You are on page 1of 14

Soil Dynamics and Earthquake Engineering 32 (2012) 42–55

Contents lists available at SciVerse ScienceDirect

Soil Dynamics and Earthquake Engineering


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

Simplified seismic analysis of soil–well–pier system for bridges


Goutam Mondal a, Amit Prashant b, Sudhir K. Jain b,n,1
a
Department of Civil Engineering, Indian Institute of Technology Kanpur, UP 208016, India
b
Department of Civil Engineering, Indian Institute of Technology Gandhinagar, GJ 382424, India

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

Article history: Seismic analysis of soil–well–pier system was carried out using three different approaches to evaluate
Received 15 February 2011 their comparative performance and associated complexities. These approaches were (a) two-dimen-
Received in revised form sional nonlinear (2D-NL), (b) two-dimensional equivalent-linear (2D-EqL), and (c) one-dimensional
21 July 2011
spring–dashpot (1D). Soil was modeled as 2D plane-strain elements in the 2D-NL and 2D-EqL
Accepted 8 August 2011
approaches, and as springs and dashpots in the 1D approach. Nonlinear behavior of soil was captured
Available online 13 September 2011
rigorously in the 2D-NL approach and approximately in the remaining two approaches. Results of the
two approximate analyses (i.e., 2D-EqL and 1D) were compared with those of the 2D-NL analysis with
the objective to assess suitability of approximate analysis for practical purposes. In the 1D approach,
several combinations of Novak’s and Veletsos’ springs were used to come up with a simplified 1D
model using three types of spring–dashpots. The proposed model estimates the displacement and force
resultants relatively better than the other 1D models available in literature.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction of the 1D approach over the 2D is that the former is very efficient
and easy to implement in design offices than the latter. However,
Well foundations (also known as caisson foundation) are most of the 1D spring–dashpot models available in the literature are
frequently adopted in the Indian subcontinent and other coun- intended mainly to the flexible pile foundation or shallow founda-
tries like Japan, USA, Thailand, etc., for the deep foundation of tion except for a few of these models [1–3], which were specifically
railway and highway bridges on rivers. Because of its large cross developed for relatively rigid embedded structures like well founda-
section and high rigidity, such foundations are often believed to tion. Japanese code [3] specified a Winkler’s spring model with six
be safe foundation systems against earthquake. However, it was types of springs for the stability check of well foundation. However,
observed during recent earthquakes that structures supported on since damping is not considered (in absence of dashpots), this model
such foundations also suffered damage during moderate to severe often overestimates the force and displacement responses of well
earthquakes mainly because of the large permanent displacement and pier. Gerolymos and Gazetas [1] proposed 1D model with four
associated with soil liquefaction. For example, many structures types of spring–dashpots considering soil and interface nonlinearity.
supported on well foundations sustained severe damage during However, since this model comprehensively considers the issues
1995 Kobe earthquake. In India, many such bridges are located in related to soil–structure interaction, it requires many parameters
the high seismic region where moderate to severe earthquakes and therefore it may not be suitable for practical purposes. Varun
are expected to occur. Therefore, seismic analysis should be et al. [2] proposed a simplified model with three types of spring–
performed for the design of new bridges and retrofitting of the dashpots for well foundation embedded in linear soil. In this
existing bridges supported on well foundation. formulation, material damping in soil was not considered. All the
In practice, well foundation is analyzed by modeling soil as two- simplified models (e.g., 1D and 2D-EqL) have some limitations,
dimensional (2D) plane-strain element or one-dimensional (1D) which stem from the assumptions associated with them. It is of
spring–dashpot (also named as 1D model herein). However, in these interest to see if these models are suitable to analyze well founda-
analyses, effect of soil nonlinearity is generally ignored or some- tion embedded in saturated cohesionless soil susceptible to liquefac-
times considered approximately. Out of these two approaches, 1D tion. For this purpose, three types of approaches were considered:
approach is widely used in practical purposes. The main advantage (a) two-dimensional nonlinear (2D-NL), (b) two-dimensional
equivalent-linear (2D-EqL), and (c) one-dimensional spring–dashpot
(1D). Results of both the approximate approaches (2D-EqL and 1D)
n
Corresponding author. Tel.: þ91 7923972574; fax: þ91 7923972622.
were compared with those of the rigorous one (i.e., 2D-NL) to study
E-mail address: skjain.iitk@gmail.com (S.K. Jain). the performance and the computational efficiency of the approximate
1
Formerly Professor, Indian Institute of Technology Kanpur, UP 208016, India. approaches. In the 1D approach, various combinations of Novak’s [4]

0267-7261/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2011.08.002
G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55 43

Table 1
Superstructure Parameters for constitutive model.

Parameters Layer 1 Layer 2 Layer 3

Pier Depth 0 m–20 m 20 m–50 m 50 m–100 m


Type of soil Medium sand Medium-dense sand Dense sand
11.35m Unit weight (t/m3) 1.9 2.0 2.1
1 1 Poisson’s ratio 0.33 0.35 0.35
G.L. 5.3 m Gra (kN/m2) 7.5  104 1.0  105 1.3  105
fb 331 371 401
gmax c 0.1 0.1 0.1
50 m Well 3.3 m fPTd 271 271 271
Contrace 0.07 0.05 0.03
Dilat1f 0.4 0.6 0.8
Dilat2f 2 3 5
Liquefac1g (kN/m2) 10 5 0
2 2 Liquefac2h 0.01 0.003 0
18 m
Liquefac3i 1 1 0

Direction of Note:
Earthquake a
Motions Gr is the reference shear modulus specified at confining pressure of
11 m
80 kN/m2.
b
f is the angle of internal friction.
c
gmax is the octahedral shear strain at which the maximum shear strength is
Fig. 1. Geometry of the well foundation and piers analyzed in the present study.
reached.
d
fPT is the phase transformation angle.
e
Contrac is a non-negative constant defining the rate of shear induced volume
and Veletsos’ [5,6] springs were considered and parametric study was contraction
performed to propose a 1D model with three types of spring– f
Dilat1 and Dilat2 are non-negative constants defining the rate of shear-
dashpots by simplifying expressions of Novak’s spring coefficients. induced volume increase.
g
Liquefac1 is the pressure below which cyclic mobility takes place.
h
Liquefac2 is the maximum amount of perfectly plastic shear strain developed
at zero effective confinement during each loading phase.
2. Geometry of the model and earthquake motions i
Liquefac3 is the maximum amount of biased perfectly plastic shear strain
accumulated at each loading phase under biased shear loading conditions.
Fig. 1 shows the geometry of the bridge substructure system
and cross-sectional dimensions of its components. The bridge
substructure system consisted of a typical double-D cellular well by 2D plane-strain representation. By comparing the response of a
foundation supporting two hollow circular piers on its top. The rigid shallow foundation of circular shape and a strip footing placed
height of each pier was 13.47 m, and the depth of well foundation on elastic half-space as obtained from the 2D and 3D analyses, they
was 50 m. Load on the pier cap due to bridge deck and live load demonstrated that it was possible to obtain approximately the
was 3000 kN. The soil profile considered in the present study natural frequencies of the system, but the radiation damping
consisted of three layers of cohesionless soil with bedrock under- associated with the low-frequency modes was significantly over-
neath. The bedrock was assumed at 100 m depth from the ground estimated, which may underestimate the response parameters.
surface. The properties of soil in these layers are shown in Table 1. Watanabe and Tochigi [9] simulated the shaking table test on
Such a bridge substructure and soil system is considered to be soil–foundation–structure system using 2D FE model and demon-
typical for bridges in alluvial rivers in India and believed to be strated that the 2D FE analysis could represent the main features of
representative of similar such structures. Because of the uncer- the system when mass and stiffness of the superstructure and
tainty associated with the design ground motion, variation of substructure were divided by the width of the foundation perpen-
ground motions was preferred over that of geometry of the dicular to the direction of excitation. Seed and Lysmer [10] com-
foundation and properties of soil layers. Therefore, nine earth- pared the seismic response of a nuclear power plant by modeling it
quake motions recorded at different geographical locations were using 2D and 3D FE formulations. They showed that the 2D analysis
selected for the seismic analysis (Table 2). These ground motions underestimates the response of superstructure and substructure.
represent different source mechanisms and epicentral distances. However, all the above studies were for shallow foundations
These were recorded at ground level at rock-outcrop as free-field resting on linear and elastic soil. Therefore, further research is
motions during strong earthquakes with magnitude 6.5 and needed to verify these limitations for deep foundations and
above. The ground motions with PGA ranges from 0.1 g to 0.3 g, inelastic soil. Despite the best effort, a similar study comparing
0.3 g to 0.5 g, and 0.5 g to 0.7 g were scaled to PGA values of 0.2 g, the response of 2D versus 3D modeling for deep foundation with
0.4 g, and 0.6 g, respectively. These three sets of ground motions inelastic soil could not be found in the literature. Moreover,
were termed as low, moderate, and severe ground motions, as seismic analysis of a 3D FE model of the SWP system with soil
shown in Table 2. Seismic motion was assumed to be along the nonlinearity is very cumbersome and computationally expensive.
direction of traffic (i.e., in the longitudinal direction). Jeremic et al. [11] performed seismic analysis of 3D FE model of a
pile-supported viaduct considering linear–elastic soil. Gerolymos
and Gazetas [12] performed 3D analysis of a small size circular
3. Two-dimensional (2D) nonlinear analysis well foundation (depth 6 m and diameter 3 m) considering soil
nonlinearity; however, coarse mesh was used since they did static
Seismic analysis of the soil–well–pier (SWP) system is inherently and dynamic analyses by applying sinusoidal motion at the top of
a three-dimensional (3D) problem. It is known that 2D modeling the well. Recently, Elgamal et al. [13] performed seismic analysis
underestimates the dynamic-spring coefficient and overestimates of a bridge supported on pile foundation using a ‘‘pilot 3D FE’’
the radiation damping [7]. Luco and Hadjian [8] studied the model considering nonlinear soil and with coarse mesh and the
feasibility of modeling 3D linear soil-structure-interaction problem analysis took 40 h to complete. Therefore, considering the
44 G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55

Table 2
Ground motions considered for the seismic analysis of soil–well–pier system.

Level Earthquake Mwa Station PGA (g) Component Epicentral distance (km) Symbol for ground motion

Low Northridge, 1994 6.7 LA-Griffith Park Observatory, (USGS 141) 0.289 270 25.4 L1
Turkey, 1999 7.1 Lamont 531 0.117 90 27.7 L2
Uttarkashi, 1991 7.0b Bhatwari, India 0.246 355 19.3 L3
Moderate Chi-Chi, 1999 7.6 CWB 99999 CHY041 0.302 90 51.2 M1
Loma Prieta, 1989 6.9 Corralitos (CDMG 57007) 0.479 90 7.2 M2
Northridge, 1994 6.2 UCSB 99999 LA 00 0.388 90 14.4 M3
Severe Chi-Chi, 1999 7.6 CWB 99999 TCU071 0.655 0 15.4 S1
Northridge, 1994 6.7 DWP 77 Rinaldi Receiving Station 0.633 318 10.9 S2
Petrolia, 1992 7.8 CGS-89156 0.685 90 4.5 S3

Note:
a
Mw ¼ Moment magnitude.
b
Surface wave magnitude (Ms).

Pier plane-strain condition. Well foundation and piers were discretized


Rigid massless using two-noded linear beam-column elements with three DOF at
outrigger each node. Massless rigid beam-column elements were added to
0 the embedded part of the well foundation to account for the width
Well of the well foundation along the earthquake motion. Interface
nonlinearity (i.e., gapping and sliding at the soil–well interface)
was ignored since it does not have significant influence in the
-25
design force resultants and displacements of well foundation [15].
In other words, the soil–well interfaces were assumed to be
perfectly-bonded. Viscous boundary proposed by Lysmer and
-50 Kuhlemeyer [16] was used as radiation boundary at the two vertical
boundaries and at the base of the FE model. Ground motion was
applied at the base of the FE model in the form of equivalent shear
force proportional to velocity of incident wave motion [17]. Seismic
-75
Radiation analysis of the SWP system was performed in open source code
Y
boundary OpenSees, the Open System for Earthquake Engineering Simulation,
Z X developed specially to simulate the performance of structural and
100 geotechnical systems subjected to earthquake motions [18].
0 25 50 75 100 125 150 In two-dimensional nonlinear (2D-NL) analysis, the soil
v3 v2 response was simulated using elasto-plastic pressure-dependent
3 multi-yield surface (nested-yield surface) constitutive model
u3 2 u
2 (‘‘PressureDependMultiYield’’ model) available in OpenSees
v1 v2 [18–20]. In this model, a set of Drucker–Prager nested yield
u4 u1 θ1 θ2
1 u1 u2 surfaces with a common apex and different sizes were used to
4
v4 v1 1 2 simulate nonlinear behavior of undrained cohesionless soil. The
values of the parameters required for the constitutive model are
Fig. 2. Details of the finite element model of soil–well–pier (SWP) system. taken from the table given in the user’s manual [19]. Soil damping
(a) Finite element discretization of the SWP system, (b) Plane-strain element for was primarily captured through hysteretic energy dissipation,
soil, (b) Plane-strain element for at the boundaries and (d) Beam-column element
for well and pier.
and therefore, no other damping was considered.
The numerical model in OpenSees was validated with
SHAKE2000 [21] to check the adequacy of the mesh size and
complexity of analyzing rigorous 3D FE model, 2D representation time step considered for the seismic analysis, and to verify the
of the foundation soil was used in the present study by assuming effectiveness of Lysmer–Kuhlemeyer (L–K) radiation boundaries.
plane-strain condition. Further, this 2D model is able to capture, Since SHAKE2000 can perform only free-field analysis (i.e., only
at least qualitatively, the key aspects of the effects of soil soil layers in absence of structure), the FE model in OpenSees was
nonlinearity (liquefaction) on the overall seismic response modified by excluding the piers and replacing the well foundation
mechanism of the SWP system. One may note that even though by soil elements. Verification was carried out under the north–
this 2D model may have somewhat underestimated the seismic south component of the horizontal acceleration (PGA¼0.314 g)
response of pier and well, it has captured many important aspects recorded at El Centro during 1940 Imperial Valley earthquake
of the seismic behavior. (Fig. 3a). Effectiveness of the L–K boundary was verified by
Fig. 2 shows the FE discretization of half of the SWP system. In assuming soil to be linear–elastic and undamped to check the
the dynamic behavior of a bridge substructure, the superstructure performance of L–K boundary in the worst scenario. Further
stiffness does not contribute significantly [14], and hence only the details of the verification can be found elsewhere [22].
mass of the superstructure was applied at the pier cap. Hydro- Fig. 3b–d shows comparison of the acceleration response
dynamic mass was not considered in the present study. The mass histories obtained from OpenSees and SHAKE2000 at ground
of water and sand inside the well foundation was though surface, 25 m depth and 50 m depth, respectively. The accelera-
incorporated. In the 2D FE model, soil domain was discretized tion response histories obtained from both the procedures
using four-noded, bilinear, isoparametric finite elements with two matched satisfactorily throughout the duration of the motion.
degrees of freedom (DOF) at each node. Soil was analyzed under However, peak horizontal acceleration (PHA) values were
G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55 45

0.8

Acceleration (g)
Input Motion
0.4
0
-0.4 PGA = 0.319g
-0.8
0 5 10 15 20 25 30
Time (s)
0.8 Peak Horizontal Acceleration (g)
Acceleration (g)

At 0 m
0.4 0 0.2 0.4 0.6 0.8
0
0
-0.4 -10

-0.8 -20
0.8
Acceleration (g)

At -20 m -30
0.4
0 -40

Depth (m)
-0.4 -50
-0.8
-60
0.8
Acceleration (g)

At -50 m
0.4 -70

0 -80
-0.4
-90
-0.8
0 5 10 -100
Time (s) OpenSees
SHAKE

Fig. 3. Comparison of acceleration response of soil obtained from free-field analysis in SHAKE and OpenSees for the El-Centro motion (a) input motion (b)–(d) acceleration
response histories at specific points (e) variation of peak horizontal acceleration with depth.

underestimated in OpenSees by 6%, 1%, and 2%, respectively, at were zero-length elements with two nodes; one node was
ground surface, 25 m depth and 50 m depth. Hence, such small connected to the boundary node and the other node was fixed
variations in the peak responses were considered acceptable. in space. Finally, seismic analysis of the SWP system was per-
Moreover, the variation of PGA along depth as obtained from formed by applying horizontal seismic excitation in the form of
the two models also matched well (Fig. 3e). Therefore, it was effective nodal forces applied at the base of the computational soil
concluded that dimensions of the finite elements and time step domain.
considered during the seismic analysis were adequate, and the The six response parameters, namely, the maximum displace-
L–K boundaries were effective in eliminating reflection of spur- ment at pier top (DPT) and well top (DWT), the maximum shear force
ious waves. in pier (SFP) and well (SFW), and the maximum bending moment in
Analysis was performed in several steps to simulate the initial pier (BMP) and well (BMW) were obtained from the 2D-NL analysis
condition of the SWP model. In the first step, gravity analysis was and compared with those obtained from the 2D-EqL and 1D
performed for the self-weight of soil and weight of the embedded analyses, as discussed in the subsequent sections to evaluate the
portion of well by assuming the soil as linear–elastic. In this step, performance of the simpler models (i.e., 2D-EqL and 1D models).
the vertical boundaries of soil domain were restrained in
horizontal direction only, and the base of FE model was restrained
in both vertical and horizontal directions to develop the desired 4. Two-dimensional (2D) equivalent-linear analysis
confining pressure to all the soil elements. In the second step, the
soil constitutive model was switched from linear–elastic to 4.1. Method of analysis
elasto-plastic. The new equilibrium state under gravity was
obtained iteratively. In the third step, self-weight of pier, well In the two-dimensional equivalent-linear approach (2D-EqL),
cap, and other gravity loads were applied statically to the non- the FE model was similar to the 2D-NL model except the analysis
linear soil model. Reaction forces at the boundary nodes were being linear in the former. In this approach, the soil nonlinearity
obtained at the end of gravity analysis. In the fourth step, all the was considered indirectly using the equivalent-linear (or effec-
restraints along the boundary nodes were removed and the tive) soil properties (shear modulus G and damping z). These soil
reaction forces obtained from the gravity analysis of the SWP properties were obtained from the free-field analysis of soil
system were statically applied at the corresponding nodes. This column (without the foundation structure) subjected to a given
state of the model was assumed to be the initial condition of input earthquake motion at the column base using the commer-
the SWP system for the seismic analysis. In the fifth step, both cial software SHAKE2000 [21]. SHAKE2000 uses 1D wave propa-
horizontal and vertical radiation dampers were added at the gation theory to iteratively calculate the level of maximum
nodes of lateral boundaries and base boundary. These dampers strain in soil and determines the equivalent-linear properties of
46 G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55

soil at any desired depth. The above two steps of the 2D-EqL similar type of soil. In the present study, these were obtained
approach (i.e., 1D analysis in SHAKE and 2D linear analysis in from the procedure proposed by Zhang et al. [24]. Fig. 4 shows the
OpenSees) significantly reduce the computational time from the variation of shear modulus (normalized with the maximum shear
single step 2D-NL analysis. modulus Gmax) and damping ratio with shear strain at different
Since the properties G and z depend upon the shear strain level levels of confining pressure (p0 ). Ideally, these curves should be
in soil, their variations with shear strain are needed during the calculated for each layer corresponding to p0 of that layer.
free-field analysis in SHAKE. These variations are generally However, for simplicity, only five sets of G/Gmax and damping
evaluated from laboratory test. However, in absence of laboratory ratio curves representing p0 6 kN/m2, 25 kN/m2, 80 kN/m2,
test, these can be estimated from empirical relationships available 200 kN/m2, and 500 kN/m2 were used in this study by following
in the literature [23–26], or from the experimental results of a the method suggested by Stokoe et al. [27]. The parameter Gmax
(in kN/m2), which is required to estimate G from the G/Gmax curve,
can be obtained from Eq. (1):
1  0 d
p
Gmax ¼ Gr 0 ð1Þ
pr
0.8
where Gr is the shear modulus at reference confining pressure of
0.6 pr0 (¼80 kPa), and d ( ¼0.5) is a positive constant defining the
G/Gmax

variation of G as a function of instantaneous effective confining


2
p’ = 6 kN/m (0m−2m) pressure p0 .
0.4 2
p’ = 25 kN/m (2m−7m)
2 In linear time-domain analysis, the stress–strain curve of soil
p’ = 80 kN/m (7m−20m)
0.2 2
p’ = 200 kN/m (20m−50m)
is linear, and therefore, hysteretic energy dissipation does not
p’ = 500 kN/m2 (50m−100m) occur. However, the energy dissipation in soil was approximately
captured through the equivalent-linear damping zeff in soil
0
10−4 10−3 10−2 10−1 100 obtained from the SHAKE analysis. This damping in soil was used
Shear Strain (%) in the form of mass and stiffness proportional Rayleigh damping.
Rayleigh damping coefficients were determined by considering
two target modes, ith and jth with damping ratio zeff following the
20
p’ = 6 kN/m2 (0m−2m) concept proposed by Hudson et al. [28]. Damping in structure was
2 assumed to be 5% of the critical damping. The ground motions in
p’ = 25 kN/m (2m−7m)
Damping Ratio (%)

15 p’ = 80 kN/m2 (7m−20m) this analysis were applied by following the same procedure as in
p’ = 200 kN/m2 (20m−50m) the 2D-NL analysis.
2
p’ = 500 kN/m (50m−100m)
10
4.2. Comparison of 2D equivalent-linear and 2D nonlinear analyses

5 Displacement and force responses at different locations in pier


and well foundation from the 2D-EqL analysis were compared with
those from the 2D-NL analysis. The percentage error is shown in
0 Fig. 5. It was observed that the error in the maximum absolute
10−4 10−3 10−2 10−1 100
displacement at pier top (DPT) and well top (DWT) was satisfactorily
Shear Strain (%)
estimated by the 2D-EqL analysis with a maximum error within 10%
Fig. 4. Shear–strain dependent (a) shear modulus reduction curves, and except in S1 motion. In S1 motion, the observed error was 24% and
(b) damping curves used in the 2D-EqL in SHAKE [24]. 16% in DPT and DWT, respectively. However, the average absolute

30
L1 M1 S1
15
0
−15
−30
30
L2 M2 S2
15
Error (%)

0
−15
−30
30
L3 M3 S3
15
0
−15
−30
Response Parameters

DPT DWT SFP SFW BMP BMW

Fig. 5. Bar chart of percentage error in the maximum responses (i.e., maximum displacement at pier top (DPT) and well top (DWT); maximum shear force in pier (SFP) and
well (SFW), and maximum bending moment in pier (BMP) and well (BMW)) in the 2D-EqL analysis as compared to the 2D-NL analysis.
G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55 47

Shear Force (MN) Bending Moment (MN−m)


0 25 50 75 100 0 500 1000 1500
0

−5 S1 S1

−10

−15

−20
Depth (m)

−25

−30

−35

−40

−45

−50
2D−NL 2D−Eql 1D−Proposed

Fig. 6. Comparison of envelopes of the maximum force resultants (a) shear force and (b) bending moment along the depth of well foundation obtained from
two-dimensional nonlinear (2D-NL) model, two-dimensional equivalent-linear (2D-EqL) model, and 1D proposed model for severe motion S1.

error and the associated standard deviation (SD) were 8.9% and 5.7%, Superstructure mass
respectively, for DPT, and 7.4% and 4.5%, respectively, for DWT. The
Distributed translation Pier
maximum error in the force resultants (both shear force and bending
springs (kx, cx)
moment) in both pier and well foundation were within 27% while
the average absolute error was about 13% with SD of 7%. The
maximum error in all the response parameters considered in this .
study was limited to 27% while the average absolute error was 11% .
.
with SD of 7%. Moreover, envelopes of the maximum shear force Distributed
Displacement

rotational
time histories

along the depth of well foundation match considerably (Fig. 6(a)). springs (k, c)
Similar match was also observed in the maximum bending moment .
.
envelopes of well foundation (Fig. 6(b)). It is worth mentioning that .
a point in an envelope is obtained by estimating the maximum
response parameter (e.g., shear force) at that location during the
Well
earthquake; thus any two points (i.e., responses) on the envelopes .
may not occur at the same instant of time. In general, it can be . Rotational base
.
inferred that equivalent-linear analysis can satisfactorily estimate springs (kb, cb)
response of well foundation for small to severe earthquakes in spite
of the approximations involved in the analysis. Translational base
spring (kbx, cbx)

Fig. 7. 1D model with four types of spring–dashpot connected in parallel. Note: all
5. One-dimensional (1D) spring–dashpot model
the springs are connected with dampers in parallel. However, dampers are not
shown for clarity.
5.1. Method of analysis

In the 1D spring–dashpot model, a model combining Novak’s


and Veletsos’ springs was used as a basic model. These two where Kx and Ky are the dynamic impedance of soil for transla-
springs were used in this study since Novak’s springs are widely tional and rotational vibrations, respectively; r0 is the radius of
acceptable for the deep foundations and Veletsos’ springs are the circular cylinder or equivalent radius of the foundation with
often used for rigid footings resting on homogeneous half-space. other cross sections; a0 ð¼ r0 o=Vs Þ denotes the dimensionless
Several combinations of these springs were used for the para- frequency; o is the circular frequency; Vs represents the shear
pffiffiffiffiffiffiffi
metric study. Details of the models are described below. wave velocity of soil; i ¼ 1; Sx1 and Sx2 are the real and the
Novak et al. [4] proposed the following dynamic impedances imaginary parts, respectively, of the dimensionless complex
or complex stiffness of unit length of soil when an infinitely long stiffness for horizontal vibration; similarly, Sy1 and Sy2 are the
rigid cylinder is embedded in homogeneous soil and subjected to real and the imaginary parts, respectively, of the dimensionless
translational and rotational modes of vibration independently complex stiffness for rocking; n is Poisson’s ratio, and D is the
assuming plane-strain condition along the length of the cylinder: material damping. The real (kx and ky) and the imaginary parts
(cx and cy) of the impedance functions represent the stiffness and
Kx ¼ G½Sx1 ða0 , n,DÞ þiSx2 ða0 , n,DÞ ð2Þ damping coefficients (both material damping of soil and radiation
damping), respectively, of the complex spring (spring and dashpot
Ky ¼ Gr02 ½Sy1 ða0 ,Ds Þ þ iSy2 ða0 ,Ds Þ ð3Þ connected in parallel). Therefore, the following stiffness and
48 G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55

damping coefficients are obtained from Eqs. (2) and (3): where
(
) 10:2a0 for a0 r 2:5
kx ¼ GSx1 ða0 , n,DÞ k0by ¼
G ð4Þ 0:5 for a0 4 2:5
cx ¼ o Sx2 ða0 , n,DÞ
The parameter n is the Poisson’s ratio of soil at well base. The
9
ky ¼ Gr02 Sy1 ða0 ,Ds Þ = above four-spring–dashpots model consisted of (a) distributed
Gr 2 ð5Þ translational spring–dashpots along the well shaft, (b) distributed
cy ¼ o0 Sy2 ða0 ,Ds Þ ; rotational spring–dashpots along the well shaft, (c) the concen-
trated base translational spring–dashpot, and (d) the concen-
These coefficients are frequency (o) dependent; however, for trated base rotational spring–dashpot. One end of all the springs
simplicity, the stiffness and damping coefficients were evaluated was connected to the well foundation. The other end of all the
corresponding to the predominant frequency of input motion rotational springs (i.e., springs b and d) was fixed in space while
(i.e., o ¼ oip). that of the translational springs (i.e., springs a and c) was
In the present study, the springs and dashpots with coeffi- subjected to earthquake motions in the form of displacement
cients obtained from Eq. (4) were connected in parallel and history corresponding to that soil layer estimated from the
distributed throughout the embedment depth of well shaft as equivalent-linear analysis in SHAKE. In fact, the fixed end of the
shown in Fig. 7. Similarly, Eq. (5) was used to obtain the rotational springs should be subjected to dynamic rotation ye(t)
coefficients of distributed rotational springs along the embed- (¼ due(t)/dz, where due(t) is the differential displacement of the
ment depth of well shaft (Fig. 7). top and bottom nodes of the layer and dz is the thickness of the
For a well foundation of finite depth embedded in a layered layer) imposed by the horizontal displacement profile. However,
soil, the coefficients in Eqs. (4) and (5) were assumed to be valid. for simplicity, small amount of dynamic rotation at the end of the
In other words, the stiffness and damping coefficients per unit rotational springs was neglected in this study.
depth of a particular layer were obtained by assuming that a rigid The spring and dashpot coefficients used in Eqs. (2)–(7) are
well foundation of the same cross section but infinite depth was based on linear soil. In order to consider soil nonlinearity,
embedded in a homogeneous half-space with property of that approximately, the spring coefficients were estimated using the
layer. The same assumption has been made by Novak and Aboul- equivalent-linear properties of soil (G and z) obtained from 1D
Ella [29] for flexible pile embedded in layered soil. wave propagation analysis in SHAKE, as discussed earlier.
Unlike pile where the lateral soil reaction is the sole resisting
mechanism during earthquake, in well foundation base reaction
5.2. Comparison of 1D and 2D-NL analyses
also plays an important role along with the lateral reaction as the
resisting mechanism. Since, solution of Novak et al. [4] was for the
Table 3 illustrates nine different 1D models obtained by
infinitely long foundation, concept of base springs was not there.
different combinations of the springs and dashpots as discussed
Therefore, springs at the well base were introduced in this study
in the previous section. In 1D-1, all the four sets of springs were
to model the base resistance mechanism of the finite depth of
considered and dashpots were neglected while in 1D-2 all the
well foundation considered in this study. The following expres-
springs and dashpots were considered. The models 1D-3 and 1D-4
sions for the dynamic impedance of rigid circular foundation
were developed from 1D-2 by reducing the damping coefficients
resting on surface of elastic homogeneous medium proposed by
to 10% and 20%, respectively, of that considered in 1D-2. The
Veletsos and co-authors [5,6] were used as the coefficients of the
reduction in damping is due to the presence of frequency cut-off
concentrated translational and rotational spring–dashpot system
for radiation damping when natural period of the SWP system is
at the well base:
more than the fundamental period of the soil strata [12,30,31].
9 In such a case, the radiation damping is significantly small as
8Gr0 =
kbx ¼ 2n compared to its half-space value.
8Gr0 0:6a0 ð6Þ
cbx ¼ 2n U o
; Fig. 8 shows the percentage error in the maximum displace-
ment and force resultants in pier and well foundation obtained
from these analyses. The percentage error was estimated based
9 on the 2D-NL analysis. About 75% error was observed in the
8Gr03 0 >
kby ¼ 3ð1nÞ Ukby = response parameters in model 1D-1, which neglected dashpots.
8Gr03 0:35a30 > ð7Þ
cby ¼ The model 1D-2, which considered full damping, resulted in
3ð1nÞ U oð1 þ a2 Þ ;
0
about 50% error in the maximum force and displacement

Table 3
Types of 1D spring–dashpot models analyzed in the present study.

Springs type Kx Cx Ky Cy Kbx Cbx Kby Cby Remarks

1D-1 NV – NV – VS – VS – No dashpot
1D-2 NV NV NV NV VS VS VS VS All springs and dashpots
1D-3 NV 0.1 NV NV 0.1 NV VS 0.1 VS VS 0.1 VS Dashpot coefficients are 10% of the half-space value
1D-4 NV 0.2 NV NV 0.2 NV VS 0.2 VS VS 0.2 VS Dashpot coefficients are 20% of the half-space value
1D-5 NV 0.2 NV NV 0.2 NV VS 0.2 VS – – No base rotational spring and dashpot
1D-6 NV 0.2 NV NV 0.2 NV – – VS 0.2 VS No base translational spring and dashpot
1D-7 NV 0.2 NV NV 0.2 NV – – – – No base springs and dashpots
1D-8 NV 0.2 NV – – VS 0.2 VS VS 0.2 VS No distributed rotational springs and dashpots
1D-Proposed PR PR PR PR VS 0.2 VS – – Proposed distributed springs with no base rotational spring

Note: NV¼ Novak springs; VS ¼Veletsos springs; PR¼ Proposed.


G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55 49

200 L1 M1 S1
150
100
50
0
−50
200 L2 M2 S2
150
Error (%)

100
50
0
−50
200 L3 M3 S3
150
100
50
0
−50
1D−1 1D−2 1D−3 1D−4 1D−1 1D−2 1D−3 1D−4 1D−1 1D−2 1D−3 1D−4
Types of Analysis
DPT DWT SFP SFW BMP BMW

Fig. 8. Bar chart of percentage error in the maximum responses (i.e., maximum displacement at pier top (DPT) and well top (DWT); maximum shear force in pier (SFP) and
well (SFW), and maximum bending moment in pier (BMP) and well (BMW)) obtained from the analysis using four types of 1D models.

Shear Force (MN) Bending Moment (MN−m)


0 25 50 75 100 0 500 1000 1500
0

−5 S1 S1

−10

−15

−20
Depth (m)

−25

−30

−35

−40

−45

−50
2D−NL 1D−2 1D−3 1D−4

Fig. 9. Comparison of envelopes of the maximum force resultants, (a) shear force and (b) bending moment, along the depth of well foundation obtained from the two-
dimensional nonlinear (2D-NL) model and three one-dimensional spring–dashpot models (1D-1, 1D-2, and 1D-3) for severe motion S1.

responses of pier and well foundation. However, the models subsequent analysis to study the effect of individual springs on the
1D-3 and 1D-4 with reduced damping indicated about 730% response of well foundation. Since the ground motion was applied in
and 720% maximum error, respectively. Envelopes of the max- the horizontal direction only, the distributed horizontal springs were
imum force resultants (shear force and bending moment) of well dominant over the other springs. Therefore, the individual signifi-
obtained from the 1D models were also compared with those cance of the other three types of spring was studied.
obtained from the 2D-NL model. Fig. 9 shows such comparison for
the case of S1 motion. The best agreement was observed in case of 5.3. Effect of base rotational spring–dashpot
model 1D-4.
From the above discussion, it was concluded that the coefficient The effect of rotational spring–dashpot (kby, cby) 1D-5 at well
of dashpots should be limited to 20% of that for foundation base on the response of well and pier was studied by comparing
embedded in half-space when the natural period of the SWP system the results of models 1D-5 and 1D-4. The model 1D-5 was
is more than the fundamental period of the soil profile, which prepared by removing the base rotational spring and dashpot
is used. Therefore, the model 1D-4 was used as a basis for the from the model 1D-4. It was observed that the magnitudes of
50
Error with respect to 1D−4 (%) G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55

Shear Force (MN) Bending Moment (MN−m)


40 0 10 20 30 40 0 200 400 600 800
0
20
−10
0
−20
−20
−30
−40
1D−5 1D−6 1D−7 1D−8 −40
Types of Analysis
DPT DWT SFP SFW BMP BMW −50
0
Fig. 10. Effect of different spring components on the maximum responses (i.e.,
maximum displacement at pier top (DPT) and well top (DWT); maximum shear −10
force in pier (SFP) and well (SFW), and maximum bending moment in pier (BMP)
and well (BMW)). −20

−30
error in the maximum force and maximum displacement
−40
responses of well foundation and piers were less than 5% in

Depth (m)
absence of base rotational spring–dashpot (Fig. 10). Comparison −50
of shear force envelopes obtained from these two models (1D-4 0
and 1D-5) shows that the envelopes almost overlapped each
other (Fig. 11(a)). Similarly, the maximum bending moment −10
envelopes of well foundation obtained from these two models
also compared well with each other except near the base of the −20
well foundation where the model 1D-5 estimated slightly lower
value of the maximum bending moment as compared to model −30
1D-4 (Fig. 11(b)). Therefore, it can be inferred that the base
rotational spring and dashpot can be removed without causing −40
significant error in the response parameters. Hence, one can use a
model with only three types of spring–dashpots. −50
0

5.4. Effect of base translational spring–dashpot −10

To study the influence of translational spring and dashpot at −20


well base (kbx, cbx), the maximum displacement and maximum
force responses of model 1D-6 were compared with those of the −30
model 1D-4. The model 1D-6 was prepared by removing the base
translational spring and dashpot from model 1D-4. It was −40
observed that the displacement responses (i.e., DPT and DWT)
were reduced by only 5% and the maximum force resultants in −50
well foundation and pier were reduced by 10% to 13% in absence 1D−4 1D−5 1D−6
of base translational spring and dashpot (Fig. 10). However, the 1D−7 1D−8
envelopes of the maximum shear force were highly underesti-
mated near the base of the well foundation (Fig. 11c). Therefore, Fig. 11. Effect of individual spring–dashpots on the maximum shear force and
one should not neglect the base translational spring during the maximum bending moment envelopes of well foundation: (a–b) base rotational
spring–dashpot (Kby  Cby), (c–d) base translational spring–dashpot (Kbx  Cbx),
seismic analysis of the SWP system. (e–f) both translational and rotational base spring–dashpots (Kbx  Cbx, Kby  Cby),
(g–h) rotational distributed spring–dashpots (Ky–Cy).

5.5. Combined effect of base translational and rotational


spring–dashpots translational and rotational base spring–dashpots is significant
and cannot be neglected.
The combined effect of the base spring–dashpots (both trans-
lational and rotational) were studied by analyzing the well 5.6. Effect of distributed rotational spring–dashpots along well shaft
foundation using only distributed spring–dashpots (kbx, cbx and
kby, cby) along the well shaft (in model 1D-7) and comparing the The effect of distributed rotational spring–dashpots (ky, cy)
results with those of the model 1D-4. In absence of base spring– was studied by analyzing the model 1D-8 and comparing the
dashpots (both translational and rotational), the maximum values results with those of the model 1D-4. The model 1D-8, which was
of the displacement at pier cap (DPT) and well cap (DWT) were prepared by excluding the distributed rotational spring–dashpots
reduced by 8% to 10%. Similarly, magnitudes of the maximum from model 1D-4, resulted in overestimation of the maximum
shear force and bending moment were reduced by about 9% in displacement of pier cap and well cap by about 20% and the
pier and about 15% in well foundation. Moreover, the maximum maximum force resultants in well foundation and pier by about
shear force and bending moment below the central half of well 12% (Fig. 10). The envelopes of the maximum shear force in well
foundation were underestimated if base springs were neglected foundation obtained from the analysis using the models 1D-8 and
(Fig. 11(e) and (f)). Therefore, the combined effect of the 1D-4 vary significantly (Fig. 11(g) and (h)). Therefore, it can be
G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55 51

5 20

4
15
ν = 0.4
3 ν = 0.3

Sx1

Sx2
10
2 ν = 0.3
ν = 0.4
5
1

0 0
5 6

4 5

4
3

Sθ2
Sθ1

3
2
2
1 1

0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
a0 = ωr0/Vs a0 = ωr0/Vs
ζ = 5% ζ = 10% Proposed

Fig. 12. Stiffness (Sx1 and Sy1) and damping parameters (Sx2 and Sy2) for translational and rotational springs proposed by Novak et al. [4] along with the simplified
parameters proposed in the present study.

concluded that the distributed rotational springs and dashpots Superstructure mass
have significant influence in the response of well foundation Pier
Distributed translational
and pier. springs (kx, cx)

6. Proposed 1D model

In the previous section, it was concluded that when the natural Distributed
period of the SWP system is more than the fundamental period of rotational
springs (k, c)
Displacement
time histories

the soil strata, the coefficient of dashpots should be reduced to


about 20% of that for foundation embedded in half-space. It was
also observed that the rotational base spring and dashpot may be
neglected without causing significant error in the estimation of the
maximum displacement and force resultants in well foundation
and pier. Well
Based on the above findings, a spring–dashpot model consisting
of (1) distributed translational spring–dashpots, (2) distributed
rotational spring–dashpots, and (3) a base translational spring–
dashpot is proposed (Fig. 13). The spring and dashpot parameters
Translational base
in Eqs. (4) and (5) have been simplified by some linear approxima- spring (kbx, cbx)
tion of these parameters with dimensionless frequency (a0). Fig. 12
shows that the translational spring stiffness parameter Sx1 proposed Fig. 13. Proposed 1D model with three types of spring–dashpots. Note: all the
by Novak et al. [4] depends on the dimensionless frequency (a0) and springs are connected with dampers in parallel; however, dampers are not shown
for clarity.
n. However, in the frequency range of interest (0.5oao1.5), Sx1 can
be assumed to be independent of a0 and n. Hence, Sx1 is crudely 9
approximated to be equal to 3.5. However, the translational damp- ky ¼ Gr02 ð3:00:75a0 Þ =
ing parameter Sx2 almost linearly varies with a0. Neglecting the 0:6Gr02 a0 ð9Þ
cy ¼ ;
o
effect of n on Sx1, it is proposed that Sx2 E10a0. Similarly, the
rotational spring and damping parameters have been simplified as 9
8Gr0 =
Sy1 E3.00.75a0 and Sy2 E3a0. The proposed simplified expressions kbx ¼ 2n
8Gr0 0:12a0 ð10Þ
of the Novak’s spring coefficients are plotted in Fig. 12 along with cbx ¼ 2n U o
;
the original expressions. Summary of the proposed spring–dashpots
coefficients is shown in Eqs. (8)–(10) Results of analysis of the well foundation using the proposed
) three-spring model (model 1D-Proposed) were compared with those
kx ¼ 3:5G
ð8Þ obtained from the 2D-NL analysis. The values of percentage error in
cx ¼ 2Ga
o
0
the response parameters were also compared with those for 2D-EqL
52 G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55

60

30

−30
L1 M1 S1
−60
60

30
Error (%)

−30
L2 M2 S2
−60
60

30

−30
L3 M3 S3
−60
2D−EqL 1D−4 1D−Proposed 2D−EqL 1D−4 1D−Proposed 2D−EqL 1D−4 1D−Proposed
Types of Analysis

D D SF SFW BM BMW
PT WT P P

Fig. 14. Bar chart of percentage error in the maximum responses (i.e., maximum displacement at pier top (DPT) and well top (DWT); maximum shear force in pier (SFP) and
well (SFW), and maximum bending moment in pier (BMP) and well (BMW)) obtained from the various approaches.

and 1D-4 analyses (Fig. 14). It was observed that the proposed model Performance of the above spring–dashpot models was checked
satisfactorily estimates the maximum force and maximum displace- by analyzing the SWP system using these models and comparing
ment responses in pier and well foundation; the maximum percen- results with those of the 2D-NL analysis. The coefficient cy
tage error in the response parameters was within 30% except for the proposed by Gerolymos and Gazetas [1] and Varun et al. [2]
motion S3. Moreover, the values of the percentage error for the was assumed to be zero when it attained a negative value. It was
proposed model and the model 1D-4 match well with each other. observed that the proposed model performed considerably well in
Therefore, it can be concluded that the effects of simplification of comparison to the other models considered in this study. Sig-
Novak’s [4] spring and dashpot coefficients, and the base rotational nificant error (more than 50%) was observed in the displacement
spring on the maximum force and displacement responses of pier and and force response parameters of well foundation and pier for
well foundation are not significant. models proposed by Gerolymos and Gazetas [1] and Varun et al.
[2] (Fig. 16). In a few cases, the JRA model, which does not
consider dashpots, also estimated the responses satisfactorily.
However, it significantly overestimated (error4100%) the max-
7. Performance of some other existing 1D models imum shear force and maximum bending moment in pier for the
motions L2, M1, S1, and S2.
As stated earlier, there are only a few spring–dashpot models,
available in the literature, which were developed specifically for
the seismic analysis of well foundation. In the previous discussion,
the distributed translational and rotational springs and dashpots 8. Computational efficiency of different approaches
proposed by Novak et al. [4] were used to analyze the SWP system
and a simplified model with three types of springs was proposed. Computational efficiency of the 2D-NL, 2D-EqL, and 1D
In this section, three other 1D models were studied; one model approaches was compared to study the acceptability of the
was based on the recommendation of Japanese code [3], and the available approaches in design offices. It was observed that the
other two were proposed by Gerolymos and Gazetas [1] and Varun computational time required for the above three approaches were
et al. [2]. Fig. 15 illustrates the comparison of spring and dashpot 15 h, 6 h, and 5 min for the L1 motion. Hence, one can imagine
coefficients of the soil layer considered in this study. It was that the 3D SWP model with appropriate model parameters may
observed that the value of translational spring coefficient kx take significantly long time for the execution under seismic load-
specified in JRA [3] was significantly higher in comparison to those ing. Although there was no significant difference in the estimated
of the other three models. Values of the coefficients ky, cx, and cy response parameters, computational efficiency of the 2D-EqL
obtained from these models significantly varied from each other. It analysis was about three times better than that of the 2D-NL
is worth noting that the dashpot coefficient of the rotational analysis. On the other hand, the 1D spring–dashpot model req-
springs (cy) proposed by Gerolymos and Gazetas [1] and Varun uired negligible computational time as compared to the 2D-NL
et al. [2] attained negative values. The negative values of cy may and 2D-EqL analyses. However, the 1D approach is comparatively
sometimes cause convergence problem during time history analy- less accurate than the 2D-EqL model. Therefore, one should
sis. Since these models were developed specifically for frequency choose correct approach depending on the importance of the
domain analysis, their application in the time domain may not be problem. In general, 1D approach can be chosen for first hand
suitable. answer and 2D-EqL approach may be chosen for detailed analysis.
G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55 53

Kx (kN/m) × 106 Cx (kN.s/m) × 104


0 4 8 12 0 1 2 3 4 5
0

−10

−20

−30

−40
Depth (m)

−50
Kθ (kN.m) × 108 Cθ (kN.m.s) × 107
0 1 2 3 −3 −2 −1 0 1
0

−10

−20

−30

−40

−50
Novak et al. [4] JRA [3]
Gerolymos and Gazetas [1] Varun et al. [2]

Fig. 15. Comparison of spring and dashpot coefficients of distributed side springs obtained from various 1D models.

9. Summary and conclusions specifically developed for well foundation were also studied. Based
on the present study, the following conclusions can be made:
Seismic response of soil–well–pier (SWP) system was
performed by three approaches, namely, two-dimensional non- 1) The 2D-EqL approach, which is simpler and computationally
linear (2D-NL), two-dimensional equivalent-linear (2D-EqL), and more efficient than the 2D-NL approach, can be very useful for
one-dimensional spring–dashpot (1D) approaches. Out of these the seismic analysis of well foundation though it considers soil
three approaches, the 1D approach is the simplest one and it is nonlinearity approximately. Such analysis may cause up to
easy to implement in design offices. Nine earthquake motions with 30% error in design displacement and force resultants in pier
peak ground acceleration ranging from 0.2 g to 0.6 g were used to and well foundation. However, computationally, it is about
represent the variations in ground motion parameters. Displace- three times more efficient than the 2D-NL analysis.
ment and force responses obtained from the 2D-EqL analysis were 2) For routine work in design offices, engineers may use the 1D
compared with those determined by the rigorous 2D-NL analysis. spring–dashpot model, which requires negligible computational
In the 1D approach, various 1D models combining springs pro- time as compared to the 2D analyses (both 2D-NL and 2D-EqL).
posed by Novak et al. [4], and Veletsos and co-authors [5,6] were In this model, radiation damping should be considered otherwise
prepared and used to analyze the SWP system. Results obtained displacement and force responses may be overestimated.
from these models were compared with the results of 2D-NL 3) For long span bridges supported on well foundation, the
model. Based on this study, the coefficients of Novak’s springs natural period of the SWP system is generally more than the
were further simplified and a three-spring–dashpot model was fundamental period of the soil. In this case, the use of radiation
proposed. The proposed model consisted of distributed transla- damping for half-space may lead to underestimation of dis-
tional and rotational spring–dashpots along the well shaft and placement and force resultants in well foundation and pier.
concentrated base translational spring–dashpot. The proposed In such a situation, it is recommended that the coefficient of
model estimated relatively well the seismic response of the SWP dashpots should be reduced to 20% of that for foundation
system. Performances of some other 1D spring–dashpot models embedded in half-space. The spring and damping coefficients
54 G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55

200
L1 M1 S1

100

−100
200
L2 M2 S2
Error (%)

100

−100
200
L3 M3 S3

100

−100
Proposed JRA [3] Gerolymos Varun Proposed JRA [3] Gerolymos Varun Proposed JRA [3] Gerolymos Varun
and et al. [2] and et al. [2] and et al. [2]
Gazetas [1] Gazetas [1] Gazetas [1]
Types of Analysis
DPT DWT SFP SFW BMP BMW

Fig. 16. Bar chart of percentage error in the maximum responses (i.e., maximum displacement at pier top (DPT) and well top (DWT); maximum shear force in pier (SFP) and
well (SFW), and maximum bending moment in pier (BMP) and well (BMW)) obtained from the 1D analysis.

should be estimated based on the equivalent-linear properties References


of soil.
4) One can use the proposed model, which consists of three types [1] Gerolymos N, Gazetas G. Winkler model for lateral response of rigid caisson
of spring–dashpots and ignores base rotational spring–dashpot foundations in linear soil. Soil Dyn Earthquake Eng 2006;26:347–61.
[2] Varun, Assimaki D, Gazetas G. A simplified model for lateral response of large
to analyze approximately the well foundation embedded in
diameter caisson foundations—linear elastic formulation. Soil Dyn Earth-
soil. The proposed model estimated the displacement and quake Eng 2009;29:268–91.
force resultants in pier and well foundation relatively better [3] JRA. Specifications for Highway Bridges, Part IV: Substructures. Japan: Japan
than the available 1D models. In a few cases, the JRA springs Road Association; 2002.
[4] Novak M, Nogami T, Aboul-Ella F. Dynamic soil reactions for plane strain case.
[3] also estimated the responses satisfactorily. J Eng Mech Div, ASCE 1978;104:953–9.
[5] Veletsos AS, Wei YT. Lateral and rocking vibration of footings. J Soil Mech
Found Div, ASCE 1971;97:1227–48.
In the present study, seismic analysis was performed on a [6] Veletsos AS, Verbic B. Basic response functions for elastic foundations. J Eng
Mech Div, ASCE 1974;100:189–202.
typical well foundation embedded in a layered cohesionless soil.
[7] Wolf JP, Meek JW. Insight on 2D versus 3D modelling of surface foundations
Considering the fact that such foundations may be located in via strength of materials solutions for soil dynamics. Earthquake Eng Struct
different seismic zones and may be subjected to different inten- Dyn 1994;23:91–112.
sities of earthquake motion, the present study focuses on the [8] Luco JE, Hadjian AH. Two-dimensional approximations to the three-
dimensional soil–structure interaction problem. Nucl Eng Des 1975;31:
variation of ground motions. Further comprehensive research is 195–203.
needed to study the influence of the geometry of the foundation and [9] Watanabe H, Tochigi H. Model vibration test concerning the dynamic
properties of soil layers on the conclusions drawn in this paper. interaction of soil structure system followed by sliding and separation and
their numerical simulation. Proc JSCE 1986:319–27. [368-I-5].
Seismic analysis of the SWP system is a 3D problem, which has
[10] Seed BH, Lysmer J. Soil–structure interaction analyses by finite elements–
been represented as a 2D plane-strain problem in this paper. State of the art. Nucl Eng Des 1978;46:349–65.
Despite the limitations of the 2D model, it is able to capture (at [11] Jeremic B, Kunnath S, Xiong F. Influence of soil–foundation–structure inter-
least qualitatively) the key aspects of the effects of soil non- action on seismic response of the I-880 viaduct. Eng Struct 2004;26:391–402.
[12] Gerolymos N, Gazetas G. Static and dynamic response of massive caisson
linearity (liquefaction) on the overall seismic response mechan- foundations with soil and interface nonlinearities – validation and results.
ism of the SWP system. More comprehensive research may be Soil Dyn Earthquake Eng 2006;26:377–94.
needed in future for carrying out seismic analysis of 3D SWP [13] Elgamal A, Yan L, Yang Z, Conte JP. Three-dimensional seismic response of
Humboldt Bay bridge-foundation-ground system. J Struct Eng 2008;134:
model with nonlinear soil.
1165–76.
[14] Chang CY, Mok CM, Wang ZL, Settgast R, Waggoner F, Ketchum MA, et al.
Dynamic soil–foundation–structure interaction analysis of large caissons,
MCEER-00-0011. New York: Multidisciplinary Centre for Earthquake Engi-
neering Research, University of Buffalo; 2000.
Acknowledgements [15] Mondal G, Prashant A, Jain SK. Significance of interface nonlinearity on seismic
response of well-pier system in cohesionless soil. Earthquake Spectra, in press.
[16] Lysmer J, Kuhlemeyer RL. Finite dynamic model for infinite media. J Eng
Authors gratefully acknowledge Poonam and Prabhu Goel
Mech Div, ASCE 1969;95:859–77.
Foundation at Indian Institute of Technology Kanpur for the financial [17] Joyner WB, Chen ATF. Calculation of nonlinear ground response in earth-
support in conducting the present study. quakes. Bull Seismol Soc Am 1975;65:1315–36.
G. Mondal et al. / Soil Dynamics and Earthquake Engineering 32 (2012) 42–55 55

[18] Mazzoni S, McKenna F, Scott MH, Fenves GL, et al. Open system for [25] Seed HB, Idriss IM. Soil moduli and damping factors for dynamic response
earthquake engineering simulation: user command-language manual, analyses. Report no. EERC 70-10. Berkeley, California: Earthquake Engineer-
version 1.7.3. Berkeley: Pacific Earthquake Engineering Center, University ing Research Center, University of California; 1970.
of California; 2006. /http://opensees.berkeley.eduS. [26] Sun JI, Golesorkhi R, Seed HB. Dynamic moduli and damping ratios for
[19] Yang Z, Lu J, Elgamal A. OpenSees soil models and solid-fluid fully coupled cohesive soils. Report no. EERC-88/15. Berkley: Earthquake Engineering
elements: user’s manual, Version 1. San Diego: University of California; 2008. Research Center, University of California; 1988.
[20] Yang Z. Numerical modeling of earthquake site response including dilation [27] Stokoe KH II, Hwang SK, Darendeli MB, Lee NJ. Correlation study of nonlinear
and liquefaction. PhD thesis. New York: Department of Civil Engineering and dynamic soil properties. Final Report to Westinghouse Savannah River
Engineering Mechanics, Columbia University; 2000. 255 p.
Company, Aiken, SC; 1995.
[21] Ordóñez GA. User’s manual of SHAKE-2000: a computer program for the 1-D
[28] Hudson M, Idriss IM, Beikae M. User’s manual for QUAD4M: a computer
analysis of geotechnical earthquake engineering problems. GeoMotions;
program to evaluate the seismic response of soil structures using finite
2004, 310 p. /http://www.shake2000.comS.
[22] Mondal G. Seismic analysis of soil–well–pier system for bridges. PhD thesis. element procedures and incorporating a compliant base. Davis, USA: Uni-
Kanpur: Department of Civil Engineering, Indian Institute of Technology versity of California; 1994.
Kanpur; 2011. [29] Novak M, Aboul-Ella F. Impedance functions of piles in layered media. J Eng
[23] Darendeli MB. Development of a new family of normalized modulus reduc- Mech Div, ASCE 1978;104:643–61.
tion and material damping curves. PhD thesis. Austin: University of Texas at [30] Wolf J, Song C. To radiate or not to radiate. Earthquake Eng Struct Dyn
Austin; 2001. 1996;25:1421–32.
[24] Zhang J, Andrus RD, Juang CH. Normalized shear modulus and material [31] Meek J, Wolf J. Insights on cutoff frequency for foundation on soil layer.
damping ratio relationships. J Geotech Geoenviron Eng 2005;131:453–64. Earthquake Eng Struct Dyn 1991;20:651–65.

You might also like