Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

Received: 15 September 2020 Revised: 10 January 2021 Accepted: 11 January 2021

DOI: 10.1002/eqe.3427

RESEARCH ARTICLE

Effect of superstructure modeling assumptions on the


seismic performance of seismically isolated buildings

Shoma Kitayama Ph.D. Michael C. Constantinou

Department of Civil, Structural, and


Environmental Engineering, University at Abstract
BuffaloState University of New York, Recent studies on the seismic performance of seismically isolated buildings with
Buffalo, New York, New York
braced frames designed using the minimum criteria of the ASCE 7-16 standard
Correspondence resulted in observations that seismically isolated buildings have unacceptable
Shoma Kitayama, Department of Civil, collapse performance due to either insufficient isolator displacement capacity
Structural and Environmental Engineer-
ing, University at Buffalo, State University or excessive inelastic structural deformations. The collapse performance and the
of New York, Buffalo, NY 14260, USA. overall performance could be improved by increasing the displacement capac-
Email: shomakit@buffalo.edu
ity of the isolators and by limiting inelastic action in the superstructure while
still providing ductile detailing. These observations may have been dependent
on the modeling assumptions for the superstructure. This article investigates
the effect of various modeling methods that are commonly used in research and
design practice on the seismic performance of a six-story braced seismically iso-
lated building with two different isolation systems, and a comparable nonisolated
six-story building, all designed to meet the minimum criteria of ASCE 7-16. Per-
formance measures considered in this study include story drift ratios, residual
story drift ratios, floor accelerations, isolator horizontal displacement, and col-
lapse. The study found that for seismically isolated buildings and for the mod-
eling assumptions considered, only the inclusion of the effect of gravity frames
has a beneficial effect of reducing the occurrence of small to moderate peak floor
accelerations while there is insignificant effect on the other performance mea-
sures and particularly the computed probabilities of collapse in the MCER . Also,
the different modeling assumptions were found to have a more important effect
on the seismic performance of nonisolated buildings than that of comparable
seismically isolated buildings.

KEYWORDS
ASCE/SEI 7 standard, conditional spectra, modeling assumptions, probabilistic assessment,
Seismic isolation, structural and nonstructural damage

1 INTRODUCTION

Nonlinear analysis, including response history analysis, is an important step of performance-based seismic design and
assessment process.1 Structural models used in the analysis affect the results of the design and assessment.2,3 The effect of

[Correction added on 30 January 2021 after online publication: Replaced the word seis12mically to seismically]
[Correction added on 30 January 2021 after online publication: Rearranged the reference citation]

Earthquake Engng Struct Dyn. 2021;1–19. wileyonlinelibrary.com/journal/eqe © 2021 John Wiley & Sons Ltd. 1
2 KITAYAMA and CONSTANTINOU

modeling assumptions on the engineering demand parameters (EDP) (eg, deformation of structure) of conventional non-
isolated structures obtained in nonlinear response history analysis has been shown to be significant.4,5 By recognizing the
importance of structural modeling and the fact that there is no specific guidance for creating nonlinear structural models,
NIST1 developed a series of guidelines for establishing consistent modeling parameters and assumptions for conventional
structural systems.
Many studies modeled seismically isolated buildings by single, two, or multiple degrees of freedom simplified stick
models,6–9 etc.). Such simplified modeling approaches are justified based on the assumption that the behavior of seis-
mically isolated buildings under seismic ground motions is dominated by the behavior of the isolation system and only
minimally affected by the behavior of the superstructure. Few studies investigated the effect of modeling assumptions
on the performance of seismically isolated buildings but were limited to linear elastic superstructure models and did
not consider their inelastic behavior.10,11 Similarly, analytical models used in the analysis and design of seismically iso-
lated buildings often utilize highly detailed but elastic superstructure models as the analysis is limited to seismic ground
motions that cause no or limited inelastic action in the superstructure.12,13
The seismic performance assessment of structures is nowadays typically based on calculations of the structural response
under seismic ground motion of a wide range of intensity, characterized by frequent to rare seismic events.14 While under
low intensity frequent seismic events an isolated superstructure behaves elastically, it may undergo inelastic action under
rare events and may even collapse.15–21 Thus, modeling the superstructure of seismically isolated buildings may have an
effect on the seismic performance assessment. In fact, the aforementioned studies and other studies22–24 on the seismic
performance of seismically isolated buildings showed significant differences in collapse and other performance indicators
between seismically isolated moment and braced frames. These differences, however, are due to differences in the stiffness
of the superstructures (affecting drift and acceleration) and differences in their ability to deform (the story drift ratio at
collapse for special moment resisting frames is larger than for special concentrically braced frames [SCBF]). There is not
a systematic study that investigated the effect of the superstructure modeling on the seismic performance of seismically
isolated buildings.
This study utilizes a probabilistic procedure based on NIST25 and Lin et al.26 and investigated the effect of the superstruc-
ture modeling on the seismic performance of a six-story seismically isolated building. Each of the superstructure models
used is a legitimate model of increasing complexity by comparison of a baseline model for which its bases are available
in Kitayama and Constantinou.18–21 The evaluation of seismic performance included information on floor accelerations
and story drifts, as being key EDP affecting structural damage and nonstructural damage to mechanical, electrical, and
plumbing equipment, ceilings and partitions, and building contents (Section C17.2 in Ref. [27]. The residual story drift
was also included as it is related to the need to repair or demolish a damaged building after an earthquake.14,28–30 Finally,
this study considered collapse as an important performance indicator. Specifically, this study computed the mean annual
frequency of exceeding the EDP of peak story drift ratio, peak floor acceleration, peak residual story drift ratio, and peak
isolator horizontal displacement (which is important in assessing failure of the seismic isolation system and collapse of
the building). Also, collapse fragility curves for the building were constructed and used to assess the importance of the
modeling assumption on its collapse performance.
Analysis was conducted with 400 ground motions which were selected to be compatible with conditional spectra (CS)
which were constructed for the site of each of the analyzed buildings. The CS represented 10 different seismic intensi-
ties characterized by return periods of 43 to 10 000 years. Forty ground motion records were selected for each seismic
intensity (based on the return period) with their conditional mean and their conditional standard deviation being con-
sistent with the period of each of the analyzed structures.25,26 The analyzed building was the six-story perimeter frame
building previously studied in Kitayama and Constantinou18–21 as a seismically isolated and as a conventional building.
The analytical model of the building was modified for this study to account for the following: (i) shear tab modeling
in seismic force resisting frame; (ii) gravity frame modeling; (iii) gusset plate modeling; and (iv) brace fracture model-
ing. The design of the superstructure followed the minimum requirements of ASCE 7-1627 and was originally reported in
Kitayama and Constantinou.18–21 The isolation system consisted of triple Friction Pendulum (FP) isolators having a dis-
placement capacity at initiation of stiffening equal to DM (isolator maximum displacement in the MCER per ASCE 7-16).
Additionally, a second isolation system was studied consisting of Double Concave (DC) isolators. These isolators lacked
a displacement restrainer and had a displacement at collapse just a little over the displacement DM . The two systems had
essentially the same behavior up to displacement DM . A third isolation system was also studied consisting of the triple FP
isolators and with a moat wall placed so that it was engaged when the displacement reached the limit of DM . The isola-
tion systems and the superstructure designs met the minimum requirements of ASCE 7-16 but were found in Kitayama
and Constantinou18–21 to have unacceptable probabilities of collapse in the MCER due to either insufficient isolator
KITAYAMA and CONSTANTINOU 3

F I G U R E 1 Plan and braced frame elevation of


analyzed building

TA B L E 1 Sections of columns and braces


SCBF Gravity frames Isolated &
Story Isolated Nonisolated Nonisolated
RI = 2.0 RI = 6.0 –
6 Column W12 × 35 W12 × 35 W12 × 35
Brace HSS4 × 0.237 HSS5 × 0.25 –
5 Column W12 × 35 W12 × 35 W12 × 35
Brace HSS5 × 0.5 HSS6 × 0.5 –
4 Column W12 × 58 W12 × 58 W12 × 58
Brace HSS6 × 0.5 HSS7 × 0.5 –
3 Column W12 × 58 W12 × 58 W12 × 58
Brace HSS6 × 0.5 HSS7 × 0.5 –
2 Column W12 × 96 W12 × 96 W12 × 96
Brace HSS6.625 × 0.5 HSS8.625 × 0.5 –
1 Column W12 × 96 W12 × 96 W12 × 96
Brace HSS6.625 × 0.5 HSS8.625 × 0.5 –

displacement capacity or excessive story drift. The analyzed buildings had two distinctly different fundamental periods
(3.66 s effective period at displacement DM for the isolated building in the lower bound friction conditions, and 0.524 s for
the nonisolated building) so that two sets of CS were constructed and two sets of 400 ground motions each were selected
to represent the CS.

2 DESCRIPTION OF ANALYZED BUILDING

The plan and a braced frame elevation of the analyzed building are shown in Figure 1. The original design of this building
was presented in SEAONC Volume 5 Seismic Design Manual31 and McVitty and Constantinou.32 The seismic weight at
each floor was based on previous studies.31,32 The building was assumed located on soil class D in San Francisco, CA (Lat-
itude 37.783◦ , Longitude -122.392◦ ) with Risk-Targeted Maximum Considered Earthquake (MCER 27 spectral acceleration
values of SMS = 1.5 g and SM1 = 0.9 g. For the Design Earthquake (DE),27 the parameters are SDS = 1.0 g and SD1 = 0.6 g. The
seismic force resisting frames for the isolated buildings were designed for RI = 2 and detailed as SCBF. The frames are min-
imally compliant with the requirements in ASCE 7-16.27 Sections for the columns and beams in the gravity frames were
originally selected in SEAONC.31 The comparable nonisolated building was designed based on Section 12.8.4.2 in ASCE
7–16.27 A two-dimensional model of this building was previously studied in Kitayama and Constantinou.18–21 Tables 1
and 2 present the sections of columns, braces, and beams for the isolated and nonisolated buildings. Table 2 also includes
information of seismic floor weight. Note that the girder sections above the isolators (first floor) are listed in the tables but
neither in this study nor in the previous studies,18–21,31,32 the size of the girders has been accounted for in the analysis as
4 KITAYAMA and CONSTANTINOU

TA B L E 2 Sections of beams
SCBF Gravity frames Floor weight (kN)
Floor Isolated Nonisolated Isolated & Nonisolated Isolated Nonisolated
RI = 2.0 RI = 6.0 – RI = 2.0 RI = 6.0
Roof W16 × 26 W16 × 26 W21 × 44 3400 3400
6 W16 × 26 W16 × 26 W21 × 44 8100 8100
5 W16 × 26 W16 × 26 W21 × 44 8100 8100
4 W16 × 26 W16 × 26 W21 × 44 8100 8100
3 W16 × 26 W16 × 26 W21 × 44 8200 8200
2 W16 × 26 W16 × 26 W21 × 44 8200 8200
1 W33 × 118 – W33 × 118 8900 –

FIGURE 2 Representation of gravity frames and seismic force-resisting frames of isolated building

the analysis models constrained uplift of the isolators and did not consider failure of the girders. The sectional properties
of these girders used in the models were arbitrarily increased by a large factor to suppress uplift displacement.33 Uplift is
a phenomenon which is highly dependent on details of the structural system analyzed and therefore is not included in
the assessment of performance. Engineers modify the structural system and employ massive girders above the isolators to
eliminate or reduce uplift to manageable levels. Modifications of the bearings (increasing the height of the restrainer rings)
are then employed and tested to prevent instability.12 The steel columns were spliced at the mid-height of odd-numbered
stories except for the first story. Note that the elevator shaft and stairway space were omitted for simplicity.

3 MODELING OF SUPERSTRUCTURE AND ISOLATION SYSTEM

Analysis was performed in program OpenSees34 using models of two-dimensional representations of the concentrically
braced frames (placed in three-dimensional space) as illustrated in Figure 2 for the seismically isolated building. A similar
model was used for the nonisolated building. Two seismic force-resisting frames in a principal direction (ie, building’s
strong axis) were modelled with details that will be described below. Seismic ground motion was considered only in the
principal direction. The seismic force resisting frames in the perpendicular direction were not considered. Each of the
columns of the seismic force-resisting frame was supported by one exterior isolator.32 Each of the columns of the gravity
frames was supported by one interior isolator. The exterior and interior isolator were identical in geometry but had differ-
ent frictional properties as described in Section 3.2 of this article due to differences in the gravity loads. The gravity and
seismic force-resisting frames were placed parallel to each other. The in-plane stiffness of slabs at all floors was accounted
KITAYAMA and CONSTANTINOU 5

FIGURE 3 Backbone curve used to model shear tab connections and definition of pinching parameters

for by using rigid truss elements connecting the two ends of each beam element and the opposite corners of each slab of
every bay, whereas the out-of-plane stiffness was neglected except at base level (which was assumed to be rigid). Seismic
weights (and mass) were lumped at the nodes (ie, at all beam-column intersections) in proportion to each node’s tributary
area.

3.1 Modeling of superstructure and isolation system

The modeling of superstructure was based on the use of concentrated plasticity elements (nonlinear spring hinges) for
the beams and columns and of distributed plasticity elements (fiber sections) for the braces.2 This modeling provides for
numerically efficient formulations and is, thus, suitable for studies that require a large number of nonlinear response his-
tory analyses.3,35 The elastic stiffness of beam-column elements between the concentrated plasticity springs was selected
based on Ibarra and Krawinkler36 so that the equivalent stiffness of the “rotational spring—elastic element— rotational
spring” assembly was equivalent to the stiffness of the actual frame member. The following four types of detailed modeling
were considered in this study: (i) shear tab modeling in the seismic force resisting frames; (ii) gravity frame modeling; (iii)
gusset plate modeling; and (iv) brace fracture modeling. These models are described below.

3.1.1 Shear tab modeling in seismic force resisting frames

There are several ways to model the shear tab connections in concentrically braced frames. They have been modeled as
pins or zero-rotational stiffness elements31,37–39 or as rigid connections capable of transferring moment17–21,40,41 Studies
have shown that the beam-to-column connections in concentrically braced frames have bending strength of 30 to 70% of
the expected plastic bending moment of the associated steel beam.38,42,43 Models to simulate these shear tabs for braced
frames were developed by Karamanci and Lignos.44 In this study, the shear tab connections were first modelled as pins
(reference case) and then more accurately using the model developed by Karamanci and Lignos44 and the two models
were compared. The model used to simulate the moment-rotation backbone curve of shear tabs (with and without gusset
plates that connect the brace to beams or columns in SCBF) is shown in Figure 3. The behavior of shear tabs was sim-
ulated by using the modified Ibarra-Medina-Krawinkler (IMK) deterioration model with pinched hysteretic response.44
The parameters for this modified IMK deterioration model used in this study are listed in Table 3. Note that while tests
conducted by Stroakes and Fahnestock38 did not include the concrete slab, the model used in this study was calibrated
using their test results because of lack of any other test results to use. Alternatively, one may ignore the effect of the gusset
plate on the behavior of the shear tabs.48 When modelling the shear tabs as pinned connections in this study, the modified
IMK model was replaced by a linear elastic rotational spring element with a very small rotational stiffness (10−10 kN-
m/rad). Figure 4A and B compares the simulated hysteresis (properties per Table 3) of a shear tab connection to results
of tests with and without gusset plates. From this figure, it is seen that the modeling approach used for the gusset plate
reproduces well the observed behavior and the model may be considered validated.
6 KITAYAMA and CONSTANTINOU

TA B L E 3 Properties for shear tab (modified IMK deterioration model with pinched hysteresis)
Parameter Without gusset plate With gusset plate
Mmax + 0.38Mp,Exp *4 My + +αs + K0 ∙θp +
− + *4
Mmax −0.62Mmax My − +αs − K0 ∙θp −
K0 Mslip + /θslip +*5 My + /0.004*3
+ *4
as (strain hardening ratio) 0 0.107*3
− *4
as (strain hardening ratio) 0 0.107*3
My + Mmax + *4 0.5Mp,Exp *3,6
− − *4
My Mmax −My + *3
Fpr+ (for pinching) 0.75*1 0.75*1
− *1
Fpr (for pinching) 0.75 0.75*1
APinch (for pinching) 0.75*1 0.75*1
*1
Λ (all) 0.90 0.90*1
c (all) 1.00*2 1.00*2
+ + + *5
θp θmax -θi 0.055*1
θp − θmax − -θi − *5 0.055*1
+ +
𝜃𝑑𝑟𝑜𝑝 −𝜃𝑚𝑎𝑥
θpc + + + +
*5
0.18*1
1−𝑀𝑑𝑟𝑜𝑝 ∕(𝑀𝑑𝑟𝑜𝑝 −𝑀𝑀𝑎𝑥 )
− −
𝜃𝑑𝑟𝑜𝑝 −𝜃𝑚𝑎𝑥
θpc − − − −
*5
0.18*1
1−𝑀𝑑𝑟𝑜𝑝 ∕(𝑀𝑑𝑟𝑜𝑝 −𝑀𝑀𝑎𝑥 )

κ+ 0.55*5 0 *1
− *5
κ 0.53 0 *1
θu + θbind + +0.04 *7 θmax + +θpc +
− −
θu θbind -0.04 *7
θmax − +θpc −
D+ 1.00*8 1.00*3
− *8
D 1.00 1.00*3
*1
Values based on Karamanci and Lignos.44
*2
Values based on Lignos and Krawinkler.45
*3
Values based on description in Karamanci and Lignos44 and test results in Stroakes and Fahnestock.38
*4
Values from Elkady and Lignos.46
*5
Mslip + = 0.25Mmax + , θslip = 0.004 rad, Mdrop + = 0.55Mmax + , Mdrop − = 0.53Mmax − , θdrop + = 0.04 rad, θdrop − = 0.060 rad, θmax + = 0.03 rad, θmax − = 0.02 rad.46,47
*6
Mp,Exp = 1.1Ry Fy Zx per Stoakes and Fahnestock.38 Note, Ry = 1.1, Fy = 50 ksi for ASTM A992 steel.
*7
θbind + (θbind − ) are rotation at beam binding between the beam flange and the column face. θbind + (θbind − ) = g/df , where g = gap between beam flange and column;
and df = distance from the mid-height of the shear tab to furthest beam flange.47 Conveniently, df is taken as half of beam depth, and g = 1 inch are used.46
*8
Values based on description in Karamanci and Lignos,44 and Elkady and Lignos46 ; and test results in Liu and Astaneh-Asl.42,43

3.1.2 Gravity frame modeling

The effect of gravity columns can be included, in the simplest form, by leaning columns connected to the seismic force
resisting frame by rigid truss elements and with zero rotational stiffness at the connections to the floors (ie, all beam-to-
column connections or shear tab connections are considered pinned).2 While this approach is simple and can simulate
gravity load or P-Δ effects, it ignores the resistance of the gravity frames to lateral loads. Elkady and Lignos46 showed that
when the gravity frames are properly included in the analytical model, the overall base shear strength of steel moment
frame buildings could increase by as much as 50%. Other studies found other beneficial effects of including the gravity
frames in the analytical models such as reduction of drift concentration along the height of buildings49 and improvements
of collapse probability of nonisolated steel moment frames buildings.46,50 Studies of the performance of seismically isolated
buildings by Shao et al.,17 and Kitayama and Constantinou18–21 used single or multiple sets of leaning columns but have
ignored the seismic force resistance provided by the gravity framing. This study investigates the effect of these ignored
effects of the gravity frame on the seismic performance of seismically isolated buildings. The shear tab connections were
modelled as composite following Elkady and Lignos46 who investigated bare and composite shear tab connections and
found that the latter provides for greater modification to the shear tab behavior. The modeling of shear tabs in the gravity
frame followed the hysteretic rule of Figure 3 with properties for the case of “without gusset plate” in Table 3.
KITAYAMA and CONSTANTINOU 7

180 300 1500


Test data Test data Test data
120 Simulation Simulation 1000 Simulation
Moment (kN−m) 150

Moment (kN−m)

Force (kN)
60 500
0
0 0
−150
−60 −500

−120 −300 −1000


−0.15 −0.075 0 0.075 0.15 −0.07 −0.035 0 0.035 0.07 −100 −50 0 50 100
Rotation (rad) Rotation (rad) Axial deformation (mm)
(A) (B) (C)

F I G U R E 4 Comparisons of simulated and test results: (A) shear tab hysteretic loop without gusset plate (data from Ref. [42,43]; (B) shear
tab hysteretic loop with gusset plate (data from Ref. [38]; (C) Pipe5STD brace (with fracture of brace; test data from Figure A.1 in Ref. [55]

3.1.3 Gusset plate modeling

Gusset plates that are located at the end of braces deform when the braces are bent out-of-plane. Many studies of seismi-
cally isolated buildings modelled these connections as pins18–22,24 However, a study by Hsiao et al.51 found that use of pins
to model the gusset plate underestimates the compressive strength of braces and the story shear strength. Other studies by
Hsiao et al.41 and Hwang and Lignos48 used the model proposed by Hsiao et al.51 to investigate the seismic performance
of nonisolated structures with braced frames. In this study, the gusset plate was modeled as a pin connection (reference
case) and then again as a nonlinear rotational spring that can simulate elastic flexibility and yielding of the gusset plate
as proposed by Hsiao et al.51
The gusset plate was designed using the “balanced design procedure” in Roeder et al.52 and Lehman et al.53 The non-
linear rotational spring material “Steel02 (Giuffré-Menegotto-Pinto Model with Isotropic Strain Hardening)” was used to
model yielding of the gusset plate based on Hsiao et al.51 and with parameters R0 = 20, cR1 = 0.925, and cR2 = 0.15.34 When
the gusset plate was modeled as a pin, the Steel02 material was replaced with a linear elastic rotational spring with a very
small rotational stiffness (10−10 kN-m/rad).

3.1.4 Brace fracture modeling

The model of the braces of the SCBF used in this study consisted of eight displacement-based fiber elements that are able
to trace flexural buckling (out-of-plane direction). Uriz and Mahin54 observed in testing of braces that fracture initiates
in the mid-span of braces due to low-cycle fatigue. Chen and Mahin40 concluded in an analytical study that the effect of
inclusion of fracture modeling has an important effect on the computed residual story drift.
In this study, braces were modeled by a model that captures flexural buckling and fracture, and then again with a model
that only captures flexural buckling (reference case). Flexural buckling can be simulated in available software for structural
analysis31,34 ; and has been previously utilized in the studies of Kitayama and Constantinou.18–21 However, modeling of
brace fracture is more challenging.31 In this study, the flexural buckling and fracture initiation by low-cycle fatigue was
modeled based on the recommendations of Karamanci and Lignos44 using the model in program OpenSees for simulating
fracture that was originally developed by Uriz and Mahin.54 This model is based on a linear strain accumulation rule in
accordance with the Coffin-Manson relationship55 in the logarithmic domain as described by:
( )𝑚
𝜀𝑖 = 𝜀0 𝑁𝑓 (1)

In this equation, ε0 is a material parameter indicating the strain amplitude εi at which one complete cycle of the undam-
aged material will cause fracture. Exponent m is a material parameter that relates the dependency of the total strain
amplitude of the material to the number of cycles to fracture (Nf ). Information on the elements and materials for model-
ing braces in OpenSees is presented in Table 4. When the fracture of brace is ignored, the “fatigue material” (see Table 4)
is removed from the model. A comparison of simulation results (parameters of Table 4 and with pins at the ends of the
8 KITAYAMA and CONSTANTINOU

TA B L E 4 Brace element and material properties in OpenSees


Number of segments along the length 8
Number of integration point 5
Number of fibers along the outside perimeter 12
Number of fibers through the thickness 4
Steel02 Material (Giuffré-Menegotto-Pinto Model) Yield stress (expected) 441 MPa
Young’s modulus 2 00 000 MPa
Strain hardening ratio 0.005
R0 24.0
cR1 0.925
cR2 0.250
a1 0.02
a2 1.00
a3 0.02
a4 1.00
Fatigue Material ε0 *
m −0.30
*ε0 is calculated based on Equation ( 5) in Karamanci and Lignos.44

TA B L E 5 Geometric and upper and lower bound friction properties of TFP and DC isolators
Interior isolators Exterior isolators
Outer Inner Outer Inner
Isolator type Quantity surface surface surface surface
Triple friction pendulum Radius Reff1 = Reff4 (mm) 2120 2120 2120 2120
Radius Reff2 = Reff3 (mm) 229 229 229 229
Upper bound friction coef. μ1 = μ4 0.087 0.022 0.101 0.022
Lower bound friction coef. μ1 = μ4 0.042 0.015 0.042 0.015
Double concave Radius Reff1 = Reff4 (mm) 2120 2120 2120 2120
Upper bound friction coef. μ 0.080 NA 0.093 NA
Lower bound friction coef. μ 0.039 NA 0.039 NA

brace) and test results (data from Ref. [56] is shown in Figure 4C. There is good agreement between the two sets of results
to conclude that the model is valid.

3.2 Design and modeling of seismic isolation system

The seismic isolation system considered in this study consisted either of triple FP isolators with stiffening behavior
or equivalent DC isolators without stiffening behavior. Identical bearings were placed at each of the columns of the
building but because of the differences in gravity load, the isolators were modelled with different frictional properties
at the inner and outer (perimeter) locations. McVitty and Constantinou32 and Kitayama and Constantinou18 presented
details of the selection of the properties of the isolators based on prototype bearing tests and application of bounding
analysis principles. Table 5 presents the geometric (effective radii) and upper and lower bound frictional properties of
the isolators. Figure 5 presents drawings and force-displacement curves of the triple FP (TFP) and the DC isolators. The
DC isolators lack restrainer rings, which is nowadays a common practice in applications in Europe but not the United
States.57 It should be noted that the current version of the European Standard EN15129 (CEN, 2018; Section 8.3.1.2.3)58
prohibits the use of restrainer rings while ASCE 7-1627 does not have such requirement. Also, the TFP isolators lack
an interior restrainer ring which also is a common practice.59 The lack of the interior rings does not affect the ultimate
displacement capacity of the isolator.
KITAYAMA and CONSTANTINOU 9

FIGURE 5 Drawings and force-displacement loops of triple FP and DC isolators

The isolator displacement capacity was selected to meet the minimum requirements of standard ASCE 7-16.27 That is,
quantity DCapacity = 518 mm shown in Figure 5 is equal to the maximum displacement DM calculated by the procedures of
ASCE 7-16 for the maximum earthquake in the lower bound conditions, which include simplified and nonlinear response
history analysis procedures. For the triple FP isolator, this displacement was also selected to be the displacement at which
initiation of stiffening occurs (for lower bound friction conditions). The triple FP isolators collapse at a displacement
DUltimate = 683 mm whereas the DC isolators collapse at displacement DUltimate = 526 mm as they are smaller in plan
dimensions than the equivalent triple FP isolators. The ultimate displacement capacity was computed for a configuration
in which the internal components of the isolators become unstable and verified computationally using the theory of Sarlis
and Constantinou.60,61
The same TFP isolators are used in a third isolation system in which a moat wall is placed at location to be reached
when the isolation system displacement equals to DM . The behavior of this isolation system is depicted in Figure 5 with
a higher stiffness when the displacement reaches the value DM . Details of the moat wall modeling are presented later in
Section 3.3.
Kitayama and Constantinou18,21 have shown that the isolated building with these three isolation systems has unac-
ceptable probability of collapse when the superstructure is modelled by the representation defined in this article as the
“reference” case. The interest in this article is to investigate the effect of the modeling assumptions on the calculated
collapse and other performance indicators.
The model of Dao et al.62 in OpenSees (version 2.5.0, rev. 6345) was used in this study. Friction was assumed to be of
the Coulomb type so that the values of friction coefficient in Table 5 were assumed to be valid for all velocities and to be
independent of the instantaneous value of the vertical load. The vertical force-deformation relationship of each element
was simulated using a bilinear elastic spring material (“Elastic Uniaxial Material” in OpenSees) with compressive stiffness
calculated per Sarlis and Constantinou63 by kVC = ASlider ∙(E/2)/HTFP , where ASlider is area of slider, E is Young’s modulus
of steel ( = 200 000 MPa) and HTFP is height of the bearing ( = 320 mm). In tension, the isolators do not have resistance
and a very small stiffness was assigned in lieu of zero (kVT = 17.5 N/mm or 0.1 kip/inch). The vertical damping is included
in each element representing a bearing to have damping ratio of β = 0.1 leading to a damping constant for each isolator
element cv,c = 2∙β∙√(kVC ∙W/g), where g is the acceleration of gravity and W is the gravity load (which differs depending on
the location). The addition of vertical damping in the isolator was found to be useful for achieving numerical convergence
without affecting the horizontal seismic response (studied by varying the value of β). This was possible in this study as
isolator uplift, rocking and impact were not considered. When such phenomena occur, vertical damping could not be
modelled using the approach used and instead vertical damping in isolator bearings should be small and functional only
in compression.
10 KITAYAMA and CONSTANTINOU

3.3 Consideration of pounding to moat wall

The moat wall was modeled by the Hertz model64 with nonlinear damping previously used in the areas of robotics and
multibody systems.65 The Hertz model has been used in recent investigations of seismic response analysis of seismically
isolated buildings impact (eg, impact to restraining rim in sliding isolation bearings by Bao et al.66 ; impact to moat wall
by Masroor and Mosqueda.15 ). Other studies used similar models to simulate impact to a moat wall or adjacent structures
(modified Kelvin-Voight model by Pant and Wijeyewickrema67,68 ; linear viscoelastic model by Komodromos et al.69 ).
The application of the Hertz model to impact simulation was detailed in Muthkumar and DesRoches.70 A correction to
the Muthkumar and DesRoches approach by Ye et al.71 was used in this study. For the Hertz model, the impact force FC
is given by
{
𝑘h 𝛿 𝑛 + 𝑐h 𝛿̇ (𝛿 > 0)
𝐹C = (2)
0 (𝛿 ≤ 0)

where n is the Hertz coefficient, kh is the nonlinear impact spring constant, ch is the nonlinear damping constant, δ is the
relative penetration ( = u1 -ug ; u1 is the relative displacement of the structure with respect to the ground and ug is the gap
distance from the structure to the wall) and 𝛿̇ is the penetration velocity. The values of n = 1.5 and kh = 1576 kN/mm were
used based on Masroor and Mosqueda.15
The damping constant ch in Equation (2) was calculated as:

𝑐h = 𝜁𝛿 𝑛 (3)

where ζ is the damping ratio, determined per Ye et al.71 as:

8 𝑘h (1 − 𝑒)
𝜁= ⋅ (4)
5 𝑒𝛿̇ 0

where e is the coefficient of restitution and 𝛿̇ 0 is initial penetration velocity. In this study, e = 0.6 was used based on
Masroor and Mosqueda.15
Note that any contribution of the soil backfill on the moat wall was not considered as this would have required detailed
design of the moat wall.72 Also, frictional forces at the contact and torsional response, which may affect the engagement of
the moat wall have not been considered. The uniaxial material element in OpenSees was utilized to simulate pounding on
the basis of the Hertz model described above. The uniaxial material element incorporating the Hertz model was coupled
with the Two Node Link Element and used in the building model based on Pant and Wijeyewickrema.66 Note that the moat
wall model is based on physical considerations without realistic experimental validation (ex. 67,68) or small-scale labora-
tory testing and/or detailed finite element modeling (ex. 15,72 ). Also, use of different moat wall models (eg, Section 10.3.3
in Ref. [73] may alter the results presented in this article. These are the limitations of this study that are worthy of future
investigation.

3.4 Model of inherent damping for nonlinear response history analysis

Inherent damping in the isolated buildings was modelled using the “modal damping” approach74 to provide a damping
ratio of 0.02 in all modes but the first “isolated” mode, for which the damping ratio was set equal to zero per directions in
Sarlis and Constantinou.63 For the comparable nonisolated buildings, the damping ratio was specified as 0.02 in all modes.
In the modal damping model, the elastic properties of the structure were used to construct a damping matrix, which was
then used in the response history analysis. The application of the procedure for the isolated building model is complicated
by the fact that the isolated building models are extremely stiff in the elastic range (the elastic stiffness is the friction force
at initiation of motion divided by some very small “yield” displacement, which in the model was set equal to 1 mm). To
construct the damping matrix for the postelastic stiffness of isolators in accordance with Sarlis and Constantinou,60,61
horizontal, vertical, and rotational springs were placed above each isolator element to represent the desired stiffness in
the horizontal direction. These springs were restrained to only deflect by 0.025 mm so that under inelastic conditions in
the nonlinear response history analysis they did not affect the behavior of the isolators.18,19
KITAYAMA and CONSTANTINOU 11

TA B L E 6 Analyzed cases
Shear tab in seismic Gravity frame Gusset plate Brace
Case force resisting frame model model fracture System
1* Not modeled Not modeled Not modeled Not modeled TFP
2 Modeled Not modeled Not modeled Not modeled TFP
3 Not modeled Modeled Not modeled Not modeled TFP
4 Not modeled Not modeled Modeled Not modeled TFP
5 Not modeled Not modeled Not modeled Modeled TFP
*
6 Not modeled Not modeled Not modeled Not modeled TFP with Moat Wall
7 Modeled Not modeled Not modeled Not modeled TFP with Moat Wall
8 Not modeled Modeled Not modeled Not modeled TFP with Moat Wall
9 Not modeled Not modeled Modeled Not modeled TFP with Moat Wall
10 Not modeled Not modeled Not modeled Modeled TFP with Moat Wall
11* Not modeled Not modeled Not modeled Not modeled DC
12 Modeled Not modeled Not modeled Not modeled DC
13 Not modeled Modeled Not modeled Not modeled DC
14 Not modeled Not modeled Modeled Not modeled DC
15 Not modeled Not modeled Not modeled Modeled DC
*
16 Not modeled Not modeled Not modeled Not modeled Nonisolated
17 Modeled Not modeled Not modeled Not modeled Nonisolated
18 Not modeled Modeled Not modeled Not modeled Nonisolated
19 Not modeled Not modeled Modeled Not modeled Nonisolated
20 Not modeled Not modeled Not modeled Modeled Nonisolated
*Denotes a reference case.

3.5 Analyzed cases

Table 6 summarizes the studied cases. In the study, each of the modeling assumptions was progressively evaluated against
the reference case. Note that all isolated buildings had the same superstructure which was designed per ASCE 7-1627 for
RI = 2.0 (maximum allowed RI ). A total of 20 cases were analyzed, each using 400 ground motions for a total of 8000
nonlinear response history analyses.
Note that each of the four modeling cases in Table 6 was individually studied and that models combining the effects
were not considered as validated models for the combined effects do not exist. However, we attempted to combine some
of the effects in one model but found that the computation took significant time, and the results were difficult to interpret
and with frequent numerical convergence issues without any other indicators of potential failure like large drift and brace
fracture. We abandoned the combined models as lacking validity.

3.6 Definition of collapse

Collapse of the analyzed buildings was defined as the occurrence of any of the following:

a. The maximum story drift ratio exceeds 0.05.41,75


b. The isolator displacement exceeds the ultimate displacement (DUltimate = 683 mm for the TFP isolators with or without
a moat wall; DUltimate = 523 mm for the DC isolators) (collapse of the isolator),
c. Instability detected by termination of the analysis program (numerical convergence problem).

Numerical convergence problems were observed in the analysis of all cases in Table 6. In general, numerical conver-
gence problems were observed when large displacements, large story drifts, fracture of braces, and combinations of these
occurred, and they were more frequent when TFP isolators with moat walls were used. We investigated the effect of the
solution algorithm (we used the Krylov Newton algorithm and then replaced it with the NewtonLineSearch34 and varied
12 KITAYAMA and CONSTANTINOU

the time step of integration to achieve numerical convergence). In some cases, numerical convergence was not achieved
and this was considered an indication of collapse.
Collapse of the isolators was considered as failure of the structure. The behavior of the isolation interface after collapse
of the isolators is unknown and largely dependent on details of construction, which are not topics investigated in this
article. It is possible to provide the isolation system with another fail-safe system to prevent collapse of the building but
such systems can only function for elastomeric isolators.8,76 Moreover, while such backup systems have been proposed
some two decades ago, they only found application in the first seismic isolation application to a new building in the US-the
Foothill Communities Law and Justice Center in California which features an elastomeric isolation system (Section 10 in
Ref. [77].

4 SEISMIC PERFORMANCE ASSESSMENT

4.1 Selection and scaling of ground motions

The motions used are identical to those used in Kitayama and Constantinou18,20 where details on the site-specific response
spectra and on the selection and scaling of the motions can be found. Background information can be found in NIST25 and
Lin et al.26 A total 800 motions were selected and scaled to represent 10 different seismic intensities as measured by the
earthquake return period of 43, 144, 289, 475, 949, 1485, 2475, 3899, 7462, and 10 000 years, and for two different building
periods (3.66 s for the isolated and 0.524 s for the nonisolated buildings).
The multiple stripe analysis technique78 was used for the analyses as it allowed for the use of different sets of hazard-
consistent ground motions at each intensity level (ie, return period). The results of the study are presented in the form of
relationships between specific values of EDP and the annual frequency of exceeding these EDP values. The EDP examined
are the maximum story drift ratio, the maximum residual story drift ratio, the peak floor acceleration, and the peak isolator
displacement. These EDP are indicators of damage to the structural and nonstructural systems and the contents of the
buildings.14 Collapse was also assessed.

4.2 Effect of superstructure modeling on seismic performance of seismically isolated


buildings

Mean annual frequencies of the EDP exceeding a value y, λ(EDP > y), for each of the 20 cases in Table 6 were calculated.
Results are presented in terms of the ratio of λ(EDP > y) of a case to the corresponding λ(EDP > y) for the reference
case, 𝜆∕𝜆Ref erence . When this ratio equals to unity, the studied modeling assumption has no effect. When the ratio is less
than unity, the studied modeling assumption results in improvement in the mean annual frequency and vice versa. Values
of the mean annual frequency for the refences cases λReference are also presented.
The calculation of λ(EDP > y) was based on the following18,20,25,26 :
𝑛

𝜆 or 𝜆Ref erence = 𝜆 (𝐸𝐷𝑃 > 𝑦) = 𝑃 (𝐸𝐷𝑃 > 𝑦| (𝑆𝑎(𝑇1 ) = 𝑥𝑖 ) 𝜆 (𝑆𝑎 (𝑇1 ) = 𝑥𝑖 ) (5)
𝑖=1
{ ( )}
ln 𝑦 − 𝜇ln 𝐸𝐷𝑃
𝑃 (𝐸𝐷𝑃 > 𝑦| (𝑆𝑎(𝑇1 ) = 𝑥𝑖 ) = 𝑃 (𝐶) + (1 − 𝑃 (𝐶)) 1−Φ (6)
𝜎ln 𝐸𝐷𝑃

𝜆 (𝑆𝑎 (𝑇1 ) = 𝑥𝑖 ) = 0.5 ⋅ {𝜆 (𝑆𝑎 (𝑇1 ) > 𝑥𝑖−1 ) − 𝜆 (𝑆𝑎 (𝑇1 ) > 𝑥𝑖+1 )} (7)

where P(EDP > y|Sa (T1 ) = xi ) is the probability of EDP exceeding y given a ground motion intensity (in terms of spectral
acceleration at fundamental period T1 ) with Sa(T1 ) = xi and seismic hazard curve λ(Sa(T1 ) > xi ). λ(Sa(T1 ) = xi ) is the
rate of observing Sa(T1 ) amplitudes in a small range represented by the discrete amplitude xi . Also, P(C) is the probability
of collapse. While Lin et al.26 and NIST25 computed the logarithmic mean and standard deviation, μlnEDP and σlnEDP
in Equation (6), for the peak floor acceleration using all data, including both collapse and noncollapse cases, this study
excluded collapse cases in computing the mean and standard deviation. This is due to the observation that the inclusion
of collapse cases in the computation of these values generated biased results with some unrealistically small mean values
for large seismic intensities.
KITAYAMA and CONSTANTINOU 13

−2 −3 −3 −2
[10 ] [10 ] [10 ] [10 ]
2 2 2 2 2 2 2 4
Gusset plate
Shear tab
λ / λReference Brace fracture

λ / λReference

λ / λReference

λ / λReference
λReference

λReference

λReference

λReference
Gravity frame
1 1 1 1 1 1 1 2

Reference case DUltimate=683mm


0 0 0 0 0 0 0 0
0.1 1 2 0.001 0.01 0.05 0.001 0.01 0.02 1 10 100
Peak floor acceleration (g) Peak story drift ratio Peak residual story drift ratio Peak isolator dispacement (mm)
TFP isolation system
−2 −3 −3 −2
[10 ] [10 ] [10 ] [10 ]
2 2 2 2 2 2 2 4
Gusset plate
Shear tab
λ / λReference

λ / λReference

λ / λReference

λ / λReference
Brace fracture
λReference

λReference

λReference

λReference
Gravity frame
1 1 1 1 1 1 1 2

Reference case DUltimate=683mm


0 0 0 0 0 0 0 0
0.1 1 2 0.001 0.01 0.05 0.001 0.01 0.02 1 10 100
Peak floor acceleration (g) Peak story drift ratio Peak residual story drift ratio Peak isolator dispacement (mm)
TFP with moat wall isolation system
−2 −2 −2 −2
[10 ] [10 ] [10 ] [10 ]
2 2 2 2 2 2 2 4
Reference case
λ / λReference

λ / λReference

λ / λReference

λ / λReference
λReference

λReference

λReference

λReference
1 1 1 1 1 1 1 2
Gusset plate
Shear tab
Brace fracture
Gravity frame DUltimate=526mm
0 0 0 0 0 0 0 0
0.1 1 2 0.001 0.01 0.05 0.001 0.01 0.02 1 10 100
Peak floor acceleration (g) Peak story drift ratio Peak residual story drift ratio Peak isolator dispacement (mm)
DC isolation system

FIGURE 6 Mean annual frequency ratios and values for reference cases for seismically isolated buildings

Figure 6 presents the results of the analysis for the seismically isolated buildings. Colored solid lines represent
λ/λReference as computed by Equation (5). The dashed lines represent λReference . Evidently, there are minor effects of the
modeling assumptions on the calculated mean annual frequencies of exceeding limits of the selected EDP. Only the
inclusion of the gravity frame in the analysis model consistently resulted in some noticeable beneficial effect of reducing
the occurrence of the peak floor acceleration for values less than 1 g. The likely reason for this is the increase in the
superstructure stiffness when the gravity frame is included in the model. This may have an impact on the performance
of acceleration-sensitive nonstructural components.14 Shear tabs also resulted in similar effect but the effect is smaller
because the number of shear tabs in the seismic force resisting system is less than the number of shear connections in
the gravity frame (see Figure 1). Also, the inclusion of the brace fracture in the model had some small detrimental effect
on the mean annual frequency of exceeding limits of large story drift and large residual story drift when moat walls
were used. Note that effect of variation in superstructure models when DC bearings are used is smaller than the effect
when other isolation systems are used as the superstructure does not exhibit nonlinear behavior (collapse occurs due to
failure of the isolators). These effects should be expected to be more important when DC isolators of larger displacement
capacities are used. In summary, the results suggest that modeling of isolated buildings should include the gravity frames
to accurately simulate peak acceleration response, especially for isolation systems with stiffening behavior.

4.3 Effect of superstructure modeling on seismic performance of nonisolated buildings

Figure 7 presents the results of the analysis for the conventional nonisolated buildings. The modelling assumptions have a
significantly larger effect than for the comparable seismically isolated buildings. The inclusion of the gravity frame effects
14 KITAYAMA and CONSTANTINOU

−2 −2 −3
[10 ] [10 ] [10 ]
2 4 2 4 2 4
Gusset plate
λ / λReference Shear tab

λ / λReference

λ / λReference
Brace fracture

λReference

λReference

λReference
Gravity frame
1 2 1 2 1 2

Reference case

0 0 0 0 0 0
0.1 1 2 0.001 0.01 0.05 0.001 0.01 0.02
Peak floor acceleration (g) Peak story drift ratio Peak residual story drift ratio

FIGURE 7 Mean annual frequency ratios and values for reference cases for nonisolated buildings
Probability of collapse (%/100)

Probability of collapse (%/100)

Probability of collapse (%/100)

Probability of collapse (%/100)


1.0 1.0 1.0 1.0

0.5 Reference 0.5 Reference 0.5 Reference 0.5 Reference


Gusset plate Gusset plate Gusset plate Gusset plate
Shear tab Shear tab Shear tab Shear tab
Brace fracture Brace fracture Brace fracture Brace fracture
Gravity frame Gravity frame Gravity frame Gravity frame
0.0 0.0 0.0 0.0
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5 0.0 3.0 6.0 9.0
Sa(T1) (g) Sa(T1) (g) Sa(T1) (g) Sa(T1) (g)
TFP TFP with moat wall DC Non−isolated
Probability of collapse, P(C)

FIGURE 8 Effect of modeling assumptions on collapse fragility curves

results in important reduction of the occurrence of peak story drift ratio and residual story drift ratio. Also, the gusset
plate modeling resulted in reduction of the occurrence of large peak floor acceleration (exceeding 1 g) but resulted in a
marked increase of the occurrence of the peak story drift ratio. Modeling of brace fracture had minor effects in increasing
the occurrence of large values of story drift ratio and of large residual story drift ratio. The latter observation is consistent
with the study in Chen and Mahin40 where it was demonstrated that models of concentrically braced frames without
brace fracture resulted in smaller median maximum residual story drift than a model in which both buckling and fracture
were considered.

4.4 Effect of superstructure modeling on collapse probabilities of buildings

4.4.1 Collapse fragility curves

Fragility curves were constructed by fitting the empirical collapse data with a lognormal distribution. The empirical col-
lapse data were in the form of the probability of collapse (number of collapse cases divided by the number of analyses)
versus 10 distinct values of the seismic intensity as measured by the spectral acceleration at the fundamental period (for
return periods of 43 to 10 000 years). Collapse was defined in Section 3.6. Figure 8 presents collapse fragility curves of
the isolated and nonisolated buildings as represented by the five different models of Table 6. Each of the graphs in this
figure contains five curves for a total of 20 fragility curves. Each of the fragility curves is characterized by the median,
𝑆ˆ𝑎Collapse (𝑇1 ) (value of spectral acceleration for which the probability of collapse is 0.5), and the dispersion, βRTR , which
reflects the dispersion in the results due to the record-to-record variability (uncertainty). The results of Figure 8 show that
the modeling assumptions have no effect on the collapse fragility curves for the isolated buildings without the moat wall.
When a moat wall is used, there is a small increase in the probability of collapse by comparison to the reference case when
the gravity frames and when brace fracture were included in the analysis models.
For the nonisolated building, a notable reduction of the probability of collapse is observed when the gravity frame
is included in the analysis model. This was expected as modeling of the gravity frame results in a stiffer and stronger
KITAYAMA and CONSTANTINOU 15

superstructure (this is also true but to a much lesser effect for the other modelling assumptions except that of the brace
fracture). However, the opposite was observed for the seismically isolated building with moat walls where modeling of
the gravity frame resulted in small increases in the probability of collapse. The likely reason of this result is the effect of
both the very stiff moat wall and the stiff gravity frames that caused an increase in the number of numerical instability
problems. Changes to the solution algorithm and the integration time step resulted in improvements of the numerical
solutions but still there were cases that did not reach convergence. It is likely that the increase in the probability of failure
when moat walls are used is partially due to inability of the analysis program to capture impact on the moat walls. That
is, more robust moat wall models and/or solution algorithms are needed.

4.4.2 Evaluation of probability of collapse at MCER

While the fragility curves in Figure 8 show small differences when different models are used in the case of the seismically
isolated buildings, it is useful to compute probabilities of collapse given the occurrence of the maximum considered earth-
quake (MCER ). These probabilities may be affected by the modeling assumptions depending on the value of the Sa(T1 )
in the MCER spectrum.27 In the approach used, and unlike the approach of FEMA,73 the calculation of the probabili-
ties of collapse given the occurrence of the MCER was done directly from the fragility curves without the need to make
corrections for spectral shape effects.18,19 That is, the adjusted collapse margin ratio (ACMR)73 was directly computed as

𝑆ˆ
𝑎Collapse (𝑇1 )
𝐴𝐶𝑀𝑅 = (8)
𝑆𝑎MCE (𝑇1 )

where SaMCE (T1 ) is the spectral acceleration of the MCER spectrum27 at the fundamental period of structure T1 . Values of
the MCER spectral acceleration at the fundamental period are SaMCE (T1 = 3.66 s) = 0.246 g for the isolated buildings and
SaMCE (T1 = 0.524 s) = 1.5 g for the nonisolated building.18,19 Uncertainties from the various sources were then considered
to compute the total uncertainty βTOT per FEMA73 :

2 2 2 2
𝛽TOT = 𝛽RTR + 𝛽DR + 𝛽TD + 𝛽MDL (9)

In Equation (9), βDR is the design requirements-related collapse uncertainty, βTD is the test data-related collapse uncer-
tainty, and βMDL is the modeling-related collapse uncertainty. For the seismically isolated buildings, the following quality
ratings and related uncertainties were used: good with βMDL = 0.2 for modeling; good with βTD = 0.2 for test data; and
superior with βDR = 0.1 for design requirements (same assumptions made in Ref. [15]. For the nonisolated structure,
βMDL = 0.2, βTD = 0.2, and βDR = 0.2 were used based on Chen and Mahin40 and NIST.79 The probability of collapse at the
MCER , PCollapse,MCE is computed by
[ 2
]
1
1 (ln 𝑠 − 𝐴𝐶𝑀𝑅)
𝑃Collapse,MCE = √ exp − 𝑑𝑠 (10)
∫0 𝑠𝛽TOT 2𝜋
2
2𝛽TOT

Figure 9 presents the ratio of computed value of PCollapse,MCE in each of the 20 analyzed cases to the probability of
collapse for reference case (see Table 6). In general, the modelling assumptions have insignificant effect on the probability
of collapse in the MCER for the isolated buildings without a moat wall. When a moat wall is used, the probabilities of
collapse increase by as much as 1.5 when including the gravity frame effects. As noted earlier, the computation of the
probabilities included cases with numerical convergence problems. We believe that in order for the results to be consistent
with theoretical expectations, such numerical convergence problems need to be solved. They present a challenge in the
analysis of seismically isolated buildings with moat walls, particularly when stiff superstructures are considered. Also, the
insensitivity of the probability of collapse for the seismically isolated buildings without moat walls in Figure 9 is due to
the fact that collapse is mostly due large isolator displacements rather than due to excessive superstructure deformation
so that changes in the superstructure model have insignificant effects. For a discussion on this issue and for a presentation
of representative collapse modes, the reader is referred to Kitayama and Constantinou.21
For the nonisolated building, there is marked improvement in the probability of collapse given the MCER when the
gravity frames are included in the model and the opposite effect when brace fracture is included. Both of these effects
were expected and discussed earlier in Section 4.4.1.
16 KITAYAMA and CONSTANTINOU

PCollapse, MCE / PCollapse, MCE, Reference


TFP TFP with moat wall DC Non−isolated

PCollapse, MCE, Reference PCollapse, MCE, Reference PCollapse, MCE, Reference PCollapse, MCE, Reference
= 21.79% = 12.53% = 40.58% = 8.40%
0
10 20

te
Sh te

te

Sh te
re

re
b

b
e

e
ur

ur
m

m
ta

ta

ta

ta
la
la

la

la
tu

tu
fra

ct

fra

ct

fra

fra
tp

tp

tp
tp

ar

ar

ar

Br ear
ac

ac
fra

fra
e

e
se

se

se
se

r
ty

ty

ty

ty
Sh

Sh
ef

ef
us

us

us
us

ce

ce
vi

vi

vi

vi
ac

ac
G

G
G

ra

ra

ra

ra
a

a
Br

Br

Br
G

G
FIGURE 9 Ratios of probability of collapse and collapse probabilities for the reference cases in the MCER

Earlier studies by authors have found the probabilities of collapse for the minimally designed seismically isolated build-
ings to be unacceptably large.18–21 This is evident in Figure 9 where the values of the probability of collapse for the refer-
ence cases are reported. The results show that the modelling assumptions, however, complex and advanced, provide no
relief to the earlier predictions. That is, changing the details in superstructure model of seismically isolated buildings may
not result in improvements in the computed collapse performance. Improvements in the collapse performance requires
increases in the superstructure strength and/or in the displacement capacity of isolators.

5 CONCLUSIONS

This study examined the effect of different superstructure modeling methods on the seismic performance of seismically
isolated and conventional buildings with steel SCBF, which were designed using the minimum requirements of ASCE
7-16.27 The methods examined included: (i) Shear tab modeling in the seismic force resisting frames, (ii) Gravity frame
modeling; (iii) Gusset plate modeling; and (iv) Brace fracture modeling.
Isolation systems with stiffening triple FP isolators and DC isolators without stiffening behavior were designed to have
the minimum displacement capacity required by the ASCE standard. In an alternate configuration, these isolation systems
were supplied with a moat wall placed at the minimum distance required by the ASCE standard to prevent collapse of the
isolators. The superstructure was designed with the largest value of the R factor allowed for steel special braced frames
(RI = 2 for the isolated and R = 6 for the nonisolated buildings). The properties of the isolation systems were considered to
be within bounds determined by considering heating effects, aging, and manufacturing variability. The presented results
are for the lower bound conditions for which the probabilities of collapse were the highest. The assessment of performance
was based on the calculation of mean annual frequencies of exceeding specific values of engineering demand parameters
(story drift, residual story drift, and floor acceleration) and collapse fragility curves constructed using a suite of 400 motions
to represent seismic intensities ranging from 43 to 10 000 years of return period for a particular site in California. The
fragility curves were then used to compute probabilities of collapse given the MCER . The findings of the study are:

a. The effect of the superstructure modeling methods on seismic performance was significantly more pronounced for the
nonisolated buildings than the comparable seismically isolated buildings.
b. The effect of the modeling methods on the calculated mean annual frequencies of exceeding limits of the selected
engineering demand parameters for the isolated buildings were minor except for the case of the gravity frames and
only for the case of the peak floor acceleration. The inclusion of the gravity frame in the analysis model resulted in a
noticeable beneficial effect of reducing the occurrence of the peak floor acceleration for values less than 1 g.
c. The inclusion of the gravity frame in the analysis model of nonisolated buildings resulted in significant reduction of
the occurrence of the peak story drift ratio and the residual story drift ratio. This was not observed in the comparable
seismically isolated buildings.
d. The collapse fragility curves of seismically isolated buildings and the probabilities of collapse in the MCER were practi-
cally unaffected by the modeling methods studied except for the case of the isolation system with moat wall and gravity
frame modeled, when the probability of collapse increased by factor of 1.5. It is likely that the increase in the probability
of collapse when moat walls and gravity frame are used is partially due to inability of the analysis program to capture
impact on the moat walls. That is, more robust moat wall models and/or solution algorithms are needed.
KITAYAMA and CONSTANTINOU 17

e. The collapse fragility curves of the nonisolated building and the probabilities of collapse in the MCER were significantly
improved when the gravity frame was included in the modeling of the structure.
f. Changes in the modeling of the superstructure may not result in any improvement in the collapse performance. Rather,
improvements of performance of isolated buildings require reduction of the RI factor and/or increases in the displace-
ment capacity of the isolators beyond the minimum requirements of ASCE 7-16.18–21

D A T A AVA I L A B I L I T Y S T A T E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

ORCID
Shoma Kitayama Ph.D. https://orcid.org/0000-0001-9416-3772

REFERENCES
1. NIST. Guidelines for nonlinear structural analysis for design of buildings. Part II–General. NIST GCR 17-917-46v1. Gaithersburg, MD:
National Institute of Standards and Technology; 2017.
2. NIST. NEHRP seismic design technical brief No. 4: nonlinear structural analysis for seismic design: a guide for practicing engineers. NIST
GCR 10-917-5. Gaithersburg, MD: National Institute of Standards and Technology; 2010.
3. Ugurhan B, Baker JW, Deierlein GG. Uncertainty estimation in seismic collapse assessment of modern reinforced concrete moment frame
buildings. Tenth U.S. National Conference on Earthquake Engineering. Frontiers of Earthquake Engineering. July 21–25. Anchorage, Alaska:
EERI; 2014.
4. Elnashai AS, McClure DC. Effect of modelling assumptions and input motion characteristics on seismic design parameters of RC bridge
piers. Earthquake Eng Struct Dyn. 1996;25(5):435-463.
5. Terzic V, Schoettler MJ, Restrepo JI, Mahin SA. Concrete column blind prediction contest 2010: outcomes and observations. Technical
Report PEER 2015/01. Berkeley, CA: University of California; 2015.
6. Kikuchi M, Tamura K, Wada A. Safety evaluation of base-isolated structures. Journal of Structural and Construction Engineering (Trans-
actions of AIJ). 1995;60:65-73.(in Japanese).
7. Sayani PJ, Ryan KL. Comparative evaluation of base-isolated and fixed-base buildings using a comprehensive response index. J Struct Eng.
2009;135(6):698-707.
8. Nakazawa N, Kishiki S, Qu Z, Miyoshi A, Wada A. Fundamental study on probabilistic evaluation of the ultimate state of base isolated
structures. 8CUEE Conference proceedings, 8th International conference on urban earthquake engineering, March 7-8. Tokyo, Japan: Tokyo
Institute of Technology; 2011.
9. Chimamphant S, Kasai K. Comparative response and performance of base-isolated and fixed-base structures. Earthquake Eng Struct Dyn.
2016;45(1):5-27.
10. Kikuchi M, Kitamura Y, Inoue K, Ueda M. An analytical model for seismic isolation devices with hysteretic damping. Journal of Structural
and Construction Engineering (Transactions of AIJ). 2003;68:63-71. in Japanese.
11. Giammona AP, Ryan KL, Dao ND. Evaluation of assumptions used in engineering practice to model buildings isolated with triple pendu-
lum isolators in SAP2000. Earthquake Spectra. 2015;31(2):637-660.
12. Sarkisian M, Lee P, Hu L, Doo C-S, Tsui A. Enhanced seismic performance of the new San Bernardino justice center. SEAOC 2011 Con-
vention proceedings. Las Vegas, NV: SEAOC; 2011.
13. Huang S, Islam S, Skokan M, Wang M, Oguzmert M, (2017). “Base-isolated intermodal transportation center serves dual purpose as emer-
gency response nerve center.” 16th World Conference on Earthquake, 16WCEE 2017. Santiago Chile, January 9th to 13th 2017. Paper #:
4632.
14. FEMA. (2012). Seismic performance assessment of buildings. Washington, DC: Report FEMA P-58, Federal Emergency Management
Agency; 2012.
15. Masroor A, Mosqueda G. Assessing the collapse probability of base-isolated buildings considering pounding to moat walls using the FEMA
P695 methodology. Earthquake Spectra. 2015;31(4):2069-2086.
16. Bao Y, Becker TC, Sone T, Hamaguchi H. To limit forces or displacements: collapse study of steel frames isolated by sliding bearings with
and without restraining rims. Soil Dyn Earthquake Eng. 2018;112:203-214.
17. Shao B, Mahin SA, Zayas VA. Achieving targeted levels of reliability for low-rise seismically isolated structures. Soil Dyn Earthquake Eng.
2019;125:105744.
18. Kitayama S, Constantinou MC. Seismic performance assessment of seismically isolated buildings designed by the procedures of ASCE/SEI
7. MCEER 18-0004. Buffalo, NY: Multidisciplinary Center for Earthquake Engineering Research; 2018.
19. Kitayama S, Constantinou MC. Collapse performance of seismically isolated buildings designed by the procedures of ASCE/SEI 7. Eng
Struct. 2018;164:243-258.
20. Kitayama S, Constantinou MC. Probabilistic seismic performance assessment of seismically isolated buildings designed by the procedures
of ASCE/SEI 7 and other enhanced criteria. Eng Struct. 2019;179:566-582.
21. Kitayama S, Constantinou MC. Effect of displacement restraint on the collapse performance of seismically isolated buildings. Bull Earth-
quake Eng. 2019;17(5):2767-2786.
18 KITAYAMA and CONSTANTINOU

22. Lin AN, Shenton HW. Seismic performance of fixed-base and base-isolated steel frames. J Struct Eng. 1992;118(5):921-941.
23. Sayani PJ, Erduran E, Ryan KL. Comparative response assessment of minimally compliant low-rise base-isolated and conventional steel
moment-resisting frame buildings. J Struct Eng. 2010;137(10):1118-1131.
24. Erduran E, Dao ND, Ryan KL. Comparative response assessment of minimally compliant low-rise conventional and base-isolated steel
frames. Earthquake Eng Struct Dyn. 2010;40(10):1123-1141.
25. NIST. Selecting and scaling earthquake ground motions for performing response-history analysis. NIST GCR 11-917-15. Gaithersburg, MD:
National Institute of Standards and Technology; 2011.
26. Lin T, Haselton CB, Baker JW. Conditional spectrum-based ground motion selection. Part I: hazard consistency for risk-based assessments.
Earthquake Engineering and Structural Engineering. 2013;42(12):1847-1865.
27. ASCE/SEI 7. Minimum Design Loads and Associated Criteria for Buildings and Other Structures. Reston, VA: American Society of Civil
Engineers; 2017.
28. Christopoulos C, Pampanin S, Priestley MJN. Performance-based seismic response of frame structures including residual deformations.
Part I: single-degree of freedom systems. J Earthquake Eng. 2003;7(1):97-118.
29. Ruiz-Garcia J, Miranda E. Evaluation of residual drift demands in regular multi-story frames for performance-based seismic assessment.
Earthquake Eng Struct Dyn. 2006;35(13):1609-1629.
30. Erochko J, Christopoulos C, Tremblay R, Choi H. Residual drift response of SMRFs and BRB frames in steel buildings designed according
to ASCE 7-05. J Struct Eng. 2011;137(5):589-599.
31. Structural Engineers Association of California. SEAOC structural/seismic design manual. Volume 5: examples for seismically isolated
buildings and buildings with supplemental damping. Washington, DC: International Code Council, 2014.
32. McVitty WJ, Constantinou MC, (2015). Property modification factors for seismic isolators: design guidance for buildings. MCEER-15-0005,
Multidisciplinary Center for Earthquake Engineering Research, Buffalo, NY.
33. Christopoulos C, Filiatrault A . Principles of passive supplemental damping and seismic isolation. Pavia, Italy: IUSS Press, Intituto Uni-
versitario di Studi Superiori di Pavia; 2006.
34. McKenna FT, (1997). Object-oriented finite element programming: frameworks for analysis, algorithms and parallel computing. Ph.D.
thesis, University of California, Berkeley, CA.
35. Deierlein GG, Reinhorn AM, Willford MR. Nonlinear structural analysis for seismic design. A guide for practicing engineers. NEHRP
Seismic Design Technical Brief No. 4. NIST GCR 10-917-5. Gaithersburg, MD: NIST: National Institute of Standards and Technology; 2010.
36. Ibarra LF, Krawinkler H, (2005). Global collapse of frame structures under seismic excitations. John A. Blume Earthquake Engineering
Center Technical Report 152. Stanford Digital Repository. http://purl.stanford.edu/dj885ym2486. Accessed Jan. 19, 2021.
37. Marino EM, Nakashima M, Seismic performance and new design procedure for chevron-braced frames. Earthquake Engineering and
Structural Dynamics. 2005;35 (4). 433-452.
38. Stoakes CD, Fahnestock LA. Cyclic flexural testing of concentrically braced frame beam-column connections. J Struct Eng. 2011;137(7):739-
747.
39. Del Gobbo GM, Williams MS, Blakeborough A. Seismic performance assessment of Eurocode 8-compliant concentrically braced frame
buildings using FEMA P-58. Eng Struct. 2018;155:192-208.
40. Chen CC, Mahin SA. Performance-based seismic demand assessment of concentrically braced steel frame buildings. Technical Report PEER
2012/103. Berkeley, CA: University of California; 2012.
41. Hsiao PC, Lehman DE, Roeder CW. Evaluation of the response modification coefficient and collapse potential of special concentrically
braced frames. Earthquake Eng Struct Dyn. 2013;42(10):1547-1564.
42. Liu J, Astaneh-Asl A. Cyclic testing of simple connections including effects of slab. J Struct Eng. 2000;126(1):32-39.
43. Liu J, Astaneh-Asl A. Cyclic tests on simple connections, including effects of the slab. Report No. SAC/BD-00/03. Background document.
SAC Joint Venture. June 2000.
44. Karamanci E, Lignos DG. Computational approach for collapse assessment of concentrically braced frames in seismic regions. J Struct
Eng. 2014;140(8):A4014019.
45. Lignos DG, Krawinkler H. Deterioration modeling of steel components in support of collapse prediction of steel moment frames under
earthquake loading. J Struct Eng. 2011;137(11):1291-1302.
46. Elkady A, Lignos DG. Effect of gravity framing on the overstrength and collapse capacity of steel frame buildings with perimeter special
moment frames. Earthquake Eng Struct Dyn. 2015;44(8):1289-1307.
47. Liu J, Astaneh-Asl A. Moment-rotation parameters for composite shear tab connections. J Struct Eng. 2004;130(9):1371-1380.
48. Hwang SH, Lignos DG. Effect of modeling assumptions on the earthquake-induced losses and collapse risk of steel-frame buildings with
special concentrically braced frames. J Struct Eng. 2017;143(9):04017116.
49. Ji X, Kato M, Wang T, Hitaka T, Nakashima M. Effect of gravity columns on mitigation of drift concentration for braced frames. J Constr
Steel Res. 2009;65(12):2148-2156.
50. Flores FX, Charney FA, Lopez-Garcia D. Influence of the gravity framing system on the collapse performance of special steel moment
frames. J Constr Steel Res. 2014;101:351-362.
51. Hsiao PC, Lehman DE, Roeder CW. Improved analytical model for special concentrically braced frames. J Constr Steel Res. 2012;73:80-94.
52. Roeder CW, Lumpkin EJ, Lehman DE. A balanced design procedure for special concentrically braced frame connections. J Constr Steel
Res. 2011;67:1760-1772.
KITAYAMA and CONSTANTINOU 19

53. Lehman DE, Roeder CW, Herman D, Johnson S, Kotulka B. Improved seismic performance of gusset plate connections. J Struct Eng.
2008;134(6):890-901.
54. Uriz P, Mahin SA. Toward earthquake-resistant design of concentrically braced steel-frame structures. Technical Report PEER 2008/08.
Berkeley, CA: University of California; 2008.
55. Manson SS. Fatigue: a complex subject—Some simple approximations. Exp Mech. 1965;5(4):193-226.
56. Fell BV, Kanvinde AM, Deierlein GG. Large-scale testing and simulation of earthquake induced ultra-low cycle fatigue in bracing members
subjected to cyclic inelastic buckling. Blume Earthquake Engineering Center. Report No. 172. Stanford, CA: Stanford University; 2010.
57. Ponzo FC, Cesare AD, Leccese G, Nigro D, Shake table testing on restoring capability of double concave friction pendulum seismic isolation
systems.Earthquake Eng Struct Dyn 2017;46 (14). 2337-2353.
58. CEN, European Committee for Standardization. Brussels, Belgium: European Standard EN15129: Anti-seismic devices. (English version);
2018.
59. Lee D, Constantinou MC. Combined horizontal–vertical seismic isolation system for high-voltage–power transformers: development, test-
ing and validation. Bull Earthquake Eng. 2018;16:4273-4296.
60. Sarlis AA, Constantinou MC. Model of triple friction pendulum isolators for general geometric and frictional parameters and for uplift
conditions. Report MCEER-13-0010. Buffalo, NY: Multidisciplinary Center for Earthquake Engineering Research; 2013.
61. Sarlis AA, Constantinou MC. A model of triple friction pendulum bearing for general geometric and frictional parameters. Earthquake
Eng Struct Dyn. 2016:1837-1853.
62. Dao N, Ryan KL, Sato E, Sasaki T. Predicting the displacement of triple pendulum™ bearings in a full-scale shaking experiment using a
three-dimensional element. Earthquake Eng Struct Dyn. 2013;42(11):1677-1695.
63. Sarlis AA, Constantinou MC. Modeling triple friction pendulum isolators in program SAP2000. Report distributed to the engineering
community together with example files. University at Buffalo, Buffalo, NY; 2010.
64. Hertz H, Gesammelte Werke. Vol 1, Leipzig, 1985.
65. Lankarani HM, Nikravesh PE. A contact force model with hysteresis damping for impact analysis of multibody systems. J Mech Des.
1990;112(3):369-376.
66. Bao Y, Becker TC, Hamaguchi H. Failure of double friction pendulum bearings under pulse-type motions. Earthquake Eng Struct Dyn.
2016;46(5):715-732.
67. Pant DR, Wijeyewickrema AC. Structural performance of a base-isolated reinforced concrete building subjected to seismic pounding.
Earthquake Eng Struct Dyn. 2012;41(12):1709-1716.
68. Komodromos P, Polycarpou PC, Papaloizou L, Phocas MC. Response of seismically isolated buildings considering poundings. Earthquake
Eng Struct Dyn. 2007;36(2):1605-1622.
69. Pant DR, Wijeyewickrema AC. Influence of near-fault ground motions on the response of base-isolated reinforced concrete buildings
considering seismic pounding. Advances in Structural Engineering. 2013;16(12):1973-1988.
70. Muthkumar S, DesRoches R. A hertz contact model with non-linear damping for pounding simulation. Earthquake Eng Struct Dyn.
2006;35(7):811-828.
71. Ye K, Li L, Zhu H. A note on the Hertz contact model with nonlinear damping for pounding simulation. Earthquake Eng Struct Dyn.
2008;38(9):1135-1142.
72. Hughes PJ, Mosqueda G, Evaluation of uniaxial contact models for moat wall pounding simulations. Earthquake Eng Struct Dyn. 2020;49
(12). 1197-1215.
73. FEMA. Quantification of building seismic performance factors. Washington, DC: Report FEMA P695. Federal Emergency Management
Agency; 2009.
74. Chopra AK, McKenna F. Modeling viscous damping in nonlinear response history analysis of buildings for earthquake excitation. Earth-
quake Eng Struct Dyn. 2016;45(2):193-211.
75. Sabelli R, Roeder CW, Hajjar JF, (2013). Seismic design of steel special concentrically braced frame systems—a guide for practicing engi-
neers. NEHRP Seismic Design Technical Brief No. 8, NIST GCR 13-917-24.
76. Kelly JM, Beucke KE. A friction damped base isolation system with fail-safe characteristics. Earthquake Eng Struct Dyn. 1983;11(1):33-56.
77. Tarics AG, Way D, Kelly JM. The implementation of base isolation for the foothill communities law and justice center. County of San
Bernardino, California. Washington, DC: National Science Foundation; 1984. NSF/CEE-84041. November. Section 10.
78. Jalayer F, (2003). Direct probabilistic seismic analysis: implementing non-linear dynamic assessments. PhD Thesis, Stanford University.
79. NIST. Evaluation of the FEMA P-695 methodology for quantification of building seismic performance factors. NIST GCR 10-917-8. Gaithers-
burg, MD: National Institute of Standards and Technology; 2010.

How to cite this article: Kitayama S, Constantinou MC. Effect of superstructure modeling assumptions on the
seismic performance of seismically isolated buildings. Earthquake Engng Struct Dyn. 2021;1–19.
https://doi.org/10.1002/eqe.3427

You might also like