0% found this document useful (0 votes)

6 views

A Return Map Algorithm For General Isotr

Uploaded by

anailpereslontsi

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

6 views

A Return Map Algorithm For General Isotr

Uploaded by

anailpereslontsi

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 38

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING

Int. J. Numer. Meth. Engng 2004; 60:461–498 (DOI: 10.1002/nme.970)

A return map algorithm for general isotropic elasto/visco-plastic

materials in principal space

Luciano Rosati1,∗,† and Nunziante Valoroso2

1 Dipartimento di Scienza delle Costruzioni, Università di Napoli Federico II, via Claudio 21,
Napoli 80125, Italy
2 Istituto per le Tecnologie della Costruzione, Consiglio Nazionale delle Ricerche, Viale Marx 15,
Roma 00137, Italy

SUMMARY
We describe a methodology for solving the constitutive problem and evaluating the consistent tangent
operator for isotropic elasto/visco-plastic models whose yield function incorporates the third stress
invariant J3 . The developments presented are based upon original results, proved in the paper, con-
cerning the derivatives of eigenvalues and eigenprojectors of symmetric second-order tensors with
respect to the tensor itself and upon an original algebra of fourth-order tensors A obtained as second
derivatives of isotropic scalar functions of a symmetric tensor argument A. The analysis, initially
referred to the small-strain case, is then extended to a formulation for the large deformation regime;
for both cases we provide a derivation of the consistent tangent tensor which shows the analogy
between the two formulations and the close relationship with the tangent tensors of the Lagrangian
description of large-strain elastoplasticity. Copyright 䉷 2004 John Wiley & Sons, Ltd.

KEY WORDS: J3 plasticity and viscoplasticity; principal space formulation; return map algorithm;
consistent tangent

1. INTRODUCTION

Computational plasticity and viscoplasticity have been characterized by signiﬁcant advancements

in the last two decades mainly after the work by Simo, whose fundamental contributions
in the ﬁeld have been collected in two recent monographs [1, 2]. The general framework
outlined in these books and in several other papers on the subject, see e.g. References [3–7]
among others, has addressed isotropic elasto(visco-) plastic materials, i.e. materials whose
elastic and (visco)plastic behaviours are described through an isotropic stored energy function
and an isotropic yield function.

∗ Correspondence to: Luciano Rosati, Dipartimento di Scienza delle Costruzioni, Università di Napoli Federico II,
via Claudio 21, 80125 Napoli, Italy.
† E-mail: [email protected]

Contract/grant sponsor: Italian National Research Council (CNR)

Contract/grant sponsor: Italian Ministry of Education, University and Research (MIUR)
Received 2 October 2002
Revised 26 July 2003
Copyright 䉷 2004 John Wiley & Sons, Ltd. Accepted 5 August 2003
462 L. ROSATI AND N. VALOROSO

As a matter of fact, the approaches outlined in References [1, 2] for small and large-strain
isotropic plasticity are quite different. Actually, in the former case the solution of the constitutive
problem and the evaluation of the consistent tangent are usually addressed in terms of (intrinsic)
tensor quantities. On the contrary, most of the solution strategies presented in the literature
[8–11] for elastoplastic models in the large-strain regime, substantially derived from Reference
[12], heavily hinge on the isotropy of the involved functions. Indeed, the assumption of isotropy
is used to establish that the return mapping algorithm takes place at fixed principal axis so
that only the principal values of the state variables need to be iterated upon.
In the present work, we show how a fully tensorial approach naturally supplies a unified
framework for small- and large-strain isotropic plasticity which, in addition, allows for a deeper
insight in the formulation and finite element implementation of the solution strategy both at
the constitutive and structural level.
This is obtained by illustrating the relationships existing between the first and second
derivatives of isotropic scalar functions of a symmetric rank-two tensor A, assigned ei-
ther as function of the eigenvalues of A or of its invariants and by proving some basic
results concerning the differentiation of the eigenprojectors Ak of A with respect to the ten-
sor A itself. Moreover, the principal space representation and inversion of rank-four positive-
definite symmetric tensors A obtained as the second derivative of with respect to A is
discussed.
The developments carried out in the paper provide proper evidence to three main issues. First,
the yield function can be assigned either in terms of eigenvalues or of the invariants of the stress
tensor without affecting the derivation of the tensor quantities entering the return mapping and
the tangent operator. Second, the usual procedure of expressing the material tangent operator in
the principal reference frame and then transforming it to the given reference frame via matrix
manipulations can be by-passed since the expression of the tangent operator in the global
co-ordinate system can be directly constructed. Third, the terms entering the final expression
of the consistent tangent operator can be clearly identified.
The results of tensor analysis alluded to above are employed to carry out a return mapping
algorithm for associative isotropic elasto/visco-plastic models whose yield function depends
upon the J3 stress invariant.
Yield criteria of this type are representative of a wide class of engineering materials such
as concrete and geomaterials, see e.g. References [13–21], for specific applications and [22]
for a detailed account of failure criteria incorporating the J3 invariant. Recently, the use of J3 -
dependent surfaces has also been advocated, among others, in References [23, 24], for describing
phase transitions in shape-memory alloys within the superelastic range.
The solution of structural problems endowed with J3 -dependent yield functions represents
a considerable challenge, by far more severe than the one associated with the classical J2
(Von Mises) model. For this reason, several approaches have been proposed in the literature
[25–30]; in some cases [31], implicit procedures are abandoned in favour of explicit ones
[32, 33] because of their complicated algorithmic structure.
Within the framework of closest-point projection algorithms, two additional solution strategies
for J3 -dependent plasticity models have been recently proposed by the authors [34, 35] by
providing an explicit representation formula for the inverse of the elasto(visco)plastic compliance
tensor G entering the exact algorithm linearization. In order to allow a direct extension to a
principal space formulation, the inversion of G is here carried out on the basis of the general
properties of positive-definite rank-four tensors discussed in the paper.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 463

It is shown in particular that only the dyadic part of G, expressed as linear combination
of dyadic tensor products S⊗ ij = Si ⊗ Sj between the eigenprojectors Sk of the stress deviator
S, is strictly needed in the computation of the return map solution since it is associated with
the algorithm linearization at constant eigenvectors. On the contrary, the non-dyadic part of
G, which is expressed as linear combination of the square tensor products S⊠ ij = Si ⊠Sj
between the eigenprojectors of S, is associated with the differentiation of the eigenprojectors
with respect to the stress tensor and plays no active role in the local Newton iteration scheme.
These circumstances suggest to express the return mapping algorithm directly in terms of
principal values by making reference, at least formally, to a yield function assigned in terms
of stress eigenvalues.
Nonetheless, irrespective of the expression originally assumed for the yield function, all
tensorial quantities entering the principal return mapping and the expression of the consistent
tangent tensor are directly expressed in the global co-ordinate system.
This approach carries over in a natural way to the case of isotropic hyperelastoplasticity at
ﬁnite strains based on the multiplicative decomposition since, in this case, the stored energy
function is usually given in terms of principal stretches and the solution of the constitutive
problem can be computed in terms of principal values [12]. With reference to this last case,
an original derivation of the consistent tangent tensor, allowed by the results provided in
the paper, is presented. It highlights the perfect analogy of the relevant expression with the
ones pertaining to the continuum and algorithmic tangent tensors obtained for the Lagrangian
description of large-strain elastoplasticity [36] and the close relationships with the expression
of the small-strain elastoplastic tangent.
Finally, a numerical example shows the performances of the proposed implementation.

2. ELEMENTS OF TENSOR ALGEBRA

We provide a brief review of the basic properties of linear transformations (tensors) on a

three-dimensional inner product space V over the real ﬁeld.
Denoting by Lin (Lin) the space of all second- (fourth-)order tensors on V, we shall
also illustrate the properties of tensor products between elements of Lin. Unless differently
stated, we shall use boldface normal (capital) letters to indicate elements of V (Lin) and
BOLDBLACKBOARD symbols to denote fourth-order tensors.
Given a, b, c, d ∈ V and A ∈ Lin the following relations can be proved:
(a ⊗ b) · (c ⊗ d) = (a · c)(b · d)
(a ⊗ b)(c ⊗ d) = (b · c)(a ⊗ d)
(1)
A(a ⊗ b) = Aa ⊗ b
(a ⊗ b)A = a ⊗ AT b

with the superscript T standing for transpose.

Given A, B ∈ Lin the tensor product A ⊗ B, usually termed dyadic product, is the element
of Lin such that
(A ⊗ B)C = (B · C)A = tr(BT C)A ∀C ∈ Lin (2)
where the symbol tr(·) denotes the trace operator.

More recently Del Piero [37] has introduced an additional tensor product A⊠B between
second-order tensors deﬁned by
(A ⊠ B)C = ACBT ∀C ∈ Lin (3)
it will be referred to in the sequel as the square tensor product on account of the symbol
originally proposed by Del Piero.
The previous product allows one to represent the identity tensor I in Lin as
I = 1⊠1 (4)
where 1 is the identity tensor in Lin.
The following composition rules can be shown to hold:
(A ⊠ B)(C ⊠ D) = (AC) ⊠ (BD)

(A ⊠ B)(C ⊗ D) = (ACBT ) ⊗ D (5)

(A ⊗ B)(C ⊠ D) = A ⊗ (CT BD)

for every A, B, C, D ∈ Lin. Furthermore, if A and B are invertible, one can show that such is
their square product:
(A ⊠ B)−1 = A−1 ⊠ B−1
Composition formulas between elements of Lin and Lin follow immediately from the previous
deﬁnitions and properties:
(a ⊗ b) ⊠ (c ⊗ d) = (a ⊗ c) ⊗ (b ⊗ d)

(A ⊗ B)(c ⊗ d) = [B · (c ⊗ d)]A = (BT c · d)A (6)

(A ⊠ B)(c ⊗ d) = A(c ⊗ d)BT = Ac ⊗ Bd

for every A, B ∈ Lin and a, b, c, d ∈ V.
Analogous to the standard deﬁnition adopted for second-order tensors, we remind that the
transpose of a fourth-order tensor is deﬁned by
AB · C = B · AT C ∀B, C ∈ Lin
so that A is symmetric whenever A = AT . For the translation of the previous tensor formalism
into the matrix one the reader is referred to Appendix A.

2.1. Eigenprojectors of second-order symmetric tensors

Let A be a second-order symmetric tensor, i.e. an element of the subspace Sym ⊆ Lin.
According to the spectral theorem [38] there is at least one orthonormal basis for V consisting
of eigenvectors of A; hence, denoting by âi , {i = 1, 2, 3}, the eigenvalues of A, supposed to
be distinct, and by ai the associated unit eigenvectors, it turns out to be
3
3

A= âi ai ⊗ ai = âi Ai (7)
i=1 i=1

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 465

where Ai is the eigenprojector corresponding to âi , i.e. the orthogonal projection operator onto
the null space of A − âi 1. In particular formula (1)1 yields

1 if i = j
Ai · Aj = (8)
0 otherwise
Additional properties of the eigenprojectors that will be extensively used in the sequel are

Ai if i = j
Ai Aj = Aj Ai = (9)
0 otherwise
and
A1 + A2 + A3 = 1 (10)
Immediate consequence of (7) and (9) is that
Ai A = AAi = âi Ai (11)
Remark 2.1
The eigenvalues of A are not necessarily distinct. Hence, if A has exactly two distinct eigen-
values â1 and â2 and a1 is a unit vector belonging to the characteristic space of â1 , the spectral
representation of A reads
A = â1 a1 ⊗ a1 + â2 (1 − a1 ⊗ a1 ) (12)
while, for three coincident eigenvalues, one has
A = â1
being â the unique eigenvalue of A.
The presence of coalescent eigenvalues does not represent a real problem from the com-
putational standpoint since, even in this situation, it is possible to deﬁne three orthogonal
eigenprojectors; as an alternative, one can use a perturbation of the possibly repeated eigen-
values; see e.g. References [2, 39, 40].

A basic result [39, 41] setting in correspondence the generic eigenvalue with the associated
eigenprojector is contained in the following:
Lemma 1
Let âi be an eigenvalue of A ∈ Sym. The derivative of âi with respect to A is provided by
dA âi = ai ⊗ ai = Ai (13)
Proof
By deﬁnition of eigenvalue it turns out to be
Aai = âi ai (14)
so that
âi = Aai · ai = A · (ai ⊗ ai ) (15)
since the eigenvectors ai have been assumed of unit length.

The directional derivative of (15) in the direction H ∈ Lin yields

dA âi · H = H · (ai ⊗ ai ) + A · [(dA ai )H ⊗ ai ] + A · [ai ⊗ (dA ai )H] (16)

by applying the product rule [42].

Further, from ai · ai = 1 we obtain

2(dA ai )H · ai = 0

so that, recalling (1)2 , we obtain

A · [(dA ai )H ⊗ ai ] = tr {A[(dA ai )H ⊗ ai ]} = tr {[(dA ai )H ⊗ ai ]A}

= tr [(dA ai )H ⊗ Aai ] = (dA ai )H · âi ai = 0

Following a similar path of reasoning it is easy to show that the last term in (16) is zero;
hence, from the arbitrariness of H in (16) we ultimately infer the result.

Remark 2.2
In the previous lemma it has been implicitly assumed that the tensor A has three distinct
eigenvalues âi and univocally associated eigenprojectors ai ⊗ ai .
However, it is immediate to verify that the stated results hold true also in the case of two
or three coincident eigenvalues provided that each of them is associated with a number of
orthonormal eigenprojectors equal to its algebraic multiplicity.
For instance, if A has two distinct eigenvalues â1 and â2 having in turn multiplicity 1
and 2, see e.g. (12), the proof of the lemma is obtained by associating with â1 the unique
eigenprojector a1 ⊗ a1 and with â2 the two eigenprojectors a2 ⊗ a2 and a3 ⊗ a3 , with a2
and a3 being a pair of mutually orthogonal unit vectors belonging to the (two-dimensional)
characteristic space of â2 .
Analytical details can be found in References [41, 43] and references therein.

A basic property concerning the tensor products of eigenprojectors is given by

Lemma 2
Let Ai be an eigenprojector of A ∈ Sym. Then

Ai ⊗ Ai = Ai ⊠ Ai (17)

Proof
It trivially follows from (6)1 .

A crucial result concerning the evaluation of the derivative of the eigenprojector of a rank-two
tensor with respect to the tensor itself is provided by the following:

Lemma 3
Let Ai be an eigenprojector of A ∈ Sym. Then
Ai ⊠ Aj + Aj ⊠ Ai Ai ⊠ Ak + Ak ⊠ Ai
dA Ai = + , i = j, k, i, j, k ∈ {1, 2, 3} (18)
âi − âj âi − âk

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 467

Proof
Start by considering the identity:
(A ⊠ 1 + 1 ⊠ A − 2âi I)Ai = Ci Ai = 0 (19)
which immediately follows from (11). By differentiating (19) and invoking (10) and (17) the
following relationship is obtained:
Ci dA Ai = −2 sym(Ai ⊠ Aj + Ai ⊠ Ak ), i = j, k (20)
where j and k denote cyclic permutations of i.
Being
−dA Ak = dA Ai + dA Aj (21)
by virtue of (10), a manipulation of (20) yields
1
dA Aj = [2 sym(Ai ⊠ Aj + Aj ⊠ Ak + Ak ⊠ Ai ) − (âi − âk )dA Ai ] (22)
(âj − âk )
substituting the previous expression into (20), referred to the index j , and subtracting (20)
gives then

(âi − âk )(Cj − Ci )dA Ai = 2(âi − âk ) sym(Ai ⊠ Aj + Ai ⊠ Ak )

+ 2(âj − âk ) sym(Ai ⊠ Aj + Aj ⊠ Ak )

+ 2Cj sym(Ai ⊠ Aj + Aj ⊠ Ak + Ak ⊠ Ai ) (23)

Invoking (19) and the composition rules (5) one gets the identity:
Cj − Ci = 2(âi − âj )I
and the three additional ones
2Cj sym(Ai ⊠ Ak ) = 2(âi + âk − 2âj )2 sym(Ai ⊠ Ak )
Substitution of the previous identities in (23) yields ﬁnally the result.

A more detailed proof of the previous lemma and additional results concerning the derivatives
of isotropic tensor functions in the case of coalescent eigenvalues can be found in Reference [44]
which the interested reader may refer to.

Remark 2.3
The expression of the derivative of the eigenprojector provided in the previous lemma can be
proven to be fully equivalent to the existing one due to Carlson and Hoger which, for the
three-dimensional case and distinct eigenvalues, reads [41]:
1
dA Ai = [1 ⊠ A + A ⊠ 1 − (âj + âk )1 ⊠ 1
(âi − âj )(âi − âk )

− (2âi − âj − âk )Ai ⊠ Ai − (âj − âk )(Aj ⊠ Aj + Ak ⊠ Ak )], i = j, k

see e.g. Reference [44]. However, due to its considerably simpler form, expression (18) will
be retained for the ensuing developments.

2.2. Some properties of isotropic scalar functions of a symmetric tensor argument

Let us consider an isotropic scalar function of a symmetric tensor argument A ∈ Sym. Owing
to isotropy, such a function can be expressed either as function of the invariants of A or as
symmetric function of the eigenvalues of A in the equivalent form [42]
ˆ âi ),
(IA , II A , III A ) = ( i = 1, 2, 3 (24)
where
IA = tr A, II A = 21 [(tr A)2 − tr A2 ], III A = det A
are the principal invariants of A.
In particular we are interested in the connections existing between the ﬁrst and second
derivatives of and ˆ with respect to A. Using the chain rule of differential calculus [45], the
derivative of (24) can be expressed as
ˆ i,
dA = dâi A i = 1, 2, 3 (25)
which shows that dA admits a spectral decomposition whose eigenvalues are exactly the
derivatives of the scalar function ˆ with respect to the eigenvalues âi of A.
Differentiation of (25) supplies then
2
dAA = dâ2i âj A
ˆ i ⊗ Aj + dâi d
ˆ A Ai , i, j = 1, 2, 3 (26)
where the last term on the right-hand side can be evaluated by means of (18).
It is worth noting that the previous relationships allow one to make reference to the derivatives
of through a different function , ˆ although the explicit expression of this last function does
not need to be actually computed.
In particular, recalling the well-known formulas [46]
dA IA = 1, dA II A = IA 1 − A, dA III A = (det A)A−T (27)
we get
dA = 1 (A)1 + 2 (A)A + 3 (A)A2 , i = 1, 2, 3 (28)
where 1 , 2 and 3 are isotropic scalar functions of A.
Substitution into the previous expression of the spectral decomposition of A yields
dA = 1 (A)A1 + 2 (A)A2 + 3 (A)A3 (29)
whose coefﬁcients are given by
i = 1 + 2 âi + 3 ai2 = dâi ,
ˆ i = 1, 2, 3 (30)
as it can be inferred from (25), (28) and (29).
In a similar way, differentiation of (28) and use of (27) allows one to express the rank-four
2 as linear combination of terms of the type A ⊗A +A ⊗A and A ⊠ A +A ⊠ A
tensor dAA
with , ∈ {0, 1, 2}. The connection existing between this expression and the one reported on
the right-hand side of (26) is presented in the next section.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 469

3. REPRESENTATION OF SYMMETRIC FOURTH-ORDER TENSORS

IN PRINCIPAL SPACE

Let us consider a symmetric fourth-order tensor A expressed as linear combination of tensor

products of second-order symmetric tensors in the form
2
2
⊗
(A ⊗ A + A ⊗ A ) + p
⊠
(A ⊠ A + A ⊠ A )]

A= [p
=0 =0

⊗ ⊠ are isotropic scalar functions.

where A0 = 1 and p and p
By virtue of Rivlin’s identities for tensor polynomials [47], A can be equivalently ex-
pressed as
2
2 1
1
⊗
(A ⊗ A + A ⊗ A )] + ⊠
(A ⊠ A + A ⊠ A )]

A= [q [q (31)
=0 =0 =0 =0

⊗ ⊠ depend upon p ⊗ and p ⊠ and the invariants of A.

where q and q
We are interested to investigate on the expression of A resulting from the spectral rep-
resentation of A; in this respect it is easy to show that A is amenable to the following
representation:
3 3
aij⊗ Ai ⊗ Aj + aij⊠ Ai ⊠ Aj = [A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ]

A=
i,j =1 i,j =1
i =j

⊗ ⊗ ⊗
a11 a12 a13 A1 ⊗ A1 A1 ⊗ A2 A1 ⊗ A3
   
 ⊗ ⊗ ⊗ 
  
=
 a12 a22 a23  ·  A2 ⊗ A1 A2 ⊗ A2 A2 ⊗ A3 

⊗ ⊗ ⊗ A3 ⊗ A1 A3 ⊗ A2 A3 ⊗ A3
a13 a23 a33
⊠ ⊠
  
0 a12 a31 O A1 ⊠A2 A1 ⊠A3

  
 ⊠ ⊠

+  a12 0 a23 · A2 ⊠A1 O A2 ⊠A3  (32)

   
⊠ ⊠ A3 ⊠A1 A3 ⊠A2 O
a31 a23 0

in which the entries aij⊗ and aij⊠ are polynomial expressions of the eigenvalues of A and of
⊗ ⊠ while the dots between arrays [ ] · [ ] used in the previous formula
the coefﬁcients q and q
indicate sum of the products of the elements having the same position.
As an example of the representation formula (32), we consider the rank-four identity tensor I.
Recalling (4) and (10) we have

I = A1 ⊗ A1 + A2 ⊗ A2 + A3 ⊗ A3

+ A1 ⊠A2 + A2 ⊠A1 + A2 ⊠A3 + A3 ⊠A2 + A3 ⊠A1 + A1 ⊠A3 (33)

on account of Lemma 2. Hence,

A1 ⊗ A1 A1 ⊗ A2 A1 ⊗ A3
   
1 0 0
I = [1⊗ ] · [A⊗ ] + [1⊠ ] · [A⊠ ] = 
   
0 1  ·  A2 ⊗ A1
0 A2 ⊗ A2 A2 ⊗ A3 


0 0 1 A3 ⊗ A1 A3 ⊗ A2 A3 ⊗ A3

O
  
A1 ⊠A2 A1 ⊠A3

0 1 1
   
+
1 0  ·  A2 ⊠A1
1  O A2 ⊠A3 
 (34)
1 1 0 A3 ⊠A1 A3 ⊠A2 O

A further example is represented by the second derivative of an isotropic scalar function

expressed in terms of principal invariants of a rank-two tensor A ∈ Sym. Recalling the
considerations reported at the end of the previous section, the right-hand side of (26) can be
written as
 2
dâ1 â1 ˆ dâ21 â2 ˆ dâ21 â3 ˆ
 
A1 ⊗ A1 A1 ⊗ A2 A1 ⊗ A3

  
2
=  dâ21 â2 ˆ dâ22 â2 ˆ dâ22 â3 ˆ  · 

dAA A2 ⊗ A1 A2 ⊗ A2 A2 ⊗ A3 
 
   
dâ21 â3 ˆ dâ22 â3 ˆ dâ23 â3 ˆ A3 ⊗ A1 A3 ⊗ A2 A3 ⊗ A3

⊠ ⊠
  
0 d12 ˆ d13 ˆ O A1 ⊠A2 A1 ⊠A3

  
 ⊠ ⊠

+  d12 ˆ 0 d23 ˆ  ·  A2 ⊠A1 O A2 ⊠A3  (35)

   
⊠ ⊠ A3 ⊠A1 A3 ⊠A2 O
d13 ˆ d23 ˆ 0

where
dâi ˆ − dâj ˆ
dij⊠ ˆ = , i = 1, 2, 3, j = 2, 3, 1
âi − âj
In the case of coalescent eigenvalues, the above formulas need to be modified as detailed
in Reference [44].
We shall now prove the first of two basic results which play a paramount role in the
developments that follow.
Lemma 4
Let A = [A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ] be a symmetric rank-four tensor. Then:
⊗
[A ] is positive definite
A positive-definite ⇐⇒
[A⊠ ] has positive off-diagonal entries
Proof
By definition:
A positive definite ⇔ AB · B > 0 ∀B = 0, B ∈ Lin

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 471

Let A be positive deﬁnite. Expressing B in the Cartesian frame represented by the eigen-
vectors of A:
3

B= bkl ak ⊗ al (36)
k,l=1

and observing that the following composition rules:

Ai if j = k = l
(Ai ⊗ Aj )(ak ⊗ al ) =
0 otherwise
(37)
ai ⊗ aj if i = k and j = l
(Ai ⊠ Aj )(ak ⊗ al ) =
0 otherwise
hold true on account of (6), one obtains
⊗ 2 ⊗ 2 ⊗ 2 ⊗ ⊗ ⊗
AB · B = a11 b11 + a22 b22 + a33 b33 + 2a12 b11 b22 + 2a23 b22 b33 + 2a13 b33 b11
⊠ 2 2 ⊠ 2 2 ⊠ 2 2
+ a12 (b12 + b21 ) + a23 (b23 + b32 ) + a31 (b31 + b13 )>0 ∀B = 0

hence, setting [b] = [b11 , b22 , b33 ]T , one has ﬁnally

3
AB · B = [A⊗ ][b] · [b] + aij⊠ bij
2

(38)
i,j =1
i =j

Since the previous inequality must hold for every B ∈ Lin, it must necessarily be
3
[A⊗ ][b] · [b] > 0, aij⊠ bij
2

> 0, ∀B = 0
i,j =1
i =j

By considering a tensor B whose associated matrix [B] in the principal reference frame for
A is diagonal, the ﬁrst relation allows one to infer that [A⊗ ] is positive-deﬁnite. Further, by
choosing a [B] having only one non-zero off-diagonal term, it turns out to be
aij⊠ > 0, i, j ∈ {1, 2, 3}, i = j

Conversely, if [A⊗ ] is positive-deﬁnite and each aij⊠ > 0 (i = j ) one trivially infers from
(38) that A is positive-deﬁnite. The result is thus proved.

3.1. Inversion of symmetric fourth-order tensors

Let us suppose that a positive deﬁnite tensor A is given in form (31), or equivalently in form
(32), and that its inverse A−1 has to be computed.
This can be accomplished by virtue of two lemmas which, although proved in the sequel for
the case of interest of positive-deﬁnite tensors A, hold in general for any invertible fourth-order
tensor, see e.g. Reference [48].

Lemma 5
Given a symmetric positive-definite fourth-order tensor A in the form
A = [A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ]
its inverse A−1 admits an analogous representation, i.e.
A−1 = [C⊗ ] · [A⊗ ] + [C ⊠ ] · [A⊠ ]
where [C⊗ ] and [C ⊠ ] are 3 × 3 matrices.
Proof
In order to establish a representation formula for the fourth-order tensor A−1 consider the
tensor equation:
AX = H (39)
in the unknown X. Clearly, X is a function of H and of the tensor A entering the definition
(31) of A:
X = F(A, H) = A−1 H (40)
Using the composition rules between tensor products (5), it is not difficult to prove that F
is an isotropic function of A and H simultaneously:
QF(A, H)QT = F(QAQT , QHQT ) ∀Q ∈ Orth (41)
i.e. that QXQT is the solution of (39) provided that A and H are replaced by QAQT and
QHQT , respectively.
To provide a concise proof of the final result we further assume H to be symmetric. The
most general proof can be obtained by following the same path of reasoning illustrated in
Reference [49] with reference to the simpler case of A = A ⊠ 1 + 1 ⊠ A.
Owing to the symmetry of H and that of A, the solution of (39) is symmetric as well so
that, being F linear in H, the representation theorem for isotropic tensorial functions of two
symmetric tensor arguments, see e.g. Reference [47], yields

A−1 = aI + b(A⊠1 + 1⊠A) + c(A2 ⊠1 + 1⊠A2 )

+ d(1 ⊗ 1) + e(A ⊗ 1 + 1 ⊗ A) + f (A2 ⊗ 1 + 1 ⊗ A2 )

+ g(A ⊗ A) + h(A2 ⊗ A + A ⊗ A2 ) + i(A2 ⊗ A2 )

Using the spectral decomposition (7) the principal space representation is

A−1 = [C⊗ ] · [A⊗ ] + [C ⊠ ] · [A⊠ ] (42)
and the result is demonstrated.
We can now prove the next:
Lemma 6
Given a symmetric positive-deﬁnite fourth-order tensor A in the form
A = [A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ]

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 473

it turns out to be
A−1 = [A⊗ ]−1 · [A⊗ ] + [A⊠ ]−1 · [A⊠ ]
where [A⊗ ]−1 is the inverse matrix of [A⊗ ] and [A⊠ ]−1 is the matrix whose components are
the reciprocals of the non-zero entries of [A⊠ ].

Proof
By invoking representation (32) and the previous lemma, the matrices [C⊗ ] and [C ⊠ ] can be
determined by enforcing the condition:
AA−1 = I
Recalling (34) one has
([A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ])([C⊗ ] · [A⊗ ] + [C ⊠ ] · [A⊠ ]) = [1⊗ ] · [A⊗ ] + [1⊠ ] · [A⊠ ]
or, explicitly, by using the convention of repeated indices:

aij⊗ ckl
⊗ ⊗ ⊗
Aij Akl + aij⊗ ckl⊠ A⊗ ⊠ ⊠ ⊗ ⊠ ⊗ ⊠ ⊠ ⊠ ⊠ ⊗ ⊗ ⊠ ⊠
ij Akl + aij ckl Aij Akl + aij ckl Aij Akl = [1 ] · [A ] + [1 ] · [A ]
(43)

The computation of the previous expression requires the composition of dyadic and square
tensor products of the eigenprojectors of A; invoking (5) and (11), the following composition
rules can be established:

Ai ⊗ Al if j = k
⊗ ⊗
Aij Akl = (Ai ⊗ Aj )(Ak ⊗ Al ) =
O otherwise

Ai ⊗ Aj if j = k = l
A⊗ A ⊠
ij kl = (Ai ⊗ Aj )(Ak ⊠ A l ) =
O otherwise
(44)
Ak ⊗ Al if i = j = k
Aij⊠ A⊗
kl = (Ai ⊠ Aj )(Ak ⊗ Al ) =
O otherwise

Ai ⊠ Aj if i = k and j = l
⊠ ⊠
Aij Akl = (Ai ⊠ Aj )(Ak ⊠ Al ) =
O otherwise
Hence, the left-hand side of (43) becomes
aij⊗ cj⊗l A⊗ ⊠ ⊠ ⊠ ⊗ ⊗ ⊠ ⊠
il + aij cij (Aij )i =j = aij cj l Ai ⊗ Al + aij cij (Ai ⊠ Aj + Aj ⊠ Ai )i =j

since the composition of dyadic and square tensor products of eigenprojectors yields non-zero
tensors only when the factors of the square product do coincide, see e.g. (42)2,3 , a circumstance
which is, however, ruled out in the deﬁnition of [A⊠ ] given by (32).
The product aij⊗ cj⊗l is the il-entry of the matrix [A⊗ ][C⊗ ] while aij⊠ cij⊠ represents the
ij -entry of the product of the matrices [A ⊠ ] and [C ⊠ ] performed componentwise, i.e. by
multiplying the elements of the two matrices having the same position.

This operation, denoted by the symbol ∗ in the sequel, is usually termed Hadamard’s product
in the literature [50].
In conclusion, (43) simpliﬁes to

[A⊗ ][C⊗ ] · [A⊗ ] + ([A⊠ ] ∗ [C ⊠ ]) · [A⊠ ] = [1⊗ ] · [A⊗ ] + [1⊠ ] · [A⊠ ]

from which we infer

[A⊗ ][C⊗ ] = [1⊗ ] [A⊠ ] ∗ [C ⊠ ] = [1⊠ ]

whence the result recalling the deﬁnition of [1⊗ ] and [1⊠ ] in formula (34).

3.2. Composition of fourth-order tensors with second-order tensors

Additional operations which are useful in the applications are given by the composition of a
fourth-order tensor A with a second-order symmetric tensor B coaxial with the constituent A
of A, i.e.:

B = b1 A1 + b2 A2 + b3 A3 = [b] · [EA ] (45)

where
  
A1

b1
   
[b] = 
 b2 
 and [EA ] = 
 A2 
 (46)
b3 A3

are the vectors collecting in turn the eigenvalues of B and the eigenprojectors of A.
Speciﬁcally, we need to specialize operations such as AB, AB · B and AB ⊗ AB when a
principal representation of A is assigned. These results are contained in the following three
lemmas.

Lemma 7
Let A be a symmetric fourth-order tensor given as

A = [A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ]

If B ∈ Sym is coaxial with A, it turns out to be

AB = aij⊗ bj Ai = ([A⊗ ][b]) · [EA ]

where [b] is the vector of the principal values bj of B.

Proof
According to (45), we can write

AB = ([A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ])bk Ak = aij⊗ A⊗ ⊠ ⊠

ij bk Ak + aij Aij bk Ak

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 475

On the other hand, it turns out to be

Ai if j = k
A⊗
ij Ak = (Ai ⊗ Aj )Ak =
0 otherwise

Ai if i = j = k
Aij⊠ Ak = (Ai ⊠ Aj )Ak =
0 otherwise

by virtue of (8), (9) and of the deﬁnitions of dyadic and square tensor products. Hence,

AB = aij⊗ bj Ai = ([A⊗ ][b]) · [EA ] (47)

since the elements of Aij⊠ are square tensor products of eigenprojectors with different indices
(i = j ).

The previous result allows us to establish the following:

Lemma 8
Given a fourth-order tensor A in the form

A = [A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ]

and B ∈ Sym coaxial with A, it turns out to be

AB · B = aij⊗ bj bi = ([A⊗ ][b]) · [b]

AB ⊗ AB = aij⊗ bj akl
⊗
bl Ai ⊗ Ak = {([A⊗ ][b]) ⊗ ([A⊗ ][b])} · A⊗

The proof of these last two results immediately follows from (47).
It is thus apparent from the previous two lemmas that only [A⊗ ] does play an effective role
in the composition of A with a second-order symmetric tensor B coaxial with the constituent
A of A.

4. CONSTITUTIVE MODEL

We shall start our considerations with reference to the case of isotropic plasticity and visco-
plasticity in the small deformation regime by summarizing the basic governing equations for the
continuum problem and the relevant discrete formulation. The extension to the large deformation
case will be discussed later in this section.

4.1. Continuum formulation. Small deformations

Let us consider an elasto/viscoplastic structural model undergoing a quasi-static loading process
whose events are ordered by a pseudo-time scalar parameter t.

Under the assumption of small transformations the kinematics of the deformation is charac-
terized by the inﬁnitesimal strain measure and by the additive decomposition:

=e+p (48)

having denoted by e and p the elastic and (visco)plastic shares, respectively. Addressing the
purely mechanical case, we shall make reference to a stored energy function given in fully
decoupled form as [51]

= el (e) + h () (49)

where el and h are isotropic functions, both assumed to be twice differentiable with positive
definite Hessian, and is a strain-like scalar hardening variable.
The constitutive relations for the stress and the thermodynamic affinities are identified by
making use of the classical thermodynamic argument [57]. Namely, from the isothermal dissi-
pation inequality one has

= de el (e); q = d h () (50)

where is the Cauchy stress and q the stress-like internal variable accounting for the evolution
of the yield locus in the stress space. This last one is deﬁned in terms of a general isotropic
scalar function which is assumed to be convex and smooth:

(, q) = (I1 , J2 , J3 ) − q − Y0 (51)

where Y0 depends upon the initial uniaxial yield limits of the material

I1 = tr(); J2 = 21 tr S2 ; J3 = 13 tr S3

are invariants of and S = − 31 tr()1 is the stress deviator.

The evolutionary equations for the (visco)plastic strain and the hardening variable are pro-
vided by the principle of maximum plastic dissipation as

ė = ˙ − ˙ d (, q); ˙ = −˙dq (, q) = ˙ (52)

where
* * *
d (, q) = 1+ S+ S2 − 23 J2 1 = n1 1 + n2 S + n3 S2 (53)
*I1 *J2 *J3
is the gradient of the yield function and ˙ represents the continuum consistency parameter.
In rate-independent plasticity ˙ is characterized as a Lagrange multiplier obeying the load-
ing/unloading conditions in Kuhn–Tucker form

(, q) 0, ˙ 0, ˙ (, q) = 0 (54)

The class of viscoplasticity models based on the overstress law, originally introduced by
Perzyna [52], can be incorporated within the same formalism by appealing to the regularized
version of the maximum dissipation principle. Actually, the viscoplastic evolution equations can

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 477

be expressed in the same way as in (52) by replacing the loading/unloading conditions (54)
with the relationship [53]
1
˙ = ((, q)) (55)

where ∈ (0, +∞) represents a fluidity-like parameter, · is the ramp function, defined as
x = (x + |x|)/2, and denotes the flow function, which has to be taken convex, C (1) ,
monotone non-decreasing and vanishing on ℜ− 0.
The above-defined equations provide a constrained problem of evolution which, in order to
advance the solution within a typical time step [tn , tn+1 ], has to be transformed into a sequence
of constrained optimization problems via a suitable integration algorithm.

4.2. Discrete formulation and return map algorithm. Small deformations

As a consequence of the additive structure of the evolution problem, an update algorithm is
classically based on an elastic–plastic operator split [1]. In particular, the decomposition of
Equations (52) is as follows:

ė = ˙ ; ˙ = 0 elastic part (56)

ė = −˙d (, q); ˙ = ˙ plastic part (57)

Equations (56) along with the initial conditions

e(tn ) = (tn ) − p(tn ) = e0 ; (tn ) = 0 (58)
give an initial value problem amenable to exact solution
e = etr = e0 + ; = tr = 0 (59)
The corresponding values of the stress-like variables
tr = de el (etr ); q tr = d h (tr ) (60)
identify the trial stress state to be checked for plastic consistency.
As shown in Reference [1], the algorithmic statement of the loading/unloading conditions
can be formulated exclusively in terms of the trial state. Accordingly, if tr 0 the trial state
equals the one at solution; otherwise, plastic consistency has to be restored by solving the
plastic equations for which Equations (59) provide the initial conditions. In this last case, the
time integration of (52) by means of the fully implicit (backward Euler) scheme provides
e − etr = −d (, q); − tr = (61)
where is the discrete algorithmic consistency parameter. It satisﬁes, for the rate-independent
case, the discrete consistency conditions in Kuhn–Tucker form
(, q) 0; 0; (, q) = 0 (62)
or the relation

(, q) = −1 (63)
t

which expresses the discrete consistency for Perzyna-type viscoplasticity replacing the condition
(, q) = 0, recovered in the limit as → 0.
For plastic loading (tr > 0) one has the residuals
re(k) = e(k) − etr + (k) d (k)
(k)
r = (k) − tr − (k) (64)
(k) (k)
r = ((k) , q (k) ) − −1 = (k) − (k)
t
whose linearization around the kth estimate of the solution yields [36]
     (k+1) 
re(k) G(k) 0 d (k) (k)
   
 (k)   −1   (k+1) 

 r  +  (k)
   0 (H ) −1   q(k)  =0 (65)
    
(k) (k+1)
r (d (k) )T −1 −d (k) (k)
2 ((k) ) the tangent hardening modulus and G(k) the rank-four tensor:
being H (k) = d h

2 −1 2 (k)
G(k) = (dee el (e(k) )) + (k) d (66)
It turns out to be positive-deﬁnite due to the positive deﬁniteness of the elastic tangent and
the convexity of the yield function; accordingly, system (65) yields
(k) −1 (k) (k)
(k+1)
r − (G(k) ) re · d (k) + H (k) r
(k) = −1
(67)
(G(k) ) d (k) · d (k) + H (k) + d (k)
The above-outlined procedure describes a general iterative solution scheme for the stress
update. The algorithm for the rate-independent case is trivially obtained by the previous one
by ruling out the viscosity-dependent term (k) .

4.3. The tensor G and its inverse

Equation (67) immediately reveals that a major computational burden in the solution of the
boundary value problem in elasto(visco)plasticity is associated with the inversion of the fourth-
order compliance tensor (66) since its inverse needs to be computed at each yielded Gauss
point of the structural model for every (local) constitutive iteration of each (global) equilibrium
iteration.
An effective procedure for the computation of G−1 has been presented in Reference [34]
by providing a representation formula for G−1 whose coefﬁcients can be computed, basically,
by solving a linear system of order three. This procedure was developed for the usual intrinsic
formulation of small strain plasticity and yield functions expressed in terms of stress invariants.
A different approach, based on the spectral decomposition of S:
S = ŝ1 S1 + ŝ2 S2 + ŝ3 S3 (68)
has been envisaged in Reference [35] in order to implement the representation formula
for G−1 .

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 479

In order to establish the relationship between the principal space formulation and the intrinsic
one and to address indifferently yield functions expressed either in terms of stress invariants
or of principal values, the evaluation of G−1 is here carried out by making use of Lemma 6.
We start computing the explicit expression of G by considering the quadratic elastic potential
deﬁned as

el (e) = 1
2 K − 23 G (tr e)2 + G tr(e2 ) (69)

where K and G denote the bulk and shear moduli, respectively; accordingly, the elastic tangent
is given by

2
dee el = 2GI + K − 23 G (1 ⊗ 1) = d1 I + d2 (1 ⊗ 1) (70)

Since

S1 + S2 + S3 = 1

see (10), the elastic tangent is amenable to the following representation:

2
dee el = [D⊗ ] · [S⊗ ] + [D⊠ ] · [S⊠ ]

d1 + d2 S1 ⊗ S1 S1 ⊗ S2 S1 ⊗ S3
   
d2 d2
   
=
 d2 d1 + d2 d2  ·  S2 ⊗ S1
  S2 ⊗ S2 S2 ⊗ S3 

d2 d2 d1 + d2 S3 ⊗ S1 S3 ⊗ S2 S3 ⊗ S3

O
  
S1 ⊠ S2 S1 ⊠ S3

0 d1 d1
   
+
 d1 0  ·  S2 ⊠ S1
d1   O S2 ⊠ S3 

d1 d1 0 S3 ⊠ S1 S3 ⊠ S2 O

on account of (32). By virtue of Lemma 6 it turns out to be

2
(dee el )−1 = [D⊗ ]−1 · [S⊗ ] + [D ⊠ ]−1 · [S⊠ ] = [D̃⊗ ] · [S⊗ ] + [D̃ ⊠ ] · [S⊠ ]

where the entries d̃ij⊗ of [D̃⊗ ] and d̃ij⊠ (i = j ) of [D̃ ⊠ ] are

d1 + 2d2
d̃ii⊗ = no sum on i
d1 (d1 + 3d2 )
d2
d̃ij⊗ = − , i = j
d1 (d1 + 3d2 )
1
d̃ij⊠ = , i = j
d1

The evaluation of G through (66) requires the second derivative of the yield function; in
the general case (51) one has
2
d = (n2 d S + n3 d S2 + 1 ⊗ d n1 + S ⊗ d n2 + S2 ⊗ d n3 )

= e1 I + e2 (S⊠1 + 1⊠S)

+ e3 (1 ⊗ 1) + e4 (S ⊗ 1 + 1 ⊗ S) + e5 (S2 ⊗ 1 + 1 ⊗ S2 )

+ e6 (S ⊗ S) + e7 (S2 ⊗ S + S ⊗ S2 ) + e8 (S2 ⊗ S2 ) (71)

where the coefﬁcients are given by

e1 = d2 = n2 , e2 = d3 = n3

e3 = (− 31 n2 + d11 − 43 J2 d13 + 49 J22 d33 )

e4 = (− 23 n3 + d12 − 23 J2 d23 ), e5 = (d13 − 23 J2 d33 )

e6 = d22 , e7 = d23 , e8 = d33

and
2
* *
di = , dij = ; Pi , Pj ∈ {I1 , J2 , J3 } (72)
*Pi *Pi *Pj
It is worth noting that the presence of the square tensor products entries in (71) is directly
related to the dependence of the yield function upon the third invariant J3 .
The principal space representation of (71) reads
2
d = [F⊗ ] · [S⊗ ] + [F⊠ ] · [S⊠ ] (73)
where
⊗ ⊗ ⊗
f12⊠ f13⊠
 
f11 f12 f13 0
 
 ⊗  
⊗ ⊗
[F⊗ ] =  [F⊠ ] =  f12⊠ f23⊠ 

 f12 f22 f23 and 0
 
  
⊗ ⊗ ⊗
f13 f23 f33 f13⊠ f23⊠ 0
the entries being given by
fii⊗ = e1 + e3 + 2(e2 + e4 )ŝi + (2e5 + e6 )ŝi2 + 2e7 ŝi3 + e8 ŝi4 no sum on i

fij⊗ = e3 + e4 (ŝi + ŝj ) + e5 (ŝi2 + ŝj2 ) + e6 ŝi ŝj + e7 (ŝi2 ŝj + ŝi ŝj2 ) + e8 ŝi2 ŝj2 (74)

fij⊠ = e1 + e2 (ŝi + ŝj ) i = j

In conclusion, recalling (66), we can write
G = ([D̃⊗ ] + [F⊗ ]) · [S⊗ ] + ([D̃ ⊠ ] + [F ⊠ ]) · [S⊠ ] = [G⊗ ] · [S⊗ ] + [G⊠ ] · [S⊠ ] (75)

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 481

where
gii⊗ = d̃ii⊗ + fii⊗ no sum

gij⊗ = d̃ij⊗ + fij⊗ , i = j (76)

gij⊠ = d̃ij⊠ + fij⊠ , i = j

On account of Lemma 6, the computation of the inverse of the fourth-order tensor G is

equivalent to the inversion of the matrix [G⊗ ] collecting the coefﬁcients of its dyadic part and
to the term-by-term inversion of the matrix of the non-dyadic part [G⊠ ].

4.4. Principal space return mapping

In the algorithmic framework discussed in Section 4.2 the return mapping is entirely carried
out in intrinsic form, i.e. by making reference to tensor quantities.
As ﬁrst observed in Reference [12], for the isotropic case of interest an efﬁcient alternative
implementation of the stress update algorithm can be obtained by using a principal axis for-
mulation; moreover, Lemmas 7 and 8 show that in the computation of the return map solution
only the dyadic part of G−1 , i.e. the one pertaining to the linearization with respect to the
principal stresses, is strictly needed. These circumstances suggest to express the return mapping
algorithm in principal space by making reference, at least formally, to a yield function given
in terms of stress eigenvalues.
Denoting by ˆ the yield function expressed in terms of principal stresses and observing that
and etr share the same eigenvectors, we can write
ˆ i
(êi − êitr )Si = −di S (77)
ˆ
by virtue of (25). Collecting the eigenvalues of e, etr and d in the vectors ê, êtr and dˆ ,
respectively, the return mapping can thus be reformulated as follows:
ˆ
ê = êtr − dˆ (78)
where, according to formula (30) and (53), it turns out to be
ˆ = n1 + n2 ŝi + n3 ŝ 2 ,
di i = 1, 2, 3
i

The linearization of (77) can be computed by invoking formula (35) with reference to the
yield function . In particular a comparison of formula (35) with (73) shows that
ˆ = [F⊗ ],
[d2i j ] i, j = 1, 2, 3, ˆ = f ⊠,
dij⊠ i, j = 1, 2, 3, i = j
ij

Note that the second term on the right-hand side of (35), stemming from the analogous one
in formula (26), accounts for the change of the eigenvectors of the relevant tensor. However,
this circumstance is ruled out since all tensors in the non-linear equation (78) are coaxial; this
means that the linearization of this relationship can be carried out at constant eigenvectors, i.e.
ruling out the matrix [F ⊠ ] and making reference directly to the 3 × 3 positive-deﬁnite matrix:
2 ˆ −1 (k)
[Ĝ(k) ] = [dêê el (ê(k) )] + (k) d2ˆ ˆ ˆ = [G⊗ ](k) (79)

Accordingly, setting
(k) (k)
ˆ
rê = ê(k) − êtr + (k) dˆ (80)
one obtains the following expression for the iterative increment of the plastic parameter:
(k) −1 (k) (k) (k)
r ˆ − [Ĝ(k) ] rê ˆ
· dˆ + H (k) r
(k+1)
(k) = −1 (k) (k)
(81)
[Ĝ(k) ] ˆ
dˆ ˆ
· dˆ + H (k) + d (k)
and the remaining state variables can be updated according to the scheme of Table I.
In summary, not only the plastic correction phase can be carried out at ﬁxed eigenvectors
but the relevant tensorial quantities can be obtained exactly in the same way irrespective of
the fact that the yield function is assigned in terms of invariants or of the eigenvalues of the
stress tensor.

4.5. Continuum formulation. Large deformations

The phenomenological description of large-strain plasticity and viscoplasticity considered here
follows the kinematic hypothesis of Lee [54] and Mandel [55] relying upon the notion of
intermediate relaxed (stress-free) configuration.
Accordingly, we assume the multiplicative decomposition
F = Fe Fp (82)
where the superscripts e and p are used to denote the elastic and plastic parts of the local
deformation gradient F with Jacobian J = det F > 0.
The total, elastic and plastic strain tensors are defined in terms of the corresponding de-
formation gradients; for the developments that follow we introduce the right Cauchy–Green
tensors as
C = FT F; Cp = Fp,T Fp (83)
which act on the reference configuration; the left Cauchy–Green tensors are defined on the
current (deformed) configuration and are given by
b = FFT ; be = Fe Fe,T (84)
thus, be is the push-forward [56] of the inverse of the plastic right Cauchy–Green tensor to
the current configuration:
be = (F⊠F)(Cp )−1 (85)
As in the linearized theory, we shall assume that the elastic behaviour is unaffected by the
spread of inelastic deformation; accordingly, given the functional restrictions imposed by the
principle of frame invariance [46] and on account of elastic isotropy, the free energy potential
in the spatial description is taken as
˜ el (Ii (be )) + h ()
= el (be ) + h () = (86)
where Ii (be ), i ∈ {1, 2, 3}, are the principal invariants of be .

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 483

Table I. Principal space return mapping.

1. Initialize: k = 0; ê(k) = etr ; (k) = tr ; (k) = 0; set tolerance tol.

2. Compute trial state
ˆ tr = dê
ˆ êtr ; q tr = d ˆ (tr ); ˆ tr = (ˆ tr , q tr )
el h
3. Check yield:
tr
ˆ > 0) THEN
IF (
GOTO 4
ELSE
Set (·) = (·)tr and EXIT
ENDIF
4. Compute residuals and residual norm
(k)
rê ˆ (k) ;
= ê(k) − êtr + (k) dˆ r
(k)
= (k) − tr − (k) ;
(k)
rˆ ˆ (k)
=

(k) (k) (k) (k)
R (k) = [(r ˆ )2 + rê ·rê + (r )2 ]1/2

5. Convergence test
tr
ˆ ) EXIT
IF (R (k) tol
6. Solve for the consistency parameter
(k) −1 (k) (k) (k)
r ˆ −[Ĝ(k) ] rê ·dˆ ˆ +H (k) r
(k+1)
(k) =
−1 (k) (k)
[Ĝ(k) ] dˆ ˆ ·dˆ ˆ +H (k) +d (k)
7. Update solution
(k+1) = (k) + (k+1)
(k)
−1 (k) (k+1) (k)
ˆ (k+1) = ˆ (k) − [Ĝ(k) ] (rê + (k) ˆ
dˆ )
(k) (k+1)
q (k+1) = q (k) − H (k) (r − (k) )
(k+1)
ˆ = (ˆ (k+1) , q (k+1) )
(k+1) (k+1) (k+1) (k+1) T
ˆ
dˆ ˆ
= [dˆ 1 ˆ
, dˆ 2 ˆ
, dˆ 3 ]
−1
2
ê(k+1) = ê(k) + [dêê ˆ (ê(k) )] (ˆ (k+1) − ˆ (k) )
el
−1
(k+1) = (k) + (H (k) ) (q (k+1) − q (k) )
8. Increment iteration counter: k = k + 1
9. GOTO 4

Use of standard continuum thermodynamics [46] yields, after straightforward manipulations

of the expression of the local dissipation function, the constitutive relations for the Kirchhoff
stress and the static internal variable q as
= [2dbe el (be )]be ; q = d h () (87)
˙ being Lv (be )
and identiﬁes the local (visco)plastic ﬂow with the rates (− 21 Lv (be )(be )−1 , ),
the Lie derivative with respect to the spatial velocity v of the elastic left Cauchy Green tensor

deﬁned as
˙
Lv (be ) = (F ⊠ F)[(Cp )−1 ] (88)
Following [12], the yield function is postulated as in (51), with the only exception that the
stresses and S are now replaced by the Kirchhoff stress measure and the relevant devia-
tor, and the principle of maximum plastic dissipation is assumed. Accordingly, the following
evolution equations are obtained:
b˙e = lbe + be lT − [2˙d (, q)]be ; ˙ = −˙dq (, q) = ˙ (89)
where l = ḞF−1 is the gradient of the spatial velocity v while ˙ is the consistency parameter
subjected to constraints (54) or (55).

4.6. Discrete formulation. Large deformations

Apart from minor modiﬁcations, the considerations developed in Section 4.2 for the inﬁnites-
imal theory do also apply to the case at hand. In particular, the problem of evolution (89) can
be split as follows:

b˙e = lbe + be lT ; ˙ = 0 elastic part (90)

b˙e = −[2˙d (, q)]be ; ˙ = ˙ plastic part (91)

for fixed intermediate configuration and fixed actual configuration, respectively.

Equations (90) along with the initial conditions
be (tn ) = be0 , (tn ) = 0 (92)
are amenable to exact solution. Actually, (90) follow from (89) by keeping fixed the plastic
right Cauchy–Green tensor so that they can be rephrased as
˙
b˙e = [lbe + be lT + Lv (be )]|Cp =Cp (tn ) = [be,tr ]; ˙ = 0 (93)
whose solution amounts to the update of the current configuration
be = be,tr = Fe,tr (Fe,tr )T = (f⊠f)be0 ; = tr = 0 (94)
In the previous formula Fe,tr is the trial elastic deformation gradient
Fe,tr = F|t=tn+1 (Fp )−1 |t=tn = f|t=tn+1 Fe |t=tn (95)
and f the relative deformation gradient
f|t=tn+1 = F|t=tn+1 F−1 |t=tn (96)
For tr > 0, Equations (91), along with the initial conditions (94), are used to restore plastic
consistency. Specifically, for fixed actual configuration time integration of the flow equations
(91) yields the non-linear equations:
be = exp[−2d (, q)]be,tr ; = tr + (97)
where the implicit exponential approximation has been employed.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 485

The use of the exponential map for the time integration of (89)1 , as originally proposed
in Reference [57], is well known to entail some important algorithmic properties namely, the
preservation of the plastic volume for pressure-insensitive yield criteria and a particularly simple
form of the return mapping algorithm in principal space [12]. This last property stems from
the fact that, under the assumption of isotropy, all terms in (97)1 commute so that the return
mapping takes place at ﬁxed eigenvectors which are obtained from the symmetric eigenvalue
problem:
e,tr 2
[be,tr − ( tr
i ) 1]si =0 (98)

where ie,tr are the principal elastic trial stretches and sitr the relevant eigenvectors.
Hence, by expressing (97)1 with reference to principal axes and taking the logarithm of
both sides one obtains the same equation as in (78) with ê and êtr being now replaced by the
vectors of principal values of the elastic logarithmic strain tensors:

e= 1
2 ln(be ); etr = 1
2 ln(be,tr ) (99)

With this result at hand, and noting that the isotropy assumption for the elastic potential
entails a constitutive relation for the principal values of the Kirchhoff stress completely similar
to the one holding in the inﬁnitesimal case, i.e.: ˆ = dê ˆ el (ê), it is immediate to recognize
that the principal return mapping does possess the same structure of that of the inﬁnitesimal
theory. In particular, taking the Hencky potential:

3

ˆ el (ê) = 1 K − 2 G (ln J e )2 + G (ln
e 2
(100)
i)
2 3 i=1

where J e = det (Fe ) is the elastic Jacobian and K and G are the usual elastic moduli, the
spectral return mapping algorithm discussed in Section 4.4 applies with no modiﬁcation.

4.7. Consistent tangent operator

An essential ingredient for the efﬁcient solution of the discretized boundary value problem
arising in elasto(visco)plasticity via Newton’s method is the computation of the consistent
tangent tensor [4]. In particular, the objective of this section is that of evaluating the constitutive
or material tangent; as opposite to the local algorithm linearization, in which no change in the
principal vectors is accounted for, the material contribution to the tangent tensor requires the
full linearization of the discretized constitutive relations.
We shall ﬁrst examine the large deformation case for which, in the adopted formulation, the
devised relation is [8]

◦ = Etan d (101)

where ◦ is the Truesdell rate [56] of the Cauchy stress, d is the rate of deformation tensor
and Etan is the material contribution to the spatial elasto(visco)plastic tangent.
In order to work out the expression of Etan we start by computing the tangent moduli tensor
relative to the intermediate relaxed conﬁguration, which is in turn obtained as the gradient of

the symmetric stress tensor T deﬁned as

−1
T = (Fe,tr ⊠ Fe,tr ) (102)

i.e. the pull-back of the Kirchhoff stress to the intermediate conﬁguration at t = tn .

Recalling that the spectral form of the Kirchhoff stress is
3
3

= ˆ i (si ⊗ si ) = ˆ i Si (103)
i=1 i=1

the spectral decomposition of T can be obtained as

3 3
tˆi (ti ⊗ ti ) = tˆi Ti

T= (104)
i=1 i=1

where the principal values tî and the unit principal vectors ti are given by
î 1 e,tr e,tr −1
tî = e,tr 2 ; ti = e,tr (Fe,tr )T si = i (F ) si (105)
( i ) i

These last ones also eigenvectors of the trial elastic right Cauchy–Green tensor

Ce,tr = (Fe,tr )T Fe,tr (106)

as one can easily verify upon premultiplication by (Fe,tr )T of the eigenvalue statement (98);
hence, the eigenprojectors Ti of T:
e,tr 2 e,tr −1
Ti = ( i ) (F ⊠ Fe,tr ) Si (107)

are also eigenprojectors of Ce,tr .

By introducing the strain rate tensor relative to the ﬁxed intermediate conﬁguration at t = tn :

Ce,tr = 2 (Fe,tr ⊠ Fe,tr )T d = 2 sym[(Fe,tr )T Fe,tr ] (108)

the directional derivative of the stress tensor (104) along Ce,tr is given by
3 3
dCe,tr T Ce,tr = [dCe,tr (tˆi ) · Ce,tr ]Ti + tˆi dCe,tr (Ti ) Ce,tr

(109)
i=1 i=1

By exploiting the chain rule the derivative dCe,tr (tˆi ) is computed as

3
dCe,tr (tˆi ) = dêtr (tˆi )dCe,tr (êjtr )

j
j =1

e,tr 2
3 1 tr
3 2ˆ i ij ( j )
dCe,tr (êjtr )

= d tr
e,tr 2 êj (ˆ i )dCe,tr (ê j ) − e,tr 4 (110)
j =1 ( i ) j =1 ( i )

e,tr
where êjtr = ln j and ij is the Kronecker symbol.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 487

e,tr 2
Since ( j ) are eigenvalues of Ce,tr , Lemma 1 supplies
e,tr 2
dCe,tr (( j ) ) = Tj (111)

from which we infer

1
dCe,tr (êjtr ) = e,tr 2 Tj (112)
2( j )

Accordingly, the ﬁrst term on the right-hand side of (109) is

3 3 1
[dCe,tr (tˆi ) · Ce,tr ]Ti = dêtr (ˆ i )(Ti ⊗ Tj )Ce,tr

e,tr 2 e,tr 2 j
i=1 i,j =1 2( i ) ( j )

3 ˆi
⊗ Ti )Ce,tr

− e,tr 4 (Ti (113)
i=1 ( i )

Further, invoking relation (18) one has

Ti ⊠ Tj + Tj ⊠ Ti Ti ⊠ Tk + Tk ⊠ Ti
dCe,tr Ti = e,tr 2 e,tr 2 + e,tr 2 e,tr 2 , i = j, k (114)
( i ) −( j ) ( i ) −( k )

so that the second term on the right-hand side of (109) becomes

3 3 tˆi − tˆj
tˆi dCe,tr (Ti ) Ce,tr = ⊠ Tj )Ce,tr

e,tr 2 e,tr 2 (Ti
i=1 i,j =1 ( i ) −( j )
i =j

e,tr 2 e,tr 2
3 1 ˆi ( j ) − ˆj ( i )
⊠ Tj )Ce,tr

= e,tr 2 e,tr 2 e,tr 2 e,tr 2 (Ti (115)
i,j =1 ( i ) ( j ) ( i ) −( j )
i =j

The spatial tangent deﬁned by (101) can now be given an explicit expression by exploiting
the push-forward relations [56]:
1 1
◦ = Lv () = (Fe,tr ⊠ Fe,tr )dCe,tr TCe,tr (116)
J J
2d = Lv (1) = (Fe,tr ⊠ Fe,tr )−T Ce,tr (117)

Accordingly, use of (116) and (117) yields

e,tr 2
1 3 3 3 ˆ i ( j ) − ˆ j ( ie,tr )2
Etan =

dêtr (ˆ i )(Si ⊗ Sj ) − 2ˆ i (Si ⊗ Si ) + 2 e,tr 2 e,tr 2 (Si ⊠ Sj )
J i,j =1 j
i=1 i,j =1 ( i ) −( j )
i =j
(118)

By virtue of (6)1 the previous expression coincides with formula (A.33) reported in Reference
[8]. It is then immediate to verify that such formulas have the same structure as for the
Lagrangian description of multiplicative elastoplasticity; see e.g. Equations (34)–(35) and (46)
of Reference [8].
To obtain the ﬁnal expression of the spatial tangent it remains to evaluate the term
3
dêtr (ˆ i )(Si ⊗ Sj ) = [dêtr ˆ ] · [S⊗ ]

j
i,j =1

i.e. the dyadic part of the derivative detr .

This term can be computed from the linearization of the principal return mapping algorithm
at the local converged state (t = tn+1 ) by considering the system
 
[G⊗ ] ˆ

−êtr 0 dˆ dêtr ˆ
 
   −1
 
 0 + 0 [H ] −1   dêtr q =0 (119)

  

0 ˆ T
(dˆ ) −1 −d dêtr

whose solution with respect to dêtr ˆ yields

[G⊗ ]−1 dˆ ˆ

ˆ ⊗ [G⊗ ]−1 dˆ
dêtr ˆ = [G⊗ ]−1 − (120)
ˆ · dˆ
[G⊗ ]−1 dˆ ˆ + H + d

that exactly coincides with the dyadic part of the consistent tangent of the inﬁnitesimal theory,
see e.g. Reference [35].
With reference to this last case, by following a path of reasoning identical to the one
exploited before, one can compute the derivative of the Cauchy stress with respect to the
inﬁnitesimal trial elastic strain etr as
3
3 ˆ i − ˆ j
detr = dêtr (ˆ i )(Si ⊗ Sj ) + tr tr (Si ⊠ Sj ) (121)
i,j =1 êi − êj
j
i,j =1
i =j

in which the coefﬁcients multiplying the terms Si ⊠ Sj can be easily shown to coincide with
the reciprocals of the terms gij⊠ deﬁned in (76). In this respect, by observing that

ˆ i − ˆ j ŝi − ŝj
tr tr = 2G tr (122)
êi − êj ŝi − ŝjtr

where ŝitr is the ith principal value of the deviator of the trial Cauchy stress, it is immediate
to verify that the return map solution for ŝi − ŝj is given by

ŝi − ŝj = ŝitr − ŝjtr − 2G(ŝi − ŝj )[n2 + n3 (ŝi + ŝj )] (123)
with n2 , n3 given as in (53); accordingly
ŝi − ŝj 1 1 1
= = (124)
ŝitr − ŝjtr 1 + 2G[n2 + n3 (ŝi + ŝj )] 2G gij⊠

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 489

As clearly shown by the previous formulas, the square tensor product terms in the expression
of the tangent tensor are due to the change in the eigenvectors; in the small deformation case this
is simply accounted for by the non-dyadic part of the tensor G while in the ﬁnite deformation
case the non-dyadic part of the tangent tensor, which originates from the push-forward of (115),
is slightly more complicated due to the non-linearity of the strain measure.
As a ﬁnal remark, we emphasize that the above expressions provide the consistent tangent
operator directly in the given reference frame. Indeed, the spectral representation of the stress
tensor implies that the matrix [Si ] = [si ⊗ si ] of the generic eigenprojector is assigned in the
same reference frame as that of [S] so that no co-ordinate transformation between the principal
frame and the reference Cartesian one is needed for implementing the consistent tangent.

5. NUMERICAL EXAMPLE

To demonstrate the performance of the solution scheme discussed in the previous sections a
numerical example is presented. The simulation refers to an isotropic model deﬁned through
the following pressure-insensitive yield function:

[2J2 ]1/2
(J2 , J3 ) = − fc′ (125)
r( )

where fc′ is the magnitude of the limit stress in uniaxial compression and r( ) is a function
deﬁning the shape of the failure surface in the deviatoric plane:

2(rc2 − rt2 ) cos − (rc − 2rt )[4(rc2 − rt2 ) cos2 + 5rt2 − 4rc rt ]1/2
r( ) = rc (126)
4(rc2 − rt2 ) cos2 + (rc − 2rt )2

The parameter is the Lode angle deﬁned as

√
1 3 3 J3
= arccos , ∈ 0, (127)
3 2 J 3/2 3
2

where the constants rt and rc are non-dimensional material parameters of the model which
are representative of the shear strength on the tensile and compressive
√ meridians; they
√ are
computed from the uniaxial tensile and compressive limits as rt = 2/3ft′ /fc′ and rc = 2/3.
The deviatoric section of the yield locus defined by (125) is plotted in Figure 1.
Equation (126) has been originally proposed by Willam and Warnke in Reference [21] as a
part of the celebrated five parameter model, which is one of the most successful descriptions
of the triaxial failure envelope of concrete; moreover, the same expression has been used for
modelling soil behaviour [58]. This equation defines a smooth elliptic interpolation between
the tensile ( = 0) and compressive ( = /3) meridians which results in a surface possessing
convexity within the range 21 rt /rc 1. For a detailed account on the derivation of (126) the
reader may refer to Reference [22].
For rt /rc = 1 the influence of the J3 invariant through the Lode angle is dropped out
and the yield surface collapses to the Von Mises cylinder, while the minimum shape factor

Figure 1. Deviatoric sections of the yield locus.

1
rt /rc = 2 corresponds to a triangular shape of the deviatoric section described by
1/2
2
(J2 , J3 ) = [2J2 ]1/2 2 cos − fc′ (128)
3

5.1. Elastic–plastic plate with ﬂat hole

The numerical example here considered has been first analysed in Reference [59] to test the
behaviour of stabilized mixed finite element methods in the framework of finite elasticity.
The problem consists of the large-strain analysis of a square plate with a central slot under
plane strain which is subject to normal boundary conditions along the whole outer edge and
to imposed displacements in the vertical direction; see Figure 2. For the finite element
analysis only a quarter of the plate has been modelled with 366 Q1/P0 plane elements by
applying the appropriate symmetry boundary conditions. The loading history is of cyclic type,
see Figure 3, and has been assigned using three different sets of initial time increments:
t = 0.4, 0.5, 1.0. Elastic-perfectly plastic behaviour has been considered; the elastic properties
are taken as K = 8333.3 and G = 3846.15 and the data set for the yield limits are:

(i) ft′ = fc′ = 100.0;

(ii) ft′ = 100.0; fc′ = 142.86;
(iii) ft′ = 100.0; fc′ = 185.71;

which correspond to a shape factor rc /rt for the yield surface equal to 1, 10/7, 13/7, re-
spectively; see also Figure 1. The relevant load–deﬂection curves are reported in Figure 4

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 491

Figure 2. Square strip with a slot. Model problem and ﬁnite elements mesh.

Figure 3. Plate with ﬂat hole. Loading data.

and refer to the solution obtained for t = 0.4. Computations have been carried out by
using a customized version of the ﬁnite element code FEAP rel. 7.1b [60]; all of them
have been successfully completed in 125, 100 and 50 load steps by using a local linear line
search scheme to render the return map algorithm globally convergent. In the numerical ex-

Figure 4. Plate with ﬂat hole. Load–deﬂection curves. t = 0.4.

amples described below use has been made of a global termination criterion expressed in
terms of the incremental energy norm [11] with a tolerance = 10−16 . The convergence
behaviour is illustrated in Tables II and III in terms of this norm for some typical load
steps.

6. CONCLUDING REMARKS

A uniﬁed approach for the formulation and implementation of isotropic elasto(visco)plastic

models endowed with yield functions of arbitrary type has been presented. The algorithmic
framework detailed in the paper is exactly the same irrespective of the fact that the yield
function is assigned in terms of the set of stress invariants (I1 , J2 , J3 ) or of stress eigenvalues.
The result has been achieved by exploiting some original results, proved in the paper, concerning
rank-four tensors A obtained as second derivatives of isotropic scalar functions of a symmetric
rank-two tensor A and the differentiation of the eigenprojectors Ak of A with respect to the
tensor A.
In particular, the derivative of the eigenprojectors of rank-two symmetric tensors with respect
to the tensor itself has been provided by an original formula, more compact than the existing
ones. The contributed expression clariﬁes the meaning and the role of the square tensor products
entering the expression of the material tangent tensor which is used in the solution of the
elasto(visco)plastic boundary value problem via Newton’s method.
In the detailed framework the state variables and the tangent tensor are always referred to
the global co-ordinate system so that no co-ordinate transformation is needed for the numerical
implementation.
The specialization of the proposed approach to the signiﬁcant class of plane stress problems
will be pursued in a forthcoming paper.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
Copyright 䉷 2004 John Wiley & Sons, Ltd.

Table II. Plate with ﬂat hole. Energy norms for typical load steps for rc /rt = 10/7.

Time 6.00E + 00 1.00E + 01 1.60E + 01 2.60E + 01 3.00E + 01 3.60E + 01 4.60E + 01 5.00E + 01

displacement 2.40E − 02 4.00E − 02 1.60E − 02 3.00E − 02 5.00E − 02 2.00E − 02 3.60E − 02 6.00E − 02
iteration

1 2.954E − 01 2.940E − 01 3.096E − 01 4.678E − 01 4.648E − 01 4.914E − 01 6.829E − 01 6.793E − 01

ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS

2 8.271E − 03 1.660E − 02 2.176E − 03 1.178E − 02 2.599E − 02 5.580E − 03 1.689E − 02 3.230E − 02
3 4.948E − 04 1.119E − 03 1.813E − 05 8.700E − 04 2.165E − 03 9.569E − 05 1.495E − 03 3.981E − 03
4 6.831E − 06 1.469E − 05 1.277E − 08 4.214E − 06 4.154E − 05 2.016E − 07 1.025E − 05 1.568E − 04
5 2.497E − 09 5.968E − 09 3.533E − 14 7.040E − 10 3.520E − 08 4.055E − 12 3.880E − 09 8.413E − 07
6 1.772E − 15 1.900E − 15 1.555E − 24 4.680E − 17 3.957E − 14 2.224E − 20 3.057E − 15 2.486E − 12
7 5.981E − 27 1.559E − 25 5.359E − 24 2.960E − 26 7.416E − 22
t 4.00E − 01

1 4.618E − 01 4.594E − 01 4.834E − 01 7.315E − 01 7.262E − 01 7.674E − 01 1.068E + 00 1.061E + 00

2 1.238E − 02 2.484E − 02 3.308E − 03 1.699E − 02 3.889E − 02 8.454E − 03 2.540E − 02 4.803E − 02
3 9.613E − 04 2.188E − 03 3.427E − 05 1.410E − 03 4.119E − 03 1.876E − 04 2.526E − 03 7.016E − 03
4 1.944E − 05 4.403E − 05 3.728E − 08 1.129E − 05 1.159E − 04 5.299E − 07 2.597E − 05 3.923E − 04
5 1.629E − 08 4.319E − 08 2.596E − 13 3.755E − 09 2.254E − 07 1.913E − 11 1.681E − 08 5.750E − 06
Int. J. Numer. Meth. Engng 2004; 60:461–498

6 2.176E − 14 4.204E − 13 8.133E − 23 1.081E − 17 1.322E − 12 4.497E − 19 2.052E − 14 2.178E − 09

7 3.661E − 23 6.524E − 23 4.577E − 22 3.648E − 25 8.682E − 18
t 5.00E − 01

1 1.852E + 00 1.838E + 00 1.927E + 00 2.939E + 00 2.904E + 00 3.060E + 00 4.287E + 00 4.245E + 00

2 4.238E − 02 8.279E − 02 1.176E − 02 5.167E − 02 1.298E − 01 2.956E − 02 9.152E − 02 1.588E − 01
3 6.553E − 03 1.463E − 02 2.774E − 04 8.878E − 03 2.319E − 02 1.510E − 03 2.114E − 02 3.533E − 02
4 3.211E − 04 8.725E − 04 1.907E − 06 1.889E − 04 1.773E − 03 1.050E − 05 4.971E − 04 3.794E − 03
5 2.179E − 06 8.478E − 06 2.530E − 08 2.891E − 07 2.918E − 05 5.859E − 09 1.679E − 06 9.691E − 05
6 2.057E − 10 2.005E − 09 1.573E − 11 1.548E − 12 1.615E − 08 2.358E − 14 5.613E − 10 1.445E − 07
7 7.046E − 17 3.958E − 14 7.096E − 18 8.896E − 21 3.828E − 13 6.518E − 21 1.952E − 13 1.045E − 12
8 4.008E − 21 9.123E − 22 1.222E − 20 2.361E − 18
t 1.00E + 00

493
Copyright 䉷 2004 John Wiley & Sons, Ltd.

494
Table III. Plate with ﬂat hole. Energy norms for typical load steps for rc /rt = 13/7.

Time 6.00E + 00 1.00E + 01 1.60E + 01 2.60E + 01 3.00E + 01 3.60E + 01 4.60E + 01 5.00E + 01

displacement 2.40E − 02 4.00E − 02 1.60E − 02 3.00E − 02 5.00E − 02 2.00E − 02 3.60E − 02 6.00E − 02
iteration

1 2.946E − 01 2.909E − 01 3.068E − 01 4.639E − 01 4.572E − 01 4.859E − 01 6.737E − 01 6.671E − 01

2 5.434E − 03 9.696E − 03 1.319E − 03 6.096E − 03 1.667E − 02 3.363E − 03 1.350E − 02 2.293E − 02
3 5.860E − 04 2.333E − 03 5.761E − 06 5.307E − 04 4.345E − 03 6.320E − 05 1.627E − 03 6.000E − 03
4 4.349E − 06 2.984E − 05 1.705E − 08 3.123E − 06 7.748E − 05 1.488E − 07 1.960E − 05 1.126E − 04
7.695E − 10 2.594E − 08 3.586E − 12 3.205E − 10 8.161E − 08 2.813E − 10 9.689E − 09 1.150E − 07

L. ROSATI AND N. VALOROSO

5
6 1.910E − 17 7.083E − 14 3.038E − 19 3.883E − 18 2.231E − 13 2.150E − 15 1.983E − 12 3.979E − 13
7 1.441E − 19 4.824E − 20 3.057E − 23 4.836E − 19 8.546E − 19
t 4.00E − 01

1 4.605E − 01 4.545E − 01 4.790E − 01 7.256E − 01 7.144E − 01 7.585E − 01 1.053E + 00 1.042E + 00

2 8.276E − 03 1.471E − 02 2.015E − 03 9.069E − 03 2.541E − 02 5.121E − 03 2.033E − 02 3.499E − 02
3 9.753E − 04 3.814E − 03 1.200E − 05 8.503E − 04 7.062E − 03 1.057E − 04 2.469E − 03 9.615E − 03
4 1.038E − 05 7.130E − 05 3.675E − 08 6.794E − 06 1.775E − 04 2.562E − 07 4.116E − 05 2.506E − 04
5 3.588E − 09 1.151E − 07 1.211E − 11 1.203E − 09 3.523E − 07 5.243E − 10 2.790E − 08 4.596E − 07
Int. J. Numer. Meth. Engng 2004; 60:461–498

6 2.284E − 16 7.437E − 13 4.546E − 17 7.944E − 17 3.316E − 12 5.946E − 15 4.467E − 14 4.086E − 12

7 3.022E − 19 4.709E − 18 2.424E − 20 5.214E − 19 2.122E − 17
t 5.00E − 01

1 1.848E + 00 1.820E + 00 1.908E + 00 2.918E + 00 2.860E + 00 3.021E + 00 4.234E + 00 4.172E + 00

2 3.008E − 02 5.321E − 02 7.113E − 03 2.989E − 02 9.209E − 02 1.796E − 02 6.612E − 02 1.280E − 01
3 4.576E − 03 1.650E − 02 1.023E − 04 3.059E − 03 2.889E − 02 6.069E − 04 1.014E − 02 3.939E − 02
4 1.182E − 04 8.338E − 04 4.812E − 07 5.929E − 05 1.877E − 03 2.319E − 06 4.616E − 04 2.561E − 03
5 2.444E − 07 6.876E − 06 1.070E − 09 9.452E − 08 2.376E − 05 9.812E − 09 8.469E − 07 2.332E − 05
6 3.250E − 12 9.670E − 10 2.178E − 13 2.159E − 13 1.131E − 08 1.405E − 11 8.657E − 10 8.271E − 09
7 1.200E − 18 9.029E − 17 9.216E − 18 7.663E − 18 7.747E − 17 1.822E − 16 3.353E − 16 3.041E − 13
8 3.312E − 17
t 1.00E + 00
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 495

APPENDIX A

For the sake of completeness, we provide hereafter the matrix representations of the fourth-
order tensors referred to in the body of the paper. Additional details on this topic can be found
in Reference [61].
The Cartesian components of the dyadic and the square tensor products between rank-two
tensors A, B ∈ Lin are

(A ⊗ B)ij kl = Aij Bkl (A ⊠ B)ij kl = Aik Bj l ∀ A, B ∈ Lin

Consistent with the tensor-to-matrix mapping commonly employed in the computational me-
chanics literature [9, 11], stress-(T) and strain-like (D) rank-two symmetric tensors are expressed
in vector form as follows:

T = [T11 , T22 , T33 , T12 , T23 , T31 ]

D = [D11 , D22 , D33 , 2D12 , 2D23 , 2D31 ]

Accordingly, given A, B ∈ Sym, a fourth-order tensor A ⊗ B mapping strain to stress tensors

is amenable to the following matrix representation:

A11 B11 A11 B22 A11 B33 A11 B12 A11 B23 A11 B31
 

 A22 B11 A22 B22 A22 B33 A22 B12 A22 B23 A22 B31 
 
 
 A33 B11 A33 B22 A33 B33 A33 B12 A33 B23 A33 B31 
 
[A ⊗ B] =  
 A12 B11 A12 B22 A12 B33 A12 B12 A12 B23 A12 B31 
 
 
A B A23 B22 A23 B33 A23 B12 A23 B23 A23 B31 
 23 11 
A31 B11 A31 B22 A31 B33 A31 B12 A31 B23 A31 B31

which trivially follows from the definition of the dyadic tensor product. Analogously, from the
definition of the square tensor product, it is not difficult to show that
 A11 B12 + A12 B11 A12 B13 + A13 B12 A13 B11 + A11 B13 
A11 B11 A12 B12 A13 B13
 2 2 2 
 

 A21 B21 A21 B22 + A22 B21 A22 B23 + A23 B22 A23 B21 + A21 B23 
A22 B22 A23 B23 

 2 2 2 

A31 B32 + A32 B31 A32 B33 + A33 B32 A33 B31 + A31 B33
 
 
 A31 B31 A32 B32 A33 B33 
 2 2 2 
[A ⊠ B] =  
 A11 B22 + A12 B21 A12 B23 + A13 B22 A13 B21 + A11 B23 
 A11 B21 A12 B22 A13 B23
 
2 2 2

 
 

 A21 B31 A21 B32 + A22 B31 A22 B33 + A23 B32 A23 B31 + A21 B33 
A22 B32 A23 B33 

 2 2 2 

A31 B12 + A32 B11 A32 B13 + A33 B12 A33 B11 + A31 B13
 
A31 B11 A32 B12 A33 B13
2 2 2

and I = 1 × 1 = diag[1, 1, 1, 1/2, 1/2, 1/2].

ACKNOWLEDGEMENTS
The authors acknowledge the Italian National Research Council (CNR) and the Italian Ministry of
Education, University and Research (MIUR) for the ﬁnancial support.
This work has been partly carried out within the framework of the activities of the Laboratoire
Lagrange, a European research group gathering CNR, CNRS, Università di Roma ‘Tor Vergata,’
Université de Montpellier II, ENPC, LCPC.

REFERENCES
1. Simo JC, Hughes TJR. Computational Inelasticity. Springer: New York, 1998.
2. Simo JC. Numerical analysis and simulation of plasticity. In Handbook of Numerical Analysis, Ciarlet PG,
Lyons J-L (eds), vol. VI. Elsevier: Amsterdam, 1998.
3. Hofstetter G, Simo JC, Taylor RL. A modified cap model: closest point solution algorithms. Computers and
Structures 1993; 46:203–214.
4. Simo JC, Taylor RL. Consistent tangent operators for rate-independent elastoplasticity. Computer Methods in
Applied Mechanics and Engineering 1985; 48:101–118.
5. Simo JC, Taylor RL, Pister KS. Variational and projection methods for the volume constraint in finite
deformation elasto-plasticity. Computer Methods in Applied Mechanics and Engineering 1985; 51:177–208.
6. Simo JC, Taylor RL. A return mapping algorithm for plane stress elastoplasticity. International Journal for
Numerical Methods in Engineering 1986; 22:649– 670.
7. Simo JC, Ju W, Pister KS, Taylor RL. An assessment of the cap model: consistent return algorithms and
rate-dependent extensions. Journal of Engineering Mechanics (ASME) 1988; 144:191–218.
8. Bonet J, Wood RD. Nonlinear Continuum Mechanics for Finite Element Analysis. Cambridge University
Press: Cambridge, MA, 1997.
9. Crisfield MA. Non-linear Finite Element Analysis of Solids and Structures. Vol I: Essentials. Wiley: Chichester,
1991.
10. Crisfield MA. Non-linear Finite Element Analysis of Solids and Structures. Vol II: Advanced Topics. Wiley:
Chichester, 1997.
11. Zienkiewicz OC, Taylor RL. The Finite Element Method (5th edn). Butterworth-Heinemann: London, 2000.
12. Simo JC. Algorithms for static and dynamic multiplicative plasticity that preserve the classical return
mapping schemes of the infinitesimal theory. Computer Methods in Applied Mechanics and Engineering
1992; 99:61–112.
13. Argyris JH, Faust G, Szimmat J, Warnke EP, Willam KJ. Recent developments in finite element analysis of
prestressed concrete reactor vessels. Nuclear Engineering and Design 1974; 28:42–75.
14. Argyris JH, Faust G, Willam KJ. Limit load analysis of thick-walled concrete structures: a finite element
approach to fracture. Computer Methods in Applied Mechanics and Engineering 1976; 8:215 –243.
15. Aubertin M, Li L, Simon R, Khalfi S. Formulation and application of a short-term strength criterion for
isotropic rocks. Canadian Geotechnical Journal 1999; 36:947–960.
16. Hsieh SS, Ting EC, Chen WF. A plastic fracture model for concrete. International Journal of Solids and
Structures 1982; 18:181–197.
17. Labbane M, Saha NK, Ting EC. Yield criterion and loading function for concrete plasticity. International
Journal of Solids and Structures 1993; 30:1269–1288.
18. Menetrey P, Willam KJ. Triaxial failure criterion for concrete and its generalization. ACI Structural Journal
1995; 92:311–318.
19. Ottosen NS. A failure criterion for concrete. Journal of Engineering Mechanics (ASCE) 1977; 103:527–535.
20. Wang J, Bathe KJ, Walczak J. A stress integration algorithm for J3-dependent elasto-plasticity models. In
Computational Fluid and Solid Mechanics. Proceedings of the 1st M.I.T. Conference, Cambridge, MA, Bathe
KJ (ed.). Elsevier: Amsterdam, 2001.
21. Willam KJ, Warnke EP. Constitutive models for triaxial behaviour of concrete. International Association
for Bridges and Structural Engineering, Seminar on Concrete Structures Subject to Triaxial Stresses, Paper
III-01, Bergamo, 1974; 1–30.
22. Chen WF, Saleeb AF. Constitutive Equations for Engineering Materials. Wiley: New York, 1982.
23. Gillet Y, Patoor E, Berveiller M. Calculation of pseudoelastic elements using a non symmetrical
thermomechanical transformation criterion and associated rule. Journal of Intelligent Materials Systems and
Structures 1998; 9:366 –378.

Copyright 䉷 2004 John Wiley & Sons, Ltd. Int. J. Numer. Meth. Engng 2004; 60:461–498
ISOTROPIC ELASTO/VISCO-PLASTIC MATERIALS 497

24. Raniecki B, Lexcellent C. Thermodynamics of isotropic pseudoelasticity in shape memory alloys. European
Journal of Mechanics A/Solids 1998; 17:185 –205.
25. Armero F, Pérez-Foguet A. On the formulation of closest-point projection algorithms in elastoplasticity. Part I:
The variational structure. International Journal for Numerical Methods in Engineering 2002; 53:297–329.
26. Auricchio F, Taylor RL. A return map algorithm for general associative isotropic elasto-plastic materials in
large deformation regimes. International Journal of Plasticity 1999; 15:1359–1378.
27. Fuschi P, Dutko M, Peric D, Owen DRJ. On numerical integration of the five parameter model for concrete.
Computers and Structures 1994; 53:825 –838.
28. Matzenmiller A, Taylor RL. A return mapping for isotropic elastoplasticity. International Journal for Numerical
Methods in Engineering 1994; 37:813–826.
29. Pérez-Foguet A, Armero F. On the formulation of closest-point projection algorithms in elastoplasticity.
Part II: Globally convergent schemes (with application to deviatoric and pressure-dependent plastic models).
International Journal for Numerical Methods in Engineering 2002; 53:331–374.
30. Peric D, De Souza Neto EA. A new computational model for Tresca plasticity at finite strains with an
optimal parametrization in the principal space. Computer Methods in Applied Mechanics and Engineering
1999; 171:463– 489.
31. Luccioni LX, Pestana JM, Rodriguez-Marek A. An implicit integration algorithm for the finite element
implementation of a nonlinear anisotropic material model including hysteretic nonlinearity. Computer Methods
in Applied Mechanics and Engineering 2000; 190:1827–1844.
32. Luccioni LX, Pestana JM, Taylor RL. Finite element implementation of non-linear elastoplastic constitutive
laws using local and global explicit algorithms with automatic error control. International Journal for
Numerical Methods in Engineering 2001; 50:1191–1212.
33. Sloan SW, Abbo AJ, Sheng D. Refined explicit integration of elastoplastic models with automatic error
control. Engineering Computations 2001; 18:121–154.
34. Palazzo V, Rosati L, Valoroso N. Solution procedures for J3 plasticity and viscoplasticity. Computer Methods
in Applied Mechanics and Engineering 2001; 191:903–939.
35. Palazzo V, Rosati L, Valoroso N. Computational issues of general isotropic elastoplastic models, In ECCOMAS
2000 International Conference, Barcelona, Oñate E, Bugeda G, Suàrez B (eds). CIMNE: Barcelona, 2000.
36. Betsch P, Steinmann P. Derivation of the fourth-order tangent operator based on a generalized eigenvalue
problem. International Journal of Solids and Structures 2000; 37:1615 –1628.
37. Del Piero G. Some properties of the set of fourth-order tensors, with application to elasticity. Journal of
Elasticity 1979; 3:245 –261.
38. Halmos P. Finite-Dimensional Vector Spaces. Van Nostrand: New York, 1958.
39. Miehe C. Computation of isotropic tensor functions. Communications in Numerical Methods in Engineering
1993; 9:889–896.
40. Simo JC, Taylor RL. Quasi-incompressible finite elasticity in principal stretches. Continuum basis and
numerical algorithms. Computer Methods in Applied Mechanics and Engineering 1991; 85:273–310.
41. Carlson DE, Hoger A. The derivative of a tensor-valued function of a tensor. Quarterly of Applied Mathematics
1986; XLIV:409– 423.
42. Gurtin ME. An Introduction to Continuum Mechanics. Academic Press: New York, 1981.
43. Ball JM. Differentiability properties of symmetric and isotropic functions. Duke Mathematical Journal 1984;
51:699–728.
44. Rosati L, Valoroso N. Derivatives of isotropic tensor functions. Comptes Rendus de l’Académie des Sciences—
Série IIb—Mécanique 2002, in press.
45. Dieudonné J. Foundations of Modern Analysis. Academic Press: New York, 1960.
46. Truesdell C, Noll W. The Non-linear field theories of mechanics. In Handbuch der Physik, band III/3,
Flügge S (ed.). Springer: Berlin, 1965.
47. Rivlin RS. Further remarks on the stress-deformation relations for isotropic materials. Journal of Rational
Mechanics and Analysis 1955; 4:681–701.
48. Rosati L, Valoroso N. Evaluation of conjugate stresses to Seth’s strain tensors. Technische Mechanik 2002;
22:1–10.
49. Rosati L. A novel approach to the solution of the tensor equation AX + XA = H. International Journal of
Solids and Structures 2000; 37:3457–3477.
50. Styan GPH. Hadamard products and multivariate statistical analysis. Linear Algebra and its Applications
1973; 6:217–240.

51. Lubliner J. Plasticity Theory. MacMillan: New York, 1990.

52. Perzyna P. Fundamental problems in viscoplasticity. Advances in Applied Mechanics 1966; 9:243–377.
53. Alfano G, De Angelis F, Rosati L. General solution procedures in elasto/visco-plasticity. Computer Methods
in Applied Mechanics and Engineering 2001; 190:5123–5147.
54. Lee EH. Elastic-plastic deformation at finite strains. Journal of Applied Mechanics (ASME) 1969; 36:1– 6.
55. Mandel J. Plasticité Classique et Viscoplasticité, CISM Courses and Lectures—No. 97. Springer: Wien,
New York, 1972.
56. Marsden JE, Hughes TJR. Mathematical Foundations of Elasticity. Prentice-Hall: Englewood Cliffs, NJ, 1983.
57. Weber G, Anand L. Finite deformation constitutive equations and a time integration procedure for isotropic,
hyperelastic-viscoplastic solids. Computer Methods in Applied Mechanics and Engineering 1990; 79:173–202.
58. Macari-Pasqualino EJ, Runesson K, Sture S. Response prediction of granular materials at low effective
stresses. Journal of Geotechnical Engineering Division (ASCE) 1994; 120:1252–1268.
59. Klaas O, Maniatty A, Shepard M. A stabilized mixed finite element method for finite elasticity. Formulation
for linear displacement and pressure interpolation. Computer Methods in Applied Mechanics and Engineering
1999; 180:65 –79.
60. Taylor RL. FEAP User Manual, rel. 7.1, University of California at Berkeley, 1999.
61. Nadeau JC, Ferrari M. Invariant tensor-to-matrix mappings for evaluation of tensorial expressions. Journal
of Elasticity 1998; 52:43– 61.

The Good Earth Study Guide
100% (1)
The Good Earth Study Guide
4 pages
Modeling The Hysteretic Response of Mechanical Connections For Wood Structures
No ratings yet
Modeling The Hysteretic Response of Mechanical Connections For Wood Structures
11 pages
A Finite Element Formulation For Nonlinear Incompressible Elastic and Inelastic Analysis
No ratings yet
A Finite Element Formulation For Nonlinear Incompressible Elastic and Inelastic Analysis
53 pages
Complementary Mixed Finite Element Formulations For Elastoplasticity
No ratings yet
Complementary Mixed Finite Element Formulations For Elastoplasticity
30 pages
Dimas2015 PDF
No ratings yet
Dimas2015 PDF
10 pages
Finite Element Implementation of An Orthotropic PL
No ratings yet
Finite Element Implementation of An Orthotropic PL
29 pages
Numerical Analysis of Dynamic Behaviour of Elastoplastic Beams
No ratings yet
Numerical Analysis of Dynamic Behaviour of Elastoplastic Beams
19 pages
2321-Article Text-10181-1-10-20110622
No ratings yet
2321-Article Text-10181-1-10-20110622
23 pages
Resine Epoxy
No ratings yet
Resine Epoxy
16 pages
Mean Square Design
No ratings yet
Mean Square Design
12 pages
Manuscript
No ratings yet
Manuscript
20 pages
3-A O - Vatulyan,-S A - Nesterov
No ratings yet
3-A O - Vatulyan,-S A - Nesterov
15 pages
Mixed FEM For Elastoplasticity
No ratings yet
Mixed FEM For Elastoplasticity
30 pages
Rheological Modelling of Complex Fluids. I. The Concept of Effective Volume Fraction Revisited
No ratings yet
Rheological Modelling of Complex Fluids. I. The Concept of Effective Volume Fraction Revisited
9 pages
Composite Materials Abaqus Analytical Solutions Eshelby 13
No ratings yet
Composite Materials Abaqus Analytical Solutions Eshelby 13
12 pages
FPFComposite JPCSA4 R1
No ratings yet
FPFComposite JPCSA4 R1
9 pages
Constitutive Modeling and Computer PDF
No ratings yet
Constitutive Modeling and Computer PDF
25 pages
Effective Medium Theory For Dispersions of Dielectric Ellipsoids
No ratings yet
Effective Medium Theory For Dispersions of Dielectric Ellipsoids
18 pages
Kausel
No ratings yet
Kausel
16 pages
Elastic-Plastic Large Deformation Analysis Using SPH
No ratings yet
Elastic-Plastic Large Deformation Analysis Using SPH
5 pages
art3A10.10072Fs11012-015-0185-1
No ratings yet
art3A10.10072Fs11012-015-0185-1
19 pages
Modeling of Large Deformations of Hyperelastic Materials
No ratings yet
Modeling of Large Deformations of Hyperelastic Materials
4 pages
Lecture 12
No ratings yet
Lecture 12
19 pages
RodriguesBurgosMartha-MethodsX-2021
No ratings yet
RodriguesBurgosMartha-MethodsX-2021
37 pages
024 - An Exact Modal Analysis Approach To Vibration
No ratings yet
024 - An Exact Modal Analysis Approach To Vibration
51 pages
Statistics of ideal randomly branched polymers in a semi-space
No ratings yet
Statistics of ideal randomly branched polymers in a semi-space
11 pages
PEER Stage2 10.1088/0022-3727/43/45/455406
No ratings yet
PEER Stage2 10.1088/0022-3727/43/45/455406
23 pages
Preprint - Accelerated Vibro-Acoustics of Porous Domains Via A Novel Multiscale Finite Element Method
No ratings yet
Preprint - Accelerated Vibro-Acoustics of Porous Domains Via A Novel Multiscale Finite Element Method
15 pages
Implementation of The Strain Gradient Plasticity With Linear Finite Elements and Reduced Integration
No ratings yet
Implementation of The Strain Gradient Plasticity With Linear Finite Elements and Reduced Integration
2 pages
2003 Europm Elastoplasticsimulation PDF
No ratings yet
2003 Europm Elastoplasticsimulation PDF
7 pages
3D Cylindrical Finite Element Model For Thick Curved
No ratings yet
3D Cylindrical Finite Element Model For Thick Curved
21 pages
Eshelby Tensors and Overall Properties of Nano-Composites Considering Both Interface Stretching and Bending Effects
No ratings yet
Eshelby Tensors and Overall Properties of Nano-Composites Considering Both Interface Stretching and Bending Effects
34 pages
Evaluation and Verification of An HSDT-L PDF
No ratings yet
Evaluation and Verification of An HSDT-L PDF
60 pages
Khazana Mil Gaya
No ratings yet
Khazana Mil Gaya
45 pages
Practical Aspects of Finite Element Simulation
No ratings yet
Practical Aspects of Finite Element Simulation
11 pages
Ortiz 1985
No ratings yet
Ortiz 1985
16 pages
Finite-Element Formulations For Problems of Large Elastic-Plastic Deformation
0% (1)
Finite-Element Formulations For Problems of Large Elastic-Plastic Deformation
16 pages
1304 5516
No ratings yet
1304 5516
22 pages
Use of Logarithmic Strains To Evaluate "True" Cauchy Stresses in Finite Deformation Problems
No ratings yet
Use of Logarithmic Strains To Evaluate "True" Cauchy Stresses in Finite Deformation Problems
14 pages
Finite Deformation Continuum Model For Single-Walled Carbon Nanotubes
No ratings yet
Finite Deformation Continuum Model For Single-Walled Carbon Nanotubes
9 pages
Finite Element Analysis For Setting Shrinkage Stress of Recycled-PET Polymer Concrete
No ratings yet
Finite Element Analysis For Setting Shrinkage Stress of Recycled-PET Polymer Concrete
8 pages
Pstruct 2021 113645
No ratings yet
Pstruct 2021 113645
8 pages
Solid Effects On Solid Objects PDF
No ratings yet
Solid Effects On Solid Objects PDF
11 pages
Elastic-Plastic Stress Distribution in A Plastically Anisotropic Rotating Disk
No ratings yet
Elastic-Plastic Stress Distribution in A Plastically Anisotropic Rotating Disk
8 pages
Pro076 010
No ratings yet
Pro076 010
9 pages
Nuka La 2006
No ratings yet
Nuka La 2006
31 pages
Plastic Damage Model For Progressive Failure Analysis of Composite Structures
No ratings yet
Plastic Damage Model For Progressive Failure Analysis of Composite Structures
6 pages
Bursi1997 Benchmark
No ratings yet
Bursi1997 Benchmark
26 pages
Elastic-Plastic Analysis of R/C Coupled Shear Walls: The Equivalent Stiffness Ratio of The Tie Elements
No ratings yet
Elastic-Plastic Analysis of R/C Coupled Shear Walls: The Equivalent Stiffness Ratio of The Tie Elements
8 pages
Investigation of Pencil Lead Break
No ratings yet
Investigation of Pencil Lead Break
13 pages
1981 Tvergaard
No ratings yet
1981 Tvergaard
19 pages
Shafiqu Jeas 0608 104
No ratings yet
Shafiqu Jeas 0608 104
7 pages
A Generalized Differential Quadrature Solution For Laminated Composite Shells of Revolution
No ratings yet
A Generalized Differential Quadrature Solution For Laminated Composite Shells of Revolution
2 pages
High-performance-unsymmetric-8-node-hexahedral-el_2023_International-Journal
No ratings yet
High-performance-unsymmetric-8-node-hexahedral-el_2023_International-Journal
16 pages
On The Formulation of Closest-Point Projection Algorithms in Elastoplasticity - Part I
No ratings yet
On The Formulation of Closest-Point Projection Algorithms in Elastoplasticity - Part I
38 pages
Radial Return Method Applied in Thick-Walled Cylinder Analysis
No ratings yet
Radial Return Method Applied in Thick-Walled Cylinder Analysis
15 pages
Modeling of Lamb Waves in Composites Using New Third-Order Plate Theories
No ratings yet
Modeling of Lamb Waves in Composites Using New Third-Order Plate Theories
15 pages
Return Mapping Algorithms and Consistent Tangent Operators in Ferroelectroelasticity
No ratings yet
Return Mapping Algorithms and Consistent Tangent Operators in Ferroelectroelasticity
43 pages
Analytical Fracture Mechanics
From Everand
Analytical Fracture Mechanics
David J. Unger
No ratings yet
Continuum Mechanics
From Everand
Continuum Mechanics
A. J. M. Spencer
3/5 (1)
Fracture Mechanics and Crack Growth
From Everand
Fracture Mechanics and Crack Growth
Naman Récho
No ratings yet
A Desertion: by Stephen Crane
No ratings yet
A Desertion: by Stephen Crane
4 pages
Scenario #1: IP Allocation in A MAN: Bahria University (Karachi Campus)
100% (1)
Scenario #1: IP Allocation in A MAN: Bahria University (Karachi Campus)
6 pages
ENGLISH 2ND PAPER 4TH SEMESTER
No ratings yet
ENGLISH 2ND PAPER 4TH SEMESTER
10 pages
Idioms
No ratings yet
Idioms
8 pages
Australia Corporate Annex Careers & Latest Job Vacancy List
No ratings yet
Australia Corporate Annex Careers & Latest Job Vacancy List
7 pages
SK Utilization Report
100% (8)
SK Utilization Report
2 pages
SBI Contra Fund Factsheet September 2024
No ratings yet
SBI Contra Fund Factsheet September 2024
1 page
Modern School Magazine
No ratings yet
Modern School Magazine
8 pages
Ethernet Cables - RJ45 - Colors & Crossover
No ratings yet
Ethernet Cables - RJ45 - Colors & Crossover
2 pages
Waring Blender
No ratings yet
Waring Blender
2 pages
Introduction To Psychology PPT and Report Saniya Lanjekar
No ratings yet
Introduction To Psychology PPT and Report Saniya Lanjekar
16 pages
Harmonics power systems and smart grids Second Edition De La Rosa - Get the ebook in PDF format for a complete experience
No ratings yet
Harmonics power systems and smart grids Second Edition De La Rosa - Get the ebook in PDF format for a complete experience
79 pages
REFERENCE LIST - Purdue Owl
100% (1)
REFERENCE LIST - Purdue Owl
2 pages
Fordyce MW 1983
No ratings yet
Fordyce MW 1983
46 pages
BT619A 2 Assignmentt
No ratings yet
BT619A 2 Assignmentt
4 pages
Assignment 3
No ratings yet
Assignment 3
9 pages
Typescript Short Tutorial
No ratings yet
Typescript Short Tutorial
7 pages
Portal Frames
No ratings yet
Portal Frames
76 pages
CBE Online Application System (COAS)
No ratings yet
CBE Online Application System (COAS)
1 page
Samson Fekadu PDF
No ratings yet
Samson Fekadu PDF
92 pages
Literary Analysis - Catch-22
No ratings yet
Literary Analysis - Catch-22
12 pages
Stavanger, Norway: Finn Tengs Christensen
No ratings yet
Stavanger, Norway: Finn Tengs Christensen
6 pages
Daily Checklist: 120/80 MMHG 36.5 67 22 1500 CC 1400 CC
No ratings yet
Daily Checklist: 120/80 MMHG 36.5 67 22 1500 CC 1400 CC
53 pages
The Characteristics of Believers
No ratings yet
The Characteristics of Believers
7 pages
[Lyric Translation] G.E.M. 鄧紫棋 - City Zoo 摩天動物園 - Asian Pop Weekly
No ratings yet
[Lyric Translation] G.E.M. 鄧紫棋 - City Zoo 摩天動物園 - Asian Pop Weekly
7 pages
A Phenomenology of Tourist Experiences
100% (1)
A Phenomenology of Tourist Experiences
24 pages
LG Commission-Working-1
No ratings yet
LG Commission-Working-1
27 pages
Datasheet Flatpack2 483000 HE
No ratings yet
Datasheet Flatpack2 483000 HE
2 pages
RESEARCH
No ratings yet
RESEARCH
17 pages

A Return Map Algorithm For General Isotr

Uploaded by

A Return Map Algorithm For General Isotr

Uploaded by

INTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING

Int. J. Numer. Meth. Engng 2004; 60:461–498 (DOI: 10.1002/nme.970)

A return map algorithm for general isotropic elasto/visco-plastic

Luciano Rosati1,∗,† and Nunziante Valoroso2

Computational plasticity and viscoplasticity have been characterized by signiﬁcant advancements

Contract/grant sponsor: Italian National Research Council (CNR)

2. ELEMENTS OF TENSOR ALGEBRA

We provide a brief review of the basic properties of linear transformations (tensors) on a

with the superscript T standing for transpose.

(A ⊠ B)(C ⊗ D) = (ACBT ) ⊗ D (5)

(A ⊗ B)(C ⊠ D) = A ⊗ (CT BD)

(A ⊗ B)(c ⊗ d) = [B · (c ⊗ d)]A = (BT c · d)A (6)

(A ⊠ B)(c ⊗ d) = A(c ⊗ d)BT = Ac ⊗ Bd

2.1. Eigenprojectors of second-order symmetric tensors

The directional derivative of (15) in the direction H ∈ Lin yields

dA âi · H = H · (ai ⊗ ai ) + A · [(dA ai )H ⊗ ai ] + A · [ai ⊗ (dA ai )H] (16)

by applying the product rule [42].

so that, recalling (1)2 , we obtain

A · [(dA ai )H ⊗ ai ] = tr {A[(dA ai )H ⊗ ai ]} = tr {[(dA ai )H ⊗ ai ]A}

= tr [(dA ai )H ⊗ Aai ] = (dA ai )H · âi ai = 0

A basic property concerning the tensor products of eigenprojectors is given by

(âi − âk )(Cj − Ci )dA Ai = 2(âi − âk ) sym(Ai ⊠ Aj + Ai ⊠ Ak )

+ 2(âj − âk ) sym(Ai ⊠ Aj + Aj ⊠ Ak )

+ 2Cj sym(Ai ⊠ Aj + Aj ⊠ Ak + Ak ⊠ Ai ) (23)

− (2âi − âj − âk )Ai ⊠ Ai − (âj − âk )(Aj ⊠ Aj + Ak ⊠ Ak )], i = j, k

2.2. Some properties of isotropic scalar functions of a symmetric tensor argument

3. REPRESENTATION OF SYMMETRIC FOURTH-ORDER TENSORS

Let us consider a symmetric fourth-order tensor A expressed as linear combination of tensor

⊗ ⊠ are isotropic scalar functions.

⊗ ⊠ depend upon p ⊗ and p ⊠ and the invariants of A.

+ A1 ⊠A2 + A2 ⊠A1 + A2 ⊠A3 + A3 ⊠A2 + A3 ⊠A1 + A1 ⊠A3 (33)

on account of Lemma 2. Hence,

A further example is represented by the second derivative of an isotropic scalar function

and observing that the following composition rules:

hence, setting [b] = [b11 , b22 , b33 ]T , one has ﬁnally

3.1. Inversion of symmetric fourth-order tensors

A−1 = aI + b(A⊠1 + 1⊠A) + c(A2 ⊠1 + 1⊠A2 )

+ d(1 ⊗ 1) + e(A ⊗ 1 + 1 ⊗ A) + f (A2 ⊗ 1 + 1 ⊗ A2 )

+ g(A ⊗ A) + h(A2 ⊗ A + A ⊗ A2 ) + i(A2 ⊗ A2 )

Using the spectral decomposition (7) the principal space representation is

[A⊗ ][C⊗ ] · [A⊗ ] + ([A⊠ ] ∗ [C ⊠ ]) · [A⊠ ] = [1⊗ ] · [A⊗ ] + [1⊠ ] · [A⊠ ]

from which we infer

[A⊗ ][C⊗ ] = [1⊗ ] [A⊠ ] ∗ [C ⊠ ] = [1⊠ ]

3.2. Composition of fourth-order tensors with second-order tensors

B = b1 A1 + b2 A2 + b3 A3 = [b] · [EA ] (45)

A = [A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ]

If B ∈ Sym is coaxial with A, it turns out to be

AB = aij⊗ bj Ai = ([A⊗ ][b]) · [EA ]

where [b] is the vector of the principal values bj of B.

AB = ([A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ])bk Ak = aij⊗ A⊗ ⊠ ⊠

On the other hand, it turns out to be

AB = aij⊗ bj Ai = ([A⊗ ][b]) · [EA ] (47)

The previous result allows us to establish the following:

A = [A⊗ ] · [A⊗ ] + [A⊠ ] · [A⊠ ]

and B ∈ Sym coaxial with A, it turns out to be

AB · B = aij⊗ bj bi = ([A⊗ ][b]) · [b]

4.1. Continuum formulation. Small deformations

= el (e) + h () (49)

= de el (e); q = d h () (50)

(, q) = (I1 , J2 , J3 ) − q − Y0 (51)

are invariants of and S = − 31 tr()1 is the stress deviator.

ė = ˙ − ˙ d (, q); ˙ = −˙dq (, q) = ˙ (52)

(, q) 0, ˙ 0, ˙ (, q) = 0 (54)

4.2. Discrete formulation and return map algorithm. Small deformations

ė = ˙ ; ˙ = 0 elastic part (56)

ė = −˙d (, q); ˙ = ˙ plastic part (57)

Equations (56) along with the initial conditions

4.3. The tensor G and its inverse

see (10), the elastic tangent is amenable to the following representation:

on account of (32). By virtue of Lemma 6 it turns out to be

where the entries d̃ij⊗ of [D̃⊗ ] and d̃ij⊠ (i = j ) of [D̃ ⊠ ] are

+ e6 (S ⊗ S) + e7 (S2 ⊗ S + S ⊗ S2 ) + e8 (S2 ⊗ S2 ) (71)

where the coefﬁcients are given by

e1 = d2  = n2 , e2 = d3  = n3

e3 = (− 31 n2 + d11  − 43 J2 d13  + 49 J22 d33 )

e4 = (− 23 n3 + d12  − 23 J2 d23 ), e5 = (d13  − 23 J2 d33 )

− (2âi − âj − âk )Ai ⊠ Ai − (âj − âk )(Aj ⊠ Aj + Ak ⊠ Ak )], i = j, k

= el (e) + h () (49)

= de el (e); q = d h () (50)

(, q) = (I1 , J2 , J3 ) − q − Y0 (51)

ė = ˙ − ˙ d (, q); ˙ = −˙dq (, q) = ˙ (52)

(, q) 0, ˙ 0, ˙ (, q) = 0 (54)

ė = ˙ ; ˙ = 0 elastic part (56)

ė = −˙d (, q); ˙ = ˙ plastic part (57)

where the entries d̃ij⊗ of [D̃⊗ ] and d̃ij⊠ (i = j ) of [D̃ ⊠ ] are

e1 = d2 = n2 , e2 = d3 = n3

e3 = (− 31 n2 + d11 − 43 J2 d13 + 49 J22 d33 )

e4 = (− 23 n3 + d12 − 23 J2 d23 ), e5 = (d13 − 23 J2 d33 )

e6 = d22 , e7 = d23 , e8 = d33

fij⊠ = e1 + e2 (ŝi + ŝj ) i = j

gij⊗ = d̃ij⊗ + fij⊗ , i = j (76)

gij⊠ = d̃ij⊠ + fij⊠ , i = j

1. Initialize: k = 0; ê(k) = etr ; (k) = tr ; (k) = 0; set tolerance tol.

b˙e = lbe + be lT ; ˙ = 0 elastic part (90)

b˙e = −[2˙d (, q)]be ; ˙ = ˙ plastic part (91)

i.e. the pull-back of the Kirchhoff stress to the intermediate conﬁguration at t = tn .

Ce,tr = 2 (Fe,tr ⊠ Fe,tr )T d = 2 sym[(Fe,tr )T Fe,tr ] (108)

i.e. the dyadic part of the derivative detr .

whose solution with respect to dêtr ˆ yields

[G⊗ ]−1 dˆ ˆ