Beruflich Dokumente
Kultur Dokumente
a r t i c l e i n f o a b s t r a c t
1. Introduction
In classical small deformation theories of metal plasticity, when attempting to model the Bausch-
inger phenomenon (Bauschinger, 1886) and other phenomena associated with cyclic loading, kine-
0749-6419/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijplas.2008.11.008
1834 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
matic strain hardening is often invoked by introducing a symmetric and deviatoric stress-like tensorial
internal variable Tback called the back-stress, which acts to oppose the Cauchy stress T in the formu-
lation of an appropriate yield criterion and a corresponding flow rule for the material. The simplest
model for the evolution of such a back-stress is Prager’s linear kinematic-hardening rule (Prager,
1949):
1
Also recently published as Frederick and Armstrong (2007).
2
Lion uses the notation FM F, Fe Fe , Fin Fp , Fine Fpen , Finp Fpdis . Based on a statistical-volume-averaging argument for the
response of a polycrystalline aggregate, Clayton and McDowell (2003, Eq. (13)) have also introduced a Lion-type decomposition
Fp ¼ F ~i Fp a ‘‘meso-incompatibility tensor.”
~p ; they call F
~i F
en
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1835
their model; one such integration procedure involves the solution of a system of 13 highly non-linear
equations – cf., Eq. (53) of Vladimirov et al. (2008). Motivated by the work of these authors, it is the
purpose of this paper to:
(1) Develop a thermodynamically consistent rate-dependent plasticity theory with combined isotro-
pic and kinematic hardening of the Armstrong–Frederick type, based on the dual decomposi-
tions F ¼ Fe Fp and Fp ¼ Fpen Fpdis . The limit m ! 0, where m is a material rate-sensitivity
parameter in our theory, corresponds to a rate-independent model. Importantly, we develop
the theory based on the principle of virtual power and carefully lay down all assumptions
and specializations that we have adopted so that it may in the future be possible to generalize
the theory with other, more flexible isotropic and kinematic-hardening models.
(2) Develop a simple, stable, semi-implicit time-integration procedure for our rate-dependent con-
stitutive theory for implementation in displacement-based finite element programs. The proce-
dure that we develop is ‘‘simple” in the sense that it involves the solution of only one stiff non-
linear equation,3 rather than a system of non-linear equations. We show that our time-integration
procedure is stable for relatively large time steps, is first-order accurate, and is objective.
2. Kinematics
We consider a homogeneous body B identified with the region of space it occupies in a fixed refer-
ence configuration, and denote by X an arbitrary material point of B. A motion of B is then a smooth one-
to-one mapping x ¼ vðX; tÞ with deformation gradient, velocity and velocity gradient given by4
F ¼ rv; v ¼ v;
_ _ 1 :
L ¼ grad v ¼ FF ð2:1Þ
To model the inelastic response of materials, we assume that the deformation gradient F may be
multiplicatively decomposed as (Kröner, 1960)
F ¼ Fe Fp : ð2:2Þ
As is standard, we assume that
J ¼ det F > 0;
and consistent with this we assume that
def def
J e ¼ det Fe > 0; J p ¼ det Fp > 0; ð2:3Þ
e p
so that F and F are invertible. Here, suppressing the argument t:
Fp ðXÞ represents the local inelastic distortion of the material at X due to a ‘‘plastic mechanism.” This
local deformation carries the material into – and ultimately ‘‘pins” the material to – a coherent struc-
ture that resides in the structural space5 at X (as represented by the range of Fp ðXÞ);
Fe ðXÞ represents the subsequent stretching and rotation of this coherent structure, and thereby rep-
resents the corresponding ‘‘elastic distortion.”
3
With reference to the power-law model (9.35) in Section 9.3, the stiffness of the equations depends on the strain-rate-
sensitivity parameter m, and the stiffness increases to infinity as m tends to zero, the rate-independent limit. For small values of m
special care is required to develop stable constitutive time-integration procedures; cf., e.g., Lush et al. (1989).
4
Notation: We use standard notation of modern continuum mechanics. Specifically: r and Div denote the gradient and
divergence with respect to the material point X in the reference configuration; grad and div denote these operators with respect to
the point x ¼ vðX; tÞ in the deformed body; a superposed dot denotes the material time-derivative. Throughout, we write
Fe1 ¼ ðFe Þ1 , Fp> ¼ ðFp Þ> , etc. We write trA, symA, skwA, A0 , and sym0 A, respectively, for the trace, symmetric, skew, deviatoric,
and symmetric–deviatoric
pffiffiffiffiffiffiffiffiffiffiffi parts of a tensor A. Also, the inner product of tensors A and B is denoted by A : B, and the magnitude of A
by jAj ¼ A : A.
5
Also sometimes referred to as the ‘‘intermediate” or ‘‘relaxed” local space at X.
1836 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
L ¼ Le þ Fe Lp Fe1 ; ð2:4Þ
with
As is standard, we define the total, elastic, and plastic stretching and spin tensors through
9
D ¼ symL; W ¼ skwL; >
=
e e e e
D ¼ symL ; W ¼ skwL ; ð2:6Þ
>
;
Dp ¼ symLp ; Wp ¼ skwLp ;
so that L ¼ D þ W, Le ¼ De þ We , and Lp ¼ Dp þ Wp .
The right and left polar decompositions of F are given by
F ¼ RU ¼ VR; ð2:7Þ
where R is a rotation (proper-orthogonal tensor), while U and V are symmetric, positive-definite ten-
sors with
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
U¼ F> F; V¼ FF> : ð2:8Þ
Also, the right and left Cauchy–Green tensors are given by
(i) First, we make the standard assumption that plastic flow is incompressible, so that
J e ¼ J: ð2:15Þ
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1837
(ii) Second, from the outset we constrain the theory by limiting our discussion to circumstances
under which the material may be idealized as isotropic. For isotropic elastic–viscoplastic theo-
ries utilizing the Kröner decomposition it is widely assumed that the plastic flow is irrotational
in the sense that6
Wp ¼ 0: ð2:16Þ
p p
Then, trivially, L D and
F_ p ¼ Dp Fp ; with tr Dp ¼ 0: ð2:17Þ
Thus, using (2.1), (2.4), (2.5), and (2.17), we may write (2.4) for future use as
Next following Lion (2000), in order to account for energy storage mechanisms associated with plastic
flow, we introduce an additional kinematic constitutive assumption that Fp may be multiplicatively
decomposed into an energetic Fpen and a dissipative part Fpdis as follows:
Fp ¼ Fpen Fpdis ; det Fpen ¼ det Fpdis ¼ 1: ð2:19Þ
We call the range of Fpdis ðXÞ
the local substructural space. Thus Fpen
maps material elements from the
p
substructural space to the structural space (which is the range of F ðXÞ). Lion (2000) calls the substruc-
tural space as the ‘‘intermediate configuration of kinematic hardening.” A ‘‘physical interpretation” of
such an ‘‘intermediate configuration” is not completely clear7; here we simply treat it as a mathematical
construct resulting from Lion’s presumed decomposition (2.19).
Then, from (2.5)
6
A note on role of the plastic spin in isotropic solids: There are numerous publications discussing the role of plastic spin in theories
based on the Kröner decomposition (cf., e.g., Dafalias, 1998). Based on the classical principle of frame-indifference (cf., Section 3),
Green and Naghdi (1971) (cf., also Casey and Naghdi, 1980), introduced the notion of a change in frame of the intermediate structural
space. This notion leads to transformation laws for Fe and Fp of the form
F
p
¼ QFp ; F
e
¼ Fe Q > ;
in which Q ðX; tÞ is an arbitrary time-dependent rotation of the structural space. Unfortunately, Green and Naghdi viewed struc-
tural frame-indifference as a general principle; that is, a principle that stands at a level equivalent to that of conventional frame-
indifference. This view has been refuted by many workers (cf., e.g., Dashner, 1986, and the references therein). While we agree
with the view that structural frame-indifference is not a general principle, this hypothesis may be used to represent an important
facet of the behavior of a large class of polycrystalline materials based on the Kröner decomposition. Indeed, for polycrystalline
materials without texture the structural space is associated with a collection of randomly oriented lattices, and hence the evo-
lution of dislocations through that space should be independent of the frame with respect to which this evolution is measured.
Using the notion of a change in frame of the structural space, Gurtin and Anand (2005, p. 1714) have shown that within a frame-
work for isotropic plasticity based on the Kröner decomposition (which we follow here), one may assume, without loss in gener-
ality, that the plastic spin vanishes Wp ¼ 0.
Here, we adopt the assumption Wp ¼ 0 from the outset; we do so based solely on pragmatic grounds: when discussing finite
deformations for isotropic materials a theory without plastic spin is far simpler than one with plastic spin.
7
However, see Clayton and McDowell (2003) for a ‘‘statistical-volume-averaging” argument for the response of a polycrystalline
aggregate.
1838 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
)
Dpen ¼ symLpen ; Wpen ¼ skwLpen ;
ð2:22Þ
Dpdis ¼ symLpdis ; Wpdis ¼ skwLpdis ;
The right and left polar decompositions of Fpen and Fpdis are given by
Fpen ¼ Rpen Upen ¼ Vpen Rpen ; Fpdis ¼ Rpdis Updis ¼ Vpdis Rpdis ; ð2:29Þ
where Rpen and Rpdis are rotations, while Upen , Vpen , Updis , Vpdis are symmetric, positive-definite tensors with
qffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Upen ¼ Fp> p
en Fen ; Vpen ¼ Fpen Fp> en ; Updis ¼ Fp> p
dis Fdis ; Vpdis ¼ Fpdis Fp> dis : ð2:30Þ
Also, the corresponding right and left Cauchy–Green tensors are given by
3. Frame-indifference
Changes in frame (observer) are smooth time-dependent rigid transformations of the Euclidean
space through which the body moves. We require that the theory be invariant under such transforma-
tions, and hence under transformations of the form
v ðX; tÞ ¼ Q ðtÞðvðX; tÞ oÞ þ yðtÞ; ð3:1Þ
with Q ðtÞ a rotation (proper-orthogonal tensor), yðtÞ a point at each t, and o a fixed origin. Then, under
a change in frame, the deformation gradient transforms according to
F ¼ QF; ð3:2Þ
and hence
C ¼ C that is; C is invariant; ð3:3Þ
also F_ ¼ Q F_ þ Q_ F, and by (2.1)3 ,
Thus,
Fe ¼ Re Ue ! QFe ¼ QRe Ue ;
Fe ¼ Ve Re ! QFe ¼ QVe Q > QRe ;
and we may conclude from the uniqueness of the polar decomposition that
Upen ; Cpen ; Vpen ; Bpen ; Rpen ; Updis ; Cpdis ; Vpdis ; Bpdis ; Rpdis :
Following Gurtin and Anand (2005), the theory presented here is based on the belief that
the power expended by each independent ‘‘rate-like” kinematical descriptor be expressible in terms of an
associated force system consistent with its own balance.
However, the basic ‘‘rate-like” descriptors, namely, v, Le , Dp , Dpen , and Dpdis are not independent, since
by (2.18) and (2.28) they are constrained by
and it is not apparent what forms the associated force balances should take. It is in such situations that
the strength of the principle of virtual power becomes apparent, since the principle of virtual power
automatically determines the underlying force balances.
1840 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
We denote by Pt an arbitrary part (subregion) of the deformed body with n the outward unit nor-
mal on the boundary @Pt of Pt . The power expended on Pt by material or bodies exterior to Pt results
from a macroscopic surface traction tðnÞ, measured per unit area in the deformed body, and a macro-
scopic body force b, measured per unit volume in the deformed body, each of whose working accom-
panies the macroscopic motion of the body; the body force b presumed to account for inertia; that is,
granted the underlying frame is inertial,
b ¼ b0 qv;
_ ð4:3Þ
with b0 the noninertial body force and q the mass density in the deformed body. We therefore write
the external power as
Z Z
Wext ðPt Þ ¼ tðnÞ vda þ b vdv: ð4:4Þ
@Pt Pt
Assume that, at some arbitrarily chosen but fixed time, the fields v, Fe , Fpen (and hence F, Fp , Fpdis ) are
known, and consider the fields v, Le , Dp , Dpen , and Dpdis as virtual velocities to be specified independently
in a manner consistent with the constraints (4.1) and (4.2). That is, denoting the virtual fields by v ~, L~e ,
~ pen , and D
~ p, D
D ~ p to differentiate them from fields associated with the actual evolution of the body, we
dis
require that
~ ¼ L~e þ Fe D
grad v ~ p Fe1 ; ~ p ¼ 0;
trD ð4:7Þ
and
D ~ p þ sym Fp D
~p ¼ D ~ p p1 ; ~ p ¼ 0:
~ p ¼ trD
en en dis Fen trD en dis ð4:8Þ
respectively, for the external and internal expenditures of virtual power, the principle of virtual power is
the requirement that the external and internal powers be balanced. That is, given any part Pt ,
Wext ðPt ; VÞ ¼ Wint ðPt ; VÞ for all generalized virtual velocities V: ð4:10Þ
Also, since the change in frame is arbitrary, if we choose it such that Q is an arbitrary time-independent
rotation, and that Q_ ¼ 0, we find from (4.13) that
h i
ðQ > Se Q Þ Se : L~e þ J 1 ðTp Tp Þ : D en en dis
~ p ¼ 0:
~ p þ S Sdis : D
~ p þ S Sen : D
dis
~e , D
Since this must hold for all L ~ p, D ~ p , we find that the stress Se transforms according to
~ pen , and D
dis
Se : Q_ ¼ 0;
or that the tensor Se is symmetric,
Se ¼ Se> : ð4:16Þ
e
Thus, the elastic stress S is frame-indifferent and symmetric.
1842 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
L~e ¼ grad v
~: ð4:17Þ
For this choice of V, (4.10) yields
Z Z Z
~da þ
tðnÞ v ~dv ¼
bv Se : grad v
~dv: ð4:18Þ
@Pt Pt Pt
T ¼ T> ð4:23Þ
as the standard macroscopic Cauchy stress and (4.21) as the local macroscopic force balance. Since we are
working in an inertial frame, so that (4.3) is satisfied, (4.21) reduces to the local balance law for linear
momentum,
divT þ b0 ¼ qv;
_ ð4:24Þ
with b0 the noninertial body force.
v ~ p 0;
~ 0 and D ð4:25Þ
dis
L~e ¼ Fe D
~ p Fe1 ~p ¼ D
and Den
~ p; ð4:26Þ
def
Me ¼ JFe> TFe> ; ð4:28Þ
which in general is not symmetric. Then, on account of our choice (4.25)1 the external power vanishes,
so that, by (4.10), the internal power must also vanish, and satisfy
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1843
Z
Wint ðPt ; VÞ ¼ J 1 ðTp þ Sen Me Þ : D
~ p dv ¼ 0:
Pt
~ p , a standard argument
Since this must be satisfied for all Pt and all symmetric and deviatoric tensors D
yields the first microforce balance
ð4:29Þ
v ~ p 0;
~ 0 and D ð4:30Þ
~ p arbitrarily, and let
also, choose the virtual field D dis
~ p p1
~ p ¼ sym Fp D
D en en dis Fen ð4:31Þ
ð4:34Þ
5. Free-energy imbalance
We consider a purely mechanical theory based on a second law requiring that the temporal increase
in free energy of any part Pt be less than or equal to the power expended on Pt . The second law therefore
takes the form of a dissipation inequality
Z _
wJ 1 dv 6 Wext ðPt Þ ¼ Wint ðPt Þ; ð5:1Þ
Pt
with w the free energy, measured per unit volume in the structural space.
Since J 1 dv with J ¼ det F represents the volume measure in the reference configuration, and since
Pt deforms with the body,
Z _ Z
wJ 1 dv ¼ _ 1 dv:
wJ
Pt Pt
Thus, since Pt is arbitrary, we may use (4.6), (4.22), and (4.23) to localize (5.1); the result is the local
free-energy imbalance
1844 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
C_ e ¼ Fe> F_ e þ F_ e> Fe :
Hence, since Te is symmetric,
6. Constitutive theory
To account for the classical notion of isotropic strain hardening we introduce a positive-valued sca-
lar internal state-variable S, which has dimensions of stress. We refer to S as the isotropic deformation
resistance. Since S is a scalar field, it is invariant under a change in frame.
Next, guided by the free-energy imbalance (5.6), and by experience with previous constitutive the-
ories, we assume the following special set of constitutive equations:
9
w¼w ðeÞ ðCe Þ þ w
ðpÞ ðBp Þ; >
en >
>
>
e ðCe Þ;
Te ¼ T >
>
>
>
p >
=
Sen ¼ Sen ðB Þ; en
ð6:1Þ
Tp ¼ T p ðDp ; SÞ; >
>
>
>
dis ðDp ; dp ; SÞ;
Sdis ¼ S >
>
dis >
>
p
>
;
S_ ¼ hðd ; SÞ;
where
p def
d ¼ jDp j ð6:2Þ
p
is the scalar flow rate corresponding to D . Note that since w is the free energy per unit volume of the
ðeÞ is chosen to depend on Ce and w
structural space, w ðpÞ on Bp , because Ce and Bp are squared stretch-
en en
like measures associated with the intermediate structural space. Also note that we have introduced a
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1845
p
possible dependence of Sdis on d because for some materials we anticipate that Dpdis (and thereby Sdis )
p
may be constrained by the magnitude of d .
The energy w ðeÞ represents the standard energy associated with intermolecular interactions, and the
energy w ðpÞ is a ‘‘defect-energy” associated with plastic deformation. For metallic materials w ðpÞ may be
attributed to an energy stored in the complex local microstructural defect state, which typically in-
cludes dislocations, sub-grain boundaries, and local incompatibilities at second-phase particles and
grain boundaries. At the macroscopic continuum level, the ‘‘defect-energy” w ðpÞ leads to the develop-
ment and evolution of the energetic stress Sen , and this allows one to phenomenologically account for
strain-hardening phenomena commonly called kinematic hardening. Isotropic hardening is modeled by
the evolution (6.1)6 of the isotropic deformation resistance S.
Finally, note that on account of the transformation rules listed in the paragraph containing (5.7)
and since S is also invariant, the constitutive equations (6.1) are frame-indifferent.
From (6.1)1
ðeÞ ðCe Þ
@w ðpÞ ðBp Þ
@w
w_ ¼ e : C_ e þ en
: B_ pen ;
@C @Bpen
and, using8
Hence, satisfaction of the free-energy imbalance (5.6) requires that the constitutive equations (6.1)
satisfy
ðeÞ ðCe Þ ðpÞ p
1 e e @w en ðBp Þ 2 @ w ðBen Þ Bp
T ðC Þ e : C_ e þ S en p en : Dpen
2 @C @Ben 0
þT dis ðDp ; dp ; SÞ : Dp P 0;
p ðDp ; SÞ : Dp þ S ð6:4Þ
dis dis
and hold for all arguments in the domains of the constitutive functions, and in all motions of the body.
Thus, sufficient conditions that the constitutive equations satisfy the free-energy imbalance are that9
(i) the free-energy determine the stresses Te and Sen via the stress relations
ðeÞ e
e ðCe Þ ¼ 2 @ w ðC Þ ;
T ð6:5Þ
@Ce
ðpÞ p
en ðBp Þ ¼ 2 @ w ðBen Þ Bp
S : ð6:6Þ
en p en
@Ben 0
(ii) the plastic distortion-rates Dp and Dpdis satisfy the dissipation inequality
Tp : Dp þ Sdis : Dpdis P 0: ð6:7Þ
We assume henceforth that (6.5) and (6.6) hold in all motions of the body. We assume further that the
material is strictly dissipative in the sense that
8
Recall that Wpen ¼ 0, cf., (2.26).
9
We content ourselves with constitutive equations that are only sufficient, but generally not necessary for compatibility with
thermodynamics.
1846 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
6.2. Isotropy
The following definitions help to make precise our notion of an isotropic material (cf., Anand and
Gurtin, 2003):
þ
(i) Orth = the group of all rotations (the proper-orthogonal group);
(ii) the symmetry group GR , is the group of all rotations of the reference configuration that leaves the
response of the material unaltered;
(iii) the symmetry group GS at each time t, is the group of all rotations of the structural space that
leaves the response of the material unaltered;
(iv) the symmetry group GSS at each time t, is the group of all rotations of the substructural space that
leaves the response of the material unaltered.
Note that there is a slight difficulty in attaching a ‘‘physical meaning” to a rotation in GSS . Since the
substructural space itself is a mathematical construct defined as the range of the linear map Fpdis , a rota-
tion Q 2 GSS should be considered as a rotation in a ‘‘thought experiment” involving the substructural
space; such a rotation may never be ‘‘physically attainable.” We now discuss the manner in which the
basic fields transform under such transformations, granted the physically natural requirement of
invariance of the internal power or equivalently, the requirement that
Fp ! Fp Q ; Fpdis ! Fpdis Q and Fe and Fpen are invariant; and hence Ce and Cpen are invariant;
ð6:12Þ
so that, by (2.5), (2.12), and (2.21),
" ! #
~ ðpÞ ðI p Þ
@w Ben
¼ 2 Fp> Bpen Fp> ;
en
@Bpen en
0
" ! #
~ ðpÞ ðI p Þ
@w Ben
¼ 2 Fp> Fpen ; ð6:30Þ
en
@Bpen
0
7. Flow rules
Henceforth, in accord with common terminology we call the symmetric and deviatoric tensor Sen a
back-stress, and we denote an effective deviatoric Mandel stress by
def
Meeff 0
¼ Me0 Sen : ð7:1Þ
Then, upon using the constitutive relation (6.1)4 and the first microforce balance (4.29), together with
symmetry of the Mandel stress discussed above, a central result of the theory is the first flow rule
Meeff p ðDp ; SÞ:
¼T ð7:2Þ
0
Next, on account of the symmetry and deviatoric nature of the plastic stress Mpen (cf., (6.30)), the second
microforce balance (4.34) and the constitutive equation (6.4)5 yield the second flow rule of the theory:
dis ðDp ; d ; SÞ: p
Mpen ¼ S dis ð7:3Þ
We now make two major assumptions concerning the plastic flow for isotropic materials:
Guided by (7.5), we assume henceforth that the dissipative flow stress Tp is parallel to and
points in the same direction as Np so that
p ðdp ; Np ; SÞ ¼ Yðdp ; Np ; SÞNp ;
T ð7:6Þ
where
p p ðdp ; Np ; SÞ : Np
Yðd ; Np ; SÞ ¼ T ð7:7Þ
represents a scalar flow strength of the material.
Similarly we assume that dissipative flow stress Sdis is parallel to and points in the same direc-
tion as Npdis so that
dis ðdp ; Np ; dp ; SÞ ¼ Y dis ðdp ; Np ; dp ; SÞNp ;
S ð7:8Þ
dis dis dis dis dis
where
p p dis ðdp ; Np ; dp ; SÞ : Np
Y dis ðddis ; Npdis ; d ; SÞ ¼ S dis dis dis ð7:9Þ
represents another scalar flow strength of the material.
We refer to the assumptions (7.6) and (7.8) as the codirectionality hypotheses.10
10
These assumptions corresponds to the classical notion of maximal dissipation in Mises-type theories of metal plasticity.
1850 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
Thus, using (7.10) and (7.6), the flow rule (7.2) reduces to,
p
ðMeeff Þ0 ¼ Yðd ; SÞNp ; ð7:12Þ
which immediately gives
ðMeeff Þ0
Np ¼ ; ð7:13Þ
jðMeeff Þ0 j
and
p
jðMeeff Þ0 j ¼ Yðd ; SÞ: ð7:14Þ
p
When jðMeeff Þ0 j
and S are known, (7.14) serves as an implicit equation for the scalar flow rate d .
Next, using (7.11) and (7.8), the flow rule (7.3) reduces to,
p p
Mpen ¼ Y dis ðddis ; d ; SÞNpdis ; ð7:15Þ
which gives
Mpen
Npdis ¼ ; ð7:16Þ
jMpen j
and
p p
jMpen j ¼ Y dis ðddis ; d ; SÞ: ð7:17Þ
p
When jðMeeff Þ0 j
and S are known, (7.17) serves as an implicit equation for the scalar flow rate ddis .
Finally, using (7.6), (7.10), and (7.14), as well as (7.8), (7.11), and (7.17) the dissipation inequality
(6.7) reduces to
p p
jðMeeff Þ0 jd þ jMpen jddis P 0: ð7:18Þ
e
where ICe and IBpen are the lists of the principal invariants of C and Bpen , respectively.
(2) Cauchy stress.
and is symmetric.
The back-stress is given by
!
~ ðpÞ ðI p Þ
@w Ben
Sen ¼ 2 Bpen ; ð8:4Þ
@Bpen
0
F_ p ¼ Dp Fp ; ð8:7Þ
p
with D given by
p ðMeeff Þ0
Dp ¼ d Np ; Np ¼ ; ð8:8Þ
jðMeeff Þ0 j
p
where the scalar flow rate d is obtained by solving the scalar strength relation
p
jðMeeff Þ0 j ¼ Yðd ; SÞ; ð8:9Þ
p
for given ðMeeff Þ0 and S, where Yðd ; SÞ is a rate-dependent flow strength. The evolution equation
for Fpdis is
p Mpen
Dpdis ¼ ddis Npdis ; Npdis ¼ ; ð8:11Þ
jMpen j
p
where the scalar flow rate ddis is obtained by solving the scalar strength relation
p p
jMpen j ¼ Y dis ðddis ; d ; SÞ; ð8:12Þ
p p p
for given Mpen ,
d , and S, where Y dis ðddis ; d ; SÞ is another rate-dependent flow strength.
(5) Evolution equation for S.
p
S_ ¼ hðd ; SÞ: ð8:13Þ
The evolution equations for Fp , Fpdis , and S need to be accompanied by initial conditions. Typical ini-
tial conditions presume that the body is initially (at time t ¼ 0, say) in a virgin state in the sense that
FðX; 0Þ ¼ Fp ðX; 0Þ ¼ Fpdis ðX; 0Þ ¼ 1; SðX; 0Þ ¼ S0 ð¼ constantÞ; ð8:14Þ
e p p
so that by F ¼ F F and F ¼ Fpen Fpdis e
we also have F ðX; 0Þ ¼ 1 and Fpen ðX; 0Þ ¼ 1.
In this section, based on experience with existing recent theories of viscoplasticity with isotropic
and kinematic hardening, we specialize our constitutive theory by imposing additional constitutive
assumptions.
1852 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
where ðre1 ; re2 ; re3 Þ are the orthonormal eigenvectors of Ce and Ue , and ðke1 ; ke2 ; ke3 Þ are the positive eigen-
values of Ue . Instead of using the invariants ICe , the free-energy wðeÞ for isotropic materials may be
alternatively expressed in terms of the principal stretches as
ðeÞ ðke ; ke ; ke Þ:
wðeÞ ¼ w ð9:2Þ
1 2 3
ðeÞ ðke ; ke ; ke Þ
@w X3 ðeÞ ðke ; ke ; ke Þ @ke X
@w 3 ðeÞ ðke ; ke ; ke Þ @ xi
1 @w
Te ¼ 2 1
e
2 3
¼2 1
e
2 3 i
e ¼ e
1 2 3
: ð9:3Þ
@C i¼1
@ki @C i¼1
ki @kei @Ce
Assume that the squared principal stretches xei are distinct, so that the xei and the principal directions
rei may be considered as functions of Ce ; then
@ xei
¼ rei rei ; ð9:4Þ
@Ce
and, granted this, (9.4) and (9.3) imply that
X3
1 @w ðeÞ ðke ; ke ; ke Þ
Te ¼ e
1 2 3
rei rei : ð9:5Þ
i¼1
k i @kei
Next, since Me ¼ Ce Te (cf., (8.3)), use of (9.1) and (9.5) gives the Mandel stress as
X
3 ðeÞ ðke ; ke ; ke Þ
@w
Me ¼ kei 1 2 3
rei rei : ð9:7Þ
i¼1
@kei
Let
def
X
3
Ee ¼ ln Ue ¼ Eei rei rei ; ð9:8Þ
i¼1
w ðeÞ ðEe ; Ee ; Ee Þ;
ðeÞ ðke ; ke ; ke Þ ¼ w ð9:10Þ
1 2 3 1 2 3
We consider the following simple generalization of the classical strain-energy function of infinitesimal
isotropic elasticity which uses a logarithmic measure of finite strain11
T ¼ J 1 Re Me Re> : ð9:15Þ
where ðl1 ; l2 ; l3 Þ are the orthonormal eigenvectors of Bpen and Vpen , and ða1 ; a2 ; a3 Þ are the positive eigen-
values Vpen . Instead of using the invariants IBpen , the free-energy wðpÞ may alternatively be expressed as
ðpÞ ða1 ; a2 ; a3 Þ:
wðpÞ ¼ w ð9:17Þ
Then, by the chain-rule
ðpÞ ða1 ; a2 ; a3 Þ X
@w 3 ðpÞ ða1 ; a2 ; a3 Þ @ai
@w 1X 3 ðpÞ ða1 ; a2 ; a3 Þ @bi
1 @w
p ¼ p ¼ : ð9:18Þ
@Ben i¼1
@a i @B en 2 i¼1
ai @ai @Bpen
Assume that bi are distinct, so that the bi and the principal directions li may be considered as functions
of Bpen . Then,
@bi
¼ li li ; ð9:19Þ
@Bpen
and, granted this, (9.18) implies that
ðpÞ ða1 ; a2 ; a3 Þ 1 X
@w 3 ðpÞ ða1 ; a2 ; a3 Þ
1 @w
p ¼ li li : ð9:20Þ
@Ben 2 i¼1
ai @ai
Also, use of (9.16) and (9.20) in (8.4) gives the deviatoric back-stress as
!
X
3 ðpÞ ða1 ; a2 ; a3 Þ
@w
Sen ¼ ai li li : ð9:21Þ
i¼1
@ai
0
Let
def
X
3
Epen ¼ ln Vpen ¼ ln ai li li : ð9:22Þ
i¼1
11
Vladimirov et al. (2008) employ a Neo-Hookean free-energy function modified for elastic compressibility (cf., their Eq. (28)).
Here, as an alternate, we use the classical strain-energy function for infinitesimal elasticity, but employ the large deformation
Hencky-logarithmic elastic strain Ee . Anand (1979, 1986) has shown that the simple free-energy function (9.12) is useful for
moderately large elastic stretches. Additionally, use of the logarithmic strain helps in developing implicit time-integration
procedures based on the exponential map (cf., Appendix A).
1854 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
denote a logarithmic energetic strain with principal values ðln ai Þ. Note that since trVpen ¼ a1 a2 a3 ¼ 1,
trðEpen Þ ¼ ln a1 þ ln a2 þ ln a3 ¼ lnða1 a2 a3 Þ ¼ 0; ð9:23Þ
and hence the strain tensor Epen is traceless, Next, consider the following simple defect energy:
h i
ðpÞ ðEp Þ ¼ BjEp j2 ¼ B ðln a1 Þ2 þ ðln a2 Þ2 þ ðln a3 Þ2 ;
w ð9:24Þ
en en
with B > 0, which, since Epen is deviatoric, is analogous to (9.12). Then using (9.21), (9.22), and (9.23)
Sen ¼ 2BEpen ; ð9:25Þ
we call the positive-valued constitutive parameter B the back-stress modulus.
Then the plastic Mandel stress (4.33) becomes
Mpen ¼ Fp> p> p> p p p1
en Sen Fen ¼ R en V en ð2BEen ÞV en Rpen ;
or
Mpen ¼ Rp> p
en Sen R en : ð9:26Þ
12
pffiffiffi
We absorb inconsequential factors of 2 in writing (9.30).
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1855
where m > 0, a constant, is a rate-sensitivity parameter and m0 > 0, also a constant, is a reference flow-
rate. The power-law function allows one to characterize nearly rate-independent behavior, for which m
is very small. Further, granted the power-law function (9.34), the expression (9.33) has the specific
form
m1
s
mp ¼ m0 : ð9:35Þ
S
Next consider the strength relation
p p
jMpen j ¼ Y dis ðddis ; d ; SÞ; ð9:36Þ
appearing in (8.12). Define another equivalent shear stress by
1
sen def
¼ pffiffiffi jMpen j; ð9:37Þ
2
and another equivalent shear strain rate by
pffiffiffi pffiffiffi
mpdis def p
¼ 2ddis ¼ 2jDpdis j; ð9:38Þ
13
respectively. Using the definitions (9.37) and (9.38), we rewrite the strength relation (9.36) as
sen ¼ Y dis ðmpdis ; mp ; SÞ: ð9:39Þ
p p
A specialization of Y dis ðm dis ; m ; SÞ which leads to Armstrong and Frederick (1966) type kinematic hard-
ening is
B
Y dis ðmpdis ; mp ; SÞ ¼ mpdis ; ð9:40Þ
cm p
where B is the back-stress modulus (cf., (9.25)), and c P 0 is a dimensionless constant. Using (9.40),
(9.39) becomes
B
sen ¼ mpdis ; ð9:41Þ
cmp
so that the term ðB=cmp Þ represents a ‘‘pseudo”-viscosity (Dettmer and Reese, 2004). The relation
(9.41) may be inverted to give
sen
mpdis ¼ cmp ; ð9:42Þ
B
which shows that mpdis ¼ 0 when either c ¼ 0 or when mp ¼ 0. Thus note that the constitutive assump-
tion (9.42) constrains the theory such that Dpdis ¼ 0 whenever Dp ¼ 0, and this in turn will lead to no
change in the back-stress Sen when Dp ¼ 0, that is when there is no macroscopic plastic flow.
The evolution equation for the isotropic deformation resistance is specialized in a rate-independent
form as
S_ ¼ hðSÞmp ð9:43Þ
with hðSÞ a hardening function, which we take to be given by (Brown et al., 1989):
( a
h0 1 SS for S0 6 S 6 S ;
hðSÞ ¼ ð9:44Þ
0 for S P S ;
13
pffiffiffi
Again absorbing inconsequential factors of 2 in writing (9.39).
1856 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
where S , a, and h0 are constant moduli with S > S0 , a P 1, and h0 > 0. The hardening function (9.44)
is strictly decreasing for S0 6 S 6 S and vanishes for S P S .
Finally, using (9.28), (9.29), (9.37), (9.38), and (9.42), the dissipation inequality (7.18) reduces to
c
s þ ðsen Þ2 mp P 0: ð9:45Þ
B
In this section, we summarize the specialized form of our theory, which should be useful in appli-
cations. The theory relates the following basic fields:
T ¼ J 1 Re Me Re> ; ð9:55Þ
where
Me ¼ 2GEe0 þ KðtrEe Þ1; ð9:56Þ
is an elastic Mandel stress.
The symmetric and deviatoric back-stress is defined by
Sen ¼ 2BEpen ; ð9:57Þ
p
and the driving stress for D is the effective stress given by
ðMeeff Þ0 ¼ Me0 Sen : ð9:58Þ
The corresponding equivalent shear stress is given by
1
s ¼ pffiffiffi jðMeeff Þ0 j: ð9:59Þ
2
The driving stress for Dpdis is the stress measure
Mpen ¼ Rp> p
en Sen R en : ð9:60Þ
The corresponding equivalent shear stress is given by
1
sen ¼ pffiffiffi jMpen j: ð9:61Þ
2
(3) Flow rules.
The evolution equation for Fp is
9
F_ p ¼ Dp Fp ; Fp ðX; 0Þ ¼ 1; >
>
>
e =
ðM Þ
Dp ¼ mp 2effs 0 ; ð9:62Þ
>
>
>
;
mp ¼ m0 sS 1=m ;
where m0 is a reference plastic strain rate with units of 1=time, and m is a strain-rate-sensitivity
parameter.
The evolution equation for Fpdis is
9
F_ pdis ¼ Dpdis Fpdis ; Fpdis ðX; 0Þ ¼ 1; >
>
p =
M
Dpdis ¼ mpdis 2senen ; ð9:63Þ
>
>
;
mpdis ¼ c sBen mp ;
where c P 0 is a dimensionless constant.
1858 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
First, in order to demonstrate that the constitutive theory captures the various isotropic and kine-
matic-hardening effects, we carried out a numerical simulation of a symmetric strain-cycle in simple
tension and compression. The material parameters used in our numerical study are:
G ¼ 80 GPa; K ¼ 175 GPa;
m0 ¼ 0:001 s1 ; m ¼ 0:02;
ð10:1Þ
S0 ¼ 100 MPa; h0 ¼ 1250 MPa; a ¼ 1 S ¼ 250 MPa;
B ¼ 40 GPa; c ¼ 400;
the value of m ¼ 0:02 is intended to represent a nearly rate-independent material.
The axial strain-cycle was imposed between true strain limits of ¼ 0:02 at an absolute axial true
strain rate of j_ j ¼ 0:01 s1 ; a fixed time step Dt ¼ 0:01 s was used in the numerical simulations. The cal-
culations were performed for: (i) no hardening, h0 ¼ 0 and B ¼ 0; (ii) isotropic hardening only, h0 –0 and
B ¼ 0; (iii) kinematic hardening only, h0 ¼ 0 and B–0; and finally (iv) combined kinematic and isotropic
hardening, h0 –0 and B–0. Plots of the axial stress r versus the axial strain are shown in Fig. 1(a). With no
hardening, we obtain an elastic-perfectly-plastic response, as expected. When isotropic hardening only is
allowed, we observe an increase in the flow stress, and the stress level at which plastic flow recommences
on strain reversal is equal in magnitude to the stress level from which the reversal in strain was initiated.
In the case of kinematic hardening only, we again observe an increase in the flow stress, but this time the
magnitude of the stress level at which plastic flow recommences on strain reversal is substantially smaller
than the stress level from which the reversal in strain direction was initiated; a clear manifestation of the
Bauschinger effect. When isotropic hardening is combined with kinematic hardening, we see the contin-
ual evolution in the flow stress associated with combined isotropic and kinematic hardening during for-
ward straining, and a clear Bauschinger effect due to the kinematic hardening upon strain reversal.
Next, we concentrate on numerical solutions related to homogeneous simple shear. With respect to
a rectangular Cartesian coordinate system with origin o and orthonormal base vectors fei ji ¼ 1; 3g, a
simple shearing motion is described by
x ¼ X þ ðmtÞX 2 e1 ; ð10:2Þ
with m a shear strain rate, and C ¼ mt is the amount of shear. Throughout this section we use the mate-
rial parameters summarized in (10.1).
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1859
a 600
400
−400
−600
−0.02 −0.01 0 0.01 0.02
Axial strain, ε
b 300
200
(MPa)
100
No hardening
Isotropic only
12
0
Stress, T
Kinematic only
Combined
−100
−200
−300
0 0.01 0.02 0.03 0.04 0.05
Amount of shear, Γ
Fig. 1. (a) Comparison of axial stress r versus axial strain for various types hardening in a symmetric strain-cycle simulation
in simple tension and compression. (b) Comparison of shear stress T 12 versus shear strain C for various types of hardening in
reversed simple shear.
The specific goals of our numerical solutions related to simple shear are as follows:
To further demonstrate that the constitutive theory captures the major features of isotropic and
kinematic hardening in the case of simple reversed shear.
To investigate the convergence and stability properties of the time-integration scheme in simple
reversed shear.
To demonstrate that the constitutive equations do not suffer from oscillations in the stress response
at large shear strains.
To demonstrate that the constitutive equations and the time-integration procedure are indeed
objective.
1860 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
First, we wish to further demonstrate that the constitutive theory captures the various isotropic
and kinematic-hardening effects. To this end, we consider simple shear, as described above, for a time
of 5 s at a shear strain rate of m ¼ 0:01 s1 for a total shear strain C ¼ mt ¼ 0:05, and then at a shear
strain rate of m ¼ 0:01 s1 for another 5 s to complete one reversal of strain, while using a fixed time
step Dt ¼ 0:01 s. As for simple tension/compression, the calculations were performed for: (i) no hard-
ening, h0 ¼ 0 and B ¼ 0; (ii) isotropic hardening only, h0 –0 and B ¼ 0; (iii) kinematic hardening only,
h0 ¼ 0 and B–0; and finally (iv) combined kinematic and isotropic hardening, h0 –0 and B–0, and the
results are shown in Fig. 1(b). Again, we observe the important qualitative features of each case in sim-
ple reversed shear, i.e. elastic-perfectly-plastic response in the case of no hardening, a continual in-
crease in the flow stress in cases involving isotropic hardening, and a clear Bauschinger effect in
cases involving kinematic hardening.
Next, we investigated the stability and convergence properties of our time-integration procedure.
Since our integration procedure is semi-implicit rather than fully-implicit, one might be concerned
that there may be a restrictive time step constraint for stability. To ascertain any time step restrictions,
we performed a convergence test. Reversed simple shear to a maximum shear strain of C ¼ 0:05 at a
shear strain rate of jmj ¼ 0:01 s1 was considered while increasingly coarsening the time step. The spe-
cific time steps Dt of
300
200
(MPa)
100
Δ t = 1×10−3 s
12
−2
0 Δ t = 1×10 s
Stress, T
−1
Δ t = 1×10 s
Δt=1s
−100
−200
−300
0 0.01 0.02 0.03 0.04 0.05
Amount of shear, Γ
Fig. 2. Comparison of solutions for reversed simple shear for various time steps.
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1861
8.5
8 1
1
7.5
log(E)
6.5
6
−2.5 −2 −1.5 −1 −0.5 0 0.5
log(Δ t)
Fig. 3. Convergence diagram for the time-integration procedures demonstrating first-order accuracy.
We wish next to demonstrate that the constitutive equations do not produce shear-stress oscilla-
tions at large strains – a feature which has plagued numerous previous large deformation constitutive
theories for kinematic hardening. To this end, we performed a monotonic simple shear simulation to a
very large final shear strain of C ¼ 10, at a shear strain rate of m ¼ 0:01 s1 using a fixed time step
Dt ¼ 0:1 s. The shear T 12 and normal components, T 11 and T 22 , of the Cauchy stress T are plotted in
Fig. 4(a). Clearly there are no oscillations in the T 12 versus C response; the shear stress rises and satu-
rates (because of the chosen value of the material parameters) at a shear strain of C
1:5. We have also
carried out a similar calculation with all other material parameters the same as before, but by setting the
dynamic recovery parameter c ¼ 0; this corresponds to the case of ‘‘linear” kinematic hardening. The
variation of the shear and normal components of the Cauchy stress T for this case are plotted in
Fig. 4(b). Here, since there is no dynamic recovery for the kinematic hardening, the stress levels for this
physically unrealistic case get quite large, but the T 12 versus C curve still does not show any oscillations;
the peak in this curve is due to the use of a logarithmic energetic strain lnðVpen Þ and not due to any oscil-
latory behavior in the evolution of the back-stress.
Finally, in order to demonstrate the frame-indifference of the constitutive equations and numerical
algorithm, we have carried out a numerical calculation for simple shear with superposed rigid rota-
tion, which is described by Weber et al. (1990):
x ¼ Q ðtÞ½X þ ðmtÞX 2 e1 ; ð10:3Þ
with
Q ðtÞ ¼ ðe1 e1 þ e2 e2 Þ cosðxtÞ þ ðe1 e2 e2 e1 Þ sinðxtÞ þ e3 e3 ; ð10:4Þ
a rotation about the e3 -axis. In performing our numerical calculation, we used the material parameters
listed in (10.1), and the calculation was performed with x ¼ 0:1p radians per second and m ¼ 0:01 per
second, for t 2 ½0; 20, so that h xt 2 ½0; 2p and C mt 2 ½0; 0:2. The calculation was carried by using
a fixed time step of Dt ¼ 0:1 s, which corresponds to a shear strain increment of DC ¼ 0:001 and a rota-
tion increment of Dh ¼ 3:6 . Snapshots of the initial and deformed geometry at a few representative
stages of the simple shear plus superposed rotation are shown in Fig. 5. Let the solution for the stress cor-
responding to the motion (10.3) be denoted by T ðtÞ. The result of the shear component of the ‘‘unrotat-
ed” Cauchy stress ðQ ðtÞ> T ðtÞQ ðtÞÞ versus the amount of shear C, obtained from the numerical
calculation is shown in Fig. 6. Also shown in this figure is T 12 versus C for the motion (10.3) with
Q ðtÞ ¼ 1. The two shear stress versus shear strain curves are indistinguishable from each other; the
1862 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
a 400
350
300
50
−50
0 2 4 6 8 10
Amount of shear, Γ
b 1.5
5
x 10
T
12
T11
1
T
22
0.5
Stress (MPa)
−0.5
−1
−1.5
0 2 4 6 8 10
Amount of shear, Γ
Fig. 4. Simple shear to large shear strain (a) with dynamic recovery c–0; and (b) without dynamic recovery, c = 0. Note that
there are no shear-stress oscillations in either case.
excellent agreement between the two calculations verifies the objectivity of the constitutive equations
and time-integration procedure.
Next, in order to exercise the algorithmic tangent in a case of inhomogeneous deformation, we consider
the cyclic deformation of a curved bar of circular cross-section. A finite element mesh for the curved bar is
shown in Fig. 7(a). The bar, which has a diameter of 50 mm, curves through a radius of 100 mm for 90 , and
then extends straight for another 100 mm. The mesh consists of 2184 Abaqus-C3D8R elements. The nodes
on the face denoted as ‘‘fixed”-face in Fig. 7(a) have prescribed null displacements, u1 ¼ u2 ¼ u3 ¼ 0, while
the boundary conditions on the face denoted as the ‘‘moved”-face were specified as
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1863
Fig. 5. Simple shear with superposed rigid rotation. The deformed geometry at each stage is shown with a solid line, while the
dashed line represents the initial geometry.
occurring at any integration point in the model during an increment. The value of Dcpmax was kept close
to a specified nominal value Dcps . From Fig. 2, we note that the accuracy of the solution begins to
1864 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
350
Shear and rotation
Shear only
300
250
Stress, T12 (MPa)
200
150
100
50
0
0 0.05 0.1 0.15 0.2
Amount of shear, Γ
Fig. 6. Comparison of results for T 12 versus C curves from shear with superposed rotation against shear with no superposed
rotation.
degrade for shear strain increments of around 1 103 ; accordingly, we chose Dcps to be half this va-
lue, 0:5 103 . The automatic time-stepping algorithm operated to keep the ratio
Dcpmax
R ð10:6Þ
Dcps
close to 1.0 by adjusting the size of the time increments.14
As the curved bar is cyclically displaced in the ðe2 ; e3 Þ-plane at the ‘‘moved-face”, the section of
the bar near the ‘‘fixed”-face is subjected to reversed-torsion, while the part of the bar towards the
‘‘moved”-face is subjected to reversed-bending, resulting in states of shear, tension, and compres-
sion in various parts of the body. With increasing displacement magnitude at the ‘‘moved”-face,
inhomogeneous plastic deformation initiates and spreads in various disparate regions of the curved
bar. The integrated reaction force in the 2-direction on the ‘‘moved”-face is plotted against the pre-
scribed u2 -displacement in Fig. 7(b). This resulting overall cyclic load–displacement curve is the re-
sult of the complex evolution of the internal variables controlling the combined isotropic and
kinematic hardening at the various sections of the bar which are undergoing plastic deformation.
The overall load versus displacement response is smooth, and one can clearly see the effects of
both kinematic hardening in the prominent Bauschinger effect and isotropic hardening in the con-
tinual increase in the magnitude of the load with cyclic deformation.
This example shows that our semi-implicit integration procedure, when coupled with our heuris-
tic automatic time-stepping algorithm and our approximate algorithmic tangent, is effective in
14
After an equilibrium solution for a time increment Dtn ¼ t nþ1 tn was found, the value of the ratio R was checked to determine
whether this solution would be accepted or not. If R was greater than 1.25, then the solution was rejected, and a new time
increment was used that was smaller by the factor ð0:85=RÞ. If R 6 1:25, then the solution was accepted, and the value of R was
used to determine the first trial size of the next time increment. The following algorithm was used:
Note that the measure Dcpmax was allowed to exceed the user specified value Dcps by up to 25%. This was done to avoid having to
recalculate increments that came out just slightly above the specified nominal value, but were otherwise essentially acceptable.
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1865
b 150
100
50
Load (kN)
−50
−100
−150
−20 −10 0 10 20
Displacement, u , (mm)
2
Fig. 7. (a) Undeformed and deformed meshes for the cyclic loading of a curved bar. (b) Resulting tip load versus tip
displacement.
obtaining an accurate, stable, and efficient solution in an inhomogeneous large deformation implicit
finite element calculation.
As our final numerical example, we consider a simple plane-strain sheet forming operation in
which an aluminum sheet is plastically bent about a mandrel of fixed radius and then unloaded.
We use the recent experimental data of Cao et al. (2008) for AA6111-T4 sheets to estimate the mate-
rial parameters in our model. The fit of the model to the experimental data is shown in Fig. 8, and the
estimated material parameters are listed in (10.8):
400
Experiment
Model
300
100
−100
−200
−300
−400
0 0.02 0.04 0.06 0.08 0.1 0.12
True axial strain
Fig. 8. Fit of the model to experimental data for AA6111-T4 sheet material. The data is from Cao et al. (2008); note that the kink
in the reversed loading data for the smallest strain amplitude is actually reported by these authors, and is probably an artifact of
the experimental difficulties associated with performing reversed loading experiments on sheet materials.
The initial configuration for our plane-strain sheet forming simulation is shown in Fig. 9(a). We
consider a 1 mm thick sheet with a length of 75 mm, which is to be to bent to a radius of 50 mm
between a pair of matched-rigid dies; accordingly, the radius of the top and bottom dies is pre-
scribed to be 50.5 mm and 49.5 mm, respectively. Due to the symmetry of the problem, we con-
sider only half of the geometry with suitable boundary conditions at the symmetry-plane. The
sheet is modeled using a mesh consisting of 1880 Abaqus-CPE4H plane-strain elements with seven
elements through the sheet thickness, and the dies are modeled using rigid surfaces. Contact be-
tween the sheet and the dies was modeled as frictionless. In the first step of the sheet forming
simulation, the top die is moved downward to bend the sheet completely around the bottom
die. Fig. 9(b) shows the geometry of the initial and fully bent geometries of the sheet. In the sec-
ond step, the top die is moved back upwards, and the sheet is allowed to spring-back. The un-
loaded sprung-back geometry is also shown in Fig. 9(b).
It is widely believed that plasticity models which do not account for kinematic hardening tend to
incorrectly predict the amount of spring-back in simulations of sheet-metal forming operations (cf.,
e.g., Zhao and Lee, 2001). Thus, our large deformation theory which not only includes both kinematic
and isotropic hardening but also accounts for large rotations, should be useful in improved modeling
of spring-back phenomena in sheet forming operations.
In this paper, we have formulated a large deformation constitutive theory for combined isotropic
and kinematic hardening based on the dual decomposition F ¼ Fe Fp and Fp ¼ Fpen Fpdis . We now show
that a similar kinematic-hardening theory may also be constructed without using the decomposition
Fp ¼ Fpen Fpdis . Recall the relation (6.3) for the evolution of Bpen ,
b 5
0
x2 (mm)
−5
Undeformed
−10
Fully bent
Sprung−back
−15
−20
0 10 20 30 40
x (mm)
Fig. 9. (a) Initial configuration for plane-strain matched-die sheet bending simulations. (b) Undeformed and fully bent sheet
geometries, together with the sprung-back geometry.
Dp ¼ Dpen þ sym Fpen Dpdis Fp1
en : ð11:2Þ
p
Bmdis
¼ Dp ðln Bpen Þ;
2s
en
1 p
¼ Dp cm ðln Bpen Þ ðusing ð9:42ÞÞ: ð11:4Þ
2
Using (11.4) in (11.1) gives
1 p 1
B_ pen ¼ Dp cm ðln Bpen Þ Bpen þ Bpen Dp cmp ðln Bpen Þ ;
2 2
or
Acknowledgements
This work was supported by the National Science Foundation under Grant Nos. CMS-0555614 and
CMMI-062524.
Appendix A
Cenþ1 ¼ Fe> e
nþ1 Fnþ1 ; Cetrial ¼ Fe> e
trial Ftrial : ð12:5Þ
Thus, using (12.3),
Cenþ1 ¼ exp DtDpnþ1 Cetrial exp DtDpnþ1 : ð12:6Þ
To proceed further we make our first approximation:
e1 p :
Ctrial Dnþ1 Cetrial ¼ Dpnþ1 : ð12:10Þ
Using (12.10), we find that (12.8) reduces to
:
Cenþ1 ¼ Cetrial 1 2DtDpnþ1 : ð12:11Þ
Further, using the symmetry of Cenþ1 we note that the symmetric tensors Cetrial and 1 2DtDpnþ1 com-
mute, and hence share the same principal directions. Thus taking the logarithm of (12.11) we obtain
1870 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
: 1
Eenþ1 ¼Eetrial þ ln 1 2DtDpnþ1 ; ð12:12Þ
2
where
1 1
Eenþ1 ¼ ln Cenþ1 and Eetrial ¼ ln Cetrial : ð12:13Þ
2 2
Next, we make our third approximation:
(A3) Recalling the series representation of the logarithm of a tensor, we assume that
:
ln 1 2DtDpnþ1 ¼ 2DtDpnþ1
for small DtDpnþ1 . Thus, under the assumptions (A1)–(A3) of small Dt and small jEeG
trial j, (12.6) yields the
important relation
:
Eenþ1 ¼Eetrial DtDpnþ1 : ð12:14Þ
Next, let
def 1
C ¼ 2G I 1 1 þ K1 1 ð12:15Þ
3
denote the elasticity tensor, and let
Menþ1 ¼ C½Eenþ1 and Metrial ¼ C½Eetrial ; ð12:16Þ
respectively, denote the elastic Mandel stress at the end of the time step, as well as its trial value.
Then, operating on (12.14) by C gives
Menþ1 ¼ Metrial DtC½Dpnþ1 : ð12:17Þ
15
Subtracting ðSen Þnþ1 from the left-hand side and
ðSen Þtrial ðSen Þn ; ðtrial values are evaluated with plastic flow frozenÞ
from the right-hand side of (12.17), and writing
def
ðMeeff Þtrial ¼ Metrial ðSen Þtrial ; ð12:18Þ
gives
ðMeeff Þnþ1 ¼ ðMeeff Þtrial DtC½Dpnþ1 : ð12:19Þ
Next, from (9.62) we have
1 ððMeeff Þnþ1 Þ0 1
Dpnþ1 ¼ pffiffiffi mpnþ1 Npnþ1 ; Npnþ1 ¼ pffiffiffi ; snþ1 ¼ pffiffiffi jððMeeff Þnþ1 Þ0 j; ð12:20Þ
2 2s nþ1 2
where
pffiffiffi
mpnþ1 ¼ 2jDpnþ1 j: ð12:21Þ
Using (12.15) and (12.20) in (12.19) we obtain
pffiffiffi
ðMeeff Þnþ1 ¼ ðMeeff Þtrial 2GðDt mpnþ1 ÞNpnþ1 : ð12:22Þ
Since Npnþ1 is deviatoric, the deviatoric and spherical parts of (12.22) give
15
Subtracting ðSen Þn rather than ðSen Þnþ1 from the right-hand side of (12.17) to arrive at (12.19), is a critical step in our time-
integration procedure. It allows us to decouple the update of Fp and Fpdis into to two separate, yet otherwise implicit updates.
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1871
pffiffiffi
ðMeeff Þnþ1 0
¼ ððMeeff Þtrial Þ0 2GðDt mpnþ1 ÞNpnþ1 ;
trðMeeff Þnþ1 ¼ trðMeeff Þtrial : ð12:23Þ
Using (12.20)2 , (12.23)1 may be arranged as
pffiffiffi
2 s nþ1 þ GðDtmpnþ1 Þ Npnþ1 ¼ ððMeeff Þtrial Þ0 : ð12:24Þ
Next, defining
We turn next to updating Sen and Fpdis . The evolution equation for F_ pdis ¼ Dpdis Fpdis is also integrated by
using an exponential map and (9.63) as
c
ðFpdis Þnþ1 ¼ exp DtðDpdis Þnþ1 ðFpdis Þn with ðDpdis Þnþ1 ¼ mp ðMp Þ ; ð12:37Þ
2B nþ1 en nþ1
the inverse of ðFpdis Þnþ1 is then
ðFpdis Þ1 p 1 p
nþ1 ¼ ðFdis Þn exp DtðDdis Þnþ1 : ð12:38Þ
Hence, using Fpen ¼ Fp Fp1
dis ,
p
the energetic part of F at the end of the step is given by
ðFpen Þnþ1 ¼ ðFpen Þtrial exp DtðDpdis Þnþ1 ; ð12:39Þ
where
def
ðFpen Þtrial ¼ Fpnþ1 ðFpdis Þ1
n ð12:40Þ
is a trial value of Fpen at the end of the step.
The tensors ðFpen Þnþ1 and ðFpen Þtrial admit the polar decompositions
ðFpen Þnþ1 ¼ ðRpen Þnþ1 ðUpen Þnþ1 ; ðFpen Þtrial ¼ ðRpen Þtrial ðUpen Þtrial : ð12:41Þ
Using (12.41) in (12.39) and rearranging, we obtain
ðRpen Þnþ1 ðUpen Þnþ1 exp DtðDpdis Þnþ1 ¼ ðRpen Þtrial ðUpen Þtrial : ð12:42Þ
Next, from (9.57) and (9.60)
def
ðSen Þnþ1 ¼ ðSen Þtrial cðDtmpnþ1 ÞðSen Þnþ1 ; where ðSen Þtrial ¼ 2B lnðVpen Þtrial : ð12:49Þ
Thus, the back-stress Sen is updated as
!
1
ðSen Þnþ1 ¼ ðSen Þtrial : ð12:50Þ
1 þ cðDtmpnþ1 Þ
Correspondingly,
Remark. Due to the use of the exponential map in integrating the evolution equations for Fp and Fpdis ,
the constraint of plastic incompressibility is exactly maintained by our time-integration procedure.
This is easily verified by recalling the identity detðexp AÞ ¼ expðtrAÞ and recognizing the deviatoric
nature of Dp and Dpdis in (12.1) and (12.37), respectively.
In typical ‘‘implicit” finite element procedures utilizing a Newton-type iterative solution method,
one needs to compute an algorithmically consistent tangent, often called the Jacobian matrix. We ob-
tain an estimate for our Jacobian matrix below. From the outset we note that Jacobian matrices are
used only in the search for the global finite element solution, and while an approximate Jacobian
might affect the rate of convergence of the global iteration scheme, it will not impair the accuracy
of our constitutive time-integration algorithm.
Consider the Cauchy stress at the end of the increment:
as Abaqus/Standard (2008), in which we have implemented our constitutive model by writing a user
material subroutine (UMAT).16. Thus, here we concentrate on evaluating the variation DMenþ1 in (12.78),
which is computed from
DMenþ1 ¼ C DEetrial ; ð12:79Þ
where the fourth-order tensor
def @Menþ1
C¼ ð12:80Þ
@Eetrial
is the important constitutive contribution to the global Jacobian matrix.
Recall that the time-integration procedure gives
pffiffiffi
Menþ1 ¼ Metrial 2GðDtmpnþ1 ÞNptrial : ð12:81Þ
Using (12.27)2 , (12.81) may be written as
pffiffiffi
Menþ1 ¼ Metrial þ 2ðs trial ÞNptrial :
nþ1 s ð12:82Þ
By the product rule
pffiffiffi
@Metrial pffiffiffi @Nptrial p @s
nþ1 @ strial
C¼ þ 2 ð s
nþ1 s
trial Þ þ N 2 : ð12:83Þ
@Eetrial @Eetrial trial
@Eetrial @Eetrial
Since
Metrial ¼ C Eetrial ; ð12:84Þ
we have
@Metrial
¼ C: ð12:85Þ
@Eetrial
Next, since
ððMeeff Þtrial Þ0
Nptrial ¼ pffiffiffi ; ð12:86Þ
2s trial
we have
rffiffiffi
@Nptrial 1 1 @ððMeeff Þtrial Þ0 ððMeeff Þtrial Þ0 @ s
trial
¼ e : ð12:87Þ
@Eetrial 2 s trial e
@Etrial strial
2
@Etrial
Now,
ðMeeff Þtrial ¼ Metrial ðSen Þn
¼ C Eetrial ðSen Þn ð12:88Þ
1
¼ 2G I 1 1 þ K1 1 Eetrial ðSen Þn :
3
Hence,
1 e
ððMeeff Þtrial Þ0 ¼ 2G I 1 1 Etrial ðSen Þn ; ð12:89Þ
3
16
Abaqus uses a hypoelastic constitutive equation for the stress of the type
_
T ¼ TWe _
TþTWe ¼ TWTþTW p
|fflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflffl} ¼ C½DD :
for Wp ¼0
1876 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
and therefore
@ððMeeff Þtrial Þ0 1
e ¼ 2G I 1 1 : ð12:90Þ
@Etrial 3
Further,
rffiffiffi
1
strial ¼ ððMeeff Þtrial Þ0 : ð12:91Þ
2
Therefore, by the chain-rule
rffiffiffi >
@s
trial 1 @ððMeeff Þtrial Þ0 @jððMeeff Þtrial Þ0 j
¼ ð12:92Þ
@Eetrial 2 @E e e
@ððMeff Þtrial Þ0
rffiffiffi trial >
1 1 ððMeeff Þtrial Þ0
¼ 2G I 1 1 ; ð12:93Þ
2 3 jððMeeff Þtrial Þ0 j
pffiffiffi
1 ððMeeff Þtrial Þ0
¼ 2G I 1 1 ð12:94Þ
3 jððMeeff Þtrial Þ0 j
pffiffiffi ððMe Þ Þ
eff trial 0
¼ 2G ð12:95Þ
jððMeeff Þtrial Þ0 j
pffiffiffi
¼ 2GNptrial : ð12:96Þ
Thus, using (12.90) and (12.96) in (12.87) we obtain
@Nptrial 2G 1
¼ pffiffiffi I 1 1 Nptrial Nptrial ; ð12:97Þ
@Eetrial 2s trial 3
and substituting (12.85) and (12.97) in (12.83), we have
snþ1 1
C ¼Cþ 1 2G I 1 1 Nptrial Nptrial
strial 3
pffiffiffi
nþ1 pffiffiffi p
@s
þ Nptrial 2 e 2GNtrial : ð12:98Þ
@Etrial
Noting that Eetrial enters the equations through s
trial , we have, using (12.96)
@s
nþ1 @ s nþ1 @ s nþ1 pffiffiffi p
trial @ s
e ¼ e ¼ 2GNtrial : ð12:99Þ
@Etrial @ strial @Etrial @ s
trial
Finally, substituting (12.99) into (12.98), using (12.15), and rearranging, we have
~ 2G snþ1 @ snþ1 Np Np
C¼C ð12:100Þ
strial @ strial trial trial
where
~ I 1 1 1 þ K1 1 with G
~ ¼ 2G ~ def snþ1
C ¼ G: ð12:101Þ
3 strial
Thus, it remains to determine the derivative
@s
nþ1
@s
trial
in (12.100). First, we rewrite the updates for s
nþ1 ; (12.27)2 , and Snþ1 , (12.69), as
snþ1
snþ1 ¼ strial GDtf ;
Snþ1
ð12:102Þ
snþ1
Snþ1 ¼ Sn þ DthðSn Þf ;
Snþ1
D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878 1877
References
17
Here I is the second-order identity matrix.
1878 D.L. Henann, L. Anand / International Journal of Plasticity 25 (2009) 1833–1878
Cao, J., Lee, W., Cheng, H.S., Seniw, M., Wang, H.-P., Chung, K., 2008. Experimental and numerical investigation of combined
isotropic-kinematic hardening behavior of sheet metals. International Journal of Plasticity, 10.1016/j.ijplas.2008.04.007.
Casey, J., Naghdi, P.M., 1980. A remark on the use of the decomposition F ¼ Fe Fp in plasticity. Journal of Applied Mechanics 47,
672–675.
Chaboche, J.L., 2008. A review of some plasticity and viscoplasticity constitutive theories. International Journal of Plasticity 24,
1642–1693.
Clayton, J., McDowell, D., 2003. A multiscale multiplicative decomposition for elastoplasticity of polycrystals. International
Journal of Plasticity 19, 1401–1444.
Dafalias, Y., 1998. Plastic spin: necessity or redundancy? International Journal of Plasticity 14, 909–931.
Dashner, P.A., 1986. Invariance considerations in large strain elasto-plasticity. Journal of Applied Mechanics 53, 55–60.
Dettmer, W., Reese, S., 2004. On the theoretical and numerical modelling of Armstrong–Frederick kinematic hardening in the
finite strain regime. Computer Methods in Applied Mechanics and Engineering 193, 87–116.
Frederick, C.O., Armstrong, P.J., 2007. A mathematical representation of the multiaxial Bauschinger effect. Materials at High
Temperatures 24, 1–26.
Green, A.E., Naghdi, P.M., 1971. Some remarks on elastic–plastic deformation at finite strain. International Journal of
Engineering Science 9, 1219–1229.
Gurtin, M.E., Anand, L., 2005. The decomposition F ¼ Fe Fp , material symmetry, and plastic irrotationality for solids that are
isotropic-viscoplastic or amorphous. International Journal of Plasticity 21, 1686–1719.
Håkansson, P., Wallin, M., Ristinmaa, M., 2005. Comparison of isotropic hardening and kinematic hardening in thermoplasticity.
International Journal of Plasticity 21, 1435–1460.
Kröner, E., 1960. Allgemeine kontinuumstheorie der versetzungen und eigenspannungen. Archive for Rational Mechanics and
Analysis 4, 273–334.
Lemaitre, J., Chaboche, J.L., 1990. Mechanics of Solid Materials. Cambridge University Press, Cambridge.
Lion, A., 2000. Constitutive modelling in finite thermoviscoplasticity: a physical approach based on nonlinear rheological
models. International Journal of Plasticity 16, 469–494.
Lush, A.M., Weber, G., Anand, L., 1989. An implicit time-integration procedure for a set of internal variable constitutive
equations for isotropic elasto-viscoplasticity. International Journal of Plasticity 5, 521–549.
Menzel, A., Ekh, M., Runesson, K., Steinmann, P., 2005. A framework for multiplicative elastoplasticity with kinematic hardening
coupled to anisotropic damage. International Journal of Plasticity 21, 397–434.
Prager, W., 1949. Recent developments in the mathematical theory of plasticity. Journal of Applied Physics 20, 235–241.
Ristinmaa, M., Wallin, M., Ottosen, N.S., 2007. Thermodynamic format and heat generation of isotropic hardening plasticity.
Acta Mechanica 194, 103–121.
Rosakis, P., Rosakis, A.J., Ravichandran, G., Hodowany, J., 2000. A thermodynamic internal variable model for the partition of
plastic work into heat and stored energy in metals. Journal of the Mechanics and Physics of Solids 48, 581–607.
Vladimirov, I.N., Pietryga, M.P., Reese, S., 2008. On the modelling of non-linear kinematic hardening at finite strains with
application to springback – comparison of time integration algorithms. International Journal for Numerical Methods in
Engineering 75, 1–28.
Wallin, M., Ristinmaa, M., 2005. Deformation gradient based kinematic hardening model. International Journal of Plasticity 21,
2025–2050.
Weber, G., Anand, L., 1990. Finite deformation constitutive equations and a time integration procedure for isotropic,
hyperelastic–viscoplastic solids. Computer Methods in Applied Mechanics and Engineering 79, 173–202.
Weber, G.G., Lush, A.M., Zavaliangos, A., Anand, L., 1990. An objective time-integration procedure for isotropic rate-independent
and rate-dependent elastic–plastic constitutive equations. International Journal of Plasticity 6, 701–744.
Zhao, K.M., Lee, J.K., 2001. Material properties of aluminum alloy for accurate draw-bend simulation. Journal of Engineering
Materials and Technology 123, 287–292.