Mechanics of Materials 46 (2012) 113–122

Contents lists available at SciVerse ScienceDirect

Mechanics of Materials
journal homepage: www.elsevier.com/locate/mechmat

Thermodynamic analysis of fatigue failure in a composite laminate

M. Naderi, M.M. Khonsari ⇑
Department of Mechanical Engineering, Louisiana State University, Baton Rouge, LA 70803, USA

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

Article history: We put forward a thermodynamic approach for analyzing fatigue failure in a composite
Received 3 March 2011 laminate. We show that fatigue is an irreversible progression of increasing entropy that
Received in revised form 15 December 2011 accumulates until it reaches a critical value called the fracture fatigue entropy (FFE) at
Available online 24 December 2011
the onset of failure. Extensive series of fatigue tests are carried out that involve load-con-
trolled tension-tension, and displacement controlled fully-reversed bending fatigue with
Keywords: three different stress ratios as well as constant- and variable-loading. The role of hysteresis
Fatigue
energy in the entropy generation is investigated. FFE values are calculated based the exper-
Fracture
Woven Glass/Epoxy laminate
imental data obtained for temperature and hysteresis energy of a woven Glass/Epoxy (G10/
Thermodynamic entropy FR4) laminate. The concept of tallying entropy accumulation and the use of FFE are useful
for determining the fatigue life of composite laminates undergoing cyclic loading.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction gue strength failure criterion and related it to the number

Composite materials—owing to high strength and stiff-
ness, low weight, and high fatigue life—are widely used
in the industry and their applications are steadily growing.
These materials are anisotropic and inhomogeneous and
thus exhibit significantly more complicated behavior than
their metal counterparts, particularly during repetitive cyc-
lic loadings. In general, the fatigue degradation behavior of
a composite laminate is characterized by a combination of
matrix cracking, delamination, fiber/matrix debonding, and
fiber breakage (Mao and Mahadevan, 2002; Varvani-Farah-
ani et al., 2007). Thus, the complexities of multi-mode fati-
gue damage mechanisms in composite laminates have
hampered the development of a general fatigue failure cri-
terion (Varvani-Farahani et al., 2007).
A survey of literature reveals that there are a number of
notable research works devoted to the prediction of fatigue
life in composite laminates. One of the earliest models of fa-
tigue life prediction is due to the work of Hashin and Rotem
(1973) followed by Sims and Brogdon (1977) who based
their analyses on the conventional S–N curve method. They
replaced the static strength failure criterion with the fati-
⇑ Corresponding author. Tel.: +1 225 578 9192; fax: +1 225 578 5924.
E-mail address: khonsari@me.lsu.edu (M.M. Khonsari).
of cycles to failure. Fawaz and Ellyin (1994) reported a
semi-log linear relationship between the applied stress
and the number of cycles to failure. Their analysis was
capable of predicting the fatigue life of a composite lami-
nate for different load ratios and load directions by the
use of a reference S–N curve. According to Philippidis and
Vassilopoulos (1999), however, this criterion is very sensi-
tive to the choice of the reference curve. Huang (2002) pre-
sented a fatigue life prediction model for plain-woven glass
fabric–reinforced polyester composites subjected to biaxial
repetitive loading. There are a number of other investiga-
tors who have also attempted to characterize the fatigue
failure based on S–N curve method for woven fabric–rein-
forced composites. See, for example, Owen and Griffiths
(1978), Amijima et al. (1991), and Khan et al. (2002).
One of the promising approaches for estimating the fa-
tigue life of composite materials is based on energy consid-
eration. Ellyin and Elkadi (1990) postulated that the strain
energy can be utilized to predict fatigue failure and
showed that the fatigue life is related to the cyclic en-
ergy—a combination of number of cycles at failure and
material constants—through a power-law relation. Using
the critical plane approach (the plane in which cracks ini-
tiate and propagate), Plumtree and Cheng (1999) and Pet-
ermann and Plumtree, (2001) developed a fatigue failure

0167-6636/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mechmat.2011.12.003
114 M. Naderi, M.M. Khonsari / Mechanics of Materials 46 (2012) 113–122

criterion model. Natarajan et al. (2005) also put forward a

fatigue model based on the strain energy density by per-
forming tension–tension and bending fatigue experiments
on fiber–reinforced polymeric composites. Later, Shokrieh
and Taheri-Behrooz (2006) proposed a fatigue life model
based on the energy method for unidirectional polymer
composite laminates under tension–tension and compres-
sion–compression fatigue loading. To account for the effect
of off- and on-axis strain energy on fatigue life, Varvani-
Farahani et al. (2007) proposed a fatigue life model for
three different damage stages: matrix cracking, fiber ma-
trix interface delamination, and fiber breakage.
Most of the fatigue failure prediction models discussed
above are typically performed under constant amplitude
and simplified loading conditions. To date, a unified fatigue
criterion for a composite laminate subjected to variable
loading remains lacking. Fig. 1. Energy balance, entropy production and entropy flow in an
The premise of this paper is that the strain energy density arbitrary control volume.
associated with permanent degradation during fatigue must
be accompanied by irreversible entropy gain in accordance respectively. By convention, energy added to the control
with the second law of thermodynamics (Bryant et al., volume is assumed to be positive.
2008). Thus, it is hypothesized that entropy—a fundamental In terms of specific quantities, Eq. (1) can be written as
parameter that characterizes disorder—offers a natural follows:
measure for assessment of deterioration (Basaran and Yan, du
q ¼ w  r  Jq ð2Þ
1998; Doelling et al., 2000; Amiri and Khonsari, 2010, dt
2012; Beheshti and Khonsari, 2010; Aghdam et al., 2011; where w is the rate of volumetric mechanical work, Jq is
Naderi and Khonsari, 2011). More recently, Naderi et al. heat flux across the boundary, q is the density, and u de-
(2010). have demonstrated that by tallying the entropy notes the specific internal energy.
production in two metallic specimens (Al 6061-T6 and SS The time rate of internal energy, q du , is composed of two
dt
304) up to the fracture point, one arrives at a unique mate- terms: (i) energy associated with thermal capacity, qc @h ,
@t
rial property—a so-called fracture fatigue entropy (FFE). where c denotes specific heat and h represents tempera-
Within the range of experimental conditions, this parameter ture; (ii) damage energy, ed. The energy associated with
was found to be independent of geometry, load, and thermal capacity is responsible for rising the temperature
frequency. In other words, the necessary and sufficient of the specimen while damage energy is responsible for cre-
condition for a metal to fracture due to cyclic loading is that ation of new surfaces, internal cracks, changes in material
its accumulated production of entropy reaches a certain micro-structures (Taylor and Quinney, 1934; Halford,
level. In this paper, we quantify the mechanical entropy gain 1966; Kaleta et al., 1990). Equation (2) can be written as:
in a displacement-controlled bending fatigue and load-  
@h
controlled tension–tension fatigue in a Glass/Epoxy (G10/ w ¼ r  J q þ qc þ ed ð3Þ
@t
FR4) laminate subjected to either constant- or variable-
amplitude loading and determine the associated FFE. For an open system, which is allowed to exchange heat
The outline of the paper is as follows. In Section 2, we with its environment (Fig. 1), the total exchange of entropy
present the fundamental of the theory associated with en- (dtS) consists of two terms: the entropy flow into or out of
tropy accumulation and introduce the concept of FFE. In the system, deS, and the entropy production within the sys-
Section 3, experimental procedure to assess fatigue life tem, diS (Prigogine, 1967; Basaran and Nie, 2004), i.e.,
prediction using entropy concept is presented. Results of dt S di S de S
entropy calculation are discussed in Section 4, followed ¼ þ ð4Þ
dt dt dt
by the conclusions.
The entropy exchange with the surroundings, deS, can be
either positive or negative. But according to the second
2. Theory and formulation law of thermodynamics (Clausius–Duhem inequality) the
rate of entropy generation within the system must be
The first law of thermodynamics states that the internal non-negative, i.e.,
energy within an arbitrary control volume changes only if
di S P 0 ð5Þ
energy flows into (or out of) the control volume (Fig. 1) (Pok-
rovski, 2005) in accordance to the following expression: In terms of the specific quantities, Eq. (4) can be written as
(Naderi et al., 2010):
dE dW dQ qs_ ¼ r  J s þ c_ ð6Þ
¼ þ ð1Þ
dt dt dt
where s represents the total entropy per unit mass. Js is the
where E is total energy, t denotes time, Q and W are heat entropy flow and c denotes the volumetric entropy
flow and work across the boundary of the control volume, production.
 M. Naderi, M.M. Khonsari / Mechanics of Materials 46 (2012) 113–122
Using the definition of Helmholtz free energy,
W = u  h  s, the energy balance and the entropy produc-
tion Eqs. (2) and (6) can be rewritten as (Naderi et al.,
2010)
qðW_ þ hs _ ¼ r  J q þ w
_ þ hsÞ
_ þ hsÞ
w qðW _ J q  rh
c_ ¼   P0
h h h2
By introducing the specific heat capacity, c, and the Fou-
rier’s heat conduction law (Jq = k r h) into Eqs. (7) and
(8), one can obtain the general form of the heat conduction
equation and entropy production inequality as follows
(Naderi et al., 2010 ; Voyiadjis and Faghihi, 2011).):
kh;ii þ w ¼ qch_
w ed k
c_ ¼ þ þ 2 h;i h;i P 0
h h h
where k is the thermal conductivity.
Eq. (10) consists of a combination of three terms: the
mechanical  dissipation due to permanent deformation
c_ mech ¼ wh , internal variable evolution due to, for example,
 
damage energy c_ int¼ ehs and the thermal dissipation due
to heat conduction c_ cond ¼ hk2 h;i h;i .
The fracture fatigue entropy (FFE), cf, can be obtained by
integrating Eq. (10) up to the time tf when failure occurs
(Naderi et al., 2010):
Z tf  
w ed k
cf ¼ þ þ 2 h;i h; i dt
0 h h h
Eq. (11) reveals that the cyclic hysteresis energy (w) is a
crucial parameter for determining FFE. In metals, w—the
so-called cyclic plastic energy—is a function of fatigue
parameters such as cyclic strain hardening, fatigue ductil-
ity coefficient, and fatigue strength coefficient. To estimate
the FFE for a metallic component, one can make use of the
cyclic plastic energy formula such as the Morrow’s equa-
tion (Morrow, 1965) for w. Theoretical procedure and
experimental verification of thermodynamic fracture en-
tropy was recently reported by the authors (Naderi et al.,
2010; Naderi and Khonsari, 2010). However, in composite
materials, the hysteresis energy variation is much more
complicated than their metal counterparts and, to the best
of our knowledge, there are no expressions available to
relate cyclic hysteresis energy to composite fatigue proper-
ties. It should also be pointed out that in metals, internal
variables evolution represent only 5–10% of the entropy
generation owing to mechanical dissipation and is often
negligible (Clarebrough et al., 1955a,b; Halford, 1966).
However, in composite materials, the evolution of internal
variables represents nearly 30–50% of mechanical entropy
generation (Gamstedet et al., 2002). Due to the importance
of internal variable term, the effect of damage energy on
the fracture fatigue entropy must be considered. However,
in this work, only the experimentally-determined hystere-
sis energy is used for evaluating the FFE. The detailed of
this approach is described in Section 4.1.1.
115
3. Material and experimental procedure
The material studied is Glass/Epoxy (G10/FR4) which is
an unbalanced woven fabric composite with plain weave
and aligned configuration and consists of a continuous fil-
ð7Þ ament glass cloth with an epoxy resin binder. The plain
woven glass fabric is stacked in 15 and 24 layers with
two different thicknesses of 3 and 4.85 mm. Each woven
ð8Þ layer has two unidirectional layers stacked in [0°/90°]. This
type of a composite offers high tensile and flexural
strength (see Table 1) and thus finds use in a variety of
applications such as electrical equipment, aerospace struc-
tures, and rocket structural components. Specimens are
prepared with on- and off-axis stacking sequences. In on-
axis stacking, the warp and weft directions are aligned
with the load direction. The former is called lengthwise
ð9Þ (0°) while the latter is called crosswise (90°). In off-axis
stacking, the angle between the warp and the load direc-
tion is (a = 15°, 30°, 45°, 60° and 80°, in this study); See
ð10Þ Fig. 2a. Fig. 2a also presents a schematic diagram of the
experimental setup for tension–tension fatigue test. The
apparatus used is MTS 810 servohydraulic single actuator.
Sinusoidal fatigue loads are defined in Multipurpose Test-
Ware (MPT) software which controls the fatigue test and
applies with a frequency in the range of 5–15 Hz and a load
ratio of 0 and 0.1. Constant and variable loads are applied
in both the on-axis (0° and 90°) and the off-axis (15, 30,
45, 60 and 80°) directions. Variable loading tests include
both high-to-low and low-to-high sequences.
Fig. 2b shows the specimen dimensions used in a fully
reversed bending fatigue test. The apparatus is a compact,
bench-mounted unit with a variable-speed motor, vari-
ð11Þ
able-throw crank connected to the reciprocating platen,
with a failure cut-off circuit in a control box, and a cycle
counter. The variable throw crank is infinitely adjustable
from 0 to 50.8 mm to provide different levels of stress
amplitude.
Test conditions are summarized in Table 2. Tension–
tension fatigue tests are carried out with a total of 62 con-
stant loads and 16 variable loads (high-to-low and low-to-
high) with 3 and 4.85 mm-thick specimens in the range of
2–10 kN load amplitude. For the bending fatigue tests, 16
constant amplitude loads with the 3 mm-thick specimens
are performed.
The instrumentation includes a high-speed, high-reso-
lution infrared (IR) thermography used to record the tem-
perature evolution of the specimen during the entire
experiment. The IR camera is a MIKRON M7500 with tem-
perature range between 0 °C and 500 °C, resolution of
320  240 pixel, accuracy of ±2% of reading, sensitivity/
NETD of 0.08 °C at 30 °C, and image update rate of 7.5 Hz.
Before fatigue testing, the surface of the specimen is coated
with black paint to increase the thermal emissivity of the
specimen surface.
4. Results and discussion
A series of tension–tension and bending fatigue tests
are performed under constant and variable amplitude
loading. All tests are conducted until fracture occurs to
116 M. Naderi, M.M. Khonsari / Mechanics of Materials 46 (2012) 113–122

Table 1
Mechanical properties of G10/FR4.

Tensile strength (MPa) Flexural strength (MPa) Elastic modulus in flexure (GPa)
Lengthwise Crosswise Lengthwise Crosswise Lengthwise Crosswise
275 240 380 310 18 15

Fig. 2. (a) and (b) Specimen geometry and schematic diagram of the experimental setup in tension–tension fatigue test and specimen geometry in bending
fatigue test (all dimensions are in mm).

Table 2
Fatigue test conditions.

Fatigue test Frequency (Hz) Specimen thickness (mm) Number of tests Load amplitude
Constant load Variable load
Tension–tension 5, 10, 15 3, 4.85 62 10 2–10 (kN)
Bending 10 3 16 – 20–50 (mm)

determine the fatigue fracture entropy (FFE) associated
with each specimen. In the following sections the results
of tension–tension and bending fatigue tests are discussed.
4.1. Tension–tension fatigue
In the following sections, we present the results of
experimentally-obtained hysteresis energy and tempera-
ture evolution that are necessary for thermodynamic anal-
ysis followed by experimental entropy generation
evaluation.
4.1.1. Hysteresis energy
Fig. 3 shows the area within the hysteresis loop—the so-
called strain energy—increases during a series of tension–
tension fatigue tests. The hysteresis area of each cycle is
obtained using a MATLAB™ code that utilizes the experi-
mental load and displacement data between loading and
unloading paths. Typical experimentally-determined hys-
teresis curves at 5% of the total life, 50% of the total life,
and just before failure are also presented in Fig. 3. Exami-
nation of the strain energy reveals three distinct stages.
The first stage is limited to about 5–10% of the total life;
 M. Naderi, M.M. Khonsari / Mechanics of Materials 46 (2012) 113–122 117

tion of Fig. 5a and b reveals that similar to the temperature

Fig. 3. Accumulation of hysteresis energy per cycle during tension–
tension fatigue test at different constant load amplitudes, frequency of
10 Hz, and R = 0. The hysteresis energy evolution has three phases: initial
increase, gradual and slow rise and abrupt rise.
during the second stage, gradual and slow increase in
strain energy occurs and lasts about 70–75% of total life;
and the third stage the fibers tend to break and high strain
energy is released.
rise, the entropy production undergoes three separate
stages: an initial rise (Stage I), slow and steady increase
(Stage II), and a drastic increase prior to failure (Stage
III). During Stage I, entropy accumulates due to three
mechanisms: the high energy released at the tips of mi-
cro-cracks in multiple locations in the matrix between lay-
ers, debonding at the cross-over points between warp and
weft fibers and matrix, and the breakage of some fibers
with low strength. Degradation grows and entropy gener-
ation piles up from the beginning of the test and continues
to the onset of Stage II. Due to a comparatively small num-
ber of cycles in this phase, the accumulation of entropy is
not appreciable. As soon as the deterioration in the mate-
rial reaches the saturation level, the trend of degradation
changes. At this stage, the process of entropy production
reduces due to a lower rate of heat generation and a slower
damage process. However, entropy accumulation in Stage
II is more pronounced than that of the previous stage due
to a higher span of fatigue life there (around 55% of total
life).

4.1.2. Temperature, entropy evolution, and entropy

production
Fig. 4 shows the surface temperature recorded using an
IR camera plotted versus the number of cycles-to-failure
for a series of tension–tension fatigue test at frequency of
10 Hz and R = 0. The temperature plotted in this figure is
measured at the cross section where failure occurs. It
shows that, under the conditions tested, temperature in-
creases rapidly over the first 20% of the total life, followed
by a more gradual increase. Finally, a sudden increase of
temperature takes place and the specimen fractures. A
similar trend of the temperature evolution for woven and
Polypropylene/Glass fiber composite is reported by Toubal
et al. (2006), Paulo et al. (2010), Naderi et al. (2011).
We now turn our attention to the entropy production
rate and entropy accumulation calculated based on Eqs.
(10) and (11) for tension–tension fatigue test. The results
of entropy generation are plotted in Fig. 5a and b. Examina-

Fig. 5. (a) Entropy production during tension-tension fatigue test of

Fig. 4. Evolution of surface temperature increase during tension-tension 5.25 kN load amplitude at the frequency of 10 Hz, and R = 0, respectively.
fatigue test at different constant load amplitudes, frequency of 10 Hz, and The entropy production undergoes three stages: initial increase, gradual
load ratio R = 0. Temperature increases initially, continues with slow and and slow rise and abrupt rise. (b) Entropy accumulation. Entropy
steady increase, and rises prior to failure. accumulates during fatigue until reaches the critical value at the fracture.
118 M. Naderi, M.M. Khonsari / Mechanics of Materials 46 (2012) 113–122

At the beginning of Stage III, the fatigue damage behav-

ior and the degree of irreversibility vary due to the change
of degradation modes from matrix cracking to fiber break-
age. More energy is released due the occurrence of fiber
breakage during this stage. Both entropy production and
its accumulation also increase with time until failure oc-
curs. It is also to be noted that changing the degradation
mode affects the heat dissipation rate and consequently
the rate of entropy production.
Fig. 6a–c depicts the results of FFE versus number of cy-
cles for tension-tension fatigue tests. Tests are carried out
for two different load ratios of R = 0 and 0.1. The results
of entropy accumulation subjected to variable loading
(high-to-low and low-to-high load) are also shown in
Fig. 6a–c. It is to be noted that according to an order of
magnitude analysis, entropy generation owing to the
mechanical dissipation is dominant and the entropy gener-
ation due to heat conduction is negligible (see Appendix A).
Hence, in the calculation of FFE, heat conduction effect is
neglected. The Results cover different frequencies within
the range of 5 to 15 Hz, constant- and variable-load ampli-
tude, and off-axis angles of 0°, 15°, 30°, 45°, 60°, 80°, and
90°. Within the range of operating conditions tested, FFE
is independent of frequency, load and fiber orientation.
The accumulated fracture energy (total hysteresis en-
ergy) and FFE are plotted in Fig. 7a and b for all the exper-
imental tests performed at different load ratios, load
amplitudes, and load angles. The accumulated fracture en-
ergy (AFE) is calculated by summing up the hysteresis en-
ergy per cycle
  to the total number of cycles at failure
PN f
wt ¼ j¼1 wi . Fig. 7a and b shows that the predictions
of the fatigue life based on AFE and FFE results have similar
trends. Fig. 7 a also reveals that AFE of on-axis and off-axis
angles close to warp and weft directions have different
trends than those of the off-axis angles far from warp
and weft directions. Failure in the weft and the warp direc-
tions and in off-axis angles close to on-axis are mostly
dominated by the normal stress. However, the shear failure
is mostly dominated by shear stress at the angles far from
the warp and the weft directions.

4.2. Bending fatigue

A series of fully reversed bending fatigue tests is carried

out for two different fiber orientations (0° and 90°) at the
frequency of 10 Hz. In these series of tests, the dissipated
strain energy—which is later utilized to determine the en-
tropy accumulation—is estimated using an experimental
procedure described by Meneghetti (2007) and Meneghetti
and Quaresimin (2011). This procedure involves measuring
the cooling rate after a sudden interruption of the fatigue
test. Following the work of Meneghetti (2007) and Mene-
ghetti and Quaresimin (2011), the statement of energy bal-
ance reads:
@h
w ¼ h þ qc þ ed
@t
Fig. 6. Fracture fatigue entropy versus the number of cycles to failure for
different tension–tension fatigue tests of Glass/Epoxy (G10/FR4) laminate
with different specimen thicknesses, frequencies and loads. a) Fracture
fatigue entropy in crosswise (90°) direction. Constant loads are in the
range of 4.9–9.5 kN and variable loads are 5.15–6 kN and 6–5.15 KN. (b)
Fracture fatigue entropy in lengthwise (0°) direction. Constant loads are
in the range of 5–13.5 kN and variable loads are 5.7–6 KN and 6–5.7 kN.
Fracture fatigue entropy is within the range of 1–3 MJ m3 K1. (c) Off-
axis fracture fatigue entropy of 15°, 30°, 45°, 60° and 80°. Loads are in the
ð12Þ
range of 2.5–5 kN.

where w is the input mechanical energy. The parameter h

denotes the rate of the dissipated thermal energy in a unit the rate of variation of internal energy which consists of
volume due to conduction, convection and radiation. The two terms. The first term depends on the temperature
remaining two terms on the right-hand side account for variation of the material. The second term, ed, is the time
 M. Naderi, M.M. Khonsari / Mechanics of Materials 46 (2012) 113–122 119

Fig. 8. Determination of dissipated thermal energy by experimental

measurement of the cooling rate.

rate of variation of the so-called stored energy of cold

work, which is the part of the mechanical input energy
responsible for creation of new surfaces, internal cracks,
and changes in material micro-structures (Bledzki et al.,
1997; Halford, 1966; Kaleta et al., 1991; Taylor and Quin-
ney, 1934).
Referring to Fig. 8, upon suddenly stopping the fatigue
test at time t⁄ when the surface temperature reaches h⁄,
the mechanical input power and the rate of variation of
cold work, ed, in Eq. (12) drop to zero. Hence, just after
t = (t⁄)+ Eq. (12) can be written as:


@h
Fig. 7. Comparison of fatigue life prediction models at various load ratios h ¼ qc ð13Þ
(R = 0, and 0.1), and load/fiber angles (a = 0°, 15°, 30°, 45°, 60°, 80°, 90°). @t t¼ðt Þþ
(a) Accumulated fracture energy concept. (b) Fracture fatigue entropy
criterion.
Measuring the temperature change as a function of time
during the cooling period after stopping the fatigue test

Fig. 9. (a) Comparison of cyclic strain energy based on the MTS results and the results of measuring the cooling curve for tension–tension fatigue test at the
frequency of 10 Hz, R = 0, and 5.75 kN load amplitude. (b) Dissipated energy variation during bending fatigue of G10/FR at the frequency of 10 Hz and
displacement amplitude of 38.1 mm.
120 M. Naderi, M.M. Khonsari / Mechanics of Materials 46 (2012) 113–122

provides the right-hand-side of Eq. (13). The thermal en- 4.2.1. Validation of hysteresis energy estimation
ergy per unit volume per cycle, q, is: In order to examine the validity of Eqs. (12)–(14) for
determining the hysteresis energy, we first apply the pro-
h
q¼ ð14Þ cedure described above to a tension–tension fatigue test
f
and then compare the results to the hysteresis energy ob-
where f is the frequency of the test. tained directly using the MTS instrument.

Fig. 10. SEM images of the surface of the specimen and the fracture surface for bending fatigue test at the frequency of 10 HZ and displacement amplitude
of 38.1 mm. (a) and (b) SEM image of the surface of the specimen around 85% of total life. (c) and (d) SEM image of the surface of the specimen around 95%
of total life. (e) SEM image of the fracture cross section of woven Epoxy/Glass laminate. SEM images show different failure mechanism such as matrix
cracking, debonding between fiber strands and matrix, delamination between warp and weft fiber strand at the cross-over points, and fiber breakage.
 M. Naderi, M.M. Khonsari / Mechanics of Materials 46 (2012) 113–122
Fig. 9a shows the strain energy evolution of tension–
tension fatigue test at the frequency of 10 Hz, R = 0, and
5.75 kN load amplitude. The curve-fitting toolbox of MAT-
LAB™ is used to generate the cooling curves from which
the rate of drop in temperature, @h @t
, is calculated and used
in Eq. (14) to evaluate the cyclic strain energy. The results
show that dissipated energy obtained by stopping the
fatigue test is in good agreement with the experimental
results of MTS. Since the bending fatigue apparatus is not
capable of measuring hysteresis loop, the method is ap-
plied for bending tests to determine the strain energy per
cycle for different series of tests.
4.2.2. Hysteresis energy in bending fatigue
The results of cyclic dissipated energy obtained by stop-
ping the fatigue test versus number of cycles to failure is
plotted in Fig. 9b for bending fatigue test at the frequency
of 10 Hz and displacement amplitude of 38.1 mm. To dem-
onstrate the failure mechanisms based on the distinct
stages in Fig. 9b, SEM images of the surface of the specimen
as well as the fracture cross section are shown in Fig. 10a–e
for bending fatigue test of G10/FR4 at the frequency of
10 Hz and displacement amplitude of 38.1 mm.
The SEM images of Fig. 10a–e also show the details of
matrix cracking, debondings, and delamination of the weft
and the warp strands. Fig. 10a and b shows the SEM images
of the specimen surface captured after completing approx-
imately 85% of the total life and Fig. 10a and b shows the
SEM images of the specimen taken after about 95% of total
life is depleted. The SEM image of the fracture cross section
is shown in Fig. 10e. Examination of Fig. 10a and b reveals
that some of the debonded regions join each other at inter-
face between the warp and the weft yarns. Debondings con-
tinue to propagate and delamination takes place between
warp and weft bundles at the fabric cross-over points as
well as between the adjacent layers in the Glass/Epoxy lam-
inate. As fatigue process progresses to around 95% of the to-
tal life (Figs. 10c–e), breakages at the delaminated areas
occur and fiber strands are separated from the surface. Deb-
ondings and delaminations are the origin of more cracks in
the pure matrix region. These cracks also propagate into
some of the fiber bundles as shown in the SEM image of
the fracture cross section (Fig. 10e).
The results of dissipated energy for different series of
bending fatigue for two different on-axis fiber directions
along with temperature data are then used to calculate
the FFE presented in Fig. 11. The abscissa of Fig. 11 shows
the entropy generation at fracture, cf, and the ordinate
shows the number of cycles to failure. Examination of the
results reveals that by using the FFE parameter, all the re-
sults of the experimental data are independent of load
amplitude and fiber orientation.
The results presented in Figs. 6 and 11 demonstrate the
validity of the hypothesis that the cumulative entropy at
failure for Glass/Epoxy (G10/FR4) laminate for tension–
tension and bending fatigue tests are independent of load-
ing direction, and loading type (constant load and variable
amplitude load).
A possible application of the proposed hypothesis of the
constant entropy gain at failure is in the development of a
methodology for prevention of the failure of composite
121
Fig. 11. Fracture fatigue entropy versus the number of cycles to failure for
different fully reversed bending fatigue tests of Glass/Epoxy (G10/FR4)
laminate for different loads (30–50 mm displacement amplitudes) and
frequency of 10 Hz.
laminates subjected to fatigue load. As demonstrated in
this work (Figs. 6 and 11), the entropy generation increases
during the fatigue life toward a final value of cf. Thus, frac-
ture fatigue entropy, FFE can be utilized as an index of fail-
ure. Successful implementation of this concept for
monitoring the fatigue fracture of metals has been recently
demonstrated (Naderi and Khonsari, 2010).
5. Conclusions
A thermodynamic approach for characterization of wo-
ven Glass/Epoxy (G10/FR4) laminate degradation is pro-
posed which utilizes the entropy generated during the
entire life of the specimens undergoing tension–tension
and fully-reversed bending fatigue tests subjected to con-
stant and variable loadings. The results reveal that fatigue
failure prediction by tallying entropy criteria is indepen-
dent of loading direction, constant load and variable ampli-
tude load (high-to-low and low-to-high load), and loading
frequency. Both accumulated fracture energy and accumu-
lated fracture entropy have the similar variation and can
be considered as a methodology for prediction of fatigue
life. However, the results of entropy accumulation tend
to be more unified than that of accumulated fracture en-
ergy due to the consideration of the temperature in its for-
mulation (Eq. (11)). The implication of these findings is
that the evolution of entropy generation can be utilized
to assess the severity of degradation of the specimen and
guard against fatigue failure.
Appendix A. An order of magnitude analysis
An order of magnitude analysis is performed to show
that entropy generation due to mechanical dissipation is
dominant and that entropy generation due to heat conduc-
tion is negligible. Referring to Eq. (10)
 2
w k Dh
c_  þ 2  c_ mech þ c_ cond ðA:1Þ
h h Dy
where Dh represents the temperature difference between
two cross sections at a distance, Dy with one of the
sections being the failure section (Fig. A.1). The tension–
122 M. Naderi, M.M. Khonsari / Mechanics of Materials 46 (2012) 113–122
Fig. A.1. Schematic of temperature gradient in scale analysis.
Table A.1
Order of magnitude of tension-tension fatigue test.
% of total life (Nf) Dh (°C) h1 (°C) w (kJ m3) c_ mech
c_ cond
5 0.1 22.4 42.4 66,782
60 0.3 47.7 54 10,253
90 3 62.5 75 150
tension fatigue test of 5 kN load amplitude, the load ratio
of zero, and the frequency of 10 Hz is considered for this
analysis. The hysteresis energy data obtained using MTS
instrument, temperature data from IR camera and
Dy = 1.2 mm are substituted in the Eq. (A.1) to calculate
the effect of mechanical dissipation and thermal dissipa-
tion on the entropy production. Table A.1 summarizes
the scale analysis in details.
Examination of the results show that entropy genera-
tion due to heat conduction is negligibly small to that of
the mechanical dissipation. Hence, Eq. (A.1) reduces to
w Natarajan, V., GangaRao, H.V.S., Shekar, V., 2005. Fatigue response of fabric–
c_   c_ mech
h
References
Aghdam, A.B., Beheshti, A., Khonsari, M.M., 2011. On the fretting crack
nucleation with provision for size effect. Tribol. Int., doi:10.1016/
j.triboint.2011.10.001.
Amijima, S., Fujii, T., Hamaguchi, M., 1991. Static and fatigue tests of a
woven glass fabric composite under biaxial tension torsion loading.
Composites 22, 281–289.
Amiri, M., Khonsari, M.M., 2010. On the thermodynamics of friction and
wear – A review. Entropy 12, 1021–1049.
Amiri, M., Khonsari, M.M., 2012. On the role of entropy generation in
processes involving fatigue. Entropy 14, 24–31. doi:10.3390/
e14010024.
Basaran, C., Yan, C.Y., 1998. A thermodynamic framework for damage
mechanics of solder Joints. Trans. ASME, J. Electr. Packag. 120, 379–384.
Basaran, C., Nie, S., 2004. An irreversible thermodynamic theory for
damage mechanics of solids. Int. J. Damage Mech. 13, 205–223.
Beheshti, A., Khonsari, M.M., 2010. A thermodynamic approach for
prediction of wear coefficient under unlubricated sliding condition.
Tribol. Lett. 38, 347–354.
Bledzki, A.K., Gassan, J., Kurek, K., 1997. The accumulated dissipated
energy of composites under cyclic-dynamic stress. Exp. Mech. 37,
324–327.
Bryant, M.D., Khonsari, M.M., Ling, F.F., 2008. On the thermodynamics of
degradation. Proc. R. Soc. A 464, 2001–2014.
Clarebrough, L.M., Hargreaves, M.E., Head, A.K., West, G.W., 1955a. Energy
stored during fatigue of copper. Trans. Am. Inst. Mining Metallurg.
Eng. 203, 99–100.
Clarebrough, L.M., Hargreaves, M.E., West, G.W., Head, A.K., 1955b. The
Energy Stored in Fatigued Metals. Proc. R. Soc. A 242, 160–166.
Doelling, K.L., Ling, F.F., Bryant, M.D., Heilman, B.P., 2000. An experimental
study of the correlation between wear and entropy flow in machinery
components. J. Appl. Phys. 88, 2999–3003.
Ellyin, F., Elkadi, H., 1990. A fatigue failure criterion for fiber reinforced
composite laminate. Compos. Struct. 15, 61–74.
Fawaz, Z., Ellyin, F., 1994. Fatigue failure model for fiber–reinforced
materials under general loading conditions. J. Compos. Mater. 28,
1432–1451.
Gamstedet, E., Redon, O., Brondsted, P., 2002. Fatigue dissipation and
failure in unidirectional and angle-ply glass/fiber/carbon fiber hybrid
laminates. Key Eng. Mater. 221–222, 35–48.
Halford, G.R., 1966. Energy required for fatigue. J. Mater. 1, 3–18.
Hashin, Z., Rotem, R., 1973. A fatigue failure criterion for fiber reinforced
materials. J. Compos. Mater. 7, 448–464.
Huang, Z.M., 2002. Fatigue life prediction of a woven fabric composite
subjected to biaxial cyclic loads. Compos. Part A – Appl. Sci. Manuf.
33, 253–266.
Kaleta, J., Blotny, R., Harig, H., 1990. Energy stored in a specimen under
fatigue limit loading conditions. J. Test. Eval. 19, 326–333.
Kaleta, J., Blotny, R., Harig, H., 1991. Energy stored in a specimen under
fatigue limit loading conditions. J. Test. Eval. 19, 326–333.
Khan, R.H., Khan, Z., Al-Sulaiman, F., Merah, N., 2002. Fatigue life
estimates in woven carbon fabric/epoxy composites at non-ambient
temperatures. J. Compos. Mater. 36, 2517–2535.
Mao, H., Mahadevan, S., 2002. Fatigue damage modelling of composite
materials. Compos. Struct. 58, 405–410.
Meneghetti, G., 2007. Analysis of the fatigue strength of a stainless steel
based on the energy dissipation. Int. J. Fatigue 29, 81–94.
Meneghetti, G., Quaresimin, M., 2011. Fatigue strength assessment of a
short fiber composite based on the specific heat dissipation. Compos.
Part B – Eng. 42 (2), 217–225.
Morrow, J., 1965. Cyclic plastic strain energy and fatigue of metals. ASTM
STP 378, pp. 45–87.
Naderi, M., Amiri, M., Khonsari, M.M., 2010. On the thermodynamic
entropy of fatigue fracture. Proc. R Soc. A 466, 423–438.
Naderi, M., Khonsari, M.M., 2010. Real-time fatigue life monitoring based
on thermodynamic entropy. Struct. Health Monitor., doi:10.1177/
1475921710373295.
Naderi, M., Kahirdeh, A., Khonsari, M.M., 2011. Dissipated thermal energy
and damage evolution of Glass/Epoxy using infrared thermography
and acoustic emission. Composite Part B, doi:10.1016/
j.compositesb.2011.08.00.
Naderi, M., Khonsari, M.M., 2011. A comprehensive fatigue failure
criterion based on thermodynamic approach. J. Compos. Mater.,
doi:10.1177/0021998311419540.
ðA:2Þ reinforced polymeric composites. J. Compos. Mater. 39, 1541–1559.
Owen, M.J., Griffiths, J.R., 1978. Evaluation of biaxial stress failure surfaces
for a glass fabric reinforced polyester resin under static and fatigue
loading. J. Mater. Sci. 13, 1521–1537.
Paulo, N.B.R., Ferreira, J.A.M., Richardson, M.O.W., 2010. Fatigue damage
characterization by NDT in polypropylene/glass fibre composites.
Appl. Compos. Mater., doi:10.1007/s10443-10010-19172-10449.
Petermann, J., Plumtree, A., 2001. A unified fatigue failure criterion for uni-
directional laminates. Compos. Part A – Appl. Sci. Manuf. 32, 107–118.
Philippidis, T.P., Vassilopoulos, A.P., 1999. Fatigue strength prediction
under multiaxial stress. J. Compos. Mater. 33, 1578–1599.
Plumtree, A., Cheng, G.X., 1999. A fatigue damage parameter for off-axis
unidirectional fiber reinforced composites. Int. J. Fatigue 21, 849–856.
Pokrovski, V.N., 2005. Extended thermodynamics in a discrete-system
approach. Eur. J. Phys. 26, 769–781.
Prigogine, I., 1967. Etude Thermodynamique des Processus Irreversible,
fourth ed. BelgiumLiege, Desoer.
Shokrieh, M.M., Taheri-Behrooz, F., 2006. A unified fatigue life model
based on energy method. Compos. Struct. 75, 444–450.
Sims, D.F., Brogdon, V.H., 1977. Fatigue behavior of composites under
different loading modes. In: K.L. Reifsnider, K.N. Lauraitis (Eds.), Fatigue
of Filamentary Composite Materials, ASTM STP 636, pp. 185–205.
Taylor, G.I., Quinney, H., 1934. The latent energy remaining in a metal
after cold working. Proc. R. Soc. Lon. Ser. – A 143, 0307–0326.
Toubal, L., Karama, M., Lorrain, B., 2006. Damage evolution and infrared
thermography in woven composite laminates under fatigue loading.
Int. J. Fatigue 28, 1867–1872.
Varvani-Farahani, A., Haftchenari, H., Panbechi, M., 2007. An energy-
based fatigue damage parameter for off-axis unidirectional FRP
composites. Compos. Struct. 79, 381–389.
Voyiadjis, G.Z., Faghihi, D., 2011. Thermo-Mechanical strain gradient
plasticity with energetic and dissipative length scales. Int. J. Plasticity,
doi:10.1016/j.ijplas.2011.10.007.