1 s2.0 S0749641913002180 Main

International Journal of Plasticity 58 (2014) 154183
Contents lists available at ScienceDirect
International Journal of Plasticity

journal homepage: www.elsevier.com/locate/ijplas
Thermo-mechanical-metallurgical modeling for hot-press

forming in consideration of the prior austenite
deformation effect
Hyun-Ho Bok a, JongWon Choi a,b, Frdric Barlat a, Dong Woo Suh a, Myoung-Gyu Lee a,
a
Graduate Institute of Ferrous Technology, Pohang University of Science and Technology (POSTECH), San 31, Hyoja-dong, Nam-gu, Pohang-si,
Gyeongsangbuk-do 790-784, South Korea
b
POSCO, 699 Geumho-dong, Gwangyang-si, Jeollanam-do 545-875, South Korea
a r t i c l e i n f o a b s t r a c t
Article history: In this study, a prior austenite grain renement model was incorporated into semi-empir-
Received 21 June 2013 ical diffusive transformation kinetics for application to hot-press forming. In particular, the
Received in nal revised form 16 September kinetics equations were modied to include the effects of boron addition and austenite
2013
deformation on transformation behaviors during forming. To simulate the hot-press form-
Available online 17 December 2013
ing process, a thermo-mechanical-metallurgical model was formulated implicitly and
implemented into the nite element program ABAQUS using the user subroutines UMAT
Keywords:
and UMATHT. This nonconventional nite element modeling is appropriate to consider
Hot-press forming
Austenite grain renement
thermal- and transformation-associated strains. The proposed model was validated
Thermo-mechanical-metallurgical through simple nite element simulation examples, i.e., dilatometry simulation with and
simulation without external loading, and hot torsion and quenching of a rod. Finally, the hot-press
A. Phase transformation forming of a U-channel-type part was simulated to study the effect of austenite deforma-
tion on the phase kinetics, hardness and residual stress. The simulation results showed that
the austenite deformation had considerable inuence on the nal strength and residual
stress distribution in the hot-press formed sheet, which resulted from an increase in ferritic
phases due to the modied kinetics. In particular, the austenite deformation effect was
more noticeable in the side-wall region of the U-channel where plastic deformation was
the most severe.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
The hot-press forming1 (HPF) process has drawn considerable attention in the automotive industry because parts produced
by this technology can meet the requirements for passenger safety and fuel efciency. The HPF process consists of high-tem-
perature forming of a boron steel sheet at around 800 C and subsequent quenching inside the tools (die quenching). For die
quenching, specialized tools containing cooling channels are used to rapidly lower the temperature of the formed part. An al-
most fully martensitic microstructure can be obtained in the formed part, which results in a strength of 1500 MPa with 22MnB5
steels. In principle, the process is equivalent to conventional heat treatment, but HPF uses direct metal-to-metal contact be-
tween the hot sheet and the surrounding tools. Therefore, a wide range of part strengths can be obtained by controlling the
hardenability of the steel sheet and the cooling capability of the tools.
Corresponding author. Tel.: +82 54 279 9034; fax: +82 54 279 9299.
E-mail address: mglee@postech.ac.kr (M.-G. Lee).
1
Hot-press forming is referred to by various terms, e.g., hot stamping, press hardening, and press quenching. However, a distinction should be made
with warm forming or forming at elevated temperature because hot-press forming (or equivalent terms) is associated with a phase transformation.
0749-6419/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijplas.2013.12.002
H.-H. Bok et al. / International Journal of Plasticity 58 (2014) 154183 155
In order to investigate material behavior of steels subjected to solid-to-solid phase transformation in transient thermal
led, various thermo-mechanical-metallurgical constitutive models have been proposed and implemented (Kim et al.,
2005; Andrade-Campos, 2010; Moumni et al., 2011; Fischlschweiger et al., 2012; Mahnken et al., 2012). Similarly, a number
of nite element (FE) simulations of HPF processes considering phase transformation have been carried out to optimize the
quality of the HPF product, such as the part strength, shape xability and formability (kerstrm and Oldenburg, 2006;
Turetta et al., 2006; Hoffmann et al., 2007; Tekkaya et al., 2007; Lee et al., 2009; Bok et al., 2010, 2011; Choi et al., 2012,
2013; Kim et al., 2013; Park et al., 2013).
In general, microstructural defects introduced by the plastic strain enhances the diffusive transformation kinetics, while
displacive martensitic kinetics is suppressed by strain hardening of the deformed austenite, i.e., mechanical austenite stabil-
ization (Tamura et al., 1988; Bhadeshia and Honeycombe, 2006). This phenomenon has frequently been observed in modern
steels during hot rolling (Umemoto et al., 1992; Senuma et al., 1992). For instance, slabs in a hot-rolling mill are subjected to
a series of severe plastic deformations in the austenite state. Prior austenite grains can be rened from grain diameters of
102 lm to 10 lm. The drastic grain renement is mainly attributed to dynamic/static austenite recrystallization above
the austenite recrystallization stop temperature Tnr. As the thickness of a slab is reduced during hot rolling, the temperature
rapidly decreases below Tnr; however, renement can also take place even without recrystallization. In this case, austenite
grains are directly rened through grain elongation or the formation of deformation bands, which provide additional nucle-
ation sites of diffusive trnasformation (Kern et al., 1992; Zrnik et al., 2003). Therefore, the rened austenite enhances phase
transformations and determines the grain size of the nal ferritic phases (Sun et al., 2002; Maalekian et al., 2007; Kruglova
et al., 2007; Eghbali, 2010).
As for boron-added steels, many observations have also been reported on this acceleration of diffusive kinetics due to
plastic strain in the prior austenite. For example, a drastic loss in the hardness of hot pre-strained (up to 17%) and
quenched 22MnB5 steel was observed; this was attributed to the enhanced diffusive transformation by grain renement
(Barcellona and Palmeri, 2009). Somani et al. (2001) reported that the phase fraction of ferrite in 0.22C1.1Mn0.2Cr
0.0034B steel was increased by 22% and 48% with pre-strains of 0.16 and 0.39, respectively, under a cooling rate of
50 C/s. An increase in the bainite start temperature and decrease in the martensite start temperature resulting from
the hot deformation and dynamic recrystallization of prior austenite were observed in 22MnB5 steel (Nikravesh et al.,
2012). The bainite transformation started earlier, while the transformation of martensite was delayed. Fan et al.
(2008) reported a shift of the nose of transformation curves to earlier time region in the continuous cooling transforma-
tion (CCT) diagram of a boron-bearing CMn steel when hot compression was applied before quenching. Consequently,
ferrite and bainite transformations were enhanced at low and intermediate cooling rates, respectively. Ti added
22SiMn2TiB boron steel promoted diffusive transformation and suppressed displacive transformation after the quenching
of non-isothermally deformed austenite (Shi et al., 2012). A uniaxial stress eld in austenite is known to accelerate both
diffusive (Jepson and Thomson, 1949; Bhattacharyya and Kehl, 1955) and displacive (Kulin et al., 1952; Patel and Cohen,
1953) martensite transformation kinetics.
However, few simulations have considered the effect of austenite deformation on the phase transformation in spite of the
importance of this issue (Esaka et al., 1986; Min et al., 2012). In the present study, a coupled thermo-mechanical-metallur-
gical FE model including the austenite deformation effect on subsequent diffusive transformation was developed to better
predict the microstructure, mechanical properties and residual stress distribution in a hot-press formed part. For the inter-
action between deformation and phase transformation, a dislocation-density-based austenite grain renement model was
introduced into a semi-empirical diffusive transformation kinetics model. This is the main idea of the present work. The con-
stitutive models were implemented in the implicit FE software ABAQUS/Standard using UMAT and UMATHT for mechanical
and thermal user material subroutines, respectively. Related stress integration and microstructure evolution algorithms
were also developed. The overall developed material description was applied to several applications: (1) dilatometry test
to validate fundamental kinetics without external loading; (2) dilatometry test to validate the kinetics under external load-
ing, which introduces transformation plasticity; (3) quenching distortion; (4) hot torsion and quenching; and (5) hot-press
forming of a U-channel shaped part.
2. Material modeling
2.1. Microstructure evolution in low-alloy steels
2.1.1. Austenite grain renement

Accurate estimation of the austenite grain size (AGS) at hot working temperatures is important because the austenite
grain boundary area (i.e., number of nuclei) is known to have a considerable inuence on the phase transformation kinetics
(Cahn, 1956; Kirkaldy et al., 1978). Therefore, the AGS is a factor that controls the nal mechanical properties of steel parts
after heat treatment. Austenite deformation is associated with grain elongation and the increase in dislocation density along
austenite grain boundaries. Therefore, nucleation rates of ferritic phases at grain boundaries increase considerably due to the
dislocations; this enhances the mobility of diffusing atoms in austenite. Newly generated deformation bands and annealing
twins become additional nucleation sites for ferritic phases inside austenite grains (Tamura et al., 1988; Walker and Honey-
combe, 1978; Umemoto et al., 1984). Consequently, the overall transformation rate increases considerably.
156 H.-H. Bok et al. / International Journal of Plasticity 58 (2014) 154183
A dislocation-density-based effective austenite grain size concept (EAGS, deff) was proposed by Suehiro et al. (1987) to
account for these effects through the grain size of austenite.
n 1
deff d0 1 ad qdd 2:1
where ad and nd are constant coefcients. In this study, these constants refer to the experimental work from Suehiro et al.
(1987), where a low-carbon steel [0.15%C0.3%Si1.1Mn] was austenitized at 1000 C and deformed at a temperature of
800900 C, i.e., ad of 1.0 1011 and nd of 1.154. In Eq. (2.1), d0 and qd are the initial AGS and dislocation density, respec-
tively. Yoshie et al. (1996) proposed the following dislocation accumulation model after a thermo-mechanical analysis of
carbon steels.

be_ ce
qd 1 exp q0 2:2
c e_
where b and c are variables associated with the dislocation density accumulation and decrease due to dynamic recovery.
These two variables can be expressed as a function of the effective plastic strain e:

m Qb
b b00 expfb C Nb d0 b e_ nb exp 2:3
RT

m Qc
c c00 expfc C Nb d0 c e_ nc exp 2:4
RT
where b00, c00, mb, mc, nb, nc, Qb, and Qc are material coefcients. R is the gas constant, 8.314 J/mol K, and T is the temperature
in kelvins. In Eqs. (2.3) and (2.4), fb(CNb) and fc(CNb) are functions that introduce the effects of niobium. However, since this
material does not contain niobium, then fb C Nb fc C Nb 0 in this work. By rearranging Eq. (2.2) without the term related
to the niobium effect, the dislocation density becomes

b00 mb mc _ 1nb nc Qb Qc m Qc
qd d0 e exp 1 exp c00 d0 c e_ nc 1 e exp q0 2:5
c00 RT RT
Therefore, the evolution of the dislocation density may be calculated from above as

@ qd @ qd _ @ qd
dqd de
de @T _ dT
2:6
@ e e_ ;T @ e_ e;T e;e
All of the required material constants in Eqs. (2.3) and (2.4) are listed in Table 2. Note that these were taken from Yoshie et al.
(1996), who obtained the data from low-carbon steel grades.
The static recovery model for the dislocation density immediately after deformation is as follows:

T rev nrev
qrev qd exp crev exp t rev 2:7
T
where the constants crev, Trev and the exponent nrev were determined to be 90 s-nrev, 8000 K and 0.7, respectively, from a pre-
vious article (Totten et al., 2004). The recovery period trev is obtained by the subtraction of the deformation nish time tfor
from the total time t. The rate form of this density may be calculated by differentiation

@ qrev @q
dqrev dtrev rev dT 2:8
@t rev T @T trev
Thus, the dislocation density can be updated incrementally in the simulations using
qd jn1 qd jn dqd jn1 if 0 6 t 6 tfor 2:9
or
Table 1
Chemical composition in mass percent.
C Mn Si P Ni Cr W Cu Al Ti B
0.25 1.55 0.25 0.01 0.002 0.009 0.027 0.01 0.03 0.04 0.002
Table 2
Material constants for dislocation density evolution.
b00 (m2) c00 (s1) mb mc nb nc Qb (J/Kmol) Qc (J/Kmol)

13
1.33 10 144 0.207 0.182 0.105 1.02 34100 18200
qd jn1 qd jnfor dqrev jn1 if t > tfor 2:10
where ()|n and ()|n+1 represent the values at time steps n and n + 1, respectively. qd jnfor is the dislocation density at tfor. As
Eqs. (2.10) and (A.8) indicate, the dislocation density during recovery is updated incrementally throughout a time span after
tfor. By Eq. (A.8), the dislocation density increment for recovery dq rev is always negative and thus the total dislocation density
in Eq. (2.10) decreases when dq rev is nonzero. Also, the empirical model (Totten et al., 2004) considers temperature effect,
which leads to an exponential decrease of recovery reaction as temperature decreases (see Eq. (2.7)). Therefore, the temper-
ature history is automatically considered in time steps imposed in the general FE process. In fact, the time for initiating the
static recovery tfor can be determined for each integration point (or, equivalently, element) by checking the mechanical
deformation rate (or the strain rate). In other words, static recovery is initiated if the temperature condition is met when
the deformation is static (or the strain rate becomes close to zero), and the time trev is activated from this point. In this study,
a simplication was adopted by considering the specic characteristic of the HPF process, in which the sheet part is formed
at a given high temperature and, when nished, becomes static. Therefore, the time for recovery initiation can be clearly
specied as the forming nish time in the nite element modeling. Of course, more elaborated numerical treatment might
be necessary as a future work.
Note that the initial dislocation density q0 in Eq. (2.2) was assumed to be zero in this work. The partial derivatives in Eqs.
(2.6) and (2.8) are provided in Appendix A.
To consider the possible annihilation of dislocations as recrystallization progresses, the mean effective AGS d was pro-
posed to be
X RX dRX 1 X RX d
d 2:11
eff
where dRX and XRX stand for the grain size of the fully recrystallized austenite and volume fraction, respectively.
Recrystallization occurs either dynamically (dynamic recrystallization or DRX) during the plastic deformation stage or stat-
ically (static recrystallization or SRX) after deformation. In some cases, a mixed dynamic-static form of recrystallization
(meta-dynamic recrystallization, or MDRX) takes place when dynamically generated nuclei exist even after deformation.
The Avrami-type kinetics model describing the recrystallizations and corresponding grain size models are outlined in
Appendix A.
The recrystallization stop temperature Tnr in austenite is usually determined by the mean ow stress (MFS) analysis,
which uses a series of hot torsion tests for rod-type specimens (Ryan and McQueen, 1986; Sun et al., 1993; Maccagno
et al., 1994). In the present study, the specimens were soaked at 1200 C for 600 s followed by 21 passes of torsion during
which they were cooled at a rate of 1 C/s. The torsion tester is schematically shown in Fig. 7 and additional details are given
at the beginning of Section 3. The time interval between passes was 30 s, which led to a temperature drop of 30 C between
each pass. A nominal shear strain of 0.2 was applied with a rate of 2 s1 during each pass to deform the austenite and induce
Re
recrystallization. The MFS at each torsion step was eb ea 1 eab rav de, where rav is the average stress in each torsion step
for a strain range of ea to eb. Tnr was assumed to be located at the intersection point between two linear-tted lines on the
MFS vs. 1000/T plot. Fig. 1 presents the value normalized by the MFS at 1200 C. The slope change in the MFS was attributed
to strain hardening in the non-recrystallized regime below Tnr. The measured Tnr of the boron steel was determined to be
962 C. This temperature is much higher than the usual hot-press forming temperature, which is around 800 C. However,
Tnr can be changed by the process conditions, such as the cooling rate, interpass time and degree of deformation because
Tnr depends on the precipitation kinetics, solute drag effect and other parameters (Hodgson and Gibbs, 1992; Senuma
et al., 1992; Sun et al., 1993). In general, recrystallization is retarded by ne precipitation and the solute drag effect; this
eventually increases Tnr. Fe23(C, B)6 iron-borocarbide precipitations were observed in boron-added steel (Melloy et al.,
1973; Maitrepierre et al., 1975; Lanier et al., 1994). When the boron content was higher than 0.002 mass%, the precipitation
became predominant in low-carbon steels with an average carbon content of about 0.2 mass% (Melloy et al., 1973). Boron
was also reported to retard recrystallization in NbTi added IF steels (Zahiri et al., 2004; Kim et al., 2005). In this case, how-
Fig. 1. Mean ow stress of the steel and determination of recrystallization stop temperature.
ever, a different suppression mechanism was likely to be activated because the boron content was lower than 0.001 mass%;
since boron atoms segregated in austenite grain boundaries, B-rich precipitation was not predominant. Regardless of these
observations, other studies have reported recrystallization in hot deformed boron steels (Fan et al., 2008; Nikravesh et al.,
2012). Further study might be necessary to devise an optimized test method for obtaining a more accurate Tnr. Nevertheless,
Tnr as measured by the hot torsion test was regarded as a reasonable material parameter in this study.
Once the mean EAGS is determined, the conversion relationship to the ASTM grain size number G becomes
3 G1 1=2
25:4 10 2
d

2:12
100
This relationship is used in the following sections to associate the grain renement model with a semi-empirical diffusive
transformation model.
2.1.2. Transformation kinetics models

The Kirkaldy-type semi-empirical equation (Kirkaldy and Venugopalan, 1984) has been frequently used to model diffu-
sive transformation kinetics for low-alloy steels when timetemperature-transformation (TTT) and CCT diagrams are not
available. On the other hand, the JMAK model (Kolmogorov, 1937; Johnson and Mehl, 1939; Avrami, 1939, 1940, 1941) is
preferable when the experimental TTT diagram is available. In general, the Kirkaldy model predicts isothermal kinetics of
steels based on the ZenerHillert model (Zener, 1946; Hillert, 1957), where diffusive transformations from austenite to fer-
rite, pearlite, and bainite are presented separately. The original Kirkaldy model based on various TTT diagrams presented in
the US Steel Atlas was further modied by Li et al. (1998). The Li model can predict the kinetics more accurately than the
original model, which usually underestimates the hardenability. The main advantage of Kirkaldy-type models is that the
kinetics equations require only the chemical composition data and AGS. The general kinetics can be written as
u_ k fk G; Tfk Comp1 fk uk 2:13

where / and Comp denote the phantom phase fraction and chemical composition in mass%, respectively. The subscript k
denotes the type of reaction, namely, k = 1, 2, and 3 for ferrite, pearlite and bainite reactions, respectively. The functions fk for
each reaction are

Q k
fk G; T Bk GT k TNk exp 2:14
RT
where Bk(G) are the AGS related terms: B1 = 20.41G, B2 = 20.32G, and B3 = 20.29G. Based on the grain renement model discussed
in the previous section, G becomes a function of the strain, strain rate, temperature, initial AGS and recrystallized fraction.
The transformation start temperatures T1, T2 and T3 are equal to the Ae3, Ae1 and Bs temperatures of the alloy. The recom-
mended exponents are 3 for N1 and N2 and 2 for N3 (Kirkaldy and Venugopalan, 1984; Li et al., 1998). The activation energies
Q1, Q2, and Q3 are 27,500 cal/mol as suggested in the original literature (Li et al., 1998). The gas constant R in Eq. (2.14) is
1.987 cal/mol K. The chemical composition function fk(Comp) is described by

fk Comp exp bk aCk C C aMn Si Ni Cr Mo
k C Mn ak C Si ak C Ni ak C Cr ak C Mo 2:15
where CM is the chemical composition in mass% of chemical species M. The coefcients in fk are listed in Table 3. The phase
fraction function that captures the sigmoidal behavior of diffusive transformation is shared by all kinds of diffusive reactions.
n0 1uk 00
fk uk ukk 1 uk nk uk 2:16
In this study, kinetics exponents n0k and n00k in the original work (Li et al., 1998) were modied as a function of temperature for
each kth transformation to improve the hardenability predictability for the boron-added steels. The procedure to construct
the exponents is described in detail in Section 2.1.3. The temperature ranges for the activation of the transformations for
k = 1, 2, and 3 are (Ae3 > T > Bs ), (Ae1 > T > Bs ), and (B>
s T > Ms), respectively. The calculations of the equilibrium temperatures
Ae3 and Acm, and the bainite start temperature Bs, are explained in Appendix B. The eutectoid temperature Ae1 is located at the
intersection made by the carbon contents CC (Ae3 ) and CC (Acm ), which are quadratic functions of the variables Ae3 and Acm,
respectively. The phantom fractions /1 and /2 should be converted to the actual fractions using the following relationships:
u1 ua1 =ueq
1 and u2 ua2 =1 ueq
1 2:17
Table 3
Coefcients of chemical composition function, f k (Comp).
k bk aCk aMn
k aSi
k aNi
k aCr
k aMo
k
1 1 6.31 1.78 0.31 1.12 2.7 4.06

2 4.25 4.12 4.36 0.44 1.71 3.33 5.19CMo1/2
3 10.23 10.18 0.85 0 0.55 0.9 0.36
where the superscripts a and eq indicate the actual and equilibrium fractions, respectively. The equilibrium ferrite fraction is
determined by applying the lever rule.
ueq
1 C C c T C C =C C c T C C1 2:18
The carbon content CCc in austenite varies with temperature and can be obtained by replacing CCc(T) with the function CC
(Ae3 ) if the temperature is above Ae1 or CC (Acm ) if the temperature is below Ae1. The carbon content in ferrite CC1 was assumed
to be zero considering its negligible amount compared with CCc .
The diffusionless martensite transformation begins to take place below Ms. This phase decomposition depends on the
undercooling from Ms; thus, the phase fraction of martensite /4 can be expressed as a function of T. Considering the amount
of retained austenite fraction /5 at Ms, the kinetics for the martensite is given by
u4 u5 1 expnMs TN 2:19
with
X
3
u5 1 uk at T Ms 2:20
k1
The widely accepted values for the constant n and exponent N are 0.011 and 1, respectively, as suggested by Koistinen and
Marburger (1959).
2.1.3. Modication of transformation kinetics

To improve the predictability of the overall transformation kinetics, a series of cooling tests was conducted with axial
strain measurement. The specimen had the geometry of a 10 mm length and 1.2 mm diameter. Specically, a single martens-
ite transformation could be clearly analyzed by applying the lever rule on the strain change, as depicted in Fig. 2a. The lever
rule is used to estimate the phase fraction using the dilatation and thermal strains of a single phase in a transformation
Fig. 2. Determination of martensite kinetics: (a) lever rule; (b) martensite model.
temperature range based on the relation uT bT=aT. Fitting the curve for /4(T) Z, as schematically shown in Fig. 2b,
produced an S-curve for the martensite evolution. However, the conventional KoistinenMarburger (KM) model, described
in Eq. (2.19), does not represent an S-curve. To capture this curve, an additional exponential curve was combined with the
original KM model:
u4 u5 u4 I HI u4 II HII 2:21
with the following two kinetics components
u4 I b expnI DT I 2:22
u4 II 1 expnII DT NII 2:23

and the Heaviside step functions

1 if M s I P T P T c
HI 2:24
0 otherwise

1 if HI 0
HII 2:25
0 otherwise
In Eqs. (2.22) and (2.23), the undercoolings DTI and DTII are (M s I T) and (M s II T), respectively. Tc is the temperature at
which the values of /4_I and /4_II are identical. For the material considered in this study, the measured Tc, Ms_I, and Ms_II were
353, 422 and 355 C, respectively. The parameters b, nI, and nII and exponent N were found to be 1.496 103, 6.65 102,
0.117 and 0.651, respectively, based on the best ttings to the above equations. The experimental Vickers hardnesses of indi-
vidual phases, Hvk, were expressed as a function of the imposed cooling rate
Hv 12 H0v 12 a1 T_ 12 ; Hv 3 H0v 3 a2 T_ 3 ; and

(
H0v 41 a3 T_ 4 a4 T_ 24 if 70 C=s P T_ 4 P 30 C=s
Hv 4 2:26
H0v 42 a5 T_ 4 if 500 C=s P T_ 4 > 70 C=s
where the coefcients ai and H0v i are summarized in Table 6. The average cooling rates T_ 12 , T_ 3 , and T_ 4 were (Ae3Bs)/
(tBstAe3), (BsMs)/(tMstBs) and (MsTroom)/(tTroomtMs), respectively where Troom is the room temperature. tAe3, tBs, tMs,
and tTroom are the times when the temperature reaches Ae3, Bs, Ms and Troom, respectively, during cooling. The overall
hardness of the steel was assumed to be
X
4
Hv u1 u2 Hv 12 uk Hv k 2:27
k3
Note that the contribution of austenite to the hardness is ignored in Eq. (2.27) because the austenite phase fraction u5 is
negligible after cooling.
Further optimization of transformation kinetics became available by setting the aforementioned kinetics exponents n0k
and n00k in Eq. (2.16) as
n0k T n0k0 n0k1 DT k n0k2 DT 2k 2:28
n00k T n00k0 n00k1 DT k n00k2 DT 2k 2:29

with DT k Ae3 T for k = 1 and 2 and DTk = Bs T for k = 3. The coefcients were obtained by applying an optimization pro-
cedure that minimized the following objective function for the hardness:
N T_
X ex _ i 2
Minimize Hsim _ i
v T Hv T 2:30
i1
where the superscripts sim and ex denote the simulated and experimental hardnesses, respectively. T_ i is the cooling rate for
the i-th linear cooling, and N T_ is the total number of applied cooling rates covering the range of 0.1100 C/s. The chemical
composition, required in Eq. (2.15) was listed in Table 1. The austenite grain size number was 8 in the ASTM denition. The
optimization results are presented in Fig. 3a and the corresponding coefcients are listed in Table 7. The continuous trans-
formation kinetics results are shown in Fig. 3b in the form of a CCT diagram; the experimental diagram in a previous work by
the authors (Choi et al., 2011) is superimposed for comparison purpose.
2.1.5. Non-isothermal treatment of diffusive transformation

The diffusive transformation model explained in Section 2.1.2 is based on isothermal kinetics. In many practical cases,
however, steels transform under a continuously decreasing temperature eld. In this circumstance, the incubation time
Fig. 3. Comparison of continuous cooling responses: (a) resulting hardness; (b) CCT diagram.
for the start of the transformation is delayed compared to the isothermal case. Therefore, a proper treatment to convert the
incubation from isothermal to non-isothermal is required. Scheils additivity principle (Scheil, 1935) has been widely applied
to account for this issue:
Z ts X Dt n
dt
1 2:31
0 suk ; T n suk ; T n
where ts is the incubation time in continuous cooling while s(/k, Tn) is the incubation time for / to reach 0.001 at the n-th
isothermal holding temperature Tn. The time increment Dtn is the dwelling time at each Tn along the discretized cooling
curve, as depicted in Fig. 4. The isothermal incubation time s for the Li model is now expressed as
Z uk
duk
sk fk Compfk G; T i 1 2:32
0 fk uk
When the condition in Eq. (2.31) is satised, it is assumed that the transformation can start. Then, phase evolution proceeds
and the phase fractions are updated isothermally at every Tn for Dtn as outlined in the following Section 2.1.6.
2.1.6. Microstructure update algorithm for continuous cooling

The nonlinear isothermal transformation kinetics equation in Eq. (2.13) has been solved by the explicit method (Watt et
al., 1988), RungeKutta method (Lee et al., 2009) and Newtons method (Chandran and Michaleris, 1999; kerstrm and Old-
enburg, 2006; Bok et al., 2011). As mentioned, the transformation was assumed to initiate when Scheils additivity principle
in Eq. (2.31) meets the condition. The Gauss quadrature integration method with ve integration points was used for Eq.
(2.32). Then, Newtons method was applied to update the phase fraction at the end of each discretized time step as explained
in Appendix B. The calculated volume fraction was used as the initial guess for iteration of Newtons method for the next
time step (with the updated temperature). This procedure was implemented for T > Bs when k = 1, 2 and for Bs > T > Ms when
k = 3 as long as /5 > 0. This update scheme was also applied by Fernandes et al. (1985) to the JMAK-type kinetics model. They
introduced the ctitious time concept proposed by Tzitzelkov et al. (1974) to account for the incubation period in the JMAK
model. In the present work, the JMAK model was replaced by the Li model. The explicit and Newtons methods were
Fig. 4. Incubation time conversion from TTT to CCT.
considered together to solve the kinetics equations. To save computational cost, the explicit scheme, as expressed in Eq. (B.1)
in Appendix B, was activated when the time increment tn+1 tn was smaller than the prescribed critical time step 0.1 s.
2.2. Thermo-mechanical-metallurgical constitutive model
The thermal, volumetric and transformation plasticity strains as well as the classical elasto-plastic strain were considered
in addition to the microstructural eld discussed in the previous sections to simulate the deformation of hot working of bor-
on steel in a transient thermal eld.
2.2.1. Description of deformation eld

Under the assumption of additive decomposition, the total strain increment de consists of elastic dee, plastic dep and ther-
mo-metallurgical detm parts.
de dee dep detm 2:33

The thermo-metallurgical part was decomposed into thermal (th), volumetric (vol) and transformation plasticity (trp) strain
rates as follows:
detm deth dev ol detrp 2:34
The thermal and volumetric parts combine as a function of the density of existing phases as follows:
1=3 !
q T n 1 X 5
uk
deth dev ol 1 I with 2:35
q T n1 q qk T
k1
where qk(T) is the density in kilograms per cubic meter and is described by
q5 8099:79 0:506T1 1:46 102 C C
and
q1;2;3;4 7865:96 0:297T 5:62 105 T 2 1 2:62 102 C C 2:36
for low-carbon steels (Jablonka et al., 1991). In Eq. (2.35), I is the identity tensor. For the transformation plasticity strain
increment, which accounts for the additional plastic strain during transformation even below the mechanical yield stress,
Greenwood and Johnsons (1965) model was considered.
!
X
4
3 d/uk
trp
de K 5!k u
_k hry ; r s with /uk 2 uk uk
2:37
k1
2 duk
In Eq. (2.37) the function u expresses the changing rate of transformation plasticity strain with the progress of phase trans-
formations (Desalos, 1981). K is the transformation plasticity coefcient for reactions 5 to k, and s is the stress deviator. The
coefcients K5?1, 2, 3 and K5?4 were 5.114 105 and 5.656 105, respectively, based on a previous experimental inves-
tigation of the same boron steels (Choi et al., 2011). These values were obtained using a series of dilatometry tests under
various uniaxial compressive loads. The function h proposed by Leblond et al. (1989) can consider the nonlinear stress
dependency of transformation plasticity at a relatively high stress level.
(
1 if rr < 12
hr; r
2:38
1 3:5rr 12 if rr P 12
where r
and r are the effective stress dened by the specic yield function and effective stress, respectively.
In the elasticplasticity theory, the stress state is elastic if the following criterion is satised:
Ur; q r
r ry q 6 0 2:39
where r is the yield function and ry is the evolution of the yield surface as a function of the hardening parameter q. For a von
Mises isotropic material,
p 1
r r 3J 2 with J 2 s : s 2:40
2

To account for the material behavior at high temperature, the hardening parameter q can be the effective strain v arepsilon,
_
strain rate v arepsilon and temperature T. To consider the contributions of existing phases to the yield stress, the following
expression was taken:
ry uk ry k 2:41
where ry_k were individual ow stresses for ferritic phases. The Swift hardening law as a function of temperature was used
for the ow stresses:
ry k CTk e0 k enTk for k 1; 2; 3; 4 2:42
with
CTk C 0 k C 1 k T and nTk n0 k n1 k T 2:43
For the austenite phase, a more accurate expression that considers the effects of strain rate and temperature was used be-
cause most deformation during hot-press forming takes place in the austenite state. In this study, an expression based on a
modied JohnsonCook model (Yoo et al., 2010) was used:
!
e_ T Tr T T r me_
ry k Ae0 en 1 Ce ln _ 1 sign j j for k 5 2:44
e0 Tm Tr Tm Tr
with
e_
Ce C 0 C 1 e and me_ m0 m1 ln 2:45
e_ 0
where C0, C1, m0, and m1 are material coefcients for the austenite phase. Here, the constants for the austenite ow curve
were obtained from previous literature (Yoo et al., 2010) for the same boron steel, while those of the other phases were esti-
mated using the thermodynamics database software JMatPro (Saunders et al., 2003). Table 4 summarizes the required con-
stants for the ow curves. The Youngs modulus of 22MnB5 boron steel as a function of temperature was taken from the
literature (Turetta, 2008) to be as follows:
ET 2:825 105 61:4T 0:684T 2 4:568 104 T 3 MPa 2:46
Considering the constitutive relationship above, the stress was updated thermo-elasto-plastically. If the stress state did not
satisfy the yield condition dened in Eq. (2.39), the stress state was assumed to be thermo-elastic. A predictorcorrector
algorithm based on the NewtonRaphson method was applied for the stress update (Simo and Ortiz, 1985; Soare and Barlat,
2011; Lee et al., 2012). The numerical procedure to update the stress components is summarized in Appendix C.
Table 4
Hardening constants for ow curves.
k C0_k (MPa) C1_k (MPa/C) n0_k n1_k (/C) e 0_k

For ferritic phases
1 1022 1.1033 0.5247 6 104 1.8 103
2, 3 2958 3.7276 0.4871 7 104 2 103
4 2449 1.0667 0.0742 6 105 2 103
For austenite (k = 5)
e0 e_ 0 s1
A (MPa) n C0 C1 m0 m1 Tm (C) Tr (C)
0.01 0.7 370 0.3 0.0586 0.011 0.8 0.0335 1500 730
Table 5
Thermal material properties (T appeared in this table is in the Kelvin temperature).
k DHk (kJ/kg) cpk (J/kgK) k (W/mK)

1 89.4 435 + 0.102T + 541 106T2 52.284 0.0316T 1.41049 104T2 2.78145 107T3 1.3279 1010T4
2 77.5
3 66.5
4 81.5
5 n/a 439 + 0.1142T
Table 6
Hardness coefcients.
H0v 12 (HV) H0v 3 (HV) H0v 41 (HV) H0v 42 (HV) a1 (s/C) a2 (s/C) a3 (s/C) a4 (s2/C2) a5 (s/C)
152.3 283.8 453.8 484.7 20.05 0 0.823 -0.0047 0.0499
Table 7
Optimized coefcients of diffusive kinetics exponents.
k n0 k0 n0 k1 (s/C) n0 k2 (s2/C2) n00 k0 n00 k1 (s/C) n00 k2 (s2/C2)

4 5 4
1 0.3996 2.481 10 1.341 10 0.3996 3.944 10 1.143 104
2 0.3998 4.326 105 4.114 106 4.0006 1.884 104 2.957 105
3 0.6503 8.656 104 1.966 106 0.6504 3.856 104 4.682 106
2.3. Description of temperature eld
The heat equation that calculates the temperature eld in a material can be written as
T_ r raTT qtrs T; t 0 2:47

trs
where a is a temperature-dependent material parameter and q is the latent heat generation from transformations. A micro-
structure dependent thermal constitutive modeling can be expressed as
kT kT X
5
aT P5 P5 and qtrs T; t DHk u
_k 2:48
qTcp T k1 uk qk T k1 uk cpk T k1
where cpk , k Z, and DHk Z are the specic heat, thermal conductance, and latent heat per unit mass, respectively. The values
were taken from previous literature (Lee et al., 2010) and are listed in Table 5. The phase mass densities qk, previously shown
in Eq. (2.36), were incorporated into the ABAQUS input parameters with user eld variables from another subroutine called
USDFLD. Latent heat effect can be considered through either HETVAL or UMATHT. Heat generation can be directly performed
by the former subroutine. However, this heat generation by the phase transformation can be indirectly considered through
UMATHT by the modication of the derivative of internal energy oU/oT, which is identical to cp. For this purpose, the ctitious
specic heat cp (Simsir and Gr, 2008) was introduced in UMATHT to modify the derivative in this work.
@U
cp cp T qtrs T_ 1 2:49
@T
Using the thermal constitutive model, conduction, radiation and convection were generally solved in the coupled tempera-
ture-displacement FE step for the thermo-mechanical problems in Section 3.
Thermal boundary conditions prescribing the heat ux along the normal direction n to the surface area C were estab-
lished as follows.
For contact with tools:
@T
kT hcnt T T tool 2:50
@n
For air cooling:
@T
kT hconv hrad T T air ; hrad rSB erad T 2 T 2air T T air 2:51
@n
where hcnt, hconv, and hrad are heat transfer coefcients of conduction, convection and radiation, respectively. rSB and erad are
the StefanBoltzmann constant (5.6697 108 W/m2 K4) and radiation emissivity of the material, respectively. The initial
condition for the entire material volume X was then
T 0 Tx; y; z; t0 2:52
2.4. Model summary
A comprehensive material model was proposed for the implementations in thermo-mechanical-metallurgical FE

simulations of hot-press forming processes, in which a large amount of deformation takes place an-isothermally in austenite
at elevated temperature. In particular, the effect of the austenite deformation on the subsequent diffusive phase transforma-
tions was taken into account in order to address the accelerated transformation resulting from this deformation. This con-
sideration was done by incorporating a grain renement model (Suehiro et al., 1987) based on dislocation density change
Fig. 5. Thermo-mechanical-metallurgical coupling: (a) interaction among three elds; (b) ow chart of the proposed metallurgical eld.
according to the forming conditions (Yoshie et al., 1996; Totten et al., 2004) into the semi-empirical diffusive transformation
model by Li et al. (1998).
In addition, the original Li and the KoistinenMarburger models for diffusive and displacive transformations, respectively,
were modied to better capture the overall transformation behavior of the considered boron steel.
Related thermal material properties such as specic heat and conductivity were dened as functions of existing micro-
structure constituents and temperature. The latent heat generation during the transformation was also included through
the ctitious specic heat.
Furthermore, thermal and transformation relevant strains (e.g., eth, evol, etrp) were added to the von Mises plasticity to
consider thermal stress development in a part subjected to hot-press forming.
The overall interaction among the metallurgical, mechanical and thermal elds is briey described in Fig. 5a. In this g-
ure, Mechanical Field, containing the constitutive descriptions in Section 2.2, is implemented in ABAQUS through the UMAT
subroutine for stress update. The UMATHT contains Metallurgical and Thermal Fields, in which the microstructure evolution
explained in Section 2.1 and the thermal constitutive model in Section 2.3, respectively, are implemented. Temperature, a
primary variable in the ABAQUSs temperature-displacement problem, is shared in the UMAT and UMATHT for each time
step. However, the other parameters such as effective strain, phase fraction, thermo-metallurgical strains are shared as
user-dened state variables in the subroutines.
Mechanical and Thermal elds are inter-linked as follows: thermal boundary conditions depend on the deformation,
which governs the contact conguration between the blank surface and the tools. Also, the expansion or contraction due
to temperature gradient is an important interaction between the two modules. Metallurgical eld is closely related to the
thermal and mechanical elds. Latent heat and volume variation due to phase transformation, mechanical property changes
during phase evolution and transformation plasticity are all examples of this interaction. Of course, most of the phase trans-
formations depend on the applied load or deformation. The integrated metallurgical model is specically presented with the
owchart in Fig. 5b.
Finally, the entire material constitutive model in Section 2 was incorporated into the FE simulations in this study through
the construction of UMAT and UMATHT. The former subroutine was associated with the stress update, while the latter was
mainly associated with the microstructure evolution model.
3. Model validation and application
The developed model and its implementation were validated under basic thermo-mechanical processing conditions. Sec-
tions 3.1 and 3.2 present dilatation tests with and without axial load that validated the developed constitutive description,
including the transformation kinetics of the boron steel. Section 3.3 describes the FE modeling for a shaft with a at key
groove to validate the 3D thermal distortion behavior during water quenching. Section 3.4 compares the simulation results
of a series of hot torsion tests followed by water quenching with the experimental results to validate the EAGS effect. For this
validation, experiments in Sections 3.1, 3.2, and 3.4 were conducted in the present study, while the experiment results in
Section 3.3 was obtained from the literature (Huang et al., 2000; Arimoto et al., 2009).
Finally, the draw bending of a U-channel was virtually performed as a representative problem for a real hot-press forming
process. This virtual simulation was used to predict and analyze the strength and residual stress of a hot-press formed chan-
nel; details are given in Section 3.5.
The alloy composition listed in Table 1 was used for all the simulations in the following subsections with the required
material parameters as listed through Tables 27.
For the dilatometry experiments in Sections 3.1 and 3.2, the Theta dilatometer (Theta Industries, Inc.) in Fig. 6a was
used to obtain the axial strain change under heating and cooling. As the gure shows, the specimen temperature is
controlled through heating by induction coil and cooling by He gas jet from the hollow inner coil, in which gas nozzles
are facing the specimen surface. The axial length change of the specimen is transmitted to the LVDT (Linear Variable
Differential Transducer) through push rods and the resulting strain is recorded based on the electrical signal converted
by the LVDT. High temperature oxidation of the specimen is suppressed in the vacuum chamber, which is connected to
mechanical and diffusion pumps. As an illustration, a typical strain evolution during the dilatometry test is provided in
Fig. 6b. In this gure, the strain returns to near the original state after one cycle of heating and cooling. However, it is
not perfectly reversible due to the complex behavior associated with phase transformation; i.e., anisotropy in transfor-
mation along axial and width directions, transformation plasticity by small holding force from the holding (gripping)
rods.
Hot deformation followed by water quenching in Section 3.4 were carried out with the special torsion tester, described in
Fig. 7. In this machine, a servo motor produces torsional deformation in the central region of the specimen. The correspond-
ing torque is measured by a load cell, while the engineering shear strain cLd is calculated by the rotation angle measured by a
shaft encorder. In this tester, heating is controlled by an IR furnace and cooling is controlled by either air ow or water spray-
ing according to the imposed cooling rate.
Note that the implementation in this study is based on the small strain plasticity theory, where additive decomposition of
strain components is assumed. For the strain measurement, the denition of the true strain (or logarithmic strain) is adopted
to be consistent to the strain in ABAQUS.
Fig. 6. Dilatometry test: (a) layout of dilatometer (Theta dilatometer); (b) strain change in heating and cooling in a dilatometry test.
3.1. Thermo-metallurgical dilatation without external loading
A series of dilatometry tests were simulated using the developed constitutive and kinetics models. The geometry of the
specimen was 10 mm (length) 1.2 mm (diameter), as shown in Fig. 9a. Only 1/8th of the sample was modeled in view of
the symmetry of the geometry and boundary conditions. For this dilatometry model, no external force was imposed. Three
different predened cooling rates (or temperature histories) were directly applied: 0.5, 10, and 50 C/s. These temperature
loadings were assigned as temperature histories according to the given cooling rates.
Correspondingly, the same experiments were carried out with the Theta dilatometer for comparison. The specimens were
held at 930 C for 5 min for complete austenitization before cooling. In the simulation, however, it was assumed that the
austenitization was already nished. Therefore only the cooling process was simulated with the initial temperature of
930 C.
Fig. 8a presents the simulated and experimental dilatation curves for cooling at three different rates. Overall, the simu-
lated dilatation results agree reasonably with the experiments, particularly at a cooling rate of 50 C/s. However, the simu-
lated dilatations under the two lower cooling rates deviated slightly from the measured curves. This deviation might be due
to a complicated transformation behavior that the current model cannot fully capture.
Calculated phase fractions along cooling are depicted in Fig. 8bd, f as the applied cooling rate increases. As shown in
these gures, multiple transformations took place in the cooling rates of 0.5 and 10 C/s, while only a single martensite trans-
formation occurred in the fastest cooling. This feature is exactly reected in the corresponding dilation curves; i.e., the multi-
ple uctuations in the slower cooling cases, while a monotonic change in the fastest case.
Fig. 7. Schematic of torsion tester and specimen geometry.
Fig. 8. Simulated phase transformations under linear cooling conditions: (a) comparison of dilatations; (b) calculated phase fraction with cooling rates of
0.5 C/s; (c) calculated phase fraction with cooling rates of 10 C/s; (d) calculated phase fraction with cooling rates of 50 C/s.
As explained in Section 2.1.3, the material parameters in Table 7 for the modied kinetics function in Eq. (2.16) were ob-
tained basically by a calibration of the resulting hardness through the optimization function in Eq. (2.30). This approach
might be one of the reasons for the deviation obtained in the dilatation curves.
Another optimization method using a multi-objective function involving both hardness and dilatation data might be
more accurate to reduce the deviation. This multi-objective approach, however, could lead to more difculty in getting a con-
verged solution and to an increase of computational cost.
Fig. 9. Transformation plasticity: (a) simulation condition and the geometry of the rod; (b) comparison of experimental and simulated dilatations with the
transformation plasticity effect; (c) simulated transformation plasticity strain.
Instead, a simplication consisting of adopting a unied ferrite-pearlite transformation kinetics is suggested. In this case,
it is postulated that the ferrite and pearlite transformations occur simultaneously in an average sense; i.e., not sequentially
although ferrite appears earlier during cooling. Then, the transformation kinetics is unied as a single transformation (for
example, f + p phase transformation) leading to a reduction in the number of total transformation equations. As an example,
the calculated kinetics based on this assumption is presented in Appendix D after an optimization for the experimental dila-
tation. It is shown that the S-shape monotonic transformation curve can be captured effectively in Fig. D1.
3.2. Thermo-metallurgical dilatation under compressive loading
Transformation plasticity is known to play an important role when simulating the stress relaxation and additional per-
manent deformation during transformation. Therefore, the residual stress, nal deformed shape after springback, and ther-
mal contact status between the workpiece and tool surfaces can be inuenced by this special strain component. For
verication purposes, the same steel rod, Fig. 9a, was cooled at a rate of 50 C/s under various compressive loads: 1.18,
3.08, 6.51, and 13 MPa in simulations and experiments. The same simulation strategy as described in Section 3.1 was applied
to the simulations. In addition, the predened temperature history according to the applied cooling rate was directly im-
posed to the FE model. Compressive axial loads were imposed on one end surface of the rod to reproduce transformation
plasticity effects. For the experiments, these loads were transmitted to the specimens through the push rod of the dilatom-
eter, as depicted in Fig. 6a.
Note that the magnitudes of the external loads were much lower than the yield strength of the considered material; this
implies that there should be no classical plastic deformation in austenite according to the conventional elasticplasticity
theory. Therefore, only transformation plasticity was expected as the phase transformation progressed. Fig. 9b shows the
change in dilatation curves under the three external compressive loads for both simulation and experiment. Fig. 9c shows
that this unusual plastic strain increasing with the magnitude of the applied load and induced by the phase transformation
was reasonably predicted by the simulation.
3.3. Distortion of a shaft with a at key way in water quenching
Huang et al. (2000) and Arimoto et al. (2009) simulated water quenching of a shaft made of 0.44 mass% carbon steel with
a at key groove. The initial temperatures of the shaft and water were 860 and 30 C, respectively. The special asymmetric
Fig. 10. Water-quenching of a steel shaft with a at key groove: (a) description of geometry; (b) heat transfer coefcient with temperature; (c) result of
quenching distortion and temperature eld with time (scale factor = 3); (d) comparison of curvature changes with time.
cross-sectional shape of the shaft was suitable for emphasizing the distortion due to quenching. The heat transfer coef-
cients at the edge and outer surface regions were taken from Huang et al. (2000). In this section, the same thermal boundary
condition was imposed to investigate the performance of the developed constitutive model in predicting quenching distor-
tion. Unlike the previous cases in Sections 3.1 and 3.2 in which the linear cooling conditions were enforced, the latent heat
effect option was activated to involve the heat generation during phase transformation of the key; i.e., the fully coupled ther-
mo-displacement analysis with heat transfer condition was solved.
Fig. 10a and b shows the nite element model with the dimensions of the thermo-mechanical solid elements (C3D8T) and
heat transfer coefcients for both edge and non-edge regions. The simulations were carried out with the phase transforma-
tion turned on and off. In addition, with transformation turned on, two cases were considered using the transformation
strain (TP strain) on and off.
Fig. 10c shows the changes in the simulated shape (scale factor = 3) with the temperature distribution. At t = 0.78 s, the
shaft bent upward due to the rapid contraction around the edge region: This was attributed to the greater heat transfer by
larger exposed surface area to the quenchant. In this stage, the temperature difference between the edge and center regions
was about 700 C. After t = 0.78 s, the specimen began to bend in the reverse direction as internal regions cooled down. How-
ever, after t = 1.65 s, another reversal occurred again before the curvature stabilized to a permanent value after the temper-
ature differential between surface and center was less than 28 C. Fig. 10d shows the details of the curvature changes during
cooling.
Fig. 11. Hot torsion followed by water quenching: (a) thermo-mechanical schedule; (b) effective austenite grain size with effective plastic strain; (c)
evolution of effective austenite grain size and dislocation density; (d) resulting phase fractions with effective plastic strain; (e) comparison of experimental
hardness to those simulated hardnesses with and without austenite renement effect.
The gure shows that the calculated curvatures by simulation considering phase transformation agreed reasonably well
with the experimental results in terms of the curvature change, particularly the W-shape curvature. The absolute magnitude
values were different, mainly due to the different thermo-mechanical material properties and transformation kinetics. Note
that the material used by Arimoto et al. (2009) differed from the boron steel used in this study because the exact data were
not provided in the literature.
This W-shape change may be associated with the competition between strains induced by the thermal contraction and
volume expansion from transformation of mainly martensite. The transformation plasticity inuenced the distortion in spite
of the absence of applied load in this particular example. However, only a V-shape was predicted when the transformation
was turned off.
3.4. Quenching of hot deformed rods
To validate the effect of prior austenite deformation on phase transformation, hollow rod-type specimens shown in Fig. 7
were machined from an ingot with the same chemical composition as that listed in Table 1. This ingot was produced in a lab-
scale vacuum melting furnace in the research laboratories at POSCO. Using the torsion tester in Fig. 7, theses specimens were
heated up to 930 C: This temperature was maintained for 300 s for complete austenitization. The specimens were then
cooled down to the forming temperature of 730 C at an air cooling rate of 20 C/s. At this temperature, the specimens were
deformed by torsion with nominal shear strain cLd of 0.2, 0.6, and 0.8, leading to effective plastic strains of 0.115, 0.346, and
cLd
0.462, respectively, using e p 3
. An effective strain rate of 0.0346/s was applied for all the tests. Finally, the hot deformed
specimens were moved into the cooling chamber where they were sprayed with water to induce martensite transformation.
In the FE simulation side, a cooling rate of 500 C/s was assumed to represent the effect of water quenching. Fig. 11b
shows the change in EAGS with the amount of applied strain. In the model, tensile deformation was considered instead
of torsional deformation for the sake of simplicity. The calculated effective grain size decreased remarkably for the applied
strain of 0.12 but was almost constant with larger strains. The predicted evolutions of the grain size and dislocation density
for the applied effective strain of 0.346 are shown in Fig. 11c. This gure shows that the effective grain size rapidly decreased
as dislocations accumulated during the deformation (during 1020 s). Fig. 11d shows the predicted phase fractions. The
Fig. 12. Finite element model for hot press forming of a U-channel: (a) three dimensional geometries of tool and blank in the initial stage; (b) the location of
inspection point and coordinate systems for result analyses.
Table 8
Process conditions for the hot press forming of the U-channel.
tforming tdiequenching Tblank_initial(C) Ttool (C) Tair (C) hcnt (kW/m2K) h(conv+rad) (kW/m2K) Initial ASTM number
(s) (s)
0.4 30 810 60 25 1.5 0.1 8 (22.5 lm)
martensite decreased while ferrite increased as the deformation increased or, equivalently, grain renement proceeded. Note
that the calculated volume fractions of other diffusive products such as pearlite and bainite remained unchanged. As a con-
sequence of this enhanced ferrite transformation, the predicted Vickers hardness decreased with prior deformation. More-
over, the calculated hardness was in very good agreements with the measured values, as shown in Fig. 11e.
When the grain renement effect was not included in the proposed model, the predicted hardness was almost constant
due to full martensite transformations despite the isothermal holdings of 3.3, 10, and 13.3 s to achieve the strains of 0.115,
0.346, and 0.462, respectively, before cooling. This emphasized the importance of considering the effect of grain renement
during prior austenite deformation. Based on the above analysis, it can be concluded that the developed microstructure mod-
el effectively reects the accelerated diffusive phase transformations as a result of austenite deformation before quenching.
3.5. Hot U-draw bending followed by die quenching
3.5.1. Microstructure and strength analyses

As a practical application, hot-press forming of a U-channel was simulated using the developed FE procedures. Fig. 12a
shows the FE meshes for the tools and a blank sheet. A material point was selected for inspection of the material behavior,
temperature and grain size during the process. This inspection point was positioned at the middle of side wall at the end of
the forming, as shown in Fig. 12b. An additional coordinate along the arc length of the sheet was introduced, as shown also in
Fig. 12b. A plane strain assumption was imposed by restricting the displacements of nodes on both edges of the blank to
efciently simulate a long channel. The thickness and width of the blank were 1.2 and 5 mm, respectively. The blank was
modeled with 3D continuum elements and the tools with rigid body elements. Six elements were used through the thickness
direction of the blank to accommodate the curvature in the die shoulder. The position of the blank holder was xed to main-
tain a constant gap between the die and holder and to allow the blank to be drawn smoothly. The punch velocity and stroke
were 100 mm/s and 40 mm, respectively. Die quenching was carried out for 30 s. The other conditions for the simulation are
summarized in Table 8.
The material point starting from the ange region of the blank underwent bending at the die shoulder entry: This was
followed by attening at the end of the die shoulder as the blank was drawn by the punch. Therefore, the elements facing
Fig. 13. Comparison of effective strains during forming in: (a) distribution; (b) history in the inspection point.
Fig. 14. Comparison of effective austenite grain sizes during forming.
Fig. 15. Comparison of phase fraction distributions of: (a) ferrite in x-curvilinear coordinate; (b) ferrite in thickness direction.
the die surface experienced compression followed by tension and vice versa for the elements facing the punch surface.
Fig. 13a and b shows a three-dimensional distribution of the effective plastic strain after the die quenching and the history
of the effective strain of the inspection node during the forming. Most of the deformation accumulated in the wall region,
with a maximum effective plastic strain of about 0.2 at the inspection point located at the surface of the wall center. As a
result, the grains were rapidly rened down to an EAGS of 0.94 lm in that region when the grain renement model was ap-
plied, as shown in Fig. 14. Based on this grain renement effect, all the diffusive transformations in the wall region were
greatly enhanced reducing martensite fraction. These results were compared with the results when the renement model
was turned off (see Fig. 15a and b and Fig. 16). This effect was the most inuential on the ferrite reaction among the diffusive
transformations. The ferrite fraction in the wall region predicted by the renement model turned off was also slightly in-
creased due to imperfect thermal contact by the clearance in that region. Consequently, the strength of the part was greatly
decreased in the wall region due to the austenite deformation effect, as shown in Fig. 17ac.
3.5.2. Residual stress analysis

There are more factors that inuence the shape change after hot-press forming than after conventional cold forming
process. This is because the residual stress of the hot-press formed part depends highly, not only on the strain history,
Fig. 16. Comparison of phase evolutions of ferrite and martensite at the inspection point.
Fig. 17. Resulting Vickers hardness: (a) distribution; (b) prole in x-curvilinear coordinate; (c) cross sectional prole in thickness direction.
Fig. 18. Predicted residual stress: (a) magnitude along x-curvilinear coordinate; (b) cross sectional prole; (c) evolution with time.
but also on temperature eld variations and phase transformation. Transformation plasticity is also known to play a
major role in reducing residual stress. In this U-channel example, the thermo-metallurgical strains of the part were
strongly restricted by the surrounding tools during die quenching. This was particularly predominent in the side-wall
region because the thermal contraction in the y-direction was constrained by the tools, as shown in Fig. 12b. In fact,
the side wall is the most important region in the U-shape channel for determining the nal shape after forming and
springback (NUMISHEET93, 1993; Lee et al., 2005). Fig. 18a compares simulated distributions of effective stresses for
cases with the grain renement turned on and off. A larger residual stress evolved at the side-wall region when the
grain renement was turned on. This was attributed to the early ferrite transformation at high temperature. In this
case, the transformation plasticity acted earlier to relax stress, and a large thermal stress developed during the subse-
quent cooling. Fig. 18b shows a non-uniform cross-sectional stress distribution when the renement option was on. By
the same explanation, the residual stress of the side wall without the renement model was relatively low due to the
martensite-dominant transformation, which considerably relaxed the stress after about 10 s, as shown in Fig. 18c. Con-
sequently, a strong dependency of residual stress (or springback) on the transformation kinetics, which is associated to
the austenite deformation, was reasonably analyzed through this nite element simulation. In Fig. 18c, the stress shows
signicant oscillations. This is likely due to the difculty in getting a converged solution in this fully coupled temper-
ature-displacement simulation, since the high nonlinearity is exacerbated by the contact between the blank and the
tools. Although the maximum time step size was very small, convergence, not only in force but also in thermal equi-
librium, require drastic conditions. In particular, the large friction coefcient of 0.4 might intensify this oscillation
through a stickslip type process during sliding contact.
4. Conclusions
A fully coupled thermo-mechanical-metallurgical nite element model was developed for hot-press forming simulations.
In particular, the effect of prior austenite deformation on phase transformation kinetics was newly accounted to predict the
hardenability of the hot-press formed parts more accurately. For this purpose, an effective austenite grain size (EAGS) con-
cept was introduced to account for the effects of both the inter-granular boundary areas and intra-granular nucleation sites.
The evolution of the EAGS was based on a dislocation density model for the austenite phase. The updated austenite grain size
eld was incorporated into the grain size function of the semi-empirical diffusive phase transformation kineticswhich was
modied to consider the effect of boron additionto model the effect of austenite deformation on the transformation char-
acteristics. In terms of the mechanical eld of the coupled simulation, a predictorcorrector iterative method was developed
to update the stress state. An isotropic yield function and the JohnsonCook-type hardening model, which depends on the
strain, strain rate, and temperature, were used. Regarding the metallurgical eld, transformation plasticity strains, and trans-
formation volumetric and thermal strains were modeled in the developed coupled simulations. The constitutive models
were built through the user material subroutine UMAT with the implicit nite element code, while the metallurgical eld
was coded with UMATHT of ABAQUS/Standard.
A series of validations were conducted through FE simulation and compared to the corresponding experimental results:
(1) dilatometry tests, (2) water quenching of a bar and (3) hot torsion tests. Finally, hot-press forming of the U-channel part
was simulated and analyzed in terms of the phase constituents, strength, and residual stress to study the deformation effect
of prior austenite. The following conclusions were drawn from the present modeling approach:
(1) The modied Lis model for the diffusional transformation and modied KM model for the martensite transformation
could predict the strength and dilatation characteristics of the boron-added steel under various cooling rates. The pre-
dicted results by the modied models were in good agreement with the measurements.
(2) The developed material model that considers prior austenite deformation effect on phase transformation kinetics was
validated using isothermal torsion tests in the austenite state followed by water quenching. The dramatic decrease in
Vickers hardness observed in the experiment was reasonably well captured by the simulation when the grain rene-
ment model (i.e., evolution of the EAGS) was considered.
(3) Hot-press forming of the U-channel was simulated to virtually evaluate the part strength and residual stress distribu-
tions. The simulation results showed that the hardness at the side-wall region was considerably decreased when the
EAGS evolution was turned on. This is mainly because of the enhanced diffusive transformationsequivalently, sup-
pression of martensite in the high dislocation density region due to the prior austenite deformation effect.
(4) A strength gradient was predicted along the sheet thickness direction. The hardness near the blank surface was lower
than that at the center. This was attributed to the accumulated strain gradient in the thickness direction because a
larger plastic strain accumulated near the sheet surfaces from bending and unbending.
(5) The ferrite transformation in the high temperature range induced a larger residual stress at the side wall due to the
less contribution of the transformation plasticity on stress relaxation.
Although the current modeling approach efciently represents the austenite deformation effect on various aspects of hot-
press formed part properties, there should be more validations on real hot-press forming applications; this is planned for
future work.
Acknowledgements
The authors appreciate the support by POSCO. This work was supported by the National Research Foundation of Korea
(NRF) Grant funded by the Korea government (MSIP) (No. 2012R1A5A1048294) and industrial source technology develop-
ment program (#10040078) of MKE.
Appendix A. Mathematical descriptions for austenite grain renement
Partial derivatives for qd in Eq. (2.2):

@ qd b00 _ nb Qb
e
exp C A:1
@ e e_ ;T c00 RT

@ qd _ nb nc _ 1nb nc b00 Qb Qc
1 n n e 1 B e
BD exp A:2
@ e_ e;T
b c
c00 RT

@ qd b00 mb mc _ 1nb nc Qb Qc
d0 e A1 B exp BC A:3
@T e;e_ c00 RT
with

Qc Qb Qb Qc
A exp A:4
RT 2 RT

m Qc
B exp c00 d0 c e_ nc 1 e exp A:5
RT

m Qc
C c00 d0 c e_ nc 1 eT 2 exp A:6
RT

m Qc
D c00 d0 c nc 1e_ nc 2 e exp A:7
RT
Partial derivatives for qrev in Eq. (2.7):

@ qrev T rev T rev nrev
qd crev nrev t n
rev
rev 1
exp exp crev exp t rev A:8
@t rev T T T

@ qrev T rev T rev T rev nrev
qd crev t nrerevv 2 exp exp crev exp trev A:9
@T trev T T T
Evaluation of recrystallized austenite fraction and corresponding grain size in Eq. (2.11):
The Avrami type kinetics for recrystallization according to Sellars (Sellars and Whiteman, 1979; Sellars, 1990) is given by
n
t
X RX 1 exp 0:693 A:10
t 0:5
where the subscript RX stands for the type of reaction, i.e., either SRX or MDRX. The exponent n is 1.5 and 1.0 for MDRX
and SRX, respectively. The parameter t0.5 is the time to reach 50% recrystallization for MDRX and SRX and can be described
as
!
0:8 2:3 105
tMDRX
0:5 1:1Ze_ ; T exp A:11
RT
!
15 2:5 2 2:3 105
tSRX
0:5 2:3 10 e d0 exp A:12
RT
where e and d0 are the strain and the initial AGS, respectively. The ZenerHollomon parameter Z, or temperature-compen-
sated strain rate, is expressed by
!
3:12 105
Z e_ exp A:13
RT
To determine the type of recrystallization, Sellars (1990) introduced a critical amount of strain ec in conjunction with the
strain at a peak stress level epeak, after which the material softens due to DRX. When the amount of cumulative strain exceeds
ec, steels recrystallize dynamically during hot deformation and meta-dynamically after the hot deformation. Otherwise, SRX
takes place after deformation. The critical strain is
ec 0:8epeak A:14
epeak 6:97 104 d0:3

0 Z
0:17
A:15
The rened AGS in a fully recrystallized regime can be obtained by Hodgsons model (1996) for MDRX and SRX as follows:
!
0:5 0:4 4:5 104
dMDRX 343e d0 exp A:16
RT
dSRX 2:6 104 Z 0:23 A:17
Appendix B. Phase fraction updating algorithm
When time step size is smaller than the prescribed critical value, the Forward method is applied as
uk jn1 uk jn Duk jn1 with Duk jn1 u_ k jn tn1 tn B:1

Otherwise, the following Newtons scheme was applied.
uk jn1 uk jn Du k
f uk u
_k B:2
t n1 tn Dt
The residual Rk for each diffusive transformation becomes
Du k
Rk f uk B:3
Dt
By applying the Taylor expansion to the residual after truncation of higher-order terms, the residual during the i-th iteration
is written as
i
i1 @Rk
Rk jn1 Rk jin1 i1
Duk jn1 0 B:4
@ uk n1
i
@Rk
Rk jin1 Duk ji1 with Duk ji1 i1 i
n1 uk jn1 uk jn1 B:5
@ uk n1 n1
Therefore, the increment of phase fraction for the i + 1-th iteration is obtained as
!1
1 @u _ k i
Duk ji1 i
n1 Rk B:6
Dt @ uk n1
and
i1
uk jn1 uk jin1 Duk jn1
i1
B:7
Finally, Eq. (B.7) is substituted into Eq. (B.3) to check if the residual is within the numerical tolerance. If the residual vio-
lates the prescribed tolerance, the iteration is repeated with i = i + 1. Otherwise, the solution is assigned as the initial value
for the next time step n + 2, i.e.,
i1
uk jn1 uk j0n2 B:8
To complete the derivation, the derivative in Eq. (B.6) needs to be calculated for Lis model.

@u_k @g I @g
fk G; Tfk Comp1 g II g I II B:9
@ uk @ uk @ uk
where
n0 1uj 00
g I ujk and g II 1 uj nk uj B:10
with

@g I n0 1uj 1 uj
ujk n0k lnuj B:11
@ uk uj

@g II 00 uj
1 uj nk uj n00k ln1 uj B:12
@ uk 1 uj
In connection with this, the kinetics parameters summarized in Table 7 are applied to Eq. (B.2) for the reactions k = 1, 2, 3,
which are implemented in the corresponding temperature ranges, (Ae3 > T > Bs ), (Ae1 > T > Bs ), and (Bs > T > Ms), respec-
tively. In this study, the equilibrium temperatures Ae3 and Acm are given by (Lusk and Lee, 1999)
Ae3 C C 90:91C C 16:45C Cr 23:5C Mn 10:8C Mo 14:71C Ni 275:89 C Ni 0:76C Ni 1:45C Mn 25:56
C Si 13:53 3:47C Mn 29:96C Mn 8:49C Mo 12:26C Cr 883:49 B:13
Acm C C 977:65 417:57C C 21:36C Cr 1:37C Mn 3:77C Ni 8:1C Si C Cr 2:58C Si 1:5C Cr 35:29
2:76C Mo 30:36C Si 0:95C Mn 217:5 B:14
The bainite start temperature Bs is (Steven and Haynes, 1956)
Bs 656 58C C 35C Mn 75C Si 15C Ni 34C Cr 41C Mo B:15
Appendix C. Stress integration procedure
(1) Initialize state variables:
i 0; ep j0n1 ep jn ; dcj0n1 0; dep j0n1 0 C:1

(2) Compute a trial stress assuming the thermo-metallurgical strain increment to be an initial strain increment:
rtrial jin1 rjn Ce dejn1 detm jn C:2

(3) Check the yield condition:
Utrial jin1 r
rtrial jin1 ry ejin1 ; e_ jin1 ; T n1 > 0 C:3
If satised, go to (4) to calculate plastic strain increments; otherwise, go to (6)
(4) Check the size of the residual, R
q
R kR1 k2 R22 6 Tol C:4
where R1 is the residual vector of the yield surface normal and R2 is the residual of the yield condition as follows
i
e i @ Utrial
R1 de jn1 dc ji1
n1 dejin1 C:5
@ rtrial
n1

rtrial jin1 ry ejin1 ; e_ jin1 ; T n1
R2 r C:6
If Eq. (C.4) meets the prescribed tolerance Tol (=106), go to (6); otherwise,
(5) Compute increment of plastic multiplier dc and correct stress components
i1
dcjn1 dcjin1 Ddcji1
n1 C:7
i i
@ Utrial @ Utrial
dep ji1 i1
n1 dcjn1 deji1
n1 and ep ji1 p i p i1
n1 e jn1 de jn1 C:8
@ rtrial @ rtrial
n1 n1

i1
rtrial jn1 rjn Ce dejn1 detm jn dep ji1
n1 C:9
where
i
trial
R2 jin1 R1 jin1 Njin1 @@Urtrial
Ddcji1
n1 i i
n1
i C:10
@ Utrial i @ Utrial @ ry
@ rtrial
Njn1 @ rtrial @ e
n1 n1 n1
with
0 i 11
@ 2 Utrial
Njin1 @C e1
dc jin1 A C:11
@ rtrial @ rtrial
n1
Go to (4) setting i = i + 1.
(6) Update stresses and corresponding plastic strains:
i1
rjn1 rtrial jn1 and ep jn1 ep ji1
n1 C:12
Exit.
Once this algorithm converged, the consistent tangent modulus (or algorithmic tangent modulus) can be calculated by
differentiating the consistency condition cdU = 0 in plastic state with c > 0 leading to dU = 0 in order to provide a quadratic
convergence rate in solving the global force equilibrium in the ABAQUS/Standard.

@dr Njn1 @@Ur n1
Njn1 @@Ur n1
Cep jn1 Nj C:13
@de n1 n1
@ U
Nj @ U
@r
y
@ r n1 n1 @ r n1 @de
n1
Fig. D1. An example of simulated dilatation applying ferrite-pearlite unied kinetics model.
Appendix D. Further improvement of simulated dilatation at low cooling rate
For the considered boron steel, the simulated dilatations subjected to the low cooling rates show deviations from exper-
iments not only in the transformation start temperature but also in the transforming behavior; i.e., the simulation results in
double peak dilatation, while experiment closes to the monotonous S shape. This problem might be able to be resolved by
assuming a single phase transformation of mixed ferrite and pearlite in their transformation temperature ranges (i.e., Ae3 > -
T > Bs and Ae1 > T > Bs for ferrite and pearlite, respectively). Therefore, ferrite and pearlite are assumed as an aggregation of
them. In this unied ferrite-pearlite kinetics, only the kinetics equation for ferrite (or pearlite) is used as a kinetics of the
aggregation, and calibrated directly to corresponding experimental dilatation. As a result, the simulated dilatation could well
capture the experimental behavior as presented in Fig. D1.
References
kerstrm, P., Oldenburg, M., 2006. Austenite decomposition during press hardening of a boron steel-computational simulation and test. J. Mater. Process.
Technol. 174, 399406.
Andrade-Campos, A., 2010. The inuence of the material heat treatment and phase transformations in forming processes. Int. J. Mater. Form. 3, 11391142.
Arimoto, K., Yamanaka, S., Narazaki, M., Funatani, K., 2009. Explanation of the origin of quench distortion and residual stress in specimens using computer
simulation. Int. J. Microstruct. Mater. Prop. 4 (2), 168186.
Avrami, M., 1939. Kinetics of phase change. I. General theory. J. Chem. Phys. 7, 11031112.
Avrami, M., 1940. Kinetics of phase change. II. Transformation-time relations for random distribution of nuclei. J. Chem. Phys. 8, 212224.
Avrami, M., 1941. Kinetics of phase change. III. Granulation, phase change, and microstructure. J. Chem. Phys. 9, 177184.
Barcellona, A., Palmeri, D., 2009. Effect of plastic hot deformation on the hardness and continuous cooling transformations of 22MnB5 microalloyed boron
steel. Metall. Mater. Trans. A 40A, 11601174.
Bhadeshia, H.K.D.H., Honeycombe, R.W.K., 2006. Steels: Microstructure and Properties, third ed. Elsevier.
Bhattacharyya, S., Kehl, G.L., 1955. Isothermal transformation of austenite under externally applied tensile stress. Trans. ASM 47, 351.
Bok, H.H., Lee, M.G., Kim, H.D., Moon, M.B., 2010. Stress-strain response of low alloy steel at elevated temperature and its practical application to hot
stamping. Met. Mater. Int. 16 (2), 185195.
Bok, H.H., Lee, M.G., Pavlinar, E.J., Barlat, F., Kim, H.D., 2011. Comparative study of the prediction of microstructure and mechanical properties for a hot-
stamped B-pillar reinforcing part. Int. J. Mech. Sci. 53, 744752.
Cahn, J.W., 1956. The kinetics of grain boundary nucleated reactions. Acta Metall. 4, 449459.
Chandran, B.G., Michaleris, P. 1999. Prediction and optimization of microstructure in material processing. In: Proceedings of the Third World Congress of
Structural and Multidisciplinary Optimization. Buffalo, NY.
Choi, B.K., Bok, H., Lee, M.G., Barlat, F., Son, H.S., Chung, K. 2011. Experimental investigation of hot press formed TWB sheet. In: Hirt, G., Tekkaya, A.E. (Eds.),
Proceedings of the 10th International Conference on Technology of Plasticity, Aachen, Germany.
Choi, J.W., Lee, M.G., Barlat, F., Son, H.S., Nam, J.B., 2012. Hot press forming of tailor welded blank: experiments and FE modeling. ISIJ Int. 52 (11), 20592068.
Choi, J.W., Bok, H.H., Barlat, F., Son, H.S., Kim, D.J., Lee, M.G., 2013. Experimental and numerical analyses of hot stamped parts with tailored properties. ISIJ
Int. 53 (6), 10471056.
Desalos, Y. 1981. Comportement dilatomtrique et mcanique de laustnite mtastable dum acier A 533, IRSID Report No. 95349401.
Eghbali, B., 2010. Study on the ferrite grain renement during intercritical deformation of a microalloyed steel. Mater. Sci. Eng., A 527, 34073410.
Esaka, K., Wakita, J., Takahashi, M., Kawano, O., Harada, S., 1986. The development of the precise model predicting and controlling mechanical properties.
Seitetsu-Kenkyu 321, 92104.
Fan, D.W., Kim, H.S., De Cooman, B.C. 2008. Thermo-mechanical analysis of the hot press forming of AHSS. In: International Conference on New
Developments in Advanced High Strength Sheet Steels, Orlando, USA, pp. 325337.
Fernandes, F.M.B., Denis, S., Simon, A., 1985. Mathematical model coupling phase transformation and temperature evolution during quenching of steel.
Mater. Sci. Technol. 1, 838844.
Fischlschweiger, M., Cailletaud, G., Antretter, T., 2012. A mean-eld model for transformation induced plasticity including backstress effects for non-
proportional loadings. Int. J. Plast 37, 5371.
Greenwood, G.W., Johnson, R.H., 1965. The deformation of metals under small stress during phase transformations. Proc. R. Soc. 283A, 403422.
Hillert, M., 1957. The role of interfacial energy during solid state transformations. Jernkontorets Ann. 141, 758789.
Hodgson, P.D., 1996. Microstructure modelling for property prediction and control. J. Mater. Process. Technol. 60, 2733.
Hodgson, P.D., Gibbs, R.K., 1992. A mathematical model to predict the mechanical properties of hot rolled C-Mn and microalloyed steels. ISIJ Int. 32 (12),
13291338.
Hoffmann, H., So, H., Steinbeiss, H., 2007. Design of hot stamping tools with cooling system. Ann. CIRP 56, 269272.
Huang, D., Arimoto, K., Lee, K., Lambert, D., 2000. Prediction of quenching distortion on steel shaft with keyway by computer simulation. In: Proceedings of
20th ASM Heat Treating Society Conference. ASM International, St. Louis, MO, pp. 708712.
Jablonka, A., Harste, K., Schwerdtfeger, K., 1991. Thermomechanical properties of iron and iron-carbon alloys: density and thermal contraction. Steel Res. 62,
2433.
Jepson, M.D., Thomson, F.C., 1949. Acceleration of the rate of isothermal transformation of austenite. J. Iron Steel Inst. 162, 49.
Johnson, W.A., Mehl, R.F., 1939. Reaction kinetics in processes of nucleation and growth. Trans. AIME 135, 416458.
Kern, A., Degenkolbe, J., Msgen, B., Schriever, U., 1992. Computer modelling for the prediction of microstructure development and mechanical properties of
HSLA steel plastes. ISIJ Int. 32 (3), 387394.
Kim, S.I., Choi, S.H., Lee, Y., 2005a. Inuence of phosphorus and boron on dynamic recrystallization and microstructures of hot-rolled interstitial free steel.
Mater. Sci. Eng. A 406, 125133.
Kim, J., Im, S., Kim, H.-G., 2005b. Numerical implementation of a thermo-elasticplastic constitutive equation in consideration of transformation plasticity
in welding. Int. J. Plast 21, 13831408.
Kim, H.Y., Park, J.K., Lee, M.G. 2013. Phase transformation-based nite element modeling to predict strength and deformation of press-hardened tubular
automotive part, Int. J. Adv. Manuf. Technol., in press, http://dx.doi.org/10.1007/s00170-013-5424-9..
Kirkaldy, J.S., Venugopalan, D. 1984. Prediction of microstructure and hardenability in low alloy steels. In: Marder, A.R., Goldstein, J.I. (Eds), International
Conference on Phase Transformation in Ferrous Alloys, AIME, Philadelphia, PA, pp. 125148.
Kirkaldy, J.S., Thomson, B.A., Baganis, E.A. 1978. Hardenability concepts with applications to steel. Kirkaldy, J.S., Doan, D.V. (Eds), AIME, Warrendale, PA, pp.
82125.
Koistinen, D.P., Marburger, R.E., 1959. A general equation prescribing the extent of the austenitemartensite transformation in pure ironcarbon alloys and
plain carbon steels. Acta Metall. 7, 5960.
Kolmogorov, A.N., 1937. Statistical theory of crystallization of metals. Izv. Akad. Nauk SSSR 1, 355359.
Kruglova, A.A., Orlov, V.V., Khlusova, E.I., 2007. Effect of hot plastic deformation in the austenite interval on structure formation in low-alloyed low-carbon
steel. Met. Sci. Heat Treat. 49, 1112.
Kulin, S.A., Cohen, M., Averbach, B.L., 1952. Effect of applied stress on the martensite transformation. Trans. ASM 194, 661.
Lanier, L., Metauer, G., Moukassai, M., 1994. Microprecipitation in boron-containing high-carbon steels. Mikrochim. Acta 144 (115), 353361.
Leblond, J.B., Devaux, J., Devaux, J.C., 1989. Mathematical modeling of transformation plasticity in steels: I. Case of ideal-plastic phases. Int. J. Plast 5, 551
572.
Lee, M.G., Kim, D., Kim, C., Wenner, M.L., Chung, K., 2005. Spring-back evaluation of automotive sheets based on isotropic-kinematic hardening laws and
non-quadratic anisotropic yield functions, part III: applications. Int. J. Plast 21, 915953.
Lee, M.G., Kim, S.J., Han, H.N., Jeong, W.C., 2009a. Application of hot press forming process to manufacture an automotive part and its nite element analysis
considering phase transformation plasticity. Int. J. Mech. Sci. 51, 888898.
Lee, M.G., Kim, S.J., Han, H.N., Jeong, W.C., 2009b. Implicit nite element formulations for multi-phase transformation in high carbon steel. Int. J. Plast 25,
17261758.
Lee, S.J., Pavlina, E.J., Van Tyne, C.J., 2010. Kinetics modelling of austenite decomposition for end-quenched 1045 steel. Mater. Sci. Eng. A 527, 31863194.
Lee, J.W., Lee, M.G., Barlat, F., Kim, J.H., 2012. Stress integration schemes for novel homogeneous anisotropic hardening model. Comput. Methods Appl.
Mech. Eng. 247248, 7392.
Li, M.V., Niebuhr, D.V., Meekisho, L.L., Atteridge, D.G.A., 1998. Computational model for the prediction of steel hardenability. Metall. Mater. Trans. 29B, 661
672.
Lusk, M.T., Lee, Y.K. 1999. A global material model for simulating the transformation kinetics of low alloy steels. In: Proceedings of the 7th International
Seminar of IFHT, Budapest, Hungary, pp. 273282.
Maalekian, M., Lendinez, M.L., Kozeschnik, E., Brantner, H.P., Cerjak, H., 2007. Effect of hot plastic deformation of austenite on the transformation
characteristics of eutectoid carbon steel under fast heating and cooling conditions. Mater. Sci. Eng., A 454455, 446452.
Maccagno, T.M., Jonas, J.J., Yue, S., McCrady, B.J., Slobodian, R., Deeks, D., 1994. Determination of recrystallization stop temperature from rolling mill logs and
comparison with laboratory simulation results. ISIJ Int. 34 (11), 917922.
Mahnken, R., Wolff, M., Schneidt, A., Bhm, M., 2012. Multi-phase transformations at large strains Thermodynamic framework and simulation. Int. J. Plast
39, 126.
Maitrepierre, Ph., Thivellier, D., Tricot, R., 1975. Inuence of boron on the decomposition of austenite in low carbon alloyed steel. Metall. Trans. A 6, 287
301.
Melloy, G.F., Slimmon, P.R., Podgursky, P.P., 1973. Optimizing the boron effect. Metall. Trans. 4, 22792289.
Min, J., Lin, J., Min, Y., Li, F., 2012. On the ferrite and bainite transformation in isothermally deformed 22MnB5 steels. Mater. Sci. Eng. A 550, 375387.
Moumni, Z., Roger, F., Trinh, N.T., 2011. Theoretical and numerical modeling of the thermomechanical and metallurgical behavior of steel. Int. J. Plast 27,
414439.
Nikravesh, M., Nadri, M., Akbari, G.H., 2012. Inuence of hot plastic deformation and cooling rate on martensite and bainite start temperatures in 22MnB5
steel. Mater. Sci. Eng. A 540, 2429.
NUMISHEET93 Benchmark Problem, 1993. In: Makinouchi, A., Nakamachi, E., Onate, E., Wagoner, R.H. (Eds), Proceedings of 2nd International Conference on
Numerical Simulation of 3D Sheet Metal Forming Processes-Verication of Simulation with Experiment. Isehara, Japan.
Park, J.K., Kim, Y.S., Suh, O.S., Lee, M.G., Kim, H.Y., 2013. Improved hot-stamping analysis of tubular boron steel with direct measurement of heat convection
coefcient. Int. J. Automot. Technol. 14, 717722.
Patel, J.R., Cohen, M., 1953. Criterion for the action of applied stress in the martensite transformation. Acta Metall. 1 (5), 531538.
Ryan, N.D., McQueen, H.J., 1986. Mean pass ow stresses and interpass softening in multistage processing of carbon-, HSLA-, tool- and c-stainless steels. J.
Mech. Working Technol 12, 323349.
Saunders, N., Guo, U.K.Z., Li, X., Miodownik, A.P., Schill, J.-P.h., 2003. Using JMatPro to model materials properties and behavior. JOM 55 (12), 6065.
Scheil, E., 1935. Anlaufzeit der austenitumwandlung. Arch. Eisenhuettenwes 12, 565567.
Sellars, C.M., 1990. Modelling microstructural development during hot rolling. Mater. Sci. Technol. 6, 10721081.
Sellars, C.M., Whiteman, J.A., 1979. Recrystallization and grain growth in hot rolling. Metal Sci. 13, 187194.
Senuma, T., Suehiro, M., Yada, H., 1992. Mathematical models for predicting microstructural evolution and mechanical properties of hot strips. ISIJ Int. 32
(3), 423432.
Shi, Z., Liu, K., Wang, M., Shi, J., Dong, H., Pu, J., Chi, B., Zhang, Y., Jian, L., 2012. Effect of non-isothermal deformation of austenite on phase transformation and
microstructure of 22SiMn2TiB steel. Mater. Sci. Eng. A 535, 290296.
Simo, J.C., Ortiz, M., 1985. A unied approach to nite deformation elastoplastic analysis based on the use of hyperelastic constitutive equations. Comput.
Methods Appl. Mech. Eng. 49, 221245.
Simsir, C., Gr, C.H., 2008. 3D FEM simulation of steel quenching and investigation of the effect of asymmetric geometry on residual stress distribution. J.
Mater. Process. Technol. 207, 211221.
Soare, S.C., Barlat, F., 2011. A study of the YLD2004 yield function and one extension in polynomial form: a new implementation algorithm, modeling range,
and earing predictions for aluminum alloy sheets. Eur. J. Mech. A Solids 30, 807819.
Somani, M.C., Karjalainen, L.P., Eriksson, M., Oldenburg, M., 2001. Dimensional changes and microstructural evolution in a b-bearing steel in the simulated
forming and quenching process. ISIJ Int. 41 (4), 361367.
Steven, W., Haynes, A.G., 1956. The temperature formation of martensite and bainite in low-alloy steels: some effects of chemical composition. J. Iron Steel
Inst. 183, 349359.
Suehiro, M., Sato, K., Tsukano, Y., Yada, H., Senuma, T., Matsumura, Y., 1987. Computer modeling of microstructural change and strength of low carbon steel
in hot strip rolling. Trans. Iron Steel Inst. Jpn. 27, 439445.
Sun, W.P., Militzer, M., Bai, D.Q., Jonas, J.J., 1993. Measurement and modeling of the effects of precipitation on recrystallization under multipass deformation
conditions. Acta Metall. Mater. 41 (12), 35953604.
Sun, Z.Q., Yang, W.Y., Qi, J.J., Hu, A.M., 2002. Deformation enhanced transformation and dynamic recrystallization of ferrite in a low carbon steel during
multipass hot deformation. Mater. Sci. Eng. A 334, 201206.
Tamura, I., Ouchi, C., Tanaka, T., Sekine, H., 1988. Thermomechanical Processing of High-Strength Steels. Butterworth & Co. Ltd..
Tekkaya, A.E., Karbasian, H., Homberg, W., Kleiner, M., 2007. Thermo-mechanical coupled simulation of hot stamping components for process design. Prod.
Eng. Res. Dev., 8589.
Totten, G.E. et al, 2004. Modeling and simulation for material selection and mechanical design. In: Suehiro, M. (Ed.), A Mathematical Model for Predicting
Microstructural Evolution and Mechanical Properties of Hot-rolled Steels. Marcel Dekker, New York, p. 14.
Turetta, A. 2008. Investigation of thermal, mechanical and microstructural properties of quenchable high strength steels in hot stamping operations
(Doctoral theses), Universit degli Studi di Padova.
Turetta, A., Bruschi, S., Ghiotti, A., 2006. Investigation of 22MnB5 formability in hot stamping operations. J. Mater. Process. Technol. 177, 396400.
Tzitzelkov, I., Hougardy, H.P., Rose, A., 1974. Mathematical description of the TTT diagram for isothermal transformation and continuous cooling. Arch.
Eisenhttenwes 45, 525532.
Umemoto, M., Ohtsuka, H., Tamura, I., 1984. Transformation to pearlite from work-hardened austenite. Tetsu-to-Hagane 70 (2), 238244.
Umemoto, M., Hiramatsu, A., Moriya, A., Watanabe, T., Nanba, S., Nakajima, N., Anan, G., Higo, Y., 1992. Computer modelling of phase transformations from
work-hardened austenite. ISIJ Int. 32 (3), 306315.
Walker, D.J., Honeycombe, R.W.K., 1978. Effect of deformation on the decomposition of austenite: part I The ferrite reaction. Met. Sci. 12 (10), 445452.
Watt, D.F., Coon, L., Bibby, M., Goldak, J., Henwood, C., 1988. An algorithm for modeling microstructural development in weld heat-affected zones (part A)
reaction kinetics. Acta Metall. 36 (11), 30293055.
Yoo, D.H., Kim, D.G., Ahn, K.H., Kim, H.G., Son, H.S., Kim, G.-S., Chung, K., 2010. Characterization of mechanical properties for hot press forming. Steel Res. Int.
81 (9), 857860.
Yoshie, A., Fujita, T., Fujioka, M., Okamoto, K., Morikawa, H., 1996. Formulation of ow stress of Nb steels by considering work-hardening and dynamic
recovery. ISIJ Int. 36 (4), 467473.
Zener, C., 1946. Kinetics of decomposition of austenite. Trans. AIME 167, 550583.
Zhairi, S.H., Byon, S.M., Kim, S.I., Lee, Y., Hodgson, P.D., 2004. Static and metadynamic recrystallization of interstitial free steels during hot deformation. ISIJ
Int. 44 (11), 19181923.
Zrnik, J., Kvackaj, T., Sripinproach, D., Sricharoenchai, P.P., 2003. Inuence of plastic deformation conditions on structure evolution in Nb-Ti microalloyed
steel. J. Mater. Process. Technol. 133, 236242.

1 s2.0 S0749641913002180 Main

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

1 s2.0 S0749641913002180 Main

Încărcat de

Drepturi de autor:

Formate disponibile

International Journal of Plasticity 58 (2014) 154183

Contents lists available at ScienceDirect

International Journal of Plasticity

Thermo-mechanical-metallurgical modeling for hot-press

2.1. Microstructure evolution in low-alloy steels

2.1.1. Austenite grain renement

b00 (m2) c00 (s1) mb mc nb nc Qb (J/Kmol) Qc (J/Kmol)

qd jn1 qd jnfor dqrev jn1 if t > tfor 2:10

2.1.2. Transformation kinetics models

u_ k fk G; Tfk Comp1 fk uk 2:13

1 1 6.31 1.78 0.31 1.12 2.7 4.06

2.1.3. Modication of transformation kinetics

u4 II 1 expnII DT NII 2:23

Hv 12 H0v 12 a1 T_ 12 ; Hv 3 H0v 3 a2 T_ 3 ; and

n0k T n0k0 n0k1 DT k n0k2 DT 2k 2:28

n00k T n00k0 n00k1 DT k n00k2 DT 2k 2:29

2.1.5. Non-isothermal treatment of diffusive transformation

2.1.6. Microstructure update algorithm for continuous cooling

Fig. 4. Incubation time conversion from TTT to CCT.

2.2. Thermo-mechanical-metallurgical constitutive model

2.2.1. Description of deformation eld

de dee dep detm 2:33

ry k CTk e0 k enTk for k 1; 2; 3; 4 2:42

ET 2:825 105 61:4T 0:684T 2 4:568 104 T 3 MPa 2:46

k C0_k (MPa) C1_k (MPa/C) n0_k n1_k (/C) e 0_k

k DHk (kJ/kg) cpk (J/kgK) k (W/mK)

k n0 k0 n0 k1 (s/C) n0 k2 (s2/C2) n00 k0 n00 k1 (s/C) n00 k2 (s2/C2)

2.3. Description of temperature eld

T_ r  raTT qtrs T; t 0 2:47

2.4. Model summary

A comprehensive material model was proposed for the implementations in thermo-mechanical-metallurgical FE

3. Model validation and application

3.1. Thermo-metallurgical dilatation without external loading

Fig. 7. Schematic of torsion tester and specimen geometry.

3.2. Thermo-metallurgical dilatation under compressive loading

3.3. Distortion of a shaft with a at key way in water quenching

3.4. Quenching of hot deformed rods

3.5. Hot U-draw bending followed by die quenching

3.5.1. Microstructure and strength analyses

Fig. 14. Comparison of effective austenite grain sizes during forming.

3.5.2. Residual stress analysis

Appendix A. Mathematical descriptions for austenite grain renement

Partial derivatives for qd in Eq. (2.2):

epeak 6:97 104 d0:3

dSRX 2:6 104 Z 0:23 A:17

Appendix B. Phase fraction updating algorithm

uk jn1 uk jn Duk jn1 with Duk jn1 u_ k jn tn1 tn B:1

Appendix C. Stress integration procedure

(1) Initialize state variables:

i 0; ep j0n1 ep jn ; dcj0n1 0; dep j0n1 0 C:1

rtrial jin1 rjn Ce dejn1 detm jn C:2

Appendix D. Further improvement of simulated dilatation at low cooling rate

S-ar putea să vă placă și

T_ r raTT qtrs T; t 0 2:47