See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/229485199

Experimental and analytical studies on the response of freestanding

laboratory equipment to earthquake shaking

Article  in  Earthquake Engineering & Structural Dynamics · May 2009

DOI: 10.1002/eqe.871

CITATIONS READS

62 260

2 authors:

Dimitrios Konstantinidis Nicos Makris

University of California, Berkeley Southern Methodist University
102 PUBLICATIONS   1,115 CITATIONS    245 PUBLICATIONS   6,125 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Phenomenological Models with Non-Integer Order Derivatives View project

Soil-Pile Foundation-Structure Interaction View project

All content following this page was uploaded by Nicos Makris on 23 October 2018.

The user has requested enhancement of the downloaded file.

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS
Earthquake Engng Struct. Dyn. 2009; 38:827–848
Published online 10 December 2008 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/eqe.871

Experimental and analytical studies on the response of freestanding

laboratory equipment to earthquake shaking

Dimitrios Konstantinidis1, ‡ and Nicos Makris2,3, ∗, †, §, ¶

1 Department of Civil and Environmental Engineering, University of California, Berkeley, U.S.A.
2 Department of Civil Engineering, University of Patras, Patras GR-26500, Greece
3 Earthquake Engineering Research Center, University of California, Berkeley, U.S.A.

SUMMARY
This paper presents results of a comprehensive experimental program on the seismic response of full-
scale freestanding laboratory equipment. First, quasi-static experiments are conducted to examine the
mechanical behavior of the contact interface between the laboratory equipment and floors. Based on
the experimental results, the response analysis that follows adopts two idealized contact friction models:
the elastoplastic model and the classical Coulomb friction model. Subsequently, the paper presents shake
table test results of full-scale freestanding equipment subjected to ground and floor motions of hazard
levels with corresponding displacements that can be accommodated by the shake table at the UC Berkeley
Earthquake Engineering Research Center. For the equipment tested, although some rocking is observed,
sliding is the predominant mode of response, with sliding displacements reaching up to 60 cm. Numerical
simulations with the proposed models are performed. Finally, the paper identifies a physically motivated
intensity measure and the associated engineering demand parameter with the help of dimensional analysis
and presents ready-to-use fragility curves. Copyright q 2008 John Wiley & Sons, Ltd.

Received 5 March 2007; Revised 10 September 2008; Accepted 15 September 2008

KEY WORDS: laboratory equipment; non-structural components; shake table experiments; rocking;
sliding; PBEE; fragility

∗ Correspondence to: Nicos Makris, Department of Civil Engineering, University of Patras, Patras GR-26500, Greece.
†
E-mail: nmakris@upatras.gr
‡
Postdoctoral Scholar.
§ Professor.
¶ Senior Research Engineer.

Contract/grant sponsor: Earthquake Engineering Research Centers Program of the National Science Foundation;
contract/grant number: EEC-9701568

Copyright q 2008 John Wiley & Sons, Ltd.

828 D. KONSTANTINIDIS AND N. MAKRIS

1. INTRODUCTION

During strong earthquake shaking, heavy equipment located at various floor levels of research
laboratories, hospitals, and other critical facilities may slide appreciably, slide-rock, rock, or even
overturn. Rocking response is very sensitive to the geometry of the slender object and the kinematic
characteristics of the ground. Minor variations in the input can result in overturning—catastrophe
[1–3]. Even if overturning does not occur, the high acceleration spikes that develop during impact
of the rocking equipment are a major concern, since they can result in serious damage or loss of
the equipment contents.
Of the possible modes of response, sliding is the most favorable. Nonetheless, excessive sliding
displacements may block a path or doorway that services evacuation. Large displacements of
sensitive/heavy equipment that result in impact with walls or neighboring equipment should be
avoided, since the resulting acceleration spikes endanger the contents or even the equipment itself.
In practice, excessive sliding is prevented by restraining the equipment—commonly by chaining it
to the framing of the nearby wall. Although this may succeed in reducing sliding displacements,
it substantially amplifies accelerations. The problem of equipment sliding has been studied in the
past at various scales by Shao and Tung [4], Lopez Garcia and Soong [5, 6], and Hutchinson and
Chaudhuri [7].
The research presented in this paper is part of a wider study that was set out to apply the
performance-based earthquake engineering (PBEE) methodology [8] as proposed by the Pacific
Earthquake Engineering Research (PEER) Center on a specific testbed: an actual science laboratory
building, herein referred to as UC Science Building. The UC Science Building is located on the
western part of the main UC Berkeley campus and is approximately 1 km west of the Hayward
fault. It is a modern structure, completed in 1988 in order to provide high-tech research laboratories
[9]. The PEER PBEE methodology consists of four stages [8]: hazard analysis, structural analysis,
damage analysis and loss analysis.
In this paper we present experimental and analytical studies that examine the seismic vulner-
ability of freestanding and restrained laboratory equipment located in the UC Science Building
laboratories within several floor levels. The equipment of interest includes low-temperature refrig-
erators, freezers, incubators and other heavy equipment. The study investigates the response of
equipment to moderately strong motions (50 and 10% probability of being exceeded in 50 years)
which result to peak ground displacements or peak floor displacements that the Earthquake Engi-
neering Research Center (EERC) shake table at UC Berkeley can accommodate. Additional results
are presented in the report by Konstantinidis and Makris [10]. The results of tests conducted
on the shake table were used to develop a dimensionless engineering demand parameter (EDP)
(a parameter that quantifies the response of the equipment), as a function of the intensity measure
(IM) (a parameter of the excitation that corresponds to a certain seismic hazard level). Since
uncertainties are inherent in all stages of the PEER methodology [8], a probabilistic approach was
taken. An analysis was performed to generate fragility curves, which give the probability that the
aforementioned EDP will exceed a specific limit c.

2. SEISMIC HAZARD, GROUND AND FLOOR MOTIONS

The seismic hazard study for the UC Science testbed was performed by Somerville [11]. The
seismic hazard on the UC Berkeley campus is dominated by potential ground motions generated

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 LABORATORY EQUIPMENT TO EARTHQUAKE SHAKING 829

from the Hayward fault, which is located approximately 1 km east of the site. The Hayward fault
is a strike-slip fault that has a potential to generate earthquakes having magnitudes as large as
MW = 7.0. The ground motions for the site were selected and scaled to correspond to three hazard
levels: (a) events with probability of exceedance (POE), equal to 50% in 50 years, (b) events
with POE 10% in 50 years, and (c) events with POE 2% in 50 years. For a hazard level equal
to 50% in 50 years, the largest contributions come from earthquakes in the magnitude range of
MW = 5.5 to MW = 6.0. For hazard level equal to 10% in 50 years and to 2% in 50 years, the
largest contributions come from earthquakes in the magnitude range of MW = 6.5 to MW = 7.0. It
is noteworthy that the higher 2% in 50 years do not reflect larger magnitudes (as the Hayward fault
can generate earthquakes only up to MW = 7.0) but rather stronger ground motions with the same
magnitude (with larger standard deviation above the mean) [11]. The motions listed in Table I
have been selected to satisfy (to the extent possible) the magnitude and distance combination from
a strike-slip earthquake on NEHRP-classified SC soil type for moderately strong motions (50 and
10% in 50 years).
The seismic response of the UC Science laboratory building was analyzed by Lee and Mosalam
[9, 12], who developed a sophisticated structural model of the building. Their analyses resulted in
simulated floor motions. Floor motions are of unique interest in assessing the seismic response
of building contents since they differ appreciably from ground motions. Table I lists the recorded
ground-acceleration motions and the simulated floor-acceleration motions that were used as input
for the shake table experiments conducted in this study.
All input motions used for shake table tests in this study were one-directional. Preliminary
dynamic analyses of a sliding block resting on a base subjected to horizontal and vertical earthquake
motion yielded peak sliding displacements that were only slightly amplified when the vertical
component of the excitation was included. The same observation is made in a study by Shao and
Tung [4]. This is because the high-frequency content of vertical acceleration signals in association
with their lower amplitude renders them too feeble to impart a significant amplification of the peak
sliding displacement. There are studies [5, 6], however, that indicate that certain combinations of
the horizontal and vertical components of the excitation may not be ignored.

3. GEOMETRIC AND PHYSICAL CHARACTERISTICS OF THE TEST EQUIPMENT

3.1. Geometric properties

The equipment of interest included incubators, low-temperature freezers, refrigerators and other
heavy laboratory equipment of the UC Science Building at the UC Berkeley campus. In particular,
three pieces of equipment were obtained from the building laboratories in order to examine their
mechanical properties and to perform shake table tests. Figure 1 shows pictures of the equipment,
while Table II lists their geometric and physical characteristics. Figure 2 is a schematic of a piece of
equipment that shows the geometric quantities that are listed in Table II. Each piece of equipment
has two vertical faces, designated here by W for width and D for depth. The stockiness angles
W and  D of a piece of equipment are defined by
   
−1 W −1 D
W = tan and  D = tan (1)
H H

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 830

Table I. Input motions for the shake table tests conducted on the three pieces of laboratory equipment. Recorded peak table accelerations

Copyright q
(PTA) and maximum sliding displacements (Umax ) of the freestanding equipment.
Equipment
FORMA KELVIN ASP
Face Face Face Profile
Hazard level PGA PTA Umax PTA Umax PTA Umax PTA Umax
Earthquake Record (in 50 years) (%) (g) (g) (cm) (g) (cm) (g) (cm) (g) (cm)
Aigion, Greece OTE FP 0.50 0.75 17.5 0.76 15.2 0.86 8.1 0.73 6.1
15/6/1995,
MW = 6.2, D = 5.0 km

2008 John Wiley & Sons, Ltd.

Coyote Lake, California Gilroy Array #6 FN 0.47 0.68 6.6 0.69 5.3 0.73 4.1 0.70 2.5
6/8/1979,
MW = 5.7, D = 3.0 km
Parkfield, California Cholome Array #8 FN 50 0.56 0.70 9.4 0.75 3.3 0.77 2.0 0.70 0.8
27/6/1966, GROUND
MW = 6.0, D = 8.0 km
Parkfield, California Cholome Array #8 FN 50 0.71 1.61 34.3 1.66 22.6 1.7 57.9 1.68 60.2
27/6/1966 6TH LEVEL
Coyote Lake, California Gilroy Array #6 FN 50 0.47 0.75 7.9 0.74 8.4 0.76 9.1 0.75 9.4
6/8/1979, GROUND
MW = 5.7, D = 3.0 km
Coyote Lake, California Gilroy Array #6 FN 50 0.78 0.43 2.0 0.76 18.8 0.77 23.9 0.80 20.3
6/8/1979 6TH LEVEL
D. KONSTANTINIDIS AND N. MAKRIS

Loma Prieta, California Gavilan College FN 10 0.66 0.73 9.7 0.76 5.8 0.77 4.8 0.79 6.6
17/10/1989, GROUND
MW = 7.0, D = 9.5 km
Loma Prieta, California Gavilan College FN 10 0.75 Leg failure 0.64 13.0 0.67 17.8 0.67 17.0
17/10/1989 6TH LEVEL
Tottori, Japan Kofu FN 10 0.69 1.07 35.3 1.07 42.2
6/10/2000, GROUND
MW = 6.6, D = 10.0 km
Tottori, Japan Kofu FN 10 0.54 0.81 13.2 1.08 35.3 0.85 20.3
6/10/2000 6TH LEVEL

DOI: 10.1002/eqe
Earthquake Engng Struct. Dyn. 2009; 38:827–848
 LABORATORY EQUIPMENT TO EARTHQUAKE SHAKING 831

FORMA Incubator Kelvinator Refrigerator ASP Refrigerator

Figure 1. The three pieces of freestanding heavy laboratory equipment that were obtained from the UC
Science research facility for the purposes of this study.

Table II. Geometric and physical characteristics of the equipment.

Weight H W (Wout ) D (Dout ) RW RD W D pW pD
Equipment (N) (cm) (cm) (cm) (cm) (cm) (rad) (rad) (rad/s) (rad/s)
FORMA 3780 228.6 92.7 62.2 123.4 118.4 0.39 0.27 2.44 2.49
Reach-In Incubator (96.5) (78.7)
KELVINATOR 1565 213.4 63.5 66.0 111.3 111.8 0.29 0.30 2.57 2.57
Refrigerator/freezer (83.8) (82.6)
ASP 720 180.3 76.2 58.4 97.8 94.7 0.40 0.31 2.74 2.79
Refrigerator (81.3) (67.3)

The stockiness of a block is an indicator of its disposition to enter rocking motion. The smaller the
stockiness (more slender), the more likely for the equipment to uplift, enter rocking motion, and
possibly overturn. The frequency parameters pW and p D are measures of the size of the equipment
and are given by [1–3, 13]
 
3g 3g
pW = and p D = (2)
4RW 4R D
√ √
where g is the acceleration of gravity and RW = H 2 + W 2 /2, R D = H 2 + D 2 /2 (Figure 2). The
larger the block (larger R), the smaller the p.

3.2. Friction tests

The mechanical properties of the contact interface between the equipment and the laboratory floors
were determined by conducting slow-pull tests on the equipment. The floors throughout the UC
Science Building are lined with vinyl tiles. In order to simulate the actual conditions, a 2.4×3.6 m

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
832 D. KONSTANTINIDIS AND N. MAKRIS

Figure 2. Schematic diagram of the experimental setup for the quasi-static pull tests.

pressboard surface covered with identical vinyl tiles was constructed. Atop it rested the equipment
specimen. Figure 2 shows a schematic of the experimental setup of the quasi-static pull tests
conducted on the equipment.
Figure 3 plots load–displacement curves recorded during the quasi-static pull tests on the three
pieces of equipment shown in Figure 1. The pre-yielding elasticity in the load–displacement curves
originates from the flexure of the legs of the equipment prior to sliding. This pre-yielding elasticity
of the legs and the friction force that develops along the vinyl surface combine to a yielding
mechanism of the interface. Simple idealizations of the yielding mechanism of each interface are
the elastoplastic models shown with dashed lines in Figure 3. The model parameters that define the
elastoplastic idealization are the yield displacement, u y , and the normalized strength, k = Q/mg,
where Q is the post-yield constant force. Another idealization of the contact interface is that of
classical Coulomb friction where a static friction coefficient, s , and a kinetic friction coefficient,
k , are used.
Numerical simulation studies using the elastoplastic model with k values extracted from the
slow-pull tests were conducted after the shake table tests in order to examine the validity of the
mechanical model. The analyses yielded results that were in fair agreement with the experimental
data from the shake table tests (presented in the following section). The predicted response of all
three pieces of equipment was appreciably improved when lower values of their respective friction
coefficients were used. The lower friction levels of the elastoplastic idealization are indicated in
Figure 3 with solid lines. The heavy solid lines shown in Figure 3 correspond to the Coulomb
model with reduced friction coefficients. The reduced values of s and k have been obtained by
best-fitting results from numerical simulations using the commercially available software Working
Model [14] to results obtained from shake table experiments. Table III summarizes the values of
friction coefficients and yield displacements that were obtained from the slow-pull tests and from
the best-fitting procedure.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 LABORATORY EQUIPMENT TO EARTHQUAKE SHAKING 833

Coefficient of Friction Tests Specimen: FORMA Incubator, Weight=3780N

1200
Elastoplastic (Slow Pull Tests) FORMA
uy=0.35cm, μk=0.23
1000
Lateral Load [N]

800
Elastoplastic (Best Fit)
uy=0.23cm, μk=0.15
600

400
Working Model
200 μs=0.18, μk=0.13

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Coefficient of Friction Tests Specimen: KELVINATOR refrigerator, Weight=1570N

800
Elastoplastic (Slow Pull Tests) KELVINATOR
uy=0.07cm, μk=0.28
600
Lateral Load [N]

400

200 Working Model

Elastoplastic (Best Fit)
uy=0.05cm, μk=0.20 μs=0.23, μk=0.17

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Coefficient of Friction Tests Specimen: ASP refrigerator, Weight=720N

400
Elastoplastic (Slow Pull Tests) ASP
uy=0.02cm, μk=0.31
300
Lateral Load [N]

200

100
Elastoplastic (Best Fit) Working Model
uy=0.02cm, μk=0.24 μs=0.28, μk=0.20

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Displacement [cm]

Figure 3. Recorded load–displacement plots obtained from the slow-pull tests (wavy lines);
the elastoplastic idealization with k from the slow-pull tests (dashed line); the elastoplastic
idealization with k from best-fitting (solid line); the Coulomb friction model (Working Model)
with s and k by best-fitting (heavy line).

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
834 D. KONSTANTINIDIS AND N. MAKRIS

Table III. Coefficients of friction and yield displacements obtained from slow-pull tests and from best-fitting
numerical simulation results to experimental results from the shake table tests.
Values used to fit Values used to fit Coulomb
Elasto-plastic Model (Rigid-Plastic) Model
simulation results to simulation results to
experimental results from experimental results from
Values from Slow-Pull Tests the shake table tests the shake table tests
Equipment s k u y (cm) k u y (cm) s k
FORMA 0.30 0.23 0.35 0.15 0.23 0.18 0.13
Incubator
Kelvinator 0.37 0.28 0.07 0.20 0.05 0.23 0.17
Refrigerator
ASP 0.43 0.31 0.02 0.24 0.02 0.28 0.20
Refrigerator

The static friction coefficient, s , is known to exhibit time dependence; with time an object
sinks into the surface of its base, as asperities deform, causing an increase in meniscus and
Van der Waals forces, which are macroscopically manifested by an increase in s [15]. Furthermore,
s depends on the rate of application of the pull force [16–18]. The kinetic friction coefficient,
k , can also exhibit a pressure and velocity dependence [16–18].

4. SHAKE TABLE TESTS OF FREESTANDING LABORATORY EQUIPMENT

The three pieces of equipment shown in Figure 1 were subjected to shake table tests at the UC
Berkeley Earthquake Engineering Research Center’s (EERC) Earthquake Simulator Laboratory.
The same type of pressboard surface that was used as the base for the slow-pull tests was built on
the shake table to support the equipment. Figure 4 shows a photograph of one of the freestanding
equipment resting on the shake table.
The displacement of the shake table and the equipment were measured with wire transducers
attached to a frame fixed on the laboratory floor. Figure 4 shows the locations of the wire transducers
on the test specimen with heavy white lines. Accelerometers were also installed on the positions
shown with black arrows in Figure 4 in order to capture horizontal and vertical accelerations. The
horizontal displacement capacity of the shake table at EERC is ±15 cm. Given this constraint,
experiments at full scale were run only for moderate ground motions (i.e. with hazard level equal
to 50% in 50 years and 10% in 50 years).
In several occasions the equipment while shaken along the primary direction, exhibited rotations
about its vertical axis. In some cases these plane rotations were as small as 0.005 rad, while in
others as large as 0.33 rad, indicating that even when the excitation is one-directional, the response
is in fact in three dimensions.
Table I lists the earthquake records that were used to test the freestanding equipment. Also listed
are the recorded peak table acceleration (PTA) and the equipment peak sliding displacement, Umax .
The PTA level of the records used in this experimental program ranged from 0.43 to 1.70g and

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 LABORATORY EQUIPMENT TO EARTHQUAKE SHAKING 835

Figure 4. The FORMA incubator resting atop the shake table at the UC Berkeley Earthquake Engineering
Research Center. The locations of the wire transducers are indicated with white lines, and the locations
of the accelerometers are indicated with dark arrows.

the Umax from 2 cm to as much as 60 cm. The shake table tests indicated that although some uplift
occurs, the primary mode of response of the three pieces of equipment is sliding. Criteria provided
by Shenton [19] allow the determination of the initiation of a response mode using the values of
 listed in Table II in conjunction with the values of s listed in Table III. If the values of s
obtained from the slow-pull tests are used, slide-rocking is anticipated to initiate for all three pieces
of equipment used in this study. However, sliding dominates the response. In fact, the maximum
recorded uplift rotation in all tests for the Kelvinator refrigerator was only 0.02 rad (0.069) and
only 0.005 rad (0.015) for the ASP refrigerator. The maximum recorded uplift rotation for the
FORMA incubator was 0.13 rad (0.489). Although the FORMA incubator experienced larger
uplift rotations than the other two specimens, these rotations were still well below the level for
overturning (>). Nonetheless, after about 20 tests one of the flimsy leg supports of the FORMA
incubator failed. The resulting instability caused overturning of the specimen.
Figure 5 (bottom window) plots the OTE FP ground-acceleration history recorded during the
1995 Aigion, Greece, earthquake. The graph on the window above the acceleration record plots the
resulting shake table displacement, and the third window from the top plots with a heavy solid line
the recorded sliding displacement of the FORMA incubator. The recorded sliding displacement
history shows that the equipment suddenly slides once a threshold table acceleration is exceeded
(at about 3.8 s). The top two graphs in Figure 5 that plot with heavy solid lines the equipment uplift
and the plane rotation (rotation about the vertical axis) show that the specimen does uplift and
twist just slightly even before the initiation of sliding (at about 3.65 s). In other cases, one mode of
response does not seem to trigger the other, but rather both happen simultaneously. The response
mode coupling is less pronounced for the Kelvinator refrigerator which, although demonstrating
a slightly larger coefficient of friction during the slow-pull tests than the FORMA incubator
(s = 0.37 versus s = 0.30), is stockier ( = 0.289 versus  = 0.266) and therefore less susceptible
to uplift. The Kelvinator refrigerator exhibited relatively small rotations, never exceeding 7% its

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
836 D. KONSTANTINIDIS AND N. MAKRIS

equipment: FORMA incubator motion: Aigion, OTE FP

uplift [rad] 0.1

-0.1
0 2 4 6 8 10
plane rot. [rad]

0.05

-0.05
0 2 4 6 8 10
20
sliding [cm]

15
10
5
0
-5
0 2 4 6 8 10
table disp. [cm]

15
10
5
0
-5
0 2 4 6 8 10
1
table acc. [g]

-1
0 2 4 6 8 10
time [sec]

Figure 5. Response of the FORMA incubator subjected to the OTE FP motion recorded during the 1995
Aigion, Greece, earthquake. The heavy gray lines on the bottom two graphs plot a Type-B trigonometric
pulse that approximates the main pulse of the motion record.

stockiness angle, , in any of the tests. Moreover, the Kelvinator refrigerator exhibited very small
rotation about its vertical axis. The maximum recorded plane rotation in all tests performed on the
Kelvinator refrigerator was 0.05 rad. Figure 6 plots the response of the Kelvinator refrigerator to
the computed [9, 12] 6th-floor motion of the Gavilan College FN (10% in 50 years) record of the
1989 Loma Prieta, California, earthquake.
The ASP refrigerator is considerably stockier in one direction than in the other (W = 0.400
versus  D = 0.313; see Table II and Figure 2). When the table excitation is along the more stocky
direction (W ), the configuration is designated Profile, whereas when the table excitation is along
the more slender direction (D), the configuration is designated Face. The recorded response for all
the shake table tests performed on the ASP refrigerator in the Face configuration and the Profile
configuration is almost identical. The maximum recorded uplift rotation did not exceed 0.005 rad
(1.5% of  D ). Figure 7 plots the recorded response of the ASP refrigerator (Profile configuration)
subjected to the computed [9, 12] 6th-floor motion of the Gilroy Array #6 FN (50% in 50 years)
record of the 1979 Coyote Lake, California, earthquake.
An interesting characteristic to note is the waviness of the heavy solid lines that plot the sliding
displacement. This wobbling is more pronounced for the FORMA incubator whose legs were very
flexible, less pronounced for the Kelvinator refrigerator whose legs are fairly stiff, and almost
non-existent for the ASP refrigerator whose legs are nearly rigid.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 LABORATORY EQUIPMENT TO EARTHQUAKE SHAKING 837

equipment: KELVINATOR refrigerator motion: Loma Prieta, Gavilan College FN 6th Floor (10% in 50yrs)
uplift [rad] 0.1

-0.1
0 2 4 6 8 10
plane rot. [rad]

0.05

-0.05
0 2 4 6 8 10
5
sliding [cm]

0
-5
-10
-15
0 2 4 6 8 10
table disp. [cm]

15
10
5
0
-5
-10
-15
0 2 4 6 8 10
1
table acc. [g]

-1
0 2 4 6 8 10
time [sec]

Figure 6. Response of the Kelvinator refrigerator subjected to the UC Science Building 6th-floor
motion of the Gavilan College (10% in 50 years) record of the 1989 Loma Prieta, California,
earthquake. The heavy gray lines on the bottom two graphs plot a Type-C1 trigonometric pulse that
approximates the main pulse of the motion record.

5. REGRESSION ANALYSIS AND FRAGILITY CURVES

5.1. Governing parameters during sliding

Many parameters influence the full behavior of a piece of equipment subjected to seismic motion.
However, since we are primarily concerned with sliding, the governing parameters become those
that describe (a) the mechanical characteristics of the equipment–floor interface and (b) the kine-
matic characteristics of the base motion. From the slow-pull tests performed on the equipment, the
load–displacement curves show that there is a pre-yielding elasticity due to the flexibility of the
legs. A peak value of force is reached (associated with the static coefficient of friction s ) after
which the equipment starts sliding with a relatively constant force (associated with the kinetic
coefficient of friction k ). Among these three parameters, the one parameter that best describes
the sliding resistance of the interface is the kinetic coefficient of friction k , designated hereafter
simply as .
Figures 5–7 that plot the results of the shake table tests on the freestanding equipment also plot
the results obtained by numerical simulations for two different models of the sliding interface:
an elastoplastic model (MATLAB) and a rigid-plastic (Coulomb friction) model (WM2D). It was
noted that when the values  obtained from the slow-pull tests (see Figure 3 and Table III)

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
838 D. KONSTANTINIDIS AND N. MAKRIS

equipment: ASP refrigerator (profile) motion: Coyote Lake, Gilroy Array #6 FN 6th Floor (50% in 50yrs)
uplift [rad] 0.1

-0.1
0 2 4 6 8 10
plane rot. [rad]

0.05

-0.05
0 2 4 6 8 10
5
sliding [cm]

0
-5
-10
-15
-20
-25
0 2 4 6 8 10
table disp. [cm]

15
10
5
0
-5
-10
-15
0 2 4 6 8 10
1
table acc. [g]

-1
0 2 4 6 8 10
time [sec]

Figure 7. Response of the ASP refrigerator ( profile configuration) subjected to the UC Science Building
6th-floor motion of the Gilroy Array #6 (50% in 50 years) record of the 1979 Coyote Lake, California,
earthquake. The heavy gray lines on the bottom two graphs plot a Type-C1 trigonometric pulse that
approximates the main pulse of the motion record.

were used, the numerical predictions of the elastoplastic model (plotted with dashed gray lines in
Figures 5–7) are, in general, closer to the experimental results (heavy black lines) than the numerical
prediction offered by the rigid-plastic (Coulomb friction) model (solid gray lines). Nevertheless,
even the elastoplastic model predictions are only in fair agreement with the experimental results.
A considerably improved agreement between experimental and numerical results from either model
is observed when the numerical value of  is reduced in the numerical simulations. The reader will
note that the reduction proposed for the rigid-plastic (Coulomb) model is consistently larger in all
three equipment than the reduction proposed for the elastoplastic model. Part of the reason for the
reduction of the friction coefficient in the elastoplastic model is that the friction coefficient between
the steel equipment legs and the vinyl floors may exhibit some pressure dependence that was not
captured during the quasi-static pull tests. During the quasi-static tests, the equipment specimen
was pulled from a height of approximately 15 cm from the sliding surface, which means that the
pressure on all four legs was about equal. However, during the shake table tests, the overturning
moment on the equipment results in an uneven distribution of pressure on the equipment legs,
which in turn may have resulted in lower overall frictional resistance. This would be similar to the
pressure-dependent behavior of a teflon–steel interface observed by other investigators [16, 17].
The further reduction that is needed in the rigid-plastic (Coulomb) model is due to its zero yield
displacement. Recent studies by the senior author [20, 21] have shown that for the same value of the

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 LABORATORY EQUIPMENT TO EARTHQUAKE SHAKING 839

friction coefficient (strength) the sliding displacement increases with increasing yield displacement.
Consequently, given that in all three pieces of equipment the behavior at the sliding interface is
elastoplastic (finite yield displacement), when a rigid-plastic (Coulomb) model is adopted to capture
the behavior, a smaller value of the coefficient of friction is needed. Interestingly, the amount of
reduction in the value of the coefficient of friction is proportional to the yield displacement—a
conclusion that is consistent with the aforementioned studies by the senior author.

5.2. Kinematic characteristics of pulse-type motions

Physically realizable pulses can adequately describe the impulsive character of near-fault ground
motions both qualitatively and quantitatively. The input parameters of the model have an unam-
biguous physical meaning. The minimum number of parameters is two, which are either the
acceleration amplitude, ap , and duration, Tp , or the velocity amplitude, vp , and duration, Tp [22–25].
The current established methodologies for estimating the pulse characteristics of a wide class
of records are of unique value since the product ap Tp2 is a characteristic length scale of the ground
excitation and is a measure of the persistence of the most energetic pulse to generate inelastic
deformations in structures [26]. It is emphasized that the persistence of the pulse is a different
characteristic than the strength of the pulse which is measured with the peak pulse acceleration. The
reader should recall that among two pulses with different acceleration amplitudes (say ap1 >ap2 )
and different pulse durations (say Tp1 <Tp2 ), the inelastic deformation does not scale with the
peak pulse acceleration (most intense pulse) but with the stronger length scale (larger ap Tp2 = most
persistent pulse).
The third column of Table IV lists the type and defining parameters (Tp and vp ) of trigonometric
pulses that approximate the predominant pulses of each of the earthquake motions used in this
study. A Type-A pulse, which results in a forward displacement of the ground, is defined by
ground displacement u A g (t) = 0.5vp [t −sin(p t)/p ], 0tTp [24, 27, 28]. A Type-B pulse, which
results in a forward-and-back ground displacement, is defined by u B g (t) = vp [1−cos(p t)]/p ,
0tTp [24]. The velocity and acceleration histories are obtained by differentiating the expressions
for ground displacement.
Not all near-source motions are forward or forward-and-back pulses. Figure 6 plots the 6th-floor
motion resulting from nonlinear dynamic analysis of the UC Science Building subjected to the
fault-normal component of the Gavilan College (10% in 50 years) record of the 1989 Loma Prieta,
California, earthquake [9, 12]. The bottom two windows of Figure 6 also plot the displacement
and acceleration histories of a Type-C1 pulse that approximates the one-cycle predominant pulse
of the floor motion. Near-fault ground motions for which the displacement history exhibits one or
more long-duration cycles are approximated by Type-Cn pulses with ground displacement u C g (t) =
n

vp [cos −cos(p t +)−p t sin ]/p , 0t(n +0.5−/)Tp , where  is equal to 0.0697 for
n = 1 and 0.0410 for n = 2 [2, 22, 23].

5.3. Intensity Measure and EDP

The attempt to estimate the sliding response of the heavy laboratory equipment is cast in a
probabilistic framework. The PEER PBEE methodology [8] suggests the identification of an IM.
After several considerations, a simple, physically motivated, dimensionless IM is
PTA
IM = −1 (3)
g

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 840

Table IV. Kinematic characteristic of the 50 and 10% in 50 years hazard level earthquake motions used in this study.

Copyright q
Earthquake motion Pulse, Tp (s), PTA Umax   IM EDP
Equipment (Probability of exceedance in 50 years) vp (cm/s) (g) (cm) (slow pull) (best fit) PTA/g −1 Umax 2p /PTA

FORMA Aigion, OTE B, 0.53, 61 0.75 17.4 0.23 0.13 2.26 3.34
Coyote Lake, Gilroy #6 C1, 0.70, −50 0.68 6.5 0.23 0.13 1.96 0.79
Coyote Lake, Gilroy #6 ground (50%) C1, 0.70, 50 0.75 7.8 0.23 0.13 2.26 0.85
Coyote Lake, Gilroy #6 6th floor (50%) C1, 0.64, 75 0.75 14.7 0.23 0.13 2.26 1.93
Loma Prieta, Gavilan College ground (10%) A, 0.40, −70 0.73 9.6 0.23 0.13 2.17 3.32
Kelvinator Aigion OTE B, 0.53, 61 0.76 15.2 0.28 0.17 1.71 2.87
Coyote Lake, Gilroy #6 C1, 0.70, −50 0.69 5.4 0.28 0.17 1.46 0.64
Coyote Lake, Gilroy #6 ground (50%) C1, 0.70, 50 0.74 8.3 0.28 0.17 1.64 0.92

2008 John Wiley & Sons, Ltd.

Coyote Lake, Gilroy #6 6th floor (50%) C1, 0.64, 75 0.76 18.7 0.28 0.17 1.71 2.42
Loma Prieta, Gavilan College ground (10%) A, 0.40, −70 0.76 5.7 0.28 0.17 1.71 1.90
Loma Prieta, Gavilan College 6th floor (10%) C1, 0.80, 60 0.64 12.9 0.28 0.17 1.29 1.26
Tottori, Kofu 6th floor (10%) C2, 0.45, 55 0.81 13.3 0.28 0.17 1.89 3.27
ASP Aigion OTE B, 0.53, 61 0.86 8.0 0.31 0.20 1.77 1.34
(Face) Coyote Lake, Gilroy #6 C1, 0.70, −50 0.73 4.1 0.31 0.20 1.35 0.46
Coyote Lake, Gilroy #6 ground (50%) C1, 0.70, 50 0.76 9.2 0.31 0.20 1.45 0.99
Coyote Lake, Gilroy #6 6th floor (50%) C1, 0.64, 75 0.77 23.7 0.31 0.20 1.48 3.03
Loma Prieta, Gavilan College ground (10%) A, 0.40, −70 0.77 4.7 0.31 0.20 1.48 1.54
Loma Prieta, Gavilan College 6th floor (10%) C1, 0.80, 60 0.67 17.9 0.31 0.20 1.16 1.68
Tottori, Kofu ground (10%) C2, 0.48, 47 1.07 35.3 0.31 0.20 2.45 5.76
Tottori, Kofu 6th floor (10%) C2, 0.45, 55 0.87 16.8 0.31 0.20 1.80 3.86
D. KONSTANTINIDIS AND N. MAKRIS

ASP Aigion, OTE B, 0.53, 61 0.73 6.1 0.31 0.20 1.35 1.20
(Profile) Coyote Lake, Gilroy #6 C1, 0.70, −50 0.70 2.6 0.31 0.20 1.26 0.30
Coyote Lake, Gilroy #6 ground (50%) C1, 0.70, 50 0.75 9.3 0.31 0.20 1.42 1.02
Coyote Lake, Gilroy #6 6th floor (50%) C1, 0.64, 75 0.80 20.3 0.31 0.20 1.58 2.50
Loma Prieta, Gavilan College ground (10%) A, 0.40, −70 0.79 6.6 0.31 0.20 1.55 2.10
Loma Prieta, Gavilan College 6th floor (10%) C1, 0.80, 60 0.67 17.0 0.31 0.20 1.16 1.60
Tottori, Kofu ground (10%) C2, 0.48, 47 1.07 42.2 0.31 0.20 2.45 6.90
Tottori, Kofu 6th floor (10%) C2, 0.45, 55 0.85 20.2 0.31 0.20 1.74 4.73
The motion PTA and p = 2/Tp together with the maximum recorded equipment sliding displacement, Umax , and interface friction coefficient, ,
produce the intensity measure, IM, and engineering demand parameter, EDP.

DOI: 10.1002/eqe
Earthquake Engng Struct. Dyn. 2009; 38:827–848
 LABORATORY EQUIPMENT TO EARTHQUAKE SHAKING 841

where PTA is the peak table acceleration,  is the interface kinetic friction coefficient and g is
the acceleration of gravity. Recall that the condition for sliding to occur is PTA/s g>1, and since
generally s >, it is possible for sliding to occur even for negative values of the chosen IM.
Nevertheless, the numerical simulation studies that were done in parallel with the experimental
studies showed that the coefficient of static friction s had little influence on the maximum sliding
displacement that the equipment exhibited. On the contrary, the maximum sliding displacement
was considerably sensitive to the kinetic coefficient of friction . For this reason, the expression
for IM given by (3) features  and not s .
The chosen EDP will be henceforth designated as . It is given by
Umax 2p
= (4)
PTA
where Umax is the maximum sliding displacement recorded, p = 2/Tp is the circular frequency
of the pulse that approximates the predominant pulse of the earthquake excitation and PTA is
the peak table acceleration of the earthquake excitation. The choice for the EDP emerges from
dimensional analysis in conjunction with previously published results on the response of a sliding
block [10, 29–31].
The quantity of interest Umax can be expressed as a function of independent variables ap , p ,
and g
Umax = f (ap , p , g) (5)
The dependent variable Umax and independent variables ap , p and g, involve only two dimen-
sions, those of length, L, and time, T . The quantities of interest have dimensions
[Umax ] = L , [ap ] = LT −2 , [p ] = T −1 , [g] = LT −2 (6)
Buckingham’s Pi Theorem states that a dimensionally homogeneous equation with a total of k
variables and r reference dimensions can be reduced to a relationship among k −r independent
dimensionless -products [30]. Accordingly, in this case there are 4−2 = 2 dimensionless products.
Two obvious choices for -products are
Umax 2p g
1 = and 2 = (7)
ap ap
and the two are related by a function
1 = (2 ) (8)
For the rectangular acceleration pulse, this function is [29]
 
1
1 = 2 2
−1 (9)
2
Note from Equations (3) and (9) that the IM chosen is exactly the quantity in parenthesis in (9)
except that the IM uses for simplicity PTA instead of ap . The two values are expected to be close.
For the case of trigonometric pulses such as Type-A, Type-B, and Type-Cn , the response of the
rigid-plastic system is again described by Equation (8) [10, 26], and the form of the function  is
obtained numerically.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
842 D. KONSTANTINIDIS AND N. MAKRIS

5.4. The EDP as a lognormal random variable; regression of experimental data

Figure 8 (bottom left) plots the EDP,  =Umax 2p /PTA, as a function of the IM = PTA/g −1, for
the shake table tests performed on the three pieces of equipment. The value of the coefficient of
friction, , is that obtained by the slow-pull tests. It is obvious that the data exhibit considerable
scatter, which suggests that  has to be treated as a random variable. When a random variable,
, expresses a quantity that is only positive (>0), it is common to assume that the variable is
lognormally distributed. In this study we hypothesize that the EDP, , is lognormally distributed,
and we test this hypothesis against the experimental results. The reason behind the name lognormal
is that the lognormally distributed variable  is related to a normally distributed variable X by
X = ln . Note that  attains only positive values, >0, while the corresponding X variable is
unrestricted, −∞<x<∞. It is assumed that  takes the form
 

PTA
= −1 Z = IM
Z (10)
g

where and
are regression parameters and Z is a lognormally distributed random variable.
Another important feature of the scatter in Figure 8 (bottom left) is that it increases with increasing

 from slow pull tests  from best fit with shake table test results
X = ln(Δ ) = ln(Umax  2p /PTA)

3 3
X = ln(Δ ) = ln(Umax  2p /PTA)

2 2

1 1

0 0

-1 –1

-2 –2
0 0.2 0.4 0.6 0.8 1 0 0.4 0.8 1.2 1.6 2
ln(PTA/µg−1) ln(PTA/µg−1)

8 8
EDP = Δ = Umax  2p /PTA

EDP = Δ = Umax  2p /PTA

6 6

4 4

2 2

0 0
0 0.5 1 1.5 2 2.5 0 1 2 3 4 5
IM = PTA /g −1 IM = PTA /g −1

Figure 8. Top: the normal random variable X = ln  and the linear fit that gives its mean m X . Bottom: the
associated lognormal random variable , together with the power-law fit for its mean m  (Equations (12)
and (13)). The distributions plotted indicate that  X is constant, while  increases with IM. The value
of  is that obtained by the slow-pull tests (left) and that obtained by best-fitting numerical simulation
results to shake table test results (right).

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 LABORATORY EQUIPMENT TO EARTHQUAKE SHAKING 843

IM (funneling effect), i.e. the variance is non-constant. However, the logarithmic transformation
X = ln  = ln +
ln IM+ln Z (11)
stabilizes the variance, as seen on the top left plot of Figure 8. ln Z and X are normally distributed,
and therefore ordinary least squares can be applied to compute the parameter and
. We obtain
 1.88
PTA
 = 0.663 −1 Z = 0.663 IM1.88 Z (12)
g
Figure 8 (right) also plots the EDP as a function of the IM for the shake table tests performed on
the three pieces of equipment; however, the values of  are not those obtained from the slow-pull
tests but rather those obtained from best-fitting numerical simulation time history results to results
recorded during the shake table experiments. For these ,
 1.50
PTA
 = 0.289 −1 Z = 0.289 IM1.50 Z (13)
g
The lognormally distributed variable  has the probability density function
   
1 1 ln −m X 2
f  () = √ exp − , >0 (14)
2 X  2 X
where the two defining parameters m X and  X of the distribution are in fact the mean and standard
deviation of the corresponding normally distributed variable X [32]. The cumulative distribution
function of  is given by
    2 
1 1 ln  −m X
F () = √ exp − d (15)
0 2 X  2 X

Using the substitution u = (ln  −m X )/ X , we obtain

 (ln −m X )/ X  2  
1 u ln −m X
F () = √ exp − du =  (16)
−∞ 2 2 X
where  is the cumulative distribution function of a standard normal variable (i.e. mean zero and
standard deviation of 1).
Note from Equation (11) that the mean of X is m X (IM) = ln( IM
). Also, the standard deviation
of X can be estimated from

  2
1  n i
 X = ln Z =
ln (17)

n −1 i=1 IMi
and it is  X = 0.631 for  obtained from the slow-pull tests and  X = 0.678 for  obtained from
the best-fit to experimental results. The mean, m  (IM), and standard deviation,  (IM), of  can
be shown to be related to the mean, m X (IM), and standard deviation,  X , of the normal variable
X through
 
2X
2X
m  (IM) = exp m X (IM)+ = IM exp (18)
2 2

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
844 D. KONSTANTINIDIS AND N. MAKRIS

and
   
2 2X
 (IM) = exp m X (IM)+ X exp(2X )−1 = IM
exp exp(2X )−1 (19)
2 2

Figure 8 (top) shows how X is distributed with a constant standard deviation  X at four different
IM values, while Figure 8 (bottom) shows how  is distributed with an increasing  (IM).
In order to not reject the hypothesis that  has a lognormal distribution, the Kolmogorov–
Smirnov goodness-of-fit test was performed [33]. Figure 9 plots the empirical cumulative distribu-
tion function, F∗ = i/n, and the cumulative distribution function of the hypothesized distribution,
F = ((ln −m X )/ X ), where  is obtained by slow-pull tests (top) and by best-fitting numer-
ical simulation results to shake table experimental results (bottom). The lognormal distribution

Figure 9. The empirical cumulative distribution function, F∗ (50 and 10% hazard level motions), and the
theoretical cumulative distribution function F of its hypothesized lognormal distribution when the value
of the friction coefficient  is obtained from: the slow-pull tests (top); best-fitting numerical simulation
results to shake table test results (bottom).

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 LABORATORY EQUIPMENT TO EARTHQUAKE SHAKING 845

hypothesis is shown to be acceptable even for large significance levels (0.2 or more). A detailed
description of the goodness-of-fit test is offered in the report by Konstantinidis and Makris [10].

5.5. Fragility curves

Fragility is broadly defined as a conditional probability of failure. For the problem at hand, where
we want to characterize the seismic response of heavy laboratory equipment in a probabilistic
framework, we define fragility as the probability Pf that the EDP,  =Umax 2p /PTA, for a piece
of equipment will exceed a certain threshold (capacity), c, given the IM. For the lognormally
distributed random variable ,
Pf ≡ P(>c|IM) = 1− P(<c|IM) = 1− F (c) (20)
Substituting (16) and m X (IM) = ln( IM
)
into (20) gives
   
ln c −m X (IM) 1 c
Pf = 1− = 1− ln (21)
X X IM

where  is the standard normal cumulative distribution function, and


0.631, 0.663, 1.88 for slow-pull 
{ X , ,
} = (22)
0.678, 0.289, 1.50 for best-fit 
Figure 10 shows fragility curves for four values of capacity c (0.5, 1.0, 2.0, and 3.0). Note
that each graph includes two curves; one is the fragility curve that corresponds to values of 
obtained from the slow-pull tests, while the other corresponds to values of  obtained by tailoring
numerical simulations to approximate shake table experiment results. The shaded area between the
two curves provides an envelope for the fragility. Typically, fragility curves are plotted against the
EDP. In this study, however, they are plotted against the IM, as this makes their use more direct.
Once the  X is estimated from Equation (17) and the regression parameters and
are obtained
from ordinary least-squares fit on Equation (11), the fragility curves can be generated using
Equation (22) for different capacities c. Given (or generating) a family of such fragility curves,
a practicing engineer who wants to evaluate the probability that a piece of equipment will not
exceed a threshold displacement could proceed as follows.

5.6. Example: how to use a fragility curve

Suppose that an engineer is concerned with a rigid piece of laboratory equipment that is located on
the 6th floor of a hospital near an emergency exit. The design absolute floor acceleration is 1.4g,
and the expected period of vibration is about Tp = 0.8 s (say, the first modal period of the building).
The engineer wants to estimate the probability that the equipment with  = 0.5 will exceed a
displacement of 45 cm, which will result in blocking the emergency exit. After the engineer has
confirmed that sliding is the expected mode of response (e.g. by using s and  and Shenton’s
[19] criteria for the initiation of the various response modes), can proceed as follows. By Equation
(3), IM = PTA/g −1 = 1.4 g/0.5 g −1 = 1.8 and the capacity value c that corresponds to the above
design parameters is c =Umax 2p /PTA = 45 cm(2/0.8 s)2 /(1.4×981 cm/s2 ) ≈ 2. Referring to the
fragility curves generated for c = 2 (Figure 10, top right), we read off a value of Pf between the
two curves. For the computed IM of 1.8, a conservative-average value of P(Umax 2p /PTA>2) =
P(Umax >45 cm) ≈ 0.3 is read with engineering judgment.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
846 D. KONSTANTINIDIS AND N. MAKRIS

Figure 10. Fragility curves for c = 0.5 (top left), c = 1.0 (bottom left), c = 2.0 (top right) and c = 3.0
(bottom right). Two curves are plotted in each graph. The top curve corresponds to values of  obtained
by slow-pull tests, while the bottom curve corresponds to values of  obtained by best-fitting numerical
simulation results to shake table experiment results.

6. CONCLUSIONS

In this study, a comprehensive experimental program investigating the seismic response of free-
standing laboratory equipment located within several floor levels of a research facility was under-
taken. The study followed the PBEE approach.
Results from shake table tests experiments on the freestanding equipment subjected to ground
and floor motions of 50 and 10% in 50 years hazard levels were presented. For the equipment tested,
although there was some rocking observed (particularly for the FORMA incubator), sliding was the
predominant mode of response, with sliding displacements reaching up to 60 cm. The experimental
results were then used to calibrate and validate numerical simulation models. Numerical simulation
studies with MATLAB [34] on the sliding response of the equipment using the elastoplastic model
with the values of the friction coefficient  extracted from the slow-pull tests yielded results that
were in fair agreement with the experimental results. The predicted response of the equipment
was appreciably improved when reduced values of their friction coefficients were used. Numerical
simulation studies with Working Model [14] using the Coulomb friction model also provided fair
results that were considerably improved when the values of s and k were reduced.
The IM, IM = PTA/g −1, and (random) EDP,  =Umax 2p /PTA, were identified. The results of
the shake table tests were then used to test the hypothesized lognormal distribution of the random
variable  and to arrive at simple relationships for the mean m  and standard deviation  in

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 LABORATORY EQUIPMENT TO EARTHQUAKE SHAKING 847

terms of the IM. Fragility curves, that give the probability that the EDP will exceed a specified
threshold c as a function of the IM were generated, and an example was presented to illustrate
how the fragility curves can be used.

ACKNOWLEDGEMENTS
From the UC Science Laboratory Building, we thank Barbara Duncan and Todd Laverty for providing the
equipment that was used for the shake table experiments; Eddy Kwan for providing us with invaluable
information regarding the laboratory equipment and its in situ boundary conditions in the building
laboratories. From the Earthquake Simulator Laboratory at the Earthquake Engineering Research Center,
UC Berkeley, where the shake table experiments were conducted, we greatly appreciate the technical
assistance of Don Clyde, Wes Neighbour and David Maclam. We also thank our PEER colleagues
for their collaboration. We are very grateful for their continual feedback throughout the course of this
study. This work was supported primarily by the Earthquake Engineering Research Centers Program
of the National Science Foundation under award number EEC-9701568 through the Pacific Earthquake
Engineering Research Center (PEER).

REFERENCES
1. Yim CS, Chopra AK, Penzien J. Rocking response of rigid blocks to earthquakes. Earthquake Engineering and
Structural Dynamics 1980; 8(6):565–587.
2. Makris N, Roussos Y. Rocking response of rigid blocks under near-source ground motions. Géotechnique 2000;
50(3):243–262.
3. Makris N, Konstantinidis D. The rocking spectrum and the limitations of practical design methodologies.
Earthquake Engineering and Structural Dynamics 2003; 32(2):265–289.
4. Shao Y, Tung CC. Seismic response of unanchored bodies. Earthquake Spectra 1999; 15(3):523–536.
5. Lopez Garcia D, Soong TT. Sliding fragility of block-type non-structural components. Part 1: unrestrained
components. Earthquake Engineering and Structural Dynamics 2003; 32(1):111–129.
6. Lopez Garcia D, Soong TT. Sliding fragility of block-type non-structural components. Part 2: restrained
components. Earthquake Engineering and Structural Dynamics 2003; 32(1):131–149.
7. Hutchinson TC, Chaudhuri SR. Bench-shelf system dynamic characteristics and their effects on equipment and
contents. Earthquake Engineering and Structural Dynamics 2006; 35(13):1631–1651.
8. Porter KA. An overview of PEER’s performance-based earthquake engineering methodology. Proceedings of
the Ninth International Conference on Applications of Statistics and Probability in Civil Engineering (ICASP9),
San Francisco, CA, 2003.
9. Comerio MC (ed.). PEER testbed study on a laboratory building: exercising seismic performance assessment.
Report No. PEER 2005-12, Pacific Earthquake Engineering Research Center, University of California, Berkeley,
CA, 2005.
10. Konstantinidis D, Makris N. Experimental analytical studies on the seismic response of freestanding and anchored
laboratory equipment. Report No. PEER 2005-07, Pacific Earthquake Engineering Research Center, University
of California, Berkeley, CA, 2005.
11. Somerville PG. Ground Motion Time Histories for the UC Lab Building. URS Corporation: Pasadena, CA, 2001.
12. Lee TH, Mosalam KM. Seismic demand sensitivity of reinforced concrete shear-wall building using FOSM
method. Earthquake Engineering and Structural Dynamics 2005; 34(14):1719–1736.
13. Housner GW. The behavior of inverted pendulum structures during earthquakes. Bulletin of the Seismological
Society of America 1963; 53(2):404–417.
14. Working Model. User’s Manual. MSC. Software Corporation: San Mateo, CA, 2000.
15. Gitis NV, Volpe L. Nature of static friction time dependence. Journal of Physics D: Applied Physics 1992;
25(4):605–612.
16. Mokha A, Constantinou MC, Reinhorn AM. Teflon bearings in aseismic base isolation: experimental studies
and mathematical modeling. Technical Report NCEER-88-0038, National Center for Earthquake Engineering
Research, State University of New York at Buffalo, NY, 1988.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe
 848 D. KONSTANTINIDIS AND N. MAKRIS

17. Constantinou MC, Tsopelas P, Kim YS, Okamoto S. NCEER-Taisei Corporation research program on sliding
seismic isolation systems for bridges: experimental and analytical study of a friction pendulum system (FPS).
Technical Report NCEER-93-0020, National Center for Earthquake Engineering Research, State University of
New York at Buffalo, NY, 1993.
18. Konstantinidis D, Kelly JM, Makris N. Experimental investigation on the seismic response of bridge bearings.
Report No. EERC 2008-02, Earthquake Engineering Research Center, University of California, Berkeley, CA,
2008.
19. Shenton III HW. Criteria for initiation of slide, rock, and slide-rock rigid-body modes. Journal of Engineering
Mechanics (ASCE) 1996; 122(7):690–693.
20. Makris N, Black CJ. Dimensional analysis of rigid-plastic and elastoplastic structures under pulse-type excitations.
Journal of Engineering Mechanics (ASCE) 2004; 130(9):1006–1018.
21. Makris N, Psychogios T. Dimensional response analysis of yielding structures with first-mode dominated response.
Earthquake Engineering and Structural Dynamics 2006; 35(10):1203–1224.
22. Makris N, Chang SP. Effects of viscous, viscoplastic and friction damping on the response of seismic isolation
structures. Earthquake Engineering and Structural Dynamics 2000; 29(1):85–107.
23. Makris N, Chang SP. Response of damped oscillators to cycloidal pulses. Journal of Engineering Mechanics
(ASCE) 2000; 126(2):123–131.
24. Makris N. Rigidity-plasticity-viscosity: can electrorheological dampers protect base-isolated structures from near-
source ground motions? Earthquake Engineering and Structural Dynamics 1997; 26(5):571–591.
25. Mavroeidis GP, Papageorgiou AS. A mathematical representation of near-fault ground motions. Bulletin of the
Seismological Society of America 2003; 93(3):1099–1131.
26. Makris N, Black CJ. Dimensional analysis of inelastic structures subjected to near-fault ground motions. Report
No. EERC 2003-05, Earthquake Engineering Research Center, University of California, Berkeley, CA, 2003.
27. Veletsos AS, Newmark NM, Chelepati CV. Deformation spectra for elastic and elastoplastic systems subjected
to ground shock and earthquake motions. Proceedings of the 3rd World Conference on Earthquake Engineering,
Wellington, New Zealand, 1965.
28. Jacobsen LS, Ayre RS. Engineering Vibrations, with Applications to Structures and Machinery. McGraw-Hill:
New York, NY, 1958.
29. Newmark NM. Effects of earthquakes on dams and embankments. Fifth Rankine Lecture. Géotechnique 1965;
15:139–160.
30. Barenblatt GI. Scaling, Self-similarity, and Intermediate Asymptotics. Cambridge University Press: Cambridge,
U.K., 1996.
31. Konstantinidis D, Makris N. Seismic response analysis of multidrum classical columns. Earthquake Engineering
and Structural Dynamics 2005; 34(10):1243–1270.
32. Crow EL, Shimizu K (eds). Lognormal Distributions. Marcel Dekker: New York, NY, 1988.
33. Scheaffer RL, McClave JT. Probability and Statistics for Engineers (4 edn). Duxbury Press: Belmont, CA, 1995.
34. MATLAB. High-performance Language Software for Technical Computing. The MathWorks, Inc.: Natick, MA,
2002.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2009; 38:827–848
DOI: 10.1002/eqe

View publication stats