Composites Part B 133 (2018) 129e144

Contents lists available at ScienceDirect

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

Reliability analysis of tensile strengths using Weibull distribution in

glass/epoxy and carbon/epoxy composites
K. Naresh a, K. Shankar a, R. Velmurugan b, *
a
Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai, India
b
Department of Aerospace Engineering, Indian Institute of Technology Madras, Chennai, India

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

Article history: In this study, experiments are carried out to determine the tensile strength of laminates, for three
Received 26 July 2016 different orientations [(0/90/30/-60), (0/90/45/-45) and (30/-60/60/-30)] of glass/epoxy and carbon/
Received in revised form epoxy composites, for the strain rate range of 0.0083e542 s1. Using two-parameter Weibull distribu-
21 October 2016
tion, the theoretical tensile strength values are determined for Glass Fiber Reinforced Polymer (GFRP)
Accepted 1 September 2017
Available online 6 September 2017
and Carbon Fiber Reinforced polymer (CFRP) composites for different strain rates by a linear curve fitting.
The theoretical and experimental values match well. The deviation between the theoretical and exper-
imental values is less than 12% for GFRP laminates and less than 13% for CFRP laminates. Normally the
Keywords:
Glass/epoxy
mean values of mechanical properties are sufficient to use theoretical models, whereas all tested spec-
Carbon/epoxy imen data are considered (including the mean values) in Weibull distribution. Therefore, Weibull dis-
Fabrication tribution contains more information and it will be useful for designers and composite manufacturers to
High strain rate ensure the reliability of structures. Studies show that the (0/90/45/-45) laminates have exhibited the
Scanning electron microscopy highest scale parameter increment of 80.5% for glass/epoxy and 53% for carbon/epoxy composites with
the increase of strain rate, when compared to laminates of other orientations. Using scanning electron
microscopy (SEM), the failure mechanisms such as matrix microcracking and fiber-matrix interface
microcracking are observed in quasi-static testing whereas the fiber pull-out, matrix damage and fiber-
matrix interface cracking are seen in the high strain rate tests, which ensure the increase in strength
values at high strain rate.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction from quasi-static to dynamic loading. In the recent years, many

researchers have focused on Weibull statistics [8e10] to predict the
Aircraft Radome and wing structures, helicopter rotor blades, degree of variability of mechanical properties of FRP composites
automotive bumpers, ship hull structures and sports helmets have [11]. The mechanical properties are accurate when the theoretical
been manufactured using E-glass and carbon fiber reinforced and experimental cumulative probability density (cumulative fail-
polymer composites due to their higher electrical and corrosion ure probability) curves match well. Recent studies [12,13] show that
resistance and better mechanical properties such as high specific the two-parameter Weibull distribution predicts accurately the
strength and stiffness [1e3]. However, these possess different tensile strength distribution.
mechanical properties in different directions owing to anisotropic The studies related to Weibull distribution are outlined briefly.
behavior [4]. It is difficult to predict the accurate mechanical Dirikolu et al. [14] studied the tensile strength variation of carbon/
properties of composites [5]. In particular, strength properties are epoxy composites using the Weibull distribution at a displacement
generally scattered owing to their internal structure and rate of 0.022 mm/s. It has been reported that the Weibull distri-
misalignment of fiber direction during fiber cutting and laminate bution is useful for researchers to determine the composite fracture
preparation [6,7]. Therefore, it is important to employ statistical strength distribution and it is also useful for composite manufac-
analyses for quantifying the mechanical properties of composites turers to estimate the product reliability. Zhu et al. [15] have per-
formed the flexural tests on neat carbon/epoxy and carbon/epoxy/
multi-walled carbon nanotube composites at a displacement rate of
* Corresponding author. 0.9 mm/min and the results are analyzed using the two-parameter
E-mail address: [email protected] (R. Velmurugan). Weibull distribution. It is observed that when neat carbon/epoxy

http://dx.doi.org/10.1016/j.compositesb.2017.09.002
1359-8368/© 2017 Elsevier Ltd. All rights reserved.
130 K. Naresh et al. / Composites Part B 133 (2018) 129e144

composites are replaced with nano filled carbon/epoxy composites, theoretical values which provide more information to designers
the scale parameter increases but the shape parameter decreases. and composite material manufacturers.
This indicates the high strength scatter of nanofilled composites.
Alqam et al. [16] used the two and three parameter Weibull dis- 2. Materials selection and specimen preparation
tributions to quantify the accuracy of tensile, compressive and
shear properties (strength and modulus) of pultruded composites. Materials used for characterization are GFRP and CFRP com-
They have observed that both the distributions are in good agree- posites which use E-glass fiber of woven roving mat (WRM) of 610
ment. Sakin and Ay [17] conducted fatigue experiments on the [0 / gsm and carbon fiber of woven roving mat of 450 gsm as rein-
90 ] and [± 45 ] glass/polyester composites and the results are forcement with Araldite epoxy matrix and Hardener (Aradour),
modelled using the two-parameter Weibull distribution to deter- respectively. Fiber to resin ratio of 1:1 and resin to hardener ratio of
mine the Weibull parameters (shape and scale parameters). The 10:1, by weight, are maintained. Fig. 1 depicts the different stacking
fatigue life distribution curves have been obtained using these sequences [(0/90/30/-60), (0/90/45/-45) and (30/-60/60/-30)] of
parameters. Wang and Xia [18] observed the tensile strength dis- glass/epoxy and carbon/epoxy composites which are prepared by
tribution of E-glass fibers at the strain rates of 90e1700 s1 for the using compression molding technique. Each laminate has four
temperature range of 70 to 140 C using the bimodal Weibull layers and each layer has 0.4 mm thickness. The samples are pre-
distribution. It is observed that the shape parameters are inde- pared by using water jet cutting machine.
pendent of both strain rate and temperature, whereas the scale
parameters change with stain rate as well as with temperature. Ou 3. Experimental details
and Zhu [19] have used the Weibull statistics to measure the vari-
ability of the tensile strength of GFRP composites, in the range of Quasi-static tests have been performed by using the Instron
strain rates from 1/600 to 160 s1. They have examined that the universal testing machine of capacity 30 kN at a strain rate of
tensile strength increases with strain rate. Azeko et al. [20,21] 0.0083 s1. Using the drop mass test setup, high strain rate studies
experimentally investigated the mechanical behavior of laterite/ are carried out. The gauge length of 10 mm is used for the high
cement matrix and different volume percentages (10%, 20% and strain rate test. The drop mass setup is used to perform medium
30%) of Polyethylene (PE)/laterite composites at a strain rate of 0.01 (above 10 s1) to high strain rate (below 1000 s1) studies [34,35],
s1 and the results are validated by employing the Weibull statis- which is shown in Fig. 2. This setup fills the gap between conven-
tical analysis. The results indicated that the composite having tional servo-hydraulic machine and Split-Hopkinson pressure bar.
20 vol % of PE possesses the highest strength (compressive and The high strain rate fixture [Fig. 2 (c)] indicates that when the mass
flexural) and fracture toughness as compared to other combina- is dropped on the fixture top disc (circular plate) which hit the
tions. In contrast, the laterite/cement matrix and 30 vol % of PE/ specimen assembly through the sliding guide rods to induce the
laterite composite values are almost equivalent due to more stress tensile failure. The Piezoelectric load sensor is used to obtain the
concentration in the latter which is attributed to the clustering of dynamic load data with the help of LabVIEW-National Instrument-
PE particles. Data acquisition (DAQ).
High strain rate studies involve the complex failure phenomena The dog bone geometry is used for this study, in order to avoid
owing to inertia and wave propagation effects. Therefore, it is failure in the grip portion. Typical tested samples of different ori-
difficult to obtain the reliable data in dynamic loading [22]. It re- entations of glass/epoxy and carbon/epoxy composites are shown
quires several trials to acquire the consistent values. Furthermore, in Figs. 3 and 4, respectively, which ensure that the failure has
the probability-based studies are limited for woven roving com- occurred in the gauge section.
posites of different stacking sequences subjected to dynamic
loading. Most of these studies have been used for predicting the 4. Results and discussion
quasi-static strength values [23e25]. Hence, efforts have been
made to get theoretical values in both quasi-static and dynamic Tables 1 and 2, present the tensile strength (s) values of glass/
loading by employing the two parameter Weibull distribution. The epoxy and carbon/epoxy composites of different orientations for
main advantage of using the Weibull distribution is not only for the strain rates from quasi-static to 542 s1, respectively.
validating the experimental results but it also measures the strain From Tables 1 and 2, it is apparent that as the strain rate in-
rate sensitivity of composites [26]. creases from quasi-static to 542 s1, the tensile strength increases
Harding and Welsh [27] experimentally studied the strain rate for all the laminates of different orientations of glass/epoxy and
effect of unidirectional CFRP composites and for different orienta- carbon/epoxy composites. The fiber matrix interfacial bond
tions [0 and 45 ] of woven roving GFRP composites. They strength increases with the increase of test speed. As a result,
concluded that the tensile strength of unidirectional carbon/epoxy tensile strength is enhanced at higher strain rates as compared to
composites is independent of strain rate but for both orientations of lower strain rates.
glass/epoxy composites, the tensile strength increases with the For (0/90/30/-60) glass/epoxy composites the tensile strength
increase of strain rate. In our previous work [28], it is observed that increases by 74.5% whereas for (0/90/30/-60) carbon/epoxy com-
there is not much variation of tensile strength (only 6.3% increase) posites the tensile strength increases by 50%, as the strain rate in-
for woven cross ply carbon/epoxy composites. Recent studies show creases from 0.0083 s1 to 542 s1. The corresponding increase is
that owing to shear effect, more disorientated carbon/epoxy com- 82.3% and 55%, for (0/90/45/-45) glass/epoxy and carbon/epoxy
posites are strain rate sensitive [29,30]. It is imperative to under- composites, respectively and for (30/-60/60/-30) laminates the
stand the fiber orientation effect for making carbon/epoxy values are 65% and 48%, respectively.
composites as strain rate sensitive. So, an attempt is made to Due to less fiber content in the loading direction, the (30/-60/
investigate the variability of tensile strength for different orienta- 60/-30) laminates have less tensile strength than (0/90/30/-60) and
tions of glass/epoxy and carbon/epoxy composites. (0/90/45/-45) laminates [36,37]. The increase in tensile strength
Several researchers [31e33] in the past have employed the with strain rate is significantly higher for glass/epoxy composites as
analytical models to determine the theoretical tensile strength compared to the carbon/epoxy composites. The strain rate sensi-
values by using mean experimental values. But in this paper au- tivity on the tensile strength is more for (0/90/45/-45) laminates
thors have considered all tested samples data for obtaining the than the other laminates. This trend is seen both in glass/epoxy and
K. Naresh et al. / Composites Part B 133 (2018) 129e144 131

Fig. 1. Schematic representation of different stacking sequences a) (0/90/30/-60) b) (0/90/45/-45) c) (30/-60/60/-30).

Fig. 3. Failed specimens for different GFRP composites a) (0/90/30/-60) b) (0/90/45/-

45) c) (30/-60/60/-30).

Fig. 2. Drop mass tower a) Experimental setup b) Schematic diagram (CATIA) c)

Tensile fixture.

carbon/epoxy composites.
In contrast to (0/90) carbon/epoxy laminates [Naresh et al. [28]],
the significant strength variation is observed in the present study
for carbon/epoxy composites, which is mainly due to the orienta-
tion of the layers. For theoretical determination of the cumulative Fig. 4. Failed specimens for different CFRP composites a) (0/90/30/-60) b) (0/90/45/-
probability density, tensile strength values [Tables 1 and 2] are fed 45) c) (30/-60/60/-30).
into Eq. (1) from lowest to highest for each set of samples [38].

used to get the variability of tensile strength of glass/epoxy and

4.1. Weibull distribution carbon/epoxy composites for different strain rates from 0.0083 to
542 s1. Using Weibull distribution, the experimental data in
In the present study, two parameter Weibull statistic analysis is Tables 1 and 2 are modelled by linear curve fit. Unlike other
132 K. Naresh et al. / Composites Part B 133 (2018) 129e144

Table 1
Tensile strength results for different laminate orientations of GFRP composite.

Height (m) Strain rate (s1) (0/90/30/-60) (0/90/45/-45) (30/-60/60/-30)

Tensile strength (MPa) Tensile strength (MPa) Tensile strength (MPa)

Quasi-static 0.0083 213.60 165.16 103.25

219.15 174.54 111.66
224.40 181.67 118.84
Avg 219 ± 5.4 Avg 173.5 ± 8.25 Avg 111.25 ± 7.8
0.25 221 262.68 191.14 149.00
269.16 208.76 156.00
273.98 225.86 162.55
Avg 268.6 ± 5.67 Avg 208.58 ± 17.36 Avg 155.85 ± 6.78
0.5 313 266.41 228.86 152.95
286.22 252.23 157.55
304.70 267.45 161.00
Avg 285.78 ± 19.14 Avg 249.51 ± 19.5 Avg 157.16 ± 4
0.75 384 305.61 278.50 157.80
323.88 288.52 160.45
336.75 295.96 162.56
Avg 322 ± 15.65 Avg 287.66 ± 8.76 Avg 160.27 ± 2.38
1 443 350.69 283.74 162.50
358.55 292.00 168.00
364.59 300.88 171.87
Avg 357.94 ± 7 Avg 292.2 ± 8.57 Avg 167.45 ± 4.71
1.25 495 355.48 296.26 174.16
378.78 302.85 176.20
402.17 308.00 177.60
Avg 378.81 ± 23.34 Avg 302.37 ± 5.88 Avg 176 ± 1.73
1.5 542 346.77 310.76 181.50
383.00 316.89 183.63
417.68 321.30 185.25
Avg 382.5 ± 35.45 Avg 316.32 ± 5.3 Avg 183.46 ± 1.88

Table 2
Tensile strength results for different laminate orientations of CFRP composite.

Height (m) Strain rate (s1) (0/90/30/-60) (0/90/45/-45) (30/-60/60/-30)

Tensile strength (MPa) Tensile strength (MPa) Tensile strength (MPa)

Quasi-static 0.0083 282 254.92 109.2

303 272.85 117.9
317 290.14 126.25
Avg 300.7 ± 17.62 Avg 272.63 ± 17.61 Avg 117.8 ± 8.52
0.25 221 350 344.66 135
369 353.92 148
387.17 362.58 155
Avg 368.72 ± 18.58 Avg 353.72 ± 8.96 Avg 146 ± 10.15
0.5 313 377.13 365.7 151
391.54 371.14 156
405.21 378.86 160.94
Avg 391.29 ± 14 Avg 370.9 ± 5 Avg 156 ± 4.97
0.75 384 403.06 375.85 152.91
418.93 388.94 163.5
433.11 400.1 169.82
Avg 418.37 ± 15 Avg 388.3 ± 12.14 Avg 162 ± 8.54
1 443 420.77 376.34 166.16
427.62 400 169
433.8 414.54 171
Avg 427.4 ± 6.51 Avg 396.96 ± 19.28 Avg 168.72 ± 2.43
1.25 495 429.63 395.72 167.8
445 408.52 171.8
456.72 421.63 174.83
Avg 443.78 ± 13.6 Avg 408.62 ± 12.95 Avg 171.48 ± 3.52
1.5 542 432.88 408.73 172.44
454.36 423.33 174.39
467.88 435.96 176
Avg 451.7 ± 17.65 Avg 422.67 ± 13.62 Avg 174.3 ± 1.78

theoretical models, the Weibull distribution requires all tested b

specimen data (including the average values), for the determina- s
h
tion of the Weibull parameters. F ¼1e (1)
The two parameter Weibull distribution expression for cumu-
lative probability density is given by Refs. [11,39]. where, F is the cumulative probability density function (cumulative
failure probability). s denotes the tensile strength. h represents the
K. Naresh et al. / Composites Part B 133 (2018) 129e144 133

scale parameter (characteristic strength) and b represents the

shape parameter (Weibull modulus). j 0:3
F¼ (3)
n þ 0:4
This can be written as
where, j and n are the present and total test numbers in each set,
respectively.
For each laminate orientation of glass/epoxy and carbon/epoxy
composites, 3 samples are tested from quasi-static to high strain
rate. Therefore, for each strain rate and each lay-up sequence, j ¼ 1
1
ln ln ¼ b ln s b ln h (2) to 3 (j ¼ 1 for specimen 1 and j ¼ 3 for specimen 3 in each set of
1F
samples) and n ¼ 3. Scale and shape parameters are calculated for
Cumulative probability density is expressed in terms of Median GFRP and CFRP composites of different orientations using Eq. (2)
rank formula [40,41], which is given by and the known strength values by linear curve fit, which are seen

Fig. 5. Weibull plots for the (0/90/30/-60) GFRP laminate.

Fig. 6. Weibull plots for the (0/90/45/-45) GFRP laminate.

134 K. Naresh et al. / Composites Part B 133 (2018) 129e144

in Figs. 5e7 and Figs. 8e10, respectively. For all curves [Figs. 8e10] of (0/90/30/-60), (0/90/45/-45) and
The correlation coefficient (R2) values are obtained for strain (30/-60/60/-30) carbon/epoxy composites, R2 (correlation coeffi-
rates from quasi-static to 542 s1, from the curves [Figs. 5e7] of (0/ cient) values are greater than 0.91, 0.97 and 0.92, respectively, as
90/30/-60), (0/90/45/-45) and (30/-60/60/-30) corresponding to the strain rate increases from quasi-static (0.0083 s1) to 542 s1.
glass/epoxy composites and the values are greater than 0.93, 0.95 These values reveal the best correlation between experimental and
and 0.97, respectively, which demonstrates the good correlation curve fit. The regression lines [Figs. 8e10] moves towards the right
between experimental and curve fit. It is observed from Figs. 5e7, which indicates the significant strength increment for different ply
that the regression lines are highly spread out for (0/90/45/-45) orientations of carbon/epoxy composites. The Weibull parameters
laminates while the regression lines are closer for (30/-60/60/-30) for glass/epoxy and carbon/epoxy composites of different orienta-
laminates, as the strain rate changes from quasi-static to 542 s1. tions are given in Tables 3 and 4, respectively.
These regression lines indicate that the increment of tensile From Table 3, it is observed that, as the strain rate increases from
strength with strain rate is maximum for (0/90/45/45) laminates quasi-static to 542 s1, the scale parameter increases by 77%, 80.5%
whereas it minimum for (30/-60/60/-30) laminates. and 61% for glass/epoxy composites with lay-ups of (0/90/30/-60),

Fig. 7. Weibull plots for the (30/-60/60/-30) GFRP laminate.

Fig. 8. Weibull plots for the (0/90/30/-60) CFRP laminate.

K. Naresh et al. / Composites Part B 133 (2018) 129e144 135

Fig. 9. Weibull plots for the (0/90/45/-45) CFRP laminate.

Fig. 10. Weibull plots for the (30/-60/60/-30) CFRP laminate.

Table 3
Shape and scale parameters for different GFRP laminates and at different strain rates.

Laminate Weibull parameters 0.0083 s1 221 s1 313 s1 384 s1 443 s1 495 s1 542 s1

0/90/30/-60 b 39.01 46.33 14.00 20.28 48.90 15.44 8.48

h ðMPaÞ 221.12 271.34 294.62 329.20 361.25 389.05 391.80
0/90/45/-45 b 19.56 11.45 12.74 32.60 38.19 48.90 64.97
h ðMPaÞ 176.58 215.78 258.4 291.35 294.64 304.79 318.59
30/-60/60/-30 b 13.81 21.77 39.01 63.97 36.03 97.8 94.81
h ðMPaÞ 114.49 158.88 158.98 161.42 169.51 176.76 184.25
136 K. Naresh et al. / Composites Part B 133 (2018) 129e144

Table 4
Shape and scale parameters for different CFRP laminates and at different strain rates.

Laminate Weibull parameters 0.0083 s1 221 s1 313 s1 384 s1 443 s1 495 s1 542 s1

0/90/30/-60 b 16.19 19.09 27.96 27.54 63.97 31.98 25.73

h ðMPaÞ 309.67 376.60 396.79 424.3 430.10 449.45 459.44
0/90/45/-45 b 14.95 39.01 68.95 32.6 21.32 32.6 31.98
h ðMPaÞ 279.92 357.38 373.58 393.28 404.26 414.45 427.52
30/-60/60/-30 b 13.52 13.72 24.78 18.96 63.97 46.33 97.8
h ðMPaÞ 121.12 150.31 157.95 165.9 169.69 173 175

Fig. 11. Gamma function values for different GFRP laminates.

Fig. 12. Gamma function values for different CFRP laminates.

K. Naresh et al. / Composites Part B 133 (2018) 129e144 137

Table 5
Theoretical tensile strength values for glass/epoxy composites.

Height (m) Strain rate (s1) (0/90/30/-60) (0/90/45/-45) (30/-60/60/-30)

Tensile strength (MPa) Tensile strength (MPa) Tensile strength (MPa)

Quasi-static 0.0083 218 187.5 126.5

0.25 221 266 237 171.2
0.5 313 303 281.5 167.94
0.75 384 331.61 303.3 168.86
1 443 353.9 305.37 179.40
1.25 495 397.67 314 183.90
1.5 542 420.48 327 191.80

Table 6
Theoretical tensile strength values for carbon/epoxy composites.

Height (m) Strain rate (s1) (0/90/30/-60) (0/90/45/-45) (30/-60/60/-30)

Tensile strength (MPa) Tensile strength (MPa) Tensile strength (MPa)

Quasi-static 0.0083 335.37 298.64 135.28

0.25 221 404.25 366.54 167.71
0.5 313 419.2 379 170.89
0.75 384 448.5 405.34 181.63
1 443 445.55 423.22 179.23
1.25 495 472.77 427.16 183.8
1.5 542 486.84 440.9 184

Fig. 13. Experimental and theoretical cumulative probability density vs. tensile strength of the (0/90/30/-60) GFRP laminates.

(0/90/45/-45) and (30/-60/60/-30), respectively. Similarly, the scale The average tensile strength and scale parameter values (char-
parameter increases by 48.5%, 53% and 44.5%, respectively for the acteristic strength) [40,43] are related by the expression
carbon/epoxy composites (Table 4) of same lay-ups. The shape

parameter is a function of the variation of data in each set of the 1
s0 ¼ h Gþh G (4)
samples. The higher values of shape parameter provide uniform b
distribution. The strength values are highly scattered when b values
are lower and vice versa [42]. Gilat et al. [29] reported that the where,
strain rate sensitivity is more for the [±45 ]s carbon/epoxy lami-
nates owing to shear effect induced in the matrix. It leads to the s0 is the average tensile strength; G is the gamma function.
conclusion that laminates having [±45 ] lay-ups possess more
strain rate effect. In the present study, it is observed (Tables 3 and 4) The gamma function values for GFRP and CFRP composites of
that the strain rate sensitivity on the tensile strength is more for different orientations are calculated by using Eq. (4), from the
both GFRP and CFRP laminates of (0/90/45/-45) lay-up sequence known scale parameter and mean experimental values, which are
than the other laminates. shown in Figs. 11 and 12, respectively.
138 K. Naresh et al. / Composites Part B 133 (2018) 129e144

From Figs. 11 and 12, the gamma function ðG Þ values of 0.96, match well. The difference between the theoretical and experi-
1.01, and 1.03 are obtained for the (0/90/30/-60), (0/90/45/-45) and mental average tensile strength values are less than 10.5%, 9% and
(30/-60/60/-30 glass/epoxy composites, respectively and the cor- 13% for the laminates of orientation (0/90/30/-60), (0/90/45/-45)
responding values for the carbon/epoxy composites are 1.02, 1 and and (30/-60/60/-30), respectively.
1.04. The corresponding R2 (correlation coefficient) values of 0.99 The theoretical cumulative probability density is determined
are obtained for all laminates. The theoretical tensile strength from Eq. (1) by using s, h and b values. The experimental value of
values are calculated by substituting the scale parameter; shape cumulative probability density is determined from Eq. (3) using
parameter and gamma function values in Eq. (4) and are given in known j and n values. The comparison of experimental and theo-
Tables 5 and 6, for glass/epoxy and carbon/epoxy composites, retical values vs. tensile strength of the GFRP and CFRP composites
respectively. of different orientations are shown in Figs. 13e15 and Figs. 16e18,
From Table 5, it is observed that the mean theoretical tensile respectively.
strength values of glass/epoxy composites are in good agreement From Figs. 13e15 and Figs. 16e18, it is examined that as the
with the experimental values (Table 1). The difference between the strain rate increases from quasi-static (0.0083 s1) to high strain
theoretical and mean experimental values are less than 10.5%, 11.5% rate (542 s1), the curves shift to the higher strength direction
and 12% for GFRP composites of orientation (0/90/30/-60), (0/90/ (towards right) for all the laminates of glass/epoxy and carbon/
45/-45) and (30/-60/60/-30), respectively. epoxy composites. It is worthy to note that there is a significant
From Tables 2 and 6, it is seen that the theoretical and experi- increase in the tensile strength in the glass/epoxy and carbon/
mental mean tensile strength values of carbon/epoxy composites epoxy composites for all the three orientations, as the strain rate

Fig. 14. Experimental and theoretical cumulative probability density vs. tensile strength of the (0/90/45/-45) GFRP laminates.

Fig. 15. Experimental and theoretical cumulative probability density vs. tensile strength of the (30/-60/60/-30) GFRP laminates.
K. Naresh et al. / Composites Part B 133 (2018) 129e144 139

increases from 0.0083 to 542 s1. The theoretical and experimental In general, glass fibers are abrasive by nature. They induce
cumulative probability density values are in good agreement. It has cracks in the matrix and fiber-matrix interfacial areas when loads
led to the conclusion that the tensile strength values of the GFRP are axially applied [44,45]. These cracks are propagated with in-
and CFRP composites are accurate, which can be used for designing crease in strain rate and cause fiber pull-out, matrix damage and so
the structural elements in the aerospace and automotive vehicles. on. The fiber-matrix debonding, matrix microcracking, fiber-matrix
adhesion, fiber fracture surface and resin rich areas are observed in
GFRP composites of (0/90/30/-60), (0/90/45/-45) and (30/-60/60/-
5. Scanning electron microscopy analysis
30) laminates, respectively during quasi-static testing, which are
seen in SEM images [Fig. 19(aef)].
Using Hitachi S-4800 SEM instrument, the scanning electron
The river markings, matrix damage, fiber-matrix bonding, fiber
microscopy studies are conducted on the failed tensile test speci-
pull-out, fiber breakage, areas of little matrix adhesion are observed
mens of glass/epoxy and carbon/epoxy composites for different
in the fractography analysis of glass/epoxy composites of different
laminate configurations. The specimens are gold coated prior to
laminate orientations under dynamic loading rates, which are seen
SEM analysis using an Ion sputter equipment to improve the
in Fig. 20(aef).
conductance for proper investigation of fracture surface. For each
Under quasi-static tension, progressive damage occurs in the
laminate orientation of GFRP and CFRP composites, two SEM im-
glass/epoxy composites such that the glass fibers being tightly held
ages of the same scale (50 mm) are presented for better under-
by the epoxy resin even after failure [46], which is examined in SEM
standing of failure mechanisms during quasi-static and high strain
images whereas more fiber fractures to attribute the brittle failure
rates.

Fig. 16. Experimental and theoretical cumulative probability density vs. tensile strength of the (0/90/30/-60) CFRP laminates.

Fig. 17. Experimental and theoretical cumulative probability density vs. tensile strength of the (0/90/45/-45) CFRP laminates.
140 K. Naresh et al. / Composites Part B 133 (2018) 129e144

Fig. 18. Experimental and theoretical cumulative probability density vs. tensile strength of the (30/-60/60/-30) CFRP laminates.

Fig. 19. SEM images for the glass/epoxy composites of different laminate configurations at quasi-static testing a) Fiber-matrix debonding (0/90/30/-60) b) Matrix microcracking (0/
90/30/-60) c) Fiber-matrix adhesion (0/90/45/-45) d) Fiber-matrix debonding (0/90/45/-45) e) Fiber fracture surface (30/-60/60/-30) f) Resin rich area (30/-60/60/-30).
K. Naresh et al. / Composites Part B 133 (2018) 129e144 141

at a higher test speed due to less stress relaxation time [47]. The matrix debonding, fiber-matrix bonding, broken fibers, matrix
fiber matrix debonding is seen in both quasi-static and dynamic adhered to the fibers and fiber-matrix interface microcracking in
loading. the CFRP composites of different ply orientations.
Matrix microcracks [Fig. 19(a)] are observed in (0/90/30/-60) The damage mechanisms such as fiber-matrix interface
laminates during quasi-static testing while matrix cracks in the cracking, matrix damage, bunch fiber-pullout, crack propagation,
form of river markings and matrix damage [Fig. 20(a)] are found in broken fibers and river markings are seen in CFRP laminate of
dynamic loading rates, which are seen in SEM micrographs. The different configurations in high strain rate loading, which are given
areas of fiber-matrix bonding are rarely seen in the failed speci- in SEM pictures [Fig. 22(aef)].
mens of (0/90/30/-60) GFRP laminates in high strain rate It is observed that the damage of carbon/epoxy composites
[Fig. 20(b)]. The (0/90/45/-45) laminates fail mainly due to inter- under quasi-static loading is less because enough time is available
action effects of adjacent plies. When the (0/90/45/-45) laminates for failure of the composite as compared to high strain rate. The
are axially loaded, the matrix cracks first occur in the 90 ply in primary mode of failure in quasi-static testing is shear failure.
unidirectional or (0 /90 ) ply in bi-directional laminates. These When loads are applied in the longitudinal direction, the matrix
cracks propagate across the thickness until these reach the inter- microcracks form in the epoxy matrix that leads to a shear failure.
face of 0 /90 /±45 [48,49]. Furthermore, the cracks in the ±45 ply Kim et al. [50] and Rhee et al. [51] have reported that the woven
grow with test speed and are responsible for the mode of failure carbon/epoxy composites possess typical wear mechanisms of
from very little fiber breakage with matrix adhesion [Fig. 19(c)] in fiber-matrix interface debonding, microcracks in the epoxy resin,
quasi-static loading to shear failure followed by a fiber pull-out fiber pull out and fiber breakage which are also observed in the
[Fig. 20(c)] in high strain rate. In the case of (30/-60/60/-30) present study for quasi-static loading [Fig. 21(b), (d) and (f)].
glass/epoxy laminates, the fiber fracture surfaces and resin rich In (0/90/30/-60) carbon/epoxy laminates, the tensile splitting of
areas [Fig. 19(e and f)] are found in SEM images during quasi-static fibers and fiber matrix debonding [Fig. 21(a and b)] are seen in SEM
testing whereas fiber breakage with areas of very little matrix micrographs for quasi-static testing. In contrast to this, the fiber-
adhesion [Fig. 20(e and f)] are seen in dynamic loading. matrix interface cracking and severe matrix damage [Fig. 22(a
The SEM micrographs subjected to quasi-static loading are dis- and b)] are seen in high strain rate. In quasi-static loading, for [0/90/
played in Fig. 21(aef) which indicate criss-cross lay-ups, fiber- 45/-45] CFRP laminates, the brittle failure at the fiber ends occurs

Fig. 20. SEM images for the glass/epoxy composites of different laminate configurations at high strain rate testing a) River markings (0/90/30/-60) b) Fiber-matrix bonding (0/90/
30/-60) c) Fibers criss-crossing at 45 /-45 (0/90/45/-45) d) Fiber breakage (0/90/45/-45) e) Areas of little matrix adhesion (30/-60/60/-30) f) Fiber breakage (30/-60/60/-30).
142 K. Naresh et al. / Composites Part B 133 (2018) 129e144

Fig. 21. SEM images for the carbon/epoxy composites of different laminate configurations at quasi-static testing a) Criss-cross layups (0/90/30/-60) b) fiber-matrix debonding (0/90/
30/-60) c) Fiber-matrix bonding (0/90/45/-45) d) Broken fibers (0/90/45/-45) e) Matrix adhered to fibers (30/-60/60/-30) f) Fiber-matrix interface microcracking (30/-60/60/-30).

due to good fiber-matrix interface bonding which is seen in and matrix becomes weaker which initiates failure, as is observed
Fig. 21(c and d), since the fiber-matrix interface bonding persuades by Okoli et al. [54].
the shear and tensile stresses in the laminates [52]. The bunch fiber
pull out [Fig. 22(c)] and crack propagation [Fig. 22(d)] perpendic-
6. Conclusions
ular to the loading direction are seen in [0/90/45/-45] CFRP lami-
nates for high strain rate studies. For quasi-static testing of (30/-60/
In the present work, experiments are carried out for different
60/-30) CFRP laminates, the matrix protruding from the fiber
strain rates ranging from 0.0083 s1 to 542 s1 and the results are
fracture surfaces is seen in Fig. 21(e and f), i.e, the shattered matrix
used in two parameter Weibull distribution to determine the
pieces still adhered to carbon fibers. The fiber breakage and cohe-
theoretical tensile strength values of glass/epoxy and carbon/epoxy
sive failure of matrix [Fig. 22(e and f)] are seen in dynamic loading.
composites of three different orientations [(0/90/30/-60), (0/90/
Once the matrix scattered from the fibers at high speed, complete
45/-45) and (30/-60/60/-30)]. Comparison of GFRP and CFRP lam-
laminate failure occurs instantaneously [Fig. 22(e)] due to less fiber
inates with the stacking sequences of (0/90/30/-60) and (0/90/45/-
and more resin content in the loading direction as compared to the
45) indicate that the obtained values of the tensile strength of CFRP
laminates of other orientations. Thus the failure mainly depends on
laminates are significantly higher than those of GFRP laminates but
the matrix behavior for these laminates.
for (30/-60/60/-30) lay-up sequence both materials have approxi-
SEM examinations, evidence a brittle failure in high strain rate
mately equal tensile strength due to the dominant of epoxy matrix.
loading. The fiber-matrix bonding strength increases with the in-
For the (0/90/30/-60), (0/90/45/-45) and (30/-60/60/-30) GFRP
crease of strain rate [53]. As a consequence, the tensile strength
laminates, the scale parameter (characteristic strength) increases
increases for both glass/epoxy and carbon/epoxy composites of
by 77%, 80.5% and 61%, respectively, whereas for CFRP laminates,
different laminate orientations. The increase in bonding strength is
the corresponding increase are 48.5%, 53% and 44.5%, respectively
less for (30/-60/60/-30) laminates due to less fiber content in the
with the increase of strain rate from quasi-static to 542 s1. The
loading direction when compared to other laminates. Hence, the
strain rate sensitivity of all laminate orientations is higher for GFRP
(30/-60/60/-30) laminates have less tensile strength. It is also
composites than CFRP composites. Both GFRP and CFRP composites
observed that as test speed increases, the fibers become stronger
of the (0/90/45/-45) laminates possess more strain rate sensitivity
K. Naresh et al. / Composites Part B 133 (2018) 129e144 143

Fig. 22. SEM images for the carbon/epoxy composites of different laminate configurations at high strain rate testing a) Fiber-matrix interface cracking (0/90/30/-60) b) Matrix
damage (0/90/30/-60) c) Bunch fiber-pullout (0/90/45/-45) d) Crack propagation (0/90/45/-45) e) Broken fibers (30/-60/60/-30) f) River markings (30/-60/60/-30).

than (0/90/30/-60) and (30/-60/60/-30) laminates. The theoretical Crashworthiness 2016;21:542e54.

[4] Ramesh M, Gopinath A, Deepa C. Machining characteristics of fiber reinforced
tensile and cumulative probability density values are obtained us-
polymer composites: a review. Indian J Sci Technol 2016;9:1e7.
ing Weibull parameters. These values show good agreement with [5] Lifshitz JM. Impact strength of angle ply fiber reinforced materials. J Compos
experimental results. SEM fractography analysis is used to get some Mater 1976;10:92e101.
information about strength of interaction between glass and car- [6] Andersons J, Joffe R, Hojo M, Ochiai S. Glass fiber strength distribution
determined by common experimental methods. Compos Sci Technol 2002;62:
bon fibers, and epoxy resin which provides the evidence of the 131e45.
brittle failure in dynamic loading such as matrix damage, fiber pull- [7] Pickering KL, Murray TL. Weak link scaling analysis of high-strength carbon
out and fiber breakage are found in (0/90/30/-60), (0/90/45-45) and fiber. Compos Part A 1999;30:1017e21.
[8] Weibull W. A statistical function of wide applicability. J Appl Mech 1951;18:
(30/-60/60/-30) glass/epoxy laminates, respectively whereas in 293e307.
carbon/epoxy composites of same lay-ups, the fiber-matrix inter- [9] Swolfs Y, Morton H, Scott AE, Gorbatikh L, Reed PAS, Sinclair I, et al. Syn-
face cracking, bunch fiber pull-out and river markings in the matrix chrotron radiation computed tomography for experimental validation of a
tensile strength model for unidirectional fibre-reinforced composites. Compos
are observed. Part A 2015;77:106e13.
[10] Catangiu A, Ungureanu DN, Despa V. Data scattering in strength measurement
of steels and glass/epoxy composite. Sci Bull Valahia Univ Mater Mech
Acknowledgment
2017;15:11e6.
[11] Barbero E, Fernandez-Saez J, Navarro C. Statistical analysis of the mechanical
The project is supported by ACECOST of AR&DB (Ref. No. DARO/ properties of composite materials. Compos Part B Eng 2000;31:375e81.
08/1051633/M/1), Structures Panel, India. [12] Huang Y, Talreja R. Statistical analysis of oblique crack evolution in composite
laminates. Compos Part B Eng 2014;65:34e9.
[13] Lekou DJ, Philippidis TP. Mechanical property variability in FRP laminates and
References its effect on failure prediction. Compos Part B Eng 2008;39:1247e56.
[14] Dirikolu MH, Aktas A, Birgoren B. Statistical analysis of fracture strength of
composite materials using Weibull distribution. Turk J Eng Env Sci 2002;26:
[1] Chowdhury NM, Chiu WK, Wang J, Chang P. Experimental and finite element
45e8.
studies of bolted, bonded and hybrid step joints of thick carbon fibre/epoxy
[15] Zhou Y, Pervin F, Lewis L, Jeelani S. Fabrication and characterization of carbon/
panels used in aircraft structures. Compos Part B Eng 2016;100:68e77.
epoxy composites mixed with multi-walled carbon nanotubes. Mater Sci Eng
[2] Singh H, Namala KK, Mahajan P. A damage evaluation study of E glass/epoxy
A 2008;475:157e65.
composite under low velocity impact. Compos Part B Eng 2015;76:235e48.
[16] Alqam M, Bennett RM, Zureick AH. Three-parameter vs. two-parameter
[3] Zhu P, Lu J, Ji Q, Cheng Z. Experimental study of in-plane mechanical perfor-
Weibull distribution for pultruded composite material properties. Compos
mance of a carbon/glass hybrid woven composite at different strain rates. Int J
144 K. Naresh et al. / Composites Part B 133 (2018) 129e144

Struct 2002;58:497e503. [35] Groves SE, Sanchez RJ, Lyon RE, Brown AE. High strain rate effects for com-
[17] Sakin R, Ay I. Statistical analysis of bending fatigue life data using Weibull posite materials. In: Camponeschi ET, editor. Composite materials: testing and
distribution in glass-fiber reinforced polyester composites. Mater Des design, ASTM STP 1206, vol. 11. Philadelphia: American Society for Testing
2008;29:1170e81. and Materials; 1993. p. 162e76.
[18] Wang Z, Xia Y. Experimental evaluation of the strength distribution of fibers [36] Sikarwar RS, Velmurugan R, Vemuri M. Experimental and analytical study of
under high strain rates by bimodal Weibull distribution. Compos Sci Technol high velocity impact on kevlar/epoxy composite plates. Cent Eur J Eng 2012;2:
1997;57:1599e607. 638e50.
[19] Ou Y, Zhu D. Tensile behavior of glass fiber reinforced composites at different [37] Sikarwar RS, Velmurugan R, Gupta NK. Influence of fiber orientation and
strain rates and temperatures. Constr Build Mater 2015;96:648e56. thickness on the response of glass/epoxy composites subjected to impact
[20] Azeko ST, Mustapha K, Annan E, Odusanya OS, Soboyejo ABO, Asce F, et al. loading. Compos Part B Eng 2014;60:627e36.
Statistical distribution of the strength and fracture toughness of recycled [38] McCool JI. Using the Weibull distribution: reliability, modeling and interfer-
polyethylene-reinforced laterite composites. J Mater Civ Eng 2016;28: ence. chapter 3. Wiley series in Probability and Statistics; 2012.
04015146. [39] Baojiang D, Changqin J, Ping C, Jingmin G. Study of tensile strength distribu-
[21] Azeko ST, Mustapha K, Annan E, Odusanya OS, Soboyejo WO. Recycling of tion based on composite materials for aeronautical engineering. Mod Appl Sci
Polyethylene into strong and tough earth-based composite building materials. 2012;6.
J Mater Civ Eng 2016;28:04015104. [40] Khalili A, Kromp K. Statistical properties of Weibull estimators. J Mater Sci
[22] Jacob GC, Starbuck JM, Fellers JF, Simunovic S, Boeman RG. Strain rate effects 1991;26:6741e52.
on the mechanical properties of polymer composite materials. J Appl Polym [41] Wang F, Shao J. Modified Weibull distribution for analyzing the tensile
Sci 2004;94:296e301. strength of bamboo fibers. Polymers 2014;6:3005e18.
[23] Swolfs Y, Verpoest I, Gorbatikh L. Issues in strength models for unidirectional [42] Xiaoju G, Laifei C, Dongming Y, Liangjun LI. Quasi-static and dynamic
fibre-reinforced composites related to Weibull distributions, fibre packings compressive fracture behavior of SiCf/SiC composites. J Wuhan Univ Technol
and boundary effects. Compos Sci Technol 2015;114:42e9. Mater Sci Ed 2015;30:484e8.
[24] Torres JP, Vandi LJ, Veidt M, Heiztmann MT. Statistical data for the tensile [43] Foray G, Descamps-Mandine A, R'Mili M, Lamon J. Statistical flaw strength
properties of natural fibre composites. Data Brief 2017;12:222e6. distributions for glass fibres: correlation between bundle test and AFM-
[25] Lim K, Yap AUJ, Agarwalla SV, Tan KBC, Rosa V. Reliability, failure probability, derived flaw size density functions. Acta Mater 2012;60:3711e8.
and strength of resin-based materials for CAD/CAM restorations. J Appl Oral [44] Nairn JA. Matrix microcracking in composites. Polym matrix Compos 2000;2:
Sci 2016:447e52. 403e32.
[26] Zhou LJ, Xu SL, Shi CY. Full field strain statistics and failure mechanism of [45] Jagannathan N, Gururaja S, Manjunatha CM. Probabilistic strength based
granite under dynamic and quasi-static compression. Rock dynamics: from matrix crack evolution in multi-directional composite laminates. Compos Part
research to engineering. In: Proceedings of the 2nd international conference B Eng 2016;87:263e73.
on rock dynamics and applications (ROCDYN-2); 2016. p. 239e46. Suzhou, [46] Cecen V, Seki Y, Sarikanat M, Tavman IH. FTIR and SEM analysis of polyester-
China. and epoxy- based composites manufactured by VARTM process. J Appl Polym
[27] Harding J, Welsh LM. A tensile testing technique for fibre-reinforced com- Sci 2008;108:2163e70.
posites at impact rates of strain. J Mater Sci 1983;18:1810e26. [47] Okoli OI. The effects of strain rate and failure modes on the failure energy of
[28] Naresh K, Shankar K, Rao BS, Velmurugan R. Effect of high strain rate on glass/ fiber reinforced composites. Compos Struct 2001;54:299e303.
carbon/hybrid fiber reinforced epoxy laminated composites. Compos Part B [48] Tong J, Guild FJ, Ogin SL, Smith PA. On matrix crack growth in quasi-isotropic
Eng 2016;100:125e35. laminates-I. experimental investigation. Compos Sci Technol 1997;57:
[29] Gilat A, Goldberg RK, Roberts GD. Experimental study of strain-rate- 1527e35.
dependent behavior of carbon/epoxy composite. Compos Sci Technol [49] Singh CV, Talreja R. A synergistic damage mechanics approach for composite
2002;62:1469e76. laminates with matrix cracks in multiple orientations. Mech Mater 2009;41:
[30] Sebaey TA, Costa J, Maimi P, Batista Y, Blanco N, Mayugo JA. Measurement of 954e68.
the in situ transverse tensile strength of composite plies by means of the real [50] Kim MT, Rhee KY, Lee JH, Hui D, Lau AKT. Property enhancement of a carbon
time monitoring of microcracking. Compos Part B Eng 2014;65:40e6. fiber/epoxy composites by using carbon nanotubes. Compos Part B Eng
[31] Shokrieh MM, Omidi MJ. Tension behavior of unidirectional glass/epoxy 2011;42:1257e61.
composites under different strain rates. Compos Struct 2009;88:595e601. [51] Rhee KY, Park SJ, Hui D, Qiu Y. Effect of oxygen plasma-treated carbon fibers
[32] Brown KA, Brooks R, Warrior NA. The static and high strain rate behavior of a on the tribological behavior of oil-absorbed carbon/epoxy woven composites.
commingled E-glass/polypropylene woven fabric composite. Compos Sci Compos Part B Eng 2012;43:2395e9.
Technol 2010;70:272e83. [52] Yeung P, Broutman LJ. The effect of glass-resin interface strength on the
[33] Chiem CY, Liu ZG. The relationship between tensile strength and shear impact strength of fiber reinforced plastics. Polym Eng Sci 1978;18:62e72.
strength in composite materials subjected to high strain rates. J Eng Mater [53] Okoli OI, Smith GF. High strain rate characterization of a glass/epoxy com-
Technol 1988;110:191e4. posite. J Compos Technol Res 2000;22:3e11.
[34] Gurusideswar S, Velmurugan R, Gupta NK. High strain rate sensitivity of [54] Okoli OI, Smith GF. Failure modes of fibre reinforced composites: the effects of
epoxy/clay nanocomposites using non-contact strain measurement. Polymer strain rate and fibre content. J Mater Sci 1998;33:5415e22.
2016;86:197e207.