International Journal of Non-Linear Mechanics 83 (2016) 65–77

Contents lists available at ScienceDirect

International Journal of Non-Linear Mechanics

journal homepage: www.elsevier.com/locate/nlm

The impact of the concave distribution of rolling friction coefficient on

the seismic isolation performance of a spring-rolling system
Biao Wei a,b,n, Peng Wang a,b, Weian Liu c, Menggang Yang a,b, Lizhong Jiang a,b
a
School of Civil Engineering, Central South University, Changsha 410075, China
b
National Engineering Laboratory for High Speed Railway Construction, Changsha 410004, China
c
Department of Structural Engineering, University of Californian at San Diego, La Jolla, CA 92093, USA

art ic l e i nf o a b s t r a c t

Article history: A complex yet realistic nonuniform rolling friction force distribution of a spring-rolling isolation system
Received 11 February 2015 could lead to great complexity in determining its seismic response. This paper investigates the isolation
Received in revised form performance of a spring-rolling isolation system assuming that the rolling friction force gradually and
5 April 2016
linearly increases with the relative displacement between the isolator and the ground. A series of ground
Accepted 6 April 2016
motions with different characteristics were applied to this system. The analysis results show that the
Available online 9 April 2016
considered concavely distributed friction force is capable of dissipating the earthquake energy, and it is
Keywords: also able to modify the structural natural period. These merits combined help to improve the isolating
Structure efficacy of the spring-rolling isolation system compared with scenarios with uniform distribution pat-
Isolation
tern, and more importantly lead to a relatively optimum isolation state, avoiding a sudden amplification
Concave friction distribution
of the structural seismic response, regardless of the input motion characteristics.
Spring
Seismic performance & 2016 Elsevier Ltd. All rights reserved.

1. Introduction for a wide range of disturbance frequencies and intensities. Later,

Earthquakes have been increasingly inducing serve damages to
building and bridge structures in recent decades. To mitigate the
seismic demand to the structures, a variety of isolation devices
have been developed and demonstrated its seismic resistant
merits. However, for the traditional isolation devices, e.g. lami-
nated rubber bearing and lead rubber bearing isolator, have ten-
dency to amplify the demand to the isolated structure when
subjected to high-intensity ground shaking [1,2].
In order to achieve an optimum isolation mechanism, an old
type isolation device called rolling-based isolation, has been re-
fined in the last two years (2014 and 2015). For example, Harvey
and Gavin [3] proposed a mathematical model for double rolling
isolation systems (RISs) and validated it against experimental data.
In addition, the effect of the initial conditions, the mass of the
isolated object, and the amplitude and period of the disturbance
on the system's performance were studied. Simultaneously, Har-
vey, Zehil and Gavin [4] presented a simplified model that was
applicable to RISs with any potential energy function. The model
has demonstrated its effectiveness in predicting peak responses
n
Corresponding author at: School of Civil Engineering, Central South University,
Changsha 410075, China.
E-mail addresses: weibiao@csu.edu.cn (B. Wei),
wangpeng192318@csu.edu.cn (P. Wang), wel007@ucsd.edu (W. Liu),
mgyang@csu.edu.cn (M. Yang), lzhjiang@csu.edu.cn (L. Jiang).
Harvey and Gavin [5] proposed a novel reduced order modeling
approach to examine the performance of structures incorporated
with lightly- and heavily-damped RISs. Furthermore, Ismail and
Casas [6–7] proposed and studied a roll-n-cage (RNC) isolator, and
the analysis results show that RNC is able to protect cable-stayed
bridges against seismic demand induced by near-fault ground
motions. Furthermore, Ismail [8] demonstrated that the RNC iso-
lator could be an efficient seismic design strategy particularly for
structures located at the fault zones. Wanget et al., [9] numerically
studied the sloped multi-roller isolation which was typically used
to protect equipment and facilities, and the results show an ex-
cellent in-plane seismic isolation performance. Chunget et al., [10]
suggested that an isolation system might not be very effective
when an inappropriate damping is selected, and proposed a the-
oretical method that can be used to determine the optimal fric-
tional coefficient of an isolation system. Ortiz, Magluta and Roit-
man [11] validated a numerical model developed for dynamic
analysis of buildings with roller seismic isolation bearings against
experimental results.
In addition, similar extensive series of research were conducted
on rolling-based isolation devices. Jangid and Londhe [12] devel-
oped a theoretical formulation to obtain seismic responses of a
multistory building supported by elliptical rolling rods in 1998,
and analysis results show that the device is quite effective in re-
ducing the seismic response of the system without undergoing
large base displacements. Jangid [13] investigated the stochastic

http://dx.doi.org/10.1016/j.ijnonlinmec.2016.04.001
0020-7462/& 2016 Elsevier Ltd. All rights reserved.
66 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77
response to the earthquake motion of flexible multi-storey shear
type buildings isolated by rolling rods with a re-centering device,
and the research outcome continue to illustrate the effectiveness
of the rolling rods in which help to reduce the stochastic dynamic
response of the structure. In 2004, Antonyuk and Plakhtienko [14]
analyzed the possible states of a system of interacting solids with
rolling friction and unilateral sliding friction bonds, and con-
sidered that the proposed strategy can be utilized for seismic
isolation purpose. In 2010, Lee [15,16] proposed a roller seismic
isolation bearing for use in highway bridges which combined the
rolling mechanisms, a self-centering capability, and some unique
friction devices for supplemental energy dissipation. After in-
vestigating seismic behaviors of the proposed bearing through
parametric studies, Lee suspected there was something wrong
with the calculation method in AASHTO Specifications and sug-
gested further investigations. In 2013, Harvey and Gavin [17]
presented the modeling of a rolling isolation platform built from
four pairs of recessed steel bowls, which could be used to protect
objects from the hazards of horizontal shaking. The numerical
results showed that uni-axial models could not be used to predict
responses of these systems. Simultaneously, Harvey, Wiebe and
Gavin [18] analyzed the chaotic response of a similar rolling-
pendulum vibration isolation system. Rich chaotic behavior is ex-
hibited in the case considering impacts of the ball-bearing with
the bowl lip. As to isolate an entire raised floor in a building, Cui
[19] performed a series of experiments on a concrete ball-in-cone
isolator with solid rubber and polyurethane balls in 2012, and
identified the practicability of the isolation system. Similarly, Luís
Guerreiro [20] carried out a seismic test and a numerical modeling
of a rolling-ball isolation system to protect some light structures in
2007, and the results showed an effective reduction of the accel-
eration levels induced in the isolation structures. For the purpose
of prolonging the isolation system's life span of service, Tsai [21]
proposed a static dynamics interchangeable-ball pendulum sys-
tem (SDI-BPS) in 2010. Several general steel balls provided sup-
ports to long terms of service loadings. When an earthquake
happened, these balls did not work any more, and a damped steel
ball surrounded by damping materials began to uphold the ver-
tical loads and supply additional damping to the bearing by de-
forming the damping material. Kurita [22] developed a similar
device for seismic response reduction, and the peak acceleration
amplitude was decreased by about 50–90%. In 2012, Nanda [23]
considered that the base isolation in the form of pure friction (P-F),
among all other isolation methods developed so far, was the
simplest one, which could be easily applied to low cost brick
masonry buildings. Furthermore, the P-F isolation is one of the
best alternatives for reducing earthquake energy transferred to
superstructure during strong earthquake.
In all these studies, although the rolling friction isolation device
reduces structural damage caused by earthquakes, the induced
structural displacement may be very large and difficult to control
[24,25]. Therefore, some restoring devices, such as springs, are
usually added to the isolation device to provide the restoring force,
which can eliminate excessive relative displacement and reduce
the structural residual displacement [26,27]. However, how to
combine the spring device and the rolling friction device to obtain
an optimum isolation performance needs further investigation. In
addition, the previous researches and applications usually assume
all of the rolling friction coefficients on the contact surface as a
constant value, i.e. the distribution of rolling friction coefficient on
the entire contact surface is absolutely uniform, to simplify the
calculation process of the structural seismic response. But the as-
sumption is not reasonable in some sense [28–30]. Because the
contact surface is usually rough in fact, and the according rolling
friction coefficient on the contact surface is uneven [31,32]. The-
oretically, the uneven distribution of rolling friction coefficient
M
K
2 1 0 1 2
Increment ratio of concave friction distribution
Concave distribution of friction coefficient
3 2 1 0 1 2 3
Fig. 1. The spring-rolling isolation system.
may lead to great uncertainty in the structural seismic responses
[33,34].
In view of this, how to use the uneven distribution of rolling
friction coefficient to improve the isolation performance is an in-
teresting topic. This paper artificially makes the uneven distribu-
tion of rolling friction coefficient to be concave as shown in Fig. 1.
The rolling friction coefficient in the center of the contact surface
is the smallest one, and it gradually increases when the relative
displacement between the rolling ball and the ground increases.
The friction coefficient increment of unit length is defined as the
increment ratio of the concave friction distribution. A spring is
added to a contact surface with this concave friction distribution
to form a spring-rolling isolation system as shown in Fig. 1. The-
oretically, based on the restoring force provided by the spring and
the concave friction distribution, the isolation structure may not
displace considerably during an earthquake, and the isolator may
roll back to the original position without a significant residual
displacement after an earthquake. The main purpose of this paper
is to utilize a numerical analysis to detailedly analyze the impact of
the concave friction distribution on the seismic performance of the
spring-rolling isolation system under different ground motions.
2. Numerical analysis method
Based on the motion characteristics of a spring-rolling isolation
system, this section establishes a numerical analysis procedure.
2.1. Movement characteristics
Fig. 1 schematically shows a SDOF pier system supported by
spring-rolling device where the rolling ball isolates the super-
structure from the ground. Simultaneously, the friction action of
the rolling ball dissipates the seismic energy. The spring element
represents the elastic restoring component which could help to
avoid the excessive structural relative displacement and reduce
the structural residual displacement. In this simplified system,
three important design variables include the rolling friction coef-
ficient μ, spring constant K , and the increment ratio of the concave
friction distribution. The system with different combination of
these variables could result in three different scenarios with dif-
ferent characteristics as follows:
(1) A system with a significantly small μ and significantly large K .
This indicates that the system has a large stiffness but sig-
nificantly small friction force. In this regard, this system can be
identified as a classic linear oscillator. This characteristic might
 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77
be present when the system sustains a very small relative
displacement as indicated in Fig. 1. As the ratio of the spring
constant K to the rolling friction coefficient μ becomes larger,
the system is able to return to the original position without
experiencing a significant residual displacement during an
earthquake. However, the obvious natural period of the isola-
tion structure implies a potential resonance if the low-fre-
quency components of an earthquake is close to the structural
natural period. If the response is amplified due to the
resonance, the system no longer belongs to this stage and it
may enter each of the other cases below.
(2) A system with a significantly large μ and significantly small K .
The natural period of this system is governed by the friction
force due to the small elastic stiffness. The corresponding
isolation structure has no obvious vibration period, and be-
haves like a pure rolling isolation system. Therefore, the sys-
tem may not experience response amplification due to the
resonance with the dominant period of an earthquake. This
phenomenon approximately exists in the case with the very
large relative displacement in Fig. 1, which has a large rolling
friction coefficient μ far away from the center of the isolation
device and a relatively small spring constant K . And the large
spring force due to the large relative displacement will keep
the isolation system to avoid excessive relative displacement.
(3) A system with comparable μ and K . In this case, the corre-
sponding isolation structure will behave between motion
cases of the traditional elastic isolation system and the pure
rolling isolation system.
2.2. Mathematical formulation
As for the seismic response of the spring-rolling isolation sys-
tem, both the ground and the structure are moving. In order to
clearly describe the motions of the ground and the structure in
math expressions, the absolute displacement coordinate system is
defined in Fig. 1, where d̈e , dė , de are defined as the acceleration,
absolute velocity and absolute displacement of the ground motion,
respectively. Likewise, d̈s , dṡ , ds are defined in the same way for the
structure. In addition, the relative displacement coordinate system
is also defined in Fig. 1. And d̈ , d ̇, d are defined as the relative
acceleration, velocity and displacement of the structure,
respectively.
Based on the comparison between the ground motion velocity
dė and the structural velocity dṡ , the seismic responses of the
isolation system are categorized into three cases:
̇ 0 indicates that the relative motion appears be-
(1) dė > dṡ or d>
tween the structure and the ground, and the combination
force of the spring force and the friction force, acting on the
structure, is (μmg − Kd ). The according equation of motion of
the system is
md¨s = μmg −Kd or md¨ + Kd = μmg − md¨ e
(2) dė < dṡ ord<
̇ 0 is another case that is similar to case (1). But the
combination force of the spring force and the friction force is
( − μmg − Kd ), in which the direction of the friction force is
opposite to that of case (1). The according equation of motion
of the system is
md¨s = − μmg − Kd or md¨ + Kd = − μmg − md¨ e
(3) dė =dṡ ord=
̇ 0 implies that the structure does not move relative
to the ground, and there are two possible phenomena for it: i)
both the structure and the ground are immobile, which
usually exists when the earthquake just starts; and ii) both the
structure and the ground are moving, and they sometimes
67
have the same absolute velocity. In order to forecast the next
motion characteristics of the structure, it is necessary to
compare the absolute values of the acceleration |(d¨ e +Kd/m)|
and μg .
(i) When |(d¨ e +Kd/m)|≤μg , the structure, subjected to the inertia
force by the ground motion and the spring force, will still
move along with the ground, and the structural acceleration d̈s
will be equal to the acceleration d̈e of the ground motion.
(ii) When |(d¨ e +Kd/m) >μg , the structure will move relative to the
ground at the next time, and the combination force of the
spring force and the friction force, acting on the structure, will
be ( ± μmg − Kd ), in which the sign ± will depend on the di-
rection of the acceleration d̈e of the ground motion.
In addition, as for the cases of d> ̇ 0 and d<
̇ 0, if the conditions of
̇ ¨
|d−Kd∆ti/m|≤μg ∆ti and |(de +Kd/m)|≤μg are both satisfied, the
structure is expected to move along with the ground at the next
time, i.e. d= ̇ 0, including the immobile state, where ∆t =t −t ,
i i i−1
while ti is the time at the ith step of the ground motion input.
Through combining this case with case (3) (ii), there is a de-
termined conversion relationship between the three relationships,
i.e. d>̇ 0, d <
̇ 0 and d= ̇ 0, of the structural velocity d ̇ and the ground
s
motion velocity de . ̇
2.3. Computer program
Based on the foregoing mathematical formulation, a computer
program was compiled using Tcl/Tk programming language. This
computer program was an independent program rather than a
secondary development subprogram of large software.
The computer program was composed of five modules, including
building structure model, inputting ground motion, rolling state de-
termination, non-rolling state determination and results output. On
the basis of the mathematical formulation in Section 2.2, these five
modules were integrated to create a complete analysis program.
3. Calculation process
To examine the effect of the concave friction distribution on the
seismic isolation performance, this section defines the specific
values for parameters of the spring-rolling isolation system, in-
cluding the rolling friction coefficient, the spring constant, and the
input motion. Each of these parameters was selected for different
values. Many calculation cases of the isolation system in Fig. 1
were obtained as a result of combing these parameters with dif-
ferent values.
3.1. Structure model
As demonstrated in the related studies, the superstructure is
(1)
assumed as a rigid body in the computer program since the stiff-
ness of the isolation device was far less than that of the structure
[26]. The structure mass was set to be 300t and the spring con-
stant of the isolation device adopted 200, 400, 600, 800 and
1000 kN/m, respectively. And thus all of the structural natural
periods were 7.69, 5.44, 4.44, 3.84 and 3.44 s, respectively. Because
they were much larger than the predominant period of the com-
(2)
mon earthquakes, the probability of resonance was quite little.
Shear keys were considered and designed in the isolation device to
meet the requirements under normal loadings and failure was
only permitted under earthquakes.
For the concave friction distribution, the rolling friction coef-
ficient in the center of the contact surface was the smallest one
68 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77

Fig. 2. Earthquake input.

which was assumed to be a constant value of 0.005 in this paper.

When the relative displacement between the structure and the
ground increased, the rolling friction coefficient on the contact
surface increased. The slope of the concave distribution of friction
coefficient as shown in Fig. 1 was defined as the increment ratio of
the concave friction distribution. And the increment ratio of the
concave friction distribution in Fig. 1 was assumed to be 0, 0.005,
0.010, 0.015, 0.020, 0.025 and 0.030 m 1, respectively. This linearly
varying friction coefficient could be achieved in practical applica-
tions. When a steel ball was rolling on a smooth surface made of
steel, the rolling friction coefficient was about 0.005 [35]. As the
smooth surface got rougher along the path of motion in compar-
ison to that in the center of the contact surface, the rolling friction
coefficient increased.

3.2. Ground motion

As for each response spectrum for the soil profile I, II, III, and IV,
in which the site types of soil profile were from stiff to soft, in
Chinese criteria as shown in Fig. 2(a) (JTJ 004-89) [36], one ac-
celerogram was generated by Simqke procedure to be the ground
motion input of the structural model [37]. One representative
ground motion out of four was shown in Fig. 2(b). Other motions
were not presented due to the similarity to Fig. 2(b).

3.3. Calculation cases

35 cases were obtained by combining 1 structure mass,

5 spring constants and 7 increment ratios of the concave friction
distribution.
As for each case, each accelerogram in Section 3.2 was input as
the ground motion, whose peak ground accelerations (PGA) were
adjusted to be 0.2, 0.4, 0.6 and 0.8 g, respectively. And thus 560
cases were generated for the further calculation.
Then each case was calculated by the computer program, and a
large number of seismic responses were obtained. The following
sections discussed those results, but only classical and common
results were discussed in detailed manner due to space limitations Fig. 3. Effect of the increment ratio of concave friction distribution on the struc-
while other results were considered but not listed. tural maximum acceleration.
 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77
Fig. 4. Effect of the spring constant on the structural maximum acceleration.
69
4. Peak acceleration response
This section analyzes the structural maximum acceleration re-
sponse, as it directly indicates whether the earthquake will da-
mage the isolated structure or not.
4.1. The effect of the concave friction distribution
Fig. 3 shows the effect of the increment ratio of concave friction
distribution on the structural maximum acceleration. It is ob-
served from this figure that the structural maximum acceleration
gradually decreases with the increment ratio of concave friction
distribution for most calculation cases. This decreasing trend is
more significant and more apparent when PGA is larger and the
site type of soil profile is softer, especially for calculation cases
with the larger spring constant.
4.2. The effect of the spring constant
Fig. 4 shows the influence of the spring constant on the
structural maximum acceleration. This figure illustrates that the
structural maximum acceleration gradually increases with the
spring constant, except that some fluctuations appear when the
value of the spring constant is within a certain range. In addition,
this increasing trend is generally more significant when PGA is
larger and the site type of soil profile is softer, especially for cal-
culation cases with the less increment ratio of concave friction
distribution. In contrast, as the increment ratio of concave friction
distribution increases, the fluctuation trend of the structural
maximum acceleration is more remarkable.
4.3. The effect of different ground motions
Fig. 5 demonstrates that the site type of soil profile can sig-
nificantly affect the structural acceleration response subjected to
earthquakes. The softening of soil profile of earthquakes leads to
the significant increasing of the structural maximum acceleration.
Furthermore, this increasing trend is more remarkable under the
conditions of the less increment ratio of concave friction dis-
tribution, the larger PGA and the larger spring constant.
Figs. 3, 4 and 5 also validate that the uncertain of PGA is one of
the most influence factors affecting the structural acceleration
responses. It is observed from these figures that the structural
maximum acceleration nonlinearly increases as PGA becomes
larger for most calculation cases. Simultaneously, if an isolation
system with a stronger spring is subjected to an earthquake in
softer soil, the increasing of PGA can significantly leads to a much
larger structural acceleration. Fortunately, the concave friction
distribution can effectively reduce the increasing trend of struc-
tural maximum acceleration caused by the underestimation of
PGA. On the contrary, if an isolation system with a softer spring,
which is like a linear oscillator with a longer natural period as
shown in Fig. 5(a) and (b), is subjected to an earthquake in stiffer
soil, the increasing of PGA only has weak effect on the structural
acceleration.
4.4. Discussion
For an elastic structure with a constant natural period, the ratio
of the structural maximum acceleration to PGA should remain
constant regardless of earthquake intensity. However, Figs. 3,
4 and 5 show that the acceleration ratio is dependent on the
earthquake intensity. Therefore, the rolling friction plays an im-
portant role in dissipating the earthquake energy and changing
the natural period of the system. The according system is a non-
linear system.
70 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77

Fig. 5. Effect of the site type of soil profile on the structural maximum acceleration. Fig. 6. Friction force, spring force and the combination force.

Since the structure is assumed as a rigid body in the calculating force is the combination force of the spring force and the friction
process, the structural maximum acceleration is dominated by the force as shown in Fig. 6, which are both influenced by the struc-
maximum external force acting on the structure. The external tural relative displacement for the isolation system in Fig. 1.
 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77
Generally, increasing the rolling friction coefficient can dissipate
the earthquake energy to reduce the structural relative displace-
ment and the spring force, however, the friction force increases. In
this case, if the reduction of the spring force is larger than the
increment of the friction force, the combination force of the spring
force and the friction force decreases, and accordingly, the struc-
tural maximum acceleration decreases. Otherwise, the structural
maximum acceleration increases.
As explained in Section 3.1, with regard to the concave friction
distribution except the small and even zero increment ratio cases
in Fig. 1, the friction action around the original position of the
contact surface is more insignificant comparing to that of other
positions. Under the condition of small relative displacement, the
isolation system in Fig. 1 functions like a traditional elastic isola-
tion system. In this case, increasing the rolling friction coefficient,
due to the concave friction distribution and the possibly enlarged
relative displacement under earthquakes, can efficiently dissipate
the earthquake energy and reduce the structural relative dis-
placement increment, and accordingly, the spring force increment
significantly decreases. At this time, the reduction of the spring
force increment is larger than the increment of the friction force,
and thus the structural maximum acceleration can not increase
significantly in comparison to a spring-rolling isolation system
with the uniform friction distribution under the same earthquake.
Furthermore, in the same condition, the structural relative
displacement in the same event of an earthquake will decrease
due to more damping provided by increasing the increment ratio
of concave friction distribution as shown in Fig. 3. And the re-
duction of the spring force, but not the reduction of the spring
force increment, will be larger than the increment of the friction
force, and thus the structural maximum acceleration will decrease
as shown in Fig. 6(b) and (d).
For the same isolation system in Fig. 1, a larger earthquake,
including a softer site type of soil profile and a larger PGA by
comparing Fig.6(a) and (c) with Fig. 6(b) and (d), or a larger spring
constant by comparing Fig. 6(a) and (b) with Fig. 6(c) and
(d) means that the spring performs a more important role in the
isolation system, which would function more like a traditional
elastic isolation system. In this case, more damping provided by
increasing the increment ratio of concave friction distribution will
result in a more significant decrease of the structural maximum
acceleration.
In contrast, as the spring can not dissipate any seismic energy,
increasing the spring constant of the spring-rolling isolation sys-
tem in Fig. 1 subjected to an earthquake naturally increases the
spring force. In this case, the structural relative displacement is
reduced and the friction force of the isolation device subsequently
decreases due to the concave friction distribution of the contact
surface in Fig. 1. In general, the increment of the spring force is
larger than the reduction of the friction force by comparing Fig. 6
(a) and (b) with Fig. 6(c) and (d), and accordingly, their combi-
nation force and the structural maximum acceleration increase.
However, sometimes the increment of the spring force may be less
than the reduction of the friction force, and it results in the de-
creasing of the combination force and the structural maximum
acceleration. This case mainly occurs when the spring constant,
the increment ratio of concave friction distribution and the site
type of soil profile are all large. Under these conditions, the iso-
lation layer is not easy to roll, and the isolation system nearly has
natural period components. Simultaneously, a softer site type of
soil profile means more long-period components of the corre-
sponding seismic wave. In theory, the local resonance phenom-
enon is easier to appear under these conditions, and leads to the
sudden increment of the structural relative displacement and the
according sudden increment of the combination force of the spring
force and the friction force. However, if the spring constant
71
continuously increases, the natural period of the isolation struc-
ture is changed and shifts away from the local resonance region,
which results in the sudden reduction of the structural relative
displacement and the relevant sudden reduction of the combina-
tion force of the spring force and the friction force. These sudden
increment and reduction phenomena eventually form the fluc-
tuations of the structural maximum acceleration as shown in
Fig. 4. Overall, in aspect of only reducing the structural accelera-
tion, the best choice is to reduce the spring constant. When the
spring constant approaches zero, the structural maximum accel-
eration will be only μg during the entire movement.
The earthquake is a highly unpredictable natural disaster. The
inaccurate estimations of earthquake, which are normal at present,
affect the structural acceleration response. When the site type of
soil profile is softer or PGA is larger by comparing Fig. 6(a) and
(c) with Fig. 6(b) and (d), both the spring force and the friction
force increase due to the concave friction distribution of the con-
tact surface as shown in Fig. 1. Finally, the combination force of
these two forces and the structural maximum acceleration
increase.
5. Peak relative displacement response
The structural relative displacement is also of concern for the
isolated structure, as it indicates if the structure will leave its
foundation, and if the structure will collide with other things.
Therefore, this section analyzes the structural maximum relative
displacement.
5.1. The effect of the concave friction distribution
Fig. 7 shows the effect of the increment ratio of concave friction
distribution on the structural maximum relative displacement. It is
observed from this figure that the structural maximum relative
displacement gradually decreases with the increment ratio of
concave friction distribution. This decreasing trend is more sig-
nificant when PGA is larger and the site type of soil profile is softer.
5.2. The effect of the spring constant
Fig. 8 depicts the influence of the spring constant on the
structural maximum relative displacement. Overall, as the spring
constant increases, the structural maximum relative displacement
fluctuates. In addition, this fluctuation phenomenon is generally
more remarkable when PGA is larger and the site type of soil
profile is softer, especially for calculation cases with the larger
increment ratio of concave friction distribution.
5.3. The effect of different ground motions
Fig. 9 shows the effect of the site type of soil profile on the
structural maximum relative displacement. Overall, it is observed
from this figure that the structural maximum relative displace-
ment firstly increases and then decreases with the site type of soil
profile. This variation trend is more remarkable under the condi-
tions of the larger PGA and the larger spring constant. In contrast,
when the spring constant is relatively small as shown in Fig. 9
(a) and (b), this variation only presents an increasing trend.
Figs. 7–9 also identify that the structural maximum relative
displacement becomes larger as PGA increases, especially when
PGA begins to increase from a small value. Fortunately, this in-
creasing trend can be effectively reduced by concave friction dis-
tribution and increasing its increment ratio.
72 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77

Fig. 7. Effect of the increment ratio of concave friction distribution on the struc- Fig. 8. Effect of the spring constant on the structural maximum relative
tural maximum relative displacement. displacement.
 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77 73

earthquakes in the same graph. On the other hand, it is preferable

to demonstrate the effects of the rolling friction coefficient which
will be detailedly illustrated in the following paragraph.
Theoretically, the rolling friction not only dissipates the earth-
quake energy to reduce the structural relative displacement, but
also shortens the structural natural period. The latter function can
be clearly seen from the graphs. In theory, as the elastic structure
has a fixed natural period, altering PGA of the same seismic wave
will not change the ratios of the structural maximum relative
displacement to the maximum absolute displacement of the
ground motion. However, in the graphs, these ratios are obviously
changed by altering PGA. It can be seen from these graphs that the
friction action has a great impact on the structural natural period,
and the impact is limited by the magnitude of the earthquake. In
an event of a weak earthquake, the rolling state of the isolation
system in Fig. 1 is not obvious and the rolling friction has a sig-
nificant limitation on the structural vibration. In this case, the
structural natural period dramatically and sensitively changes.
When the earthquake is too weak to overcome the rolling friction
to trigger the relative displacement of the isolation device in Fig. 1,
the natural period of the isolation structure will be equal to that of
the non-isolation structure.
In addition, based on Figs. 7–9, it is concluded that the concave
friction distribution is conducive to reducing the structural relative
displacement. This can be detailedly explained in the following
paragraph.
When ve >vs at some time during an earthquake, the friction
force μmg provides a forward acceleration component μg , refer-
ring to the coordinate system in Fig. 1, to the isolated structure. In
this case, there are two possible cases due to the influence of the
concave friction distribution:

(1) If the moving structure is on the left of its original position at

this moment, it means that the rolling friction coefficient on
the contact surface is increasing, and the according increment
of the structural acceleration μg is conducive to reducing the
structural relative displacement.
(2) In contrast, if the moving structure is on the right of its ori-
ginal position now, the rolling friction coefficient on the con-
tact surface is decreasing which is opposite to case (1). And
accordingly, the reduction of the structural acceleration μg ,
however, is also conducive to reducing the structural relative
displacement.

There are similar rules in the case of ve <vs .

In summary, the concave friction distribution exactly has a
function of reducing the structural relative displacement. And in
theory, a larger increment ratio of concave friction distribution
implies a smaller structural maximum relative displacement.
Specific effects are divided into two cases:

(1) When the increment ratio of concave friction distribution in

Fig. 9. Effect of the site type of soil profile on the structural maximum relative
displacement.
5.4. Discussion
The ordinate axis in the above Figs. 7–9, adopting the form of
‘Structural max. relative displacement/Ground's max. absolute
displacement’, is based on two considerations. On the one hand, it
is convenient to draw all of the calculation results under different
Fig. 1 increases from a little value, the corresponding friction
action is so insignificant that a slight ground motion may
result in the vibration of the structure, which is more like a
traditional elastic isolation system. In this case, increasing the
rolling friction coefficient, due to increasing the increment
ratio of concave friction distribution, can efficiently dissipate
the earthquake energy and reduce the structural relative
displacement. In addition, the increment of the rolling friction
coefficient resulted from increasing the increment ratio of
concave friction distribution shortens the structural natural
period so that it can also decrease the structural relative
displacement. Finally, it can be concluded that the interaction
of above two effects of increasing the increment ratio of
concave friction distribution from a little value efficiently
74 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77
decreases the structural relative displacement.
(2) In contrast, if the increment ratio of concave friction dis-
tribution in Fig. 1 continuously increases on the basis of a large
value which is larger than that of case (1), the vibration of the
structure is so difficult that the energy dissipation capacity of
the rolling friction becomes insignificant. In this case, as the
increment ratio of concave friction distribution increases, the
structural relative displacement continuously decreases, how-
ever, with very low speed. In addition, with the continuously
enlarged increment ratio of concave friction distribution, the
structural natural period gets shorter and is closer to some
long-period wave components of an earthquake. And accord-
ingly, the decreasing trend of the structural relative displace-
ment gradually diminishes and is even transformed into the
increasing trend. Based on the two aspects above, increasing
the increment ratio of concave friction distribution, on the
basis of a large value, finally results in slowly decreasing the
structural relative displacement.
When a larger spring constant is adopted in the isolation
structure in Fig. 1 or a larger spring force is resulted from a severe
earthquake, the spring performs a more important role in the vi-
bration of the isolation system, which is more like a traditional
elastic isolation system. In this case, increasing the increment ratio
of concave friction distribution significantly reduces the structural
relative displacement. Only when the increment ratio of concave
friction distribution increases to a much larger value, is the energy
dissipation capacity of the rolling friction significantly reduced.
And accordingly, the structural relative displacement begins to
decrease slowly.
In addition, increasing the spring constant shortens the struc-
tural natural period so that the structural relative displacement
decreases. Simultaneously, as analyzed in the above paragraph, a
larger spring constant or a severer earthquake leads the isolation
system to vibrate more like a traditional elastic isolation system.
This mechanism is validated by the phenomenon, as shown in
Figs. 7–9, that all the corresponding ratios of the structural max-
imum relative displacement to the maximum absolute displace-
ment of the ground motion related to different PGA are closer to
each other when comparing to that of other cases. In this case, if
the structural natural period approaches some long-period wave
components of an earthquake, the structural maximum relative
displacement may fluctuate up and down as depicted in Fig. 8,
which implies that resonance happens.
Most of ground motion characteristics of earthquake can not be
accurately predicted at present, and the inaccurate estimations of
earthquake inevitably affect the structural displacement re-
sponses. When other parameters of the isolation system in Fig. 1
are unchanged, the softening of soil profile extends the wave
periods of an earthquake and accordingly increases the absolute
displacement of the ground motion. Its effects on the structural
maximum relative displacement are divided into two cases:
(1) If the spring constant and the increment ratio of concave
friction distribution of the isolation system in Fig. 1 are both
small, the induced structural absolute acceleration, velocity
and displacement under earthquakes are inevitably small. As
the structural relative displacement is defined as the differ-
ence between the structural absolute displacement and the
ground absolute displacement, it increases as a result of the
opposite trends of the above two absolute displacements
induced by softening the soil profile of an earthquake and
simultaneously setting the above parameters with small
values of the isolation systems. Fortunately, because the
concave friction distribution is conducive to decreasing the
structural relative displacement, the enlarged increment ratio
of concave friction distribution weakens the increasing trend
of the structural relative displacement.
(2) In contrast, when the spring constant is set to be a large value,
the according isolation system has a relatively short natural
period and is prone to resonance under the condition of
softening soil profile to extend the wave periods of an earth-
quake. The resonance leads to the appearance of the peak
value of the structural maximum relative displacement in
Fig. 9(c) and (d). However, as the rolling friction can dissipate
the earthquake energy and the concave friction distribution
can change the structural natural period, the peak value of the
structural maximum relative displacement resulted from re-
sonance is not very large.
The inaccurate estimations of PGA also significantly affect the
structural displacement responses, but in a different way. By
keeping other parameters of the isolation system in Fig. 1 the
same, only increasing PGA increases the absolute displacement of
the ground motion more efficiently than the absolute displace-
ment of the structure. And accordingly, the structural relative
displacement, i.e. the difference between the structural absolute
displacement and the ground absolute displacement, increases.
6. Residual displacement of the system
To reduce the repair cost of the isolation structure, the isolator
is expected to roll back to the original position without a sig-
nificant residual displacement after earthquakes. Therefore, this
section analyzes the residual displacement of the isolation system
in Fig. 1 subjected to different ground motions.
6.1. The effect of the concave friction distribution
Fig. 10 shows the effect of the increment ratio of concave fric-
tion distribution on the structural residual displacement. It is ob-
served from this figure that the distribution range of the structural
residual displacement, induced by earthquakes with different soil
profiles and PGA, becomes larger with the increment ratio of
concave friction distribution. However, the corresponding envelop
value is always less than the maximum theoretical value which
will be discussed in Section 6.4.
6.2. The effect of the spring constant
Fig. 11 depicts the influence of the spring constant on the
structural residual displacement. Overall, the structural residual
displacement, induced by earthquakes with different soil profiles
and PGA, takes on a decreasing trend with the spring constant.
And the corresponding envelop value is always less than the
maximum theoretical value which will be discussed in Section 6.4.
6.3. The effect of different ground motions
Figs. 12 and 13 show the effect of different ground motions
related to different soil profiles and PGA on the structural max-
imum relative displacement. Because the earthquake has no
function of restoring the structure to the original position, the
changing of the soil profile or PGA of the earthquake only leads to
irregular variation trend of the structural residual displacement.
However, based on the fitting curves in Figs. 12 and 13, the
structural residual displacement takes on an increasing trend with
soil profile and PGA.
 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77
Fig. 10. Effect of the increment ratio of concave friction distribution on the struc-
tural residual displacement.
6.4. Discussion
Theoretically, the structural residual displacement is influenced
by the spring, the rolling friction and the earthquake. The spring of
the isolation system in Fig. 1 tends to restore the structure to the
original position, while the rolling friction prefers to accommodate
the structure to any place, and prevents the structure from rolling
back to the original position. Based on the statics equilibrium be-
tween the spring force KΔ and the friction force μmg , the structural
theoretical residual displacement Δ n Fig. 1 can be obtained from the
equation KΔ = μmg . As the rolling friction coefficient μ is the func-
tion of the residual displacement Δ due to the concave friction dis-
tribution, it is expressed as μ=0. 005 + rΔ, in which 0.005 is the
smallest rolling friction coefficient in the center of the contact surface
as described in Section 3.1 and r is the increment ratio of the concave
75
Fig. 11. Effect of the spring constant on the structural residual displacement.
friction distribution. And thus the equation KΔ = μmg is further ex-
pressed as KΔ = (0. 005 + rΔ) mg . Finally, the structural theoretical
residual displacement is Δ = 0. 005mg /(K − rmg ) or
Δ = 0.005/(K /mg − r ). It infers from this equation that the spring is
beneficial to decreasing the structural residual displacement as
shown in Fig. 11, while the rolling friction, including the rolling
friction coefficient in the center of the contact surface and the con-
cave friction distribution, is against decreasing the structural residual
displacement as shown in Fig. 10.
However, the conclusion above is only based on the statics
mechanics. The feeble ground motion before the earthquake
completely stops will break the equilibrium state above and in-
duce the structure to a new equilibrium condition in which the
residual displacement will be less than the theoretical value of
Δ = 0.005/(K /mg − r ) as shown in Figs. 10 and 11.
It is concluded from the above discussions that the structural
residual displacement is less than the value of
Δ = 0.005/(K /mg − r ) in spite of any conditions. Before the
76 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77
Fig. 12. Effect of the site type of soil profile on the structural residual displacement.
Fig. 13. Effect of PGA on the structural residual displacement.
structure completely stops, the isolator will vibrate on the contact
surface between ±0.005/(K /mg − r ) for a moment, and the influ-
ence factors on the relative displacement as analyzed in Section 5
will also affects the structural residual displacement which is the
relative displacement just when the structure stops moving.
7. Conclusions
This paper studies the effect of the concave distribution of
rolling friction force at the foundation-ground interface on its
seismic response under different ground motions through nu-
merical analyzes. The main conclusions can be drawn as follows:
(1) In the spring-rolling isolation system, the rolling friction can
not only dissipate the earthquake energy, but also change
the structural natural period. These functions of rolling
friction are significantly strengthened by the concave fric-
tion distribution.
(2) The concave friction distribution help to improve the isolating
efficacy of the spring-rolling isolation system compared with
scenarios with uniform distribution pattern. In addition, an
increase of the increment ratio of concave friction distribution
is able to effectively reduce the structural peak acceleration,
peak relative displacement, and residual displacement when
the spring constant is appropriately selected.
(3) The considered concave friction distribution pattern consistently
leads to a relatively optimum state of the spring-rolling isolation
system regardless of the unpredictable ground motion char-
acteristics. However, the friction coefficient is assumed to line-
arly increase when the relative displacement between the rolling
ball and the ground increases. The effect of friction uncertainty of
the considered concave friction surface on isolation performance
is another topic of future research.
(4) When subjected to design level earthquakes, the concave
friction distribution combined a reasonable spring constant in
a spring-rolling isolation system is able to render a reduced
peak structural relative displacement, residual displacement
and peak acceleration. The peak structural acceleration is re-
duced with a decrease of the spring constant. However, an
increase of the spring constant could effectively reduce the
peak structural relative displacement and residual displace-
ment. Furthermore, adopting system with a relatively large
increment ratio of concave friction distribution and a rela-
tively small spring constant will not sustain a sudden ampli-
fication of the structural seismic response.
Acknowledgments
This research is jointly supported by the National Natural Sci-
ence Foundations of China under Grant no. 51308549 and
51378504, the Natural Science Foundations of Hunan Province
under Grant no. 2015JJ3159, and the Innovation-driven Plan in
Central South University under Grant no. 2015CX006. The above
support is greatly appreciated.
References
[1] M. Ismail, J. Rodellar, F. Pozo, Passive and hybrid mitigation of potential near-
fault inner pounding of a self-braking seismic isolator, Soil Dyn. Earthq. Eng.
69 (2) (2015) 233–250.
[2] D.M. Siringoringo, Y. Fujino, Seismic response analyses of an asymmetric base-
isolated building during the 2011 Great East Japan (Tohoku) Earthquake,
Struct. Control Health Monit. 22 (1) (2015) 71–90.
[3] P.S. Harvey, H.P. Gavin, Double rolling isolation systems: a mathematical
model and experimental validation, Int. J. Non-Linear Mech. 61 (1) (2014)
80–92.
[4] P.S. Harvey, G.P. Zehil, H.P. Gavin, Experimental validation of a simplified model
for rolling isolation systems, Earthq. Eng. Struct. Dyn. 43 (7) (2014) 1067–1088.
[5] P.S. Harvey, H.P. Gavin, Assessment of a rolling isolation system using reduced
order structural models, Eng. Struct. 99 (2015) 708–725.
[6] M. Ismail, J. Rodellar, F. Pozo, An isolation device for near-fault ground mo-
tions, Struct. Control Health Monit. 21 (3) (2014) 249–268.
[7] M. Ismail, J.R. Casas, Novel isolation device for protection of cable-stayed
bridges against near-fault earthquakes, J. Bridge Eng. 19 (8) (2014) 50–65.
[8] M. Ismail, An isolation system for limited seismic gaps in near-fault zones,
Earthq. Eng. Struct. Dyn. 44 (7) (2015) 1115–1137.
[9] S.J. Wang, J.S. Hwang, K.C. Chang, C.Y. Shiau, W.C. Lin, M.S. Tsai, J.X. Hong, Y.
H. Yang, Sloped multi-roller isolation devices for seismic protection of
equipment and facilities, Earthq. Eng. Struct. Dyn. 43 (10) (2014) 1443–1461.
[10] L.L. Chung, P.S. Kao, C.Y. Yang, L.Y. Wu, H.M. Chen, Optimal frictional coefficient
of structural isolation system, J. Vib. Control 21 (3) (2015) 525–538.
[11] N.A. Ortiz, C. Magluta, N. Roitman, Numerical and experimental studies of a
building with roller seismic isolation bearings, Struct. Eng. Mech. 54 (3) (2015)
475–489.
[12] R.S. Jangid, Y.B. Londhe, Effectiveness of elliptical rolling rods for base isola-
tion, J. Struct. Eng. 124 (4) (1998) 469–472.
 B. Wei et al. / International Journal of Non-Linear Mechanics 83 (2016) 65–77
[13] R.S. Jangid, Stochastic seismic response of structures isolated by rolling rods,
Eng. Struct. 22 (8) (2000) 937–946.
[14] E.Y. Antonyuk, N.P. Plakhtienko, Dynamic modes of one seismic-damping
mechanism with frictional bonds, Int. Appl. Mech. 40 (6) (2004) 702–708.
[15] Y.C. Ou, J.W. Song, G.C. Lee, A parametric study of seismic behavior of roller
seismic isolation bearings for highway bridges, Earthq. Eng. Struct. Dyn. 39 (5)
(2010) 541–559.
[16] G.C. Lee, Y.C. Ou, T.C. Niu, J.W. Song, Z. Liang, Characterization of a roller
seismic isolation bearing with supplemental energy dissipation for highway
bridges, J. Struct. Eng. 136 (5) (2010) 502–510.
[17] P.S. Harvey, H.P. Gavin, The nonholonomic and chaotic nature of a rolling
isolation system, J. Sound Vib. 332 (14) (2013) 3535–3551.
[18] P.S. Harvey, R. Wiebe, H.P. Gavin, On the chaotic response of a nonlinear rolling
isolation system, Phys. D: Nonlinear Phenom. 256–257 (2013) 36–42.
[19] S. Cui, Integrated design methodology for isolated floor systems in single-
degree-of-freedom structural fuse systems (Ph.D thesis), StateUniversi-
tyofNewYork, Buffalo, 2012.
[20] L. Guerreiro, J. Azevedo, A.H. Muhr, Seismic tests and numerical modeling of a
rolling-ball isolation system, J. Earthq. Eng. 11 (1) (2007) 49–66.
[21] C.S. Tsai, Y.C. Lin, W.S. Chen, H.C. Su, Tri-directional shaking table tests of vi-
bration sensitive equipment with static dynamics interchangeable-ball pen-
dulum system, Earthq. Eng. Eng. Vib. 9 (1) (2010) 103–112.
[22] K. Kurita, S. Aoki, Y. Nakanishi, K. Tominaga, M. Kanazawa, Fundamental
characteristics of reduction system for seismic response using friction force, J.
Civ. Eng. Archit. 5 (11) (2011) 1042–1047.
[23] R.P. Nanda, P. Agarwal, M. Shrikhande, Base isolation system suitable for
masonry buildings, Asian J. Civ. Eng. (Buil. Hous.) 13 (2) (2012) 195–202.
[24] D. Kosntantinidis, N. Makris, Experimental and analytical studies on the re-
sponse of freestanding laboratory equipment to earthquake shaking, Earthq.
Eng. Struct. Dyn. 38 (6) (2009) 827–848.
[25] A.D. Lewis, R.M. Murray, Variational principles for constrained systems: The-
ory and experiment, Int. J. Non-Linear Mech. 30 (6) (1995) 793–815.
77
[26] B. Wei, G.L. Dai, Y. Wen, Y. Xia, Seismic performance of an isolation system of
rolling friction with spring, J. Cent. South Univ. 21 (I4) (2014) 1518–1525.
[27] B. Wei, Y. Xia, W. Liu, A.Lateral vibration analysis of continuous bridges uti-
lizing equal displacement rule, Lat. Am. J. Solids Struct. 11 (1) (2014) 75–91.
[28] B. Wei, R.B. Cui, G.L. Dai, Seismic performance of a rolling-damper isolation
system, J. Vibroeng. 15 (3) (2013) 1504–1512.
[29] Y.J. Wang, Q.C. Wei, J. Shi, X.Y. Long, Resonance characteristics of two-span
continuous beam under moving high speed trains, Lat. Am. J. Solids Struct. 7
(2) (2010) 185–199.
[30] C.F. Yin, B. Wei, Numerical simulation of a bridge-subgrade transition zone
due to moving vehicle in Shuohuang heavy haul railway, J. Vibroeng. 15 (2)
(2013) 1062–1068.
[31] C.J. Begley, L.N. Virgin, Impact response and the influence of friction, J. Sound
Vib. 211 (5) (1998) 801–818.
[32] D.G. Flom, A.M. Bueche, Theory of rolling friction for spheres, J. Appl. Phys. 30
(I11) (1959) 1725–1730.
[33] B. Wei, H. YangT., Z. JiangL., Influence of friction variability on isolation per-
formance of a rolling-damper isolation system, J. Vibroeng. 17 (2) (2015)
792–801.
[34] C.S. Yim, A.K. Chopra, J. Penzien, Rocking response of rigid blocks to earth-
quakes, Earthq. Eng. Struct. Dyn. 8 (I6) (1980) 565–587.
[35] E. Wu, T. Guo, A. Li, A study on the rolling friction coefficient and the hor-
izontal driving-force calculation in building monolithic moving engineering,
China Civ. Eng. J. 39 (6) (2006) 57–60 (in Chinese).
[36] Standard of the Ministry of Communications of P.R. China., JTJ004-89 Speci-
fications of Earthquake Resistant Design for Highway Engineering, China
Communications Press, Beijing, 1989 (in Chinese).
[37] Y. Fahjan, Z. Ozdemir, Scaling of earthquake accelerograms for non-linear
dynamic analysis to match the earthquake design spectra, in: Proceedings of
the 14th World Conference on Earthquake Engineering, Chinese Society for
Earthquake Engineering, 2008.