Vous êtes sur la page 1sur 13

International Journal of Fatigue 102 (2017) 171–183

Contents lists available at ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier.com/locate/ijfatigue

Fatigue of short fiber thermoplastic composites: A review of recent

experimental results and analysis
Seyyedvahid Mortazavian, Ali Fatemi ⇑
Mechanical, Industrial and Manufacturing Engineering Department, The University of Toledo, 2801 West Bancroft Street, Toledo, OH 43606, USA

a r t i c l e i n f o a b s t r a c t

Article history: Cyclic deformation and fatigue behavior of two short fiber thermoplastic composites (SFTCs) under a
Received 15 December 2016 number of loading and environmental conditions are investigated. The considered environmental effects
Received in revised form 24 January 2017 include those of low and elevated temperatures as well as moisture (or water absorption). Fatigue behav-
Accepted 25 January 2017
ior is also explored under the action of non-zero mean stress (or R ratio) in addition to fully-reversed
Available online 7 February 2017
(R = 1), as well as various cyclic loading frequencies. Material anisotropy and geometrical discontinuity
effects (i.e. stress concentration) are other aspects considered in this study. Mechanisms of fatigue failure
are also assessed under environmental effects. Based on experimental observations and analysis, a num-
Short fiber
ber of analytical and empirical models are developed for estimating fatigue behavior under different con-
Thermoplastic composites ditions. Empirical equations are presented to characterize self-heating under cyclic loading. Tsai-Hill
Modeling criterion is applied to account for the effect of fiber orientation on fatigue life. Mean stress effect is cor-
rected with several mean stress parameters and a shift factor of Arrhenius type is defined to characterize
the effect of temperature on fatigue life. Two methodologies are presented to estimate fatigue properties
based on tensile properties, in addition to approximation of strain-life curves based on load-controlled
fatigue data. Estimation of notched fatigue behavior based on smooth (un-notched) fatigue behavior is
also presented.
Ó 2017 Elsevier Ltd. All rights reserved.

1. Introduction interface [2,3]. Thermoplastic materials exhibit time dependent

properties and relatively low melting temperatures. As a result,
Application of short fiber thermoplastic composites (SFTCs) is significant effect of load frequency is observed on fatigue behavior
increasingly growing due to their remarkable properties. Light of SFTCs [4].
weight, low manufacturing cost with a high volume production Increased fatigue performance of SFTCs is a function of fiber
rate, and the capability to be molded in complex geometries are reinforcement, its orientation and distribution which in turn is
the main characteristics of SFTCs. A wide range of effects related associated with the geometry of fibers and the component, vis-
to microstructure, environment and load conditions are involved coelastic behavior of the matrix, and flow field during the injection
in fatigue design of SFTCs. However, a relatively small number of molding process [5]. A shell-core morphology across the thickness
studies have been conducted on fatigue behavior characterization of a molded part has frequently been reported for SFTCs, where
of SFTCs, although components made of these materials are typi- higher fiber alignment exists in two shell layers compared with
cally subjected to cyclic loads. Due to the complexity as well as a the core layer [6]. Recently, micro-tomography studies have been
large number of parameters influencing mechanical behavior of performed on SFTCs for their fatigue damage investigation [7].
SFTCs, fatigue behavior has been mainly investigated through However, the effect of fiber orientation effect has often been eval-
experimental techniques, while less attention has been given to uated through conducting fatigue tests on samples with different
fatigue behavior modeling [1]. thicknesses and with fibers in different orientation with respect
Under cyclic loading, a continuous softening is generally to the loading [6].
observed in SFTCs, which is typically due to initiation and growth Environmental effects including temperature and moisture on
of damage in the matrix as well as at fiber ends and fiber-matrix fatigue behavior of SFTCs have been explored in several studies.
A significant degradation of fatigue strength has been observed
⇑ Corresponding author. from temperatures below to above the glass transition tempera-
E-mail addresses: seyyedvahid.mortazavian@rockets.utoledo.edu (S. Mortaza- ture (Tg) [6]. A recent extensive survey has been performed on high
vian), afatemi@eng.utoledo.edu (A. Fatemi). temperature fatigue behavior of SFTCs in [8]. The effect of water

0142-1123/Ó 2017 Elsevier Ltd. All rights reserved.
172 S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183

absorption is a function of the polymer type and fiber-matrix cou- analysis model used to represent or estimate each effect. The con-
pling agents [9]. sidered effects include cyclic deformation, load frequency and self-
A small number of studies have been devoted to effects of mean heating, anisotropy or fiber orientation effect, moisture, tempera-
stress or R ratio and stress concentration on fatigue behavior of ture, mean stress, and stress concentration.
SFTCs. A significant effect of mean stress, which may be accompa-
nied by cyclic creep or ratcheting is observed on fatigue behavior of 2. Material, specimen geometry, and experimental method
SFTCs [10]. The reduction of fatigue strength due to mean stress
has been observed to be less for notched specimens as compared The two composite materials considered were a polybutylene
with smooth specimens, due to the presence of stress gradient near terephthalate with 30 wt% short glass fiber (here referred to as
the notch root. Modified Goodman and Gerber mean stress equa- PBT) and a polyamide-6 with about 10 wt% rubber and 35 wt%
tions have been used with the use of creep rupture strength to cor- short glass fiber (here referred to as PA6). The glass transition tem-
rect for the mean stress effect [11,12]. perature (Tg) of both materials was about 60 °C, as obtained from
In this study, a number of aspects related to fatigue behaviors of dynamics mechanical analysis. The average fiber aspect ratio was
two short fiber thermoplastic composites (SFTCs) were experimen- estimated at 26 [13,14].
tally investigated and fatigue life estimation methodologies are Materials were injection molded in rectangular plaques with
presented to account for these aspects. The materials and specimen dimensions of 100 mm  200 mm in 3 and 3.8 mm thicknesses.
geometries used, as well as the experimental procedure are To study the effect of fiber orientation, rectangular strips were
described first. Then, experimental results for the different effects machined from molded plaques at 0°, 18°, 45° and 90° angles with
considered are presented and discussed, followed by the fatigue respect to the injection mold flow direction, as shown in Fig. 1(a). A

Fig. 1. (a) Specimen cutting directions with respect to mold flow and, (b) specimen geometry designed for fatigue tests. For notched specimens a 2 mm diameter central hole
was drilled in the middle of the gage section (Kt = 2.5) (all dimensions are in mm).
S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183 173

dog-bone shape specimen with an optimized geometry was cut

from the strips, as shown in Fig. 1(b). For notched tests, a circular
hole with diameter of 2 mm was drilled in the center of the spec-
imen gage section. The elastic stress concentration factor (Kt) in the
critical section of the notched specimen was obtained to be 2.5.
Fatigue tests were conducted on a uniaxial servo-hydraulic test-
ing machine and controlled by a digital controller. A mechanical
extensometer was used to measure strain and a thermal imaging
camera was used to measure surface temperature rise in room
temperature tests. For temperature effect study, an environmental
chamber employing an electronic heating element and a liquid
nitrogen cooling system was used. For studying moisture or water
absorption effect, specimens were immersed at room temperature
water for a certain period of time prior to testing and then tested at
room temperature.
Load-controlled un-notched fatigue tests were performed in a
range of cycles to failure between 103 and 106 cycles on specimens
machined in various directions of mold flow. Fatigue tests were
conducted at 40 °C, 23 °C, and 125 °C under the stress ratios of
1, 0.1, and 0.3. Notched fatigue tests were conducted using spec-
imens in both longitudinal and transverse directions of mold flow
and under stress ratios of 1 and 0.1.

3. Experimental results, analysis, and models

3.1. Cyclic deformation behavior

Cyclic deformation behavior of polymeric materials is very dif-

ferent with that of metallic materials with grain boundaries. This is
due to formation of long chain molecules in polymeric materials. In
this study, incremental step cyclic deformation tests with blocks of
increasing stress amplitude were performed with a sinusoidal
load-controlled wave form under fully-reversed (R = 1) condition Fig. 2. (a) Progressive deformation of PBT in the transverse direction with
at 40°, 23° and 125 °C and in both longitudinal and transverse hysteresis loops shown from initial cycles to and near fracture cycles and, (b)
directions. Test continued at each stress level until a relatively sta- Cyclic stress-strain data and the corresponding Ramberg-Osgood (dashed) and
monotonic tension (solid) curves for PBT in the longitudinal direction [14].
bilized strain was reached. Test results were then used to obtain
cyclic stress-strain curves and evaluate cyclic softening.
Progressive cyclic deformation of PBT in the transverse direc- 125 °C. Similar behavior was observed for the transverse direction,
tion at a stress level corresponding to about 65% of tensile strength as well as for PA6.
is shown in Fig. 2(a). With continued cycling, the area under the Under cyclic loading, the strain is adjusted by microscopic rear-
hysteresis loops increases and the cyclic modulus decreases. Due rangement of polymer molecules. Polymeric materials often show
to higher straining in tension than in compression, the strain cyclic softening and a number of factors such as molecular struc-
amplitude and mean strain increase as cycling is continued. This ture, time, temperature and additives control the degree of soften-
behavior was observed for both materials, in both longitudinal ing. A number of cyclic deformation mechanisms which can take
and transverse directions, and at all test temperatures. The afore- place in polymers consist of homogenous mechanisms such as
mentioned changes in hysteresis loops were more pronounced at molecular chain disentanglement, reorientation or slip and crystal-
higher temperatures and at higher stress levels, such that at stress lization as well as heterogeneous mechanisms including craze and
levels corresponding to fatigue life of 103 cycles or shorter, no sta- shear band formation [16].
bilized hysteresis loop was observed.
The stress-strain response of the materials under cyclic loading 3.2. Temperature rise effect and modeling
can be quite different from that under monotonic loading. Cyclic
stresses and strains from relatively stabilized hysteresis loops were The hysteresis area in each load cycle generally represents
used to obtain the cyclic stress-strain curves. Ramberg-Osgood energy loss per unit volume of the material. At lower frequencies,
equation was used to mathematically represent the cyclic stress- a greater time is given to polymer chains to disentangle and align
strain behavior, expressed as [15]: into the load direction and, therefore, thermoplastics indicate a
ra ra lower stiffness and a higher degree of energy dissipation per cycle.
ea ¼ þ ð1Þ As frequency of test increases, the dissipated energy reduces and a
E0 K0
higher stiffness is observed. However, due to low thermal conduc-
where K0 and n0 are cyclic strength coefficient and cyclic hardening tivity of thermoplastics, the generated heat due to energy dissipa-
exponent, respectively, and obtained from the fit of true stress tion results in self-heating. Presence of fiber reinforcements can
amplitude versus true plastic strain amplitude. increase strength, stiffness, as well as thermal stability of thermo-
The cyclic stress-strain data, corresponding Ramberg-Osgood plastics, but friction between fiber and matrix and higher stress
curves, as well as and monotonic tension curves are superimposed concentrations near fiber ends may increase the degree of self-
in Fig. 2(b) for PBT in the longitudinal direction and at tempera- heating [17,18].
tures of 40, 23, and 125 °C. Significant cyclic softening is observed Fatigue life predictions for service load histories are generally
at room temperature, while small softening is observed at 40 and based on fatigue data performed with constant amplitude loading,
174 S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183

which can be substantially influenced by cyclic frequency in ther- at 1 Hz for the higher stress amplitude test and more than 35 °C
moplastics. Therefore, an appropriate frequency should be selected at 4 Hz for the lower stress amplitude test. The degree of cyclic
for obtaining constant amplitude fatigue properties of thermoplas- softening, as reflected by the increase in displacement amplitude,
tic materials. The effect of frequency on R = 1 fatigue life of PA6 in is related to the amount of increase in temperature.
the longitudinal direction is shown in Fig. 3(a). Two stress ampli- Regardless of the chosen stress amplitude and test frequency,
tudes corresponding to about 40% and 34% of ultimate tensile when the temperature rise exceeded 10 °C, the displacement
strength (Su) were used. Considerable effect of frequency on fatigue amplitude rapidly increased and thermal failure occurred. There-
life was observed by changing the frequency by a factor of 4 for the fore, low test frequencies were chosen to limit the temperature rise
higher stress amplitude test (from 0.25 Hz to 1 Hz) and by a factor to a maximum of 10 °C in fatigue tests in order to prevent signifi-
of 2 for the lower stress amplitude test (from 2 Hz to 4 Hz). cant self-heating. A relatively higher sensitivity to frequency was
Displacement amplitude versus applied cycles for these tests observed for PA6 as compared with PBT, due to a higher dissipation
are shown in Fig. 3(b), along with the increase in measured surface of energy in PA6. Due to viscoelastic nature of polymers, with
temperature relative to the beginning of the test. As can be seen increasing test frequency up to a critical frequency, fatigue life
from this figure, the temperature increase was more than 13 °C may increase [19]. This is because with increasing frequency, the
chain disentanglement is restricted and polymer stiffness
increases. However, for the considered short fiber thermoplastics
in this study, this effect was not evident.
To characterize the effects of stress level and frequency on tem-
perature rise, incremental step tests with increasing frequency at
each step were performed. At each stress level, cycles were applied
under a constant frequency until the surface temperature of spec-
imen was nearly stabilized. Then the test was stopped for a period
of time, until the surface temperature returned to the room
Transient temperature rise curves with applied cycles at differ-
ent frequencies are shown in Fig. 4(a) for a longitudinal sample of
PBT under R = 1 with stress amplitude of about 32% of Su. The sta-
bilized temperature is higher at higher frequency, as expected. At a
critical frequency, the temperature rise was more than 10 °C and
surface temperature did not stabilize.
At all stress levels of both materials and in both longitudinal
and transverse directions of mold flow, a linear relationship
between stabilized temperature rise and cycling frequency was
obtained, as seen in Fig. 4(b). As the stress level is increased, a
higher rate of temperature rise is observed with increased
Energy-based models were applied to the incremental step fre-
quency data to generalize correlation of temperature rise as a func-
tion of cycling frequency and stress amplitude for each material. A
linear relationship between the temperature rise and dissipated
energy per unit volume and time was obtained, as follows [4]:

DT ¼ Bwf ð2Þ

where f is test frequency, w is the area inside the hysteresis loop, DT

is temperature rise, and B is a material parameter. A linear correla-
tion of temperature rise data presented in Fig. 4(b) with parameter
(w f) in Eq. (2) is shown in Fig. 4(c).
Another model based on a constant energy approach was also
applied to the data. This model is based on a linear one-
dimensional conductive heat transfer and assumes negligible
energy storage in the test specimen, expressed as [20]:

DT ¼ C ra e2a f ð3Þ

where ra is stress amplitude, ea is strain amplitude, and C is a mate-

rial parameter which can be obtained by a linear fit of temperature
rise data on Eq. (3). Linearity of the relationship between tempera-
ture rise and (ra ea2 f) parameter is observed in Fig. 4(c).
Knowing the constant B in Eq. (2) or the constant C in Eq. (3),
Fig. 3. (a) Effect of testing frequency on fatigue life and, (b) displacement amplitude temperature rise as a function of the loading (stress and strain)
versus applied cycles for PA6 in the longitudinal direction at room temperature
under R = 1 condition for stress amplitudes of 40% and 34% of Su [21]. The
and the applied frequency can be estimated. These constants for
temperatures shown are measured surface temperatures relative to the beginning each material and mold flow direction can be obtained from a
of the test. small number of tests.
S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183 175

Fig. 5. Thickness and plaque location effects on fatigue behavior of PA6 in the
longitudinal and transverse directions at room temperature [14]. Test frequency is
between 0.25 and 5 Hz.

strength in edge and middle of the plaque geometry, as shown in

Fig. 1(a).
Fatigue lives of the middle specimens were reduced by about a
factor of two, compared to the edge specimens, as seen in Fig. 5.
This may be due to the relatively thinner core layer in the edge
samples, compared to the middle samples. No effect of thickness
is observed in the longitudinal direction, while in the transverse
direction fatigue lives of 3 mm samples were reduced by about a
factor of 5 for PA6, as compared to the fatigue lives of 3.8 mm sam-
ples. The effect of thickness in the longitudinal direction of PBT was
also negligible, but a factor of 4 reduction in fatigue life was
observed from 3.8 mm to 3 mm thickness.
The effect of thickness is due to a shell-core morphology
observed across the thickness of the short fiber composites. During
the injection molding process, due to the gradient of velocity of
mold, fibers in the core layer are mainly oriented perpendicular
to the injection molding direction, while they are highly in line
with the mold flow direction in the two shell layers near the walls.
The core layer comprises about 0.5 mm and 0.2 mm of the speci-
men thickness for 3.8 mm and 3 mm thickness samples, respec-
tively. Therefore, the effect of thickness on fatigue strength of the
considered SFTCs can be related to the thicknesses of the core
and shell layers.
The effect of mold flow direction was evaluated in 0°, 18°, 45°,
and 90° directions relative to the injection molding direction.
Due to the presence of fiber reinforcement, fatigue performance
was highly dependent on the sample orientation relative to the
mold flow direction. With increasing the specimen angle with
respect to the mold flow direction, fatigue strength increases such
that longitudinal specimens have about 40% higher fatigue limit
(defined as fatigue strength at 106 cycles) than transverse speci-
mens, as seen in Fig. 6(a) for PBT at R = 1 condition. A higher
degree of anisotropy is observed at 125 °C, compared with 23 °C
and 40 °C, such that fatigue limit increased about 60% from the
transverse direction to the longitudinal direction of the injection
Fig. 4. (a) Transient surface temperature rise curves at the stress amplitude of 32% mold flow.
Su, (b) stable surface temperature rise as a function of cycling frequency at different The Tsai-Hill [22] criterion was utilized to estimate the off-axis
stress amplitudes and, (c) fits of stable surface temperature rise versus parameters fatigue strength, which is commonly used for orthotropic laminate
introduced in Eqs. (2) and (3) [21].
composites, expressed as:
" #12
2 4 2
3.3. Anisotropic fatigue behavior and modeling cos2 ðhÞðcos2 ðhÞ  sin ðhÞÞ sin ðhÞ cos2 ðhÞ sin ðhÞ
rfat ðhÞ ¼ þ þ
r2L;fat ðNÞ r2T;fat ðNÞ s2LT;fat ðNÞ
Fig. 5 shows the effect of fiber orientation on fatigue behavior of ð4Þ
PA6 at 23 °C under R = 0.1 condition. Specimens in 3 mm and
3.8 mm thicknesses were machined in the longitudinal and trans- where rL,fat (N), rT,fat (N), and sLT,fat (N) are the experimental fati-
verse directions of mold flow. For samples in the longitudinal gue strengths for a specimen life of N cycles. S-N curves generated
direction, a comparison was also made between the fatigue in 0°, 45°, and 90° directions were used to determine the directional
properties of Tsai-Hill equation to estimate the S-N curves for the
176 S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183

18° direction tests. The Tsai-Hill criterion reasonably captures the

effect of mold flow direction on fatigue behavior. Comparison of
the experimental fatigue strength and Tsai-Hill criterion correla-
tions as a function of fiber orientation angle with respect to the load
direction for different fatigue lives can be observed in Fig. 6(b).

3.4. Moisture effect and modeling

Dried specimens for PBT (at 120 °C for 6 h) and PA6 (at 80 °C for
6 h) in the longitudinal direction of injection mold flow were
immersed in room temperature water. The percentage of water
absorption was periodically measured by weighting the specimen.
Water absorption variation with square root of exposure time (t1/2)
is shown in Fig. 7(a) for longitudinal samples of PA6. As seen, per-
centage of water absorption by weight linearly increases until it
reaches a plateau with a maximum percentage of water absorption
of 5.2 wt%. This behavior follows the Fick’s law commonly used for
modeling the kinetics of moisture absorption process, expressed as
Mt 8 Dt
¼ 1  2 exp  2 p2 ð5Þ
Mm p h
where D is diffusion coefficient, h is specimen thickness, t is expo-
sure time, Mt is absorbed water, and Mm is the maximum capacity
of water absorption. The diffusion coefficient was estimated by fit-
ting Eq. (5) to absorption data at short times, shown in Fig. 7(a). Dif-
fusion coefficient was estimated at 5.5  1013 m2/s. Using the
calculated diffusion coefficient and thickness of specimen, along
with the maximum water absorption capacity, Eq. (5) can be used
to compute the absorbed water with exposed time.
Fig. 6. (a) Effect of the mold flow direction on fatigue behavior and, (b) Fatigue
Stress-strain curves in the longitudinal direction of PA6 with a
strength data as a function of specimen angle and Tsai-Hill criterion for PBT at 23 °C range of absorbed water between 0.0 and 5.2 wt% are shown in
under R = 1 loading condition [14]. Test frequency is between 0.25 and 6 Hz. Fig. 7(b). The time of exposure at room temperature water for each

Fig. 7. (a) Kinetics of water absorption at room temperature water (the dashed line corresponds to Fick’s law), (b) tensile stress-strain curves obtained at room temperature at
displacement rate of 1 mm/min showing the effect of water absorption, (c) variations of tensile strength and elastic modulus with water absorption, and (d) variations of
strain at tensile strength and tensile toughness with water absorption for the longitudinal samples of PA6 [26].
S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183 177

of these tests can be calculated from Fig. 7(a), and is also shown in
the brackets in Fig. 7(b). From this figure, it can be seen that
strength, elastic modulus and ductility of PA6 are highly influenced
by the amount of absorbed water. Tensile strength and elastic
modulus decrease with water content, while ductility indicates
the maximum value after 20 days of immersion and then reduces
to a smaller value.
Fig. 7(c) indicates variations of tensile strength and elastic mod-
ulus with percentage of absorbed water for longitudinal samples of
PA6. Exponentially decaying fits are observed for these properties.
The degraded properties of samples with maximum water absorp-
tion (5.2 wt%) recovered by only 50% after 30 h of hot drying, while
it was fully-recovered by 12 h vacuum drying at 80 °C. It should
also be mentioned that the rate of water absorption for PBT was
significantly lower than for PA6, such that after four days of
immersion less than 0.1 wt% water absorption and no degradation
of tensile properties were observed. PBT is considered as a
hydrophobic material, due to presence of four methylene repeat
units. Methylene groups reduce the polarity of PBT molecules
and reduce their tendency to bond with hydrogen molecules. A
high degree of crystallinity in PBT can be another aspect decreasing
the degree of water absorption [24]. The ability of the PA6 to water
absorption can be due to amide polar groups, low degree of crys-
tallinity, and the presence of 10% rubber modifier [25].
Strain at tensile strength exponentially increases up to strain of
about 11% and then reaches a plateau, as can be seen in Fig. 7(d). This
effect can result from combined effects of plasticity and decrease of
crack initiation resistant region due to absorption of water. Variation Fig. 8. (a) S-N curves at room temperature under R = 0.1 condition showing the
of tensile toughness defined as the area under the stress-strain cure effect of water, and (b) bar chart comparing the ratio of fatigue strength to tensile
is also shown in Fig. 7(d). Tensile toughness increases up to 3 wt% strength in both longitudinal and transverse directions of PA6 for dry and wet
water absorption and then decreases. PA6 material is commonly conditions [26]. Test frequency is between 0.25 and 6 Hz.
used in applications where high toughness is required and water
absorption shows beneficial effect on tensile toughness of PA6. aT 0 ðTÞ ¼ ð6Þ
Fatigue behavior was studied in dry condition as well as wet N0f
condition with four days of immersion at room temperature water.
where N0 f is the reduced cycles to failure due to the effect of temper-
Fig. 8(a) shows the R = 0.1 S-N curves of PA6 in both longitudinal
ature. The 23 °C data were selected as references and the 125 °C and
and transverse directions. Fatigue life reduced by more than an
40 °C data were shifted to the right and left sides of 23 °C data,
order of magnitude in both LCF and HCF life regimes and in both
respectively, until sufficiently high data correlations were obtained
the longitudinal and transverse directions. At the same stress
for all the data at various temperatures. Master curves in the two
amplitude levels, significantly larger stress-displacement loops at
mold flow directions of PA6 under R = 1 condition are shown in
midlife were observed in wet samples, as compared with the dried
Fig. 9(b).
samples. Stress-life data in dry and wet conditions became closer
The log shift factor obtained from experimental data is plotted
when plotted in terms of the area inside the stress-displacement
as a function of the reciprocal of test temperature in Fig. 9(c).
loops at midlife.
The equation of line fits follows the Arrhenius equation form,
The ratios of R = 0.1 fatigue strength at 106 cycles to tensile
expressed as:
strength for both dry and wet conditions are shown in Fig. 8(b)
for longitudinal and transverse direction samples of PA6. In the Ea 1 1
transverse direction, nearly the same ratio of Sf /Su was obtained LogaT 0 ¼  ð7Þ
8:314 T T 0
in both dry and wet conditions, confirming similar effect of water
on both tensile and fatigue strengths. However, in the longitudinal where T0 is reference temperature and Ea is the activation energy
direction a higher ratio is observed for the wet condition, as com- which is different for the temperature ranges above and below
pared with the dry condition. This results from a higher effect of the glass transition temperature (Tg).
water in the tensile strength compared with the fatigue strength, Knowing Ea and T0 in the Arrhenius equation for a particular
in the longitudinal direction of PA6. The ratio of fatigue strength material, the shift factor at any temperature can be calculated from
to tensile strength is about 0.25 in both dry and wet conditions Eq. (7). Using the master curve, the fatigue life at that temperature
and in both longitudinal and transverse directions. for a particular stress level can then be estimated. Since the Arrhe-
nius equation form is identical for each material at both stress
3.5. Test temperature effect and modeling ratios (R = 1 and R = 0.1) and in both mold flow directions (L
and T), only one mold flow direction, one stress ratio, and two tem-
A significant effect of temperature was observed in both longi- peratures above and two temperatures below Tg are needed to
tudinal and transverse directions. Fatigue strength at 125 °C signif- obtain this equation.
icantly decreased compared to the 23 °C, and increased at 40 °C,
as seen in Fig. 9(a). S-N fatigue data were correlated by shifting the 3.6. Fatigue crack initiation failure mechanisms
fatigue life at various temperatures to a reference temperature (T0).
A shift factor (aTo) is defined in order to generate a master curve for Fatigue fracture surfaces of mold flow oriented PA6 samples
a particular material, direction, and stress ratio as: were analyzed under the SEM. Specimens tested at R = 0.1 high
178 S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183

Fig. 9. (a) Effect of temperature on fatigue behavior, (b) construction of master curves of stress amplitude for fully-reversed data of PA6 in both longitudinal and transverse
directions and, (c) variation of log shift factor with the reciprocal of temperature for PBT and PA6 under various test conditions [14]. Test frequency is between 0.125 and 7 Hz.

cycle fatigue at room temperature for both dry and wet conditions, increases, a larger area of fracture surface exhibits micro-ductile
as well as at 120 and 40 °C were studied. For all samples, a lighter fracture. A higher ductility is observed in fibril structure of matrix
area which comprises a small fraction of specimen cross section is compared with the test at room temperature and a higher length of
attributed to crack initiation life and a darker area which occupies fiber stubs was visible. For the 40 °C sample, matrix indicated
most of the fracture surface results from a fast crack growth mech- extensive brittleness throughout the fracture surface. Matrix areas
anism. SEM of the fracture surfaces for the crack initiation areas are were almost flat and featureless with significant fiber pull out.
illustrated in Fig. 10. Part (a) of Fig. 10 corresponds to the crack ini-
tiation region of dry sample tested at room temperature and indi- 3.7. Estimation of strain-life curves and properties
cates a micro-ductile deformation and stretch of matrix. In this
area the stretched matrix has covered the fibers and no fiber Strain-life behavior is often used in the low-cycle fatigue appli-
pull-out is observed. cations, as well as at critical locations of components such as stress
Fatigue fracture surface of a PA6 specimen with 1.6 wt% concentrations where significant plastic deformations exist. Since
absorbed water is shown in Fig. 10(b). As seen, a higher degree of performing strain-controlled testing is more complicated than
matrix micro-ductility is observed, as compared with the dried stress-controlled testing, estimating strain-controlled fatigue
sample, which is attributed to plasticizing effect of water. As mole- properties and curves from stress-controlled tests is desirable.
cules of water diffuse in the composite, they form hydrogen bonds Therefore, the cyclic stress-strain relationship expressed earlier
between polymer chains and increase their mobility. The fibers in Eq. (1) was used to develop strain-life curves from fully-
were covered with a thin layer of matrix and only a short length reversed (R = 1) load-controlled fatigue tests.
of fiber ends was not covered. A cohesive deformation of matrix The total strain amplitude value for each load-controlled test
around the debonded fiber indicates good interfacial bonding was obtained using the cyclic Ramberg-Osgood relation. Total
between the glass fibers and the coupling agent. In the crack initi- strain amplitude was then resolved into elastic and plastic compo-
ation site of the wet specimen, a larger section in shell area intro- nents, and both elastic and plastic strain amplitudes versus rever-
duced earlier, showed micro-ductile structure, as compared with sals to failure were approximated as straight lines in log-log scale.
the dried sample. The straight-line elastic behavior is represented by:
Fracture surfaces of PA6 samples tested at 125 °C and 40 °C
are shown in Fig. 10(c) and (d), respectively. As temperature ra ¼ r0f ð2Nf Þb ð8Þ
S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183 179

Fig. 10. SEM of the HCF crack initiation surface of PA6 specimens at (a) dry, (b) wet conditions tested at room temperature, and dried specimens tested at (c) 125 °C and (d)
40 °C [14].

where r0 f is the fatigue strength coefficient and b is the fatigue

strength exponent which were obtained by a power fit of the data.
These values are related to fatigue strength intercept (A) and slope
(B) of S-N line in Eq. (1) by r0 f = (2)b A and b = B.
A power fit of plastic strain amplitude values versus reversals to
failure can also be obtained, represented by the following

¼ e0f ð2Nf Þc ð9Þ

where e0 f is the fatigue ductility coefficient and c is the fatigue duc-

tility exponent. Knowing the elastic strain-life and plastic strain-life
fits, the total strain-life fit can then be expressed as:

ea ¼ ð2Nf Þb þ e0f ð2Nf Þc ð10Þ
Fig. 11 shows the fully-reversed (R = 1) superimposed strain-
life curves at 40 °C, 23 °C, and 125 °C for both PBT and PA6 in
the longitudinal direction. For each material, the differences
between strain-life curves at different testing temperatures are
much smaller than the corresponding S-N curves.
The transition fatigue life which is derived from the intersection
of elastic and plastic strain-life curves is expressed as:

e0f E0 bc

2Nt ¼ ð11Þ
For fatigue lives longer than the transition fatigue life, the
behavior is mainly elastic and for the fatigue lives shorter than
the transition fatigue life the behavior is mainly plastic. For nearly
all testing conditions, the obtained fatigue lives were greater than Fig. 11. Superimposed fully-reversed (R = 1) strain amplitude versus reversals to
the transition fatigue life, therefore, elastic deformation was dom- failure curves at 40 °C, 23 °C, and 125 °C for (a) PBT and (b) PA6 in the longitudinal
inant for all testing conditions. direction.
180 S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183

3.8. Mean stress or R ratio effect and modeling ent R-ratios was not corrected by this parameter. The k values for
different testing conditions varied between 0.86 and 1.44.
Fig. 12(a) shows the effect of mean stress on fatigue behavior of The Walker equation can be expressed as [28]:
PA6 in the transverse direction at 23 °C, 40 °C, and 125 °C. A sig-
nificant decrease of fatigue strength is observed under R = 0.1 load- SNf ¼ ðSa þ Sm Þ1c ðSa Þc ð13Þ
ing condition, as compared to R = 1, at all test temperatures. The where Sa, Sm, and SNf are stress amplitude, mean stress, and fully-
effect of tensile mean stress was more pronounced in the LCF reversed stress amplitude, respectively, and c is the mean stress
regime, as compared to the HCF regime. Smaller or no difference parameter. The value of c was determined by the best fits obtained
in fatigue lives is observed between R = 0.1 and R = 0.3 conditions for R = 0.1 and 0.3 data to the fully-reversed data, for each material,
in the HCF regime, as compared with in the LCF regime. mold flow direction, and temperature. A low c value indicates a
Many mean stress parameters have been applied to assess the higher mean stress sensitivity and a value of c = 1 indicates no
effect of mean stress. The Walker equation and a general fatigue mean stress sensitivity.
life estimation model showed more accurate correlations of mean The fits of experimental data for PA6 in the transverse direction
stress data for different test conditions considered in this study. based on the Walker equation are shown in Fig. 12(c), indicating
Modified Goodman is a common mean stress parameter used to reasonable mean stress correction for all the temperatures consid-
estimate the fatigue strength under the effect of mean stress. This ered. For both materials at 23 °C and 40 °C, the c value is nearly
parameter is expressed as [27]: constant at about 0.47, with a range between 0.4 and 0.55, while
at 125 °C, c values indicated the following relationship with the
tensile strength (MPa) in the corresponding direction and
Sa Sm Sa Su temperature:
þ ¼ 1 or SNf ¼ ð12Þ
SN f Su Su  Sm c ¼ 0:0043Su þ 0:364 ð14Þ
In this equation the tensile strength for each testing condition A general fatigue life estimation model based on a strength
was obtained from the average of duplicate tension tests at the dis- degradation concept under constant amplitude loading was also
placement rate of 1 mm/min. The fits of experimental data for PA6 applied to the experimental data. This model was applied to con-
in the transverse direction are shown in Fig. 12. In some cases, this tinuous fiber composites and is expressed as [29]:
equation correlated the fully-reversed and mean stress data well. Su  Smax ¼ aS1n n n b
u Smax ð1  RÞ ðN f  1Þ ð15Þ
In other cases, more than an order of magnitude difference
between the fully-reversed and equivalent mean stress fatigue where a and b are material constants and n is a function of stress
lives are observed. ratio and mold flow direction, expressed as:
An improvement to the modified Goodman parameter may be
n ¼ 1:6  R sin h ð16Þ
obtained by adding a correction factor exponent (k) to the mean
stress term. For k = 2, this equation is known as Gerber parabola In this equation, h is the angle between the mold flow direction
[15]. Although this correction factor provided a better correlation and the loading direction and R is the stress ratio. The simplified
of data compared to the modified Goodman, the inaccuracy of form of the equation arranged in an equivalent stress form is
modified Goodman with regards to the slope of S-N lines for differ- expressed as [29]:

Fig. 12. (a) Effect of R-ratio or mean stress on fatigue behavior for PA6 in the transverse direction, and correlation of mean stress data using (b) Modified Goodman equation,
(c) the Walker equation and, (d) the general fatigue life estimation model [30]. Test frequency is between 0.125 and 7 Hz.
S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183 181

Su  1R
Seq ¼ A þ1 ð17Þ
aSRu sin h0:6 ðDSÞ 1:6R sin h

where A and B are the intercept and slope of the fully-reversed S-N
line, respectively, DS is the stress range, R is the stress ratio, h is
fiber orientation angle, and a and b are the model parameters. Good
correlations of equivalent stress values versus fatigue life for differ-
ent stress ratios of PA6 in the transverse direction and at different
temperatures are shown in Fig. 12(d).
The value of a was nearly independent of stress ratio, but varied
with temperature and mold flow direction. In the longitudinal
direction of both PBT and PA6 and at all test temperatures,
a = 0.135 was suggested. In the transverse direction of both PBT
and PA6, a = 0.074 was suggested for both 23 °C and 40 °C tests
and a = 0.1 was suggested for 125 °C tests. It should be mentioned
that although the data and mean stress correlations are shown for
PA6 in the transverse direction, similar behaviors and correlations
were obtained for the longitudinal direction, as well as for PBT.

3.9. Estimation of fatigue strength from tensile strength

Fully-reversed (R = 1) fatigue strengths at fatigue lives of 103

and 106 cycles linearly correlated well with the corresponding ten-
sile strength at different temperatures and in different mold flow
directions. These correlations between these fatigue strengths
and tensile strength were observed to be relatively independent
of material, temperature, and mold flow direction. Therefore,
fully-reversed S-N fatigue line was estimated based on the extrap-
olation of fatigue strengths at 103 and 106 cycles, as:

¼ 1:08ðNf Þ0:085 ð18Þ
Correlation of fatigue strength data normalized by tensile Fig. 13. (a) Normalized fatigue strength with tensile strength and, (b) equivalent
strength is shown in Fig. 13(a). Therefore, in the absence of fatigue stress amplitude versus fatigue life using the general fatigue life model for fully-
data, this relation may be used to roughly estimate fatigue life at a reversed fatigue data of PBT and PA6 at 40 °C, 23 °C, and 125 °C and in 0°, 18°, 45°,
given stress amplitude for a particular material, temperature, and and 90° mold flow directions [14].
mold flow direction, based on the tensile strength for the material
and conditions.
Fatigue strength sensitivity of a material to a notch can be char-
The general fatigue life estimation model introduced in Eq. (17)
acterized by the notch sensitivity factor, expressed as:
can be also be used in a simpler form to correlate the fully-reversed
fatigue data at of the two materials at different temperatures and Kf  1
in different mold flow directions, expressed as: q¼ ð20Þ
Kt  1
Su  D2S where Kt is the elastic stress concentration factor and Kf is the fati-
Seq ¼ 96:3 þ1 ð19Þ
a 0:6
Su ðDSÞ1:6 gue notch factor (defined as unnotched fatigue strength divided by
notched fatigue strength at 106 cycles). Notch sensitivity factor
This equation has only one material parameter a, which is takes values between zero for no notch sensitivity and one for full
dependent on fiber orientation and can be approximated as notch sensitivity.
a = 0.0005 h + 0.118, where h is the specimen angle with respect This factor was obtained from experimental results for fully-
to the mold flow direction in degrees. A better correlation of data reversed fatigue strength to be about 0.5 in both longitudinal and
was obtained using Eq. (19), as compared to Eq. (18), as can be seen transverse directions. In LCF regime, a lower notch sensitivity
in Fig. 13(b). was observed in both mold flow directions, such that smooth and
notch specimen S-N lines nearly converged at one cycle. This is
3.10. Stress concentration effect and analysis due to notch plastic deformation in the LCF regime reducing notch
sensitivity and is similar to the behavior typically observed for
As mentioned earlier, the specimen geometry used for notched ductile metallic materials [15].
fatigue tests was the same as the geometry shown in Fig. 1, except The local strain or stress approach can be utilized for notch
a 2 mm diameter circular notch was drilled in the center of gage behavior predictions by using finite element analysis (FEA) results,
section (Kt = 2.5). A significant decrease of fatigue life was observed or by using a notch deformation rule such as the commonly used
in notched specimens, as compared with smooth (i.e. unnotched) Neuber’s rule for metallic materials. To consider the effect of stress
specimens, as seen in Fig. 14(a) for PBT in the transverse direction gradient, Kf rather than Kt is often used in Neuber’s rule for life pre-
under R = 1 loading condition. Nominal stress amplitude for dictions, and an approach such as the theory of critical distance
notched specimens plotted in this figure was obtained by dividing (TCD) can be used in conjunction with the FEA results. These
the applied load amplitude by the net cross section area of the approaches were used for the notched fatigue data in this study,
specimen. as detailed in [30].
182 S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183

 Cycling frequency resulting in self-heating can lead to cyclic

softening and a significant detrimental effect on fatigue behav-
ior. Energy-based models were used to characterize the temper-
ature rise as a function of loading and frequency. Such models
can be used to estimate the amount of temperature rise when
conducting fatigue tests.
 An effect of thickness on fatigue behavior was observed in the
transverse specimens, resulting from a core–shell morphology
produced during the injection molding process. The effect of
fiber orientation with respect to the loading on fatigue strength
was significant. Tsai–Hill criterion could represent the effect of
mold flow direction on fatigue strength reasonably well.
 A significant detrimental effect of moisture was found on tensile
and fatigue strengths of PA6, while the effect was negligible for
PBT. Water absorption by weight linearly increased with square
root of time, until it reached a plateau of 5.2 wt% for PA6. The
effect of moisture on fatigue strength reduction in both LCF
and HCF and in both longitudinal and transverse directions of
PA6 was nearly identical. The degraded properties can be recov-
ered, at least partially, by drying.
 A significant effect of temperature on fatigue behavior was
observed in both longitudinal and transverse directions of both
materials. Cold temperature had a beneficial effect and elevated
temperature had a detrimental effect on fatigue life, as com-
pared to at room temperature. Fatigue data at different temper-
atures were correlated by a shift factor represented by
Arrhenius equation, the form of which was independent of
stress ratio and mold flow direction.
 A method was presented to obtain strain-life curves from load-
Fig. 14. (a) Effect of notch (Kt = 2.5) on fully-reversed fatigue behavior of PBT in the controlled fatigue data at various temperatures. Strain-life
transverse direction under R = 1 condition and, (b) smooth and notched fatigue curves at different temperatures become closer to each other,
data correlations based on local stress approach and FEA, Neuber’s rule, and TCD
as compared with the stress-life curves.
methods [30]. Test frequency is between 0.25 and 6 Hz.
 A significant decrease of fatigue strength was observed under
tensile mean stress for both materials, all mold flow directions,
and at all test temperatures. The effect of tensile mean stress
Notched fatigue life data correlations based on the local stress was more pronounced in the LCF regime, as compared to the
approach using FEA, Neuber’s rule, and TCD method for PBT are HCF regime. The Walker equation and a more general fatigue
shown in Fig. 14(b). As can be observed from this figure, FEA notch life estimation model correlated the mean stress data at all tem-
stress results in overly conservative life estimations by several peratures and in both mold flow directions reasonably well.
orders of magnitude. This is because the effect of stress gradient  Two methods were evaluated for estimating fatigue strength of
at the notch is also an important factor in controlling notched fati- SFTCs for different conditions (i.e. material, fiber orientation
gue behavior. Better correlations are obtained by using Neuber’s direction, temperature) as a function of cycles to failure based
rule or TCD. on ultimate tensile strength for the corresponding condition.
These methods may be used as a reasonable first estimate of
4. Summary and conclusions fatigue strength, in the absence of fatigue data.
 A significant decrease of fatigue life was observed in notched
Application of short fiber thermoplastic composites (SFTCs) has specimens, as compared with smooth (i.e. unnotched) speci-
seen significant growth in recent years. A wide range of aspects mens. Lower notch sensitivity was observed in LCF regime, as
related to microstructure, environment and load conditions affect compared to HCF regime, due to significant notch plastic defor-
fatigue behavior of these materials. A number of these effects were mation in LCF. Notched fatigue life assessments based on the
experimentally investigated by conducting uniaxial constant local stress approach using FEA results were overly conserva-
amplitude load-controlled tests for two short fiber thermoplastic tive, while the use of Neuber’s rule or TCD provided relatively
composites in this study. The considered effects included cyclic accurate predictions.
deformation, load frequency and self-heating, anisotropy or fiber
orientation, moisture, temperature, mean stress or R ratio, and
stress concentration. Fatigue analysis models for representing or
estimating each effect were also presented. Based on the experi-
mental observations and the analyses conducted, the following
Financial support of this study was provided by General Motors.
conclusions can be made:

 Cyclic deformation behavior indicated progressive modulus References

reduction and unsymmetrical straining in tension and compres-
sion. Cyclic softening was observed at room temperature, while [1] Mortazavian S, Fatemi A. Fatigue behavior and modeling of short fiber
little or no softening was observed at 40 °C and 125 °C. The reinforced polymer composites: a literature review. Int J Fatigue
cyclic stress-strain curves for different conditions could be rep- [2] Wang SS, Chim EM. Fatigue damage and degradation in random short-fiber
resented by the Ramberg-Osgood equation. SMC composite. J Compos Mater 1983;17(2):114–34.
S. Mortazavian, A. Fatemi / International Journal of Fatigue 102 (2017) 171–183 183

[3] Launay A, Marco Y, Maitournam MH, Raoult I, Szmytka F. Cyclic behavior of [16] Cahn RW, Cahn RW, Haasen P, Kramer EJ. Materials SCIENCE AND
short glass fiber reinforced polyamide for fatigue life prediction of automotive TECHNOLOGY: A COMPREHENSIVE TREatment. Structure and properties of
components. Proc Eng 2010;2(1):901–10. nonferrous alloys, vol. 8. P.O. Box 10 11 61, Weinheim, D-69451, Germany:
[4] Bernasconi A, Kulin RM. Effect of frequency upon fatigue strength of a short VCH Verlagsgesellschaft mbH; 1996. [837].
glass fiber reinforced polyamide 6: a superposition method based on cyclic [17] Sun CT, Chan WS. Frequency effect on the fatigue life of a laminated composite.
creep parameters. Polym Compos 2009;30(2):154–61. In Composite materials: testing and design (fifth conference); 1979, ASTM STP
[5] Shaharuddin SIS, Salit MS, Zinudin ES. A review of the effect of moulding [Vol. 674, pp. 418–430].
parameters on the performance of polymeric composite injection moulding. [18] Kharrazi MR, Sarkani S. Frequency-dependent fatigue damage accumulation in
Turk J Eng Environ Sci 2006;30:23–34. fiber-reinforced plastics. J Compos Mater 2001;35(21):1924–53.
[6] De Monte M, Moosbrugger E, Quaresimin M. Influence of temperature [19] Eftekhari M, Fatemi A. On the strengthening effect of increasing cycling
and thickness on the off-axis behaviour of short glass fibre reinforced frequency on fatigue behavior of some polymers and their composites:
polyamide 6.6–cyclic loading. Compos Part A Appl Sci Manuf 2010;41 experiments and modeling. Int J Fatigue 2016;87:153–66.
(10):1368–79. [20] Krause O. Frequency effects on lifetime. Optimat Blades Project. DLR, doc.
[7] Arif MF, Meraghni F, Saintier N, Chemisky Y, Fitoussi J, Robert G. Fatigue OB_TX_N003 rev. 1, Wieringerwerf, The Netherlands; 2002.
damage investigation of PA66/GF30 by X-ray microtomography. In: 16th [21] Mortazavian S, Fatemi A, Mellott SR, Khosrovaneh A. Effect of cycling
European conference on composite materials; 2014. frequency and self-heating on fatigue behavior of reinforced and
[8] Eftekhari M, Fatemi A. Tensile, creep and fatigue behaviours of short fibre unreinforced thermoplastic polymers. Polym Eng Sci 2015;55(10):2355–67.
reinforced polymer composites at elevated temperatures: a literature survey. [22] Agarwal BD, Broutman LJ, Chandrashekhara K. Analysis and performance of
Fatigue Fract Eng Mater Struct 2015;38(12):1395–418. fiber composites. 3rd ed. New Jersey: Jon Wiley & Sons; 2006.
[9] Hoppel CP. The effects of tension-tension fatigue on the mechanical behavior [23] Cai LW, Weitsman Y. Non-Fickian moisture diffusion in polymeric composites.
of short fiber reinforced thermoplastics. Proceedings American Society for J Compos Mater 1994;28(2):130–54.
Composites; 1992. [24] Thomas S, Visakh PM. Handbook of engineering and specialty thermoplastics:
[10] Sonsino CM, Moosbrugger E. Fatigue design of highly loaded short-glass-fibre volume 3: polyethers and polyesters. John Wiley & Sons; 2011. [Vol. 64].
reinforced polyamide parts in engine compartments. Int J Fatigue 2008;30 [25] Carrascal I, Casado JA, Polanco JA, Gutiérrez-Solana F. Absorption and diffusion
(7):1279–88. of humidity in fiberglass-reinforced polyamide. Polym Compos 2005;26
[11] Oka H, Narita R, Akiniwa Y, Tanaka K. Effect of mean stress on fatigue strength (5):580–6.
of short glass fiber reinforced polybuthyleneterephthalate. Key Eng Mater [26] Mortazavian S, Fatemi A, Khosrovaneh A. Effect of water absorption on tensile
2007;340:537–42 [Trans Tech Publications]. and fatigue behaviors of two short glass fiber reinforced thermoplastics. SAE
[12] Mallick PK, Zhou Y. Effect of mean stress on the stress-controlled fatigue of a Int J Mater Manuf 2015;8:435–43.
short E-glass fiber reinforced polyamide-6, 6. Int J Fatigue 2004;26(9):941–6. [27] Goodman J. Mechanics applied to engineering. Green: Longmans; 1918.
[13] Mortazavian S, Fatemi A. Effects of fiber orientation and anisotropy on tensile [28] Walker K. Effects of environment and complex load history on fatigue life,
strength and elastic modulus of short fiber reinforced polymer composites. ASTM STP 462. In: American Society for Testing and Materials; 1970. p. 1–14.
Compos Part B: Eng 2015;72:116–29. [29] Epaarachchi JA, Clausen PD. An empirical model for fatigue behavior
[14] Mortazavian S, Fatemi A. Fatigue behavior and modeling of short fiber prediction of glass fibre-reinforced plastic composites for various stress
reinforced polymer composites including anisotropy and temperature effects. ratios and test frequencies. Compos Part A Appl Sci Manuf 2003;34(4):313–26.
Int J Fatigue 2015;77:12–27. [30] Mortazavian S, Fatemi A. Effects of mean stress and stress concentration on
[15] Stephens RI, Fatemi A, Stephens RR, Fuchs HO. Metal fatigue in fatigue behavior of short fiber reinforced polymer composites. Fatigue Fract
engineering. John Wiley & Sons; 2000. Eng Mater Struct 2016;39(2):149–66.