A Mechanical Model For Elastomeric Seismic Isolation Bearings Including The Influence of Axial Load

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS
Earthquake Engng Struct. Dyn. 2009; 38:157180

Published online 26 August 2008 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/eqe.847
A mechanical model for elastomeric seismic isolation bearings

including the influence of axial load
Sachie Yamamoto1, , , Masaru Kikuchi1 , Masaiki Ueda1 and Ian D. Aiken2

1 Department of Architectural and Structural Design, Graduate School of Engineering, Hokkaido University,
Kita-13, Nishi-8, Kita-ku, Sapporo 060-8628, Japan
2 Seismic Isolation Engineering, Inc., 1144 65th Street, Unit D, Emeryville, CA 94608, U.S.A.
SUMMARY
For the purpose of predicting the large-displacement response of seismically isolated buildings, an analyt-
ical model for elastomeric isolation bearings is proposed. The model comprises shear and axial springs
and a series of axial springs at the top and bottom boundaries. The properties of elastomeric bearings
vary with the imposed vertical load. At large shear deformations, elastomeric bearings exhibit stiffening
behavior under low axial stress and buckling under high axial stress. These properties depend on the
interaction between the shear and axial forces. The proposed model includes interaction between shear
and axial forces, nonlinear hysteresis, and dependence on axial stress. To confirm the validity of the
model, analyses are performed for actual static loading tests of leadrubber isolation bearings. The results
of analyses using the new model show very good agreement with the experimental results. Seismic
response analyses with the new model are also conducted to demonstrate the behavior of isolated buildings
under severe earthquake excitations. The results obtained from the analyses with the new model differ in
some cases from those given by existing models. Copyright q 2008 John Wiley & Sons, Ltd.
Received 20 February 2008; Revised 8 July 2008; Accepted 15 July 2008
KEY WORDS: seismic isolation; elastomeric isolation bearing; large deformation; axial force; nonlinear
hysteresis model; seismic response analysis
INTRODUCTION
Seismic isolation is the most effective technology for protecting structures from the damaging
effects of earthquakes. It has been extensively used worldwide over the last three decades. The
widespread use of seismic isolation has necessitated better understanding of some of the more
Correspondence to: Sachie Yamamoto, Department of Architectural and Structural Design, Graduate School of
Engineering, Hokkaido University, Kita-13, Nishi-8, Kita-ku, Sapporo 060-8628, Japan.
E-mail: sachi615@eng.hokudai.ac.jp
Contract/grant sponsor: Publishing Arts Research Council; contract/grant number: 98-1846389
Copyright q 2008 John Wiley & Sons, Ltd.

158 S. YAMAMOTO ET AL.
complex aspects of isolation device behavior, such as under large shear deformations or high
compressive stresses. Leadrubber bearings and high-damping rubber bearings are two commonly
used types of seismic isolation device. Elastomeric isolation bearings exhibit stiffening or buckling
behavior, influenced by the imposed compressive stress at large shear deformations. The change in
shear stiffness due to high compressive stress is an important behavior to consider for elastomeric
bearings when isolated buildings experience extreme earthquake shaking.
The actual behavior exhibited by an elastomeric isolator depends on the interaction between
shear and axial forces. Taking the shearaxial interaction into account, Koh and Kelly proposed a
simple mechanical model (two-spring model) for elastomeric bearings [1]. The two-spring model
is a combination of a shear spring and a rotational spring and a rigid column connected above the
springs. The results of seismic response analyses with the two-spring model show that neglecting
P effects can lead to considerable errors for bearings with low buckling safety factors [2].
An improvement of the two-spring model was refined to give the so-called KohKelly model,
which has rotational springs at the top and bottom and the rotational springs are connected by
rigid columns with a shear spring at mid-height [3]. Since the KohKelly model consists of linear
springs, it is not suitable for the analysis of elastomeric bearings that exhibit nonlinear behavior at
large shear displacements. In order to better predict the large-displacement behavior of elastomeric
bearings, Iizuka introduced finite deformation and nonlinear hysteresis into the two-spring model
to produce the so-called macroscopic model [4]. The macroscopic model successfully predicts
skeleton curves enveloping the shear forcedisplacement relationships obtained from cyclic loading
tests. Ryan et al. examined the effects of shear strain level and applied vertical load on the
properties of elastomeric isolation bearings using response data from characteristic tests of three
different types of bearings and proposed an analytical model for leadrubber bearings by the
modifying the KohKelly model with a bilinear hysteretic relationship and an empirical equation
for bearing yield strength [57]. The authors have also proposed an analytical model for elastomeric
isolation bearings under large shear deformations, through the introduction of nonlinear hysteretic
relationships to the KohKelly model [8].
The actual hysteretic behavior of isolation bearings under a structure subjected to severe earth-
quake shaking is not the same as that obtained from cyclic shear tests with constant axial stress
because of the variation of vertical load due to overturning forces [9]. For the response of isolated
structures, especially for those structures in which large overturning forces occur, the influence of
the variation of vertical load on bearing behavior may be significant. Buckle and Liu examined
the relationship between loss of lateral stiffness due to buckling effects of individual isolation
bearings and the overall stability of a system of bearings, taking into account variations in vertical
load due to overturning forces [10]. The important aspect of varying vertical load is its influence
on the bending characteristics of a bearing. To predict the response of isolated structures under
large bearing shear displacements, an analytical method that accounts for both nonlinear bearing
behavior in large shear deformations and the effect of varying vertical load on bearing properties,
especially bending properties, is needed. It is beyond the scope of any of the previously proposed
models to incorporate both of these effects in a single response analysis. Two of the authors have
previously developed a nonlinear shear hysteresis model for elastomeric bearings [11]. That model
can accurately predict bearing shear behavior including large deformation, stiffening properties,
but does not include vertical load effects. An analytical model that includes the previously devel-
oped, accurate shear hysteresis model and also incorporates the effect of varying vertical load
would allow more accurate prediction of bearing behavior under severe demands. In this paper, a
new analytical model for elastomeric bearings with these capabilities is proposed. The new model
Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:157180
DOI: 10.1002/eqe
A MECHANICAL MODEL FOR ELASTOMERIC SEISMIC ISOLATION BEARINGS 159
includes the interaction between shear and axial forces as a function of varying vertical load on a
bearing during earthquake loading.
A NEW MECHANICAL MODEL FOR ELASTOMERIC BEARINGS
In this section, the formulation of the new mechanical model for elastomeric seismic isolation
bearings and the associated numerical analysis procedure are described.
Model formulation
Figure 1 shows the proposed new mechanical model to incorporate interaction between shear and
axial forces as a function of varying vertical load. The model comprises shear and axial springs
and two series of axial springs at the top and bottom boundaries. Each spring in the series of
axial springs represents an individual strip of the bearing cross-sectional area. The rigid columns,
which represent the height of the bearing, are combined between the series of axial springs and the
mid-height shear and the axial springs. Each spring in the model is a uniaxial, nonlinear spring.
When this collection of springs is combined in the model, the nonlinear, interaction behavior is
achieved. The nodes, a, m, n, and b, in Figure 1 are the points where the components of the
mechanical model inter-connect. The nodes are classified into two types. One is an external node
(a or b), which has mass and connects the outer elements. The other is an internal node (m or n),
Figure 1. Multi-spring mechanical model.
DOI: 10.1002/eqe
which has no mass and connects the inner elements of the model. The external nodes, a and b, have
displacements in three directions (horizontal, vertical, and rotational) and the internal nodes, m
and n, have displacements in two directions (vertical and rotational). The horizontal displacements
of the internal nodes are equal to those of the external nodes. The node pairs a and m and b and n
are spatially coincident. The forces and displacements on the model are shown in Figure 2.
By using incremental displacements of nodes a and m, the incremental normal deformation of
the ith spring in the series of axial springs between a and m, i N a , is expressed as

va

a

i N a = [1 i la 1 i la ] (1)

vm

m
where i la is the length between the ith spring and the center of the series of axial springs, which
represents the distance from the centroid of the ith strip to the center of the cross-sectional area
of the bearing.
The incremental normal restoring force of the ith spring in the series of axial springs between
nodes a and m, i f N a , and the incremental bending moment generated from the normal restoring
force of the ith spring, i m am , may be expressed as
i f N a = i K N a i N a (2)
i m am = i la i f N a =i la i K N a i N a (3)
where i K N a is the stiffness of the ith axial spring located between nodes a and m.
Figure 2. Forces and displacements on the mechanical model.
DOI: 10.1002/eqe
As the incremental forces on nodes a and m are the sums of the incremental restoring forces
on the series of axial springs between a and m, the incremental vertical force and the bending
moment on a and m may be expressed as follows:
f va = i f N a
m a = i m am
(4)
f vm = i f N a
m m = i m am
Equations (1)(4) may be expressed by the following matrix equation:

f va va

m a
a

= Kam (5)

f vm

vm

m m m
where

1 ka 2 ka 1 k a 2 k a

3 ka 2 k a 3 k a
Kam =

(6)
1 ka 2 ka
symm. 3 ka
and
1 ka = i K N a
2 ka = i K N a i la
3 ka = i K N a i la2
The incremental vertical force and bending moment on nodes n and b are the sums of restoring
forces on the series of axial springs between n and b. The same procedure used for getting the matrix
equation of the forcedisplacement relationship on nodes a and m, Equations (5) and (6), is also
applied to derive the forcedisplacement relationship on nodes n and b. The forcedisplacement
relationship on nodes n and b may be expressed by the following matrix equation:

f vb vb

m b
b

= Knb (7)

f vn

vn

m n n
DOI: 10.1002/eqe
where

1 kb 2 kb 1 k b 2 k b

3 kb 2 k b 3 k b
Knb =

(8)
1 kb 2 kb
symm. 3 kb
and
1 kb = i K N b
2 kb = i K N b i lb
3 kb = i K N b i lb2
and i K N b is the stiffness of the ith spring in the series of axial springs between nodes n and b
and i lb is the length between the ith spring and the center of the series of axial springs.
Now consider the forcedisplacement relationships for the shear and axial springs at the mid-
height of the model. The forcedisplacement relationship on nodes n and m in Figure 2, which
exclude the rigid columns, may be expressed as follows:
fmn = Kmn umn (9)
fmn = { f ha

f vm m m f hb

f vn m n }T (10)
umn = {u a vm

m u b vn n }T (11)

KS 0 0 K S 0 0

0 KN 0 0 K N 0

0 0 KR 0 0 K R
Kmn =

(12)
K S 0 0 KS 0 0

0 K N 0
0 0 KN
0 0 K R 0 0 KR
where K S and K N are the stiffnesses of the mid-height shear and axial springs, respectively. K R
is the stiffness of a rotational spring connected between nodes m and n , which is not shown in
Figure 2. This spring is a supplementary element, which provides rotational flexibility at the mid-
height and gives the new model the ability of handling various distributions of bending moment.
The assumption that K R is infinity is made in the present development.
In order to transform the forcedisplacement relationship on nodes m and n , expressed by
Equation (9), to that on nodes m and n, which includes the rigid columns, a transformation matrix,
T, is employed. Taking the geometrical relationships of the deformations, the force-equilibrium
condition, and the P effect into account gives the transformation matrix, T. The geometrical
relationships of the deformations and forces are shown in Figure 3. Small deformation kinematics
is assumed in the construction of the transformation matrix, T. Since the shear stiffness of an
DOI: 10.1002/eqe
Figure 3. Geometrical relationships of the deformations and forces.
elastomeric bearing is much smaller than its axial stiffness (typically K S /K N is of the order of
103 or even less), it is assumed that any additional contribution of horizontal forces to axial
forces, and of axial deformations to horizontal displacements, may be neglected. By using the
matrix T, the transformation of the displacement and the force on nodes m and n into those on
nodes m and n may be expressed as follows:
umn = Tumn (13)
fmn = TT fmn (14)
umn = {u a vm m u b vn n }T (15)

fmn = { f ha f vm m m f hb f vn m n }T (16)

h
1 0 0 0 0
2

u b u a

m 1 0 0 0
2

0 0 1 0 0 0

T= (17)
h
0 0 0 1 0
2

u b u a

0 0 0 n 1
2
0 0 0 0 0 1
where h is the total height of the bearing.
DOI: 10.1002/eqe
Substituting Equations (9) and (13) into Equation (14), the forcedisplacement relationship on
nodes m and n may be expressed as
fmn = Kmn umn (18)

Kmn = TT Kmn T (19)
Evaluating Equation (19) gives the elements of the stiffness matrix, Kmn :

1 kmn 4 kmn 6 kmn 3 kmn 4 kmn 6 kmn

8 kmn 9 kmn 5 kmn 8 kmn 9 kmn

10 kmn 7 kmn 9 kmn 11 kmn
Kmn =

(20)
2 kmn 5 kmn 7 kmn

9 kmn
symm. 8 kmn
10 kmn
where
1 kmn = K S + K N 2m
2 kmn = K S + K N 2n
3 kmn = K S + K N m n
4 kmn = K N m
5 kmn = K N n
h K S K N m (u b u a )
6 kmn = +
2 2
h K S K N n (u b u a )
7 kmn = +
2 2
8 kmn = KN
K N (u b u a )
9 kmn =
2
h 2 K S K N (u b u a )2
10 kmn = + +KR
4 4
h 2 K S K N (u b u a )2
11 kmn = + KR
4 4
The overall stiffness matrix for an elastomeric bearing, Kab , is obtained by arranging the elements
of the partial stiffness matrices, Kam , Knb , and Kmn (Equations (6), (8), and (20)), into a 1010
DOI: 10.1002/eqe
matrix:

fex uex
= Kab (21)
fin uin
where
fex = { f ha f va m a f hb f vb m b }T (22)
fin = { f vm m m f vn m n }T (23)
uex = {u a va a u b vb b }T (24)
uin = {vm m vn n }T (25)
and fex and fin are the incremental forces on the external nodes a and b and on the internal
nodes m and n, respectively, and uex and uin are the incremental displacements on the external
nodes and on the internal nodes, respectively. Each element of each of the partial stiffness matrices
is placed in the 1010 matrix, Kab , based on the node at which the element participates.
Numerical analysis procedure

Since nonlinear hysteresis springs are used in the model, unbalanced forces are generated at the
internal nodes and hence a process for resolving these unbalanced forces is required for equilibrium.
The overall stiffness matrix, Kab , is divided into four sub-matrices, K11 , K12 , K21 , and K22 . The
boundaries between the four sub-matrices are defined between the elements for the external nodes
and the elements for the internal nodes. By using the sub-matrices, the relationship of the external
and internal forces on both the external and internal nodes may be expressed as follows:

Fex K11 K12 uex fex
= + (26)
Fin K21 K22 uin fin
where Fex and Fin are the forces acting on the external and internal nodes, respectively. Since
this new bearing model is intended to be used with other structural analytical elements, such as
columns or beams, the reduced stiffness matrix expressed by only the external node elements
is derived. The internal nodes, m and n, are not affected by the inertia force caused by ground
acceleration as the internal nodes have no mass. For this reason, Fin = 0 is assumed. Substituting
Fin = 0 into Equation (26) and solving for Fex gives
Fex = r Kab uex +fex (27)
where
r Kab = K11 K12 K1

22 K21 (28)
fex = K12 K1
22 fin +fex (29)
The incremental displacement on the internal nodes uin , may be expressed as
uin = K1 1
22 K21 uex K22 fin (30)
DOI: 10.1002/eqe
The apparent inner force vectorwhich acts on the external nodesin Equation (27), fex , is
calculated from Equation (29). Equation (29) includes the sub-stiffness matrices and the force
vectors of the external and internal nodes. The elements of the sub-stiffness matrices and the
force vectors are found by comparing the deformations with the nonlinear forcedeformation
relationship of each nonlinear spring. The nonlinear forcedeformation relationships are described
in the following section. The deformations at the mid-heightthe shear deformation, S , and
the normal deformation, N ,and the normal deformations of the ith spring in the series of
axial springs, i N b and i N a , at the top and bottom, respectively, are calculated using the
elements of the incremental displacement vectors of the external nodes, uex , and the internal
nodes, uin :
S = (u b u a )+ h2 (n +m )

(u b u a )
N = (vn vm )(u b n u a m ) (n +m )
2 (31)
i N a = (vm va )+(i la m i la a )
i N b = (vb vn )+(i lb b i lb n )
Comparing the nonlinear spring hysteresis models with the deformations, S , N , i N b , and
i N a , gives the restoring forces and the tangent stiffness of the springs. By using the restoring
forces, f S , f N , i f N a and i f N b , the elements of the inner force vectors that acts on the external
and internal nodes, fex and fin , may be expressed as follows:
f ha = f S + f N m
f va = i f N a
m a = i f N a i la
(32)
f hb = f S f N n
f vb = i f N b
m b = i f N b i lb
f vm = i f N a f N
h (u u a )
m m = i f N a i la + fS b fN M
2 2
(33)
f vn = i f N b + f N
h (u u a )
m n = i f N b i lb + fS b fN + M
2 2
Finally, by using the inner force vector of the external and internal nodes, fex and fin , the apparent
inner force vector in Equation (27), fex , is calculated from Equation (29). Equilibrium of forces
between nodes m and n at each time step is achieved by internal iteration within the time step.
DOI: 10.1002/eqe
This method facilitates stable numerical analyses and provides automatically resolution of the
unbalanced forces at the internal nodes during the analyses.
HYSTERESIS MODELS
In order to predict the large-deformation behavior of elastomeric bearings, nonlinear hysteresis

relationships are used for the springs in the new mechanical model. In this section, new hysteresis
relationships for the shear and axial springs of the model are developed.
Shear spring
Two of the authors previously developed a hysteresis model for the shear deformation of elastomeric
isolation bearings, which is capable of predicting the behavior of high-damping rubber bearings
under large shear deformations [11]. The following presents a modification of this earlier model,
to extend applicability to leadrubber bearings. The shear forcedeformation relationship of the
bearing is expressed by the combination of an elastic component of force, F1 , and a hysteretic
component, F2 , and is given by the following equations (Figure 4(a)):
F = F1 + F2 (34)
F1 = 12 (1u)Fm {x |x|n } (35)
F2 = u Fm {12ea(1x) +b(1 x)ec(1x) } (36)
where Fm is the peak force on the skeleton curve (defined by Equation (35)), x is the normalized
displacement (x = X/ X m ), and X m is the peak displacement on the skeleton curve. In Equation (35),
the parameter n specifies the stiffening. The parameter u is the ratio of force at zero displacement,
(a) (b)
Figure 4. Hysteresis models: (a) shear hysteresis models and (b) axial stressstrain behavior
for top and bottom series of axial springs.
DOI: 10.1002/eqe
Fu , to Fm (u = Fu/Fm ). The parameters a and b are calculated from the following equations,
which are derived assuming that the analytical and experimental hysteresis loop areas are equal:
1e2a 2u h eq
= (37)
a 2u

2 h eq 2 2a
b=c 2+ (e 1) (38)
u a
where h eq is the equivalent viscous damping ratio. The parameter c is a pre-selected constant,
which specifies the shape of the hysteresis loop.
The equivalent viscous damping ratio, h eq , as a function of shear strain, is well established
for high-damping rubber bearing and obtained from regression analysis of test results. Bilinear
modeling is usually used for leadrubber bearings; therefore, h eq cannot be defined explicitly as
a function of shear strain. Instead, design equations for the bilinear model have been developed
[12]. The equivalent viscous damping ratio of the bilinear hysteresis loop is evaluated using
Equation (39). The value of parameter b in Equation (38) can be obtained by replacing h eq in
Equation (38) with h eq,bl given by Equation (39). The values of the hysteresis loop parameters are
updated when the displacement exceeds the previous peak value:

2Fd Fd
h eq,bl = 1 (39)
Fm (K e K d )X m
where Fd is the force at zero displacement, and K e and K d are the elastic stiffness and the yielding
stiffness, respectively, of the bilinear model. The values of these parameters are easily calculated
by using of the design equations for leadrubber bearings as mentioned above.
Axial springs
The new mechanical model has two kinds of axial springs, a series of axial springs at the top
and the bottom boundaries and an axial spring at mid-height. In the present work, it is assumed
that the series of springs at the top and bottom represents the axial properties of the bearings, and
the stiffness of the mid-height axial spring is assumed infinite.
The axial stiffness of an elastomeric bearing may be calculated from the following equations [13]:
EC A
KV = (40)
hr
E 0 (1+2S21 )E
EC = (41)
E 0 (1+2S21 )+ E
where A is the cross-sectional area of the bearing, h r is the total height of rubber in the bearing,
E 0 is Youngs modulus of rubber, is a constant related to rubber hardness, S1 is the rubber
layer shape factor, and E is the bulk modulus of rubber. Figure 4(b) shows the stressstrain
relationship for the series of axial springs. The stressstrain relationship consists of the following
DOI: 10.1002/eqe
four loading states:

(1) Loading and unloading in the elastic region (oab):
= E init (42)
2K V l
E init = (43)
A
where E init is the initial Youngs modulus and l is the length of each axial spring in the
series of axial springs. Since the series of axial springs is between the spatially coincident
external internal nodes, the length, l, is an imaginary length. One half of the total height of
rubber in the bearing is used for this imaginary length.
(2) Loading after yielding (bc):
= y + E y (y ) (44)
Ey = 1
500 E init (45)
y = E init y (46)
where y is the tension yielding stress, E y is the post-yield modulus, and y is the tension
yielding strain. A value of 1.0 MPa is used for the tension yielding stress, y .
(3) Unloading after tension yielding (cda):
r l
= r (r ) (l r ) (47)
r l
= E init (<l ) (48)
where r and r are the stress and strain of the most recent point of load reversal, and l
and l are the stress and strain of the target point during unloading after tension yielding.
(4) Over compression yield stress region (ae):
= c E comp (c ) (49)
E comp = 12 E init (50)

where c and c are the stress and strain of the compression yield point. A value of
100.0 MPa is used for the compression yield stress, c . When re-loading occurs, even after
the compression yield point has been passed, the modulus is still E init .
The stiffness of each axial spring in the series of axial springs, i K N a or i K N b , is calculated from
the following equation:
i E i Adiv
i KNa or i K N b = (51)
l
where i E is Youngs modulus of each axial spring in the series of axial springs, which is obtained
by referring to the hysteresis model, and i Adiv is the area of each divided part. The axial hysteresis
model is established based on the shear forcedeformation relationship obtained from cyclic shear
tests of two leadrubber bearings, which are described in the following section.
DOI: 10.1002/eqe
BEARING TESTS AND SIMULATION ANALYSES
Bearing tests were conducted to verify the new model. In this section, results of bearing tests
to investigate the behavior of leadrubber bearings up to shear strain deformations of 400%
with various different vertical load conditions are presented, along with corresponding simulation
analyses using the new model.
Bearing tests
Two types of cyclic shear tests of leadrubber bearings were conducted to identify their mechanical
characteristics under large deformations at different axial loads. The first were cyclic shear tests
with constant vertical loads. The bearing tested is shown in Figure 5(a). The bearing comprised
24 layers of 2.0 mm thick rubber at 250 mm diameter (first shape factor, S1 = 31.3 and second
shape factor, S2 = 5.2), with a 50.0 mm diameter lead plug, and was connected by bolting directly
to top and bottom end plates. The tests consisted of sinusoidal horizontal displacement-controlled
loading, with four fully reversed cycles of loading at shear strain amplitudes of 50, 100, 200, 300,
and 400%. The loading velocity was 1.5 cm/s. Vertical load on bearing was maintained constant
during each test, and tests were performed at compressive stresses of = 0, 5, 10, 20, and
30 MPa.
The second type of test comprised cyclic shear tests with varying vertical loads. An advantage
of the proposed model is the capability to predict a variety of forcedisplacement relationships for
an isolation bearing with a single model for the bearing. To confirm this capability, cyclic shear
tests of leadrubber bearings under varying vertical load were conducted. The bearings tested are
shown in Figures 5(a) and (b). Two types of bearing were tested, a bearing of the same design as
used for the constant vertical load tests described above (S2 = 5, Figure 5(a)) and a bearing with a
second shape factor, S2 , of 4 (Figure 5(b)). The tests of the S2 = 5 bearing comprised cyclic shear
tests with the vertical load varying (during the shear cycling) between 0 and 30 MPa, and the tests
of the S2 = 4 bearing comprised cyclic shear tests with the vertical load varying (during the shear
cycling) between 0 and 20 MPa. The shear and axial loading histories for both bearing types are
shown in Figure 6.
(a) (b)
Figure 5. Leadrubber bearing designs tested: (a) S2 = 5 bearing and (b) S2 = 4 bearing.
DOI: 10.1002/eqe
(a)
(b)
Figure 6. Loading histories for combined shear and varying vertical load tests: (a) shear and compressive
loading histories for S2 = 5 bearing and (b) shear and compressive loading histories for S2 = 4 bearing.
Constant vertical load

The hysteresis loops obtained from the cyclic shear tests of the leadrubber bearing under constant
vertical loads are shown in Figure 7. The bearing exhibited obvious shear stiffening behavior
beyond approximately 300% shear strain at low axial stresses of = 0, 5, and 10 MPa. However,
a deterioration of horizontal stiffness is seen in the hysteresis loops for the higher compressive
stresses ( = 20 and 30 MPa). The bearing exhibited significant negative stiffness for a stress of
30 MPa. The results show that compressive stress strongly influences the shear hysteretic behavior
of the leadrubber bearing tested.
Simulation analyses of the tests were conducted to verify the validity of the proposed mechanical
model and the spring hysteresis models. The following material properties were used for the
analyses: Youngs modulus of rubber, E 0 = 1.44 MPa, the bulk modulus of rubber, E = 1960 MPa,
and the constant related to rubber hardness, = 0.85. The stiffnesses for the mid-height shear spring
and the series of axial springs at the top and the bottom are not changed for any of the analyses
(compressive stresses of 0, 5, 10, 20, and 30 MPa). The only parameter varied in the analyses is the
compressive stress on the bearing. Figure 8 shows the hysteresis loops predicted by the analytical
model for the same loading histories and compressive stress as the experimental hysteresis loops
shown in Figure 7. Comparing Figures 7 and 8, the experimental and analytical results show good
agreement under all five levels of compressive stress. The results show that the proposed model can
accurately predict complex nonlinear behaviors for leadrubber bearings under large deformations,
as a function of the imposed axial stress on the bearing.
To estimate the accuracy of the proposed model, the point of buckling (the point on the shear
forcedisplacement relation at which the tangent stiffness is zero) and the effective shear stiffness,
K eff , obtained from the experimental and analytical hysteresis loops are examined. Figure 9(a)
shows a comparison of the buckling point obtained from the experimental and analytical hysteresis
loops and the buckling limit line [10, 14]. In Figure 9(a), the compressive stress on the bearing, , is
expressed in terms of its ratio to the critical load, cr , which is calculated by the Haringx solution.
The point of buckling given by the Haringx solution combined with the area reduction formula
DOI: 10.1002/eqe
Figure 7. Leadrubber bearing (S2 = 5) shear forcedisplacement hysteresis loops for cyclic shear tests with
different vertical loads: (a) = 0 MPa; (b) = 5 MPa; (c) = 10 MPa; (d) = 20 MPa; and (e) = 30 MPa.
(a) (b) (c)
(d) (e)
Figure 8. Analytical results for leadrubber bearing cyclic shear tests: (a) = 0 MPa; (b) = 5 MPa;
(c) = 10 MPa; (d) = 20 MPa; and (e) = 30 MPa.
DOI: 10.1002/eqe
(a) (b) (c)
Figure 9. Comparison of buckling limit and effective shear stiffness, K eff : (a) buckling limit
line; (b) K eff (100% shear strain); and (c) K eff (300% shear strain).
(which reduces the buckling stress according to the ratio of the net overlap area to the plan area of
the bearing) and that given by the proposed model are very similar; however, the point of buckling
determined by experiment is seen to be less. Figures 9(b) and (c) show the effect of compressive
stress level on the bearing effective shear stiffness, K eff . The solid lines in Figures 9(b) and (c)
show the reduction in horizontal stiffness calculated by the Haringx solution. Since the stiffness at
zero compressive stress given by the Haringx solution is not equal to K eff , the lines are modified
by replacing the horizontal stiffness at zero compressive stress with the value of K eff at zero
compressive stress and the corresponding shear strain. At 100% shear strain, the experimental and
analytical values of K eff and the horizontal stiffness given by the Haringx solution all show good
agreement. At 300% shear strain, however, the experimentally observed reduction in K eff with
increasing compressive stress is more rapid than that given by the Haringx solution. At this large
strain, the proposed model shows better agreement with experiment than does the Haringx solution.
Evaluation of Figure 9 indicates that although the new model does not predict the experimental
buckling point with more accuracy than does the Haringx solution, the nonlinear reduction in
horizontal stiffness as a function of compressive stress and material nonlinearity is well expressed
by the new model.
Varying vertical load

The hysteresis loops obtained from the tests of leadrubber bearings under varying vertical load
are shown in Figure 10. These loops are not symmetric, rather they show both shear stiffening
behavior and a deterioration of horizontal stiffness depending on the level of vertical load on the
bearing. It is also observed, comparing the hysteresis loops of the S2 = 4 and 5 bearings, that the
S2 = 4 bearing loop shows more significant negative stiffness than the S2 = 5 bearing loop under
high compressive loading (quadrants three and two of the plots). Figure 11 shows the analytical
hysteresis loops given by the new model for the tests of the two bearing types. The analytical
results for both the S2 = 5 and 4 bearings show very good agreement with the experimental results
(Figure 10). This agreement demonstrates that the proposed model can accurately predict the
influence of varying vertical loads on these types of elastomeric isolation bearings and can also
account for second shape factor effects. This capability is useful for dynamic analyses in which
it is desired to include the effect of varying vertical load on isolation bearings during earthquake
loading.
DOI: 10.1002/eqe
(a) (b)
Figure 10. Leadrubber bearing shear forcedisplacement hysteresis loops for cyclic shear tests with
varying vertical load: (a) S2 = 5 bearing and (b) S2 = 4 bearing.
(a) (b)
Figure 11. Analytical results for leadrubber bearing cyclic shear loading with varying vertical load:
(a) S2 = 5 bearing and (b) S2 = 4 bearing.
SEISMIC RESPONSE ANALYSES
The influence of the varying vertical load on the behavior of seismic isolation bearings and the
influence of this behavior on the response of isolated buildings during earthquake loading are
investigated through a series of nonlinear time-history analyses. In general, overturning forces
developed in an isolated structure during earthquake excitation cause variations in the vertical
load on bearings located at the corners of the building, or those under major lateral load-resisting
frames. This response is more evident in taller buildings, but may also occur in shorter buildings
with concentrated lateral load resisting systems. Nonlinear time-history analyses using the new
model are conducted to gain better insight to the seismic behavior of tall isolated buildings, or
those in which substantial overturning forces may be developed.
Structure model
Building models used for the seismic response analyses are shown in Figure 12. Stick models
are developed to represent typical frames of 15- and 20-story superstructures and are supported
by two leadrubber bearings connected by a rigid beam. The superstructure consists of masses
DOI: 10.1002/eqe
Figure 12. Structure models for seismic response analyses.
Table I. Dimensions of leadrubber bearings used for seismic response analyses.

Bearing Lead plug Total rubber Shape Aspect
diameter (mm) diameter (mm) thickness (mm) factor S1 ratio S2
15 Story 900 180 204 (6.034 layers) 37.5 4.4
20 Story 1100 220 203 (7.029 layers) 39.3 5.4
connected by linear beam elements. Stiffness proportional damping is assumed, with a damping
ratio of 2% of critical at the fundamental period of the fixed-base structure. For the numerical
analyses, the Newmark beta method is used with a value of of 14 and a time increment of
0.01 s. The dimensions of the leadrubber bearings used for the analyses are shown in Table I. The
new bearing mechanical model is used to represent these leadrubber bearings and the two other
mechanical models previously discussed are also used to help in understanding the effect of the
varying vertical load on system response. The first is the uncoupled model, which is the typical
modeling approach used by structural designers. The uncoupled model consists of uncoupled shear
and axial springs and does not include any interaction between the horizontal and vertical forces.
The second is the nonlinear KohKelly model, which was previously proposed by the authors [8].
The nonlinear KohKelly model is the KohKelly model with the enhancement of nonlinear shear
and rotational springs and includes the effect of interaction between horizontal and vertical forces.
A limitation of the nonlinear KohKelly model is that it assumes constant imposed vertical load in
order to define rotational behavior of bearings and it cannot capture the variations in the vertical
load on bearings during an earthquake excitation.
Earthquake ground motions

The earthquake ground motions used for the nonlinear time-history analyses are the 1940 El Centro
(NS), 1968 Hachinohe (NS), and 1958 Taft (EW) records. To gain insight to the behavior of
isolated buildings subject to large overturning forces caused by earthquake excitations ranging from
DOI: 10.1002/eqe
design to ultimate levels, each of these motions is input at four different intensities, corresponding
to peak ground velocities of 25, 50, 75, and 100 cm/s. The 25 and 50 cm/s velocity levels are the
two design levels specified by the Japanese seismic design code for ordinary as well as seismically
isolated buildings [15]. The two higher velocities, 75 and 100 cm/s, are assumed as ultimate level
excitations in this study and are sometimes also used for beyond design level analyses of isolation
systems in Japan.
Analysis results
Table II gives peak response values for the 15- and 20-story building models obtained from the
nonlinear time-history analyses for the three earthquake ground motions at PGV of 100 cm/s. In
most cases, the response values in Table II for the uncoupled model, the nonlinear KohKelly
model, and the new model are almost same, and the peak response displacements of the first
floor (D1F ) are less than 500 mm (corresponding to a leadrubber bearing shear strain of about
Table II. Peak response values from 15- and 20-story model analyses.
Uncoupled Nonlinear KohKelly
model model New model
15 Story model A1F (g) 0.218 0.204 0.216
El Centro, 1940 Aroof (g) 0.299 0.261 0.276
(NS) D1F (mm) 448 457 462
PGV = 100 cm/s S15F 0.298 0.261 0.276
15 Story model A1F (g) 0.137 0.391 0.181
Hachinohe, 1968 Aroof (g) 0.176 0.515 0.237
(NS) D1F (mm) 584 570 597
PGV = 100 cm/s S15F 0.176 0.518 0.237
15 Story model A1F (g) 0.187 0.192 0.173
Taft, 1958 Aroof (g) 0.233 0.249 0.227
(EW) D1F (mm) 488 489 495
PGV = 100 cm/s S15F 0.233 0.249 0.227
20 Story model A1F (g) 0.277 0.286 0.290
El Centro, 1940 Aroof (g) 0.412 0.364 0.397
(NS) D1F (mm) 420 397 402
PGV = 100 cm/s S20F 0.411 0.364 0.396
20 Story model A1F (g) 0.178 0.335 0.232
Hachinohe, 1968 Aroof (g) 0.264 0.456 0.349
(NS) D1F (mm) 518 507 524
PGV = 100 cm/s S20F 0.264 0.452 0.347
20 Story model A1F (g) 0.253 0.214 0.229
Taft, 1958 Aroof (g) 0.322 0.290 0.307
(EW) D1F (mm) 466 475 475
PGV = 100 cm/s S20F 0.321 0.290 0.305
A1F or Aroof , peak response acceleration of the first floor or the roof, respectively; D1F , peak
response displacement of the first floor; S15F or S20F , peak story shear ratio of the 15th or 20th

floor, respectively; Story shear ratio of the ith floor of an N -story building is SiF = Q i / Nj=i+1 w j ,
where Q i is the shear force acting on the ith floor and wi is the weight supported by the ith floor.
DOI: 10.1002/eqe
250%). However, for both buildings subjected to the 1968 Hachinohe (NS) motion at PGV of
100 cm/s, the response values obtained using the three isolator models are different. The peak
response displacements of the first floor (D1F ) given by the new model are 597 and 524 mm
and the leadrubber bearing shear strain is about 290 and 260%, respectively, for the 15- and
20-story buildings. These levels of shear strain reach the region where the influence of shearaxial
interaction appears, as previously shown in Figures 7 and 10.
Figure 13 shows the right-side isolator shear hysteresis loops for the two buildings and the
three different models for the 1968 Hachinohe (NS) motion at PGV of 100 cm/s. These are the
largest isolation system displacements of all of the analysis cases. In Figure 13, the three models
show generally similar hysteresis loop shapes; however, the loops obtained from the nonlinear
KohKelly model and the new model show some indication of shear stiffness deterioration, and
the loops given by the nonlinear KohKelly model also show some buckling behavior.
Figure 14 shows story acceleration, displacement, and story shear ratio response profiles for the
following three cases: the 1968 Hachinohe (NS) and 1958 Taft (EW) motions at 100 cm/s PGV
input to the 15-story building model and the 1968 Hachinohe (NS) motion at 100 cm/s PGV
input to the 20-story building model. In Figure 14(b), the superstructure response profiles given
by the three different isolator models are similar because the peak response displacement of the
isolation bearings does not reach the stiffening or buckling region. In Figures 14(a) and (c), the
superstructure response profiles given by the three different isolator models are clearly different.
In these two cases, comparing the superstructure response values obtained using the three isolator
models, the results obtained from the two models that include the axialshear interaction show
(a)
(b)
Figure 13. Hysteresis loops for right-side isolator from seismic response analyses:
(a) 15-story model, 1968 Hachinohe (NS) PGV = 100 cm/s input and (b) 20-story
model, 1968 Hachinohe (NS) PGV = 100 cm/s input.
DOI: 10.1002/eqe
(a)
(b)
(c)
Figure 14. Acceleration, displacement and story shear response profiles: (a) 15-story model, 1968
Hachinohe (NS) PGV = 100 cm/s input; (b) 15-story model, 1958 Taft (EW) PGV = 100 cm/s
input; and (c) 20-story model, 1968 Hachinohe (NS) PGV = 100 cm/s input.
DOI: 10.1002/eqe
larger responses than those given by the uncoupled model, and the superstructure response values
obtained from the new model are less than those obtained from the nonlinear KohKelly model.
The differences seen in the response values from the three different isolator models are caused by
the interaction between the shear and axial forces and varying vertical load on bearings.
The results obtained from the nonlinear time-history analyses indicate that the differences in the
responses given by the different isolator models appear when the leadrubber bearing shear strain
level reaches the stiffening or buckling region (greater than about 250%), but when the leadrubber
bearing shear strain level is less than the stiffening or buckling strain, the responses given by the
nonlinear KohKelly model and the new model are similar to those given by the uncoupled model.
It is also observed that the results given by the nonlinear KohKelly model tend to be larger than
those given by the new model. The nonlinear KohKelly model cannot capture the effect of varying
vertical load during dynamic response (since the model implicitly assumes that the applied vertical
load on bearings during earthquake loading is constant). This assumption results in excessive
bearing rotational angle and response values of superstructure. These results indicate that isolator
axial-shear behavior at large shear deformations and large vertical loads should be considered for a
more accurate prediction of individual isolator behavior and for improved prediction of the overall
response of isolated buildings in which overturning response is important.
CONCLUSIONS
A new analytical model for elastomeric seismic isolation bearings is proposed to more accurately
predict the large-displacement response of isolated buildings. The proposed model includes the
interaction between shear and axial forces, nonlinear hysteresis, and dependence on varying vertical
load. The hysteresis models for the shear spring and the series of axial springs in the new model
were developed separately to appropriately represent the large deformation behaviors of elastomeric
bearings.
Cyclic shear tests of a leadrubber bearing under constant vertical load were performed as
part of the development of the new model. The leadrubber bearing tested exhibited stiffening
or buckling behavior influenced by axial stress under large deformations. Through comparison
with the test results, the new model is shown to successfully predict a variety of complex bearing
forcedisplacement relationships under a wide range of vertical load conditions. Cyclic shear tests
under varying vertical load were performed to assess the influence of varying vertical load on
bearing hysteretic behavior. The hysteresis loops obtained from the tests were not symmetric and
showed both stiffening and deteriorating shear stiffness, due to the varying vertical load. The
analytical results given by the new model show very good agreement with the bearing test results.
The simulation analyses also indicate that the new model can successfully predict the influence
of different bearing aspect ratios (S2 ) on large shear deformation behavior under varying vertical
loads.
Nonlinear time-history analyses using the new model were performed to investigate the influence
of shearaxial interaction and of varying vertical load at large deformations on the behavior of
isolation bearings. The results obtained using the new model differ in some cases from those given
by an uncoupled shear and axial spring model and the nonlinear KohKelly model, especially
in the case of bearings experiencing extreme shear deformations (over 250% shear strain) and
large variations of vertical load. The new analytical model is a significant advance toward more
accurate nonlinear time-history analysis of tall isolated structures, and isolated structures in which
DOI: 10.1002/eqe
significant overturning forces cause large variations in vertical load on isolation bearings during
earthquake shaking. Consideration of bearing shear behavior influenced by vertical load under
large deformations should be given for the more accurate prediction of the response of these types
of seismically isolated buildings.
REFERENCES
1. Koh CG, Kelly JM. A simple mechanical model for elastomeric bearings used in base isolation. International
Journal of Mechanical Sciences 1988; 30:933943.
2. Koh CG, Balendra T. Seismic response of base isolated buildings including P effects of isolation bearings.
Earthquake Engineering and Structural Dynamics 1989; 18:461473.
3. Kelly JM. Earthquake-resistant Design with Rubber (2nd edn). Springer: London, 1997.
4. Iizuka M. A macroscopic model for predicting large-deformation behaviors of laminated rubber bearings.
Engineering Structures 2000; 22:323334.
5. Ryan KL, Kelly JM, Chopra AK. Experimental observation of axial load effects in isolation bearings. Proceedings
of the 13th World Conference on Earthquake Engineering, Vancouver, BC, Canada, 2004; Paper No. 1707.
6. Ryan KL, Kelly JM, Chopra AK. Nonlinear model for leadrubber bearings including axial-load effects. Journal
of Engineering Mechanics 2005; 131(12):12701278.
7. Ryan KL, Chopra AK. Estimating the seismic response of base-isolated buildings including torsion, rocking
and axial-load effects. Report No. UCB/EERC-2005/01, Earthquake Engineering Research Center, University of
California, Berkeley, CA, 2005.
8. Kikuchi M, Yamamoto S, Aiken ID. An analytical model for leadrubber bearings under large deformations.
Proceedings of the 8th Pacific Conference on Earthquake Engineering, Singapore, 2007; PaperID 222.
9. Hwang JS, Hsu TY. Experimental study of isolated building under triaxial ground excitations. Journal of Structural
Engineering 2000; 126(8):879886.
10. Buckle IG, Liu H. Stability of elastomeric isolation systems for buildings. International Workshop on Recent
Developments in Base-Isolation Techniques for Buildings, Tokyo, 1992; 153169.
11. Kikuchi M, Aiken ID. An analytical hysteresis model for elastomeric seismic isolation bearings. Earthquake
Engineering and Structural Dynamics 1997; 26:215231.
12. Ikenaga M, Suzuki A, Miyazaki M. Examples which applied anti-seismic devices to existing buildings and the
new approval system for seismic isolator in Japan. Proceedings of the 7th International Seminar on Seismic
Isolation, Passive Energy Dissipation and Active Control of Vibrations of Structures, Assisi, Italy, 2001.
13. Gent AN, Lindley PB. The compression of bonded rubber blocks. Proceedings of the Institution of Mechanical
Engineers 1959; 173:111122.
14. ISO 22762-3. Elastomeric seismic-protection isolators part 3: applications for buildings-specifications. Reference
Number ISO 22762-3:2005(E), 2005.
15. Higashino M, Okamoto S. Response Control and Seismic Isolation of Buildings. Taylor & Francis: London, 2006.
DOI: 10.1002/eqe

A Mechanical Model For Elastomeric Seismic Isolation Bearings Including The Influence of Axial Load

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

A Mechanical Model For Elastomeric Seismic Isolation Bearings Including The Influence of Axial Load

Hochgeladen von

Copyright:

Verfügbare Formate

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2009; 38:157180

A mechanical model for elastomeric seismic isolation bearings

Sachie Yamamoto1, , , Masaru Kikuchi1 , Masaiki Ueda1 and Ian D. Aiken2

Received 20 February 2008; Revised 8 July 2008; Accepted 15 July 2008

Contract/grant sponsor: Publishing Arts Research Council; contract/grant number: 98-1846389

Copyright q 2008 John Wiley & Sons, Ltd.

A NEW MECHANICAL MODEL FOR ELASTOMERIC BEARINGS

Figure 1. Multi-spring mechanical model.

Figure 2. Forces and displacements on the mechanical model.

Equations (1)(4) may be expressed by the following matrix equation:

fmn = Kmn umn (9)

umn = {u a vm

Figure 3. Geometrical relationships of the deformations and forces.

umn = Tumn (13)

fmn = TT fmn (14)

umn = {u a vm m u b vn n }T (15)

where h is the total height of the bearing.

fmn = Kmn umn (18)

Numerical analysis procedure

r Kab = K11 K12 K1

S = (u b u a )+ h2 (n +m )

In order to predict the large-deformation behavior of elastomeric bearings, nonlinear hysteresis

F2 = u Fm {12ea(1x) +b(1 x)ec(1x) } (36)

four loading states:

 = c E comp (c ) (49)

E comp = 12 E init (50)

BEARING TESTS AND SIMULATION ANALYSES

Constant vertical load

(a) (b) (c)

(a) (b) (c)

Varying vertical load

SEISMIC RESPONSE ANALYSES

Figure 12. Structure models for seismic response analyses.

Table I. Dimensions of leadrubber bearings used for seismic response analyses.

Earthquake ground motions

Das könnte Ihnen auch gefallen

= c E comp (c ) (49)