You are on page 1of 7

15th International Conference on Experimental Mechanics

PAPER REF: 3071

INDENTIFICATION OF JOHNSON-COOK CONSTITUTIVE MODEL


PARAMETERS FOR HIGHT STRAIN RATE DEFORMATION
Xicheng Huang1(*), Jianshi Zhu2, Wenjun Hu1
1
Institute of Structural Mechanics, China Academy of Engineering Physics
2
Institute of Applied Physics and Computational Mathematics, China
(*)
Email: huangxc@caep.ac.cn

ABSTRACT
This work presents a methodology to determine material parameters Johnson-Cook
constitutive model. The direct method and optimization method identifying constitutive
model parameters are proposed. The thermal effect and thermal-softening parameter in the
model are also discussed. The methodology is applied in predicting Johnson-Cook
constitutive model parameters of SS21-6-9.
Keywords: Johnson-Cook model, parameter determination, high strain rate, parameter
identification
INTRODUCTION
The numerical simulation of dynamic deformation of metals have become very important in
many fields such as automotive industry and military application in order to predict dynamic
events in advance and also decrease the cost of experimental investigations. However,
accuracy and reliability of these predictions rely on numerical algorithm, physical
simplification and material constitutive model which describes the flow stress as function of
strain, strain rate, temperature, etc. The five-parameter empirical Johnson-Cook (JC)
constitutive model is widely used in simulation of high strain rate deformation, for instance
hypervelocity impact and explosively driven fracture. The model represents the flow stress
( p , , T ) in terms of three functions: strain hardening f1 ( p ) , strain rate strengthening

f 2 ( ) and thermal-softening f3 (T ) , which can be written as follows (Johnson, 1983, 1993)

( p , , T ) = f1 ( p ) f 2 ( ) f 3 (T ) = ( A + B pn )(1 + C ln  / 0 )(1 T *m )

(1)

where is the equivalent stress; p and  are the equivalent strain and strain rate,
respectively; 0 is a normalizing reference strain rate; A and B are the strain hardening
parameters, whereas C is a dimensionless strain rate strengthening coefficient. Parameters n
and m are power exponents of the strain hardening and thermal softening terms;
T * = (T Tr ) / (Tm Tr ) is homologous temperature, in which Tm is the melting temperature
while Tr refers to a (reference) transition temperature at or below which the response is not
temperature dependent. For the details about JC model and parameters, see a previous study
and the similar work (Liu, 2007; Huang 2010). Determination and identification of JC
constitutive material model parameters considering the deformation characteristics of wide
range of strain rate are crucial. Through universal testing machine and Split Hopkinson
Pressure Bar (SHPB) at different temperature levels and at different stress states, the flow
stress data can be obtained for wide range of strain rate. For the direct method determining
ICEM15

Porto/Portugal, 22-27 July 2012

material parameters A, B and n of strain-hardening function in JC model, the three typical


tests of tension, compression and torsion are suggested. And for the thermal softening
parameter m, the adiabatic heating should be considered. For parameter identification method,
the advanced computational methods such as conjugate direction methods in optimization are
used.
DETERMINATION OF PARAMETERS IN STRAIN-HARDENING FUNCTION

The strain hardening function f1 ( p ) = A + B pn describes the flow stress of material under
conditions of  = 0 and T = Tr . Hence, parameters A, B and n refer to initial yield stress,
hardening modulus and strain hardening exponent respectively under the conditions. For the
different reference strain rate 0 , the different constitutive equation will be obtained. It is
advised that the strain hardening parameters should be determined by quasistatic tests at first
step, then slightly modified by SHPB experiment at the strain rate of practical application, e.g.
103 s-1~104 s-1.
In quasi-static tests uniaxial tensile test and uniaxial torsional test were conducted. If necking
phenomenon occurs in tension, the modification should be made according to Bridgman
formula (Bridgman, 1964). In data processes torsional stress-strain data should be converted
to equivalent stress and strain. Two sets of parameters { A, B, n}Tension and { A, B, n}Torsion can be
obtained from the quasistatic tension and torsion data. The material parameters under
quasistatic condition can be calculated by averaging
A = ( ATension + ATorsion ) / 2

(2)

B = ( BTension + BTorsion ) / 2

(3)

n = (nTension + nTorsion ) / 2

(4)

If the satisfactory data can be obtained from uniaxial compressive test by avoiding the
instability and nonuniform of deformation, the average values can be used
A = ( ATension + Acomp + ATorsion ) / 3

(5)

B = ( BTension + Bcomp + BTorsion ) / 3

(6)

n = (nTension + ncomp + nTorsion ) / 3

(7)

Fig. 1 shows the typical tensile test data of stainless steel 2196. For the material A=451MPa,
B=1951MPa, and n=0.77.
DETERMINATION OF THERMAL SOFTENING PARAMETER

The thermal term in JC model f3 (T ) = [1 T *m ] = {1 [(T Tr ) / (Tm Tr )]m } characterizes the


material softening duo to adiabatic heating, not the environmental temperature. However, we
often use isothermal quasistatic tests at different temperature to determine the temperature
effect on material behaviours. The simple method of determination of parameter m is to
obtain the following discrete points through quasistatic tensile tests at different temperatures

15th International Conference on Experimental Mechanics

f3 (T ) = 1 T *m =

(T * )
(Tr )

(8)

1200

1.0

1000

0.9

0.8

800

Stress ratio f3

True stress/ MPa

By fitting the curve we can obtain the parameter m. Fig. 2 shows the stress ratio f3 (T ) of
stainless steel 2196. For the material m=0.65.

600

0.7

0.6

400

0.5
200

0.4
0
0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.0

0.35

0.1

0.2

0.3

Plastic strain

Fig.1 Tensile test data of SS2169

0.4

0.5

0.6

Fig.1 Stress ratio f 3 (T ) of SS2169

For the material that adiabatic heating is apparent, we should consider the adiabatic
phenomenon. From equation (1) we have

T Tr

m = ln 1
ln

n
( A + B p )(1 + C ln  / 0 ) Tm Tr

(9)

The temperature term T in (9) is calculated as


k
T = T0 +
cp

(10)

Define a parameter Wp as
p

Wp = d p

(11)

Substituting (11) into (9), we have

T0 Tr

m = ln 1
ln
n
( A + B p )[1 + C ln  / 0 ] Tm Tr

kW p
1 +
(T0 Tr ) c p

(12)

If c p is a function with temperature, i.e.


c p = f (T )

(13)

Then temperature T is described as


T = T0 +

ICEM15

kW p

f (T )

(14)

Porto/Portugal, 22-27 July 2012

Under general condition we can not solve the temperature analysis term. Numerical solving
the following non-linear equation can get T
(T T0 ) f (T ) kW p = 0

(15)

Let c p be a linear function with temperature, i.e.


c p = c p 0 + T

(16)

Then we have the expression


T = 12 (T0 c p 0 / ) + (c p 0 / T0 ) 2 / 4 + (c p 0T0 + kW p ) /

(17)

DETERMINATION OF STRAIN RATE STRENGTHENING COEFFICIENT

From equation (1) we have

1
C=
1

(ln  / 0 )
n
*m

(
A
B
)(1
T
)
+

(18)

For given strain rate  , the parameter C could be determined by strain p and corresponding
stress . If we choice p = 0 at room temperature, then we get


C = 1 (ln  / 0 ) 1
A

(19)

or the relation ~ ln()

1 = (ln  / 0 )C

(20)

By dynamic compression, tension and torsion experiments, we can obtain parameters CComp ,
CTension and CTorsion , then average
C = (CTension + CComp + CTorsion ) / 3

(21)

PARAMETER IDENTIFICATION METHOD

In the experimental methods for determining material parameters discussed above, there are
some disadvantages, such as not all points in the experimental data are considered, the lack of
links between each curve. In order to overcome these shortcomings, we use the optimization
algorithm to identify the material parameters, that is, the iterative solution in the parameter
space, to find the optimal point to minimize the deviation between the experimental data with
model predictions(Ralf, 2000). For simplicity, we will not consider the thermal softening
parameter m, and only to identify the parameters A, B, n and C.
Let parameter vector be z = { A, B, n, C}T , the objective function based on stress residual error
is as follows

15th International Conference on Experimental Mechanics

f ( z) =
i =1

di ( z )2
( i )2
exp

(i ) (i ) ( z )
= exp (i )cal
exp
i =1

(22)

(i )
(i )
is experimental value, cal
( z ) model prediction, N the number of experimental
in which exp

data, di the error between experimental and predicted value. The objective is min f ( z ) .
Using the conjugate gradient method (Hestenes, 1980), we have

f ( z ) = 0

(23)

The gradient vector is


f ( z )
f ( z ) f ( z ) f ( z ) f ( z )
g=
,
,
,
=

B
n
C
z j j = A, B , n ,C A

(24)

If the thermal softening is not considered, then we have

= B pn (1 + C ln  )
A

(25)

= pn (1 + C ln  )
B

(26)

= B pn ln p (1 + C ln  )
n

(27)

= A + B pn ln 
C

(28)

And the terms in equation (24) can be written as


N ( A + B ( i ) n )(1 + C ln  ) ( i )
f ( z )
p
exp
= 2(1 + C ln  )
( i )2
A
exp
i =1
(i )
N

( A + B (pi ) n )(1 + C ln  ) exp


f ( z )
(i ) n

= 2(1 + C ln ) p

( i )2
B
exp
i =1

(29)

(30)

(i )
N

( A + B p(i ) n )(1 + C ln  ) exp


f ( z )
(i ) n
(i )
= 2 B(1 + C ln  ) p ln p

( i )2
n

i =1

exp

(31)

(i )
N

( A + B p(i ) n )(1 + C ln  ) exp


f ( z )
(i ) n

= 2 ln ( A + B p )

( i )2
exp
C
i =1

(32)

The calculation flow is as follows.


Step 0: Let z0 n , if f ( z0 ) = 0 then stop;
Step 1: i = 0 , H 0 = Id n (or any nonnegative symmetric matrix), g 0 = f ( z0 ) ;
Step 2: hi = H i gi ;
Step 3: f ( zi + i hi ) = min{ f ( zi + hi ) | 0}| ;
ICEM15

Porto/Portugal, 22-27 July 2012

Step 4: f ( zi + i hi ) ;
Step 5: if f ( zi + i hi ) = 0 , then stop, or else
zi +1 = zi + i hi ; gi +1 = f ( zi +1 ) ; gi = gi +1 gi ; zi = zi +1 zi ;
H i +1 = H i

H i gi H iT gi zi (zi )T
+
;
(gi )T H i gi (zi )T gi

Step 6: i = i + 1 ; goto step 2.


The initial value z0 can be set from the above-mentioned method so that the iteration steps
can be reduced.
We can also set the objective function based on residual error of the interface force in SHPB
system as follows (Shigeru, 2001; Gavrus,2001, 2003)
N

f ( z) =

]2
[ Fi calj ( z ) Fi exp
j
j =1

[F
j =1

exp 2

ij

[F
j =1

cal
oj

( z ) Foexp
]2
j

(33)

[F
j =1

exp 2
oj

where Fi cal ( z ) is the calculated force on interface IB-S, Fi exp ( z ) the experimental value of
the force on interface IB-S, Focal ( z ) is the calculated force on interface OB-S, Foexp ( z ) the
experimental value of the force on interface OB-S. The subscript i and o refer to input and
output; IB-S means input bar and specimen interface 1, OB-S output bar and specimen
interface 2, shown in Fig.3(Sharpe, 2008).

Fig.3 Schematic of SHPB system

RESULTS AND CONCLUSION

The strain rate strengthening parameter C is dependent on selection of reference strain rate.
For the application of moderate or low strain rate, the reference strain rate should be lower;
for high strain rate deformation such as penetration, the reference strain rate should be higher
as 104s-1.
For parameters A, B and n, quasistatic tests of tension and torsion should be conducted, and
value of each parameter is determined on average, e.g. A=(A tension +A torsion)/2.
If the adiabatic heating phenomenon is obvious, the temperature rise duo to the plastic
deformation should be considered. The thermal softening parameter m can be determined by

15th International Conference on Experimental Mechanics

averaging values obtained through dynamic experiments at different initial temperature and at
various strain rates.
The presented methodology is applied in predicting Johnson-Cook constitutive model
parameters of SS21-6-9.
ACKNOWLEDGMENTS

This work was supported by a grant from the National Basic Research Program of China (973
Program) (No. 2010CB832702) and by Science Foundation of China Academy of
Engineering Physics (Grant No. 2009A0201008). We would like to thank professors Hao
Zhiming, Zhao Feng and Hu Haibo at CAEP for many insights they had shared with us into
dynamic material behaviors and numerical modeling methods.
REFERENCES

Bridgman PW. Studies in Large Plastic Flow and Fracture. Harvard University Press,
Cambridge, Massachusetts, 1964
Gavrus A, Caestecker P and Ragneau E. Dynamic compression test analysis using an inverse
method, EUROMECH2001, Cachan, 24-27 september 2001.
Gavrus A, Caestecker P, Ragneau E and Davoodi B. Analysis of the dynamic SHPB test using
the finite element simulation, J. Phys. IV, 2003, France 110.
Hestenes M R. Conjugate direction methods in optimazation, App. Math. 12, Springer-Verlag,
1980
Huang Xicheng. Analysis of Mechanical States and Failure modes of Shells Subjected to
Implosive and Explosive Loadings. Ph Dissertation of CAEP, 2010
Handbook experimental solid mechanics. Ed. William N. Sharpe, Jr., Springer Science +
Business Media, LLC New York, 2008
Johnson GR, Cook WH. A constitutive model and data for metal subjected to large strains,
high strain rates and high temperatures. In: Proceedings of the Seventh Symposium on
Ballistics, The Hague, The Netherlands; 1983
Johnson GR. Material Characterization for Warhead Computations. In Tactical Missile
Warheads, edt by Joseph Carleone, American Institute of Aeronautics and Astronautics,
Inc.,1993, Washington, DC, p768
Liu Xuhong, Huang Xicheng, Chen Yuze, etc.: A review of constitutive models of metals
subjected to intense loading. Advance in Mechanics, 2007. [in Chinese]
Meryers MA. Dynamic behavior of materials. Johhn Wiley & Sons, Inc. 1994
Ralf Mohrmann, Automated identification and calculation of the parameters of nonlinear
material models. In IUTAM symposium on field analyses for determination of material
parameters -- experimental and numerical aspects. Kluwer Academic Publishers, 2003
Shigeru Aoki, Kenji Amaya, Tomohiro Noguchi, Identification of elastic/visco-plastic
material constants under impact loading. Transactions of the Japan Society of Mechanical
Engineers. A, 2001, Vol.67, N0.653, p.1-7

ICEM15

You might also like