Soil Dynamics and Earthquake Engineering 48 (2013) 162–174

Contents lists available at SciVerse ScienceDirect

Soil Dynamics and Earthquake Engineering

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

Seismic soil–foundation–structure interaction observed in geotechnical

centrifuge experiments
H.B. Mason a,n, N.W. Trombetta b, Z. Chen c, J.D. Bray d, T.C. Hutchinson b, B.L. Kutter e
a
School of Civil & Construction Engineering, Oregon State University, 220 Owen Hall, Corvallis, OR 97331, USA
b
Department of Structural Engineering, University of California, San Diego, La Jolla, CA 92903, USA
c
Department of Civil & Mechanical Engineering, University of Missouri, Kansas City, Kansas City, MO 64110, USA
d
Department of Civil & Environmental Engineering, University of California, Berkeley, CA 94720, USA
e
Department of Civil & Environmental Engineering, University of California, Davis, CA 95616, USA

a r t i c l e i n f o a b s t r a c t

Article history: This paper reports the experimental results from two geotechnical centrifuge tests examining seismic
Received 9 March 2012 soil–foundation–structure interaction of frame structures. In the first test, two three-dimensional frame
Received in revised form structures were placed apart, and in the second test, the same frame structures were located adjacent
14 August 2012
to each other. One of the frame structures was founded on shallowly embedded spread footings, and
Accepted 24 January 2013
Available online 5 March 2013
the other was founded on a deeper basement. During the second test, seismic structure–soil–structure
interaction was also examined. In this paper, the experimental set-up is described in detail, some
kinematic interaction observations are made, and seismic footing response results are reported. The
results of the tests indicate that the seismic response of the shallow footings of a flexible frame
structure is complex. This result is important because these complexities are not well incorporated in
current soil–foundation–structure interaction analysis procedures. The adjacency of two frame
structures, and thus structure–soil–structure interaction, further complicates analysis procedures.
In the second test, the spread footings rock, slide, and settle less when they are directly adjacent to a
deep basement. These results imply that SSSI can be beneficial or detrimental, depending on the
earthquake motion and the structural system. In a companion paper, the seismic response of the frame
superstructures and the footing force demands are also examined.
& 2013 Elsevier Ltd. All rights reserved.

1. Introduction (DCI). A related phenomenon, usually referred to as ‘‘site-city

effects’’, examines SSSI effects on a city scale [9]. Relative to SFSI,
During earthquakes, civil structures interact with the sur- SSSI has received relatively little attention in the literature,
rounding soil at their foundations, which is referred to as soil– although there are a few seminal theoretical studies regarding
foundation–structure interaction (SFSI). SFSI causes the seismic the subject published by Luco and Contesse [10], Lee and Wesley
response of a soil–foundation–structure (SFS) system to differ [11], and Wong and Trifunac [12]. A full literature review on SFSI
from the usually assumed fixed-base condition. Although seismic and SSSI principles is beyond the scope of this paper. Interested
SFSI has received much attention in the earthquake engineering readers are referred to Kausel [13] for a comprehensive discussion
literature [1–8], the majority of SFSI studies consider SFS systems of SFSI research and to Menglin et al. [14] for an up-to-date
in isolation (i.e., a single structure with no neighboring struc- survey of SSSI literature.
tures). However, in urban areas this assumption is not valid. This paper focuses on the results of two centrifuge tests
Instead, SFS systems are located in close proximity to each other designed to examine both SFSI and SSSI effects. The tests were
and the interactions between adjacent systems become impor- performed at the University of California at Davis Center for
tant. These interactions are referred to as structure–soil–structure Geotechnical Modeling (NEES@UC Davis) at a centrifugal accel-
interaction (SSSI) or sometimes as dynamic cross-interaction eration of 55 g. Two steel moment-resisting frame structures,
founded on dry, dense sand were subjected to a variety of
earthquake motions during these tests. The first structural model,
n
Corresponding author. Tel.: þ1 541 737 2014; fax: þ1 541 737 3052. referred to herein as MS1F_SF80, is a single-story frame founded
E-mail addresses: [email protected] (H.B. Mason),
[email protected] (N.W. Trombetta), [email protected] (Z. Chen),
on shallowly embedded spread footings. The second structural
[email protected] (J.D. Bray), [email protected] (T.C. Hutchinson), model, referred to herein as MS3F_B, is a three-story frame
[email protected] (B.L. Kutter). founded on a large, deep basement. Fig. 1 shows pictures of the

0267-7261/$ - see front matter & 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.soildyn.2013.01.014
 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174 163

two structural models. Both models feature reduced cross- MS1F_SF80 spread footings are examined. All results reported in
sections near the ends of their beams and at the base of their these sections are reported in the prototype scale, unless other-
columns to ensure the localization of inelastic structural wise noted. Finally, SFSI and SSSI observations are made and
response. MS1F_SF80 and MS3F_B were designed to capture the conclusions are drawn. Readers are directed to a sequential
predominant dynamic characteristics of a 3-story and a 9-story companion paper for discussion of the structural model designs,
prototype structure, respectively. In Test-1 [15], the structural seismic foundation forces, and the seismic performance of the
models were placed as far apart as possible within the constraints superstructures during both tests.
of the centrifuge container, and the baseline nonlinear SFSI
response of each structure was recorded. During Test-2 [16], the
structural models were placed directly adjacent to each other to 2. The need for experimental SFSI and SSSI research
investigate SSSI effects. Fig. 2 shows pictures of the Test-1 and
Test-2 set-up. In these pictures, the models are fully constructed There are three broad categories of SFSI and SSSI research:
and on the centrifuge arm. (1) theoretical; (2) numerical; and (3) experimental. Within this
Within this paper, the experimental set-ups for both Test-1 section, research in these three categories is described, and the
and Test-2 are discussed in detail. This includes a discussion of pressing need for experimental laboratory work is emphasized.
centrifuge scaling laws and centrifuge boundary effects. The The goal of theoretical SFSI research is to develop closed-form
earthquake motion plan used for both tests is also described. A solutions for a structural response quantity (e.g., base shear, over-
variety of sequential strong earthquake motions, including both turning moment, roof displacement) for a SFS system subjected to a
forward-directivity and ordinary motions, were applied to the base input motion. The theoretical approach is advantageous,
model during each test (i.e., 17 motions during Test-1 and 21 because the provided solutions are grounded in computational
motions during Test-2). The selected earthquake motions are mechanics, giving insight into the physical mechanisms underlying
applicable to seismically-active areas characterized by shallow, SFSI effects. Developing closed-form solutions usually involves
crustal earthquake motions, such as in the Los Angeles and San solving governing partial differential equations. Accordingly, most
Francisco areas. Inelastic and transient displacements of the closed-form solutions involve several simplifying assumptions
regarding soil response, structural system response and boundary
conditions, which is a key disadvantage of the approach. As a result,
theoretical SFSI solutions have difficulty capturing nonlinear soil,
nonlinear foundation (i.e. foundation uplift), or inelastic structural
(i.e., yielding of structural components) response. Therefore, the
provided solutions are rarely compared to either experimental data
or well-documented case histories from strong shaking events.
Seminal SSSI theoretical studies include Luco and Contesse [10],
Lee and Wesley [11], Wong and Trifunac [12], and Kobori and
Kusakabe [17,18].
The goal of numerical SFSI work is to use a numerical procedure,
such as finite or boundary element method, to examine elastic or
nonlinear/inelastic SFSI effects. Usually, the numerical work is com-
pared to theoretical SFSI work, experimental SFSI work, or both. A key
advantage of the numerical approach is that once a robust numerical
model is developed, extensive parametric analyses can be performed.
Fig. 1. Models used in Test-1 and Test-2: (left) MS1F_SF80 and (right) MS3F_B. These parametric analyses can lead to insights into SFSI effects and

Fig. 2. The overall set-up of Test-1 (left) and Test-2 (right). Both models are fully instrumented and on the centrifuge arm.
164 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174

can test the validity of the theoretical results. It is often difficult to Table 1
construct a numerical model that fully accounts for the complexities Scaling factors for converting N  g scaled-model measurements to
prototype values.
of the SFS system under examination and to do so requires the direct
approach [19]. To alleviate the difficulties with modeling an entire Quantity Prototype/model
SFS system, some researchers use the substructure approach [19],
wherein the SFS system is divided into its subcomponents and the Length N
dynamic response of each subcomponent is examined separately. Area N2
Volume N3
A key disadvantage of the substructure approach is that linearity Mass N3
must be assumed for the global system response to equal the sum Density 1
of the subcomponent responses. In reality, the soil, the soil– Stress 1
foundation interface and often the superstructure respond Strain 1
Gravitation 1/N
nonlinearly during an earthquake. Accordingly, results from the
Acceleration 1/N
substructure approach require validation using either high-quality Dynamic time N
laboratory tests or case history data. Recent computational studies Dynamic frequency 1/N
have addressed a variety of issues regarding adjacent foundations,
SSSI, and city-site effects [20–24].
The goal of experimental SFSI work is to measure SFSI effects study have been converted to prototype scale unless otherwise
in either a well-controlled laboratory experiment, or by instru- noted. Readers can convert important results contained within
menting SFS systems in the field. In the field, researchers can this paper by employing these scaling factors.
instrument SFS systems with seismometers and passively moni-
tor for earthquakes, but it is impossible to know when an earth-
3.2. Experimental set-up
quake will affect the instrumented site. For this reason, micro-
tremor analysis can also be employed [25], with the results
The centrifuge tests were performed at a centrifugal accelera-
extrapolated to larger earthquakes. In addition, forced vibration
tion of 55 g (N ¼55), and one-dimensional shaking was applied
of the SFS system can be used [26,27]. For both these approaches,
along the long dimension of the soil container (herein referred to
the resulting SFS system demand is small, and all components of
as the north–south direction). The Test-1 and Test-2 soil models
the system stay well within their elastic ranges. To the authors’
were constructed with Nevada sand that was dry-pluviated to a
knowledge, only one well-documented case history illustrating
target relative density (Dr) of 80%. The prototype scale depth of
the response of adjacent instrumented field structures to a large
both model sites was 29.5 m. In Test-1, the two model buildings
earthquake exists in the literature [28,29].
were located with a north–south foundation edge-to-foundation
In addition to field case history data, laboratory experiments
edge spacing of 31.4 m, and a north–south foundation edge-to-
that examine SFSI effects can be performed. The advantage of
container wall distance of approximately 15.5 m each. The static
laboratory testing compared to field testing is that the site
contact pressure of each MS1F_SF80 structural model footing was
conditions and structural systems are well defined and can be
1.87 atm, and the static contact pressure of the MS3F structural
examined under high-intensity earthquake motions. A popular
model basement mat was 1.67 atm.
laboratory device for SFSI testing is the geotechnical centrifuge,
In Test-2, the same two model buildings were placed adjacent
because it allows the researcher to measure the simultaneous
to one another, with only 0.27 m of foundation edge-to-
seismic response of the soil, foundation, and structure [30]. The
foundation edge separation. The static contact pressure of each
disadvantage of laboratory testing is the need to extrapolate the
MS1F_SF80 structural model footing was 1.87 atm (same as Test-
limited number of results to additional cases. Although a few
1), and the static contact pressure of the MS3F basement mat was
examples of physical SSSI modeling exist in the literature [31–33],
1.46 atm. Schematics of both experimental models are provided
the excitations were low amplitude and only elastic structural
in Fig. 3, which includes important dimensions and the locations
behavior was considered. The centrifuge experiments described in
of select instrumentation. The soil and model structures were
the following section represent the first, to the authors’ knowl-
instrumented with accelerometers, displacement transducers,
edge, experimental study of SSSI to include nonlinear SFSI
and strain gauges. These instruments recorded the seismic
coupled with inelastic structural response.
response of the soil and structures during the test program.
Detailed descriptions of the experimental set-ups, including full
instrumentation plans, are presented in the data report for each
3. Experimental set-up
test [15,16].
3.1. Centrifuge modeling principles
3.3. Earthquake motions
Model similitude is an important consideration in experimen-
tation using reduced-scale models that are intended to capture The earthquake motions used during Test-1 and Test-2 were
the response of field-scale prototypes. Herein, to remain consis- selected to reflect the wide variety of scenarios possible in
tent with centrifuge modeling literature, the nomenclature pro- downtown Los Angeles, CA (N34.082, W118.224). An interactive
totype scale and model scale is used to refer to full-scale and seismic hazard deaggregation tool developed by the United State
reduced-scale, respectively [34]. During a centrifuge test, the Geologic Survey (USGS) was employed to develop site-specific
stress and strain measured at model scale is the same as the target spectra at different return periods. In addition, the Next
stress and strain experienced by the prototype. Soil has stress- Generation Attenuation (NGA) relationships [38–41] were also
dependent strength, deformation, and volume change properties; used to create site-specific target spectra. The chosen earthquake
therefore, the centrifuge is a useful tool for examining geotechni- motions were chosen from the Pacific Earthquake Engineering
cal earthquake engineering problems. Additional information on NGA ground motion database. Time-domain scaling factors were
geotechnical centrifuge scaling laws can be found in Schofield employed to fit the chosen earthquake motions to the different
[35,36], Kutter [34], and Garnier et al. [37]. Table 1 contains a list target spectra. A more detailed account of the earthquake motion
of scale factors useful for this paper. All results presented in this selection procession is given in Mason et al. [42].
 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174 165

Fig. 3. Centrifuge models with important soil instrumentation demarcated for Test-1 (left) and Test-2 (right).

Table 2 Table 3
Earthquake motions used for Test-1 and Test-2. Earthquake motion order performed during Test-1 and Test-2.

Name Earthquake Station Type Test-1 Test-2

JOS 1992 Landers Joshua Tree 090 Ord Motion ID Earthquake Motion ID Earthquake
LCN 1992 Landers Lucerne 260 FD
PTS 1987 Sup. Hills Parachute Test Site 315 Ord 1 JOS_L_1 1 JOS_L_1
RRS 1994 Northridge Rinaldi Receiving Station 228 FD 2 TCU_L 2 TCU_L_1
SCS 1994 Northridge Sylmar Converter Station 052 FD 3 RRS 3 RRS_1
TCU 1999 Chi Chi TCU078-E Ord 4 PTS 4 PTS_1
WPI 1994 Northridge Newhall-W. Pico Canyon 046 FD 5 SCS_L_1 5 SCS_L_1
WVC 1989 Loma Prieta Saratoga-W. Valley Coll. 270 FD 6 LCN 6 LCN_1
PRI 1995 Kobe Port Island (79 m) FD 7 JOS_L_2 7 JOS_L_2
8 SCS_L_2 8 TCU_L_2
9 WVC_L 9 RRS_2
The Los Angeles site is located near active faults, so near-fault 10 SCS_H 10 PTS_2
11 JOS_H 11 SCS_L_2
earthquake motions are important. Forward-directivity (FD) effects 12 WPI_L 12 LCN_2
are observed in earthquake motions recorded near and in the 13 JOS_L_3 13 JOS_L_3
direction of fault rupture and are characterized by an intense, 14 WPI_H 14 WVC_L
short-duration velocity pulse(s). Accordingly, earthquake motions 15 PRI 15 SCS_H
16 TCU_H 16 JOS_H
exhibiting FD effects are potentially more damaging to the built
17 WVC_H 17 WPI_H
environment than ordinary earthquake motions recorded more 18 PRI
than 20–30 km from the earthquake rupture [43]. Both ordinary 19 TCU_H
and FD motions were utilized during this study. The FD earthquake 20 WVC_H
motions were selected (1) by updating the target spectra to include 21 JOS_L_4

the effects of directionality [44–46] and (2) by seeking FD earth-

quake motions from databases [44–48]. Detailed information the motion intensity variations. When this was the case, a suffix of
regarding earthquake selection, filtering, and actuator calibration ‘‘_H’’ or ‘‘_L’’ is employed to denote high or low intensity, respec-
procedures is provided in Mason et al. [49]. The final selected tively. (e.g. JOS_L_1 is a lower-intensity version of JOS_H, and
earthquake motions are given in Table 2. In this table, the ‘‘type’’ of JOS_L_1 was the first application of the JOS_L motion during that
earthquake motion refers to whether it has forward-directivity test). Table 3 shows the order of earthquake motions performed
characteristics (‘‘FD’’) or is an ordinary earthquake motion (‘‘Ord’’). during both Test-1 and Test-2.
Some of the earthquake motions shown in Table 2 were Fig. 4 shows the 5% damped elastic acceleration and displace-
performed multiple times per test. When this is the case, a number ment response spectra developed from representative base accel-
suffix is added to the earthquake name. In addition, some of the eration and free-field surface recordings of each of the earthquake
motions were performed at varying intensities. Applying a linear motions employed during Test-1 and Test-2 (i.e., JOS_L was
scale factor to the acceleration–time series of the motion resulted in performed multiple times, but is only illustrated with a single
166 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174

Fig. 4. Pseudo-acceleration (left) and displacement (right) response spectra of the earthquake motions used during Test-1 and Test-2. All spectra are developed
considering a damping ratio of 5%.

set of spectral plots because the resulting JOS_L recordings were indicate that the period lengthening was on the order of þ0.02 to
nearly identical). The spectra in Fig. 4 are developed from the þ0.05 s for only the most intense motions. Accordingly, we used a
actual recordings; i.e., the accelerometer data recorded during the site period value of 0.6 s for all earthquake motions.
centrifuge tests. In this figure, the word ‘‘base’’ indicates the
recording at the bottom of the soil layer (i.e., similar to a bedrock 3.4. Centrifuge boundary conditions
recording). From this figure, it can be seen that a variety of
possible earthquake scenarios were investigated. Earthquake The finite dimensions of the container used to house the
motions with a relatively short return period (i.e., JOS_L, PTS) centrifuge model can lead to unwanted ‘‘boundary effects,’’ which
and earthquake motions with a longer return period (i.e., PRI, affect the experimental results. Examples of unwanted boundary
SCS_H) were applied to the models. Accordingly, these series of effects include: (1) increased lateral resistance caused by the
motions represent adequately the seismic hazard at the site of container walls, (2) soil arching at the soil–container boundary,
interest. (3) a large impedance contrast between the rigid bottom boundary
Important earthquake motion intensity measures are also and overlying soil, (4) complex reflections and refractions of waves
examined. Fig. 5 shows the peak ground acceleration (PGA), peak off the container and back into the soil model, and (5) total stress
ground velocity (PGV), significant duration (D5-95), Arias intensity change at the soil–container interface; i.e., the ‘‘silo effect’’ [50,51].
(Ia), and 5%-damped spectral acceleration at the soil’s site period. In this subsection, units of length are reported in ‘‘centrifuge scale’’
The data in Fig. 5 are from recordings in both the free-field surface dimensions; accordingly, structural model dimensions are reported
and base. The soil’s fundamental site period is taken as approxi- in model scale.
mately 0.6 s, which is consistent with the observed peaks in the The NEES@UC Davis centrifuge facility employs several tech-
ratio of spectral acceleration from the base to the free-field nologies to reduce boundary effects and their potentially negative
ground surface, as shown in Fig. 6, and reasonable for a level site affect on experimental results [34,52]. First, the centrifuge is
that is 29.5 m deep with an average measured shear wave capable of carrying large payloads in relatively large centrifuge
velocity (Vs) of approximately 200 m/s. The peaks occasionally containers. The interior dimensions of the centrifuge container
observed in the spectral ratios at around 0.25 s are considered to employed during Test-1 and Test-2 are 1652 mm by 790 mm by
be a result of higher modes or other effects. The fundamental site 584 mm. If most of the important instrumentation is placed far
period, which is initially at around 0.6 s, does lengthen slightly from the container boundaries, then the boundary effects described
during the more intense earthquake motions (see Fig. 6). The data above can be minimized. Second, the container used for Test-1 and
 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174 167

Fig. 5. The peak ground acceleration (PGA), peak ground velocity (PGV), significant duration (D5-95), Arias intensity (Ia), and 5%-damped spectral acceleration at the site
period [Sa(Ts,5%)] recorded for each earthquake motion performed during Test-1 (left column) and Test-2 (right column) at the surface in the free-field and at the base.

structural influence. The free-field earthquake motion is often

used as a baseline motion, and motions recorded within the
structure and in the near-field (i.e., in the soil surrounding the
structure) are compared to it in both the time-domain and
frequency-domain. Trifunac [55] defined a free-field earthquake
motion as a motion recorded at least a distance that is an order of
magnitude larger than the characteristic dimension of the
structure-of-interest. In the case of the MS1F_SF80 structure,
the model footings are 80 mm square; therefore, the free-field
motion would need to be recorded at least 800 mm away. The
MS3F_B model structure has a 200 mm footprint, so the free-field
motion would need to be recorded at least 2000 mm away to
satisfy the Trifunac criterion. Such separation distances were not
Fig. 6. Ratio of spectral ordinates (surface to base) for the earthquake motions possible given the dimensions of the centrifuge container. Thus,
recorded during Test-1. the term ‘‘free-field’’ used herein indicates that these sensors are
not truly free from the effects of the structure and model
Test-2 is a flexible shear beam container with installed ‘‘shear container, but instead are separated as far as possible in the large
rods’’ [34]. The shear rods are steel threaded rods, stiff in compres- NEES@UC Davis centrifuge container.
sion and tension but flexible laterally, which can transfer comple- Figs. 7 and 8 show recordings from accelerometers located in two
mentary shear stresses to the container base. This design reduces different free-field locations at the ground surface during Test-1 and
the unrealistically high stress concentrations at the container’s end Test-2 for the JOS_L and SCS_H motions, respectively (see Fig. 3 for
walls, and reduces the silo effect. Specific boundary conditions of locations). The JOS_L motion is a lower-intensity motion, and the
the NEES@UC Davis apparatus, as well as the shaking table– SCS_H motion is a higher-intensity, forward-directivity motion. The
container–soil model interaction, are discussed in Ilankatharan recordings from the accelerometers are expressed in the time-
and Kutter [53]. In addition, surface wave reflections off the domain (acceleration–time series) and the frequency-domain (Four-
centrifuge container boundaries are discussed in Choy [54]. ier amplitude spectra). As shown in Figs. 7 and 8, there is little
SFSI analysis procedures often rely on a reliable definition of a discernible difference between the two free-field recordings for
‘‘free-field’’ earthquake motion. The free-field earthquake motion these two very different motions. As the two recordings are located
is defined as the motion recorded away from the presence of any at different locations relative to the structural models, if the
168 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174

Fig. 7. The JOS_L motion motions recorded during Test-1 and Test-2 using different free-field accelerometers in the time-domain and frequency-domain.

Fig. 8. The SCS_H motions recorded during Test-1 and Test-2 using different free-field accelerometers in the time-domain and frequency-domain.

structural models were affecting the free-field recordings, the two free-field surface and foundation-level motions is attributed to
recordings would differ. They do not. Hence, both recordings are kinematic SFSI. This difference between the free-field surface and
judged to be adequate, if imperfect, representations of the free-field foundation-level motion is important, because most earthquake
motion. engineers use the free-field surface motion for dynamic analyses,
when realistically, they should use the foundation-level motion.
3.5. Soil-model response to repeated seismic loadings Kinematic interaction effects are particularly important for short-
period buildings, for buildings with larger mat foundations, and
Seventeen and 21 earthquake motions were performed during for buildings founded on deeply embedded foundations [56].
Test-1 and Test-2, respectively. As the earthquake motions were The foundation-level motion is usually referred to as the
performed, the soil-model was progressively densified. However, foundation-input motion (FIM). The strictest definition of a FIM,
calculations based on the total cumulative surface settlement however, requires that the foundation and superstructure are
show that the change in relative density between the start of the both massless. This is a theoretical construct, not applicable to
testing (Dr 80%) (i.e., when the model was in a pristine condi- interpretation of experimental data. Accordingly, when examin-
tion) and the end of the testing (Dr  84%) was on the order of 3– ing experimental kinematic interaction results, it is impossible to
4% for both tests. In summary, the soil model was very dense at completely separate out the effects of inertial interaction caused
the beginning and end of testing, so the soil-model damage is by the vibrations of the superstructure and foundation from
expected to have little influence on the results presented herein. kinematic interaction effects. This does not render experimental
kinematic interaction results useless, however. Stewart [57] notes
that, ‘‘the effects of structural inertia on foundation motions tend
4. Kinematic interaction to be concentrated near the first-mode structural frequency, so
kinematic effects can be approximately evaluated across the
When seismic waves travel to the ground surface, they can remainder of the frequency spectrum.’’ Within this paper, the
either reach a free-field or a built environment. If they reach a term ‘‘foundation-level motion’’ is used in lieu of FIM to clearly
built environment, then embedded foundations change the demarcate an experimentally measured motion from a theoretical
waves’ characteristics with respect to waves recorded in the construct.
free-field. Thus, an earthquake motion recorded at the bottom Kinematic interaction changes the earthquake motion primarily
of a building’s foundation is different than an earthquake motion by two mechanisms: (1) base slab averaging and (2) embedment
recorded at the surface in the free-field. This difference between effects [58]. Base slab averaging occurs when incoherent, non-
 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174 169

Fig. 9. Acceleration–time series zoomed-in to strong shaking to show the difference between the free-field motion and basement-level motion. Note that (1) the time axes
have been shifted such that each subplot starts at zero, and (2) the acceleration scale is different for the JOS_L_1 motions to show more detail.

vertically propagating waves are ‘‘averaged’’ over the length of the surface in the free-field and a sensor located 55 cm (prototype
foundation. This averaging effect is facilitated by the impedance scale) under the basement slab of the MS3F_B structural model.
contrast between the foundation and the surrounding soil [3]. Base Unfortunately, sensors located at the bottom of the basement
slab averaging decreases the observed foundation translation; directly on the basement slab malfunctioned, so for the purpose
however, foundation rocking and torsion are exacerbated. Within of this paper, it has been assumed that the basement slab and the
the geotechnical centrifuge environment, base slab averaging may ground immediately below it respond nearly identically during an
be masked, because in the centrifuge environment, the waves are earthquake motion.
more coherent and vertically-propagating and the soil is more
homogeneous than typical field cases. Earthquake motions
recorded at the bottom of a foundation usually have smaller 5. Footing response
amplitude than earthquake motions recorded at the surface free-
field. Accordingly, embedment effects can be important for deeply The seismic response of shallow foundations has been studied at
embedded foundations, such as a basement [5]. the centrifuge scale more extensively in recent years [30,59,60]. These
In the experimental results, base slab averaging effects are experiments have revealed insights into the inelastic response of
observed in the data as a change in frequency content. Specifi- shallow foundations below simplified inelastic structural components
cally, the foundation-level motion contains less high-frequency (i.e., including footing uplift or soil yielding beneath a rocking mat
motion than the free-field surface motion, because of the aver- foundation). However, no centrifuge-based studies have investigated
aging (i.e., filtering) effect. Additionally, embedment effects work foundation response below flexible frame structures. Gelagoti et al.
to reduce the amplitude of the foundation-level motion. Fig. 9 [61,62] numerically show that the seismic response of individual
shows zoomed-in acceleration–time series for three different shallow foundations below flexible frame structures is modified
earthquake motions recorded during Test-1 and Test-2, and the significantly from the case of a similar foundation below a single-
data in this figure clearly show the aforementioned kinematic column structure. Importantly, Gelagoti et al. [61,62] concluded that
interaction effects. Fig. 9 contains data for a sensor located at the the dynamic variation in vertical footing loads due to superstructure
170 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174

frame action resulted in asymmetric foundation response. MS1F_SF80 ment box was slightly modified between Test-1 and Test-2. All
represents, to the authors’ knowledge, the first centrifuge model these deviations should have relatively minor, though non-zero,
designed to elicit the effect of inelastic frame action on the seismic effects on the results presented in this paper. More details about
response of individual spread footings. the structural model differences between the two tests are given
The structural model footings were instrumented with: in the companion paper by Trombetta et al. [63].
(1) horizontal and vertical accelerometers to capture the dynamic The foundation deformation is expressed in terms of the total
acceleration as well as transient velocity and displacement caused roof displacement (ur) of MS1F_SF80, which is described by
by each earthquake motion; and (2) horizontal and vertical
ur ¼ ug þuf ,avg þ hs ðyf ,avg þ aÞ þ us ð1Þ
displacement gauges to capture the permanent displacement
caused by each earthquake motion. The measurements from where ug is the ground displacement, uf,avg is the average footing
these instruments are utilized to investigate the seismic response sliding, hs is the height of the structure, yf is the average footing
of spread footings attached to a three-dimensional, inelastic, rotation, a is the angle of the differential settlement, and us is the
moment-resisting frame structure (i.e., MS1F_SF80). structural deformation. Eq. (1) can be formulated by replacing a
Experimental results from the two centrifuge tests yield SFS system with a multi-degree or single-degree-of-freedom
insights into SFSI of inelastic frame-structures and SSSI (when oscillator supported on a rigid, shallow foundation [2,7].
comparing results between the two tests). In this paper, insights Fig. 10 shows an elevation view schematic of the structural
into SFSI and SSSI are explored in terms of the relationships model footprints for Test-1 and Test-2 along with important
between foundation deformation and earthquake motion inten- nomenclature used throughout the rest of this paper. In Test-1,
sity measures (IM). The free-field peak ground velocity (PGV) the MS3F_B structural model is located away from the MS1F_SF80
recorded at the surface as well as the spectral displacement at the structural model. Accordingly, it is assumed that the seismic
flexible-base period and flexible-base damping ratio of the response of the MS3F_B structural model does not affect the
MS1F_SF80 structural model are useful IMs for characterizing response of the MS1F_SF80 footings. Because this is the case, all
earthquake motions. the MS1F_SF80 footings are considered ‘‘free’’ during Test-1.
Importantly, there are some differences between Test-1 and In Test-2, the MS3F_B structural model is located directly adja-
Test-2, which have some influence on the test-to-test compar- cent to the MS1F_SF80 structural model. Accordingly, it is
isons. First, the earthquake motion order is different between the assumed that the seismic response of the MS3F_B structural
two tests, as presented in Table 3. Second, the structural models model does affect the response of the MS1F_SF80 footings. This
were slightly different between the two tests. In particular, the is contrary to the Test-1 set-up, and this difference allows SSSI
structural configurations with respect to the earthquake motions effects of inelastic frame structures to be examined. The footings
performed were different. Third, the configuration of the base- closest to the MS3F_B structural model are considered

Fig. 10. Schematic of the footing nomenclature for Test-1 and Test-2.

Fig. 11. Transient roof displacement of the MS1F_SF80 structural model versus peak ground velocity recorded at the surface in the free-field for Test-1 (left) and Test-2
(right).
 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174 171

‘‘restrained’’ during Test-2, and the footings furthest from the three times larger than sliding induced roof displacements. During
MS3F_B structural model are considered to be still ‘‘free’’ during both tests, the footing rocking response correlated well with the
Test-2. Importantly, the Test-2 ‘‘free’’ footings are closer to the calculated spectral displacement values, and on average, correlated
MS3F_B structural model than the Test-1 ‘‘free’’ footings, so it is better than with the measured PGV. During Test-2, it was also
likely that the Test-2 ‘‘free’’ footings are affected somewhat by the observed that the ‘‘restrained’’ footings’ sliding contributed less to
seismic response of the MS3F_B structural model during Test-2, roof displacement than their ‘‘free’’ counterparts. During Test-1, the
and this is considered in the evaluation of footing response. footings had nearly identical measurements of peak sliding through-
However, there is a significant corresponding difference between out all levels of shaking. In terms of rotation, the ‘‘free’’ and
the south ‘‘free’’ footing of Test-1 and the ‘‘restrained’’ footing of ‘‘restrained’’ footings responded less erratically during Test-2 com-
Test-2, which is invaluable for investigating SSSI effects. Finally, pared to Test-1 through all levels of shaking. Even though peak
the MS1F_SF80 structural model has four footings, as shown in rotation measurements were nearly identical during Test-2, the
Fig. 3. In subsequent figures, the footings are labeled F1, F2, F3, moment–rotation responses of the ‘‘free’’ and ‘‘restrained’’ footings
and F4, which correspond to the northeast, southeast, northwest, were not. The seismic footing displacements recorded during both
and southwest footings, respectively. Accordingly, during Test-2, tests were greatly influenced by the superstructure above, and could
F1 and F3 are ‘‘free footings,’’ whereas F2 and F4 are ‘‘restrained not be predicted by available seismic foundation displacement
footings.’’ theories.
In Figs. 11 and 12 the transient peak contributions to roof Fig. 13 shows the cumulative residual settlement measured
displacement from individual sliding (uf,ind) and individual footing after each earthquake motion during both Test-1 and Test-2. The
rotation (hs  yf,ind) are plotted, respectively, versus peak ground residual settlements are calculated by subtracting the measured
velocity (PGV) and spectral displacement [Sd(T1SSI,b1SSI)]. The peak free-field surface settlement from the measured total footing
displacement values are calculated from double-integrated accel- settlement (measured with displacement transducers). Residual
erometer measurements recorded at the footings, and are thus displacements are not shown for some of the earthquake motions,
termed transient measurements, because they do not include because the necessary instruments malfunctioned during those
information regarding residual displacements. Peak transient motions. The x-axes of Fig. 13a and b are labeled ‘‘Motion ID,’’
sliding was calculated by taking the absolute maximum differ- which corresponds to the order of earthquake motions performed
ence between the recorded transient horizontal footing displace- during Test-1 and Test-2, as shown in Table 3.
ment measurement and the recorded transient free-field The MS1F_SF80 structural model is located at a distance from
displacement. Transient footing rotation was calculated by taking the MS3F_B structure; therefore, both the North and South Foot-
the difference between two vertical accelerometers (one each on ings are considered ‘‘free’’ footings during Test-1. From Fig. 13a, it
the north and south side of the footing) measurements on each can be seen that the residual settlement due to each earthquake
footing and normalizing by the footing width. The peak absolute motion is nearly the same for both the North and South footings
rotation measurements were then multiplied by the height of the during Test-1. This indicates that both footings are settling at
structure, yielding an estimate of roof displacement due to rock- approximately the same rate during the earthquake motions.
ing. As a result, both sliding and rocking are calculated in the For Test-2 (Fig. 13b), the ‘‘free’’ footing refers to the footing
units of displacement (mm), so they can be compared. Figs. 11 furthest from the MS3F_B structural model, and the ‘‘restrained’’
and 12 do not present information regarding the ‘direction’ of the footing refers to the footing directly adjacent to the MS3F_B
peak displacement demands, as they are absolute values. structural model. As shown in Fig. 13b, the residual displacement
Although the flexibility of the frame structure resulted in footing of the ‘‘restrained’’ footing is less than the residual displacement
displacements that were, at times, erratic (i.e., the footings did not of the ‘‘free’’ footing for every earthquake motion. The disparity
slide and rock in phase with each other during all time steps), between the residual displacements for the two footings becomes
several important observations can be gleaned from Figs. 11 and 12. larger for the later, and generally more intense, earthquake
First, the relative contributions of rocking and sliding to MS1F_SF80 motions.
roof displacement are immediately apparent. At all levels of shaking, Fig. 13c compares the cumulative displacement of the north
the rocking rotation of the spread footings contributes more to roof footing and the south footing recorded during both Test-1 and
displacements than sliding. For high intensity earthquakes, esti- Test-2. For Test-1, a linear regression yields a slope of 1.01, which
mated rocking-induced roof displacements were observed to be is nearly identical to the one-to-one line. This corroborates the

Fig. 12. The transient roof displacement versus the spectral displacement at the flexible-base period and damping ratio of the MS1F_SF80 structural model for Test-1 (left)
and Test-2 (right).
172 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174

Fig. 13. The total residual settlement following each earthquake motion in (a) Test-1 and (b) Test-2. (c) The cumulative settlement of the South footing versus the North
footing recorded during Test-1 and Test-2.

Fig. 14. The final locations of the free footing (footing 1) and restrained footing (footing 2) after all the earthquake motions had been performed during Test-2.

Fig. 15. Schematic illustrating the effect of a basement on footing settlement.

observations from Fig. 13a. When observing the Test-2 data, possible that the very stiff basement reduces cyclic shear strains
however, it can be seen that the slope is 0.44. This indicates that and settlement of the adjacent soil, while also increasing the
the South footing, which is the footing adjacent to the MS3F_B confining stress (and hence soil stiffness) below the footing.
structural model, settles less during each earthquake motion than Regardless of the underlying mechanism, it is clear that the
the North footing. This corroborates the observations from Fig. 13b. basement restrains the settlement of the adjacent footing, which
Comparing the cumulative settlement of the MS1F_SF80 foot- leads to increased differential settlement across the building
ings between Test-1 and Test-2 gives some insights into SSSI of frame and potentially higher moments in the column supported
frame structures founded on spread footings. First, it is instructive by the ‘‘restrained’’ footing.
to examine the final locations of the north (free) footing and south
(restrained) footing at the end of Test-2. This is shown in Fig. 14.
The ‘‘free’’ footing settled permanently more than the
‘‘restrained’’ footing; it slid permanently in the direction opposite 6. Conclusions
the MS3F_B structural model more than the ‘‘restrained’’ footings;
and it rotated counterclockwise more than the ‘‘restrained’’ There is a need for experimental soil–foundation–structure
footing. These observations have important implications for the interaction and structure–soil–structure interaction research. Cen-
seismic response of the inelastic superstructure. Differential trifuge testing is a useful tool for performing experimental SFSI and
settlement of the footings with differing restraint conditions SSSI research. Researchers can design, build, instrument, and test
produces increased demands within the superstructure that three-dimensional, inelastic frame structures with realistic founda-
may not be properly accounted for in an analysis that does not tions on a variety of soil conditions with realistic earthquake
consider SSSI effects. Additionally, the restrained rotation of the motions.
‘‘restrained’’ footing could lead to higher moments in the sup- Kinematic interaction results in a difference between the surface
ported column, which could be damaging if not properly free-field ground motion and the motion recorded at the foundation-
accounted for in the analysis. Further discussions about higher level. This difference is oftentimes neglected in engineering design,
moments developing in the supported column are given in the where most designers use the surface free-field motion directly as
companion paper. input into dynamic analyses. Kinematic interaction effects are
The schematic shown in Fig. 15 illustrates one possible effect expressed typically by a reduction in higher-frequency content and
of a basement on settlement of a neighboring footing. The very a reduction in amplitude of the foundation-level motion when
strong, stiff basement impedes the development of a bearing compared to the surface free-field motion. Results from these two
capacity mechanism, as shown in the right side of Fig. 15. It is also SFSI centrifuge tests validate that previous methods developed to
 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174 173

account for kinematic interaction provide reasonable assessments of [12] Wong HL, Trifunac MD. Two-dimensional, antiplane, building-soil-building
SFSI effects. interaction for two or more buildings and for incident planet SH waves.
Bulletin of the Seismological Society of America 1975;65:1863–85.
Settlement, uplift, rocking and sliding are important SFSI and [13] Kausel E. Early history of soil–structure interaction. Soil Dynamics and
SSSI phenomena. Importantly, during Test-2, the adjacency of the Earthquake Engineering 2010;30:822–32.
two structures created a ‘‘restrained’’ footing condition as [14] Menglin L, Huaifeng C, Xi C, Yongmei Z. Structure–soil–structure interaction:
literature review. Soil Dynamics and Earthquake Engineering 2011;31:
opposed to what is normally considered, which is an unrestrained 1724–31.
or ‘‘free’’ footing condition. The ‘‘free’’ footing, which is not [15] Mason H, Trombetta N, Chen Z, Choy B, Bolisetti C, Bray J, et al.. Seismic
adjacent to the basement, settled, uplifted, rocked and slid more peformance assessment in dense urban environments. In: Centrifuge data
report for HBM02. Davis, CA: Center for Geotechnical Modeling; 2011.
than the ‘‘restrained’’ footing. The differing responses of the two
[16] Mason H, Trombetta N, Chen Z, Choy B, Bolisetti C, Bray J, et al.. Seismic
footing conditions induce additional demands on the superstruc- peformance assessment in dense urban environments. In: Centrifuge data
ture. Thus, neglecting the interaction between two adjacent report for HBM03. Davis, CA: Center for Geotechnical Modeling; 2011.
structures can be unconservative, and in these cases SSSI effects [17] Kobori T, Kusakabe, K. Dynamic cross-interaction between two embedded
structures. In: Fifth Japan earthquake engineering symposium, Tokyo, Japan;
should be considered. 1978.
Finally, these centrifuge tests, to the authors’ knowledge, represent [18] Kobori T, Kusakabe K. Cross-interaction between two embedded structures in
the first attempt to characterize SFSI and SSSI effects of inelastic earthquakes. In: Seventh world conference on earthquake engineering,
Istanbul, Turkey; 1980.
frame structures founded on shallowly embedded footings. The [19] Wolf JP. Dynamic soil–structure interaction. Upper Saddle River, NJ: Prentice-
footing response, at times, is complex, and cannot be described by Hall, Inc.; 1985.
simplified theoretical models. More work is needed to understand [20] Clouteau D, Aubry D. Modifications of the ground motion in dense urban
areas. Journal of Computational Acoustics 2001;9:1659–76.
footing response with respect to SFSI and SSSI effects. [21] Ghergu M, Ionescu IR. Structure–soil–structure coupling in seismic excitation
and city effect. International Journal of Engineering Science 2009;47:
342–54.
[22] Mulliken JS, Karabalis DL. Discrete model for dynamic through-the-soil
Acknowledgments coupling of 3-D foundations and structures. Earthquake Engineering &
Structural Dynamics 1998;27:687–710.
[23] Qian J, Beskos DE. Dynamic interaction between 3-D rigid surface founda-
This material is based upon work supported by the National tions and comparison with the ATC-3 provisions. Earthquake Engineering &
Science Foundation under Grant no. CMMI-0830331. Any opinions, Structural Dynamics 1995;24:419–37.
findings, and conclusions or recommendations expressed in this [24] Wirgin A, Bard P-YY. Effects of buildings on the duration and amplitude of
ground motion in Mexico City. Bulletin of the Seismological Society of
material are those of the authors and do not necessarily reflect the America 1996;86:914–20.
views of the National Science Foundation. Experiments were [25] Trifunac MD. Comparisons between ambient and forced vibration experi-
conducted at the Center for Geotechnical Modeling at the University ments. Earthquake Engineering & Structural Dynamics 1972;1:133–50.
[26] Jennings PC. Distant motions from a building vibration test. Bulletin of the
of California at Davis, which is supported by the NSF George E. Brown,
Seismological Society of America 1970;60:2037–43.
Jr. Network for Earthquake Engineering Simulation (NEES) program [27] Luco JE, Trifunac MD, Wong HL. Isolation of soil–structure interaction effects
under Award no. CMMI-0402490. The authors gratefully acknowl- by full-scale forced vibration tests. Earthquake Engineering & Structural
edge the assistance of the UCD staff and in particular Dan Wilson, Dynamics 1988;16:1–21.
[28] Celebi M. Seismic responses of two adjacent buildings: data and analyses.
Ray Gerhard, and Anatoliy Ganchenko. We would also like to give Journal of Structural Engineering 1993;119:2461–76.
credit for the substantial assistance of the other researchers involved [29] Celebi M. Seismic responses of two adjacent buildings: interaction. Journal of
in this research project, but not listed in the authorship: Benjamin Structural Engineering 1993;119:2477–92.
[30] Gajan S, Kutter BL, Phalen JD, Hutchinson TC, Martin GR. Centrifuge modeling
Choy, Gregg Fiegel, Katherine Jones, Roshani Patel, Clayton Proto, of load-deformation behavior of rocking shallow foundations. Soil Dynamics
Chandrakanth Bolisetti, Andrew Whittaker, and Robert Reitherman. and Earthquake Engineering 2005;25:773–83.
Lastly, we would like to acknowledge the contributions of our [31] Kitada Y, Hirotani T, Iguchi M. Model test on dynamic structure–structure
interaction of nuclear power plant buildings. Nuclear Engineering and Design
professional practice committee: Marshall Lew, Mark Moore, Farzad 1999;192:205–16.
Naeim, Farhang Ostadan, Paul Somerville, and Michael Willford. [32] Mizuno H. Effects of structure–soil–structure interaction during various
excitations. In: Seventh world conference on earthquake engineering, Istan-
bul, Turkey; 1980.
[33] Nakagawa S, Kuno M, Naito Y, Nozawa T, Momma T, Mizuno J, et al. Forced
References
vibration tests and sumulation analyses of a nuclear reactor building. Nuclear
Engineering and Design 1998;179:145–56.
[1] Gazetas G. Formulas and charts for impedances of surface and embedded [34] Kutter BL. Recent advances in centrifuge modeling of seismic shaking. In:
foundations. Journal of Geotechnical Engineering 1991;117:1363–81. Proceedings of the third international conference on recent advances in
[2] Jennings PC, Bielak J. Dynamics of building–soil interaction. Bulletin of the geotechnical earthquake engineering and soil dynamics, St. Louis, MO; 1995.
Seismological Society of America 1973;63:9–48. [35] Schofield AN. Cambridge geotechnical centrifuge operations. Geotechnique
[3] Kramer SL, Stewart JP. Geotechnical aspects of seismic hazards. In: Bozorgnia 1980;30:227–68.
Y, Bertero VV, editors. Earthquake engineering: from engineering seismology [36] Schofield AN. Dynamic and earthquake geotechnical centrifuge modelling. In:
to performance based engineering. Boca Raton, FL: CRC Press; 2004. International conference on recent advances in soil dynamics and earthquake
[4] Luco JE. Dynamic interaction of a shear wall with the soil. Journal of the engineering, St. Louis, Missouri; 1981.
Engineering Mechanics Division 1969;95:333–46. [37] Garnier J, Gaudin C, Springman SM, Culligan PJ, Goodings D, Konig D, et al.
[5] Stewart JP, Fenves GL, Seed RB. Seismic soil–structure interaction in build- Catalogue of scaling laws and similitude questions in geotechnical centrifuge
ings. I: analytical methods. Journal of Geotechnical and Geoenvironmental modelling. International Journal of Physical Modelling in Geotechnics 2007;3:
Engineering 1999;125:26–37. 1–23.
[6] Stewart JP, Seed RB, Fenves GL. Seismic soil–structure interaction in build- [38] Abrahamson N, Silva W. Summary of the Abrahamson & Silva NGA ground–
ings. II: empirical findings. Journal of Geotechnical and Geoenvironmental motion relations. Earthquake Spectra 2008;24:67–97.
Engineering 1999;125:38–48. [39] Boore DM, Atkinson GM. Ground-motion prediction equations for the
[7] Veletsos AS, Meek JW. Dynamic behaviour of building-foundation systems. average horizontal component of PGA, PGV, and 5%-damped PSA at spectral
Earthquake Engineering & Structural Dynamics 1974;3:121–38. periods between 0.01 s and 10.0 s. Earthquake Spectra 2008;24:99–138.
[8] Veletsos AS, Nair VVD. Seismic interaction of structures on hysteretic [40] Campbell KW, Bozorgnia Y. NGA ground motion model for the geometric
foundations. Journal of the Structural Division 1975;101:109–29. mean horizontal component of PGA, PGV, PGD and 5\% damped linear elastic
[9] Taborda R. Three dimensional nonlinear soil and site-city effects in urban response spectra for periods ranging from 0.01 to 10 s. Earthquake Spectra
regions, in: civil and environmental engineering. Pittsburgh, PA: Carnegie 2008;24:139–71.
Mellon University; 2010. [41] Chiou BS-J, Youngs RR. An NGA model for the average horizontal component
[10] Luco JE, Contesse L. Dynamic structure–soil–structure interaction. Bulletin of of peak ground motion and response spectra. Earthquake Spectra
the Seismological Society of America 1973;63:1289–303. 2008;24:173–215.
[11] Lee TH, Wesley DA. Soil–structure interaction of nuclear reactor structures [42] Mason, HB, Bray, JD, Jones, KC, Chen, Z, Hutchinson, TC, Trombetta, NW, et al..
considering through-soil coupling between adjacent structures. Nuclear Earthquake input motions and seismic site response in a centrifuge test
Engineering and Design 1973;24:374–87. examining SFSI effects. In: Proceedings of the fifth international conference
174 H.B. Mason et al. / Soil Dynamics and Earthquake Engineering 48 (2013) 162–174

on recent advances in geotechnical earthquake engineering and soil [53] Ilankatharan M, Kutter B. Modeling input motion boundary conditions for
dynamics, University of Missouri, Rolla, Rolla, MO; 2010. simulations of geotechnical shaking table tests. Earthquake Spectra 2010;26:
[43] Alavi B, Krawinkler H. In: John A, editor. Effects of near-fault ground motions on 349–69.
frame structures. Stanford, CA: Blume Earthquake Engineering Center; 2001. [54] Choy BY. Forced vibration test in a centrifuge test examining SSSI effects.
[44] Huang Y-H, Whittaker AS, Luco N. Maximum spectral demands in the near- Davis, CA: University of California; 2011.
fault region. Earthquake Spectra 2008;24:319–41. [55] Trifunac MD. Interaction of a shear wall with the soil for incident plane SH
waves. Bulletin of the Seismological Society of America 1972;62:63–83.
[45] Somerville PG, Smith NF, Graves RW, Abrahamson NA. Modifications of
[56] FEMA-440. Improvement of nonlinear static seismic analysis procedures.
empirical strong ground motion attenuation relations to include the ampli-
Washington, DC: Federal Emergency Management Agency; 2005.
tude and duration effects of rupture directivity. Seismological Research
[57] Stewart JP. An empirical assessment of soil–structure interaction effects on
Letters 1997;68:199–222. the seismic response of structure. Berkeley, CA: University of California;
[46] Spudich P, Chiou BSJ. Directivity in NGA earthquake ground motions: 1996.
analysis using isochrone theory. Earthquake Spectra 2008;24:279–98. [58] Stewart JP, Seed RB, Fenves GL. Empirical evaluation of inertial soil–structure
[47] Baker JW. Quantitative classification of near-fault ground motions using interaction effects. Berkeley, CA: Pacific Earthquake Engineering Research
wavelet analysis. Bulletin of the Seismological Society of America Center; 1998.
2007;97:1486–501. [59] Chang, BJ, Raychowdhury, P, Hutchinson, TC, Gajan, S, Kutter, BL, Thomas, J.
[48] Bray JD, Rodriguez-Marek A. Characterization of forward-directivity ground Centrifuge testing of combined frame–wall–foundation structural systems.
motions in the near-fault region. Soil Dynamics and Earthquake Engineering In: Proceedings of the national conference on earthquake engineering. San
2004;24:815–28. Francisco, CA: Earthquake Engineering Research Institute; 2006.
[49] Mason, HB, Bray, JD, Kutter, BL, Wilson, DW Choy, BY. Earthquake motion [60] Deng L, Kutter B. Characterization of rocking shallow foundations using
selection and calibration for use in a geotechnical centrifuge. In: Proceedings centrifuge model tests. Earthquake Engineering & Structural Dynamics
of the seventh international conference on physical modelling in geotechnics, 2012;41:1043–60.
[61] Gelagoti F, Kourkoulis R, Anastasopoulos I, Gazetas G. Rocking-isolated frame
Taylor & Francis, London, England; 2010.
structures: margins of safety against toppling collapse and simplifed design
[50] Dashti S. Toward developing an engineering procedure for evaluating building
approach. Soil Dynamics and Earthquake Engineering 2012;32:87–102.
performance on softened ground. Berkeley, CA: University of California; 2009.
[62] Gelagoti F, Kourkoulis R, Anastasopoulos I, Gazetas G. Rocking isolation of
[51] Hausler EA. Influence of ground improvement on settlement and liquefac- low-rise frame structures founded on isolated footings. Earthquake Engineer-
tion: a study based on field case history evidence and dynamic geotechnical ing & Structural Dynamics 2012;41(7):1177–97.
centrifuge tests. Berkeley, CA: University of California; 2002. [63] Trombetta, N.W., Mason, H.B., Chen, Z., Hutchinson, T.C., Bray, J.D., Kutter, B.L..
[52] Kutter, BL, Idriss, IM, Kohnke, T, Lakeland, J, Li, XS, Sluis, W, et al. Design of a Nonlinear dynamic foundation and frame structure response observed in geo-
large earthquake simulator at UC Davis. In: Centrifuge 94, A.A. Balkema, technical centrifuge experiments. Soil Dynamics and Earthquake Engineering,
Rotterdam; 1994. p. 169–75. in press.