Download as pdf or txt
Download as pdf or txt
You are on page 1of 31

August 5, 2021 1:20:17pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Journal of Earthquake and Tsunami


(2022)
.c World Scienti¯c Publishing Company
#
DOI: 10.1142/S1793431122500038

Phase-Structure in Conditional Simulation of


Spatially Varying Ground Motion and Possible
In°uence on Structural Demand

Gopala Krishna Rodda, Narsiram Gurjar and Dhiman Basu*


Department of Civil Engineering
Indian Institute of Technology Gandhinagar, India
*dbasu@iitgn.ac.in

Received 21 April 2021


Accepted 5 July 2021
Published

Recorded ground motion is nonstationary in both intensity and frequency contents. Two
methodologies were reported by the authors elsewhere for generating spatially varying ground
motion (SVGM), namely, (i) auto-spectral density (ASD)-based framework, and (ii) evolu-
tionary power-spectral density (EPSD)-based framework. While the former framework
imparts nonstationarity through a uniform modulation (that accounts for nonstationarity
only in intensities), the latter framework accounts for nonstationarity in both intensity and
frequency contents. Reported EPSD-based framework was modeled through a decay function
and a random component and was investigated only in the context of horizontal ground
motion. Reported EPSD-based framework made two strong assumptions that need further
investigation: (i) spatial variation of the random component was assumed to be frequency
independent; and (ii) phase-structure of the ground excitation simulated around the reference
station (with seed motion) was assumed to be same as that of the seed motion. This paper
investigates the possible impact of these two assumptions on the simulated SVGM through
appropriately revising the framework and introducing the phase-structure accordingly. Pos-
sible e®ects of the phase-structure on structural demand are investigated through an idealized
long-span bridge. Revised EPSD-based framework is next assessed against the vertical
recordings of SMART1 array along with the auto-spectral density (ASD) framework. Though
spectral representation is nearly identical in both the frameworks, the acceleration time series
simulated using the revised EPSD-based framework matches the recorded data better when
compared with the ASD-based framework. Possible e®ect of spatially varying vertical ground
motion on the seismic design is investigated through the same idealized bridge model. Sig-
ni¯cant increase in the demand of axial force in piers and mid-span moment in the deck are
observed. Although these inferences are contingent on the idealized example considered for
illustration, the spatially varying vertical ground motion is expected to contribute signi¯-
cantly to the seismic design of long-span bridges.

Keywords: Nonstationary vertical ground motion; spatially varying ground motion; Hilbert
transform; modulating function.

*Corresponding author.

2250003-1
August 5, 2021 1:20:17pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

1. Introduction
Owing to the paucity of the recorded ground motion, structural engineers often need
to rely on a simulated ground motion for the analysis of important structures.
Simulating the stationary ground motion was the focus of initial studies [Housner,
1955; Rosenblueth, 1956; Bycroft, 1960]. However, recorded ground motions dem-
onstrate that the ground acceleration is always nonstationary and hence, the use of
stationary ground motion is unrealistic. Several researchers [Bolotin, 1960; Amin and
Ang, 1968; Shinozuka and Sata, 1967; Jennings et al., 1968; Preumont, 1985]
reported procedures that could impart the nonstationarity to a stationary ground
motion by multiplication with a temporally varying modulating function and these
procedures are referred to uniformly modulated process. The recorded ground motion
exhibits nonstationarity in both intensity and frequency contents whereas these
uniformly modulated procedures simulate only the nonstationarity in intensity. One
way of incorporating the frequency nonstationarity is by segmenting the ground
motion into di®erent parts and modeling each segment as an independent uniformly
modulated process [Shinozuka and Sata, 1967; Liu, 1970]. Kameda [1975] and Li and
Kareem [1993] made use of evolutionary power spectral density (EPSD) and multi-
¯lter techniques to simulate the nonstationary ground acceleration. Der Kiureghian
and Crempien [1989] generated the nonstationary ground acceleration through the
superposition of uniformly modulated time series corresponding to the disjoint fre-
quency bands. Zeldin and Spanos [1996] simulated the nonstationary SVGM
using the wavelet transform. Use of Hilbert transform is gaining attention in
recent times for simulating nonstationary ground motions [Wen and Gu, 2004;
Sgobba et al., 2011].
Amplitude as well as the phase associated with the ground excitation do change as
the seismic waves propagate over a spatial footprint and it is referred to the spatially
varying ground motion (SVGM). Coherency-based methods, often referred to the
inverse method, are the commonly used frameworks for simulating the SVGM. Sev-
eral lagged coherency models were reported for modeling the spatial variability [Loh,
1985; Harichandran and Vanmarcke, 1986; Luco and Wong, 1986; Loh and Yeh,
1988; Hao et al., 1989; Abrahamson et al., 1991; Hao, 1996; Zerva and Harada, 1997;
Rodda and Basu, 2018a,b; Wang et al., 2019]. These coherency-based methods are
generally classi¯ed as: (1) Cross-Spectral Density (CSD)-based frameworks; and (2)
ASD-based frameworks. CSD-based frameworks model the spatial variability using
CSD/evolutionary CSD (ECSD) and assuming the spatially uniform ASD/
EPSD. Hence, these frameworks account for only the phase variation. However,
Rodda and Basu [2018b] reported an ASD-based framework that simulates the
SVGM using a calibrated coherency model and mapping of ASD, with due consid-
eration of both the amplitude and phase variabilities.
CSD-based frameworks, in general, simulate the nonstationary SVGM through
the uniform modulation [Hao et al., 1989; Bi and Hao, 2012; Zhang et al., 2013; Di
Paola and Zingales, 2000; Wu et al., 2011; Shinozuka and Zhang, 1996; Kameda and

2250003-2
August 5, 2021 1:20:17pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

Morikawa, 1992]. However, some CSD-based frameworks generate the nonstationary


SVGM without using uniform modulation and instead of one of the following tech-
niques: (1) dividing the acceleration into several stationary segments [Vanmarcke
and Fenton, 1991; Vanmarcke et al., 1993; Konakli and Der Kiureghian, 2012]; (2)
decomposing the ECSD matrix [Deodatis, 1996; Cacciola and Deodatis, 2011; Hu et
al., 2012; Shields, 2015]; (3) use of phase and duration spectra on CSD matrix
[Shrikhande and Gupta, 1998]; and (4) using either wavelet transform [Huang, 2014;
Sarkar et al., 2016] or Hilbert transform [Wen and Gu, 2004; Yao et al., 2017]. Rodda
and Basu [2020a] revisited the ASD-based framework and reported an EPSD-based
framework that generates the nonstationary SVGM in two steps: (i) calibrating the
coherency model; and (ii) mapping of EPSD over a spatial footprint. This EPSD-
based framework made two strong assumptions: (i) spatial variation of the random
component was assumed to be independent of the frequency; and (ii) phase of the
ground excitation simulated around the reference station (with seed motion) was
assumed to be same as that of the seed motion.
First objective of this paper is to address these two limitations and making ap-
propriate recommendations in a revised EPSD-based framework. SVGM simulated
using both EPSD- and revised EPSD-based frameworks will be compared against the
recorded ground motion to understand the e®ect of phase correlation and frequency
dependency of the contribution from random components. Possible in°uence on the
structural demands will also be investigated in the context of an idealized long-span
bridge.
The SVGM frameworks reported in the prior art reviewed above do not di®er-
entiate between the vertical and horizontal components of ground motion although
rarely explored with the former. Probably the same frameworks can be used in the
case of vertical component as well. Vertical SVGM is likely to alter the expected
demands in certain structures when compared with uniform seismic input
[Harichandran and Wang, 1990; Zerva, 1994; Harichandran et al., 1996; Lupoi et al.,
2005; Adanur et al., 2016; Efthymiou and Camara, 2017]. Hence, the simulation of
SVGM for vertical components is critical to the seismic design of certain structures.
Rodda and Basu [2020b] reported further details in this context.
Second objective of this paper is to assess the possible contribution of spatially
varying vertical ground motion on the structural demand. The same idealized long-
span bridge is again studied against the spatially varying horizontal ground motion
with and without considering the spatially varying vertical components. Also com-
pared is the SVGM of vertical component constructed in this paper against that
generated using the ASD-based framework reported by Rodda and Basu [2018b].

2. Description of the Seismic Array (SMART1)


Ground motion recorded at SMART1 array (Lotung, Taiwan) (Fig. 1) is considered
in this study. SMART1 array consists of 37 accelerometers that are arranged in three
circles with one station at the center (C00). Each circle consists of twelve recording

2250003-3
August 5, 2021 1:20:18pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

Fig. 1. SMART1 array [Zerva and Zervas, 2002].

stations. Radii of inner (I) circle is 0.2 km. Similarly, radii of middle (M) and outer
(O) circles is 1 km and 2 km, respectively. Four seismic events (source: PEER NGA
database   https://ngawest2.berkeley.edu) considered for the present study are
described in detail elsewhere [Rodda and Basu, 2020a].

3. Revisiting the EPSD-based Framework by Rodda and Basu


[2020a] for Possible Improvement
Seismic waves are assumed to be propagating as plane waves (with velocity c Þ and
Fig. 2 provides a schematic representation. Consider two stations, namely, a and b
(at site AÞ separated by a distance u  along the direction of wave propagation. Let the
arrival time perturbation between stations b and a be to . Assuming the soil medium
to be a linear system with hðt; uÞ as the impulse response function, the acceleration
¯eld can be described by
u
xb ðt; uÞ ¼ hðt; uÞ  xa ðt  to Þ þ rðt; uÞ where to ¼ : ð1Þ
c
Here,  denotes convolution; u and c are the magnitudes of the position vector and
wave velocity, respectively; and rðt; uÞ is a zero-mean random component. Despite

2250003-4
August 5, 2021 1:20:19pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

Fig. 2. Seismic wave propagation.

the nonstationarity, it is reasonable to assume the ground excitation as stationary


over a small time window and hence, the associated Fourier transform can be
computed around a speci¯ed time. With this understanding, Fourier transform of
Eq. (1) around time t may be written as

Xb ðf; t; uÞ ¼ Hðf; uÞei2fto Xa ðf; tÞ þ Rðf; t; uÞ: ð2Þ

Here, Xa ; Xb , and R are the Fourier transforms (around time tÞ of ground motions at
stations a, b, and the random component, respectively. Hðf; uÞ is the amplitude
decay and hence, representing the transfer function of medium between the stations
a and b.
Amplitude decay, by de¯nition, is a function of both frequency and distance, but
for simplicity, assumed here as a function of distance only ½Hðf; uÞ  HðuÞ [Refer to
Rodda and Basu [2018b] for the relevant supporting details]. HðuÞ is referred to
decay function in the reminder of this paper.
Assuming the random component to be uncorrelated with the ground accelera-
tion, EPSD and ECSD between stations a and b can be related through
Sbb ðf; t; uÞ ¼ H 2 ðuÞSaa ðf; tÞ þ Srr ðf; t; uÞ;
ð3Þ
Sab ðf; t; uÞ ¼ HðuÞei2fto Saa ðf; tÞ:
Here, Saa , Sbb and Srr denote the EPSD of accelerations at stations a, b and random
component, respectively. Sab ðf; t; uÞ is the ECSD between the ground accelerations
at stations a and b.

2250003-5
August 5, 2021 1:20:21pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

Further, EPSD of the random component is approximated as


Srr ðf; t; uÞ ¼ LðuÞSo ðf; tÞ: ð4Þ
Here, So ðf; tÞ is the EPSD at in¯nite distance; and LðuÞ is the spatial variation of
random component.
For the purpose of quantifying the spatial variability, lagged coherency ½jðf; uÞj
is modeled by that reported in Rodda and Basu [2018b]:
 0:5
LðuÞ
jðf; uÞj ¼ 1 þ g2 ðfÞ
HðuÞ2
Pm !0:5
i u cosðk uÞÞni Xn
i¼1 Pi ð1  e 2 2
¼ 1þ i
Aj eðfj Þ =2 j ;
ð1 þ a1 uÞ2a2 j¼1
X
m
 Pi ¼ 1: ð5Þ
i¼1

Here, g2 ðfÞ represents the frequency variation of lagged coherency and Rodda and
Basu [2018b] may be referred to for detailed interpretation of the parameters.
Mapping of EPSD is given by
Sbb ðf; t; uÞ ¼ ½H 2 ðuÞ þ LðuÞg2 ðfÞSaa ðf; tÞ: ð6Þ
After estimating the EPSD at some distance away from reference station, the as-
sociated nonstationary ground motion is generated using a \backward process" re-
quiring three inputs, namely, (1) EPSD at an away station; (2) associated
modulating functions and (3) phase of the associated frequency bands. Given a
frequency band, the modulating functions across the SMART1 array are compared
and the reported variation is negligible. Hence, the modulating functions associated
with the seed motion are recommended over a spatial footprint. Phase of ground
motion over the entire array is considered same as that of the seed motion for the
sake of simplicity.
It is evident from the discussion presented above that the EPSD-based framework
was developed under two strong assumptions: (1) frequency independent spatial
variation of the random component [Eqs. (4) and (6)]; and (2) same phase as that of
the seed motion over the entire footprint. Possible impact of these two assumptions
on the simulated SVGM is investigated here by assessing against the horizontal
acceleration recorded over SMART1 array. Spatial variation of the random com-
ponent in di®erent frequency bands is explored in the next section followed by the
phase di®erence between the away- and reference-stations.

3.1. Revisiting the spatial variation of the random component


While investigating the spatial variation of random component against frequency,
mapping of EPSD over the spatial footprint can also be performed in di®erent
narrow frequency bands instead of the entire frequency contents. In line with Eq. (6),

2250003-6
August 5, 2021 1:20:22pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

EPSD mapping over the kth frequency band can be re-written as


k
S bb ðf; t; uÞ ¼ ½H 2 ðuÞ þ Lk ðuÞg2 ðfÞS aa
k
ðf; tÞ: ð7Þ
Here, f of kth frequency band refers to the frequencies only in the range fk  fkþ1 .
Lk ðuÞ is computed at every frequency band using Eq. (7) in nonlinear least-square
sense.
Coherency model is calibrated against the data available. The functions, g2 ðfÞ
and HðuÞare taken from the calibrated coherency model. Direction of arrival (DOA)
is identi¯ed as the line joining station C00 (center of the array) with the epicenter.
Stations are arranged into di®erent bins based on their distance (along DOA) from
the reference station (@ 500 m). EPSD at the reference station is noted and median
EPSD for each bin is computed. Using, (1) decay function ½HðuÞ; (2) frequency
variation of lagged coherency ½g2 ðfÞ; (3) EPSD at the reference station; and (4)
median EPSD for each bin, Lk ðuÞ is computed in each frequency band employing
Eq. (7).
Figure 3 presents a sample illustration of Lk ðuÞ computed at various frequency
bands along with the statistically important median, lower and upper envelops.
Logarithmic scale is used for better clarity. Lk ðuÞ associated with the individual
frequency bands is presented in colored plots while the median, lower and upper
envelops are presented in black. Frequency dependency of the spatial variation of
random component is evident from the di®erence between lower and upper envelopes
and the scatter at any distance. Therefore, this paper recommends mapping of EPSD
in narrow frequency bands instead of the entire frequency content. EPSD associated
with kth frequency band at an away station is now estimated using Eq. (7), through
the four inputs: (1) decay function; (2) frequency variation of lagged coherency; (3)
EPSD of kth band at the reference station; and (4) Lk ðuÞ. In other words
k
S away ðf; t; uÞ ¼ ½H 2 ðuÞ þ Lk ðuÞg2 ðfÞS reference
k
ðf; tÞ: ð8Þ

Fig. 3. Lk ðuÞ at all frequency bands.

2250003-7
August 5, 2021 1:20:25pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

3.2. Phase-structure of the ground motion ¯eld


Phase of the entire spatially varying ground motion ¯eld was assumed to be identical
with that of the seed/reference acceleration in the EPSD-based framework reported
by Rodda and Basu [2020a] for simplicity. This may be true over a separation
distance of few meters and for the extremely high wave velocities (such as the dense
rock). However, Zerva [2016] studied the arrival time perturbations over SMART1
array using cross correlation between the ground accelerations. Though the reported
arrival time perturbations were arbitrary, one may not ignore it altogether. This
paper investigates the phase variation over a spatial footprint through the ground
motion ¯eld considered in Eq. (2).
Let us consider the kth band with frequency content in the range space of fk 
fkþ1 and the associated ground motion between stations a and b are related though
X bk ðf; t; uÞ ¼ Hðf; uÞei2fto X ak ðf; tÞ þ Rk ðf; t; uÞ: ð9Þ
Here, X ak , X bk and Rk are the Fourier transforms of ground excitations at stations a,
b, and random motion of kth frequency band around time t, respectively.
As discussed earlier, EPSD of random motion in kth frequency band is given by
k
S rr ðf; t; uÞ ¼ Lk ðuÞS ok ðf; tÞ: ð10Þ
Further, EPSD at an in¯nitely away station is given by
S ok ðf; tÞ ¼ g2 ðfÞS aa
k
ðf; tÞ: ð11Þ
From Eq. (10), Fourier amplitude of the random component at time t is given by
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Rk ðf; t; uÞ ¼ S rrk
ðf; t; uÞT eir ðfÞ : ð12Þ

Here, T is the ground motion duration; r is the random phase at each frequency,
uniformly distributed in the interval 0  2.
Substituting Eqs. (10) and (11) into Eq. (12), one may write that
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Rk ðf; t; uÞ ¼ Lk ðuÞg2 ðfÞS aa k
ðf; t; uÞT eir ðfÞ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ Lk ðuÞg2 ðfÞjX ak ðf; tÞjeir ðfÞ : ð13Þ
p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Here, jX ak ðfÞj ¼ S aa k ðf; uÞT is the amplitude of X .
a
Substituting Eq. (13) into Eq. (9), Fourier transform of ground motion at
station b is
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X bk ðf; uÞ ¼ Hðf; uÞei2fto X ak ðfÞ þ Lk ðuÞg2 ðfÞjX ak ðfÞjeir ðfÞ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
¼ Hðf; uÞei2fto jX ak ðfÞjei a ðfÞ þ Lk ðuÞg2 ðfÞjX ak ðfÞjeir ðfÞ : ð14Þ
k

Here,  ka ðfÞ is the phase of ground motion at station a. Further, Eq. (14) can be
simpli¯ed to
0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X bk ðf; t; uÞ ¼ jX ak ðf; tÞj½Hðf; uÞei a;k ðfÞ þ Lk ðuÞg2 ðfÞeir ðfÞ : ð15Þ

2250003-8
August 5, 2021 1:20:25pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

Here,  0a;k ¼  ka  i2fto . Further simpli¯cation of Eq. (18) leads to


" pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #
i 0a;k ðfÞ Lk ðuÞg2 ðfÞ ir ðfÞ
X b ðf; t; uÞ ¼ jX a ðf; tÞjHðf; uÞ e
k k
þ e
Hðf; uÞ
0
¼ jX ak ðf; tÞjHðf; uÞ½ei a;k ðfÞ þ a ko ðf; uÞeir ðfÞ : ð16Þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Lk ðuÞg2 ðfÞ
Here, a ko ðf; uÞ ¼ Hðf;uÞ . It is evident from Eq. (16) that
0
Phase ½X bk ðf; t; uÞ ¼ Phase ½ei a;k ðfÞ þ a ko ðf; uÞeir ðfÞ : ð17Þ
This is applicable in all the frequency bands. It is evident from Eq. (17) that phase of
the ground acceleration at station b requires four inputs: (1) phase at station a; (2)
arrival time perturbation, to ; (3) the parameter, ao , which is derived from the con-
stituting functions of the coherency model; and (4) uniformly distributed random
phase. Inputs 1, 3 and 4 are known from the seed motion, coherency model, and
random phase, respectively. The arrival time perturbation (input 2) can be estimated
using the knowledge of wave velocity over the footprint of the array. Zerva [2016]
reported the arrival time perturbation as random rather than conforming to the
plane wave propagation. Hence, a di®erent approach is recommended for the arrival
time perturbation as follows.
Using the recorded data, arrival time perturbations are computed in this paper for
all the possible station pairs and maximum arrival time perturbation is noted.
Maximum separation distance between the stations along the DOA is noted. Arrival
time perturbation at any location is then estimated as the weighted average of (1)
arrival time perturbation corresponding to the given separation distance (with re-
spect to the reference station); and (2) random value uniformly distributed over the
interval between zero and maximum arrival time perturbation. In other words:
to ¼ w 1 t 1 þ w 2 t 2 ;
maximum arrival time perturbation ð18Þ
t1 ¼  given separation distance:
maximum separation distance
Here, to is the arrival time perturbation at a location; t1 is the arrival time pertur-
bation corresponding to the given separation distance; t2 is the random value uni-
formly distributed over the interval between zero and maximum arrival time
perturbation; w1 and w2 ¼ 1  w1 are the associated weights.
Equal weight (=0.5) is considered in this paper and hence, the arrival time per-
turbation at any location is
to ¼ 0:5ðt1 þ t2 Þ: ð19Þ
This completes the estimation of phase-structure over the spatial footprint.
To summarize, the revised EPSD-based framework recommends two steps: (1)
estimating the EPSD at an away station though band-wise mapping; and (2) band-
wise estimation of the phase-structure. Once the EPSD at an away station, and
phases in all frequency bands are estimated, associated nonstationary ground

2250003-9
August 5, 2021 1:20:26pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

acceleration are computed through the backward process (Refer to Rodda and
Basu [2020a]).

3.3. Assessment using the horizontal ground motion


Nonstationary horizontal ground acceleration ¯eld is generated through the revised
EPSD-based framework. A target spectrum is de¯ned using the recorded ground
acceleration as follows for the purpose of assessment. Stations are arranged into bins
based on their distance (along the DOA) from the reference station. Spectral ac-
celeration at each station in a bin is computed at any speci¯ed period followed by the
computation of median spectral ordinate at the same period. Repeating this for all
the periods leads to the target spectrum (Readers may refer to Rodda and
Basu [2018b] for the relevant details).
Ground acceleration ¯eld is also generated using the EPSD-based framework
reported by Rodda and Basu [2020a] for the purpose of comparison. In other words,
SVGM is simulated here using two frameworks: (1) Case I  
 EPSD-based frame-
work by Rodda and Basu [2020a]: assuming LðuÞ is independent of the frequency
bands and the phase of the ground motion is same as that of the reference station;
and (2) Case II  
 Revised EPSD-based framework: Calibrating Lk ðuÞ at each
frequency band and using the phase-structure per Eq. (17).
Figure 4 presents a sample comparison of the spectra from both the cases against
the target spectra: Di®erence between the spectra is noted to be negligible. Figure 5
presents the comparison between the acceleration time series: both the ground
motions appear to be nearly identical for all practical purposes. It is instructive to
note that this observation is limited to the dataset considered in this paper and is
likely to be di®erent in another seismic event. Phase di®erence (wrapped within
0  2Þ between the simulated motion at an away station using Case-II and the seed

(a) Bin 1000–1500 m (b) Bin 2000–2500 m

Fig. 4. Assessment of the simulated spectra.

2250003-10
August 5, 2021 1:20:30pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

(a) Bin 500–1000 m

(b) Bin 3000–3500 m

Fig. 5. Comparison of the ground motion (Event 1 


 Along the DOA).

motion at reference station is next compared with that computed from the recorded
data. Figure 6 presents the sample comparison at two away stations. Though di®ers
somewhat from the recorded data, Case-II clearly represents better phase-structure
than Case-I (with zero phase di®erence).
Possible in°uence of this phase-structure on structural demand cannot be assessed
by comparing either the response spectrum or acceleration time series at one loca-
tion. Therefore, e®ect of the SVGM simulated from both Case-I and Case-II is next
investigated using a long-span bridge as described in what follows.

(a) Distance 500–1000 m (b) Distance 3000–3500 m

Fig. 6. Phase di®erence with respect to the reference station.

2250003-11
August 5, 2021 1:20:52pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

3.4. E®ect on the structural response


3.4.1. Bridge geometry
A 2-Lane bridge (with 11 m deck width) is considered here and supported by 41
identical piers at the spacing of 100 m ðLÞ with two abutments at 6m from the end
piers. The cross-section of the deck is provided with four longitudinal T-beam girders
as shown in Fig. 7(a). Each pier is of 10 m height ðHÞ and 2.25 m diameter ðDÞ. M30
concrete (characteristic cube compressive strength is 30 MPa) and Fe500 steel (yield
strength is 500 MPa) are considered for the purpose of material characterization.
After analyzing against the dead load and standard live loads, an equivalent gravity
load of 200 kN/m is considered to be acting prior to the seismic event.

3.4.2. Natural period of numerical model


Fundamental period of the bridge for horizontal vibration as per IS1893-P3-2014 is
pffiffiffiffiffiffiffi
given by Th ¼ 2 =g where,  is the horizontal displacement at pier cap due to a
horizontal force w ¼ Mg and M is the total mass lumped at the pier cap contributed
from adjacent spans. Elastic sti®ness of the pier is given by 3EIp =H 3 where, EIp is
the °exural rigidity of pier based on gross cross-section. Fundamental period of
vertical vibration for a simply supported deck with uniformly distributed mass ðmÞ is

(a) Deck cross-section (c) Load-deformation behavior of the pier

(b) Numerical model of the bridge

Fig. 7. Bridge description.

2250003-12
August 5, 2021 1:20:56pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion


qffiffiffiffiffiffi
given by Tv ¼ 2 L2 EI m
d
where, EId is the °exural rigidity of deck. The resulting
natural periods are calculated as 0.88 and 0.10 s for the horizontal and vertical
vibration, respectively.

3.4.3. Numerical model


The bridge is considered with six degrees of freedom at each node as presented
schematically in Fig. 7(b). Abutments are considered as restraint against transla-
tions. Base of the piers is considered as rigidly held with the ground. The deck is
modeled as simply supported beams with °exural rigidity as 8E+10 kN-m2 to match
the vertical period. Each pier is modeled as a bilinear spring element with (i) elastic
sti®ness 103,359 kN/m as discussed above, (ii) post-elastic sti®ness ratio as 0.05 and,
(iii) yield displacement of 20 mm as illustrated in Fig. 7(c).

3.4.4. Ground motion input


The bridge is considered to be aligned along the diameter O06  O12 of SMART1
array for the purpose of analysis against the ground motion ¯eld recorded during
Event-4. The station O06 is closest to the epicenter and hence, chosen as the reference
station. Pier 1 is located at O06 and Pier 41 at O12 (Fig. 7(b)). In other words, DOA is
aligned along the span of the bridge. As-recorded components (EW-NS) are rotated
along and normal to the DOA. Rotated pair at the reference station is considered as
the seed motion and applied at the base of Pier-1. In line with the spacing of bridge
piers, horizontal SVGM is generated using Case-I and Case-II at the interval of 100 m
(i.e. at the location of each pier). If the recorded ground motion is available at any
pier location, the simulated pair is replaced by the recorded data. This is the case
with piers- 1, 11, 19, 21, 23, 31 and 41. Ground motion at the abutment is considered
the same as that of the nearest pier. Randomness is involved in two stages while
simulating the SVGM, namely, (1) calibration of nonlinear coherency model
(Eq. (5)); and (2) random phase contribution (Eq. (17)). Therefore, 10 di®erent sets
of SVGM are simulated to account for this randomness while using Cases-I and
II. The structure is analyzed for these sets of ground motion. Gravity load is applied
in 10 steps and the time history analysis against each set of SVGM is continued from
state at the end of gravity load application.

3.4.5. Structural response


Displacement contours are considered here for the purpose of assessment. Figure 8
presents a sample illustration for Set 2 at some of the piers. Maximum resultant
displacement from Case-I and Case-II are denoted as RI and RII , respectively, and
schematically compared in Appendix at all the pier locations. Substantial di®erence
is noted between RI and RII but without any de¯nite trend. Ratio of maximum
displacements between Case-II and Case-I ðDR ¼ RII =RI Þ is considered for further
quantitative assessment (Table 1). Maximum and minimum DR associated with each
set are highlighted in bold. At some of the piers, Case II leads to higher displacements

2250003-13
August 5, 2021 1:20:58pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

(a) Pier 6 (b) Pier 7

(c) Pier 15 (d) Pier 16

Fig. 8. Displacement contours at selected piers.

Table 1. Ratio of maximum displacements ðDR Þ.

Set 1 Set 2 Set 3 Set 4 Set 5 Set 6 Set 7 Set 8 Set 9 Set 10

Pier 1 1.13 1.13 1.01 1.00 1.50 1.06 1.14 1.28 1.43 1.55
Pier 2 0.84 1.06 0.88 0.52 0.96 0.65 0.70 0.80 0.60 0.75
Pier 3 0.93 1.44 1.00 0.93 1.07 1.68 1.40 0.76 0.92 0.97
Pier 4 1.12 0.97 0.78 0.74 0.77 1.02 1.00 0.97 1.15 0.92
Pier 5 0.92 0.79 0.65 0.73 0.75 0.73 1.24 0.94 0.91 0.74
Pier 6 0.96 1.11 0.80 0.88 0.81 0.86 0.82 1.01 1.29 1.14
Pier 7 1.04 0.94 0.85 0.69 1.03 1.05 1.04 1.07 0.99 0.68
Pier 8 1.09 0.99 1.08 1.05 1.31 1.08 0.94 1.12 1.24 1.26
Pier 9 1.03 1.19 1.05 0.83 0.91 0.72 0.72 1.12 0.63 0.81
Pier 10 1.06 0.86 0.81 0.82 1.01 0.88 0.90 1.07 1.04 1.22
Pier 11 1.25 1.10 1.18 0.82 1.33 1.25 1.09 0.96 0.74 0.90
Pier 12 0.95 0.84 1.25 0.69 1.11 1.01 0.79 0.95 0.89 0.81

2250003-14
August 5, 2021 1:21:23pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

Table 1. (Continued )

Set 1 Set 2 Set 3 Set 4 Set 5 Set 6 Set 7 Set 8 Set 9 Set 10

Pier 13 0.73 0.79 1.05 0.65 0.86 0.60 0.90 1.09 0.88 0.94
Pier 14 0.93 0.81 1.17 0.75 1.04 0.92 0.84 0.78 1.47 0.89
Pier 15 1.49 1.02 0.93 0.91 1.08 0.97 0.90 1.06 0.93 1.07
Pier 16 1.04 1.08 0.96 1.00 1.07 1.10 0.88 0.77 1.00 0.99
Pier 17 1.14 0.71 0.72 0.68 0.80 0.68 1.06 0.89 0.85 0.71
Pier 18 0.85 1.24 1.12 0.66 0.91 0.71 0.97 0.90 0.85 0.96
Pier 19 0.72 0.86 1.03 0.87 0.86 0.77 0.90 0.85 0.78 1.09
Pier 20 1.15 1.01 1.04 1.14 0.99 0.99 1.22 0.91 1.05 1.25
Pier 21 1.22 0.82 0.83 0.81 0.76 0.85 0.71 0.77 0.78 0.76
Pier 22 1.00 1.04 1.14 0.87 1.19 0.76 1.03 0.85 1.00 0.99
Pier 23 1.31 0.95 1.33 1.05 0.95 1.09 1.32 1.26 1.24 1.23
Pier 24 1.20 0.84 0.80 0.60 0.86 0.92 1.02 0.88 0.82 0.75
Pier 25 0.86 0.62 0.81 0.78 0.76 0.77 0.96 0.76 0.83 0.69
Pier 26 0.65 1.38 0.77 0.92 1.24 1.32 0.88 1.04 0.93 0.91
Pier 27 0.76 0.63 0.89 0.84 0.82 0.81 1.03 1.26 0.87 0.79
Pier 28 0.85 1.04 0.91 0.81 1.43 0.77 1.20 0.85 0.95 0.69
Pier 29 0.72 0.76 0.58 0.85 0.62 0.60 0.73 0.56 0.53 0.54
Pier 30 1.47 1.84 1.69 0.67 1.31 1.25 0.93 0.93 1.10 0.78
Pier 31 0.70 1.03 0.57 0.73 0.86 0.80 0.59 0.80 0.71 0.72
Pier 32 1.14 1.26 1.44 0.94 1.48 1.09 1.33 0.88 1.13 1.12
Pier 33 0.88 0.71 0.69 0.63 0.57 0.65 0.48 0.69 0.67 0.48
Pier 34 1.40 0.90 1.14 0.93 0.98 0.81 1.05 0.87 0.84 0.98
Pier 35 0.93 0.82 0.70 0.90 0.80 0.76 1.01 1.19 0.92 0.81
Pier 36 0.92 1.32 0.84 0.89 0.68 0.61 0.77 1.10 0.92 0.97
Pier 37 1.01 1.03 1.03 1.31 1.29 0.99 1.51 0.90 0.86 0.89
Pier 38 1.19 1.02 0.71 0.88 0.65 1.43 0.93 0.84 1.01 1.00
Pier 39 1.55 1.36 0.75 0.75 0.82 0.89 0.94 0.89 0.97 0.71
Pier 40 0.92 1.00 1.09 0.88 1.23 0.97 0.81 0.86 0.90 0.67
Pier 41 1.04 1.15 0.94 1.19 1.04 1.22 0.92 1.02 0.90 1.06

Table 2. Ratio of maximum displacements ðDR Þ.

Set 11 Set 12 Set 13 Set 14

Pier 1 0.87 1.30 0.95 0.82


Pier 2 0.89 0.72 0.54 0.83
Pier 3 0.85 1.17 0.67 1.35
Pier 4 1.21 0.97 0.64 1.17
Pier 5 1.07 1.05 0.88 0.98
Pier 6 1.02 0.88 0.71 0.92
Pier 7 1.05 0.98 0.70 0.97
Pier 8 0.91 1.29 0.98 0.94
Pier 9 1.05 0.96 1.05 1.02
Pier 10 0.93 0.77 0.83 1.13
Pier 11 0.81 0.91 1.01 0.75
Pier 12 0.93 1.00 0.75 0.80
Pier 13 0.88 0.99 0.90 1.20
Pier 14 0.77 1.06 0.83 0.76
Pier 15 0.76 0.88 0.91 0.82
Pier 16 0.68 0.94 0.90 0.64

2250003-15
August 5, 2021 1:21:24pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

Table 2. (Continued )

Set 11 Set 12 Set 13 Set 14

Pier 17 0.76 0.96 0.73 1.19


Pier 18 0.76 0.76 0.65 0.81
Pier 19 0.80 0.77 0.89 0.79
Pier 20 0.56 0.95 0.70 0.79
Pier 21 1.12 0.83 0.78 0.87
Pier 22 0.68 0.88 1.02 1.00
Pier 23 0.56 0.72 0.65 0.91
Pier 24 1.30 1.00 0.68 1.34
Pier 25 0.83 0.66 0.72 0.60
Pier 26 0.85 0.69 0.60 0.78
Pier 27 0.65 0.85 0.59 0.84
Pier 28 0.73 0.94 0.78 1.04
Pier 29 0.95 0.79 0.70 0.74
Pier 30 0.78 0.71 0.60 0.70
Pier 31 0.59 0.70 0.69 1.18
Pier 32 1.02 0.78 0.58 0.63
Pier 33 1.00 0.77 0.63 0.80
Pier 34 0.95 0.65 1.42 0.88
Pier 35 0.84 0.82 0.76 0.75
Pier 36 0.75 0.92 0.81 0.98
Pier 37 0.97 1.31 0.89 1.15
Pier 38 0.86 0.97 0.62 0.67
Pier 39 0.87 0.84 0.84 0.84
Pier 40 0.52 0.48 0.47 0.77
Pier 41 1.19 1.08 0.87 1.52

(as high as 84%) than Case I and it is other way round in other piers (Case-II
underestimates as much as 52%). Pier associated with the maximum DR is also noted
as di®erent from one set to another.
Unlike the present problem, recorded data are not often available at multiple pier
locations from the same seismic event and hence, the bridge is considered again
subjected to the SVGMs generated above but without replacing the simulated data
by that recorded at the pier locations other than the reference station (i.e. pier- 11,
19, 21, 31 and 41). Four sets of SVGMs are considered for this purpose and the
associated DR is presented in Table 2 with maximum and minimum values
highlighted in bold. Observations and inferences remain the same.
To summarize, even though the resulting SVGMs do not di®er much in terms of
time history and spectral representation, the revised EPSD-based framework has a
stronger theoretical background than the EPSD-based framework as discussed in
Secs. 3.1 and 3.2. Precisely, the revised EPSD-based framework degenerates to the
EPSD-based framework [Rodda and Basu, 2020a] under three assumptions: (1)
mapping of EPSD and computation of spatial variation of random component are
achieved over the entire frequency range as opposed to the narrow frequency bands
(i.e. Lk ðuÞ ¼ L1 ðuÞ ¼ LðuÞÞ; (2) the arrival perturbation at any location is zero; and
3) the contribution from random phase is zero. Revised EPSD-based framework

2250003-16
August 5, 2021 1:21:24pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

developed in this paper is therefore recommended for simulating the nonstationary


SVGM.

4. Spatially Varying Nonstationary Vertical Ground Motion


Nonstationary vertical SVGM is generated through the revised EPSD-based
framework and assessed against the recorded vertical data. SVGM is also simulated
using the ASD-based framework by Rodda and Basu [2018b] and denoted as Case A
for comparison. While the results of revised EPSD-based framework is denoted as
Case B, that from the recorded event is de¯ned as the target.

4.1. Results and discussion


Target spectrum for the vertical components is constructed using the procedure
described in Sec. 3.3. Simulated response spectra (from Cases -A and -B) and the
associated target spectrum are compared in Fig. 9: spectra from Cases -A, -B and the
target spectrum are noted to be close.

(a) Bin 3500–4000 m Event 1 (b) Bin 0–500 m Event 2

(c) Bin 1000–1500 m Event 3 (d) Bin 2500–3000 m Event 4

Fig. 9. Assessment of response spectra from both the frameworks.

2250003-17
August 5, 2021 1:21:32pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

Table 3. Spectral contrast angles between the target and simulated spectra.

Event 1 Event 2 Event 3 Event 4

Bin Case-A Case-B Case-A Case-B Case-A Case-B Case-A Case-B

1 15 11 15 14 15 12 15 15
2 17 14 15 13 14 10 20 15
3 14 12 15 13 15 11 16 12
4 22 15 14 11 15 11 15 12
5 11 11 15 14 16 11 17 12
6 16 11 18 13 13 11 19 15
7 16 11 12 12 16 12 16 12
8 15 11 17 13 11 10 15 11

Spectral contrast angle based on distance correlation [Rodda and Basu, 2018c] is
employed here to quantify the similarity between two response spectra. A close to
zero spectral contrast angle indicates two nearly identical spectra whereas close to
90 illustrates no similarity. Based on a study of large database [Rodda and Basu,
2018c], a spectral contrast angle of 20 was reported to be the boundary di®erenti-
ating the similar and non-similar spectra. Spectral contrast angles between the target
and simulated (Case-A and Case-B) spectra are presented in Table 3: Case-A and
Case-B both exhibit close resemblance with the target spectrum for all practical
purposes. In other words, ASD-based framework [Rodda and Basu, 2018b] and the
revised EPSD-based framework both take into account the amplitude and phase
variabilities.
Simulating the nonstationary SVGM and assessing them against the recorded
accelerations is the primary objective in this paper. For this purpose, ground
accelerations generated using the revised EPSD-based framework and ASD-based
framework are compared with the recorded. Figure 10 presents some sample illus-
trations. Ground motions generated using the revised EPSD-based framework are
noted closer to the recorded data when compared with the ASD-based framework.
Therefore, unlike the ASD-based framework, the revised EPSD-based framework
accounts for the nonstationarity of recorded vertical ground motion. Possible e®ect
of the spatially varying vertical ground motion on structural response is next in-
vestigated in the following section.

4.2. E®ect of spatially varying vertical ground motion


on structural response
The same bridge described in Sec. 3.4 is considered here for the purpose of illustrating
the e®ect of vertical SVGM on structural response. Horizontal SVGM simulated in
Sets 13 and 14 (in Sec. 3.4) are considered here. Vertical SVGM is generated using
the revised EPSD-based framework at the interval of 100 m for two di®erent sets to
account for the inherent randomness. Response of the structure subjected to si-
multaneous horizontal and vertical SVGM is compared with that due to only

2250003-18
August 5, 2021 1:21:32pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

(a) Event 1

(b) Event 2

(c) Event 4

Fig. 10. Assessment of SVGM in time series from.

horizontal SVGM. Axial force of the piers and mid-span moment in deck are con-
sidered as the demand parameters of this comparison. One must note that, horizontal
and vertical SVGMs are simulated in the revised EPSD-based framework indepen-
dently, without accounting for any cross-directional correlation. Hence, four load
combinations (Table 4) are considered here for structural analysis to make any
meaningful comparison. Here, H1 and H2 in Table 4 denote the horizontal SVGM
from sets 13 and 14, respectively; V1 and V2 denote the two sets of vertical SVGM.

Table 4. Load combinations considered.

S. No Load combination Assessed against

C-I H1 +V1 H1
C-II H1 +V2 H1
C-III H2 +V1 H2
C-IV H2 +V2 H2

2250003-19
August 5, 2021 1:22:03pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

Table 5. Ratio of structural response.

Axial force Mid-span moment

C-I C-II C-III C-IV C-I C-II C-III C-IV

Pier 1 1.53 1.65 1.53 1.65 Span 1 1.67 1.83 1.66 1.82
Pier 2 1.49 1.72 1.49 1.72 Span 2 1.68 1.92 1.68 1.92
Pier 3 1.63 1.78 1.63 1.78 Span 3 1.94 1.98 1.94 1.97
Pier 4 1.90 1.85 1.90 1.85 Span 4 2.23 2.15 2.23 2.15
Pier 5 2.14 2.01 2.13 2.01 Span 5 2.33 2.34 2.33 2.34
Pier 6 2.15 2.15 2.15 2.15 Span 6 2.46 2.42 2.46 2.42
Pier 7 2.28 2.26 2.28 2.26 Span 7 2.50 2.40 2.50 2.40
Pier 8 2.38 2.21 2.38 2.21 Span 8 2.58 2.57 2.58 2.57
Pier 9 2.36 2.41 2.36 2.40 Span 9 2.63 2.59 2.63 2.59
Pier 10 2.41 2.35 2.41 2.35 Span 10 2.69 2.60 2.69 2.60
Pier 11 2.50 2.42 2.50 2.41 Span 11 2.73 2.68 2.73 2.68
Pier 12 2.51 2.44 2.51 2.44 Span 12 2.80 2.75 2.79 2.75
Pier 13 2.59 2.48 2.58 2.48 Span 13 2.81 2.85 2.80 2.85
Pier 14 2.59 2.67 2.59 2.67 Span 14 2.94 2.98 2.94 2.97
Pier 15 2.56 2.58 2.56 2.58 Span 15 2.86 2.85 2.85 2.84
Pier 16 2.68 2.60 2.68 2.60 Span 16 3.04 3.02 3.04 3.02
Pier 17 2.70 2.69 2.70 2.68 Span 17 2.96 3.22 2.96 3.22
Pier 18 2.75 2.86 2.75 2.86 Span 18 3.10 2.97 3.10 2.96
Pier 19 2.74 2.83 2.74 2.83 Span 19 2.99 3.24 2.99 3.24
Pier 20 2.73 2.82 2.73 2.82 Span 20 3.12 3.19 3.12 3.19
Pier 21 2.85 2.80 2.85 2.80 Span 21 3.12 3.12 3.12 3.11
Pier 22 2.80 2.81 2.80 2.81 Span 22 3.19 3.05 3.19 3.05
Pier 23 2.80 2.74 2.80 2.74 Span 23 3.00 3.08 2.99 3.08
Pier 24 2.76 2.71 2.76 2.71 Span 24 2.97 3.01 2.97 3.00
Pier 25 2.68 2.72 2.68 2.72 Span 25 3.01 3.08 3.01 3.07
Pier 26 2.64 2.61 2.63 2.61 Span 26 2.87 2.87 2.87 2.86
Pier 27 2.52 2.47 2.51 2.47 Span 27 2.81 3.04 2.81 3.03
Pier 28 2.54 2.72 2.53 2.72 Span 28 2.74 3.12 2.74 3.12
Pier 29 2.59 2.57 2.59 2.57 Span 29 2.94 2.95 2.94 2.95
Pier 30 2.64 2.73 2.64 2.72 Span 30 2.94 2.93 2.94 2.92
Pier 31 2.49 2.76 2.48 2.76 Span 31 2.83 3.32 2.83 3.32
Pier 32 2.59 2.82 2.59 2.82 Span 32 2.75 2.91 2.75 2.91
Pier 33 2.57 2.56 2.57 2.56 Span 33 2.78 2.79 2.78 2.79
Pier 34 2.45 2.41 2.45 2.41 Span 34 2.59 2.60 2.59 2.60
Pier 35 2.35 2.45 2.35 2.45 Span 35 2.52 2.78 2.52 2.78
Pier 36 2.19 2.32 2.19 2.32 Span 36 2.24 2.49 2.24 2.49
Pier 37 2.01 2.23 2.01 2.23 Span 37 2.34 2.48 2.34 2.48
Pier 38 2.01 2.22 2.01 2.22 Span 38 1.98 2.48 1.98 2.47
Pier 39 1.79 1.93 1.79 1.93 Span 39 1.97 2.12 1.97 2.12
Pier 40 1.68 1.97 1.68 1.97 Span 40 1.75 2.18 1.75 2.18
Pier 41 1.59 1.93 1.59 1.93

Axial forces at individual piers subjected to (1) combined action of horizontal and
vertical SVGM and (2) only horizontal SVGM are noted. Ratio of peak axial forces in
two cases at each pier is computed for all the load combinations (Table 4) and is
reported in Table 5. The ratio exceeds unity in most of the pier and is as high as 2.86.
In other words, inclusion of vertical SVGM leads to increase of axial load in pier as

2250003-20
August 5, 2021 1:22:03pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

high as 186%. Similar comparison for the midspan moment in the deck is also in-
cluded in Table 5 and the ratio is noted as high as 3.32. Hence, inclusion of vertical
SVGM leads to an increase of span moment in deck as high as 232%. Although these
results are contingent on the idealized example considered here for illustration, the
spatial varying vertical ground motion is expected to contribute signi¯cantly to the
seismic design of long-span bridges and hence, a reliable framework for simulating
the ground motion is of paramount importance.

5. Summary and Conclusions


Rodda and Basu [2020a] reported a state-of-the-art evolutionary power spectral
density (EPSD) based framework for simulating the nonstationary spatially varying
ground motion (SVGM) ¯eld. This framework models the spatial variability through
a decay function and a random component and is based on two strong assumptions:
(1) spatial variation of the random component is assumed to be independent of the
frequency; and (2) phase of the ground excitation simulated around the reference
station (with seed motion) is assumed to be same as that of the seed motion.
This paper investigates the impact of these two assumptions on the simulated
ground motion ¯eld and proposes a revised EPSD-based framework for the possible
improvements. First, spatial variation of the random component is noted to be
varying signi¯cantly from one (narrow) frequency band to another. Hence, the re-
vised EPSD-based framework proposes the mapping of EPSD over the individual
frequency bands rather than over the entire usable frequency range. Next, phase of
the ground motion across the footprint of the array is noted to be signi¯cantly
di®erent than that of the seed motion. Hence, the revised EPSD-based framework
proposes a phase-structure accounting for the variability in phase over the spatial
footprint. Proposed phase-structure requires four inputs: (1) phase of the ground
motion at reference station; (2) arrival time perturbation; (3) a parameter ao , which
is derived from the constituting functions of the coherency model; and (4) a uni-
formly distributed random phase.
Framework by Rodda and Basu [2020a] and the revised EPSD-based framework
proposed in this paper produce nearly identical horizontal SVGM for all practical
purposes when assessed in terms of magnitude (i.e. peak ground acceleration) and
spectral representation. However, phase of the ground motion ¯eld is signi¯cantly
di®erent in both the frameworks. Possible e®ect of this phase-structure cannot be
observed by comparing either the response spectrum or the acceleration time series at
one location. The e®ect of horizontal SVGM simulated from both the frameworks,
namely, Case-I [Rodda and Basu, 2020a] and Case-II (revised EPSD), is investigated
using an idealized long span bridge. Ten sets of SVGM ¯eld are simulated for the
individual framework to account for the randomness involved in the coherency cal-
ibration and phase-structure. Ratio of maximum displacements between Case-II and
Case-I is noted to be signi¯cantly di®erent than unity (0.52–1.84). The pier a®ected
with maximum variation is also noted as di®erently from one set to another. Revised

2250003-21
August 5, 2021 1:22:03pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

EPSD-based framework is therefore recommended over that reported by Rodda and


Basu [2020a] due to stronger theoretical background.
Revised EPSD-based framework is next assessed against the vertical recordings of
SMART1 array along with the auto-spectral density (ASD) framework reported by
Rodda and Basu [2018b]. Though spectral representation is nearly identical in both
the frameworks, the acceleration time series simulated using the revised EPSD-based
framework matches the recorded data better when compared with the ASD-based
framework. This indicates that the revised EPSD-based framework accounts for the
nonstationarity along with the amplitude and phase variabilities. Possible e®ect of
spatially varying vertical ground motion on the seismic design is investigated
through the same idealized bridge model. Signi¯cant increase in the demand of axial
force in piers and mid-span moment in the deck are observed. Although these results
reported here are contingent on the idealized example considered for illustration, the
spatially varying vertical ground motion are expected to contribute signi¯cantly to
the seismic design of long-span bridges and hence, a reliable framework for simulating
the ground motion such as that proposed in this paper is of paramount importance.

Acknowledgment
This research is funded by MoES, Government of India, under Grant No. MoES/P.
O.(Seismo)/1(370)/2019 and the ¯nancial support is acknowledged.

Declarations
Code availability: Standard licenced software is used. Custom code is developed in
MATLAB environment and not available for sharing.
Authors' contributions: DB proposed the idea, interpret the results, prepared the
¯nal draft, acquire the funding, and manage the overall research. GR processed the
data, analyzed with custom code, generated and interpreted results, and prepared
the ¯rst draft. NG analyzed with software, generated and interpreted results, and
assisted in preparing the ¯rst draft.

2250003-22
August 5, 2021 1:22:03pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

Fig. A.1. Displacement contours at all piers.


Appendix A

2250003-23
August 5, 2021 1:22:45pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

Fig. A.1. (Continued )

2250003-24
August 5, 2021 1:23:27pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

Fig. A.1. (Continued )

2250003-25
August 5, 2021 1:24:11pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

Fig. A.1. (Continued )

2250003-26
August 5, 2021 1:24:54pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

Fig. A.1. (Continued)

2250003-27
August 5, 2021 1:25:36pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

Fig. A.1. (Continued )

2250003-28
August 5, 2021 1:25:42pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

References
Abrahamson, N. A., Schneider, J. F. and Stepp, J. C. [1991] \Empirical spatial coherency
functions for application to soil-structure interaction analyses," Earthq. Spectra 7(1),
1–27.
Adanur, S., Altunişik, A. C., Soyluk, K., Bayraktar, A. and Dumanoğlu, A. A. [2016]
\Multiple-support seismic response of bosporus suspension bridge for various random
vibration methods," Case Stud. Struct. Eng. 5, 54–67.
Amin, M. and Ang, A. H. S. [1968] \Nonstationary stochastic models of earthquake motions,"
J. Eng. Mech. Div. 94(2), 559–584.
Bi, K. and Hao, H. [2012] \Modelling and simulation of spatially varying earthquake ground
motions at sites with varying conditions," Probabil. Eng. Mech. 29, 92–104.
Bolotin, V. V. [1960, July] \Statistical theory of the aseismic design of structures," in Proc.
2nd World Conf. Earthquake Engineering, Tokyo, Vol. 2, pp. 1365–1374.
Bycroft, G. N. [1960] \White noise representation of earthquakes," J. Eng. Mech. Div. 86(2),
1–16.
Cacciola, P. and Deodatis, G. [2011] \A method for generating fully nonstationary and
spectrum-compatible ground motion vector processes," Soil Dyn. Earthq. Eng. 31(3),
351–360.
Deodatis, G. [1996] \Non-stationary stochastic vector processes: Seismic ground motion
applications," Probabil. Eng Mech. 11(3), 149–167.
Der Kiureghian, A. and Crempien, J. [1989] \An evolutionary model for earthquake ground
motion," Struct. Safety 6(2–4), 235–246.
Di Paola, M. and Zingales, M. [2000] \Digital simulation of multivariate earthquake ground
motions," Earthq. Eng. Struct. Dyn. 29(7), 1011–1027.
Efthymiou, E. and Camara, A. [2017] \E®ect of spatial variability of earthquakes on cable-
stayed bridges," Proc. Eng. 199, 2949–2954.
Hao, H. [1996] \Characteristics of torsional ground motions," EESD 25(6), 599–610.
Hao, H., Oliveira, C. S. and Penzien, J. [1989] \Multiple-station ground motion processing and
simulation based on SMART-1 array data," Nucl. Eng. Des. 111(3), 293–310.
Harichandran, R. S. and Wang, W. [1990] \Response of indeterminate two-span beam to
spatially varying seismic excitation," Earthq. Eng. Struct. Dyn. 19(2), 173–187.
Harichandran, R. S. and Vanmarcke, E. H. [1986] \Stochastic variation of earthquake ground
motion in space and time," J. Eng. Mech. 112(2), 154–174.
Harichandran, R. S., Hawwari, A. and Sweidan, B. N. [1996] \Response of long-span bridges to
spatially varying ground motion," J. Struct. Eng. 122(5), 476–484.
Housner, G. W. [1955] \Properties of strong ground motion earthquakes," Bull. Seismol. Soc.
Am. 45(3), 197–218.
Hu, L., Xu, Y. L. and Zheng, Y. [2012] \Conditional simulation of spatially variable
seismic ground motions based on evolutionary spectra," Earthq. Eng. Struct. Dyn. 41(15),
2125–2139.
Huang, G. [2014] \An e±cient simulation approach for multivariate nonstationary
process: Hybrid of wavelet and spectral representation method," Probabil. Eng. Mech. 37,
74–83.
IS-1893 [2014] Indian standard criteria for earthquake resistant design of Structures, Part 3:
Bridges and retaining walls, in Bureau of Indian Standards, New Delhi.
Jennings, P. C., Housner, G. W. and Tsai, N. C. [1968] Simulated earthquake motions.
Technical Report, Earthquake Engineering Research Laboratory, California Institute of
Technology, Pasadena, CA.
Kameda, H. [1975] \Evolutionary spectra of seismogram by multi¯lter," J. Eng. Mech. Div.
101(6), 787–801.

2250003-29
August 5, 2021 1:25:42pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

G. K. Rodda, N. Gurjar & D. Basu

Kameda, H. and Morikawa, H. [1992] \An interpolating stochastic process for simulation of
conditional random ¯elds," Probabil. Eng. Mech. 7(4), 243–254.
Konakli, K. and Der Kiureghian, A. [2012] \Simulation of spatially varying ground motions
including incoherence, wave-passage and di®erential site-response e®ects," Earthq. Eng.
Struct. Dyn. 41(3), 495–513.
Li, Y. and Kareem, A. [1993] \Simulation of multivariate random processes: Hybrid DFT and
digital ¯ltering approach," J. Eng. Mech. 119(5), 1078–1098.
Liu, S. C. [1970] \Evolutionary power spectral density of strong-motion earthquakes," Bull.
Seismol. Soc. Am. 60(3), 891–900.
Loh, C. H. [1985] \Analysis of the spatial variation of seismic waves and ground movements
from smart-1 array data," Earthq. Eng. Struct. Dyn. 13(5), 561–581.
Loh, C. H. and Yeh, Y. T. [1988] \Spatial variation and stochastic modelling of seismic
di®erential ground movement," Earthq. Eng. Struct. Dyn. 16(4), 583–596.
Luco, J. E. and Wong, H. L. [1986] \Response of a rigid foundation to a spatially random
ground motion," Earthq. Eng. Struct. Dyn. 14(6), 891–908.
Lupoi, A., Franchin, P., Pinto, P. E. and Monti, G. [2005] \Seismic design of bridges ac-
counting for spatial variability of ground motion," Earthq. Eng Struct. Dyn. 34(4–5),
327–348.
Preumont, A. [1985] \The generation of non-separable arti¯cial earthquake accelerograms for
the design of nuclear power plants," Nucl. Eng. Des. 88(1), 59–67.
Rodda, G. K. and Basu, D. [2018a] \Coherency model for translational and rotational ground
motions," Bull. Earthq. Eng. 16(7), 2687–2710.
Rodda, G. K. and Basu, D. [2018b] \Spatial variation and conditional simulation of seismic
ground motion," Bull. Earthq. Eng. 16(10), 4399–4426.
Rodda, G. K. and Basu, D. [2018c] \Apparent translational component for rotational ground
motions," Bull. Earthq. Eng. 16(1), 67–89.
Rodda, G. K. and Basu, D. [2020a] \A novel framework for conditional simulation of fully-
nonstationary spatially varying ground motion ¯eld," Earthq. Eng. Struct. Dyn.,
doi: 10.1002/eqe.3343.
Rodda, G. K. and Basu, D. [2020b] \Spatially correlated vertical ground motion for seismic
design," Eng. Struct. 206, 110191.
Rosenblueth, E. [1956, June] \Some applications of probability theory in aseismic design," in
World Conf. Earthquake Engineering, Berkeley, California.
Sarkar, K., Gupta, V. K. and George, R. C. [2016] \Wavelet-based generation of spatially
correlated accelerograms," Soil Dyn. Earthq. Eng. 87, 116–124.
Sgobba, S., Sta®ord, P. J. and Marano, G. C. [2011] \A seismologically consistent husid
envelope function for the stochastic simulation of earthquake ground-motions," in
Computational Methods in Stochastic Dynamics. Springer, Dordrecht, pp. 229–246.
Shields, M. D. [2015] \Simulation of spatially correlated nonstationary response spectrum–
compatible ground motion time histories," J. Eng. Mech. 141(6), 04014161.
Shinozuka, M. and Sata, Y. [1967] \Simulation of nonstationary random process," J. Eng.
Mech. Div. 93(1), 11–40.
Shinozuka, M. and Zhang, R. [1996] \Equivalence between Kriging and CPDF methods for
conditional simulation," J. Eng. Mech. 122(6), 530–538.
Shrikhande, M. and Gupta, V. K. [1998] \Synthesizing ensembles of spatially correlated
accelerograms," J. Eng. Mech. 124(11), 1185–1192.
Vanmarcke, E. H. and Fenton, G. A. [1991] \Conditioned simulation of local ¯elds of earth-
quake ground motion," Struct. Safety 10(1–3), 247–264.
Vanmarcke, E. H., Heredia-Zavoni, E. and Fenton, G. A. [1993] \Conditional simulation of
spatially correlated earthquake ground motion," J. Eng. Mech. 119(11), 2333–2352.

2250003-30
August 5, 2021 1:25:42pm WSPC/238-JET 2250003 ISSN: 1793-4311 2nd Reading

Phase-Structure in Conditional Simulation of Spatially Varying Ground Motion

Wang, D., Wang, L., Xu, J., Kong, F. and Wang, G. [2019] \A directionally-dependent
evolutionary lagged coherency model of nonstationary horizontal spatially variable seis-
mic ground motions for engineering purposes," Soil Dyn. Earthq. Eng. 117, 58–71.
Wen, Y. K. and Gu, P. [2004] \Description and simulation of nonstationary processes based on
Hilbert spectra," J. Eng. Mech. 130(8), 942–951.
Wu, Y., Gao, Y. and Li, D. [2011] \Simulation of spatially correlated earthquake ground
motions for engineering purposes," Earthq. Eng. Eng. Vibr. 10(2), 163.
Yao, E., Miao, Y. and Wang, G. [2017] \Synthesis of spatially correlated earthquake ground
motions based on Hilbert transform," Model. Simul. Eng. 2017, Article ID 2614769, 1–9.
Zeldin, B. A. and Spanos, P. D. [1996] \Random ¯eld representation and synthesis using
wavelet bases," J. Appl. Mech. 63(4), 946–952.
Zerva, A. [1994] \On the spatial variation of seismic ground motions and its e®ects on life-
lines," Eng. Struct. 16(7), 534–546.
Zerva, A. [2016] Spatial Variation of Seismic Ground Motions: Modeling and Engineering
Applications. CRC Press.
Zerva, A. and Harada, T. [1997] \E®ect of surface layer stochasticity on seismic ground motion
coherence and strain estimates," Soil Dyn. Earthq. Eng. 16(7–8), 445–457.
Zerva, A. and Zervas, V. [2002] \Spatial variation of seismic ground motions: An overview,"
Appl. Mech. Rev. 55(3), 271–297.
Zhang, D. Y., Liu, W., Xie, W. C. and Pandey, M. D. [2013] \Modeling of spatially correlated,
site-re°ected, and nonstationary ground motions compatible with response spectrum,"
Soil Dyn. Earthq. Eng. 55, 21–32.

2250003-31

You might also like