[Link]

ca

The Failure behavior of an epoxy resin

subject to multiaxial loading

Hu, Y.; Xia, Z.; EllyinF.

NRCC-48667

A version of this document is published in

/ Une version de ce document se trouve
dans: ASCE 2006 Pipeline Conference,
Chicago, Illinois, July 30, 2006, pp. 1-8
 The Failure Behavior of an Epoxy Resin Subject to Multiaxial Loading

Y. Hu1 *, Z. Xia2 and F. Ellyin3

1
Centre for Sustainable Infrastructure Research, Institute for Research in Construction, National
Research Council, Suite 301, 6 Research Drive, Regina, SK S4S 7J7; PH (306) 780-5432; FAX (780)
780-3421; email: [Link]@[Link]
2
Department of Mechanical Engineering, University of Alberta, Edmonton, AB T6G 2G8; PH (780)
492-3870; FAX (780) 492-2200; email: [Link]@[Link]
3
Department of Mechanical Engineering, University of Alberta, Edmonton, AB T6G 2G8; email:
[Link]@[Link]

Abstract

The failure behavior of an epoxy resin (EPON 826) was investigated by a series of tests on thin-walled
tubular specimens and the results are presented in this paper. The tests were designed such that a
constant octahedral shear strain rate was applied along different loading paths. The rate effect was also
studied by changing the octahedral shear strain rate. The experimental investigation showed that a
change in stress state influences the failure strengths and failure modes. The different failure strengths
and failure modes are attributed to the influence of the hydrostatic component of stress. Changing the
strain rate influenced the failure stress, but not the failure characteristics.

Introduction

One of the biggest challenges facing our cities today is the deterioration of underground piping
systems (e.g., water and sewer systems). In the past, excavation was used exclusively to replace
deteriorated water and sewer pipes. However, as cities became more congested, the disruption and
expense associated with excavation has become less acceptable. Therefore, polymer-based linings
were developed permitting pipe rehabilitation with little or no excavation. Of these, epoxy-based
cured-in-place pipe (CIPP) linings are becoming popular because of their high strength and toughness,
good corrosion resistance and low shrinkage during curing. Epoxy polymers are also widely used as
adhesives or matrices in other applications in a variety of industries, e.g., automotive, aerospace, oil
and gas, and marine industries. With the increasing use of such epoxy resins in critical-load-bearing
structures and with the increased application of limit states design methods in the design of these
structures, it is imperative to understand the failure behavior of the polymeric materials.

The stress state in many applications is complex and multiaxial because of the constraints imposed by
other components and reinforcement, in combination with the loading conditions (Asp et al. 1996). For
metals and alloys (elasto-plastic materials), yielding is governed by the shear stress component and the
effect of the hydrostatic stress component is negligible. For polymers, however, earlier studies (Ward

* This research was part of the researches toward a Ph.D. degree and conducted at the University of Alberta.

1
1971a, Anderson 1988, Asp et al. 1995, 1996, Kody and Lesser 1997, among others) have shown that
the effect of a multiaxial stress state on the failure behavior of polymers includes two aspects, failure
mode and failure stress. Ductile yield and brittle fracture are two failure modes common to most
polymers and a change in stress state may affect the failure behavior from ductile yield to brittle
fracture or vice versa. The influence of a multiaxial stress state on yield strength may be described by a
modified von Mises criterion or by a modified Tresca criterion (Ward 1971a, Raghava et al. 1973,
among others). Both modified failure criteria recognize that an appropriate failure criterion for many
polymers must consider the hydrostatic stress effects on failure behavior.

Although a large number of experiments have been done to verify these criteria and/or to determine
their constants, the design of the experiments has often been inadequate. In most experiments, multiple
sample geometries were used to arrive at the different stress states (see e.g., Ward 1971a, Asp et al.
1996). As pointed out by Kody and Lesser (1997), this approach severely limits the number of different
stress states that can be investigated and convolutes the results with effects of fabricating different
specimen geometries. In studies where single geometries were used (Broutman, et al. 1970, Sultan and
McGarry 1973, among others), the time effect and/or strain rate effect of the viscoelastic materials was
not fully considered. Kody and Lesser (1997) overcame these problems by maintaining a constant
octahedral shear strain rate, ε&oct , along the octahedral shear planes for all stress states and investigated
different stress states using one specimen geometry and test configuration.

This paper presents the results of an experimental investigation whose aim was to elucidate the yield
and fracture response of polymers and relate this response to the state of stress and strain rate. EPON
826/EPI-CURE 9551 was chosen for this purpose because it has excellent mechanical and thermal
properties and, therefore, is suitable for CIPP and other high performance applications. An attempt was
made to determine the multiaxial failure behavior of the EPON 826 epoxy system and the relative
importance of deviatoric and hydrostatic stress components in the failure behavior of the epoxy. Rate
effect was also studied by changing the octahedral shear strain rates.

Specimen Preparation and Test Facility

Specimen fabrication
EPON 826 and EPI-CURE Curing Agent 9551 are a bisphenol-A epoxy resin and a non-MDA
(methylene dianiline) polyamine system, respectively. EPON 826 and hardener were mixed with 2.3:1
(volume) or 100:36 (weight) ratio, stirred thoroughly and poured into a tubular mould. The samples
were then cured for 2 hours at 50°C with subsequent post cure for 2.5 hours at 120°C and cooled to
room temperature in the oven prior to machining.

Specimens were manufactured from the samples on a computer numerical controlled (CNC) lathe that
provides for precise machining of any transition profile geometry. The machined geometry is shown in
Fig.1. The specimen had a uniform gauge length of 20 mm. The outside and inside diameters in the
gauge length were 35.56 mm and 32.56 mm, respectively. The ratio of radius over wall thickness was
about 11.4 and, therefore, the specimen could be treated as a thin-walled tubular specimen. Aluminum
end tabs were glued to the tube ends so that the specimen could be inserted into the gripping system of
the test machine without damaging the specimen extremities during gripping process.

2
Test Facility
All experiments were conducted in a servo-hydraulic,
computer-controlled, multiaxial testing machine (see
Ellyin and Wolodko (1997) for detailed description). This
test facility could apply a combination of axial load,
differential pressure (inside and outside pressures) and
torsion on tubular specimens, and was capable of
generating principal stress and strain ratios in all
stress/strain quadrants, as well as change in the direction
of the applied principal stresses and strains. Stress/strain Figure 1. Thin-walled tubular
controlled, monotonic/cyclic, and in-phase/out-of-phase specimen geometry and end tabs
(non-proportional) tests under ambient temperature
conditions, could be performed with this system. The associated computer controlled hardware and
software allows for a number of complex user-defined loading and acquisition schemes.

Loading Scheme and Stress State

The loading scheme and stress state for tubular specimens are shown in Fig.2(a). Axial load, pressure
differential between internal and external pressures, and torsion could be applied to the specimen
simultaneously or in stages, or individually. This then induced axial, hoop and shear stresses within the
tube wall, Fig.2(b). (The radial stress component, σ r , was smaller by an order of magnitude and could,
therefore, be neglected in comparison to the other two principal stress components). The axial, hoop,
and shear stresses imposed on a thin-walled tube, subject to a combined axial load F, internal pressure
Pi , external pressure Pe , and torque T, could be calculated by
F (P − Pe )D
σa = + i (1)
πDt 4t
(P − Pe )D
σh = i (2)
2t
2T
τ= (3)
Dt
where D is the mean diameter and t is the thickness of the tube, σ a , σ h and τ are the axial, hoop and
shear stresses, respectively.

(a) (b)
Figure 2. Loading scheme and strain gauge layout (a) and stress state (b)

In order to compare the mechanical behavior of epoxy resins under different stress states, an equivalent
stress is often used. Here the octahedral shear stress, τ oct , was chosen which is defined as:

3
 2
τ oct =
σ a2 − σ aσ h + σ h2 + 3τ 2 (4)
3
Another important parameter for polymers is the hydrostatic stress, σ m , which is given by:

σ m = (σ a + σ h )
1
(5)
3

Strain Gauge Layout

A pair of strain gauges, one in the axial direction and another in the circumferential direction, served as
an axial and hoop strain transducers (Fig.2a). Two additional pairs of strain gauges mounted at ±45° to
the axial direction were used for shear strain measurement.

CEA-13-062UW-350 strain gauges from Micro-Measurements Co., were used for measuring the
strains. These gauges were attached to the tubular specimens with the M Bond, AE-10/15 adhesive
system. The strain ranges for the axial and shear strains were all calibrated to ±5% strain with a
resolution of 0.0005%.

Test Procedure

The failure behavior of this epoxy resin was studied by a loading scheme shown in Fig.3. Four
different loading paths, labeled O1, O2, O3 and O4, were chosen in the principal strain space. These 4
paths correspond to uniaxial tensile and compressive ( ε a / ε h = −υ ), equi-biaxial tensile ( ε a / ε h = 1 )
and pure shear stress ( ε a / ε h = −1 ) states , respectively. Due to material isotropy, the stress values at
three other points, 1', 3' and 4' were assumed to be the same but with different signs (material
symmetry). No data were obtained in the third quadrant because of the occurrence of buckling in the
range of applied strains. A constant octahedral shear strain rate,
εh
ε&oct , was imposed along all loading paths of Fig. 3. The 1'
octahedral shear strain rate, ε&oct , is defined as: 3 2

υ
ε&oct =
2
ε& a2 − ε& a ε& h + ε& h2 + (ε&a + ε&h )2 + 3 γ& (6)
4
εa
3 (1 − υ ) 2
4
O

where ε&a , ε&h and γ& are the axial, hoop and shear strain rates,
1

respectively and υ is Poisson’s ratio. 3'

4'
One specimen was utilized for each stress path at a given
octahedral shear strain rate. The rate effect was investigated by Figure 3. Loading paths in
loading specimens up to failure at three different octahedral principal strain space
shear strain rates of 10 −3 s −1 , 10 −4 s −1 , and 10 −5 s −1 .

All tests were performed at ambient environments (25°C and 50% humidity).

4
Test Result

Uniaxial tensile
Figure 4 shows the tensile stress – strain curves up to failure at three different octahedral shear strain
rates of 10 −3 s −1 , 10 −4 s −1 and 10 −5 s −1 . All these three stress – strain curves have similar patterns to
failure with a highly nonlinear response prior to failure. The failure mode is a brittle fracture (Ward
1971b). An increase in the strain rate results in an increase of the failure stress, from 72.3 MPa to 81.0
MPa, however, the failure strains decrease (c.f. Fig.4).

Uniaxial compressive
Figure 5 depicts the uniaxial compressive stress-strain behavior of this epoxy at three different strain
rates. These tests were conducted at a constant displacement rate control using a solid cylinder, as
shown in the insert of the figure, due to the buckling of the tubular specimens. The diameter and height
of the solid cylinders were 12.7 mm and 25.4 mm, respectively (in conformity with ASTM D 695-96).
Compared with the tensile failure curves depicted in Fig. 4, the compressive tests have higher failure
stresses and strains. Another difference is the failure mode: for the compressive tests, the material
manifested a ductile failure (Ward 1971b), with peak stresses at approximately 6.25% strain.
100

80
80
Axial Compressive Stress σ (MPa)

Ductile Yielding
Axial Stress (MPa)

60
60
Brittle Fracture

12.7
40 40
Uniaxial Compressive
-3 -1
-3
εoct = 10 s
-1
ε = 10 s
-4 -1 25.4 -4
ε = 10 s
-1
εoct = 10 s
20 20
-5 -1 -5 -1
εoct = 10 s ε = 10 s

0
0
0 1 2 3 4 5 0.0 2.5 5.0 7.5 10.0 12.5

Axial Strain (%) Axial Compressive Strain e (%)

Figure 4. Tensile stress-strain curves to failure Figure 5. Compressive stress-strain curves to

at three different octahedral shear strain failure at three different octahedral shear
rates: 10 −3 s −1 , 10 −4 s −1 and 10 −5 s −1 strain rates: 10 −3 s −1 , 10 −4 s −1 and 10 −5 s −1

Pure shear
The pure shear stress-strain curves to failure at three different octahedral shear strain rates are shown
in Fig.6. These curves have maximum stresses prior to their final failures, thus indicating ductile yield
behavior. It is noted that the failure stress increases with the increasing strain rate.

Equibiaxial tensile
The equibiaxial stress-strain curves to failure at three different octahedral shear strain rates are shown
in Fig.7. It is observed that these typical biaxial stress-strain curves have some common features as the
uniaxial curves shown in Fig.4. For example, linear behavior can be observed at low stress levels in
both axial and hoop stress-strain curves. With the increased strain, the rate of increase in stress

5
decreases in both loading directions. The axial and hoop stress-strain curves were slightly different
possibly due to the asymmetry of the specimen’s geometry. The equibiaxial tests also failed in a brittle
manner and the failure stresses increased with the increase of the in applied strain rate.
50 80

40
60

Axial & Hoop Stress (MPa)

Ductile Yielding
Shear Stress τ (MPa)

Brittle Fracture
30

40
εa : εh = 1 : 1
20 -3 -1
-3
εoct = 10 s
-1
εoct = 10 s
-4 -1
Axial Hoop
εoct = 10 s
-4 -1
-5 -1
20 εoct = 10 s
εoct = 10 s
10 Axial Hoop
-5 -1
εoct = 10 s
Axial Hoop
0 0
0 1 2 3 4 5 6 7 8 9 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Shear Strain γ (%) Axial & Hoop Strain (%)

Figure 6. Shear stress-shear strain curves to Figure 7. Equibiaxial stress-strain curves to

failure at three different octahedral shear strain failure at three different octahedral shear
rates: 10 −3 s −1 , 10 −4 s −1 and 10 −5 s −1 strain rates: 10 −3 s −1 , 10 −4 s −1 and 10 −5 s −1

Hydrostatic stress effect

The octahedral shear stress versus octahedral shear strain curve is used to compare different loading
paths (Fig.8). For metals and alloys (elastic-plastic materials), these curves for different proportional
loading paths tend to collapse to one curve (Ellyin 1997). Figure 8 shows four octahedral shear stress –
octahedral shear strain curves for the uniaxial tensile ( σ a > 0 and σ h = 0 ), uniaxial compressive
( σ a < 0 and σ h = 0 ), equibiaxial tensile ( σ a = σ h > 0 ), and pure shear ( σ a = −σ h > 0 ) loading
paths, respectively. It is obvious that, for the epoxy resin, the curves are generally different from one
loading path to another, with the uniaxial compressive curve the highest and the equibiaxial tensile
curve being the lowest one. In between are the uniaxial tensile and pure shear curves with the uniaxial
tensile curve slightly higher than that of the pure shear. The difference can be attributed to the
influence of the hydrostatic stress.

For the uniaxial tensile test, σ a > 0 , consequently σ m > 0 , while for the uniaxial compressive test,
σ a < 0 and hence σ m < 0 ; as for the pure shear test, σ a = −σ h , therefore, σ m = 0 . Since both σ a > 0
and σ h > 0 for equibiaxial tensile test, the corresponding σ m value should be the highest among the
four stress states. Therefore, a higher hydrostatic component of stress corresponds to the lowest stress
response and a lower hydrostatic component of stress corresponds to a higher stress response. The
relationship between the failure stress and the hydrostatic stress was described by a modified von
Mises or modified Tresca equation (Kody and Lesser 1997, Asp et al., 1996):

τ yoct = τ yoct0 − μσ m (7)

6
where τ y0
oct
is the failure shear stress in the absence of any overall hydrostatic stress and μ is commonly
referred to as the coefficient of internal friction. If σ m has no effect on τ oct
f , then μ is equal to zero, the

failure is said to follow a typical von Mises or Tresca behavior.

The quantitative relationship between the failure stress and the hydrostatic stress is shown in Fig.9. In
this figure, the experimental failure data are compared with the modified Tresca criterion (eq. (7))
(only the Tresca failure envelope for 10 −4 s −1 is depicted in this figure), here,

τ yoct =
1
(σ 1 − σ 3 ) (σ 1 〉σ 2 〉σ 3 ) (8)
2

It is observed that the modified Tresca criterion fits the experimental data fairly well. For comparison,
a Tresca envelope is also included in the figure with dotted lines. No experimental data was obtained
for biaxial compressive failure stress in the third quadrant due to buckling of the tubular specimens.
Those data, if available, will further check the aforementioned criterion.

90
σh
-3 -1
50 εoct = 10 s
-4 -1
εoct = 10 s 60
-5 -1
εoct = 10 s
40
Octahedral Shear Stress τoct (MPa)

Tresca 30
Modified Tresca

30 0
σa
-120 -90 -60 -30 0 30 60 90
-30
20
-4 -1
εoct = 10 s
Equibiaxial
Uniaxial Tensile -60
10 Pure Shear
Uniaxial Compressive
-90
0
0 1 2 3 4 5
-120
Octahedral Shear Strain εoct (%)

Figure 8. Octahedral shear stress versus octahedral Figure 9. Comparison of experimental

shear strain curve for different loading paths at data with the modified Tresca criterion
a fixed strain-rate

Comparison of Figs 5-8 shows that the failure modes are different: the failure mode for uniaxial tensile
and equibiaxial tensile tests is brittle fracture while pure shear and uniaxial compressive tests show
ductile yielding behavior. Since the uniaxial tensile and equibiaxial tensile tests have positive
hydrostatic stress and the pure shear and uniaxial compressive tests have zero or negative hydrostatic
stress, a high positive hydrostatic component of stress tends to produce to a brittle fracture failure
while ductile yielding prevails when the hydrostatic stress is zero or negative.

Conclusions

7
Specimens were loaded to failure under specified octahedral shear strain rates to investigate the failure
behavior of the EPON 826/EPI-CURE 9551 epoxy system subject to multiaxial stress state. From the
results presented in this experimental investigation, the failure behavior was found to be dependent on
the stress state as manifested by different octahedral shear stress -- octahedral shear strain curves. The
failure modes were also different depending on the stress state. The different failure curves and failure
modes are attributed to the influence of the hydrostatic component of stress. A high positive
hydrostatic component of stress tends to produce to a brittle fracture failure while ductile yielding
prevails when the hydrostatic stress is zero or negative. A modified Tresca criterion can described the
failure stress envelope quite well when the hydrostatic component of stress is included in the
modification criterion. Changing strain rate only influences the failure stress, but not the failure
characteristics.

Acknowledgment

The work presented here is part of a general investigation of the behavior of polymeric composites
under complex loading and adverse environment. This research was supported by the NOVA/NSERC
Senior Industrial Research Chair Program (F.E.) and NSERC research grants to F.E. and Z.X. The
authors are also indebted to Dr. J. Wolodko for his assistance in the software modification and valuable
suggestions.

References

Anderson, T.L. (1991). Fracture Mechanics: Fundamentals and Applications, CRC Press.
Asp, L.E., Berglund, L.A., and Gudmundson, P. (1995). “Effects of a Composite-like Stress State on
the Fracture of Epoxies.” Composites Science and Technology, 53, 27-37.
Asp, L.E., Berglund, L.A., and Talreja, R. (1996). “A Criterion for Crack Initiation in Glassy Polymers
Subjected to a Composite-like Stress State.” Composites Science and Technology, 56, 1291-1301.
Broutman, L.J., Krishnakumar, S.M., and Mallick, P.K. (1970). “Combined Stress Failure Tests for a
Glassy Plastic.” Journal of Applied Polymer Science, 14, 1477-1489.
Ellyin, F. (1997). Fatigue Damage, Crack Growth and Life Prediction, Chapman & Hall, London.
Ellyin, F., and Wolodko, J.D. (1997). “Testing Facilities for Multiaxial Loading of Tubular
Specimens.” The American Society for Testing and Materials, Philadelphia, PA, ASTM STP 1280, 7-24.
Kody, R.S., and Lesser, A.J. (1997). “Deformation and Yield of Epoxy Networks in Constrained States
of Stress.” Journal of Materials Science, 32, 5637-5643.
Stassi-D'Alia, F. (1969). “Limiting Conditions of Yielding for Anisotropic Materials.” Meccanica, 4,
349-364.
Sultan, J.N., and McGarry, F.J. (1973). “Effect of Rubber Particle Size on Deformation Mechanisms in
Glassy Epoxy.” Polymer Engineering and Science, 13, 29-34.
Ward, I.M. (1971a). “Review: The Yield Behaviour of Polymers.” Journal of Materials Science, 6,
1397-1417.
Ward, I.M. (1971b). Mechanical Properties of Solid Polymers, Wiley-Interscience, New York.