You are on page 1of 25

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/247990834

Comparison of viscoplasticity formats and algorithms

Article in Mechanics of Cohesive-Frictional Materials · January 1999


DOI: 10.1002/(SICI)1099-1484(199901)4:13.0.CO;2-4

CITATIONS READS

17 714

3 authors, including:

Matti Ristinmaa
Lund University
167 PUBLICATIONS 3,789 CITATIONS

SEE PROFILE

All content following this page was uploaded by Matti Ristinmaa on 09 May 2018.

The user has requested enhancement of the downloaded file.


MECHANICS OF COHESIVE-FRICTIONAL MATERIALS
Mech. Cohes.-Frict. Mater. 4, 75—98 (1999)

A comparison of viscoplasticity formats and algorithms

Kenneth Runesson*, Matti Ristinmaa and Lennart Mähler


Division of Solid Mechanics, Chalmers University of Technology, S-412 96 Göteborg, Sweden
Department of Solid Mechanics, University of Lund, Box 118, S-221 00 Lund, Sweden

SUMMARY
Algorithmic issues for the two thermodynamically consistent viscoplastic formulations of Perzyna and
Duvaut—Lions are discussed. It is shown that it is simple to avoid the numerical problems associated with
a small relaxation time without resorting to elaborate perturbation techniques, as suggested in the literature.
A systematic numerical investigation of the efficiency of Newton iterations, that employ the Algorithmic
Tangential Stiffness (ATS) tensor, as compared to various approximations, is carried out for a cohesive-
frictional model with non-linear isotropic hardening. Generally, the ATS-tensor is formulated in such an
explicit fashion that its tensorial structure resembles that of the underlying rate-independent con-
tinuum stiffness. For both the Perzyna and the Duvaut—Lions format, it appears that the ATS-tensor is
obtained by a proper augmentation of the corresponding rate-independent ATS-tensor. Copyright  1999
John Wiley & Sons, Ltd.
KEY WORDS: Algorithmic Tangential Stiffness tensor; viscoplasticity; Closest-Point-Projection Method

1. INTRODUCTION
Viscoplasticity can conveniently be considered as a special class of the classical ‘standard
material’, if the evolution equations for the internal variables are associated with the (quasi-
static) yield function. It is well-known that this situation correponds to ‘maximal dissipation’
characteristics. Nonassociative flow and hardening rules are thus ‘non-standard’ features.
After integration of the evolution equations, relations between stresses and strains are often
obtained on incremental form, whose (non-linear) characteristics depend on the used integration
method. A ‘core-algorithm’ that employs the strain-driven format may be used within any
iterative procedure to solve the equilibrium equations in a FE-strategy (or whenever a stress
component is prescribed, such as in the case of plane stress). Newton iterations are frequently
used, which means that the Algorithmic Tangent Stiffness (ATS) tensor is required.
The classical (and most frequently used) formulation of viscoplasticity is based on the ideas by
Perzyna. Implicit integration of the rate equations gives an incremental format that reminds
strongly about the rate-independent counterpart, as is shown in Section 2.3. It seems that the
ATS-tensor was first established for the Closest-Point-Projection Method (CPPM), applied to
perfect plasticity by Runesson and Booker, and then to various classes of plasticity models with
hardening by Ortiz and Popov, Simo and Taylor, and Runesson et al.. A wealth of literature

* Correspondence to: K. Runesson, Division of Solid Mechanics, Chalmers University of Technology, S-41296, Göteberg,
Sweden.

CCC 1082-5010/99/010075—24$17.50 Received 29 May 1997


Copyright  1999 John Wiley & Sons, Ltd. Revised 11 March 1998
76 K. RUNESSON E¹ A¸.

has thereafter been devoted to the characteristics of implicit integration rules. For example, the
issue of symmetry of the ATS-tensor was addressed by Ortiz and Martin, Simo and Gondvin-
djee and Runesson and Larsson. It has been established that the ATS-tensor is always
symmetrical (and can be derived as the Hessian of an incremental strain energy density) when the
evolution rules are fully associative. This is only a sufficient condition, since exceptions exist
(which is discussed in this paper). However, to the author’s knowledge, the ATS-tensor has not
been formulated in closed form such that it shows the tensorial structure as compared to the
Continuum Tangent Stiffness (CTS) tensor.
In recent years, the format of viscoplasticity of Duvaut and Lions has been advocated, e.g.
by Simo and Hughes, Simo et al., Simo and Gondvindjee, and it has been shown that
the ATS-tensor is composed additively (with certain weights) of the elastic stiffness and the
ATS-tensor of the underlying rate-independent solution. To the author’s knowledge, the
generalization to non-associative flow and hardening rules has attended less interest. As shown
in this paper, it turns out that the weights are not scalars but fourth-order tensors in general.
We may also mention that attempts have been made to use the consistent linearized format
(which is represented by the ATS-tensor) in order to assess the possibility for localized deforma-
tion in the presence of viscoplastic deformation. For example, Münz et al., while considering the
classical Bingham model in the Duvaut—Lions format, investigated numerically the conditions for
suppressing localization when the underlying plastic response exhibits isotropic linear softening.
For both the Perzyna and Duvaut—Lions formulations, the appropriate ATS-tensor is for-
mulated in such a fashion that its tensorial structure becomes transparent, which requires
elimination of the hardening (internal) variables. As a consequence, a more explicit format than
that employed by e.g. Simo and Hughes and Peric is obtained. It is also shown that the
ATS-tensor for the Perzyna format can be obtained by a very simple augmentation of the
hardening modulus in the corresponding rate-independent ATS-tensor.
The paper is concluded by a systematic numerical evaluation of the two viscoplastic formula-
tions for a cohesive-frictional model involving the Drucker—Prager quasistatic yield criterion. It is
shown in the paper how numerical difficulties for very small relaxation time (when the rate-
independent solution is retrieved), as reported by Peric, can be completely avoided by very
simple manipulation. The efficiency of proper Newton iterations is compared with various
approximations.

2. CONSTITUTIVE RELATIONS IN VISCOPLASTICITY—PERZYNA’S


FORMULATION
2.1. Basic relations

It is possible to consider elastic—viscoplastic material behaviour as a particular subclass of the


‘standard dissipative material’, if this class of material behaviour is extended to include
‘non-standard’ features (as defined later). Adopting, for simplicity, a decoupled format of elastic
and plastic properties, we may introduce the free energy density (per unit volume) as

1 1
t(e, i )" e : E : e# H i i , e"e!e (1)
? 2 2 ?@ ? @
?@
Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 77

where the following notation was introduced: E is the positive-definite (constant) tensor of elastic
stiffness moduli, H is taken as a positive-definite (constant)* matrix of hardening moduli
?@
(corresponding to strict hardening), e is the strain, whereas e and i are the interval variables
?
(plastic part of the strain and hardening/softening variables, respectively). In order to simplify
notation, i are treated as scalars in this paper (unless they are defined explicitly).
?
From the second law of thermodynamics, the following constitutive relations are considered:

r"*e t"E : e; D"r : e #


K iR *0 (2)
? ?
?
where D is the rate of dissipation. The ‘dissipative’ stresses due to hardening/softening K , that are
?
energy-conjugated to i , are defined as
?
K "!* t"!
H i (3)
? G? ?@ @
@
For general non-associative plastic flow and hardening rules, we introduce the dissipative
potential F* in such a fashion that

1 1
e " (F)*r F*; i " (F)* F* (4)
t . ? t . )?
* *
where t is the natural relaxation time, and (F) is a nondimensional monotonically increasing
* .
overstress function with the properties

d
(F)'0 and . (F)'0 for F'0, (F)"0 for F)0 (5)
. dF .

where F(r, K ) is the convex quasistatic yield function. The elastic domain is thus defined by the
?
convex setR

B"+(r, K )"F(r, K ))0, (6)


? ?
and it is assumed that (0, 0)3B.
Upon differentiating the constitutive equations (2) and (3) with respect to time, and introducing
the flow and hardening rules in (4), we obtain the constitutive relations of viscoplasticity on rate
form as follows:

1
r # (F)E : *r F*(r, K )"r  with r "E : e (7)
t . A
*
1
KQ # (F) H * F*(r, K )"0, a"1, 2, 2. (8)
? t . ?@ )@ A
* @
which can be integrated, for given strain history e(t) and initial conditions r(0)"r, K (0)"0.
?

* That H is constant is not a serious restriction; non-linear hardening can still be modelled (as shown later).
?@
R To avoid technical complications, it is assumed in this paper that the dissipative potential F* (as well as the yield
function F) are smooth.

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
78 K. RUNESSON E¹ A¸.

Special case: Rate-independent plasticity is obtained in the special case when t P0, which
*
gives (F)"0 and, consequently, F"0 (in the case of loading). Hence, we make the substitution
.
j"(1/t ) (F)'0 and we propose the loading (complementary) conditions
* .
j*0, F)0, jF"0 (9)
where we have also included the situation of elastic unloading defined by j"0, F(0. It is clear
that (9) are the classical Kuhn—Tucker conditions in the case of fully associative flow and
hardening rules.
To complete the picture of elasto-plasticity, we also give the Continuum Tangent Stiffness
(CTS) tensor E from the relation
1
r "Ee with E"E! E : *r F*  *r F : E (10)
A
where
A"*r F : E : *r F*#h, h" * FH * F* (11)
)? ?@ )@
?@
and h is the ‘generalized plastic modulus’. Since we have assumed strict hardening, then h'0 and
it follows that E is positive definite under very mild restrictions on F*, cf. Reference 15. Clearly,
E is nonsymmetrical if E : *r F*OE : *r F.

2.2. Thermodynamic admissibility


As to the satisfaction of the dissipation inequality, in (2) , we first note that it is trivially

satisfied in the special case of associative flow and hardening rules, defined by F"F* (like in the
case of rate-independent plasticity). However, even when FOF* it is possible to show that D*0
for many models of practical importance. For example, let us define
F*"F# G K K (12)
?@ ? @
?@
where the matrix G is assumed to be (at least) positive semidefinite. With this choice of F*, we
?@
can represent common forms of nonlinear quasistatic hardening, as discussed in Section 4.
We now obtain from (2) , (4), and (12) that


 
1
D" (F) r : *r F# K * F# G K K
t . ? )? ?@ ? @
* ? ?@ (13)

 
1
* (F) r : *r F# K * F
t . ? )?
* ?
where the inequality follows from the assumed properties of G . Moreover, since F is assumed to
?@
be a convex (smooth) function at the point (r, K ), it follows directly that D*0.
?
In general, non-associative flow rules will violate the dissipation inequality. However, there are
exceptions, such as the situation when F is of the Drucker—Prager type with a dilatancy angle that
is less than the friction angle.* To include such possibilities, we shall henceforth consider such

* Thermodynamics admissibility may be conditional on the actual state of stress and hardening, as discussed by Cannmo.

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 79

forms of non-associativity for which there exists a symmetric positive-definite tensor E* and
a positive-(semi)definite matrix H* that satisfy
?@
E : *r F*"E* : *r F, H * F*" H* * F (14)
?@ )@ ?@ )@
@ @
In general, E* and H* are state-dependent (although E and H are assumed constant).
?@ ?@
However, for the model considered in Section 4, based on the Drucker—Prager quasistatic yield
criterion, E* turns out to be constant (for a constant dilatancy angle).

Remark. The special case of plastic incompressibility is not allowed in this context, since it
would infer that E* is positive semidefinite. More precisely, E* would be a purely deviatoric
tensor in that case.
As to the choice of H* , let us consider an explicit expression that is associated with the choice
?@
of F* in (12). Obviously, the choice of H* is not unique. One possibility is the following: H* is
?@ ?@
obtained from H as the rank-one augmentation
?@

 
\
H* "H #ag g with a" g * F '0 with g  " H G K (15)
?@ ?@ ? @ ? )? ? ?@ @A @
? @A
which holds when g O0.
?
Using (14), we may now rewrite the constitutive relations (7) and (8) in the following
‘associative’ form:
1
r # (F)E* : *r F(r, K )"r  (16)
t . A
*
1
KQ # (F) H* * F(r, K )"0 (17)
? t . ?@ )@ A
* @
2.3. Implicit integration—incremental constitutive relations
Applying the backward Euler rule to equations (7) and (8), we obtain the incremental relations

L>r"L>r!*j(L>F)E : *r F*(L>r, L>K ) with L>r  "Lr#E : *e


A (18)
L>K "L>K!*j(L>F) H * F*(L>r, L>K ) with L>K  "LK
? ? ?@ )@ A ? ?
@
where L>r is the ‘elastic’ or ‘trial’ stress, and where we have introduced the function
*t t
*j(F)" (F) or * *j(F)" (F) (19)
t . *t .
*
In the special case that L>F)0, then *j"0 and it follows from (18) that the ‘elastic’ solution
L>r"L>r and L>K "L>K ("LK ) is obtained in the considered time increment. Hence,
? ? ?
the loading condition is the same as in plasticity. With L>F"(L>r, L>K ), we conclude that
?
L>F"L>F(0 signals elastic unloading, whereas L>F'0 signals a viscoplastic solution. In
this latter case an iterative procedure is (in general) required in practice to obtain the solution.

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
80 K. RUNESSON E¹ A¸.

When L>r and L>K have been found, it is possible to compute the increments *e and *i
? ?
from (2)—(4) as
*e"*j(L>F)*r F*"(E)\(L>r!L>r)
(20)
*i "*j(L>F)* F*" (H )\(L>K !L>K )
? )? ?@ @ @
@
where (18) was used.
Special case: Rate-independent plasticity is also defined by (18) if *j is interpreted as a Lagran-
gian multiplier, which is determined by the incremental version of the loading condition in (9). To
be more explicit, we may introduce (14) to obtain
L>r"L>r!*jE* : *r F(L>r, L>K )
A
L>K "L>K!*j L>H* * F(L>r, L>K ) (21)
? ? ?@ )@ A
@
*j*0, L>F)0, *jL>F"0
In the case that E* and H* are constant, then (21) are, indeed, the incremental relations pertinent
?@
to the Closest-Point-Projection Method (CPPM), which may also be characterized by the
following non-linear, convex, minimization problem:

(L>r, L>K )"arg


? 
min E (L>r!r, L>K!K )
 Y )Y? Z r
B
 ? ? (22)

with the generalized complementary energy norm E defined as



E (r, K )"r : (E*)\ : r# K (H* )\K (23)
 ? ? ?@ @
?@
From the properties of E*, H* and B, it is clear that the solution (L>r, L>K ) is unique.
?@ ?
2.4. Algorithmic tangent stiffness tensor
The ATS-tensor, E (for any material response), is defined by the relation
d(*r)"E : d(*e) (24)
for variations d(*e) of the current strain increment *e. (In order to simplify the notation, we drop
the superscript n#1 subsequently, since all values are evaluated at the updated state.)
For elastic loading, it follows trivially that E"E. For viscoplastic loading, i.e. when *j*0,
the pertinent expression of E is derived in Appendix A as
1
E"E! E : *r F*  *r F : E (25)
A
where the following notation has been introduced: E is the (symmetric) Algorithmic Elastic
Stiffness (AES) tensor defined as

 
\
E" (E)\#*j*rr F*!(*j) *r F*(H ) * r F* (26)
)? ?@ )@
?@
Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 81

Moreover, *r F and *r F* are the algorithmic gradients of F and F*, respectively, and are defined
as

*rF "*rF!*j *K F(H ) *K rF*


? ?@ @
?@ (27)
*rF*"*rF*!*j *rK F*(H ) *K F*
? ?@ @
?@
whereas A is defined as

 
t d \
A"A# * . (28)
*t dF

The quantity A, which is the algorithmic version of A in (11) pertinent to rate-independent
plasticity, is defined as

A"*rF : E : *rF*#h with h " *K F(H ) *K F* (29)


? ?@ @
?@
Moreover, the matrix (H ) , which is the ‘algorithmic’ version of the hardening matrix H , is
?@ ?@
defined as

(H ) "[(H )\#*j*K K F*]\ (30)


?@ ?@ ? @

Remark. The similarity in mathematical structure of E (25) and the CTS-tensor E (10) is
striking. Moreover, the only difference between E, according to Perzyna’s formulation of
viscoplasticity, and E , pertinent to rate-independent plasticity, is the second term in (28) for A.
Since the rate-independent plasticity is defined by t P0, it follows readily that APA and,
* 
consequently, EPE, in this extreme case, which should be expected.

It appears readily that E is non-symmetrical in general. In order for E to be symmetrical, it
must be required that E : *rF "E : *rF*. A trivial sufficient condition for this equality to hold is
*rF "*rF* . It is noted that even if the flow rule is of the associative type, i.e. *rF"*rF*, we may
still have *rF O*rF* . In order to guarantee that *rF "*rF* we must require that both the flow
and hardening rules are of associative type, i.e. the format of the ‘‘Standard Dissipative Material’’,
cf. Reference 1, must be employed. However, as we shall see later, the condition
E : *rF "E : *rF* can be satisfied under weaker conditions.

3. CONSTITUTIVE RELATIONS IN VISCOPLASTICITY—DUVAUT—LIONS’


FORMULATION
3.1. Basic relations

The basic thermodynamic relations (1)—(3) are the same as for Perzyna’s theory whereas the
flow rule and the hardening rule (pertinent to the quasistatic yield surface) are conveniently
chosen in the spirit of Duvaut—Lions as

1 1
e " (r) (E)\ : (r!r), iR " (r) H\ (K !K ) (31)
t " ? t " ?@ @ @
* * @
Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
82 K. RUNESSON E¹ A¸.

The pair (r, K ) is ‘quasistatically admissible’ in the sense that it is defined as the solution to
?
r"r!jE : *rF*(r, K )
A
K "K !j H *K F*(r, K ) (32)
? ? ?@ @ A
@
j*0; F(r, K ))0; jF(r, K )"0
A A
where it was tacitly assumed that F and F* are smooth.
Again, for constant E* and H* , (32) is the KT-problem associated with the CPPM-projection
?@
of the current state (r, K ) onto the convex set B, i.e.
?

(r, K )"arg
?  min E (r!r, K !K )

(r, K )3B
?
? ?  (33)

where the norm E was defined already in (23).



Remark. The complete resemblance of (r, K ) and the updated solution (L>r, L>K ) in
? ?
rate-independent plasticity upon using CPPM for the integration is noteworthy.

The metric r is defined as

r"E (r!r, K !K ) (34)


 ? ?
i.e. r is the minimum value of E in (33) for given values (r, K ) when (r, K )3B. In this fashion,
 ? ?
the original formulation by Duvaut—Lions, cf. Reference 11, is generalized to the (new)
overstress function (r) with the following properties:
"
d
(r)*1, " (r)'0 for r'0 (35)
" dr

We remark that and are different functions (or their respective arguments). Moreover, the
. "
relaxation time t does not necessarily take the same value in the two formats.
*
It remains to combine the flow and hardening rules (31) with the constitutive relations (2) to
yield the rate equations

1
r # (r) (r!r)"E : e
t "
*
(36)
1
KQ # (r) (K !K )"0
? t " ? ?
*
Special case: Rate-independent plasticity is obtained when (the penalty parameter) t P0.
*
Moreover, the classical version of the Duvaut—Lions formulation is defined by (r)"1.
"

3.2. Thermodynamic admissibility


Next, we investigate under which conditions the suggested theory is thermodynamically
consistent, i.e. when the dissipation inequality D*0, expressed in (2) , is satisfied. To this end, we

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 83

use (31) to rewrite


1
r : e " (r) [r#(r!r)] : (E)\ : (r!r)
t "
*
1
* (r) r : (E)\ : (r!r)
t "
*
where it was used that E is a positive-definite tensor. Upon solving for r!r from (32) and
inserting into the expression above, we obtain
1
r : e * (r)jr : *rF*(r, K )
t " A
*
In the same fashion, we conclude that
1
K iR * (r)j K *K F*(r, K )
? ? t " ? ? A
? * ?
We thus obtain

 
1
D* (r) j r : *rF*(r, K )# K *K F*(r, K ) (37)
t " A ? ? A
* ?
Since (r, K )3B, it is clear that D*0 in the case of a standard material, i.e. whenever F*"F.
?
Moreover, D*0 when the special assumption for hardening in (12) is employed. Clearly, these
situations represent only sufficient conditions for D*0.

3.3. Backward Euler integration—incremental constitutive relations


Applying the backward rule to (36), we obtain after some rearrangement, the updated values
1
L>r" (L>r#*j(L>r)L>r) with L>r"Lr#E : *e
1#*j(L>r)
(38)
1
L>K " (L>K#*j(L>r)L>K ), a"1, 2, 2
? 1#*j(L>r) ? ?
where *j(r) is defined by
*t t
*j(r)" (r) or * *j(r)" (r) (39)
t " *t "
*
and L>K"LK . Now, (L>r, L>K ) is obtained by CPPM-projection of (L>r, L>K ) onto
? ? ? ?
the quasistatically admissible set B. Upon inserting (38) into (32), we conclude that (32) can be
replaced by the rate-independent plasticity problem defined as
L>r"L>r!jE : *rF*(L>r, L>K )
A
L>K "L>K!j H *K F*(L>r, L>K ) (40)
? ? ?@ @ A
@
j*0; F(L>r, L>K ))0; jF(L>r, L>K )"0
A A
Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
84 K. RUNESSON E¹ A¸.

where
j"j(1#*j(L>r)) (41)
Here we have used that (1#*j) is a positive quantity, i.e. we are allowed to multiply (40c) and

(40c) by this factor.

Remark. It appears that the underlying plasticity problem (40) can be solved (for each time
increment) without any coupling to the viscoplastic properties.
With j known from the solution of (40), we may use (41) to calculate *j(L>r) as follows: Upon
combining (32) and (40) with (41), we obtain the relation
L>r"L>r(1#*j(L>r)) (42)
where we used the notation
r"E (r!r, K !K ), r"E (r!r, K!K ) (43)
 ? ?  ? ?
Equation (42) may be solved for L>r (by iteration in general), and *j(L>r) is found. It is then
possible to calculate the updated solution (L>r, L>K ) from (38).
?
3.4. Algorithmic tangent stiffness tensor
The ATS-tensor can be calculated from the expression (38) in a quite straightforward manner
as (dropping the superscript n#1)
1
E" [(I!R ) : E#(*jI#R ) : E #R ] (44)
1#*j   

where E is the ATS-tensor corresponding to the CPPM-projection problem in (40). The


appropriate expression was given in (25) with t "0. Moreover, we have introduced the
*
transformation tensors R and R defined as
 
R "crr  rr : (E*)\ with rr"r!r
 (45)
R "crr  rK (H* )\(R ) with rK "K !K
 ? ?@ @ ? ? ?
?@
where *k(r) is defined by
*t d (r) 2*k
*k(r)" " ; c" (46)
t dr 1#*j#2r*k
*
Details of the derivation are found in the appendices. In the particular case that is independent
"
of the state, we obtain R "R "0. Moreover, when t P0 (*jPR) we obtain EPE .
  *

4. COHESIVE/FRICTIONAL VISCOPLASTIC MODEL


4.1. Preliminaries

As a proponent for a class of models showing both cohesive and frictional characteristics, we
choose the quasistatic yield surface of Drucker—Prager type with non-associative flow and

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 85

hardening rules, accounting for dilatancy and nonlinear isotropic hardening.* In its
rate-independent form such a model is classical in soil and powder mechanics. However, it may
also be used to represent the response of certain metallic materials, which show pressure
dependence. Such an example is grey-cast-iron, cf. Reference 17, in which case the friction angle is
relatively small (in the order of 10). In this paper we restrict our attention to such a situation,
whereby it is possible to presume that only regular solutions are encountered (in order to comply
with the adopted theoretical framework and to avoid further difficulties in conjunction with
possible apex solutions). At this junction, we remark that issues related to integration and the
consistent linearization of non-smooth yield criteria of the Mohr—Coulomb type were recently
discussed by Larsson and Runesson.
The pertinent (class of ) models can be derived from the following form of Helmholtz’s free
energy function (per unit volume)

t" e : E : e# Hi (47)


 
together with the following form of the field and potential functions:

s K
F"p #kp !s(p #K), F*"p #k*p !s(p #K)# (48)
    2K

where p and p are the mean stress and equivalent stress, respectively, which are defined as



1 3
p " tr(r), p " "r ", r "r!p d (49)
3  2  

In (47), i is the ‘strain-like’ internal variable associated with the drag-stress K, whereas H is the
initial hardening modulus in uniaxial stress. Moreover, the friction coefficient k and the
parameter s can be computed as

3(.!1) 2. p
k" , s" with ." '1 (50)
.#1 .#1 p

where p '0 and p '0 are the (initial) yield values in uniaxial compression and tension,
 
respectively. It appears that von Mises criterion is retrieved when ."1, in which case k"0 an
s"1. Finally, a non-associative flow rule is obtained when k*(k, where k* is the dilatancy
coefficient.
We shall assume isotropic elasticity, whereby E and E* are given as

k*
E"2GI #K d  d, E*"2GI #K dd (51)
    k

where G is the shear modulus, K is the bulk modulus, and I is the deviatoric projection tensor,
 
that projects any second order tensor onto its own deviatoric. This tensor is defined as

I "I! d  d (52)
 

* The restrictions to isotropic hardening is made merely to avoid unnecessary technical complexity.

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
86 K. RUNESSON E¹ A¸.

The constitutive relations are found from (2) and (3) as



r"E : e, K"!Hi (53)

Finally, in this preliminary section, we give the CTS-tensor E pertinent to the rate-independent
behaviour (for completeness):

9G
E"E! u*  u (54)
Ap

where

K kp K k*p
u"r #   d, u*"r #  d (55)
 3G  3G

 
K
A"3G#K kk*#h, h"sH 1! (56)
 K


4.2. Perzyna+s formulation


The pertinent constitutive rate equations take the form

1 3G
r # (F) r "2Ge
 t . p  
* 
1
pR # (F)K k*"K eR (57)
t .   
*

 
1 K
KQ ! (F)sH 1! "0, K(0)"0
t . K
* 

Remark. If k*"0, then we obtain from (57b) that pR "K eR , corresponding to only elastic
 
volumetric change.

Upon backward Euler integration of (57) we obtain, after rearrangement of terms, the
expressions


3G*j 3
L>r "c (*j)L>r , c (*j)"1! , p" "L>r "
    L>p  2 

L>p "L>p !*jK k* (58)


 
sH*j \
L>K"a (*j) (L>K#sH*j), a (*j)" 1#
) ) K

where the ‘trial elastic’ stress is decomposed as

L>r "Lr #2G*e ; L>p "Lp #K *e (59)


    
Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 87

In the case that viscoplastic loading has been detected, i.e. when
 L>p#kL>p !s(p #LK)'0
L>F" (60)
 
then the updated solution must be obtained in an iterative fashion via the solution of the equation
t
* *j! (L>F(*j))"0 (61)
*t .

with the function L>F(*j) given as


F(*j)"L>FI (*j)!AI (*j)*j (62)
where
L>FI (*j)"L>p#kL>p !s(p #a (*j)LK)
  )
(63)
AI (*j)"3G#K kk*#a (*j)sH
 )

Remark. The special case of rate-independent response, i.e. when t "0, does not present any
*
extra difficulty in the way the problem has been formulated here.

As to the ATS-tensor, it can first be shown that the AES-tensor takes the form
3G*j/p
E"E!2GbQ with b"  (64)
1#(3G*j/p )

where Q is the idempotent tensor defined as*
3
! r r
Q"I (65)
 2p  

The technicalities in obtaining (64) are somewhat lengthy and are therefore omitted here. As to
A defined in (28), we may show that

 
K
A"3G#K kk*#h , h "sHa 1! (66)
 ) K


Remark. It appears that AOA in general, since A"A only if K "R (linear

hardening).
Finally, we may establish E (*e) as follows:
9G
E"E! u*  u (67)
Ap

with E given in (64) and A defined by (66) and (28).

* Q is singular with the eigenvalues r and d corresponding to zero eigenvalue.




Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
88 K. RUNESSON E¹ A¸.

Remark. The only source of non-symmetry is the non-associative flow rule, whereas the
non-associative (non-linear) isotropic hardening does not destroy the symmetry. Hence, this is an
example that a fully associative rule is only a sufficient (but not necessary) condition for symmetry
of E (and E ). However, it can be shown that non-linear kinematic hardening does indeed
introduce non-symmetry.

4.3. Duvaut—Lions+ formulation


The yield and the potential functions are still defined by (48). The pertinent constitutive rate
equations now become

1
r # (r) (r !r )"2Ge
 t "   
*
1
pR # (p !p )"K e (68)
t "  
*
1
KQ # (r) (K!K)"0, K(0)"0
t "
*
where the quasistatically admissible state (r, K) is defined by the CPPM-projection of (r, K)
onto B.
Upon backward Euler integration of (68), we obtain

1
L>r " (L>r #*j(L>r)L>r )
 1#*j(L>r)  

1
L>p " (L>p #*j(L>r)L>p ) (69)
1#*j(L>r)

1
L>K" (L>K#*j(L>r)L>K)
1#*j(L>r)

where the quasistatic solution (L>r, L>K) is the CPPM-projection of (L>r, L>K) onto B as
follows:

L>r "c (j)L>r


  
L>p "L>p !jK k* (70)

L>K"a (j) (L>K#sHj)
)
The plastic multiplier j is defined by the KT-conditions

j*0, F(r, K))0, jF(r, K)"0 (71)

As to the explicit expression of the ATS-tensor, we first note that E is identical to the expression
for E for the Perzyna model given in (67) with t "0. The pertinent expression for E is then
*
Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 89

given by (44). In the special case that "constant, then we obtain the simpler relation
"
1
E" (E#*jE ) (72)
1#*j

Remark. Like for the Perzyna formulation, the case t "0 can be simply dealt with. Rewriting
*
(69) to contain the term t *j solves the problem when approaching the rate independent limit.
*
Actually, no problems are encountered if t "0 is used in the formulation.
*

5. COMPARISON OF THE PERZYNA AND DUVAUT—LIONS FORMULATIONS


5.1. General

Upon combining the quasistatic solution (r, K ) from (32) with the rate equation in (31), we
?
obtain the following rate laws for the Duvaut—Lions formulation:
1 1
e " (r)j*rF*(r, K ); iR " (r)j*K F*(r, K ) (73)
t " A ? t " ? A
* *
These expressions may now be compared with those in (4) for the Perzyna formulation, to show
that, for the formulations to represent the same behavior, a scalar k (that may be state-dependent)
must exist such that
*rF*(r, K )"k*rF*(r, K ); *K F*(r, K )"k*K F*(r, K ) (74)
A A ? A ? A
where k is given by
(F)
k" . (75)
(r)j
"
Clearly, if the proportionality conditions (74) are satisfied for some k, then it is possible, for given
(F), to choose (r) according to (75). It is, however, obvious that the proportionality
. .
conditions (74) can not be satisfied in general. An important exception is discussed below.

5.2. Cohesive/frictional model


The issue whether the formulations of Perzyna and Duvaut—Lions can be brought into
equivalence for the cohesive/frictional model, defined in Section 4, will be discussed next. As to
the (non-associative) flow rule, we note that
3 k* 2 k*
*rF*(r, K)" r # d; *rF*(r, K)" r # d (76)
2p  3 3p  3
 
Since (69) and (70) shows that r and r are proportional, it follows trivially that k"1, as
 
defined in (75). However, considering the (non-associative) isotropic hardening, we obtain the
relations

   
K K
* F*(r, K)"! 1! ; * F*(r, K)"! 1! (77)
) K ) K
 
Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
90 K. RUNESSON E¹ A¸.

and it follows kO1 in general (since KOK in the general situation). Hence, no unique value of
k exists unless the hardening is linear (K "R), in which case k"1.

In the particular case of linear hardening (when k"1), we obtain from (75) and (32) that

(F) F
. "1 and j" with A"3G#K kk*#sH (78)
(r)j A 
"

Remark. The solution j"F/A of (32) is the counterpart of the solution j"F/A of (40) in
the particularly simple case of linear hardening.

Hence, from (78) we obtain the relation

A
(r)" (F) (79)
" . F

which can be satisfied only if both sides are constant, i.e. independent of the state. This is the case
for the classical Bingham model defined by the linear function and constant as follows:
. "
1F2 K kk* H
(F)" "1#  #s (80)
. 3G " 3G 3G

in which case no iterations are necessary for obtaining the updated solution. For example, in the
Perzyna model, we obtain from (62) and (63) that

 
*t 1 A *t \ F
*j" 1# FP when t P0 (81)
t 3G 3G t A *
* *
Moreover, A becomes

 
t
A"3G 1# * #K kk*#sH (82)
*t 

6. NUMERICAL EXAMPLES
The non-linear isotropic hardening cohesive/frictional viscoplastic material has been
implemented, whereby the following values of the material properties were used in the
calculations:

E H K
"400, "4, "0)2, l"0)3, k"tan 10°, k*"tan 5°
p p p
  
Here, E and l are Young’s modulus and Poisson’s ratio, respectively. Plane strain conditions will
be assumed in all examples. As to the overstress function, we adopt the simple model defined by
(80) (in which case it was shown above that either the Perzyna or the Duvaut—Lions formulation
can be used to describe the same model in the special case of linear hardening).

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 91

In the finite element calculations we shall use the residual norm as the termination criterion:

""R(a )""(tol ""R(a )"" (83)


G> 
where R is the residual and a is the incremental displacement vector in the ith iteration. The value
G
10\ of tol will be adopted in the calculations presented below.
A specimen containing a cylindrical hole is studied, as shown in Figure 1, and due to the double
symmetry of the specimen only one-quarter needs to be modelled. The used finite element mesh,
as shown in Figure 1(b), comprises 384 CST elements. The geometry is defined by the width ¼,
the height ¸, the thickness h, and the radius of the hole R; the ratio ¸/¼ is set to 2 whereas ¼/R is
set to 6. Symmetric boundary conditions are imposed on the edges x "0 and x "0.
 
A prescribed constant velocity uR "uR is imposed on the edge located at x "¸/2, while the
 
remaining boundaries are assumed to be traction free.

6.1. Cohesive/frictional model (Perzyna format)


In Figure 2, the load—displacement curves are shown for the different values of uR t /¸, which can
*
be interpreted as different loading rates or different values of the natural relaxation time. We note
that the rate-independent response is recovered for t "0. To show that the numerical strategy
*
can handle the case where the rate independent limit is approached uR t /¸"8;10\ is used,
*
which corresponds to a small value of the loading rate uR or a small value of the natural relaxation
time t . For each loading rate, two different load increments have been used in the calculations:
*
*u/¸"8;10\ and *u/¸"16;10\, which are indicated by a ( * ) and (䡩), respectively, cf.
Figure 2.

Figure 1. The geometry of the specimen containing a cylindrical hole, and the discretization used in the finite element
calculations. Each quadrilateral is built up by four Constant Strain Triangle (CST) elements

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
92 K. RUNESSON E¹ A¸.

Figure 2. Load-displacement curves for the non-linear hardening classical Bingham model (Perzyna format). ( * ) indicates
results for the load increment *u/¸"8;10\ whereas (䡩) indicates results for *u/¸"16;10\. The smooth curves are
obtained by taking very small load increments. The curves correspond to different loading rates or different values of the
realization time

The convergence properties were investigated for the following choice of iteration tensors in the
equilibrium iterations: The ATS-tensor E (corresponding to full Newton), the symmetric
ATS -tensor E  (the symmetric part of E ), the corresponding elastic-plastic ATS -tensor
 
E (which is obtained from the ATS-tensor E upon setting t "0), and the elastic ETS-tensor
*
E. We note that E and E are non-symmetric tensors.

Remark. E is still evaluated at the actual state and should not be confused with E for the
Duvaut—Lions model. Moreover, E is the proper tangent tensor only in the case of
rate-independent response.
The number of iterations to satisfy the convergence criterion (83) is shown in Table I for
uR t /¸"8;10\. The importance of using the ATS-tensor, pertinent to the consistent Newton
*
scheme with quadratic convergence, is clearly demonstrated.* An interesting observation is that
the behavior of the ATS -tensor is extremely poor, despite the apparently ‘innocent’ difference

between E and E.

* However, it appears that the symmetric part of the ATS-tensor is equally efficient (for this particular example), which is
a clear advantage in terms of reduced solution time when a conventional direct solver is used.

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 93

6.2. Cohesive/frictional model (Duvaut—Lions formulation)


Since we consider non-linear hardening (according to the choice of parameters above), the
results of this model will not coincide with those from the Perzyna format. However, the
difference is not great (no load—displacement curves are shown). As discussed earlier, the Perzyna
and Duvaut—Lions model degenerate to the same model when K PR. Moreover, for this case

(of linear hardening) it can be confirmed explicitly that the ATS-tensor for the Duvaut—Lions
format (72) equals the ATS-tensor for the Perzyna format (67).
The convergence behaviour is displayed in Table II for the loading rate uR t /¸"8;10\. We
*
compare the performance by the ATS-tensor E, the symmetrized ATS -tensor E  , the

elastic—plastic ATS -tensor E (which is evaluated at he static solution), and the elastic

ETS-tensor E. The results follow closely those in the previous section.

Table I. Number of iterations to convergence for the non-linear hardening Bingham model (Perzyna
format), uR t /¸"8;10\ and tol"10\. Two different displacement increments are used: *u/¸"8;10\
*
and *u/¸"16;10\. The sign — indicates that convergence could not be achieved

Stiffness Displacement u/¸;10\


tensor
8 16 24 32 40 48 56 64 72 80

ATS 1 5 6 4 5 5 5 5 5 5
ATS 1 8 9 9 9 9 9 9 9 9

ATS 1 38 —

ETS 1 75 136 149 151 151 151 153 154 155
ATS 5 7 5 5 6
ATS 9 11 11 11 11

ATS 18 —

ETS 166 255 264 267 271

Table II. Number of iterations to convergence for the non-linear hardening Bingham model (Duvaut—Lions
format), uR t /¸"8;10\ and tol"10\. Two different displacement increments are used: *u/¸"8;10\
*
and *u/¸"16;10\. The sign — indicates that convergence could not be achieved

Stiffness Displacement u/¸;10\


tensor
8 16 24 32 40 48 56 64 72 80

ATS 1 5 6 4 5 5 5 5 5 5
ATS 1 8 9 9 9 9 9 9 9 9

ATS 1 46 — — — — — — — —

ETS 1 75 135 149 151 151 152 153 155 156
ATS 5 7 5 5 6
ATS 9 11 11 11 11

ATS 19 — — — —

ETS 106 256 265 268 272

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
94 K. RUNESSON E¹ A¸.

7. CONCLUSIONS
The two viscoplasticity formulations of Perzyna and Duvaut—Lions were investigated with regard
to algorithmic formulation including their respective consistent linearization (defined by the
ATS-tensor). Both formulations employ the CPPM-concept for the underlying rate-independent
format, and it was concluded that the ‘rate-dependent ATS-tensor’ has a mathematical structure
that is remarkably similar to the CTS-tensor. Moreover, the precise conditions under which this
(rate-independent) ATS-tensor is symmetrical was investigated for cohesive/frictional model with
non-linear isotropic hardening. It turns out that this type of hardening (although it is of
non-associative type) does not introduce non-symmetry. For both the Perzyna and
Duvaut—Lions formulations, the total ATS-tensor is obtained by quite simple (but totally
different) augmentation of the rate-independent correspondent. Moreover, contrary to what has
been claimed in the literature, it is possible to avoid totally any numerical problems associated
with a small relaxation time without resorting to elaborate perturbation technique.
From the limited numerical investigation in this paper, it was found in a consistent fashion that
the symmetric part of the ATS-tensor can be used without serious detriment to the expected
quadratic rate of convergence. However, it was found totally unacceptable to use only the
rate-independent part (for the considered loading rate), since it may even lead to divergent
iterations.

APPENDIX A: ATS-TENSOR IN PERZYNA’S FORMULATION


Assume viscoplastic loading in the time-step, and consider the relations defined by (18), rewritten
in the more convenient form

(E)\ : *r!*e#*j*rF*(r, K )"0 (84)


A
H\*K #*j*K F*(r, K )"0 (85)
?@ @ ? A
@
*t
y"*j! (F(r, K ))"0 (86)
t . A
*
Upon differentiating (84), we obtain

(E)\ : d(*r)!d(*e)#d(*j)*rF*#*jd(*rF*)"0 (87)

where

d(*rF*)"*rrF* : d(*r)# *rK F*d(*K )"0 (88)


? ?
?
Moreover, by differentiating (85), we obtain

H\d(*K )#d(*j)*K F*#*jd(*K F*)"0 (89)


?@ @ ? ?
@
where

d(*K F*)"*K rF* : d(*r)# *K K F*d(*K ) (90)


? ? ? @ @
@
Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 95

Upon introducing (90) into (89) we obtain

d(*K )"! (H ) *K F*d(*j)!*j (H ) *K rF* : d(*r) (91)


? ?@ @ ?@ @
@ @
where

(H ) "[(H )\#*j*K K F*]\ (92)


?@ ?@ ? @

The expression (88) may now be combined with (91) to yield


d(*rF*)" *rrF*!*j
?@
? ?@ @ 
*rK F*(H ) *K F* : d(*r)! *rK F*(H ) *K rF*d(*j)
?@
? ?@ @
(93)

which, in its turn, may be inserted into (87) to yield, after rearrangement, the expression

d(*r)"E : (d(*e)!*rF*d(*j)) (94)

where

 
\
E" (E)\#*j*rrF*!(*j) *rK F*(H ) *K rF* (95)
? ?@ @
?@
*rF*"*rF*!*j *rK F*(H ) *K F* (96)
? ?@ @
?@
Upon inserting (94) and (91), we obtain

d(*K )"!
?
@
 (H ) *K F*!*j
?@ @
@
?@ @ 
(H ) *K rF* : E : *rF* d(*j)

!*j (H ) *K rF* : E : d(*e) (97)


?@ @
@
It remains to calculate d(*j) from the (convergence) condition dy"0 (at viscoplastic loading),
which yields

 
*t d
d(*j)! . *rF : d(*r)# *K Fd(*K ) "0 (98)
t dF ? ?
* ?
Upon inserting (94) and (97) into (98), we obtain

*t d
d(*j)! . (*rF : E : d(*e)#A d(*j))"0 (99)
t dF
*
where

*rF "*rF!*j *K F(H ) *rK F* (100)


? ?@ @
?@
A"*rF : E : *rF*#h ; h " *K F(H ) *K F* (101)
? ?@ @
?@
Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
96 K. RUNESSON E¹ A¸.

Since we require that '0 during viscoplastic loading, we may solve for d(*j) from (99) in
.
terms of (the prescribed value of ) d(*e) as

1
d(*j)" *rF : E : d(*e) (102)
A

where

 
t d \
A"A# * . (103)
*t dF

Finally, upon inserting d(*j) into (94), we obtain the ATS-tensor E pertinent to Perzyna’s
formulation as

1
E"E! E : *rF*  *rF : E (104)
A

Remark. It is also possible to establish the ‘ATS-tensor’ (R )


?
d(*K )"(R ) : d(*e) (105)
? ?
by inserting (102) into (97). In fact, this relation will be useful in deriving the ATS-tensor for the
Duvaut—Lions formulation, as discussed in Appendix B.

APPENDIX B: ATS-TENSOR IN DUVAUT—LIONS’ FORMULATION


Upon differentiating the incremental relations (38):

1
r" (r#*jr)
1#*j
(106)
1
K" (K#*jK )
? 1#*j ? ?

we obtain

1
d(*r)" [d(*r)#*jd(*r)!rrd(*j)] with rr"r!r
1#*j
(107)
1
d(*K )" [*jd(*K )!rK d(*j)] with rK "K !K
? 1#*j ? ? ? ? ?

Moreover, from (39) we obtain

*t d (r)
d(*j)"*k dr with *k" " (108)
t dr
*
Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
VISCOPLASTICITY FORMATS AND ALGORITHMS 97

and


dr"2 rr : (E*)\ : (d(*r)!d(*r))# rK (H* )\(d(*K )!d(*K ))
?@
? ?@ @ @  (109)

Upon introducing (107) into (109), it follows that (108) takes the form


d(*j)"c rr : (E*)\ : (d(*r)!d(*r))! rK (H* )\d(*K )
?@
? ?@ @  (110)

where we have introduced the scalar


2*k
c" (111)
1#*j#2r*k
We shall now introduce the relations
d(*r)"E : d(*e), d(*r)"E : d(*e), d(*K )"(R ) : d(*e) (112)
? ?
The tensors E and (R ) are obtained from (104) and (105), respectively, upon setting t "0.
? *
Clearly, they are the appropriate ATS-tensors for the rate-independent problem corresponding to
the static solution (defining by superscript s).
Let us also introduce the notation
R "crr  rr : (E*)\ and R "crr  rK (H* )\(R ) (113)
  ? ?@ @
?@
This allows us to solve for d(*r) from (107), and obtain the final expression for the ATS-tensor
E pertinent to the Duvaut—Lions formulation as
1
E" [(I!R ) : E#(*jI#R ) : E #R ] (114)
1#*j   

REFERENCES
1. B. Halphen and Q. S. Nguyen, ‘Sur les materiaux standards generalises’, J. Mechanique, 14, 39—63 (1975).
2. P. Perzyna, ‘Fundamental problems in viscoplasticity’, Adv. Appl. Mech., 9, 243—377 (1966).
3. K. Runesson and J. R. Booker, ‘On mixed and displacement finite element methods in perfect elasto-plasticity’, Proc.
of the 4th Int. Conf. in Australia on Finite Element Methods, 1982, pp. 85—89.
4. M. Ortiz and Popov, ‘Accuracy and stability of integration algorithms for elastoplastic constitutive equations’, Int. J.
Numer. Meth. Engng., 21, 1561—1575 (1985).
5. J. C. Simo and R. L. Taylor, ‘Consistent tangent operators for rate independent elasto-plasticity’, Comput. Meth. Appl.
Mech. Engng., 48, 101—118 (1985).
6. K. Runesson, A. Samuelsson and L. Bernspras ng, ‘Numerical technique in plasticity including solution advancement
control’, Int. J. Numer. Meth. Engng., 22, 769—788 (1986).
7. M. Ortiz and J. B. Martin, ‘Symmetry-preserving return mapping algorithms and incrementally external paths:
a unification of concepts’, Int. J. Numer. Meth. Engng., 28, 1839—1853 (1989).
8. J. C. Simo and S. Govindjee, ‘Non-linear B-stability and symmetry preserving return mapping algorithms for
plasticity and viscoplasticity’, Int. J. Numer. Meth. Engng., 31, 151—176 (1991).
9. K. Runesson and R. Larsson, ‘Properties of incremental solutions for dissipative material’, J. Engng. Mech., 119,
647—666 (1993).
10. G. Duvaut and J. L. Lions, ¸es Inequations en Mechanique et en Physique, Dunod, Paris, 1982.
11. J. C. Simo and J. R. Hughes, Elastoplasticity and »iscoplasticity. Computational Aspects’, Springer, Berlin, 1988, to be
published.

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)
98 K. RUNESSON E¹ A¸.

12. J. C. Simo, J. G. Kennedy and S. Govindjee, ‘Non-smooth multisurface plasticity and viscoplasticity.
Loading/unloading conditions and numerical algorithms’, Int. J. Numer. Meth. Engng., 26, 2161—2185 (1988).
13. T. Münz, S. Bakhrebah and K. Willam, ‘Viscoplastic adoption with regard to discontinuous bifurcation’, in S. N.
Atluri, G. Yagawa and T. A. Cruse (eds) Proc. Int. Conf. on Comp. Engineering Science (Computational Mechanics’ 95),
Springer, Berlin, 1995, pp. 1259—1264.
14. D. Peric, ‘On a class of constitutive equations in viscoplasticity: formulation and computational issues’, Int. J. Numer.
Meth. Engng., 36, 1365—1393 (1993).
15. K. Runesson and Z. Mroz, ‘Note on nonassociated plastic flow rules’, Int. J. Plasticity, 5(6), 639—658 (1989).
16. P. Cannmo, ‘A damage-based interface model: application to the degradation and failure of a polycrystalline
microstructure’, Ph.D. ¹hesis, Department of Solid Mechanics, Chalmers Univ. of Tech., Göteborg, Sweden, 1997.
17. B. L. Josefson, U. Stigh and H. E. Hjelm, ‘Nonlinear kinematic hardening model for elastoplastic deformations in grey
cast iron’, J. Engng. Mater. ¹echnol., 117(2), 145—150 (1995).
18. R. Larsson and K. Runesson, ‘Implicit integration and consistent linearization for yield criteria of the
Mohr—Coulomb type’, Mech. Cohesive-Frictional Mater., 1, 367—383 (1996).

Copyright  1999 John Wiley & Sons, Ltd. Mech. Cohes.-Frict. Mater., 4, 75—98 (1999)

View publication stats

You might also like