takaki, T., Hirouchi, T., Hisakuni, Y., Yamanaka, A., Tomita, Y., 2008. Multi-phase-eld model to simulate microstructure evolutions during dynamic
recrystallization. Mater. Trans. 49 (11), 25592565.
Takaki, T., Hirouchi, T., Hisakuni, Y., Yamanaka, A., Tomita, Y., 2009. Multi-phase-eld simulations for dynamic recrystallization. Comput. Mater. Sci. 45 (4),
881888.
Takaki, T., Tomita, Y., 2010. Static recrystallization simulations starting from predicted deformation microstructure by coupling multi-phase-eld method
and nite element method based on crystal plasticity. Int. J. Mech. Sci. 52, 320328.
Takaki, T., Yamanaka, A., Tomita, Y., 2011. Multi-phase-eld simulations of dynamic recrystallization during transient deformation. ISIJ Int. 51 (10), 1717
1723.
Tanner, A.B., McDowell, D.L., 1999. Deformation, temperature and strain rate sequence experiments on OFHC Cu. Int. J. Plast. 15 (4), 375399.
Tanner, A.B., McGinty, R.D., McDowell, D.L., 1999. Modeling temperature and strain rate history effects in OFHC Cu. Int. J. Plast. 15 (6), 575603.
Terada, K., Miura, T., Kikuchi, N., 1997. Digital image-based modeling applied to the homogenization analysis of composite materials. Comput. Mech. 20,
331346.
Terada, K., Kikuchi, N., 1997. A class of general algorithms for multi-scale analyses of heterogeneous media. Comput. Methods Appl. Mech. Eng. 190 (4041),
54275464.
Tomita, Y., 1994. Simulation of plastic instabilities in solid mechanics. Appl. Mech. Rev. 47 (6), 171205.
Wang, K.L., Fu, M.W., Lu, S.Q., Li, X., 2011. Study of the dynamic recrystallization of Ti6.5Al3.5Mo1.5Zr0.3Si alloy in b-forging process via nite element
method modeling and microstructure characterization. Mater. Des. 32 (3), 12831291.
Won, H.L., Im, Y.-T., 2010. Numerical modeling of dynamic recrystallization during nonisothermal hot compression by cellular automata and nite element
analysis. Int. J. Mech. Sci. 52, 12771289.
Wusatowska-Sarnek, A.M., Miura, H., Sakai, T., 2002. Nucleation and microtexture development under dynamic recrystallization of copper. Mater. Sci. Eng.
A323 (12), 177186.
Wusatowska-Sarnek, A.M., 2005. The new grain formation during warm and hot deformation of copper. J. Eng. Mater. Technol. 127 (3), 295301.
Xiao, N., Zheng, N., Li, D., Li, Y., 2008. A simulation of dynamic recrystallization by coupling a cellular automaton method with a topology deformation
technique. Comput. Mater. Sci. 41 (3), 366374.
Yanagimoto, J., Karhausen, K., Brand, A.J., Kopp, R., 1998. Incremental formulation for the prediction of ow stress and microstructural change in hot
forming. J. Manuf. Sci. Eng. 120 (2), 316322.
Yeom, J.T., Lee, C.S., Kim, J.H., Park, N.-K., 2005. Finite-element analysis of microstructure evolution in the cogging of an Alloy 718 ingot. Mater. Sci. Eng.
A449451, 722726.
Zheng, C., Xiao, N., Li, D., Li, Y., 2008. Microstructure prediction of the austenite recrystallization during multi-pass steel strip hot rolling: a cellular
automaton modeling. Comput. Mater. Sci. 44 (2), 507514.

Takaki 2013

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Takaki 2013

Uploaded by

Copyright:

Available Formats

International Journal of Plasticity xxx (2013) xxxxxx

Contents lists available at ScienceDirect

International Journal of Plasticity

Multiscale modeling of hot-working with dynamic

2.1. Micro eld model

2.2. Macro eld model

2.3. Coupling of micro and macro elds

3. Computational conditions for cylinder compression simulation

b 2:56 1010 m, A1 2:0 1044 , A2 7:6,

4.1. Uniform compression simulation

=DZ (dini = 50.0 lm).

4.2. Nonuniform compression simulation

=DZ as obtained by the MPFFE-DRX

Fig. 8. Macroscopic mechanical behavior depending on initial grain size.

4.3. Effects of initial grain size

You might also like