ORIGINAL PAPER

S. Klinkel C. Sansour W. Wagner

An anisotropic bre-matrix material model at nite
elastic-plastic strains
Received: 1 June 2004 / Accepted: 4 October 2004 / Published online: 10 December 2004
Springer-Verlag 2004
Abstract In this paper a constitutive model for aniso-
tropic nite strain plasticity, which considers the major
eects of the macroscopic behaviour of matrix-bre
materials, is presented. As essential feature matrix and
bres are treated separately, which allows as many
bundles of bres as desired. The free energy function is
additively split into a part related to the matrix and in
parts corresponding to the bres. Usually the free energy
function is dened by the integrity basis of the main
variable and structural tensors, which leads to compli-
cated numerical schemes. Here, the continuum is con-
sidered as superimposed of the isotropic matrix and
further one-dimensional continua each of them repre-
sents one bundle of bres. The deformation gradient
applies to all continua introducing a constraint, which
links the dierent continua. One of the most striking
features of the model is its suitability for a numerical
treatment.
Keywords Anisotropic plasticity Finite strains
Material modelling Fibre-Matrix material
1 Introduction
Anisotropic eects in the elastic-plastic deformation
range may occur either due to oriented internal struc-
tures, see e.g. [1], or to bre matrix compositions. The
paper is a contribution to analyse bre-matrix materials
using a simple constitutive model which capture the
characteristic macroscopic behaviour of the anisotropic
material. An accurate modelling is important for the
design of materials like e.g. brous Boron-Aluminium
composites [2] or bre reinforced composites [3].
The description of anisotropic materials is very often
based on averaged quantities. That is, the anisotropic
behaviour of the material is captured in an averaged
sense. The elasticity tensor is the most profound example
of such description, where two tensorial (second order)
quantities are related through an anisotropic fourth or-
der tensor. However, very often tensor-valued relation-
ships can be derived from a scalar function (energy, ow
rule), which means that the scalar functions themselves
are to be described in an anisotropic manner, that takes
possible material symmetries into account. The mathe-
matical means for this is the notion of structural tensors,
which is a general tool for describing certain classes of
anisotropy and encompasses well established models of
elastic anisotropy and anisotropic ow rules, such as the
orthotropic yield condition early published in the clas-
sical work of Hill [4]. Structural tensors are dyadic
products of the privileged directions of the material.
Essentially, the anisotropic function under consideration
(internal energy, ow rule) is formulated in terms of the
so-called integrity basis, which consists of invariants of
the main variable (strain or stress) together with invar-
iants of the tensor products of the main variable with the
structural tensors. The resulting function is supposed to
describe the average behaviour of the system, irrespec-
tive of the physical nature of anisotropy itself which may
be a result of the crystal structure of the material or a
result of the fact that the material is composed of dif-
ferent components such as a matrix and bres. Examples
are provided in [5] for the case of elasticity, and in [6
10], for the formulation of anisotropic yield criteria, to
mention few.
Recently a considerable eort has been devoted into
the extension of isotropic nite strain models of inelas-
ticity to the anisotropic case. In this regard a clear dis-
tinction can be drawn between models based on additive
decompositions of suitable strain measures as in [11],
[12], or [13], and those which are based on the multi-
plicative decomposition of the deformation gradient into
elastic and inelastic parts such as in [1410], or [17]. An
Comput Mech (2005) 35: 409417
DOI 10.1007/s00466-004-0629-2
S. Klinkel (&) W. Wagner
Institut f. Baustatik, Universita t Karlsruhe (TH),
Kaiserstr. 12, Karlsruhe, 76131,
Germany
E-mail: sk@bs.uni-karlsruhe.de, ww@bs.uni-karlsruhe.de
C. Sansour
School of Petroleum Engineering,
University of Adelaide, SA 5005,
Australia
E-mail: carlo.sansour@adelaide.edu.au
important issue in relation to these dierent models is
the eort to be considered with regard to numerical
implementations. In fact, anisotropic schemes based on
averaged description are very involved. Of the above
mentioned, those based on the multiplicative decompo-
sition lead to especially involved and tedious schemes
and the numerical eort to obtain solutions can be very
considerable.
This paper is concerned with a simple constitutive
model for bre reinforced material at nite elastic-plas-
tic strains. The material is assumed to consist out of a
matrix and one or more bre bundles. The model is
characterized by its conceptual simplicity as well as its
numerical eciency. The numerical eort essentially can
be compared with that known from purely isotropic
computations. The main idea is to consider the contin-
uum as superimposed of the matrix and further one-
dimensional continua each of them represent one family
of bres. The bres are assumed to be equally distrib-
uted over the body. The dierent continua are linked by
the kinematic constraint that the deformation gradient
applies to all of them. The elastic-plastic model is based
on the multiplicative decomposition of the deformation
gradient into an elastic and plastic part, for both the
matrix and the bre. The free energy function is addi-
tively decomposed into an isotropic part for the matrix
and into parts for each bre bundle, which induce the
anisotropic behaviour.
The greatest advantage of such formulation, as
compared to an averaged description, lies in the pos-
sibility of treating each constituent separately. While
the matrix can be considered isotropic and well
established schemes can be applied, a specic formu-
lation must be developed for the one-dimensional
continua which are supposed to capture the behaviour
of the bres. In fact, it turns out that a simple for-
mulation can be found rendering the whole approach
extremely ecient. The essential features and novel
aspects of the presented formulation are summarized
as follows:
(i) The additive split of the free energy function leads
to separated constitutive equations for the matrix
and the bres. For the matrix we use an isotropic
elastic-plastic model for large strains, which was
proposed in [18]. In contrast to [18] the plastic
constitutive model is implemented in a Lagrangian
setting, for details see [19], [20]. It is noted that the
Eulerian model used in [18] and the Lagrangian
model are physically equivalent.
(ii) For the bre a one-dimensional material model is
introduced. The governing equations are derived
with the principle of maximum dissipation and the
yield condition as a constraint.
(iii) The introduced constitutive model is able to con-
sider various bres with arbitrary directions. A
formulation of an anisotropic yield condition,
which captures all material symmetries, is not
necessary.
(iv) An implicit integration method yields a scalar-
valued return mapping algorithm for each bre,
which, along with a consistent tangent modulus, is
very ecient for numerical implementation.
(v) To obtain quadratic convergence in the frame of
the Newton-Raphson method, the consistent elas-
tic-plastic tangent modulus for the bre is derived.
The tangent modulus for the matrix material may
be found in [21].
The paper is organized as follows: In Sect. 2 the aniso-
tropic constitutive model is proposed. First the kine-
matics of multiplicative elastoplasticity are summarized.
The free energy function is introduced. On this basis the
elastic and plastic constitutive models are derived along
standard arguments in rational continuum mechanics.
In Sect. 3 an implicit integration algorithm for the rate
independent plasticity formulation is presented. On this
framework the algorithmic elastic-plastic tangent mod-
ulus is derived. Finally, numerical examples in Section 4
demonstrate the anisotropic behaviour of the proposed
model.
2 Anisotropic constitutive model
2.1. Kinematics
Let the deformation /(X; t) be a point map form the
body B with X B to the actual conguration B
t
at the
time t. The tangent to / is denoted the deformation
gradient F(X) : TB TB
t
. If X and x are the co-ordi-
nates of the body B and the actual conguration B
t
, the
deformation gradient is given by
F = Grad
X
(x) : (1)
A fundamental assumption is now that the above kine-
matics applies to a body which exhibits more then one
constituent. The constituents dier not only with respect
to their material behaviour but in their geometry as well.
In our specic case, we assume that one constituent will
always be an isotropic classical continuum. Each of the
other constituents will be dened as a continuum which
deforms and carries loads only in one specic direction,
the direction of a corresponding family of bres. Of such
continua there will be as much as there are families of
bres.
Let us denote a typical direction of a bre at the
reference conguration by V
F
, where the vector is
considered to be normalized. Let v
F
= FV
F
be the map
of V
F
at the actual conguration. The fact that the
behaviour of the bre continuum will be one-dimen-
sional necessitates the introduction of a one-dimensional
bre-related deformation gradient as
F
F
= v
F
V
F
: (2)
By introducing a structural tensor M
F
according to
M
F
:= V
F
V
F
, Eq. (2) can be rewritten as
F
F
= FM
F
: (3)
410
From the last relation it becomes obvious that F
F
is
nothing but the projection of F in the direction of the
bre (note that F is a two-point tensor).
The approach to inelasticity starts with the multipli-
cative decomposition of F into an elastic part F
e
and a
plastic part F
p
F = F
e
M
F
p
M
; (4)
which may be motivated by a micro-mechanical view of
plastic deformations. As the decomposition is assumed
to apply in full for the matrix, the index
M
is added.
Likewise the one-dimensional bre-related deformation
gradient is assumed to undergo a similar decomposition
according to
F
F
= F
e
F
F
p
F
: (5)
The well known multiplicative decomposition implies a
stress free intermediate conguration for the matrix and
for each bre.
The following quantities of the right CauchyGreen
type can be dened:
C = F
T
F; C
e
M
= F
e
M
T
F
e
M
; C
p
M
= F
p
M
T
F
p
M
: (6)
Likewise similar denitions apply for the bre-contin-
uum:
C
F
= F
T
F
F
F
; C
e
F
= F
e
F
T
F
e
F
; C
p
F
= F
p
F
T
F
p
F
: (7)
By denition, the last quantities are one-dimensional.
Note also that the rotational part of the motion is
included in the elastic part of the deformation gradient.
This is general praxis in isotropic nite strain inelasticity
and is followed here as well.
The quantity C
e
M
denes the elastic deformation
tensor related to the matrix, which will enter the de-
nition of a stored energy function. To identify a similar
quantity for the bre-continuum we rst note that F
p
F
must be of the one-dimensional form
F
p
F
= k
p
M
F
; (8)
with the initial condition k
p
(t = 0) = 1, where t is a time-
like parameter and t = 0 relates to a reference state. The
relation immediately gives
C
p
F
= k
p2
M
F
: (9)
With (5), (7)
2
and (8), the bre-related right Cauchy
Green tensor reads
C
F
= k
p2
M
F
F
e
F
T
F
e
F
M
F
: (10)
Noting that M
F
is idempotent, we conclude that F
e
F
M
n
F
denes, for an arbitrary power n, an equivalent class of
elastic parts of the deformation gradient. We consider a
representative for the class by dening
~
F
e
F
= F
e
F
M
F
: (11)
Correspondingly,
~
C
e
F
=
~
F
e
T
F
~
F
e
F
=
1
k
p2
M
F
CM
F
; (12)
denes a representative elastic strain measure. The last
relation follows from (10) and the observation that
C
F
= M
F
CM
F
. In what follows and for the sake of
brevity, we refrain from using and refer to C
e
F
and F
e
F
as
representatives of the equivalent class.
Finally we consider the time derivatives of the cor-
responding decompositions. First, the time derivative of
the deformation gradient reads
_
F = LF = FL ; (13)
where l and L denote the left and the right rate. For the
plastic deformation we dene a right rate according to
_
F
p
M
= F
p
M
L
p
M
;
_
F
p
F
= F
p
F
L
p
F
; (14)
which is the suitable choice as our formulation is a fully
material one.
2.2. Free energy function and dissipation
The free energy function is introduced by
W = W
M
(C
e
M
; Z
M
)
X
nF
F =1
W
F
(C
e
F
; Z
F
) ; (15)
where nF is the total number of bres. The quantities
Z
M
, Z
F
are the internal plastic variables for the matrix
and a bre. The localized form of the dissipation
inequality for an isothermal process reads
T = N : L q
0
M
_
W
M

X
nF
F =1
q
0
F
_
W
F
_ 0 : (16)
Here q
0M
and q
0F
are the densities of the matrix and
bre materials in the reference conguration and N is
the Mandel stress tensor, which is related to the Kirch-
ho stress s and the 2
nd
-Piola-Kirchho stress S by
N = F
T
sF
T
= CS, see e.g. Sansour and Kollmann [22].
The rates of the free energy may be derived as follows:
_
C
e
M
= F
p
M
T
L
T
CF
p
M
1
F
p
M
T
CLF
p
M
1
F
p
M
T
L
p
M
T
CF
p
M
1
F
p
M
T
CL
p
M
F
p
M
1
(17)
and
_
C
e
F
=
1
k
p2
M
F
L
T
CM
F

1
k
p2
M
F
CLM
F
2
_
k
p
k
p3
M
F
CM
F
:
(18)
Considering the denitions
N
M
:= 2q
0
M
CF
p
1
M
oW
M
oC
e
M
F
p
T
M
;
N
F
:= 2q
0
F
1
k
p2
CM
F
oW
F
oC
e
F
M
F
(19)
Y
M
:= q
0
M
oW
M
oZ
M
; Y
F
:= q
0
F
oW
F
oZ
F
(20)
in Eq. (16), the dissipation inequality reduces to
411
T = N N
M

X
nF
F =1
N
F
!
: L N
M
: L
p
M
Y
M
_
Z
M

X
nF
F =1
N
F
: M
F
_
k
p
k
p
Y
F
_
Z
F
!
_ 0: (21)
Since inequality (21) must hold for arbitrary processes in
the material, standard arguments in rational thermody-
namics yield the equations
N = N
M

X
nF
F =1
N
F
(22)
T = N
M
: L
p
M
Y
M
_
Z
M
N
F
: M
F
_
k
p
k
p
Y
F
_
Z
F
_ 0 : (23)
Now, we impose the more restrictive assumption that
contributions to the dissipation of each continuum must
be positive on its own right. That is, we postulate the
validity of the relations
T
M
= N
M
: L
p
M
Y
M
_
Z
M
_ 0 (24)
T
F
= N
F
: M
F
_
k
p
k
p
Y
F
_
Z
F
_ 0 : (25)
For the matrix material we assume isotropic behaviour.
Therefore Eq. (24) may be rewritten as T
M
= s
M
:
1
2
L
(b
e
)b
e
Y
M
_
Z
M
_ 0, where b
e
M
is the elastic left Cauchy-
Green tensor of the matrix and L(b
e
M
) is the Lie time
derivative. This equation is identical to the formulation
in Simo [18].
Remark. Naturally, the volume of bres present in the
body must be considered to generate the physically
correct response. This can be done either by explicitly
including a factor which describes the volume fraction or
by generating the material constants through a homog-
enization process. Such a process automatically consid-
ers the density of the brous material and its volume
fraction. For instance the E-modulus of the bre con-
tinuum will be determined by considering the E-modulus
of the single bre and the fraction of the surface of the
brous material per unit surface. We have chosen the
second approach. That is, it is assumed that all material
parameters are generated such that the volume fraction
(or surface fraction) of the real brous material per unit
volume (or per unit surface) is considered.
2.3 Elastic constitutive model
For the matrix material we assume that the elastic strain
energy is a quadratic function of the logarithmic elastic
principal stretches. The Youngs modulus and the
Poisson ratio are denoted by E
M
and m
M
, respectively.
To introduce the elastic constitutive model for the
bre, Eq. (19) is rewritten as S
F
= C
1
N
F
=
(2q
0F
1
k
p2
oW
F
oC
e
F
: M
F
)M
F
. With the denition
r
F
:= E
F
1
2
tr(ln[C
e
F
[)
= E
F
1
2
(tr(ln[M
F
CM
F
[) 2ln[k
p
[) ; (26)
where E
F
denotes Youngs modulus of the bre-contin-
uum, the 2
nd
-Piola-Kirchho stress tensor reads
S
F
= r
F
M
F
: (27)
2.4. Plastic constitutive model
The plastic constitutive model is determined by the yield
condition, which denes the elastic domain. For the
matrix material we consider an isotropic yield criteria of
the von Mises type
f
M
:=

3
2
s
d
M
: s
d
M
r
(Y
M0
Y
M
) _0
with Y
M
:=h
M
Z
M
(Y
M
Y
M0
)(1exp[g
M
Z
M
[) :
(28)
Here s
d
M
means the deviatoric part of the Kirchho
stress of the matrix and h
M
, g
M
, Y
M
, Y
M0
are plastic
parameters. Alternatively, a physically equivalent for-
mulation can be given in terms of the material stress
tensor N
M
as follows:
f
M
:=

3
2
tr( N
d
M
)
2
r
(Y
M0
Y
M
) _0
with Y
M
:=h
M
Z
M
(Y
M
Y
M0
)(1exp[g
M
Z
M
[) :
(29)
The evolution equations of the plastic variables are
_
C
p
M
= 2^ cC
p
F
1
of
M
o s
M
F

= 2^ cC
p
of
M
o N
M

_
Z
M
= ^ c
of
M
oZ
M
:
(30)
where ^ c denotes the Lagrangian multiplier. For details
see [18], [20].
For the bre we introduce the yield criterion
f
F
:=

tr(M
F
N
F
M
F
)
2
q
(Y
F 0
Y
F
) _ 0
and Y
F
:= h
F
Z
F
(Y
F
Y
F 0
)(1 exp[g
F
Y
F
[) ;
(31)
with the plastic parameters h
F
, g
F
, Y
F
, Y
F 0
. The prin-
ciple of the maximum plastic dissipation along with the
fulllment of the yield condition results in the optimi-
zation problem
T
F
( N
F
; Y
F
) ^ cf
F
( N
F
; Y
F
) _ 0 : (32)
A partial derivation to the unknown variables N
F
, Y
F
yields the evolution equations
_
k
p
M
F
= k
p
^ c
of
F
o N
F
_
Z
F
= ^ c
of
F
oY
F
; (33)
412
along with the loading and unloading conditions f
F
_ 0,
^ c _ 0 and ^ cf
F
= 0. Taking into account that

tr(M
F
N
F
M
F
)
2
q
=

r
2
F
p
tr(M
F
CM
F
) =

r
2
F
p
tr(CM
F
)
one obtains the evolution equations as
_
k
p
M
F
= k
p
^ csign(r
F
)M
F
_
Z
F
= ^ c :
(34)
3 Implicit integration algorithm
For the isotropic matrix material we apply an implicit
exponential integration algorithm to integrate the plastic
strains (30), see Simo [18]. The plastic parameter is de-
ned by c = (t
n1
t
n
)^ c and the increment of the plastic
stretch is given as Dk
p
= (t
n1
t
n
)
_
k
p
, where [t
n1
; t
n
[ is a
typical time step. An implicit integration of the evolu-
tion equations of the bre (34) yields
Dk
p
= k
p
n1
c sign(r
F
)
Z
F n1
= Z
F n
c :
(35)
Considering k
p
n1
= k
p
n
Dk
p
, these equations are
rewritten as
c =
Dk
P
k
P
n
Dk
P
sign(r
F n1
)
Z
F n1
= Z
F n

Dk
P
k
P
n
Dk
P
sign(r
F n1
)
(36)
The unknown plastic stretch Dk
P
is iteratively deter-
mined by fulllment of the yield condition (31) using a
Newton iteration scheme
f
k1
F
n1
= f
k
F n1

of
k
F n1
oDk
P
DDk
P
~ 0, where k denotes the
iteration step. With
of
F n1
oDk
P
=
of
F n1
or
F n1
or
F n1
oDk
P

of
F n1
oY
F n1
oY
F n1
oZ
F n1
oZ
F n1
oDk
P
= sign(r
F n1
)

E
F
k
P
n
Dk
P
tr(C
n1
M
F
)
Y
F
/
n1
k
P
n
(k
P
n
Dk
P
)
2
!
(37)
and
Y
F
/
n1
=
oY
F n1
oZ
F n1
= h
F
(Y
F
Y
F 0
) g
F
exp[g
F
Z
F n1
[
(38)
the incremental update of Dk
P
k1
= Dk
P
k
DDk
P
is
given as
DDk
P
= sign(r
F n1
)

f
F
n1
E
F
k
P
n
Dk
P
tr(C
n1
M
F
) Y
F
/
n1
k
P
n
(k
P
n
Dk
P
)
2
: (39)
3.1. Algorithmic consistent tangent modulus
The algorithmic consistent tangent tensor is dened by
C
n1
:= 2
d S
n1
dC
n1
: (40)
Considering the additive split of the stresses in Eq. (22)
one obtains
C
n1
= 2
dS
Mn1
dC
n1

X
n
F
F =1
2
dS
F n1
dC
n1
= C
Mn1

X
n
F
F =1
C
F n1
: (41)
Since the numerical treatment of the isotropic case
including the derivation of the tangent modulus is well
established in the literature, see e.g. [18], we restrict
ourselves to the numerical treatment of the bre con-
tinuum which is to the best of our knowledge new. For
each bre we take into account Eq. (27) in which the
structural tensor M
F
is constant. The stress r
F n1
is a
function of the right Cauchy-Green tensor C
n1
and the
plastic parameter c
n1
. The fulllment of f
F n1
(r
F n1
;
Y
F n1
) = 0 implies that c is a function of C. Thus the
derivation
dr
F n1
dC
n1
reads
dr
F n1
dC
n1
=
or
F n1
oC
n1

or
F n1
ok
p
n1
ok
p
n1
oC
n1
: (42)
Here the unknown
oc
n1
oC
n1
may be derived by the use of
the implicit derivation of f
F n1
(C
n1
; c
n1
(C
n1
)) = 0,
which is
of
F n1
oC
n1

of
F n1
or
F n1
or
F n1
oC
n1

or
F n1
ok
p
n1
ok
p
n1
oC
n1

of
F n1
oY
F n1
oY
F n1
oZ
F n1
oZ
F n1
ok
p
n1
ok
p
n1
oC
n1
= 0 : (43)
With
of
F n1
oC
n1
=

(r
F n1
)
2
q
M
F
;
of
F n1
or
F n1
= sign(r
F n1
)tr(C
F n1
M
F
)
(44)
or
F n1
oC
n1
=
E
F
2tr(C
n1
M
F
)
M
F
;
or
F n1
ok
p
n1
= E
F
=k
p
n1
(45)
of
F n1
oY
F n1
= 1 ;
oZ
F n1
ok
P
n1
= sign(r
F n1
)
k
P
n
(k
P
n1
)
2
(46)
it follows
ok
P
n1
oC
n1
=
r
F n1

1
2
E
F
tr(C
F n1
M
F
)E
F
Y
F
/
n1
k
P
n
k
P
n1
k
P
n1
M
F
: (47)
413
Inserting the latter equation in (42), the tangent modulus
appears in the form
C
F n1
=

E
F
tr(C
F n1
M
F
)

E
F
(E
F
2r
F n1
)
tr(C
F n1
M
F
)E
F
Y
F
/
n1
k
p
n
k
p
n1
!
M
F
M
F
:
(48)
4 Numerical examples
Some numerical examples demonstrate the main char-
acteristics of the proposed model. Therefore the con-
stitutive model is implemented into a 3D hexahedral
element suggested in [20]. In the rst example an element
formulation is used, which is numerically integrated with
one integration point and stabilized with the method
proposed by Reese et al. [23]. For all other examples an
element formulation suggested in [20] is considered. The
following examples are calculated with an extended
version of the program FEAP [24] and illustrate the
anisotropic eects, which occur during large elastic-
plastic deformations.
4.1. Necking of a specimen
In the rst example the inuence of the bre direction h
on the overall behaviour of the structure is discussed.
Therefore a necking problem of a specimen with a plane
strain condition is investigated. The specimen, the geo-
metrical data and the boundary conditions are shown in
Fig. 1. The nodal displacement u along the loaded edge
is linked, a free contraction of the strip is allowed. The
specimen is modelled with 80 20 elements in plane and
one element through the thickness t = 0:1. The material
data are summarized in Table 1. Only one bre is con-
sidered in this example.
With a displacement driven computation the speci-
men is stretched up to 22:5% in longitudinal direction.
The deformed congurations with a plot of the equiva-
lent plastic strains of the bre and the matrix are shown
in Fig. 2, 3 for dierent bre directions h.
During the deformation the right edge of the speci-
men is moving upwards or downwards depending on the
bre direction. Furthermore Figs. 2, 3 show, that the
appearance of necking in the middle of the specimen is
strongly inuenced by the bre direction.
4.2. Circular blank
The next example, which was introduced in [11], dem-
onstrates the dierence between an elastoplastic bre
and an elastoplastic matrix.
A circular blank with a concentric circular hole is
deep-drawn into a cup, see Fig. 4. To simulate the
drawing process without contact elements, the inner
circular boundary is pulled uniformly inwards in a radial
direction up to a maximum displacement of 120, while
the outer edge is free. A plane strain condition is con-
sidered for the calculation. The circular blank is mod-
elled with 10 40 elements in plane and one element
through the thickness.
For the material behaviour it is distinguished between
an elastic matrix with an elastic-plastic bre material
and an elastic-plastic matrix with an elastic bre mate-
rial. The yield stress either for the bre or for the matrix
is given by 0.45. No hardening is assumed for the plastic
behaviour. The Youngs-moduli of the bre and the
matrix are identical. Two bres with the bre directions
of H = 45

degree are assumed.

In Fig. 5 the deformed congurations with the
equivalent plastic strains of the bres are depicted, the
matrix material is assumed to be elastic. Whereas in
Table 1 Material data for the matrix and the bre
matrix E
M
= 206:9, m
M
= 0:29,
Y
M0
= 0:450, Y
M
= 0:715, h
M
= 0:12924, g
M
= 16:930
bre E
F
= 206:9,
Y
F 0
= 2:25, Y
F
= 3:15, h
F
= 0:12924, g
F
= 16:930
Fig. 1 Geometry and boundary conditions of the specimen
Fig. 2 Equivalent plastic strains of the bre
Fig. 3 Equivalent plastic strains of the matrix
414
Fig. 6 the deformed congurations with the equivalent
plastic strains of the matrix are shown and the bre is
assumed to be elastic.
The equivalent plastic strains of the bre occur
mainly in the bre direction under the angle of
H = 45

. In opposite to that, the plastic strains of the

matrix arise mainly along the x and y directions. Here,
the deformed structure is characterized by the so called
earing which occurs at the free edge. The same char-
acteristic was obtained in [11], [15], where an anisotropic
yield function has been used.
4.3. Punching of a conical shell
The last example demonstrates the strong anisotropic
response of a structure due to the presence of bres. A
conical shell is loaded eccentrically at the upper outer
rim and supported at the lower outer rim. For the
geometry data and the material parameters see Fig. 7.
One quarter of the shell structure is discretized
with 8 8 elements and one element through the
thickness.
For the bre material it is distinguished between a
sti and a soft material set. The bre direction is indi-
cated with the arrows shown in Fig. 7 and is supposed to
be constant for each quarter of the shell. Only linear
hardening is taken into account for the matrix and the
bre material. The non-linear behaviour is computed
using an arclength method with displacement control,
which leads to the load deection curve shown in Fig. 8.
In the diagram the load p is plotted over the vertical
displacement w of the upper rim of the conical shell for
dierent material sets. The isotropic model consists out
of the matrix material without any bre and is denoted
by the solid line. The bre-matrix material with the soft
Fig. 4 Geometry of the circular
blank and its material data
Fig. 5 Deformed congurations
with a plot of the equivalent
plastic strains of the bre
Fig. 6 Deformed congurations
with a plot of the equivalent
plastic strains of the matrix
415
bres, which is denoted by the line with circles, shows
almost the same characteristic through the hole loading
process as the isotropic material.
The load is increased until the elastic limit load is
reached at w ~ 0:04. Hence a rolling process starts at the
top of the shell until a second stability point at w ~ 0:5 is
traced. Here a global snap through behaviour of the shell
is observed. At the local minimum w~ 1:6 the stable path
is reached and the load is increased until w = 2:1.
The sti model, displayed by the line with squares,
leads to a complete dierent characteristic of the load
deection curve. Here the local minimum at w ~ 0:5 is
not reached, which means that the rolling up and the
snap through phenomena arise in one step.
The deformed meshes with a plot of the equivalent
plastic strains are shown in Figs. 9, 10 at characteristic
points for the sti bre-matrix model. Here, the rst
plastic strains emerge when the elastic limit load is
reached. At this point one obtains that the rolling up
process starts form the upper edge. Obviously the de-
formed structure at w = 0:5 is wavily and the distribu-
tions of the equivalent plastic strains for the bre as well
as for the matrix are not rotationally symmetric. At
w = 1:21 until w = 1:6 the snap through phenomenon is
dominating the structural behaviour. The equivalent
plastic strains attain a maximum at w = 2:1. The maxi-
mumof the equivalent plastic strain of the matrix is 206%
and the plastic strain of the bres increases to a maximum
of ~ 5%. The sti bre matrix model shows strong
anisotropic eects through the hole loading process.
Fig. 7 The geometry of the con-
ical shell and its material data
Fig. 8 Load deection curve of the conical shell
Fig. 9 Deformed congurations
with a plot of the equivalent
plastic strains of the bre for
the sti bre-matrix material
model
416
5 Conclusion
In this paper a computational framework for the anal-
ysis of bre reinforced materials at nite elastic-plastic
deformations is presented. A constitutive model for
anisotropic nite strain plasticity was introduced. The
additive split of the free energy function leads to sepa-
rated constitutive equations for the matrix and for the
bres. The material behaviour of the matrix is assumed
to be isotropic. The anisotropic eect is induced by the
bres, which are described by macroscopic one-dimen-
sional material models. The governing equations were
derived with the principle of maximum dissipation along
with the yield condition as a constraint. A set of
numerical examples demonstrated the numerical e-
ciency of the whole approach.
References
1. Boehler JP. (1987) Introduction to the invariant formulation of
anisotropic constitutive equations. In: Boehler J.P. (ed),
Application of Tensor Functions in Solid Mechanics, CISM
Courses and Lectures No. 292, pages 1364, Wien/New York,
Springer-Verlag.
2. Dvorak GJ, Bahei-El-Din YA, Macheret Y, Liu CH. (1998).
An experimental study of elastic-plastic behavior of a brous
boron-aluminum composite. J. Mech. Phys. Solids, 36:665687
3. Spencer AJM. (1992). Plasticity theory for bre-reinforced
composites. J. Eng. Math., 26:107118
4. Hill R. (1950) The Mathematical Theory of Plasticity.
Clarendon Press, Oxford
5. Reese S, Raible T, Wriggers P. (2001) Finite element modelling
of orthotropic material behaviour in pneumatic membranes.
Int. J. of Sol. and Struc., 38:95259544
6. Spencer AJM. (1987) Anisotropic invariants and additional
results for invariant and tensor representations. In: J.P. Bo-
ehler, (ed), Application of Tensor Functions in Solid
Mechanics, CISM Courses and Lectures No. 292, pages 171
186, Wien/New York, Springer-Verlag.
7. Oller S, Car E, Lubliner J. (2003) Denition of a general im-
plicit orthotropic yield criterion. Comp. Meth. Appl. Mech.
Eng., 192:895912
8. Rogers TG. (1990) Yield criteria, ow rules, and hardening in
anisotropic plasticity. In: J.P. Boehler, (ed), Yielding, Damage,
and Failure of Anisotropic Solids, EGF5, London, Mechanical
Engineering Publications.
9. Steinmann P, Miehe C, Stein E. (1996) Fast transient dynamic
plane stress analysis of orthotropic hill-type solids at nite
elastoplastic strains. Int. J. of Sol. and Struc., 33:15431562
10. De Borst R, Feenstra PH. (1990) Studies in anisotropic plas-
ticity with reference to hill criterion. Int. J. Num. Meth. Eng.,
29:315336
11. Papadopoulos P, Lu J. (2001) On the fomulation and numer-
ical solution of problems in anisotropic nite plasticity. Comp.
Meth. Appl. Mech. Eng., 190:48894910
12. Schro der J, Gruttmann F, Lo blein J. (2002) A simple ortho-
tropic nite elasto-plasticity model based on generalized stress-
strain measures. Comp. Mech., 30:4864
13. Miehe C, Apel N, Lambrecht M. (2002) Anisotropic additive
plasticity in the logarithmic strain space. modular kinematic
formulation and implementation based on incremental mini-
mization principles for standard materials. Comp. Meth. Appl.
Mech. Eng., 191:53835425
14. Sansour C, Bocko J. (2003) On the numerical implications of
multiplicative inelasticity with an anisotropic elastic constitu-
tive law. Int. J. Num. Meth. Eng., 58:21312160
15. Gruttmann F, Eidel B. (2002) On the implementation of
anisotropic nite strain plasticity. In: Mang, HA Rammer-
storfer, FG Eberhardsteiner J (eds) Fifth Wolrd Congress on
Computational Mechanics, WCCM V, Vienna Austria
16. Menzel A, Steinmann P. (2003) On the spatial formulation of
anisotropic multiplicative elasto-plasticity. Comp. Meth. Appl.
Mech. Eng., 192:34313470
17. Reese S. (2003) Meso-macro modelling of bre-reinforced
rubber-like composites exhibiting large elastoplastic deforma-
tions. Int. J. of Sol. and Struc., 40:951980
18. Simo JC. (1992) Algorithms for static and dynamic multipli-
cative plasticity that preserve the classical return mapping
schemes of the innitesimal theory. Comp. Meth. Appl. Mech.
Eng., 99:61112
19. Miehe C. (1998) A constitutive frame of elastoplasticity at
large strains based on the notion of a plastic metric. Int. J. of
Sol. and Struc., 35:38593898
20. Wagner W, Klinkel S, Gruttmann F. (2002) Elastic and plastic
analysis of thin-walled structures using improved hexahedral
elements. Comput. & Struct., 80:857869
21. Simo JC Hughes TJR. (1998) Computational Inelasticity,
volume 7, Mechanics and Materials. Springer-Verlag, Inter-
disciplinary Applied Mathematics, New York
22. Sansour C, Kollmann FG(1998). Large viscoplastic deforma-
tions of shells. Theory and nite element formulation. Comp.
Mech., 21:512525
23. Reese S, Wriggers P. (2000) A stabilization technique to avoid
hourglasing in nite elasticity. Int. J. Num. Meth. Eng., 48:79
109
24. Taylor RL. Feap - manual.
http://www.ce.berkeley= ~ rlt/feap/manual.pdf, .
Fig. 10 Deformed congura-
tions with a plot of the equiva-
lent plastic strains of the matrix
for the sti bre-matrix mate-
rial model
417