Académique Documents
Professionnel Documents
Culture Documents
343-351, 1995
~ U T T E R W 0 ~-I • H
I T'J E M A i"J rq
Copyright © 1995 Elsevier Science Limited
Printed in Great Britain. All rights reserved
(H42-1123/95/$10.00
Recent experimental efforts have established the significance of the fibre-matrix interface/interphase
in the long-term behaviour of polymeric composites. Results indicate that small alterations at the
interface level could translate into orders-of-magnitude changes in fatigue life. However, there is no
model currently available in the literature to predict these changes. In this paper, a micromechanics
model that includes the effects of the fibre-matrix interface is used in a simple cumulative damage
scheme to predict the tensile fatigue behaviour of composite laminates. A new parameter called the
'efficiency of the interface' is used to model the degradation of the interface under fatigue loading.
A rate equation that describes the changes in interfacial efficiency as a function of cycles is estimated
using experimentally determined stiffness reduction data. The influence of this interfacial efficiency
parameter on the tensile strength of unidirectional laminates is assessed using a micromechanics
model. The effect of damage on the stiffness of the laminate is estimated by solving a boundary value
problem associated with the particular damage mode (e.g. transverse matrix cracking). The fatigue
life of the laminate is estimated by considering changes in stiffness due to creep and damage in the
subcritical elements, and changes in strength associated with the critical element (0 ° ply). The
influence of a fibre-matrix interface is included in the model by considering the degradation in the
interface (interfacial efficiency) under fatigue loading. Changes in the interface property are used in
the micromechanics model to estimate changes in the in-situ tensile strength of the 0° ply. The stress
state and the strength of the 0° ply, calculated including the effects of damage, are then used in a
maximum strain failure criterion to determine the fatigue life of the laminate. Predictions from this
model are compared with experimental data. The predicted fatigue life and failure modes agree very
well with the experimental data.
The fatigue behaviour of composite materials has been materials and the prediction of the degradation of
a subject of active research in recent years. The properties under cyclic loading have received a great
damage process in laminated composites subjected to deal of attention in the last few years. Talreja 7, Allen
fatigue loading is significantly different from that et al. 8, and Joshi and Frantziskonis 9 have developed
observed in conventional materials. Four main damage models based on internal variables, to predict damage
modes have been observed in laminated composites accumulation under fatigue loading. N u m e r o u s semi-
under fatigue loading: matrix cracking, fibre-matrix empirical models have been proposed to predict the
debonding, delaminations, and fibre fracture t,2. Typi- stiffness degradation and fatigue life of laminated
cally, matrix cracking and delamination occur early in composites. Lee and Daniels l° have used experimen-
the life, while f i b r e - m a t r i x debonds and fibre fractures tally determined S - N curves for 0 ° and 90 ° laminates
initiate during the beginning of the life and accumulate to predict the property degradation and fatigue life of
rapidly towards the end, leading to final failure 3. It cross-ply laminates. Hashin and R o t e m ~l assumed that
has been observed that the stiffness of the laminate failure could be described by a macroscopic failure
reduces during the process of damage accumulation in function in terms of the applied load. The failure
laminated composites 4.5. Reifsnider and Stinchcomb 6 function contains unknown parameters, which are
have investigated the concept of using stiffness change determined from simple experiments. The failure
as a non-destructive fatigue damage parameter. In functions are dependent on the stress ratio, applied
general, they found that stiffness change can be load frequency and cycles. Charewicz and Danieis ~2
quantitatively related to the fatigue life and residual have proposed a model based on the assumption that
strength of composite laminates through various models the rate of reduction of residual strength is a function
based on the observed microdamage. of the life fraction. H a h n and Kim ~3 assumed that
The analytical modelling of damage in composite the residual strength reduction rate is inversely
343
344 S. Subramanian et el.
(Tsai-Hill or maximum stress type) to estimate the Figure 1 Schematic of the critical element model
fatigue life of unidirectional laminates. None of these
models explicitly considers the influence of the
fibre-matrix interface. reduced properties, the stress state in the critical
Recent experimental results 2°'2~ indicate that the element is calculated as a function of cycles. Reduction
region near the fibre surface, called the interphase, in the strength of the critical element as a function of
could possess elastic properties that are significantly cycles is typically estimated from the S - N curve of
different from those of the bulk matrix material. Small the 0 ° ply and from rate equations that describe any
changes in properties of the interphase could lead to time-dependent degradation such as environmental
significantly altered fatigue performance. Swain -'1 and attack. Knowing the stress state and strength of the
Subramanian et al. 2° have shown that by varying the critical element as a function of cycles, failure of the
local properties at the fibre-matrix interface level, the critical element is determined using a suitable failure
fatigue life can be altered significantly. They have also criterion (maximum strain, Tsai-Hill etc. ). The fatigue
reported varied damage modes and failure mechanisms life is estimated by calculating the number of cycles
in cross-ply laminates, which have the same fibres and or time required for the failure of the critical element.
matrix, but different fibre-matrix interphases. In order The scheme described above has been used success-
to predict the fatigue response of these composites, a fully in numerous earlier investigations 23"24 to predict
micromechanics-based model is appropriate. In this the fatigue life of laminated composites. However, in
paper, a micromechanics model is used in conjunction this study, a micromechanics model described in ref.
with the critical element scheme ~ to predict the tensile 22 is used in the above scheme to study the effects of
fatigue life of laminated composites. The predictions the interface on the tensile fatigue behaviour of cross-
from this model are compared with experimental ply laminates. This is done by varying a parameter
results reported in ref. 22. associated with the fibre-matrix interface, as a function
of cycles, and using a micromechanics model to
estimate the effect of this change on the in-situ tensile
D E S C R I P T I O N OF T H E M O D E L
strength of the 0 ° ply.
A cumulative damage scheme based on the critical It is well known that the static tensile strength of
element m o d e l proposed by Reifsnider and Stinch- unidirectional laminates is related to fibre-matrix
comb 14 is used to predict the fatigue behavior of bonding. Numerous recent experimental efforts 2-',25,26
laminates. Figure 1 shows a schematic of the critical have demonstrated this effect clearly. The influence
element model. In this model, the laminate is assumed of the fibre-matrix interface on the tensile strength
to be composed of subcritical and critical elements. of unidirectional laminates has been predicted by
The primary load-carrying members (e.g. fibres at the many investigators using micromechanics models -'7.2~.
fibre/matrix level, 0 ° ply at the lamina level) are Recently, Subramanian et al. 22 have introduced a new
normally the critical elements. The secondary, non- parameter called the efficien~;v o f the interface, 71, to
load-carrying members (e.g. matrix at the fibre/matrix predict the influence of the interface on the tensile
level, off-axis plies at the lamina level) are called the strength of unidirectional composites. They have used
subcritical elements. Non-catastrophic damage such as two parameters to describe the interface: the interracial
matrix cracking, local delaminations etc. is typically efficiency ~ and interfacial shear strength ri. Both of
associated with subcritical elements. Strength of the these parameters can be easily determined through
laminate is associated with failure of the critical simple experiments: namely, unidirectional tensile
element. The reduction in stiffness as a function of modulus measurement and single fibre fragmentation
cycles, caused by damage in the subcritical elements, tests respectively. The interfacial efficiency determines
is calculated by solving the boundary value problem how well the load is transferred from the matrix to
associated with the specific damage mode. Using the the fibre. It is defined as the ratio between the
Fatigue life of composite laminates 345
displacement in the fibre and the matrix at the interface it is advantageous to use this parameter to describe
region (7 -- Uf/Um). If the bonding is perfect and there the degradation of the interface under fatigue loading.
is efficient load transfer from the matrix to the fibre, The exact form of this degradation equation can be
7/--~ 1. If the bonding is imperfect and there is no estimated from the stiffness reduction in the 0 ° ply
load transfer from the matrix to the fibre then ~ ~ 0. of the laminate under cyclic loading. Knowing the
In the tensile strength model, the stresses in the fibre interracial degradation rate (changes in interracial
and the matrix are determined using shear lag analysis. shear strength and/or efficiency) as a function of
These stresses are used in an improved version of cycles, the influence of the interphase on tensile fatigue
Batdorf's analysis L9 to predict tensile strength. behaviour can be predicted using the critical element
Initiation of interfacial debonding is determined by model in conjunction with the micromechanics model
comparing the average shear stress at the interface described in ref. 22.
with the interfacial shear strength ri. The matrix is It must be mentioned here that the direct experimental
assumed to behave in an elastic-perfectly plastic determination of interfacial shear strength degradation
manner. The model predicts the tensile strength and as a function of cycles is difficult because most
the failure mode (elastic/plastic) of the unidirectional interfacial strength measurement tests require model
composite. Failures accompanied by interracial debond- composites. Also, there is no non-destructive testing
ing are called plastic failures while the stress-concen- technique currently available to estimate the degra-
tration-dominated failures with no interracial debond- dation of interfacial strength under fatigue loading.
ing are termed elastic failures. The micromechanics Hence the use of a parameter such as interracial
model shows that the tensile strength of a unidirectional efficiency (that can be directly estimated from the
laminate increases as 7/ reduces, when the failure is stiffness reduction data of the laminate) to describe the
elastic (when there is no interracial debonding prior interfacial degradation would be greatly advantageous.
to final failure). However, when the failure is plastic Reduction in the strength of the 0 ° ply is estimated
(accompanied by interfacial debonding), the tensile using the S - N , curve which is typically expressed in
strength reduces as T/reduces. This effect is shown in the following form:
Figure 2. The transition from elastic to plastic failure
results in a sharp increase in tensile strength.
The authors 22 have described a simple scheme ~PP = A . - B. [log(N)]",, (1)
~ttlt
to determine the efficiency parameter 7/ using the
experimentally measured tensile modulus of unidirec- where f t . , B,, P, are material constants, ~,pp is the
tional laminates in a concentric cylinders model. applied external stress, (r,t,, is the static tensile strength
Results indicate that as r/--~ 1 the modulus of the and N is the fatigue life. Note that the static tensile
unidirectional laminate approaches the value predicted strength of the 0 ° laminate is a function of the
by the rule of mixtures, and as ~--* 0 it approaches interracial efficiency parameter rl, and varies as a
the matrix modulus. Thus the interracial efficiency function of cycles. It is claimed that the constants in
manifests itself in the form of a modulus reduction in the tensile strength reduction curve of a unidirectional
unidirectional laminates. laminate are unaffected by changes in the interface.
In this paper, the interfacial efficiency is used to However, the in-situ tensile strength of the unidirec-
represent changes in the interfacial bonding condition tional laminate is influenced by the interracial bonding
in the 0 ° ply of the laminate, under tensile fatigue condition, which is characterized by two parameters:
loading. Since the interfacial efficiency is directly the interracial efficiency ~ and interracial shear strength
related to the modulus of the unidirectional laminate, "q. It is postulated that the interfacial bonding condition
changes with cycles, in the 0 ° ply of the cross-ply
laminate.
O. 60 Knowing the interfacial strength/efficiency degra-
M Weibull shape factor for the fiber
ct o Weibull location parameter for the fiber
dation rate, the in-situ tensile strength of the unidirec-
tional laminate can be estimated as a function of cycles
0.55: I ---ii..~ using the micromechanics model described in ref. 22.
This tensile strength is used in the S - N curve of the
8.50-
0/~ o
S ===== M = 7
0° ply to estimate the in-situ tensile strength of the 0°
ply and the corresponding life, N, as a function of
cycles. The influence of the interface on the tensile
fatigue behaviour is characterized using different
interfacial degradation rates in the 0 ° ply of the
0.45- laminates in the micromechanics model. In this analysis,
the degradation in interfacial strength is not considered.
It is assumed that the changes in the interface can be
O. 40- represented completely by rate equations that describe
the changes in the interfacial efficiency rt, under tensile
fatigue loading.
[ Elas~
0.3B l l l l l l l till JJlll J l l l ] l ] l , r l l ] t l l ~ l ~ l I F l i l f f l l l l l l
0.20 ~, 40 ~, 60 O. 80 1 . O0 1 • 20 PREDICTION OF TENSILE F A T I G U E LIFE OF
CROSS-PLY L A M I N A T E S WITH D I F F E R E N T
INTERPHASES
Figure 2 Variation of normalized tensile strength with the efficiency The cumulative damage scheme outlined in the previous
factor r/(ref. 22) section is used to predict the influence of an interface
346 S. Subramanian et al.
on the tensile fatigue life of cross-ply laminates. To Figure 4 shows the stiffness reduction curves for the
illustrate the application of the model, the experimental three material systems at 80% load level. The figure
results reported in ref. 20 are used. shows that damage, as measured by stiffness reduction
in the laminate, is different in these material systems.
Experimental data The 810 A and 820 A laminates reveal less damage,
Subramanian et al.22 have systematically altered the while the 810 O laminate shows significantly greater
fibre-matrix interface, and studied the effects on the damage. The failure of the 810 A and 820 A laminates
tensile fatigue behavior of (0,903)s cross-ply laminates. is sudden, while there is a significant increase in the
A brief description of the material systems used in stiffness reduction rate in the 810 O laminate prior
their study and a summary of the results is provided to final failure. Results indicate that the damage
in this section. Three material systems, designated as mechanisms and failure modes in these materials are
810 A, 820 A and 810 O, were used to study the different under tensile fatigue loading.
effects of interphase. The same Apollo fibres and HC In addition to the results reported in ref. 20, the
9106-3 toughened epoxy matrix were used in all three variation in the transverse crack density in these
material systems. However, the fibres used in the 810 laminates was also measured as a function of cycles,
A and 820 A systems received 100% and 200% and is reported in ref. 31. A master curve (independent
industry standard surface treatment respectively, and of applied load level) of transverse crack density
were sized with Bisphenol-A unreacted epoxy material. variation was obtained for the three material systems by
The fibres in the 810 O system received 100% surface multiplying the transverse crack density by (1 - ~r..... /
treatment and were sized with polyvinylpyrrolidone O'static ). These normalized data were fitted with a
(pvp), a thermoplastic material. The amount of sizing power law curve, and are shown in Figure 5. Room
material used in these systems was less than l wt%. temperature tensile creep test results on (90)12 lami-
The authors have verified the formation of different nates of the three material systems are also reported
interphases in these material systems using a perman- in ref. 31. These are shown in Figure O. The figure
ganic etching technique 3°. Some preliminary tensile indicates that the 810 O system, with a thermoplastic
fatigue life and damage analysis results for (0,903).~ sizing, exhibits the greatest creep compliance changes,
cross-ply laminates have been reported in ref. 20. and the 810 A and 820 A systems show similar creep
However. for completeness, the results have been behaviour. The data for each material system were
reproduced here. fitted with a logarithmic curve and are shown as
The S - N curves for the three material systems are dashed and continuous lines in the figure. These
shown in Figure 3. The fatigue tests were performed experimental results were used to generate the inputs
at R = 0.1 and 10 Hz frequency• The figure indicates necessary for the fatigue life prediction model.
that the fatigue lives of cross-ply laminates with altered
interphases are vastly different. Small changes in the Fatigue life prediction
interphase results in fatigue lives that are different by A schematic of the cumulative damage scheme
orders of magnitude. Increase in surface treatment (based on the critical element model) used to predict
level shifts the S - N curve to the right and increases life of (0,903)~ cross-ply laminates is shown in Figure
the slope of the curve by a small amount. The use of 7. This model could be generalized very easily for
pvp sizing material instead of Bisphenol-A epoxy shifts other laminates containing 0° plies. The model requires
the S - N curve to the right and also increases the slope rate equations that describe damage evolution in the
of the curve significantly. material as a function of cycles. These include the
transverse crack density and creep compliance vari-
ation, strength degradation in the 0° ply and the
iI'001~".
"-.- Z &%,. . . . . . _~ ._
Z ~ "- -
Ld 1""---.. ", "..
OC
F-
t
]
, ~
~
, ~..
\
,'.ooo
" ~ &90
\
o -. D
hJ
N
H
J \
F- ,i , * , , B10A -. <0.~
0')0.70. ALAA_~B100 "" E
0 # @ B20A
0 ~ 8100
0 ** ~ 820A
2"
"0-_ t
Figure 3 S - N curves for the 810 A, 810 O and 820 A cross-ply Figure 4 Variation of normalized stiffness reduction in the 810 A,
laminates 2° 810 O and 82(1 A cross-ply laminates at 80% load level-'"
Fatigue life of composite laminates 347
4
N
eo 3
e"
/
Total Stiffness / To;;IuSct~r::gth
u
Reduction
\
IIIII BI@O
O0_OO_OB2gA
Stress state
I lIllln I i i iilliq I i nlllq i ilIllll I i I ]l/Ill I I I I I]llq I J [[,ll
0 ° ply
1~ 18 2 I@ 3 I@' I@ = 1~ ~' I@'
C Y C L E S (N)
~l Failure function ]
Figure 5 Variation of normalized transverse crack density with k (Maximum Strai~
cycles in the three material systemP ~
1
! FAT,G0 ',FE !
] i ; ~)
I i
~11 1 • lf=
610 A Figure 7 Schematic of the cumulative damage scheme used to
Hlf/ L' predict damage and fatigue life of cross-ply laminates
8:7", I A
i-
Table 1 Equations for the 810 A. 820 A and 810 O material system used in the fatigue damage and life prediction mudel
Process
modelled 810 A 810 O 820 A
Crack n[1 - ((r,,,,,~ler,,,,)J = 2 (N) ...... nil - (~r,,,,,,/,r,,,)] 2 (N) ....... nil (c%.~/cru,,) [ 0.4 {A,'),,.
density vs
cycles
S - N curve e,/"' = ~(," ( 1.06 0.012 ln(N)"") e,,'.... e,,~' ( 1.06 (1.012In(N)"") E,,'.... e,," ( 1.06 0.012 In(N)" ")
of 0° ply
Efficiency 7/ = 0.95 ~ = 0.76 -- 3e 7.x rt = 0,90
vs cycles
Stiffness E~,)~'p = E,,,[1.03 (t) .......] E,,,)c'P= E.,,,[1.09 (t) .....~"1 E.,,/~P= E,,,[ 1.03 (t) ....... {
reduction in
90° ply due
to creep
n = number of transverse cracks/in; N - number of cycles: t - time (s); r/ = efliciencv factor: L,,, = unidirectional transverse stiffness (msi):
E~(( ~p = unidirectional transverse stiffness including creep effects (msi - psi × 1(]": psi lbf in :)
then used in a m a x i m u m strain failure criterion to ply discount theory. Also, the stiffness continues to
determine the fatigue life of the laminate. reduce after the initial stage, and accelerates towards
It must be pointed out here that the equations the end of the life of the laminate. This leads one to
describing crack density and creep compliance vari- believe that there is significant stiffness reduction in
ations were o b t a i n e d experimentally for these material the 0 ° ply of the laminate during fatigue loading. Since
systems, H o w e v e r , the S - N curve for the 0 ° laminate the stiffness of the 0 ° ply is directly related to the
is not available for these material systems. H e n c e a efficiency of the interface, it is claimed that there is
logarithmic e q u a t i o n that is typical of g r a p h i t e / e p o x y appreciable degradation in the interfacial efficiency
unidirectional laminates is used. The same S - N curve during tensile fatigue loading in the 810 O system.
is used for the three material systems because the Based on the stiffness reduction data, a rate equation
same fibre and matrix were used in these materials. A n of the following form is used to describe the change
additional rate e q u a t i o n that describes the interfacial in interfacial efficiency with cycles:
degradation as a function of cycles is used to represent
the role of the interphase u n d e r fatigue loading. As
m e n t i o n e d earlier, the interfacial efficiency is directly 1/ = O. 76 - e 7 I , , g l ,, ) (2)
related to the longitudinal m o d u l u s of the 0 ° ply. The equations used to represent the degradation in
H e n c e the rate e q u a t i o n representing the change in rt as a function of cycles for the three material systems
interracial efficiency as a function of cycles was are shown in F i g u r e 8. Using the rate equations
estimated from the experimentally d e t e r m i n e d stiffness described in T a b l e 1 (and discussed a b o v e ) , the tensile
reduction curves. fatigue behaviour of cross-ply laminates with different
The stiffness reduction in the 810 A and 820 A interphases is predicted. The predicted values were
laminates is small ( < 10%), and there is no significant c o m p a r e d with the experimental data reported in
stiffness reduction in these laminates after the first ref. 20.
10% of the life. This indicates that most of the stiffness
reduction is due to d a m a g e in the 90 ° ply, and there
is very little stiffness reduction in the 0 ° ply of the
I, 88-
laminates. This idea is further confirmed by the simple
ply discount t h e o r y calculations that indicate that a
total of 12% stiffness reduction will occur in the
8.88
laminate if the 90 ° ply is discounted completely. It
must be a d d e d that, since the d y n a m i c stiffness was
m e a s u r e d during fatigue tests, it is difficult to correlate
the dynamic stiffness reduction with the reduction in 0.60
<
secant m o d u l u s predicted by the shear lag model and
the ply discount theory. LJ
In addition to this, the d a m a g e analysis results 0.40 --810 A
--81@ @
r e p o r t e d in ref. 20 indicate that there is very little -- 828 A
interfacial d a m a g e in the 0 ° ply of the 810 A and 820
A laminates, while there is substantial interfacial 0.20
d e b o n d i n g in the 810 O laminate. With this in mind,
it is a s s u m e d that the interfacial efficiency does not
vary u n d e r fatigue loading in the 810 A and 820 A 0.00 J ttlH, I ; ~tlmq i I~I~HI I t i~tHlr I r Tll;rH I
1.00- 1.00-
z I
L~0.90 ~0.90
z z
w w
8c
l-
co F
\0.80 03
o1 ,,0.83.
oi o1
Ld 03
OC w * l
F
010.70 , . , , , EXPT b
~AAAAPRED (/)0.70 ***** EXPT
AAAnA PRED
0, 6 8 - - ~ l ~ ~ r , i,,,,, I L , , I , , u ~ T,;,,
10 2 1~ ~
i i TJ~Iq~T~I
1@'
Illlllq
1~ ~
I I JIMIq
10 ~
I Ib,ll.
1~ 7
D(n)= I -\
(EE,,n) ) (3)
LIFE (CYCLES)
Figure 10 Comparison of experimental and predicted fatigue lives where D(n) and E(n) are damage and stiffness after
of 820 ,k laminates n cycles, and E,, is the initial stiffness of the laminate.
350 S. Subramanian et al.
0.3( 0.40-
***** E)O~T ( BS~ ***** E~T (B6s)
..i,-- E)C'T (Bes) AAa~AE)~T (B~v3
E~OT (75s) 00000 E)~°T (75s)
0,2 ~
. _ PREP(BB~ __ PREO (B6s)
PREP(B~ _ _lED (8(}i)
PREI)( ~
~,3~ _ PREP¢75~
0.20
Ld
0
<
3-0,15 A
<I~ AI
D • a
$
0.10 . _ I ---~ --
~A 0 0
0.10~
0.Og