Polymer Testing 46 (2015) 134e143

Contents lists available at ScienceDirect

Polymer Testing
journal homepage: www.elsevier.com/locate/polytest

Creep loading response and complete loadingeunloading

investigation of industrial anti-vibration systems
Robert Keqi Luo a, b, Xiaoping Wu b, c, *, Limin Peng b
a
Trelleborg Industrial Anti-Vibration Systems, Leicester LE4 2BN, UK
b
Department of Railway Engineering, School of Civil Engineering, Central South University, Changsha, Hunan 410075, China
c
Centre for Transport Studies, University College London, London WC1E 6BT, UK

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

Article history: This article presents engineering approaches to evaluate creep loading response and a complete loading
Received 17 June 2015 eunloading procedure for rubber components used as anti-vibration applications. A damage function for
Accepted 27 July 2015 creep loading and a rebound resilience function for mechanical unloading are introduced into hypere-
Available online 29 July 2015
lastic models independently. Hence, a hyperelastic model can be extended for both creep and unloading
evaluations. A typical rubber product and a dumbbell specimen were selected to validate the proposed
Keywords:
approaches. It has been demonstrated that the predictions offered by the new models are consistent with
Creep
the experimental data. In addition, a loading procedure using the same final value, with and without
Rebound energy
Hyperelasticity
involving unloading, prior to a creep test can produce different results. The proposed approach can
Rubber unloading capture this phenomenon which was observed in the literature. The proposed approach can also be easily
Design incorporated into commercial finite element software (e.g., Abaqus). It is demonstrated that the proposed
method may be used for anti-vibration products at an appropriate design stage.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction Creep of rubbers easily exceeds structural limitations, and fracture

Rubber springs are widely used in industry as anti-vibration
components and typically provide many years of service. When a
constant stress is imposed on a rubber spring, the resulting defor-
mation increases continuously with time; this is known as creep,
drift [1] or strain relaxation [2].
The creep of vulcanized rubbers is usually classified into two
processes: primary creep, which is due to the physical rearrange-
ment of molecular positions in the final stages of the approach to
equilibrium; and secondary creep, which is chiefly due to oxi-
dization. Primary creep dominates at the beginning of the loading
time, while secondary creep becomes evident and eventually
dominates after a significant length of time (e.g. in years) has
elapsed [1,3]. Sometimes, a third or tertiary stage is referred to as
creep failure occurring at an increasing rate and terminating in
fracture [4]. It is known that the relatively poor creep resistance of
rubbers is unfavourable to their application as structural materials.
* Corresponding author. Centre for Transport Studies, University College London,
London WC1E 6BT, UK.
E-mail addresses: xpwu@csu.edu.cn, xpwucsu@foxmail.com (X. Wu).
due to creep often occurs due to long-term loading. Because long-
term creep tests are very expensive and time-consuming, many
accelerated methods for creep tests on polymers have been
developed to predict long-term creep behaviour based on short-
term experiments. Based on the fact that higher stresses increase
the creep or relaxation rate of viscoelastic materials, which is
similar to the effect of higher temperatures, several time-
etemperatureestress superposition principles (TTSSP) have been
proposed using Boltzmann's superposition principle and its modi-
fied forms; these principles have predicted creep for longer periods
based on shorter-period tests [5e8]. Similarly, Starkova et al. [9]
applied timeestress superposition (TSS) to construct smooth
master curves by considering the nonlinearity of viscoelastic
behaviour and by introducing a stress reduction function into an
exponential creep kernel. Gupta and Raghavan [10] and Achereiner
et al. [11] used the time-temperature-superposition principle
(TTSP) to obtain master creep curves for a time period beyond the
experimental time window. In Kolarik and Pegoretti [12,13], a
concept was adopted so that the non-linearity of tensile creep is
primarily caused by the strain-induced increment of the free vol-
ume. Consequently, the traditional stressestrain linearity range can
be viewed as an artificial limit related to the limited accuracy of

http://dx.doi.org/10.1016/j.polymertesting.2015.07.009
0142-9418/© 2015 Elsevier Ltd. All rights reserved.
 R.K. Luo et al. / Polymer Testing 46 (2015) 134e143
measurements at low stresses and strains. The internal time-tensile
compliance superposition of non-linear creep data was applied to
construct creep master curves, which corresponded to a pseudo-
iso-free-volume state; a Boltzmann-like superposition principle
for multistep nonlinear tensile creep was used with three types of
polypropylene. For alternative methods to the Boltzmann super-
position principle, statistical analyses have been performed for
creep prediction. Gnip et al. [14] built long-term creep curves using
the data obtained in direct experiments with a duration of 122e183
and 1100e1432 days, and demonstrated that creep compliance
could be approximated by power and exponential regression
equations.
A number of factors affecting polymer creep have been inves-
tigated in previous studies. To ensure that the equipment settings
were correct for the creep experiment, some improvements were
made. Nitta and Maeda [15], for example, designed a creep appa-
ratus in which the external load was controlled by the transient
data of the cross-sectional area of the deformed specimen to make
sure that a constant stress level was maintained. Tomlins et al. [16]
obtained tensile creep data for vinyl chloride at a number of tem-
peratures ranging from 21.5 to 58
C. It was shown that neither the
ageing rate nor the shape of the distribution of retardation times
was affected by temperature. Further studies (e.g., Dean [17]) found
that errors in long-term properties derived from the application of
TTS could arise unless changes in creep behaviour caused by
physical ageing were considered. Printer et al. [18] found that times
to failure increase continuously with the age of the solutions on the
degradation of the surface's active environment. Creep failure could
also be predicted based on a thermally activated rate process where
the creep rate reaches its minimum value [19]. The effect of tem-
perature may also be important and is probably different for sec-
ondary creep. Additionally, oxidation is probably diffusion-
controlled at certain temperatures. Attempts to study long-term
creep behaviour by accelerated tests on small samples require,
therefore, careful study to avoid false conclusions [3]. In parallel
with above methods mentioned, viscoelasticity models were also
developed to predict polymer creep. The theoretical background of
the framework of viscoelasticity for polymer creep behaviour has
been well documented [4,20e22].
Mechanical loadingeunloading processes strongly affect rub-
ber's creep behaviour. Drozdov [23] demonstrated that the loading
process significantly influences rubber creep due to unloading. In
his experiments, specimens were subjected to stretching up to
various maximum strains followed by retraction to various smaller
strains, creep tests were then performed and it was found that the
creep curves were strongly affected by the unloading increments.
Additionally, Khan and Yeakle [24] found that a loading history
prior to the creep test consisting of loading to a maximum stress or
strain followed by partial unloading to a target stress could
significantly modify strainetime behaviour, which was consistent
with the findings from Omans and Nagode [25]. It was established
that unloading the specimen between two different stress levels
had a substantial influence on creep results.
For mechanical loadings on anti-vibration components, rubbers
are widely modelled as hyperelastic materials based on strain en-
ergy density [22,26,27] (e.g., the Mooney-Rivlin, Ogden and Yeoh
models). These models are generally used to predict the loading
response and fatigue life of rubber products used in industry (Luo
et al. [28e34] and Zerrin and Fatemi [35]). However, these models
can only be used for the loading portion of a loadingeunloading
cycle. To include unloading effects on the hyperelastic models, a
single additional (i.e., softening) variable was added to model a
complete loadingeunloading process (Ogden and Roxburgh [36,37]
and Luo et al. [38e40]). Additionally, two more variables were used
in the energy function so that it captured the observed softening
135
and residual strain response (Dorfmann and Ogden [41] and Luo
et al. [42,43]). Despite the progress described above, current
hyperelastic models cannot be used for creep response because
these models are not referenced to the elapsed loading time.
Rubber creep as a time dependent deformation is one of the
critical factors when considering engineering design and applica-
tions for rubber springs. An important requirement is to control a
rubber product to not exceed its structural limitation and to avoid
early failure due to creep over its required service life. In industry,
reliable prediction of creep for rubber structures over their entire
service lives is essential at the design stage. For many anti-vibration
applications, rubber components must be checked at specified time
periods for safety issues; for example, rubber suspension systems
used in rail vehicles must be inspected and adjusted during their
operating service life.
In this study, we introduce a damage concept into hyperelastic
models to predict creep for rubber materials. In addition, a dissi-
pating function is added to capture the observed stress softening
response for unloading during a cyclic loading process. Conse-
quently, hyperelastic models based on strain energy potential can
be extended to predict rubber unloading and creep performance.
The remainder of this article is outlined as follows. Experimental
results for a rubber spring and a concept for rebound energy as well
as its measurement are presented in Section 2. The hyperelastic
models including unloading and creep effects with necessary
equations are contained in Section 3. Next, simulation and valida-
tion results are demonstrated in Section 4. Finally, findings from
this investigation are summarized in Section 5.
2. Experiment on rubber anti-vibration product
In engineering design and applications, rubber structures are
usually loaded under compression and shear conditions. A typical
industrial product (Metacone type spring) was selected for exper-
iment and validation. Two experiments (i.e., mechanical loading
and creep loading) were performed. The rubber compound used
was a filled synthetic high cis polyisoprene with a shear modulus
1.2 MPa.
2.1. Mechanical loading experiment
Mechanical loading was performed before each creep experi-
ment began to ensure that the product was compliant with real
engineering applications and had the correct material properties.
This component, which measured maximum 230 mm in diameter
Fig. 1. A Metacone spring.
136 R.K. Luo et al. / Polymer Testing 46 (2015) 134e143
and 140 mm in height, is used in industrial mounts, as shown in
Fig. 1. Three layers of rubber and two metal interleaves are used in
this part. Compression loading in the vertical direction was applied
to the component. In this procedure, a force of approximately 58 kN
was loaded along the vertical direction within a specified working
compression range at a loading rate of 10 mm per minute. Fig. 2 (a)
shows the load deflection response of this rubber component for
this loadingeunloading procedure. Different paths between
loading and unloading were observed. A reduction in the load
during unloading (i.e., stress softening) was observed at a given
deflection between the loading and unloading conditions.
2.2. Creep experiment
A creep experiment was performed in a controlled laboratory.
The ambient temperature was maintained at 27 ± 1
C. An
approximately constant load of 46 kN along the vertical direction
was applied to the rubber spring. The creep experiment lasted 553
days, and the creep result was recorded into a database. The creep-
time history is shown in Fig. 2(b). In the first few days, the gradient
of the creep curve was maximum in the loading response. After that
period, the slope of the creep curve changed gradually to a lower
value as time progressed.
Fig. 2. Experimental results for the Metacone spring.
2.3. Measurement of rebound energy: rubber resilience test
Luo et al. [38e40] introduced rubber rebound energy (i.e., resil-
ience) as a single function in a strain energy density expression. The
rebound energy or rebound resilience, which is the ratio between
the rebound energy and the initial loading energy, is a property of
rubber materials and a parameter for a given rubber. It can be used to
predict characteristics of the material during unloading after cyclic
loadingeunloading processes. Measurement of the rebound resil-
ience (i.e., the resilience test) is standardised in BS ISO 4662:2009
[44]. The test method involves hitting a standard test piece with a
falling mass, which is free to rebound after impact. There are two
methods to determine the rebound resilience of rubber: the
pendulum method and the tripsometer method. In this study, the
tripsometer method was used, and a layout of the equipment is
shown in Fig. 3. In this method, a flat test piece was impacted with a
hemisphere. The hemisphere was mounted on the periphery of a
disc, which was supported on an axle and caused to rotate by an off-
axis mass. The kinetic energy of the impacting mass was measured
immediately before and after impact. The impact angle was set to
45
. The ratio of the rebound height to the drop height, which is
expressed as a percentage, is equal to the rebound resilience of the
tested rubber compound. More details can be found in Refs. [44,45].
 R.K. Luo et al. / Polymer Testing 46 (2015) 134e143 137

More details on the above equations, including how these

equations are obtained, can be found in Refs. [22,26,27]. Equation
(1) has been widely used to predict material response under the
loading portion of a loadingeunloading process. However, it is not
suitable for prediction of the response during the unloading
portion. Equation (1) must be modified to consider the unloading
process.

3.2. Rebound energy approach model for unloading prediction

The unloading effect on products is not considered when the

mechanical properties of rubber are represented in terms of a
general straineenergy function. A model is proposed to account for
the unloading condition based on a rebound energy concept. A
function q(b) is added to modify the first part (WI) of equation (1) to
account for the unloading condition:


W ¼ qðbÞWI I þ WJ ðJel Þ (5)

We define q(b) as the rebound resilience, which is the ratio of

the rebound energy to the initial loading energy or the ratio be-
tween the energy input after loading and the energy returned
during unloading. The rebound resilience, as described in 2.3, is a
function of the loading variable b, which is a nominal scale variable
for a given loadingeunloading process with an index a during
loading and unloading. The boundary conditions for q(b) are as
Fig. 3. Tripsometer. follows:

During loading: q(b) ¼ 1

3. Rubber material models
During unloading:
Start: b ¼ 0, q(0) ¼ 1
3.1. General hyperelastic models of rubber material
Finish: b ¼ 1, q(1) ¼ q0
There are several hyperelastic material models commonly used
q0 is an engineering constant for a given rubber material, which
to describe rubber and other elastomeric materials based on strain
represents the total rebound resilience and can be determined
energy potential or strain energy density (Ogden [22] and Bower
using the tripsometer rebound test or similar methods, as
[26]). These hyperelastic models for rubber materials can be
described in 2.3.
expressed in a general form:
The above description of q(b) is a general framework that allows

 for considerable flexibility in choosing a specific form of q(b) if the
W ¼ WI I þ WJ ðJel Þ (1)
above-mentioned boundary conditions are satisfied. We define a
formulation to include the change in the rebound resilience during
where WI ðIÞ is the deviatoric part of the strain energy density of the
unloading:
primary material response and WJ(Jel) is the volumetric part of the
strain energy density. qðbÞ ¼ 1  ð1  q0 Þba (6)
I can be further expanded into I 1 and I 2 , which are the first and
second deviatoric strain invariants, which are defined as: where a is an index variable that can be considered the rate of
2 2 2
energy change. When a ¼ 1, equation (6) becomes a linear model
I1 ¼ l1 þ l2 þ l3 (2) for the change in the rebound resilience:

2 2 2 qðbÞ ¼ 1  ð1  q0 Þb (7)
I 2 ¼ l1 þ l2 þ l3 ; (3)
From a mathematical perspective, the function of the proposed
where li are the deviatoric stretches and Jel is the elastic volume model is continuous at both loading and unloading points. From an
ratio. engineering perspective, the required parameters above are phys-
The corresponding stresses may be written as: ically meaningful concepts.
2
   
2 41 vW vW vW vW dij 3.3. Damage approach model for creep prediction
sij ¼ þ þ I 1 B ij  I 1 þ 2 I 2
Jel J 23 v I1 v I2 v I2 v I2 3
el To determine the creep behaviour of the rubber materials under
3
1 vW vW constant loading, an additional function is introduced into equation
 4 Bik Bkj 5 þ d (1). This function should relate the creep behaviour with both the
J 3 v I 2 v Jel ij
el material structure change under loading and the time elapsed.
(4) Hence, two variables are required. A time-decay function Rc with a
damage concept is proposed and incorporated into equation (1).
where Bij is the component of the left Cauchy-Green deformation The principle is to link the rubber creep to the material loading
tensor B, and dij is the Kronecker delta. condition and to the elapsed loading time. It is assumed that the
138 R.K. Luo et al. / Polymer Testing 46 (2015) 134e143

creep effect from loading on the material is related to invariant I 1 ,
I 2 and the elapsed loading time T. Therefore, a hyperelastic model
is not only related to the loading condition of the polymer material
through the invariant I 1 , I 2 but also to the elapsed time T:



W ¼ WI I þ WJ ðJel Þ þ Rc I; T (8)
loadingeunloading processes under the experimental conditions. A
polynomial form of the strain energy density function was selected
for demonstration. The proposed approach was incorporated into
the selected N ¼ 3 model.
Equation (1) without unloading capability can be expressed by
the following polynomial series:

where:

W ¼ WI I þ WJ ðJel Þ


 X
N
i
j XN
1
Rc I; T ¼ kT z I 1 þ I 2 (9) ¼ Cij I 1  3 I 2  3 þ ðJel  1Þ2i (10)
iþj¼1 i¼1
D i
where T is the time elapsed from the end of a mechanical loading,
T ¼ 0 when an initial loading ends, and k and z are parameters to be where Cij and Di are material constants for a given rubber
defined. compound.
From a mathematical perspective, the proposed function (9) is If N ¼ 3, equation (10) becomes:
continuous. From a scientific perspective, the above parameters



2


2
W ¼ C10 I 1  3 þ C01 I2  3 þ C20 I 1  3 þ C11 I1  3 I 2  3 þ C02 I 2  3

3
2


2
3
þC30 I 1  3 þ C21 I 1  3 I 2  3 þ C12 I 1  3 I 2  3 þ C03 I 2  3 (11)
1 1 1
þ ðJ  1Þ2 þ ðJel  1Þ4 þ ðJel  1Þ6
D1 el D2 D3

have physically meaningful concepts. k is an amplitude parameter After integrating the resilience approach model, equation (11)
that describes the strength of the creep damage, z can be consid- becomes:
ered to be a hardening parameter that describes how quickly creep

h


2


2
3
W ¼ ½1  ð1  q0 Þba  C10 I 1  3 þ C01 I 2  3 þ C20 I 1  3 þ C11 I 1  3 I 2  3 þ C02 I 2  3 þ C30 I 1  3

2


2
3 i
þC21 I 1  3 I 2  3 þ C12 I 1  3 I 2  3 þ C03 I 2  3 (12)
1 1 1
þ ðJ  1Þ2 þ ðJel  1Þ4 þ ðJel  1Þ6
D1 el D2 D3

damage develops with the elapsed time T, I 1 and I 2 are the
deformation state of rubber material. Thus, we can define k as an
amplitude indicator for creep damage and define z as a hardening
index for creep damage. The application of the proposed approach
is shown in the following section.
4. Simulation and validation
Mechanical and creep experiments were simulated based on the
constitutive models described above with the finite element soft-
ware Abaqus. Due to symmetry of the components' geometries and
the loading conditions, models for the rubber components can be
simplified to perform the simulation more efficiently.
We will investigate the mechanical loadingeunloading model
and the creep model in detail in the following sub-sections.
4.1. Mechanical loading simulation and validation
The proposed model based on the rebound energy concept
was used to simulate the response of the mechanical
The material data (Cij and Di), which are typically either deter-
mined by a company or externally contracted, were obtained from
the manufacturer's database, whose accuracy and suitability for
engineering design and applications have been verified. The ma-
terial coefficients may also be directly obtained by testing the real
products (Luo et al. [46]). Rebound resilience q0 was set equal to 0.7
from the tripsometer experiment.
For the Metacone rubber spring, an axisymmetric model was
generated due to the geometry and loading condition. The total
number of elements was approximately 660, and the total number
of degrees of freedom was approximately 1900. The rubber ele-
ments used a four-node bilinear axisymmetric quadrilateral and
hybrid elements with constant pressure. The type of metal ele-
ments used was a four-node bilinear axisymmetric quadrilateral
element. A full finite element model of the component is shown in
Fig. 4 (a). The main dimension of the cross-section of the compo-
nent is shown in Fig. 4(b). During mechanical loading, the
maximum deflection along the vertical direction was approxi-
mately 40 mm. Fig. 4(c) shows this deformed shape; for clarity, the
results of the entire component are displayed. The rubber parts of
 R.K. Luo et al. / Polymer Testing 46 (2015) 134e143
Fig. 4. Unloaded and loaded finite element model of the Metacone spring.
the product were compressed (i.e., height decreased) and bulged
upwards due to the deformation.
For the unloading portion of the loadingeunloading process, we
assume that q(b) varies linearly with unloading (i.e., index a ¼ 1), as
indicated in Equation (7). The complete load-deflection histories of
the simulations and experiments are compared in Fig. 5(a). The two
results are consistent, which implies that the model can predict the
loadingeunloading condition reasonably well, even when the
predicted forces are marginally above those observed in the
experiment for the beginning of the unloading. This result indicates
that the primary characteristics of the unloading are captured using
this approach, and further improvement may be possible. To obtain
the best possible simulation result, a ¼ 0.7 was arrived at by opti-
misation based on the experimental data. The entire load-
deflection histories between the simulations and experiments are
compared again in Fig. 5(b). The predicted unloading forces are
nearly identical to those observed in the experiments.
139
Fig. 5. Mechanical loadingeunloading comparisons between the simulation and
experimental data for the Metacone spring.
The load-deflection curve indicates that different stress values
are possible at the same deflection. As indicated by Equation (2),
different strain energy values are expected because of the change
in the function q(b), which leads to different stress levels. Hence,
the stress profiles are compared to verify this expectation. Fig. 6
shows two stress profiles with identical deflection values at
loading and unloading. The two stress patterns look notably
similar, but the value during unloading is approximately 20%
below that during loading. This result indicates “stress softening”
at the same deflection during unloading, which is consistent with
engineering principles and the observed load-deflection curve
(Fig. 2 (a)).
To verify the approach further, a type of simple standard
dumbbell specimen (BS ISO 37:2011) shown in Fig. 7 (a) was used.
The thickness of the specimen is 2 mm. The rubber compound is
synthetic high cis polyisoprene with shear modulus 0.7 MPa. The
rebound resilience measured using the tripsometer method was
80% A three-dimensional finite element model was generated to
simulate a loadingeunloading experiment. 1200 rubber elements
with 8-node linear brick, hybrid with constant pressure were used.
The total number of degrees of freedom was approximately 8400.
The finite element model is shown in Fig. 7 (b). We kept the same
parameter value a ¼ 0.7 as used for the Metacone rubber spring.
The total load-deflection responses between the simulation and
experiment are compared in Fig. 8. The predicted unloading forces
agree well with those observed in the experiment, which implies
the suitability of the parameter proposed in the model.
140 R.K. Luo et al. / Polymer Testing 46 (2015) 134e143

Fig. 8. Mechanical loadingeunloading comparisons between the simulation and

experimental data for the dumbbell specimen.

4.2. Creep loading simulation and validation

We utilize the polynomial series (N ¼ 3) in equation (11) and the

damage function presented in equation (9) to perform the creep
simulation:



2
W ¼ C10 I 1  3 þ C01 I2  3 þ C20 I 1  3



2
3
þC11 I 1  3 I 2  3 þ C02 I 2  3 þ C30 I 1  3

2

þC21 I 1  3 I 2  3


2
3
þC12 I 1  3 I 2  3 þ C03 I 2  3
1 1 1

þ ðJel  1Þ2 þ ðJel  1Þ4 þ ðJel  1Þ6 þ kT z I 1 þ I 2
Fig. 6. Stress profiles of the Metacone spring during loadingeunloading. D1 D2 D3
(13)

Fig. 7. Dumbbell specimen and finite element model. Fig. 9. Stress profiles of the Metacone spring during the creep process.
 R.K. Luo et al. / Polymer Testing 46 (2015) 134e143
Fig. 10. Creep comparisons between the simulation and experimental data for the
Metacone spring.
The material data Cij and Di are unchanged during the me-
chanical loading simulation. The two parameters of the damage
function for creep evaluation are k ¼ 0.01785 and damage hard-
ening parameter z ¼ 0.12, which were determined from the creep
experimental data. A Fortran code was used to transform equation
(13) into a user subroutine and integrated with Abaqus. The creep
simulation was performed in accordance with the laboratory
Fig. 11. Creep comparisons between the simulation and experimental data for the dumbbell specimen.
141
experiment. A constant force of approximately 46 kN was applied
and held for 553 days, as described in Section 2. We specified that
time ¼ 0 corresponded to the point where the initial mechanical
loading finished. Fig. 9 demonstrates stress profiles for the rubber
part of the Metacone spring at time ¼ 0 and time ¼ 553 days. The
patterns of the two stress profiles look similar, but the value at
time ¼ 553 days was approximately 14.5% above that at time ¼ 0.
These results indicate “stress hardening” during the creep process
due to “strain hardening” (i.e., further geometric deformation from
creep) in addition to the initial mechanical deformation, which is
consistent with engineering principles. The creep history for the
spring between the simulations and experiments is compared in
Fig. 10. It is shown that the predicted creep curve is near the
measured data from the experiments. The comparison has indi-
cated that the proposed approach yields very good predictions for
creep evaluation.
To investigate the parameters used in the model, we applied this
approach to the dumbbell specimens. Taking a reference on the
industrial products used in anti-vibration applications, a tensile
loading equivalent to stress level 1.5 MPa was applied to the
dumbbell specimen in a laboratory experiment. In the simulation
procedure, the amplitude parameter k was optimized as a value
of 0.00184 and a value of 0.306 was taken as the hardening
parameter z. Fig. 11(a) shows the creep comparison between the
simulation and the experimental data for the dumbbell specimen at
1.5 MPa stress state. The creep history from the simulation agrees
142 R.K. Luo et al. / Polymer Testing 46 (2015) 134e143
well with the experimental data. To test the parameter in an
extreme loading condition, a tensile loading equivalent to stress
level 3.0 MPa was applied to the dumbbell specimen. k ¼ 0.01785
and z ¼ 0.12 were assigned to the model. The creep simulation
result was compared with the experimental data, as shown in
Fig. 11 (b). The comparison demonstrated good agreement between
the two results except for some deviations at the very beginning of
the test period.
The patterns of the creep curves from the simulations for the
dumbbell specimen are very similar to that from the Metacone
spring. Furthermore, the overall creep behaviour predicted was
qualitatively consistent with the experimental data using dumbbell
specimens in literature (see Refs. [5,11,13,25]). These comparisons
indicate the flexibility of the parameters suggested in the model.
The proposed approach may need to be further verified using more
engineering cases.
4.3. Creep loading simulation under mechanical unloading
condition
From experimental observations, it has been found that the
creep response of rubber structures is different between loading
and unloading conditions [23e25]. This phenomenon can be
simulated and explained using the proposed approach.
To combine creep with an unloading process, equations (5) and
(8) are merged to form the following expression for the strain en-
ergy potential:



W ¼ qðbÞWI I þ WJ ðJel Þ þ Rc I; T
To keep the analyses consistent, the same polynomial expression
(N ¼ 3) and the same values for WI and Rc ðI; TÞ are used. Creep
loading is approximately 23 kN, which is 50% of the maximum load.
Fig. 12 shows the creep prediction under these two conditions: one
is loading from 0 to 23 kN, and the other is from 0 to 46 kN and then
unloading to 23 kN. The creep response from the unloading creep is
found to be above that of the loading creep; an approximately 30%
difference between the two responses at day 553 is shown. This
result is qualitatively consistent with the experimental observations
from the published literature [23e25] and verifies the suitability of
the combined approach for creep prediction during unloading.
Fig. 12. Creep predictions between 50% loading and unloading conditions for the
Metacone spring.
5. Conclusions
This article presents engineering approaches to evaluate the
creep loading response and the loadingeunloading procedure for
rubber components used as anti-vibration applications. For the
unloading modelling, the theory is based on the elastic energy (i.e.,
rebound energy) remaining in the rubber material during unload-
ing. The form of the rebound resilience function q(b) ¼ 1(1q0)ba
is suitable for the unloading prediction and validated with the
experimental result.
For creep modelling, there is no universal method to predict
creep response for rubber anti-vibration products. The hyperelastic
models based on strain energy density are widely used to predict
the mechanical loading response without reference to the time
effect. To extend the capability of the hyperelastic models and to
provide a method for creep evaluation of rubber suspension sys-
tems, the damage function Rc ðI; TÞ ¼ kT z ðI 1 þ I 2 Þ is introduced into
the strain energy density expression. The theory is based on the
elastic energy change of the rubber material over the time elapsed
and on the concept of material damage. The proposed approach
was validated by considering a Metacone rubber component, which
is not a simple component. The method has been shown to provide
good agreement between predicted and experimental results. In
addition, the loading condition of the rubber product is neither
uniaxial tension nor uniaxial compression; instead, the loading
condition is a combination of shear and compression, which is
typical of a real application of anti-vibration systems. One impor-
tant outcome from this proposed approach is the combination of
unloading and creep, which has revealed that a creep loading
procedure may produce different results even with the same load is
applied; this agrees with the published literature. The proposed
approach can also be easily incorporated with commercial finite
element software (e.g., Abaqus). It is shown that the proposed
(14) method may be used for anti-vibration products at an appropriate
design stage. However, the proposed approach should be verified
further using more engineering cases.
Funding
The authors would like to acknowledge the funding support of
the National Natural Science Foundation of China (51078361) and
the Key Grant Science and Technology Research Planning Projects
of the Ministry of Railway of the Peoples Republic of China
(2014G005-A; 2008G017-C).
Acknowledgements
Permission to reproduce extracts from BS ISO 4662:2009 is
granted by BSI British Standards Limited. No other use of this ma-
terial is permitted. British Standards can be obtained in PDF or hard
copy formats from the BSI online shop: www.bsigroup.com/Shop or
by contacting BSI Customer Services for hard copies only: Tel: þ44
(0)20 8996 9001, Email: cservices@bsigroup.com.
The authors would like to thank Mr. J. Simmonds, Mr. J. Patel and
Mr. J. Foster at Trelleborg IAVS (U.K.) for the technical discussions
and assistance. The reviewers' helpful comments are also greatly
appreciated.
References
[1] A.N. Gent, J. Appl. Polym. Sci. vi (22) (1962) 433e441.
[2] P. Du, Mater. Des. 10 (5) (September/October 1989) 264e265.
[3] A.N. Gent, J. Appl. Polym. Sci. vi (22) (1962) 442e448.
[4] Findley N.W., Lai J.S., Onaran K., ISBN 0-468-66016-8, US, 1989.
[5] Said Jazouli, Wenbo Luo, Fabrice Bremand, Toan Vu-Khanh, Polym. Test. 24
(2005) 463e467.
[6] R.A. Schapery, Polym. Eng. Sci. 9 (4) (1969) 295e310.
[7] S.C. Yen, F.L. Williamson, Compos. Sci. Technol. 38 (1990) 103e118.
[8] Wenbo Luo, Chuhong Wang, Xiaoling Hu, Tingqing Yang, Acta Mech. Solida
Sin. 25 (6) (December, 2012) 571e578.
[9] Olesja Starkova, Jinglei Yang, Zhong Zhang, Compos. Sci. Technol. 67 (2007)
 R.K. Luo et al. / Polymer Testing 46 (2015) 134e143
2691e2698.
[10] Abhishek Gupta, J. Raghavan, Compos. Part A 41 (2010) 1289e1300.
[11] Frede
ric Achereiner, Kurt Engelsing, Polym. Test. 32 (2013) 447e454.
[12] J. Kolarik, J. Pegoretti, Polymer 47 (2006) 346e356.
[13] J. Kolarik, J. Pegoretti, Polym. Test. 27 (2008) 596e606.
[14] I.Y. Gnip, S. Vaitkus, V. Kersulis, S. Ve' jelis, Polym. Test. 27 (2008) 378e391.
[15] Koh-hei Nitta, Hirohito Maeda, Polym. Test. 29 (2010) 60e65.
[16] P.E. Tomlins, B.E. Read, G.D. Dean, Polymer 35 (20) (1994) 4376e4381.
[17] G.D. Dean, Polym. Test. 30 (2011) 229e235.
[18] G. Pinter, M. Haager, R.W. Lang, Polym. Test. 26 (2007) 700e710.
[19] G. Spathis, E. Kontou, Compos. Sci. Technol. 72 (2012) 959e964.
[20] Ferry JD., third edition, Wiley 1980.
[21] Lakes RS, CRC Press, US, 1998.
[22] Bower A.F. CRC Press, USA, 2010.
[23] A.D. Drozdov, Int. J. Solids Struct. 47 (2010) 3221e3233.
[24] Fazeel Khan, Colin Yeakle, Int. J. Plast. 27 (2011) 512e521.
[25] S. Oman, M. Nagode, Mater. Des. 60 (2014) 451e457.
[26] Ogden R.W. John Wiley & Sons, UK, 1984.
[27] Dassult Systems, Abaqus User Manual, Dassult Systems, USA, 2013.
[28] R.K. Luo, W.X. Wu, W.J. Mortel, J. Mater. Des. Appl. 217 (No.4) (2003)
287e290.
[29] R.K. Luo, P.W. Cook, W.X. Wu, W.J. Mortel, J. Rail Rapid Transit 217 (3) (2003)
237e240.
143
[30] R.K. Luo, W.X. Wu, P.W. Cook, W.J. Mortel, J. Rail Rapid Transit 218 (2) (2004)
173e177.
[31] Luo R. K, Wu W. X., vol. 13, No. 1, 2006,110e116.
[32] R.K. Luo, W.J. Mortel, X.P. Wu, Eng. Fail. Anal. 16 (5) (2009) 1366e1378.
[33] R.K. Luo, W.J. Mortel, P.W. Cook, J. Lake, Plast., Rubber & Compos. 4 (40) (2011)
155e160.
[34] R.K. Luo, X.P. Wu, W.J. Mortel, Proc. IMechE Part F J Rail Rapid Transit 227 (1)
(2013) 103e112, http://dx.doi.org/10.1177/0954409712450255.
[35] Touhid Zerrin-ghalami, Ali Fatemi, Proc. IMechE Part D. J. Automob. Eng. 227
(5) (2013) 691e703.
[36] R.W. Ogden, D.G. Roxburgh, Proc. R. Soc. Lond. A 455 (1999) 2861e2878.
[37] R.W. Ogden, D.G. Roxburgh, in: A. Dorfmann, A. Muhr (Eds.), Constitutive
Models for Rubber, Balkema, Rotterdam, The Netherlands, 1999, pp. 23e28.
[38] R.K. Luo, X.P. Wu, J. Strain Anal. Eng. Des. 49 (2) (February 2014) 76e85.
[39] R.K. Luo, X.P. Wu, W.J. Mortel, J. Polym. Test. 34 (April 2014) 49e57.
[40] R.K. Luo, W.J. Mortel, X.P. Wu, L.M. Peng, Proc. IMechE Part F J Rail Rapid
Transit (2014), http://dx.doi.org/10.1177/0954409714524748.
[41] A. Dorfmann, R.W. Ogden, Int. J. Solids Struct. 41 (2004) 1855e1878.
[42] R.K. Luo, Int. J. Damage Mech. 24 (2) (2015) 153e167.
[43] R.K. Luo, Peng Li Min, Wu Xiaoping, W. Mortel, J. Polym. Test. 40 (2014)
304e312.
[44] BS ISO 4662:2009, BSI British Standards Limited, 2009.
[45] Nagdi K., Hanser Publishers, Germany, 1993.