Models for Filled Rubber in Simple Shear

E. Tubaldi, H.R. Ahmadi, A.H. Muhr & J. Kingston

Tun Abdul Razak Research Centre, Brickendonbury, Hertford, Hertfordshire, UK

ABSTRACT: Laminated natural rubbersteel bearings employed in seismic isolations are often subjected to a
biaxial simple shear, so there is a need to model their behaviour under such deformations. In this paper we
first review the available models, experimental techniques and experimental data for filled rubber in simple
shear, drawing on the rather separate bodies of work from the seismic protection and the rubber research
communities. We then report and discuss results of tests carried out at TARRC to evaluate the multi-axial
response of rubber and provide insight into the nature of softening of the virgin rubber subjected to bidirectional simple shear.

1 INTRODUCTION
The work is motivated by the need for a biaxial
time-domain model for laminated natural rubber
steel seismic isolation bearings. Such bearings need
to have fairly high damping at low frequency
(~0.5Hz) and high shear strain (typically at least
100%), which is achieved in part by using a high
loading of reinforcing filler.
The bearings consist of multiple horizontal layers
of rubber, bonded between planar shims of steel
thick enough to be effectively inextensible. During
an earthquake the bearings accommodate the large
differential horizontal displacement between ground
and isolated structure, which imposes essentially a
uniform simple shear strain in the rubber, parallel to
the shims.
The evaluation of the seismic reliability of
structural systems isolated with rubber bearings
entails the analysis of the response to a stipulated set
of actual or synthetic transient ground motions after
a long period of rest, calling for a time-domain
model of the rubber in biaxial simple shear that
includes effects of strain history, such as the Mullins
effect. It is customary to carry out time history
analyses by assuming that the bearing responses
along two orthogonal axes are uncoupled, i.e, the
bearing response along one axis is not affected by
the input along the other orthogonal axis. However,
some studies (eg Huang 2002) have shown that
rubber bearings exhibit a coupled response in shear,
and it is important to evaluate the consequences of
this effect on the response of the isolated system

In this paper we first review the available models,

experimental techniques and experimental data for
filled rubber in simple shear, drawing on the rather
separate bodies of work from the seismic protection
and the rubber research communities. We then
report and discuss some results of a campaign of
tests carried out at TARRC to evaluate the multiaxial response of rubber and provide insight into the
nature of softening of the virgin rubber.
2 SIMPLE SHEAR DEFORMATION
Parallel motion of the metal plates (here taken to be
discs of radius a) between which a rubber layer (of
thickness h) is bonded (Figure 1) imposes a state of
nominal uniform simple shear on the rubber.
z
a
x

y
Figure 1. Single rubber layer bonded between shims.

We shall consider general biaxial, but not torsional,

relative motion of the plates. The approximation to
uniform simple shear becomes closer as the shape
factor S = a/2h of the rubber layer is increased,
because the imperfection in the boundary conditions
due to the stress-free edges of the rubber becomes

less significant. For seismic isolation bearings S is

typically of the order of 10, while standards for
testing rubber in simple shear stipulate S > 4. For
our current purposes we shall assume the rubber to
be in uniform simple shear.
3 REVIEW OF EXPERIMENTAL STUDIES AND
MODELS
3.1 Introduction
Many experimental investigations have been
conducted in the last 50 years of the response of
rubber layers and rubber isolators in uniaxial and
biaxial simple shear, and different models have been
proposed for describing the test results and the
rubber constitutive behaviour. Due to space
constraints, this review focuses on biaxial shear.
Gent (1960) devised a test machine to evaluate the
dynamic properties of a rotating simple shear
testpiece of rubber. Dusi et al. (1999) also conducted
some shaking table tests on laminated rubber
bearings and used the results to calibrate the
properties of a parallel elasto-plastic model to be
used within finite element (FE) codes.
Huang (2002) carried out a series of bi-directional
displacement-controlled experiments on an isolated
rigid block supported by four laminated Bridgeston
KL301 high damping rubber bearings. Different
displacement orbits were imposed to evaluate the
path- and strain-history dependent response of the
rubber as well as the effects of scragging. The
experimental results indicate that horizontal
restoring forces exhibit a coupling effect. A
constitutive model for the bearings in biaxial shear
was proposed by employing the concept of the
bounding surface and its parameters were calibrated
against experimental results. This model was later
extended and improved by Grant et al. (2004).
Abe et al. (2004a) conducted a series of cyclic
experiments on three types of laminated rubber
bearings under multiaxial loading state. The results
showed the importance of accounting for the
coupling effects in dynamic structural analyses. A
mathematical model of laminated rubber bearings
under multiaxial loading was then proposed on the
basis of the three-dimensional constitutive law of the
Ozdemir model (Abe et al. 2004b).
Kikuchi et al. (2010) carried out cyclic tests on
square laminated rubber bearings under different
directions of loading and proposed a threedimensional phenomenological model for describing
the ultimate behaviour by also considering the
effects of axial loads.
Aside from the phenomenological models
specifically developed for describing the bidirectional bearing behaviour, a number of fully

three-dimensional large strain continuum mechanics

models have been proposed.
Ahmadi et al. (1999) critically analyzed the
results of the tests of Gent (1960) and of Dusi et al.
(1999), and considered introduction of a plasticity
model in parallel to a hyperviscoelastic model to
describe the behaviour of filled rubber in simple
shear. Following Austrell (1997), this simple but
comprehensive modelling approach makes use of
material models already available in commercial FE
software and captures the amplitude-dependence of
the dynamic stiffness as well as hyperelasticity and
relaxation behaviour, although some difficulties of
implementation as a 3D continuum model in
ABAQUS still remain (Gough et al, 2015).
The major difficulty relates to the description of
the rate-independent behaviour, introduced to
simulate the Payne effect in filled rubber. The use to
this end of a continuum of elasto-perfectly-plastic
elements (Ahmadi et al. Part 1) has some
unsatisfactory features:
The tracking of a multitude of different yield
criteria, or alternatively but equivalently the use of
Masing's rule, is computationally expensive in terms
of numerical time-domain integration;
the elasto-plastic contribution can only describe
hysteretic behaviour with a monotonically
decreasing loading stress, so cannot capture the
effect of strain crystallisation which results in an
increase in stiffness and hysteresis at high strain;
retraction for a uniaxial deformation results in a
discontinuity in incremental stiffness, not seen
experimentally.
Instead, many researchers have investigated the
application of smoothed (e.g. bounding surface)
plasticity models (Abe 2004b, Grant et al. 2004),
although these models are usually not available in
commercial FE software.
4 EXPERIMENTAL CAMPAIGN FOR BIAXIAL
SIMPLE SHEAR RESPONSE
A first experimental campaign has been carried out
at TARRC to evaluate the behaviour of different
types of rubber in biaxial simple shear. In particular,
cylindrical testpieces were moulded for three
different natural rubber materials, two of which
(EDS19 and EDS16) being unfilled and filled
natural rubber (NR) respectively, formulated as
given in Table 1 of Besdo et al. (2003), and the
third (HDNR25) being a proprietary material for
seismic isolation bearings. The test pieces have a
cylindrical form, with a single rubber layer of
thickness h equal to about 8 mm, and radius a = 25
mm bonded between two steel cylinders.
Figures 2 and 3 show the test rig used for
performing the test. Two plates sliding on horizontal
orthogonal linear bearings are employed to give the

bidirectional horizontal displacements to the

specimens without rotation and torsion. Each plate is
attached to a horizontal actuator, whose maximum
load and stroke capacities are respectively 10 kN
and 150 mm. A sliding plate on a vertical linear
bearing can be used to impose a vertical load, or
fixed at a given displacement.

amplitude of laminated rubber bearings employed

for seismic isolation.
Path "b" traces a circular strain orbit of radius
max = 1 or 2 in the shear strain plane and it is
obtained by imposing in the two orthogonal
direction harmonic motions at the same frequency of
0.5 Hz and with a phase lag of /2.

y [-]
max

x [-]

max

max

y [-]

max

b)
max

a)

max

max

max

x [-]

max

max

y [-]
max

x [-]

max

y [-]
max

Figure 2. Biaxial test rig: showing actuators and triaxial load

cell.

x [-]

d)
max

c)

max

max

Figure 4. Imposed strain biaxial orbits: a) cruciform orbit, b)

circular orbit, c) figure 8 orbit, d) square orbit.

Throughout the experiments, the displacements in

the horizontal directions are measured by
displacement transducers in the actuators whereas
the forces, free of the rolling resistance of the linear
bearings, are measured by a triaxial load cell placed
below the test piece.
A series of cyclic displacement orbits of different
shapes and amplitudes (Figure 4) have been imposed
on the rubber layers under an initial vertical pressure
of 2 MPa.
Path "a" traces a cruciform orbit in the strains
plane and permits evaluation of the rubber response
along two orthogonal directions under temporally
separated cyclic harmonic displacements; 0.5 Hz and
amplitudes max = 1 or 2 were chosen, for
consistency with the design isolation frequency and

2
1.5

x
y

1
0.5

[-]

Figure 3. Biaxial test rig: showing rubber sample in shear

Path "c" traces a figure 8-shaped path, obtained by

imposing harmonic displacement at a frequency of
0.5 Hz along one direction, and at twice this
frequency along the other direction, for a maximum
strain amplitude max = 1.
Finally, Path "d" traces a square orbit at the strain
amplitude max = 1. The total number of cycles for
the different tests varies between 8 and 20.
Here we report only results for EDS16 rubber in
the last cycle, when the softening per cycle has
greatly diminished.

0
-0.5
-1
-1.5
-2
-2.5 -2

-1.5 -1

-0.5

0.5

1.5

2.5

[-]

Figure 5. Stable hysteretic response to cruciform strain orbit.

Figure 5 shows the hysteretic response of the rubber

layer along the x and y directions under the

0.8
0.6
0.4

y [MPa]

cruciform strain orbit, and confirms that the material

is isotropic. Obviously, when the rubber is subjected
to a shear strain in one direction, it exhibits a
response only on that direction, whereas the resisting
force in the orthogonal direction is zero. As expected
from the presence of reinforcing filler in the
material, the hysteresis loops are not elliptical but
have characteristic rotational symmetry, indicative
of significant amplitudes of odd harmonics in the
periodic stress, with an enhanced retraction stiffness,
consistent with the Payne effect (Ahmadi et al.
1997).
Figures 6 and 7 show the response to the circular
orbit. Remarkably, the behaviour is much as would
be expected for a viscoelastic material: in Figure 7,
the hysteresis loops are nearly elliptical, suggesting
the Payne effect has been suppressed by the
presence of the orthogonal component. However, the
hysteresis, compared to Figure 5, has increased.

-0.4
-0.6
-0.8
-0.8

-0.4

-0.2

0.2

0.4

0.6

0.8

0.8
0.6

x
y

0.4

[MPa]

0.2

0.5

[MPa]

-0.6

x [MPa]
Figure 8. Stable stress orbit response to cruciform strain orbit.

0
-0.2
-0.4

-0.5

-0.6
-0.8
-1.5

-1

-1

-0.5

0.5

1.5

[-]
-1.5

-1

-0.5

0.5

1.5

x [MPa]
Figure 6. Stable stress orbit response to circular strain orbit.
1.5
1

x
y

Figure 9. Stable hysteretic response to figure 8 strain orbit.

For the square orbit (Figures 10 and 11), there is an

apparently instantaneous drop in stress when the
straining direction changes, consistent with
expectation if there were an element of behaviour
akin to plasticity. Moreover, the drop appears to
have some contribution from the stress relaxation.
0.8

0.5

[MPa]

0
-0.2

1.5

-1.5

0.2

0.6
0

0.4

y [MPa]

-0.5
-1

0.2
0
-0.2

1.5
-2.5 -2

-1.5 -1

-0.5

0.5

1.5

2.5

[-]

Figure 7. Stable hysteretic response to circular strain orbit.

A coupling effect is also evident for the figure of 8

results, shown in Figures 8 (note the asymmetry
about the x = 0 axis) and 9.

-0.4
-0.6
-0.8
-0.8

-0.6

-0.4

-0.2

0.2

0.4

0.6

0.8

x [MPa]
Figure 10. Stable stress orbit response to square strain orbit.

time t0 and then another 6 half cycles at the angle

180. Waiting times of 0.3s, 2.5s (as shown in
Figure 14), and 60s were used.

0.8
0.6
0.4

x
y

a)
1.5

[-]

[MPa]

0.2

-0.2
-0.4

-1.5
0

-0.6
-0.8
-1

10

15

-0.8 -0.6 -0.4 -0.2

0.2 0.4 0.6 0.8

[-]

Figure 11. Stable hysteretic response to square strain orbit.

30

35

40

25

30

35

40

0
-1.5
0

10

15

20

t [-]

Figure 12. Imposed strain histories: a) 20 full cycles, b) 12 half

cycles, separated by a pause t0 = 0.3s at zero strain.
a)

[MPa]

1,1+

1,10

b)

[MPa]

10

15

20

25

30

35

40

25

30

35

40

t [-]

2,1

1,1

2,7
0

10

15

20

t [-]

Figure 13. Stress histories corresponding to the strain histories

of Figure 12.
a)

[MPa]

10

15

20

25

30

35

40

25

30

35

40

t [-]

b)
2

[MPa]

Cylindrical double shear testpieces (described by

Ahmadi et al., 2008a) were installed in a special jig
in a uniaxial servohydraulic testing machine. This
jig permitted rotation of the testpiece about its axis,
such that all three metal pieces go through the same
angle so as not to induce torsion, between discrete
tests along the machine axis. During these uniaxial
tests, the metal pieces were firmly clamped to
eliminate any backlash or tendency to rotate.
Starting from zero shear strain, the testpieces were
subjected to a sequence of cycles in uniaxial shear
first in one direction, then in a different direction.
For 180 rotation, the sign of the strain was changed,
rather than rotating the testpiece.
Due to space constraints, only some results are
shown concerning the response of two different
double shear pieces of virgin HDNR25 rubber
subjected to the strain histories shown in Figure 12.
The strain rate between direction changes was 2 s-1.
The corresponding stress histories are shown in
Figure 13.
It is observed that the repeated cycling induces a
softening in the rubber, as expected. However, the
maximum stress 2,7 obtained in the first half cycle at
180 following the 6 cycles at 0 in strain history b)
is very close to the corresponding stress value 1,1attained by imposing full cycles, in correspondence
of the second part of the first full cycle. In other
words, this figure suggests that the scragging at 0
does not influence the scragging at 180. Thus,
existing isotropic continuum damage models may
not describe accurately the evolution of softening.
The second set of results discussed here concerns
the response of four other virgin rubber samples
subjected to 12 half cycles at the angle 0 and at the
amplitude 1.5. The other three strain histories consist
of 6 half cycles at the angle 0 followed by a waiting

25

1.5

[-]

5 EXPERIMENTAL CAMPAIGN FOR

CHARACTERIZING SOFTENING UNDER
SIMPLE BIAXIAL SHEAR RESPONSE

20

t [-]

b)

10

15

20

t [-]

Figure 14. Stress histories corresponding to similar strain

histories to Figure 12b, except for the waiting time t0 = 2.5s.

In order to measure the evolution of the softening

following the imposition of a strain history, the following parameter is defined:

Edi d

(1)

which represents the energy dissipated in a strain

cycle. Figure 15 plots the variation of Edi with the
number of cycles for two different strain histories
leading to the stress histories represented in Figure
14. It can be observed that the energy dissipated in a
cycle due to softening in the strain history (a),
characterized by cycles of deformation imposed only
at 0, always decreases with increasing number of
cycles. On the other hand, the energy dissipated in a
cycle of the other strain histories increases after the
direction of straining is changed from 0 to 180,
and the value of Edi for i=7 is very close to the value
of for i=1. Thus, the softening at one angle does not
influence significantly the softening at the other
angle. Using the jig permitting rotation of double
shear testpieces between tests, it was established that
the softening decreases progressively as the angle of
test is varied from 0 to 180.
2
history (a)
history (b) t0=0.3 s
history (b) t0=2.5 s
history (b) t0=60 s

1.8
1.6

Edi [-]

1.4
1.2
1
0.8
0.6
0.4
0.2
0

10

11

12

cycle [-]
Figure 15. Stress histories corresponding to the strain histories
of Figure 14

CONCLUSIONS
In this paper, experiments carried out to characterize
the behaviour of natural rubber in simple shear are
reviewed, together with existing models developed
to describe this behaviour.
Some preliminary results concerning the response
of different types of rubber subjected to simple shear
are shown. These results are currently being used to
develop a new phenomenological model for
describing the behaviour of laminated rubber
bearings subjected to the earthquake excitation and
the relevant effects of softening exhibited during the
first cycles of deformation.

ACKNOWLEDGMENTS
Ian Stephens is thanked for help with jig design.
REFERENCES
Abe, M., Yoshida, J., Fujino, Y. 2004a. Multiaxial behaviors of
laminated rubber bearings and their modeling. I:
Experimental study, Journal of Structural Engineering,
130(8): 1119-1132.
Abe, M., Yoshida, J., Fujino, Y. 2004b. Multiaxial Behaviors
of Laminated Rubber Bearings and their Modeling. II:
Modelling, Journal of Structural Engineering, 130(8): 11331144.
Ahmadi, H.R., Kingston, J.G.R. & Muhr, A.H. 2008a.
Dynamic properties of filled rubber Part I: Simple model,
experimental data and simulated results. Rubber Chemistry
and Technology. 81(1): 1-18.
Ahmadi, H.R. & Muhr, A.H. 1997. Modelling dynamic
properties of filled rubber. Plastics, Rubber and Composites
Processing and Applications. 26: 451-461.
Ahmadi, H.R. & Muhr, A.H. 2011. Dynamic properties of
filled rubber Part II: Physical basis of contributions to the
model. Rubber Chemistry and Technology. 84: 24-40.
Ahmadi, H.R., Besdo, D., Ihlemann, J., Kingston, J.G.R. &
Muhr, A.H. 2005. Transient response of inelastic materials
to changes of amplitude. In Austrell & Kari (eds.)
Constitutive Models for Rubber IV, Leiden: Balkema.
Ahmadi, H.R., Ihlemann, J. & Muhr, A.H. 2007. Timedependent effects in dynamic tests of filled rubber. In
Boukamel et al. (eds.) Constitutive Models for Rubber V,
Leiden: Balkema.
Austrell, P.E. 1997. Modeling of elasticity and damping for
filled elastomers. PhD Thesis, Lund Institude of
Technology Report TVSM-1009, University of Lund,
Sweden.
Besdo, D., Ihlemann, J., Kingston, J.G.R. & Muhr A.H. 2003.
Modelling inelastic stress-strain phenomena and a scheme
for efficient experimental characterisation. In Busfield &
Muhr (eds.), Constitutive Models for Rubber IV, Leiden:
Balkema.
Dusi A., Bettinali F., Rebecchi V., Bonacina G. (1999). High
Damping Laminated Rubber Bearings (HDLRB): a
simplified non linear model with exponential constitutive
law. Model description and validation through experimental
activities. In Dorfmann and Muhr (eds.) Constitutive
Models for Rubber I, Leiden: Balkema.
Gent, A.N. 1960. Simple rotary dynamic testing machine.
British Journal Appl. Physics, 11(4): 165-167.
Gough J, Ahmadi H.R, Harris A.J, Kingston J.G.R, Whear R,
Kilgour A, Seng T.C; 2015 Modelling the dynamic
behaviour of an automotive bush with a visco-hyperelasticelastoplastic model, Constitutive Models for Rubber 9
Grant D N, Fenves G L& Whitaker A S 2004; Bidierctional
modelling of HDR bearings J Earthquake Engineering, 8,
161 - 185
Huang, W-H 2002. Bidirectional Testing, Modelling & System
Response of Seismically Isolated Bridges, PhD thesis,
University of California, Berkeley.
Kingston, J.G.R. & Muhr, A.H. 2011. Effect of scragging on
parameters in viscoplastic model for filled rubber. Plastic,
Rubbers and Composites, 40(4): 161-168.
TARRC, 1979-1986. Natural Rubber Engineering Data Sheets
(EDS). TARRC, Hertford, UK .