Anda di halaman 1dari 28

7.

FATIGUE OF COMPOSITE MATERIALS

This section is aimed at introducing the basics of fatigue testing nomenclature and
procedures. It is not by any means intended to cover all the basics of fatigue in general or for
composite either for that matter, but rather to serve as an introduction to the concept and some
initial thoughts about fatigue in composite laminates. The basics of fatigue theories and
procedures are described in textbooks such as Fuchs and Stephens (1980), Suresh (1991) and
Reifsnider (1991).

First of all, the fatigue life of classical engineering materials are difficult to predict. For
composites it is then not surprisingly even more difficult. There are several reasons for this. A
composite lamina or laminate has many failure modes and failure mechanisms. These
mechanisms will respond differently to fatigue loading. The stress distribution in a composite
lamina may be several orders of magnitude different in different directions even tough the
strains may be of the same order, due to a strong anisotropy. This in turn highlights the
problem of using some generalised stress measure, like the von Mises stress commonly used
for metals, in predicting the fatigue life.

7.1 The Stress Life Approach


The basic fatigue characteristics for an emerging material or material combination are
essential to gain confidence in the durability of the material and its structural application. The
stress life approach to fatigue was first introduced in the 1860s by Wöhler. Out of his work
evolved the concept of an “endurance” limit, which characterises the applied stress amplitude
below which a material is expected to have an infinite fatigue life. This empirical method has
found widespread use in fatigue analysis although it does not account for plastic deformation
during the cyclic loading. The basis for the creation of Wöhler curves or S/N curves is
constant amplitude testing of smooth specimens, i.e. the test specimens are cyclically loaded
between a maximum and minimum stress (or strain) level, Smax and Smin until failure occurs.
The notation is illustrated in Fig.7.1, where the mean stress and the stress amplitude are
defined as
Smax − Smin
Samp = (7.1)
2

Smax + Smin
Smean = (7.2)
2

7.1
Foundations of Fibre Composites

Smax
Samp
Smean
Smin
0
load cycle time

Figure 7.1 Nomenclature for constant stress amplitude loading

The S/N diagrams, where S is the applied stress and N is the number of load cycles can be
plotted in a logarithmic or in a semi-logarithmic diagram (generally is S on a linear scale and
N on a logarithmic) as shown in Fig.7.2. In a S/N diagram the total specimen life is plotted,
where the total life implies the number of load cycles necessary to initiate fatigue cracks in
the smooth specimens plus the number of cycles to propagate the dominant fatigue crack to
failure. The stress, S, in eqs.(7.1-2) may be replaced with strain or even a stress intensity
factor.
S

Region I Region II Region III


Low cycle fatigue High cycle fatigue Endurance limit

Figure 7.2 Typical S/N diagram with the line showing a piece-wise linear representation of the fatigue
function.

Under constant amplitude loading many engineering materials exhibit a plateau in the stress-
life plot typically beyond about 106 fatigue cycles, which also seems to be valid for the
sandwich core materials investigated herein. This load level below which the specimen may
be cycled an infinite number of times without showing any crack initiation, or propagation of
an existing crack, is called the endurance limit or threshold level. Tests performed below this
level are generally interrupted and their corresponding result representation in the S/N
diagram is accompanied by an arrow indicating a non-failed test, as illustrated in Fig.7.2. One
usually divides the S/N into three regimes. The first regime is commonly denoted the Low
cycle fatigue regime which is indicated by a high maximum stress level in the load cycle and
low number of cycles to failure. The maximum stress is usually near or above the plastic yield
stress of the material. The second regime is usually called the High cycle fatigue regime. In
this regime the log(stress) vs. log(cycles) plot commonly lies on a straight line as indicated in
Fig.7.2. This regime is commonly valid for in the regime up to 106 or 107 load cycles to
failure. The maximum stress in the load cycle is now well within the elastic regime. The final
regime is the so called Endurance limit. At fatigue stress levels below a certain value no

7.2
Fatigue of Composite Materials

failure can obtained whatever number of load cycles are applied. Typically, the number of
load cycles applied in testing to obtain this limit is in the order of 106 to 109 depending on
application. One can debate whether there is such a thing as an endurance limit and for some
materials it is argued that such a limit does not really exist. Many components are actually
designed based on the stress level near the endurance limit, particularly components that are
subjected to many load cycles in its predicted life-life, one examples is rail-way wheel axles.

The fatigue life of a material or a component may differ dramatically under constant
amplitude loads with a maintained maximum applied load but changed minimum load or
rather, changed amplitude. Therefore is the load ratio, R, introduced as
Smin
R= (7.3)
Smax

where R<0 corresponds to a load cycle with both compression and tension loading or, as in
the majority of the investigations in this thesis, positive and negative shear. The interval
0<R<1 represent tests under tension/tension loads and R>1 corresponds to
compression/compression loading. The characterisation of a new material generally involves
test series at different load ratios. The fatigue life of most materials will decrease with
increasing mean stress level and thus increasing R-value. The schematic effect of increased
load ratio is illustrated in Fig.7.3.
σ σcr

Figure 7.3 S/N diagram for differnt R-values

The time to failure may be divided into two phases, the nucleation and formation of small
cracks and then when a one or several dominating macroscopical cracks have formed, the
crack propagation phase. The major part of the fatigue life is often the damage nucleation.
Since this nucleation phase can vary significantly between, not only material types and
qualities, but even between specimens from the same material batch. Therefore the data
scatter must be considered when fatigue data is interpreted. A significant parameter in fatigue
is the statistical distribution of fatigue data.

7.2 Fatigue Life Representation


For emerging material systems and combinations, applications normally precede and drive the
development of life prediction methods. As a result, many empirical models have been used
and are still used to characterise the fatigue life of sandwich structures and the materials of
the core and faces. Some of these are Suresh (1991)

S( N ) = S cr N a (7.4)

7.3
Foundations of Fibre Composites

S( N ) = S cr − b log N (7.5)

where S is the applied maximum cyclic load, N is the number of cycles, Scr is the static
strength and a and b are material parameters. Eq.(7.4) is the classical power-law fatigue
criterion producing a linear S/N curve in a logarithmic plot while Eq.(7.5) is a linear
representation of the fatigue data in a semi-logarithmic plot. A more sophisticated model
including the fatigue threshold, Sth, is the following smooth fatigue data prediction curve
b
S( N ) = Sth + ( Scr − Sth )e − log( N / a ) (7.6)

This equation was proposed by Weibull (1951, 1961) to describe statistical variations in e.g.
mechanical testing of materials. The purpose of a curve fit is to obtain a simple representation
of the stress life behaviour, which can be used in the design process. The material parameters
are found as a best curve fit by i.e. minimising the quadratic error between the test data and
the theoretical value of eq.(7.6).

Mean Stress Effect


It is possible to plot the test results with their curve fits (or prediction curves) for different
load ratios, R, in one S/N diagram. But more generally is the effect on the fatigue life as a
function of the mean stress or the stress amplitude of interest. The mean stress effects in
fatigue can also be represented in terms of constant-life diagrams, as shown in Fig.7.4. In
these models different combinations of the stress amplitude and mean stress are plotted to
provide a constant life. Most well known among these models are those due to Goodman
(1899)
S amp S mean
= 1− , (7.7)
S fs S$

Gerber (1874),
2
S amp S 
= 1 −  mean  (7.8)
S fs  S$ 

and Soderberg (1939)

S amp S 
= 1 −  mean  (7.9)
S fs  S yield 

where Sfs is the fatigue strength (for a fixed life) for fully reversed loading (Smean=0 and R=−1)
and Syield is the yield strength of the material. The constant amplitude diagram is usually used
when only the static strength and the fatigue strength have been evaluated or are accessible.

7.4
Fatigue of Composite Materials

S amp S fs
1 Gerber
Goodman

Soderberg

S yield 1 S mean S cr

Figure 7.4 Constant life diagram

If a sufficient number of tests are conducted, a Haigh diagram as shown in Fig.7.4, can be
constructed for constant lifetime curves. Again the mean stress is represented on the x-axis
and the stress amplitude is given on the y-axis. If this type of diagram is to be fully
constructed numerous test results for different combinations of mean stresses and amplitudes
are required. When a Haigh diagram has been used in the preceding chapters a short
description of its construction is of interest.

The mean stress is a linear function of the stress amplitude for a fixed R and these ratios are
plotted in Fig.7.5 as solid lines. Actual test results may be included in a Haigh diagram but
the interpretation is made easier by using data from the curve fits (Eq.(7.6)) based on the test
results. Data points for a fixed interval, one per decade, of load cycles are then plotted in the
diagram. These data points are not actual test results but are based on the best curve fit and
are thus a good representation of the actual test results. By connecting these points with a
curve or a line a visual representation of the relation between the mean stress and the stress
amplitude is possible.
s amp
R = -0.5
Low cycle fatigue
R=0
2
10 R = 0.5
6
High cycle fatigue 10

scrit smean

Figure 7.5 Principle of Haigh diagram

The prediction curves in Figs.7.3-4 can be extrapolated to the left of the ordinate axis to
represent the effect of compressive mean stresses.

7.3 Material degradation due to Fatigue Loading


When performing fatigue tests at constant stress amplitudes the cyclic softening or hardening
of the material is accompanied by an increase or a decrease in the strain amplitude. The cyclic
softening behaviour is shown schematically in Fig.7.6. Consequently, if a constant strain
amplitude is applied, an increase or a decrease of the stress level in the material will take
place.

7.5
Foundations of Fibre Composites

σ ε

t t

(a) (b)

Figure 7.6 (a) Stress controlled loading and (b) the strain response due to cyclic softening.

The cyclic softening during the fatigue life can be expressed graphically in stress-strain loops,
as shown in Fig.7.7.
σ σ
Cyclic stress-strain curve

E
∆σ ε ε

εp εe

∆ε

(a) (b)
Figure 7.7 Schematic representation of (a) stable stress-strain hysteresis and (b) cyclic stress-strain
curve drawn through the tips of the stable loops.

Stress-strain hysteresis is particularly important for polymeric materials, the matrix material,
and less important for the fibres.

7.4 Crack Growth Approach


A material that has some sort of sharp notch or crack will behave differently under fatigue
loading. An un-notched specimen of part of a structure can usually be subjected to a certain
number of load cycles without any significant change in the behaviour. The stiffness and
strength remains unchanged. However, after a critical number of load cycles, called the
initiation time, or number of cycles to initiation, defects or cracks will form which will grow
for each load cycle. The time or number of load cycles it takes for these to grow so large that
the structure will fail is usually much shorter than the initiation time, it may only be a handful
of load cycles.

A structure containing a defect or crack with some macroscopic dimension will respond to
fatigue in a quite different manner. The defect will now rather grow, a small distance for each
load cycle. The crack propagation rate, da/dN, is usually plotted against the stress intensity
range, ∆K, where the crack propagation is divided into three phases: initiation, stable crack

7.6
Fatigue of Composite Materials

growth and unstable crack growth, as illustrated in Fig.7.8. The first part deals primarily with
non continuum failure processes, where the increment of average crack growth is less than 10-
6
mm (cycle)-1. In this regime, the stress intensity factor range approaches the fatigue crack
growth threshold, Kth. The intermediate part is also called the Paris regime and is generally
the most interesting since there is a linear relation in the log-log plot between the crack
propagation rate and the stress intensification range due to the notch. This corresponds to a
stable crack growth and this relation is written as
da
= c∆K m (7.10)
dN
where m is the slope of the curve and c is the point where an extension of the curve will
intersect with ∆K=1 MPa m. Regime III, at very high ∆K values, the fatigue crack growth
rates are significantly higher than those observed in regime II, Paris regime. A higher
sensitivity of crack growth to microstructure, load ratios and stress ratios is also noticed in
regime III.

Additionally in Fig.7.8 the effect of load ratio is shown. The enhanced influence of load ratio
is a consequence of the critical condition that the maximum stress intensity factor value for
the fatigue cycle, Kmax, approaches the fracture toughness of the material, Kc. Since the ∆K
values at which Kmax begins to approach Kc are lower for high R ratios, catastrophic fatigue
failure occurs at lower ∆K values with increased load ratio.
log da high R low R
dN

I II III

∆ K th Kc
log ∆ K

Figure 7.8 Fatigue crack growth da/dN versus stress intensity amplitude ∆K.

Pre-cracked specimens are primarily used to study the crack propagation rate in the
considered materials. These types of specimen have one or two initial cracks from which
propagation will occur. Two standard types of such specimens are shown in Fig.7.9: the
single edge notched bending specimen, SENB, and the compact tension specimen, CT. Both
are used for Mode I crack growth investigations and the test methods and the geometry of
these specimens are thoroughly described for example in ASTM standards.

7.7
Foundations of Fibre Composites

P P

W a 1.2W
a

W
4W
P
(a) (b)
Figure 7.9 (a) The single edge notched bend (SENB) and (b) the compact tension (CT) specimens for
measuring the Mode I crack propagation rate (the dashed lines indicate the crack growth direction).

For metals, the compact tension (CT) specimen, shown in Fig.7.8(b) is commonly used to
measure the fatigue crack growth in the Paris’ regime (regime II) and the test procedure
specifications are found in ASTM specification E647-93. The stress intensity factor for this
configuration, as function of crack length, is given by the relation

∆P a
∆K = ⋅ φ  (7.11)
BW 1/ 2
W

where φ is a finite width correction factor, approximately given by the relation


a
2+  a 
2 3 4
 a W a a a
φ  = 3 
0.886 + 4.64  − 13.32  + 14.72  − 5.60   (7.12)
 w  a  2  W W W  W  
1 − 
 W

For composite laminates, crack propagation is common in practice as the growth of


delaminations, which may occur in both Mode I and Mode II. The specimens used to test this
are the same as for static fracture mechanics testing described in Chapter 9.9.

A typical Paris’ law relation is shown in Fig.7.10. This one is, however, not obtained for a
classical engineering material, but for a closed cell polymer foam. The data is given for crack
propagation in both mode I (opening) and mode II (shear). As seen, the both the crack
propagation rate and slope of the Paris’ law curve is higher in mode I than in mode II.

7.8
Fatigue of Composite Materials

0.1
(a)

0.01
H100 (mode I)
C=3995
m=7.59

log (da/dN), mm/cycle


0.001

0.0001

10-5

H100 (mode II)


C=0.17
10-6 m=5.68

10-7
0.02 0.06 0.1
log ∆K, MPa √m

Figure 7.10 Paris’ law curves for crack propagation in mode I and mode II for a closed cell polymer
foam (Divinycell H100)

7.5 Fatigue of Solid Polymers


The characterization of the fatigue performance of polymers is performed in many ways done
as for metallic materials. Typical Wöhler-curves are based on constant strain or stress fatigue
testing. As seen in Fig.7.2 the curve consists of three distinct regions. The existence of a
region I is dependent on whether crazes are formed and whether these crazes causes micro
cracks to nucleate under the high loading or not. If the maximum tensile stress in the very first
cycles is not sufficiently large to form crazes, this region I will not exist. The region I will
mere be an extrapolation of the slope of the S/N curve in region II. At the high end of region
II the dominating damage mechanisms is the slow growth of crazes and their transformation
into cracks. Region III forms the endurance limit for the polymer, the fatigue life in this
region is controlled by the incubation time for the nucleation of microscopic flaws. Three
different, irreversible damage mechanisms can occur in polymeric solids; crazing, slip-band
formation, and thermal degradation. If the damage is introduced by crazing, slip-band
formation can accumulate under either monotonic or cyclic loading.

Polymeric materials undergo cyclic softening when subjected to fatigue loads as illustrated in
Fig.7.7. This is most pronounced for ductile polymers, but amorphous, semi-crystalline and
polymer-matrix composites also exhibit cyclic softening. Changes in the extent of cristallinity
mainly affect the degree and the rate of the cyclic softening.

Crazes
Crazes occur in glassy polymers that are subjected to tensile stress at low temperatures.
Above a certain stress level, striations appear in planes perpendicular to the direction of the
loading. Glassy polymers are e.g. polystyren (PS), polymethylmetacrylate (PMMA),

7.9
Foundations of Fibre Composites

polypropylen (PP). There is a continuity of material across a craze whereas the faces of a
Griffith crack in a in a brittle solid are fully separated. A craze contains fibrils of highly
orientated molecules separated by porous regions. The density of the material in the craze is
some 40-60% of that of the polymer itself. Cyclic deformation, fatigue, in many polymers is
dictated by the nucleation, growth and breakdown of crazes. Crazes will only form under
local tensile loading, cyclic softening is only observed in the tensile portion of fatigue, and
hysteresis loop remains stable in the compression portion. Craze zones ahead of a crack in a
polymer is somewhat analogous to the plastic zone ahead of a crack in a brittle solid.
Slipband
At stress levels lower than the strength of the glassy polymer, ‘plastic’ deformation can be
initiated by the formation of shear bands, or slip-bands.
Heating
A major aspect in fatigue of polymers is the question of adiabatic heating, which can lead to
failure due to thermal heating. Energy is absorbed under cyclic loading, either by a high
surrounding temperature or by internal micro-crack friction. This heating will weaken the
polymeric chains and thus reduce the resistance to deformation. At high strain rates, i.e. high
frequencies the fatigue resistance will decrease due to the internal heating as illustrated in
Fig.7.11. The fatigue resistance is further dependent on the volume to surface ratio of the
specimen, the heat loss is through the external surface of the specimen. A larger specimen
will suffer from increased internal heating and temperature raise compared with a smaller
specimen with the same cross section geometry and stress. Not only will the geometry affect
the heating, different polymeric materials have different visco-elastic damping. The fatigue
strength of a material is increased with increased visco-elastic damping.

f
stress

log N

Figure 7.11 Effect of cyclic frequency, f, on the fatigue strength.

7.6 The Fatigue Process in Composite Lamina – Loading Parallel to the Fibres
The fatigue mechanisms in composites are similar to that for metals in that they both involve
a fatigue crack initiation phase and crack propagation phase. In another sense it is very
dissimilar, containing a complicated micro-mechanical process. The fatigue process described
in this chapter is more thoroughly described by Talreja (1987).

Firstly, we should consider plotting the standard stress-life (S-N) diagram in a somewhat
different form. It appears to be much more logical to use strain versus number of cycles than
stress. Now, why is that so? There are two ways to reason; firstly, imagine two unidirectional
lamina of identical material constituents but with different fibre volume fractions. The one

7.10
Fatigue of Composite Materials

with high fibre volume fraction will have a much higher strength, but the strain to failure is
governed by fibre properties and will be (usually!) the same. The fatigue life should also then
be same in terms of strain, but not stress. If we then make a laminate with layers other than
zero plies, the strength will decrease compared to the unidirectional laminate, but since final
failure is governed by fibre fractures the strength will be different, but not the strain to failure.
A similar argumentation can be applied to fatigue.

Before actually looking at fatigue data to see what happens, let’s make a phenomenological
discussion about fatigue. The discussion is basically taken from the book by Talreja (1987)
and most of the figures and graphs are from the same book (reproduced with permission).

Fatigue at high loads, near the static strength


We all know that material fracture to some extent is statistical, i.e., the fracture stress is not a
sole given number, but varies slightly between two identical specimens. Our composite
laminates consists of a large number of fibres, each fibre having slightly different strength and
the strength may even vary somewhat along the fibre itself. At high load levels, near the static
strength of the laminate, some fibres will fail during the first load cycle, providing the stress
is beyond the strength of weakest fibre in the laminate. This will happen everywhere in the
laminate and independently of each other. At the next load application (second load cycle),
the stress state will be somewhat different due to the already failed fibres, and some more
fibres will fail, still independently of each other. The same thing will happen at the third load
cycle, and so on. One can argue that this process is unaffected by any other type of failures
occurring, like matrix cracking and that the fatigue life is totally governed by this random
fibre fracture process. When the number of fibre failure becomes large enough in a local area,
the load redistribution in such an area implies that more fibre fractures will occur due to the
stress concentration created and a crack will be formed that will grow very fast during the
next few load cycles, leading to final failure. This is schematically illustrated in Fig.7.12.

Figure 7.12 Fibre breakage in unidirectional composites under loading parallell to fibres.(reproduced
with kind permission of Prof. R. Talreja)

According to this discussion, the fatigue life of the composite can vary between a few load
cycles only (actually the specimen can fail already in the first load cycle) or at basically any
number of cycles, since the process of fibre fracture chaotic in nature and does not involve
any growth mechanism until the very last few load cycles. One can then argue that the fatigue

7.11
Foundations of Fibre Composites

life according to this mechanism short produce a horizontal scatter band in the fatigue life
diagram, as shown in Fig.7.13. The strain at this level is a scatter band around the strength of
composite, denoted εc.

Figure 7.13 Fatigue life diagram for uni-directional composites under loading parallell to
fibres.(reproduced with kind permission of Prof. R. Talreja)

The process of fibre breakage is non-progressive since there is no damage zone growing from
the early stages of fatigue loading, i.e., the damage progression cannot be traced. Due to this,
the scatter in the fibre breakage is not dependent on the fatigue life, but the apparent fatigue
life dependency in this band comes from the fact that the probability of finding a cross-section
with enough broken fibres to cause failure under the applied maximum strain increases with
the number of cycles.

Fatigue at low loads


We can continue this discussion for the case of very low loads. Assume that the strain during
fatigue is so that during the first load cycle no failures occur at all – no fibre break, there are
no crack developing in the matrix. At the second load cycle, nothing has changed, there is no
active failure mechanism, no energy dissipating mechanisms has occurred. Thus, the second
load cycle is identical to the first and there will be no failures even in the second load cycle.
For such loads, the composite will not suffer from fatigue, which is called the fatigue limit.
Actually, this load level is found, in most cases, to be the same as the fatigue limit of the
matrix material, denoted εm, and will appear in the fatigue life diagram as another horizontal
line (see Fig.7.13).

We can look at this in a slightly different manner too. The fatigue mechanism in a polymeric
matrix is similar to that of metals, involving crack propagation normal to the tensile load.
Under cyclic loading, the constrained matrix is subjected to strain controlled fatigue and if the
strain is above the failure strain the matrix cracks. At high strains, these cracks could grow,
while at low strains, they could stop at the fibre/matrix interfaces. If the load is low enough,
matrix cracks could form, but they would all stop due to the inhomogeneous nature of the
material and a fatigue limit is reached. Thus, although there may exist an energy dissipating
mechanism at the early stages of fatigue loading, it will cease to act if all matrix cracks will

7.12
Fatigue of Composite Materials

stop propagating. Then, from load cycle n to cycle n+1 there will be no more fracture surfaces
created and we are below the fatigue limit.

Fatigue at intermediate loads


Firstly we must recognise that for cracks or damage to grow during fatigue loading, there
must exist some kind of irreversible mechanism. In the case of metals, this energy dissipating
mechanism is plastic yielding, e.g. in front of a growing crack tip. Thus, for cracks and
damage to grow, the state of stress and strain in the material must change from one load cycle
to the next. If there are no such changes, no energy is dissipated and there will be no growth.

Under cyclic loading the matrix is subjected to strain controlled fatigue since the matrix is
constrained between the load bearing fibres, (remember the assumptions made when the
deriving E1 using the rule of mixtures in chapter 2.1). If this strain exceeds the fatigue (strain)
limit for the matrix (not the composite!) matrix cracks will be induced. If the strain level is
low, these matrix cracks will stop at a fibre interface, and we have reached the fatigue as
discussed above. If, however, the strain level is high these crack may continue to grow during
the next load cycle – there is damage growth, energy dissipated and the material will continue
to degrade.

In this stage, matrix cracks will form in the early stages of fatigue loading. If, however, the
strain level is high the stress intensity at the crack tip might exceed the fracture stress of the
fibre and hence this might lead to fibre failure. Now, a macro crack is formed, which will
propagate in an opening mode to the next interface. Either this fibre also will fail and the
crack will continue to propagate in an opening mode, or the shear stress at the crack tip will
cause the crack to propagate in a shearing mode in the interface between the fibre and the
surrounding matrix (see Fig.7.14). The debonding length depends on the shear strength of the
interface and is usually small, of the order of a few fibre diameters. Again, the load will be
redistributed to other fibres, which may again break, and more interfacial cracking will
appear. The material will in this manner degrade until there are so many fibre failures that the
entire laminate will fracture. Since there is slow degradation mechanism, the fatigue life of
the material will depend on the load and the number of load cycles. In this regime there will
thus be a dependence between the applied load (or strain) and the number of cycles creating
an inclined relation in the fatigue life diagram, as illustrated in Fig.7.13.

Figure 7.14 Fatigue damage mechanisms in unidirectional composite laminate (a) fibre breakage with
interfacial debonding, (b) matrix cracking and (c) interfacial shear failure.

Of course, all three types of damage mechanisms, matrix cracking, fibre fracture and interface
cracking, may occur simultaneously. However, the predominant mechanism leading to failure
would be effective in a limited range of the applied strain, i.e. the fatigue life at a specific

7.13
Foundations of Fibre Composites

strain level will be dependent of one of the three damage mechanisms. Based on this
assumption it would be expected that the fatigue-life diagram would consist of different
components, each corresponding to the underlying damage mechanism.

Fig.7.15 shows an idealised schematic of fatigue damage mechanisms in unidirectional


composites under tensile loading parallel to the fibres. The resulting fatigue-life diagram is
shown in Fig.13 where the maximum applied strain is plotted against the logarithmic of the
number of cycles to failure.

Figure 7.15 Fatigue damage mechanisms in unidirectional composite under loading parallal to fibers
(reproduced with kind permission of Prof. R. Talreja)

The damage mechanisms in tensile fatigue may be divided into three types. The scatter band
centred around the composite failure strain corresponds to fibre breakage. The sloping scatter
extending from this band to the horizontal line representing the fatigue limit corresponds to
matrix cracking and interfacial shear failure. The fatigue limit is defined as the strain
corresponding to the boundary between the non-propagating matrix cracks and the
propagating matrix cracks at 106 cycles.

Effect of composite stiffness on fatigue performance


Recall the strength prediction of composites discussed in chapter 4.3 and especially Fig.4.6.
The strength of the composite is related to the lowest of either the matrix strain at failure or
the fibre strain at failure. In most practical cases the fibre strain at failure is lower than the
matrix strain at failure. For high modulus fibres this is most certainly the case. The fatigue life
of unidirectional composites loaded in the fibre direction is also dependent on the modulus of
the fibres. Let’s study two specific cases, one having a very stiff fibre with long strain to
failure and the other being a less stiff fibre with higher strain to failure but still lower than the
matrix strain to failure. Schematic stress-strain relations of these are shown in Fig.716.

7.14
Fatigue of Composite Materials

σ σ fibre
fibre
composite
composite

matrix
matrix

ε ε
εm εc εm εc

Fig.7.16 Stress-strain relation for (a) low stiffness fibres and ( b) high stiffness fibres

We can now start to discuss the effect of these assumptions made. By plotting the strain
versus the number of cycles in the S-N diagram (S is here interpreted as the strain, rather than
stress), the effect of the fibre volume fraction should disappear, although the laminates will
have very different strengths depending on the fibre volume fraction. This is indeed seen from
Fig.7.17 where a unidirectional glass/epoxy laminate of various fibre volume fractions are
shown.

Figure 7.17 Fatigue life diagram for glass-epoxy unidirectional composite.(reproduced with kind
permission of Prof. R. Talreja)

The next issue is to discuss the effect of the different failure mechanisms. Assume first that
we have a unidirectional laminate subjected to fatigue loading parallel to the fibres and that
the fibres are very stiff and have a low strain to failure. If the fibre failure strain is lower than
the strain at which matrix cracks develop, or at least so low that matrix cracks cannot grow,
then the fatigue life is governed by the process of fibre breakage, and all data should lie
within the scatter band corresponding to this failure mode. Data for such a unidirectional
laminate are shown in Fig.7.18. The data is obtained for a high-modulus carbon fibre with a
very low strain to failure (0.5%). In this case, and the cases discussed following this, the
fatigue limit for matrix material (epoxy) is taken as 0.6% strain.

7.15
Foundations of Fibre Composites

Figure 7.18 Fatigue life diagram for high modulus carbon-epoxy unidirectional composite.(reproduced
with kind permission of Prof. R. Talreja)

Data like the one in Fig.7.18 came out rather early in the development, the data in Fig.7.18 is
from 1973 (recall that carbon fibres first were produced in a lab in 1961). Indeed all fatigue
data points appear to be on, or near the scatter band extending from quasi-static tensile
failure. As a result of such data the general opinion was that composites do not suffer from
fatigue degradation, a view sometimes still heard, and a quite misleading conclusion.

As the development of carbon fibres continued, more ductile carbon fibres were produced
with higher strain to failure. An example of a fatigue life diagram for an intermediate stiffness
carbon fibre composite is shown in Fig.7.19, a fibre with a strain to failure closer to 1%.
According to the discussion above, there should be a region of progressive damage growth
and degradation in the strain regime between the fibre failure scatter band and the fatigue
limit of the matrix. As seen in Fig.7.19 a small such region seems to exist.

Figure 7.19 Fatigue life diagram for intermediate modulus carbon-epoxy unidirectional
composite.(reproduced with kind permission of Prof. R. Talreja)

As the stain to failure for the fibre increases, the gap between the fibre fracture scatter band
and the matrix fatigue limit increases and there should be a more pronounced region of
fatigue degradation. Fatigue data for a high strain to failure carbon fibre laminate is shown in

7.16
Fatigue of Composite Materials

Fig.7.20 that seems to verify this hypothesis. The data for the glass/epoxy laminate shown in
Fig.7.17 again points to the same thing.

Figure 7.20 Fatigue life diagram for high strain to failure carbon-epoxy unidirectional
composite.(reproduced with kind permission of Prof. R. Talreja)

7.7 The Fatigue Process in Composite Lamina – Loading Off-Axis


The strength of a unidirectional composite loaded at an off-axis angle to the fibre will be
drastically reduced compared to the on axis strength. The predominate failure mode will be
matrix shear or delamination between fibres, as shown in Fig.7.21. The crack initiation in a
unidirectional laminate subjected to off-axis fatigue loading will most probably be in this
region, the interface between the matrix and the fibre. Such a crack will be subjected to a
mixed mode loading; an opening and an in-plane (parallel to the fibres) shear loading. The
opening part of the mix-mode loading will increase with the off-axis angle. The opening
mode is the more critical of the two and hence the limiting strain under which no crack
propagation will occur will decrease with decreasing off-axis angle.

There is one main difference between the off-axis case and the case of loading along the
fibres; any matrix cracks forming in the case of loading parallel to the fibres could arrest
when meeting a matrix-fibre interface. In the off-axis case there are no discontinuities in the
material that a growing crack can intercept. Thus, in the off-axis case one can assume that the
fatigue limit, still being a matrix controlled property, will be lower in the off-axis case, being
without any arrest mechanisms.

7.17
Foundations of Fibre Composites

Figure 7.21 Matrix and interfacial cracking under off-axis of unidirectional composites: (a) mixed
mode crack growth, 0 < θ < 90° (b) opening mode crack growth, θ = 90°. .(reproduced with kind
permission of Prof. R. Talreja)

In Fig.7.22 the fatigue life diagram for off-axis fatigue of unidirectional composites is shown.
At an off-axis angle of 90°, i.e. when the applied loading is perpendicular to the fibres and
hence the crack growth will be in the opening mode only. This will lead to debonding
between the fibre and matrix at a low strain and with a low fatigue limit.

Figure 7.22 Fatigue-life diagram for off-axis fatigue of unidirectional composites. Dotted lines
correspond to the fatigue-life diagram for on-axis fatigue.(reproduced with kind permission of Prof. R.
Talreja)

Once the off-axis angle decreases from 90 degrees, the matrix will grow in a mixed mode I-
mode II state, which requires a higher strain and also gives a higher fatigue limit. At
sufficiently low off-axis angle matrix cracks will have to grow almost entirely in mode II and
the crack arresting mechanism may again be present. The fatigue limit as function of off-axis
angle is shown in Fig.7.23.

7.18
Fatigue of Composite Materials

Figure 7.23 Fatigue limit as function of off-axis angle for unidirectional glass-epoxy composite.
(reproduced with kind permission of Prof. R. Talreja)

The lower limit of the curve in Fig.7.23 is usually denoted εd – fatigue limit for transverse
fibre debonding which is the limiting strain value for the formation of interface cracks in a
90-degree lamina.

7.8 The Fatigue Process in Composites Laminates


Angle-plied laminates
Angle-plied laminates consist of multiple layers of off-axis layer and thus the fracture
mechanism is the same as in off-axis laminates, i.e. fibre matrix delamination. A new
mechanism is added for angle-plied laminates, delamination between layers. The
delamination is caused by the interlaminar stresses acting between the layers. This region is
matrix dominated and thus the failure is governed by the matrix properties. There is one other
important point to consider as well; as a fibre-matrix crack in one layer has grown all across
the width of the specimen, the specimen will NOT fail completely, as it would have done for
a unidirectional laminate. The reason is simply that there is another layer in another direction
that will cross-over the crack and carry the load. Thus, one crack is not sufficient to break the
laminate. As one crack in one layer has grown completely across the laminate, other cracks
may start growing. However many of these crack will grow completely across the laminate, in
either ply, will still not break the laminate. For complete failure of the laminate, delamination
between the layers must develop. The fibre-matrix cracks in one layer create stress
concentrations that will promote delamination (compare with discussion in Chapter 6 in
delamination stresses appearing at free edges). For a [+θ/−θ]s where θ is a small angle, the
cracks will grow mainly in mode II (slow) and very large delaminations must develop prior to
complete fracture. For large angles, the reversed can be argued. For small angles, the fatigue
limit will then be close to that of the matrix itself, εm, whereas for large angles it will
approach the fatigue limit for the formation of delaminations - εd,l. An angle-ply laminate
cannot have a fatigue limit lower than that of a unidirectional off-axis laminate. The
mechanism of delamination required for failure will make angle-ply laminates much better, at
least for small angles. This is schematically shown in Fig.7.24.

7.19
Foundations of Fibre Composites

Figure 7.24 Variation of the fatigue limit with fibre angle in a symmetric angle-ply laminate of glass-
epoxy (reproduced with kind permission of Prof. R. Talreja)

Cross-ply laminates ([0/90]-type laminates)


In a cross-ply laminate ([0/90]-type laminates) the first event of failure is fibre matrix
interface debonding in the 90-degree layers. These cracks then grow toward the layer
interfaces and upon meeting them create stress concentration. These stress concentrations
create delamination in a similar way as in the angle-ply laminates, essentially reducing the
laminate to a unidirectional lamina. The difference between the cross-ply and the
unidirectional laminate is that the fatigue limit for cross-ply laminates now being governed by
the fatigue limit for delamination growth, rather than the matrix material fatigue limit.

A fatigue life diagram for a cross-ply carbon-epoxy laminate is shown in Fig.7.25. The top
scatter band corresponding to fibre breakage is essentially the same as for a unidirectional
laminate. The fatigue limit is given by the strain under which no transverse debonding cracks
are formed. The progressive damage in the 90-degree layers leading to delamination is
primarily responsible for the sloping scatter band.

Figure 7.25 Fatigue life diagram for cross-ply carbon epoxy laminate (reproduced with kind
permission of Prof. R. Talreja)

7.20
Fatigue of Composite Materials

Other laminates
The most common type of laminates consists of 0, 90 and ±45-degree layers, where the 0-
degree layers are in the main loading direction. The mechanism of damage if found to be
failure of the 90-degree plies (transverse fibre debonding) as in the cross-ply laminates, which
leads to delamination and overloading of the 0-degree fibres. The fatigue life diagram is thus
anticipated to show features as for unidirectional laminates, angle-ply and cross-ply
laminates. The upper part of the fatigue life diagram (high strains) will then show a fibre
breakage scatter band. This is followed by a sloping scatter band corresponding to matrix
cracking, interfacial fracture and delamination. The lower limit will again be governed by the
minimum strain causing delamination due to debonding in the 90-degree layers. For the
epoxy resin used in the tests presented in Figs.7.26 and 7.27, this strain was found to be
0.46%.

Figure 7.26 Fatigue life diagram for [0/±45/90]s glass epoxy laminate (reproduced with kind
permission of Prof. R. Talreja)

Figure 7.27 Fatigue life diagram for [0/45/90/-452/90/45/0]s carbon epoxy laminate (reproduced with
kind permission of Prof. R. Talreja)

Recapitulating and summarising the development of damage in composite laminates under


fatigue: In the initial stage matrix cracks will form in the layers with an off-axis lay-up (not
parallel to the principal loading). These cracks will increase with the number of load cycles

7.21
Foundations of Fibre Composites

and will eventually form macroscopic cracks. These macroscopic cracks will show a classical
crack propagation pattern. This stage is called the Characteristic Damage State, CDS, and has
found to be independent of loading history but determined by the laminate properties, i.e.
stacking sequence and ply stiffness. The matrix cracks will generally run through the ply
thickness and also the ply width. These cracks will initiate microcracks in adjacent plies. In
the ply interfaces close to the macroscopic and microscopic cracks strong interlaminar
stresses develop which lead to separation between the plies locally. At this stage the rate of
damage progression increases rapidly and will soon lead to an area where the local stresses
reaches a level above the critical and a fracture is initiated. This is illustrated in the damage
progression-time diagram in Fig.7.28. Two dominating stages can be clearly identified: the
formation of local, independent matrix micro cracks and a second stage where various types
and orientations of cracks interact with increasing rates which finally will lead to failure.

Figure 7.28. Development of damage in composite laminates under fatigue (reproduced with kind
permission of Prof. R. Talreja)

7.9 Fatigue Design Criteria


In the drive for more optimised structures and products, the development is towards lighter,
stronger and more reliable structures. Some of these requirements are in contradiction with
each other and hence there is a demand for good and reliable design criteria. Different fatigue
design methods are used depending of the application as described in e.g. (Fuchs, 1980).
Infinite-Life
A part or product designed using this approach part should not break during its entire
estimated life. The stress level applied should be so low that they are safely below the fatigue
threshold. This approach is used on components difficult to inspect or where an over-sized
design is not crucial, e.g. valves in engines. This is the oldest criteria.

7.22
Fatigue of Composite Materials

Safe-Life
This part should withstand a certain number of load cycles at a specified loading. The safe-
life design often uses safety factors based on stress-life or strain-life calculations, either the
results are taken from S/N diagrams or are results from spectrum fatigue tests which
correspond to the load conditions of the part. When a part has reached its service life, it is
replaced, whether it has failed or not. The safe life of a part has many uncertainties, such as
changes in load conditions, scatter in the test results, and variations in the materials. By
selecting large margins of safety or “safety factor” a safe operating life can be guaranteed. In
the strive for high performance and low cost this design procedure is very conservative and
maybe not optimum. The part may still fail due to initial flaws producing cracks which will
propagate under fatigue loading and invoke a premature failure.
Fail-Safe Design
The weight penalty for the high safety factors used in the safe life designs call upon a new
design criteria. In the fail-safe criteria fatigue cracks are tolerated but should not cause failure
before they are detected and can be repaired or kept under supervision. Another way of
achieving structural integrity is the usage of multiple load paths, i.e. if a part fails other parts
will carry the load. The implementation of crack stoppers is also common in parts where a
fail-safe design strategy has been used. The approach is commonly used in e.g. the aircraft
industry (wings and fuselages).
Damage Tolerant Design
The approach assumes that fatigue cracks are present and uses fracture mechanics analyses
and tests to check whether such cracks will grow large enough to cause failure before they are
detected during a periodical inspection. In order to use this very sophisticated method
designers have to evaluate all possible locations for fatigue crack onset. This is generally
performed using high accuracy finite element evaluation. Of course this procedure requires
massive resources and hence this design approach is almost only used in the aerospace
industry.

7.10 Fatigue Testing


Some specimens used for fatigue testing are described in chapter 9.8 of this text. These
specimens are unnotched and thus used for the extraction of S/N-curves. The specimen is
usually subjected to a constant amplitude sinusoidal load variation with a given R-value and
run until failure. One specimen will then provide one point for the S/N-curve. Testing at high
load values, close to the static strength of the specimen, is difficult. Tests at low load values
are very time consuming. The bulk specimens tested are usually at intermediate load values at
fatigue lives in the range of 103-106 cycles to failure.

Testing for crack growth data, or Paris’ law data, can be done using the same specimens for
fracture toughness testing, as described in chapter 9.9. The loading is again commonly
sinusoidal, with the load or the displacement controlled. The crack growth is monitored and
from and knowledge of the load (or displacement) the stress intensity factor can be calculated
and the Paris’ law curve created. In theory, the load or displacement can be varied so that the
entire Paris’ law can be obtained with just one specimen, but usually at least a few specimens
are required. Testing can be done in both mode I (e.g. with the DCB-specimen) or in mode II
(e.g. the ENF specimen).

7.23
Foundations of Fibre Composites

7.11 Fatigue Life Estimation Theories


Fatigue prediction model by Hashin and Rotem
Hashin and Rotem (1973) were one of the first to attempt to formulate a criterion for the
prediction of the fatigue life of composite laminates. It is based on classical lamination theory
and takes on a form similar to the Tsai-Hill criterion. They study a symmetric laminate (i.e. B
= 0) subjected to uni-axial in-plane loading. The global co-ordinate system for the laminate is
as usual denoted (x,y) and the local by (1,2). In a general state of plane stress, the global
stresses transform into local stresses through the use of the rotation angle θ, which is the
angle between the local (lamina fixed) and global (laminate fixed) co-ordinate systems. For a
uni-axial loading, the global stress vector can be written as

σ = (σx, σy, τxy)t = (σ, 0, 0)t (7.13)


The stress in a lamina rotated an angle θ from the global co-ordinate system is then obtained
through stress transformation to

σl = (σ cos2θ, σ sin2θ, σ cosθ sinθ)t (7.14)


From tests it has been observed that there are basically two different failure modes; for small
θ (smaller than 1 to 2°) specimens fail by cumulative fibre failure. For larger angles the
failure mode is a crack through the matrix, parallel to the fibres. By first studying the static
case one can first assume that the observed failure modes are independent. Hashin and Rotem
(1973) then suggest a failure criterion of the kind
σ 1 = σ̂ 1 and (7.15)

F (σ 2 , τ 12 ) = 1

where F is function approximated by

Aσ 22 + Bσ 2τ 12 + Cτ 122 = 1

Since the material should be insensitive to the direction and sign of the shear stress, the
constant B must be taken as zero. Then, for the criterion to be applicable for stress in
transverse loading alone (2-direction) and pure shear loading alone, the constants A and C
must equal the square of the strengths under those conditions. Next, one must also realise that
the failure stress may be different for tensile and compressive loading. By using this, the static
criterion of eq.(7.15) may be written

σ 1 = σ̂ 1t when σ1 > 0 and σ 1 = σ̂ 1c when σ1 < 0 (7.16a)

2 2 2 2
 σ2   τ 12   σ2   τ 12 
  +   = 1 when σ2 > 0 and   +   = 1 when σ2 < 0 (7.16b)
 σˆ 2t τ
  12 ˆ  σˆ 2c τ
  12 ˆ

In the case of cyclic fatigue loading we first define

σmax – maximum applied stress during the load cycle

7.24
Fatigue of Composite Materials

σmin – minimum applied stress during the load cycle

N - number of load cycles to failure


σ mzx
R= (7.17)
σ min

Hashin and Rotem (1973) then assumed that one could use the same criterion in fatigue
loading. Thus,

σ 1 = σ 1N (7.18a)

2 2
 σ2   τ 12 
 N  +  N  = 1 (7.18b)
σ 2   τ 12 

If one then writes the fatigue strength (with superscript N) for the lamina principal axes as
function of the static strength, the load ratio R and the number of cycles as

σ 1N = σˆ 1 f 1 ( R, N )

σ 2N = σˆ 2 f 2 ( R, N ) (7.19)

τ 12N = τˆ12 f 12 ( R, N )

where f are material fatigue functions, which basically describes the function for the strength
as function of number of load cycles, i.e., the Wöhler curves in the three principal stress
components. In the case the maximum and minimum cyclic stresses have the same sign, R is
positive. For R = 1, there is no cyclic loading and the static case is obtained for which
f1(1,N) = f2(1,N) = f12(1,N) = 1
We can use this in the static strength criterion to become

σˆ f ( R, N ) when σ 1 > 0


σ 1N =  1 1 (7.20)
σˆ 1 g1 ( R, N ) when σ 1 < 0

σˆ f ( R, N ) when σ 2 > 0


σ 2N =  2 2
σˆ 2 g 2 ( R, N ) when σ 2 < 0

τ 12N = τˆ12 f 12 ( R, N )

If the stresses change sign in the cycles, i.e. for R < 0, then the choice of σ̂ 1 and σ̂ 2 becomes
ambiguous since it is no longer clear whether the tensile or compressive stresses should be
chosen. Still, these parameters can be regarded as stress parameters, which are determined by
curve fitting to experimental results.

7.25
Foundations of Fibre Composites

Assume for simplicity that σmax and σmin have the same sign. We can then in analogy with the
static case find the fatigue function for the case by applying a loading in the fibre direction to
extract f1(R,N). Then we proceed to test with a pure tensile or compressive fatigue load
transverse to the fibre direction of the lamina to get f2(R,N) and finally for a pure shear load
case to get f12(R,N).

The testing necessary to implement this theory is then a set of static tests of the lamina
oriented in different direction to the fibre direction. This can appear as shown in Fig.7.29.
σ

θc θ
0 10 20 30 40 50 60 70 80 90

Figure 7.29 Static off-axis strength as function of fibre angle for a unidirectional lamina.

The cross-over point, θc, is the fibre angle for which the failure mode changes from fibre
dominated failure to matrix dominated failure. In the fatigue case we must then find the
functions f1(R,N), f2(R,N) and f12(R,N), through fatigue tests in the fibre direction, transverse
the fibre direction and for pure shear loading.

If we now proceed to discuss off-axis loading, the following can be applied; Suppose that the
specimen fails in the fibre dominated mode. Then, the applied fatigue stress can be written as
a function of the static strength of same specimen in analogy with eq.(7.18a) and (7.19) as

σ N = σˆ (θ ) ⋅ f ' ( R, N ,θ ) (7.21)

where the fatigue function f' may take on different form depending on the sign of the applied
stress as in eq.(7.20). In the case of static loading, R = 1, then we must have that
f (1, N ,θ ) = 1 .

Suppose now that the specimen fails in the a fibre dominated mode, then we must have that

σ 1N = σˆ 1 f1 ( R, N ) = σˆ cos 2 θ ⋅ f ' ( R, N ) (7.22)

where f' is the fatigue function for off-axis loading. However, following eqs.(7.21) and (7.22)
the following must hold
f 1 ( R, N ) = f ' ( R, N ,θ ) = f ' ( R, N ) (7.23)

7.26
Fatigue of Composite Materials

Next, suppose that the specimen fails in a matrix dominated mode.. We can then write

σ N = σˆ (θ ) ⋅ f ' ' ( R, N ,θ ) (7.24)

where f'' is the off-axis fatigue function for fibre dominated failure. Then the failure criterion
is that of eq.(7.18b). Therefore,
2 2 2 2
 σ2   τ 12   σˆ sin 2 θ ⋅ f ' ' ( R, N ,θ )   σˆ sin θ cosθ ⋅ f ' ' ( R, N ,θ ) 
 N  +  N  =   +   = 1
σ 2   τ 12   σˆ 2 ⋅ f 2 ( R, N )   τˆ12 ⋅ f12 ( R, N ) 

or

 sin 2 θ   sin θ cosθ  


2 2

{σˆ ⋅ f ' ' ( R, N ,θ )}  2


 +    = 1 (7.25)
 σˆ 2 ⋅ f 2 ( R, N )   τˆ12 ⋅ f12 ( R, N )  

In the static case, when R = 1, this becomes


2 2
 σˆ   σˆ 
  sin 4 θ +   sin 2 θ cos 2 θ = 1 (7.26)
 σˆ 2   τ 12 
ˆ

It then follows from eqs.(7.25) and (7.26) that


2 2
   ˆ 
{ f ' ' ( R, N ,θ )}  Nσ sin θ  +  σN sin θ cosθ  = 1
2
2 ˆ
 σ 2 ⋅ f 2 ( R, N )   τ 12 ⋅ f12 ( R, N ) 

and by substituting

σˆ 2τˆ12
σˆ =
τˆ12 sin θ + σˆ 2 in 2θ cos 2 θ
4

from eq.(7.26), the fatigue function f'' can be obtained as


2
 τˆ 
1 +  12  tan 2 θ
f ' ' ( R, N ) = f12  σˆ 2  (7.27)
2
 τˆ f 
1 +  12 12  tan 2 θ
 σˆ 2 f 2 

Equation (7.27) can now be used either to predict the fatigue strength of an off-axis specimen
through eq.(7.24), providing the functions f2 and f12 are known for the specific load ratio R, or
to find the functions f2 and f12, by means of testing off-axis specimens in at least two different
off-axis angles.

The transition from fibre dominated to matrix dominated failure may be defined by a critical
angle θc at which both criteria are valid simultaneously at some applied stress σNC. It follows
then from eqs.(7.14) and (7.16b) that

σ NC cos 2 θ c = σ 1N

7.27
Foundations of Fibre Composites

2 2
 σ NC   σ NC 
 N  sin 4 θ +  N  sin 2 θ cos 2 θ = 1
 σ2   τ 12 

Elimination of σNC in the two above equations leads to the expression

1  σ 2N   (τ N ) 2  
tan 2 θ c =    1 +  N12 N  − 1 (7.28)
2  τ 12N   σ 2 σ1  

Since σ 1N in general is much larger than both σ 2N and τ 12N , the later being of the same order
of magnitude, the squared term under the root is very small. By series expansion of the square
root, the expression becomes approximately equal to

τ 12N τˆ12 f12 ( R, N )


tan θ c ≈ = (7.29)
σ 1N σˆ 1 f1 ( R, N )

The next issue is to actually extract usable data for these models. Some data from Hashin and
Rotem are illustrated (without detail) in Fig.7.30. Note that the y-axis is drawn to different
scales.
σ σ
θ=0 θ=10
1000
θ=15
10
θ=20
θ=30
θ=5
θ=60
0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
log N log N

Figure 7.30 Tensile on-axis and off-axis fatigue failure stress vs. number of cycles. All values
extracted at R = 0.1

7.28

Anda mungkin juga menyukai