You are on page 1of 14

Soil–Structure Interaction Effects on Seismic Performance

and Earthquake-Induced Losses in Tall Buildings


Luis G. Arboleda-Monsalve, M.ASCE 1; Jaime A. Mercado, S.M.ASCE 2;
Vesna Terzic 3; and Kevin R. Mackie, F.ASCE 4

Abstract: This paper evaluated the seismic performance of hypothetical tall buildings by estimating intensity measures, engineering demand
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

parameters, and earthquake-induced losses using a soil–structure interaction (SSI) numerical framework. Numerical models for 40-story
buildings were developed using OpenSees to study their seismic performance under the following modeling and building configuration
conditions: (1) fixed-base structural model, (2) model including SSI effects, (3) fixed-base model with shear walls, and (4) model including
shear walls and considering SSI effects. The buildings were assumed to be supported on subsurface conditions typical of downtown
Los Angeles. The natural period of the soil profile was parametrically studied; the larger its natural period, the lower were the seismic
demands of the building. The inclusion of shear walls caused a reduction of the natural period of the building and computed settlements
in relation to the buildings without a shear wall system. Considering SSI effects in the modeling approach changed the computed seismic
demands of the tall buildings in terms of maximum interstory drifts, peak story horizontal accelerations, and seismic-induced settlements.
Computed median direct economic losses for the 2,475-year mean return period increased as much as 33% by considering SSI effects in the
numerical analyses in relation to building losses ignoring those effects. DOI: 10.1061/(ASCE)GT.1943-5606.0002248. © 2020 American
Society of Civil Engineers.

Introduction and described the importance of including SSI in the numerical


simulations as opposed to using conventional fixed-base building
Civil infrastructure experiences significant damage quantified in models. SSI effects are properly accounted for in the analysis when
terms of direct and indirect losses after an earthquake strikes. soil strength and stiffness reduction (i.e., strength–stress–strain
Earthquake-induced losses are a major concern to owners and in- soil response), soil damping evolution characteristics, and soil–
surance companies because major earthquakes result in loss of life foundation–structure responses are modeled as a coupled system
and create significant societal disruption. The performance-based (i.e., direct approach). The seismic response of soils is oversimpli-
earthquake engineering (PBEE) framework, described by Miranda fied by considering lumped spring systems to model strength and
and Aslani (2003), is used to quantify the probabilistic response of stiffness reduction (softening) and damping evolution characteris-
buildings using performance metrics that include not only engineer- tics of most soils. The PBEE approach provides a probabilistic
ing demand parameters (EDPs), but also individual building com- framework to quantify EDPs as a function of the intensity of the
ponent damage and losses. PBEE evaluation generally is performed seismic hazard: intensity measures (IM). Then, computed probabi-
for low- and midrise buildings using fixed-base building models, listic seismic demand models are coupled with individual component
thus ignoring the influence of the foundation soils in the structural damage fragilities and repair scenarios to estimate earthquake-
seismic response (e.g., Jayaram et al. 2012; Molina Hutt et al. 2016; induced losses. Several PBEE studies have evaluated the influence
Kolozvari et al. 2018). The influence of soil–structure interaction of different structural configurations on loss estimation. For example,
(SSI) on the dynamic response of buildings needs to be assessed Kolozvari et al. (2018) studied earthquake-induced losses of a
to avoid unrealistic computation of EDPs. Mercado et al. (2019) 15-story building, evaluating the influence of shear walls on its seis-
evaluated SSI effects in tall buildings using direct approaches mic performance. Median repair costs up to 24% were computed for
1
earthquakes with a 2,475-year mean return period. Molina Hutt et al.
Assistant Professor, Dept. of Civil, Environmental, and Construction (2016) assessed the seismic losses and repair time of 40-story build-
Engineering, Univ. of Central Florida, Orlando, FL 32816 (corresponding
ings constructed in the mid-1970s and 1980s using the PBEE meth-
author). ORCID: https://orcid.org/0000-0002-2977-2544. Email: luis
.arboleda@ucf.edu
odology. They estimated median expected direct economic losses for
2
Graduate Research Assistant, Dept. of Civil, Environmental, and Con- a 40-story steel moment-resisting frame structure up to 34% of the
struction Engineering, Univ. of Central Florida, Orlando, FL 32816. Email: replacement building cost at the 475-year mean return period hazard
jamercado@knights.ucf.edu level. Jayaram et al. (2012) evaluated the performance of new tall
3
Associate Professor, Dept. of Civil Engineering and Construction En- buildings (i.e., 20-, 40-, and 42-story structures) and estimated
gineering Management, California State Univ., Long Beach, CA 90840. the losses associated with the building performance, finding direct
Email: vesna.terzic@csulb.edu losses of as much as 27% of the replacement building cost at the
4
Professor, Dept. of Civil, Environmental, and Construction Engineer- 500-year mean return period hazard level. None of the aforemen-
ing, Univ. of Central Florida, Orlando, FL 32816. ORCID: https://orcid
tioned studies included SSI effects on the loss estimation.
.org/0000-0003-1287-6520. Email: Kevin.Mackie@ucf.edu
Note. This manuscript was submitted on May 6, 2019; approved on
This paper focused on the influence of SSI on the seismic
December 4, 2019; published online on March 11, 2020. Discussion period response and estimation of losses for hypothetical tall building
open until August 11, 2020; separate discussions must be submitted for models using site-specific seismic hazard analyses and engineer-
individual papers. This paper is part of the Journal of Geotechnical ing demands computed with direct and continuum soil–structure
and Geoenvironmental Engineering, © ASCE, ISSN 1090-0241. interaction approaches. It was hypothesized herein that the response

© ASCE 04020028-1 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


of tall buildings to strong earthquake motions and the quantification sediments from the Pleistocene and Holocene ages, and some local
of losses largely are affected by the geotechnical characteristics of formations from the Pliocene and Miocene ages are present in the
the foundation soils. Four tall buildings models [i.e., (1) a fixed-base downtown Los Angeles area. The Pliocene and Miocene sedimen-
building model, (2) a building model including SSI effects, (3) a tary materials, commonly referred to as Fernando and Puente for-
fixed-base building model with shear walls acting as the main lateral mations, respectively, were identified on the geologic maps by
load-resisting system, and (4) a building model including shear walls Lamar (1970) and Dibblee and Ehrenspeck (1991). These forma-
and SSI effects] were developed in OpenSees version 2.5.0 tions provide firm-ground conditions to the Los Angeles downtown
(McKenna et al. 2000) to assess the influence of SSI on the seismic area. The subsurface soil stratigraphy and geotechnical character-
performance of tall buildings. These numerical models were devel- istics used herein corresponded to downtown Los Angeles and were
oped in two dimensions, considering torsion and eccentricity effects based on field exploration data in terms of standard Penetration
on structures, and three-dimensional (3D) radiation patterns in the tests (SPTs), cone penetration tests (CPTs), and shear wave velocity
soil are not considered. The analyses presented herein consider soundings (V s ). The collected information is publicly available in
earthquake-induced soil strength and stiffness reduction, damping the GeoDOG online tool developed by the California Department
evolution characteristics of the supporting soils, and coupled soil– of Transportation and was supplemented with geotechnical data for
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

foundation–structure system response. Quantifying discrepancies the construction of three underground train stations in downtown
on the structural response arising from the proposed modeling strat- Los Angeles (AMEC 2013).
egies is key for the estimation of earthquake-induced losses in tall Fig. 1 presents the field investigation data used to summarize
buildings. The buildings were supported on highly nonlinear- the subsurface conditions up to 40 m under the ground surface
inelastic deformable materials. The baseline soil conditions of this in the area of study, mainly consisting of artificial fills composed
study corresponded to downtown Los Angeles; a generalized soil of mixtures of silty to clayey sands and gravels deposited on top of
profile was presented. These soil conditions were determined from dense alluvium materials. Alluvium materials are denoted Qal1 and
field investigation public records obtained from the California De- Qal2 in the figure. Soil layer Qal2 has more fine-grained soils than
partment of Transportation Digital Archive Geotechnical Data does layer Qal1. In addition to slight differences in particle size
(GeoDOG) and complemented with site investigations developed between Qal1 and Qal2, the alluvium deposit layer from depths
for the construction of Metro stations in downtown Los Angeles 5.5 to 15 m was divided to account for the approximate location
(AMEC 2013). of the water table: 10.5 m from the ground surface. The relative
The IMs considered in this study were peak ground acceleration density of each soil layer was computed using the relationship pro-
(PGA), peak ground velocity (PGV), Arias intensity (AI), and posed by Meyerhof (1957), which correlates the results of SPT
Housner intensity (HI). Three earthquake hazard levels were con- blow counts (N SPT ) corrected for energy ratio, borehole diameter,
sidered, 50%, 10%, and 2% probability of exceedance in 50 years, sampling method, rod length, and overburden pressure [i.e., ðN 1 Þ60
which resulted in 72, 475, and 2,475-year mean return periods, re- values]. The Qal2 layer has an average relative density (Dr ) larger
spectively. The EDPs considered herein were peak story horizontal than 80%; thus, liquefaction phenomenon may not occur. However,
accelerations, maximum interstory drifts, plastic rotation of the Karimi et al. (2018) showed that dense soils can soften signifi-
shear walls, and maximum vertical displacements at the base of cantly, affecting the engineering demands in the superstructure and
the building. As part of the SSI analyses, the pressure-dependent the response of the foundation in terms of settlement and tilt. At
multiyield-surface (PDMY02) constitutive soil model was used. approximately 15.5 m from the ground surface, a very dense clayey
The main structural components were modeled using linear elastic silt transitioning to siltstone is found. Soil contents in the figure
elements, but the shear walls were modeled using the nonlinear were obtained from sieve analyses in the geotechnical data of
shear-flexure interaction multiple vertical line element model AMEC (2013). Based on V s measurements of the topmost 30 m,
(SFI-MVLEM) coded in OpenSees and proposed by Kolozvari et al. the computed natural frequency of the soil deposit was 3.2 Hz
(2015a, c). Parametrically, fixed-base building models and those (a natural period of 0.31 s). The soil profile was classified as site
including SSI effects were compared in this research and conclu- class C (ASCE 2017) according to the V s measured for the topmost
sions were drawn regarding the seismic response of the analyzed 30 m of the soil profile. The figure also compiles friction angle
structures. Transfer functions also were presented to provide insight values from laboratory direct shear and triaxial tests developed for
into the SSI influence on period lengthening characteristics of the soils in the area and values determined using the correlation
the analyzed systems when foundation soils are included in the with N SPT proposed by Hatanaka and Uchida (1996).
analyses. The natural period of the soil was parametrically varied
to support the idea that SSI affects the seismic performance of tall
buildings and to assess the influence of different soil profile char- Constitutive Soil Model and Structural Model
acteristics on the structural response. The contribution of structural Characteristics
and nonstructural building components to earthquake-induced di-
rect losses was studied in four building models placed on a soil The numerical simulations were developed using OpenSees. The
profile typical of downtown Los Angeles. The Performance Assess- soil domain was modeled using nine-node quadrilateral plane-
ment Calculation Tool (PACT version 3.1.1) (ATC 2012) was used strain elements capable of simulating the dynamic response of a
in this paper to estimate losses. solid–fluid fully coupled material based on the Biot (1962) theory
for porous media. The constitutive soil model used was PDMY02
(Elgamal et al. 2002; Yang et al. 2003, 2008). This constitutive soil
Local Geology and Subsurface Conditions model is an elastoplastic model used to simulate the monotonic and
cyclic response characteristics (i.e., dilatancy or contraction and
The Los Angeles basin is a sedimentary basin located south of nonflow liquefaction) of soils depending on the confining pressure.
the Santa Monica mountains and within the Peninsular Range geo- The PDMY02 model defines the multiyield criterion as the number
morphic province of southern California. This province trends of open conical yield surfaces (i.e., Drucker–Prager-type yield sur-
northward and is characterized by northwest trending faults. faces) with a common apex at the origin. The outermost surface
The Los Angeles basin floor is characterized by unconsolidated defines the shear strength envelope of the material. This soil model

© ASCE 04020028-2 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Summarized subsurface conditions corresponding to downtown Los Angeles area.

Table 1. PDMY02 constitutive soil parameters computed for proposed soil conditions
Parameter Description Fill Qal1 Qal2 Siltstone
k (m=s) Hydraulic conductivity 4.0 ×10−5 2.0 × 10−5 2.3 ×10−5 3.0 × 10−5
V s (m=s) Shear wave velocity 228.6 280.1 396.2 487.6
Dr (%) Relative density 75 96 91 85
e Void ratio 0.65 0.5 0.55 0.6
ρ (Mg=m3 ) Saturated soil mass density 1.93 1.98 2.06 2.01
Gmax;1;oct (MPa) Octahedral low-strain shear modulus 124.0 190.9 137.3 587.9
Br (MPa) Bulk modulus 342.0 417.1 998.1 1,547.7
ϕTXC (degrees) Friction angle under triaxial compression conditions 29.3 34.6 31.3 30.3
ϕPT (degrees) Phase transformation angle 24.3 29.6 26.3 25.3
ϒmax;r Maximum octahedral shear strain 0.1 0.1 0.1 0.1
Pr0 (kPa) Reference confining pressure 101 101 101 101
d Pressure dependency coefficient 0.5 0.5 0.5 0.5
c1 Contaction rate parameter 0.005 0.001 0.002 0.004
c2 Fabric damage parameter 0.5 0.1 0.25 0.5
c3 Overburden stress effect parameter 0 0 0 0
d1 Dilation rate parameter 0.15 0.8 0.6 0.4
d2 Fabric damage parameter 3 3 3 3
d3 Overburden stress parameter 0 0 0 0
NYS Number of yield surfaces 20 20 20 20
liquefac1 Control liquefaction-induced perfectly plastic shear strain 1 1 1 1
liquefac2 Control liquefaction-induced perfectly plastic shear strain 0 0 0 0

uses nonlinear kinematic hardening principles and a nonassociative computed using V s measurements. The model friction angle (ϕ), as-
flow rule to reproduce the dilative or contractive behavior of most sumed as the friction angle under triaxial compression conditions,
soils. The plastic flow in this model is purely deviatoric, and thus was determined using N SPT correlations and compared with a data-
no plastic volume changes take place under a constant stress ratio. base of laboratory direct shear and triaxial test results developed for
Although the soils in this study were not susceptible to liquefaction the soils in the Los Angeles downtown area. The parameters control-
because of their large relative density, the model was used to ac- ling the mechanisms of plastic shear strain accumulation and rate of
count for the soil stiffness degradation and shear strength mobili- shearing-induced volume changes were selected based on previous
zation when subjected to earthquake-induced cyclic loading and to studies developed of medium-dense to dense sands (Khosravifar
study soil–foundation–structure interaction effects on the seismic et al. 2014; Karimi and Dashti 2015) and recommended values in
response of tall buildings. the OpenSees manual (Yang et al. 2008).
The constitutive model parameters were determined using the Four numerical models were developed to evaluate the seismic
soil characteristics from the geotechnical field investigation program. response of tall buildings including SSI effects. The first two models
Table 1 presents the PDMY02 soil parameters and a brief description corresponded to a 40-story building numerically simulated assuming
of their physical meaning. The soil parameters Gmax;1;oct and Br were fixed-base conditions or modeling the soil domain realistically in a

© ASCE 04020028-3 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Soil–structure–foundation interaction finite-element models of the tall buildings.

continuum and direct form. The other two models corresponded to The buildings initially were modeled assuming fixed-base con-
the same hypothetical tall building but included shear walls in the ditions. Then the soil domain was added, and the mat foundation
middle bay acting as the main lateral load-resisting mechanism, thus nodes were tied to the soil nodes using master/subordinate node
causing a reduction in the building flexibility. constraints, which implied zero relative movement (i.e., no gapping
Fig. 2 presents the finite-element discretization of the two soil– or sliding) and perfect friction at the interface (i.e., no interface
structure interaction models developed in OpenSees. The buildings elements). On the left and right sides of the model, tied degrees
had 40 stories and 5 bays. Both buildings had height and width of of freedom were applied to the nodes at the same elevation. The
120 and 30 m, respectively. Beams and columns had lengths of 6.0 bottom boundary of the soil domain was modeled assuming that
and 3.0 m, respectively, and were modeled using linear elastic el- the nodes were fixed against displacements in the vertical direction.
ements (i.e., elasticBeamColumn elements in OpenSees). Cross At the bottom left corner of the model, a Lysmer and Kuhlemeyer
sections were assumed to be W27 × 161 and W24 × 104 for col- (1969) dashpot was utilized to account for the compliance of the
umns and beams, respectively, to assign specific values of area and underlying elastic medium. The dashpot properties were selected
moment of inertia to the structural elements in the model. A 1.5-m- to represent the siltstone material found from the geotechnical test-
thick concrete mat foundation was added to the base of the model ing reports (i.e., V s ¼ 700 m=s and mass density ¼ 2.5 Mg=m3 ).
as a linear elastic element with an elastic modulus of 31 MPa. Mat The dashpot coefficient was 1,750 Mg=s. The finite-element mesh
foundation and structural elements were connected using the same was discretized to have element sizes varying from 1.0 to 2.5 m.
node; therefore, they shared the same three degrees of freedom. This represented 8–10 elements/wavelength to avoid numerical
The shear walls were modeled using the SFI-MVLEM proposed dispersion. The finer parts of the mesh were used for the soil clus-
by Kolozvari et al. (2015a, c). This model captures the interaction ters underneath and surrounding the buildings. Although the model
between axial-flexural and shear behavior of RC structural walls considered elastoplastic soil behavior and built-in constitutive
and columns under cyclic loading. The SFI-MVLEM uses ad- model damping, a 5% Rayleigh damping proportional to the mass
vanced constitutive relationships for concrete and steel materials and committed elements stiffness was supplemented to account for
available in OpenSees, such as ConcreteCM (Chang and Mander damping at small strain values of the soil. Ground motions were in-
1994) and the Giuffre–Menegotto–Pinto uniaxial strain-hardening put to the model using the uniform excitation technique in OpenSees.
material model Steel02 (Filippou et al. 1983). The shear wall dis-
cretization in the vertical and horizontal directions consisted of two
elements per story height and six macrofibers, respectively. This Intensity Measures
model has been used successfully for nonlinear dynamic analysis
of RC fixed-base buildings (e.g., Zhang and Li 2017; Kolozvari Earthquake ground motion time histories of various levels of inten-
et al. 2018) and is presented in this study for the first time in sity were selected to evaluate the structural response of the hypo-
coupled SSI systems. The shear walls reduce the natural period thetical buildings. The ground motions, used subsequently for
of the structure as a result of adding a stiff lateral load-resisting the computation of EDPs, were selected following FEMA P-58
mechanism to the structure. Further characteristics of this model, (FEMA 2012). Ten ground motion records per hazard [72, 475,
shear wall design, and model parameters adopted herein were given and 2,475-year mean return periods (i.e., 50%, 10%, and 2% prob-
by Kolozvari et al. (2015b) and Kolozvari and Wallace (2016). abilities of exceedance in 50 years, respectively)] were selected

© ASCE 04020028-4 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


using the PEER ground motion database tool (PEER 2014) to computed based on the associated M, R, and ε for each hazard
match the conditional mean spectrum (CMS) proposed by Baker level. Table 2 presents the selected earthquakes from the PEER
(2011). The CMS relates the spectral amplitude at a given period ground motion database and its characteristics such as M, R, scale
associated to an earthquake magnitude (M), distance from the factor based on CMS, mechanism (i.e., type of faulting), PGA,
earthquake to the site of interest (R), and normalized residual from PGV, AI, HI, and predominant natural frequency for each return
a ground-motion model prediction (ε) of the pseudospectral accel- period. Based on the CMS, the PGAs were approximately 0.16,
eration (Sa). These can be obtained from a deaggregation analysis 0.28, and 0.38 g for the 72, 475, and 2,475-year mean return peri-
to predict ground motions at the site of interest, based on a prob- ods, respectively.
abilistic seismic hazard analysis (PSHA). Local effects associated For each ground motion, the intensity measures PGA, PGV, AI,
with the acceleration amplification of earthquakes with periods and HI were computed (Table 2) and correlated with the Sa at 4.0 s
larger than 2.0 s in the Los Angeles Basin (Graves and Aagaard of the CMS. The CMS Sa values at 4.0 s were 0.04, 0.16, and
2011) were not included. Fig. 3 presents the deaggregation analysis 0.31 g for the 72, 475, and 2,475-year mean return periods, respec-
for the Los Angeles downtown area showing the contribution of tively. Fig. 5 shows the relationship for the site-specific soil con-
active and potential sources to the seismic hazard. The analyses ditions considered for different IM levels. The figures show the
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

were developed using a Site class C based on ASCE 7.16 (2017) Pearson’s r correlation coefficient of the relationship between the
according to the V s soundings. The deaggregation analyses were IM and the Sa at 4.0 s of the CMS. This linear correlation coef-
performed using the USGS tool (2008). The results show that ficient measures the strength and the direction of a linear relation-
the 72-year mean return period has an approximate M ¼ 6.52 ship between the two variables. The worst fit in terms of intensity
and R ¼ 22.89 km. The 475-year mean return period is represented measures are found for the PGA. PGAs are less suitable to relate the
by M ¼ 6.74 and R ¼ 12.78 km. The 2,475-year mean return frequency content and duration of an earthquake and tend to under-
period has an approximate M ¼ 6.88 and R ¼ 9.51 km. Fig. 4 estimate the ground motion intensity. Accumulated measures such
shows the ground motions selected and scaled to match the com- as AI and HI are more representative of the earthquake intensity
puted target CMS at a period of 4.0 s. This period was chosen than are peak acceleration values of an earthquake because they
to approximately match the expected first-mode natural period contain the information of the entire time history (Housner 1975;
of the hypothetical 40-story structures studied herein. CMS was Newmark 1975).

(a) (b) (c)

Fig. 3. Probabilistic seismic hazard deaggregation analysis for downtown Los Angeles for different hazard levels: (a) 72 years; (b) 475 years; and
(c) 2,475 years.

(a) (b) (c)

Fig. 4. Conditional mean spectra at a given Sa (T ¼ 4.0 s) for 10 selected ground motions per return period: (a) 72 years; (b) 475 years; and
(c) 2,475 years.

© ASCE 04020028-5 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


Table 2. Selected horizontal earthquake acceleration time histories from PEER ground motion database
Return Scale factor
period NGA based on R PGA PGV AI HI Frequency V s30
(year) No. Earthquake Year record CMS Mw (km) Mechanism (g) (m=s) (m=s) (m) (Hz) (Period s) (m=s)
72 1 Denali Alaska 2002 2,115 1.33 7.9 126.3 Strike slip 0.1 0.13 0.48 0.34 12.9 (0.07) 376
2 Chi-Chi Taiwan 1999 1,560 2.8 7.62 63.8 Reverse oblique 0.12 0.18 0.87 1.01 2.89 (0.34) 463
3 Duzce Turkey 1999 1,613 1.87 7.14 25.7 Strike slip 0.1 0.11 0.15 0.31 5.73 (0.17) 782
4 Hector Mine 1999 1,776 2.05 7.13 56.4 Strike slip 0.14 0.15 0.47 0.63 3.33 (0.30) 359
5 Hector Mine 1999 1,836 3.39 7.13 42.0 Strike slip 0.22 0.24 0.8 0.69 6.92 (0.14) 635
6 Chalfant Valley-02 1986 549 1.03 6.19 14.3 Strike slip 0.26 0.2 0.53 0.77 3.94 (0.25) 303
7 San Fernando 1971 68 0.69 6.61 22.7 Reverse 0.16 0.15 0.32 0.55 5.39 (0.18) 316
8 Superstition Hills-01 1987 718 1.31 6.22 17.5 Strike slip 0.17 0.17 0.45 0.66 7.69 (0.13) 179
9 Loma Prieta 1989 802 0.49 6.93 7.5 Reverse oblique 0.25 0.2 0.35 0.87 6.13 (0.16) 380
10 Landers 1992 850 1.75 7.28 21.8 Strike slip 0.3 0.34 2.16 1.06 5.46 (0.18) 359
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

475 1 Kobe Japan 1995 1,100 2.08 6.9 24.8 Strike slip 0.46 0.44 2.51 1.32 2.55 (0.39) 256
2 Chi-Chi Taiwan 1999 1,196 2.43 7.62 41.9 Reverse oblique 0.13 0.51 1.23 0.85 2.89 (0.34) 210
3 Chi-Chi Taiwan 1999 1,237 2.94 7.62 58.4 Reverse oblique 0.23 0.3 2.6 1.9 2.25 (0.44) 180
4 Hector Mine 1999 1,762 1.62 7.13 41.8 Strike slip 0.29 0.45 2.27 1.48 7.82 (0.12) 382
5 Chi-Chi Taiwan-04 1999 2,746 3.31 6.2 33.0 Strike slip 0.17 0.36 1.06 1.49 4.2 (0.23) 253
6 Taiwan SMART1(45) 1986 570 2.25 7.3 56.0 Reverse 0.16 0.16 0.91 0.72 6.1 (0.16) 309
7 Denali Alaska 2002 2,113 2.37 7.9 53.0 Strike slip 0.13 0.27 1.15 0.57 12.2 (0.08) 382
8 Chi-Chi Taiwan-06 1999 3,313 3.74 6.3 58.8 Reverse 0.33 0.29 1.12 1.54 2.72 (0.36) 221
9 Taiwan SMART1(45) 1986 576 2.78 7.3 55.1 Reverse 0.29 0.27 1.95 0.88 6.1 (0.16) 327
10 Superstition Hills-02 1987 729 1.31 6.54 23.8 Strike slip 0.23 0.41 1.31 1.43 7.82 (0.12) 179
2,475 1 Chi-Chi Taiwan 1999 1,476 1.93 7.62 28.0 Reverse oblique 0.3 0.72 2.14 1.44 4.73 (0.21) 406
2 Chi-Chi Taiwan 1999 1,534 3.02 7.62 15.9 Reverse oblique 0.39 1.0 10.8 4.38 1.55 (0.64) 409
3 Chi-Chi Taiwan 1999 1,199 3.64 7.62 35.4 Reverse oblique 0.32 0.99 6.28 3.41 1.55 (0.64) 192
4 Imperial Valley-06 1979 171 1.96 6.53 0.07 Strike slip 0.62 1.4 3.33 3.53 1.46 (0.68) 264
5 Chi-Chi Taiwan-03 1999 2,507 3.45 6.2 24.4 Reverse 0.5 1.0 5.0 4.41 3.71 (0.26) 258
6 Superstition Hills-02 1987 729 2.57 6.54 23.8 Strike slip 0.46 0.81 5.03 2.8 7.82 (0.12) 179
7 Kobe Japan 1995 1,121 3.08 6.9 27.7 Strike slip 0.49 0.65 10.13 3.73 1.21 (0.82) 256
8 Kobe Japan 1995 1,114 1.77 6.9 3.31 Strike slip 0.62 1.6 5.53 6.0 0.84 (1.19) 198
9 Chi-Chi Taiwan 1999 1,541 2.08 7.62 12.3 Reverse oblique 0.39 0.88 6.77 2.93 3.28 (0.30) 493
10 Chi-Chi Taiwan 1999 1,496 3.43 7.62 10.4 Reverse oblique 0.54 1.5 10.4 3.53 1.99 (0.50) 403

(a) (b)

(c) (d)

Fig. 5. Relations between Sa at 4.0 s of the CMS and each IM: (a) PGA; (b) PGV; (c) AI; and (d) HI.

© ASCE 04020028-6 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


Natural Periods of Proposed Numerical Models participation of the first-mode shows that the computed period
lengthening is possible due to the small participation of higher
Fig. 6 presents elastic transfer functions computed to identify peri- modes, which is in agreement with Stewart et al.’s (1999) findings.
ods and modes of vibration of the four building models. These
transfer functions were determined as the ratio of the frequency
response of nodes on top of the structure (i.e., 40th story) to the Engineering Demand Parameters for Hypothetical
input time history frequency response. Frequency responses were Tall Buildings
computed by applying the fast Fourier transform (FFT) to the out-
put acceleration time histories. A total of 30 earthquakes were used For the aforementioned building models and selected ground mo-
to compute the transfer functions, and the average trendline results tions, the EDPs evaluated, to quantifying potential discrepancies on
are highlighted in the figure. the structural response, were the interstory drift ratio (IDR), peak
The hypothetical tall building modeled as a fixed-base structure story horizontal acceleration, and building settlement. Those EDPs
had a predominant natural period of 4.2 s (i.e., natural frequency ¼ were selected herein to describe the performance of the structure
0.24 Hz). This period was obtained from an eigenvalue analysis and study their importance when quantifying damage states and
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

performed in OpenSees and was confirmed by the transfer func- losses using the PBEE methodology. Fig. 7(a) shows means values
tions. Higher modes also can be identified with the transfer func- of computed maximum interstory drifts. The interstory drifts do not
tions; the second and third modes were 1.3 and 0.7 s, respectively represent postearthquake residual drifts. These curves represent the
[Fig. 6(a)]. When adding the soil domain, which consisted of pre- computed mean values of the 10 earthquakes for each hazard level.
dominantly competent alluvium deposits (Fig. 1), the building Fig. 7(b) shows mean values of computed peak story horizontal
slightly elongated its predominant natural period to approximately accelerations for the four building models and three hazard levels.
4.6 s (i.e., natural frequency ¼ 0.21 Hz). This represents a 10% The accelerations presented were obtained using absolute recorders
~
period lengthening (i.e., T=T ¼ 1.1). Higher modes had negligible from OpenSees. As expected, the 2,475-year mean return period
period lengthening. Fig. 6(b) shows the transfer function of the hazard produced the largest demands on the building models: maxi-
building when shear walls were added in its middle bay. This stiff- mum 1.75% and 1.2 g for computed mean interstory drifts and
ening effect caused a 14% reduction of the natural period of the mean peak story horizontal accelerations, respectively. Comparing
building (i.e., natural period decreased from 4.2 to 3.6 s). Higher fixed-base models only (i.e., ignoring SSI effects), significant re-
periods also were reduced. Adding the soil domain to the building duction of interstory drift distribution along building height was
with shear walls caused a 20% period lengthening, i.e., T=T ~ ¼ 1.2 computed as a result of adding the shear wall system. This stiffen-
[Fig. 6(d)]. Given the competent soil conditions of this study ing effect reduced 14% the natural period of the building. Negli-
(i.e., Dr > 85% for the fill, Qal1, and Qal2 soil layers), the numeri- gible changes in the interstory drifts over the building height for
cal output of the fixed-base building model and the corresponding the three hazard levels were computed when modeling explicitly
model adding the competent soil domain could lead to similar dy- the continuum soil domain. The inclusion of the continuum soil
namic responses. NIST (2012) stated that period lengthening for domain affected the computed building response, as reflected by
tall buildings can be neglected, but in this study, given the soil con- the increase in maximum peak story horizontal acceleration distri-
ditions and adopted structural configuration, period lengthening of bution along building height. The building model without shear
as much as 20% seems plausible. Stewart et al. (1999) stated that walls and considering SSI effects (Fig. 7, dashed-dotted line cor-
for high-rise structures with significant higher-mode response, responding to largest natural period of 4.6 s) caused the largest peak
period lengthening can be neglected. For the fixed-base tall build- story horizontal accelerations. Conversely, the fixed-base model
ing model analyzed herein, the modal mass participation ratios (Fig. 7, solid line) resulted in the smallest computed peak story
were computed as 74.6%, 14.3%, and 3.8% for the first, second, horizontal accelerations. The inclusion of SSI effects in the numeri-
and third vibration modes, respectively. The large modal mass cal modeling changed the distribution of the peak story horizontal

20 20
1st Mode = 4.2s 1st Mode = 3.6s
rd 3rd Mode = 0.5s
15 3 Mode = 0.7s 2nd Mode = 1.3s
Transfer Function (Roof/input history)

15

10 10 2nd Mode = 1.0s

5 5

(a) 0 (b) 0
25 0 1 2 3 4 5 6 25 0 1 2 3 4 5 6
∼ ∼
nd
2 Mode = 1.3s T/T = 1.1 nd
2 Mode = 1.1s T/T = 1.2
20 20
rd 1st Mode = 4.6s
15 3 Mode = 0.7s 15
3rd Mode = 0.5s 1st Mode = 4.3s
10 10
5 5
0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6
(c) Period (s) (d) Period (s)

Fig. 6. Transfer functions for: (a) fixed-base building; (b) fixed-base building with shear walls; (c) building including SSI effects; and (d) building
with shear walls including SSI effects.

© ASCE 04020028-7 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


(a)
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

(b)

Fig. 7. Computed seismic EDPs along the height for each building model and hazard level: (a) envelope of maximum interstory drifts ; and (b) peak
story horizontal accelerations.

(a) (b) (c)

Fig. 8. EDPs at Sa (T ¼ 4.0 s) of the CMS: (a) interstory drifts; (b) peak horizontal acceleration; and (c) settlements.

accelerations and the interstory drifts along the building height. or II according to ASCE 7.16 Table 12.12.1 (ASCE 2017). The
This was more noticeable for the models considering SSI effects 2,475-year mean return period hazard levels may lead to unaccept-
in the 2,475-year hazard level. able performance of this hypothetical structure in terms of inter-
Fig. 8 shows the maximum computed EDPs among all the sto- story drift ratios, which warrants the addition of the shear wall
ries of the building models for each set of ground motions grouped system to reduce the natural period of the building and thus reduce
in terms of Sa at T ¼ 4.0 s of the CMS. These relations are rep- the interstory drift ratios [Fig. 8(a), dashed and dotted lines below
resented for each building model along with a power-law functional solid and dashed-dotted lines for interstory drift]. For the build-
fit to the data. The parameters of the functional fit [i.e., EDP ¼ ing models considered herein, the maximum computed interstory
aðSaÞb ], a and b, were determined using least-squares analysis. drift levels barely changed by considering SSI effects; however,
The largest computed interstory drift ratio among the set of earth- noticeable changes in their distribution along the building height
quakes was 2.5%. For comparison purposes, maximum allowable were computed [Fig. 7(a)]. The peak seismic-induced horizontal
interstory drifts are 2.0% of the story height for a Risk Category I accelerations increased when SSI effects were considered in the

© ASCE 04020028-8 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


Table 3. Parameters for power regression equations relating EDPs with pseudoacceleration at 4.0 s of CMS for each tall building model
Interstory drift Peak horizontal acceleration Settlement
Model type a b R2 σln x a b R2 σln x a b R2 σln x
Fixed-base model 7.29 1.07 0.77 0.86 3.11 1.11 0.67 0.72 N/A N/A
Model including SSI effects 7.26 1.09 0.72 0.79 3.12 0.85 0.67 0.62 0.75 1.42 0.70 1.04
Fixed-base shear walls model 5.38 1.03 0.81 0.76 2.25 0.75 0.73 0.59 N/A N/A
Model with shear walls including SSI effects 4.59 0.92 0.84 0.67 3.75 0.83 0.68 0.6 0.33 1.01 0.65 1.02

numerical model. The addition of a stiff shear wall system to the The coefficient of correlation, R2 , for each EDP-hazard level rela-
building models increased their peak horizontal acceleration re- tionship also is presented in the Table. To describe the dispersion of
sponse compared with those of the building models without shear the variables, the lognormal standard deviation is presented in the
walls. Table. These equations are presented to describe the variation of
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

A geotechnical EDP that is related to possible damages on the EDPs with the Sa of the CMS for each hazard level. Parameter
building is the maximum vertical displacement (i.e., settlement) a represents a scaling factor of the fitting curve, which shifts ver-
computed at the base of the mat foundation. Fig. 8(c) presents the tically as a increases. Parameter b determines the rate of growth or
variation of maximum settlement computed for each building decay of the function. The larger the flexibility of the building, re-
model including SSI effects as a function of Sa (T ¼ 4.0 s) of the flected in large computed natural vibration periods, generally the
CMS. Maximum settlements of 0.24 m were computed for the larger was the a value.
2,475-year mean return period condition. The addition of a shear
wall system (which reduced the building’s natural period) reduced
the building settlements to 0.18 m (i.e., 25% reduction). This out- Parametric Variation of Soil Profile
come is applicable to the soil and structural conditions considered
in this paper, because seismic-induced settlements depend on nu- To understand the role that the strength characteristics of the soil
merous factors, including presence of liquefiable soils, input mo- domain plays in EDP-IM relationships at the Sa (T ¼ 4.0 s) of the
tion characteristics, in situ conditions and stress history of the soil CMS, a variation of soil parameters was performed to modify the
deposit, and configuration of the soil–foundation–structural system. natural period of the soil profile. Fig. 9 shows the parametric varia-
The soils studied herein were not susceptible to liquefaction (Fig. 1), tion of shear wave velocities and friction angles used to further
as verified by computed excess pore-water pressure ratios less than 1 study the influence of soil properties in SSI effects on the building
for all building models. Several authors, such as Yang et al. (2003), response; V s and ϕ were parametrically varied by 100 m=s and 5°,
Howell et al. (2015), and Karimi and Dashti (2016), have shown that respectively, from the values adopted in the previous section. This
the PDMY02 model tends to underpredict seismic-induced settle- variation yielded three soil profiles in terms of the predominant soil
ments due to underestimation of the coefficient of the volumetric natural period: (1) Soil type C with T ¼ 0.3 s, (2) Soil type C with
compressibility. T ¼ 0.4 s (i.e., soil profile adopted in the remaining sections of the
Table 3 presents both parameters for the power-fit equations, a paper), and (3) Soil type D with T ¼ 0.6 s. These natural periods
and b, which were obtained using power-fit regression analyses. were calculated according to the V s measured for the topmost 40 m
of the soil profile, which is consistent with the maximum depth
considered in numerical simulations. Constitutive soil parameters
were recalculated for the cases of predominant soil natural period
T ¼ 0.3 s and 0.6 s and are presented in Table 4.
The tall building models without shear walls were used to quan-
tify SSI in IM-EDP relationships given the assumed parametric
variation of the soil profile. Fig. 10 shows for each hazard level
the computed EDPs for the Sa of the CMS at 4.0 s. The stiffer
the soil, i.e., small natural period of the soil profile (T ¼ 0.3 s

Table 4. PDMY02 constitutive soil parameters recomputed for


parametrically modified soil conditions
Soil Parameter Fill Qal1 Qal2 Siltstone
Soil type C, V s (m=s) 328.6 380.1 496.2 587.6
T ¼ 0.3 s Dr (%) 84 98 96 99
Gmax;1;oct (MPa) 256.3 351.5 623.2 853.5
Br (MPa) 560.1 606.2 1,240.0 1,780.1
ϕTXC (degrees) 34.6 40.8 36.9 35.8
ϕPT (degrees) 29.6 35.84 31.9 30.8
Soil type D, V s (m=s) 128.6 180.1 296.2 387.6
T ¼ 0.6 s Dr (%) 70 85 82 89
Gmax;1;oct (MPa) 39.2 78.9 222.1 371.5
Fig. 9. Parametric variation of shear wave velocities and friction angles Br (MPa) 137.1 217.6 705.5 1,238.2
to study role of soil domain on EDP-IM relationships at Sa (T ¼ 4.0 s) ϕTXC (degrees) 24.4 29.2 26.3 25.3
of the CMS. ϕPT (degrees) 19.4 24.2 21.3 20.3

© ASCE 04020028-9 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


(a) (b) (c)

Fig. 10. EDPs at Sa (T ¼ 4.0 s) of the CMS arising from changes in soil profile from Class C to Class D: (a) interstory drifts; (b) peak horizontal
acceleration; and (c) settlements.
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

in the figure), the larger were the computed interstory drift ratios compared with and adjusted from similar studies of earthquake-
and peak horizontal accelerations (Fig. 10, dashed lines above induced losses in tall buildings, such as Ramirez and Miranda (2009)
dashed-dotted lines). Smaller interstory drifts and peak story hori- and Molina Hutt et al. (2016). The selected fragility functions were
zontal accelerations were computed by changing the soil type from grouped depending on the EDPs used to define damage. Interstory
C to D. Maximum interstory drift ratios of 3.1% were calculated drifts and peak story horizontal accelerations were selected as the
for the soil with T ¼ 0.3 s, which represents an 8% increase with main engineering input demands in the PBEE evaluation conducted
respect to those in Fig. 8. In this parametric study, by lowering on the four tall building models. For the model including the shear
shear wave velocities (i.e., increasing natural vibration period of wall system, shear wall rotation also was considered as an EDP. The
the soil profile), large degradation of the shear modulus and large curvature of the wall was obtained by linearizing the strain between
mobilized shear strains in the soil caused by the earthquake were the top and bottom of the element. The curvature was assumed to be
expected. This implies large hysteretic damping in the soil domain constant over the height of the wall; therefore the wall rotation was
due to the large shear strains (Seed and Idriss 1970), and thus, computed as the curvature multiplied by the height. Large values
smaller acceleration of the ground motion through the structure and of peak story horizontal accelerations tend to affect nonstructural
reduction of seismic structural demands. The large damping asso- components (e.g., suspended ceilings and pendant lighting) more
ciated with a flexible foundation soil (i.e., Soil type D in this case) than structural components, which are negatively impacted by large
causes a reduction of the computed seismic demands in the building values of interstory drifts (e.g., exterior walls and structural steel con-
(Fig. 10, dashed-dotted line). Slight differences were computed in nections). Computed losses related directly to seismic-induced set-
the peak story horizontal accelerations for the soil conditions cor- tlements were not included in the analyses. Foundation deformations
responding to T ¼ 0.3 and 0.4 s. Increasing the flexibility of the represent a source of losses that is not considered in the typical PBEE
soil, reflected in larger natural period [Fig. 10(c)], caused maxi- loss analysis. Larger losses than those computed herein are expected
mum computed settlements up to 0.5 m for the 2,475-year mean because settlements can affect structural and nonstructural building
return period condition, which implies an increase of 150% with components.
respect to the values in Fig. 8. Excessive differential settlements The loss analyses were performed using PACT version 3.1.1
leading to large interstory drifts can trigger architectural, func- (ATC 2012), which performs Monte Carlo simulations to calculate
tional, and structural damage of structural and nonstructural com- losses. PACT can be used to account for various uncertainties in
ponents of tall buildings, leading to large seismic-induced losses loss estimations using Monte Carlo simulation procedure. In this
that affect the PBEE decision-making framework. Because typical study, 2,000 realizations were required to reach stable loss assess-
downtown Los Angeles area soil profiles period are closer to 0.4 s ment results. For each realization, fragility functions and EDPs
(for the topmost 40 m), this was adopted in the remaining sections were used to compute damage states. PACT uses simulation results,
of the paper. demand models, and component damage fragilities to determine
damage states for each structural and nonstructural building com-
ponent. Subsequently, PACT assigns quantities necessary to re-
Loss Analyses for Tall Building Models mediate damage in terms of costs and mobilization times using
consequence functions. Repair costs for building components at
The PBEE methodology evaluates at each story level the response each hazard level are summed for each realization to obtain the cu-
of structural and nonstructural components susceptible to seismic mulative distribution function of losses. Repair time can be com-
damage. These structural and nonstructural components have an puted using the assumption that the floors are repaired in series or
associated fragility function, defined as the probability of exceed- parallel.
ing a particular damage stage at a certain level of the EDP. FEMA In addition to information related to EDPs and fragility func-
(2012) compiled fragility functions and consequence data based on tions of building components, input parameters such as estimated
occupancy models, including multiunit residential, commercial, hos- total replacement cost and time of the building, typical occupancy,
pitality, and education (FEMA 2012). Selected structural and non- floor area, and story height need to be input to perform loss analy-
structural components considered for the tall building models studied ses. The total replacement cost of the building was assumed as
herein along with their quantities are given in Table 5. Structural and $130 million, based on $3,550=m2 ($330=ft2 ) replacement cost
nonstructural components and quantities were selected based on case typical for tall buildings with floor area of approximately 900 m2.
histories and typical values from buildings with similar use and oc- The total replacement time was estimated as 730 days: 2 weeks=
cupancy reported in FEMA P-58 (FEMA 2012). Quantities were story plus 200 days of demolition and redesign (Molina Hutt 2017).

© ASCE 04020028-10 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


Table 5. Selected fragility functions and quantities related to engineering demand parameters
Quantity per direction
Fragility number Component description per FEMA P-58 Floor level Units L T ND EDP
B1031.001 Bolted shear tab gravity connections All EA 40 40 — IDR (rad)
B1031.001a Bolted shear tab gravity connections All EA 56 56 — IDR (rad)
B1031.011c Steel column base plates, column W > 300 plf First EA 12 12 — IDR (rad)
B1035.021b Post-Northridge connection beam depth ≤ W27 All EA 4 4 — IDR (rad)
B1035.031b Post-Northridge connection beam depth ≤ W27 All EA 8 8 — IDR (rad)
B2022.036 Midrise stick-built curtain wall All 32 ft2 64.55 64.55 — IDR (rad)
C1011.001b Wall partition All 100 linear ft 3.87 3.87 — IDR (rad)
C1011.001c Wall partition, full height All 100 linear ft 0.97 0.97 — IDR (rad)
C2011.001f Prefabricated steel stair that accommodate drift All EA 1 — — IDR (rad)
C3011.001b Wall finishes All 100 linear ft 0.37 0.37 — IDR (rad)
B1031.021b Welded column splices, column 150 plf < W < 300 plf c
EA 12 12 — IDR (rad)
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

B1044.102aa Slender concrete wall, 18 in. thick, 12 ft high, 20 ft long First 7 floors 144 ft2 13.45 13.45 — PHR
C3027.002 Raised access floor, seismically rated All 100 ft2 — — 7.20 PFA
C3032.003a Suspended ceiling (Ip ¼ 1.0), A < 250 ft2 All 250 ft2 — — 8.71 PFA
C3032.003b Suspended ceiling (Ip ¼ 1.0), 250 < A < 1,000 ft2 All 600 ft2 — — 3.63 PFA
C3032.003c Suspended ceiling (Ip ¼ 1.0), 1,000 < A < 2,500 ft2 All 1,800 ft2 — — 2.42 PFA
D1014.011 Traction elevator First EA — — 11.00 PFA
D2021.023a Cold or hot potable water piping, D > 2.5 in: All 1,000 linear ft — — 0.15 PFA
D2021.023b Cold or hot potable water piping, D < 2.5 in: All 1,000 linear ft — — 0.15 PFA
D2022.013a Heating hot water piping, D < 2.5 in: All 1,000 linear ft — — 0.81 PFA
D2022.013b Heating hot water piping, D < 2.5 in: All 1,000 linear ft — — 0.81 PFA
D2022.023a Heating hot water piping, D > 2.5 in: All 1,000 linear ft — — 0.29 PFA
D2022.023b Heating hot water piping, D > 2.5 in: All 1,000 linear ft — — 0.29 PFA
D2031.013b Sanitary waste piping All 1,000 linear ft — — 0.55 PFA
D3031.013o Chiller capacity: 350 to <750 t First 500 t — — 1.00 PFA
D3031.023o Cooling tower capacity: 350 to <750 t First 500 t — — 1.00 PFA
D3041.011c HVAC galvanized sheet metal ducting, A < 6 ft2 All 1,000 linear ft — — 0.73 PFA
D3041.012c HVAC galvanized sheet metal ducting, A > 6 ft2 All 1,000 linear ft — — 0.19 PFA
D3052.013o Air handling unit capacity: 25,000 to <40,000 ft3 =min First 30,000 ft3 — — 5.00 PFA
D4011.023a Fire sprinkler water piping All 1,000 linear ft — — 1.94 PFA
D4011.053a Fire sprinkler drop standard threaded steel All 100 — — 0.87 PFA
D5012.013o Motor control center First EA — — 8.00 PFA
D5012.023n Low voltage switchgear capacity: 100–350 A All 225 A — — 2.00 PFA
D3031.013n Chiller capacity: 350 to <750 t Roof 500 t — — 1.00 PFA
D3031.023n Cooling tower capacity: 350 to <750 t Roof 500 t — — 1.00 PFA
D3052.013n Air handling unit capacity: 25,000 to <40,000 ft3 =min Roof 30,000 ft3 — — 5.00 PFA
D5012.013n Motor control center Roof EA — — 8.00 PFA

Note: EA = each; L = longitudinal; T = transverse; ND = nondirectional; PFA = peak floor acceleration (g); plf = pound per linear foot; and PHR = plastic hinge
rotation. 1 ft = 0.3048 m; 1 in. = 0.0254 m; and 1 lb = 0.45 kg.
a
These fragility functions were updated and included in the model with shear walls.
b
These fragility functions were not included in the model with shear walls.
c
This fragility function was assigned to the following floors: 4th, 7th, 10th, 13th, 16th, 19th, 22th, 25th, 28th, 31th, 34th, and 37th.

The population type for the 40-story buildings was assumed as not vary significantly for the 72-year hazard level. Less than 0.3%
commercial office, and the quantities for the fragility functions in direct losses were computed among the building models for 50th
were determined at the 50th percentile level. For each hazard level percentile cumulative distribution function. The same analyses
and building model, economic losses were computed by integrating yielded 50th percentile losses up to 0.6% of the total building
repair losses over all components at each story of the building replacement costs for the 475-year hazard level when SSI effects
(FEMA 2012). Large postearthquake residual drifts were not com- were included in the tall building model. The values in Fig. 11
puted due to the predominantly elastic structural behavior of the tall slightly underestimate losses in relation to other reported fixed-base
building models. Thus, collapse scenarios and mechanisms are not structural models that include material and geometric nonlinearities
discussed. Cost of engineering designs, demolition, and other field in the structural behavior (e.g., Jayaram et al. 2012; Molina Hutt
activities prior to construction were not included in the computation et al. 2016; Kolozvari et al. 2018). Large differences in losses were
of building replacement cost because this paper focuses only on computed for the 2,475-year hazard level. Maximum median losses
estimating differences in direct losses from the proposed numerical computed for building models subjected to 2,475-year mean return
tall building models including soil–structure interaction effects on period earthquakes were 5.4% of the total replacement cost of the
their seismic response. building. By including SSI effects, the median direct economic
Fig. 11 shows cumulative distribution functions of the direct losses increased up to 33% in relation to building losses ignoring
losses for each tall building model and seismic hazard level. Those SSI with shear walls effects for the 2,475-year mean return period.
direct losses were calculated based on the fragility functions in For a given probability level in Fig. 11(c), the largest losses were
Table 5. Indirect losses associated with repair time were not explic- computed for the building models without shear walls because
itly accounted for in the analyses. The computed direct losses did those buildings had the largest engineering demands for such a

© ASCE 04020028-11 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


(a) (b) (c)

Fig. 11. Cumulative distribution functions of earthquake-induced losses for each tall building model and seismic hazard level: (a) 72-year mean
return period; (b) 475-year mean return period; and (c) 2,475-year mean return period.
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

separating structural (i.e., interstory drift and plastic rotation-


sensitive) and nonstructural (i.e., acceleration-sensitive) building
components. The results computed for the 72-year mean return
period showed that the main contribution to the median losses arises
from damage of nonstructural components. This is because very
small interstory drift ratios (up to 0.6%) were computed for this
hazard level. For the 2,475-year mean return period, structural com-
ponents dominated the computed median losses regardless of the
adopted tall building model. The inclusion of a shear wall lateral
load-resisting system, modeled using the SFI-MVLEM, reduced
the computed total repair losses (compare Bar graphs 1 and 3 with
Bar graphs 2 and 4 for the 2,475-year mean return period). Further-
more, inclusion of a shear wall system to the tall building consid-
erably reduced the structural contribution to the losses (observe
Fig. 12. Median total repair direct losses for each tall building model the size reduction of the shaded regions of Models 1 and 2 in relation
and seismic hazard level. to Models 3 and 4 in Fig. 12) but increased the nonstructural con-
tribution (observe the increase in size of the unshaded regions of
Models 1 and 2 in relation to Models 3 and 4). Considering SSI ef-
large earthquake [Fig. 8(a), interstory drifts from CMS Sa fects in the tall building models triggered larger median total repair
(T ¼ 4 .0 s) equal to 0.31 g]. Regardless of the seismic hazard level, losses (compare Bar graphs 1–2 and 3–4 for any return period).
including SSI effects in the analyses and the lack of a lateral load- Fig. 13 shows cumulative distribution functions of repair time
resisting shear wall system triggered the largest earthquake-induced for each tall building model and seismic hazard level. Building re-
losses. Overall, the smallest losses were computed for the tall build- pair time represents the necessary time to achieve building reoccu-
ings with lateral load-resisting shear wall systems modeled using a pancy and full operational level of the building after an earthquake.
fixed-base approach. In practical application, this conventional mod- The repair times were computed assuming that floors are repaired
eling approach can lead to unconservative designs or unrealistic pre- in series. This assumption is conservative (i.e., it tends to overesti-
dictions of earthquake-induced direct losses. These findings are mate repair time), and the results are presented to comparatively
consistent with the computed transfer functions and inherent flexi- evaluate repair times computed with each tall building model rather
bility associated with the natural periods of each building model. than to provide accurate predictions of repair time. The maximum
The contribution of structural and nonstructural components to median (i.e., 50th percentile) computed repair time for the 72-year
the seismic-induced losses is presented in Fig. 12 for each tall build- mean return period was 27 days for the tall building model with a
ing model and earthquake hazard level. This was accomplished by shear wall system and considering SSI effects. The 475-year mean

(a) (b) (c)

Fig. 13. Cumulative distribution functions for repair time for each tall building model and seismic hazard levels: (a) 72-year mean return period;
(b) 475-year mean return period; and (c) 2,475-year mean return period.

© ASCE 04020028-12 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


return period earthquake caused tall building repair times of as computed seismic demands of the tall building without shear
much as 45 days, for the tall building models considering SSI ef- walls. The smaller the natural period of the soil, the larger were
fects. Computed median recovery time for the 2,475-year mean re- the computed interstory drift ratios. The results also showed that
turn period earthquake level was 340 days for the building with no the larger the natural period of the soil, the smaller were the
shear walls. In terms of direct losses and repair time, it was shown computed peak story horizontal accelerations. Increasing the
that neglecting the effects of SSI and the lateral load-resisting sys- natural vibration period of the soil caused large damping and
tem in the tall building models yields a response that is not degradation of the shear modulus (i.e., large mobilized shear
representative of the actual seismic behavior. In Fig. 13, the strains in the soil caused by the earthquake). As expected, large
fixed-base tall building model (i.e., solid line) shifts from being earthquake-induced settlements were computed for flexible
the model with the shortest predicted repair times to the model with soils (i.e., those with large natural periods). These findings also
the largest estimated repair times as the seismic hazard level in- can be expected with soils that can experience similar earth-
creases. In general, the inclusion of SSI in the numerical modeling quake performance and period lengthening during earthquakes.
provides larger and perhaps more-realistic estimates of direct losses 5. Regardless of the seismic hazard level, including SSI effects in
and repair time. the analyses and the lack of a lateral load-resisting shear wall
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

system triggered the largest earthquake-induced losses. In gen-


eral, the smallest losses were computed for the tall buildings
Summary and Conclusions with a lateral load-resisting shear wall system modeled using
a fixed-base approach. This modeling approach of tall buildings
This paper presented the results of numerical simulations of could lead to unrealistic predictions of earthquake-induced di-
hypothetical tall building models used to determine the earthquake- rect losses. For the specific conditions assumed herein, com-
induced losses associated with three engineering demand parame- puted median direct economic losses for the 2,475-year mean
ters: interstory drift ratios, peak story horizontal accelerations, and return period increased as much as 33% when considering
shear wall rotations. The reduction of the natural period of the SSI effects in the numerical analyses in relation to building
coupled soil–structure system by adding a lateral load-resisting losses ignoring those effects. Academics and practitioners are
shear wall also was studied in this paper. A parametric variation encouraged to include SSI effects when analyzing seismic per-
of the soil profile was performed to assess its contribution to the formance and losses of tall building in high seismicity areas.
seismic response of the tall buildings. The numerical simulations These findings are consistent with the computed transfer func-
considered three seismic hazard levels and were performed using tions and natural periods of each building model.
OpenSees and the PDMY02 constitutive model for foundation soils 6. Interstory drift- and plastic rotation-sensitive structural compo-
and the SFI-MVLEM model for the shear wall system. From this nents dominated the computed median losses regardless of the
study, the following conclusions can be drawn: adopted building model for strong earthquakes. Small-intensity
1. Computed transfer functions from acceleration output and input earthquakes produced damage mostly to nonstructural compo-
time histories were used in this study to identify the natural per- nents as computed with the median losses. By adding a shear
iods and vibration modes for the tall building models. The in- wall system to the tall building models, interstory drifts de-
clusion of the soil domain in a tall building model changed its creased but peak horizontal accelerations increased. This com-
predominant natural period from 4.2 to 4.6 s. First-mode natural bined effect caused a general decrease of total direct losses, but
period lengthening was obtained for all models due to the in- an increase of nonstructural damage in relation to tall buildings
clusion of SSI effects. Higher vibration modes had negligible without a lateral load-resisting shear wall system. In general, the
impact on the computed period lengthening. Inclusion of the addition of SSI effects to the tall building models triggered lar-
shear walls decreased by 14% the natural period of the building ger losses compared with other fixed-base models. The authors
(i.e., natural period reduction from 4.2 to 3.6 s). recommend considering SSI effects in tall building models to
2. Amplification of peak story horizontal accelerations was com- avoid underestimation of earthquake-induced direct losses.
puted for the tall building models by considering SSI effects. This
was obtained regardless of the seismic hazard level. The com-
puted interstory drift levels slightly decreased by considering SSI Data Availability Statement
effects in the building models. Maximum seismic-induced settle-
ments of approximately 0.18 m were computed for the building Some or all data, models, or code generated or used during the
with stiff lateral load-resisting shear wall system. Considering SSI study are available from the corresponding author by request.
effects on the seismic response of tall buildings is desirable to
avoid unrealistic predictions of earthquake-induced EDPs.
3. Considering a stiff lateral load-resisting shear wall system in the Acknowledgments
building models increased the acceleration response and re-
Financial support for this work was provided by National Science
duced significantly the interstory drift ratios in relation to the
Foundation Grant No. CMMI-1563428. The support of Dr. Joy
building models without shear walls. Tall building models with
Pauschke and Dr. Richard Fragaszy, program directors at the
shear walls reduced the building settlements by approximately
National Science Foundation, is greatly appreciated. The authors
25%. Current practice tends to favor the inclusion of core shear
acknowledge Ahmadreza (Reza) Mortezaie and Sharid K. Amiri
walls to stiffen the lateral loading response behavior of tall
at Caltrans for their valuable comments and encouragement to ac-
buildings. Thus, modeling shear walls leads to a reduction of
complish the goals of this paper.
the seismic-induced interstory drift ratios and settlements. How-
ever, for those buildings with a large number of acceleration-
sensitive components, shear walls should be used and analyzed References
carefully.
4. Parametric variations of the initial site natural period of the AMEC. 2013. Final geotechnical data report. Regional connector transit
downtown Los Angeles soil profile changed significantly the corridor project. Contract No. E0119. Los Angeles: Metro.

© ASCE 04020028-13 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028


ASCE. 2017. Minimum design loads and associated criteria for buildings Kolozvari, K., V. Terzic, R. Miller, and D. Saldana. 2018. “Assessment of
and other structures. ASCE 7.16. Reston, VA: ASCE. dynamic behavior and seismic performance of a high-rise rc coupled
ATC (Applied Technology Council). 2012. Performance assessment com- wall building.” Eng. Struct. 176 (Apr): 606–620. https://doi.org/10
putation tool (PACT). Redwood City, CA: ATC. .1016/j.engstruct.2018.08.100.
Baker, J. W. 2011. “Conditional mean spectrum: Tool for ground-motion Kolozvari, K., T. A. Tran, K. Orakcal, and J. W. Wallace. 2015c. “Modeling
selection.” J. Struct. Eng. 137 (3): 322–331. https://doi.org/10.1061 of cyclic shear-flexure interaction in reinforced concrete structural
/(ASCE)ST.1943-541X.0000215. walls. II: Experimental validation.” J. Struct. Eng. 141 (5): 04014136.
Biot, M. A. 1962. “Generalized theory of acoustic propagation in porous https://doi.org/10.1061/(ASCE)ST.1943-541X.0001083.
dissipative media.” J. Acoust. Soc. Am. 34 (9A): 1254–1264. https://doi Kolozvari, K., and J. W. Wallace. 2016. “Practical nonlinear modeling of re-
.org/10.1121/1.1918315. inforced concrete structural walls.” J. Struct. Eng. 142 (12): G4016001.
Chang, G. A., and J. B. Mander. 1994. Seismic energy based fatigue dam- https://doi.org/10.1061/(ASCE)ST.1943-541X.0001492.
age analysis of bridge columns. Part 1: Evaluation of seismic capacity. Lamar, D. L. 1970. Geology of the Elysian Park-Repetto Hills area, Los
NCEER Technical Rep. No. NCEER-94-006. New York: National Angeles County, California. Special Rep. No. 101, Map Scale 1:24,000.
Center for Earthquake Engineering Research. Sacramento, CA: California Division of Mines and Geology.
Dibblee, T. W., and H. E. Ehrenspeck. 1991. Geologic map of the Holly- Lysmer, J., and R. L. Kuhlemeyer. 1969. “Finite dynamic model for infinite
Downloaded from ascelibrary.org by Columbia University on 03/11/20. Copyright ASCE. For personal use only; all rights reserved.

wood and Burbank (south 1/2) quadrangles, Los Angeles, California. media.” Eng. Mech. Div. 95 (1882): 167–188.
Map Scale: 1:24,000. Los Angeles: Dibblee Geological Foundation. McKenna, F., G. L. Fenves, M. H. Scott, and B. Jeremic. 2000. Open sys-
Elgamal, A., Z. Yang, and E. Parra. 2002. “Computational modeling of tem for earthquake engineering simulation (OpenSees). Berkeley, CA:
cyclic mobility and post-liquefaction site response.” Soil Dyn. Earth- Pacific Earthquake Engineering Research Center, Univ. of California,
quake Eng. 22 (4): 259–271. https://doi.org/10.1016/S0267-7261(02) Berkeley.
00022-2. Mercado, J. A., L. G. Arboleda-Monsalve, and V. Terzic. 2019. “Seismic
FEMA (Federal Emergency Management Agency). 2012. Seismic perfor- soil-structure interaction response of tall buildings.” In Proc., Geo-
mance assessment of buildings. FEMA P-58. Washington, DC: FEMA. Congress 2019, 118–128. Reston, VA: ASCE. https://doi.org/10.1061
Filippou, F. C., E. P. Popov, and V. V. Bertero. 1983. Effects of bond /9780784482100.013.
deterioration on hysteretic behaviour of reinforced concrete joints. Re- Meyerhof, G. 1957. “Discussion on research on determining the density of
port to the National Science Foundation. Richmond, CA: Earthquake sands by penetration testing.” In Vol. 1 of Proc., 4th Int. Conf. on Soil
Engineering Research Center. Mechanics and Foundation Engineering, 110. London: Butterworths
Scientific.
Graves, R. W., and B. T. Aagaard. 2011. “Testing long-period ground-
Miranda, E., and H. Aslani. 2003. Probabilistic response assessment for
motion simulations of scenario earthquakes using the Mw 7.2 El
building-specific loss estimation. PEER Rep. No. 2003/03. Berkeley,
Mayor-Cucapah mainshock: Evaluation of finite-fault rupture charac-
CA: Pacific Earthquake Engineering Research Center.
terization and 3D seismic velocity models.” Bull. Seismol. Soc. Am.
Molina Hutt, C. 2017. Risk-based seismic performance assessment of
101 (2): 895–907. https://doi.org/10.1785/0120100233.
existing tall steel framed buildings. London: Univ. College London.
Hatanaka, M., and A. Uchida. 1996. “Empirical correlation between pen-
Molina Hutt, C., I. Almufti, M. Willford, and G. Deierlein. 2016. “Seismic
etration resistance and internal friction angle of sandy soils.” Soils
loss and downtime assessment of existing tall steel-framed buildings
Found. 36 (4): 1–9. https://doi.org/10.3208/sandf.36.4_1.
and strategies for increased resilience.” J. Struct. Eng. 142 (8):
Housner, G. W. 1975. “Measures of severity of earthquake ground shak-
C4015005. https://doi.org/10.1061/(ASCE)ST.1943-541X.0001314.
ing.” In Proc., 1st US National Conf. on Earthquake Engineering. Oak- Newmark, N. 1975. “Seismic design criteria for structures and facilities
land, CA: Earthquake Engineering Research Institute. trans-Alaska pipeline system.” In Proc., 1st US National Conf. on
Howell, R., E. M. Rathje, and R. W. Boulanger. 2015. “Evaluation of sim- Earthquake Engineering. Oakland, CA: Earthquake Engineering Re-
ulation models of lateral spread sites treated with prefabricated vertical search Institute.
drains.” J. Geotech. Geoenviron. Eng. 141 (1): 04014076. https://doi NIST. 2012. Soil-structure interaction for building structures. Rep. No.
.org/10.1061/(ASCE)GT.1943-5606.0001185. NIST/GCR 12-917-21. Gaithersburg, MD: NIST.
Jayaram, N., N. Shome, and M. Rahnama. 2012. “Development of earth- PEER (Pacific Earthquake Engineering Research). 2014. “PEER ground
quake vulnerability functions for tall buildings.” Earthquake Eng. motion database.” Accessed February 20, 2019. https://ngawest2
Struct. Dyn. 41 (11): 1495–1514. https://doi.org/10.1002/eqe.2231. .berkeley.edu.
Karimi, Z., and S. Dashti. 2015. “Numerical and centrifuge modeling of Ramirez, C. M., and E. Miranda. 2009. Building-specific loss estimation
seismic soil–foundation–structure interaction on liquefiable ground.” methods & tools for simplified performance-based earthquake engi-
J. Geotech. Geoenviron. Eng. 142 (1): 04015061. https://doi.org/10 neering. Rep. No 171. Stanford, CA: John A. Blume Earthquake En-
.1061/(ASCE)GT.1943-5606.0001346. gineering Center.
Karimi, Z., and S. Dashti. 2016. “Seismic performance of shallow founded Seed, H. B., and I. M. Idriss. 1970. Soil moduli and damping factors
structures on liquefiable ground: Validation of numerical simulations us- for dynamic response analyses. Technical Rep. No. EERRC-70-10.
ing centrifuge experiments.” J. Geotech. Geoenviron. Eng. 142 (6): Berkeley, CA: Univ. of California, Berkeley.
04016011. https://doi.org/10.1061/(ASCE)GT.1943-5606.0001479. Stewart, J. P., R. B. Seed, and G. L. Fenves. 1999. “Seismic soil-structure
Karimi, Z., S. Dashti, Z. Bullock, K. Porter, and A. Liel. 2018. “Key pre- interaction in buildings. II: Empirical findings.” J. Geotech. Geoenviron.
dictors of structure settlement on liquefiable ground: A numerical para- Eng. 125 (1): 38–48. https://doi.org/10.1061/(ASCE)1090-0241(1999)
metric study.” Soil Dyn. Earthquake Eng. 113 (Mar): 286–308. https:// 125:1(38).
doi.org/10.1016/j.soildyn.2018.03.001. USGS. 2008. 2008 interactive deaggregation, v3.3.1. Washington, DC:
Khosravifar, A., R. W. Boulanger, and S. K. Kunnath. 2014. “Effects of USGS.
liquefaction on inelastic demands on extended pile shafts.” Earthquake Yang, Z., A. Elgamal, and E. Parra. 2003. “Computational model for cyclic
Spectra 30 (4): 1749–1773. https://doi.org/10.1193/032412EQS105M. mobility and associated shear deformation.” J. Geotech. Geoenviron.
Kolozvari, K., K. Orakcal, and J. W. Wallace. 2015a. “Modeling of cyclic Eng. 129 (12): 1119–1127. https://doi.org/10.1061/(ASCE)1090
shear-flexure interaction in reinforced concrete structural walls. I: -0241(2003)129:12(1119).
Theory.” J. Struct. Eng. 141 (5). https://doi.org/https://doi.org/10 Yang, Z., J. Lu, and A. Elgamal. 2008. OpenSees soil models and solid-
.1061/(ASCE)ST.1943-541X.0001059. fluid fully coupled elements: User’s manual. San Diego: Univ. of
Kolozvari, K., K. Orakcal, and J. W. Wallace. 2015b. Shear-flexure inter- California San Diego.
action modeling for reinforced concrete structural walls and columns Zhang, Q., and Y. Li. 2017. “The performance of resistance progressive
under reversed cyclic loading. PEER Rep. No. 2015/12. Berkeley, CA: collapse analysis for high-rise frame-shear structure based on Open-
Pacific Earthquake Engineering Research Center. Sees.” Shock Vib. 2017: 1–13. https://doi.org/10.1155/2017/3518232.

© ASCE 04020028-14 J. Geotech. Geoenviron. Eng.

J. Geotech. Geoenviron. Eng., 2020, 146(5): 04020028

You might also like