Original Article

J Strain Analysis
2015, Vol. 50(1) 4–14
Ó IMechE 2014
Modified creep constitutive equation Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
for an epoxy-based adhesive with DOI: 10.1177/0309324714554965
sdj.sagepub.com

nonlinear viscoelastic behavior

Mohammad Zehsaz, Farid Vakili-Tahami and

Mohammad-Ali Saeimi-Sadigh

Abstract
This article presents the procedure to obtain constitutive equations to express the creep behavior of Araldite 2015
epoxy adhesive using the results of the uniaxial creep tests. The experimental data show that the examined adhesive has
a nonlinear viscoelastic behavior. Three types of constitutive equations have been used: (a) Bailey–Norton equation, (b)
generalized time-hardening model and (c) rheological model which is a series combination of springs and dampers
known as Maxwell and Zener combination. The first two models have vast application in commercial finite element–
based software. It is shown that the generalized time-hardening model can simulate the creep behavior of the adhesive
better than the Bailey–Norton model. However, this model is less accurate at the elevated temperatures. Therefore, an
empirical equation based on Maxwell and Zener’s was proposed which demonstrates a very good consistency with the
results of the experimental data over the assumed range of stress and temperatures.

Keywords
Adhesive, creep, nonlinear viscoelastic material

Date received: 8 May 2014; accepted: 2 September 2014

Introduction model the nonlinear behavior of adhesives. Many vis-

coelastic and viscoplastic models have been defined to
Adhesive joints are being used widely in engineering perform stress analysis of adhesively bonded structures.
industries. This is because of increasing demand to The simplest viscoelastic models are the classical models
design lightweight structures such as aircrafts and vehi- of Voigt and Maxwell.3 Other improved viscoelastic
cle body frames. In addition, the ability of adhesives to and viscoplastic models are available in the literature.4,5
join dissimilar materials makes them popular in joining Feng et al.6 established a method to study the long-term
different components. To design and for proper use of creep behavior of epoxy adhesives based on series of
adhesively bonding, knowledge of the stress and strain short-term accelerated tests. They investigated the tem-
distribution in the adhesive joint is required. Accurate perature and moisture effects by means of mechanical
predictions are obtained by means of models that pre- response and they used a coupling model to analyze
cisely describe the deformation behavior of adhesives. creep behavior in epoxies. An important attribute of
The behavior of adhesives, like most polymers, is time- this model is its ability to physically describe the charac-
dependent. They exhibit viscoelastic and viscoplastic
teristics of the molecular mobility by means of coupling
behavior when subjected to relatively high temperature
parameters. Dean7 modeled the uniaxial tension creep
and high stress levels. Thus, close attention needs to be
behavior of polypropylene using stretched exponential
paid to the creep behavior of the adhesive joints in
structural applications to increase durability and relia-
bility of them.1 Department of Mechanical Engineering, The University of Tabriz, Tabriz,
Experiments reveal that deformation behavior of Iran
epoxies is approximately linear below moderate stress
Corresponding author:
levels; however, with increasing time under load, the Mohammad-Ali Saeimi-Sadigh, Department of Mechanical Engineering,
stiffness of material decreases and it exhibits nonlinear The University of Tabriz, 29 Bahman Str., No2, 5166-16471 Tabriz, Iran.
behavior.2 Thus, extensive studies were carried out to Email: saeimi.sadigh@tabrizu.ac.ir

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015
Zehsaz et al. 5

function with four elements. He modeled the nonlinear effects of time and temperature was evaluated. Choi et
behavior by means of the parameters called relaxation al.20 studied the rheological modeling and finite element
time. Yu et al.8 presented a complete range of exponen- simulation of epoxy adhesive creep in fiber-reinforced
tial data for a typical adhesive system including con- polymers (FRPs) in reinforced concrete beam. They
stant strain rate, creep and recovery in tension and have shown that significant creep occurs at the concrete
compression. They used Bailey–Norton model, unified FRPs. Hamed and Chang21 studied the effect of creep
theory and viscoelastic models to model the creep on the edge de-bonding failure of FRP-strengthened
response of epoxy adhesives. Majda and Skrodzewicz9 reinforced beams and showed that depending on the
have developed a modified creep model of epoxy in material properties, creep can increase or decrease the
ambient temperature. They modified Burgers mechani- maximum load that leads to edge de-bonding.
cal model which is composed of Maxwell and Kelvin– Jhin et al.22 have studied the relationship between
Voight models to model nonlinear behavior of the crack path and test method by comparing the perfor-
epoxy adhesives. Yu et al.10 derived a model-based mance of individual adhesive–adherend combinations
empirical method to determine properties like time- in quasi-static fracture, mixed-mode fatigue to the
dependent creep compliance function of viscoelastic threshold and creep crack growth. In mixed-mode
adhesive contact models. They used the parcel swarm loading, crack paths in creep were similar to those in
optimization method to find solution to the nonlinear quasi-static fracture, even though the crack growth
equations of this model. Roseley et al.11 have studied rates in creep were much smaller.
the creep response of three epoxy-based thixotropic Most of the research work in this subject is focused
adhesives that are less heavily cross linked and have on the mechanical behavior of the adhesives. There
transition temperature values between 30 °C and 60 °C. are also studies in the literature which model the
They first used Kelvin–Voight equation to model creep adhesively boned joints using finite element tech-
strain, but this model was not successful in modeling niques. Most of these studies assume that the adhe-
the behavior in temperature range above transition tem- sive behaves linear elastic, so they do not address the
perature. Therefore, they improved the model over the viscoelastic behavior of the adhesive. However, as it
complete range of temperature using a series of two- has been shown in Jhin et al.,22 this effect cannot be
element Kelvin–Voight models. Although these models disregarded in the analysis.
provide appropriate fit to the experimental data, due to Since most of the researches that have been carried
their complex form, it is difficult to obtain the multi- out recently study the creep behavior of epoxy materials
axial form of them. Since the actual behavior of the at ambient temperature, this article intends to investi-
adhesively bonded structures is multi-axial, it is reason- gate the creep behavior of adhesive at higher tempera-
able to use simple models that can be extended to tures. This research presents experimental results of the
multi-axial form. uniaxial creep tests on the Adhesive epoxy adhesive.
Various analytical and finite element techniques This adhesive shows nonlinear viscoelastic behavior
have been used to study viscoelastic behavior of adhe- and does not obey the Boltzmann’s superposition prin-
sively bonded single lap joints.12–16 In addition, a num- ciple. The aim is to propose a creep constitutive model
ber of articles deal with creep constitutive modeling of that is able to simulate the creep characteristics of this
the bulk adhesive. Pandey and Narasimhan17 have ana- adhesive at different stress and temperature levels. With
lyzed the adhesively bonded joints by modeling the the methodology presented in this article, both primary
adhesive layer with elasto-viscoplastic behavior and the and secondary creep strains of the adhesive are pre-
adherend as linearly elastic material. They used a non- dicted using a developed constitutive equation.
linear three-dimensional (3D) model to carry out finite
element analysis. Su and Mackie1 have performed a
two-dimensional creep analysis to investigate the stress Materials and creep tests
distribution in the adhesive layer over time. They have
used multi-linear model to predict the creep behavior In this research, the uniaxial creep test has been carried
of the adhesive based on the experimental data. out on the Adhesive (Huntsman Advanced Materials)
Khalili et al.18 have studied the effect of fiber rein- which is a two-component epoxy adhesives. Table 1
forcing the adhesive on the creep behavior of single lap shows the mechanical properties and curing conditions
joints. They experimentally showed that at a tempera- of the adhesive provided by its manufacturer. In addi-
ture above the glass transition temperature of the adhe- tion, in order to obtain the tensile properties of the
sive, the failure time and initial strain for all specimens adhesive, bulk specimen has been manufactured based
were elevated compared to un-reinforced adhesive on ASTM D 638-10 standards. Tensile tests were con-
joints. Ferrier et al.19 investigated the creep behavior of ducted on bulk specimens in a Zwick tensile test
epoxy adhesives, used to strengthen steel-reinforced machine with 1000 N capacity and 0.01 N accuracy
concrete structures. The durability and performance of load cell. The specimens were tested at 5 mm/min cross-
these reinforcements, which were dependent on the head displacement rate. Contact extensometers with
rheological behavior, were investigated, and permissible accuracy of 61 mm were used to record specimen
shear stress of the interface taking into account the elongation.

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015
6 Journal of Strain Analysis 50(1)

Figure 1 shows the tensile diagram of this adhesive. changed according to the Amsler creep test machine
According to this figure, Young’s modulus is 1.91 GPa grips (Figure 2). Then this plate is located on the top of
and tensile yield strength and tensile failure strength are another plate as shown in this figure to form a mold for
equal to 12.5 and 22.09 MPa, respectively. Comparing producing adhesive samples. The adhesive was poured
these results and the values given in Table 1 shows that into the mold and cured in the furnace at 60 °C for
adhesive specimen that has been provided for this 35 min. The advantage of this method is that the resi-
research meets the manufacturer’s specifications. dual stresses resulting from machining process do not
In order to conduct uniaxial tensile creep tests, in exist in the specimen.
most researches the bulk specimen was prepared using The uniaxial creep tests have been carried out using
computer numerical control (CNC) machining.9 This Amsler testing machine (Figure 3). The force is applied
operation causes residual stresses in the specimen. In through the weights located on the arm of the machine.
order to eliminate the effect of residual stresses, a differ- To obtain more accuracy in controlling the applied
ent approach is used in this study; first, the shape of the load on the specimen, a load cell device is used which
specimen is formed in a 3-mm silicon sheet using laser measures the applied load with accuracy of 0.1 kg. The
cutting machine. The specimen was prepared based on specimen elongation measurements have been carried
ASTM D 638-10. However, the tab of the specimen was out using contact extensometers and with accuracy of
0.5 mm. An interface system is used to record all
Table 1. Mechanical properties and curing temperature given data including applied load, displacement and tempera-
by adhesive manufacturer. ture and save them in a computer file every 1 s. A
programmable logic controller (PLC) system is
Araldite 2015 Properties employed to maintain a constant temperature within
the furnace with accuracy of 60.5 °C. The specimens
1.85 ± 0.21 Young’s modulus, E (GPa)
0.33 Poisson’s ratio, n were tested at 25 °C, 40 °C and 55 °C. Table 1 presents
12.63 ± 0.61 Tensile yield strength, sy (MPa) the variation in lap shear strength of the Araldite 2015
21.63 ± 1.61 Tensile failure strength, sf (MPa) with temperature which is provided by the manufac-
14.6 ± 1.3 Shear yield strength at 25 °C, t y (MPa) turer of the adhesive. Based on the information which
21 Lap shear strength at 25 °C, t f (MPa) has been provided for the adhesive in technical data
19 Lap shear strength at 40 °C, t f (MPa)
15.5 Lap shear strength at 55 °C, t f (MPa) sheet, at stress levels below 60% of lap shear strength,
60/35 Curing temperature/time (min/°C) adhesive shows elastic behavior. In addition, in this
87 Glass transition temperature (°C) article, we have conducted uniaxial tensile test at room
temperature to justify the above-mentioned assertion.
This assumption is justified by noting that at 25 °C the
lap shear strength is 21 MPa and yield strength is
12.63 MPa (see Table 1) which is about 60% of lap
shear strength or 11.55 MPa. However, we have con-
ducted our tests at 55% of lap shear strength to make
sure no initial plasticity occurs during creep tests.
Figure 4(a) shows the scanning electron microscope
(SEM) micrographs of the fracture surfaces of the bulk
specimen after creep test. After mixing two reactive
parts, air entrapment during mixing and/or application
process often produces micro-voids in the adhesive
bond-lines and influences the strength of the bonded
joints. The presence of these vacancies and nucleation
of voids at elevated temperatures are the reasons of
Figure 1. Stress–strain diagram of the adhesive bulk specimen. higher creep straining and lower creep life. In this

Figure 2. (a) Schematic of the silicon mold prepared by laser cutting machine and (b) creep sample dimensions.

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015
Zehsaz et al. 7

linear, but the effective modulus decreases with time so

that the stress–strain curves at different times are
separated from one another. However, the curves in
Figure 5 show the nonlinear viscoelastic behavior of
adhesive, and therefore, nonlinear formulations should
be used to model the creep behavior of the material.
This can be achieved using certain combination and
layout springs and dampers, which will be explained in
the following.

Creep constitutive equations

The application of more general models, which are
capable to explain more accurately the creep behavior
of different materials, is favored by researchers. Since
these models can be used in industrial applications such
as 3D problems, some of these models are defined in
commercial software for modeling the creep phenom-
Figure 3. Uniaxial creep test machine: (a) sample and fixtures ena. These equations are capable of modeling the pri-
and (b) testing machine. mary creep and secondary creep as well as combined
primary + secondary creeps of various materials.
Many of these equations are obtained through generali-
research, two-part epoxy has been mixed using dual- zations of the time-hardening and strain-hardening
cartridge and static-mixer to produce a specimen with equations. For example, ANSYS24 provides 13 models
smaller and lesser number of micro voids.23 As can be to explain the creep behavior of the materials.
seen in Figure 4, the majority of the micro-voids in the In this article, the creep models available in ANSYS
fracture surface were between 20 and 120 mm in size software are used to model the creep behavior of the
which have been highlighted using dashed and solid adhesive. In the first step, the famous Bailey–Norton
lines in Figure 4(b) and (c), respectively. equation (equation (2)), which is available in most of
the finite element software and is widely used for pri-
mary creep modeling, is employed
Experiment results and modeling
ecr = C1 sC2 tC3 + 1
Uniaxial creep tests on bulk specimens of adhesive at ð2Þ
temperatures of 25 °C, 40 °C and 55 °C are performed. C3 + 1
The elongation of specimens is recorded and the true In this equation, s is the stress, t is the time and C1–C3
strain of specimen is calculated. The maximum applied are constitutive parameters. In order to verify the ade-
stress on the specimens is lower than the yield stress; quacy of the Bailey–Norton model for the creep strain
hence, the total strain is sum of elastic and creep strains. characteristic of the adhesive, the constitutive para-
Creep strain is calculated using equation (1) meters (C1–C3) have been calculated at different tem-
peratures and stress levels, using nonlinear regression.
ec (t) = etotal (t)  ee ð1Þ
Table 2 shows the values of C1–C3 for temperature lev-
where ec is the creep strain, ee is the initial elastic strain els of 25 °C, 40 °C and 50 °C. Also, the values of R2
and etotal is the total strain. Both creep strain and total (explained variance) are presented to show the quality
strain are time (t) dependent in this equation. As of regression. The value of R2 close to unity testifies
observed, in the uniaxial creep test of the adhesive, the good fitting of the model to the experimental data.
higher straining and lower creep life of the adhesive Figure 6(a)–(c) compares the creep strain obtained by
occur with increasing temperature. experiments with those which are calculated using
Some polymeric materials demonstrate nonlinear Bailey–Norton model.
variation in creep strain with respect to stress. This can As it can be seen from Figure 6 and the values of R2,
be observed using isochronous creep curves as pre- Bailey–Norton model is not well fitted to the experimental
sented in Figure 5 for the creep tests at 25 °C. This fig- data. Therefore, generalized time-hardening model with
ure shows that the examined adhesive is a medium five constitutive parameters (D1–D5) is used to simulate
nonlinearly viscoelastic material and does not obey the the creep behavior of the adhesive (equation (3))
Boltzmann’s linear superposition principle,3 in the ecr = (D1 s + D2 s2 + D3 s3 )t(D4 + D5 s) ð3Þ
range of loads which were used in the investigations.
Therefore, the isochronous diagram of these materials Table 3 shows the constitutive parameters obtained
deviates from linearity at a certain stress level. For lin- for generalized time-hardening model. The calculated
ear viscoelastic materials, the isochronous curves are values of R2 are close to unity as shown in Table 3

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015
8 Journal of Strain Analysis 50(1)

Figure 4. Fracture surface examination of specimen after creep test at 3105, 3505 and 33000 magnifications.

Table 2. Parameters of Bailey–Norton model.

Norton power law C1 C2 C3 R2

Temperature 25 °C 1.4646e28 3.7451 20.5 0.986

Temperature 40 °C 1.1660e28 4.1840 20.51 0.974
Temperature 55 °C 5.4175e207 3.4718 20.6250 0.975

which demonstrate the acceptable quality of the regres- As it can be concluded from the results depicted in
sion. Figure 7(a)–(c) compares the creep strain obtained Figures 6 and 7 and also the values of R2 tabulated in
by experiments with those which are calculated using Tables 2 and 3, the generalized time-hardening model
generalized time-hardening model. has the better consistency of the computational

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015
Zehsaz et al. 9

Table 3. Parameters of generalized power law model.

Generalized time hardening D1 D2 D3 D4 D5 R2

Temperature 25 °C 21.1906e24 3.4867e25 21.9460e26 0.035 0.035 0.994

Temperature 40 °C 24.3120e24 1.1284e24 26.3443e26 0.035 0.035 0.987
Temperature 55 °C 21.037e23 3.5789e24 22.5963e25 0.035 0.035 0.975

separately for each stress level. Comparing the values

obtained for each parameter at different stress level,
they have observed a major effect of the stress level on
the parameters. So they proposed a modified model in
which the material parameters of the model are func-
tions of stress.
In this study, a similar method is used to model the
creep behavior of Adhesive. For this purpose, the
model shown in Figure 8, which is a series combination
of two Zener and Maxwell models, is implemented.
This model was proposed based on the combination of
spring and damper elements. Such model is particularly
attractive, similarly as other mechanical models in the
Figure 5. Isochronous stress–creep strain curves for Araldite
analysis of creep phenomena, because its parameters
2015 adhesive at 25 °C.
can be identified with the characteristic periods of
creep. In this model, the constant E1 as the modulus of
longitudinal elasticity at the initial moment of loading
characteristics with the experimental characteristics
can be identified with the instantaneous strain resulting
achieved in different stress and temperature levels than
from the Hook’s law, E2 and EN are the moduli of elas-
Bailey–Norton model. However, the fitting accuracy
ticity of relaxation response and h1 and h2 are the coef-
decreases with increasing temperature. Therefore, an
ficients of dynamic viscosity.
empirical model based on combination of the Maxwell
The compliance was obtained using the usual solu-
and Kelvin–Voight models was proposed to simulate
tion method for solving ordinary linear differential
adhesive creep behavior in the elevated temperatures.
equations with constant coefficients. Therefore, the
adhesive strain (etotal(t)) could be determined as a func-
tion of time (t) and average tensile stress (s0)
Rheological modeling of the material
The spring and damper configuration can be used to etotal (t) =
 
model the short time elastic or un-relaxed response 1 t 2 E0  E‘ ut E0  E‘ ut
s0 + +  e 1 e 2
of the material. It can also predict long-term E1 h1 E‘ E0 E0
response of the material with the limited deforma- ð4Þ
tion in fully relaxed state.7 Short-term tests on
epoxies show that their time-dependent deformation where E0 = EN + E2; u1 = h2 E0 =½E‘ (E0  E‘ ) and
is almost linear at low and moderate stress levels. u2 = h3 E0 =½E‘ (E0  E‘ )
However, in long-term periods, the stiffness of mate- In this equation, E0, E1, h1, h2 and h3 are constitu-
rial decreases with time. The rate of change in stiff- tive parameters. Using nonlinear regression on the
ness increases with stress which reflects the nonlinear creep test data, these constants can be calculated. As
behavior of the material.7 reported in the literature,9,19 for an adhesive with non-
The viscoelastic behavior of the polymers can be linear viscoelastic behavior, these constants are non-
modeled quantitatively using appropriate number/com- linear function of temperature and stress which leads to
bination of the elastic and damping elements. This can more complex rheological model and increase in the
be achieved by employing several Maxwell elements in number of constants in the equation. To overcome this
parallel or Kelvin–Voight elements in series. Majda problem and to decrease the number of unknown para-
and Skrodzewicz9 performed a creep test in ambient meters, equation (5) can be written as
temperature for an epoxy-based adhesive under four  
l
stress levels. They have proposed a model composed of etotal (t) = ee + j + at  (eg1 t + eg2 t ) ð5Þ
2
Maxwell and Kelvin–Voight models connected in series
in order to model nonlinear viscoelastic behavior. This where ee = s0 =E1 , j = 2s0 =E‘ , a = s0 =h1 , l=2 = s0
model is known as Burgers model. They have calcu- ((E0  E‘ )=E0 ), g 1 = 1=  u1 and g2 = 1=  u2 . In this
lated three parameters of constitutive equation equation, ee is the elastic strain; j, a, l, g 1 and g 2 are

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015
10 Journal of Strain Analysis 50(1)

0.05
Stress=6.95 MPa
Experimental: Stress=8.91 MPa
0.04 Stress=11.64 MPa
Stress=6.95 MPa
Calculated: Stress=8.91 MPa
Stress=11.64 MPa
Creep Strain

0.03

0.02

0.01

0
(a) 0 2 4 6 8 10 12 14 16 18
Time (sec) x 104
0.06
Stress=7 MPa
Experimental: Stress=8.18 MPa
0.05 Stress=10.5 MPa
Stress=7 MPa
Calculated: Stress=8.18 MPa
0.04
Stress=10.5 MPa
Creep Strain

0.03

0.02

0.01

0
(b) 0 2 4 6 8 10 12 14 16 18
Time (sce) 4
0.035 x 10

0.03

0.025
Creep Strain

0.02

0.015 Stress=6.13 MPa

Experimental: Stress=7 MPa
0.01 Stress=8.5 MPa
Stress=6.13 MPa
Calculated: Stress=7 MPa
0.005
Stress=8.5 MPa
0
(c) 0 2 4 6 8 10 12
Time (sec) 4
x 10

Figure 6. Comparison of experimentally determined creep characteristics with those calculated using the Norton power law
model for temperatures (a) 25 °C, (b) 40 °C and (c) 55 °C.

constitutive material constants. In order to obtain of constants and simplify the rheological model. In
material constants first, a nonlinear regression was this equation, ee is the elastic strain; a and b are first-
used separately for each stress level. Table 4 shows the and second-order functions of stress, respectively
results of regression for each stress level at 25 °C. The h i
^ ^
results show that l1 = 2l2 and j has very small value etotal (t) = ee + at^  b(ea5 t + ea5 t )
which can be neglected. Thus, a formulation in the ð6Þ
= eelastic + at^+ b sinh (a6 t^)
form of equation (6) is proposed to reduce the number

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015
Zehsaz et al. 11

0.05
Sress=6.95 MPa
Experimental: Stress=8.95 MPa
0.04 Stress=11.64 MPa
Sress=6.95 MPa
Creep Strain Calculated: Stress=8.95 MPa
0.03 Stress=11.64 MPa

0.02

0.01

0
(a) 0 2 4 6 8 10 12 14 16 18
Time (sec) 4
x 10
0.06
Stress=7 MPa
Experimental: Stress=8.18 MPa
0.05 Stress=10.5 MPa
Stress=7 MPa
Calculated: Stress=8.18 MPa
0.04
Creep Strain

Stress=10.5 MPa

0.03

0.02

0.01

0
(b) 0 2 4 6 8 10 12 14 16 18
Time (sec) 4
x 10
0.035
Stress=6.13 MPa
Experimental: Stress=7 MPa
0.03 Stress=8.5 MPa
Stress=6.13 MPa
Calculated: Stress=7 MPa
0.025 Stress=8.5 MPa
Creep Strain

0.02

0.015

0.01

0.005

0
(c) 0 2 4 6 8 10 12
Time (sec) 4
x 10

Figure 7. Comparison of experimentally determined creep characteristics with those calculated using the generalized time-
hardening model for temperatures (a) 25 °C, (b) 40 °C and (c) 55 °C.

Table 4. The results of regression for each stress level and temperature 25 °C.

Parameters j a l g1 g2 R2

s = 6.95 MPa 1.0430e28 4.7631e24 1.699e24 0.927 20.927 0.99

s = 8.91 MPa 2.0360e29 25.3710e23 0.0276 0.3986 20.3986 0.98
s = 11.64 MPa 2.5302e28 24.548e22 0.3294 0.2592 20.2592 0.99

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015
12 Journal of Strain Analysis 50(1)

a = a1 + a2 s
b = a3 + a4 s 0 + a5 s 2
t^= log 10 (t)

The parameters ai are material constants and define

the material behavior. These parameters are obtained
using nonlinear regression method based on the experi-
Figure 8. Rheological model composed of one Maxwell and mental data. Table 5 shows the values of the parameters
two Zener models. obtained for 25 °C, 40 °C and 55 °C. Also, this table

0.05
Stress=6.95 MPa
Experimental: Stress=8.91 MPa
0.04 Stress=11.64 MPa
Stress=6.95 MPa
Calculated: Stress=8.91 MPa
Creep Strain

0.03 Stress=11.64 MPa

0.02

0.01

(a) 0
0 2 4 6 8 10 12 14 16 18
Time (sec) 4
x 10
0.06
Stress=7 MPa
Experimental: Stress=8.18 MPa
0.05 Stress=10.5 MPa
Stress=7 MPa
Calculated: Stress=8.18 MPa
0.04
Creep Strain

Stress=10.5 MPa

0.03

0.02

0.01

(b) 00 2 4 6 8 10 12 14 16 18
Time (sec) 4
x 10
0.035

0.03

0.025
Creep strain

0.02

0.015 Stress=6.13 MPa

Experimental: Stress=7 MPa
Stress=8.5 MPa
0.01
Stress=6.13 MPa
Calculated: Stress=7 MPa
0.005 Stress=8.5 MPa

(c) 0
0 2 4 6 8 10 12
Time (sec) 4
x 10

Figure 9. Comparison of experimental data for creep strain with those calculated using the proposed rheological model (equation
(5)), for temperatures (a) 25 °C, (b) 40 °C and (c) 55 °C.

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015
Zehsaz et al. 13

Table 5. Coefficients of rheological model.

a1 a2 a3 a4 a5 a6 R2

Temperature 25 °C 9.4801e24 29.6501e25 21.0340e24 22.4210e26 2.9240e26 1.17 0.999

Temperature 40 °C 5.6501e23 28.6610e24 21.9420e23 2.6604e25 4.7305e25 0.75 0.989
Temperature 55 °C 3.4307e22 27.8303e23 20.2060 0.0510 21.5201e23 0.27 0.991

gives the value of R2 for each regression. It can be seen Funding

that R2 value is close to 1, which reveals the good accu- This research received no specific grant from any fund-
racy of regression. ing agency in the public, commercial or not-for-profit
Figure 9(a)–(c) compares the creep strain values sectors.
obtained in experimental test with those calculated
using nonlinear regression at different stress and tem- References
perature levels. It can be seen that there is a good agree- 1. Su N and Mackie RI. Two-dimensional creep analysis of
ment between experimental data and calculated values structural adhesive joints. Int J Adhes Adhes 1993; 13:
using the proposed rheological model. Comparing the 33–40.
results of R2 tabulated in Tables 2–5 and Figures 6–8, it 2. Dean G, Crocker L, Read B, et al. Prediction of defor-
can be seen that the proposed rheological model can mation and failure of rubber-toughened adhesive joints.
simulate the experimental data over the range of stress Int J Adhes Adhes 2004; 24: 295–306.
and temperature levels more accurately. 3. Brinson HL. Polymer engineering science and viscoelasti-
city. New York: Springer, 2008.
4. Ward IM and Hadley DW. An introduction to the
mechanical properties of solid polymers. New York:
Conclusion
Wiley, 1993.
In this article, uniaxial creep tests were carried out on 5. Lubliner J. Plasticity theory. Berkeley, CA: University of
Adhesive epoxy-based adhesive in three different stress California, Berkeley, 1990.
and temperature levels. The experimental data show 6. Feng CW, Keong CW, Hsueh YP, et al. Modeling of
that the examined adhesive has a nonlinear viscoelastic long-term creep behavior of structural epoxy adhesives.
Int J Adhes Adhes 2005; 25: 427–436.
behavior and does not obey the Boltzmann’s superposi-
7. Dean G. Modelling non-linear creep behaviour of an
tion principle. Also, using these data, constitutive para- epoxy adhesive. Int J Adhes Adhes 2007; 27: 636–646.
meters for three different types of models have been 8. Yu XX, Crocombe AD and Richardson G. Material
obtained employing Levenberg–Marquardt method. modeling for rate-dependent adhesives. Int J Adhes
First, two forms of creep constitutive equations, Adhes 2001; 21: 197–210.
Bailey–Norton and generalized time hardening, have 9. Majda P and Skrodzewicz J. A modified creep model of
been considered which are used in commercial finite epoxy adhesive at ambient temperature. Int J Adhes
element software. The results revealed that creep para- Adhes 2009; 29: 396–404.
meters should be expressed as a function of stress to 10. Yu H, Li Z and Wang QJ. Viscoelastic-adhesive contact
modeling: application to the characterization of the
increase the degrees of fitting. Therefore, generalized
viscoelastic behavior of materials. Mech Mater 2013; 60:
time-hardening model can predict the creep behavior of
55–65.
the adhesive more accurately than the Bailey–Norton 11. Roseley ASM, Rojo E, Ansell MP, et al. Creep response
model. However, it was observed that the accuracy of of thixotropic ambient temperature cure adhesives mea-
this model decreases at higher temperatures, and there- sured by DMTA in static tension and shear. Int J Adhes
fore, an empirical model was proposed. In this step, a Adhes 2011; 31: 575–582.
rheological model based on the combination of spring 12. Chiu WK and Jones R. Unified constitutive model for
and damper was proposed to simulate the creep beha- thermoset adhesive, FM73. Int J Adhes Adhes 1995; 15:
vior of the adhesive. Original form of this model con- 131–136.
sists of five parameters which are nonlinear functions 13. Duncan BC and Maxwell AS. Measurement methods for
time dependent properties of flexible adhesives. NPL report
of stress and temperature. In order to reduce the num-
no CMMT(A)178, April 1999. Middlesex: NPL.
ber of parameters, an improved form of the model 14. Dean GD and Broughton WR. A review of creep modeling
using sine-hyperbolic equation was proposed with three for toughened adhesives and thermoplastics. NPL report
parameters. This model was well fitted to the experi- DEPC-MPR 011, January 2005. Middlesex: NPL.
mental data in the range of different stress and tem- 15. Pandey PC, Shankaragouda H and Singh AK. Nonlinear
perature levels. analysis of adhesively bonded lap joints considering visco-
plasticity in adhesives. Comput Struct 1999; 70: 387–413.
16. Mortensen F and Thomsen OT. Analysis of adhesive
Declaration of conflicting interests bonded joints a unified approach. Compos Sci Technol
The authors declare that there is no conflict of interest. 2002; 62: 1011–1031.

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015
14 Journal of Strain Analysis 50(1)

17. Pandey PC and Narasimhan S. Three-dimensional non- 21. Hamed E and Chang Z. Effect of creep on the edge de-
linear analysis of adhesively bonded lap joints consider- bonding failure of FRP strengthened RC beams—a theo-
ing viscoplasticity in adhesives. Comput Struct 2001; 79: retical and experimental study. Compos Sci Technol 2013;
769–783. 74: 186–193.
18. Khalili SMR, Jafarkarimi MH and Abdollahi MA. Creep 22. Jhin G, Azari S, Ameli A, et al. Crack growth rate and
analysis of fibre reinforced adhesives in single lap joints— crack path in adhesively bonded joints: comparison of
experimental study. Int J Adhes Adhes 2009; 29: 656–661. creep, fatigue and fracture. Int J Adhes Adhes 2013; 46:
19. Ferrier E, Michel L, Jurkiewiez B, et al. Creep behavior 74–84.
of adhesives used for external FRP strengthening of RC 23. Katnam KB, Stevenson JPJ, Stanley WF, et al. Tensile
structures. Constr Build Mater 2011; 25: 461–467. strength of two-part epoxy paste adhesives: influence of
20. Choi K, Mahmoud M and Taha R. Rheological model- mixing technique and micro-void formation. Int J Adhes
ing and finite element simulation of epoxy adhesive creep Adhes 2011; 31: 666–673.
in FRP-strengthened RC beams. J Adhes Sci Technol 24. ANSYS use’s manual version 12. Canonsburg, PA:
2013; 27: 523–535. ANSYS, Inc., 2012.

Downloaded from sdj.sagepub.com at COLORADO STATE UNIV LIBRARIES on July 20, 2015