A mechanical model for ber reinforced composite

materials with elasto-plastic matrix and interface debonding

Roberto Brighenti
*
Department of Civil Engineering, Environment and Architecture, University of Parma, Parco Area delle Scienze 181/A,
43100 Parma, Italy
Received 19 February 2003; received in revised form 24 October 2003; accepted 13 December 2003
Abstract
The behaviour of ber-reinforced composites is examined through the formulation of a non-linear constitutive law
obtained through a physical-based approach. The elasto-plastic macro constitutive equations for such a class of
materials, composed by a matrix phase with elasto-plastic behaviour and several ber-reinforcing phases, is obtained by
considering an imperfect bounds between the matrix and the bers, i.e. a certain amount of sliding is assumed to be
present at the matrixber interface. The bers are assumed to have only axial stiness with negligible relative matrix
ber displacements orthogonal to the bers axis. The amount of sliding has been determined through energetic con-
siderations by taking into account the evolution of the shear stress distribution along a single ber during an increasing
load history applied to the composite. The mechanical constitutive law has been implemented in a F.E. code and some
numerical simulation have been performed in order to assess the reliability of the proposed model.
2004 Elsevier B.V. All rights reserved.
Keywords: Fiber-reinforced composite (FRC); Homogenisation; Elasto-plastic materials behaviour; Debonding; Fibers sliding; Finite
elements
1. Introduction
The increasing importance of composite mate-
rials in many application elds, has determined the
necessity to describe with great accuracy their
overall mechanical behaviour. In order to obtain
suitable mechanical model for such materials,
various approaches can be used such as microme-
chanical models (physically based approach [1]),
homogenisation models (mathematically based ap-
proach [24]), etc.
From a mathematical point of view, the prob-
lem of the equilibrium of a multi-phase material
involves a sets of equations with rapidly vary-
ing elastic parameters which are related to the
mechanical properties of the individual phases of
the composite material. Such a boundary-value
problem is rather complex and an approximate
solution with averaged properties should be
sought for practical applications.
Since the material is composed by two or more
phases, the problem of phases-interaction from a
mechanical point of view should be taken into
Computational Materials Science 29 (2004) 475493
www.elsevier.com/locate/commatsci
*
Tel.: +39-0521-905-910; fax: +39-0521-905-924.
E-mail address: brigh@unipr.it (R. Brighenti).
0927-0256/$ - see front matter 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.commatsci.2003.12.006
account [5]; even if the single components behaves
in a linear elastic fashion, the possibility of an
imperfect bonds between the constituents give
raise to non-linear overall behaviour of the com-
posite material. In fact the perfect bonds between
dierent phases is seldom veried and damage
phenomena at the matrix-inclusion interfaces
heavily aects the whole behaviour of the material.
In the present paper the attention will be fo-
cused on ber reinforced composites (FRC) with
an elasto-plastic matrix mixed with a ber phase
joint together with (eventually) imperfect bonds.
Nomenclature
A
p
f
cross section area of the bers belong-
ing to the pth ber phase
b body force vector eld
C
m
x elastic tensor of the matrix
C
0
m
x tangent elastic tensor of the matrix
C
eq
homogenised elastic tensor of the com-
posite
C
0
eq
tangent homogenised elastic composite
tensor
d characteristic microscopic length
D characteristic macroscopic length
E
f
bers Young modulus
E
0
f
tangent bers Young modulus
E
m
pre-yielding stressstrain slope of the
matrix
E
t
post-yielding stressstrain slope of the
matrix
F r; k
1
; . . . ; k
n
0 yield function
G
m
shear modulus of the matrix
i unit vector parallel to the generic ber
axis
k
1
; . . . ; k
n
hardening parameters

k stiness of the bermatrix interface

2L
p
f
length of the bers belonging to the pth
ber phase
R.E.V. Reference Elementary Volume
se
m
f
sliding function such that: se
fm
t e
m
f

1 se
m
f

se
m
f
mean value of the sliding function along
a single ber
~u virtual displacement eld
V volume of the composite REV
V
m
volume of the matrix phase present in
the REV
V
f;p
volume of the pth ber phase fraction
present in the R.E.V.
x position vector
e, r strain and stress tensors
e,
_
e virtual strain and strain rate elds
se
fm
t strain jump between the ber and the
matrix (parallel to bers axis) in the
case of imperfect bond
e
f
uniaxial bers strain
e
m
f
uniaxial matrix strain measured at the
location and in the bers direction
e
m
f
matrix mean strain measured in the
bers direction
e
p;f
limit strain for the pth bermatrix
interface
_
k plastic multiplier
_ r
f
stress rate tensor at the location of the
generic ber belonging to the pth phase
v
p
x point function denoting the presence of
a ber, belonging to the pth phase, at
the location x
jx point function denoting the presence of
the matrix at the location x
l V
m
=V REV matrix volume fraction
g
p
V
p
=V REV ber volume fraction of the pth
ber phase, p 1; . . . ; q
s bermatrix interface shear stress
s
au
maximum bermatrix interface shear
stress
s
fu
friction bermatrix interface shear
stress
C C
t
[ C
u
boundary of the solid
C
t
portion of the boundary on which
tractions are prescribed
C
u
portion of the boundary on which dis-
placements are prescribed
476 R. Brighenti / Computational Materials Science 29 (2004) 475493
The incremental equilibrium equation is estab-
lished and solved for simple cases. The imperfect
bonds has been studied by considering the sliding
between the matrix and the ber, obtained using
the shear stress distribution evolution along a sin-
gle ber during an increasing load applied to the
composite. The obtained mechanical constitutive
law has been implemented in a F.E. code and nu-
merical simulation have been performed in order
to assess the reliability of the proposed model.
2. Statement of the problem
The equilibrium of a body, can be stated as
follows:
divr b 0 in B
n r t on C
t
u u
0
on C
u
1
where u, r, b and n are the displacement eld, the
stress tensor, the body forces and the unit outward
normal to the boundary, respectively, while t and
u
0
represent the prescribed traction and displace-
ments on the portion C
t
and C
u
of the boundary
(Fig. 1a) where traction and displacements are
prescribed.
The linearised straindisplacement equation is
given by the well-known relation:
e symgrad u symru

1
2
grad u grad u
T
2
where e is the strain tensor and the symbols r
and stand for gradient and tensor operator
product, respectively.
The microscopic incremental elastic constitutive
relation can be written as
_ rx C
0
x : _ ex C
0
x : symgrad _ u
C
0
x : symr _ u 3
where the apex denotes an increment when applied
to the stress or strain tensors, while it denotes the
fourth-order elastic tangent tensor, C
0
x, that
depends on the location of the point inside the
body, identied by the position vector x. In a
composite material this tensor presents high vari-
ability with the position vector.
We now suppose that the heterogeneous mate-
rial is composed by one matrix phase, denoted by
the subscript m, and by q dierent ber phases
embedded in the matrix (Fig. 1b). We assume that
the bers belonging to the same phase p have the
same length 2L
p
f
and cross section A
p
f
, while their
Fig. 1. Body made of a generic composite material (a) and ber reinforced composite material (b).
R. Brighenti / Computational Materials Science 29 (2004) 475493 477
orientation are randomly distributed in the solid
angle, so the assumption that each direction
is equally represented, can be made. The ber
direction can be identied by the unit vector i
parallel to the ber axis; using the polar angles /, h
(Fig. 1b), such vector can be written as
i f i
1
i
2
i
3
g
f sin h cos / sin h sin / cos h g 4
with respect to the Cartesian reference system.
The assumptions that each ber phase is
homogeneously distributed inside the matrix and
that a reference elementary volume (R.E.V.), with
characteristic length d, has the same average
compositionand consequently the same mechan-
ical propertiesof the composite, will be made in
the following.
Furthermore we assume that the body charac-
teristic length D is much more greater than the
microscopic characteristic length d, e.g. d=D ( 1.
Focusing our attention on the R.E.V., the fol-
lowing volume fractions of each component can be
dened:
l V
m
=V matrix volume fraction
g
p
V
f;p
=V fibers volume fraction of the
pth fiber phase
5
where V , V
m
and V
f;p
are the R.E.V. volume, the
matrix and the pth ber phase volume present in
the R.E.V., respectively.
Obviously the following relation holds:
V V
m

q
p1
V
f;p
6
or equivalently in dimensionless form: 1 l

q
p1
g
p
.
3. Equivalent homogenised material
The averaged properties of the composite
material can be dened by equating the virtual
power _ w (work rate) made by the composite
material with those of the homogenised equivalent
one [68]. Introducing a generic virtual displace-
ment eld ~u (and the corresponding strain tensor e
and strain rate tensor
_
e) as follows:
~u; e symgrad~u symr~u
_
e symr
_
~u
7
The stated condition can be written, for a not too
small generic composites volume V , as
_ w
_
V
jx
_
e : rdV

q
p1
_
V
v
p
x
_
e
f;p
r
f;p
dV
..
composites work rate

_
V
_
e : r
eq
dV
..
homogenised materials work rate
8
where
_
e
f;p
, r
f;p
are the virtual strain rate and the
stress in the generic ber belonging to the pth
phase, while the point scalar functions jx, v
p
x
are dened as
jx
1 if x 2 V
m
0 if x 62 V
m
_
and
v
p
x
1 if x 2 V
f;p
0 if x 62 V
f;p
_ 9
Computing the ber stress by mean of its elastic
modulus E
f;p
multiplied by the ber strain e
f
, and
computing the stress tensor of the homogenised
equivalent material r
eq
by mean of the equiva-
lent elastic tensor C
eq
and the actual strain tensor e
as
e
f
i i : e;
_
e
f
i i :
_
e
r
f
E
f;p
i i : e; r
eq
C
eq
: e
10
we have
_ w
_
V
jx
_
e : C
m
x : e dV

q
p1
_
V
v
p
i i :
_
e E
f;p
i i : e dV

_
V
_
e : C
eq
x : e dV 11
where e
f
is the strain tensor at the ber location.
Thus the previous relation can be rewritten:
_
V
_
e : jx C
m
x
_

q
p1
v
p
xE
f;p
QQ
_
: edV

_
V
_
e : C
eq
x : edV 12
478 R. Brighenti / Computational Materials Science 29 (2004) 475493
where the second-order tensor Q i i has been
introduced.
Being the equivalent material macroscopically
homogeneous at the scale of the R.E.V., the elastic
tensor C
eq
x can be assumed to be constant with
respect to the position vector x, i.e. C
eq
x C
eq
.
By comparing the rst and the second member
of Eq. (12), the equivalent tangent elastic tensor,
C
eq
x, can be recognised; since we assume
C
eq
x C
eq
, its volume mean value can be ob-
tained:
C
eq

1
V

_
V
jx C
m
_

q
p1
v
p
x E
f;p
QQ
_
dV
l C
m

q
p1
g
p
E
f;p

_
V
QQdV 13
where the denition of the ber and matrix volume
fraction (5):
l
1
V

_
V
jx dV
V
m
V
g
p

1
V

_
V
v
p
x dV
V
f;p
V
14
has been used.
In an incremental situation (referred to the ac-
tual matrixs strain) the tangent elastic properties
of the materials should be used to calculate the
increment of stresses in the matrix, in the ber and
in the equivalent material:
_ re C
0
m
e : _ e
_ r
f
e
m
f

dr
f
de
f

de
f
de
m
f
_ _
p
_ e
f
_ r
eq
e C
0
eq
e : _ e
15
where C
0
m
and C
0
eq
are the tangent elastic tensor of
the matrix and of the equivalent material, respec-
tively. Remembering equation (14), and indicating
with e
m
f
the strain of the matrix measured in the
bers direction, and with e
f
the actual bers
strain, the tangent stiness tensor C
0
eq
e of the
equivalent homogenised material can be obtained
as
C
0
eq

1
V

_
V
jx C
0
m
_

q
p1
v
p
x

dr
f
de
m
f
_ _
p
QQ
_
dV
l C
0
m

q
p1
g
p
dr
f
de
m
f
_ _
p

_
V
QQdV
C
0
eq
l C
0
m

q
p1
g
p
C
0
f
16
In the following the hypothesis of a single bers
phase presents in the composite will be made in
order to simplify the notation, so the summation
over the index p will be omitted.
Up to now a perfect bound hypothesis between
the bers and the matrix as been assumed, i.e. the
actual ber strain can be written in the simple form
e
f
i i : e.
To evaluate the tangent stiness tensor C
0
eq
, the
last integral of Eq. (16) can be performed on an
hemisphere having volume V
em
, normalised with
respect to the hemisphere volume itself, in order to
represent all the possible bers orientations in the
composite:
1
V
em
_
V
em
QQdV

1
V
em
_
R
0
_
2p
0
_ p
2
0
QQr d/r sin hdhdr

R
3
3
1
2
3
pR
3
_
2p
0
_ p
2
0
QQ d/sin hdh

1
2p
_
2p
0
_ p
2
0
QQ d/sin hdh 17
In the case of a ber phase homogeneously
distributed and oriented in the matrix, the contri-
bution C
f
of the ber to the elastic tensor C
eq
of
the composite is isotropic (if the matrix material is
isotropic too), and for sake of clarity some of the
corresponding elasticity tensor C
f
terms are re-
ported below:
R. Brighenti / Computational Materials Science 29 (2004) 475493 479
C
f
1111

1
5
gE
f
; C
f
1122

1
15
gE
f
; C
f
1133

1
15
gE
f
C
f
1112
C
f
1123
C
f
1113
0
C
f
1212

1
15
gE
f
; . . .
where E
f
is the elastic modulus of the bers.
As an example in Fig. 2 the dimensionless
composites elastic modulus (Fig. 2a) and Poissons
ratio (Fig. 2b) against bers volume fraction g for
dierent values of the Poissons ratio of the matrix
m
m
and Youngs modulus ratio E
m
=E
f
are dis-
played.
4. Elasto-plastic behaviour of the matrix
As stated above, the matrix is assumed to be-
have as an elasto-plastic material [9].
In incremental form, the well-known elasto-
plastic constitutive equations can be expressed by
_ e _ e
el
_ e
pl

_
C
1
m
_ r _ e
pl
18
where _ e
el
, _ e
pl
stand for the elastic and plastic part
of the strain rate tensor, while _ r is the stress rate
tensor.
Assuming that the elastic behaviour is dened
by a yield function F r; k
1
; . . . ; k
n
0 with k
1
; . . . ;
k
n
n hardening parameters, we have the elastic
behaviour when: (i) F r; k
1
; . . . ; k
n
< 0, while
plastic ow occurs when (ii) F r; k
1
; . . . ; k
n
0
and F
0
r; k
1
; . . . ; k
n
0; when (iii) F r; k
1
; . . . ;
k
n
0 and F
0
r; k
1
; . . . ; k
n
< 0 we have elastic
unloading.
The yielding condition for a perfectly plastic or
a hardening material can be written:
F r; k
1
; . . . ; k
n
0 and
dF
0

oF
0
or
dr 0 perfectly plastic
material
dF
oF
or
dr
oF
ok
h
dh
h
0 hardening material
_

_
19
A standard associative ow rule has been as-
sumed to describe plastic strain increment:
_ e _ e
el
_ e
pl

_
C
1
m
: _ r
_
k
oF
or
with
_
k
f : C
0
m
: _ e
H f : C
0
m
: f
20
where f oF =or
ij
, H is the hardening parameter
(as will be explained later), while
_
k P0 if F 0
and dF 0,
_
k 0 if F 0 and dF < 0 or F < 0.
The plastic multiplier
_
k can be calculated elimi-
nating the stress rate tensor from the equations
dF r; k
1
; . . . ; k
n
0 and the rst of Eq. (20).
Standard plastic criteria has been considered in
the following. The well-known denition of the
considered plastic criteria are reported in Table 1.
Yielding functions depending upon a single yield-
ing parameter k will be considered. The yielding
parameter k will be identied with the uniaxial
Fig. 2. Dimensionless composites elastic modulus (a) and Poissons ratio (b) against bers volume fraction g for dierent values of the
Poissons ratio of the matrix m
m
and Youngs modulus ratio E
m
=E
f
.
480 R. Brighenti / Computational Materials Science 29 (2004) 475493
yield stress r
0
, of the material F r; k F r;
r
0
0.
The evolution of the yielding parameter k has
been written in the simple form: r
0
r
Y
He
pl
where H is the hardening parameter that can be
related to the post-yielding stressstrain slope E
t
,
as H E
t
=1 E
t
=E
m
; for softening materials this
parameter assumes negative values (Fig. 3).
5. Imperfect matrixber interface
5.1. General considerations
In the case of the existence of an imperfect bond
between the reinforcing ber and the matrix, we
assume that a local strain jump, se
fm
t can be
present at the matrixber interface. The strain
jump can be written in the following form:
se
fm
t e
m
f
e
f
i i : e 1 se
m
f

e
m
f
1 se
m
f
21
The scalar function s measures the local sliding
degree between the ber and the matrix. When
s 0 the sliding is complete, i.e. se
fm
t e
m
f
and
no shear stress is transferred between the matrix
and the ber, the matrixber interface has not
stiness and the ber doesnt have any stress, that
is to say the composite material behaves as an
elastic material, with inclusions of voids having the
shape of the bers. When s 1 the interface is
perfect, i.e. no strain jump occurs so se
fm
t 0;
this case represent the maximum shear stress
transfer between the matrix and the ber, and the
ber carries the maximum possible load, giving the
maximum contribution to the composite load
carrying capacity (Fig. 4a).
We would like to refer the bers strain to the
matrixs strain; thus the actual bers strain can be
written:
e
f
e
m
f
se
fm
t i i : e se
m
f

e
m
f
se
m
f
22
It is reasonable to assume that the value of the
function s depends on the amount strain e
m
f
of the
matrix, the stiness of the ber and of the interface;
Table 1
Standard plastic criterion adopted for the matrix material
Criterion Yielding surface equation, F 0 Uniaxial materials parameters
Tresca F
1
2
maxjr
I
r
II
j; jr
II
r
III
j; jr
III
r
I
j k 0 k
r
0
2
s
0
r
I
, r
II
, r
III
: principle stresses s
0
: materials yielding shear stress
Mises F J
2
k
2
0 k r
0
=

3
p
s
0
J
2
: second deviatoric stress tensor invariant s
0
: materials yielding shear stress
Mohr-Coulomb F jsj c r tg/ 0 r
t

2c cos /
1sin/
r
c

2c cos /
1sin /
c, /: materials coesion and friction angle r
t
, r
c
: materials yielding tension and compression stresses
Drucker F a I
1


J
2
p
k 0 r
t

3
p
k
1

3a
p
r
c

3
p
k
1

3a
p
I
1
: rst stress tensor invariant
J
2
: second deviatoric stress tensor invariant r
t
, r
c
: materials yielding tension and compression stresses
k, a: materials parameters
r
0
is the uniaxial yield stress (for symmetric materials), r
t
, r
c
are the uniaxial yield stresses in tension and compression respectively (for
non-symmetric materials).
Fig. 3. Simplied uniaxial stressstrain curve adopted for the
elasto-plastic matrix.
R. Brighenti / Computational Materials Science 29 (2004) 475493 481
a simple model could be, for instance, a piecewise
linear dependence between s and e
m
f
as shown in Fig.
4b (dened by two strain characteristic parameters,
e
m
f ;1
, e
m
f ;2
). The sliding function has been dened lo-
cally along the ber; if we make the assumption of
very short ber, it reasonable to assume s as a
constant function equal to the mean value of se
m
f

along the single ber, as will be shown later.
If the bers material behaves in a linear fashion
(with Youngs modulus E
f
), the tangent elastic
modulus of the ber with respect to the matrix
strain can be calculated as follows:
E
f
e
m
f

dr
f
de
m
f

d
de
m
f
E
f
e
m
f
se
f m
t
..
e
f
se
m
f
_

_
_

_
E
f
se
m
f

_
e
m
f

dse
m
f

de
m
f
_
23
The tangent stiness tensor C
0
eq
e of the
equivalent homogenised material becomes now
(see Eq. (16)):
C
0
eq
l C
0
m
g
p
E
f
se
m
f

_
e
m
f

dse
m
f

de
m
f
_
p

_
V
QQdV 24
5.2. Equilibrium of the bermatrix system
Taking into account a cylinder of material
around a single ber, the equilibrium equation
along the bers direction can be written (Fig. 5a):
df x
dx
c

k u
R
u
r
25
where c 2 p r p d, f x,

k, u
R
, u
r
are the
bers perimeter, the total axial force sustained by
the ber, the stiness of the bermatrix inter-
face,
1
the displacement measured in the bers
direction far from the ber and at the interface
(which is assumed to coincide with the displace-
ment measured on the bers axis due to the neg-
ligibility of the bers radius with respect to the
radius of the cylinder of material under study),
respectively, while r and d are the bers radius and
diameter, Fig. 5a.
Considering the following relations:
du
r
x
dx
% e
f
x
r
f
x
E
f

r
f
x
E
f

A
f
A
f

f x
E
f
A
f
du
R
x
dx
% e
m
x
r
m
x
E
m
26
and deriving Eq. (25) with respect to x we get [10]
d
2
f x
dx
2
c

k
r
m
x
E
m
_

f x
E
f
A
f
_
0 27
or in an equivalent form:
d
2
f x
dx
2
c

k
P r
f
x A
f
E
m
A
m
_

f x
E
f
A
f
_
c

k
P f x
E
m
A
m
_

f x
E
f
A
f
_
28
being
r
f
x A
f
r
m
x A
m
P
) r
m
x A
m
P r
f
x A
f
f x r
f
x A
f
where P is the total force sustained by the cylinder
of composite under study.
Fig. 4. A possible simplied relation (b) assumed to describe imperfect bonds between ber and matrix (a).
1
The stiness

k of the bermatrix interface can be calculated

writing the shear stress value at a generic distance R from the
bers center: sR
G
rlnR=r
u
R
u
r


k u
R
u
r
.
482 R. Brighenti / Computational Materials Science 29 (2004) 475493
Making the positions
a
1
E
m
A
m

1
E
f
A
f
; b

k a
_
F
P
a E
m
A
m
Eq. (28) can be rewritten in the following form:
d
2
f x
dx
2
b
2
f x b
2
F 0 29
Integrating the previous second-order dieren-
tial equation and imposing the boundary condi-
tions f x L 0 (the ber is assumed to have
length 2L), the solution of (29) becomes:
f x F 1
_

coshb x
coshb L
_
30
while the shear stress sx at the interface and the
force mx sustained by the matrix can be calcu-
lated as:
sx
1
c

df x
dx

F b
c

sinhb x
coshb L
_ _
mx P f x
F a E
m
A
m
_
1
_

coshb x
coshb L
__
31
5.3. Sliding function obtained by energetic consid-
eration
In order to obtain the mean value se
m
f
of
the sliding function se
m
f
which relates the mean
Fig. 5. Cylinder of matrix material around a single ber under tension (a). Stress distributions around a single ber in the pre-
debonding stage (b), during the debonding stage (c) and during the complete debonding stage (d).
R. Brighenti / Computational Materials Science 29 (2004) 475493 483
matrix and the mean bers deformation in an
average fashion along the ber, e
m
f
, we can deter-
mine the actual elastic energy stored in the ber
and make it equal to its averaged value. The actual
elastic density energy is:
W
1
2 V
f

_
V
f
r
f
x e
f
x dV
f

1
2 V
f

A
f
E
f

_
L
L
r
2
f
x dx

1
4 L E
f

_
L
L
r
2
f
x dx 32
where the bers volume has been written as V
f

2 L A
f
.
On the other hand the mean value of the ma-
trixs strain e
m
f
in the bers direction can be
computed as:
e
m
f

1
2L

_
L
L
e
m
f
x dx

1
2L

_
L
L
i i : ex dx 33
while the mean bers elastic energy density W is
W
1
2
r
f
e
f

1
2
E
f
se
m
f
e
m
f

2
34
where e
f
se
m
f
e
m
f
and the mean bers stress is
r
f
E
f
se
m
f
e
m
f
.
The function se
m
f
referred to the mean strain
can be obtained by solving Eq. (34):
se
m
f

1
e
m
f

2 W
E
f

35
Assuming that W W we get the function
sought:
se
m
f

1
e
m
f

2 W
E
f

36
and taking into account Eq. (32):
se
m
f

1
e
m
f
E
f

_
L
L
r
2
f
x
2L

1
E
f
2L

_
L
L
e
m
f
x dx

_
L
L
r
2
f
x dx
2L

37
with
e
m
f
x mx=E
m
A
m

P f x=E
m
A
m

F a
_

1
E
m
A
m
1
_

coshb x
coshb L
__
38
In the following, Eq. (37) will be calculated
using the proper ber stress law distribution in the
pre-, post- and ultimate stage.
5.4. Pre-debonding stage (s (x L)< s
au
)
In the pre-debonding stage the stress in the ber
can be expressed by using f x obtained from Eq.
(30): r
f
x f x=A
f
(Fig. 5b).
In this case the mean value of the sliding func-
tion is:
se
m
f

1
e
m
f
E
f

_
L
L
F
A
f
1
coshbx
coshbL
_ _ _ _
2
dx
2L

2
p
e
m
f
E
f
A
f

_
L
L
1
coshbx
coshbL
_ _
2
dx
L

_
39
and solving the previous integral, nally the slid-
ing function se
m
f
in the pre-debonding stage be-
comes:
se
m
f

F

2
p
e
m
f
E
f
A
f

2
1
L cosh
2
b L
L
sinh2b L
2b
_ _

4 sinhb L
Lb coshb L

constant 40
484 R. Brighenti / Computational Materials Science 29 (2004) 475493
where
e
m
f

1
2L

_
L
L
e
m
f
x dx

1
2L

_
L
L
F a
_

1
E
m
A
m
1
_

coshb x
coshb L
__
dx
F a
_

1
E
m
A
m

sinhb L
b L E
m
A
m
coshb L
_
41
5.5. Debonding initiation
Increasing the load applied to the composite,
the debonding phenomena appears rst at the -
bers extremities when the shear stress transmitted
from the matrix to the ber becomes equal to the
ultimate shear stress s
au
; this condition can be
stated as sx L s
au
. Since the sliding func-
tion must be expressed with respect to e
m
f
, it is
necessary to obtain such a quantity corresponding
to this situation.
Recalling the rst of Eq. (31), the debonding
condition can be simply identied:
jsLj
F b
c

sinhb L
coshb L
_ _

F b
c
tanhb L s
au
) F
deb

s
au
c
b tanhb L
42
where the subscript deb stands for debonding;
F
deb
indicates the minimum value of the total force
applied to the composite cylinder under study
that causes the starting of the debonding phe-
nomenon.
The corresponding matrix mean strain mea-
sured in the bers direction, e
m
f
, becomes:
e
m
f;deb

1
2L

_
L
L
e
m
f
xdx

s
au
c
b tanhb L

a b L E
m
A
m
b L 1
b L E
m
A
m
_ _

s
au
c
b tanhb L
a
_

1
E
m
A
m

sinhb L
b L E
m
A
m
coshb L
_
43
5.6. Partial debonding stage
The debonding stage occurs after the shear
stress has reached the value s
au
and the friction
shear stress s
fu
begins to be present in the de-
bonded zones at the extremities of the ber for a
generic length indicated as L l (Fig. 5c).
Integrating the dierential equation (29) with
the boundary conditions f x l c s
fu
Ll,
where c represents the bers perimeter, the solu-
tion becomes:
f x
r
f;deb
A
f
c s
fu
L jxj
for L6x 6 l and l 6x 6L
r
f
x A
f
F 1
cs
fu
LlF
F

coshbx
coshbl
_ _
for l 6x 6l
_

_
44
The shear stress at the matrixber interface is
sx
1
c

df x
dx

s
fu
for L6x 6 l and l 6x 6L
b s
fu
L l
F
c
_

sinhbx
coshbl
for l 6x 6l
_
_
_
45
Since sx l s
au
, we can determine the value
of the corresponding F F
deb
l, F
deb
l
cs
au
b

coshbl
sinhbl
. The bers force and the shear stress
distribution in the non-debonded zone can be
written:
f x
r
f;deb
A
f
c s
fu
Ljxj for L6x6 l and l6x6L
r
f
A
f
F
deb
l 1
cs
fu
lLF
deb
l
F
deb
l

coshbx
coshbl
_ _
F
deb
l 1M
coshbx
coshbl
_ _
for l6x6l
_
46a
R. Brighenti / Computational Materials Science 29 (2004) 475493 485
and
sx
1
c

df x
dx

s
fu
for L 6x 6 l and l 6x 6L
b s
fu
L l
F
deb
l
c
_ _

sinhbx
coshbl
for l 6x 6l
_

_
46b
Remembering Eq. (37), the mean sliding func-
tion se
m
f
in this stage can be obtained as follows:
where the following terms can be acknowledged:
_
l
l
r
2
f
xdx
F
2
A
2
f
2l
_

M
2
cosh
2
b l
l
_

sinh2b l
2b
_

4M sinhb l
b coshb l
_
;
and
2
_
L
l
r
2
f;deb
x dx
2
3

p
2
d
2
s
2
fu
A
2
f
L l
3
where we have indicated with M the following
quantity: M
cs
fu
lLF
deb
l
F
deb
l
. In this stage the ma-
trixs strain in the bers direction can be obtained:
e
m
f
x
mx
E
m
A
m

F
deb
l a
cs
fu
Ljxj
E
m
A
m
for L6x6 l and l 6x6L
F
deb
l a
1
E
m
A
m
1
cs
fu
lLF
deb
l
F
deb
l

coshbx
coshbl
_ _ _ _
for l 6x6l
_

_
48
while its mean value along the ber is:
e
m
f

1
2L

_
L
L
e
m
f
x dx

1
2L

_
l
l
e
m
f
x dx
_
2
_
L
l
e
m
f
x dx
_
e
m
f
F
deb
l a
_

1
L E
m
A
m
l
_

2
b

c s
fu
L l F
deb
l
coshb l
sinhb l
c s
fu

L
2
l
2
2
_
L l
___
49
The previous equations can be used in the fol-
lowing way:
Given the geometric bers characteristics L, d
and the mechanical parameters of the interface
s
fu
, s
au
from Eq. (46b), since at x l the shear
stress must be equal to s
au
, the corresponding
distance l can be calculated.
From Eq. (49) the value of F
deb
l can be ex-
presses in function of e
m
f
.
Finally substituting the result obtained for
F
deb
l in Eq. (47) the value of the sliding func-
tion s can be obtained.
5.7. Complete debonding stage
When the debonding zone extends for the whole
length of the ber (L
1
0, i.e. l L) the shear
stress transmitted from the matrix to the ber is
equal to the friction shear stress s
fu
everywhere
(Fig. 5d).
In this stage the forces carried by the ber at a
generic distance x from the bers midpoint is ex-
pressed by:
se
m
f

1
e
m
f
E
f

_
L
L
r
2
f
x dx
2L

1
e
m
f
E
f

_
l
l
r
2
f
x dx 2
_
L
l
r
2
f;deb
x dx
2L

F
deb

2
p
e
m
f
E
f
A
f

2l
M
2
cosh
2
bl
l
sinh2bl
2b
_ _

4Msinhbl
bcoshbl

2
3

p
2
d
2
s
2
fu
F
2
deb
L l
3
L

_
47
486 R. Brighenti / Computational Materials Science 29 (2004) 475493
f x c s
fu
L jxj for L6x 6L 50
while the shear stress at the interface keep a con-
stant value equal to sx s
fu
.
The sliding function se
m
f
in this stage can be
obtained simply by putting l 0 in Eq. (47):
se
m
f

1
e
m
f
E
f

2
_
L
0
r
2
f
x dx
2L

1
3
_

c s
fu
L
A
f

1
E
f
e
m
f
51
6. F.E. implementation
The proposed model has been implemented in a
F.E. program in order to verify its capability in
predicting the behaviour of ber reinforced com-
posite materials characterised by an elasto-plastic
matrix reinforced with bers having the possibility
of debonding. The key feature of the constitutive
model is the calculation of the mean sliding
function se
m
f
at every Gauss integration point; to
obtain such a function the mean matrix strain e
m
f
measured in the bers direction is needed. Since
the main assumption is the random distribution of
the bers in the matrixs material, i.e. each direc-
tion of the solid angle is equally represented, the
mean matrix strain can be simply obtained by
computing the hydrostatic strain from the strain
tensor of the matrix, e
m
f
tr e=3.
The initial stiness matrix method has been used
and the integration of the incremental equilibrium
equations has been made by an implicit Euler
backward algorithm.
Fig. 6. Elasto-plastic ber reinforced cylinder under internal pressure: model with 24 elements in plane strain condition (a), and model
with 16 elements in axisymmetric condition (b) (dimensions in millimetres).
R. Brighenti / Computational Materials Science 29 (2004) 475493 487
7. Applications
Some numerical tests has been conducted to
assess the reliability of the proposed constitutive
mechanical model to predict the overall behaviour
of several composite simple structures, as will be
shown in the following.
7.1. Unreinforced and ber reinforced thick-walled
cylinder under internal pressure
The rst numerical example considered is the
simulation of the mechanical behaviour of an el-
asto-plastic cylinder, with metal matrix obeying
the Mises criterion, reinforced with a random
Fig. 7. Radial displacement of the inner face of the cylinder in the case of plane strain model (a) and axisymmetric model (b) for
unreinforced and ber reinforced materials. Dots represent the theoretical solution [11].
Fig. 8. Contour map of the sliding function se
m
f
at the nal load level for the axisymmetric model: (a) 4 nodes elements with
s
fu
0:1 s
au
; (b) 8 nodes elements with s
fu
0:1 s
au
; (c) 8 nodes elements with s
fu
0:5 s
au
.
488 R. Brighenti / Computational Materials Science 29 (2004) 475493
distribution of bers (stier than the matrix) sub-
jected to internal pressure. Two dierent model
has been considered: the plane strain and the axi-
symmetric one and 4 nodes and 8 nodes elements
have been employed as shown in Fig. 6. The
maximum value of the applied internal pressure (in
10 equal steps) has been 180 MPa and the obtained
inner face radial displacement against the load
Fig. 9. Contour map of the hoop plastic deformation at the nal load level for the axisymmetric model: (a) 8 nodes elements, without
bers; (b) 8 nodes elements with V
f
10% and s
fu
0:1 s
au
; (c) 8 nodes elements with V
f
10% and s
fu
0:5 s
au
.
Fig. 10. Fiber reinforced concrete specimen under pure shear: considered discretisation (a) (dimensions in millimetres), shear stress
against vertical displacement for dierent amount of ber volume fraction and length (b).
R. Brighenti / Computational Materials Science 29 (2004) 475493 489
factor is displayed in Fig. 7. As can be noted the
nal displacement decrease increasing the bers
volume fraction with respect to the unreinforced
case [11] both for the plane strain and the axi-
symmetric models. The sliding function se
m
f
for
the axisymmetric model for s
fu
0:1 s
au
and
s
fu
0:5 s
au
is reported in Fig. 8; in proximity of
the inner face the mean value of the sliding func-
tion is greater (dark zones) with respect to its
values in proximity of the outer face, because of
the constraint eect given by the compressive ra-
dial stress that does not allows the debonding
phenomena to take place heavily in the material.
Furthermore it is possible to note that the eect
of the bers (apart the reduction of the nal dis-
placements), with respect to the unreinforced case,
is not the decreasing in the maximum plastic
strains values in the matrixs material but the
limitation of the zone where such a plastic defor-
mations occur (Fig. 9).
7.2. Pure shear test on a ber reinforced concrete
specimen
The second numerical example is the simulation
of a pure shear test on a steel ber reinforced
concrete specimen [12] as shown in Fig. 10a. The
steel reinforcing bers used were straight but not
with perfectly cylindrical shape; two dierent
length were considered: 16 mm and 32 mm with an
equivalent diameter equal to 0.55 mm. Since the
mechanical parameters necessary to dene the
Fig. 11. Contour map of the sliding function se
m
f
at the nal load level for: (a) V
f
0:5%, 2L
f
16 mm, s
fu
0:2 s
au
; (b) V
f
1:5%,
2L
f
16 mm, s
fu
0:2 s
au
; (c) V
f
1:5%, 2L
f
32 mm, s
fu
0:2 s
au
; (d) V
f
1:5%, 2L
f
32 mm, s
fu
0:02 s
au
.
490 R. Brighenti / Computational Materials Science 29 (2004) 475493
proposed model where not explicitly reported in
[12], some hypothesis has been made to perform
the present numerical simulation, such as the limit
shear stress and the friction shear stress that have
been assumed equal to 0.5 and 0.1 MPa, respec-
tively. In Fig. 10b the vertical displacement against
the uniform shear stress on the central vertical
section is displayed for dierent situations: speci-
men without bers, bers volume fraction V
f

0:5% and length 2L
f
16 mm, bers volume
fraction V
f
1:5% and length 2L
f
16 mm and
bers volume fraction V
f
1:5% and length 2L
f

32 mm. The matrixs material has been assumed to
behaves following the Drucker elasto-plastic cri-
terion with linear strain softening both in tension
and compression (the post yielding elastic modulus
has been assume to be E
t
0:05 E
m
) and with
Poissons ratio equal to 0.11.
In Fig. 11 the contour maps of the sliding
function se
m
f
in the specimen are reported in the
nal loading situation for the three cases ((a)(c))
with dierent bers volume content and length but
with constant interface characteristic shear stress
ratio, s
fu
0:2 s
au
, and for the fourth case similar
to case (c) except for the assumed value of the
friction shear stress, s
fu
0:02 s
au
. As can be
noted the sliding occurs mainly around the high
shear stress band in the centre of the specimen
(dark zones), while outside this area the sliding
phenomena is less pronounced or even absent
Fig. 12. Contour map of the shear plastic deformation at the nal load level for: (a) V
f
0:0%; (b) V
f
0:5%, 2L
f
16 mm,
s
fu
0:2 s
au
; (c) V
f
1:5%, 2L
f
32 mm, s
fu
0:2 s
au
; (d) V
f
0:5%, 2L
f
32 mm, s
fu
0:02 s
au
.
R. Brighenti / Computational Materials Science 29 (2004) 475493 491
(white zones, cases (c) and (d)) for long bers
reinforcing. Case (d) represents the practical situ-
ation of sliding without friction: in such a case the
sliding function se
m
f
assumes values near to zero
(black central zones) since the shear stress transfer
after complete debonding is absent.
Finally, in Fig. 12 the contour maps of the
plastic part of the shear strains are displayed in the
nal loading situation: due to the low value of
the bers content, the dierences are not so evident
between dierent cases. The last situation (d) is
particularly interesting: in the case of frictionless
sliding, once the debonding phenomena takes
place completely along the bers, the composite
behaves as a solid weakened by inclusions (with
the same volume of the debonded bers) so the
eect of ber reinforcing disappears. In such a case
the plastic shear strain deformation is even higher
than in the unreinforced case. The eect of the -
bers is thus negative on the load carrying capacity
of the specimen; only in the pre-debonding stage
the contribution to the composite stiness is
eective (only if the bers Young modulus is
greater of the matrixs one).
8. Conclusions
The behaviour of ber-reinforced composites
has been examined through a non-linear consti-
tutive law obtained through a physical-based
approach. The elasto-plastic macro constitutive
equations for such a class of materials has been
developed by considering the matrix phase with
elasto-plastic behaviour and the ber-reinforcing
phase with an imperfect bond with the matrix. The
debonding phenomena has been taken into ac-
count by the use of a scalar function, the sliding
function s, determined by considering the evolu-
tion of the shear stress distribution along a single
ber during the loading process. The proposed
mechanical constitutive law has been implemented
in a F.E. code and some numerical simulation
have been performed in order to assess its reli-
ability in the simulation of the mechanical behav-
iour for such a class of materials.
The obtained results show the capability of
the proposed mechanical model, despite its sim-
plicity, to describe the main non-linear phenom-
ena occurring in ber reinforced composite
materials.
The main eect of the ber phase has been
observed to be: in the pre-debonding phase the
increasing in the materials stiness, the reduction
of the displacements and the limitation of the
plastic zone extension inside the matrix material
(generally this reduction depends on the mechan-
ical characteristics of the interface).
The described behaviour heavily depends,
apart the matrixs mechanical behaviour, on the
bers volume fraction, on the value of the
interfaces stiness, on the bers elastic modulus,
on the bers length and section (bers aspect
ratio) and on the tangential stresses characteristic
of the bermatrix interface, i.e. the limit
shear stress and the friction bermatrix shear
stress.
The whole behaviour of the composite can be
completely controlled by determining the proper
values of the above listed parameters (which are all
physically-based parameters) that can be simply
identied in practical situations.
Acknowledgements
The author gratefully acknowledges the re-
search support for this work provided by the
Italian Ministry for University and Technological
and Scientic Research (MIUR) and the Italian
National Research Council (CNR).
References
[1] M. Hori, S. Nemat-Nasser, On two micromechanics
theories for determining micromacro relations in hetero-
geneous solids, Mech. Mater. 31 (1999) 667682.
[2] B. Balendran, S. Nemat-Nasser, Bounds on elastic moduli
of composites, J. Mech. Phys. Solids 43 (1995) 18251853.
[3] B. Hassani, E. Hinton, A review of homogenization and
topology optimization Ihomogenization theory for media
with periodic structure, Comput. & Structures 69 (1998)
707717.
[4] B. Hassani, E. Hinton, A review of homogenization and
topology optimization IIanalytical and numerical solu-
tion for homogenization equations, Comput. & Structures
69 (1998) 719738.
492 R. Brighenti / Computational Materials Science 29 (2004) 475493
[5] P. Bisegna, R. Luciano, Bounds on the overall properties
of composites with debonded frictionless interfaces, Mech.
Mater. 28 (1998) 2332.
[6] A.L. Kalamkarov, H.Q. Liu, A new model for the
multiphase bermatrix composite materials, Composites
B 29 (1998) 643653.
[7] A.L. Kalamkarov, H.Q. Liu, D.O. McDonald, Experi-
mental and analytical studies of smart composite rein-
forcement, Composites B 29 (1998) 2130.
[8] A.L. Kalamkarov, A.G. Kolpakov, Design problems for
the ber-reinforced composite materials, Composites B 27
(1996) 485492.
[9] D.R.J. Owen, E. Hinton, Finite Elements in Plasticity:
Theory and Practice, Pineridge Press, Swansea, UK, 1980.
[10] A. Ngolle, J. Pera, Microstructural based modelling of the
elastic modulus of ber reinforced cement composites,
Adv. Cem. Based Mater. 6 (1997) 130137.
[11] P.G. Hodge, G.N. White, A quantitative comparison of
ow and deformation theories of plasticity, J. Appl. Mech.
17 (1950) 180184.
[12] R. Khaloo, K. Nakseok, Inuence of concrete and ber
characteristics on behavior of steel ber reinforced concrete
under direct shear, ACI Mater. J. (NovemberDecember)
(1997) 592601.
R. Brighenti / Computational Materials Science 29 (2004) 475493 493