Resine Epoxy

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Composites Science and Technology 38 (1990) 371-386

Micromechanical Characterization of the Non-linear


Viscoelastic Behavior of Resin Matrix Composites
Jacob A b o u d i
Department of Solid Mechanics, Materials and Structures, Faculty of Engineering,
TeI-Aviv University, Ramat-Aviv, 69978, Israel

(Received 2 August 1989; revised version received 6 November 1989:


accepted 16 November 1989)

A BS TRA C T
A micromechanical analysis is proposed for the determination of the overall
response of unidirectional composites with non-linearly viscoelastic resins.
The non-linear behavior is modeled by the single-integral constitutive
equations developed by Schapery. The method is implemented for the
prediction of the non-linear viscoelastic behavior of unidirectional
glass/epoxy and graphite/epoxy composites subjected to various types of
loading conditions. The predicted response is compared with a finite element
solution, and good agreement between the two methods is shown to exist.

INTRODUCTION
Polymer matrix composites exhibit time-dependent effects, i.e. viscoelastic
behavior. The magnitude of the viscoelastic deformations induced in the
composite by a given loading condition is dependent upon various factors
such as stress level, temperature and humidity. In particular, for high stress
levels, the composite may exhibit non-linear viscoelastic behavior. As a
result, the stress-strain relationship is not linear and the Boltzmann
superposition principle is no longer applicable. In some cases, the stress
level at which the material becomes non-linear appears to decrease with
increasing time. 1 The implication is that extremely long time predictions
may require the use of non-linear viscoelastic constitutive equations even for
very low stress levels. Consequently, non-linear stress-strain relations must
be established in such a way that the magnitude of the applied load (stress or
strain) is taken into account.
371
Composites Science and Technology 0266-3538/90/$03"50 1990 Elsevier Science Publishers
Ltd, England. Printed in Great Britain

372

J. Aboudi

A method for the representation of the constitutive equations of nonlinear viscoelastic materials is given by a polynomial expansion of a multiple
integral expression {see, e.g. Lockett-' and references cited therein).
Experimental determination of the material functions involved in this
description requires a large number of tests. For this reason approximate
single-integral constitutive relations which incorporate the non-linear
behavior were considered by several investigators. 2
Schapery 3 has developed single-integral non-linear constitutive equations, derived from the thermodynamic theory of irreversible processes.
The only time-dependent properties these equations contain are the linear
viscoelastic properties (e.g. compliances). Furthermore, for uniaxial loading,
only four stress- (or strain-) dependent non-linear properties enter, which
characterize the non-linear effects.
The single-integral representation of Schapery was used by several
investigators, e.g. Refs 4-8, to describe the non-linear viscoelastic behavior
of composite materials. Owing to the existing anisotropy, the creep
compliance kernel includes several components to characterize the response
of the viscoelastic anisotropic material in the various directions. This
necessitates the determination of the various viscoelastic parameters in
various directions (e.g. from 0", 90 , 10 specimens, etc.; however, the number
of tests needed is limited). 9 If a different fiber-matrix combination is
required, this characterization must be carried out independently of the
previous one.
A micromechanical approach for the analysis of composite materials
predicts their behavior from the knowledge of the properties of the fiber and
matrix materials. Such an analysis relies on the study of the detailed
interaction between the fibers and matrix. As the fibers can usually be
considered as perfectly elastic (i.e. with no time dependence), and the
matrices generally exhibit isotropic behavior, the characterization of the
composite from micromechanical considerations would be relatively easier.
To this end, it would be necessary to determine the non-linear viscoelastic
behavior of the unreinforced isotropic matrix. Methods for the characterization of non-linear viscoelastic polymers were given b y Peretz and
Weitsman ~ and Bruller, 1L for example. As only the individual constituent
properties of the composite need to be determined, the micromechanics
approach is extremely cost-effective, especially if various fiber-matrix
combinations need to be evaluated.
Schaffer and Adams 12'13 performed a two-dimensional finite-element
micromechanical analysis for the prediction of the non-linear viscoelastic
behavior of unidirectional composites with epoxy resins. The non-linear
characterization of the polymeric matrix was obtained from a series of
experiments using Schapery single-integral representation. The computed

Non-linear viscoelastic behavior of resin matrix composites

373

response of the composite was compared with measured data obtained in


various situations.
In recent years, the author presented an analytical micromechanical
theory for the prediction of the overall behavior of fiber-reinforced
composites. This theory is based on the analysis of a repeating cell in a
unidirectional composite in which the fibers are distributed uniformly in the
matrix. The ability of the theory to provide the overall response of elastic,
thermoelastic and linearly viscoelastic composites was reviewed in Ref. 14.
Initial yield surfaces of metal matrix composites and their elastoplastic
response were also predicted from this micromechanical theory. Recently, ~5
the theory was used to predict the overall viscoelastic behavior of
unidirectional composites with thermorheologically complex resin matrices.
In the present paper, the micromechanical method of cells is implemented
for the prediction of the non-linear viscoelastic behavior of unidirectional
composites. The non-linearly viscoelastic matrix is modeled by the Schapery
single-integral representation for isotropic materials. To assess critically the
prediction of the present micromechanical method, comparisons with the
finite-element solution of Ref. 12 are given. To this end, glass and graphite
fibers were used to reinforce Hercules 3501-6 epoxy resin, the non-linear
viscoelastic characterization of which was given in Ref. 12. It is shown that
good correspondence between the present and finite-element prediction
exists, which supports the reliability and accuracy of the method. The
present micromechanical approach can be easily applied to predict the
response of unidirectional composites subjected to various types of
combined normal and shear loadings. The simplicity of the method results
from the fact that the procedure for the determination of the composite
response, which is computed progressively in a stepwise manner, involves
small systems of algebraic equations which are solved at each time step.
Generalizations of the proposed method for the determination of the nonlinear viscoelastic response of composite laminates and composite
structures are possible. Furthermore, in conjunction with the approach
given in Ref. 15, it should also be possible to generalize the present
micromechanical analysis to the prediction of the overall behavior of
unidirectional composites with non-linear thermoviscoelastic resin matrices.
An experimental characterization of this type of unreinforced polymer was
given in Ref. 16.

M I C R O M E C H A N I C A L ANALYSIS
The continuum model for unidirectional fiber-reinforced materials is based
on the assumption that the continuous square fibers extend in the x 1

374

J. Aboudi

direction and are arranged in a doubly periodic array in the x, and x 3


directions (see Fig. l(a)). The cross-section of the square fiber is h~, and h 2
represents its spacing in the matrix. As a result of this periodic arrangement,
it is sufficient to analyze a representative cell as shown in Fig. l(b). This
representative cell contains four subcells /~,7 = 1,2. Let four local coordinate systems t~c
~- 1 , - ~(P)
2 ,~~(m
3 / be introduced, all of which have origins that
are located at the center of each subcell.
As the average behavior of the composite is sought, it turns out that a firstorder theory in which the displacement in the subcell is expanded linearly in
terms of the distances from the center o f the subcell (i.e. in terms of (2~) and
~(;'h/ is sufficient for this purpose. Second-order theories were used in wave
--3
propagation problems where average properties are certainly not sufficient
for the prediction of dispersion and attenuation. The following first-order
displacement expansion in each subcell is considered:
ul~;',- = wl~;" + .~(2~)~Iz-.
~;') T -'3c(~')'/'(~;')v'i

i= 1,2,3

(I)

where wl~r) are the displacement components of the center of the subcell, and
the microvariables ~bl.#;), ~,I#~') characterize the linear dependence of the
IX2

fibers

/ f X,
X3

(a)
iX2

-F
JL I t . v :]_Y: 2___

_. ~

".'is.2

:s,z

ii,.,

,,.2

L/

/~'Xl
X3
(b)

Fig. 1. (a) Unidirectional fiber composite. (b) Representative unit cell.

Non-linear viscoelastic behavior of resin matrix composites

375

displacements on the local coordinates 2(2~', 2~3~'. In eqn (1) and the sequel,
repeated fl or y do not imply summation.
The c o m p o n e n t s of the small strain tensor are given by

e!~.;"
tj = x2[Olu}~'~'+ ~iul ~~']

i,j = 1, 2, 3

(2)

where ?l = ~/Ox~, ~2 = 3/02(2a' and 03 = d/O2~~'.


In the present paper, it is assumed that the fibers can be represented by
linearly elastic and anisotropic materials, whereas the matrix is modeled as a
non-linear viscoelastic material. It should be noted, however, that the
present analysis can easily be generalized to viscoelastic fibers as well.
For transversely isotropic elastic fibers, with x~ being the direction of
anisotropy, the stresses are related to the strains in the form
atP~) = C(%(t~r'

fl = 7 = 1

(3)

where
O.1#7, = [-0-(107,, 0.(202,',, fig/I;'),33 0"(1#~I' 0"(1/13/}'0"~')]

~{a~.,= [-e]aie,, e(2a~/,,da;.,,332e'ta~'', 2dae'13,2e(2a3e']


and the elasticity matrix of the fiber phase (denoted by 'f') is given by

c?i

c?;
c(2f;

0
o
0

o
0
o

0
0
0

c~f~
,,4

0
"~f'
t- 4.`*

0
0

c(2.

C (f,

symmetric

(4)

.(,-,(f )
,~(f) "~
q'22 -- '-23!

The elements -ij


rtr~ --in (4) can be easily expressed in terms of the engineering
constants of the fibers.
The matrix (fl + 7 # 2) is considered as a non-linearly viscoelastic material.
To this end, the single-integral non-linear viscoelastic constitutive equations
derived by Schapery 3 for an isotropic material under isothermal conditions
are used. For a uniaxial loading, the following relation is obtained:

e(t) = goDoa(t) + g, f l AD( - #') 0-~7[gza(,')] dr'

(5)

where D Ois the initial (time-independent) value of the compliance and AD(#)
is the transient (time-dependent) value of the compliance, Which represents
the linear viscoelastic response. The reduced time ~(t) is defined by
=

['
30

dr/a.

(6)

376

J. Aboudi

where a, is a time-scale shift factor, and ~' = ~(t'). The functions go, gt, g2 and
a, depend on the stress level a. The Boltzmann representation for linearly
viscoelastic materials under isothermal conditions can be recovered from (5)
by setting go = gt = g2 = a,, = 1.
Generalization of the constitutive eqn (5) to multiaxial loading was given
by Schapery. 3 For isotropic materials, the representation involves two
independent functions instead of the single function AD(~) in eqn (5).
Following Ref. 12, we assume in the present analysis that the only coupling
between applied stresses is that due to the Poisson effects. This assumption
leads to the following multiaxial non-linear viscoelastic relationships for the
isotropic matrix (fl + 7 - 2):
e!#.~)
= (1 + v(t)){D}trl]~' - v(t){D}a~k#k~)6ij
U

(7)

where the operation /furj ~ a is defined by the right-hand side o f e q n (5), and v(t)
is the time dependent Poisson's ratio of the matrix material.
The average stresses, 6i~, in the composite are determined from the average
stresses in the subcells, S~] ~, in the form
2

(8)

= -y
fl,7=l

where t:~. = h#h,., V = (h i +


and

h2) 2, which

is the area of the representative cell,

laa-2 lh>./2

S[~7 ) =--1
Ufl7 J-hfl/2 J-by~2

,-,--2 ,~--3

(9)

The conditions for the continuity of tractions, which are imposed along
the interfaces of the subcells of the representative cell in an average sense,
lead to~ v
S~le~
2i
~ c~2e)
'-'2i
S(fll)
3i
=

(10)
(11)

C(/~2)
'-'3i

The continuity of the displacements at the interfaces between the subcells


and neighboring cells are given by t7
w~~~'= wla2' = w12~' = wl z z ' -

w,

(12)

hl~bl'~' + h2~2~' = (ht + h 2 ) ~ x 2 W i

(13)

0
h,~'l #1,+ h2O~#2,= (/h + h2)7---wi
CX3

(14)

Non-linear viscoelastic behavior of resin matrix composites

377

The average strain in the composite is given by


2

Iy
gij -= --~

l...d

,,

~#~.)

(15)

'~13","~ij

#.'/= 1

Using eqns (2), (13) and (14), we readily obtain that

I/'&'i wfk

gii=-~t~xj+ ~ )

(16)

The micromechanics analysis establishes the relationships between the


average stresses, ~;j, and the average strains, g~j,in the composite, which leads
to the determination of its overall behavior.
Substitution of (3) in (9) leads, in conjunction with (2), to the following
relations in the fiber region fl = 7 = 1:
,.,{f) .I,(11 )

(17b)

c]f~2g11 + ..(r).ll
1~"i"-'..it)
,/,{t tl
~'23W2
-227'3

(17c)

,-,(f) ,4d t 1)

S(11)
33

="

= t4a.-tO.\. 1 "{-

(17d)

sp3"=c4,L\~--~-i + o?)"

(17e)

l ))

(17f)

12

S (23
ll)

{,4~111
(21
~-- .(f~
t- 66tW'
3 ) 0"1"-

In the matrix region (fl + 7 g: 2), the following relations are obtained from
eqns (7) and (9):
gl~ = (1 + v(t)){O}S~ "e' - vtt~s
1[

I~'kk

(18a)

~b~2aT'=(1 + v(t)){D}S~a~ :' - ,,t,~fn~c


- v , t ' - ' s o k k#',

(18b)

~3Pr' = (1 + v(t)){D}S~PJ , - v(t){D}S~':'

(18c)

0't'2 q~]#~'=
v(t)){D} S~:2:'
8x---7+
2(1 +
-C'W3
-+

~fllBT) =2(1 +

v(t)){D}S~#3 ''

~b~3#~'+ ~2P7' = 2(1 + v(t)){D}S~2#J '

18d)

(18e)

(18f)

It is readily seen that in the matrix region (fl + 7 ~ 2), eqns (18) form a system
of non-linear equations, as a result of which it is not possible to express the
subcell stresses S~ff~ directly in terms of the microvariables. Consequently,

3". Aboudi

378

the solution procedure is incremental in the sense that it is applied step by


step in time until the values of the field variables are obtained at the
requested time. The functions go, g~, g2 and a~ to be used in the equation {D}
in (18) are taken to depend on the equivalent stress S~ ~) (rather than the
single stress which exists in the uniaxial relationship (5)). The equivalent
stress is defined by
=

>

'+ '

-" _

~,(//;,)/2
'~i K,(~;,)~2) 1 2
+ 3(S{tt2"))a + R(
~ "-'13
, + o,'-'z3
, ~

(19)

Let us consider the case in which the composite is subjected to normal


stresses #1 l, 6,2, 033, for which results are given in this paper and a finite
solution is available. 12 At any time t the integral expressions involved in
eqns (18) are approximated as follows. Let a typical integral be represented
in the form
I(t) =

fo

A D ( ~ -- ~') -Of q(t ) d t

g"

'

'

(20)

This integral can be approximated at the current time t as follows:


I(t) ~- [AD(~(t) - ~(At)) + A D ( ~ ( t ) -- ~(0))][q(At) -- q(0)]

+
+ lAD(0) + A D ( ~ ( t ) - 4(t - A t ) ) ] [ q ( t ) - q(t - At}]

(21)

where At is a time increment. The function ~(t) (which is given by eqn (6)) can
be evaluated by the trapezoidal rule. As the function q(t) in (21) involves the
subcell stresses S{~ ;'), S~fl~') and S]~') (13 + 7 = 2) which are as yet unknown at
time t, their values at the previous time step t - At are used. This provides a
sufficient accuracy for small values of At. Consequently, at each time step a
system of 23 linear algebraic equations is solved in the eight non-trivial
microvariables q~t~.), ~,~t~r),three composite strains g~t, g2_,, g33 and 12 subcell
stresses S]e~~'), S~2e2'') and S~t~3"~).The 23 equations are given by eqns (10) (with
i = 2), eqns (11) (with i = 3), eqns (13) (with i = 2), eqns (14) (with i = 3), eqns
(17a)-(17c), eqns (18a)-(18c), and the given values of the applied stresses ~ ~,
#22 and Oa3 at time t. Similar treatment can be applied for the case of applied
shear stresses.
APPLICATIONS
To examine the predicted response of the unidirectional composite by the
proposed micromechanical approach, extensive comparisons with the finiteelement analysis o f Schaffer and Adams 12 are given, for Hercules 3501-6

379

Non-linear viscoelastic behavior o f resin m a t r i x composites

TABLE 1
Elastic Constants of Glass and Graphite Fibers

Glass
Graphite

EA (GPa)

va

E r (GPa)

vT

86"9
221

0"22
0.2

86-9
13"8

0-22
0-25

E, and v~, denote the axial Young's modulus and Poisson's ratio, and
Er and vr are the transverse Young's modulus and Poisson's ratio.

epoxy resin reinforced by perfectly elastic glass and AS graphite fibers. The
reinforcement volume ratio is 63% unless mentioned otherwise. The
properties of the material properties of the fibers are given in Table 1. It
should be noted that the fibers are assumed to behave time-independently,
i.e. to show no viscoelastic behavior.
The initial value of the Young's modulus of the isotropic epoxy matrix is
E(t = 0 ) = 4 . 1 4 G P a and its Poisson's ratio is assumed in Ref. 12 to be time
independent, with a value of v = 0-34.
The functions go, gt, g2 and a, of the Hercules 3501-6 epoxy resin were
determined in Ref. [12] from a series of experimental results in conjunction
with the technique developed by Lou and Schapery. '~ To this end, the power
law was used for the representation of the transient creep compliance
function AD(~), i.e.
AD(~) = C~ ~

(22)

where C is a constant and n is the exponent.


The function go(a) = 1 for a < 17 MPa. For o > 17 M P a it was found to
rise linearly from go = 1 at a = I 7 M P a
to g o = l . 0 9 at a = 1 3 8 M P a .
Similarly, the function gt(o) = 1 for 0 < 17 MPa, and for o > 17 M P a it rises
linearly to gt = 3"38 at 0 = 138MPa.
The graphs of the function g2(0) and the shift factor a,, are shown in Figs 2
and 3. Cubic splines can be easily constructed to represent these functions.
In Fig. 4 the creep response of the unreinforced epoxy matrix due to an
applied uniaxial stress of a = 13-8 M P a is shown. The figure gives the
experimental response, the finite-element solution obtained in the present
situation of neat resin ~2 and the present results. The last are based on the
following values: C = 6-293/(MPa - hours"), and n = 0"26 to be used in eqn
(22). It can be seen that good agreement exists between the three cases. It
should be noted that the strain scale in the figure is expanded, and that the
initial strain e(t = 0) for the present case of a relatively low applied stress, in
which non-linearity is negligible, has the value o f 0.33%.

380

J. Aboudi
1.0

0.8

0.6
O~

0.4

0.2

20

40

60

80

I00

120

14(9

O" (MPQ)
Stress dependence of the function g2 which characterizes the non-linear behavior of
the Hercules 3501-6 epoxy resin. ~z

Fig. 2.

In Figs 5 and 6 the response of the unreinforced resin is shown for applied
stresses of 55'2 and 131 MPa, respectively. Here, the effect of non-linearity is
appreciable and the comparison between the measured and predicted
response by the finite-element procedure t2 and the present approach is fair.
In both figures the non-linear characterization functions g o, g=, g2 and a are
operative.
Having established the accuracy of the modeling of the unreinforced
matrix (which is characterized by the constants Do, C and n and the
I.O

0.8

0.6

b
o
0.4

0.2

20

40

60

80

[ " - ~
I00

120

140

O" (MPo,)

Fig. 3.

l.)
Stress dependence of the shift factor, a~,, o f the Hercules 3501-6 epoxy resin.
-

0.575 --

0.350

~V

0.525

o.3oc

,5

!
I.O

I
1.5

2.0

t (hours)

Fig. 4. Plot of strain against time of the 3501-6 epoxy resin subjected to a uniaxial stress
loading = 13-SMPa. (
) Measured response, (. . . . ) predicted response by a finiteelement method, Z2 (___) present method (e(0)= 0-33%).

1.6

1.4

. . . . . . . . . . . . . . . . . . . . .
p

1.2

1.0

I
.5

1
1.0

I
1.5

I
2.0

t (hours)

Fig. 5. Plot of strain against time of the 350l-6 epoxy resin subjected to a uniaxial stress
loading tr = 55'2 MPa. (
) Measured response, (. . . . ) predicted response by a finite-element
method, t~" ( - - - ) present method (el0)= 1.36%).

4.00 -

3.75 ~

3.50 .o
W

3.25

3.o~

I
.5

1
I.O

I
1.5

I
2.0

t (hours)

Fig. 6. Plot of strain against time of the 3501-6 epoxy resin subjected to a uniaxial stress
loading tr = 131 MPa. (
) Measured response (. . . . ) predicted response by a finite-element
method, t2 ( - - - ) present method (e(0)= 3"43%).

382

J. Aboudi

1.0

0.9

0.8

0'70

I0

20

30

40

50

t(hours)

Fig. 7. Average transverse creep strain of the glass, epoxy composite against time, due to a
transverse uniaxial stress 62, = 137.9MPa. (
) Measured values, ( . . . . ) finite-element
solution, l-" 1 - - - ) present method (g22(0)= 0.77%).

functions go, g~, g2 and a,,), we can proceed to the micromechanical


prediction of the composite creep response to applied loadings.
In Fig. 7, the overall creep response of a unidirectional glass/epoxy
composite subjected to a transverse stress of ~22 = 137.9 M P a is shown. In
the same figure, measured data and the finite-element solution ~2 are
included. It can be readily observed that the present micromechanical
prediction deviates from the finite-element solution by not more than 8%.
In Fig. 8 the predicted, finite-element and measured response are shown
for a unidirectional glass/epoxy composite subjected to a transverse stress
622 = 158.6 MPa. The response is recorded over a period of 14 h, and two
1.2
I.I

1.0

t~ 0.9
0.8
0 "70

I0

12

14

t (hours)

Fig. 8. Average transverse creep strain of the glass/epoxy composite against time, due to a
transverse uniaxial stress 622 = 158.6 MPa. (
) Measured values, ( . . . . ) finite-element
solution, - ( - - - ) present method (g.,2(0)= 0.9%).

Non-linear viscoelastic behavior of resin matrix composites

383

0,6--

0.56

0.52

I~ 0.4e
0.44

0.4

aO

60

90

120

150

t (hours)

Fig. 9. Average transverse creep strain of the glass epoxy composite against time, due to a
) Finite-clement
biaxial transverse stress loading 6., 2 = 103"4 MPa and 633 = 34.5 MPa. (
s o l u t t o n , - 1 - - - ) present prediction (gz2(O)= 0-52%t.

sets of measured data tz are shown. The reliability of the present


micromechanical prediction can be clearly seen.
Let us consider a case in which the unidirectional glass/epoxy composite is
subjected to a combined biaxial loading. In Fig. 9, the average creep response
of the composite subjected to transverse stresses 622 = 103.4 MPa and
G33 = 34-5 MPa is shown. The figure gives a comparison between the present
and finite element tz prediction. The m a x i m u m relative deviation is less than
7/0.
The effect of fiber volume ratio on the creep response of the unidirectional
glass/epoxy composite is shown in Fig. 10. Here the composite is subjected to
a transverse stress loading 622 = 137.9 MPa, and results for two values of
1.5

......

50%

I
oJ

0.9

63%

0.7

o.=

30

60
t (hours)

90

120

150

Fig. 10. Average transverse creep strain of the glass/epoxy composite for two different fiber
) Finitevolume ratios 0"63 and 0-5, due to a uniaxial transverse stress 02z = 137-9 MPa. (
element s o l u t t o n , - ( - - - ) present prediction (gz2(0)= 0.77; 1"13%).

J. Aboudi

384

0.8

. . . . . . . .

0.4--

-0.4

--

_0.8

f
2

4
t (hours)

Fig. I I. Average transverse creep strain of the glass/epoxy composite against time, due to
uniaxial transverse stress #,., = ___ 137.9 M Pa applied in 2' l-h intervals. (
) Finite-element
s o l u t i o n , l z ( _ _ _ ) present prediction.

reinforcement volume ratios, 50 and 63%, are given. The figure exhibits the
finite-element solutions t-' and the present predictions, and the agreement is
satisfactory.
So far the glass/epoxy composite was subjected to constant applied
stresses. Let us consider the case in which a cyclic loading is applied. The
response of a glass/epoxy unidirectional composite to a transverse cyclic
loading Y_,2= -+ 137-9 MPa, applied at intervals of 2.1 h, is shown in Fig. 11.
The figure gives both the present and finite-element prediction,12 and good
agreement between the two methods is clearly observed.
2.4--

2.3

j~
~

2.2

~
a.t

I,~

/'

e.o

e*

2.0

I. .

Y t
2

I
4

I
6

I
8

P
IO

f
12

I
14

t (hours)

Average transverse creep strain of the graphite/epoxy composite against time, due
to uniaxial transverse loading 52z = 186" 1 M P a . (
) Measured values, ( . . . . ) finite-element
solution, t2 ( - - - ) present prediction (g22(0)= 2-09%).

Fig. 12.

Non-linear viscoelastic behavior of resin matrLv composites

385

As a final comparison between the finite-element solution and the present


micromechanical approach, let us consider a graphite/epoxy unidirectional
composite. Here the AS graphite fibers are transversely isotropic and their
properties are given in Table 1 (the reinforcement volume ratio is 0.63). The
predicted and measured overall creep response of this type of composite is
shown in Fig. 12. The response shown in the figure is due to an applied
transverse stress ff2z = 186-1 MPa. The agreement between the numerical
and present method is excellent (the maximum relative deviation is less than
3.6%).

CONCLUSIONS
A micromechanical prediction of the non-linear behavior of unidirectional
fibrous composites is presented. The prediction relies on the knowledge of
the properties of bulk (unreinforced) constituents. The response of the
composite, obtained by the proposed micromechanics theory, is shown to
correlate well with a finite-element computation. The prediction of the
actual non-linear behavior of the composite might require the knowledge of
the in-situ behavior of the constituents, and, in particular, of the viscoelastic
resin matrix characters. If such information is not available, the use of the
proposed micromechanics equations to establish the requested properties
should be considered. A significant advantage of this approach stems from
the simplicity of its application.

ACKNOWLEDGEMENT
The author thanks Professors C. T. Herakovich and M. J. Pindera for a
fruitful discussion.

REFERENCES
1. Brinson, H. F., Durability predictions of adhesively bonded composite
structures using accelerated characterization methods. In Composite Structures,
Vol. 3, ed. I. H. Marshall, Elsevier Applied Science Publishers, London, 1985,
pp. 1-18.
2. Lockett, F. J., Nonlinear Viscoelastic Solids. Academic Press, New York, 1972.
3. Schapery, R. A., Further development of a thermodynamic constitutive theory:
stress formulation. Report No. 69-2, School of Aeronautics, Astronautics, and
Engineering Sciences, Purdue University, Lafayette, IN, 1969.

386

f. Aboudi

4. Lou, Y. C. & Schapery, R. A., Viscoelastic characterization of a nonlinear fiberreinforced plastic. J. Comp. Mat., 5 (1971) 208-34.
5. Heil, C. C., Cardon, A. H. & Brinson, H. F., The nonlinear viscoelastic response
of resin matrix composite laminates. NASA Contractor Report 3772, 1984.
6. Tuttle, M. E. & Brinson, H. F., Prediction of the long-term creep compliance of
general composite laminates. Exp. Mech., 26 (1986) 89-102.
7. Howard, C. M. & Hollaway, L., The characterization of the nonlinear
viscoelastic properties of a randomly oriented fiber/matrix composite.
Composites, 18 (1987) 317-23.
8. Xiao, X., Studies of the viscoelastic behavior of a thermoplastic resin composite.
Comp. Sci. TechnoL, 34 (1989) 163-82.
9. Brouwer, H. R., Mechanical behavior of polymeric matrix composites in biaxial
stress fields. In ICCM VI & ECCM 2, ed. F. L. Matthews, N. C. R. Buskell,
J. M. Hodgkinson & J. Morton, Vol. 4. Elsevier Applied Science, London, 1987,
pp. 4.243-51.
10. Peretz, D. & Weitsman, Y., Nonlinear viscoelastic characterization of FM-73
adhesive. J. Rheology, 26 (1982) 245-61.
I 1. Bruller, O., On the characterization of the long term behavior of polymeric
materials. Polymer Eng. Sci., 27 (1987) 144-8.
12. Schaffer, B. G. & Adams, D. F., Nonlinear viscoelastic behavior of a composite
material using a finite element micromechanical analysis. Dept. Report
UWME-DR-001-101-1, Dept. of Mech. Eng., University of Wyoming, 1980.
13. Schaffer, B. G. & Adams, D. F., Nonlinear viscoelastic analysis of a
unidirectional composite material. J. Appl. Mech., 48 (1981) 859-65.
14. Aboudi, J., Micromechanical analysis of composites by the method of cells.
Appl. Mech. Rev., 42 (1989) 193-221.
15. Sadkin, Y. & Aboudi, J., Viscoelastic behavior of thermorheologically complex
resin matrix composites. Comp. Sci. Technol., 36(4) (1989) 351-65.
16. Peretz, D. & Weitsman, Y., The nonlinear thermoviscoelastic characterization
of FM-73 adhesives. J. Rheology, 27 (1983) 97-114.
17. Aboudi, J., Elastoplasticity theory for composite materials. Solid Mech. Arch.,
11 (1986) 141-83.

You might also like