A polynomial analytical model of rubber bearings based on series of tests

Igor Gjorgjiev

, Mihail Garevski
1
Institute for Earthquake Engineering and Engineering Seismology-IZIIS, Ss Cyril and Methodius University, Salvador Aljende 73, P.O. Box 101, 1000 Skopje, The Former
Yugolav Republic of Macedonia
a r t i c l e i n f o
Article history:
Received 6 September 2012
Revised 27 February 2013
Accepted 9 April 2013
Available online 29 June 2013
Keywords:
Rubber bearings
Production of bearings
Testing of bearings
Polynomial model
a b s t r a c t
Rubber bearings are among the most frequently applied devices in seismic isolation. Although the behav-
ior of rubber bearings under strong earthquakes is nonlinear, it is often interpreted through the bilinear
constitutive law. The investigations presented in this paper represent a successful attempt to simulate
nonlinear forcedisplacement relationships. The rst part of the paper covers production and testing of
rubber bearings and results obtained. The second part deals with development of a simple nonlinear
mathematical model of a rubber bearing involving a polynomial function and eight additional parameters
obtained from biaxial tests. The polynomial model can simulate the behavior of natural rubber bearings
in case of small and large deformations. The model is capable of covering the strengthening of the rubber
in conditions of large deformations, including the loading history effect. Based on comparison between
the analytical and experimental results, it is concluded that the proposed polynomial model is capable
enough to simulate the forcedisplacement relationship of rubber bearings.
2013 Published by Elsevier Ltd.
1. Introduction
One of the most effective techniques for design of earthquake
resistant structures is application of seismic isolation. It is a collec-
tion of isolation units, which are to reduce the transfer of seismic
energy to the upper structure. Over the last century, many isolation
devices have been invented, but only a few of them have become
popular. Out of these, rubber bearings are commonly used for iso-
lation of buildings and bridges. The rst application of a rubber iso-
lation systemtook place in the sixties of the last century. A primary
school in Macedonia was isolated by a system referred to as Swiss
Full Base Isolation [22]. The basic concept of this system was pro-
viding a full 3D isolation by use of non-reinforced rubber bearings.
Seismic isolation has practically become a reality with the develop-
ment of multi-layered elastomeric bearings produced by simulta-
neous vulcanization of rubber and gluing of steel plates. The
investigations done by Kelly [15,16], Naeim and Kelly [20] and
Skinner et al. [21] have given a huge contribution to acquiring
knowledge on the behavior of base isolated structures and differ-
ent types of bearings. The application of base isolation of structures
has been increased particularly after the Kobe (1995) earthquake
[6], when good behavior of base isolated structures was observed
[7]. Unfortunately, due to the high cost of the isolators, their appli-
cation has long been limited to economically developed countries.
However, today, even developing countries like Macedonia [8,9],
and Armenia [18] are successfully producing and applying these
devices.
To promote application of base isolation in R. Macedonia and
the remaining countries in the Balkan, the Institute of Earthquake
Engineering and Engineering Seismology (IZIIS) proposed a project
on development of low-cost rubber bearings [9], which was -
nanced by NATO through the Science for Peace (SfP) Programme.
Within this project, a number of raw rubber recipes and a technol-
ogy for production of rubber bearings were developed for the rst
time in the region. Within this project, ample experimental tests
were performed on bearings produced of 18 different rubber
compounds. These tests enabled creation of a large database of
forcedisplacement relationships for different rubber compounds
that was later used for development of a new analytical model of
rubber bearings.
The laboratory tests on rubber bearings which have been car-
ried out within these investigations and the investigations per-
formed by other authors [13,17,23], have pointed out their
nonlinear behavior. The stressstrain behavior of rubber bearings
is manifested by high horizontal stiffness under low shear strains,
low stiffness under moderate strains, and an increasing shear mod-
ulus under higher strains. Usually, rubber hardening begins at
125% shear strain and continues until failure.
Lateral bearing behavior is quite complex because of strain-rate
dependence and the presence of the Mullins effect [19], [12]. Other
factors that have an inuence upon the hysteretic behavior of
0141-0296/$ - see front matter 2013 Published by Elsevier Ltd.
http://dx.doi.org/10.1016/j.engstruct.2013.04.006

Corresponding author. Address: Department for Geotechnics and Special

Structures, Institute for Earthquake Engineering and Engineering Seismology-IZIIS,
Ss Cyril and Methodius University, Salvador Aljende 73, P.O. Box 101, 1000
Skopje, The Former Yugolav Republic of Macedonia. Tel.: +389 2 3107 701.
E-mail addresses: igorg@pluti.iziis.ukim.edu.mk (I. Gjorgjiev), garevski@pluti.
iziis.ukim.edu.mk (M. Garevski).
1
Tel.: +389 2 3107 701.
Engineering Structures 56 (2013) 600609
Contents lists available at SciVerse ScienceDirect
Engineering Structures
j our nal homepage: www. el sevi er . com/ l ocat e/ engst r uct
rubber bearings are the axial load, ambient temperature, ageing ef-
fects and compounding. All these factors have resulted in complex-
ity of hysteretic behavior of rubber bearings. Because of such
complex behavior, the generally accepted modeling of bearings
in horizontal direction by a bilinear model [5,20] cannot com-
pletely cover the actual bearing behavior. However, on the other
hand, if all these factors are included in the analytical model of a
rubber isolator, then this model will be ideal but also very
complex.
To enable a more realistic dynamic analysis of structures iso-
lated with rubber bearings, several advanced analytical models
have been proposed in literature. One of the approaches is repre-
senting the nature of damping as hysteretic [17]. This model in-
cludes the high nonlinearity of shear strains but it does not
include the strain rate and the effect of variation of the axial load
upon the hysteretic properties of the bearing. The other approach
includes implementation of the viscoelastic theory where the
forces depend on the strain rate [13]. This proposed model is de-
ned by ten parameters that are functions of the number of factors
inuencing the behavior of the elastomer. This model includes the
Mullins effect, the strain rate and temperature dependence. The
model proposed by Tsai et al. [23] also includes the effect of the
strain rate. This model is based on the Bouce-Wens model [2]
which has been extended to seven material parameters and it
can simulate high strain rate at low strains and strengthening of
rubber at high strains. Another model where the hysteresis has
the typical buttery shape is proposed by Dall Asta and Ragni
[3]. In this model, the dependence on the strain rate and the Mul-
lins effect are included, as well. The rate dependence has also been
investigated and included in the model proposed by Jankowski
[14].
The rst phase of the investigations covered the development of
rubber bearings and their testing. A comprehensive study was per-
formed to develop the technological process of production of rub-
ber bearings reinforced by steel shims. Providing quality of the end
product and development of an appropriate recipe were one of the
key tasks since such bearings were produced for the rst time in
the Balkan region. The quality of the end products was veried
through a series of dynamic tests. In this part, the results from
the vertical and horizontal tests carried out on a few characteristic
bearings are presented through forcedisplacement diagrams.
The analytical polynomial model proposed in the second part of
this paper was developed based on a series of tests on rubber iso-
lators. Presented further in this part are the most characteristic
ones. The proposed model can simulate the behavior of natural
rubber bearings in conditions of small and large deformations.
The model is dened by a polynomial function and eight parame-
ters and it is able to cover the strengthening of the rubber in con-
ditions of large deformations. It includes the loading history effect
and enables adaptation to different shapes of loading/unloading. A
least-square regression was used to determine the best coef-
cients in order to minimize the sum of the squares in an nth order
polynomial model. System property modication factors for the ef-
fects of aging, temperature and scragging of rubber bearing [1] can
also be included, if necessary. The consideration of the system
property modication factors in this analytical model enables
extension of the bilinear simplication [1] to the range of large
deformations. This model can substitute the bilinear simplication
of rubber bearings in the design procedure and was implemented
in the nite element nonlinear program [11].
2. Production and testing of natural rubber bearings
One of the main goals of the NATO SfP project was development
of a high damping rubber (HDR) compound, adoption of produc-
tion of rubber bearings and their testing and implementation.
The production of rubber bearings with different rubber com-
pounds was carried out in a small local workshop in Macedonia.
Eighteen natural rubber compounds were developed and used for
production of more than 100 bearings. Two types of bearings were
produced. The rst was of a square shape, side length of 200 m and
total height of 75 mm (Fig. 1b). The second was of a circular shape,
external diameter of 150 mm and total height of 100 mm (Fig. 1c).
The bearings pertaining to each type were produced without inter-
nal steel plates and with different number of steel plates.
Prior to the very process of vulcanization, the raw rubber was
calendared to obtain rubber sheets with the required thickness.
The internal steel plates, referred to as shims, were proportioned
190/190 mm and 140 mm. The rst phase of treatment of the
shims consisted of cleaning the sandblasted areas with medical
gasoline. This helped remove all the dirt occurring from the mo-
ment of sandblasting to the moment of vulcanization. The second
phase involved coating of the areas with an appropriate adhesive.
This coating protects the steel areas against any external effects.
After 2 h, the third phase took place, i.e., treatment of the steel
areas with the second coat. As in the case of the rst coat, after
the second coating, the plates were also left to dry at room temper-
ature for about an hour.
Upon completion of the preparation works, the tailored rubber
and the treated reinforcement were inserted in a mold, which was
previously heated to 120 C. The rubber was vulcanized for 50 min
at a temperature of 150 C (Fig. 1a). In the course of the vulcaniza-
tion, the rubber was constantly exposed, from the upper and the
lower side, to external pressure of 150 bars. Once the vulcanization
was over, the element was dismantled from the mold and left to
cool at room temperature for 24 h.
After the production of the bearings, two types of tests were
performed, namely vertical and biaxial tests. All the tests on rubber
bearings were carried out in the Dynamic Testing Laboratory of the
Institute of Earthquake Engineering and Engineering Seismology
(IZIIS), Skopje. To perform the tests, the existing biaxial dynamic
frame with a dynamic load capacity of F
MAX
= 100 kN and stroke
of D = 75 mm, was used. The static load capacity of the testing
frame is F
MAX
= 200 kN and the stroke is D = 50 mm. The set-up
of the vertical and biaxial test is shown in Fig. 2a and b.
A detailed description of the selected tested specimens is given
in Table 1.
Prior to each testing of the isolators, the materials were stabi-
lized by keeping them under room temperature of (22 2 C) with
a duration of 24 h. First, the isolators were pre-loaded with three
cycles at low frequency. Such a pre-loading is practiced in testing
rubber elements for the purpose of eliminating the Mullins effect
[19]. Pre-loading is carried out exclusively for elements not loaded
in the course of the preceding 24 h.
The procedure for the axial tests involved loading of the ele-
ment with vertical compressive force at different axial stress levels
(Fig. 3). The specimens were monotonically loaded up to the
achievement of the necessary stress level when harmonic excita-
tion was applied. The purpose of these tests was to dene the effect
of the number of internal layers upon the vertical stiffness of the
bearings at different load levels.
Fig. 4 shows the vertical forcedisplacement relationship for the
rubber bearings produced by different number of internal steel
plates. The presented graph provides a thorough insight into the
behavior of the bearings with and without layers. As expected,
the bearing with the greatest number layers (four shims) exhibited
the highest vertical stiffness.
For this series of produced bearings, the vertical stiffness at dif-
ferent axial load levels has been calculated (Table 2).
The biaxial tests consisted of application of horizontal load
while the element was exposed to vertical load. The bearings were
I. Gjorgjiev, M. Garevski / Engineering Structures 56 (2013) 600609 601
tested under two different harmonic loads: (1) a sinusoid function
with constant amplitude dened by two parameters: amplitude
and frequency (Fig. 5a), (2) a sinusoid function with a linear
increasing amplitude (Fig. 5b) and a linear decreasing amplitude
(Fig. 5c). The biaxial tests were performed to dene the behavior
of the bearings under different loading conditions. The obtained
data on the behavior of the bearings were later used in the devel-
opment of analytical lateral force displacement relationships.
These tests provided an accurate information about the hori-
zontal stiffness and the equivalent damping of the bearing for each
hysteretic cycle. Fig. 6 shows comparison of the behavior of the
rubber bearings made of eight different rubber compounds ex-
posed to horizontal sinusoidal excitation with constant ampli-
tudes. The graph provides a clear insight into the difference in
the damping abilities and the difference in shear behavior. From
the graphs presented in Fig. 6, it can be concluded that the shear
behavior mainly depends on the rubber compound. Analyzing
these types of behavior, it can be said that some bearings start to
behave in the nonlinear range at lower strains, while others start
to behave nonlinearly at higher strains. During unloading, almost
all bearings exert high nonlinearity accompanied by different
damping values.
A similar behavior of the specimens was observed under loads
of increasing and decreasing amplitudes (Fig. 7).
The effective shear modulus and hysteretic damping of the
bearings were computed within the second hysteretic cycle, under
Fig. 1. (a) Vulcanization of isolator, (b) squared isolators and (c) circular isolators.
Fig. 2. (a) Set-up of axial test of squared bearing and (b) set-up of biaxial test of two circular bearings.
Table 1
Specimen dimensions and rubber properties.
No. Label Shape Size b/d/h or
D/h (mm)
No. layers G,
50
(kPa) b
eff,50
(%)
1 spec-1 Squared 200/200/70 3 890.0 0.00
2 spec-2 Squared 200/200/70 4 1100.0 2.50
3 spec-3 Squared 200/200/70 0 1100.0 2.50
4 spec-4 Squared 200/200/70 3 1200.0 4.50
5 spec-5 Squared 200/200/70 3 370.0 4.50
6 spec-6 Squared 200/200/70 3 1500.0 7.20
7 spec-7 Squared 200/200/70 3 1000.0 7.80
8 spec-8 Squared 200/200/70 3 750.0 8.50
9 spec-9 Squared 200/200/70 3 700.0 8.60
10 spec-10 Squared 200/200/70 0 640.0 9.80
11 spec-11 Circular 200/100 5 465.0 8.70
12 spec-12 Circular 200/100 4 1350.0 12.50
Fig. 3. Vertical excitation.
Fig. 4. Comparison of axial behavior of rubber bearings.
602 I. Gjorgjiev, M. Garevski / Engineering Structures 56 (2013) 600609
an excitation with a constant amplitude and shear strain of c = 0.5
(Table 1). The results given in the enclosed table show that the
tested rubber bearings have different mechanical characteristics.
Namely, the effective shear modulus ranges within (0.371.5)
Mpa, while the hysteretic damping is within the range of 0.0% to
12.5%. It can be said that the specimens with zero damping be-
haved almost as a linear elastic material.
To enable a clear presentation of the shear behavior of the rub-
ber bearings, the forcedisplacement curves were divided into two
parts. The rst part involved the loading state (Fig. 8a) whereas the
second involved the unloading state (Fig. 8b). From the considered
behavior under loading, it was concluded that the bilinear approx-
imation was valid up to a moderate shear strain (c 125%). In the
range of high strains, strengthening of the rubber takes place and it
starts to behave with a higher degree of nonlinearity.
Horizontal tests with variable amplitude were conducted to
analyze the effect of the loading history (Fig. 9a). The graph shows
that, while horizontal deformation decreases linearly, horizontal
force at zero horizontal deformation (D
h
= 0) deteriorates nonlin-
early. In addition to the variation of horizontal force at D
h
= 0, there
is also a variation in horizontal stiffness (Fig. 9b).
To get an insight into the effect of displacement upon horizontal
(tangent) stiffness, a horizontal stiffness-displacement graph was
established (Fig. 10a). The horizontal tangent stiffness was calcu-
lated according to expression Kh = DF/DD. The force increment
DF and the displacement increment DD were obtained from the ac-
tual hysteretic curve. This gure shows the horizontal stiffness-dis-
placement relationship during the second and the last loading/
unloading cycle. In the case of loading in the second cycle (D
h
= 70-
vv), the stiffness was almost unchanged until displacement of 40vv
(c 60%). After this point, the stiffness continuously increased.
During the last cycle (D
h
= 18 vv), the stiffness continuously deteri-
orated. This happened since, during the test, the bearing was
loaded starting from a larger to a lower amplitude of excitation.
In the tests performed with an opposite order of loading (from
smaller to larger amplitudes), the stiffness tended to increase with
the increase of the horizontal displacement (Fig. 10b).
In the case of unloading in the second cycle (D
h
= 70 vv), the
stiffness considerably deteriorated from the beginning of unload-
ing until displacement of 40 vv (c 60%). After this point, the
stiffness remained constant. During the last cycle (D
h
= 18vv), the
stiffness continuously deteriorated. In both loading cycles, the hor-
izontal stiffness was almost equal at the beginning of the unload-
ing, whereas in the case of D
h
= 0, the stiffness was higher during
the loop with smaller displacements.
The effect of the previously experienced deformation upon the
horizontal stiffness of the bearings was investigated through shak-
ing table testing of a base isolated liquid storage tank [10]. In these
investigations, the base isolated structures were exposed to excita-
tion of different intensity. Also, in this case, it was concluded that
the bearings behaved in almost the same nonlinear manner. Start-
ing from these two investigations, additional parameters control-
ling the stiffness of the bearings depending on the previously
experienced deformation were implemented in the proposed ana-
lytical model.
The results from the performed experimental investigations
were also additionaly used for analysis of the Mullins effect
(Fig. 11). The Mullins effect was not only observed in a non-previ-
ously loaded material but also in a previously loaded material that
was not exposed to any load for a day. This phenomenon was also
previously investigated by Dall Asta and Ragni [3], who concluded
that the behavior of the bearings during the rst loading cycle de-
pended on the time elapsed between the application of the loads.
The inuence of Mullins (scragging) effect on the response of a
base isolated structure was investigated by Grant et al. [4]. In their
work, two nite element models were analized. In the rst model,
the rubber bearings were modeled as scragged isolators, while in
the second one, the scragging effect was included. From the pre-
sented results, it can be seen that the inclusion of scragging in
the model clearly results in much larger peak displacements. How-
ever, the peak shear forces are not signicantly different. If the pro-
posed model is applied for design purposes, the neglecting of the
scragging effect is usually on the conservative side and is therefore
acceptable. On the other hand, in the case of seismic analysis per-
formed by consideration of near fault conditions, taking the Mul-
lins effect into account could enable a more accurate analysis. If
system property modication factors [1] are included in the pro-
posed model, it can also be applicable for running near fault earth-
quake analysis.
3. Mathematical formulation of the proposed model
Based on the performed tests on the bearings made of 18 differ-
ent natural rubber compounds, an analytical model was proposed
to mathematically describe the behavior of the natural rubber
bearings. The model includes the linear-elastic behavior, the
post-elastic behavior at loading and the post-elastic behavior at
unloading (Fig. 12).
The linear elastic state includes elastic behavior of the bearing
at low strains and it is presented by the following expression:
F K
1
D 1
Table 2
Vertical stiffness of rubber bearing at various vertical loads.
Specimen F1
(kN)
K1
(kN/m
0
)
F2
(kN)
K2
(kN/m
0
)
F3
(kN)
K3
(kN/m
0
)
F4
(kN)
K4
(kN/m
0
)
spec-2 28 40,900 46 45,900 64 46,900 84 49,400
spec-3 16 7000 29 7800 48 9900 76 20,100
spec-4 17 24,700 40 32,500 67 37,800 86 40,600
spec-11 24 12,400 32 15,100 51 20,150 84 28,200
spec-12 22 22,900 33.5 25,000 59 36,500 75 45,400
(a) (b) (c)
Fig. 5. Horizontal excitations: (a) constant amplitudes, (b) increasing amplitudes and (c) decreasing amplitudes.
I. Gjorgjiev, M. Garevski / Engineering Structures 56 (2013) 600609 603
where F is the horizontal force, D is the horizontal deformation and
K
1
is the elastic (initial) stiffness of the bearing. The linear elastic
state of this model takes place at the beginning of loading of the
bearing and at the transition from unloading into loading conditions
(Fig. 13). The input parameters for this state are: elastic stiffness of
the bearing K
1
and yielding deformation.
The post-elastic state at loading involves the behavior of the
bearing after the yielding point. Once the yielding point is
exceeded, the forcedisplacement relationship is dened by the
polynomial function given in the following equation:
F
i
F
i
om
a
1
D
i
a
2
D
2
i
a
3
D
3
i
::: 2
where F
i
is the horizontal force, D
i
the horizontal deformation and
F
i
om
, a
1
, a
2
, a
3
, . . . are the coefcients of the polynomial function
which are calculated for best tting the experimental curve.
(a) (b) (c)
(d) (e) (f)
(g) (h)
Fig. 6. Comparison of horizontal behavior of bearings made of eight different rubber compounds.
(a)
-1.5 -1 -0.5 0 0.5 1 1.5
Shear strain [m/m]
-1500
-1000
-500
0
500
1000
1500
S
h
e
a
r

s
t
r
e
s
s

[
k
P
a
]
spec-4
f=0.3Hz
excitation-3
-1.5 -1 -0.5 0 0.5 1 1.5
Shear strain [m/m]
-1500
-1000
-500
0
500
1000
1500
spec-4
f=0.3Hz
excitation-4
(b)
S
h
e
a
r

s
t
r
e
s
s

[
k
P
a
]
Fig. 7. Comparison of horizontal behavior of bearings under excitation with (a) increasing amplitudes (b) decreasing amplitudes.
604 I. Gjorgjiev, M. Garevski / Engineering Structures 56 (2013) 600609
The polynomial coefcient F
i
om
BF
i
o
is dened based on the cur-
rent and previously experienced horizontal deformation. The
dependence of F
i
o
on the previously achieved deformation is de-
ned as linear and it is given by expression (3).
F
i
o

F
min
o
when D
KP
i
< D
min
o
F
min
o
F
max
o
F
min
o

D
KP
i
D
min
o
D
max
o
D
min
o
when D
max
o
< D
KP
i
< D
min
o
F
max
o
when D
KP
i
> D
max
o
8
>
>
>
<
>
>
>
:
3
where
D
KP
i
maxD
KP
; D
i
4
Forces F
min
o
and F
max
o
and deformations D
min
o
and D
max
o
are parameters
dened through tests on bearings.
The post-elastic state at unloading includes the behavior of the
bearing when |D
i1
| > |D
i
|, where |D
i1
| is the previous horizontal
deformation, |D
i
| is the current horizontal deformation of the bear-
ing. The mathematical forcedisplacement relationship for post-
elastic behavior at unloading is formulated by the same polyno-
mial function for post-elastic behavior at loading which is given
in Eq. (2). To include different types of behavior of the bearings
at unloading, modication is made in the computation of coef-
cient F
i
om
given by:
F
i
om
F
i
o
2 kF
i
o
1 5
(a) (b)
D [mm]
0
10
20
30
40
spec-8
spec-9
spec-10
loading part
0 15 30 45 60 75 0 15 30 45 60 75
D [mm]
-10
0
10
20
30
40
spec-8
spec-9
spec-10
unloading part
F

[
k
N
]
F

[
k
N
]
Fig. 8. Comparison of horizontal behavior of bearings under (a) loading and (b) unloading.
(a) (b)
D [mm]
-20
-10
0
10
20 spec-12
K
d70
K
d18
-75 -50 -25 0 25 50 75 -75 -50 -25 0 25 50 75
D [mm]
-20
-10
0
10
20
spec-12
F
0
F
n
F

[
k
N
]
F

[
k
N
]
Fig. 9. (a) Effect of previously achieved deformation upon horizontal force at zero deformation and (b) effect of previously achieved deformation upon horizontal force at zero
deformation during the rst and the last loading cycles.
0 25 50 75
D [mm]
0
500
1000
K

[
k
N
/
m
]
second cycle
last (8) cycle
spec-12
excitation-3
unloading
part
unloading
part
loading
part
loading
part
-20 -10 0 10 20
D [mm]
-10
-5
0
5
10
F

[
k
N
]
excitation-4
excitation-3
spec-9
(a) (b)
Fig. 10. (a) Horizontal (tangent) stiffness-displacement relationship during the second and the last cycle and (b) bearing behavior under increasing and decreasing amplitude.
I. Gjorgjiev, M. Garevski / Engineering Structures 56 (2013) 600609 605
where kF
i
o
is the decay coefcient and it is computed according to
the following equation:
kF
i
o

e
eD
KP
i

D
i
D
KP
i

1
e
eD
KP
i

1
6
where e is the natural logarithmic base (e = 2.71828. . .), D
KP
i
is the
horizontal deformation of the isolator at the moment of beginning
of unloading and eD
KP
i
is the exponent at the moment of beginning
of unloading.
The value of eD
KP
i
is computed according to Eq. (7) where eD
KP
i
is
linearly dependent on D
KP
i
. The remaining variables are input
parameters and they are dened on the basis of horizontal tests.
eD
KP
i
eD
min
eD
max
eD
min

D
KP
i
D
min
e
D
max
e
D
min
e
7
-40 -20 0 20 40
D [mm]
-40
-20
0
20
40
first test
time span 20'
spec-7
F

[
k
N
]
Fig. 11. Bearing behavior in a time span of 20 min.
D
F
Fo
MIN
D3
F3 Fo
MAX
F2
Fo
i
D2 D1 Dy
1
1
2
3
F1 Fy
Fi=K Di
1
polynomial
function
modified
polynomial
function
KP1
KP3
KP2
Fig. 12. Polynomial analytical model of bearings made of rubber.
D
F
1
1
Fi=K
1
Di
Yield
Point
Fi=K
1
Di
Fig. 13. Linear-elastic state at beginning of loading and at transition from
unloading to loading.
Fig. 14. (a) Dependence of the exponent on the horizontal deformation of the bearing and (b) exponential function.
(a)
(b)
Fig. 15. Post-elastic state at unloading for different exponents.
-0.06 -0.04 -0.02 0 0.02 0.04 0.06
D [m]
-30
-20
-10
0
10
20
30
F

[
k
N
]
analytical
biaxial test
spec-9
Fig. 16. Forcedeformation relationship obtained from test and the analytical
solution.
606 I. Gjorgjiev, M. Garevski / Engineering Structures 56 (2013) 600609
(a) (b)
-0.02 -0.01 0 0.01 0.02
D [m]
-10
-5
0
5
10
F

[
k
N
]
analytical
experiment
K
2
K
2
(D=8)
spec-9
-0.06 -0.04 -0.02 0 0.02 0.04 0.06
D [m]
-30
-20
-10
0
10
20
30
F

[
k
N
]
analytical
experiment
K
2
K
2
D=65
spec-9
Fig. 17. Forcedisplacement relationship in conditions of (a) low deformations (b) large deformation.
(a)
D [m]
-50
-25
0
25
50
F

[
k
N
]
analytical
experiment
spec-7
-45
-30
-15
0
15
30
45 analytical
experiment
spec-1
-30
-15
0
15
30
analytical
experiment
spec-10
-0.06 -0.04 -0.02 0 0.02 0.04 0.06
-0.075 -0.05 -0.025 0 0.025 0.05 0.075
-0.075 -0.05 -0.025 0 0.025 0.05 0.075
-0.075 -0.05 -0.025 0 0.025 0.05 0.075
-30
-20
-10
0
10
20
30
analytical
experiment
spec-12
10
5
0
-5
-10
-0.075 -0.05 -0.025 0.025 0 0.05 0.075
F

[
k
N
]
F

[
k
N
]
F

[
k
N
]
F

[
k
N
]
D [m]
D [m]
D [m]
D [m]
analytical
experiment
spec-11
(b)
(c)
(d)
(e)
Fig. 18. Verication of the analytical model with the most characteristic forcedeformation relationship obtained by experimental testing.
I. Gjorgjiev, M. Garevski / Engineering Structures 56 (2013) 600609 607
where eD
min
is exponent at deformation D
min
e
and eD
max
is exponent
at deformation D
max
e
.
This expression is valid only when deformation D
KP
i
is within the
range of [D
min
e
; D
max
e
]. In the case the deformation at the character-
istic point is beyond the above domain, then it is assigned the limit
values (D
min
e
or D
max
e
) of the interval (Fig. 14a). The graphical inter-
pretation of the dependence of eD
KP
i
on D
KP
i
is shown in Fig. 14a.
The computation of the decay coefcient kF
i
o
includes four
parameters (D
i
, D
KP
i
, eD
KP
i
and F
i
o
) that directly depend on the cur-
rent deformation of the bearing and the history of deformations.
The current deformation is included in parameter D
i
and its value
at unloading ranges between D
KP
i
and zero. The history of deforma-
tion of the bearing is included in parameters D
KP
i
and eD
KP
i
. Param-
eter D
KP
i
is the maximum achieved deformation at loading, while
parameter eD
KP
i
directly depends on D
KP
i
and it is calculated accord-
ing Eq. (7). Parameter F
i
o
is dened according to expression (3) and
depends on the current deformation and the history of
deformation.
From Eq. (6), it can be seen that the value of the decay coef-
cient kF
i
o
ranges within the following limits:
for
D
i
0 ) kF
i
o
0
D
i
D
KP
i
) kF
i
o
1
(
Accordingly:
when
D
i
0 then F
i
om
F
i
o
D
i
D
KP
i
then F
i
om
F
i
o
(
With this, it is proved that force F
i
om
satises both ultimate states:
When there is no deformation of the bearing, the force is equal
to the value of coefcient F
i
o
.
At the beginning of unloading, the force at loading and that at
unloading are equal.
To get an insight into the effect of eD
KP
i
upon the horizontal
force in the bearing, three curves with different values of
D
KP
i
2; 5 and 10 were derived (Fig. 15). Fig. 15 shows that, in
the case of curves with higher value of eD
KP
i
, horizontal force F dur-
ing unloading, decreases faster and more intensively. In all three
curves, the force has an identical value at the beginning and at
the end of unloading which points to the fact that the boundary
conditions are satised.
The subsequently presented graph in Fig. 15 shows the depen-
dence of kF
i
o
on the (D
i
=D
KP
i
) ratio. From the enclosed curves
(eD
KP
i
2; 5; and 10), it can be concluded that, at higher values of
eD
KP
i
, there is a considerable decay of the value of kF
i
o
when the cur-
rent displacement is within the limits of D
i
= [0.80 1.0] D
KP
i
. This
characteristic of kF
i
o
enables control over the total restoring force at
each deformed position.
4. Verication of the model
The aim of the verication was to prove the stability of the
model in different strain conditions. Three states were veried:
elastic, post-elastic at loading and post-elastic at unloading. To val-
idate the proposed model, the values of forces and displacements
obtained from the polynomial model were compared to the exper-
imental ones. The presented experimental results were obtained
from biaxial tests carried out on square and circular bearings pro-
duced by use of different rubber compounds. In horizontal direc-
tion, the bearings were loaded with harmonic excitation of both
constant and increasing amplitude (Fig. 5a and b).
The graph in Fig. 16 shows the forcedeformation relationship
obtained from the rubber bearing test spec-9 and that obtained
in the analytical solution. The dotted curve represents the data
from the test, whereas the solid curves represent the analytical
solution.
In order to investigate the behavior of the polynomial model at
low and large strains, the rst and the last two cycles for specimen
spec-9 were separated (Fig. 17a and b). From the presented com-
parison of the results, it could be concluded that the polynomial
model was not able to completely follow the behavior of the bear-
ing at low deformations. This was due to the parameters of the
polynomial model which were tted to large strains and the model
was not able to completely follow the behavior of the bearing at
low strains. In the case of large deformations (Fig. 17b), the polyno-
mial model completely followed the behavior of the bearing. The
other loading cycles (>2) were simulated by the analytical model
with an acceptable accuracy.
The applicability of the proposed analytical model for the bear-
ings produced of different rubber is presented through comparison
of four forcedisplacement relationships (Fig. 18a through
Fig. 18e). Fig. 18a shows the nonlinear behavior of the rubber with-
out damping, while Fig. 18be show different forms of behavior in
the case when the compound possesses a certain internal damping.
From the presented graphs, it can be concluded, that the poly-
nomial model is sufciently exible to be applied for modeling of
bearings made of different rubber compounds. The input parame-
ters of the proposed polynomial model cover sufciently well, the
main characteristics of behavior of rubber bearings. The model
shows satisfactory accuracy not only in the case of high damping
rubber bearings but also in the case of low damping rubber
bearings.
The stability and accuracy of the polynomial model was also
veried by nonlinear dynamic analysis of a seismically isolated
steel reservoir [11]. The time histories of acceleration analytically
obtained from several earthquake excitations [11] were compared
with the experimental ones, whereat the analytical model proved
to be sufciently precise.
5. Conclusions
This paper deals with production of rubber bearings, their test-
ing and development of an analytical model of a rubber bearing.
The process of production involved development of high damping
rubber compound and establishment of a stable bond between the
rubber and the steel surfaces. Vertical and biaxial tests were car-
ried out for square and circular bearings produced of 18 different
rubber compounds. Based on the performed horizontal tests, a
database was created and used to develop a polynomial analytical
model of a rubber bearing. From the tests on the bearings, it was
concluded that the behavior of the bearings was bilinear until
moderate shear strain. At higher strains, the rubber bearings be-
haved in the nonlinear range and were characterized by an in-
creased stiffness. It was also concluded that the horizontal
stiffness at zero horizontal deformation of the bearing depended
on the loading history. In the case of change of horizontal deforma-
tion at the beginning of the unloading, the force at zero deforma-
tion changes, as well.
The proposed analytical model of a rubber bearing is based on a
polynomial function. The behavior of the rubber bearing is pre-
sented through eight parameters plus the polynomial coefcients.
The analytical model and the methodology presented in this paper
are based on establishment of empirical model parameters by
matching the actual bearing test results. This analytical model
was developed on the basis of the results of experimental tests
on bearings made of different rubber compounds. The model in-
cluded large strain behavior and loading history effect. The Mullins
effect, the strain rate and the dependence on the axial load were
not taken into account in these investigations.
608 I. Gjorgjiev, M. Garevski / Engineering Structures 56 (2013) 600609
The parameters of the material for the proposed model were de-
ned through a simple biaxial test on a rubber isolator. Using the
least square method, the coefcients of the polynomial function
were dened. From the performed comparison between the ana-
lytical and the experimental results, it can be concluded that the
polynomial model is exible to be applied for modeling of isolators
made of different rubber compounds.
The proposed analytical model can be used instead of bilinear
simplication of rubber bearing for the purpose of better simula-
tion of nonlinear behavior of rubber. The system property modi-
cation factors for the effects of aging, temperature and scragging
dened in [1] code provisions can be included in this model by
changing the polynomial coefcients.
Acknowledgement
The authors are indebted to Prof. James Kelly from the Univer-
sity at Berkeley, Civil Engineering Department, California, for the
unselsh professional assistance provided during the realization
of the NATO SfP project 978028. Gratitude is also accorded for
the nancial assistance obtained through the NATO Science for
Peace Programme which enabled the performance of the presented
ample investigations.
References
[1] AASHTO. Guide specications for seismic isolation design. 3rd ed. Washington,
DC: American Association of State Highway and Transportation Ofcials; 2010.
[2] Bouc T. Modele mathematique dhysteresis. Acustica 1971;24:1625.
[3] Dall Asta A, Ragni A. Experimental tests and analytical model of high damping
rubber dissipating devices. Eng Struct 2006;28:187484.
[4] Grant Damian N, Fenves Gregory L, Auricchio Ferdinando. Bridge isolation with
high-damping rubber bearings analytical modelling and system response. In:
13th World conference on earthquake engineering (p. Paper No. 1002).
Vancouver, BC, Canada: 13WCEE Secretariat; 2004.
[5] EC8. Eurocode 8: Design of structures for earthquake resistance Part 1: General
rules, seismic actions and rules for buildings (EN 19981:2004). European
Committee for Standardization; 2004.
[6] Fujita T. Seismic isolation of civil buildings in Japan. Prog Struct Eng Mater
1998;1(3):295300.
[7] Fujita, T. Demonstration of effectiveness of seismic isolation in the Hanshin-
Awaji earthquake and progress of applications of base-isolated buildings. In:
Report on 1995 kobe earthquake by INCEDE ERC and KOBnet IIS. Tokyo:
University of Tokyo; 1999.
[8] Garevski M. Development, production and implementation of low cost rubber
bearings. In: Ansal A, Garevski M, editors. Earthquake engineering in Europe:
geotechnical, geological, and earthquake engineering, vol. 17, Part 5. Springer;
2010. p. 41137.
[9] Garevski Mihail, Kelly James. Development of low-cost rubber bearings for
seismic safety of structures in macedonia and balkan. Skopje: IZIIS; 2008.
[10] Gjorgjiev I. Analytical model for fabric reinforced rubber bearings based on
experimental results. Ss Cyril and Methodius. Skopje: Institute for Earthquake
Engineering and Engineering Seismology (IZIIS); 2011.
[11] Gjorgjiev Igor, Garevski Mihail. Monitoring of rubber bearing behavior during
shaking table tests. In: Garevski M, editor. Earthkuake and health monitoring
of civil structures. Springer Science+Business, Media; 2013. p. 287306.
[12] Govindjee S, Simo J. Transition from micro-mechanics to computationally
efcient phenomenology: carbon black lled rubbers incorporating mullins
effect. J Mech Phys Solids 1992;40(1):21333.
[13] Hwang JS, Wu JD, Pan TC, Yang G. A mathematical hysteretic model for
elastomeric isolation bearings. Earthquake Eng Struct Dynam 2002:77189.
[14] Jankowski R. Nonlinear rate dependent model of high damping rubber bearing.
Bull Earthquake Eng 2003:397403.
[15] Kelly JM. Basic concepts and potential applications for base isolation. ASCE
Struct Congress; 1986.
[16] Kelly JM. Base isolation/passive control theoretical aspects. Structural
applications of protective. Buffalo, New York: National Center for Earthquake
Engineering Research; 1987.
[17] Kikuchi M, Aiken ID. An analytical hysteresis model for elastomeric seismic
isolation bearings. Earthquake Eng Struct Dynam 1997;26:21531.
[18] Melkumyan M. Armenia is the one of the world leaders in development and
application of base isolation technologies. MENSHIN J 2006;54:3841.
[19] Mullins L. Softening of rubber by deformation. Rubber Chem Technol
1969:33962.
[20] Naeim F, Kelly JM. Design of seismic isolated structures: from theory to
practice. John Wiley & Sons; 1999.
[21] Skinner IR, Robinson WH, McVerry GH. An introduction to seismic
isolation. John Wiley & Sons; 1993.
[22] Staudacher K. Protection for structures in extremeearthquakes:
fullbaseisolation (3-D) by the Swiss seismaoat system. Nucl Eng Des
1985;84(3):34357.
[23] Tsai CS, Chiang T, Chen B, Lin Sh. An advanced analytical model for high
damping rubber bearings. Earthquake Eng Struct Dynam 2003:137387.
I. Gjorgjiev, M. Garevski / Engineering Structures 56 (2013) 600609 609