The Ratio of Flexural Strength to Uniaxial Tensile Strength in Bulk
Epoxy Resin Polymeric Materials
M. Yekani Farda,*, B. Rajib, and A. Chattopadhyaya
a
b

School for Engineering of Matter, Transport and Energy, Arizona State University, AZ 85287, USA

Pipe Reconstruction Inc, 2800 North Central Avenue, Phoenix, AZ 85004, USA

Abstract
The flexural behavior of epoxies was investigated by performing mechanical tests and applying
statistical Weibull theory and analytical methods to the results. The effects of loading systems
and environmental conditions were also considered. Three kinds of epoxies were studied: Epon
E863, PRI 2002, and PR520. In total, 53 three-point-bending (3PB) Epon E863 samples and 26
3PB PR520 were tested immediately after curing. The authors also tested 26 four-point-bending
(4PB) PRI2002 samples stored at 60°C and 90% Rh for 48 weeks. The Weibull parameters were
estimated using both linear regression and the moments method. The statistical character of the
Weibull model leads to uncertainty in the evaluated parameters, even for a large number of
experiments. This study analyzed the ratio of flexural strength to tensile strength in bulk epoxy
resin polymers. An analytical method previously developed by the authors to study the
relationship between uniaxial tension/compression stress-strain curves and flexural loaddeflection response was used to obtain the ratio. The results show that the Weibull model
overpredicted the aforementioned ratio in different load arrangements.
Keywords
Epoxy resin, flexural strength, nonlinear behavior, experimental mechanics, Weibull model,
analytical simulation
1. Introduction
Epoxy resin polymeric materials are one of the most important classes of thermosetting polymers
and are becoming increasingly popular for domestic and industrial uses. They are used as
adhesively bonded joints for structural members in different applications, such as civil
infrastructure (e.g., pipe coatings) and transport (e.g., aerospace and automotive industries). They
are also extensively used in dentistry and medical orthopedic prostheses. Despite their wide
range of potential applications, the use of epoxy resin polymers is still limited due to their
variable mechanical properties. In addition, these highly cross-linked networks are inherently
semi-brittle. In practice, the material system is a two-part thermoset epoxy resin in which an
epoxy component is mixed with a hardener to initiate curing and create a chemical reaction. The
Corresponding Author. Department of Mechanical and Aerospace Engineering, School for Engineering of
Matter, Transport and Energy, Arizona State University, AZ 85287, USA. Tel: +1 (480) 370 – 9786,
Email: Masoud.yekanifard@asu.edu

hardened material contains various defects that control fracture initiation, and their efficiency as
crack initiators is dependent on their size and shape [1-4]. The Weibull model has been widely
used to describe the statistical distribution of mechanical properties in brittle materials [5-7].
This model is based on the “weakest link” hypothesis, in which the most serious flaw controls
the strength. The estimation of Weibull parameters has been an interesting theme of scientific
research in the last 50 years [8–12]. Much effort was spent on studying the biasing of the
parameters evaluated by different methods, including linear regression, the method of moments,
and the maximum likelihood method. There are still two major limits to the description of
mechanical properties using the Weibull model. (i) Due to its statistical nature, there is
uncertainty in parameters obtained from a limited number of experiments. An infinite number of
samples should be tested to obtain the true values of Weibull parameters. (ii) Complex flaw
distribution within a material can have various effects on different superimposed Weibull
distributions. A large number of studies have been undertaken to assess the strength dispersion of
different epoxy resin systems using the Weibull theory [13 – 22]. Some examples include the
flexural strength distributions of unmodified and thermoplastic modified epoxy resin [13], size
effects on flexural strength of poly(methyl methacrylate)-based bone cements [14], flexural
behavior of sepiolite-modified epoxy resin [15], effects of voids on tensile strength and Young’s
modulus of structural adhesives [16], viscoelastic properties of epoxy-phenolic reinforced with
multi-walled carbon nanotubes (MWCNTs) [18], strength distribution of E-glass/epoxy MWCN
composites [19], and tensile failure of unidirectional composites reinforced with flax [20]. In
most of these studies, the most serious flaw controls material failure.
The complicating factor in determining the strength of bulk epoxy resin materials is their
complex failure mechanisms. As in brittle materials, the volume of the body subject to stress can
influence the measured strength in semi-brittle polymeric materials. This phenomenon is termed
size and stress gradient effects [23-25]. In general, the strength of an epoxy resin is determined
by the presence and interactions of defects (e.g., voids and micro-cracks), the generation of the
tensile stresses at these defects, and the stress gradient along the fracture path. Flexural testing is
considered an appropriate measure of the strength because it combines elements of compression,
tension and shear, which more closely mimics in vivo stresses than either compression or tension
testing alone. The effect of the stress gradient on material behavior is shown in Figure 1.
The ratio of the flexural strength to the uniaxial tensile strength has not been fully examined in
previous studies. The results presented in this paper are a part of a much broader investigation
into the mechanical behavior of bulk epoxy resin polymeric materials [26–30]. The main
objective of the present study is three-fold: (i) to investigate the effect of loading systems on
mean strength using the Weibull model, (ii) to obtain the ratio of mean strength measured in a
flexural test to the mean strength in a uniaxial tension test using both analytical [30] and
statistical approaches that reflect stress gradient effects on strength, and (iii) to study the effects
of environmental conditions and aging on this ratio.

2. Materials, Characterization, and Methods
Epon E863 and hardener EPI-CURE 3290 with a 100/27 weight ratio, PRI 2002-3-R-A and
hardener PRI2000-5-HR-B with a 100/22 weight ratio, and PR520 dyed with Nigrazine were
used in this study. PRI has a low glass transition temperature and a wide application for civil
infrastructure repair and rehabilitation due to its high workability. Epon E863 and PR520 have
relatively high glass transition temperatures and have wide applications in the aerospace
industry. The PRI2002 samples stored at 60°C and 90% Rh for almost 48 weeks were
conditioned and tested. The two-part thermoset epoxies were hand mixed using spatulas and
deposited into moulds while minimizing the enclosure of large defects (air bubbles). The same
casting procedure was used to make all the tension, compression and bending samples to ensure
that the method of fabrication did not influence the mechanical properties. All samples were
cured according to the suppliers’ recommendations and were machined from the plates. ASTM
standards D638 [31], D695 [32], D790 [33], and D 6272 [34] were considered for tension,
compression, and three-point-bending (3PB) and four-point-bending (4PB) flexural tests,
respectively. All the tests were conducted at room temperature, in displacement control and at a
loading speed of 493 µstr/sec. The loading rates corresponding to the desired strain rates were
calculated based approximately on the elastic linear assumption and geometry of the samples.
Beams of Epon E863 and PR520 with average dimensions of 3.2 mm (width) × 3 mm (thickness)
× 25 mm (length) and 20-mm span were tested in the 3PB setup. Each 3PB sample had a groove
with an average radius of 0.5 mm at the middle of the beam. The detail of the notch in PR520
was different compared to the shape of the groove in the other two material systems. 4PB beams
of PRI2002 with average dimensions of 4.2 mm (width) × 4 mm (thickness) × 25 mm (length
between two supports) and 12 mm (middle span) were tested. Tension and compression stressstrain curves were required to complete the analytical study; therefore, tensile tests were
conducted using dumbbell-shape samples with average dimensions of 3.18 mm (thickness) × 3.5
mm (width) × 16mm (length). The average dimensions of the prismatic compression samples
were 3 mm × 3 mm × 8 mm. Each sample’s thickness and width were carefully measured prior
to testing. The strain field was determined using ARAMIS [35], a 3D digital image correlation
(DIC) system that enables non-contact measurement of displacement and strain fields. The 2D
surface strains in the sample were extracted using continuum mechanics principles. Using DIC,
we examined the heterogeneous behavior of the materials and the effects of stress concentrations
on premature failure. The tension, compression, and flexural behavior of epoxy resin Epon E 863
[26], PR520, and PRI subject to 60°C and 90% Rh were obtained. Sample sizes of 20 or greater
are advised to obtain results that can be considered representative of the statistical analysis [5,9].
After testing, the surfaces of epoxy resins were examined by scanning electron microscopy
(SEM) to investigate fracture behavior. Figure 2 (a,b) shows smooth fracture surfaces of
PRI2002 and E863, which are typical of a glassy material. The surface is flat with some quasistraight, quasi-parallel lines that indicate a relatively brittle failure. Figure 2 (b) shows unstable

failure with less crack propagation interruption after initiation in the tension test and voids in the
cured E863.
3. Estimation of Weibull Parameters
Because we observed a quite large dispersion of the flexural strength, a reasonable hypothesis is
that a statistical model can describe this dispersion. The Weibull distribution model was analyzed
to assess its capacity to represent the modulus of rupture stress (σMOR) data and to estimate the
ratio of mean flexural strength to mean tensile strength. The subscript MOR denotes the flexural
strength. Therefore, the purpose of this part of the study was to determine the scale parameter
and Weibull modulus, m, of the Weibull distribution associated with the flexural strength of the
tested materials. Weibull statistics link the failure probability, Pf, to σMOR by means of the
modified Weibull model [8,9,36].
(1)

m
 σ


Pf =1 − exp −  MOR  Ve 
  σ0 




where σMOR is the flexural strength (rupture stress), σ0 is a normalizing factor, m is the Weibull
modulus that characterizes the scattering in the measured values of strength, and Ve is the
volume of the material subjected to a uniaxial tension that would have the same probability of
failure as the sample. The Weibull modulus is a statistical value without any physical meaning.
However, it could be linked to the flaw size distribution; a higher m normally leads to a steeper
function and a lower dispersion. σ0 is related to the mean flexural strength of the distribution.
Monte Carlo simulation was used to compare the different estimation methods [5,8,9], and the
results revealed that the method of linear regression with an appropriately selected estimator and
at least 20 samples provides a reasonable estimate for the Weibull model. However, in this paper,
linear regression and the method of moments were used for a more comprehensive comparison
between the statistical and analytical methods.
3.1 Method of Moments
In this method, the first and second moments of a set of data are assumed to be the mean value
and standard deviation of the whole distribution (i.e., and infinite number of samples),
respectively [5]. The coefficient of variation (Equation 2) for the Weibull distribution is only a
function of m, which was solved for using the Newton-Rhapson method.

 =





 









where Γ is the gamma function.
3.2 Method of Linear Regression

(2)

Linear regression is a special case of the least squares method. This method is based on the
minimization of the squares of deviation of the data from the fitting function. Taking the natural
logarithm of equation (1) twice gives the linear equation
 1
ln ln 
1 − Pf


 σ 
 = m ln(σ MOR ) − m ln 10/ m 
 Ve



(3)

In this method, the slope of the Weibull model determines the Weibull modulus, and the yintercept is equal to –mln(σ0/Ve1/m). As the true value of failure probability for each σMOR is not
known, we employed the most common estimators [9].
 =

.


(a),  =





(b),  =

.

 .

where i is the rank of the measured strength.

(c),  =

.

 .

(d)

(4)

4. Statistical Approach
Table 1 shows the σMOR of all the bending samples ordered from the lowest to highest stresses
using the definition of Pi1. The σMOR of PR520 was not available for the statistical analysis due to
premature sample failure. The flexural load-deflection response of PR520 is described in the
“Analytical Simulation” section. The data for Epon E863 for all the estimators were plotted as
shown in Figure 3. Table 2 presents the least square equations and the Weibull modulus for Epon
E863. Very close agreement was observed between all the estimators. By comparing the method
of moments with the results from linear regression, it was clear that the definition of Pi1 yielded
the best results of all the estimators. Therefore, estimator Pi1 was used to analyze the rest of the
samples. Figure 4 illustrates the Weibull plot for Epon E863 and PRI2002. The linear plots in
Figure 4 are nearly parallel to each other, which indicates that the epoxies have a similar Weibull
shape parameter. The Weibull shape parameter does not vary much; therefore, it can be
considered to be independent of the type of epoxy resin. The mean flexural strength of the
Weibull distribution σmean is given by:
 =

σ0

Γ

! "

1+





(5)

The Weibull distribution parameters, m (Weibull modulus), and σmean (mean flexural strength)
for all samples are listed in table 3. The correlation coefficients R2 were above 0.85 and thus
considered acceptable. Closer examination of Figure 4 shows no bimodal distribution or
“extrinsic” defects related to poor production or human error, such as poor mixing or inadequate
mould filling. The “intrinsic” flaws are related to features within the material engendered by
uncontrollable aspects of production of the final product, such as segregation, small-scale
porosity due to the escape of bubbles within the polymer, or any centers that give rise to
incompatible deformation. These flaws are in variable sizes, shapes and orientations. In one
sample, the largest crack may be normal to the applied load, while in another sample it may be at
an angle to the applied stress. Obviously, the latter has a higher strength than the former;

therefore, variable defects with respect to the applied load can be accounted for the flexural
strength scatter . The morphological features developed during epoxy resin curing as shown in
Figure 2(b) can be considered as the cause of the scatter. The ratio of the mean flexural strength
to the mean strength in uniaxial tension based on the Weibull model is
σ MOR,3 PB
2
= 2(m + 1)
σ UTS

[

]

1/ m

(a);

&'(),+,&./0

= 1

2 3 4
 

5

/

(b)

(6)

Table 3 indicates that the above ratio based on the Weibull model was between 1.37 and 1.47.
The Weibull model is based on an empirical formula that relates the probability of failure to the
rupture stress. The number of specimens tested has a large influence on the calculated Weibull
modulus. The first limit in describing the fracture behavior of brittle and semi-brittle materials by
the Weibull model is due to the statistical nature of the model itself. This limitation leads to an
uncertainty in the parameters obtained from a limited number of experiments. The true values of
the Weibull parameters can only be obtained from an infinite number of samples. For any small
number, only an estimate, not the actual value, can be achieved. In addition, the Weibull model
is based on the “weakest-link hypothesis,” which means that the most serious flaw controls the
strength. However, there are quite complex flaw distributions in the material, which may affect
two superimposing Weibull distributions. It should be noted that σMOR used in the Weibull model
was obtained using the approximate linear stress distribution.
5. Analytical Simulation
The main output of flexural testing is the load-deflection curve, which does not provide much
direct information regarding the stress distribution or the constitutive law. The assumption of
linear elastic stress distribution also does not match with the nonlinear behavior of semi-brittle
materials. To overcome these problems, Yekani Fard et al. [28] developed a novel analytical
technique based on a piecewise-linear strain-softening tension and a compression stress-strain
model to simulate the flexural load-deflection response of different epoxy resin structural
systems. The developed model can be used for both forward and inverse analysis as shown in
Figure 5. This model is based on the general shapes of the stress-strain curves in tension and
compression in epoxy resin materials [26,37-40]. As shown in Figure 6 (a,b), all the samples
appeared to follow similar tension and compression stress-strain evolution. The stress-strain
curves in epoxy resins in Figure 6 (a) are highly nonlinear and exhibit the following distinctive
features (Figure 6b): linearly elastic, nonlinearly ascending, yield-like (peak) behavior, strain
softening, and nearly perfect plastic flow. The details of the model are described in our previous
publications [29, 30]. Flexural load was used instead of approximate linear flexural stress to
avoid flexural stress calculation errors. When a material with strain-softening behavior, such as
Epon E863, is loaded beyond the peak, the increase in deformation decreases the load-carrying
capacity. If no premature failure occurs for resins, the load-deflection curve exhibits a deflectionsoftening behavior in the post-peak part of the response. Figure 7 (a) illustrates the load
deflection curves of the “Avg.” and “Avg. – Std.” of E863 experiments. Figure 7(b) shows the

mean load-deflection curve of PRI2002 in the aged condition (48 weeks, 60°C and 90% Rh).
Figure 7(c) illustrates the load-deflection response of PR520 in the reference condition. The
stress concentration due to the right angle of the notch caused lower failure strength and
premature failure in PR520 samples. However, all of the simulations show that the direct use of
tension and compression stress-strain curves underestimated the flexural load-deflection
response, as expected. An inverse analysis of the load-deflection curves shows that the ratio of
mean flexural strength to the mean tensile strength for the epoxy resin polymers is between 1.18
(PRI2002) and 1.21 (E863 and PR520).
6. Conclusions
Mechanical tests were conducted on three structural epoxy resins under different environmental
and loading system conditions. The results indicate that the direct use of tension and
compression data leads to underestimation of flexural strength. The Weibull model was used to
describe the data scatter in the flexural strengths of Epon E863 and PRI 2002. The true value of
the Weibull parameter m and consequently the ratio of flexural strength to tensile strength C1
could only be obtained for an infinite number of specimens. For any smaller number, only an
estimate can be achieved. Our comparison of the different Weibull methods and estimators
revealed that there are no statistical advantages in using the moments method compared to the
simple linear least squares method. For thermoset epoxy resins with a Weibull modulus of about
13.8, the Weibull model predicts a mean flexural strength about 37% higher than the tensile
strength. However, analytical simulations of 3PB and 4PB demonstrated that this ratio is around
1.20 for all tested epoxy resin polymeric materials. Due to the intrinsic limitations of the
statistical models, the Weibull model can only estimate the ratio. It should be noted that the
flexural strengths provided in commercial data sheets of epoxies are usually much higher,
sometimes by as much as 50%, than the corresponding values from tensile tests. This
discrepancy can be explained in terms of the underlying assumptions of the Weibull model that
do not fit with the semi-brittle nature of thermoset epoxy resins and the unrealistic assumption of
linear stress distribution.
Acknowledgements
Financial support from Pipe Reconstruction, Inc. (research grant no. 13-105330) is gratefully
acknowledged. Appreciation is extended to Honeywell Aerospace for providing PR520 samples.
Epoxy resin PRI 2002 was kindly supplied by PRI, Inc.
References
[1] P.Huang, S. Zheng, J. Huang, Q. Guo. Miscibility and mechanical properties of epoxy
resins/polysulphone blend. Polymer 1997; 38: 5565.
[2] G. Di Pasculae, O. Motta, A. Recca, J.T. Carter, P.T McGrail, D. Acierno. New highperformance thermoplastic toughened epoxy thermosets. Polymer 1997; 38: 4345.

[3] C.B Bucknall, A.H Gilbert. Toughening tetrafunctional epoxy resins using polyetherimide.
Polymer 1989; 30: 213.
[4] E.M Woo, L.D Bravenec, J.C Seferis. Morphology and properties of an epoxy alloy system
containing thermoplastic and reactive rubber. Polymer Eng & Sci 1994; 34: 1664.
[5] C.C Riccardi, C. I Vallo. Estimation of Weibull parameters for the flexural strength of
PMMA-based bone cements. Polymer Eng & Sci 2002; 42: 1260.
[6] A. Jayatilaka, K. Trustrum. Statistical approach to brittle fracture. Journal of Material Science
1977; 12: 1426.
[7] K. Trustrum, A. Jayatilaka. Applicability of Weibull analysis for brittle materials. Journal of
Material Science 1983; 18: 2765.
[8] H. Peterlik. The validity of Weibull estimators. Journal of Material Science 1995; 30: 1972.
[9] A. Khalili, K. Kromp. Statistical properties of Weibull estimators. Journal of Material
Science 1991; 26: 6741.
[10] B. Bergman. On the estimation of the Weibull modulus. Journal of Material Science Letters
1984; 3: 689.
[11] W.V Harper, T.G Eschenbach, T.R James. Concerns about maximum likelihood estimation
for the three-parameter Weibull distribution: case study of statistical software. The American
Statistician 2014; 65 (1): 44.
[12] C. Mercadier, P. Soulier. Optimal rates of convergence in the Weibull model based on
kernel-type estimators. Statistics and Probability Letters 2012; 82: 548.
[13] M.I Giannotti, M.J Galante, P.A Oyanguren, C.I Vallo. Role of intrinsic flaws upon flexural
behavior of a thermoplastic modified epoxy resin. Polymer Test 2003; 22: 429.
[14] C.I Vallo. Influence of load type on flexural strength of a bone cement based on PMMA.
Polymer Test 2002; 21: 793.
[15] A. Nohales, L. Solar, I. Porcar, C.I Vallo, C.M Gomez. Morphology, flexural, and thermal
properties of sepiolite modified epoxy resins with different curing agents. European Polymer
Journal 2006; 42: 3093.
[16] N. Ben Salem, G. Bresson, J. Jumel, M.ER Shanahan, S. Bellut, F. Lavelle. Weibull
analysis of stiffness and strength in bulk epoxy adhesives reinforced with particles. Journal of
Adhesion Science and Technology 2013; 27: 2278.
[17] E.M Odom, D.F Adams. Specimen size effect during tensile testing of an unreinforced
polymer. Journal of Materials Science 1992; 27: 1767.

[18] J. PM Arias, M. Escobar, A.Vazquez. Modeling of dynamic mechanical properties of epoxy
and epoxy-phenolic reinforced with multi-wall carbon Nanotubes. Journal of Composite
Materials 2013; DOI: 10.1177/0021998313494096.
[19] M.M Rahman, M. Hosur, S. Zainuddin, N. Jahan, E.B Miller-Smith, S. Jeelani. Enhanced
tensile performance of epoxy and E-glass/epoxy composites by randomly-oriented aminofunctionalized MWCNTs at low contents. Journal of Composite Materials 2014; DOI:
10.1177/0021998314525977.
[20] G. Coroller, A. Lefeuvre, A. Le Duigou, A. Bourmaud, G. Ausias, T. Gaudry, C. Baley.
Effect of flax fibers individualization on tensile failure of flax/epoxy unidirectional composite.
Composites: Part A. 2013; 51: 62.
[21] G. Polizos, E. Tuncer, I. Sauers, K.L More. Physical properties of epoxy resin/titanium
dioxide nanocomposites. Polymer Eng & Sci 2011; DOI: 10.1002/pen.21783.
[22] T. Hobbiebrunken, B. Fiedler, M. Hojo, M. Tanaka. Experimental determination of the true
epoxy resin strength using micro-scaled specimens. Compos Part A 2007; 38: 814.
[23] J.N Goodier. Concentration of stress around spherical and cylindrical inclusions and flaws.
Trans Am Soc Mech Eng. 1933; 55: 39.
[24] B. Paul, L. Mirando. Stresses at the surface of a flat three-dimensional ellipsoidal cavity. J
Eng Mater Technol 1976; pp 164-172.
[25] E.Z Lajtai. Effect of tensile stress gradient on brittle fracture initiation. Int. J. Rock Mech.
Min. Sci. 1972; 9: 569.
[26] M. Yekani Fard, Y. Liu, A. Chattopadhyay. Characterization of epoxy resin including strain
rate effects using digital image correlation system. J. Aerosp. Eng. 2012; 25(2): 308.
[27] M. Yekani Fard, A. Chattopadhyay, Y. Liu. Multi-linear stress-strain and closed-form
moment curvature response of epoxy resin materials. International Journal of Mechanical
Sciences 2012; 57(1): 9.
[28] M. Yekani Fard, Y. Liu, A. Chattopadhyay. A simplified approach for flexural behavior of
epoxy resin materials. Journal of Strain Analysis for Engineering Design 2012; 47(1): 18.
[29] M. Yekani Fard, Y. Liu, A. Chattopadhyay. Analytical solution for flexural response of
epoxy resin materials.” J. Aerosp. Eng. 2012; 25(3): 395.
[30] M. Yekani Fard. Nonlinear inelastic mechanical behavior of epoxy resin polymeric
materials. Ph.D dissertation, 2011.
[31] ASTM D638-10. Standard test method for tensile properties of plastics. 2010.

[32] ASTM D695-10. Standard test method for compressive properties of rigid plastics. 2010.
[33] ASTM D790-10. Standard test method for flexural properties of unreinforced and reinforced
plastics and electrical insulations materials. 2010.
[34] ASTM D6272-10. Standard test method for flexural properties of unreinforced and
reinforced plastics and electrical insulating materials by four-point bending. 2010.
[35] Aramis. User’s manual for 3-D image photogrammetry, GOM mbH, mittelweg,
Braunschweig, Germany, 2006.
[36] W. Weibull. A statistical theory of the strength of materials. Ing Vetensk Akad Proc. 1939;
151: 1.
[37] C. G’Sell, A. Souahi. Influence of cross linking on the plastic behavior of amorphous
polymers at large strains. Journal of Engineering Material and Technology 1997; 119: 223.
[38] M.C Boyce,. E.M Arruda. An experimental and analytical investigation of the large strain
compressive and tensile response of glassy polymers. Polymer Engineering and Science 1990;
30: 1288.
[39] C.P Buckley, J. Harding. Deformation of thermosetting resins at impact rates of strain, part
I: Experimental study. Journal of mechanics and Physics of Solids 2001; 49(7): 1517.
[40] S. Behzadi, F. Jones. Yielding behavior of model epoxy matrices for fiber reinforced
composites: Effect of strain rate and temperature. J Macromol Sci Phys. 2005: 993.

Figure Captions
Figure 1. Stress gradient and nonlinear stress distribution (C1 is the ratio of the flexural strength
to the tensile strength).
Figure 2. SEM micrograph of fracture surfaces (a) PRI 2002 flexural sample; (b) voids present in
the cured Epon E 863 tension sample.
Figure 3. Weibull plot for Epon E863 with various estimators.
Figure 4. Weibull model for Epon E863 and PRI2002 using estimator P1.
Figure 5. Analytical model for the simulation of flexural load-deflection response in epoxy resin
polymeric materials.
Figure 6. (a) Tension and compression stress-strain curves of epoxy resin polymers under
different environmental conditions (the stress-strain curve for E863 is from Yekani Fard et al.
[26]); (b) general features of stress-strain curves (CYS: compressive yield strength, UTS:
ultimate tensile strength).
Figure 7. Simulation of the flexural load-deflection response for (a) 3PB Epon E863 in the
reference condition, (b) 4PB PRI2002 in the aged condition (48 weeks, 60°C and 90% Rh), and
(c) 3PB PR520 in the reference condition.

Table 1. Flexural strength data for the thermoset epoxy resins Epon E863 and PRI2002. “i” is the
sample rank, and Pi is the failure probability calculated from Equation 4(a).
i
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26

PiE863
0.01
0.029
0.048
0.067
0.087
0.107
0.125
0.144
0.163
0.183
0.202
0.221
0.24
0.26
0.279
0.298
0.317
0.337
0.356
0.375
0.394
0.413
0.432
0.452
0.471
0.49

PiPRI
0.019
0.058
0.096
0.135
0.173
0.212
0.25
0.288
0.327
0.365
0.404
0.442
0.481
0.519
0.558
0.596
0.635
0.673
0.712
0.75
0.788
0.827
0.865
0.904
0.942
0.981

σE863 (MPa)
86.5
93.5
94.4
99.5
100.2
100.4
100.5
100.9
101.6
102.0
102.5
103.5
103.8
103.9
104.8
105.5
105.6
107.1
107.5
107.8
108.0
108.9
109.1
109.7
109.8
109.9

σPRI (MPa)
67
68.5
69
69.5
70.1
70.6
71.3
71.8
72.4
73
74
76.8
77
77.3
77.5
78
79.1
79.6
81.2
81.5
83
84.2
85
86.4
88
91.5

i
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52

PiE863
0.51
0.529
0.548
0.567
0.587
0.606
0.625
0.644
0.663
0.683
0.702
0.721
0.74
0.76
0.779
0.798
0.817
0.837
0.856
0.875
0.894
0.913
0.933
0.952
0.971
0.99

σE863 (MPa)
110.3
110.4
110.5
111.1
111.4
111.5
111.9
113.3
113.7
114.5
115.3
116.3
116.6
117
117.7
118.1
118.6
119.4
119.8
120.0
120.1
120.7
120.8
121.3
121.4
121.8

Table 2. Linear regression analysis and method of moment results for Epon E863.
Estimator
Pi1
Pi2
Pi3
Pi4
Method of moment

Least square equation
16.49 ln(σMOR) – 78
15.35 ln(σMOR) – 72.7
15.97 ln(σMOR) – 75.6
16.40 ln(σMOR) – 77.6
N/A

Weibull modulus
16.49
15.35
15.97
16.40
16.72

Table 3. Weibull parameters for Epon E863 and PRI2002.
Material

Epon E863
PRI2002

MM
m

σmean
(MPa)

Storage
Condition

Number of
Samples

MLR
m

C1

Room temperature
(ref.)
60°C & 90% Rh for
48 weeks

52

16.49

1.47 16.72 1.46 109.7

26

13.7

1.38 13.8

C1

1.37 77.03