Materials Science & Engineering A 634 (2015) 103115

Contents lists available at ScienceDirect

Materials Science & Engineering A

journal homepage: www.elsevier.com/locate/msea

A physically-based constitutive model for SA508-III steel: Modeling

and experimental verication
Dingqian Dong a, Fei Chen a,b,n, Zhenshan Cui a,n
a
b

National Die & Mold CAD Engineering Research Center, Shanghai Jiao Tong University, 1954 Huashan Rd., Shanghai 200030, China
Department of Mechanical, Materials and Manufacturing Engineering, University of Nottingham, Nottingham NG7 2RD, UK

art ic l e i nf o

a b s t r a c t

Article history:
Received 28 September 2014
Received in revised form
10 March 2015
Accepted 10 March 2015
Available online 18 March 2015

Due to its good toughness and high weldability, SA508-III steel has been widely used in the components
manufacturing of reactor pressure vessels (RPV) and steam generators (SG). In this study, the hot
deformation behaviors of SA508-III steel are investigated by isothermal hot compression tests with
forming temperature of (9501250)1C and strain rate of (0.0010.1)s
1, and the corresponding ow
stress curves are obtained. According to the experimental results, quantitative analysis of work
hardening and dynamic softening behaviors is presented. The critical stress and critical strain for
initiation of dynamic recrystallization are calculated by setting the second derivative of the third order
polynomial. Based on the classical stressdislocation relation and the kinetics of dynamic recrystallization, a two-stage constitutive model is developed to predict the ow stress of SA508-III steel.
Comparisons between the predicted and measured ow stress indicate that the established
physically-based constitutive model can accurately characterize the hot deformations for the steel.
Furthermore, a successful numerical simulation of the industrial upsetting process is carried out by
implementing the developed constitutive model into a commercial software, which evidences that the
physically-based constitutive model is practical and promising to promote industrial forging process for
nuclear components.
& 2015 Elsevier B.V. All rights reserved.

Keywords:
SA508-III steel
Work hardening
Dynamic recrystallization
Flow stress
Forging

1. Introduction
Knowledge of the high-temperature deformation behavior of
metals and alloys is very important for the numerical modeling of
many industrial processes. Prerequisite for a successful modeling of
hot working processes by means of numerical techniques, such as
nite element and nite difference methods, one of the most
important items is a precise establishment of the constitutive equations which describe the dependence of the ow stress on strain,
strain rate and temperature. Therefore, an accurate constitutive
equation for the work-hardening and softening behavior is essential.
Meanwhile, the understanding of ow behavior of metals and alloys at
high temperature is of great importance for designers for hot forming
process, such as hot forging, rolling and extrusion. In the last two
decades, considerable investigations have been attempted to develop
constitutive equations of materials from the experimentally measured
data to describe the hot deformation behavior [15].

n
Corresponding authors at: National Die & Mold CAD Engineering Research
Center, Shanghai Jiao Tong University, 1954 Huashan Rd., Shanghai 200030, China.
E-mail addresses: feechn@gmail.com, fei.chen@nottingham.ac.uk (F. Chen),
cuizs@sjtu.edu.cn (Z. Cui).

http://dx.doi.org/10.1016/j.msea.2015.03.036
0921-5093/& 2015 Elsevier B.V. All rights reserved.

RPV and SG are the key components of nuclear power equipment which dominate the lifespan of nuclear power plants. At
present, ASME SA508-III steel has been extensively used as the
materials of RPV and SG due to its high strength and toughness to
prevent failure under severe working conditions [6]. As a key
equipment of the ultra-super-critical generator set, the RPV and SG
should have good mechanical properties through hot forging. In
the past, many investigations have been carried out on the
behavior of SA508-III steel. Kim et al. investigated the failure
behaviors of the weld heat-affected zones, strain aging and fatigue
crack propagation under certain experiment conditions [7]. At the
same time, a new heat treatment process was developed to
improve the toughness for SA508-III steel by Kim et al. [8]. Lee
et al. studied the relationship of the composition, structure and
mechanical properties for the steel [9]. Liu et al. studied the
fracture toughness of SA508-III steel in the temperature range
from room temperature to 320 1C using the J-integral method [10].
Based on the literature reviewed above, it is found that the
previous studies mainly focus on the service performance of
SA508-III steel. However, there are only few reports on the
properties of SA508-III steel during hot forming process. Therefore,
in order to study the workability and to optimize the hot forging
processing parameters, it is highly necessary to investigate the

104

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

Table 1
The chemical composition of SA508-3 forging billet (wt%).
C
0.18

Si
0.17

Mn
1.4

Cr
0.14

Mo
0.51

Ni
0.79

Cu
0.04

S
0.003

P
0.005

V
0.005

Al
0.022

N
0.008

Co
0.0008

As
0.004

Sn
0.0042

O
0.006

H
0.0003

Sb
0.00005

1250C, 300s

Temperature (C)

10C/s
Deformation
30s
10C/s

Deformation temperature:
950-1250C with 100C interval
Strain rate:0.001,0.005,0.01,0.1/s
Strain:0.6

2. Experimental procedures

Water quenching
Time (s)
Fig. 1. Schematic representation of hot deformation process used in the experiments.

In this investigation, the chemical compositions (wt%) of the

commercial SA508-III steel are summarized in Table 1. Cylindrical
specimens were machined with a diameter of 10 mm and a height
of 15 mm. The hot compression tests were carried out on a
Gleeble-3500 thermo-mechanical simulator according to the schedule illustrated in Fig. 1. The specimens were heated to 1250 1C at

160

160

140

140

120

120

100

True stress(MPa)

True stress(MPa)

effects of different thermomechanical parameters on the ow

behaviors for this steel.
This paper mainly focuses on the high-temperature deformation behavior of the steel during hot compression at the temperatures from 950 1C to 1250 1C and strain rates from 0.001 s
1 to
0.1 s
1 using Gleeble-3500 thermo-mechanical simulator. Based
on the experimental results, the ow stress constitutive equations
are established by introducing the classical stressdislocation
relation and the kinematical recrystallization theory. Then through
secondary development of the subroutine, the developed ow
stress model is integrated into commercial software DEFORM3D.
To verify the validity and effectiveness of the developed program,
rstly the isothermal compression tests are simulated to predict
the loading force under experimental conditions, and then a threedimensional hot forging process in the industrial production
environment is performed to compare the loading force obtained
by the numerical simulation by using the developed constitutive model.

950C
1050C
1150C

80
60

1250C

950C
1150C
80

40

20

20
0.1

0.2

0.3

0.4

0.5

0
0.0

0.6

1250C

60

40

0
0.0

1050C

100

0.1

0.2

160

160

140

140

120

120

950C
1050C

100

1150C
1250C

80
60

20
0.3

True strain

0.6

0.4

0.5

0.6

950C

1050C

1150C

1250C

60

20
0.2

0.5

80

40

0.1

0.4

100

40

0
0.0

0.3

True strain

True stress (MPa)

True stress(MPa)

True strain

0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

True strain

Fig. 2. Flow stressstrain curves of the SA508-III steel in different temperatures at the same strain rates (a) 0.001 s
1, (b) 0.005 s
1, (c) 0.01 s
1 and (d) 0.1 s
1.

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

750

750
950C
1050C
1150C
1250C

450
300
150
0
-150
10

20

30

40

50

60

950C
1050C
1150C
1250C

600
Work hardening Rate( )

Work hardening Rate ( )

600

450
300
150
0
-150
15

70

30

45

60

75

90

True stress (MPa)

True stress (MPa)

900
950C
1050C
1150C
1250C

400
300
200
100
0
-100
20

30

40

50

60

70

80

90

100

Work hardening Rate( )

500

Work hardening Rate ( )

105

950C
1050C
1150C
1250C

700

500

300

100

-100
20

40

60

80

100

120

140

True stress (MPa)

True stress (MPa)

Fig. 3. Work hardening rate versus true stress on the strain rate (a) 0.001 s
1, (b) 0.005 s
1, (c) 0.01 s
1 and (d) 0.1 s
1.

a heating rate of 10 1C/s and held for 5 min and then cooled to the
test temperature at the cooling rate of 10 1C/s. Then, the specimens
were held at the forming temperature for 30 s to get a uniform
temperature distribution. The tests were performed at 950 1C,
1050 1C, 1150 1C and 1250 1C and strain rates of 0.001 s
1,
0.005 s
1, 0.01 s
1 and 0.1 s
1, respectively. The true stressstrain
curves were recorded automatically in the isothermal compression
process. All specimens were compressed to a true stain of 0.6 and
then instantly quenched into cold water in order to preserve the
hot deformation microstructure. In order to minimize the friction
between the specimen and the die during hot compression, the
at ends of the specimens were covered by a lubricant consisting
of graphite powder and machine oil. Finally, the quenched specimens were sliced along the axial section. All specimens were
sliced along the axial section and etched with (5 g saturated picric
acid 4 g SDBS 100 mL H2O) solution at 65 75 1C, and then
etching time varied from 10 s to 60 s to reveal austenite grain
boundaries. The optical micrographs were recorded. Microstructure evolution was observed by using Axio Imager M2m (Zeiss).
The grain size was determined by using the line intercept method
in two vertical directions as per ASTM E112-88 [11]. The distance
between adjacent parallel lines was 10 m, and the total true
length of the measuring line exceeded 1000 mm. The average
grain size characterized as equivalent circle diameter was

calculated by multiplying the intercept length by a factor of

(4/)1/2. The grain size distribution was obtained by dividing the
number of grains in a certain range of the average grain diameter
by the total number of measured grains.

3. Experimental result and discussion

3.1. Flow stress behavior
As shown in Fig. 2, a series of typical true stresstrue strain
curves are obtained by hot compression tests under different
strain rates and temperatures. Obviously, the ow stress is
signicantly inuenced by the forming temperature and strain
rate. In the initial deformation stage, the stress increases rapidly
due to the inuence of work hardening (WH), then the increasing
rate of curves decreases with the increase of strain for the
occurring of DRX or dynamic recovery (DRV) until reaching peak
stress p . When the softening rate exceeds hardening rate, the
stress gradually reduces to a steady state due to a new balance
between softening and hardening. According to the characteristics
of these curves, it can be seen that the values of peak stress p
and static ow stress decrease with the increases of deformation
temperature and decrease of strain rate. As shown in Fig. 2,

106

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

-1

-2

-2

-3

-3

-4

-4

ln

ln

-1

-5

-5
-6
-7
-8
3.0

-6

T=950C
T=1050C
T=1150C
T=1250C
3.2

3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

T=950C
T=1050C
T=1150C
T=1250C

-7
-8

5.0

20

40

60

lnp

100

120

140

2.0

-1

1.5

-2

ln(sinh( p))

-3

ln

-4
-5
-6
-7
-8
-1.0

80

p(MPa)

-0.5

0.0

0.5
1.0
ln(sinh( p))

1.0
0.5
0.0

T=950C
T=1050C
T=1150C
T=1250C

-0.5

1.5

-1.0
6.5x10 -4

2.0

0.001s -1
0.005s -1
0.01s -1
0.1s -1
7.0x10 -4

7.5x10 -4

8.0x10 -4

8.5x10 -4

1/T
0

Fig. 4. Plots used for calculation of hot deformation constants (a) , (b) n , (c) n, and (d) value of slope.

36
34
32
30
lnZ

majority of the curves show a single peak stress followed by a

gradual fall to a steady state stress which indicates that the typical
DRX behavior occurs with the increase of strain.
At a given strain rate, the ow stress and the peak strain (p )
clearly drop with the increase of the deformation temperatures
range from 950 1C to 1250 1C, respectively. However, when deformation temperature remains the same value, p and p will
progressively increase when the strain rate increases. At temperatures lower than 950 1C and strain rates higher than 0.1 s
1, p is
not so apparent. When the deformation test is carried out at
1250 1C within the range of the strain rates from 0.001 s
1 to
0.1 s
1, p always appears and the ow stress curves fall fully
under the type of DRX.

28
26

T=950C
T=1050C
T=1150C
T=1250C

24
3.2. Critical strain for DRX
During hot deformation process, it is generally accepted that
only when strain reaches a critical value can DRX occur [12]. The
values of work hardening rate are obtained by using the
following equation [1315]:

i
1 _
j _ ; T: i 1
j ; T:

i 1
i
1

where strain rate _ and temperature (T) are constants in each

ow curve. By using the method proposed by Poliak and Jonas
[13], the critical stress c for initiation of DRX were determined.

22
-1.0

-0.6

-0.2

0.2

0.6

1.0

1.4

1.8

ln(sinh( p))
Fig. 5. Plot used regression determinate the relationship ln Z versus ln sinh p .

The inection point is detected by tting a third order polynomial

to the
curves up to the peak point.

A1 3 A2 2 A3 A4

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

107

-0.5

-1.0
-1.0

-1.5
-2.0

ln c

ln p

-1.5

-2.5

-2.0

-3.0

T=950C
T=1050C
T=1150C
T=1250C

-2.5

-3.0
22

24

26

28

30

32

34

T=950C
T=1050C
T=1150C
T=1250C

-3.5
-4.0
22

36

24

26

28

lnZ

30

32

34

lnZ

2.0
1.5
1.5

ln (sinh( ss ))

ln (sinh(sat ))

1.0
1.0
0.5
0.0

T=950C
T=1050C
T=1150C
T=1250C

-0.5
-1.0
22

24

26

28
lnZ

30

32

34

0.5
0.0
-0.5

T=950C
T=1050C
T=1150C
T=1250C

-1.0

36

-1.5
22

24

26

28

30

32

34

36

lnZ

Fig. 6. Plot used regression determinate the relationship ln p (a), ln c (b), ln sinh sat  (c), and ln sinh ss  (d) versus ln Z.

where A1, A2, A3 and A4 are constants for a given deformation

conditions. Differentiation of Eq. (2) with respect to results in:

3.7

d
3A1 2 2A2 A3
d

d
6A1 2A2
d 2

3.5

ln 0

3.9

3.3
3.1

Then the minimum point of the second derivative of Eq. (4)

corresponds to the critical stress c . When equals to zero, and
the critical stress can be expressed as

2.9
2.7
21

A
c
2
3A1

T=950C
T=1050C
T=1150C
T=1250C
23

25

27

29

31

33

35

lnZ
Fig. 7. Relationship between ln 0 and ln Z.

Then the critical work hardening c can be obtained at the

critical stress point from Eq. (2) as following equation:

2A32
27A21

A2 A3
A4
3A1

0c can be expressed as
0c A3

A22
3A1

108

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

Fig. 8. Microstructure of tested steel at strain rate 0.1 s
1on different deformation temperature (a) 950 1C, (b) 1050 1C, (c) 1150 1C and (d) 1250 1C; (e) strain rate 0.01 s
1
and (f) strain rate 0.001 s
1 at deformation temperature 1150 1C.

It is obvious that the saturation stress s is the intersection of

the tangent extended curve with the line which corresponds to
0. And then the saturation stress expression can be obtained
using linear regression.
!
!

2A32 A2 A3
A22
A2
8

A

A
4
3
3A1
3A1
27A21 3A1

9A1 A2 A3
2A32
27A21 A4

27A21 A3
9A1 A22

A2
3A1

According to this method, the critical points for the initiation of

DRX can be obtained from the derivative of the third order
polynomial versus true stress. Based on the ow curves, the values
of critical strain were determined. The relationship between and

at the same strain rate and different temperatures are illustrated

in Fig. 3. Moreover, the values of characteristic points such as peak
stress p , steady-state stress ss and saturation stress s are
also obtained as shown in Fig. 3.

3.3. The deformation parameters of determination from experiment

results
In hot deformation of metallic materials, the Arrhenius equations are widely used to describe the relationship of the strain
rates, temperature and ow stress [1618]:



Q act
10
_ AF exp
RT

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

where F is a function of ow stress with the following

equations:
8 n0
o0:8
>
<
Z1:2
F exp
11
>
: sinh n for all
where _ is strain rate, R is the gas constant (8.314 J mol
1 K
1), T is
the absolute temperature. Qact is the activation energy of deformation. is the ow stress, and ,,n0 , n are the constants  =n0 .
The effects of the temperature, strain and strain rate on the
deformation behaviors can also be formulated by ZenerHollomon
parameter Z. The change of Z parameter is in accordance with the
variety of the p . The relationship between p and Z is usually
expressed by following equation [1921]:
 

n
Q
A sinh p
Z _ exp
12
RT
In order to determine the constant parameters, p was used as
the input data for the linear regression process. Substitution of
F from Eqs. (10)(12) and taking the natural logarithms for both
sides of the resulting equations, partial differentiation of these
equations yields to the following equations:


ln _
13
n0
ln p T constant


ln _
p T constant

p 1:384  10
3 Z 0:1715

19

c 0:0004  Z 0:19602

20


s 52:687  sinh
1 0:003646Z 0:2092



21

22

The initial stress 0 at various forming temperatures and

strain rates can be directly obtained from the true stressstrain
curves, and in this study it is taken as the initial stress 0 0:2% .
Fig. 7 illustrates that a good linear relationship exists between
ln 0 and ln Z. Then the initial stress 0 can be expressed as a
function of ZenerHollomon parameter (Z),

0 2:1829Z 0:08901

23

22

14

20

ln _


n
ln sinh p

ss 52:687  sinh
1 0:002805Z 0:2070

18
15

16

T constant

As shown in Fig. 4a and b, the slope of ln _ versus ln p and the

slope of ln _ versus p are obtained. The average values are 6.6387


and 0.1261 for n0 and , respectively. The value of  =n0 is 0.019.
There is a linear relationship between ln sinh p and ln _ when
T remains constant. In Fig. 4c, the value of
 n can be calculated from
the relationship of ln _ and ln sinh p , and the average value of
n is 4.795.
The hot deformation Qact can be regarded as a signicant index
of the deformation difculty degree. The Qact value of SA508-III
steel can be expressed as following:




ln sinh p
16
Q act Rn
1=T
_ constant


According to the relationship of ln sinh p versus 1=T, the
linear slope increases with the increase of strain rate, the average
value of these slopes is 9434.037 as shown in Fig. 4d. According to
Eq. (16), the Qact of SA508-III steel during hot deformation can be
obtained as 376,088 J/mol.


Fig. 5 illustrates the relationship between ln Z and ln sinh p .
Eq. (17) can be obtained by taking the logarithm of both sides of Eq.
(12).


ln Z n ln sinh p ln A
17
By regression from the resultant equation in Fig. 5, a new
equation was obtained as following:
!
376:088  103
Z _ exp
8:5983  1011
RT

4:6899
 sinh0:019 p
18
In order to establish the expression of feature parameters, the
values of characteristic points such as peak strain p , critical
strain c , saturation stress s and steady-state stress ss can be
easily obtained from the ow stress curves. Fig. 6 ad shows that
the linear relation exist between ln p , ln c , lnsinh s , and

lnki

"

lnsinh ss versus ln Z, respectively. So, by tting the result of

the experiment in Fig. 6, p , c , s and ss can be expressed as a
function of the ZenerHollomon parameter (Z), respectively.

14

lnk1

12

lnk2
k1 value Line Fit

10

k2 value Line Fit

8
6
4
2
0
22

24

26

28

30

32

34

36

lnZ
Fig. 9. Plot used regression determinate the relationship between ln ki and ln Z.

6
5
4
3
ln(-ln(1-Xdrx))

109

T=1050C,0.01s
T=1150C,0.01s
T=1250C,0.01s
T=1050C,0.1s
T=1150C,0.1s
T=1250C,0.1s

T=1050C,0.001s
T=1150C,0.001s
T=1250C,0.001s
T=1050C,0.005s
T=1150C,0.005s
T=1250C,0.005s

1
0
-1
-2
-3
-4
-5
-6
-3

-2

-1

0
ln((c)/p)

Fig. 10. Linear t value of kd and nd by regressed DRX parameters.

110

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

4. Model establishment for dynamic recrystallization

4.1. DRX microstructures
The microstructural evolution of hot deformed SA508-III steel
were investigated at a strain of 0.6 under various forming
temperatures and strain rates. Fig. 8ad illustrates the optical
microstructures after deformation at strain rate of 0.1 s
1
under the temperatures of 950 1C, 1050 1C, 1150 1C and 1250 1C,
respectively. It is obvious that the deformed austenite grain size
increases with the increase of the forming temperature. The
average grain sizes were measured as 35 m, 53 m, 81 m and
117 m for the forming temperatures of 950 1C, 1050 1C, 1150 1C
and 1250 1C, respectively. The measured grain size results agree
reasonable with the characteristic of the ow stress curve. As
shown in Fig. 2d, it can be seen that the ow stress peak is not so
apparent at strain rate 0.1 s
1 at 950 1C, and the ow stress curve
exhibits the characteristics of dynamic recovery, and then the
dynamic recrystallization does not occur. However, it can be
observed that the dynamical recrystallization occurs with the
increase of the forming temperature. The main reason for this
phenomenon is that higher forming temperatures lead to the
increase of deformation stored energy. And then the dynamical

80

4.2.1. Kinetics of DRX

DRX is an important phenomenon for controlling microstructure and mechanical properties in hot working. The onset of DRX
mainly depends on the distribution and density of dislocation
during deformation. Based on the classical stressdislocation

100

-1

0.1s
-1
0.01s
-1
0.005s
-1
0.001s

60

40

20

0
0.0

4.2. Mathematical Model for DRX

Volume fraction of DRX, Xdrx

Volume fraction of DRX, Xdrx

100

recrystallization grain size is rened with decreasing the deformation temperature.

Fig. 8c, e and f illustrates the optical microstructures of test
steel after deformation at 1150 1C under strain rates of 0.1 s
1,
0.01 s
1 and 0.001 s
1, respectively. It can be easily found that the
higher the strain rates, the ner the grain sizes, which were
measured as 81 m, 113 m and 144 m under the strain rates
of 0.1 s
1, 0.01 s
1 and 0.001 s
1, respectively. It can be clearly
seen that the recrystallized grain size becomes ner with the
increase of the strain rate. This phenomenon is mainly attributed
to two reasons [12]: higher strain rate increases the work hardening rate and reduces the time for austenite recrystallization,
which leads to insufcient time for austenite grain to grow up at a
certain strain, in other words the recrystallized nucleus under a
higher strain rate has insufcient time to fully grow up.

0.1

0.2

0.3

0.4

0.5

80

60

40

20

0
0.0

0.6

-1

0.1s
-1
0.01s
-1
0.005s
-1
0.001s

0.1

0.2

Strain

80

60

40

20

0
0.0

0.1

0.4

0.5

0.6

100

-1

0.1s
-1
0.01s
-1
0.005s
-1
0.001s

Volume fraction of DRX, Xdrx

Volume fraction of DRX, Xdrx

100

0.3
Strain

0.2

0.3
Strain

0.4

0.5

0.6

80

60

40
-1

0.1s
-1
0.01s
-1
0.005s
-1
0.001s

20

0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

Strain

Fig. 11. Dynamic recrystallization fraction for SA508- steel under the different strain rates with deformation temperature: (a) 950 1C, (b) 1050 1C, (c) 1150 1C and
(d) 1250 1C.

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

160

160

Predicted Strain rate:0.001s-1

Experimental

140

120

True stress(MPa)

True stress(MPa)

140

950C

100

1050C
80

1150C
1250C

60

1150C

0.4

0.5

0
0.0

0.6

0.1

0.2

140

1150C

1250C

80
60

80
60

20

20

0.2

0.3

0.4

0.5

0.6

1150C 1250C

100

40

0.1

0.6

120

40

0.0

Predicted Strain rate:0.1s

Experimental 950C

1050C

1050C
100

0.5

-1

950C

120

0.4

160

-1

Predicted Strain rate:0.01s

Experimental

True stress (MPa)

True stress(MPa)

140

0.3

True strain

True strain
160

1250C

60

20
0.3

1050C

80

20
0.2

950C

100

40

0.1

Predicted Strain rate:0.005s-1

Experimental

120

40

0
0.0

111

0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

True strain

True strain

Fig. 12. Comparisons experimental and predicted ow stress value of SA508- steel under various hot deformation conditions.

relation and the kinematical recrystallization proposed by Laasraoui and Jonas [22,23], the ow stress constitutive equations of
the WH-DRV period and DRX period were established, respectively. According to effects of the WH and DRV, the evolution of
the dislocation density with strain is generally considered to
depend on the following two components [2426]:
p
d
k1
k2
d

24

where d=d is the rate of increase of dislocation density with

strain, k1 represents the coefcient of WH, and k2 is the coefcient
of DRV. When 0, 0 , where 0 is the initial dislocation
density.

2
k2
k
k
p k2
1
1 e
2 0 e
2
25
k2 k2
when
 2
d
k1
0; s
d
k2

26

s is the saturation dislocation density corresponding to the

saturation stress s . Previous studies have shown that the effective
stress is negligible compared to the internal stress at high

Fig. 13. Correlation between the experimental and predicted ow stress data from
the developed constitutive equation.

temperature and the mechanical strength of the obstacles to

dislocation glide, so the applied stress can be related directly to

112

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

Start
Input data
t=t+t
Read simulation results
Strain increment
Yes

No

Yes
No
Output data

End
Fig. 14. Flowchart of the secondary development of the ow stress subroutine.

WH s 0
s e

k

22

; o
c

27

According to the strain and stress determined by the experiment ow stress curves, k2 can be calculated under deformation
conditions. Therefore, Eq. (27) is derived as


2
0
s
28
k2 ln

WH
s
p
k1 can also be calculated from Eq. (26) and b .
Considering the shear modulus changes with deformation
temperature, the equation [28] can be expressed as
0
k1 @

0 1
0:91 

T
300
Tm

A s k2

29

25000
Strain rate:0.1s-1
20000
Load Force, N

the square root of the dislocation density [14]. The structural

p
parameter can be selected as b . is the material
constant, is the shear modulus and b is the distance between
atoms in the slip direction [27]. Therefore, considering the impact
of WH and DRV during hot deformation, the ow stress as a
function of strain can be given as the following expression.

15000
1050C
10000
1150C
5000
1250C
0
0.0

k2 A1 Z n1

1.5

3.0

k2 319:1847Z
0:0788

31

At high temperatures and low strain rates the DRX phenomenon is more obvious. The volume fraction of DRX (Xdrx) can be
determined by [2527,29]:


 

c ndrx
X drx 1
exp
kdrx
; Z c

32

6.0

7.5

Fig. 15. Variation of anvil load force with stroke for SA508-III steel under strain
rates 0.1 s
1with deformation temperature: (a) 950 1C, (b) 1050 1C, (c) 1150 1C and
(d) 1250 1C.

Meanwhile, the progress of the DRX, Xdrx, can also be written as

WH

; Z c
s
ss

33

By combining Eqs. (28) and (29), the ow stress during DRX

period can be given by the following expression:




c ndrx
; Z c
WH
s
ss 1
exp
kdrx

34

30

Regression equation is obtained as follows:

4.5
Stroke,mm

X drx
where 0 is the shear modulus when temperature is 300 K, Tm is
melting point. Therefore, regression model can be established in
the Eq. (24). A1 and n1 are constants, so the value of k1, k2 can be
determined in Fig. 9.

950C

Experiment
Simulation

where X drx is the volume fraction of DRX. kdrx and ndrx are DRX
parameters depending on chemical composition and hot deformation conditions.
Then, Eq. (32) can be rewritten as following,



c
ln
ln1
X drx ln kdrx ndrx ln
35



The relationship between ln
ln1
X drx and ln
c =p
is used to determine kdrx and ndrx . As shown in Fig. 10, the kinetic
model of DRX can be acquired with the average value of kdrx and

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

113

Fig. 16. The forming process of experiment (a) preheating of the die, (b) initial forging state, (c) nal 80% reduction forming and (d) dimension measurement.

Table 2
Simulation conditions for the hot upsetting experiments.
Process condition

Value of parameter

Initial temperature of workpiece

Initial temperature of dies
Material of dies
Number of workpiece units
Number of Up-die units
Number of Bottom-die units
Thermal conductivity of environment
Thermal conductivity between die and workpiece
Friction factor
Upper die velocity
Increment of time
Increment of step

1200 1C
300 1C
H13
56018
16958
16866
0.02 N/s/mm/1C
5 N/s/mm/oC
0.3
2 mm/s
0.1 s/step
0.5 mm/step

summarized as
8
k

22
>
; o c
>
> WH s 0
s e
>
n
h
io
>
>
c ndrx
>
; Z c
> WH
s
ss 1
exp
kdrx

>
p
>
>
>
>
0:08901
>
0 2:1829Z
>
>
>
>
>
c 0:0004  Z 0:19602
>
>
>
>
>
< p 0:001384  Z 0:17152 ;



1
>
0:003646Z 0:2092
> s 52:687  sinh
>
>


>
>
>
>
ss 52:687  sinh
1 0:002805Z 0:2070
>
>
>
>
>
>

0:0788
>
>
> k2 319:1847Z
>
>
>
kdrx 0:6949; ndrx 1:7916
>
>
>


>
: Z _ exp 376088
RT

37

ndrx which are 0.6949 and 1.7916, respectively.



c
X drx 1
exp
0:6949
P

1:7916 !
36

Fig. 11 shows the dynamic recrystallization volume fraction of

SA508-III steel calculated by Eq. (36) under different strain rates
and temperatures. As shown in the gures, the volume fraction of
DRX increases with the increasing of the strain. For a certain strain,
the volume fraction of DRX is higher at higher deformation
temperature or lower strain rate. Therefore, under the formation
conditions such as higher deformation temperatures and lower
strain rates, DRX tends to be complete; in other words, the volume
fraction of DRX is close to 100%.
Therefore, the physical-based constitutive relation of SA508-III
steel during the WH-DRV period and DRX period can be

4.2.2. Verication of the developed constitutive equation

In order to verify the developed constitutive models, comparisons between the experimental data and predictions have been
performed in Fig. 12ad. It can be easily found that there is a good
agreement between the predicted value and the experimental
data. It illustrates that the proposed constitutive equation gives a
reasonable estimate of the ow stress for this steel.
In order to further evaluate the prediction accuracy of the
developed constitutive model, the predictability of the developed
equation can be quantied in terms of standard statistical parameters such as correlation coefcient (R) and the average absolute
relative error (AARE) as shown in Fig. 13. The AARE is calculated by
comparing the relative errors and therefore is an unbiased
statistical parameter for determining the predictability of the
equation. The correlation coefcient (R) provides information on

114

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

Fig. 17. The results of numerical simulation (a) strain-effective and (b) stress-effective.

1200000
Experiment
Simulation

Load Force, N

1000000
800000
600000
400000
200000
0

10

20
30
Stroke,mm

40

50

Fig. 18. Comparison between the load of FEM simulation and the experimental
results.

the strength of linear relationship between the experimental and

the predicted values. These are expressed as following equations:
PN
i 1 X i
XY i
Y
q
R q
38
PN
PN
2
2
X
i
X
i1
i 1 Y i
Y
AARE



1 XN Y i
X i 
  100%
i1 X
N
i

39

where Xi, Yi are the measured and predicted ow stress, respectively. X, Y are the mean values of Xi and Yi, respectively. N is the
total number of data used in this investigation. The calculated
correlation coefcient (R) is 0.997, which indicates that there is a
good correlation between the predicted and measured data.
Meanwhile, the average absolute relative error (AARE) is only
3.26%, which illustrates the good prediction capability of the
developed model. Therefore, the developed constitutive equation
can be applied to numerical simulation for describing the ow
behavior of the SA508- steel under the elevated forming
temperature.
4.3. Experimental verication
In order to carry out numerical simulation of hot forging, the
developed consitituve euqations were implemented into three
dimensional DEFORM commercial software. Fig. 14 gives the
owchart of secondary development of the ow stress subroutine.

To verify the validity and effectiveness of the developed program,

the nite element method simulations were carried out to
calculate the loading force under different experimental conditions. Fig. 15 shows the variation of loading force with stroke
under four different deformation temperatures, 950 1C, 1050 1C,
1150 1C and 1250 1C at the strain rate of 0.1 s
1. As shown in the
gure, the maximum relative error is about 5.44% with deformation temperature 950 1C at strain 0.6. Generally speaking it
illustrates that the loading force simulated by FEM ts in well
with the results of compression test. In other words, the simulation results conrm that the developed constitutive equations well
describe the ow behavior of SA508-III steel during hot forming
process.
Further, a three-dimensional hot upsetting experiment in the
industrial production environment was performed. The cylinder
workpiece with a diameter of 40 mm and a height of 60 mm was
used in the experiment. The workpiece was heated to 1200 1C and
held for 300 min to obtain a homogeneous initial temperature by
using the vacuum furnace. The 320 ton capacity hydraulic press
was adopted in the experiment and the experiment was carried
out at a ram speed of 2 mms
1 till 80% of reduction was reached.
Fig. 16ad illustrates the whole hot forging processes including
preheating of the die, initial forging state, nal forging state and
dimension measurement. The experimental conditions and FEM
simulation parameters for the upsetting process are listed in
Table 2. Fig. 17a and b shows the strain and stress elds obtained
by using DEFORM3D. The comparison between the load of the
updie during FEM simulation and the experiment load is shown in
Fig. 18. Clearly, there is a good agreement between the results
obtained by experiment and FEM simulation. It demonstrates that
the proposed deformation constitutive equations can be used to
analyze the numerical simulation during hot forging process for
SA508- steel in the industrial production environment.

5. Conclusions
In this study, the hot deformation behavior of SA508-III steel
has been investigated by means of the compression test over a
practical range of temperatures and strain rates. The ow stress
constitutive equations for the work hardening-dynamical recovery
period and dynamical recrystallization period were developed for
SA508-III steel. By comparing the predicted and measured results,
it is conrmed that the proposed constitutive equations can be
used in numerical simulation in hot forming process of SA508-III
steel. By integrating the developed ow stress model into DEFORM3D, a three-dimensional hot upsetting experiment in the

D. Dong et al. / Materials Science & Engineering A 634 (2015) 103115

industrial production environment was simulated. It indicates that

the developed numerical simulation platform is very effective to
be used in heavy forging industrial application for SA508-III steel.
Acknowledgments
This work was nancially supported by National Basic Research
Program of China (Grant no. 2011CB012903) and National Science
and Technology Major Project (Grant no. 2011ZX04014-051).
References
[1] X.J. Duan, T. Sheppard, Mater. Sci. Eng. A 351 (2003) 282.
[2] J.R. Cho, H.S. Jeong, D.J. Cha, W.B. Bae, J.W. Lee, J. Mater. Process. Technol. 160
(2005) 1.
[3] W.G. Jiang, G.C. Wang, S.Q. Lu, J.W. Li, J. Mater. Process. Technol. 182 (2007)
274.
[4] Y.C. Lin, M.S. Chen, J. Mater. Process. Technol. 209 (2009) 4578.
[5] L.Z. Wu, X.S. Li, J. Chen, J. Iron Steel Res., Int. 17 (2010) 51.
[6] B.S. Lee, M.C. Kim, J.H. Yoon, J.H. Hong, Int. J. Pres. Ves. Pip. 87 (2010) 74.
[7] Y.S. Ahn, H.D. Kim, T.S. Byun, Y.J. Oh, G.M. Kim, J.H. Hong, Nucl. Eng. Des. 194
(1999) 161.

115

[8] B.J. Lee, H.D. Kim, J.H. Hong, Metall. Mater. Trans. A 29 (1998) 1441.
[9] K.H. Lee, S.G. Park, M.C. Kim, B.S. Lee, Mater. Sci. Eng. A 534 (2012) 75.
[10] J. Liu, L. Wang, Y. Liu, X. Song, J. Luo, D. Yuan, Int. J. Miner., Metall. Mater. 21
(2014) 1187.
[11] Y. Estrin, Unied Constitutive Laws of Plastic Deformation, Academic Press,
New York (1996) 69.
[12] F. Chen, Z.S. Cui, S.J. Chen, Mater. Sci. Eng. A 528 (2011) 5073.
[13] E.I. Poliak, J.J. Jonas, Acta Mater. 44 (1996) 127.
[14] Y.C. Lin, X.M. Chen, Mater. Des. 32 (2011) 1733.
[15] E.I. Poliak, J.J. Jonas, ISIJ Int. 43 (2003) 684.
[16] Y. Deng, Z.M. Yin, J.W. Huang, Mater. Sci. Eng. A 528 (2011) 1780.
[17] A. Laasraouia, J.J. Jonas, Metall. Trans. A 22 (1991) 151.
[18] C. Zener, H.J. Hollomon, Appl. Phys. 15 (1944) 22.
[19] C.M. Sellars, Acta Met. 14 (1966) 1136.
[20] J. Jonas, C.M. Sellars, Int. Metal. Rev. 14 (1969) 1.
[21] H. Shi, A.J. McLaren, C.M. Sellars, R. Shahani, R. Bolingbroke, Mater. Sci.
Technol. 13 (1997) 210.
[22] A. Laasraouia, J.J. Jonas, Metall. Mater. Trans. A 22 (1991) 1545.
[23] H. Merking, U.F. Kocks, Acta Metall. 29 (1981) 1865.
[24] U.F. Kocks, A. Argon, M. Ashby, Prog. Mater. Sci. 19 (1975) 1.
[25] C.M. Sellars, Acta Met. 14 (1966) 1136.
[26] J.H. Beynon, C.M. Sellars, ISIJ Int. 32 (1992) 359.
[27] Z.Y. Zeng, et al., J. Mater. Sci. Technol. 27 (2011) 913.
[28] N.M. Xiao, C.W. Zheng, D.Z. Li, Y.Y. Li, Comput. Mater. Sci. 41 (2008) 366.
[29] Y.C. Lin, X.M. Chen, D.X. Wen, M.S. Chen, Comput. Mater. Sci. 83 (2014) 282.