EERI DISTINGUISHED LECTURE 2001

Seismic Isolation Systems for Developing

Countries
James M. Kelly,a) M.EERI
This paper describes an experimental and theoretical study of the feasibility of using fiber reinforcement to produce lightweight low-cost elastomeric isolators for application to housing, schools and other public buildings
in highly seismic areas of the developing world. The theoretical analysis covers the mechanical characteristics of multi-layer elastomeric isolation bearings where the reinforcing elements, normally steel plates, are replaced by a
fiber reinforcement. The fiber in the fiber-reinforced isolator, in contrast to
the steel in the conventional isolator (which is assumed to be rigid both in
extension and flexure), is assumed to be flexible in extension, but completely
without flexure rigidity. This leads to an extension of the theoretical analysis
on which the design of steel-reinforced isolators is which accommodates the
stretching of the fiber-reinforcement. Several examples of isolators in the
form of long strips were tested at the Earthquake Engineering Research Center Laboratory. The tested isolators had significantly large shape factors, large
enough that for conventional isolators the effects of material compressibility
would need to be included. The theoretical analysis is extended to include
compressibility and the competing influences of reinforcement flexibility and
compressibility are studied. The theoretical analysis suggests and the test results confirm that it is possible to produce a fiber-reinforced strip isolator
that matches the behavior of a steel-reinforced isolator. The fiber-reinforced
isolator is significantly lighter and can be made by a much less laborintensive manufacturing process. The advantage of the strip isolator is that it
can be easily used in buildings with masonry walls. The intention of this research is to provide a low-cost lightweight isolation system for housing and
public buildings in developing countries. [DOI: 10.1193/1.1503339]
INTRODUCTION
The recent earthquakes in India, Turkey, and South America have again emphasized
the fact that the major loss of life in earthquakes happens when the event occurs in developing countries. Even in relatively moderate earthquakes in areas with poor housing,
many people are killed by the collapse of brittle, heavy, unreinforced masonry or poorly
constructed concrete buildings. Modern structural control technologies such as active
control or energy dissipation devices can do little to alleviate this, but it is possible that

a)

Univ. of California, Pacific Earthquake Engineering Research Center, 1301 S. 46th St., Richmond, CA 94804

385

Earthquake Spectra, Volume 18, No. 3, pages 385406, August 2002; 2002, Earthquake Engineering Research Institute

386

J. M. KELLY

seismic isolation could be adapted to improve the seismic resistance of poor housing and
other buildings such as schools and hospitals in developing countries.
The theory of seismic isolation (Naeim and Kelly 1999) shows that the reduction of
seismic loading by an isolation system depends primarily on the ratio of the isolation
period to the fixed-base period. Since the fixed-base period of a masonry block or brick
building may be around 1/10 second, an isolation period of 1 second or longer would
significantly reduce the seismic loads on the building and would not require a large isolation displacement. For example, the current UBC code for seismic isolation (ICBO
1997) has a formula for minimum isolator displacement which, for a 1.5 second system,
would be around 15 cm (6 inches).
The problem with adapting isolation to developing countries is that conventional isolators are large, expensive, and heavy. An individual isolator can weigh one ton or more.
To extend this earthquake-resistant strategy to housing and commercial buildings, the
cost and weight of the isolators must be reduced.
The primary weight in an isolator is that of the steel reinforcing plates used to provide the vertical stiffness of the rubber-steel composite element. A typical rubber isolator has two large end-plates around 25 mm (1 inch) thick and 20 thin reinforcing plates
around 3 mm (1/8 inch) thick. The high cost of producing the isolators reflects the labor
involved in preparing the steel plates and laying-up of the rubber sheets and steel plates
for vulcanization bonding in a mold. The steel plates are cut, sand blasted, acid cleaned,
and then coated with bonding compound. Next, the compounded rubber sheets with the
interleaved steel plates are put into a mold and heated under pressure for several hours to
complete the manufacturing process. Both the weight and the cost of isolators could be
reduced if the steel reinforcing plates were eliminated and replaced by fiber reinforcement. As fiber materials are available with an elastic stiffness of the same order as steel,
the reinforcement needed to provide the vertical stiffness may be obtained by using a
similar volume of very much lighter material. The cost savings may be possible if the
use of fiber allows a simpler, less labor-intensive manufacturing process.
If fiber reinforcement were used, it would then be possible to build isolators in long
rectangular strips, with individual isolators cut to the required size. All isolators are currently manufactured as either circular or square. Rectangular isolators in the form of
long strips would have distinct advantages over square or circular isolators when applied
to buildings where the lateral-resisting system is walls. When isolation is applied to
buildings with structural walls, additional wall beams are needed to carry the wall from
isolator to isolator. A strip isolator would have a distinct advantage for retrofitting masonry structures and for isolating residential housing constructed from concrete or masonry blocks.
The vertical stiffness of a steel-reinforced bearing is approximated by assuming that
each individual pad in the bearing deforms in such a way that horizontal planes remain
horizontal and points on a vertical line lie on a parabola after loading. The plates are
assumed to constrain the displacement at the top and bottom of the pad. Linear elastic
behavior with incompressibility is assumed, with the additional assumption that the normal stress components are approximated by the pressure. This leads to the well-known
pressure solution that is generally accepted as an adequate approximate approach for

EERI DISTINGUISHED LECTURE 2001: SEISMIC ISOLATION SYSTEMS FOR DEVELOPING COUNTRIES

387

calculating the vertical stiffness. The extensional flexibility of the fiber reinforcement
can be incorporated into this approach, and the resulting vertical stiffness calculated.
A number of carbon fiber-reinforced rubber strip isolators were tested on a small
isolator test machine. The tests show that the concept is viable. The vertical and horizontal stiffnesses of the strip isolator are less than those for the equivalent steelreinforced isolator, but still adequate, and they proved to be easy to cut with a standard
saw, in contrast to steel-reinforced isolators that are difficult to cut and need special
saws. They were light and in use could be put in place without the use of lifting equipment.
Much recent discussion has focused on so-called smart rubber bearings or intelligent
base isolation systems as the new thrust in seismic isolation research. While there may
be a role for these adaptive systems for large expensive buildings in advanced economies, the development of lightweight, low-cost isolators is crucial if this method of seismic protection is to be applied to a wide range of buildings, such as housing, schools,
and medical centers in earthquake-prone areas of the world.
VERTICAL STIFFNESS OF FIBER-REINFORCED BEARINGS
The essential characteristic of the elastomeric isolator is the very large ratio of the
vertical stiffness relative to the horizontal stiffness. This is produced by the reinforcing
plates, which in current industry standard are thin steel plates. These plates prevent lateral bulging of the rubber, but allow the rubber to shear freely. The vertical stiffness can
be several hundred times the horizontal stiffness. The steel reinforcement has a similar
effect on the resistance of the isolator to bending moments, referred to as the bending
stiffness. This important design quantity makes the isolator stable against large vertical
loads.
COMPRESSION OF PAD WITH RIGID REINFORCEMENT

A linear elastic theory is the most common method used to predict the compression
and the bending stiffness of a thin elastomeric pad. The first analysis of the compression
stiffness was done using an energy approach by Rocard (1937); further developments
were made by Gent and Lindley (1959) and Gent and Meinecke (1970). A very detailed
description of the theory is given by Kelly (1996).
The analysis is an approximate one based on a number of assumptions. The kinematic assumptions are as follows:
(i)

points on a vertical line before deformation lie on a parabola after loading

(ii) horizontal planes remain horizontal

We consider an arbitrarily shaped pad of thickness t and locate a rectangular Cartesian
coordinate system, (x,y,z), in the middle surface of the pad, as shown in Figure 1a. Figure 1b shows the displacements, (u,v,w), in the coordinate directions under assumptions
(i) and (ii):

388

J. M. KELLY

Figure 1. Coordinate system (a) and constrained rubber pad (b).

ux,y,zu0x,y 1

4z2
t2

vx,y,zv0x,y 1

4z2
t2

wx,y,zwz

(1)

This displacement field satisfies the constraint that the top and bottom surfaces of the
pad are bonded to rigid substrates. The assumption of incompressibility produces a further constraint on the three components of strain, xx , yy , zz , in the form

xxyyzz0

(2)

and this leads to

4z2
u0,xv0,y 1 2 w, z0
t
where the commas imply partial differentiation with respect to the indicated coordinate.
When integrated through the thickness this gives

3 3

2t 2 c

u0,xv0,y

(3)

where the change of thickness of the pad is (0 in compression), and c/t is the
compression strain.

EERI DISTINGUISHED LECTURE 2001: SEISMIC ISOLATION SYSTEMS FOR DEVELOPING COUNTRIES

389

The other assumptions of the theory are that the material is incompressible and that
the stress state is dominated by the pressure, p, in the sense that the normal stress components can be taken as p. The vertical shear stress components are included but the
in-plane shear stress is assumed to be negligible. Linear elastic behavior is assumed.
These assumptions and the equations of stress equilibrium lead to the pressure solution

p, xxp,yy2p

12G
12G
3 2 c
t
t

(4)

The boundary condition, p0, on the perimeter of the pad completes the system for the
pressure distribution, p(x,y), across the pad.
The desired result is the effective compression modulus Ec of the pad. This is obtained by computing, p(x,y), in terms of the compression strain c , integrating, p(x,y),
over the area A of the pad to determine the resultant load P. The effective compression
modulus Ec is then given by

P
Ec
Ac

(5)

The value of Ec for a single rubber layer is controlled by the shape factor, S, defined as

loaded area
free area

which is a dimensionless measure of the aspect ratio of the single layer of the elastomer.
For example, in an infinite strip of width 2b, and with a single layer thickness of t, S
b/t, and for a circular pad of diameter and thickness t, S/(4t), and for a square
pad of side a and thickness t, Sa/(4t).
The vertical stiffness, Kv , of a rubber bearing made of n pads is given by

EcA
tr

Kv

where the trnt is the total thickness of rubber in the bearing.

In this paper we are only interested in the theory for and the testing of a bearing in
the form of a long strip when the effects of the ends can be neglected and the strip taken
to be infinite. For an infinite strip of width 2b (see Figure 2), Equation 3 reduces to

2p

d2p
12G
2 2 c
dx
t

which, with p0 at xb, gives

6G 2 2
b x c
t2

390

J. M. KELLY

Figure 2. Infinitely long rectangular pad showing dimensions.

In this case the load per unit length of the strip, P, is given by

8Gb3
P
pdx 2 c
t
b
b

(6)

Since the shape factor, Sb/t, and the area per unit length is A2b,

P
4GS2
E c
Ac

(7)

This result is reasonably accurate for shape factors in the range 5 to 15, but for large
shape factors the predicted value of the compression modulus begins to approach the
bulk modulus, K, of the material, which for natural rubber is usually taken as 2000 MPa
(300,000 psi). An analysis that includes compressibility is reviewed in a later section.
COMPRESSION STIFFNESS WITH FLEXIBLE REINFORCEMENT

Developing the solution for the compression of a pad with rigid reinforcement is algebraically simple enough to be treated in two dimensions and for an arbitrary shape.
The problem for the pad with flexible reinforcement is more complicated, however; for
simplicity, the derivation will be developed for the long, rectangular strip isolator. A very
detailed derivation of the solution is given by Kelly (1999) where the influence of the
flexibility of the reinforcement on the various quantities of interest is studied. In this
approach as before, the rubber is assumed incompressible and the pressure is assumed to
be the dominant stress component. The kinematic assumption of quadratically variable
displacement is supplemented by an additional displacement that is constant through the
thickness and is intended to accommodate the stretching of the reinforcement. Thus in
this case the displacement pattern given in Equation 1 is replaced by

4z2
ux,zu0x 1 2 u1x
t

(8)

EERI DISTINGUISHED LECTURE 2001: SEISMIC ISOLATION SYSTEMS FOR DEVELOPING COUNTRIES

391

wx,zwz
The constraint of incompressibility Equation 2 remains, leading to

3
3
u0,x u1,x
2
2t

(9)

The only equation of stress equilibrium in this case is xx,xxz,z0, and the assumption
of elastic behavior means that

xzGxz

(10)

which with

xz

8z
u
t2 0

(11)

from Equation 8, gives

8Gu0
t2

xx,x

which with the assumption that xxzzp provides the sole equation of equilibrium
as

p, x

8Gu0
t2

(12)

The individual fibers are replaced by an equivalent sheet of reinforcement of thickness

t f . The internal force, F(x), per unit width of the equivalent reinforcing sheet is related
to the shear stresses on the top and bottom of the pad by

dF


0
dx xz zt/2 xz zt/2
as shown in Figure 3. The shear stresses on the top and bottom of the pad are given by

xzzt/2

8Gu0
8Gu0
; xzzt/2
2t
2t

leading to

dF
8Gu0

dx
t

(13)

The extensional strain in the reinforcement f is related to the stretching force through
the elastic modulus of the reinforcement E f and the thickness t f such that

F
f u1, x
Eftf

(14)

392

J. M. KELLY

Figure 3. Force in equivalent sheet of reinforcement.

which when combined with Equation 14, gives

u1,xx

8G
u
Eftft 0

The complete system of equations is

p, x

8Gu0
t2

(15)

3
3
u0, x u1, x
2
2t

(16)

8G
u
Eftft 0

(17)

u1, xx

The boundary conditions used are the vanishing of the pressure, p, and the reinforcement force, F, at the edges of the strip, xb. The results for p and F from Kelly (1999)
are

Eftf
cosh x/b
1
c
t
cosh

FxE f t f 1

cosh x/b
c
cosh

(18)

(19)

where

12Gb2

Eftft
2

(20)

EERI DISTINGUISHED LECTURE 2001: SEISMIC ISOLATION SYSTEMS FOR DEVELOPING COUNTRIES

393

Figure 4. Normalized effective compression modulus as a function of (12Gb2/E f t f t)1/2.

The load per unit length of the strip, P, is given by

Eftf
2
t

x/b
2E t
btanh
dx
1 cosh

cosh
t
b

f f

This result can be interpreted as an effective compression modulus, Ec , given by

Eftf
tanh
1
t

Ec

(21)

We note that when 0, i.e., E f , we have Ec4GS2 as before. The formula also
shows that Ec4GS2 for all finite values of E f .
The effect of the elasticity of the reinforcement on Ec can be illustrated by normalizing the compression modulus, Ec , by dividing by 4GS2, giving from Equation 21

Ec
3
tanh
2 2 1
4GS

(22)

which is shown in Figure 4 for 05. Note how the stiffness decreases with decreasing E f . It is worthwhile to note that also depends on the shape factor, S, through

212

G 2t
S .
Ef tf

In the case of carbon fiber reinforcement the ratio G/E f will be extremely small although

394

J. M. KELLY

S2 and t/t f are certain to be large. If S is large and the influence of the stretching of the
fiber is significant it will be necessary to take compressibility of the elastomer into account.
COMPRESSION STIFFNESS WITH COMPRESSIBILITY OF THE ELASTOMER

When the estimated value of Ec from Equation 7 is comparable to the bulk modulus,
K, of the elastomer, it is necessary to modify the approach described in the earlier section to include the influence of compressibility; the required detailed derivation for the
single pad of arbitrary plan-form is given in Kelly (1996). The essential features of the
analysis for the incompressible pad are retained but the constraint of incompressibility
(Equation 2) is replaced by

xxyyzz

p
K

(23)

and, as shown in Kelly (1996), the equation for the pressure becomes

2p

12G p 12G
2 c
t2 K
t

and this is solved as before with p0 on the edges of the pad.

For the infinite strip bxb the solution is

pKc 1

cosh x/b
cosh

where

12Gb2 12G 2
S

Kt2
K

(24)

Integrating p over the area (per unit length) of the pad gives

P2Kcb 1

tanh

and

EcK 1

tanh

Normalizing this by 4GS2 gives

Ec
3
tanh
2 2 1
4GS

There are two points to note from these two formulas. When 0, i.e., either K
or S0, we have

EERI DISTINGUISHED LECTURE 2001: SEISMIC ISOLATION SYSTEMS FOR DEVELOPING COUNTRIES

395

Ec
3 2 24
24 GS2

1
4GS2 2 3 15
5 K
and as , i.e., S becomes very large

EcK 1

K
12G

1/2

1
S

showing that K is an upper bound to Ec .

FLEXIBLE REINFORCEMENT AND COMPRESSIBILITY

In cases of large shape factors, to estimate Ec including compressibility in a manner

consistent with the assumptions of the previous analysis the equation of incompressibility Equation 2 is replaced by

xxzz

p
K

(25)

where K is the bulk modulus. Integration through the thickness leads to the amended
form of Equation 3

2
p
u0,xu1,x c
3
K

(26)

This is then supplemented by the same equation of stress equilibrium and by the
equation for the forces in the reinforcement, Equation 14. The system of equations for
the combined effects of reinforcement flexibility and compressibility is now

p, x

8Gu0
t2

u1, xx

8Gu0
Eftft

p
2
u0, xu1, x c
3
K

(27)

(28)

(29)

Two dimensionless parameters, and , as defined in previous sections, determine

the comparative significance of flexibility in the reinforcement and compressibility in
the elastomer.
In terms of and , Equations 27 and 28 become

2
p/K,x 2u0b2
3

(30)

396

J. M. KELLY

2
u1, xx 2u0 /b2
3

(31)

Differentiation of Equation 29 once and substitution of p and u1 from Equations 30

and 31 gives

u0, xx

22
u00
b2

from which we have

u0A cosh x/bB sinh x/b

where

222
In turn, using Equations 27 and 28 gives solutions for p and u1 in the forms

2 2
2 2
u1 2 A cosh x/b 2 B sinh x/bC1xD
3
3
and

2 2
2 2
A sinh x/b
B cosh x/bC2
p/K
3 b
3 b
The constants of integration are, of course, not independent of each other but are
related through the basic equations. Substitution of the three solutions into Equation 29
gives

2
2 2
A sinh x/bB cosh x/b
A sinh x/bB cosh x/bC1
3b
3 b
2 2

A sinh x/bB cosh x/bC2c

3
The coefficients of sinh x/b and cosh x/b vanish and the result is C1C2c .
For the particular problem of the compression of the strip it is useful to consider the
obvious symmetries in the solutions. Thus u0 and u1 are anti-symmetric and p is symmetric on bxb. It follows that A0 and D0 giving

u0B sinh x/b

u1

2 2
B sin x/bC1x
3 2

p
2 2

B cosh x/bcC1
K
3 b

EERI DISTINGUISHED LECTURE 2001: SEISMIC ISOLATION SYSTEMS FOR DEVELOPING COUNTRIES

397

The boundary conditions for B and C1 are that the pressure p at the edges xb is
zero and that the stress in the reinforcement E f u1,x also vanishes at the edges. Thus

2 2
B cos C10
3 b

2 2
B cosh C1c

3 b
giving

3
1

2
2b
2 cosh c

2
C1 2 2 c

and the solution becomes

3 sinh x/b

u0 b
2 cosh c

2 x sinh x/b
u1b 2 2

b cosh c
and

p
2
cosh x/b
2 2 1
c
K
cosh

(32)

Integration of p from Equation 32 above and use of Equation 5 gives

2
tanh
EcK 2 2 1

(33)

If the effect of compressibility is negligible then 0 and and we have

12Gb2
t2

K2
giving

Eftf
tanh
1
t

Ec

(34)

which is the same as the result in the previous section. On the other hand, if the flexibility of the reinforcement is negligible then 0 and , giving

398

J. M. KELLY

Table 1. Test specimens

Name

Length
(mm)

Width
(mm)

Height
(mm)

Area
(mm2)

DRB1
DRB2
DRB3
DRB4
DRB5
DRB6
DRB7
DRB8

735
750
740
365
390
377
377
730

183
190
190
190
190
183
183
185

105
105
105
105
105
105
105
105

134505
142500
140600
69350
74100
68991
68991
135050

Presence of rubber cover

Comments

cut from 190755105

cut from 190755105
cut from 183755105
cut from 183755105
kept as sample

East

West

North

South

Yes
Yes
No
No
Yes
No
No

No
No
Yes
No
No
Yes
No

No
Yes
Yes
Yes
Yes
No
No

No
No
No
No
No
No
No

Notes:
1) All test specimens were composed from 33 layers of 3-mm thick rubber and 30 layers of 0.27 mm fiber.
2) The in-plane test machine imposed shear in the east-west direction (representing 0 direction).
3) The location angle of the specimen was measured from the west direction counterclockwise (i.e., at 90 the
former west side points south).
4) The rubber cover of bearing sides reduces the effective work area of the bearing in the vertical direction. The
thickness of the rubber cover varies from 5 mm to 9 mm on the long side of the bearing (north or south) and
varies from 1 mm to 3 mm on the short side of the bearing (east or west).

EcK 1

tanh

(35)

If the compression modulus is normalized by 4GS2, then from Equation 33 we have

Ec
3
tanh221/2

1
4GS2 22
221/2

(36)

which demonstrates how the vertical stiffness is reduced by both compressibility in the
elastomer and flexibility in the reinforcement.
EXPERIMENTAL RESULTS
Several fiber-reinforced bearings were constructed and tested in compression and
shear. Six specimens manufactured by Dongil Rubber Belt Co., Ltd. (Pusan, Korea)
were shipped in the form of strips with slightly varying dimensions as given in Table 1.
The width-to-height ratio was very close to 2 and length-to-height ratio was around 7.5.
Each bearing had 99 mm of rubber and was reinforced by 30 plane sheets of carbon
fiber 0.27 mm thick.
The effective vertical stiffness of the bearing was obtained from a compression test
conducted in the following way. The specimen was monotonically loaded up to the target
value of vertical pressure and then three cycles of vertical loading with small amplitude
about this target value were performed. The shear stiffness of a specimen was obtained
from sets of shear cycles with step-wise increasing amplitude. These shear tests were
conducted for various values of vertical pressure and for three angles between the testing
direction and the longitudinal direction of the strip. All tests were conducted on bearings
that were not bonded to the test machine.

EERI DISTINGUISHED LECTURE 2001: SEISMIC ISOLATION SYSTEMS FOR DEVELOPING COUNTRIES

399

Figure 5. Testing setup.

These tests were carried out on a test machine, shown in Figure 5, designed to produce in-plane vertical and horizontal cyclic loading. The vertical load was applied
through a stiff frame to the specimen by two hydraulic actuators. The horizontal load
was applied by a hydraulic actuator. The test machine had a displacement capacity of
254 mm (10 inches) in the horizontal direction and a load capacity of 1,140 kN (260
kips) in the vertical direction. The photograph in Figure 6 shows a global view of a test
in progress.
TEST PROGRAM

Specimen DRB1 was tested under vertical load control monotonically to 1.73 MPa
(250 psi) of vertical pressure and three fully reversed cycles with 0.35 MPa (50 psi)
amplitude were performed, then monotonically unloaded. A similar test at 3.45 MPa
(500 psi) vertical pressure with 0.35 MPa (50 psi) amplitude was performed to study
the vertical stiffness at the greater vertical load.
The horizontal test was performed under horizontal displacement control. Specimen
DRB1 was tested in cyclic shear, with three fully reversed cycles at four peak strain levels of 25%, 50%, 75%, and 100% (based on 99 mm rubber thickness) applied at a vertical pressure of 1.73 MPa (250 psi). The vertical pressure was increased to 3.45 MPa
(500 psi) and the shear test was repeated. The peak value of shear deformation was increased by 1.5 and the test repeated. The shear tests were done for the following sequence of the angle between the testing direction and the longitudinal direction of the
strip: 0, 90, and 45. Specimens DRB2 and DRB3 were tested under the same test
program, but with a different sequence of the angle between the testing direction and the
longitudinal direction of the strip. For specimen DB2 this sequence was 45, 0, and 90,
and for specimen DRB3 it was 90, 45, and 0.

400

J. M. KELLY

Figure 6. Test in progress.

In order not to exceed the capacity of the testing machine, a specimen was cut in two
equal halves, denoted as DRB4 and DRB5. The half-length specimens DRB4 and DRB5
were used to study behavior at higher levels of vertical load and larger shear deformation. The value of vertical pre-load was increased to 6.90 MPa (1000 psi) and the shear
deformation was doubled to study the behavior at a larger shear strains. The angle between the testing direction in shear and the longitudinal direction of the strip was 0 for
specimen DRB4 and 90 for specimen DRB5.
The possibility of increasing of shear capacity of the bearings by stacking them (one
on top of the other) was studied by testing DRB8 and DRB9 in this way. The joint specimen was vertically loaded up to 3.45 MPa (500 psi) vertical pressure and then was tested
in shear at 50%, 100%, 150%, and 200% peak shear strains based on the single bearing
thickness.

Figure 7. Specimen DRB6 at 100% shear deformation (90).

EERI DISTINGUISHED LECTURE 2001: SEISMIC ISOLATION SYSTEMS FOR DEVELOPING COUNTRIES

401

Figure 8. Specimen DRB6 at 100% shear deformation (45).

Finally, specimen DRB9 was cut in two equal halves, and these were designated as
specimens DRB6 and DRB7. They differed in that DRB6 had rubber cover at one end,
whereas specimen DRB7 had no side rubber cover. They were tested under the same test
program at three levels of vertical pressure: 0.87 MPa (125 psi), 1.73 MPa (250 psi), and
3.45 MPa (500 psi). Figure 7 is a photo of specimen DRB6 under 100% shear deformation testing at 90 to the longitudinal direction. The deformation with the same magnitude at 45 to the longitudinal direction is shown in Figure 8 and the specimen under
100% shear deformation in the longitudinal direction is shown in Figure 9.
DISCUSSION OF EXPERIMENTAL RESULTS

Horizontal Test Results

The manufacturer of the test isolators gave the nominal shear modulus of this natural
rubber compound as 0.690 Mpa (100 psi). The three full-length uncut specimens had an
average area of 0.140 m2 and a total rubber thickness of 0.099 m. The horizontal stiffness, KH , of a conventional isolator is given by

Figure 9. Specimen DRB6 at 100% shear deformation (0).

402

J. M. KELLY

KHGA/tr
and for these values KH is

KH970 kN/m
At 100% shear strain and a pressure of 1.73 Mpa (250 psi) the average horizontal
stiffness in the longitudinal loading direction is 1280 kN/m, in the lateral loading direction is 863 kN/m and at 45 is 1120 kN/m.
The hysteresis loops for the longitudinal loading direction tend to stiffen when the
shear strain is increased from 100% to 150%, whereas in the lateral direction the loops
turn over so that the instantaneous tangent stiffness is negative at the larger strains. However, this effect is reduced at higher pressure levels. The 45 loading does not produce
either stiffening or softening but gives values intermediate between the 0 and 90 loadings.
The value of the stiffness at 100% shear strain in the longitudinal direction is slightly
higher than would be expected from the nominal value of the shear modulus but in the
transversal loading direction the stiffness is lower. At 45 the stiffness is intermediate
between the other two. If we assume that the layout of the strip isolator is orthogonal
with roughly the same number in each direction, the average between 0 and 90 is close
to the value at 45 so that the system will have the same period in any direction of movement.
The period, T, can be roughly estimated using the pressure and the effective shear
modulus. The period is given by

T2

ptr
Gg

If the average pressure over the system is 3.45 Mpa (500 psi) as in the tests and the
modulus is 0.690 Mpa (100 psi) with 99 mm of rubber, we have a period of 1.4 seconds.
From the code formula this would produce a displacement of 143 mm (5.64 inches) and
a shear strain of 1.41. Adjusting the values to correspond to the measured stiffness at
1.5, we find that the period increases to 1.5 seconds and the displacement to 150 mm
(6 inches).
This suggests that if the period of 1.5 seconds is acceptable as the target value for the
design of the building, the strip isolators as tested would be adequate, providing that the
average pressure can be at least 3.45 Mpa (500 psi). A longer period can be obtained by
having one isolator on top of another. This leads to a period of 2 seconds, a code displacement of 200 mm (8 inches) and a 1.0. It is clear that a wide range of practical
objectives is possible. If it is necessary to have an average pressure of less than 3.45
MPa (500 psi) it is possible to use a softer compound. Compounds with shear moduli, at
100% strain, down to 0.40 Mpa (60 psi) are available.

EERI DISTINGUISHED LECTURE 2001: SEISMIC ISOLATION SYSTEMS FOR DEVELOPING COUNTRIES

403

Table 2. Vertical test results for 1.73 MPa vertical pressure

Specimen
No.

Area
(m2)

Imposed
Load (kN)

Average Pressure
(MPa)

Average Stiffness
Kav (kN/M)

Compression modulus Ec (MPa)

DRB1
DRB2
DRB3
DRB4
DRB5
DRB6
DRB7

0.135
0.143
0.141
0.069
0.074
0.069
0.069

233.6
233.6
233.6
N/A
N/A
120.2
120.2

1.73
1.63
1.66
N/A
N/A
1.74
1.74

550853.9
602975.0
597053.3
N/A
N/A
251687.8
278983.5

404
417
419
N/A
N/A
361
400

Vertical Test Results

The vertical test results are shown in Tables 2, 3, and 4. Since the dimensions of the
bearings are slightly different in each case it is useful to tabulate the vertical stiffness in
terms of the effective compression modulus, Ec , as defined by Equation 5. The fulllength bearings DRB1/2/3 are quite consistent with Ec at around 414 Mpa. The two sets
of half-length bearings have lower values of Ec at the same vertical pressures of testing.
The pair denoted by DBR4/5 was not tested at 1.73 Mpa (250 psi) but at 3.45 Mpa (500
psi) and 6.90 Mpa (1000 psi). The set denoted by DBR6/7 was tested at 0.87 Mpa (125
psi), 1.73 Mpa (250 psi), and 3.45 Mpa (500 psi). At the common test pressure of 3.45
Mpa (500 psi) the average of the two values of Ec of DBR4/5 was the same as that of
DBR6/7, so that we can interpret the effect of variation of the target pressure over an
eight-fold range. The fact that at the same pressures the Ec values for the full-length
bearings are higher than for the half-length bearings is most likely due to the larger influence of free ends in the latter case. The theoretical analysis is developed for the infinite length strip and for the full-length bearings the length-to-height ratio of 7.5 is
large enough that this assumption is valid. At half this value end effects can be expected
to have some influence. The vertical stiffness of an elastomeric isolation bearing is always difficult to measure since the displacements at the vertical loads corresponding to
practical use are extremely small and a great deal of scatter is to be expected.

Table 3. Vertical test results for 3.45 MPa vertical pressure

Specimen
No.

Area
(m2)

Imposed
Load (kN)

Average Pressure
(MPa)

Average Stiffness
Kav (kN/M)

Compression Modulus Ec (MPa)

DRB1
DRB2
DRB3
DRB4
DRB5
DRB6
DRB7

0.135
0.143
0.141
0.069
0.074
0.069
0.069

467.3
467.3
467.3
253.7
253.7
240.3
240.3

3.46
3.27
3.31
3.68
3.43
3.48
3.48

791048.8
849319.3
752785.8
349938.19
352040.6
328721.9
351392.3

580
588
529
502
471
472
504

404

J. M. KELLY

Table 4. Vertical test results for extreme values of vertical pressure

Specimen
No.

Area
(m2)

Imposed
Load (kN)

Average Pressure
(MPa)

Average Stiffness
Kav (kN/M)

Compression Modulus Ec (MPa)

DRB1
DRB2
DRB3
DRB4
DRB5
DRB6
DRB7

0.135
0.143
0.141
0.069
0.074
0.069
0.069

N/A
N/A
N/A
507.3
507.3
60.1
60.1

N/A
N/A
N/A
7.35
6.86
0.87
0.87

N/A
N/A
N/A
467372.6
445858.5
175617.3
167190.4

N/A
N/A
N/A
671
596
252
240

It is clear, however, that there is for all test specimens a systematic increase in stiffness and Ec when the central value of the pressure around which the load is cycled increases. As the pressure is doubled the average value of Ec for the three full-length isolators increases by 37% to 565 MN/m2. The half-length bearing tests show an increase in
Ec over the complete range of pressure. The increase is not linear in pressure but tends to
decrease with increasing pressure from 55% at the lowest level to 15% at the highest.
This is consistent with the type of carbon fiber used in the bearings. The fiber is woven,
two directional, and epoxied into a thin sheet. As the pressure increases the in-plane fiber sheet tension increases and tends to straighten out the fiber strands, thus increasing
the effective fiber modulus.
To assess the effect of the various parameters on the vertical stiffness it is necessary
to estimate the actual shear modulus from the tests on shear. At a vertical pressure of
3.45 Mpa (500 psi) the average shear stiffness of the first three bearings when tested in
the longitudinal direction 0 is 1.278 MN/m, which with an average area of 0.140 m2
and a thickness of rubber of 99 mm implies a shear modulus of 0.904 Mpa (131 psi),
which is considerably larger than the nominal modulus. This use of the longitudinal tests
to provide an estimate of the modulus is warranted by the fact that this case will have the
least influence from the roll-off due to the unbonded condition.
A steel-reinforced isolator with this shear modulus, this area of rubber and this thickness would, if compressibility effects were ignored, have an effective compression
modulus Ec given by Equation 5, of 3738 MN/m2. When compressibility is taken into
account the effective modulus is considerably reduced as given by Equation 35. To estimate , we use K2000 Mpa (290,000 psi), giving 25.3 and from Equation 35 we
have 1150 MN/m2. The average measured value, 563 MN/m2, can be used to deduce the
effect of the extensibility of the carbon fiber reinforcement. For this purpose we now
turn to Equation 33 and assume that 25.3. From the results we have Ec /K0.28 and
from the equation we determine that (22)1/23.7, implying 28.4. When the
known values of the various parameters are inserted into the definition of 2 we estimate
E f as 14,000 Mpa (2,000,000 psi).
There is no equipment in the laboratory to measure modulus of the carbon fiber
sheet directly nor has the fabricator provided a value. The result is somewhat lower than
others quoted for carbon/epoxy sheets and the reason is not clear, but the sheets appear

EERI DISTINGUISHED LECTURE 2001: SEISMIC ISOLATION SYSTEMS FOR DEVELOPING COUNTRIES

405

to be a very poor quality fiber with many spaces and the portion of the thickness that is
fiber cannot be determined from visual analysis. It is certainly possible that the quality
of the reinforcing could be improved but it is clear that even this poor quality sheet is
adequate for these bearings. An effective modulus Ec of 563 MN/m2 at an average pressure of 3.45 Mpa (500 psi) implies a vertical vibrational frequency of 20 Hz, which is
more than is necessary in any isolation application. The conclusion is that although the
fiber is suspected to be very low quality, and presumably low-cost, it has sufficient stiffness and strength for application to low-cost isolators.
CONCLUSIONS
The test results show that the concept of the strip isolator reinforced with carbon
fiber is viable. The fact that the isolator can be made in long, wide sheets and cut to the
required width means that the cost of the isolator can be reduced to a level that is acceptable for low-cost public housing. The tests also show that loading in the direction of
the strip across a cut surface can be a source of delamination. In practice this should not
be a problem since the width of the manufactured sheet will be used as the length of the
strip isolator and the ends will be finished edges. Loading in the transverse direction
(90), where the edges are cut, is not so severe, as the rolling of the strip tends to produce lower forces in this direction. The most vulnerable direction of loading is at 45,
which appears to put a very distorted pattern of displacement on the isolator and for this
reason it may be advisable to use either a better cutting method such as water jet, which
will leave a smoother finished surface than a steel saw, or to finish the edge by a coldbonded surface layer.
The carbon fiber appeared to be very poor quality. The fibers are laid out in only two
directions, 0 and 90, in a woven sheet with many spaces. Nevertheless the isolators
still functioned acceptably. It would be possible to make a much better isolator with a
better quality carbon fiber at little increase in cost.
The unresolved issue from the test program is that of overall system behavior,
namely, can an isolation system made of strip isolators laid out in an orthogonal grid
protect the masonry wall superstructure above. Isolators loaded in the longitudinal direction stiffen with increasing displacement and those loaded across the strip will soften
with displacement above a certain threshold. The question is can the unbalanced shear
be accommodated by the wall system. Further research is needed to study this effect and
the best way to develop a reliable procedure would be to have a masonry block house
model on isolators on a large shake table.
It is important to recall that the benefit of isolation is achieved primarily through the
ratio of the isolated period of the building to its fixed-base period. For a constant velocity spectrum the base shear of the fixed-base building is reduced by this factor when the
same building is isolated. A masonry wall structure will have an extremely short fixedbase period, in the range of 0.10 second. A reduction of a factor of 10 can be obtained
with an isolation period of 1 second and this is not difficult to obtain. In fact, the code
formula for isolation system displacement that has persisted through all versions of the
seismic isolation codes in the United States since the earliest in 1986 would predict a
displacement of 15 cm. (6 inches) for a 1.5-second period system, and the isolators

406

J. M. KELLY

tested here were tested to displacements of 15 cm. (6 inches) and more. However, if
larger displacements are needed, the tests showed that stacking two isolators on top of
each other was possible, which would allow even larger displacements.
A shake table test using a full size masonry block house would help clarify details of
access across the isolation interface, the disadvantages if any of not having the bottom
floor slab-on-grade isolated, and the extent to which a concrete tie-strip is needed between the isolators and the block wall. When these remaining uncertainties are resolved
this valuable new technology can be implemented in many highly seismic areas in the
developing world.
ACKNOWLEDGMENTS
The sample isolators were made by Dongil Rubber Belt Co. Ltd. of Pusan, Korea.
The test program was conducted at the Structural Test Facility of the Pacific Earthquake
Engineering Research Center, University of California, Berkeley. This research work
was partly supported by the Engineering Research Center for Met-Shape and Die Manufacturing of Pusan National University, Pusan, Korea, which is gratefully acknowledged.
REFERENCES
Gent, A. N., and Lindley, P. B., 1959. The compression of bonded rubber blocks, Proc. Inst.
Mech. Eng. 173 (3), 111117.
Gent, A. N., and Meinecke, E. A., 1970. Compression, bending and shear of bonded rubberblocks, Polym. Eng. Sci. 10 (2), 4853.
International Conference of Building Officials (ICBO), 1997. Earthquake regulations for seismic isolated structures, Uniform Building Code, Appendix Chapter 16, Whittier, CA.
Kelly, J. M., 1996. Earthquake-Resistant Design with Rubber, 2nd edition, Springer-Verlag,
London.
Kelly, J. M., 1999. Analysis of fiber-reinforced elastomeric isolators, J. Seismic Engrg. 2 (1),
1934.
Naeim, F., and Kelly, J. M., 1999. Design of Seismic Isolated Structures, John Wiley, New York.
Rocard, Y., 1937. Note sur le calcul des proprietes elastique des supports en caoutchouc adherent, J. Phys. Radium 8, 19.

(Received 13 March 2002; accepted 6 April 2002)