Professional Documents
Culture Documents
CITATIONS READS
19 1,025
2 authors:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Mijia Yang on 06 June 2014.
INTRODUCTION
ANALYSIS OF FATIGUE
Non-dimensional Analysis
To develop a general model for fatigue life predictions of composites, two types of
parameters, i.e., the testing and material related ones, should be considered. For an ideal
coupon sample subjected to constant amplitude and constant frequency cyclic loading,
the testing parameters, i.e., the stress amplitude (or stress ratio), maximum cyclic stress
(or mean stress), and frequency, are included in the analysis (Figure 1); while the material
parameters, such as the elastic Young’s modulus, mass, and ultimate strength, are taken
into account.
Based on the above parameters in fatigue modeling and the nature of dynamic
effect on fatigue due to cyclic loading, a general model for the fatigue life (N) can be
given as
1 cycle R = σmin/σmax
σa
Stress ∆σ
σmax
σm
σmin
Time
where is the stress amplitude, and max is the maximal cyclic stress (Figure 1); f is
the frequency of applied cyclic loading; E is the elastic Young’s modulus; m is the mass of
the composite material; and u is the uniaxial tensile strength of the composite material.
In this study, a global characterization approach was used, in which the laminates were
tested under the tension–tension uniaxial fatigue loading, and their mechanical responses
(e.g., stress and strain) along the loading direction under cyclic loading were measured.
The effects of fiber volume fraction, composite lay-up, and micro-level failure mechanism
were not studied. Thus, the initial Young’s modulus E0 along the tensile loading direction
is chosen as the Young’s modulus (E), and the static uniaxial tensile strength u0 is taken
as the uniaxial tensile strength ( u) of the composite material, assuming that the dynamic
strength under low frequency is very close to the static strength.
Normalizing Equation (1) with respect to the material parameters, the fatigue life is
expressed as a function of nondimensional factors as
max Af 2
N¼F , , ð2Þ
u u E
where is the density of composite material, and A is the cross-sectional area of the
specimen.
Sendeckyj [15] suggested that the residual strength after the constant amplitude cycles
was related to the initial static strength ( u0) by a deterministic equation. Similar to the
one by Caprino and D’Amore [16], the proposed fatigue model in this study begins with a
deterministic equation for the rate of strength degradation with inclusion of nondimen-
sional factor as
dðr =u Þ
¼ C1 nm1 ð3Þ
dn
where r is the residual strength after n cycles; C1 and m1 are the material constants; and
n is the number of fatigue cycles.
Based on the results postulated by Sendeckyj [15] and Hetrzberg and Manson [17],
C1 can be formulated in terms of nondimensional factors as
r m2
Af 2
C1 ¼ a ð4Þ
u E
where and m2 are the material constants; and the non-unit constants and are
a
¼ , ¼ 1 m1
1 m1
Fatigue Life Prediction of Pultruded E-glass/Polyurethane Composites 819
Note that the effects of the testing parameters, particularly, the frequency, and material
properties are included in Equation (5).
Now considering the general fatigue model proposed in Equation (5) and applying the
log transformation over both sides of the equation, we can obtain
Af 2 max
log N 1 þ log þ log þ m2 log ¼ log 1 ð6Þ
u E u
The constants for the given influential parameters (i.e., stress amplitude, frequency, and
maximum stress) can be respectively obtained, and their corresponding effects can be
analyzed.
The so-called ‘Goodman Line’ [19] is commonly used to describe the effect of mean
stress on fatigue strength. The fatigue strength for a given cyclic life time depends on
the cyclic load amplitude ( a) and the mean stress ( m) (Figure 1). Thus, the classical
Goodman relationship [19] is expressed as
a m
¼1 ð7Þ
N u
where N is the fatigue strength under the cyclic loading with the stress ratio of R ¼ 1.0.
By substituting a ¼ =2 and m ¼ max =2 (see Figure 1) into Equation (7), the
Goodman relationship thus becomes
u 1 max
¼1 ð8Þ
u 2N 2 u
To compare the proposed model in Equation (5) with the classical Goodman Line
equation (i.e., Equations (7) or (8)) and further simplify it to approximate the classical
S–N curves, a constant frequency is applied (i.e., f is a constant) and ¼ 1 is taken.
For each given fatigue life N (e.g., 103, 105, . . .), Equation (5) is then reduced to
max
B ¼1 ð9Þ
u u
where
m
Af 2 2
B¼ ðN 1Þ
E
Equation (9) indicates that the proposed model of Equation (5) has a similar form as the
classical Goodman relationship; however, the present model incorporates three testing
parameter effects (i.e., stress amplitude, frequency, and maximum stress) into one equa-
tion (Equation (5) or Equation (9) for the modified Goodman Line). Thus, the proposed
820 P. QIAO AND M. YANG
model represents a more generic and improved one compared to the classical S–N curve
in the literature, and it has a clear physical meaning on fatigue life prediction.
Equation (9) is further written as
0a m
B ¼1 ð10Þ
N u
where B0 ¼ ð2B þ 1ÞðN =u Þ. Thus, for a given fatigue life (N), a =N is linearly related to
m =u .
Now including the term of fatigue life (N) in the model, Equation (9) is rewritten as
max
C N 1 ¼ 1 ð11Þ
u u
where C ¼ ðAf 2 =EÞm2 . Considering the stress ratio R ¼ min/ max (Figure 1) and the
following relationship:
max
¼ ð1 RÞ ð12Þ
u u
Equation (13) has a clear physical meaning. For R ¼ 1, the fatigue strength N ¼ max.
By substituting max ¼ N and R ¼ 1 into Equation (13), we could find
u 1
CðN 1Þ ¼ ð14Þ
2N 2
Considering the relationship given in Equation (14) which is derived from Equation (5)
and comparing it with Equations (7) and (8), we can conclude that the proposed model
(Equation (5)) is coincident with the classical Goodman Line equation (Equation (7))
at R ¼ 1 (i.e., along the vertical axis of the Goodman Line diagram as illustrated later).
Similarly, the following relationships exist for R ¼ 1,
m ¼ u ; ¼ 0 ð15Þ
where N and u are the constants for each Goodman line of a given material. The
conditions in Equation (15) represent the horizontal axis of the Goodman Line diagram
as indicated later.
In summary, the present model (Equation (5)) can be simplified to the classical Goodman
Line model based on the above derivations. However, the proposed model has a strict
Fatigue Life Prediction of Pultruded E-glass/Polyurethane Composites 821
physical and mathematical basis compared to the classical Goodman Line model. Also
it is noticed that if r ¼ 1/2, we could obtain the Gerber formula [17]. Thus, the present
fatigue model represents a more generic and improved fatigue model, in which the effects
of frequency ( f ), load level (i.e., the maximum cyclic stress or mean stress), and stress
ratio (R) (or stress amplitude) are included. The proposed model is used next to study the
effects of stress ratio, frequency, and maximum cyclic stress to the fatigue life prediction
of newly developed E-glass/polyurethane composites and other common polymer matrix
composites. Further, it is interesting to observe that the proposed model can dramatically
reduce the fatigue test samples required by the ASTM standard, as illustrated later in the
experimental section.
As an application and validation, the proposed fatigue model (Equation (5)) is applied
to analyze the existing available data in the literature and also the fatigue data obtained
in this study for E-glass/polyurethane composites (more detailed material and experi-
mental information is introduced later). The proposed model is also used to fit the
experimental data in the S–N curves, and the experimental fatigue life and the material
constants in Equation (5) are given in Table 1 for four different E-glass fiber-reinforced
plastic composites. In order to compare with the classical S–N curve, the four-parameter
nonlinear model (Equation (5)) developed in this study is used to predict the fatigue life N
and the corresponding stress amplitude, of which the predicted S–N curve is obtained.
Compared to the classical S–N curve (the linear experimental data) fitting, the proposed
model (the predicted results) captures the trend of S–N data from the experiments
and provides a consistent prediction of fatigue life (see Figures 2–5). Besides capturing
the effects of stress ratio (or stress amplitude) and maximum stress (or mean stress), the
proposed model is capable of studying the influence of frequency, as illustrated later
for pultruded E-glass/polyurethane composites in this study. In the model proposed
by Epaarachchi and Clausen [18], the effect of frequency was included, but the reason
for changing the integration from the frequency domain to the time domain was not
given. When compared to the model provided by Caprino and D’Amore [16], the
proposed model can be applied to analyze the fatigue of general composites with different
material architectures and under different loading conditions. Hence, the fatigue model
proposed is an improved and generic model with a better predictability and clear physical
meaning.
E-glass/vinyl ester
Lay-up: [0]5 R ¼ 0.1 / u ¼ 0.7905–0.0694 log(N) ¼ 0.55027 Tension–tension [7]
u ¼ 581 MPa f ¼ 0.1–15 Hz ¼ 0.72798
E ¼ 21,000 MPa ¼ 270–430 MPa m2 ¼ 0.028152
A ¼ 25.0 3.68 mm2 ¼ 0.0067362
¼ 1.95 g/cc* COC** ¼ 0.939
Vf ¼ 0.27
P. QIAO
E-glass/polyurethane
[CSM/0/90/0/90/0/CSM] max ¼ 0.2–0.8 u / u ¼ 0.8751–0.1428 log(N) or ¼ 0.0269 Tension–tension Test data from the
u ¼ 748.294 MPa R ¼ 0.05–0.9 log(/ u) ¼ 0.5196–0.3264 log(N) ¼ 0.367 present study
AND
E ¼ 28,826 MPa f ¼ 1–5 Hz m2 ¼ 0.068
A ¼ 25.4 3.175 mm2 ¼ 0.228
M. YANG
¼ 1.85 g/cc COC** ¼ 0.773
*An estimated density is used, due to unavailability of the data in the reference.
**COC ¼ the coefficient of correlation.
Fatigue Life Prediction of Pultruded E-glass/Polyurethane Composites 823
1
Experimental data [5]
0.9
Predicted results
0.6
0.5
0.4
0.3
0.2
0 1 2 3 4 5 6 7
log(N)
0.39
0.37
0.35
0.33
∆s/s u
Predicted results
0.29
Linear (experimental data [10])
0.27
Linear (predicted results)
0.25
0 1 2 3 4 5 6 7 8
log(N)
Materials
The material used for this study was the E-glass/polyurethane composites manufactured
by the pultrusion process. The lay-up of the material is [CSM/0/90/0/90/0/CSM] including:
(1) two plies of Nexus veil stitched to 1 oz/ft2 continuous strand mats (CSM), (2) two
plies of [0/90] stitched fabrics (18 oz/yd2), and (3) one unidirectional layer consisting
824 P. QIAO AND M. YANG
0.8
0.7
0.6
0.5
Ds/su
0.4
0.3
Experimental data [7]
0
0 1 2 3 4 5 6 7 8
log(N )
Predicted results
0.8
Linear (experimental data)
0.6
Ds/s u
0.5
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7
log(N)
Figure 5. Comparison of S–N curves for E-glass/polyurethane composites without considering the unbroken
sample data.
of continuous glass roving of 113 yield (73 tows). The plate panels were cut with a
saw-shear machine, and the coupon samples with dimensions of 25.4 254 3.175 mm3
(1.0 10.0 0.125 in.3) were obtained. These coupon samples were then polished, and
the plastic tabs were bonded to the ends of each coupon using the epoxy adhesive.
Fatigue Life Prediction of Pultruded E-glass/Polyurethane Composites 825
The static behavior of the material was first established by testing five specimens
using the ASTM standard and under the displacement control of loading (loading rate ¼
2.0 mm/min), and the stress–strain curves of the corresponding specimens were recorded
and analyzed. The static experimental tensile Young’s modulus (E0) and strength ( u0)
of the composite is given in Table 2. The common failure modes exhibited during the static
tests included the fiber breaking and interlamina delamination (see Figure 6).
In this study, the effects of stress ratio, frequency, and maximum stress on fatigue life
prediction of E-glass/polyurethane composites were investigated. To obtain the data neces-
sary for establishing the relationship between the low cycle fatigue life and the respective
testing parameters, a range of test combinations with R ¼ 0.05, 0.1, 0.5, and 0.9;
f ¼ 1, 3, and 5 Hz; and max ¼ 0.20, 0.50, and 0.80 u0 was conducted. A load control mode
was adopted, and the fatigue failure is defined as the complete breaking/separation of the
samples. For the classical S–N model to the proposed test combination, a total of 12 S–N
curves is needed corresponding to the combined frequency ( f ) and load ratio (R), resulting
in a total number of 72 tests if a minimum number of six samples is used for each curve.
As per ASTM D 3479 requirements (see Table 1 in [20]), the minimum number of
specimens required for each S–N curve is six for preliminary and exploratory study, 12 for
826 P. QIAO AND M. YANG
research and development testing, and 24 for design allowable and reliability data. Based
on the present model, of which the testing parameters (e.g., the frequency and the load
ratio) are interrelated as shown in Equation (5), all the tests under different frequen-
cies and load ratios can be used to obtain a particular S–N curve, which reduces the
total number of tests to 36. Thus, a significant amount of time and effort could be saved
compared to the classical S–N model using more test samples.
The experimental process involved: (1) let the loading be ramped to the mean stress (the
stress set point); (2) let the load run cyclically to the full amplitude, and count the transi-
tion cycle numbers; and (3) monitor the peak and valley values for the test, and record
the instance of 2% deviations. A total of 36 samples were tested, and their fatigue life (N)
under a given testing parameter combination was recorded (Table 3). No significant
temperature increase was observed during the fatigue experiments, indicating that the
temperature effect on fatigue life could be excluded.
The classical S–N curves for individual stress ratios and frequencies are plotted in
Figure 7. As shown in Figure 7, the stress ratio has an influential effect on the fatigue
life prediction. Using the regression technique, the classical S–N curve equations
and the related coefficient of correlation (COC) for individuals given stress ratios and
frequencies are:
log ¼ 0:2129 logðNÞ þ 0:3707 for R ¼ 0:05, f ¼ 1 Hz ðCOC ¼ 0:9851Þ ð16aÞ
u
log ¼ 0:216 logðNÞ þ 0:4083 for R ¼ 0:05, f ¼ 3 Hz ðCOC ¼ 0:9770Þ ð16bÞ
u
log ¼ 0:2111 logðNÞ þ 0:3255 for R ¼ 0:05, f ¼ 5 Hz ðCOC ¼ 0:9754Þ ð16cÞ
u
log ¼ 0:2718 logðNÞ þ 0:4785 for R ¼ 0:1, f ¼ 1 Hz ðCOC ¼ 0:9997Þ ð16dÞ
u
log ¼ 0:1965 logðNÞ þ 0:2433 for R ¼ 0:1, f ¼ 3 Hz ðCOC ¼ 0:9548Þ ð16eÞ
u
log ¼ 0:1586 logðNÞ þ 0:1255 for R ¼ 0:1, f ¼ 5 Hz ðCOC ¼ 0:9914Þ ð16f Þ
u
log ¼ 0:1597 logðNÞ 0:0704 for R ¼ 0:5, f ¼ 1 Hz ðCOC ¼ 0:9647Þ ð16gÞ
u
log ¼ 0:1520 logðNÞ 0:0491 for R ¼ 0:5, f ¼ 3 Hz ðCOC ¼ 0:9952Þ ð16hÞ
u
log ¼ 0:1645 logðNÞ þ 0:0695 for R ¼ 0:5, f ¼ 5 Hz ðCOC ¼ 0:9953Þ ð16iÞ
u
log ¼ 0:3234 logðNÞ þ 0:2949 for R ¼ 0:9, f ¼ 1 Hz ðCOC ¼ 0:9396Þ ð16jÞ
u
828 P. QIAO AND M. YANG
(a) 0.4
0.2
0
0 1 2 3 4 5 6 7 8
-0.2
Log (∆σ/σu)
y = -0.2111x+0.3255
-0.4
R2=0.9754(R=0.05,f=5)
-0.6 R=0.05,f=1
R=0.05,f=3 y = -0.216x+0.4083
-0.8 R2 = 0.977(R=0.05,f=3)
R=0.05,f=5
Linear (R=0.05,f=1)
-1
Linear (R=0.05,f=3)
y = -0.2129x+0.3707
-1.2 Linear (R=0.05,f=5)
R2 = 0.9851(R=0.05,f=1)
-1.4
log(N)
0.6
(b)
0.4
0.2
0
-1 0 1 2 3 4 5 6 7 8
Log (∆σ/σu)
-0.2
-0.4
y = -0.1586x+0.1255
R=0.1,f=1
-0.6 R2 = 0.9914(R=0.1,f=1)
R=0.1,f=3
-0.8 R=0.1,f=5
Linear (R=0.1,f=1) y = -0.1965x+0.2433
-1 R2 = 0.9548(R=0.1,f=3)
Linear (R=0.1,f=3)
-1.2 Linear (R=0.1,f=5) y = -0.2718x+0.4785
R2 = 0.9997(R=0.1,f=5)
-1.4
log(N)
0
(c) 0 1 2 3 4 5 6 7 8 9
-0.2
-0.4
Log (∆σ/σu)
-0.6
y = -0.1645x+0.0695
-0.8
R2 = 0.9953(R=0.5,f=5)
R=0.5,f=1
-1
R=0.5,f=3
R=0.5,f=5
-1.2
Linear (R=0.5,f=1) y = -0.1597x-0.0704
Linear (R=0.5,f=3) R2 = 0.9647(R=0.5,f=1) y = -0.152x-0.0491
-1.4
R2 = 0.9952(R=0.5,f=3)
Linear (R=0.5,f=5)
-1.6
log(N)
Figure 7. Effect of stress ratios on the S–N curves of E-glass/polyurethane composites: (a) R ¼ 0.05;
(b) R ¼ 0.1; and (c) R ¼ 0.5.
Fatigue Life Prediction of Pultruded E-glass/Polyurethane Composites 829
For f ¼ 3, 5 Hz and R ¼ 0.9, the fatigue life is usually longer than 1.0 106 cycles, which
is the endurance limit for the material from the test data, and it belongs to the high cycle
fatigue failure.
The corresponding S–N curves predicted by the present model are compared with
the tested data for each stress ratio and frequency (Figure 8). As shown in Figure 8,
(a) 0
0 1 2 3 4 5 6
-0.1
-0.2
log(∆σ/σu)
-0.3
-0.4
R=0.05,f=1, predicted
-0.5 R=0.05,f=3, predicted
R=0.05,f=5, predicted
-0.6 R=0.05,f=1, experimental data
R=0.05,f=3, experimental data
-0.7 R=0.05,f=5, experimental data
-0.8
log(N)
(b) 0
0 1 2 3 4 5 6
-0.1
-0.2
-0.3
log(∆σ/σu)
-0.4
R=0.1,f=1, predicted
-0.5 R=0.1,f=3, predicted
R=0.1,f=5, predicted
-0.6 R=0.1,f=1, experimental data
(c) 0
0 1 2 3 4 5 6 7
-0.2
-0.4
log(∆σ/σu)
R=0.5,f=1, predicted
-0.6
R=0.5,f=3, predicted
Figure 8. Comparisons of predicted S–N curves with testing data for E-glass/polyurethane composites:
(a) R ¼ 0.05; (b) R ¼ 0.1; and (c) R ¼ 0.5.
830 P. QIAO AND M. YANG
1
E-glass/ductile epoxy, R=0.1,
unidirection
0.9
E-glass/brittle epoxy, R=0.1,
unidirection
0.8
[1], R=0.05
0.7
[2], R=0.025
0.6
Ds/s u
[3], R=0.1
0.5
0.4 R=0.05,f=1
0.3 R=0.1,f=1
0.1
Linear (R=0.1,f=1)
0
0 1 2 3 4 5 6 7
log(N )
Figure 9. Comparisons of the S–N curves of E-glass/polyurethane composites with other common
composites.
the predicted results are better related to the testing data and could be used to replace the
classical S–N curves.
As expected and also indicated in Equation (16) and Figures 7 and 8, the larger
the stress ratio, the longer the fatigue life proceeds [21]. Compared to other E-glass fiber-
reinforced plastic composites [5,7,10] (see Table 1), the S–N curve of E-glass/polyurethane
composites has a higher value of slope and a comparable interception, which indicates
that the E-glass/polyurethane composites are more fatigue sensitive. The comparisons
of the S–N curves of E-glass/polyurethane composites with other fiber-reinforced plastic
composites are also shown in Figure 9.
As shown in Figure 10 and Equation (16), we could observe that the test frequency
has a pronounced influence on the S–N curve. The slope of the S–N curve decreases as
the frequency increases, which indicates that the increased frequency extends the fatigue
life of the composites as expected.
Besides the effects of stress ratio and frequency, the fatigue life is also dependent on the
mean stress at any given load amplitude. The effect of mean stress can be described by the
modified Goodman Line concept as suggested by the present model. Based on the experi-
mental data, the modified Goodman master diagram is developed and given in Figure 11
for f ¼ 1 Hz. As shown in Figure 11, the fatigue life of E-glass/polyurethane composites
decreases as the mean stress increases. Similarly, as the stress amplitude increases, the
fatigue life is correspondingly reduced.
Fatigue Life Prediction of Pultruded E-glass/Polyurethane Composites 831
0
0 2 4 6 8
f = 1 Hz
-0.2 f = 3 Hz
f = 5 Hz
f = 1 Hz, E-glass/polyester [7]
-0.4 f = 15 Hz, E-glass/polyester [7]
Linear (f = 3 Hz)
Linear (f = 5 Hz)
-0.6 Linear (f = 1 Hz)
log(Ds/s u)
-0.8
-1
-1.2
-1.4
-1.6
log(N )
Figure 10. Effect of frequencies on the S–N curves of E-glass/polyurethane composites for all stress ratios.
0.6
R=0.05
R=0.1
0.5 R=0.5
Normalized stress amplitude
R=0.9
0.4
100
0.3
0.2 103
0.1
105
0
0 0.2 0.4 0.6 0.8 1
Normalized mean stress
Figure 11. Modified Goodman Line diagram of E-glass/polyurethane composites for f ¼ 1 Hz (both the mean
stress and the stress amplitude are normalized by the ultimate static strength).
832 P. QIAO AND M. YANG
As the cyclic loading progresses, the tensile fatigue modulus (Ef) decays from its initial
values E0 (given in Table 2) to the critical fatigue modulus. The fatigue modulus is
measured during the unloading portion of a tension–tension cycle (or tensile part of
the fatigue cycle), which is the slope of the line between the cyclic maximum stress/strain
and the cyclic minimum stress/strain. The initial modulus is measured during the unload-
ing portion of the first fatigue cycle. The normalized fatigue modulus (Ef /E0) is defined
as the ratio between the tensile fatigue modulus at any given fatigue life and the initial
tensile fatigue modulus. Dharan [22] observed that the decay of normalized fatigue
modulus for axial tensile fatigue tests of a unidirectional E-glass/epoxy composite
at f ¼ 5 Hz, R ¼ 0 is about 40% when the critical normalized fatigue modulus was reached.
Agarwal and James [5] also studied the decay of the fatigue modulus from the axial tensile
fatigue tests of an E-glass/epoxy cross-ply composite under stress or strain control of
f ¼ 0.01–2.0 Hz and R ¼ 0.05. Their data indicated that the increasing decay of fatigue
modulus was from 3 to 15% with respect to the decreasing stress range. Jessen and
Plumtree [23] showed that the decay of about 4% in the normalized fatigue modulus
occurred before the critical normalized fatigue modulus was reached for an E-glass/
polyester composite under a constant load level (50–80% ultimate strength) at f ¼ 5 Hz
and R ¼ 0.05.
In this study, the decay of fatigue modulus is also observed (see Figure 12). However,
a moderate degradation of fatigue modulus (about 8%) was observed before the critical
normalized fatigue modulus was reached compared to the published cases in [5,22].
The macroscopic fatigue damage process was observed physically. The physical
damage process of E-glass/polyurethane composites under cyclic loading initiated as the
1.05
0.95
0.9
Ef/E0
0.85
0.8
0.75
0.7
0 20,000 40,000 60,000 80,000 100,000 120,000
N
Figure 12. Degradation of fatigue modulus of E-glass/polyurethane composites at R ¼ 0.05, max ¼ 0.2 u
and f ¼ 5 Hz.
Fatigue Life Prediction of Pultruded E-glass/Polyurethane Composites 833
Figure 13. Failure modes of E-glass/polyurethane composites from the fatigue tests.
interlaminar delamination occurred, followed by partial fiber breaking from one side of
the specimen and eventually extended to the full fiber breaking across the width of the
specimen. The typical failure modes (i.e., interlaminar delamination and fiber breaking)
could be vividly observed in Figure 13.
Due to the scattering fatigue data, a statistical analysis is necessary to represent the test
results. In assessing the reliability of composite structures, the two-parameter Weibull
distribution is commonly used to represent the constant amplitude fatigue life data [20,24].
To facilitate the analysis of the present fatigue data, a set of assumptions for the purpose
of simplification was made: (1) the relationship between the log transformations of the
measured life (N cycles) and the stress is in a linear function and (2) the distribution
834 P. QIAO AND M. YANG
function of the residuals about the mean line is homogeneous, i.e., it does not depend
on the stress level. A two-parameter Weibull distribution cumulative function for the
fatigue life is given as
N
FðNÞ ¼ 1 exp ð17Þ
Using a linear regression program, the best estimations of and are computed as
¼ 63:9464
¼ 0:9434
The bounds with 95% confidence for the Weibull distribution parameters (Figure 14)
and the S–N curves (Figure 15) for the fatigue data of E-glass/polyurethane composites in
this study are, respectively, obtained based on the maximum likelihood technique [25].
3
Predicted data
0
ln(ln(1/(1-F)))
1 2 3 4 5 6
-1
-2
-3
-4
-5
ln(N )
Figure 14. The bounds for the Weibull distribution curve with 95% confidence.
Fatigue Life Prediction of Pultruded E-glass/Polyurethane Composites 835
0.9
Experimental data
0.8
Lower bound with 95%
confidence
0.7
Upper bound with 95%
confidence
0.6 Prediction data
0.5
∆σ/σu
0.4
0.3
0.2
0.1
0
0 1 2 3 4 5 6 7
log(N)
Figure 15. The bounds for the S–N curves of E-glass/polyurethane with 95% confidence.
CONCLUSIONS
ACKNOWLEDGEMENTS
This study is partially supported by the Creative Pultrusions (CP), Inc., Alum Bank,
PA. We thank the technical input provided by Mr Dustin Troutman of CP.
836 P. QIAO AND M. YANG
REFERENCES
23. Jessen, S.M. and Plumtree, A. (1989). Fatigue Damage Accumulation in Pultruded
Glass/polyester Rods, Composites, 20(6): 559–567.
24. ASTM E739-91 (1998). Standard Practice for Statistical Analysis of Linear or Linearized
Stress-Life (S-N) and Strain-Life ("-N) Fatigue Data, ASTM, West Conshohocken, PA.
25. Talreja, R. (1981). Estimation of Weibull Parameters for Composite Material Strength and
Fatigue Life Data, In: Fatigue of Fibrous Composite Materials, ASTM STP 723, American
Society for Testing Materials (ASTM), West Conshohocken, PA, pp. 291–311.