Professional Documents
Culture Documents
Talreja - Fatigue of Composite Materials (2003)
Talreja - Fatigue of Composite Materials (2003)
Ramesh Talreja
Department of Aerospace Engineering
Texas A&M University
College Station, Texas 77843-3141
Abstract This chapter summarizes part of the six lectures, pertaining to fatigue of
composite materials, presented at the session, "Modem Trends in Composite Lami-
nates Mechanics" at CISM in Udine. The summary provided here is of introductory
nature aimed at a reader who is not an expert in the subject. Ample references are
given to help the reader pursue the subject further.
1 Introduction
The subject of damage and failure of composite materials under static and time varying
loads is broad with many aspects ranging from the mechanisms involved to the mechanics
methods of analysis to failure criteria and life prediction. The six lectures prepared to
provide introduction and overview of the subject were organized as follows.
Lecture -1 Damage Mechanisms and Fatigue Life Diagrams- Part I
- Unidirectional Composites- PMCs
Lecture -2 Damage Mechanisms and Fatigue Life Diagrams- Part II
- Unidirectional Composites- MMCs and CMCs
- Fiber Architecture Effects
Lecture -3 Damage Mechanics
- Micromechanics vs. Continuum Damage Mechanics
- Damage Characterization
- Stiffness-Damage Relationships
Lecture -4 Damage Evolution
- Matrix Microcracking in Static and Fatigue Loadings
- Strength vs. Fracture Mechanics Approach
- A Mechanisms Based Model
Lecture -5 Damage and Failure from Notches
- Notch Stresses in Anisotropic Composites
- Failure Criteria
- Fatigue Damage From Notches
Lecture -6 Fatigue Testing and Evaluation
Cyclic Tension, Compression and Shear Testing
- Observation and Monitoring of Damage
H. Altenbach et al. (eds.), Modern Trends in Composite Laminates Mechanics
© Springer-Verlag Wien 2003
282 R. Talreja
Region I
Region III
LogN
metal fatigue, but is a mechanisms based plot of characteristic regions in the fatigue
behavior of a given composite under specific loading conditions. Figure 2 shows the
fatigue life diagram of the baseline case of a unidirectional composite under cyclic tension.
The three regions shown in Figure 2 represent the governing mechanisms within the
specified ranges of the composite strain. As explained in detail elsewhere Talreja (1981),
Talreja (1993), Talreja (2000), Region I operates within the scatter band of the composite
failure strain, etc., and the fiber failure mechanism here is non-progressive; thus the
failure can occur in any number of cycles. Region II represents the main progressive
mechanism of fiber failure occurring largely between the planes of a fiber-bridged crack
(Figure 3), but also in the parts of the volume affected by this cracking and aided by
fiber/matrix debonding, Gamstedt and Talreja (1999). Finally, Region III is the region
of damage that does not lead to failure in the preselected large number of cycles, e.g.
106 , and lies below the so-defined fatigue limit. The level of strain marked Em in the
diagram stands for the fatigue limit of the matrix material, as this limit is often also the
fatigue limit of the composite.
284 R. Talreja
Jl JUl
t~
Figure 3. Fiber-bridged matrix crack.
Region I
Fiber stiffuess
Fatigue
limit
Region III
LogN
Figure 4. Trends in the fatigue life diagram induced by constituent properties.
ten this aspect is considered in selecting the constituents, i.e. assuring that the thermal
expansion mismatch produces a compressive normal stress on the fiber/matrix interface.
The frictional sliding at the interface then forms the main irreversible mechanism that
plays a key role in fatigue of CMCs, as discussed in the following.
Before considering fatigue it is useful to ask what happens when a tensile axial load
is applied to a CMC. Let us first think of a load that is too low to cause any cracking.
Assuming the matrix failure strain is lower than the fiber failure strain, as is commonly
the case, the fibers remain intact at such a load level. If the same load is reapplied
after unloading, then the fibers would remain intact, as they would be strained to the
same value as in the first load application, assuming fibers as well as matrix are elastic
materials. This consideration suggests that the minimum condition to satisfy for fatigue
to occur is that the matrix failure strain (which is also the composite failure strain) is
reached. Once matrix cracking has occurred, several scenarios can be envisioned for fiber
failure progression after the first application of load. These scenarios are sketched in
Figure 5. The first scenario from left in Figure 5 corresponds to the case when the first
load has caused cracks that extend across the entire specimen cross section. The extent
of fiber failures in the first load will depend on the maximum strain level reached. Thus if
this strain is within the fiber failure scatter band, then the fiber failure will be extensive.
In this case the subsequent load repetition will not cause a progressive fiber failure, as a
small increment in fiber failure in any of the cross-sections with matrix cracks will lead to
286 R. Talreja
Figure 5. Three scenarios of matrix cracking and fiber failure in unidirectional CMCs.
total composite failure. This will result in the non-progressive Region I discusses above
for PMCs.
The second damage scenario inFigure 5 results when the first load level is high enough
to cause partial matrix cracks along with some fiber failures. When the load is repeated
the interfacial sliding in the debonded regions provides the irreversible mechanism re-
sponsible for stress redistribution in fibers. Thus, as the matrix cracks grow, more stress
redistribution is potentially available and progressively the fibers fail. It is likely that the
cracks become fully developed before total failure, if the load level is high enough. At
relatively low load levels, the matrix cracks may still be partially grown at the point of
unstable crack growth. The matrix cracking and the associated progressive fiber failure
form the damage of Region II.
The last scenario illustrated in Figure 5 is for the mechanism of Region III where
matrix cracks may form but their growth does not carry sufficient impetus for fibers to
fail at rates that could cause total failure in a large number of cycles. The fatigue limit
will depend on the specific value of that number; the larger the number, lower the fatigue
limit. An infinite number of cycles will correspond to the fatigue limit where no cracks
can be formed.
l
Figure 6. Cracking in a unidirectional composite under off-axis cyclic tension (left) and
failure from growth of a crack (right).
different and dramatically simple. The single progressive mechanism depicted in Figure
6 will produce a continuous curve (and the associated scatter band) in the fatigue life
diagram starting at the first-cycle failure strain and ending asymptotically in the fatigue
limit. For each off-axis angle the curve will be different. A schematic depiction of the
fatigue life diagram is shown in Figure 7, which illustrates the fatigue life dependence on
the off-axis angle. The on-axis fatigue life diagram is shown for reference in the figure in
broken lines. As Figure 7 illustrates, Region I (horizontal scatter band) of the fatigue
life diagram does not exist when the applied cyclic load is inclined to the fiber direction.
The single scatter band starts at the composite failure strain and asymptotically ends
at the fatigue limit. The failure strain as well as the fatigue limit strain depend on the
off-axis angle, e. A set of fatigue data for a glass/epoxy composite examined in Talreja
(1981) indicated that the fatigue limit strain varied from 0.6% at e = 0 to 0.12% at
e = 90. Most of the drop in the fatigue limit occurred until the off-axis angle of 30°.
Next we consider the case of two fiber orientations. In most practical cases, fibers are
placed symmetrically about the major loading axis, in which case the plies with fibers
in the two orientations will receive equal damage. The nature of this damage is different
from the case of a single fiber orientation. When a given ply cracks along its fibers, it
does not fail (separate in two pieces), as the adjacent ply it is bonded to holds it in place.
The adjacent ply cracks as well and is prevented from separation similarly. The stress
normal to the crack planes responsible for crack formation builds up over a distance called
shear lag (because the shear stress operates over this distance) and another crack can
form when the maximum normal stress reaches a critical value. This leads to multiple
cracking in the plies, as illustrated in Figure 8. When cracks in an off-axis ply are fully
developed, they span the ply thickness as well as the width of that ply along the fiber
288 R. Talreja
.. e
IIITIIIIITIT
""
~-~L.L L.L L.L L.L
(6=0)
-~
""'--~
""-~
(O<B<90)
LogN
Figure 7. Fatigue life diagram for unidirectional composites under off-axis loading. The
diagram in broken lines is for the on-axis loading case.
direction. As the cracks approach the ply interfaces, the intense stress field along the
crack fronts fails the interfaces. This ply separation, also called internal delamination,
initiates first in the regions where crack fronts of the two off-axis orientations intersect,
and progresses along the fronts of the two cracks. Figure 8 shows delamination along one
off-axis direction for clarity, leaving to visualize similar delamination in the other off-axis
direction. Failure occurs when the individual local strip-shaped delaminations widen and
coalesce with neighboring delaminations, causing large delaminated ply regions that fail
on overloading of fibers.
The fatigue life diagram for angle ply laminates have the same features as for the
off-axis loading case illustrated in Figure 7. However, the fatigue limit for angle ply
laminates shows considerable improvement with respect to the unidirectional off-axis
loading case, for off-axis loading angles lower than approximately 400, beyond which an
abrupt decrease of fatigue limit occurs, see Talreja (1981).
When the two fiber orientations are orthogonal, such as in a cross-ply laminate, and
loading is tensile along one of the fiber directions, the sequence of damage mechanisms
is as follows. First, cracks initiate in the transverse plies at locations that have defects
from manufacturing. These cracks usually span the thickness of the transverse plies
and then grow in the fiber direction, i.e., along the width of the transverse plies. This
Fatigue of Composite Materials 289
Delamination
Matrix
crack
Figure 8. Multiple matrix cracks in off-axis ply of a laminate (left) and subsequent
delamination caused by fatigue (right). Cracks and delamination are shown for one ply
only for clarity; the other ply is indicated in broken lines.
cracking process is non-interactive in the sense that the individual cracks do not influence
one another. When a few cracks have grown fully along the thickness and width of the
transverse plies, an interactive process sets in where neighboring cracks determine the
stress distribution in the ply region between the cracks, causing another crack to initiate
(midway) between them. As the applied load increases, the number of cracks per axial
length increases. When the maximum load in a cycle is repeated, the stress distribution
in the transverse plies changes only if an irreversible process exists. Akshantala and
Talreja (2000) identified this process to be frictional sliding between locally debonded
ply interfacial surfaces. Thus further transverse cracking occurs as number of load cycles
increases. Final failure occurs when the axial ply regions separated from the transverse
plies are overstressed.
The case of cross ply laminates has been studied extensively because it can be sim-
plified to a two-dimensional case for stress analysis. Various approximate analytical
solutions to the two-dimensional case can be found in Nairn and Hu (1994) where the
process of multiple cracking is also treated. The features of the fatigue life diagram are
similar to those of a unidirectional composite, illustrated in Figure 2. Region I exists for
cross ply laminates because at applied strains close to the failure strain of the composite
(which is also the failure strain of fibers), the axial plies are essentially the load bear-
ing plies, as the transverse plies at that strain would be far into their multiple cracking
process. The fatigue limit for cross ply laminates is determined by the strain to first
cracking of the transverse plies. Region II, lying between the fatigue limit and Region
I, represents the progressive mechanisms of multiple cracking in transverse plies and the
290 R. Talreja
combination of load level and crack size for the existing crack to grow in fatigue. The
fatigue threshold for metals and other materials has been the subject of extensive studies
since the early 1g70s. In most cases an experimental determination of the threshold
is relied upon since a model-based prediction has not met with much success. In a
unidirectional composite loaded in cyclic tension along fibers, the threshold for fatigue
cracking in the matrix could be the starting point, beyond which consideration must
be given to how the presence of fibers in the path of a growing crack would affect the
threshold for composite fatigue. In doing this a fracture mechanics approach has not
been successful because of the complexities brought in by fiber bridging, fiber/matrix
debonding and the local irregularity of fiber distribution. The lack of self-similarity of
crack growth, required for the crack growth analysis by fracture mechanics, renders an
effort in this direction close to being futile. The alternative way is to simply extract
a fatigue threshold variable by experimental means. The appropriate variable, in the
context of the fatigue life diagram discussed in section 2 above, is the maximum first
cycle strain below which fatigue failure would not occur in a preselected large number of
cycles. Noting that the fatigue threshold strain for the matrix material must be exceeded
for a crack to initiate in the composite, this threshold value is a good estimate for the
fatigue limit strain for the composite. Obviously, the presence of broken fibers and
other flaws in the composite will affect this estimate. It turns out that for glass/epoxy
composite the fatigue limit strain of epoxy, which was found to be 0.6% by fatigue testing
of the unreinforced epoxy, holds fairly well for the composite as well at different fiber
volume fractions, see Dharan (1g75) and Talreja (1g81).
For laminates the threshold strain for fatigue initiation depends on the ply that would
crack first. For cross ply laminates and laminates which have goo plies, the strain to
initiation of transverse cracking provides the laminate fatigue limit. If goo plies are not
present, initiation of cracking in the most off-axis plies will govern the fatigue limit.
and Hu (1994) and Hanaff-Gardin et al. (2000). This is no different from the fracture
mechanics approach to growth of a single crack where the rate of crack length increase
with cycles is correlated with a load parameter (range of stress intensity factor). An
alternative approach to the crack density increase in cross ply laminates under cyclic
loading was offered by Akshantala and Talreja (1998) where this increase was attributed
to an irreversible increase of stress in the transverse plies. The model put forward as-
sumed a frictional shear stress in the debonded 0/90 ply interface, which was responsible
for stress redistribution in the plies.
For general laminates the progressive fatigue damage process consists of multiplication
of cracking in all off-axis plies, generally at different rates depending on the constraints of
the cracking plies. The model development for this case is hindered by the unavailability
of analysis (even approximate) of the local ply stresses.
1.5 0
o Unidirectional
0.9 0 (0, ±45 )~
• ( 0, :t 45,90 )$
0.60~---~2-----+4-----!:6:----
logN
Figure 9. Multiple matrix cracks in off-axis ply of a laminate (left) and subsequent
delamination caused by fatigue (right). Cracks and delamination are shown for one ply
only for clarity; the other ply is indicated in broken lines.
the occurrence of failure in this region. As argued above, the failure is determined by
the initial conditions, which govern the fiber failures in the first application of load.
Final failure from these initial failures occurs with "random chance" without significant
progression involved. In Region II the final failure comes from regions of fiber-bridged
cracks getting connected by interfacial debonds Gamstedt and Talreja (1999), or by
a single crack growing unstably. The role of interfaces was clarified in a comparative
study of different interfaces by Gamstedt et al. (1998). An account of the interfacial
characteristics in developing the final failure criteria thus appears to be important.
4 Conclusion
These notes have summarized the salient features of fatigue damage mechanisms in com-
posite materials, while providing a rational framework for interpreting fatigue life data
and utilizing it as guidance for modeling of the fatigue behavior. Beginning 1981, when
this author first introduced the interpretative framework for fatigue of composites, called
fatigue life diagrams, a tool has existed for assessing the roles of matrix and fiber mate-
rials, and of laminate configuration, in determining the composite fatigue performance.
However, the century old field of fatigue of metals has dominated much of the thinking of
researchers and engineers, and has slowed progress, as has isotropic materials experience
been a hindrance in composites development. These notes, written for newcomers to
294 R. Talreja
the field of fatigue of composites, are yet another effort to bring the much needed new
thinking to the this field.
Bibliography
Akshantala, N.V., and Talreja, R. (2000). A Micromechanics Based Model for Predicting
Fatigue Life of Composite Laminates. Materials Science and Engineering A285:303~
313.
Akshantala, N. V., and Talreja, R. (1998). A Mechanistic Model for Fatigue Damage
Evolution in Composite Laminates. Mechanics of Materials 29:123~140.
Dharan, C.K.H. (1975). Fatigue Failure Mechanisms in a Unidirectionally Reinforced
Composite Material. In Fatigue of Composite Materials, ASTM STP 569, Philadel-
phia: American Society for Testing and Materials, 171 ~ 188.
Daniels, H.E. (1945). Proc. R. Soc. Land. A183:405~435.
Gamstedt, E.K., Berglund, L.A., and Peijs, T. (1998). Influence of interfacial strength
on micro- and macroscopic behavior of longitudinal glass fiber reinforced polypropy-
lene. In K.L. Reifsnider, D.A. Dillard, and A.H. Cardon, eds., Progress in Durability
Analysis of Composite Systems. Rotterdam: A.A. Balkema, 137~ 142.
Gamstedt, E.K., and Talreja, R. (1999). Fatigue Damage Mechanisms in Unidirectional
Carbon-Fibre-reinforced Plastics. Journal of Materials Science 34:2535~2546.
Hanaff-Gardin, C., Goupillaud, 1., and Lafarie-Frenot, M.C. (2000). Evolution of matrix
cacking in cross-ply CFRP laminates: Differences between mechanical and thermal
loadings. In A.H. Cardon, H. Fukuda, K.L. Reifsnider, and G. Verchery, eds., De-
velopments in Durability Analysis of Composite Systems. Rotterdam: A.A. Balkema,
69~76.
Nairn, J.A., and Hu, S. (1994). Matrix Microcracking. In R. Talreja, ed., Damage Me-
chanics of Composite Materials. Amsterdam: Elsevier Science, 187~243.
Rosen, B.W. (1962). AIAA Journal2:1985-1991.
Talreja, R. (1981). Fatigue of Composite Materials: Damage Mechanisms and Fatigue
Life Diagrams. Proc. R. Soc. Land. A378:461~475.
Talreja, R. (1985). Transverse Cracking and Stiffness Reduction in Composite Laminates.
J. Camp. Mats 19:355~375.
Talreja, R. (1993). Fatigue of Fiber Composites. In T.W. Chou, ed., Materials Science
and Technology, Chapter 13, Weinheim: VCH, 584-607.
Talreja, R. (1995). A Cenceptual Framework for Interpretation of MMC Fatigue. Mate-
rials Science and Engineering A200:21~28.
Talreja, R. (2000). Fatigue of Polymer Matrix Composites. In R. Talreja and J.-A.E.
Manson, eds., A. Kelly and C. Zweben, Series eds., Comperehensive Composite Ma-
terials, Vol. 2, Oxford: Elsevier, 529~552.