Numerical Modeling of Elasto - Viscoplastic Chaboche Constitutive Equations Using MSC - Marc
Numerical Modeling of Elasto - Viscoplastic Chaboche Constitutive Equations Using MSC - Marc
Numerical Modeling of Elasto - Viscoplastic Chaboche Constitutive Equations Using MSC - Marc
NUMERICAL MODELING OF
ELASTO-VISCOPLASTIC CHABOCHE
CONSTITUTIVE EQUATIONS USING
MSC.MARC
ANDRZEJ AMBROZIAK
Department of Structural Mechanics,
Faculty of Civil and Environmental Engineering,
Gdansk University of Technology,
Narutowicza 11/12, 80-952 Gdansk, Poland
[email protected]
(Received 22 October 2004; revised manuscript received 17 January 2005)
Abstract: The aim of the present paper is to propose a special kind of finite element procedure
for dynamic and static analysis of civil engineering structures (e.g. trusses, beams, shells and spatial
structures) including elasto-viscoplastic models. The Chaboche model with damage has been chosen
from the wide range of available elasto-viscoplastic constitutive models for numerical calculations.
The main advantage of the presented approach is the possibility of interference in subroutines and
adjusting them to a particular problem (e.g. altering the form of constitutive equations). A userdefined UVSCPL subroutine has been proposed to introduce the elasto-viscoplastic model into the
MSC.Marc system.
Keywords: elasto-viscoplastic constitutive models, damage, finite element method
1. Introduction
In materials where the viscous properties within the plastic range play an
important role it is necessary to apply the elasto-viscoplastic constitutive relations. We
can cite the following elasto-viscoplastic models for example: Perzyna [1], Chaboche
[2], Aubertin [3], Lehmann-Imatani [4], Miller [5], Bodner-Partom [6], Krempl [7],
Tanimura [8], Krieg-Swearengen-Jones [9], Walker [10], Korhonen-Hannjula-Li [11],
Freed-Virrilli [12]. A detailed description of these models can be found in paper [13].
Their relatively large number shows that they are not universal and can be applied
only under certain conditions or for a limited range of materials. From the wide
range of available elasto-viscoplastic laws, the Chaboche model with damage has
been selected in the present paper. This constitutive model is suitable for analysis
of isotropic materials (mainly metals at high temperatures) in the range of small
inelastic strains.
tq209b-e/157
11 V 2005
158
A. Ambroziak
(1)
so that it is necessary to assume small strains. The relation between the stress and
strain rates can be described for an isotropic material as:
(2)
= (1 D) B : E = B : I ,
(3)
where p,
S0 and X0 are the rate of the equivalent plastic strain (the dots denote
differentiation with respect to time) and the deviatoric parts of stress and back stress
tensors, respectively. The J(a) invariant is defined by the following equation:
r
r
3
3 ij
J(a) =
a:a=
a aij .
(4)
2
2
We will consider the isotropic damage expressed by the scalar parameter 0 D 1.
The p rate in the Chaboche model with damage is defined in [15] as:
+n
* J(S0 X0 )
Rk
1D
p =
,
(5)
K
where k, D and K, n are the initial yield stress, the damage parameter and material
parameters, respectively. The angle bracket hxi has the following form:
1
hxi = (x + |x|)
2
and is called the McCauley bracket.
tq209b-e/158
11 V 2005
(6)
159
(7)
R = b(R1 R) p,
2
= a I cX p,
X
(8)
3
where a, c, b and R1 are material parameters.
According to the damage model proposed by Kachanov [17], the damage
evolution D used in Equation (5) is expressed by the following equation (see [15]):
s
Y
D=
p,
(9)
S
where S, s are the material parameters of the damage and the Y function is given as:
2
1
2
2
(1 + )eq + 3(1 2)H ,
(10)
Y =
2(1 D)2 E 3
where E is Youngs modulus, is Poissons ratio, and eq and H are the HuberMisses equivalent stress and hydrostatic stress.
In the case of an uniaxial tension or compression state, the inelastic strain rate
can be written as:
+n
* |X|
k
1D
I = p sign( X) =
sign( X)
(11)
K
while kinematic hardening has a simple form, [2]:
X = a I cX | I |.
(12)
It is worth noting that the stress tensor, S, has one non-vanishing component,
, while the S0 and X0 tensors have three non-zero components:
2
2
0
0
X
0
0
0 0
3
3
S = 0 0 0 , S0 = 0 13
0 , X0 = 0
0 . (13)
13 X
1
1
0
0
3
0
0
3X
0 0 0
4. Numerical examples
Numerical calculations have been performed for simple solid (Figure 2) and
truss (Figure 3) structures. Such tests are usually performed in a laboratory during
tq209b-e/159
11 V 2005
160
A. Ambroziak
t
j ,
Xj1 + X
2
t
[step 1]
R = 2 Rj1 + Rj ,
t
j ,
Dj1 + D
D =
2
[step 2] S0j , X0j , J(S0j X0j ), J(S0j ),
X =
"
"
S0j X0j
3
p j
2 J(S0j X0j )
[step 3] p j =
[step 4] Ij =
Xj = Xj1 + X
Rj = Rj1 + R
Dj = Dj1 + D
tr(Sj )
n #
j = 2 a I cXj p j
[step 5] X
3 j
i
h
j = b(R1 Rj ) p j
[step 6] R
s
j = Y p j
[step 7] D
S
[step 8] Bj = (1 Dj )B
[step 9] Ij = Ij tj
[step 10] Sj = Bj (j Ij )
tq209b-e/160
11 V 2005
161
and thus the strains tend to be constant throughout the element. Linear interpolation
was used in the latter for coordinates and displacements. At the same time, only the
straight two-nodes truss element with simple linear interpolation was used in the
OFC program.
The geometrical parameters of l = 1.0m (length) and A = 0.001m2 (square
cross section area for the truss and solid elements) were assumed. Due to the purely
numerical character of the work the values of the material parameters were taken
from the literature. The author is aware that sometimes different sets of parameters
can be found in the literature for the same material under the same conditions due
to the strong physical nonlinearity of the problem (see [20]). In practical applications
such sets must be carefully verified.
tq209b-e/161
11 V 2005
162
A. Ambroziak
between the results obtained from the two computer programs and the types of finite
elements used in the numerical calculations.
tq209b-e/162
11 V 2005
163
Figure 7. Change of the damage parameter in the constant stress rate tests
the analysed range of strain and stress, damage of the structure and a break in the
numerical calculations was observed in the case of the = 0.01s1 computation. The
evolution of the damage parameter is given in Figure 7; for the = 0.01s1 calculation
damage exceeds the value limit. The simulations of creep tests are presented in
Figures 8 and 9 and for the stress values of = 1000MPa and = 1500MPa. In
this case, the damage has great influence for = 2000MPa. Finally, a diagram of the
cyclic load test for = 0.001s1 is shown in Figure 10. The cyclic test was carried out
in the range of = 0.03. Like in the first example, there is no significant difference
between the results obtained from the two computer programs. It is worth pointing out
the additive character of the damage parameter. Its value increases at each successive
step, expect for the elastic range of deformation, where its value is constant due to
tq209b-e/163
11 V 2005
164
A. Ambroziak
the lack of inelastic effects (see Figures 11 and 12). The failure of the specimen after
a single cycle of loading is attributable to the relatively high strain rate. An error in
the determination of material constants is another possible explanation. For example,
the material parameter K has a very high value compared to other references in the
literature (cf. e.g. [22, 23]).
5. Conclusions
The following conclusion and remarks may now be formulated:
(1) Thanks to the possibility of including users constitutive model subroutines
into the MSC.Marc system, the Chaboche model with damage can be directly
applied in calculations.
tq209b-e/164
11 V 2005
165
Figure 11. Change of the damage parameter of the cyclic test in the time domain
Figure 12. Change of the damage parameter of the cyclic test in the strain domain
(2) A non-linear analysis of simple solid and truss structures with elastic-viscoplastic physical equations for the Chaboche model has been successfully carried
out.
(3) The MSC.Marc and OFC programs respond almost identically in all the calculated test variants.
(4) The obtained results require a revision for more complex structures, types of
analysis and kinds of load. Especially the influence of layering in the shell
elements should be investigated.
(5) Further research should focus on the development of the UVSCPL routine with
an emphasized influence of temperature.
tq209b-e/165
11 V 2005
166
A. Ambroziak
Acknowledgements
The research was performed within the Polonium 2005 Polish-French coopera
tion program administrated by KBN and EGIDE
(KBN 5598.II/2004/2005).
Calculations presented in the paper were performed at the Academic Computer
Centre in Gdansk (TASK).
The study was supported by the European Community under the FP5 Programme, key-action City of Tomorrow and Cultural Heritage (Contract No. EVK4CT-2002-80005). This support is acknowledged with gratitude.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
tq209b-e/166
11 V 2005