Composites Part B 166 (2019) 555–579

Contents lists available at ScienceDirect

Composites Part B
journal homepage: www.elsevier.com/locate/compositesb

The fatigue of carbon fibre reinforced plastics - A review T

∗
Parvez Alam , Dimitrios Mamalis, Colin Robert, Christophe Floreani, Conchúr M. Ó Brádaigh
School of Engineering, Institute for Materials and Processes, The University of Edinburgh, UK

A R T I C LE I N FO A B S T R A C T

Keywords: Engineering structures are often subjected to the conditions of cyclic-loading, which onsets material fatigue,
Carbon fibre detrimentally affecting the service-life and damage tolerance of components and joints. Carbon fibre reinforced
Polymer-matrix composites (PMCs) plastics (CFRP) are high-strength, low-weight composites that are gaining ubiquity in place of metals and glass
Mechanical properties fibre reinforced plastics (GFRP) not only due to their outstanding strength-to-weight properties, but also because
Fatigue
carbon fibres are relatively inert to environmental degradation and thus show potential as corrosion resistant
materials. The effects of cyclic loading on the fatigue of CFRP are detailed in several papers. As such, collating
research on CFRP fatigue into a single document is a worthwhile exercise, as it will benefit the engineering-
readership interested in designing fatigue resistant structures and components using CFRP. This review article
aims to provide the most relevant and up-to-date information on the fatigue of CFRP. The review focuses in
particular on defining fatigue and the mechanics of cyclically-loaded composites, elucidating the fatigue re-
sponse and fatigue properties of CFRP in different forms, discussing the importance of environmental factors on
the fatigue performance and service-life, and summarising the different approaches taken to modelling fatigue in
CFRP.

1. Introduction properties to suit an equally wide range of applications. The properties

of the individual fibres will depend on the specific manufacturing route
Carbon fibre reinforced plastics (CFRP) are attractive engineering undertaken, coupled to its raw material source. Carbon fibres are
materials, primarily because they have a high ratio of strength to nevertheless generally categorised according to their tensile strength
weight. Their stiffness and strength properties with respect to weight and modulus properties, Fig. 2. In this figure low modulus (LM) carbon
(specific stiffness and specific strength, respectively) are superior to a fibre tensile modulus values range between 40 and 200 GPa, while their
range of popular engineering materials including polymers, metals/ tensile strength values range between 1 and 3.5 GPa, standard modulus
metal alloys and foams. Fig. 1 shows a simplified log-log Ashby diagram (HT) carbon fibres have higher modulus and strength ranges
by which means specific stiffness and strength can be compared [1]. In (200–300 GPa and 2.5–5 GPa, respectively), intermediate modulus (IM)
this figure, the properties of CFRP with respect to weight, are close to carbon fibres have the highest strengths, ranging from 3.5 to 7 GPa (and
the apex of both the abscissa and the ordinate of the Ashby plot. The tensile modulus values between 280 and 350 GPa), high modulus (HM)
specific stiffness of CFRP at its highest is ca. 100 MN m/kg, while its carbon fibres exhibit strength ranges that are very similar to HT fibres
specific strength is close to 700 kN m/kg. Technical ceramics are of very (though their moduli are higher, 350–600 GPa), and ultra high modulus
high specific stiffness (ca. 400 MN m/kg), though they can have re- (UHM) carbon fibres typically do not exceed 4 GPa in strength, but
spectable specific strengths to CFRP, they are still only marginally exhibit a range of very high moduli from 600 to 950 GPa. In composites,
higher than CFRP. Metals and metal alloys have mid-range values if the carbon fibres are typically used in long or short filament fibre forms,
averaged, but they are high density materials and as such are not ideal as chopped or milled fibres, as fibre-mat, as woven fabrics, or as braids.
for low-weight applications.
CFRP can be manufactured in a variety of forms and can be tailored 1.1. Cyclic loading: theory and overview
for different applications. Carbon fibres have the added advantage in
that they can be manufactured from different raw material sources in- 1.1.1. Fundamentals of fatigue and cyclic loading
cluding; PAN (polyacrylonitrile) carbonisation and oil or coal pitch Fatigue is the progressive damage of a material when subjected to
carbonisation. As such, carbon fibres offer a very wide selection of repeated cyclic loading. Even if a material is loaded at stress levels well

∗
Corresponding author.
E-mail address: [email protected] (P. Alam).

https://doi.org/10.1016/j.compositesb.2019.02.016
Received 12 July 2018; Received in revised form 23 December 2018; Accepted 9 February 2019
Available online 19 February 2019
1359-8368/ © 2019 Elsevier Ltd. All rights reserved.
P. Alam, et al. Composites Part B 166 (2019) 555–579

damage (second stage) which in turn gives rise to damage through

micro-cracking (third stage). At this stage, the likelihood of crack nu-
cleation increases as a function of the stress intensity, KI . Critical factors
influencing KI include topography (of surfaces or interfaces), voids,
inclusions and defects. Topographical characteristics such as roughness
are important as they are directly related to the intensity and density of
existing stress concentrations. Contrarily, smooth surfaces increase the
time required for cracks to nucleate. The subsequent stage of fatigue
damage is characterised by crack growth, which occurs in a steady
manner to a critical size. On reaching a critical size, the crack propa-
gates catastrophically through the material as the remaining materials
is unable to sustain any further imposed load.
The most common fatigue testing regimen involves cyclic loading
between two predetermined levels of stress. This is termed a constant
amplitude stress and is illustrated in Fig. 4. The first form of fatigue
loading within a constant amplitude stress involves the complete re-
versal of cycling (tension-compression loading) as is shown in Fig. 4 (a).
Here, the sinusoidal loading wave traverses from a tensile loading mode
to a compressive loading mode, the maximum and minimum stresses
being equal. Fig. 4(b) depicts repeated cyclic loading (zero-to-tension)
Fig. 1. Simplified log-log Ashby diagram showing the specific stiffness against where a material carrying no load is subjected to a load, which is
specific strength for different engineering materials. subsequently removed such that the material returns to zero-load
condition in each cycle. Pure tensile cyclic loading (tension-tension
loading) is shown in Fig. 4(c) where the stress is tensile during the
entire cycle. Fig. 4(d) shows random or irregular stress cycling, which is
a function of variable amplitudes during loading. This is perhaps the
least typical mode of testing cyclic loading, and yields the least pre-
dictable fatigue lifetime of any material.
The stress range, Δσ = σmax − σmin , is the difference between the
maximum and the minimum stresses imposed on a material in fatigue.
Averaging the maximum and minimum values results in the mean
stress, σm . Half the stress range is called the stress amplitude, σa , which
is essentially the variation about the mean. Mathematical expressions
Δσ σ −σ σ +σ
for these basic definitions are: σa = 2 = max 2 min , σm = max 2 min . The
maximum, σmax , and minimum σmin stress values of the stress amplitude
can be expressed as σmax = σm + σa and σmin = σm − σa . The signs of σa
and Δσ are always positive, since σmax > σmin , where tension is con-
sidered positive. The quantities σmax , σmin and σm can be either positive
or negative. Ratios such as the stress ratio R and the amplitude ratio A
σ σ
are commonly used and expressed as: R = σmin , and A = σ a , respec-
max m
tively. A constant fatigue life (CFL) diagram [4] is usually used to the
Figure 2. Strength-modulus ranges for different types of carbon fibres graded study the effect of R in the fatigue life of a material. A common pro-
by their mechanical properties. LM - low modulus, HT - standard modulus, IM - cedure involves the use of stress/life information to predict the ex-
intermediate modulus, HM - high modulus and UHM - ultra high modulus. pected life (or expected stress for a particular probability of failure) of a
material by plotting the mean stress against the alternating stress. A
below the elastic limit, under the conditions of continuous cyclic widely used illustrative means of estimating the influence of the mean
loading, microscopic damage occurs. This micro-damage accumulates stress on the fatigue strength of a material is through a Goodman dia-
throughout the material and can grow steadily into macro-cracks, or, gram [5], Fig. 5. This diagram plots stress amplitude against mean
may cause macro-scale damage leading to the ultimate failure of the stress with the fatigue limit (endurance limit) and the ultimate tensile
material. Mechanical loading, thermal gradients, chemical ingress and strength of the material as the two extremes. The Goodman relationship
environmental conditions all contribute to the rate at which fatigue (
is expressed as σa = σfat 1 −
σm
σuts ), where σfat is the fatigue limit for
damage occurs in materials. As such, fatigue is a complex process as the completely reversed loading and σuts is the ultimate tensile strength of
lifetime of a material under cyclic loading can be affected by many the material. The area below the straight line indicates that the material
parameters in tandem. should not fail at given stresses while the area above represents the
The fatigue life (N) of a component is defined as the total number of possibility of failure. For any given mean stress, the fatigue (endurance)
stress cycles required to cause material failure. The onset of material limit can be evaluated directly as the ordinate of the lifeline at the
failure can be subdivided into distinct stages as shown in Fig. 3. In the specific value of σm . As the mean stress of a fatigue cycle is increased,
first stage of fatigue failure; microcracks and open cracks build up the number of cycles to failure and the endurance limit (if existing)
within the material. This leads to crack nucleation and minor levels of decreases. At low stress i.e. high-cycle-fatigue, the fatigue life increases
significantly. However, for polymer composites the fatigue mechanism
is rather complex and more accurate predictions of the effect of mean
stress on fatigue strength are needed. To this effect, a modified
Goodman curve proposed by Boller [6] replaces σuts by the flexural
strength of the FRP at a time corresponding to that of the cycle life at
Fig. 3. The different stages of material damage during fatigue. the x-axis intercept.

556
P. Alam, et al. Composites Part B 166 (2019) 555–579

Fig. 4. Constant amplitude cycling types: (a) completely reversed cycle of stress (b) repeated cyclic loading (c) purely tensile cycles and (d) random stress cycles.

Fatigue analyses are not always based on the stress response, but
can also be based on the number of loading cycles needed to create a
macro-crack. A distinction can therefore be made between low-cycle
fatigue (LCF) and high-cycle fatigue (HCF). LCF relates to loading
conditions where the stresses may be high enough to induce irrecov-
erable damage within the material. The transition from low-cycle fa-
tigue to high-cycle fatigue depends on the properties of the material,
but for many materials, it is typically within the range 102 to 104 cycles
[7,8]. The physical rationale is that in the case of HCF, the stresses are
sufficiently low that the material remains within its limit of elastic
proportionality. Under LCF, the stress range may surpass this limit of
elastic proportionality and is more prone to macro-cracking and/or
premature failure. As a general rule, when the magnitude of stress is
increased, the fatigue life decreases. Fatigue tests are usually re-
presented using an S − N or ε − N diagram, where S is the stress, N is
the number of cycles to failure and ε is strain. Due to the large varia-
Fig. 5. Schematic example of a Goodman diagram. tions of N, the abscissa is usually presented in the form of log (N ) . The
S − N curve in the low-cycle fatigue region (Fig. 6) is often described
by a power law, known as the Coffin-Manson law [2,3], which is
Δϵp Δϵ
mathematically represented as: 2 = ϵ′f (2N )c where p is the plastic
2
strain amplitude, ϵ′f is an empirical constant (the fatigue ductility
coefficient - i.e. the failure strain for a single reversal), 2N is the number
of reversals to failure (N cycles) and c is an empirical constant (the
fatigue ductility exponent). In the high-cycle region of stresses (Fig. 6),
the elastic component of strain is usually described as the elastic stress
amplitude, which follows a power law function of the fatigue life
Δϵe σ ′f
Δϵe
(Basquin law) [9]: 2
= E
(2N )b . Here, is the elastic strain ampli-
2
σ ′f
tude, is the fatigue strength coefficient and b is a fatigue strength
E
exponent. An important feature is that the fatigue life is expected to
increase with an increase of σ ′f and a decrease of b. It should be fur-
thermore noted that the fatigue life point (2Nt ) in the S-N curve tran-
ϵ′f E 1/ b − c
sitions as: 2Nt = ( ) ). The total strain amplitude is the sum of both
σ ′f
Fig. 6. Elastic and plastic strain-life curves to produce total strain-life, ε-N, Δϵ Δϵp Δϵ
elastic and plastic strain: 2 = 2 + 2 e . It is worth noting that the
curve of a material.
Coffin-Manson linear damage rule for low cycle fatigue accounts for the

557
P. Alam, et al. Composites Part B 166 (2019) 555–579

plasticity of the material, while Basquin's law is a strain dependent law. was plotted against the number of cycles thus yielding an ε − N curve
If for example, the strain amplitude causes sufficient material plasticity, (as shown in Fig. 9). This was deemed necessary as the composite fa-
the lifetime of the material will be short, resulting in an LCF. If how- tigue limit can be governed to a large extent by the matrix fatigue limit,
ever, the stresses are close enough to zero, we can assume that the and that failure at the first cycle occurs when the composite strain
strains are elastic and that the lifetime of the material will be long equals the failure strain of fibres, irrespective of the fibre volume
(HCF). fraction of the CFRP. The significance of plotting the maximum strain in
the first cycle lies in that this strain value provides a good reference
1.1.2. Mechanics of cyclically loaded composites point in mapping the extent of damage from the very first cycle, and
Composites have gained considerable popularity over the last five that subsequent damage (and thus the fatigue life) will depend on this
decades as they can be tailored to have outstanding mechanical, damage state. The three distinctive regions in this ε − N curve (Fig. 9)
thermal, physical and electrical properties [10–14]. The fatigue beha- can be associated with different modes of damage. In particular, region-
viour of composite materials and structures is complex since composites 1 represents static-like damage, where the stress applied is high enough
fail by a series of, or collection of damage mechanisms including; fibre to instigate the early onset of fibre breakage. Region-2 is a mid-level
breakage, matrix cracking, fibre-matrix debonding, delamination [15] state of stress representing fracture initiation and the progressive pro-
and the effect of shear-induced diffuse damage on transverse cracks pagation and coalescence of small-scale cracks in fibres and at inter-
[16,17]. Fibre fracture is highly dependent on fibre strength and matrix faces. This eventuates in debonding and ultimately, composite failure.
cracking can be retarded somewhat, if the reinforced fibres are of high Region-3 is the region of no fatigue failure (for a specific number of
stiffness (such as carbon). This is because high stiffness and strength cycles), located just below the fatigue limit of the matrix where the
fibres can endure higher loads, which limits the strain in the system, applied stress is too low to result in crack propagation.
and thus that of the matrix (under a given load). Debonding can occur if Fibres are the main load bearers in composites and they occupy
the interface between the fibre and matrix is weakly bonded, which is 30 %− 70% of the total volume. It is expected, for a given type of fibre
itself a function of (a) wetting (b) fibre topography and (c) the physical (e.g. carbon) and a matrix material that the stress-strain relationship
chemistry at fibre-matrix interfaces. If a composite is composed of alters when the fibre volume fraction and/or the fibre properties (in
multiple layers (ply-laminated) a ply-delamination failure mode may terms of stiffness) change. Commercially available carbon fibres with
occur through fatigue loading. The fatigue performance of composites is different stiffness and strain failure ranging from ca. 0.5% up to ca. 2.1%
influenced by the composite system itself. Particularly influential can significantly influence the shape of ε − N curve, and consequently,
parameters include; fibre and matrix properties, lay-up sequence the fatigue life of the composite. The stiffness and failure strain in the
[18,19], residual stresses due to the manufacturing process or due to fibre direction of a composite are primarily determined by the fibre
discontinuities [20,21] and the maximum to minimum ratio of stress properties. As described above, region-1 consists of the horizontally
endured by the material [4,22–24], ply orientation [4], the fibre/matrix area of scatter related to the composite failure strain, εc . This strain
fractions [25], and environmental conditions affecting the composite limit is a function of the static strain to failure of the fibres. It worth
components [26–28]. Exemplary works on the fatigue of composites noting that if the composite failure strain is less than or equal to the
covered from both experimental [22,29] and theoretical [4,30–32] matrix strain failure, then region-2 can be omitted. The materials se-
perspectives may assist in the development of improved methods by lection of fibres dramatically affects region-2 of the ε − N curve. More
which means composite failure can be predicted. specifically, at an equal magnitude of fibre strain, HM fibres experience
higher stresses than LM fibres. Hence, lower fatigue degradation is
1.2. Objectives expected in composites containing HM fibres. Considering that fatigue
failure cannot occur in composites until cracks are initiated in the
This review aims to familiarise CFRP materials specialists with the matrix, the use of stiffer fibres makes sense in that they can retard crack
latest research on CFRP, cyclic-loading and fatigue. To begin, we will growth more effectively.
describe the underlying fundamentals of cyclic-loading, how it differs Konur and Matthews [99] detailed the influence of the carbon fibre
from static loading conditions, and we will follow this with a section on type (with respect to modulus) on the fatigue properties of UD-CFRP.
the mechanical theory of fibre reinforced composites subjected to They summarised the S-N behaviour of composites made from different
cyclic-loads. We will then proceed to describe fundamental research on carbon fibre types; high modulus (HM), high strength (HS) and low
the fatigue properties of cyclically-loaded CFRP drawing upon issues modulus (LM), Fig. 7. It is clear from the stress-life curves shown in
such as the effects of fibre volume fraction and dimensions, fibre sizing, Fig. 7, that HM fibres in UD-CFRP are superior in fatigue in view of
fibre orientation and properties. The fatigue properties of CFRP will strength retention as a function of loading cycles [100]. The difference
then be compared against those of other materials, with an aim of in the damage mechanism between HM, HS and LM fibre UD-CFRP has
drawing benefits and disadvantages in using CFRP for different appli- also been reported [101]. HM UD-CFRP fail catastrophically and in an
cations. Since several areas within which CFRP is applied will be sub- explosive manner, progressive failure is observed in HS UD-CFRP
ject to extreme temperature variations, submersion and aging, we will
continue with a section on the effects of heat, water-ingress and
polymer aging on the fatigue properties of CFRP. The different types of
damage with respect to different conditions and CFRP types will then be
elucidated, after which we will collate the different approaches used to
model and predict the fatigue response of CFRP.

2. Cyclically loaded CFRP

2.1. Effect of fibre type on the fatigue life

In the case of composites, where two (or more) constituents are

involved the usefulness of an S-N curve may be questioned as the role of
each constituent in the fatigue life of the composite needs to be de-
termined individually. Talreja [35], proposed a new framework for the Fig. 7. Normalised stress plotted against the number of cycles for different fibre
interpretation of fatigue life of composites, where strain and not stress types in T-T fatigue. Figure inspired by the work of [99].

558
P. Alam, et al. Composites Part B 166 (2019) 555–579

Fig. 8. S-N curves of CF/PEEK for T-T (left) and T-C (right) loadings. From Ref. [33], reproduced with the permission of Elsevier.

samples, and LM UD-CFRP failure is stepwise and progressive (more so exhibited widespread debonding, crack propagation and matrix failure.
than HS). The relationship between fibre stiffness and the failure mode The fatigue life diagrams (FLD) [35] of both CF/epoxy and CF/PEEK
can be explained in terms of the energy stored. Explosive failure occurs UD systems are shown in Fig. 9. This leads to a quasi-infinite lifespan,
because there is a large amount of stored energy and this is released though over time, molecules and interfaces will undoubtedly be af-
simultaneously once the matrix fractures. Progressive failure, contra- fected by the constancy of induced loads and the composite may begin
rily, is a result of stress localisation and the gradual stepwise release of to lose stiffness [34]. Considerably more scatter can be observed in the
pockets of stored energy. CF/epoxy composites when compared against the CF/PEEK composites,
Fig. 10. The properties of CF/PEEK composites nevertheless degrade
2.2. Thermoset and thermoplastic matrix CFRP more swiftly as can be evidenced by the steeper slope in region-2. One
reason for why scatter is often noticeably high in fatigue samples could
Tai et al. [33] investigated the fatigue behaviour of PEEK based relate to the mechanisms of micro-damage. These are essentially related
CFRP under tension-tension (T-T) R = 0.1, and tension-compression (T- to factors such as defects (which may arise during the manufacturing
C) R = −0.1 fatigue, using a quasi isotropic-composite configuration. process), and the homogeneity of the strength of adhesion at fibre-
Fig. 8 from their paper indicates that samples tested in T-C are more matrix interfaces. When cyclically loaded, micro-damage will pre-
sensitive to fatigue loading than the T-T ones. This is understood to be ferentially initiate from weak spots along fibre matrix interfaces, and in
because the T-C mode of loading is considerably more damaging to places where defects increase local strain energies. Defects present at
composite materials than T-T [33]. Gamstedt and Talreja [34] com- interfaces including voids and weakly bonded spots, as well as other
pared the fatigue behaviour of CF/epoxy (thermoset matrix) to CF/ defective geometrical arrangements such as fibre misalignment guide
PEEK (thermoplastic matrix) composites aligned unidirectionally the way in which strain energies and stress intensities are distributed
(R = 0.1 T-T at 10 Hz). They noted that the thermoset composite (CF/ (and thus how they propagate) throughout a fatigue loading cycle [39].
epoxy) was more resistant to fatigue than CF/PEEK, as only a few mi- By superposing both FLD of CF/epoxy and CF/PEEK UD systems, the
crocracks were observed and crack propagation towards the interface authors of [35] inferred that the more ductile PEEK matrix exacerbates
was gradual. Contrarily, the thermoplastic composite (CF/PEEK) the rate at which microscale damage can occur, leading to a more rapid
rupture process of fibres, as represented by the steeper slope of region-
2. Additionally, region-3 damage, which should be moderate at most, is
more prominent in CF/epoxy composites. While the fatigue behaviour
is governed by the initiation of cracks and their subsequent develop-
ment (in the case of CF/epoxy samples), the fatigue properties of CF/
PEEK were notably different. In these composites, macroscopic damage
is governed by longitudinal splitting, [36]. To better comprehend why
there are clear differences between thermoset and thermoplastic matrix
CFRPs in fatigue, we will focus on how these different matrices affect
crack initiation and development. Comparing the fracture toughness of
each is often seen as a good starting point [37]. Hojo et al. investigated
the relationships between delamination, fatigue growth and matrix
toughness, using PEEK matrix and epoxy matrix CFRPs as archetypal
materials of comparison [37]. They found the fracture toughness
(static) to be considerably higher in the thermoplastic (PEEK) compo-
sites as compared to the thermoset (epoxy) composites. They related
this to the higher ductility of the PEEK as compared to the epoxy. Fibres
have an additional effect of locally constraining the matrix, which de-
creases the expanse of plastic zone formation, decreasing thus the rate
at which fatigue fractures are able to propagate through the matrix
Fig. 9. Fatigue life diagram of longitudinal composites in tension-tension fa-
component of composites [38].
tigue. Figure inspired by the work of [34].

559
P. Alam, et al. Composites Part B 166 (2019) 555–579

Fig. 10. Fatigue life diagram of CF/epoxy and CF/PEEK. Figure from Ref. [34], reproduced with the permission of Springer-Nature.

2.3. The effects of fibre volume fraction and fibre orientation micro-cracks. Hysteresis loop area is directly correlated to the dissipa-
tion of kinetic mechanical energy by the sample. This converts to
The properties of strength and stiffness in composites are usually thermal energy through molecular friction within the material. Higher
improved by fibre presence. The fibre volume fraction (FVF) is there- frequency loads cause greater levels of molecular friction, which in turn
fore a critical parameter when considering the fatigue properties of results in the release of more heat. Heat within CFRP can accrue over
composites. Hiremath et al. [40] modelled the effect of fibre volume time and is sometimes responsible for premature composite failure.
fraction on damage accumulation within the matrix of an epoxy-matrix Gornet for example [44], when investigating the influence of stacking
CFRP subjected to fatigue stress. They concluded that damage accu- sequences (in epoxy matrix CFRP) on self-heating, reported that heat
mulation within the matrix of a composite, decreases as the fibre vo- was a cause for early fatigue failure. The temperature elevation was
lume fraction increases. Brunbauer and Pinter [41] conducted an ex- reported as being higher for angle-ply laminates [+45/-45/+45/-45]
tensive experimental study to discern the effect of mean stress and FVF as compared to cross-ply laminates [0/90/0/90], and greater levels of
(30% and 55% ) on damage mechanisms in epoxy matrix UD-CFRP. The shearing was the result. This is because the visco-plastic energy released
materials were tested under both static and cyclic conditions in both T- is proportionally much higher in angle-ply laminates as compared to
T and T-C modes and oriented at 0°, 45° and 90°. A schematic summary cross-ply laminates (Fig. 12). The influence of frequency on the fatigue
of their research is provided in Fig. 11. The slopes of the S-N curves of CFRP is well described through the work of Barron et al. [46]. In
were steepest for the T-C fatigue tests. Under static loading, a higher their paper, it is noted that kinetic energy generated through fibre-
FVF (solid lines) resulted in strength improvements in both tension and matrix shearing during fatigue, will convert to heat energy within the
compression for all fibre orientations. In fatigue, when comparing composite. This heat dissipates between the fibres and matrix primarily
samples at 30% (dashed lines) and 55% FVF (solid lines), the authors through shearing, and for this reason ± 45∘ oriented CFRP are generally
found that the FVF was influential in T-T fatigue over the high cycle more affected by frequency than [0∘] oriented (UD) CFRP [46]. ± 45∘
fatigue regime for specimens oriented at 90°. The authors postulate that oriented CFRP experience greater magntiudes of deformational shear
this behaviour is due to a change in the mechanism governing com- between fibre and matrix than UD CFRP and as such convert more ki-
posite damage as it changes from fibre pull-outs at higher stress levels, netic energy into heat energy, which in turn reduces the stiffness of the
to matrix failure at lower stress levels (in 30% FVF samples) [42]. In T- matrix material. At low frequency fatigue, the convection of heat from
C, the S-N trend varies little with respect to FVF. The UD (0°) composite the sample to air may enable the matrix to stay well below its glass
damage mechanisms changed from fibre pull-outs and single fibre transition temperature. However, under high frequency fatigue more
fracture in T-T mode, to the breakage of entire fibre bundles in T-C heat is produced within the composite and the temperature of the
mode. This is deemed to be due to buckling and consecutive fibre matrix may rise above its glass transition temperature, which in turn
bundle breakage initiated by poor stress transfer. The 45° specimens softens the matrix and greatly reduces stress transfer between matrix
exhibited both matrix failure and fibre pull out damage mechanisms in and fibre, reducing thus the lifespan of the composite [47]. Montesano
T-T. In T-C, localised buckling led to fibre-crushing and the breakage of [48] evidenced that heat production in epoxy matrix CFRP fatigued
fibre bundles. over 7000 loading cycles is inhomogeneous. The samples were loaded
from 40% to 80% of the UTS and an IR (Infra Red) camera used to
2.4. Heat dissipation in CFRP monitor heat dissipation. At 80% of the UTS, a higher crack density was
noted to occur in localised areas, which is presumed to have been the
It is well known that heat affects material deformation under the instigator of failure (Fig. 13). The self-heating vs fatigue stress study
conditions of long-term loading [43]. During fatigue, heat is produced [45] on a PA6 matrix CFRP comprising 2 stacking sequences, [0]8 and
as a function of material kinematics. The build up of heat in CFRP under [45]8, revealed that under higher loading cycles, the temperature of
cyclic loads is a point of importance as heat can affect the mechanical CFRP rises more swiftly and this leads to exacerbated damage (Fig. 14).
behaviour, and can also instigate chemical changes to the polymer This is because local heat increases the mobility of polymeric chains
matrix. Thermography can be used to monitor heat build up in CFRP, and reduces the stress transfer between the fibres and matrix, allowing
[35,44,45]. When cyclically loading CFRP under high loads and fre- thus, greater straining and more pronounced damage of the CFRP. In-
quencies, hysteresis loops can be observed to increase in area. This terestingly, the stacking sequence is as important in thermoplastic
change in hysteresis is due to local stress release and the creation of matrix CFRP as it is in thermoset matrix CFRP [45]8 sequence samples

560
P. Alam, et al. Composites Part B 166 (2019) 555–579

Figure 11. Schematic summary of investigated mechanical behaviour and damage mechanisms in epoxy based UD-CFRP depending of the FVF, the fibres orientation,
the fatigue profile (T-T, R = 0.1 and T-C, R = −1) and the applied mean stress. From Ref. [41], reproduced with the permission of Elsevier.

being more affected by heat induced visco-plasticity than [0]8, once to untreated composites under high imposed cyclic stresses. At lower
again because these composites experience more internal shear. cyclic stresses, fibre treatment was noted to have a reduced effect on the
fatigue properties of CFRP.

2.5. The effects of fibre sizing, properties and dimension

2.6. Short fibre CFRP composites in fatigue
The strength of interfacial adhesion can be attributed to; adsorption
and wetting, inter-diffusion, electrostatic attraction, chemical bonding, Despite their superior properties, carbon fibres are not usually used
and mechanical adhesion. Interfacial properties affect the fatigue be- as short fibres (unlike glass fibres), since they are more expensive. As
haviour of CFRP [50]. Good adhesion at the fibre-matrix interface such, their target areas are typically higher performance structures
maximises the homogeneity of stress transfer from between the fibres where long fibre reinforcements are needed [57]. Fsuchiyama's [53]
and matrix. Logically fibre sizing, which influences the strength of the research on randomly oriented CF (T300S, average diameter = 7 μm)
interface, will also influence the fatigue lifetime of CFRP. Broyles et al. focusing on fibre length (in a polyester matrix), revealed that the
[51] considered the influence of different types of fibre sizing on the maximum flexural strength of CFRP occurred at a carbon fibre length of
tension-compression fatigue of carbon fibre/vinyl ester UD composites 25.4 mm. Hitchen et al. [55] further demonstrated that the tensile
manufactured by resin infusion. Two dissimilar sizing agents were used properties of short fibre epoxy-matrix CFRP were unaffected by the
in the research; poly(vinyl pyrrolidone) (PVP) a brittle thermoplastic, length of carbon fibres up to 5 mm within an epoxy matrix. In their
and phenoxy resin (polyhydroxyether) a ductile thermoplastic. Both work, they found there was a significant reduction of strength from
were compared to unsized CFRP composites and both sizings were 5 mm to 1 mm carbon fibre reinforcements. Caprino [54] attributed the
shown to have an important effect on the fatigue properties of CFRP. An difference in the fibre length threshold to both the dissimilar loading
increase of 60% in the fatigue life was reported when phenoxy sized modes (T-C vs T-T) and the specific matrices used (polyester vs epoxy).
fibres were used, and an improvement of 20% of the fatigue life re- Generally, in static tension, mechanical behaviour is governed by the
ported for the PVP sized fibres. Under similar loading conditions reinforcement, even though the length of the fibres may be short (up to
(207 MPa), the lifespan of phenoxy sized CFRP composites was found to 3 mm) [54–57]. Fig. 16 shows how the CF aspect ratio influences the
be 20 times greater than unsized CFRP composites, Fig. 15. Deng et al. fatigue properties of carbon short-fibre reinforced plastics (CSFRP). As
[52] reported similar behaviour when researching 0° and 90° UD-CFRP the fibres are shortened, load transfer is less effective, which results in
under tension-tension fatigue. In their study, treated carbon fibre pre- the onset of early cracking. A 10,000 fold reduction in the lifespan of
preg composites showed an improvement in fatigue life when compared CSFRP under similar loading conditions has been reported for 1 mm

561
P. Alam, et al. Composites Part B 166 (2019) 555–579

Fig. 14. Explanation of fatigue limit detection: block of loading and self-heating
curves. From Ref. [45], reproduced with the permission of Elsevier.

Fig. 12. Experimental self-heating curves for epoxy based CFRP (a) [+45/-
45]2s, (b) [0/90]2s, (c) [+45/-45/90/0]s. From Ref. [44], reproduced with the
permission of Elsevier.

Fig. 15. S-N curves of UD carbon fibre/vinyl ester composites with and without
sizing in tension-compression. From Ref. [51], modified with the permission of
Elsevier.

their manufacture through injection moulding is suitable for high ton-

nage, low cost production. Mandell et al. investigated both GSFRP and
CSFRP using various thermoplastic matrices, as well as neat PA-66
Nylon, polyamide-imide (PAI), polycarbonate (PC), polyphenylene
sulfide (PPS), and polysulfone (PSUL) [59]. CSFRP with brittle matrices
are more resilient to high-cycle fatigue as compared to GSFRP. This
Fig. 13. IR temperature distribution after stabilisation for different fatigue tendency is nonetheless, less apparent when more ductile, thermo-
stress levels. From Ref. [48], reproduced with the permission of Elsevier. plastic matrices are used. The authors postulate that the cracked regions
are more resilient to fracture in brittle matrix composites than in ductile
long carbon fibres as compared to 15 mm long carbon fibres [49]. matrix composites, when subjected to low cycle fatigue. This can be
In contrast to short fibre reinforced thermosets, short fibre re- backed up using CSFRP thermosets as an example, as they are far more
inforced thermoplastics have received considerable attention [58] since resilient than CSFRP thermoplastics under high-cycle fatigue. Recent

562
P. Alam, et al. Composites Part B 166 (2019) 555–579

temperature of the curing cycle [62,67]. The correlation between the

appearance of high levels of residual stresses/strains and the fatigue
performance of composites (CFRPs) has been reported by many
[68,86–89]. These researchers have found that the fatigue lifetime of
CFRP decreases with a higher presence of residual stresses/strains
within the composite. Other critical factors such as under-curing or
over-curing, cure-induced voids as well as thermal degradations due to
temperature overshoots, should be taken into account during the
manufacture of CFRP [88,90,91] as these can also reduce the service
life of the composite.
Fibre waviness and crimping are other problematic defects arising
through the manufacturing process. These defects affect the quality of
consolidation and thence the fatigue life [70,71,88,92–94]. Out-of-
plane waviness affects the static compressive strength and stiffness of
composites, and decreases its fatigue life (in C-C) [95–97]. The fatigue
life of axially loaded CFRP with induced out-of-plane fibre waviness,
gives rise to a severe degradation of the fatigue life properties of CFRP
Fig. 16. Effect of the fibre length on the S-N curve of random short fibre
in (T-T) and (C-C) loading [70,98].
carbon/epoxy composites. R = 0.1. From Ref. [54] using data from Ref. [55],
reproduced with the permission of Elsevier.
2.8. Effect of pre-existing damage on the fatigue properties

efforts in recycling carbon fibre has renewed a certain interest in the Post-impact damage fatigue can be a critical determinant of the
field of CSFRP for structural applications [60], as SFRP are an appro- materials selection and design of CFRP components, joints and parts
priate material to make from recycled fibres and plastics. [102]. The fatigue properties offered by CFRP, if lost through impact
damage, may negate the utility of this material in applications where
2.7. Effect of processing/manufacturing defects on the fatigue properties impact and fatigue are inherently coupled. There are various ways in
which impact will affect the structural integrity of a composite. This
The manufacturing of CFRP is somewhat complex as there are many will rely on various factors including; the materials (both fibre and
control parameters that require fine-tuning to improve the final quality matrix) making up the composite [103,104], the orientation and con-
of the composite material. To claim that the manufacture of defect-free figuration of composite laminates [105], the number of laminates, the
composites is extremely difficult is by no means an exaggeration. There thickness of laminates [106], and the impacting material: its hardness,
are manufacturing challenges to be met if the fatigue properties of CFRP mass and velocity at the point of impact.
composites are to reach a desirable theoretical level. Types of defects Nevertheless, there are certain concepts that can be taken as good
that have been identified as arising through the manufacturing process starting points in the consideration of post-impact fatigue. Following
of CFRP include; voids, curing, environmentally induced defects, and Nettles et al. [107], we can broadly say that there is an implicit re-
fibre waviness or crimped tows. These are considered amongst the more lationship between damage size, residual strength in compression and
significant manufacturing induced defects that affect the structural the number of loading cycles to failure in compression. Compression
performance and thus, the fatigue life of CFRP [61–71]. tends to result in more prominent post-impact fatigue damage evolution
Void formation is possibly the most common manufacturing defect as compared to tension [108] due to the repeated buckling of the
and is inherent to most composite manufacturing techniques including composite encouraging its delamination behaviour at the site of impact-
resin-transfer moulding (RTM), sometimes vacuum assisted (VARTM), damage [112]. Cantwell et al. [108] researched epoxy matrix CFRP laid
both of which are well described in the literature [61–64,72,73]. Air up as [+45, −45,0,0, −45, +45,0,0]s using both non-woven and mixed-
entrapment within a composite can occur during the early stages of woven layers, and found that the residual tensile strength of post-im-
manufacturing as air bubbles can become trapped in the matrix (as a pact fatigued CFRP actually improved, whereas in compression it de-
viscous liquid resin) during the impregnation and consolidation stages, creased. The effect was more apparent in non-wovens and less so in
or through poor wetting of the filaments. The resultant formation of mixed-woven CFRP. Since they also noted that the majority of damage
voids can detrimentally reduce the intra-laminar and inter-laminar occurred between [+/ −45] laminates, much of the resistance to loading
properties of the final composite. Additionally, volatile components or post-impact, falls upon the [0] oriented laminates, which are inherently
contaminants can create voids by vaporisition during the thermal cycle suited to tensile resistance over compressive resistance as they are
[74]. In polymer composites, voids have been shown to act as stress geometrically unstable in compression, while have optimal stress
concentrators and as a consequence, increase the likelihood of dela- transfer in tension. The constant life model a = f (1 − m)u (c + m) v
mination [61,64–67,73,75,76], adversely affecting the compressive predicts the alternating stress as a function of the mean stress for
strength [64,77,78], the shear strength [65,66,74,79,80], the flexural composites in fatigue. Here, a is the normalised alternating stress, m is
properties [62,67,81] and the fatigue properties [82,83] of the com- the normalised mean stress, c is the normalised compressive strength,
posite. S-N curves reveal that voids shorten the fatigue life of a com- the exponents u and v characterise the shape of the tensile and com-
posite, and increase the rate at which the properties of the composite pressive wings of the bell curve, and f is a function of the composite
degrade from cycle to cycle (in both T-T and C-C) [69,81,83–85]. laminate tensile strength. In post-impact fatigue, this model reveals a
Curing is a complex thermo-mechanical process with several asso- distinct shift in the shape of the compressive part of the bell curve,
ciated variables that may influence composite performance, both in the further clarifying the importance in defining the type of loading that a
long and short terms. During the manufacturing process of CFRP the composite will experience [110]. Certainly, it would seem that com-
resin will undergo a curing reaction and it will transform into a solid. pression-compression fatigue is the most damaging to post-impacted
This transformation can cause shrinkage of the resin and this in turn composites. Comparing post impact damaged composites under ten-
introduces internal residual stresses/strains within the composite. sion-compression (T-C) and compression-compression (C-C) fatigue,
Whereas, shrinkage is dependent on the physical properties of the Mitrovic and co-workers reported almost no difference in T-C damage
components and their reactions, the creation of residual stresses/strains mode between the undamaged and impact-damaged composites,
within a composite is ultimately a function of the duration and whereas C-C fatigue of impact-damaged composites at high stress levels

563
P. Alam, et al. Composites Part B 166 (2019) 555–579

(70–80% ) revealed extensive damage evolution [111]. Melin and Schon can of course alter the extent to which a projectile will cause damage.
[112] further stipulated the post-impact crack line follows the buckle in Tai et al. [122] for example, studying quasi-isotropic T300/976 carbon/
the CFRP material during compressive fatigue, while Ogasawara and epoxy composites, report that stiffness losses in T-T post-impact fatigue
co-workers [113] (using high strength T800S/3900-2B epoxy matrix are alleviated by increasing the composite thickness. This in turn in-
CFRP) reported that in thick (32 ply) quasi-isotropic laminates creases material costs and weight, which is unhelpful when weight
[+45,0, −45, +90,0,0]4s , that post-impact compressive fatigue advances saving is a primary goal in using CFRP.
damage, but primarily within the boundaries of the impact-damaged
region. 2.9. Effect of fibre architecture on the fatigue properties - 2D CFRP
Cantwell and co-workers [108] also found that fibre architecture composites
had a dramatic impact on the evolution of post-impact damage during
fatigue. More specifically, they reported that mixed-wovens exhibited The use of UD laminates is not practical in a plethora of real-world
less post-impact damage evolution as compared to non-wovens, which engineering composite structures. Textile composites offer higher
displayed a considerable growth from the original damaged region drapability [123], which is needed in the manufacture of curved
during fatigue. Saito and Kimpara [114] researching multi-axial stiched structures. Textiles also are also easier to handle, are consequently less
CFRP (VARTM manufactured) concluded that the meeting of post-im- laborious to work with, and automation with this material is fairly well
pact transverse cracks with cracks propagating during fatigue, gives rise established, resulting in high volumes of production. Textile composites
to stress relaxation in the composite. Stress relaxation is particularly also possess higher interlaminar fracture toughness properties than
important when environmental factors such as water and temperature, straight fibre composites. However, the fibre waviness in textiles leads
influence CFRP during post-impact fatigue [115,116]. The impact da- to the development of localised stress concentrations on loading. In the
maged area allows water to ingress more effectively by creating open case of woven composites, the static tensile strength and Young's
areas for water to wick along fibre-matrix interfaces, and moreover modulus is 15–25% and 20%, respectively, when compared against
increases the area available for water diffusion. In both cases, the im- equivalent fibre-dominated UD composites [124]. Contrarily when the
pact area becomes more compliant and this coupled with stress re- composite layups are matrix-dominated the modulus does not reduce
laxation through interconnecting cracks during fatigue, results in con- and there is in fact an observed increase of 10–15% in the static tensile
siderably reduced capacity for post-impact fatigue load-bearing, with strength when compared to UD composites [124]. This is due to fibre-
larger variances between dry and wet reported in multi-axial (T700S- overlap, which reduces in-plane shear deformations and delamination.
12K, VARTM) over plain-woven (T300B-3K, VARTM) composites In compression, the presence of crimp in textile fibres encourages the
[115]. formation of kink bands, leading to a reduction in static strength when
The type of material used to make up the composite also plays a compared against equivalent UD composites [125].
critical role in the growth and propagation of post-impact damage. Uda The fatigue response and damage mechanisms observed in textile
and co-workers [117] for example, report that impact induced dela- composites are different to those of UD composites. In woven compo-
minations are greater in epoxy matrix CFRP (UT500/Epoxy) than they sites tested in on-axis (T-T) fatigue [126–128], cracks first develop in
are in thermoplastic matrix CFRP (AS4/PEEK). This is presumably due the weft yarns, perpendicular to the loading direction. As these cracks
to the greater polymer compliance in thermoplastics as compared to grow, breakage of isolated axial fibres occurs, followed by meta-dela-
highly cross-linked thermosets. This is one of the primary reasons for minations (i.e. decohesion of warp and weft fibre bundles), which
why thermoplastic phases are considered judicious additions to ther- eventually join at ply boundaries to form interlaminar delaminations.
moset matrix CFRPs as a means to improving their impact and post- Final failure occurs abruptly when large numbers of axial fibres break at
impact performance [103]. Multi-wall carbon nanotubes (CNT) are also the same time. In the case of (T-C) fatigue [124,126,128], delamina-
considered a means to redistribution of impact/post-impact energies tions and transverse cracks develop simultaneously, making fibre
within a composite and have been detailed by Ref. [118]. Using quasi- bundles weaker in buckling. Final failure tends to occur during the
isotropic [0, +45,90, −45]2s CFRP laminates, they reported a marginally compression load of the (T-C) fatigue cycle [128]. This is because the
higher load bearing capacity for CNT infused CFRP over a compression- fibre bundles in external plies buckle from a lack of adjacent ply sup-
compression (R = 10 at 10 Hz) fatigue life, though it might be less in- port, while internal plies experience fibre bundle kinking, which in turn
dustrially applicable once the CNT costs are taken into consideration. leads to final failure. These mechanisms lead to out of plane deforma-
Gemometrical parameters [119], alongside material and loading tions and accelerates damage propagation in (T-C) fatigue as compared
parameters are also necessary considerations in post-impact fatigue to (T-T) fatigue. This is evidenced by the steeper S-N curve observed in
analyses. The energy of impact can have a critical effect on the fatigue (T-C) fatigue [124]. Very little has been reported on the off-axis fatigue
life of a composite. Symons and Davis [120] report a decade decrease in testing of woven composites [124,129,130]. Nonetheless, it has been
the fatigue life of T300/914 quasi-isotropic layup carbon fibre/epoxy shown in 45° bias fatigue, that damage initiates in the composites
when impact pre-fatigue impact energies were increased from 5J to 10J through matrix cracking in resin rich regions [124]. This is followed by
(T-C cycling at R = −1). Importantly, they note only slight variations in the formation of meta-delaminations, which subsequently coalesce, and
the damping ratio, Δ between 0.65 and 0.7 for their composites be- lead to relative displacements between warp and weft bundles. The
tween 10 cycles and 65,500 cycles. This indicates that these composites final failure is caused by bulk delamination between the warp and weft
retain much of their brittle character throughout the entire loading bundles. Matrix dominated damage leads to very high strains to failure
u
cycle. The damping ratio here is calculated as Δ = u a where ua is the (up to 20%) [130]. Kawai et al. [129] compared the fatigue properties
m
energy absorbed per cycle, ua = ∫ σdε , and um is the maximum strain of plain woven carbon/epoxy laminates at 0°,15°, 30° and 45° from the
1 1
energy stored per cycle, um = 2 σmax εmax + 2 σmin εmin . Symons and Davis warp fibre direction. The on-axis testing exhibits a straight line on a log-
also reported considerable post-impact damage evolution under fully log S-N curve up to 106 cycles, and shows no apparent fatigue limit to
reversed T-C cycling, more prominently in the case of higher pre-fatigue within that number of cycles. Off-axis laminates exhibit an S shape on a
impact energy. This is a different outcome to that reported by Mitrovic log-log S-N plot with an increase in slope in the intermediate range of
et al. [111] who researched considerably lower energy impacts (up to fatigue life (10 4 cycles), which shows a greater sensitivity to fatigue
ca. 2J) clarifying thus, the importance of recognising the extent of than the on-axis tested composites. However, in the high cycle range
impact damage prior to fatigue on damage evolution. Higher velocity (>10 4 cycles), the S-N curve flattens, demonstrating the existence of a
impacts, and thus impact energies, will cause more damage in com- fatigue limit. Cracks in woven composites can arise in the weft direction
posites [121]. Altering the geometries of impacted/fatigued materials even at just 10% of tensile strength [131]. The fatigue failure me-
chanisms of woven composites are highly complex as they vary

564
P. Alam, et al. Composites Part B 166 (2019) 555–579

depending not only on the type of loading and direction, but also on the a loss of modulus and strength in quasi static tension, bending, and
layup. compression of 3D composites as compared to their 2D counterparts. A
Knitted composites are an alternative to woven composites as they loss in the fatigue strength and fatigue life has also been reported [142]
have enhanced drapability and a higher impact resistance when com- in many cases, with some notable exceptions, e.g. Aymerich and Priolo
pared against other textile-based composites. They can also allow the working with [02/90]s graphite based CFRP with Kevlar z-reinforce-
manufacturing of composites with certain features such as holes already ment in (T-T) fatigue [143]. Mouritz and Cox claims that, if the initial
incorporated whilst maintaining continuous fibres. Unfortunately, there monotonic strength difference is discarded by normalising the S-N
are very few papers in the literature that deal with their fatigue prop- curves, the trend is then similar in some cases. This so called, knock-
erties. Pandita et al. [132] compared the fatigue properties of plain down-behaviour, was demonstrated in compression for a [0,90]s stitched
woven and knitted fabric composites under (T-T) fatigue. Initial failure CFRP [144], as well as for [±30, 90]s flexure z-pinned CFRP [145].
occurs in the form of bundle debonds. In the locations where the fibres Exceptions can be noted, such as [+45/ −45/0/90]s quasi-isotropic stit-
are off-axis to the loading direction, the meta-delaminations were the ched CFRP in tension [144], where stitched composites performed
damage modes that were found to propagate fastest. This leads to a better at high normalised fatigue stresses, and unidirectional (UD) z-
realignment of the fibres and hence a slower subsequent growth of the pinned CFRP in tension, where the normalised fatigue life was de-
debond. In some cases the debond jumped between two fibre bundles creased by increasing the pin content and diameter [146].
creating continuous matrix cracks. The final failure occurred through As expected, the fatigue properties of 3D composites differ from 2D
fibre pull out. When the S-N curves were normalised by the ultimate composites and these depend on several parameters, one of which is the
tensile stress, the knitted composites were shown to perform similarly type of reinforcement used in the z-direction. In this section, we will
to woven laminates in the on-axis direction, though much better in the focus on each of the following reinforcement types: 3D woven, 3D
bias (45°) direction. A stiffness reduction of around 10% was observed stitched, z-pinned and 3D braided.
for knitted composites regardless of the loading direction, which was 3D woven composites are standard fabrics with fibres running in the
similar to that observed for the on-axis woven specimen, but was much x-(warp) and y-(weft) in-plane directions, with a z-yarn binding the
lower than the 50% degradation observed during the 45° off-axis plain fabric layers. There are two main types of 3D woven composites: in-
woven fatigue test. The same trend was observed for the residual terlock fabric, where the z-binder is interlaced with the in-plane yarns,
strength during fatigue. and orthogonal fabric, where the z-binder holds in place the entire stack
In similtude to knitted composites, very little literature can be found of 2D fabrics, thus avoiding crimp and maximising the in-plane stiffness
on the fatigue properties of braided composites. Three different braid [139]. Carvelli et al. [147] showed that there is a dependency of these
angles (25°, 30° and 45°) were studied by Kelkar et al. [133,134]. They composites on the loading direction (when comparing 2D and 3D wo-
reported that the strength and stiffness of braided laminates decreases vens) [0,90]s orthogonal CFRP in both warp and weft directions and
as a function of increasing braid angle. Yet, the fatigue properties tested in (T-T) fatigue. The fatigue performance of the 2D samples were
(normalised by the ultimate tensile strength) were also noted to as in between weft (lower) and warp (higher), although the difference was
being independent of the braid angle. A high endurance limit of 40–50% marginal, especially when considering the propensity for scatter when
of the ultimate tensile strength was found, although only one million it comes to fatigue of CFRP. Rudov-Clark and Mouritz investigated the
loading cycles were applied and as such, a fatigue test with a higher influence of volume content of z-binder yarns on the fatigue properties
number of cycles would need to be carried out to confirm that this of 3D orthogonal woven CFRP in (T-T) [148]. Adding z-binder reduced
represents a fatigue limit. A sudden final failure mode, with no visible the normalised load cycles-to-failure, which indicates that the 2D CFRP
cracks or delamination of the plies was observed which would be pro- composites have better fatigue properties. They postulate that this de-
blematic as it would be difficult to detect damage progression. Mon- terioration was caused by the fibre arrangement, the formation of resin-
tesano et al. [135] studied the fatigue of triaxially braided polyimide rich channels, and through fibre damage due to the weaving process.
composites at room temperature and at 225°C. They found that al- 3D Stitched fabrics are manufactured by sewing a yarn through
though the static properties are not really impacted by the temperature multiple 2D fabrics layers. Sewing methods vary from the use of simple
increase, there is a reduction in fatigue life at elevated temperatures. orthogonal needles to robotic machines with needles working at dif-
Even though the higher temperature led to matrix softening (thus mi- ferent angles simultaneously [149]. The microstructures created by the
tigating the initial dominant damage mechanism of braided yarn stitching process influence the monotonic mechanical properties, and
cracking), the crack density of the composite was noticeably higher at the final 3D stitched composites sometimes display better, sometimes
the elevated temperature, leading to a shorter lifetime. worse moduli and failure strengths as compared to 2D composites
[139,142,149,150]. There is nevertheless, little doubt that the dela-
2.10. Effect of fibre architecture on the fatigue properties - 3D CFRP mination toughness [151] and impact resistance [152] always improve
composites through stitching. In fatigue, stitching can be beneficial [143], detri-
mental [142] or neither beneficial nor detrimental, depending on the
Composite mechanical properties are usually tailored in 2D via influence of the stitch on properties such as stiffness and strength. In 3D
stacking sequences, depending on the properties required by creating woven composites, resin rich regions close to the stitches introduce
the structure (e.g. shear: ± 45∘, longitudinal: UD 0°, uniform: quasi- weak spots in the composites. The fatigue properties are also influenced
isotropic). Nevertheless, 2D reinforced composites are prone to dela- by fibre related issues intrinsic to the manufacturing route such as tow
mination due to their relatively poor interlaminar properties. This is waviness, crimping and fibre damage/breakage [153]. In (C-C) fatigue,
especially true in fatigue. 3D composites may be used to tackle this the misalignment of tows accelerates micro-buckling and kinking
intrinsic weakness of 2D composites [136,137]. Industrially, the use of [154], thereby reducing the fatigue life of stitched 3D composites. In (T-
a 3D preform greatly simplifies the out-of-autoclave (RTM) manu- T) fatigue, tow waviness and fibre damage/fracture is responsible for a
facturing cycle, saving time and money [138]. A through-thickness reduced fatigue life. In both cases of (T-T) and (C-C) fatigue, the initial
fibre reinforcement, such as 3D woven, stitched, tufted, z-anchor or z- in-plane strength to failure and stiffness knockdown from the original
pinned, as referenced by Mouritz [139], improves considerably, the 2D composites contributes to the shortened fatigue life of 3D stitched
interlaminar fracture toughness [140], the damage tolerance, the im- fabrics.
pact resistance, and it limits the propagation of macro-delaminations Z-pinned composites are stacked 2D fabrics or prepreg laminates
[141] as compared to 2D material. However, the in-plane properties of reinforced through-thickness using thin rods called z-pins, made of
composites are lower than those of straight fibre engineering compo- extruded metal or pultruded fibrous composites. The most common way
sites. An extensive study of Mouritz and Cox [142] for example, reveals of inserting the pins is by use of an ultrasonic device [155]. The z-

565
P. Alam, et al. Composites Part B 166 (2019) 555–579

pinning of CFRP either decreases, or has no effect on the fatigue life in fibre). They reported that the fatigue strength of ICF-P was about 20%
both (T-T) and (C-C) fatigue [145,146,156,157]. Increasing the pin that of the ICF-A, and was slightly higher than that of the continuous
content and diameter beyond a critical threshold does however reduce CFRP. One way of avoiding macro-damage through bolting is to co-cure
the fatigue life [146], the threshold being specific to the CFRP system structures, thereby creating composite joints. Although the mechanical
pinned. As for the other 3D CFRP through-thickness reinforcements, the properties of bolted structures are usually superior to those of joints, the
z-pins cause crimping, misalignment and breaking of the fibres. Carbon one shot curing of large structures can be an economical advantage (e.g.
fibre distortion and crimping encourages microbuckling in (C-C) fatigue in renewable energy turbine blade technologies) [173].
[142,156], while waviness and fibre breakage reduces the (T-T) fatigue
strength [145,146]. 3. Environmental effects on the fatigue performance of CFRP
3D braided composites processing technologies are fairly complex
and can be grouped into three categories: 3D braided fabrics, 3D axial- As has been previously mentioned, CFRP is a material of choice that
braided fabrics, and multiaxis 3D braided fabrics, as reported in Bilisik's is gaining ubiquity in several industrial sectors including in transpor-
review [158]. From a mechanical standpoint, 3D braided CFRP are tation, renewable energy and construction. Its fatigue durability will
serious competitors as they exhibit high stiffness, strength energy ab- nevertheless be affected by environmental factors, which may be more
sorption and fatigue properties [159]. Furthermore, 3D braided CFRP or less extreme depending on where it is applied. When used in fixed-
are flexible, allowing for the easier manufacture of complex parts (near- wing aircraft for example, CFRP can be subjected to extreme variations
net-shape) than standard CFRP counterparts [137]. Quasi static tests in in temperature ranging from ca. −55°C when cruising at altitudes of
tension display an initial stiffening effect due possibly to changes in 7–10 km up to +30–50°C if landing or remaining stationary in e.g.
crystalline orientations [160] and local fibre orientations in the loading equatorial countries. CFRP used in e.g. tidal blades are in submerged
direction [159], before softening occurs due through the accumulation conditions where water ingress into the bulk of the material is in-
of damage. Carvelli et al. [161] compared the normalised S-N curves of evitable. Additionally, there are pressure fluctuations that affect sub-
UD-CFRP with 3D braided CFRP in (T-T). The results from their study merged blades and these will collectively affect the mechanical in-
are significant, as they show that 3D braided CFRP is the only CFRP tegrity of CFRP [174]. If CFRP is used over its service-life, which in tidal
architecture with properties improved upon those of standard UD- blades should be 20–25 years, then there will be a degree of aging that
CFRP. 3D non-crimp woven and non-crimp stitched are similar to affects the composite properties, which is important to factor into life-
standard UD-CFRP under low cycle fatigue, but are inferior under high time predictions as polymer aging weakens a composite and alters its
cycle fatigue [159]. molecular structure, making it more prone to interatomic peeling
modes of failure, otherwise known as fast-fracture. The following sec-
2.11. Effect of open holes and notches on the fatigue properties of CFRP tions highlight the effects that different environments and conditions
have on the mechanical and fatigue performance of CFRP.
As understood from the previous section, the fatigue properties of
3D composites are inferior to equivalent 2D composites. This is due to a 3.1. Effects of temperature on CFRP and fatigue
knockdown in the in-plane mechanical properties and extra premature
failures arising from imperfections within the composite architecture. Environmental temperatures tend to affect the bulk of the polymer
Nevertheless, when delamination is a primary mode of failure, stitches, matrix, and the interfaces of CFRP, more so than the fibres themselves.
pins and z-yarns do in fact, slow down the propagation of cracks that Carbon fibres begin the process of decomposition at ca. 300°C, which is
are initiated from edge delaminations [144,154]. This also improves usually far beyond its utility temperatures when embedded in a
the fracture toughness [140,162], and can extend the fatigue life of 3D polymer matrix to form CFRP. Using PAN based fibres, Sauder and co-
composites. Composites have a higher propensity towards delamination workers [175] report the stiffness of carbon fibre as being still above
when they have macroscopic defects [163]. Therefore notched, lap 90% of its original value at temperatures of 1600°C (remaining almost
jointed and open hole composites can actually benefit from z-direction unaffected below temperatures of 1000°C), whereas the strength of the
reinforcements [164]. fibres actually increased to an apical value of 2850 MPa at 1800°C, after
Bolted joints are an effective way of mechanically connecting which it decreased to a room temperature value at 2000°C. In further
composite parts [165]. The circular shape of bolts can be beneficial in testing on a rayon-based carbon fibre, they found the same behaviour to
delocalising stresses. Drilled holes are macro-defects and can reduce the be true, though Sauder and co-workers were unable to elucidate a
ultimate strength of a composite by ca. 50% , depending on the stacking mechanism-based reason for this phenomenon. It is possible that this
sequence [166]. Indeed, open hole CFRP are more prone to cracking phenomenon is implicitly linked to the initial heat treatment tem-
through delamination and as such, improving the through-thickness perature (HTT) of the carbon fibres, as this has been shown to alter the
performance of CFRP is recognisably important. Undamaged 3D CFRP, thermal conductivity of the fibres at different temperatures ranging
though weaker than undamaged UD CFRP, has a better resistance to from 500 to 2500 K [176]. Under reduced temperatures, carbon fibres
macro-delamination (as described above) and this results in the im- retain their resilience and unlike polymers, tend not to become more
proved fatigue performance of 3D composites over UD composites brittle. Polymers are affected by changes in temperature, becoming
when the samples contain macro-defects such as open holes 3D woven softer under higher temperatures and more rigid and brittle at lower
[167] or stitched [168]) and notches [169,170], with normalised S-N temperatures. As such, it is the polymer component (and its interface
curves. One notable exception can be found in Tsai et al. [171] who with fibres) that limits the thermomechanical effectiveness of CFRP.
showed that the superposition of normalised S-N curves for notched and Temperature cycling affects the physical continuum of a CFRP
unnotched samples of 3D woven CFRP were almost identical. Initially composite. Cycling the ambient temperature around epoxy matrix CFRP
cut fibres (ICF) CFRP have been recently scrutinised by the composite between ca. −50°C and 150°C, causes the growth of micro-cracking
community due their unique ability to form complex shapes while re- parallel to the fibre axis beyond approximately 100 cycles in uni-
taining excellent mechanical properties, close to those of continuous directional (UD) laminates and within 10 thermal cycles for angular-ply
fibre laminates. As complex structures are often required in the in- samples [177–179]. Cracking can be induced at earlier cycles by re-
dustry, new composite manufacturing routes have to be developed and ducing the temperature further [180–183]. Micro-cracking reduces the
properties have to be examined. Sudarsono et al. [172] investigated the mechanical performance of CFRP and according to Kaw [184], tends to
influence of open holes on the fatigue performance of quasi isotropic instigate from the interfaces of the composite (where matrix materials
ICF CFRP manufactured in autoclave (ICF-A) and through press and fibre sizing diffuse into one another creating a weak boundary
moulding (ICF-P), and compared against standard CFRP (continuous layer). It can moreover be logically inferred that hot and cold

566
P. Alam, et al. Composites Part B 166 (2019) 555–579

expansions and contractions, respectively, increase localised stresses matrix polymer that is used. Epoxy matrix CFRPs tend to gain more
between the fibre and matrix materials, thus damaging the interfaces weight through moisture absorption than vinyl ester and urethane ac-
[185–188]. The idea that thermal cycling can affect the composite rylate based CFRPs [204] since epoxides have molecular structures with
properties within such a short number of cycles is of importance as larger numbers of hydrogen bonding potential per unit volume than
thermal cycling in e.g. fixed wing aircraft and wind turbine blades is esters and acrylates. Similarly to heat-induced swelling, CFRP will swell
regular and the integrated design of CFRP within these structures when water enters its structure and the extent of swelling will de-
should account for the impact of thermal cycling over the expected termine the level of damage caused to the interfaces through both shear
product service-life, alongside the degradation of CFRP properties due [205] and compressive movement [206]. The logical consequence of
to fatigue. water ingress, swelling and the internal localisation of stress, is that the
The fracture behaviour of quasi-isotropic CFRP laminates (UT500/ mechanical properties of CFRP are detrimentally affected [174]. The
135 and T800S/3900-2B) in flexural fatigue is affected by temperature affect of water absorption into CFRP has been summarised with respect
in that; whereas under ambient or low temperatures fractures tend to be to damage mechanisms by Selzer and Friedrich [207]. In their paper,
localised to the outermost fibres of the flexural specimen, as tempera- they find that all forms of damage occur in water saturated CFRP at
tures increase, so too does the depth at which fractures will propagate lower stress levels than for dry materials. Cracking in 90° layers will
[189]. This brings to light the more adverse effects of high temperatures occur prior to cracks in 45° layers, which are precursors to delamina-
on the fatigue life of quasi-isotropic CFRPs as damage propagation tion. Importantly, the ratio of catastrophically failed CFRP to damage
under higher temperatures will more detrimentally affect their residual free CFRP is ca. 4.5:1 in fully water saturated composite, while it is only
load carrying capacities. These fracture patterns are very different to 3.2:1 in dry CFRP, indicating that fatigue damage tolerance is ap-
the temperature induced fracture patterns of unidirectional CFRP under proximately one-third less in wet CFRP as compared with dry CFRP.
flexure, which at lower temperatures (60°C) are tensile fracture domi- There are nevertheless, distinct differences between Mode I and Mode II
nant, at mid-range temperatures (130°C) are compressive fracture cracking of CFRP in water and dry conditions [208]. Mode II cracks
dominant and at very high temperatures (260°C) exhibit microbuckling tends to dominate with respect to water absorption time, while Mode I
[189]. Quasi-isotropic laminates are somewhat special materials as they cracking is guided by the number of loading cycles imposed on the
exhibit several fibre orientations. Fibre orientation tends to guide material. The Mode II maximum energy release rate of water condi-
fracture in composites however, fibre orientation specific fractures are tioned end notched epoxy matrix CFRP (T800/3900–2) flexure speci-
also affected by both cyclic loading, and temperature [190], with 45° mens in fatigue tends to be lower than equivalent dry samples [208]
oriented coupons showing the largest variation in failure pattern from and this presumably relates to the plasticisation of matrix material by
equivalent tensile fractured samples. What is interesting is that when water absorption [208,209]. Komai et al. [210] further hypothesise
comparing the S-N curves of [0]8 and [0,0, +45, −45]s from − 20∘C to (using T-1/347 and MM-1/982X ± 45 CFRP laminates) that the water
+ 100∘C, it can be shown that [0]8 composites have a larger overall re- induced swelling of matrix material increases the concentration of
duction in their strength properties than [0,0, +45, −45]s , though the stress at fibre-matrix interfaces and between individual laminates. This
rate of decline in strength is greater in the [0,0, +45, −45]s as compared in turn causes microcracking at the fibre-matrix and interlaminar in-
to the [0]8 composites [191]. Similarly, Wu found that the [0/45] or- terfaces, as well as interfacial debonding [211], which promotes further
ientations were hardly affected by heat (up to the curing temperature) water absorption and results in a rapid degradation of the fatigue life of
in terms of strength degradation, but that [0/90] would lose con- water-immersed CFRP as compared to dry CFRP. This said nevertheless,
siderable strength under fatigue as a result of heating [192]. As such, it the dimensions of larger scale fractures in cyclically loaded CFRP are an
would make sense to use high temperature resistant CFRP in cross-ply inverse function of water content. Chiou and Bradley [212] using
CFRP (e.g. AS4/PEEK, G40-800/5260) [193] however, even these are [+45/0/ −45/90]s epoxy matrix CFRP (IM7/TACTIX556) find that the
ultimately subject to the properties of the matrix materials, and the crack area does in fact decrease with the CFRP moisture content. This
orthotropy of CFRP laminates. Premature fatigue failure through ma- highlights a clear difference in understanding the role of fracture and
trix-cracking [194] in cross-ply CFRP at higher temperatures can their locations within CFRP. It could be postulated based on these re-
nevertheless be circumvented by increasing the thickness of the plate ports, that microcracks in the early stages of fatigue are confined to the
through the addition of extra constituent plies within the plate [196]. interfaces. As they build up they encourage greater water absorption
Miyano and co-workers [197,198] posit that the tensile fatigue life and thus, plasticisation of the matrix. At the later stages in the fatigue
of time and temperature dependent unidirectional CFRP can be pre- life (cf. Fig. 10) larger scale fractures dominate failure and the crack
dicted if four hypotheses are applied to the predictive model. Their areas of highly plasticised CFRP are reduced as there is more plastic
predictions are quite accurate for CFRP using PAN based carbon fibre/ deformation of the matrix, and thus reduced matrix cracking (as com-
epoxy composites [197] and T300/2500, T300/PEEK and XN40/25C pared with dry CFRP). The properties of the matrix materials in CFRP
composites [198]. The four hypotheses upon which their model is based are also a guide to the mechanisms of failure [213] since polymers
include: absorb water at different rates, swell to different extents, and are also
mechanically affected by water to different levels. PEEK for example is
1. That failure at a constant strain rate, in creep and in fatigue are the a semi-crystalline thermoplastic polymer with excellent properties of
same toughness. The hygrothermal conditioning of PEEK matrix CFRP is re-
2. That the same time-temperature superposition principle is used for ported to have little influence on its T-T fatigue life compared to dry
all failure strengths composites [214]. This is in stark contrast to epoxy matrix CFRP, which
3. That the model uses a linear cumulative damage law for monotonic shows not only a strong decline in properties compared to dry com-
loading posites, but are moreover highly sensitive to the loading regimen (R
4. That there is a linear dependence of fatigue strength on the stress ratio) [215–218].
ratio.
3.3. Fatigue damage in CFRP composites
3.2. Effects of water on CFRP and fatigue
Composite fatigue damage mechanisms are complicated because
Moisture and water enter CFRP composites through mixed modes of composite materials are intrinsically inhomogeneous. Fatigue can be
Fickian-diffusion [199–201], filling of voids within the matrix bulk characterised by the initiation and growth of damage. This includes
[202], and wicking at the fibre-matrix interfaces [203]. The volume of different competing modes of damage, as well as the complex interac-
water that will enter CFRP will therefore depend highly on the type of tions that occur between them [219–222]. Different stages of damage

567
P. Alam, et al. Composites Part B 166 (2019) 555–579

Fig. 19. Damage modes and the characteristic damage state (CDS) during the
fatigue life of composite laminates. Inspired by the work of [219].

Fig. 17. Fatigue damage evolution in composites. Inspired by the work of

[225]. Structural components made of CFRP are often subjected to complex
fatigue load histories. To simplify the analysis of fatigue damage, re-
searchers usually use constant stress ratios, R, frequencies and com-
parable waveforms. The fatigue life of CFRP composites is highly de-
pendent on the stress ratio [17,231–245] and this influences the shape
of the S − N curve, which typically follows the form
S = σTU (mlogN + b) , where S is the maximum fatigue stress, N is the
number of cycles to failure, σTU is the average static strength, and m and
b are constants where m describes the slope and b is the stress intercept
of the S − N curve. It is worth noting that low values of m and high
values of b indicate a high fatigue strength. Unlike the fatigue analysis
of metallic materials where a linear Goodman [246] diagram is broadly
used to identify fatigue limits, for CFRP laminates, the fatigue limit
cannot be determined using a simple linear approach. An effective
method of evaluating S-N curves for CFRP with constant amplitude
fatigue loading at different R values is through a constant fatigue life
(CFL) diagram, Fig. 20, [244,245]. CFL diagrams describe the fatigue
behaviour of CFRP plotting the alternating stress amplitude against the
Fig. 18. Representative delamination modes in FRPs: (a) intralaminar failure
mean stress. These diagrams become asymmetric due through differ-
between the fibre layers (b) delamination and (c) fibre breakage.
ences in the tensile and compressive strengths. The alternating stress
amplitude from the fatigue loading of CFRP laminates tends to take a
can be described by the Paris law [223]. In this law, the crack growth maximum value at a nonzero mean stress [241–243,247]. For this
rate correlates to the applied energy release rate of the composite. reason, many systematic studies have been carried out to identify the
However, damage initiation is still somewhat of a conundrum and has most suitable method of determining a CFL diagram for CFRP laminates
been only sparsley touched upon in the literature. A simple expression in fatigue. Many posit that these are bell-shaped curves, regardless of
reported by Corten et al. describes the phase of damage initiation as the the types of carbon fibres and matrix resins used [239,241–243,247].
time required for a crack to form and reach a detectable size [224]. The CFL diagram has been used for several different kinds of CFRP
Fig. 17 shows the evolution of cyclic damage in composites, as has been laminates [248–254]. Recently, it was shown by Kawai [255] that a
reported in literature [224–226]. Three distinct stages are shown in unidirectional carbon/epoxy composite, under (T-T) loading
Fig. 17. In the 1st stage intralaminar cracking occurs, where cracks (0 < R < 1) and (T-C) (R < 0) loading, has a steeper S-N curve than (C-
along the fibre-matrix interfaces or within the matrix can propagate C) loaded (R > 1) composites. This observation is consistent with the
parallel to the fibre orientation. The 2nd stage is characterised by in- quasistatic behaviour of the composites, which are stronger in tension
terlaminar shear stresses, resulting in crack growth between the fibre- than in compression. Additionally, Vassilopoulos et al. [256] showed
plies (delamination). In the 3rd stage, fibre breakage occurs transverse that the S-N curve at R = 0.1 was steeper than that at R = 10 due to that
to the orientation of fibres, resulting in fibre or fibre ply breakage the accumulation of matrix cracks under tensile fatigue loads was more
[224–226]. A schematic representation of the aforementioned three pronounced and more catastrophic. However, for specimens cut in the
modes of damage in FRP is provided in Fig. 18. Understanding and transverse direction (90°), the most critical loading mode seems to be in
standardising test methods related to delamination and fracture (T-T) at R = 0.1, compared to those of R = −1 and R = 10 , as reported
toughness has been a focus point in numerous research articles by Vassilopoulos et al. [256]. Here, the absence of fibres along the
[227–230]. Generally in composites, fatigue damage can take the form loading direction made the material more vulnerable to (T-T) loading.
of any or all of the following; matrix cracking and fibre-matrix inter- In addition, the authors showed that for on-axis composites that failed
facial debonding in the off-axis layers, delamination, fibre fracture, due to fibre breakage, the reversed loadings (R = −1) was more detri-
matrix failure, fibre micro-buckling and void growth. These forms of mental to the fatigue life than the tensile loadings (R = 0.1) .
damage may interact with one another to transform damage from one The degree of fatigue damage in CFRP relates to the properties of
form to another. Fig. 19 illustrates damage evolution during the fatigue the fibres and matrices. Relevant properties include fibre size, stiffness,
life of a composite laminate with 0° plies and off (loading) axis plies strength, orientation and stacking, and the viscoelastic properties of the
[219]. matrix polymer. Composites containing stiff fibres such as carbon are

568
P. Alam, et al. Composites Part B 166 (2019) 555–579

damage was governed by cyclic creep due to the absence of 0°-layers

angle-ply laminates [232]. It is well-established that the stress ratio
affects delamination and delamination growth [229,269–272]. Other
causes of delamination may arise through machining and/or the cutting
of FRP laminates, residual stresses from the manufacturing process (e.g.
curing), or material irregularities and geometrical characteristics in-
cluding edge effects, holes and ply-drop offs [20].
Delamination is a failure mode possibly requiring the least amount
of energy. It is hence a major source of concern for composite designers
[273]. The fact that this is difficult to detect during service makes this a
more pronounced problem still [274], as it reduces the load carrying
capacity of the composite and may lead to premature failure. Delami-
nations can be triggered by fatigue loads as it is almost impossible at
present, to manufacture a composite without defects generated either
during the curing process or during machining [275]. Fatigue delami-
nation in composite structures is more complex than for metals where
cracks deviate to grow predominantly in the mode I loading direction.
Indeed, crack growth is usually constrained at the interface between
plies and therefore the delamination mode occurring during the service
life will most likely be a mix of modes I, II and III. In fact, the study of
delamination in the fatigue of fibre reinforced composites has been the
subject of several recent review papers [276,277]. A brief overview of
this topic will be given here but the full study of fatigue delamination is
beyond the scope of this review.
There exists three main ways of experimentally characterising fa-
tigue delamination [276]. These are; crack propagation, which is the
measure of the rate of crack growth per cycle as a function of the stress
intensity factor or the strain energy release rate (SERR), crack onset,
which is a measure of how many cycles are required to visibly extend an
existing crack, and crack initiation, which is a measure of the number of
cycles required to generate a crack in an ideal (defect free) composite.
Fig. 20. Example of CFL Diagram for a Temperature of 80°C. From Ref. [302],
The crack initiation tests have so far not been studied extensively and
reproduced with the permission of Elsevier.
show large scatter in the results [278]. This, as mentioned previously, is
probably caused by the presence of small defects within the composite.
often considered as being more resistant to fatigue than lower stiffness Crack propagation tests are usually carried out using the ASTM stan-
fibres. This is because they are more adept at carrying load and as such, dard for metals for mode I [279] and allows the determination of the
the weaker component of composites (the polymer matrix), will extend whole crack growth rate curve. Crack onset fatigue delamination testing
less as a function of loading when the fibres are stiff, thereby reducing in mode I for composites is the only one for which a standard currently
matrix damage. The loading modes are important as they govern the exists [280]. This leads to the determination of the SERR as a function
dominant damage mechanisms in fatigue, the most damaging loading of the number of cycles required for crack growth. However, it is often
mode being acknowledged as tension-compression (T-C). Fibre failure assumed in the literature [276] that a small crack propagation is
criteria under such modes can be either tensile or compressive equivalent to fatigue crack onset and that therefore these phenomena
[30,257,258], with compressive failures often instigated by cross-shear are the same.
within the fibres, or buckling rupture of the fibres [261–264]. Tensile There are three types of delamination modes affecting the growth of
fibre failures contrarily [259,260] can be associated to fibre pull-outs a crack: mode I (normal opening), mode II (in-plane shear) and mode III
by debonding, and fibre cross-fractures. This said, fibre pull-outs by (shear scissoring or out of plane shear). Modes I and II delamination, at
debonding are in fact, a function of the bond shear strength between the least in the case of static loading, are quite well understood. However,
fibre and matrix. Compressive failure is sometimes considered a design mode III loading has not been studied extensively and it is usually as-
limitation in UD CFRP, since compressive strengths of UD CFRP are sumed that mode III and mode II delamination have the same proper-
often less than 60% of their tensile strengths. Matrix failure in fatigue ties, which is conservative, considering mode III has been shown to
may occur either within a ply, or between plies (delamination). Matrix have a higher interlaminar fracture toughness [281]. Mode II delami-
failure may also occur at the fibre/matrix interface, which in turn leads nation toughness is usually higher than mode I toughness for composite
fracture propagation away from the interface and into the matrix bulk. materials [282], although it has been shown during crack propagation
CFRP delamination occurs due to interlaminar stresses generated be- tests that the difference in crack growth rate between the different
tween the plies, and is often a result of cracks/voids in the matrix phase modes reduces [283] as the cycle count increases and eventually dis-
[229,265]. Gamstedt et al. [18] researched the influence of matrix type appears for carbon/epoxy. It was also shown that the threshold below
on fatigue life in UD CFRPs using both thermoplastic (poly- which no fatigue damage occurs is independent of the loading mode
etheretherketone, PEEK) and thermosetting matrices (epoxy toughened [284]. Therefore, the effect of mixed mode ratio on fatigue delamina-
with a thermoplastic additive). The two composite cases were loaded in tion for high cycle fatigue can be considered negligible.
tension-tension along the fibre direction and fatigue tests revealed that When the maximum SERR near a crack tip is close to its inter-
the thermoset matrix CFRP had a higher resistance to fatigue failure laminar fracture toughness, the delamination propagates rapidly and
than the thermoplastic matrix CFRP, the fatigue life being most likely the failure is similar to that observed in a quasi-static test and therefore
guided by matrix cracking [266–268]. Petermann et al. [232] studied the R ratio has no influence on the crack growth rate. However, as the
the fatigue life of carbon-epoxy laminates with ± 45∘ angle-ply or- maximum value of the SERR is reached as a loading cycle reduces, it has
ientation under tension-tension at high stress ratios (between 0.40 and been shown that the effect of the R ratio on fatigue delamination in-
0.86 times of the static tensile strength). They reported that fatigue creases [285]. The damage threshold and the exponent of the Paris law

569
P. Alam, et al. Composites Part B 166 (2019) 555–579

curve are lower for lower R ratios for mode I and mode II crack growth [302] proposed what they called an anisomorphic CFL diagram, taking
rates in carbon/epoxy composites for R = 0.1 and R = 0.5. This is also into account the asymmetry of the curve as well as the occurrence of the
true for mixed mode delamination [286]. peak envelope at a positive mean stress. It can be built using only the
Other factors have an influence on the fatigue delamination of static strength in tension and compression and a reference S-N curve for
composite structures such as the loading frequency, loading sequence, a critical stress ratio equal to the ratio of compressive to tensile
environmental factors such as moisture, water uptake and temperatures strength. It was shown to be valid for the prediction of fatigue failure in
as well as matrix properties such as the brittleness [276,287-296]. non-woven carbon/epoxy laminates [302]. The approach was extended
However, these topics are not discussed in this paper. If further in- to predict the fatigue of woven composites for different temperatures in
formation is required, the reader is directed to review papers dedicated a later study [303]. Their approach involved using the critical stress
to fatigue delamination [276,277]. ratio defined above as the ratio between compressive and static
strength χ = σC / σT . The CFL is then divided into two domains, tension-
4. Fatigue modelling of CFRP tension dominated and compression-compression dominated. These
two domains have smooth nonlinear curves connected by a point as
A few review papers have been written on the modelling of fatigue shown in Fig. 20 Therefore, the envelope of the anisomorphic CFL
in FRP [22,297–299]. While Pascoe et al. focused solely on methods diagram can be defined by the following piecewise function:
predicting the fatigue delamination growth in composites [297], De- kT
⎧ σm − σm(χ ) 2 − ψ(χ ) (χ )
grieck and van Paepegem compared a wider variety of modelling ⎪ ⎛ ⎞ , σM ≤ σM ≤ σT
methods for both UD and textile FRP composites [298,299]. They σa − σa(χ ) ⎪ ⎝ σT − σm(χ ) ⎠
− =
proposed a classification system to separate the different modelling σa(χ ) ⎨ k C
(χ ) 2 − ψ(χ )
⎪⎛ σm − σm ⎞ , σC ≤ σM < σM (χ )
techniques: (i) fatigue life models based on experimental data such as S- ⎪⎝ σC − σm(χ ) ⎠ (3)
⎩
N curves, (ii) phenomenological models which are not based on phy-
sical mechanisms but rather, on observed macroscopic properties where σm , σa , σT and σC are the mean stress, stress amplitude, tensile and
during fatigue, such as residual strength and stiffness, and (iii) pro- compressive strengths. ψχ is the fatigue strength ratio and takes a value
gressive damage models which take into account damage mechanisms in the range ψ ∈ [0,1], χ is the critical stress ratio and k is an exponent
and either, correlate them to mechanical properties (stiffness and used to transform a straight line into a parabola in tension (T) and
strength), or, measure the damage growth (delamination size, cracks compression (C), and are both based on empirical data.
per unit area). The same classification system will be used in this sec- The reference normalised S-N curve then has the following form:
tion. 〈1 − ψχ 〉a
1 1
2Nf =
Kχ (ψχ ) 〈ψχ − ψ χ (L) 〉b
n
(4)
4.1. Fatigue models - a general overview
where the angular brackets represent the function defined as
In 1973, Hashin and Rotem developed one of the first fatigue failure 〈x 〉 = max {0, x } , ψ χ (L) is the normalised fatigue limit and Kχ , a, b and n
models for composites [30]. They considered two different failure are material constants which can be determined by fitting Equation (4)
mechanisms; fibre-failure and matrix-failure. They proposed a failure to fatigue data at the critical stress ratio. This model was shown to give
criterion for each mode of failure based on three separate S-N curves; adequate results for woven carbon/epoxy laminates.
longitudinal tensile strength, transverse tensile strength and shear Fatigue life models were the first fatigue models developed. They
strength. These curves can be obtained experimentally by off-axis ten- are often used as they do not require any understanding of the physical
sile testing of coupons under uniaxial constant amplitude stresses. In damage mechanisms and are very simple to use. However, these models
Equations (1) and (2), σa represents the longitudinal fiber stress, σT the require a large amount of experimental data and are usually calibrated
transverse stress, τ the shear stress and σaS , σTS and τ S the ultimate for one specific case-study. The breadth of applicability to other pro-
tensile, transverse and shear strengths, respectively. This model is only blems is therefore very limited.
applicable to unidirectional laminates.
4.2. Phenomenological models
σa = σaS (1)

2 2
4.2.1. Residual strength models
⎛⎜ σT ⎞⎟ + ⎛ τ ⎞ = 1 Based on previous work carried out on the development of a re-
S S
⎝ σT ⎠ ⎝τ ⎠ (2) sidual stiffness model [304], and assuming a linear stress strain re-
sponse until failure, Whitworth [305] proposed a strain failure criterion
A number of fatigue models using S-N curves and various failure
to eliminate the failure stiffness from his model and obtain the fol-
criterion were developed in subsequent years to account for multiaxial
lowing fatigue life relationship:
laminates which are restricted to the two failure modes mentioned
above. As most of these models were developed before the early 2000s, m
⎧ 1 ⎡ ⎛ c1 SU ⎞ c2 ⎤⎫
they are not presented in detail in this paper. N = exp − 1⎥ − 1
⎨ h ⎢⎝ S ⎠ ⎬
Another common approach in fatigue life modelling is to use con- ⎩ ⎣ ⎦⎭ (5)
stant fatigue life (CFL) diagrams. Here, the cycles to failure are plotted where parameters h and m depend on the applied stress, loading fre-
as a function of mean stress on the x-axis and alternating stress on the y- quency and environmental conditions, c1 and c2 are constants de-
axis. They are useful because they show the mean stress sensitivity of termined experimentally, N is the number of cycles to failure, SU is the
composites, as have been observed in tests. Through extensive experi- ultimate strength and S is the maximum applied stress.
mentation on the strength and fatigue life of aerospace CFRP laminates, Whitworth further proposed the following relation for strength de-
Harris et al. [300] showed that the CFL envelopes for these materials gradation [304]:
are asymmetric and nonlinear, and that their peak positions are shifted
n
to the right of the alternating stress axis. Ramani and Williams [301] SRγ = SUγ − [SUγ − S γ ]
N (6)
experimented on the fatigue behaviour of notched and un-notched
carbon/epoxy composites. They found that the maximum of the CFL where γ is a material constant determined experimentally and SR is the
envelope does not occur at R = −1 but rather, at around R= −0.43 and residual strength. Once all the material parameters have been de-
for a positive mean stress. Based on these previous studies, Kawai et al. termined, Equations (5) and (6) can be used to determine the residual

570
P. Alam, et al. Composites Part B 166 (2019) 555–579

Fig. 21. Flowchart of Shrokrieh and Lessard Progressive Damage Model. Figure inspired by the work of [314].

strength. Failure occurs when the maximum applied stress equals the curve to failure as linear. Their model is valid for laminates with fibre-
residual strength. Whitworth also assumed that the ultimate strength dominated properties. However, in the case of laminates with matrix-
was a statistical variable that he represented with a 2 parameter Wei- dominated properties, the assumption of linearity between stress and
bull distribution. The model was validated on T300/5280 graphite/ strain to failure is incorrect. They proposed the following degradation
epoxy laminates which were tested under constant amplitude loading law:
for 50,000 cycles. The median residual strength after the test was
dE (n)
353 MPa which the author considered sufficiently close to the value of = −E (0) Qνnν − 1
dn (9)
371 MPa predicted by the model.
Yao and Himmel [306] proposed a residual strength model for FRP where E (0) is the static stiffness, Q and ν are parameters dependent on
polymers that could capture rapid degradation in the early stages of the applied stress level, stress ratio and loading frequency. Yang et al.
fatigue life. The model was also able to account for level and slow defined E (0) using a log-normal statistical distribution in their model.
degradation that occurs in the middle of the fatigue life, as well as rapid The model assumed that failure occurred when the strain in the spe-
degradation in the final cycles. For tension-tension fatigue, they pro- cimen was equal to the static ultimate strain. The model was validated
posed the following relationship: using graphite/epoxy coupons in tension-tension fatigue. The predicted
and experimental fatigue life curves were noted to correlate well with
sin(βx )cos(β − α )
R (i) = R (0) − [R (0) − S ] one another.
sin(β )cos(βx − α ) (7)
Khan et al. [313] proposed a stiffness reduction model where they
where R (i) is the residual strength at the ith
cycle, R (0) is the static created a damage variable, D, that related to stiffness in the following
strength, S is the maximum stress in the loading cycle, α and β are way:
constants to be determined experimentally and x = i/ Nf . E
For specimens in compression-compression fatigue, the following = 1 − cD
E0 (10)
degradation law was proposed:
1/2 1/2
ν
E2 1 1 ν
i
R (i) = R (0) − [R (0) − S ] ⎜⎛ ⎟⎞ c = π ⎛⎜2 A ⎟⎞ ⎡ + − A⎤
⎝ ET ⎠ ⎢ (E E )1/2 2G EA⎥ (11)
⎝ Nf ⎠ ⎣ ⎦
A T A
(8)

where ν is a strength degradation parameter defined in their study as where EA , GA , νA are the Youngs modulus, shear modulus and Poissons
0.64. By using fatigue data obtained by Harris et al. for T800/5245 ratio in the axial direction, respectively, while ET is the transverse
laminates [307], the model was validated and showed to have good modulus of the undamaged specimen. It is therefore possible to relate
predictive capability. the damage growth per cycle to experimentally measured reductions in
Compared to other types of CFRP fatigue models, only a few have stiffness. Failure is assumed to occur when the maximum strain in a
been found that model the pure residual strength of CFRP. The main cycle is equal to the static ultimate strain. The value of the damage
drawback of these models is that they are unable to account for the variable at failure can also be calculated. The cycle count to failure can
stiffness degradation (observed experimentally), though they are still be obtained by integrating the damage variable from its initial to its
able to predict the fatigue life. The primary issue of contention is thence final value:
that these models are not able to predict the deformation of composites Df dD
throughout their fatigue life. Nf = ∫Di f (Δσ , D) (12)

4.2.2. Residual stiffness models

Residual stiffness is not simple to model as there are several over- 4.3. Progressive damage models
lapping factors including the rheological responses, bulk molecular
kinematics and variations in molecular orientations at fibre interfaces, Shokrieh and Lessard [314,315] proposed a progressive fatigue
each as a function of loading time and frequency [308–311]. Yang et al. damage model combining stress analyses, failure analyses and material
[312] developed a stiffness reduction model assuming the stress-strain degradation. The modelling method used is explained in the form of a

571
P. Alam, et al. Composites Part B 166 (2019) 555–579

fatigue life. The main disadvantage with this model was that it required
considerable experimentation for each application to obtain the gradual
degradation parameters.
A similar model was developed by Papanikos et al. [316] combining
stress, failure analyses and material degradation. Seven failure modes
were considered but simpler formulations were used with no nonlinear
parameter. For delamination in tension and compression, the Ye dela-
mination criterion was used [320,321], while 3D static Hashin Criterion
[257] was used to predict matrix tensile and compressive failure as well
as fibre-matrix shearing. In previous work [322], the authors compared
Fig. 22. Bilinear constitutive law for cohesive elements (a) and its definition for the predicted elastic responses and failure loads of a bolted joint in a
mixed-mode loading (b). Figure inspired by the work of [327]. graphite/epoxy composite to experimental results. They found that
better predictions were obtained using the Maximum Stress Criterion
(MSC) for fibre failures in tension and compression than the 3D Hashin
flowchart in Fig. 21. A finite element stress analysis is first carried out
Criterion. They also used a set of rules for sudden degradation that
on a specimen, using solid elements. Edge effects are accounted for by
yielded a closer fit to their experimental data. Therefore, unlike the
using a higher element density near edges and other areas of stress
model by Shokrieh and Lessard, once failure occurs, the associated
concentration such as holes and notches. Iterative schemes are used to
stiffnesses are not reduced to zero. Also, the generalised residual
account for nonlinearity in the stress state when failure occurs. A failure
property degradation rules used were simplified, using a linear stiffness
analysis is carried out considering seven different failure modes; fibre
reduction and a second-order polynomial for strength reduction:
tension and compression, matrix tension and compression, normal
E (n) = [A (n/ Nf ) + 1) and R (n) = [B (n/Nf )2 + C (n/ Nf ) + 1] RS ; where
tension and compression, and fibre-matrix shearing. The failure criteria
E (n) and R (n) are the residual stiffness and strength, ES and RS are the
used are similar to the Hashin static failure [257] criteria but the ma-
static stiffness and strength, A, B and C are material parameters ob-
terial properties are functions of cycles, stress state and stress ratio.
tained experimentally. Only the maximum and minimum cycle loads
Additional parameters have been added to account for material non-
were modelled to account for the stress ratio effects. As in some cases,
linearity. The fibre tension fatigue criterion for σxx > 0 is:
millions of cycles can occur before failure, conducting a full stress
σxy 2 analysis at each cycle would require tremendous amounts of CPU time.
⎛ 3 4 ⎞
2 + 4 δσxy
σxx 2Exy (n, σ , κ ) Therefore, the authors proposed applying a cycle jump between each FE
gF2+ =⎛ ⎜
⎞ +⎜⎟
2
⎟
⎝ XT (n, σ , κ ) ⎠ ⎜ S xy
+
3
δS 4
xy (n , σ , κ ) ⎟ stress analysis, during which only material property degradations were
⎝ 2Exy (n, σ , κ ) 4
⎠ applied. However, the authors did not provide any recommendations on
2
⎛ σxz 3 4 ⎞ what increment to use for the cycle jump or how that affects the ac-
2Exz (n, σ , κ )
+ 4 δσxy
+⎜ ⎟ curacy of the model. The model was validated with CFRP laminates
2
Sxz 3 4 tested in tension-compression. It was shown to not only give an accu-
⎜ + δSxz (n , σ , κ ) ⎟
⎝ 2Exz (n, σ , κ ) 4 ⎠ (13)
rate prediction of the fatigue life but also predicted the onset and
and failure occurs if gF + > 1. Here, XT , Sxy , Sxz are the residual long- growth of the correct damage modes. Indeed, both the model and ex-
itudinal tensile strength, the in-plane and out-of-plane residual shear perimental results considered delamination as the main failure mode
strengths, respectively. Exy and Exz are the residual in-plane and out-of- for low stresses and fibre tensile failure for high stresses.
plane shear stiffness values, respectively. σxx , σxy and σyz are the long- Using a similar approach, Kennedy et al. [317] developed a model
itudinal, in-plane shear and out-of-plane shear stresses in the elements, by modifying the Puck Failure criterion [318] to analyse fatigue in
respectively. δ is a material nonlinearity parameter which is considered FRPs. They carried out a stress analysis at the ply level, followed by a
to be constant. For the failure criteria of the other failure modes refer to failure analysis. A separate analysis was carried out on the fibre re-
Ref. [314]. The final part of the fatigue model considers the material sponses and the matrix responses to shear and normal stresses in the
degradation, which is separated into two parts; sudden and gradual FRP primary axis. The fibre direction strength was reduced according to
degradation. Each failure mode is associated with some of the material a linear degradation rule, while the stiffness was reduced linearly as a
properties being reduced to zero. For matrix compression failure for function of the accumulated damage. Irrecoverable cyclic strains were
example, Eyy , νyz and νyz are reduced to zero. Fibre tension and com- also calculated as a function of the degraded fibre direction modulus.
pression are considered to be catastrophic failure modes and therefore Failure was considered to have occurred when the fibre direction
are followed by a reduction of all material stiffness values and Poisson's stresses were equal to the fibre direction residual strengths. For the
ratios to zero. The authors also developed a generalised residual ma- matrix response, a failure analysis was first carried out for Inter Fibre
terial property degradation as: Failure (IFF) using modified Puck stress exposure equations [319]
where the static strengths were replaced by the fatigue strengths. Here,
β 1/ α
⎡ log(n) − log(0.25) ⎞ ⎤ IFF occurs when the stress exposure was greater than 1. At each cycle,
R (n, σ , κ ) = ⎢1 − ⎜⎛ ⎟ ⎥ ⋅(RS − σ ) + σ the fatigue strengths used in the failure initiation criteria were reduced
⎝ log(Nf ) − log(0.25) ⎠
⎣ ⎦ (14) according to a linear degradation rule. Three modes of matrix failure
were considered depending on the combination of shear and normal
λ 1/ γ
⎡ log(n) − log(0.25) ⎞ ⎤ σ σ stresses, with Mode A corresponding to tensile normal stress and mode
E (n, σ , κ ) = ⎢1 − ⎜⎛ ⋅⎜⎛ES − ⎟⎞ +
Nf ) − log(0.25) ⎠ ⎥
⎟
log( εf ⎠ εf B and C to compressive normal stress. Before the occurrence of IFF, the
⎣ ⎝ ⎦ ⎝ (15)
shear modulus was reduced slowly for Mode A while the transverse
where R (n, σ , κ ) and E (n, σ , κ ) are the residual strength and stiffness, modulus was reduced gradually for modes B and C. After IFF, both the
Nf is the fatigue life, σ is the maximum applied stress, RS and ES are the shear and transverse moduli are reduced rapidly for Modes A and B,
static strength and stiffness, α, β, λ and γ are curve fitting parameters while mode C was taken as leading into an immediate catastrophic
and εf is the average strain to failure. Shokrieh and Lessard compared failure at that point. This approach required the use of material para-
their model predictions to experimental data on pin and bolt-loaded meters that shaped the moduli degradation rules, and which were ob-
graphite/epoxy laminates. These specimens were chosen for their tained experimentally. The model was validated for quasi-isotropic E-
complexity and because of the presence of stress concentrators. A good glass/epoxy samples and shown to have a good correlation with ex-
agreement was found to exist between the predicted and measured perimentally measured fatigue life and fatigue modulus degradation.

572
P. Alam, et al. Composites Part B 166 (2019) 555–579

Fig. 23. Description of CDM approach. Figure inspired by the work of [332].

However, at higher stress levels, the model did show a tendency of over introduced the concept of an initiation zone which is defined as the
predicting the modulus degradation. Although this model thus far not zone where elements have the highest SERR and where a crack forms.
been used for CFRP, the authors claim that it can be applied to any fibre This allows for the identification of the location of a crack, and offers a
reinforced composite. It would nevertheless require an experimental set new method by which means initiation and propagation can be si-
up for CFRP in order to obtain all the material parameters for the de- multaneously modelled. Allegri et al. [326] extended the model further,
gradation rules. to account for the effect of the stress-ratio as well as mode-mixity on
Delamination is the failure mode requiring the least energy. Harper delamination propagation. The model has the option to include pro-
and Hallett [321] developed a fatigue degradation law using cohesive pagation thresholds. Only three independent material parameters are
interface elements. These elements do not have a single stiffness but are needed for the model, since the crack propagation threshold is ne-
governed by a bi-linear constitutive law shown in Figure 22, which glected.
means the displacements are initially elastic until the maximum All the models proposed in Refs. [221,321,324–326] require that
strength is reached. Damage is then tracked by a variable and used to the cohesive zone length be predetermined. Kawashita and Hallett
reduce the element stiffness. In elements which exceed their linear- implemented a cohesive zone model using the commercial explicit FE
elastic range, a crack tip is formed and damage initiates. The cohesive solver LS-Dyna [327]. The SERR is calculated for each element at each
zone consists of all the damaged elements ahead of the crack tip. A time step which allows for the identification of delamination fronts (as
static damage variable was introduced by Harper and Hallet to account elements with the highest SERR). Using this method, multiple delami-
δ −δ
for irreversible damage: ds = δ m − δm, e , where δm, e is the element dis- nation fronts can be modelled simultaneously. Fatigue damage accrues
m, f m, e
placement at the maximum stress level, δm, f is the displacement at in the elements until failure occurs, after which the nearest neigh-
failure and δm is the current element displacement. The element fails bouring elements become part of the delamination front. As such, the
when ds reaches 1. Cracks are assumed to propagate according to the cohesive zone length no longer needed to be pre-determined, but ra-
∂a
model developed by Blanco [323]: ∂N = CΔG m , where a is the crack ther, only the effective element length, which is itself defined based on
length, ΔG is the change in the strain energy release rate (SERR) and C the damage state of the neighbouring elements. This method also allows
and m are defined in Ref. [323], and are obtained through fatigue ex- for the tracking the damage propagation. This model was validated
perimentation under modes I and II loading. Interface elements in the using carbon/epoxy central cut-ply specimens, and by using the NA-
cohesive zone first go through quasi-static damage only, subsequently, FEMS benchmark for circular delamination made from carbon/epoxy.
as they get closer to the crack tip, they experience fatigue damage. A The authors claimed that a good correlation with experimental data
∂df 1 − dS − df , u ∂a could be obtained using a coarse grid, hence reducing computational
fatigue damage parameter was therefore introduced: ∂N
= Lfat ∂N
,
time.
where df , u is the unwanted fatigue damage defined as the fatigue da- Tao et al. [328] proposed a similar model for fatigue delamination
mage accumulated in the cohesive zone undergoing quasi-static da- growth but a virtual fatigue damage parameter is introduced to identify
mage, ds is the static damage variable and Lfat is the length of the co- crack tip fronts using local information only and without degrading
hesive zone undergoing mostly fatigue damage defined in this paper as elements. This ensures SERR is kept constant during identification and
half the total cohesive zone length. Failure in an element occurs when propagation stages. A correction factor is introduced to reduce the mesh
the total damage defined as the sum of static and fatigue damage is sensitivity in the model. Delamination growth as calculated by this
equal to 1. The model proposed here requires the presence of initial model was shown to have good correlation with experimental data for
cracks or high initial stress concentrations. May and Hallett extended carbon/epoxy double cantilever beam and end notch flexure specimens.
the model to include a damage initiation criterion based on the material Talreja proposed a physical based model, Synergistic Damage
S-N curve [324]. Allegri et al. [325] adapted this model to the pre- Mechanics (SDM) [329], combining micro-damage mechanics with
diction of mode II fatigue delamination growth of moderately tough continuum damage mechanics (CDM) as developed in previous works
carbon/epoxy specimen. In later work, May and Hallett [221]

573
P. Alam, et al. Composites Part B 166 (2019) 555–579

[330,331]. The CDM approach consists of a two-step homogenisation. materials engineering approach to interpreting the effects of fatigue on
First the stationary (undamaged) microstructure containing the fibres and CFRP. This is also evident from the section on modelling, where mod-
matrix is homogenised into a material with generally anisotropic elling methodologies have been implemented from using macro-scale
properties. Following this, the evolving microstructure containing the physics using finite element methods. With the recent high-level de-
damage is homogenised as shown in Fig. 23. This process is carried out velopments in computational power, it is surprising that molecular
over a representative volume element which needs to carry a sufficient modelling approaches have not been used to understand the intricate
number of discrete entities to represent the effect of the homogenised materials behaviour of CFRP in fatigue at the atomic to nano scales.
response at point P. The damage entity tensor is then defined as the Molecular modelling dynamics could be used to not only better un-
dyadic product of two vectors on the surface of a defect, a and n re- derstand the finer-scale mechanisms that essentially control the beha-
presenting the influence of the point on the surrounding medium and viour of CFRP in fatigue, but also to redesign CFRP materials structures
unit normal to the surface: dij = ∫S ai nj dS . The average of the damage and interfaces based on developed fine-scale understandings. Specific
entity tensor over the RVE volume is defined as the damage mode examples of areas of critical importance to better understand and re-
1
tensor: Dij = V ∑kα (dij(α ) )k , where α is the damage mode, V is the vo- design at the low-scales include; how unloading and reloading affect
α
lume of the RVE and k α is the number of damage entities. A relationship the shapes/structures (and thus mechanical properties) of molecules in
proposed to relate stiffness reduction to the damage entity tensor, re- the matrix bulk and at fibre-matrix interfaces, the affects of water on
quired the definition of eight material constants as well as a constraint intermolecular interactions (and energies) within the matrix bulk and at
parameter, which is a measure of the influence of damage on the re- fibre-matrix interfaces, and how local heat developed through inter-
sponse of the material and is linked to the crack opening displacement. molecular friction affects the molecular mobility and mechanical per-
The eight material parameters include the modulus in longitudinal and formance of CFRPs in fatigue.
transverse direction, the shear modulus, the principal Poisson's ratio It is vital that greater research efforts are expended into under-
and four phenomenological damage parameters, which are used to es- standing how strain is distributed at fibre-matrix interfaces during fa-
timate the reduction in stiffness and can be calculated using the initial tigue. There is currently a gap in our understanding on how strain fields
elastic properties and their values at a given crack spacing. A 3D FE actually evolve over a fatigue cycle at the micro and meso scales. This
analysis was carried out at the microscale to estimate the crack opening area is more easily tackled numerically, as it is difficult to map this
displacement, and therefore the constraint parameter, κ, [333]. The reliably in fatigue given the current technologies available to us.
material constants are determined experimentally or computationally As has been noted in the review, fatigue testing results tend to show
on a reference laminate. Subsequently, the stiffness reduction is com- considerable scatter. The lack of repeatability of CFRP in fatigue is a
puted on the laminate scale (meso-scale). The laminate response to critical concern as it depreciates engineering confidence in the material.
external loading using the reduced stiffness can be calculated (macro- Though we have suggested possible reasons for why significant scatter
scale). is a typical characteristic of CFRP in fatigue, there is actually very little
In later work, Singh and Talreja extended their model to predict the research that has been conducted to support our hypotheses. This also
evolution of crack density under a quasi-static load [334]. Singh then represents a major gap in our current understanding of CFRP fatigue
improved upon previous models for damage evolution by accounting and should be critically addressed so that remediation of the problems
for the stochastic nature of crack propagation [335]. The model was leading to scatter becomes a possibility. More focused research should
validated using a variety of carbon/epoxy cross-ply laminates. Al- therefore be conducted to deduce how manufacturing, inherent mate-
though this approach has thus far been mostly used for quasi-static rial defects, cross-link density (curing) and material homogeneity will
loading, it gives a framework for estimating the reduction in stiffness affect the extent of scatter in CFRP under cyclic loads.
based on the crack opening displacement and experimentally obtained Research on the cyclic loading of CFRP has almost entirely been
material properties. Therefore, provided the damage parameters can be based on the systematic sinusoidal loading of coupons, often using
obtained experimentally and the crack opening and crack sliding dis- simple loading modes. Yet, real structures are subjected to far more
placements can be obtained either experimentally or through a micro- complex loading conditions and it could be argued that a worthwhile
scale FE, the evolution of crack density and stiffness during cyclic exercise would be to collate application specific real-life random
loading can be estimated. Haojie et al. [336] used non-destructive spectral loading data, which can then be extrapolated to a lab-testing
testing to obtain the transverse crack density of [0/± 45∘]S laminate regimen. To include more realistic multi-directional loading conditions
during a fatigue test. This allowed them to experimentally obtain the within such a regimen would increase the complexity of the output,
damage parameters. Therefore, using an SDM approach, they were able perhaps even increase the extent of scatter, but it might yield enhanced
to predict the reduction in stiffness as a function of applied load cycles. insight into the actual mechanical profile of CFRP under fatigue and
The obtained reduction in stiffness was within 5% of the experimental might positively benefit its utility in engineering structures (e.g.
results but the model was not conservative. This was because due to the through improvements in safety factors and design).
limitations of non-destructive testing, only transverse cracks in the/
± 45∘ plies were considered and therefore delamination at the crack tips Acknowledgments
was ignored.
The authors would like to acknowledge the European Union for
funding this research through the following projects: MARINCOMP,
5. Conclusions Novel Composite Materials and Processes for Marine Renewable
Energy, Funded under: FP7-People, Industry Academia Partnerships
Considerable research has been undertaken to better understand the and Pathways (IAPP), Project reference: 612531. POWDERBLADE,
behaviour of CFRP in fatigue. Many important contributions in the area Commercialisation of Advanced Composite Material Technology:
of CFRP fatigue have been described in this review, though the review Carbon-Glass Hybrid in Powder Epoxy for Large Wind Turbine Blades,
is by no means exhaustive. We have focused on topic areas in CFRP Funded under: Horizon 2020, Fast Track to Innovation Pilot, Project
fatigue of current and generic importance, which has allowed us to reference: 730747.
identify the current gaps in knowledge and understanding within this
area. We will briefly describe each of the main gaps in understanding in References
this final conclusions section.
Much of the research presented here, has dealt with CFRP fatigue [1] Ashby MF. Materials selection in mechanical design. fourth ed. Oxford, UK:
from a macro-scale view point, using primarily a mechanical and Elsevier (Butterworth-Heinemann Imprints); 2011.

574
P. Alam, et al. Composites Part B 166 (2019) 555–579

[2] Coffin Jr. LF. A study of the effects of cyclic thermal stresses on a ductile metal. environment. J Compos Mater 2014;48(23):2905–14.
Trans ASME 1954;76:93150. [40] Hiremath C, Senthilnathan K, Guha A, Tewari A. Effect of volume fraction on
[3] Manson SS. Behavior of materials under conditions of thermal stress. Natl Advis damage accumulation for a lattice arrangement of fibers in CFRP. Materials Today
Comm Aeronaut NACA TN-2933 1954. Off: Proceedings 2015;2:2671–8.
[4] Kawai M, Itoh N. A failure-mode based anisomorphic constant life diagram for a [41] Brunbauder J, Pinter G. Effects of mean stress and fibre volume content on the
unidirectional carbon/epoxy laminate under off-axis fatigue loading at room fatigue-induced damage mechanisms in CFRP. Int J Fatigue 2015;75:28–38.
temperature. J Compos Mater 2014;48:571592. [42] Brunbauer J, Stadler H, Pinter G. Mechanical properties, fatigue damage and mi-
[5] Hertzberg RW. Deformation and fracture mechanics of engineering materials. crostructure of carbon/epoxy laminates depending on fibre volume content. Int J
fourth ed. New York, Chichester: John Wiley and Sons; 1996. Fatigue 2015;70:85–92.
[6] Boller KH. Fatigue characteristics of RP laminates subjected to axial loading. Mod [43] Cali C, Cricri G, Perrella M. An advanced creep model allowing for hardening and
Plast 1964;41:14550. damage effects. Strain 2010;46:347–57.
[7] Dowling N, Siva Prasad K, Narayanasamy R. Siva Prasad Katakam, Narayanasamy [44] Gornet L, Wesphal O, Burtin C, Bailleul JL, Rozycki P, Stainer L. Rapid determi-
R, editors. Mechanical behavior of materials engineering methods for deformation, nation of the high cycle fatigue limit curve of carbon fiber epoxy matrix composite
fracture, and fatigue. fourth ed.Boston, Mass; London: Pearson; 2013. International laminates by thermography methodology: tests and finite element simulation.
Ed. contributions by. Procedia Engineering 2013;66:697–704.
[8] Lemaitre J, Sermage JP, Desmorat R. A two scale damage concept applied to fa- [45] Peyrac C, Jollivet T, Leray N, Lefebvre F, Westphal O, Gornet L. Self-heating
tigue. Int J Fract 1999;97:6781. method for fatigue limit determination on thermoplastic composites. Procedia
[9] Basquin OH. The exponential law of endurance tests. Am Soc Test Mater Proc Engineering 2015;133:129–35.
1910;10:62530. [46] Barron V, Buggy M, McKenna NH. Frequency effects on the fatigue behaviour on
[10] Azeez AA, Rhee KY, Park SJ, Hui D. Epoxy clay nanocomposites processing, carbon fibre reinforced polymer laminates. J Mater Sci 2001;36(7):1755–61.
properties and applications: a review. Compos B Eng 2013;45:30820. [47] Kharrazi MR, Sarkani S. Frequency-dependent fatigue damage accumulation in
[11] Altenbach H. Book review: P. K. Mallick, fiber-reinforced composites. Materials, fiber-reinforced plastics. J Compos Mater 1924;35(21):1953.
manufacturing, and design. Z Angew Math Mech 2009;89:921921. [48] Montesano J, Fawaz Z, Bougherara H. Use of infrared thermography to investigate
[12] Paiva MC, Nardin M, Bernardo CA, Schultz J. Influence of thermal history on the the fatigue behaviour of carbon fiber reinforced polymer composite. Compos
results of fragmentation tests on high-modulus carbon-fibre/polycarbonate model Struct 2013;97:76–83.
composites. Compos Sci Technol 1997;57:83943. [49] Crivelli D, Guagliano M, Eaton M, Pearson M, Al-Jumaili S, Holford K, Pullin R.
[13] Davies P, Germain G, Gaurier B, Boisseau A, Perreux D. Evaluation of the dur- Localisation and identification of fatigue matrix cracking and delamination in a
ability of composite tidal turbine blades. Philos Trans R Soc London A Math Phys carbon fibre panel by acoustic emission. Compos Part B 2015;74:1–12.
Eng Sci 2013:371. [50] Hull D, Clyne TW. An introduction to composite materials. second ed. Cambridge:
[14] Bathias C. An engineering point of view about fatigue of polymer matrix composite Cambridge University Press; 1996. Cambridge University.
materials. Int J Fatigue 2006;28:10949. [51] Broyles NS, Verghese KNE, Davis SV, Lesko JJL, Riffle JS. Fatigue performance of
[15] Owen MJ, Howe RJ. The accumulation of damage in a glass-reinforced plastic carbon fibre/vinyl ester composites: the effect of two dissimilar polymeric sizing
under tensile and fatigue loading. J Phys D Appl Phys 1972;5:319. agents. Polymer 1998;39:3417–24.
[16] Nouri H, Lubineau G, Traudes D. An experimental investigation of the effect of [52] Deng S, Ye L. Influence of fiber-matrix adhesion on mechanical properties of
shear-induced diffuse damage on transverse cracking in carbon-fiber reinforced graphite/epoxy composites: I. Tensile, flexure, and fatigue properties. J Reinforc
laminates. Compos Struct 2013;106:52936. Plast Compos 1999;18:1021–40.
[17] Zhang W, Zhou Z, Zheng P, Zhao S. The fatigue damage mesomodel for fiber- [53] Tsuchiyama N. In proc. ICCM-IV progress in science and engineering of compo-
reinforced polymer composite lamina. J Reinforc Plast Compos 2014;33:178393. sites. Tokyo: Japan Society for Composite Materials; 1982.
[18] Gamstedt EK, Talreja R. Fatigue damage mechanisms in unidirectional carbon- [54] Caprino G. Short-fibre thermoset composites. In: Harris B, editor. Fatigue in
fibre-reinforced plastics. J Mater Sci 1999;34:253546. composites. Elsevier (Woodhead Publishing Imprint); 2003. 978-1-85573-608-5.
[19] Dong H, Li Z, Wang J, Karihaloo BL. A new fatigue failure theory for multi- [55] Hitchen SA, Ogin SL, Smith PA. Effect of fibre length on fatigue of short carbon
directional fiber-reinforced composite laminates with arbitrary stacking sequence. fibre/epoxy composite. Composites 1995;26:303–8.
Int J Fatigue 2016;87:294300. [56] Harris B, Reither H, Adam T, Dickson RF, Fernando G. Fatigue behaviour of carbon
[20] Lasri L, Nouari M, Mansori M El. Wear resistance and induced cutting damage of fibre reinforced plastics Composites 1990;21:232–42.
aeronautical FRP components obtained by machining. Wear 2011;271:25428. [57] Shutle K, Friedrich K, Horstenkamp G. Temperature-dependent mechanical be-
[21] Lubineau G. Estimation of residual stresses in laminated composites using field haviour of PI and PES resins used as matrices for short-fibre reinforced laminates. J
measurements on a cracked sample. Compos Sci Technol 2008;68:27619. Mater Sci 1986;21:3561–70.
[22] Quaresimin M, Susmel L, Talreja R. Fatigue behaviour and life assessment of [58] Zago A, Springer GS. Constant amplitude fatigue of short glass and carbon fiber
composite laminates under multiaxial loadings. Int J Fatigue 2010;32:216. reinforced thermoplasticJournal of. Reinforced Plastics and Composites
[23] Zhang W, Zhou Z, Scarpa F, Zhao S. A fatigue damage meso-model for fiber-re- 2001;20:564–95.
inforced composites with stress ratio effect. Mater Des 2016;107:21220. [59] Mandell JF, Huang DD, McGarry FJ. Fatigue of glass and carbon fiber reinforced
[24] Mejlej VG, Osorio D, Vietor T. An improved fatigue failure model for multi- engineering thermoplastics Polymer Composites vol. 2. 1981. p. 137–44.
directional fiber-reinforced composite laminates under any stress ratios of cyclic [60] Pimenta S, Pinho ST. Recycling carbon fibre reinforced polymers for structural
loading. Procedia CIRP 2017;66:2732. applications: Technology review and market outlook. Waste Manag
[25] Brunbauer J, Pinter G. Effects of mean stress and fibre volume content on the 2011;31:378–92.
fatigue-induced damage mechanisms in CFRP. Int J Fatigue 2015;75:2838. [61] Varna J, Joffe R, Berglund LA, Lundstrom TS. Effect of voids on failure mechan-
[26] Karbhari VM, Xian G. Hygrothermal effects on high VF pultruded unidirectional isms in RTM laminates. Compos Sci Technol 1995;52:241249.
carbon/epoxy composites: moisture uptake. Compos B Eng 2009;40:419. [62] Olivier P, Cottu JP, Ferret B. Effects of cure cycle pressure and voids on some
[27] Alessi S, Pitarresi G, Spadaro G. Effect of hydrothermal ageing on the thermal and mechanical properties of carbon/epoxy laminates. Composites 1995;26:509.
delamination fracture behaviour of CFRP composites. Compos B Eng [63] Ruiz E, Achim V, Soukane S, Trochu F, Breardb J. Optimization of injection flow
2014;67:14553. rate to minimize micro/macro-voids formation in resin transfer molder compo-
[28] Barjasteh E, Nutt SR. Moisture absorption of unidirectional hybrid composites. sites. Compos Sci Technol 2006;66:475486.
Compos Part A Appl Sci Manuf 2012;43:15864. [64] Fiedler B, Schulte K. Reliability and life prediction of composite structures.
[29] Mandell JF, Samborsky DD. DOE/MSU composite material fatigue database: test Compos Sci Technol 2006;66:615.
methods, materials, and analysis. Albuquerque, NM, and Livermore, CA (United [65] Guo ZS, Liu L, Zhang BM, Du S. Critical void content for thermoset composite
States). 1997. laminates. J Compos Mater 2009;43:17751790.
[30] Hashin Z, Rotem A. A fatigue failure criterion for fiber reinforced materials. J [66] Zhang A, Zhang D. The mechanical property of CFRP laminates with voids. Adv
Compos Mater 1973;7:44864. Mater Res 2013;652654:2528.
[31] Naderi M, Khonsari MM. A comprehensive fatigue failure criterion based on [67] Liu L, Zhang BM, Wang DF. Effects of cure cycles on void content and mechanical
thermodynamic approach. J Compos Mater 2012;46:43747. properties of composite laminates. Compos Struct 2006;73:303309.
[32] Fawaz Z, Ellyin F. Fatigue failure model for fibre-reinforced materials under [68] Zangenberg J. The effects of fibre architecture on fatigue life-time of composite
general loading conditions. J Compos Mater 1994;28:143251. materials PhD-Thesis DTU Wind Energy; 2013
[33] Tai NH, Ma CCM, Wu SH. Fatigue behaviour of carbon fibre/PEEK laminate [69] Sisodia S, Gamstedt EK, Edgren F, Varna J. Effects of voids on quasi-static and
composites. Composites 1995;26:551–9. tension fatigue behaviour of carbon-fibre composite laminates. J Compos Mater
[34] Gamstedt KE, Talreja R. Fatigue damage mechanisms in unidirectional carbon- 2015;49:21372148.
fibre-reinforced plastics. J Mater Sci 1999;34:2535–46. [70] Horrmann S, Adumitroaie A, Viechtbauer C, Schagerl M. The effect of fiber wa-
[35] Talreja R. Fatigue of composite materials: damage mechanisms and fatigue-life viness on the fatigue life of CFRP materials. Int J Fatigue 2016;90:13947.
diagrams. Proc Roy Soc Lond 1981;A378:461–75. [71] Nonn S, Kralovec C, Schagerl M. Damage mechanisms under static and fatigue
[36] Curtis PT. Tensile fatigue mechanisms in unidirectional polymer matrix composite loading at locally compacted regions in a high pressure resin transfer molded
materials. Int J Fatigue 1991;13:377–82. carbon fiber non-crimp fabric. Compos Part A Appl Sci Manuf 2018;115:5765.
[37] Hojo M, Ochiai S, Gustafson CG, Tanaka K. Effect of matrix resin on delamination [72] Leclerc JS, Ruiz E. Porosity reduction using optimized flow velocity in resin
fatigue crack growth in CFRP laminates. Eng Fract Mech 1994;49(1):35–47. transfer molding. Compos Part A 2008;39:18591868.
[38] Yee A. Modifying matrix materials for tougher composites, (1987), toughened [73] Zhu H, Wu B, Li D, Zhang D, Chen Y. Influence of voids on the tensile performance
composites, STP937-EB, johnston N. West Conshohocken, PA: ASTM International; of carbon/epoxy fabric laminates. J Mater Sci Technol 2011;27:6973.
1987. p. 383–96. [74] Costa ML, Rezende MC, Almeida SFM. Influence of porosity on the interlaminar
[39] Mouritz AP. Structural properties of z-pinned carbon-epoxy T-joints in hot-wet shear strength of carbon/epoxy and carbon/bismaleimide fabric laminates. J.

575
P. Alam, et al. Composites Part B 166 (2019) 555–579

Composite Sci. Technol. 2001;61:2101–8. [106] Tai NH, Ma CCM, Lin JM, Wu GY. Effects of thickness on the fatigue-behaviour of
[75] de Almeida SFM, dos Santos Nogueira Neto Z. Effect of void content on the quasi-isotropic carbon/epoxy composites before and after low energy impacts
strength of composite laminates. Compos Struct 1994;28:139148. Composites. Sci Technol 1999;59:1753–62.
[76] Koissin V, Kustermans J, Lomov SV, Verpoest I, Van Den Broucke B, Witzel V. [107] Nettles A, Hodge A, Jackson J. An examination of the compressive cyclic loading
Structurally stitched NCF preforms: quasi-static response. Compos Sci Technol aspects of damage tolerance for polymer matrix launch vehicle hardware. J
2009;69:27012710. Compos Mater 2011;45:437–58.
[77] Huang Y, Varna J, Talreja R. Statistical methodology for assessing manufacturing [108] Cantwell W, Curtis P, Morton J. Post-impact fatigue performance of carbon fibre
quality related to transverse cracking in cross ply laminates. Compos Sci Technol laminates with non-woven and mized-woven layers. Composites 1983;14:301–5.
2014;95:100106. [110] Beheshty MH, Harris B. A constant-life model of fatigue behaviour for carbon-fibre
[78] Carraro PA, Maragoni L, Quaresimin M. Influence of manufacturing induced de- composites: the effect of impact damage. Compos Sci Technol 1998;58:9–18.
fects on damage initiation and propagation in carbon/epoxy NCF laminates. Adv [111] Mitrovic M, Hahn HT, Carman GP, Shyprykevich P. Effect of loading parameters
Manuf Polym Compos Sci 2015;1:4453. on the fatigue behaviour of impact damaged composite laminates. Compos Sci
[79] Wisnom MR, Reynolds T, Gwilliam N. Reduction in interlaminar shear strength by Technol 1999;59:2059–78.
discrete and distributed voids. Compos Sci Technol 1996;56(1):93–101. [112] Malin LG, Schon J. Buckling behaviour and delamination growth in impacted
[80] Zhu HY, Li DH, Zhang DX, Wu B, Chen Y. Influence of voids on interlaminar shear composite specimens under fatigue load: an experimental study. Compos Sci
strength of carbon/epoxy fabric laminates. Trans Nonferrous Metals Soc China Technol 2001;61:1841–52.
2009;19:470475. [113] Ogasawara T, Sugimoto S, Katoh H, Ishikawa T. Fatigue behaviour and lifetime
[81] Suhot MA, Chambers AR. The effect of voids on the flexural fatigue performance of distribution of impact-damaged carbon fibre/toughened epoxy composites under
unidirectional carbon fibre composites Proc. ICCM16, Kyoto, Japan. July 2007. compressive loading. Adv Compos Mater 2013;22:65–78.
[82] Rotem A, Nelson HG. Failure of a laminated composite under tension-compression [114] Saito H, Kimpara I. Evaluation of impact damage mechanism of multi-axial stiched
fatigue loading. Compos Sci Technol 1989;36:4562. CFRP laminate. Composites Part A: Applied Science and Manufacturing
[83] Chambers AR, Earl JS, Squires CS, Suhot MA. The effect of voids on the flexural 2006;37:2226–35.
fatigue performance of unidirectional carbon fibre composites developed for wind [115] Saito H, Kimpara I. Damage evolution behaviour of CFRP laminates under post-
turbine applications. Int J Fatigue 2005;28:13891398. impact fatigue with water absorption environment. Composites Sci Technol
[84] Gehrig F, Mannov E, Schulte K. Degradation of NCF-epoxy composites containing 2009;69:847–55.
voids. Proc. ICCM July 2009;17. [116] Im KH, Cha CS, Kim SK, Yang IY. Effects of temperature on impact damage in
[85] Seon G, Makeev A, Nikishkov Y, Lee E. Effects of defects on interlaminar tensile CFRP composite laminates. Compos B Eng 2001;32:669–82.
fatigue behavior of carbon/epoxy composites. Compos Sci Technol [117] Uda N, Ono K, Kunoo K. Compression fatigue failure of CFRP laminates with
2013;89:194201. impact damage. Compos Sci Technol 2009;69:2308–14.
[86] Parlevliet PP, Bersee HEN, Beukers A. Measurement of (post-)curing strain de- [118] Kostopoulos V, Baltopoulos A, Karapappas P, Vavouliotis A, Paipetis A. Impact and
velopment with fibre Bragg gratings. Polym Test 2010;29:291301. after-impact properties of carbon fibre reinforced composites enhanced with
[87] Tavakol B, Roozbehjavan P, Ahmed A, Das R, Joven R, Koushyar H, Rodriguez A, multi-wall carbon nanotubes. Compos Sci Technol 2010;70:553–63.
Minaie B. Prediction of residual stresses and distortion in carbon fiber-epoxy [119] Koo JM, Choi JH, Seok CS. Prediction of post-impact residual strength and fatigue
composite parts due to curing process using finite element analysis. J Appl Polym characteristics after impact of CFRP composite structures. Compos B Eng
Sci 2013;128:941950. 2014;61:300–6.
[88] Mesogitis TS, Skordos AA, Long AC. Uncertainty in the manufacturing of fibrous [120] Symons DD, Davis G. Fatigue testing of impact-damaged T300/914 carbon-fibre-
thermosetting composites: a review. Compos. Part A Appl Sci Manuf reinforced plastic. Composites Sci Technol 2000;60:379–89.
2014;57:6775. [121] Freeman B, Schwingler E, Mahinfalah M, Kellogg K. The effect of low-velocity
[89] Hosoi A, Kawada H. Fatigue life prediction for transverse crack initiation of CFRP impact on the fatigue life of sandwich composites. Compos Struct
cross-ply and quasi-isotropic laminates. Materials 2018;11:1182. 2005;70:374–81.
[90] Padmanabhan SK, Pitchumani R. Stochastic analysis of isothermal cure of re- [122] Tai NH, Ma CCM, Lin JM, Wu GY. Effects of the thickness on the fatigue-behaviour
sinsystems. Polym Compos 1999;20(1):7285. of quasi-isotropic carbon/epoxy composites before and after low energy impacts.
[91] Guo Z, Du S, Zhang B. Temperature field of thick thermoset composite laminates Compos Sci Technol 1999;59:1753–62.
during cure process. Compos Sci Technol 2005;65(34):51723. [123] Campbell FC. Manufacturing Processes for advanced composites. Manufacturing
[92] Potter K, Khan B, Wisnom M, Bell T, Stevens J. Variability, fibre waviness and Processes for advanced composites. 2003.
misalignment in the determination of the properties of composite materials and [124] Quaresimin M, Ricotta M. Fatigue response and damage evolution in 2D textile
structures. Compos Appl Sci Manuf 2008;39(9):134354. composites, Fatigue of Textile Composites. Woodhead Publishing; 2015. p.
[93] Potter K. Understanding the origins of defects and variability in composites 193–221.
manufacture. International conference on composite materials (ICCM)-17, [125] Bishop SM. Strength and failure of woven carbon-fibre reinforced plastics for high
Edinburgh, UK. 2009. performance applications. In: Chou T-W, Ko FK, editors. Textile structural com-
[94] Horrmann S, Adumitroaie A, Schagerl M. The effect of ply folds as manufacturing posites. Amsterdam: Elsevier; 1989. p. 173e207.
defect on the fatigue life of CFRP materials. Frat Ed Integrità Strutt 2016;10:7681. [126] Sevenois RDB, Van Paepegem W. Fatigue damage modeling techniques for textile
[95] Mukhopadhyay S, Jones MI, Hallett SR. Compressive failure of laminates con- composites: review and comparison with unidirectional composite modeling
taining an embedded wrinkle; experimental and numerical study. Compos Appl Sci techniques. Appl Mech Rev 2015;67(2):21401.
Manuf 2015;73:13242. [127] Daggumati S, De Baere I, Van Paepegem W, Degrieck J, Xu J, Lomov SV, Verpoest
[96] Davidson P, Waas AM, Yerramalli CS, Chandraseker K, Faidi W. Effect of fiber I. Fatigue and post-fatigue stress–strain analysis of a 5-harness satin weave carbon
waviness on the compressive strength of unidirectional carbon fiber composites. fibre reinforced composite. Compos Sci Technol 2013;74:20–7https://doi.org/10.
53rd AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics and materials 1016/j.compscitech.2012.09.012.
conference. Honolulu, Hawaii: AIAA 2012-1612; 2012. [128] Gyekenyesi AL. Isothermal fatigue, damage accumulation, and life prediction of a
[97] Wang J, Potter K, Etches J. Experimental investigation and characterisation woven PMC. NASA/1998-206593; 1998.
techniques of compressive fatigue failure of composites with fibre waviness at ply [129] Kawai M, Taniguchi T. Off-axis fatigue behavior of plain weave carbon/epoxy
drops. Compos Struct 2013;100:398403. fabric laminates at room and high temperatures and its mechanical modeling.
[98] Horrmann S, Viechtbauer C, Adumitroaie A, Schagerl M. The effect of fiber wa- Compos Appl Sci Manuf 2006;37(2):243–56https://doi.org/10.1016/j.
viness as manufacturing defect on the fatigue life of CFRP materials. Copenhagen, compositesa.2005.07.003.
Denmark: ICCM20; 2015. [130] Pandita SD, Huysmans G, Wevers M, Verpoest I. Tensile fatigue behaviour of glass
[99] Konur O, Matthews FL. Effect of the properties of the constituents on the fatigue plain-weave fabric composites in on- and off-axis directions. Compos Appl Sci
performance of composites: a review. Composites 1989;20(4):317–28. Manuf 2001;32(10):1533–9https://doi.org/10.1016/S1359-835X(01)00053-7.
[100] Jones CJ, Dickson RF, Adam T, Reiter H, Harris B. The environmental fatigue [131] Fruehmann RK, Dulieu-Barton JM, Quinn S. Assessment of fatigue damage evo-
behaviour of reinforced plastics. Proc Roy Soc Lond A 1984;369:315–38. lution in woven composite materials using infra-red techniques. Compos Sci
[101] Curtis PT, Moore BB. A comparison of plain and double waisted coupons for static Technol 2010;70(6):937–46https://doi.org/10.1016/j.compscitech.2010.02.009.
and tensile testing of UD GRP and CFRP, Second International Conference on [132] Pandita Surya D, Verpoest Ignaas. Tension–tension fatigue behaviour of knitted
Composite Structures, Paisley, Scotland, UK, September 1983, Proceedings. London, fabric composites. Compos Struct 2004;64(2):199–209https://doi.org/10.1016/j.
UK: Elsevier Applied Science Publishers; 1983. p. 383–98. compstruct.2003.08.003.
[102] Casas-Rodriguez JP, Ashcroft IA, Silberschmidt VV. Delamination in adhesively [133] Kelkar AD, Tate JS, Bolick R. Structural integrity of aerospace textile composites
bonded CFRP joints: standard fatigue. impact-fatigue and intermittent impact under fatigue loading. Mater Sci Eng B 2006;132(1–2):79–84https://doi.org/10.
Composites Science and Technology 2008;68:2401–9. 1016/j.mseb.2006.02.033.
[103] Nash NH, Young TM, McGrail PT, Stanley WF. Inclusion of a thermoplastic phase [134] Tate Jitendra S, Kelkar Ajit D, Whitcomb John D. Effect of braid angle on fatigue
to improve impact and post-impact performances of carbon fibre reinforced. performance of biaxial braided composites. Int J Fatigue
thermosetting composites - a review Materials and Design 2015;85:582–97. 2006;28(10):1239–47https://doi.org/10.1016/j.ijfatigue.2006.02.009.
[104] Kempf M, Schwagele S, Ferencz A, Altstadt V. Effect of impact damage on the [135] Montesano J, Fawaz Z, Poon C, Behdinan K. A microscopic investigation of failure
compression fatigue performance of glass and carbon fibre reinforced composites mechanisms in a triaxially braided polyimide composite at room and elevated
18th International Conference on Composites Materials (ICCM), 21th to 26th of temperatures. Mater Des 2014;53.
August, 2011, Jeju Island, Korea. 2011. [136] Dransfield K, Baillie C, Mai Y-W. Improving the delamination resistance of CFRP
[105] Alam P. Kloc M, editor. Structures and composition of the crab carapace an ar- by stitching. Compos Sci Technol 1994;50:305–17.
chetypal material in biomimetic mechanical design Results and Problems in Cell [137] Mouritz AP, Bannister MK, Falzon PJ, Leong KH. Review of applications for ad-
Differentiation. Springer-Nature; 2018. p. 569–84. vanced three-dimensional fibre textile composites. Compos Appl Sci Manuf

576
P. Alam, et al. Composites Part B 166 (2019) 555–579

1999;30(12):1445–61. cut fibers. Open J Compos Mater 2017;7:49–62.

[138] Bogdanovich AE, Mohamed MH. Three-dimensional reinforcement for composites. [173] Fernandez G, Usabiaga H, Vandepitte D. Subcomponent development for sand-
SAMPE J 2009;45:8–28. wich composite wind turbine blade bonded joints analysis. Compos Struct
[139] Mouritz AP. Fatigue of 3D textile-reinforced composites, chapter 11 in fatigue of 2017;180(11):41–62.
textile composites. Woodhead Publishing Series Composites Science and Engineering; [174] Alam P, Robert C, Ó Brádaigh CM. Tidal turbine blade composites - a review on the
2015. p. 255–74. effects of hygrothermal aging on the perties of CFRP. Compos B Eng
[140] Mouritz AP, Baini C, Herszberg I. Mode I interlaminar fracture toughness prop- 2018;149:248–59.
erties of advanced textile fibreglass composites. Compos Appl Sci Manuf [175] Sauder C, Lamon J, Pailler R. Thermomechanical properties of carbon fibres at
1999;30(7):859–70. high temperatures (up to 2000°C). Compos Sci Technol 2002;62:499–504.
[141] Gerlach R, Siviour CR, Wiegand J, Petrinic N. In-plane and through-thickness [176] Pradere C, Batsale JC, Goyheneche JM, Pailler R, Dilhaire S. Thermal properties of
properties, failure modes, damage and delamination in 3D woven carbon fibre carbon fibers at very high temperature. Carbon 2009;47:737–43.
composites subjected to impact loading. Compos Sci Technol 2012;72(3):397–411. [177] Daniel IM, Liber T. Lamination residual stresses in fiber composites. Interim report
[142] Mouritz AP, Cox BN. A mechanistic interpretation of the comparative in-plane NASA CR-134826. IITRI D6073-I; 1975.
mechanical properties of 3D woven, stitched and pinned composites. Compos Appl [178] Fahmy A, Cunningham TG. Investigation of thermal fatigue in fiber composite
Sci Manuf 2010;41(6):709–28. materials. Final report. NASA CR-2641; 1976.
[143] Aymerich F. Effect of stitching on the static and fatigue performance of Co-cured [179] Givler RC, Gillespie JW, Pipes RB. Environmental exposure of carbon/epoxy
composite single-lap joints. J Compos Mater 2004;38(3):243–57. composite material systems. ASTM STP 768; 1982. [Composites for Extreme
[144] Aymerich F, Priolo P. Sun CT, Static and fatigue behaviour of stitched graphite/ Environments].
epoxy composite laminates. Compos Sci Technol 2003;63(6):907–17. [180] Camahort JL, Rennhack EH, Coons WC. Effects of thermal cycling environment on
[145] Chang P, Mouritz AP, Cox BN. Flexural properties of z-pinned laminates. Compos graphite/epoxy composites. ASTM STP 602; 1976. [Environmental Effects on
Appl Sci Manuf 2007;38(2):244–51. Advanced Composite Materials].
[146] Chang P, Mouritz AP, Cox BN. Properties and failure mechanisms of z-pinned la- [181] Eselun SA, Neubert HD, Woff EG. Microcracking effects on dimensional stability.
minates in monotonic and cyclic tension. Compos Sci Technol SAMPE J 1979;24:1299–309.
2006;37(10):1501–13. [182] Adams DS, Bowles DE, Herakovich CT. Thermally induced transverse cracking in
[147] Carvelli V, Tomaselli VN, Lomov SV, Verpoest I, Witzel V, Van den Broucke B. graphite/epoxy cross-ply laminates. J Reinforc Plast Compos 1986;5:152–69.
Fatigue and Post-Fatigue tensile behaviour of Non-Crimp stitched and unstitched [183] Hyer MW, Cooper DW, Cohen D. Stresses and deformations in cross-ply composite
Carbon/Epoxy composites. Compos Sci Technol 2010;70(15):2216–24. tubes subjected to a uniform temperature change. J Therm Stresses
[148] Rudov-Clark S, Mouritz AP. Tensile fatigue properties of a 3D orthogonal woven 1986;9:97–117.
composite. Composites Part A 2008;39:1018–24. [184] Kaw AK. Mechanics of composite materials. New York: CRC Press; 1997.
[149] Tong L, Mouritz AP, Bannister MK. 3D fibre reinforced polymer composites. [185] Miyano Y, Nakada M, Kudoh H, Muki R. Prediction of tensile fatigue life under
Oxford: Elsevier; 2002. temperature environment for unidirectional CFRP. Adv Compos Mater
[150] Tana KT, Watanabe N, Iwahori Y. Effect of stitch density and stitch thread 1999;8:235–46.
thickness on low-velocity impact damage of stitched composites. Compos Appl Sci [186] Miyano Y, Nakada M. Formulation of time- and temperature- dependent strength
Manuf 2010;41(12):1857–68. of unidirectional carbon fiber reinforced plastics. J Compos Mater
[151] Dransfield KA, Jain LK, Mai YW. On the effects of stitching in CFRPsI. mode I 2012;47:1897–906.
delamination toughness. Compos Sci Technol 1998;58(6):815–27. [187] Peters PWM, Andersen SI. The influence of matrix fracture strain and interface
[152] Lopresto V, Melito V, Leone C, Caprino G. Effect of stitches on the impact beha- strength on cross ply cracking in CFRP in the temperature range of -100°C to
viour of graphite/epoxy composites. Compos Sci Technol 2006;66(2):206–14. +100°C. J Compos Mater 1988;23:944–60.
[153] Mouritz AP, Cox BN. A mechanistic approach to the properties of stitched lami- [188] Grogan DM, Leen SB, Semprimoschnig COA, Ó Brádaigh CM. Damage character-
nates. Composites Part A 2000;31:1–27. isation of cryogenically cycled carbon/PEEK laminates. Compos Appl Sci Manuf
[154] Yudhanto A, Watanabe N, Iwahori Y, Hoshi H. Effect of stitch density on fatigue 2014;66:237–50.
characteristics and damage mechanisms of stitched carbon/epoxy composites. [189] Miyano Y, McMurray MK, Ktade N, Nakada M. Loading rate and temperature
Compos Appl Sci Manuf 2014;60(5):52–65. dependence of flexural behaviour of unidirectional pitch-based CFRP laminates.
[155] Freitas G, Magee C, Dardzinski P, Fusco T. Fibre insertion process for improved Composites 1995;26:713–7.
damage tolerance in aircraft laminates. J Adv Mater 1994;25:36–43. [190] Kawai M, Taniguchi T. Off-axis fatigue behavior of plain weave carbon/epoxy
[156] Isa MD, Feih S, Mouritz AP. Compression fatigue properties of quasi-isotropic z- fabric laminates at room temperatures and its mechanical modeling. Compos Appl
pinned carbon/epoxy laminate with barely visible impact damage. Compos Struct Sci Manuf 2006;37:243–56.
2011;93:2222–30. [191] Khan RKZ, Al-Sulaiman F, Merah N. Fatigue life estimates in woven carbon fabric/
[157] Kelkar AD, Tate JS, Bolick R. Structural integrity of aerospace composites under epoxy composites at non-ambient temperatures. J Compos Mater
fatigue loading. Mater Sci Eng B 2006;132:79–84. 2002;36:2517–35.
[158] Bilisik K. Three-dimensional braiding for composites: a review. Textil Res J [192] Wu CML. Thermal and mechanical fatigue analysis of CFRP laminates. Compos
2013;83(13):14141436. Struct 1993;25:339–44.
[159] Carvelli V, Pazmino J, Lomov SV, Bogdanovich AE, Mungalov DD, Verpoest I. [193] Kobayashi S, Terada K, Takeda N. Evaluation of long-term durability in high
Quasi-static and fatigue tensile behavior of a 3D rotary braided carbon/epoxy temperature resistant CFRP laminates under thermal fatigue loading. Compos B
composite. J Compos Mater 2012;47(25):31953209. Eng 2003;34:753–9.
[160] Curtis G, Milne J, Reynolds W. Non-Hookean behaviour of strong carbon fibres. [194] Henaff-Gardin C, Lafarie-Frenot MC. Specificity of matrix-cracking development in
Nature 1968;220:1024–5. CFRP laminates under mechanical or thermal loadings. Int J Fatigue
[161] Carvelli V, Lomov SV. Fatigue damage evolution in 3D textile composites, chapter 2002;24:171–7.
10 in fatigue of textile composites. Woodhead Publishing Series in Composites [196] Kawai M, Maki N. Fatigue strengths of cross-ply CFRP laminates at room and high
Science and Engineering; 2015. p. 223–53. temperatures and its phenomenological modeling. Int J Fatigue
[162] Pingkarawat K, Mouritz AP. Improving the mode I delamination fatigue resistance 2006;28:1297–306.
of composites using z-pins. Compos Sci Technol 2014;92:70–6. [197] Miyano Y, Nakada M, Judoh H, Muki R. Prediction of tensile fatigue life for uni-
[163] Tan KT, Watanabe N, Iwahori Y, Ishikawa T. Influence of stitch density and stitch directional CFRP. J Compos Mater 2000;34:538–50.
thread thickness on compression after impact strength of stitched composites. 16th [198] Miyano Y, Nakada M, Muki R. Applicability of fatigue life prediction method to
international conference on composite structures, vol. 16. Porto: ICCS; 2011. polymer composites. Mech Time-Dependent Mater 1999;3:141–57.
[164] Tong L, Jain LK, Leong KH, Kelly D, Hertzberg I. Failure of transversely stitched [199] Vanlandingham MR, Eduljee RF, Gillespie JR. Moisture diffusion in epoxy systems.
RTM lap joints. Compos Sci Technol 1998;58:221–7. J Appl Polym Sci 1999;71:787–98.
[165] McCarthy MA, McCarthy CT, Lawlor VP, Stanley WF. Three-dimensional finite [200] Barrer RM. Diffusion in and through solids. New York: The Macmillan Company;
element analysis of single-bolt, single-lap composite bolted joints: part Imodel 1941. [Cambridge England: University Press].
development and validation. Compos Struct 2005;71(2):140–58. [201] Gac PYL, Arhant M, Gall ML, Davies P. Yield stress changes induced by water in
[166] OHiggins RM, McCarthy MA, McCarthy CT. Comparison of open hole tension polyamide 6: characterisation and modelling. Polym Degrad Stabil 2017:272–80.
characteristics of high strength glass and carbon fibre-reinforced composite ma- [202] Judd NCW. Absorption of water into carbon fibre composites. The British Polymer
terials. Compos Sci Technol 2008;68(13):2770–8. Journal 1977:36–40. March 1977.
[167] Dai S, Cunningham PR, Marshall S, Silva C. Open hole quasi-static and fatigue [203] Karbhari VM, Xian G. Hygrothermal effects on high VF pultruded unidirectional
characterization of 3D woven composites. Compos Struct 2015;131:765–74. carbon/epoxy composites: moisture uptake. Composites Part B 2009;40:41–9.
[168] Yudhanto A, Iwahori Y, Watanabe N, Hoshi H. Open hole fatigue characteristics [204] Dellanno G, Lees R. Effect of water immersion on the interlaminar and flexural
and damage growth of stitched plain weave carbon/epoxy laminates. Int J Fatigue performance of low cost liquid resin infused carbon fabric composites. Composites
2012;43:12–22. Part B 2012;43:1368–73.
[169] Diamantakos C, Fritz RJ. Tensile fatigue of notched carbon/epoxy specimens [205] Wang Y, Hahn TH. AFM characterisation of the interfacial properties of carbon
search for optimum model. J Reinforc Plast Compos 1988;7:165–78. fibre reinforced polymer composites subjected to hygrothermal treatments.
[170] Vieille B. Fatigue accumulated damage in notched quasi-isotropic composites Compos Sci Technol 2007;67:92–101.
under high-temperature conditions: a discussion on the influence of matrix nature [206] Collings TA, Stone DEW. Hygrothermal effects in CFRP laminates: strains induced
on the stress energy release rate. J Compos Mater 2018;52(17):2397–412. by temperature and moisture. Composites 1985;16:307–16.
[171] Tsai KH, Chiu CH, Wu TH. Fatigue behavior of 3D multi-layer angle interlock [207] Selzer R, Friedrich K. Mechanical properties and failure behaviour of carbon fibre-
woven composite plate. Compos Sci Technol 2000;60(2):241–8. reinforced polymer composites under the influence of moisture. Composites Part A
[172] Sudarsono S, Oji K. Fatigue behavior of open-holed CFRP laminates with initially 1997;28A:595–604.

577
P. Alam, et al. Composites Part B 166 (2019) 555–579

[208] Nakai Y, Hiwa C. Effects of loading frequency and environment on delamination [243] Beheshty MH, Harris BA. constant-life model of fatigue behavior for carbon-fibre
fatigue crack growth of CFRP. Int J Fatigue 2002;24:161–70. composites: the effect of impact damage. Compos Sci Technol 1998;58:918.
[209] Matsuda S, Hojo M, Ochiai S. Effect of water environment on mode II delamination [244] Kawai M. Fatigue life prediction of composite materials under constant amplitude
fatigue in interlayer-toughened CFRP. JSME International Journal 1999;42:421–8. loading. In: Vassilopoulos AP, editor. Fatigue life prediction of composites and
[210] Komai K, Minoshima K, Shibutani T, Nomura T. The influence of water on the composite structures. Cambridge, UK: Woodhead Publishing Limited; 2010. p.
mechanical properties and fatigue strength of angle-ply carbon/epoxy composites. 177219.
JSME International Journal 1989;32:588–95. [245] Kawai M. Fatigue of compositeslife prediction methods. second ed.Nicolais L,
[211] Kawada H, Kobiki A, Koyanagi J, Hosoi A. Long-term durability of polymer matrix Borzacchiello A, editors. Encyclopedia of composites, vol. 2. Hoboken, NJ: John
composites under hostile environments. Mater Sci Eng 2005;412:159–64. Wiley and Sons; 2012. p. 883923.
[212] Chiou P, Bradley WL. Effects of seawater absorption on fatigue crack development [246] Goodman J. Mechanics applied to engineering. Harlow, UK: Longman Green;
in carbon/epoxy EDT specimens. Composites 1995;26:869–76. 1899.
[213] Hojo M, Tanaka K, Gistafson CG, Hayashi R. Propagation of delamination fatigue [247] Harris B, Gathercole N, Lee JA, et al. Life-prediction for constant-stress fatigue in
cracks in CFRP in water. JSME International Journal 1989;32:292–9. carbon-fibre composites. Phil Trans Roy Soc Lond 1997;A355:12591294.
[214] Dickinson RF, Jones CJ, Harris B, Leach DC, Moore DR. The environmental fatigue [248] Kawai M, Koizumi M. Nonlinear constant fatigue life diagrams for carbon/epoxy
behaviour of carbon fibre reinforced polyether ether ketone. J Mater Sci laminates at room temperature. Compos Appl Sci Manuf 2007;38:23422353.
1985;20:60–70. [249] Kawai M, Murata T. A three-segment anisomorphic constant life diagram for the
[215] Meziere Y, Bunsell AR, Favry Y, Teissedre JC, Do AT. Large strain cyclic fatigue fatigue of symmetric angle-ply carbon/epoxy laminates at room temperature.
testing of unidirectional carbon fibre reinforced epoxy resin. Compos Appl Sci Compos Appl Sci Manuf 2010;41:14981510.
Manuf 2005;36:1627–36. [250] Kawai M, Matsuda Y. Anisomorphic constant fatigue life diagrams for a woven
[216] Kawai M, Yagihashi Y, Hoshi H, Iwahori Y. Anisomorphic constant fatigue life fabric carbon/epoxy laminate at different temperatures. Compos Appl Sci Manuf
diagrams for quasi-isotropic woven fabric carbon/epoxy laminates under different 2012;43:647657.
hygrothermal environments. Adv Compos Mater 2013;22:79–98. [251] Kawai M, Matsuda Y, Yoshimura R. A general method for predicting temperature-
[217] Zhang A, Lu H, Zhang D. Synergistic effect of cyclic mechanical loading and dependent anisomorphic constant fatigue life diagram for a woven fabric carbon/
moisture absorption on the bending fatigue performance of carbon/epoxy com- epoxy laminate. Compos Appl Sci Manuf 2012;43:915925.
posites. J Mater Sci 2014;49:314–20. [252] Kawai M, Yagihashi Y, Hoshi H, Iwahori Y. Anisomorphic constant fatigue life
[218] Guen-Geffroy AL, Gac PYL, Diakhate M, Habert B, Davies P. Lon-term durability of diagrams for quasi-isotropic woven fabric carbon/epoxy laminates under different
CFRP under fatigue loading for marine applications. MATEC Web of Conferences: hygro-thermal environments. Adv Compos Mater 2013;22:7998.
Fatigue 2018;165. 07001[2018]. [253] Kawai M, Itoh N. A failure-mode based anisomorphic constant life diagram for a
[219] Stinchbomb WW, Bakis CE. Fatigue behavior of composite laminates. Compos unidirectional carbon/epoxy laminate under off-axis fatigue loading at room
Mater 1991;4:10580. temperature. J Compos Mater 2014;48:571592.
[220] Quaresimin M, Ricotta M. Fatigue behaviour and damage evolution of single lap [254] Kawai M, Yano K. Anisomorphic constant fatigue life diagrams of constant prob-
bonded joints in composite material. Compos Sci Technol 2006;66:17687. ability of failure and prediction of PSN curves for unidirectional carbon/epoxy
[221] May M, Hallett SR. An advanced model for initiation and propagation of damage laminates. Compos Appl Sci Manuf 2016;83:323334.
under fatigue loading part I: model formulation. Compos Struct 2011;93:23409. [255] Kawai M, Yano K. Probabilistic anisomorphic constant fatigue life diagram ap-
[222] May M, Hallett SR. Damage initiation in polymer matrix composites under high- proach for prediction of PSN curves for woven carbon/epoxy laminates at any
cycle fatigue loading A question of definition or a material property? Int J Fatigue stress ratio. Compos Appl Sci Manuf 2016;80:244258.
2016;87:5962. [256] Vassilopoulos AP, Keller T. Experimental characterization of fiber-reinforced
[223] Paris P, Erdogan F. A critical analysis of crack propagation laws. J Basic Eng composite materials. 2011. p. 2567.
1963;85:528. [257] Hashin Z. Failure criteria for unidirectional fiber composites. J Appl Mech
[224] Corten HT. Composite materials: testing and design (second conference). ASTM 1980;47:32934.
International; 1972. [258] Goto K, Arai M, Nishimura M, Dohi K. Strength evaluation of unidirectional carbon
[225] DorMohammdi S, Godines C, Abdi F, Huang D, Repupilli M, Minnetyan L. Damage- fiber-reinforced plastic laminates based on tension compression biaxial stress tests.
tolerant composite design principles for aircraft components under fatigue service Adv Compos Mater 2017:114.
loading using multi-scale progressive failure analysis. J Compos Mater [259] Miyano Y, Nakada M, Kudoh H, Muki R. Determination of tensile fatigue life of
2017;51:2181202. unidirectional CFRP specimens by strand testing. Mech Time-Dependent Mater
[226] OBrien T, editor. Long-term behavior of composites. 100 barr harbor drive, PO box 2000;4:12737.
C700. West Conshohocken, PA: ASTM International; 1983. 19428-2959. [260] Nakada M, Miyano Y, Kinoshita M, Koga R, Okuya T, Muki R. Time temperature
[227] Shivakumar K, Chen H, Abali F, Le D, Davis C. A total fatigue life model for mode I dependence of tensile strength of unidirectional CFRP. J Compos Mater
delaminated composite laminates. Int J Fatigue 2006;28:3342. 2002;36:256781.
[228] Holmes JW, Liu L, Srensen BF, Wahlgren S. Experimental approach for mixed- [261] Jeong TK, Ueda M. Longitudinal compressive failure of multiple-fiber model
mode fatigue delamination crack growth with large-scale bridging in polymer composites for a unidirectional carbon fiber reinforced plastic. Open J Compos
composites. J Compos Mater 2014;48:311128. Mater 2016;6:817.
[229] Khan R, Alderliesten R, Yao L, Benedictus R. Crack closure and fibre bridging [262] Jumahat A, Soutis C, Jones FR, Hodzic A. Fracture mechanisms and failure ana-
during delamination growth in carbon fibre/epoxy laminates under mode I fatigue lysis of carbon fibre/toughened epoxy composites subjected to compressive
loading. Compos Part A Appl Sci Manuf 2014;67:20111. loading. Compos Struct 2010;92:295305.
[230] Yao L, Alderliesten R, Zhao M, Benedictus R. Bridging effect on mode I fatigue [263] Yokozeki T, Ogasawara T, Ishikawa T. Effects of fiber nonlinear properties on the
delamination behavior in composite laminates. Compos Part A Appl Sci Manuf compressive strength prediction of unidirectional carbon fiber composites.
2014;63:1039. Compos Sci Technol 2005;65:21407.
[231] Kawai M. A phenomenological model for off-axis fatigue behavior of unidirec- [264] Nakanishi Y, Hana K, Hamada H. Fractography of fracture in CFRP under com-
tional polymer matrix composites under different stress ratios. Compos Part A pressive load. Compos Sci Technol 1997;57:113947.
Appl Sci Manuf 2004;35:95563. [265] Thom H. A review of the biaxial strength of fibre-reinforced plastics. Compos Part
[232] Petermann J, Schulte K. The effects of creep and fatigue stress ratio on the long- A Appl Sci Manuf 1998;29:86986.
term behaviour of angle-ply CFRP. Compos Struct 2002;57:20510. [266] Pagano NJ, Pipes RB. The influence of stacking sequence on laminate strength. J
[233] Reis PNB, Ferreira JAM, Costa JDM, Richardson MOW. Fatigue life evaluation for Compos Mater 1971;5:507.
carbon/epoxy laminate composites under constant and variable block loading. [267] Bailey JE, Curtis PT, Parvizi A. On the transverse cracking and longitudinal
Compos Sci Technol 2009;69:15460. splitting behaviour of glass and carbon fibre reinforced epoxy cross ply laminates
[234] Kawai M, Yang K, Oh S. Effect of alternating R-ratios loading on fatigue life of and the effect of Poisson and thermally generated strain. Proc R Soc A Math Phys
woven fabric carbon/epoxy laminates. J Compos Mater 2015;49:3387405. Eng Sci 1979;366:599623.
[235] Sturgeon JB. Creep, repeated loading, fatigue and crack growth in +/-45 oriented [268] Wicaksono S, Chai GB. A review of advances in fatigue and life prediction of fiber-
carbon fibre reinforced plastics. J Mater Sci 1978;13:14908. reinforced composites. Proc Inst Mech Eng Part L J Mater Des Appl
[236] Gude M, Hufenbach W, Koch I, Koschichow R, Schulte K, Knoll J. Fatigue testing of 2013;227:17995.
carbon fibre reinforced polymers under VHCF loading. Procedia Mater Sci [269] Armanios E, Bucinell R, Wilson D, Asp L, Sjogren A, Greenhalgh E. Delamination
2013;2:1824. growth and thresholds in a carbon/epoxy composite under fatigue loading. J
[237] Noda J, Nakada M, Miyano Y. Fatigue life prediction under variable cyclic loading Compos Technol Res 2001;23:55.
based on statistical linear cumulative damage rule for CFRP laminates. J Reinforc [270] Lin CT, Kao PW. Fatigue delamination growth in carbon fibre-reinforced alumi-
Plast Compos 2007;26:66580. nium laminates. Compos Part A Appl Sci Manuf 1996;27:915.
[238] Irving C, Archer E, PE S. Polymer composites in the aerospace industry vol. 50. [271] Gustafson C-G, Hojo M. Delamination fatigue crack growth in unidirectional
Elsevier Science; 2014. graphite/epoxy laminates. J Reinforc Plast Compos 1987;6:3652.
[239] Harris B, Reiter H, Adam T, et al. Fatigue behaviour of carbon fibre reinforced [272] Dahlen C, Springer GS. Delamination growth in composites under cyclic loads. J
plastics. Composites 1990;21(3):232242. Compos Mater 1994;28:73281.
[240] Harris B, Gathercole N, Lee JA, et al. Life-prediction for constant-stress fatigue in [273] Pagano NJ, Schoeppner GA. Delamination of polymer matrix composites: pro-
carbon-fibre composites. Phil Trans Roy Soc Lond 1997;A355:12591294. blems and assessment, comprehensive composite materials. 2000. p. p433–528.
[241] Adam T, Gathercole N, Reiter H, et al. Fatigue life prediction for carbon fibre [274] Zhang Han, Bilotti Emiliano, Peijs Ton. The use of carbon nanotubes for damage
composites. Adv Compos Lett 1992;1:2326. sensing and structural health monitoring in laminated composites: a review.
[242] Gathercole N, Reiter H, Adam T, et al. Life prediction for fatigue of T800/5245 Nanocomposites 2015;1(4):167–84.
carbon-fibre composites: I. constant- amplitude loading. Fatigue 1994;16:523532. [275] Paulo Davim J, Campos Rubio J, Abrao AM. A novel approach based on digital

578
P. Alam, et al. Composites Part B 166 (2019) 555–579

image analysis to evaluate the delamination factor after drilling composite lami- [304] Whitworth HA. A stiffness degradation model for composite laminates under fa-
nates. Compo Sci Technol 2007;67(9):1939–45https://doi.org/10.1016/j. tigue loading. Compos Struct 1997;40:95101.
compscitech.2006.10.009. [305] Whitworth HA. Evaluation of the residual strength degradation in composite la-
[276] Bak BLV, Sarrado C, Turon A, Costa J. Delamination under fatigue loads in com- minates under fatigue loading. Compos Struct 2000;48:261264.
posite laminates: a review on the observed phenomenology and computational [306] Yao WX, Himmel N. A new cumulative fatigue damage model for fibre-reinforced
methods. Applied Mechanics Reviews; 2014. plastics. Compos Sci Technol 2000;60:5964.
[277] Tabiei A, Zhang W. Composite laminate delamination simulation and experiment: [307] Adam T, Gathercole N, Reiter H, Harris B. Life prediction for fatigue of T800/5245
a review of recent development. ASME. Applied Mechanics Reviews 2018;70(3). carbon-fiber composites: II - variable-amplitude loading. Int J Fatigue
030801-030801-23. 1994;16:533–47.
[278] Kevin O'Brien T, Chawan Arun D, Krueger Ronald, Paris Isabelle L. Transverse [308] Berardi VP, Perrella M, Feo L, Cricri G. Creep behavior of GFRP laminates and
tension fatigue life characterization through flexure testing of composite mate- their phases: experimental investigation and analytical modelling. Composites
rials. Int J Fatigue 2002;24(2–4):127–45https://doi.org/10.1016/S0142- Part B 2017;122:136–44.
1123(01)00104-9. [309] Ascione L, Berardi VP, D'Aponte A. Long-term behavior of PC beams externally
[279] ASTM E647-13. Standard test method for measurement of fatigue crack growth plated with prestressed FRP systems: a mechanical model. Composites Part B
rates. West Conshohocken, PA: ASTM International; 2013. 2011;42:1196–201.
[280] ASTM D6115-97. Standard test method for mode I fatigue delamination growth [310] Berardi VP, Mancusi G. A mechanical model for predicting the long term behavior
onset of unidirectional fiber reinforced polymer matrix composites. West of reinforced polymer concretes. Mech Res Commun 2013;50:1–7.
Conshohocken, PA: ASTM International; 2011. [311] Nedjar B. Directional damage gradient modeling of fiber/matrix debonding in
[281] Robinson P, Song DQ. The development of an improved mode III delamination test viscoelastic UD composites. Compos Struct 2016;153:895–901.
for composites. Compos Sci Technol 1994;52(2):217–33https://doi.org/10.1016/ [312] Yang Jn, Yang Sh, Jones DL. A stiffness-based statistical model for predicting the
0266-3538(94)90207-0. fatigue life of graphite/epoxy laminates. Composites Technology and Research
[282] Hojo M, Ando T, Tanaka M, Adachi T, Ochiai S, Endo Y. Modes I and II inter- 1989;11:129–34.
laminar fracture toughness and fatigue delamination of CF/epoxy laminates with [313] Khan Z, Al-Sulaiman FA, Farooqi JK, Younas M. Fatigue life predictions in woven
self-same epoxy interleaf. Int J Fatigue 2006;28(10):1154–65. carbon fabric/polyester composites based on modulus degradation. J Reinforc
[283] O Brien T. Wilkins D, editor. Mixed-mode strain-energy-release rate effects on Plast Compos 2001;20:377398.
edge delamination of composites, effects of defects in composite materials, [314] Shokrieh MM, Lessard LB. Progressive fatigue damage modeling of composite
STP836-EB. West Conshohocken, PA: ASTM International; 1984. p. 125–42. materials, part I: Modeling. J Compos Mater 2000;34:10561080.
[284] O Brien T, Murri G, Salpekar S. Lagace P, editor. Interlaminar shear fracture [315] Shokrieh MM, Lessard LB. Progressive fatigue damage modeling of composite
toughness and fatigue thresholds for composite materials, composite materials: materials, Part II: material characterization and model verification. J Compos
fatigue and fracture, second volume, STP1012-EB. West Conshohocken, PA: ASTM Mater 2000;34:10811116.
International; 1989. p. 222–50. [316] Papanikos P, Tserpes KI, Pantelakis S. Modelling of fatigue damage progression
[285] Martin R, Murri G. STP1059- EB, Garbo S, editors. Characterization of mode I and and life of CFRP laminates. Fatigue Fract Eng Mater Struct 2003;26:3747.
mode II delamination growth and thresholds in AS4/PEEK composites, composite [317] Kennedy CR, Ó Brádaigh CM, Leen SB. A multiaxial fatigue damage model for fibre
materials: testing and design (ninth volume). West Conshohocken, PA: ASTM reinforced polymer composites. Compos Struct 2013;106:201–10.
International; 1990. p. 251–70. [318] Puck A, Schurmann H. Failure analysis of FRP laminates by means of physically
[286] Tanaka H, Tanaka K. Mixed-mode growth of interlaminar cracks in carbon/epoxy based phenomenological models. Compos Sci Technol 2002;62:1633–62.
laminates under cyclic loading. Proceedings of the 10th international conference [319] Puck A, Kopp J, Knops M. Guidelines for the determination of the parameters in
on composite materials, vol. 1. 1995. p. 181189. Whistler, Canada. Pucks action plane strength criterion. Compos Sci Technol 2002;62:371–8.
[287] Kumar SB, Sridhar I, Sivashanker S. Influence of humid environment on the per- [320] Ye Lin. Role of matrix resin in delamination onset and growth in composite la-
formance of high strength structural carbon fiber composites. Mater Sci Eng, A minates. Compos Sci Technol 1988;33:257277.
2008;498:1748. [321] Harper PW, Hallett SR. A fatigue degradation law for cohesive interface elements -
[288] Zhong Y, Joshi SC. Impact behavior and damage characteristics of hygrothermally development and application to composite materials. Int J Fatigue
conditioned carbon epoxy composite laminates. Mater Des 2015;65:25464. 2010;32:17741787.
[289] Shenoi RA, Wellicome JF. Composite materials in maritime structures Volume 1. [322] Tserpes KI, Labeas G, Papanikos P, Kermanidis T. Strength prediction of bolted
Fundamental aspects. Cambridge University Press; 1993. joints in graphite/epoxy composite laminates. Compos B Eng 2002;33:521529.
[290] Broutman LJ. Fracture and fatigue: composite materials vol. 5. New York and [323] Blanco N, Gamstedt EK, Asp LE, Costa J. Mixed-mode delamination growth in
London: Academic Press; 1974. carbon-fibre composite laminates under cyclic loading. Int J Solids Struct
[291] Harris B, Reiter H, Adam T, Dickson RF, Fernando G. Fatigue behaviour of carbon 2004;41:42194235.
fibre reinforced plastics. Composites 1990;21:23242. [324] May M, Hallett SR. A combined model for initiation and propagation of damage
[292] Liotier P-J, Vautrin A, Beraud J-M. Microcracking of composites reinforced by under fatigue loading for cohesive interface elements. Compos Appl Sci Manuf
stitched multiaxials subjected to cyclical hygrothermal loadings. Compos Part A 2010;41:17871796.
Appl Sci Manuf 2011;42:42537. [325] Allegri G, Jones MI, Wisnom MR, Hallett SR. A new semi-empirical model for
[293] Yavuz AK, Papoulia KD, Phoenix SL, Hui CY. Stability analysis of stitched com- stress ratio effect on mode II fatigue delamination growth. Compos Appl Sci Manuf
posite plate system with delamination under hygrothermal pressure. AIAA J 2011;42:733740.
2006;44:157985. [326] Allegri G, Wisnom MR, Hallett SR. A new semi-empirical law for variable stress-
[294] Botelho EC, Pardini LC, Rezende MC. Hygrothermal effects on the shear properties ratio and mixed-mode fatigue delamination growth. Compos Appl Sci Manuf
of carbon fiber/epoxy composites. J Mater Sci 2006;41:71118. 2013;48:192200.
[295] Dickson RF, Jones CJ, Harris B, Leach DC, Moore DR. The environmental fatigue [327] Kawashita LF, Hallett SR. A crack tip tracking algorithm for cohesive interface
behaviour of carbon fibre reinforced polyether ether ketone. J Mater Sci element analysis of fatigue delamination propagation in composite materials. Int J
1985;20:6070. Solids Struct 2012;49:28982913.
[296] Jones CJ, Dickson RF, Adam T, Reiter H, Harris B. The environmental fatigue [328] Tao C, Qiu J, Yao W, Ji H. A novel method for fatigue delamination simulation in
behaviour of reinforced plastics. Proc R Soc Lond A Math Phys Sci composite laminates. Compos Sci Technol 2016;128:104115.
1984;396:31538. [329] Talreja R. A synergistic damage mechanics approach to durability of composite
[297] Pascoe JA, Alderliesten RC, Benedictus R. Methods for the prediction of fatigue material systems. Progress in durability Analysis of composite systems, cardon, fukuda
delamination growth in composites and adhesive bonds - a critical review. Eng and reifsnider. Rotterdam: Balkema; 1996. p. 117–29.
Fract Mech 2013;112113:7296. [330] Talreja R. A continuum mechanics characterization of damage in composite ma-
[298] Sevenois RDB, Van Paepegem W. Fatigue damage modeling techniques for textile terials. Proc Math Phys Eng Sci 1985;399:195216.
composites: review and comparison with unidirectional composite modeling [331] Talreja R. Internal variable damage mechanics of composite materials, yielding,
techniques. Appl Mech Rev 2015;67:21401. damage and failure of anisotropic solids. London: EGF5, Mechanical Engineering
[299] Degrieck J, Van Paepegem W. Fatigue damage modeling of fibre-reinforced Publications; 1990. p. 509–53. 1991.
composite materials: Review. Appl Mech Rev 2001;54:279. [332] Talreja R. Damage and fatigue in composites - a personal account. Compos Sci
[300] Harris B, Gathercole N, Lee JA, Reiter H, Adam T. Life-prediction for constant- Technol 2008;68:25852591.
stress fatigue in carbon-fibre composites. Phil Trans: Mathematical, Physical and [333] Singh CV, Talreja R. A synergistic damage mechanics approach for composite la-
Engineering Sciences 1997;355:12591294. minates with matrix cracks in multiple orientations. Mech Mater 2009;41:954968.
[301] Ramani S, Williams D (1076) Notched and unnotched fatigue behavior of angle- [334] Singh CV, Talreja R. Evolution of ply cracks in multidirectional composite lami-
ply graphite/epoxy composites. NASA Tech Memo, 73, 191. nates. Int J Solids Struct 2010;47:13381349.
[302] Kawai M, Matsuda Y, Yoshimura R. A general method for predicting temperature- [335] Montesano J, Singh CV. Predicting evolution of ply cracks in composite laminates
dependent anisomorphic constant fatigue life diagram for a woven fabric carbon/ subjected to biaxial loading. Compos B Eng 2015;75:264273.
epoxy laminate. Compos Appl Sci Manuf 2012;43:915925. [336] Haojie S, Weixing Y, Yitao W. Synergistic damage mechanic model for stiffness
[303] Kawai M, Yagihashi Y, Hoshi H, Iwahori Y. Anisomorphic constant fatigue life properties of early fatigue damage in composite laminates. Procedia Engineering
diagrams for quasi-isotropic woven fabric carbon/epoxy laminates under different 2014;74:199209.
hygro-thermal environments. Adv Compos Mater 2013;22:7998.

579