Engineering Structures 40 (2012) 198204

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Critical load of slender elastomeric seismic isolators: An experimental perspective

Donatello Cardone , Giuseppe Perrone
DiSGG, University of Basilicata, Via Ateneo Lucano 10, 85100 Potenza, Italy

a r t i c l e

i n f o

Article history:
Received 27 July 2011
Revised 31 January 2012
Accepted 5 February 2012
Available online 28 March 2012
Keywords:
Seismic isolation
Elastomeric bearings
Critical load
Shear strain
Buckling

a b s t r a c t
One of the most important aspects of the seismic response of elastomeric isolators is their stability under
large shear strains. The bearing capacity of elastomeric isolators, indeed, progressively degrades while
increasing horizontal displacement. This may greatly inuence the design of elastomeric isolators, especially in high seismicity regions, where slender elastomeric isolators subjected to large horizontal displacements are a common practice. In the current design approach the critical load is evaluated based
on the Haringx theory, modied to account for large shear strains by approximate correction factors.
In this paper the critical behavior of a pair of slender elastomeric devices is experimentally evaluated at
different strain amplitudes, ranging from approximately 50% to 150%. The experimental results are then
compared to the predictions of a number of semi-empirical and theoretical formulations.
The main conclusion of this study is that current design approaches are overly conservative for slender
elastomeric seismic isolators, since they underestimate their critical load capacity at moderate-to-large
shear strain amplitudes.
2012 Elsevier Ltd. All rights reserved.

1. Introduction
An elastomeric isolation bearing consists of a number of rubber
and steel layers mutually vulcanized, to provide high stiffness in
the vertical direction together with large deformability in the horizontal direction. The elastomeric isolators work like a lter lengthening the fundamental period of vibration of the structure, thus
reducing the seismic effects (interstory drifts, oor accelerations,
stresses in the structural members, etc.) generated in the superstructure. However, this reduction is accompanied by large horizontal displacements in the isolators, which may signicantly
reduce their axial load capacity [14].
The earliest theoretical approach for the evaluation of the critical
axial load of rubber bearings was introduced by Haringx [5], considering the mechanical characteristics of helical steel springs and rubber rods. Same assumptions have been made later by Gent [6]
considering multilayered rubber compression springs. Basically,
the Haringxs theory is based on a linear one-dimensional beam
model with shear deformability, within the hypothesis of small
displacements. The critical buckling load of elastomeric seismic
isolators is expressed as:

Pcr;0

2  PE
s
EIeff
1 1 4p2
GAs eff  L2

Corresponding author. Tel.: +39 0971205054.

E-mail addresses: donatello.cardone@unibas.it (D. Cardone), giuseppe.perr@
alice.it (G. Perrone).
0141-0296/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2012.02.031

in which: GAs eff and EIeff are the effective shear rigidity and effective exural rigidity, respectively, of the elastomeric isolators, computed based on the bending modulus (E) and dynamic shear
modulus (Gdyn) of rubber, moment of inertia of the bearing about
the axis of bending (I) and bonded rubber area (As);
PE is the Euler load for a standard elastic column:

PE

p2  EIeff

L2

L is the total height of rubber layers and steel plates excluding top
and bottom connecting steel plates.
Various authors proposed different relations to evaluate the
effective shear and exural rigidity of laminated rubber bearings.
In this paper, reference to the formula derived by Buckle and Kelly
[1], Koh and Kelly [2] has been made:

GAs eff Gdyn  As 

EIeff Er I 

L
te

L
te

where te is the total thickness of the rubber layers and Er is the elastic modulus of the rubber bearing evaluated based on the primary
shape factor S1 and rubber Youngs modulus E0 as:

Er E0 1 0:742  S21

The primary shape factor S1 is dened as the ratio between the

loaded area of the bearing and the area free to bulge of the single
rubber layer (S1  D/4ti for circular bearings, where D is the diameter

199

D. Cardone, G. Perrone / Engineering Structures 40 (2012) 198204

of the isolator and ti the thickness of a single rubber layer). The rubber Youngs modulus E0 is usually taken equal to 3.3 Gdyn to 4 Gdyn.
The Haringxs theory has been later applied by Naeim and Kelly
[7], with a series of simplied assumptions, for commercial elastomeric seismic isolators. According to Naeim and Kelly [7], the critical
buckling load of elastomeric seismic isolators can be expressed in
terms of the primary and secondary shape factors S1 and S2, the latter
being dened as the ratio between the maximum dimension of the
cross section of the isolator and the total height of rubber. For circular elastomeric isolator, for instance, Naeim and Kelly [7] provides:

Pcr;0 p  GAs eff  S1  S2

2 2

Pcr;0 p  GAs eff

2 3

u
P

r
0:742  E0
 S1  S2

G

(b)

2  PE
s



EI

EI

eff
eff
1 1 4p2  EA L2 GA L2
eff

L
te

In Italy, the current design approach [10] refers to a formulation of

the critical load similar to (but more conservative than) that initially proposed by Naeim and Kelly [7] (see Eq. (6)):

Pcr;0 Gdyn  As  S1  S2

11

For all the above mentioned formulations, the buckling load at the
target shear displacement (u) is evaluated as a function of the ratio
between the effective area of the inner shim plate (A) and the overlap area of the displaced bearing (Ar) (see Fig. 1):

Pcr Pcr;0 

Ar
A

12

For circular bearings, for instance, the overlap area at the target
displacement is given by:

Ar u  sin u 
with

u 2 arccos

u
D

D2
4

u
Fig. 1. (a) Schematic deformed shape of an elastomeric bearing subjected to shear
and compression; (b) effective cross section area as a function of shear
displacement.

[12,1,3,4,13]. In this paper, the critical load of a couple of slender

(low shape factors) elastomeric bearings is experimentally evaluated. Test set-up and experimental program are presented rst in
detail. Then, the experimental results are compared with the predictions of the theoretical formulations presented in the previous
paragraph.
2. Experimental tests

10

where Gdyn is the dynamic shear modulus derived from the qualication tests of the elastomeric isolator.
More recently, a less conservative variant of the Naeim and
Kelly formulation has been adopted in the new European Standard
EN11529 [11]:

Pcr;0 1:3 Gdyn  As  S1  S2

Ar

s eff

where EAeff is the effective axial stiffness of the rubber bearing,

evaluated as:

EAeff Er  A 

The secondary shape factor S2 is dened as the ratio between the

bearing maximum dimension and the total thickness of all the rubber layers (S2  D/te for circular bearings, where te is the total thickness of all the rubber layers). It is interesting to note that the critical
buckling load capacity evaluated considering the expressions (6)
and (7), differs by 1020%, depending on the value (between 3.3G
and 4G) assumed for the rubber Youngs modulus.
Lanzo [9] modied the Haringxs expression by taking into account the axial stiffness of the rubber bearing (EA)eff:

Pcr;0

P
F

Subsequently, Kelly [8] derived a more rened formulation of the

buckling load of elastomeric isolator:

(a)

13

14

Several experimental studies of the buckling behavior of elastomeric seismic isolators have been carried out in the past

2.1. Test specimens

Test specimens are a couple of 1:2 scaled circular elastomeric
bearings with 200 mm diameter and 10 rubber layers with 8 mm
thickness. Bearing geometrical properties are summarized in
Table 1.
The mechanical properties of the specimens have been derived
from a number of standard cyclic tests, performed in accordance
with the test procedure prescribed in the Italian seismic code
[10] for the qualication of elastomeric bearings. The static shear
modulus (Gstat), in particular, has been derived from a quasi-static
test consisting of ve cycles at 0.1 Hz frequency of loading and
100% shear strain amplitude. According to the NTC 2008 [10], Gstat

Table 1
Elastomeric bearings details.
Outer diameter
Inner diameter
Rubber layer thickness
Number of rubber layers
Steel shim thickness
Number of steel shims
Total height of rubber
Primary shape factor
Secondary shape factor

De (mm)
D (mm)
ti (mm)
nti
ts (mm)
nti
te (mm)
S1
S2

200
180
8
10
2
9
80
5.63
2.25

200

D. Cardone, G. Perrone / Engineering Structures 40 (2012) 198204

(a)

(b)

Fig. 2. Hysteresis cycles of the test specimens at 100% shear strain amplitude
during (a) quasi-static test for the evaluation of Gstat and (b) dynamic test for the
evaluation of Gdyn and neq.

is dened as the slope of the shearstrain curve between 27% and

58% shear deformation with reference to the third cycle of the test
(see Fig. 2a). In the case under consideration, a value of Gstat =
0.113 MPa has been obtained. The dynamic shear modulus (Gdyn)
and the equivalent damping (neq) have been derived from a
dynamic test consisting of ve cycles at 0.5 Hz frequency of loading
and 100% shear strain amplitude with reference to the third cycle
of the test (see Fig. 2b). According to the NTC2008 [10], Gdyn is
dened as the secant stiffness to the origin of the axis of the cycle
of maximum strain amplitude, neq is evaluated as the ratio between
the energy dissipated in one complete cycle (equal to the area
enclosed in the cycle) and the strain energy stored at the maximum
strain amplitude (see Fig. 2b). In the case under consideration, a
value of Gdyn = 0.37 MPa and neq = 16% have been obtained.
2.2. Test set-up
The specimens have been tested using the uniaxial bearing test
facility available at the Laboratory of Structures at the University of
Basilicata (see Fig. 3). The test apparatus constitutes a selfbalanced system designed for cyclic testing of a pair of identical
counteracting specimens. The apparatus subjects the test specimens to axial loads in the horizontal direction and shear forces
and displacements in the vertical direction. The test rig consists
of two steel stiff beams connected by three steel columns. Two columns are linked to the base beam through a double pendulum
hinge while the third column (on the left in Fig. 3) by a simple
hinge. This has been purposely done to preserve the load directions

Fig. 3. Test set-up.

and the co-planarity of the bases of the specimens during the test.
The hinge connections accommodate the axial displacement
resulting from the axial load being applied to the specimens.
The shear displacement is vertically applied by a Schenk
hydraulic actuator, connected to the specimens through a Tshaped steel plate. The actuator is served by two pumps, each with
80 l/min maximum capacity, and equipped with two servo-valves
with maximum capacity of 63 l/min each. The Shenck actuator
can apply 250 kN maximum force and 125 mm maximum displacement. The force of the actuator is measured by a load cell
mounted on the top of the cylinder. The displacement of the actuator is measured and controlled by an internal transducer.
The compression load is horizontally applied and kept constant
during the tests by an Enerpac double-effect hydraulic jack (see
Fig. 3). The hydraulic jack is served by a pump, with 5 l/min capacity, and equipped with an electro-valve with 5 l/min maximum
capacity. The hydraulic jack can apply 300 kN maximum force.
The force of the hydraulic jack is monitored and measured by a
number of load cells of different capacity mounted on the top of
the cylinder.
The shear displacements of the specimens have been measured
by two linear transducers connected to the T-shaped steel plate.
The axial displacements of the specimens have been measured
by a couple of linear transducers, positioned between the corner
of the outer plates of the two specimens, in order to evaluate possible rotations of the specimens around the out-of-plane axis.

201

D. Cardone, G. Perrone / Engineering Structures 40 (2012) 198204

interval of 10 s was allowed to elapse between one step and the

next. The test control has been carried out manually in order to
prevent, in real time, possible damages to either specimens or test
apparatus.
A total number of 6 stability test have been performed (see
Table 2), while increasing the target shear displacement from 48
to 125 mm, corresponding to six different shear strain amplitude,
equal to 60%, 80%, 100%, 120%, 140% and 160%, respectively.

u (mm)

90

45

The rst direct result of the critical load tests is the time history
of the shear force F resulting from the application of an increasing
axial load P under a given shear displacement u. As an example,
Fig. 5 shows the shear forcetime history derived from Test 3
(see Table 2), featuring an imposed shear deformation of 100%.
From Fig. 5 it is evident that as the axial force P is increased, the
shear force F decreases until it becomes negative, while the shear
displacement remains constant.
The experimental outcomes have been employed to evaluate
the critical buckling load for each imposed shear deformation. It
is worth observing that the critical load cannot be simply estimated as the value of the axial load at which the horizontal shear
force is zero because this would give a constrained evaluation of
the critical load, since the critical load is approached while the
bearing is locked against further horizontal displacements. In the
common practice (e.g. during a seismic event), instead, the horizontal displacement is unconstrained and free to increase, as necessary, when buckling takes place. Reference to the procedure
proposed by Nagarajaiah and Ferrell [4], based on equilibrium
paths, has been made to determine the unconstrained critical load.
Fig. 6 shows the variation of the shear force F with increasing
axial load P, for each test performed at a given shear deformation
c = 60%, 80%, 100%, 120%, 140% and 160%.
The diagram of Fig. 6 has been entered alternatively with a
number of F-lines and P-lines to determine a series of equilibrium
paths (i.e. a number of smooth curves passing through discrete
points) corresponding to given levels of shear force (Fig. 7) and
axial load (Fig. 9), respectively. The equilibrium paths are unstable
past a limit point, which is the critical load Pcr. This observation can
be exploited to estimate the critical load of the rubber bearings under consideration, as explained below.
Fig. 7 shows the axial load vs. shear strain curves, corresponding
to different levels of shear force in the specimens. The critical load
can be identied with the value of the axial load that must be reduced, to maintain constant the shear force in the specimens by
further increments of shear displacement. As expected, the critical
load decreases while increasing the horizontal displacement, hence
shear strain, applied (see Fig. 8).
Fig. 9 shows the shear force vs. shear strain curves, corresponding to different values of axial load applied to the specimens. As
can be seen, each curve passes through a maximum as the horizontal displacement increases, under constant axial load. The shear

P (KN)

300

2.4. Test results

150

Fig. 4. Steps of a typical stability test (for more details see Table 2).

The acquisition frequency has been set equal to 10 Hz. The signals recorded during the tests have been ltered, amplied and
converted from analog signals to 16 bit digital signals in real time
by a proper signal processing software.
2.3. Test procedure
The testing procedure followed in this study is similar, to some
extent, to that adopted by Buckle et al. [13]. All the stability tests,
in particular, were performed on the same couple of specimens, at
different levels of shear deformations.
Basically, each test consisted in the application of two consecutive ramps of axial load with a ramp of shear displacement
between them (see Fig. 4). More precisely, each test can be divided
in the following six steps:
1. Applying an initial axial load of 150 kN, corresponding to a
design compression stress of 6 MPa.
2. Applying a given target displacement (hence shear deformation).
3. Increasing the axial load until the horizontal force became
negative.
4. Decreasing the axial load to the initial value of 150 kN.
5. Decreasing the initial displacement to 0.
6. Decreasing the axial load to 0.
Both forces and displacements have been applied very slowly
during the tests, in order to avoid dynamic effects. Moreover, an

Table 2
Stability test details (see Fig. 4).
Test no.

1
2
3
4
5
6

u1 (mm)

48
64
80
96
112
125

c (%)
60
80
100
120
140
160

P1 (kN)

150
150
150
150
150
150

P2 (kN)

280
280
280
280
200
200

Step 1

Step 2

t1 (s)

t2 (s)

t3 (s)

t4 (s)

Step 3
t5 (s)

t6 (s)

Step 4
t7 (s)

t8 (s)

t9 (s)

Step 5
t10 (s)

t11 (s)

Step 6

15
15
15
15
15
15

25
25
25
25
25
25

35
38
42
45
48
51

45
48
52
55
58
61

58
61
65
68
63
66

68
71
75
78
73
76

81
84
88
91
78
81

91
94
98
101
88
91

101
108
114
121
112
117

111
118
124
131
122
127

126
133
139
146
137
142

202

D. Cardone, G. Perrone / Engineering Structures 40 (2012) 198204

125

u (mm)

(a)

100
75
50
25

t (s)

0
0

300

100

150

200

250

300

350

P (KN)

(b)

50

200

100
t (s)

0
0

100

150

200

250

300

350

20

F (KN)

(c)

50

10

t (s)

0
0

50

100

150

200

250

300

350

-10
-20

220
F = 2.2 kN

P (kN)

235

215

200

Axial load (kN)

Fig. 5. Typical results of stability test: shear displacement, axial load and shear force time histories.

195

180
175
P = 162 kN

F=0.2KN
F=1.3KN
F=1.5KN
F=2.2KN
F=4.5KN

160

155

140
40

135
0

7
F (kN)

Fig. 6. Axial load vs. shear force variations as a function of the imposed shear strain
amplitude.

force and shear strain at which the maximum occurs decrease with
increasing axial load. Finally, the horizontal tangent stiffness tends
to zero for shear strains lower than the maximum test amplitude;
moreover it decreases while increasing axial load and horizontal
displacement.

Shear strain (%)

80

120

160

Fig. 7. Axial load vs. shear strain curves as a function of shear force.

2.5. Comparison between analytical and experimental results

Fig. 10 compares the experimental values of Pcr derived following the equilibrium path procedure described in the last section,
to the numerical results provided by the formula proposed by
Haringx [5] (Eq. (1)), Naeim and Kelly [7] (Eq. (6)), Kelly [8] (Eq.
(7)), Lanzo [9] (Eq. (8)), and those adopted in the Italian seismic
code [10] (Eq. (10)) and in the EN15129 European standard [11]

D. Cardone, G. Perrone / Engineering Structures 40 (2012) 198204

Pcr (kN)

300

experimental values

200

100

Shear strain (%)

0
60

80

100

120

140

160

Fig. 8. Critical load as a function of shear strain amplitude.

Shear force(kN)

1
Shear strain(%)

0
0

40

80

120

160

300

Pcr (kN)

Fig. 9. Shear forcestrain curves as a function of axial load.

203

results, while the formula proposed by Kelly provides the highest

values of critical load. As can be seen, for the slender elastomeric
bearings under consideration (with secondary shape factor less
than 3 and rubber thickness of 8 mm), the experimental values
of Pcr result signicantly greater than those predicted by the theoretical formulas, especially at large shear amplitudes (100150%)
where values approximately 2.53.5 times greater than expected
are found. The relationship adopted in the current Italian seismic
code appears to be too conservative for low shape factors elastomeric bearings (S1 < 12, S2 < 3), such as those considered in this
study. The formulation of Pcr proposed in the new EN11529 European Standard, being less conservative and fully compatible with
many theoretical formulations, seems to be more suitable for slender elastomeric isolators.
3. Conclusions
In this paper the critical behavior of a couple of slender elastomeric seismic isolators has been experimentally evaluated at different strain amplitudes, ranging approximately from 50% to
160%. The tests were performed with the specic objectives of
nding the effect of shear strain on the critical load of slender
(low shape factors) elastomeric seismic isolators and to evaluate
the existing design approaches. During the tests, the shear displacement of the specimens was held at a specied value while
the axial load was progressively increased until critical load conditions occurred. For the purpose of this experiment, the isolator was
considered to be in critical state when the horizontal force became
zero or negative. The critical load has been evaluated, for each
strain amplitude, based on a series of equilibrium paths derived
from the experimental results.
Based on the results of this study the following conclusions can
be drawn:
(i) The critical load decreases with increasing shear strain.
(ii) The horizontal stiffness decreases with increasing axial load
and horizontal displacement.
(iii) Current design procedures seem to be overly conservative
for slender (low shape factor) elastomeric isolators because
they underestimate the experimentally determined critical
load by a factor of approximately 1.5 at low shear strain
amplitudes and approximately 3.5 at large shear strain
amplitudes.
In the near future, it is desirable to extend this experimental
investigation considering more specimens, characterized by different shape factors and/or different rubber thickness, in order to
evaluate the inuence of such parameters on the critical behavior
of slender elastomeric seismic isolators.

200

Acknowledgements
The authors are grateful to Domenico Nigro (University of Basilicata) for his valuable help in setting up the testing apparatus and
executing the experimental tests. This work has been partially
funded by MIUR, COFIN 2007.

100

Shear strain(%)

0
60

80

100

120

140

Fig. 10. Comparison between experimental results and critical loads expected
based on different theoretical and semi-empirical formulations.

(Eq. (11)). Among the theoretical formulations considered, the

formula proposed by Haringx leads to the most conservative

References
[1] Buckle IG, Kelly JM. Properties of slender elastomeric isolation bearings during
shake table studies of a large-scale model bridge deck. In: Joint sealing and
bearing systems for concrete structures (American Concrete Institute), vol. 1. p.
24769; 1986.
[2] Koh CG, Kelly JM. Effects of axial load on elastomeric bearings. Earthquake
Engineering Research Center, Rep. UCB/EERC-86/12, Univ. of California,
Berkeley, 1986.
[3] Buckle IG, Liu H. Experimental determination of critical loads of elastomeric
isolators at high shear strain. NCEER Bull 1994;8(3):15.

204

D. Cardone, G. Perrone / Engineering Structures 40 (2012) 198204

[4] Nagarajaiah S, Ferrell K. Stability of elastomeric seismic isolation bearings. J

Struct Eng 1999;125(9):94654.
[5] Haringx JA. On highly compressive helical springs and rubber rods and their
applications for vibration-free mountings. I. Philips Res Rep 1948;3:40149.
[6] Gent AN. Elastic stability of rubber compression springs. J Mech Eng Sci
1964;6(4):31826.
[7] Naeim F, Kelly JM. Design of seismic isolated structuresfrom theory to
practice. New York: John Wiley & Sons; 1999.
[8] Kelly TE. Base isolation of structures: design guidelines. Wellington, New
Zealand: Holmes Consulting Group Ltd.; 2001.

[9] Lanzo AD. On elastic beam models for stability analysis of multilayered rubber
bearings. Int J Solids Struct 2004;41(20):573357.
[10] NTC 2008 Norme Tecniche per le Costruzioni, D.M. 14/01/08, Rome, 2008 [in
Italian].
[11] EN15129 European Standard EN 15129: anti-seismic devices, 2009.
[12] Aiken D, Kelly JM, Tajirian FF. Mechanics of low shape factor elastomeric
seismic isolation bearings. Report no. UCB/EERC-89/13, 1989.
[13] Buckle I, Nagarajaiah S, Ferrell K. Stability of elastomeric isolation bearings:
experimental study. J Struct Eng 2002;128(1):311.