Está en la página 1de 546

Proceedings of

The 2nd Japan-Greece Workshop on


Seismic Design, Observation,
and Retrofit of Foundations
April 3~4, 2007, Tokyo, Japan

Proceedings of the 2nd Japan-Greece Workshop on Seismic Design, Observation, and


Retrofit of Foundations
Tokyo, Japan, April 3~4, 2007

Sponsored by
Earthquake Engineering Committee of the Japanese Society of Civil Engineers
Hellenic Society for Earthquake Engineering
Co-sponsored by
Japanese Society of Civil Engineering
Published by
Japan Society of Civil Engineers

@2007 Japan Society of Civil Engineers

All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means
electronic or mechanical, including photocopying, recording or any information storages and
retrieval system now known or to be invented, without written permission from the publisher.
978-4-8106-0620-1
Printed by WACO Co., Ltd.

CONTENTS
Organizing committee ............................................................................................................................................... i

CONFERENCE PAPERS
Tunnels, Retaining Systems, and Landslides
Deep immersed tunnel under combined major fault rupture deformation and subsequent strong seismic
shaking

Gazetas, G., Anastasopoulos, I., Gerolymos, N., Drosos, V., Kourkoulis, R. and
Georgarakos, .--------------------------------------------------------------------------------------- 1
Seismic triggering, evolution, deposition, and retaining of rapid landslides

Gerolymos, N., Gazetas G. and Vardoulakis, I. ------------------------------------------------- 27


Interaction of earthquake-triggered landslide with foundation-structure systems

Kourkoulis, R., Gelagoti, F., Anastassopoulos, I. and Gazetas, G. --------------------------- 51


Detrimental effects of urban tunnels on design seismic ground motions

Kouretzis, G., Bouckovalas, G., Sofianos, A. and Yiouta-Mitra, P. --------------------------- 64


Seismic analysis of near-surface tunnel cross-sections

Vrettos, C. -------------------------------------------------------------------------------------------- 72
Sliding of rigid block on sloping plane: the surprising role of the sequence of long-duration pulses

Garini, E. and Gazetas, G. ------------------------------------------------------------------------- 79


A replacement to the MononobeOkabe equations by stress limit analysis

Mylonakis, G., Kloukinas, P. and Papantonopoulos, C.-------------------------------------- 105


A study on ground displacement due to fault slip and its influence to underground structure

Adachi, Y., Yoshimura, S. and Nakata, T. ------------------------------------------------------ 123

Shallow Foundations and SoilStructure Interaction


Study on applicability of seismic-isolation foundation to railway structures

Luo, X., Miyamoto, T. and Imamura, T. -------------------------------------------------------- 131


A study on dynamic soil-structure interaction effect based on microtremor measurement of building and
surrounding ground surface

Iiba, M., Watakabe, M., Fujii, A., Koyama, S., Sakai, S. and Yasui, M. -------------------- 135
Damage index evaluations of SDOF system with dynamic soil-structure interaction

Kawano, K., Kimura, Y. and Nakamura, Y. ---------------------------------------------------- 143


Seismic response of rigid structures stepping on nonlinear foundation

Palmeri, A. Makris, N. ---------------------------------------------------------------------------- 150


Numerical simulations of shaking table experiments on a shallow foundation test model at PWRI, Japan

Paolucci, R., Shirato, M. and Yilmaz, M. T.---------------------------------------------------- 158

Analytical modelling of footings under large overturning moment

Apostolou, M. and Gazetas, G. ------------------------------------------------------------------ 165


Natural period and effective damping of simple structures on footings and piles

Maravas, G. Mylonakis and D. L. Karabalis -------------------------------------------------- 185


Interaction of shallow and deep foundations with a rupturing normal fault

Anastasopoulos, L. Gazetas, G.------------------------------------------------------------------ 197


The role of the soil stiffness on the dynamic impedance functions

Pitilakis, D., Clouteau, D. and Modaressi, A. ------------------------------------------------- 212


Fundamental period of sdof systems including soil-structure interaction and soil improvement

Kirtas, E., Trevlopoulos, K. , Rovithis, E. and Pitilakis, K. ---------------------------------- 225


Analytical study on seismic design of footings subjected to earthquake loading

Kosa, K., Ando, T., Adachi, Y. and Shirato, M. ------------------------------------------------ 238


Development of the sheet-pile foundation; a series of seismic proving tests

Higuchi, S., Nishioka, H., Tanaka, K., Koda, M. and Hirao, J. ----------------------------- 246
Rocking seismic isolation of bridges supported by spread foundations

Kawashima, K., Nagai, T. and Sakellaraki, D. ------------------------------------------------ 254

Deep Foundations
Effects of limit state performance of steel bearings on a bridge upper structure

Ohtomo, K., Sato, Y. and Sakai, M. ------------------------------------------------------------- 267


Large-scale model tests of shallow foundations subjected to earthquake loads

Shirato, M., Kouno, T., Nakatani, S. and Paolucci, R.---------------------------------------- 275


Numerical modeling of centrifuge cyclic lateral pile load experiments

Gerolymos, N., Drosos, V., Escoffier, S., Gazetas, G. and Garnier, J. --------------------- 299
Seismic retrofitting of the piled foundation of a reinforced concrete building

Maugeri, M. and Castelli, F. --------------------------------------------------------------------- 320


Influence of batter piles on the seismic response of pile groups

Giannakou, A., Gerolymos, N. and Gazetas, G. ---------------------------------------------- 335


Effect of the natural frequencies of the coupled soil-pile-structure system on pile dynamic response

Rovithis, E., Kirtas, E. and Pitilakis, K. -------------------------------------------------------- 344


Utilization of P-Y curves for estimating soil pile interaction under seismic loading

Rovithis1, E., Kirtas1, E. and Pitilakis1, K.---------------------------------------------------- 357


Linear elastic transient response of deep rigid foundations: From 3D finite element simulations to design

Varun, Assimaki, D. and Gazetas, G. ----------------------------------------------------------- 369


Seismic zonation, vulnerability assessment and loss scenarios in Thessaloniki

Pitilakis, K., Anastasiadis, A., Kakderi, K., Argyroudi, S. and Alexoudi, M. -------------- 386
Minimization of fixed-head pile bending at optimal radius under realistic conditions by using seismic
deformation method

Saitoh, M. ------------------------------------------------------------------------------------------- 401

Seismic response analysis of pile foundation using finite element method

Maki, T., Tsuchiya, S., Watanabe, T. and Maekawa, K. -------------------------------------- 409


Seismic behavior of single pile in cohesive soil

Tuladhar, R., Mutsuyoshi, H. and Maki, T. ---------------------------------------------------- 417


Passive and active stress wave techniques for damage assessment of railway piers

Shiotani, T., Miwa, S., Aggelis, D. G., Luo, X. and Haya, H. -------------------------------- 424

Liquefaction Densification Related Studies, and others


Centrifuge tests on remedial measure using batter piles against liquefaction-induced soil flow after quay
wall failure

Tazoh, T., Sato, M., Jang, J. and Gazetas, G.-------------------------------------------------- 431


Why do pile-supported bridge piers and not abutments collapse in liquefiable soils during earthquakes?

Kerciku, A. A., Bhattacharya, S. and H. J. Burd----------------------------------------------- 440


Effect of excess pore water pressure buildup on building damage

Dakoulas, P.---------------------------------------------------------------------------------------- 453


Numerical analysis of gravel drain performance in liquefiable soils

Papadimitriou, G., Moutsopoulou, M.-E. and Bouckovalas, G. D. ------------------------- 467


Estimating earthquake induced settlements on granular soils Application to shallow foundations

Egglezos, D. ---------------------------------------------------------------------------------------- 479


Unstable behaviour of a fine sand under cyclic loading

Georgiannou, V. N. and Tsomokos, A. ---------------------------------------------------------- 491


Numerical simulation of liquefied sand using a shear-thinning fluid model

Schenkengel, K.-U., Becker, A. and Vrettos, C.------------------------------------------------ 497


On the liquefaction resistance of silty sands

Papadopoulou, A. I. and Tika, T. M. ------------------------------------------------------------ 504


Effect of countermeasures for pile foundation under lateral flow caused by liquefaction

Matsuda, T. and Sato, K. ------------------------------------------------------------------------- 513


Effect of duration of earthquake on onset of liquefaction

Yoshida, N. ----------------------------------------------------------------------------------------- 521


A simple evaluation method for earthquake damage to the quay walls

Soejima, M., Suizu, A., Ejiri, J. and Matsuda, T. ---------------------------------------------- 527


Effect on records of seismic intensity-meter by adjacent building

Kataoka, S. ----------------------------------------------------------------------------------------- 531

Organizing Committee
Chairman

T. Katayama, Japan
G. Gazetas, Greece

Members

H. Iemura, Japan
Y. Goto, Japan
F. Miura, Japan
K. Wakamatsu, Japan
T. Tazoh, Japan
N. Yoshida, Japan
T. Matsuda, Japan
H. Kiku, Japan
M. Saitoh, Japan
K. Pitilakis, Greece
G. Bouckovalas, Greece
G. Mylonakis, Greece
A. Nikolaou, Greece
P. Dakoulas, Greece
D. Assimaki, Greece
N. Gerolymos, Greece

Tunnels, Retaining Systems, and Landslides

Deep Immersed Tunnel under Combined Major Fault Rupture


Deformation and Subsequent Strong Seismic Shaking
G. Gazetas, I. Anastasopoulos, N. Gerolymos, V. Drosos,
R. Kourkoulis, . Georgarakos
National Technical University of Athens, Greece

Abstract
Immersed tunnels are particularly sensitive to tensile deformations such as those imposed by an
earthquake normal fault rupturing underneath, and those generated by the dynamic response
due to seismic shaking. The paper investigates the response of a future 70 m deep immersed
tunnel to the combined action of a major normal fault rupture due to an earthquake occurring in
the basement rock underneath the tunnel, and a subsequent strong seismic shaking from a
different largemagnitude seismic event that may occur years later. Non-linear finite elements
model the quasistatic fault rupture propagation through the soil overlying the bedrock and the
ensuing interaction of the rupture with the immersed tunnel. It is shown that despite an imposed
baserock offset of 2 meters, net tension or excessive compression between the tunnel
segments can be avoided with a suitable design of the joint gaskets. Then, the already
(permanently) deformed structure is subjected to strong asynchronous seismic shaking. The
tunnel is modeled as a 3D flexural beam connected to the soil through properlycalibrated
nonlinear interaction springs and dashpots, the supports of which are subjected to the freefield
acceleration time histories. The latter are basically obtained with 1D wave propagation
analysis, after they are modified to account for wave passage effects. The joints between the
tunnel segments are modeled with special non-linear hyper-elastic elements, while their
longitudinal pre-stressing due to the (7 bar) water pressure is also rigorously incorporated in the
analysis. The possibility of sliding is considered through the use of special gap elements. The
influence of segment length and joint properties is explored parametrically. A fascinating
conclusion emerges in all analysed cases for the joints between tunnel segments that were
differentially deformed due to static fault rupture displacements : upon subsequent strong
seismic shaking (with ground accelerations even exceeding 0.50 g), overstressed joints decompress and understressed joints re-compress, a healing process that leads to a more
uniform deformation profile along the tunnel. This is particularly beneficial for the precariously
de-compressed joint gaskets which thus avoid the danger of getting into net tension. Hence, the
safety of the immersed tunnel improves with subsequent strong seismic shaking !

INTRODUCTION

the narrowest crossing of the Corinthian Gulf


trench, which is characterized by an extensional
slip rate of about 1 cm/yr, the tunnel will
unavoidably cross at least one active fault. It
must be capable of safely undertaking the
ensuing permanent ground deformation.
Moreover, in the 100 year design life of the
tunnel, strong seismic shaking from another
earthquake, originating at a different major fault
located not directly under the site, is a distinct
possibility given the numerous active faults in
the Corinthian and Patraicos Gulfs, as well as in
their neighboring regions. Such seismic shaking

The objective of this paper is to investigate


the behaviour of a deep immersed tunnel under
the combined action of a major fault rupturing
underneath and a subsequent strong seismic
shaking. The presented research is part of a
feasibility study for a railway immersed tunnel in
the RionAntirrion straits, in Greece. The
construction of the tunnel will be a technological
challenge, due to the combination of great
water depth (67 m) (presently the world record),
very high seismicity, and deep alluvial soils. At

of 23 m x 11 m (width x height) section is


designed to accommodate two-way rail traffic.
The Rion-Antirron Straits is the narrowest
crossing of the Corinthian Gulf trench. The
trench is associated with a tensile tectonic
environment of NS direction, expressed by a
sequence of normal faults of EW strike dipping
towards the axis of the trench. Characterized by
a slip rate (extension) of almost 1 cm / yr, and
being one of the most seismically active zones
of Greece (Fig 1b), the area has experienced
five major earthquakes of Ms > 6, during the last
30 years. The most recent event was the Ms 6.2
Aegion 1995 earthquake, with an epicentral
distance of 28 km from the site. The Straits are
also affected by even more distant rupture
zones, such as the Ionian zone which is by far
the most active seismic zone in Europe,
capable of producing earthquakes of Ms > 7.
The recent Ms 6.4 Lefkada earthquake
(1482003) was only a small reminder of this
possibility.
Making use of the abundant geotechnical
data obtained in the course of design of the
neighboring RionAntirrion bridge we were able
to construct a credible soil profile of the
crossing. It consists of alternating layers of
sandy gravel to gravel, silty sand and clay, of
medium density and stiffness. The geotechnical
exploration reached a maximum depth of 100 m
below the seabed, without encountering
bedrock. Geological studies suggested that the
total thickness of the soil sediments exceeds
500 m. This was verified by a detailed
geophysical tomography conducted for the
needs of the present study [Tselentis, 2004]. It
was revealed that limestone bedrock lies at a
depth of about 800 m below seabed.

will find the immersed tunnel already injured


from the ground surface dislocation of the first
seismic event (which will have occurred
perhaps several tens of years earlier). It is
crucial for a successful design to ensure that
the permanent tensile deformation due to the
normalfault rupture, and the superimposed
subsequent dynamic deformation during
shaking will not jeopardize the water-tightness
of the tunnel at any moment. Figure 1 sketches
the two loading situations studied in the paper.
A FUTURE RIONANTIRRION RAILWAY LINK

A proposed railway link will be located at the


RionAntirrion straits connecting Central
Greece with Peloponnesus (Fig 2 and 3). The
straits will be crossed at approximately 300
meters east from the recently built cable-stayed
road bridge. At this narrowest point, the underwater length is about 2.5 km, with a maximum
depth of 67 m.
Two tunneling alternatives were initially
proposed and investigated : a continuous bored
tunnel, and a hybrid solution, combining a
central immersed tunnel at the deepest section
of the crossing, with two bored approach
tunnels at the two sides. The first solution would
require the boring of a tunnel at a maximum
depth of 95 m below the sea surface within
relatively loose soil. Facing water pressures of
the order of 10 bars relentlessly over a length of
at least 2 km was deemed far too risky, despite
the progress in shield tunneling technology, and
therefore only the second solution was chosen
for further research. The cross sections of the
immersed and the bored tunnel are depicted in
Figs 3b and 3c, respectively. The central tube

DYNAMIC (cyclic)
Decompression + Recompression

STATIC
Tensile Opening (Decompression)

accelerations i(t)

+
Fault Rupture

Figure 1: Definition of the problem : two types of loading that may take place in the life of the tunnel.

Lake Trihonis

Active Fault
Potentially Active Fault
0

10

20

(km)

Rion
Elik
i

Patras

Fau
lt

Aegion Fault

Figure 2: RionAntirrion straits with the active and potentially active faults

(a)

RION

ANTIRRION

Depth (m)

20

40

60

Embankment

Bored T
unn

Embankment
Immersed Tunnel

Special Connecting
Segment

80

l
Tunne
Bored

el

Special Connecting
Segment

4+000

4+500

5+000

5+500

6+000

Length (km + m)

1.5 m

0.5 m
5.5 m

11 m

4.0 m

4.4 m

4.4 m

1.5 m

2.0 m

1.0 m

1.5 m
3
1

0.7 m

4.4 m

4.4 m

2.5 m
4.0 m

23 m

12.5 m

(b)

(c)

Figure 3: (a) Longitudinal Section of the proposed tunnel, (b) Immersed tunnel cross-section, and (c) Bored
tunnel cross-section

ASEISMIC DESIGN OF THE PROPOSED RIONANTIRRION IMMERSED TUNNEL

As already discussed, the concrete


segments will be constructed in a dry-dock,
floated over the pre-excavated trench, and
lowered with the help of special sinking rigs.
Each segment will be lowered close to the
previous one, and brought to contact using
special guidance techniques. Once the two
segments gain contact, the water between them
will be drained, and the Gina gasket will be
compressed (Figs 4b and c) with the help of the
hydrostatic water pressure acting only at the
free side of the segment.
In our case, this pressure will be of the order
of 7 bars, and thus the use of the largest
available Gina gasket is indispensable. One
problem with the application of the method in
such depth is the extremely high compressive
load on the gasket. For the cross-sectional area

Until today, no immersed tunnel is known to


have ever been hit by an underneath rupturing
fault. On the other hand, two immersed tunnels
have been subjected to fairly strong seismic
shaking : the Bay Area Rapid Transit (BART)
tunnel in California, and the Osaka South Port
immersed tunnel in Japan. Both tunnels
behaved exceptionally well sustaining no
measurable damage.
After reviewing alternative seismic design
techniques already applied to existing
immersed tunnels, and given the good seismic
performance of the Osaka South Port immersed
tunnel, we proposed a similar design for our
immersed section (Fig 4).

BEFORE CONTACT

(b)

(a)

Gina Profile

Gina Profile
Tendon

Omega Profile

Shear Key

Tendon

Tendon

Coupler

(c)

Omega
Profile

Coupler

AFTER CONTACT

Figure 4: Aseismic design of the immersed section of the proposed railway link : (a) schematic detail of the
immersion joint, showing the gina gasket, the omega seal, the tendons along with the couplers, and
the shear key, (b) zoom-in of the immersion joint before the contact of two consecutive segments,
and (c) after the compression of the Gina gasket, installation of the Omega seal and connection of
the tendons with the use of a special coupler.

of our tunnel, 253 m2, the total hydrostatic force


to act on the gasket will be of the order of 175
MN ( 253 m2 x 700 kPa). Given the perimeter
(about 65 m) of the Gina gasket, the total force
per running meter will be 2.7 MPa. Gina
gaskets may fail in lateral tension under
significant
compressive
loads.
The
specifications of the largest available Gina
section show that they will be capable (in the
limit) of undertaking such a pressure. However,
since this will be the first Gina gasket to be
used in such depth, special analysis and testing
will be a prerequisite for acceptance.
After compression is completed, the
secondary omega seal is installed. Contrary to
the Gina profile that requires compression to
achieve water-tightness, the omega seal is
insensitive to the magnitude of compression.
However, none of the above materials can
transmit either shear or tension. To this end,
shear keys and tendons will be installed. A
shear key at the bottom of the tunnel will
transmit transverse shear forces, while two
similar shear keys at the side-walls will transmit
vertical shear. They are all constructed after the
segments are connected, by casting concrete in
situ. Tendons will transmit longitudinal tensile
forces, if necessary, acting as a secondary line
of defense.
The scope of this paper is to analyse the
behaviour of the tunnel under combined fault
induced displacement and strong seismic
shaking. The behaviour of the tunnel subjected
to different earthquake scenarios was
investigated in more detail in Anastasopoulos et
al [2007]. In this paper we parametrically
investigate three possible segment lengths: 70
m, 100 m, and 165 m. The segment length
determines the total number of the required
immersion joints, controlling their deformation
and the overall resilience of the tunnel. The
modern trend is to use longer segments, since
this leads to significant economy in construction.
For example, in the case of the recently built
3.5km resund Tunnel between Denmark and
Sweden [Mashall, 1999] the segment length
was chosen to be 175 m to minimize the cost. A
detailed study showed that lengths below 100
m and above 200 m could not be economical.
Naturally, this kind of optimization is very sitespecific and is also a function of the available
dry-dock facilities. In our case, the combination
of the 67 m depth with the vivid tectonic

environment do not really allow for this type of


optimization : our first priority would be to
secure its feasibility and devise a sound
aseismic design.
Besides the segment length (and hence the
total number of joints), the type of the Gina
gasket was also parametrically investigated.
The longitudinal deformation of the tunnel
depends mainly on the properties of the Gina
gasket. Additionally, since the Gina profile
constitutes the primary seal of the tunnel,
ensuring
its
permeability
is
critical.
Faultinduced
extension
and differential
settlement will unavoidably decompress at least
some of the joints. Additionally, during a strong
seismic event which may occur years later, the
longitudinal vibration of the tunnel will subject
the immersion joints to cyclic (dynamic) recompression
and
de-compression.
The
magnitude of the total de-compression is critical
for the design of the tendons. If the decompression is significant, the tendons will have
to undertake large tensile forces to ensure
impermeability.
Two types of joints (Fig 5) are parametrically
investigated. Type A refers to the idealized
behavior of the largest available Gina profile,
while type B is a proposed hypothetical doublesized Gina-type gasket. The behavior of this
hypothetical gasket provides wider deformation
limits, thus permitting significant additional
compression and decompression. The hyperelastic performance of both joints is estimated
from testing results of half-sized models
[Kiyomiya, 1995]. Already in 1995, Kiyomiya
had foreseen that in the future immersed
tunnels will be built deeper and in seismically
active areas : exactly this case. Our type B
gasket is the logical projection. Of course, it has
not yet been tested.
Finally, the allowance of the shear key is
parametrically investigated. It is expected that
this allowance will affect both the longitudinal
deformation of the tunnel due to fault-induced
deformation, and its transverse deformation
during the strong seismic shaking. If this
allowance is large enough, then the joints will
allow relative displacement and rotation
between two consecutive segments. On the
contrary, reduced allowance makes the
connection more fixed. Two extreme
alternatives are explored : 5 mm and 20 mm.

25 cm

20 cm

36 cm

Gina Type - A

Gina Type - B

40 cm

50 cm

Load (MN/m)

72 cm

10

20

30

40

Compressive Displacement (cm)


Figure 5: Hyperelastic behaviour of rubber gaskets used in the analysis. Type A refers to the largest
available GINA gasket, while Type B constitutes our hypothetical logical projection. The behaviour
has been estimated based on the half-scale tests of Kiyomiya [1995].

METHODOLOGY OF ANALYSIS

The analysis is conducted in three steps.


First, in Step 0 the hydrostatic pressure is
applied statically to the end of the segments,
to simulate the initial hydrostatic longitudinal
compression.
Then, in Step 1 the fault-induced
displacement profile is applied pseudostatically on the model, while
at the final stage, Step 2, the model is
subjected to the asynchronous dynamic
vibration. The methodology for the analysis
of fault rupture propagation, along with the
resulting seabed displacement profiles, is
presented in the next section. Then, in the
following
section,
the
asynchronous
dynamic analysis methodology is discussed
in detail.

The finite element code ABAQUS is utilized to


perform non-linear static and dynamic transient
analysis of the tunnel. The layout of the model
is depicted in Fig 6. The immersed tunnel is
modeled as a series of beams connected to the
soil through interaction springs and dashpots,
and with each other through an arrangement of
hyper-elastic elements simulating the behaviour
of the immersion joints. The tunnel segments
are modeled as beam elements. Special
transitional rigid elements are connected at
both ends of each tunnel segment to represent
the segment face, where the immersion joint is
placed.

Immersion Joint

Immersion Joint

ent
Segm m)
(bea

Fy

z
y

ay(x)

a x(x)

Immersion Joint

Fx

Immersion Joint

Fy
y

ks

Segment (beam)

Profile
Gina

Passive
Failure

slider

slider

Fy

c
Fx

Friction, x

k?

indu
Step 1 : Fault-

t
lacemen
ced disp

Shear Key
Allowance
xi-1 /C

xi /C

xi+1 /C

Step 2 : Asynchronous Seismic Shaking

Figure 6: Finite Element Modeling Layout : In the first step, the fault-induced displacements are applied to the
tunnel pseudo-statically. Then, in a second step we apply the asynchronous seismic excitation.

(expressed as force over displacement per unit


length of the tunnel [kN/m/m]), as functions of
the incident wavelength. For the vertical spring
stiffnesses St. John & Zahrah [1987] utilized the
fundamental solution to the problem of a
surfaceloaded halfspace rather than of a
within-loaded full-space Flamants problem,
instead of Kelvins (see Poulos & Davis, 1974 ;
Davis & Selvadurai, 1996).
The expressions of St. John & Zahrah
give the moduli of subgrade reaction (Winkler
spring stiffnesses) as functions of the
wavelength, a fact that introduces an additional
uncertainty. It is interesting to note that with the
ingenious introduction of wavelength they
accomplished the circumvention of the
singularity of a planestrain solution (both in
fullspace and on halfspace).
However, with immersed tunnels : (a) The
embedment is not sufficient for Kelvins solution
to even approximately apply; the half-space
solutions are more appropriate. (b) he

As foresaid, each tunnel segment is


connected to the overlying soil through
interaction springs and dashpots. The
acceleration time histories are applied to the
ends of these springs-dashpots with a time lag,
as described in detail in the relevant section.
SoilTunnel Interaction Parameters
Assigning proper values to the horizontal (x, y)
and vertical (z) supportingspring constants
(Winkler moduli) is a task with substantial
uncertainty. For embedded tunnels, St. John &
Zahrah [1987] derived an expression based on
integration of Kelvins fundamental solution for
a concentrated (point) load acting within an
infinite elastic medium (full-space), assumed to
be homogeneous and isotropic. Upon
computing the settlement due to load
distributed uniformly across the width of the
tunnel and sinusoidally over a wavelength along
its axis, they derived approximate expressions
for the two horizontal spring stiffnesses

relatively high rigidity of each tunnel segment


(of aspect ratio L/B 3 to 7 in our study) with
respect to the nearsurface (usually soft) soil
will lead to deformation of the segment in
horizontal translation (in the x or y direction)
that may be closer to uniform than to sinusoidal;
hence, the rigid-foundation solution is more
appropriate. (c) The singularity of the planestrain solution exists only in the static problem,
not the dynamic. One additional factor
suppresses the singularity in this case : the
soil modulus is not constant, but increases with
depth [Gazetas, 1983].
Therefore, one can justifiably utilize the
wealth of published elastodynamic solutions for
a rigid long rectangular foundation on halfspace, to obtain not only Winklertype springs
but also dashpots to represent the elastic soil
structure interaction. In this case the
approximate
expressions
for
a
nonhomogeneous half-space proposed by
Gazetas (1991) are utilized. To this end, using
as a starting point the generic profiles of Fig. 4,
we eventually model the effective shear
modulus profile as :

These expressions are considered valid for all


frequencies a reasonable simplification for
translational modes of vibration (e.g. Gazetas,
1991). Moreover, they ignore the effect of
embedment ; this effect is added indirectly
only in the lateral (y) direction (Fig. 6), by
increasing the spring coefficient, ky , by the
elastic sidewall resistance:

ky Efill h/B

where Efill = the (average) Youngs modulus of


the backfill, and h = the effective depth of
embedment (assumed to equal 2/3 of the actual
maximum depth, 4.4 m, of Fig. 2).
A similar approach was advocated by
Vrettos [2005], who also expressed the spring
stiffnesses as constants, independent of
wavelength or frequency.
The vertical, cz , lateral, cy , and
longitudinal, cx , dashpot coefficients, reflecting
the radiation and hysteretic damping in the soil,
are similarly obtained from the expressions and
diagrams of Gazetas [1991]. In view of the
strong soil inhomogeneity and the relatively low
dimensionless frequency parameters, ao = (2
/T) B / Vso , that are of prime interest here,
these coefficients play a minor role in the
response and are not further discussed here.
The tangential contact forces transmitted
from the seabed to the tunnel are limited by
friction
and
passive
resistance.
As
schematically illustrated in Fig. 6, in the
longitudinal direction (x) the behavior of the
interface is approximated with that of a simple
slider of a friction coefficient x. In the
transverse direction (y), the interface is more
complex, with sliding accompanied by passive
type deformation of the backfill. Therefore, the
equivalent friction coefficient, y , is estimated
through 2D plane strain analysis of the tunnel
cross-section.
The (immersion) joints between the tunnel
segments are modeled with non-linear springs.
In the longitudinal direction (x), the springs refer
to the Gina gasket. As already discussed, their
hyper-elastic
restoring
forcedeformation
backbone curves (Fig. 5) are consistent with the
results of half-size model tests [Kiyomiya,
1995]. In the transverse (y) and vertical (z)
directions Gina-type gaskets cannot transfer
shear ; the drift of the tunnel depends solely on
shear key allowance. Thus, the behavior of the

z
G( z ) Go 1 +

(1)

where Go is the shear modulus at z = 0 ; B =


23.5 m is the width of the tunnel ; and , n soilmodel parameters. The three parameters, Go ,
, and n , were obtained by curve fitting not the
initial Gmax profile, but the equivalentlinear G =
profile, i.e. the profile of the shear
modulus in the last iteration of the equivalent
linear wave propagation analysis. The
distributed vertical and horizontal springs are
then obtained as the static vertical, lateral, and
axial stiffnesses of a very long tunnel.
Expressed as stiffnesses per unit length, the
Winkler moduli kz , ky , kx , (in kN/m/m) in terms
of Poissons ratio ( 0.50) are:

kz

0.73
n
Go (1 + 2 )
1 v

(2)
n

2
2
ky
Go 1 +
2 v 3

0.2
kx k y
0.75 v

(3)
n

B 1
1 Go 1 +
L 2

(5)

(4)

1994].
Based
on
the
aforementioned
geophysical investigation, the central part of the
straits, down to the depth of 800 m where the
bedrock was detected, is discretized. The
analysis is conducted in plane strain, with the
use of the FE code ABAQUS. The model is
displayed in Fig 7b, with an H = 800 m thick soil
layer, at the base of which a normal fault
dipping at an angle ruptures and produces a
downward movement of vertical amplitude h.
he total width of the model is B = 4H = 3200
m, following the recommendation of Bray [1990]
that a B : H = 4 : 1 ratio is sufficient to minimize
undesirable boundary effects. At the central
1600 m of the model, the discretisation is finer,
with the quadrilateral elements being 20 m x 20
m (width x height), while a suitable slip-line
tracing algorithm reduces the element size in
the neighborhood of the rupture path. At the two
edges of the model, where the deformation is
expected to be much smaller, the mesh is
coarser : 40 m x 20 m. The differential
displacement is applied to the left part of the
model in small consecutive steps.
Several experimental and numerical studies
have shown that soil behaviour after reaching
failure is a decisive factor in rupture
propagation.
After a thorough review of the literature, an
elastoplastic constitutive model was adopted
[Anastasopoulos & Gazetas, 2007]: MohrCoulomb failure criterion, with an isotropic strain
softening rule, applied to cohesion c , angle of
friction , and angle of dilation . Denoting f
the plastic shear strain at which soil reaches its
residual strength, we consider c, and as
linearly decreasing functions of the total plastic
strain until they reach their residual values cres,
res, and res. Typical values of f range from
5% to 15%. Equally important is the yield
strain y, which depends on both strength and
shear stiffness.
A detailed study on fault rupture propagation
conducted by Anastasopoulos & Gazetas
[2007] has successfully compared the
numerical
results
with
case-histories,
experimental data, and earlier numerical studies.
Additionally, a successful Class A prediction
was published on the Internet utilizing
centrifuge
experiments
performed
on
Fontainebleau sand at the University of Dundee,
as part of a European research project [Davies
& Bransby, 2004]. This, along with additional

joint is modeled with special gap elements,


which would only transmit shear after the shearkey allowance closes, in which case their
stiffness becomes very large, depending mainly
on the stiffness of the concrete section in the
area of the shear key (Fig. 6, bottom left).
ANALYSIS OF FAULT RUPTURE PROPAGATION

We first analyse the fault rupture


propagation through the overlying soil deposit.
As it was previously mentioned, the location
and the magnitude of the potential dislocation of
the ground surface depend not only on the type
and magnitude of the fault rupture, but also on
the geometry and material characteristics of the
overlying soils. Geotechnical exploration in
Rion-Antirrion only reached a maximum depth
of 100 m below the seabed, without
encountering bedrock. To shed more light in the
tectonic environment of the area, and to gain
better knowledge of the alluvium thickness, a
detailed
geophysical
investigation
was
conducted as part of our feasibility study
[Tselentis, 2004]. Before proceeding with the
analysis of fault rupture propagation it is
necessary to discuss briefly the most important
findings.
Given the high seismicity of the area, the
passive tomography method was applied. A
dense network of 70 seismographs was
installed in the project area, to capture seismic
waves oriGinating from micro-tremors (small
magnitude
earthquakes).
A
tomography
imaging technique was used to locate
potentially active faults. The seismographic
network indicated no seismic activity at depths
less than 1.5 km, verifying the hypothesis that
faults within the straits do not outcrop on the
seabed. As depicted in Fig 7a, five rupture
zones were detected within the straits, with
direction practically perpendicular to the tunnel
axis. Close to the side of Antirrion, the mapped
Antirrion fault was also detected. With an
extension slip rate of 0.7 mm/yr, it constitutes
the most important tectonic feature in the
immediate neighborhood.
The finite element method has proven
successful
in
analysing
fault
rupture
propagation through soil provided that certain
conditions are satisfied, such as (i) the use of a
very refined mesh in the neighborhood of the
potential rupture, and (ii) the use of a suitable
nonlinear constitutive law for the soil [Bray,

In both cases the shear modulus G was


assumed linearly increasing with depth while y
is kept constant. The dense soil reaches
failure at relatively low strains (brittle
behaviour), while the loose soil accommodates
relatively high strains before yielding (ductile
behaviour).

verification of the developed models against


case histories and experiments from the
published literature, provided confidence in
using our numerical methodology.
Conservatively assuming a dip angle of 45,
a downward displacement of vertical magnitude
h is imposed on the left half of the model (the
hanging wall). As already mentioned, the soil
profile is known only for the upper 100 m :
alternating layers of sandy gravel to gravel, silty
sand to sand, and silty clay to clay. Given these
uncertainties, we parametrically investigated
two idealized soils :
(a) dense cohesionless soil : = 45, res =
30, = 15, res = 0, and y = 1.5%
(b) loose cohesionless soil : = 30, res = 25,
= 5, res = 0, and y= 5.0 %

To be on the conservative side, despite the


optimistic geophysical results, we assumed h =
2 m as the design bedrock displacement
(vertical component). Note that for very small
values of the base fault displacement h, relative
to the soil thickness H, the rupture cannot reach
the ground surface. Given the significant 800 m
depth, a bedrock displacement of 2 m
corresponds to an h/H = 0.25 %. The required

Distance (km)
-4
0

-2
2000

Immersed Tunnel

3000

1000

5000

(a)

Vp (m/s)

Fa u
lt

2000

on

Depth (m)

4000

4000

An
tirr
i

3000

RION

ANTIRRION

B = 4H = 3200 m

H = 800 m

Hanging wall

(b)

Figure 7: (a) Mapping of active faults in the area of the crossing, based on the geophyical tomography
[Teslentis, 2004], (b) Finite element discretization for the plane strain analysis of fault rupture
propagation through the 800 m soil sediment.

10

overlying structures. The horizontal strain, x ,


along the ground surface (positive values are
for tension) is depicted in Fig 9d.
Observe that the ground surface deforms
smoothly, as the rupture cannot reach the
surface. The bending deformation, expressed
through the distortion angle (Fig 9c), reaches
its peak, 0.26%, at a distance d 200 m from
the straightline projection of the rupture. On the
other hand, the horizontal (tensile) deformation
is maximized, x = 0.25%, at a distance d 500
m (Fig 9d). The computed herein displacement
profiles (x and y) are used as the input
displacement to be imposed on the tunnel
model. The tensile deformation causes
decompression of the joints, while the bending
deformation decompresses some of the
immersion joints and possibly compresses
further some other. Since the maximum
bending and tensile deformation does not occur
at the same location, we identify two tunnel
fault rupture relative positions (Fig 9) :
(1) Position 1 : the center of the tunnel
coincides with the location of max ,
and
(2) Position 2 : the center of the tunnel
coincides with the location of max x .
The prediction of the worst-case scenario is
not straightforward. Therefore, both positions
are investigated.
Typical results of a complete analysis are
portrayed in Figure 10, in terms of joint
deformation, sliding displacement, bending
moments and axial forces. Figure 10a
corresponds to segment length L = 70 m with a

h/H ratio for the outcropping of the rupture is an


increasing function of soil ductility, expressed in
our modified MohrCoulomb constitutive law,
through y. Even for a the brittle soil (a) with y
= 1.5 %, an h/H in the order of 0.75% is
required for the fault to break out
[Anastasopoulos & Gazetas, 2007].
EFFECTS OF FAULT RUPTURE ON THE TUNNEL

Characteristic pictures of the propagating


rupture in a deposit of idealized dense sandy
soil (y = 1.5 %) is given in Fig 8 in the form of
four snapshots of the deformed mesh, along
with the accumulated plastic strain, for different
bedrock displacements. Darker regions denote
higher plastic strains. Observe that for bedrock
displacement up to h = 4 m (i.e. h/H = 0.50%),
the propagating rupture does not emerge on the
surface (seabed). Only when the imposed
bedrock displacement exceeds 6 m (i.e, h/H =
0.75 %) does the dislocation (barely) emerge
on the surface. For the design displacement of
2 m the rupture is clearly far from reaching the
seabed (model surface). The situation is even
better with the idealized loose sandy soil : an
h/H of the order of 1.5% is needed for the fault
to outcrop [Anastasopoulos, 2005].
Figs 9a and 9b portray the profiles of vertical,
y, and horizontal displacement, x, along the
ground surface, for the (conservative) dense
soil and for hd = 2 m. The horizontal distance, d,
is measured from the point of application of the
bedrock displacement. Fig 9c depicts the slope
of the ground surface a useful response
parameter in assessing the damage potential to

h=1m

h=2m

h=4m

h=6m

Figure 8: Snapshots of deformed mesh and plastic strain for a fault rupture = 45o in dip and bedrock
displacements h/H = 1, 2, 4, and 6 m, for the idealized dense sandy soil.

11

Horizontal distance d (m)


-400

400

800

-1600 -1200 -800


0

1200 1600

-0.5

-0.5

-1

-1

x (m)

y (m)

-1600 -1200 -800


0

Horizontal distance d (m)

-1.5

-400

800

1200 1600

-1.5

-2

-2

Position 2

Position 1
-2.5

-2.5
0.4

0.4

Position 1

0.3

0.3

0.2

0.2

x (%)
( )

(%)

400

0.1

Position 2

0.1
0

-0.1
-1600 -1200 -800

-400

400

800

1200

1600

Horizontal distance d (m)

-0.1
-1600 -1200 -800

-400

400

800

1200

1600

Horizontal distance d (m)

Figure 9: Fault rupture propagation analysis results for a fault rupture = 45o in dip and bedrock
displacements hd = 2 m, for the idealized dense sandy soil : (a) horizontal displacement x, (b)
vertical displacement y, (c) angular distortion , and (d) horizontal strain x at the seabed.

2. This initial slippage is greater in the middle


of the tunnel, where the maximum tensile
deformation x takes place. Increasing the
segment length only marGinally increases
the maximum sliding displacement (from 7.5
cm to 9 cm in the middle of the tunnel for
segment lengths L = 70 m and L = 100 m
respectively).
3. At first, the axial force exhibits an initial prestressing of 160 MN due to hydrostatic
compression (Step 0). Then the application
of the faultinduced deformation (Step 1)
causes longitudinal decompression of the
tunnel, reducing the axial force (N)
significantly. In the case of the 70 m
segments, N drops from 160 MN to only 30
MN near the middle. The increased
segment length makes things worse: the
initial compression is completely lost, and
even some (small) tensile stressing
develops (at point C).

type A gasket, while Figure 10b to segment


length L = 100 m with the type B gasket.
Results for L = 165 m have been also examined
but are not shown herein: with such a large
length the tunnel cannot sustain the total
developing stressing without several joints
experiencing net tension a precarious
situation indeed. Similarly, the allowance in
the shear keys plays only a minor role; thus,
results are given only for a single value (5 mm)
of this allowance. In all cases the results
presented herein correspond to the position 2
scenario since this proved to be the most critical
case.
The following conclusions are drawn:
1. The application of the faultinduced tensile
displacement decompresses all the joints,
with those near the middle of the tunnel
experiencing the greatest tension with more
pronounced being the decompression when
the segment length is 100 m (Figure 6b).

12

methods: (i) making use of the widely accepted


equivalent linear approximation code SHAKE
[Schnabel et al, 1975], and (ii) applying the
nonlinear
constitutive
model,
BWGG,
developed by Gerolymos & Gazetas [2006].
Having estimated the free-field acceleration
time histories at the seabed, we apply the
methodology of EC8 [2002] to impose the
acceleration time histories on the tunnel
supports with a time lag. Given the length of the
tunnel ( 1 km), seismic shear waves incident at
an angle different than 90o to the horizontal,
appear as traveling with a finite propagation
velocity, C , along the surface (and hence
along the tunnel). This time lag leads to
asynchronous vibration of the tunnel segments.
It is computed through the apparent wave
velocity C. If we denote with x the longitudinal
axis of the tunnel, at a distance xi from the

4. The
fault-induced
deformation
is
responsible for the development of
longitudinal bending moments, My. With L =
70 m, at point A a moment of 70 MNm is
developed while point C (at the opposite
end) is almost unaffected by the imposed
displacement (My = 5 MNm). The increase
of L to 100 m leads to an increase of the
longitudinal bending moment.
METHODOLOGY OF DYNAMIC ANALYSIS

Fig 11 illustrates schematically the methodology


used in our dynamic analysis. The immersed
tunnel is modeled as a beam connected to the
soil through interaction springs and dashpots.
Free-field acceleration time histories are
computed trough 1-D wave propagation
analyses, that were conducted with two

(a)

CC

BB

70 m segment length- type B


A gasket
Joint
Decompression (cm)

Segment Slippage
(cm)

N (MN)

M (MNm)

-130

-70

16

7.5

-30

40

10

-50

(b)
AA

C C

BB

100 m segment length- type B gasket


Joint
Decompression (cm)

Segment Slippage
(cm)

N (MN)

M (MNm)

-100

-150

21

-9

-80

-10

14

-2

10

140

Figure 10: Joint deformation, sliding displacement, axial forces and bending moments after the fault induced
displacements for (a) 70 m segments and (b) 100 m segments. Fault at Position 2, type B gasket,
and 5 mm shear-key allowance.

13

which compares quite well with the conservative


design spectrum of the neighboring Rion
Antirrion cable-stayed bridge. The (oriGinal)
Aegion 1995 and the Lefkada 2003
accelerograms are among the strongest ever
recorded in Greece. The Aegion accelerogram,
at a distance of only 28 km from the site, is
characterized by a single long-period pulse of
0.54 g one of the greatest longperiod PGAs
ever recorded in Greece. The Lefkada
accelerogram (recorded at a distance of 120 km
from the straits), is characterized by a PGA of
0.43 g and six strong cycles with acceleration in
excess of 0.30 g.
Two types of 1-D verticalwave propagation
analyses were conducted : (a) by applying the
widely accepted equivalent linear method of
analysis with the SHAKE code [Schnabel et al,
1975], and (b) by applying a nonlinear inelastic
constitutive model for the soil, coded in the
program NLDYAS [Gerolymos & Gazetas,

beginning, the seismic excitation will arrive with


a time lag ti = xi / C (Fig 11). The apparent
wave velocity was conservatively assumed
equal to 1000 m/s (Eurocode EC8 Part 2
(Bridges) suggests apparent wave velocities C
> 1500 m/s, even for the softest soil category).

Analysis of Soil Response

As depicted in Fig 12, three real earthquake


records were used for the 1-D wave
propagation analyses : (a) the JMA record in
the 1995 Ms 7.2 Kobe earthquake, (b) the
(unique) record of the 1995 Aegion Ms 6.2
earthquake, and (c) the (unique) record of the
2003 Lefkada Ms 6.4 earthquake. They were all
scaled to a peak acceleration of 0.24 g, as
specified by the Greek Seismic Code [EAK
2000] for the area of study. The first record,
Kobe JMA, has a (scaled) response spectrum
L = 980 m

x
a

xi

t
EC8

xj

xi /C

xj /C

L /C

C : apparent wave velocity


at distance xi , time lag ti = xi/C

0.48 g

SHAKE

G
G

Depth (m)

100

WGG
Dense
Sand

High Plasticity
Clay

Loose
Sand
0

0.24 g

250

500

750

1000

1250

Vs (m/sec)

Kobe JMA

Figure 11: Dynamic Analysis methodology : The bedrock acceleration is analysed in 1-D to derive the
acceleration at the seabed. Then, following the EC8 methodology, the acceleration time histories
are imposed with a time lag on the tunnel model.

14

while the Lefkada accelerogram yields 0.63 g.


Five more possible soil profiles were
parametrically investigated. From all of the
examined soilprofile and seismicexcitation
scenarios, the PGAs were found to range
between 0.40 g and 0.63 g. The three
computed seabed accelerograms of Fig 12
were used as the free-field input for the
dynamic analysis of the tunnel. Interestingly,
the design acceleration levels of the Rion
Antirrion (cablestayed) Bridge (PGA 0.50 g)
are similar to the PGA values of this figure.

2005]. The equivalent linear is definitely the


most popular method, having however certain
limitations, especially in the case of very deep
deposits (as in our case). On the other hand,
most nonlinear soil models are incapable of
simultaneously reproducing the observed shear
modulus decrease and damping ratio increase,
usually overestimating the hysteretic damping
at large strains (if the Masing rule for
unloadingreloading is used). The recently
developed NLDYAS code uses a constitutive
model, denoted BWGG, which avoids such
disadvantages, and is capable of reproducing
even some of the most complex non-linear
characteristics of cyclic behavior, such as cyclic
mobility and liquefaction.
Fig 12 illustrates the 1-D wave propagation
results for one of the worsecase scenarios :
stiff soil profile with Vs = 1300 m/s below 100
m depth. The Aegion and the Kobe JMA
records produce peak ground acceleration
(PGA) in the order of 0.50 g at the seabed,

COMBINED EFFECTS OF FAULT RUPTURE AND


(SUBSEQUENT) DYNAMIC EXCITATION

Typical results of a complete analysis are


portrayed in Figs 1316, in the form of time
histories of accelerations, internal forces,
displacements
and
deformations,
which
develop at various points along the tunnel
during the strong seismic shaking . Static

0.6

0.6

0.3

0.3

-0.3

-0.3

-0.3

a (g)

0.48 g

-0.6
0

0.3

-0.6

10

-0.6

15

0.63 g

0.6

0.50 g

t (sec)

10

15

t (sec)

250

10

15

t (sec)

Profile C (hard)

100

1300

250

500

750

1000 1250

Vs (m/sec)

0.6

a (g)

JMA Kobe 1995

0.6

Aegion 1995

0.24 g

0.6

0.3

-0.3

-0.3

-0.3

0.24 g

-0.6

-0.6

-0.6

10

t (sec)

15

Lefkada 2003
0.24 g

0.3

0.3

10

t (sec)

15

10

15

t (sec)

Figure 12: 1-D dynamic wave propagation analysis results for the worst-case scenario : stiff soil profile,
with Vs = 1300 m/s below 100 m depth.

15

does not appear to have a significant effect on


the longitudinal accelerations of the tunnel.
In the transverse direction the response is
also differentiated along the length of the tunnel.
At the two ends (B, D), the acceleration time
histories are quite similar to the input motion (A).
At these points, the immersed tunnel is forced
to follow the input excitation, as it is rigidly
connected to the bored approach tunnels. The
latter are as a first approximation assumed to
follow the imposed displacements. However,
the central part of the tunnel (C) exhibits a
different behaviour. With the exception of a few
extremely high-frequency acceleration spikes,
attributable to gap closures at the shear-keys,
the acceleration is cut-off at about 0.30 g, a
value equal to the transversal friction
coefficient y. Hence, this acceleration cut-off
implies transverse sliding of the tunnel. As
expected, this sliding takes place only near its
center, as the two ends are restrained by the
bored tunnels. Again, the segment length does
not affect significantly the lateral acceleration of
the tunnel. It is also noted (although not shown
here) that the tunnel response accelerations are
insensitive to the gasket type.
The sliding of the tunnel (longitudinally and
transversally) is evident in Fig 14. In the
longitudinal direction, the tunnel segments have
already slipped before the application of the
dynamic excitation (Step 2), due to the fault
induced longitudinal deformation (Step 1). This
initial slippage is greater in the middle of the
tunnel, where the maximum tensile deformation
x takes place, and increases with segment
length (from 7 cm in the case of the 70 m
segments, to about 12 cm for the longer
segments). The increase of segment length
also leads to increased additional dynamic
relative displacements. While for the 70 m
segments the additional sliding due to the
dynamic excitation reaches 10 cm, in the case
of the 100 m segments the maximum relative
dynamic displacement exceeds 17 cm (location
B). In all cases, however, the permanent
(residual) displacement is significantly lower
than the maximum value. In the transverse
direction, the central segments of the tunnel
experience substantial sliding, while the ends
keep pace with the overlying soil just as
was the case with the transversal acceleration
time histories in Fig 13. Increasing the segment
length (from 70 m to 100 m) only marginally

permanent deformation and distress is


assumed to have already taken place in a
preceding event of fault rupturing underneath
the site of the tunnel.
Each of these figures presents results for
two different values of segment length, 70 m
and 100 m . Results for 165 m, which was also
examined, are not shown : with such a large
length the tunnel cannot sustain the total
developing stressing without several joints
experiencing a net tension a precarious
situation indeed.
Only results for the (scaleddown) Kobe
JMA excitation are shown in the said figures.
Although using the other motions as excitation
leads to slightly different results, the qualitative
behaviour of the tunnel remains the same and
the conclusions drawn herein do not change.
Similarly, the allowance in the shear keys
proved to play only a minor role ; thus, results
are given only for a single value (5 mm) of this
allowance . Finally, only one location of the
emerging fault rupture, position 2, is examined
in the aforementioned figures. This is the most
critical of the two studied locations.
Each figure presents the results for the
longitudinal (x) and the transverse (y)
response ; the former shown above and the
latter below the sketch of the tunnel.
Specifically, the details of each figure are as
follows :
Fig 13 shows the time histories of
acceleration which develop in the middle of
three segments of the tunnel (B,C,D), and
contrasts them with the ground excitation (point
A). Observe that in the longitudinal direction (x)
the response of the tunnel differs from point to
point. The peak acceleration is amplified from
0.48 g at the base to approximately 0.60 g near
the terminal segments (at the two edges of the
tunnel, B and D). Near the center, however, the
longitudinal acceleration is cut-off to about 0.30
g, implying sliding. This differential longitudinal
response is attributed to the previous non
uniform (along the axis) displacement from the
fault rupture and the ensuing tendency of the
(springrestrained)
tunnel
segments
to
readjust under the dynamic loading. This is
explained in detail in the next section. If the
faultinduced displacement did not pre-exist
(Step 1), then the dynamic response of the
tunnel, in the longitudinal direction, would be
nearly uniform. The increased segment length

16

0.60

0.6 A (base)

0.6

0.3

0.3

0.3

0.3

ax (g) 0

-0.3

-0.3

-0.3

-0.3

0.48 g

-0.6

0.6

-0.6

-0.6

10

0.6

-0.6

10

10

10

10

10

t (sec)
B
(a)

A
0.6

0.6

A (base)

0.6 C

0.6

0.3

0.3

0.3

0.3

ay (g) 0

-0.3

-0.3

-0.3

-0.3

-0.6

-0.6
5

10

-0.6
5

10

-0.6
5

10

t (sec)
0.6

ax (g)

0.6

A (base)

0.60

0.6

0.6 D

0.3

0.3

0.3

0.3

-0.3

-0.3

-0.3

-0.3

0.48 g

-0.6
0

-0.6

-0.6
0

10

10

-0.6
0

10

t (sec)

(b)
A
0.6

ay (g)

0.6

A (base)

0.6

0.6

0.3

0.3

0.3

0.3

-0.3

-0.3

-0.3

-0.3

-0.6

-0.6
0

10

-0.6
0

10

-0.6
0

10

10

t (sec)

Figure 13: Longitudinal and transverse acceleration time histories for : (a) 70 m segments, and (b) 100 m
segments (Kobe JMA excitation, fault at Position 2, type B gasket, and 5 mm shear-key
allowance).

17

x (cm)

30 A

30 B

20

20

10
0

30
20

20

10

10

10

17 cm

-10

-10
0

30 D

-10
0

10

-10

10

10

10

t (sec)

(a)

y (cm)

10

10 B

10

10

-5

-5
0

10

-5
0

10

-5
0

10

10

10

x (cm)

t (sec)
27 cm

30 A

30

20

20

20

20

10

10

10

10

30

-10

-10
0

-10
0

10

30 D

10

-10
0

10

t (sec)

(b)

y (cm)

10 A

10 B

10 C

10 D

-5

-5
0

10

-5
0

10

-5
0

10

10

t (sec)

Figure 14: Longitudinal and transverse sliding displacements for : (a) 70 m segments, and (b) 100 m
segments (Kobe JMA excitation, fault at Position 2, Type B gasket, and 5 mm shear-key
allowance).

18

thick cross-section of the tunnel can easily


undertake such a stressing, even with the
minimum allowable reinforcement. While My is
insensitive to the thickness of the rubber gasket
(Type A versus B), the increase of joint
allowance to 2 cm relieves the segments from
roughly 10 % of the transverse bending.
The longitudinal and transversal deformation
(x and y) of the immersion joints are the most
crucial response parameters for the seismic
safety of an immersed tunnel. Their time
histories during the seismic shaking after the
fault rupture has already had its effect are
portrayed in Fig 16, for the 70 msegment
tunnel. We compare the response, of the joints
equipped with the Type A gasket, versus that of
the joints with Type B gasket. Notice that in the
longitudinal direction, the Gina gasket
experiences an initial hydrostatic compression
(Step 0) of 17 cm and 28 cm, for Type A and B
gasket, respectively. The previous application
of the faultinduced tensile displacement
opens-up all the joints, with those near the
middle of the tunnel experiencing the greatest
tension. In the case of the Type A gasket, the
17 cm hydrostatic compression is practically
completely lost at the central part of the tunnel
(point C), where the leftover compression after
the fault rupture amounts to merely 2 cm. The
situation near the terminal segments A and D is
less critical tragic, with the remaining
compression of the gasket being 14 cm and 8
cm, respectively. During shaking, the gaskets
experience
alternating
cycles
of
decompression and re-compression. While the
dynamic compression is acceptably small, there
is some tensile deformation in gaskets A and C
a rather precarious situation.
The use of Type B rubber gasket improves
substantially the performance. One of the
reasons, of course, is the 28 cm of initial
hydrostatic compression. Much greater margins
for de-compression are thus available. The
fault-ruptureinduced de-compression of the
joints, still leaves a minimum of 12 cm
compression at the critical central part of the
tunnel,
C.
As
expected,
the
larger
decompression is observed at joints near the
middle of the tunnel, where the fault-induced
tensile deformation of the seabed is maximum ;
near the end segments unloading of gaskets is
substantially less. During dynamic oscillation,
the rubber gaskets experience cycles of

increases the maximum sliding displacement


(from 9 cm to 11 cm). In all cases, the
permanent (residual) sliding displacement does
not exceed 3 cm near the center of the tunnel,
and it is practically negligible at the two ends.
The transverse sliding displacement is also
insensitive to the thickness of the rubber gasket
(Type A versus B).
Fig 15 illustrates the bending moments and
axial forces in characteristic cross-sections of
the immersed tunnel. At first, the axial force
exhibits an initial pre-stressing of 160 MN due
to hydrostatic compression (Step 0). Then the
application of the faultinduced deformation
(Step 1) causes longitudinal decompression of
the tunnel, reducing the axial force (N)
significantly. In the case of the 70 m segments,
N drops from 160 MN to only 30 MN near the
middle. The increased segment length makes
things worse : the initial compression is
completely lost, and even some (small) tensile
stressing develops (at point C). The
readjustment of the segments during dynamic
shaking (thoroughly explained in the sequel),
leads to a redistribution of the longitudinal axial
force. Segments with higher axial forces tend to
get unloaded, while those subjected to tension
(smaller compression) tend to attract higher
compression. At the end of shaking, this
redistribution leads to a nearly uniform
distribution of the axial force : about 60 MN in
the case of the 70 m segments and 40 MN for
the 100 m segments. During shaking, a
maximum net tension of 10 MN takes place with
the 100 m segments (A).
Apart from the axial forces, the fault-induced
deformation is also responsible for the
development of longitudinal bending moments,
My . With the 70 m segments the application of
the faultinduced deformation creates a
bending moment of 60 MNm, which during
seismic shaking fluctuates by barely 15 MNm.
Increasing the segment length to 100 m leads
to a significant increase of My to 150 MNm, with
no effect on the dynamic fluctuation. The
bending moments are also insensitive to the
type of the rubber gasket.
The transverse bending moments, Mx are
not plotted here. Such moments do not develop
from the (normal to the axis) faultrupture.
Increasing the segment length from 70 m to 100
m, significantly increases My from 250 MNm to
400 MNm. In all cases, the 23 m x 11 m, 1.5 m

19

tt (sec)
(sec)
-2

N (MN)

-2

10

10

-2
0

-50

-50

-50

-100

-100

-100

-150

-150

-150

-200

F*

Dynamic

F*

-200

Dynamic

10

F*

-200

Dynamic

(a)

My (MNm)

200

200

200

100

100

100

-100

-100

F*

-200

-2

Dynamic
0

-100

F*

-200
8

10

-2

Dynamic
0

F*

-200

Dynamic

10

-2

10

10

-2

10

10

t (sec)

* : fault induced displacement

t (sec)
-2

N (MN)

-2

10
0

-50

-50

-50

-100

-100

-100

-150

-150

-150

F*

-200

Dynamic

F*

-200

Dynamic

F*

-200

Dynamic

(b)

My (MNm)

200

200

Dynamic

200

100

100

100

-100
-200

-100

F*
-2

-100

F*

-200
0

10

-2

Dynamic
0

F*

-200
8

10

-2

Dynamic
0

t (sec)

Figure 15: Axial Force and Bending Moment time histories for : (a) 70 m segments, and (b) 100 m segments
(Kobe JMA excitation, fault at Position 2, type B gasket, and 5 mm shear-key allowance).

20

x (cm)

30 A

30

30

20

20

20

20

10

10

10

10

0
0

30

0
0

10

10

0
0

10

10

t (sec)

(a)

y (cm)

10

10 B

10

10

-5

-5

-5

-5

-10

-10
0

10

-10
0

10

-10
0

10

10

x (cm)

t (sec)
30 A

30 B

30 C

30 D

20

20

20

20

10

10

10

10

0
0

0
0

10

10

0
0

10

10

10

t (sec)

(b)

y (cm)

10 A

10 B

10 C

10 D

-5

-5

-5

-5

-10

-10
0

10

-10
0

10

-10
0

10

t (sec)

Figure 16: Longitudinal and transverse joint deformation for : (a) 70 m segments with Type A gasket, and
(b) 70 m segments with Type B gasket (Kobe JMA excitation, fault at Position 2, and 5 mm
shear-key allowance).

21

coinciding with the location of maximum x), the


tectonic deformation leads to a de-compression
of all the joints (Figs 17a, c, and d). On the
other hand, when the tunnel is placed at
Position 1 relative to the fault rupture (center of
the tunnel coinciding with the location of max.
), most joints de-compress, except those near
the left edge of the tunnel (Fig 16b). At this
location, the sagging deformation of the seabed
seems to be the most important, imposing some
further compression of the joints. More
specifically, in the case of the 70 m segment
tunnel with the ( thick )Type B rubber gaskets
at Position 2 (Fig 16a), the joints decompress
from their initial hydrostatic compression of 28
cm to a minimum of 12 cm (a decompression of
16 cm) near the center of the tunnel (where
maximum tensile deformation occurs). This
decompression is reduced significantly near the
left edge of the tunnel (x 26 cm), but not to
the same extent at the right (x 18 cm).
As seen in Fig 17c, the thickness of the
rubber gasket does not substantially affect the
de-compressed
profile.
But
while
decompression is practically the same, in the case
of the (slimmer) Type A rubber gasket, the initial
hydrostatic compression is only 18 cm. As a
result, the joints near the center almost loose all
their pre-stressing (x 2 cm). This is
threatening for the watertightness of the tunnel !
Increasing the segment length to 100 m
(combined with a Type B gasket) improves the
safety margin : xmin 6 cm near the center of
the tunnel (Fig 16d) : This effect can be easily
explained : while in the case of the 70 m
segments the imposed deformation is absorbed
by a total of 13 joints, in the case of the 100 m
segments the same deformation is transmitted
to only 9 joints. Obviously, this decrease of the
number of the joints, unavoidably increases the
opening of each joint. Observe that the total
de-compression (of all joints) remains
practically the same : xtotal 1.5 m.
n insight into the aforementioned dynamic
readjustment of the tunnel segments can be
developed with the help of Fig 18, which
illustrates
the
history
of
longitudinal
deformations of six characteristic (Type B) joints
along the 70 m segment tunnel, subjected to
fault rupturing at Position 2. Initially, when only
the hydrostatic compression has been applied
(Step 1), all joints are compressed by exactly
the same amount, reaching x 28 cm.

decompression and re-compression. The total


dynamic plus static compression does not, in
any case, exceed the acceptable limits of the
Type B gasket : x (max) (29 cm). More
importantly, the gasket remains always
compressed, maintaining its water tightness :
(x)max (10 cm), in all joints. A redistribution of
joint deformation is readily observed :
overstressed joints tend to de-compress;
understressed to re-compress, so that the
safety of the immersed tunnel improves after
strong shaking.
This is a fundamental conclusion, observed
in all of the parametric cases analysed.
Moreover, it is completely explainable
theoretically. Its practical significance and its
generality should not be underestimated : a
detrimental static de-compression of a few of
the joints (arising from a rupturing fault, or from
differential settlement, or from any other case)
will be relaxed from the tunnel oscillations
during a strong seismic shaking.
In the transverse direction, the deformation
of the joints is restrained by the shear-key
"allowance" of only 5 mm. Deformations greater
than this value imply elastic deformation of the
shear key concrete. The terminal joints
experience the greatest relative displacement
y, with the maximum value being practically
equal to the shear-key allowance ( 5 mm).
Near the center, the transversal displacement
never exceeds 2 mm. These displacements are
mainly due to rotation between segments. Near
the two ends of the tunnel, where the rotation is
restrained by the stiff connection to the bored
approach tunnels, the shear keys experience
closure, without significant gapping taking
place. The transversal deformation is hardly
affected by the gasket type; it depends mainly
on the shear-key allowance.
DYNAMIC RE-ADJUSTMENT OF TUNNEL
SEGMENTS : A HEALING PROCESS

As already cited, during the asynchronous


dynamic oscillation of the tunnel, a
redistribution of total deformation among joints
is observed. This section further discusses this
interesting phenomenon.
As depicted in Fig 17, the application of the
faultinduced dislocation leads to a decompression of the immersion joints. When the
tunnel is at Position 2 relative to the
propagating fault rupture (center of the tunnel

22

40

40

(a)
Initial Hydrostatic Compression

Initial Hydrostatic Compression

30

x (m)

x (m)

30

(b)

20

10

20

10
Hydrostatic

Hydrostatic

Position 2

0
0

200

Fault
Position
1

Fault

0
400

600

800

1000

200

Horizontal distance d (m)


Hydrostatic

40

400

600

800

(c)

Fault

40

Hydrostatic

(d)

Hydrostatic

Fault

Fault

Initial Hydrostatic Compression

30

x (m)

x (m)

30

Initial Hydrostatic Compression

20

1000

Horizontal distance d (m)

10

20

10

Position 2

0
0

200

Position 2

0
400

600

800

1000

Horizontal distance d (m)

200

400

600

800

1000

Horizontal distance d (m)

Figure 17: Fault-induced decompression of the immersion joints : (a) 70 m segments with Type B gasket, at
Position 2 (max tensile displacement), (b) 70 m segments with Type B gasket, at Position 1
(max bending displacement), (c) 70 m segments with Type A gasket, at Position 2, and (d) 100
m segments with Type B gasket, at Position 2. 1

time (Step 1), joint 7 is significantly


decompressed to 12 cm : point a7 in the graph.
Now look into the hyperelastic load-deformation
curves of the two joints (Fig 5b). Joint 1, after
the application of the faultoriGinating
displacement remains sufficiently compressed
to lie on the stiffer hyperelastic regime, where
the stiffness is Khyper 50 MPa. By contrast,
joint 7, having decompressed significantly, lies
on the initial soft (elastic) branch, of Kel 3
MPa. As a result, during the dynamic shaking,
joint 1 will be an order of magnitude stiffer than
joint 7 ! Since the joints are connected in series,
the stiffer ones simply force the softer to
compress. In other words, over-compressed
joints (stiffer) force the under-compressed joints
(softer) to re-compress. Of course, such a recompression leads in turn to hyperelastic
stiffening, making the latter stiffer. Finally, an
equilibrium is reached, with all joint
compressions
being
substantially

Application of the faultinduced dislocation


(Step 2) leads to a differential de-compression
of the joints, as already explained in Fig 17. At
Step 3 the dynamic oscillation tends to
redistribute the joint compressions and relocate
the tunnel segments. Observe that joint 1,
which had experienced the least decompression, now experiences the greatest decompression during dynamic oscillation. On the
contrary, joint 7 at the center of the tunnel, with
the largest faultinduced de-compression, is recompressed the most during shaking.
This surprisingly favorable performance is
partly attributable to the hyper-elastic behaviour
of the immersion joints, combined with the
ability of the tunnel segments to slide over the
seabed. Let us now look into the performance
of joints 1 and 7 in more detail. As depicted in
Fig 18, the faultinduced stressing only slightly
decompresses the rubber gasket of joint 1 (from
28 to 26 cm) : point a1 in the graph. At the same

23

safely resist :
a normal earthquake fault rupture with a
dislocation (offset) of 2 meters in the
basement rock, 800 m underneath the
tunnel.
a subsequent strong ground shaking
arising from a different significant
seismic fault not crossing the site, but
producing peak acceleration of at least
0.50 g.
The cumulative effect of the two events,
even in the worst sequence : fault
rupture followed by shaking.
[2] The initial hydrostatic compressive force is
independent of segment length, or of total
number of the joints. Increasing the total
number of the joints (or decreasing the
segment length) leads to increased total
initial
hydrostatic
compressive
deformation of the tunnel. To ensure

homogenized. At the end of shaking, the


stiffer joint 1 exhibits a residual de-compression
at x 20 cm (point b1 in the graph), while the
softer joint 7 is re-compressed to x 13 cm
(point b7). It must be pointed out that such a
behaviour would hardly be possible if it were
not for the tunnel sliding : If the segments were
fixed to the seabed, then the stiffness of the
joints would play no, or very minor, role.
SUMMARY CONCLUSIONS

Several conclusions of practical significance


can be drawn from the presented study
(including additional parametric results not
shown here for the sake of brevity) :
[1] A properly designed immersed tunnel
(suitable thick elastic gaskets, small length
of segments, shear keys with sufficient
allowance, un-stressed tendons) can

30

Step 0 :

Step 1 :

Step 2 :

Hydrostatic
Compression

Fault-induced
Displacement

Dynamic
Oscillation
1
a1

11

25
2
b1

20

3
11

15

b7

9
a7

10
Load (MN/m)

1
Load (MN/m)

x
(cm)

Khyper

4
2
0

b1

Kel
0

10

20

a1
30

Khyper

4
2
0

40

Kel
0

x (cm)

0
-6

-4

-2

a7 b7
10

20

30

40

x (cm)

10

t (sec)

Figure 18: Longitudinal joint deformation for the 70 m segment tunnel, equipped with the thick Type B
rubber gasket, and 5 mm shear-key allowance, at Position 2 (maximum tensile deformation)
relative to the fault rupture. While the fault-induced displacement in Step 1 opens the joints, the
asynchronous dynamic shaking tends to relocate the segments and allow for a redistribution of
the gasket deformations : observe the homogenization of the residual compression of the joints.

24

[3]

[4]

[5]

water-tightness, the initial compression of


the Gina gaskets must be significant.
When subjected to faultinduced tensile
displacements, their ability to remain
water-tight is highly dependent on the
extent of the initial compression to avoid
developing net tension. On the other
hand, the additional dynamic compression
must not exceed the capacity of the joints.
The dynamicallyinduced longitudinal
deformation of the immersion joints
depends on the segment length and the
thickness of the Gina gasket. Increasing
this length unavoidably increases both the
faultinduced
and
the
dynamic
deformation of the immersion joints. Since
the tunnel segments are significantly
stiffer than the Gina gaskets, they tend to
behave as rigid blocks and most of the
imposed deformation is absorbed in the
joints. Obviously, decreasing the number
of the joints, increases their deformation.
Increasing the thickness of the Gina
gasket leads to greater initial hydrostatic
compression : from 17 cm for the Type A
gasket, to 28.5 cm for the Type B. Since
the tectonic deformation is mainly tensile,
this increase of the initial compressive
deformation leads to higher margins of
safety. With seismic shaking, the dynamic
(additional) re-compression and decompression of the gaskets is not so
sensitive to the gasket thickness.
During (asynchronous) dynamic shaking,
tunnel segments tend to re-adjust
themselves by sliding longitudinally over
the seabed. This is accompanied by a
redistribution of the longitudinal joint
deformation leading to a more uniform
profile of compression along the axis. This
surprising
healing
behaviour
is
attributable (to a substantial degree) to the
hyperelasticity of the rubber gaskets,
combined with the capability of the
segments to slide on the seabed. Thus,
while the fault displacements lead to a
detrimental de-compression different from
joint to joint, upon subsequent seismic
shaking
the
most
seriously
decompressed joints become more flexible
compared to joints with minor decompression. Since the joints are
connected in series, the stiffer ones will

force the softer to re-compressa


beneficial effect of great significance.
ACKNOWLEDGEMENTS

The authors would like to acknowledge the


financing of this research project by the Greek
Railway Organization (OSE). The help over the
years by Dr. Takashi Tazoh of Shimizu
Corporation who exposed us in the Japanese
technology was invaluable. We would also like
to thank Messieurs Hans van Italie, Hendrik
Postma, Gerard H. van Raalte, P. van der Burg,
and Royal Boskalis S.A. for kindly offering their
critical comments and suggestions on different
construction-related issues.
REFERENCES AND BIBLIOGRAPHY
Anastasopoulos I., & Gazetas G. (2007),
Foundation-Structure Systems over a Rupturing
Normal Fault : I. Observations after the Kocaeli
1999 Earthquake. Bulletin of Earthquake
Engineering, Vol. 5, No. 3.
Anastasopoulos I., & Gazetas G. (2007b),
Behaviour of StructureFoundation Systems
over a Rupturing Normal Fault : II. Analyses,
Experiments, and the Kocaeli Case Histories,
Bulletin of Earthquake Engineering, Vol. 5, No. 3.
Anastasopoulos, I., Gerolymos, N., Drosos, V.,
Kourkoulis, R, Georgarakos,P., and Gazetas, G.
(2007), Behaviour of Deep Immersed Tunnel
in
Strong
Seismic
Shaking, Journal of
Geotechnical
and
Geoenvironmental
Engineering, ASCE, Vol. 133.
Anastasopoulos I., Gazetas G., Bransby M.F.,
Davies M.C.R., and El Nahas A. (2007), Fault
Rupture Propagation through Sand : Finite
Element Analysis and Validation through
Centrifuge
Experiments,
Journal
of
Geotechnical
and
Geoenvironmental
Engineering, ASCE, Vol. 133.
Bickel, J.O., Tanner, D.N. (1982), Sunken tube
tunnels, in : Bickel, J.O., Keusel, T.R. Eds.,
Tunnel Engineering Handbook, Chapter 13, Van
Nostrand Reinhold, New York, pp. 354-394.
Bray, J.D., Seed, R.B., Cluff, L.S., and Seed, H.B.
(1994), Earthquake Fault Rupture Propagation
through
Soil,
Journal
of
Geotechnical
Engineering, ASCE, Vol. 120, No.3, pp. 543-561.
Bray, J.D., Seed, R.B., and Seed, H.B. (1994),
Analysis
of
Earthquake
Fault
Rupture
Propagation through Cohesive Soil, Journal of

25

Geotechnical Engineering, ASCE, Vol. 120, No.3,


pp. 562-580.

Poulos, H.G., & Davis, E.H. (1974), Elastic Solutions


for Soil and Rock Mechanics, J. Wiley & Sons.

Davis, R.O. & Selvadurai, A.P.S. (1996), Elasticity


and Geomechanics, Cambridge University Press.

Power, M.S., Rosidi, D., Kaneshiro, J. (1996),


Screening, evaluation, and retrofit design of
tunnels, Report, National Center for Earthquake
Engineering Research, Buffalo, New York.

Davies, M.C. & Bransby F. (2004), Centrifuge


experiments for fault rupture propagation,
University of Dundee Research Report.

Schnabel, P.B., Lysmer, J., Seed, B.H. (1972),


SHAKE : a computer program for earthquake
response analysis of horizontally layered sites,
Report no. EERC 72 /12, University of California,
Berkeley, CA, USA.

EC8 (2002) Eurocode 8 : Design of structures for


earthquake resistance. Part 2 : Bridges,
European Committee for Standardisation (CEN).
Gazetas, G. (1983), Analysis of machine foundation
vibrations: state of the art, Soil Dynamics and
Earthquake Engineering, Vol. 2, No. 1, pp. 243.

St. John, C.M. and Zahrah, T.F. (1987), Aseismic


design of underground structures, Tunnelling and
Underground Space Technology, Vol. 2, No 2,
pp. 165 -197.

Gazetas, G. (1991) Foundation Vibrations,


Foundation Engineering Handbook, 2nd edition,
ed. H.Y.Fang, Van Nostrand Reinhold.

Tselentis, A. (2004), Estimation of Seismic Hazard


and Seismic Deformation of the Rion-Antirrion
Straits, Technical Report, Earth Research Ltd,
July 2004.

Gerolymos, N., and Gazetas, G. (2005),


Constitutive Model for 1-D Cyclic Soil Behaviour
Applied to Seismic Analysis of Layered Deposits,
Soils and Foundations, Vol. 45, No. 3, June.

Veletsos, A.S., and Tang, Y. (1990), Deterministic


assessment of effects of ground motion
incoherence, Journal of Engineering Mechanics,
ASCE, Vol. 116, No. 5, pp. 1109-1124.

Ingerslev, C., and Kiyomiya, O. (1997), Earthquake


Analysis, Immersed and Floating Tunnels
Working Group Report, ITA, Vol. 2, No. 2, pp.
76-80.

Vrettos, Ch. (2005), Design issues for immersed


tunnel foundations, Proc. 1st Greece-Japan
Workshop on Seismic Design, Observation, and
Retrofit of Foundations, G. Gazetas, Y. Goto, T.
Tazoh (editors), pp. 257-266.

Ishibashi, I., and Zhang, X. (1993) Unified Dynamic


Shear Moduli and Damping Ratios of Sand and
Clay, Soils and Foundations, Vol. 33, No. 1, pp.
182-191.
Johansson, J., Konagai, K. (2004), Fault Induced
Permanent Ground Deformations Simulations
and Experimental Verification, Proc. 13th World
Conference on Earthquake Engineering.
JSCE (1975), Specifications for Earthquake
Resistant Design of Submerged Tunnels,
Japanese Society of Civil Engineers, Tokyo.
JSCE (1988), Earthquake Resistant Design for Civil
Engineering Structures in Japan, Japanese
Society of Civil Engineers, Tokyo.
Kiyomiya, O. (1995) Earthquake-resistant Design
Features of Immersed Tunnels in Japan,
Tunneling and Underground Space Technology,
Elsevier, Vol. 10, No. 4, pp. 463-475.
Kuesel, T.R. (1969). Earthquake Design Criteria for
Subways. Journal Structural Division, ASCE
ST6, pp. 1213-1231.
Nakamura, S., Yoshida, N., Iwatate, T. (1996),
Damage to Daikai Subway Station During the
1995 Hyogoken-Nambu Earthquake and its
Investigation, Japan Society of Civil Engineers,
Report on Hyogoken-Nambu Earthquake, pp.
287-295.

26

Seismic Triggering, Evolution, Deposition, and Retaining


of Rapid Landslides
Nikos Gerolymos, George Gazetas, and Ioannis Vardoulakis
National Technical University, Athens, Greece

Abstract
A mathematical model is developed for the dynamic analysis of earthquaketriggered rapid
landslide, considering two mechanically coupled substructures: (a) the accelerating
deformable body of the slide, and (b) the rapidly deforming shear band at the base of the
slide. The main body of the slide is considered as an one-phase mixture of Newtonian
incompressible fluids and Coulomb solids sliding on a plane of variable inclination. The
evolution of the landslide is modeled via an extended SavageHutter model coupled with the
MohrCoulomb sliding law for the frictional deformation of the material within the shear band.
The capability of the model is tested through analysis of the HigashiTakezawa landslide,
triggered by the 2004 Niigata-Ken Chuetsu earthquake. The mechanism of material softening
inside the shear band responsible for the 100mrunoff of the landslide, is described by a set
of equations for grain crushinginduced porewater pressures (Gerolymos and Gazetas,
2007). Then, the developed model for landslide kinematics is appropriately modified to
include the dynamic interaction between a rapidly moving landslide and a retaining piled
wall. Characteristic examples highlight the influence of both the bed inclination and the
reaction thrust exerted by the sliding mass on the wall, on the de-acceleration and finally
termination of the landslide. A limiting equilibrium approach to the problem is also presented,
resulting in design diagrams of the maximum impact velocity the wall experiences as a
function of the local landslide thickness and structural strength of the wall.

INTRODUCTION

by the inertial loading it imposes, or by causing


a loss of strength in the slope materials which
may result to catastrophic landslides, as
illustrated in Fig 1.
Analysis of rapid longrun out landslides is a
difficult challenge in hazard studies because
they endanger areas situated far from the
landslide source. Prediction of run-out distance
is a key requirement for delineation of the
hazard zone. Moreover, prediction of dynamic
parameters such as flow velocity and depth are
necessary for the design of protective
measures.
There is a variety of landslide countermeasures used to control the movement of
debris slide masses without attempting to
stabilize them (Baldwin et al., 1987; Hungr et al.
1984). Among them catch fences and trap
ditches are widely used to trap rockslide debris
while reinforced concrete retaining structures
are appropriate for debris flows and landslides.

Landslides are a serious geologic hazard


causing severe damage to structural facilities
and numerous of deaths and injuries each year.
Some landslides move slowly and induce a
limited damage, whereas others move rapidly
transporting downstream sometimes huge
volumes of sediments, covering large run-out
distances and destroying everything in their
passage. Gravity is the force that drives the
landslide movement. Factors that allow the
force of gravity to overcome the frictional
resistance of the soil and to trigger the
landslide, include: saturation by water (e.g. after
periods of heavy rainfall or rapid snow melt),
steepening of slopes by erosion or construction,
frictional softening due to earthquake induced
pore water pressures. Among the possible
causes of landsliding initiation, earthquake
shaking is of particular interest. A seismic
shaking can cause a slope to become unstable

27

Fig 1: Photos of characteristic earthquakeinduced catastrophic landslides. (left) landslide triggered by the
2001 El Salvador Mw = 7.1 earthquake. The landslide mass buried hundred of residences and accounted for
over half of the nearly 700 earthquake victims (the photograph was produced by Corbis Coorporation). (Top
right) landslide triggered by the 2004 Niigata-Ken Chuetsu MJMA = 6.8 earthquake (by Koichi Kamoshida/Getty
Images). (Bottom right) Massive landslides on the outskirts of Muzaffarabad, triggered by the 2005 Pakistan
Mw = 7.6 earthquake (by Image courtesy DLR, Kathryn Cramer, Google Earth).

a versatile and economical approach and has


been proved to be effective for the backanalysis of landslides with large run-out
distance. Nevertheless, it has the following
deficiencies: It provides limited information on
the evolution of the landslide, and it is not
capable of reproducing the mechanisms of
deposition of the slope material.
On the other hand, the deformability of the
sliding mass is fully considered by an elastoplastic solid model. A variety of constitutive
laws of varying degree of accuracy are
available to reproduce the nonlinear stress
strain behaviour of the slope material. Complex
phenomena such as material softening and loss
of strength due to porewater pressure
development can be readily modeled within the
framework of elasto-plasticity. Theory of elastoplasticity has been extensively used in
predicting the failure mechanism of a potential
landslide. Sophisticated methods based on
gradient and non-local constitutive models are

Evidently, numerical modeling of landslide


retaining wall interaction systems could serve
as a tool for evaluating landslide hazard.
The models for the analysis of earthquake
triggered landslide could be classified into three
broad categories: (a) Newmark sliding block
model and its extensions, (b) Elasto-plastic
solid models, and (c) Rheologicaldepth integrated
models. Each has advantages and limitations.

The former has been widely used in


geotechnical engineering for deformation
analysis of earthquakeinduced landslides. The
sliding block method assumes the potential
mass to be rigid, even though conditions for
actual slopes certainly vary from this
assumption. The accuracy of a sliding block
analysis depends strongly on the yield
acceleration. That is the threshold acceleration
at initiation of sliding. Obviously, the model
predicts zero permanent slope displacement if
earthquake induced accelerations never exceed
the threshold value. The sliding block model is

28

model (Laigle and Coussot 1997), and the


quadratic shear stress model (O Brien and
Julien 1985). What these models have in
common is that the total dynamic frictional
resistance is the resultant of a hysteretic
(Coulomb) and of a viscous (velocity
dependent) component.
On the other hand, in a variable density
mixture, the limitations mentioned above vanish
allowing for a wider choice of rheological
models. The coarser sediments settle even
though the remaining mixture continues to flow
downstream. In a two-phase mixture model,
termination of the landslide could be achieved
even for minimal frictional resistance and nonzero bed inclination.
A considerable amount of studies has been
dedicated the last decade to the development
of rheological depthintegrated models of
varying degrees of accuracy. Sophisticated
numerical techniques have been utilized to
solve the governing equations even for arbitrary
topographies. As to the authors knowledge
none of those studies address the interesting
issue of earthquaketriggered landslides, while
most of them suggest simple hysteretic frictional
laws (e.g. constant Coulomb friction) insufficient
to reproduce complicated material softening
behaviour which usually accompanies the
deformation of a shear band.
In this paper the governing equations of
landslide motion, originally proposed by Savage
and Hutter (1989) and extended by Iverson
(1997) and Gray (1999), are reformulated to
account for inertial force due to seismic loading.
The landslide mass is considered as an onephase mixture of incompressible fluids and
solids sliding on a plane of variable inclination.
The governing equations are coupled with a
BoucWen type constitutive model (Gerolymos
and Gazetas 2005, Gerolymos et al. 2007) for
the hysteretic stressdisplacement behaviour of
the shear band in cyclic loading, in conjunction
with the MohrCoulomb failure law for the
frictional resistance, and the quadratic Chezy
law for the viscous (turbulent) shear resistance.
The proposed model is applied to the
analysis of HigashiTakezawa landslide,
triggered by the 2004 Niigata-Ken Chuetsu
earthquake. The landslide mass covered a
distance of 100 m, filled a valley and stopped a
river flow forming a large natural reservoir. The
surprisingly large and rapid runoff of the soil
mass motivated several researchers (Kokusho

capable of simulating the formation and


development of shear zones with great
accuracy and thus improving the modeling of
failure mechanisms. However, it has not yet
been examined as to whether those models are
able to describe postfailure behaviour.
Moreover, the field equations of elasto-plastic
analysis are generally formulated by coupled
equations of soil and pore water based on the
solid mechanics within the framework of
infinitesimal strain. However, the runoff distance
of a landslide usually ranges from a few meters
to a few hundred meters. And there are more
limitations: (i) Due to strain localization
processes at the base of the potential failure
mass, the landslide behaves as a nearly rigid
body after its initiation. And (ii) Erosion and
deposition processes during the landslide can
hardly be reproduced by elasto-plastic models.
Rheological models overcome most of the
limitations mentioned above. These models
assume that the sliding mass behave as a liquid
mixture of interacting fluids and solids. Solid
and fluid constituents obey mass and
momentum balances, which are summed and
depthintegrated to yield equations that
describe shallow flow of the mixture.
Rheological depthintegrated models were
originally developed to simulate debris flow, but
they have been also successively used in
modeling
post-liquefaction
behaviour
of
granular soil. The mixture can be considered
either as an one-phase fluid of constant density,
or as a two-phase mixture of variable density
composed by granular material immersed in an
interstitial fluid.
A constant density fluid model cannot,
however, simulate the erosion/deposition
process in which the coarser sediments settle in
the upper part of the alluvial fan or near
obstacles in the river bed. Moreover, it is well
known that erosiondeposition mechanism
significantly contributes to the de-acceleration
and termination of the landslide irrespectively of
the topographic inclination and the frictional
resistance on the sliding base. Evidently, the
predictive power of one-phase fluid models is
strongly influenced by the choice of an
appropriate frictional constitutive model in which
a minimum value for the soil shear strength is
assumed. Such models are: the Bingham
model (Fraccarollo 1995, Jan 1997, Jin and
Fread 1997, Chen and Lee 2002, Hadush et al.
2000, Uzuoka et al. 1998), the HerchelBulkley

29

the wall on the sliding mass, on the deacceleration and termination of the landslide.
Finally, a limiting equilibrium approach to the
problem is also provided, resulting in diagrams
of the maximum impact velocity the wall
experiences as a function of the local landslide
thickness and structural strength of the wall,
which can be readily utilized in the design of the
retaining structure.

and Ishizawa, 2005; Tsukamoto and Ishihara,


2005, Sassa et al., 2005) to study the Higashi
Takezawa
landslide,
providing
different
interpretations of the sliding process. The
questions to be answered arose on: (a) the
exact position of the sliding surface, and (b) the
mechanism of material softening behind the
accelerating landslide movement. Laboratory
tests on soil samples taken from the site of the
slip surface indicated undrained friction angles
larger than the slip inclination (Sassa et al.,
2005). Moreover, the sliding material consisting
of silt to dense silty sand was not susceptible to
liquefaction (Kokusho and Ishizawa, 2005).
However, experimental evidences revealed the
mechanical instability of the shear band
material due to grain crushing. For this, a
constitutive model (Gerolymos and Gazetas
2007) is utilized that mathematically interprets
the concept of high porewater pressure
generation by grain crushing along the sliding
surface. The constitutive model for shear band
behaviour coupled with the depthintegrated
model for landslide kinematics, reproduces
satisfactorily the field observations.
The developed depthintegrated model is
extended and further utilized to model the
interaction of a retaining piledwall with a
rapidly deforming landslide. The piledwall is
represented as a flexible beam, while the wall
soil interface is replaced with continuously
distributed nonlinear dashpots. Computation of
the impact forces exerted by the flowing soil
mass on the wall is based on analytical
formulae proposed by Morison (1950) and
Armanini and Scotton (1993). The bending
moment of the wall section is a nonlinear
function of its curvature obeying a criterion of
the extended BoucWen type (Gerolymos and
Gazetas, 2005).
The derived system of non-linear differential
equations is solved numerically applying an
explicit central difference numerical technique
in conjunction with a shockcupturing scheme
to avoid spurious oscillations originating from
abrupt variations of the model properties. The
capability of the developed model to analyse
the evolution of a landslide on a bed with
varying inclination and its interaction with a
retaining structure constructed in front of the of
the landslide, is investigated through a
parametric study. Valuable conclusions are
drawn regarding the role of both the bed
inclination and the reaction thrust exerted by

THE MODEL : EQUATIONS AND


PARAMETERS
Problem Definition

The problem studied is that of a finite


moving soil mass assembled by a number of
columns in contact with each other (Fig 2). The
columns are free to deform but retain fixed
volumes (constant density ) of solidfluid
mixtures during their movement down a slope.
The evolution of the mixture is considered to be
onedimensional with no aggradation or
degradation processes and with uniformly
distributed (depthintegrated) velocity along
each column. At the base of the sliding mass
we assume a shear band of infinite length and
of thickness db (Fig 2), subjected to an
acceleration time history uniformly imposed
along the entire length of the sliding base. The
material inside the shear band is assumed to be
free of porewater pressures. The shear zone
deforms with a velocity (t, x) equal to that of
the sliding mass.
Applying the mass and momentum
conservation laws and using Eulerian
description of motion, a system of two partial
differential equations are obtained:

h
h

+
+h
=0
t
x
x

(1)

and


(h x )
g

+ Td Tr T f = h
+
+
x
t
x
t

(2)
h is the depth in the z direction normal to the
bed (Fig 2), is the depthaveraged velocity in
the x direction parallel to the base of the
landslide, g is the seismic acceleration

30

(x)

h(x,t)
(x,t)

bedrock

dx
A(t)hdx

x +

gcos
Tr + Tf

A(t)

x
dx
x

Td

Fig 2: 1 dimensional depthintegrated model for the analysis of earthquakeinduced landslide evolution.
Stress equilibrium is referenced to a local coordinate system that is fitted to the underlying topography.

imposed at the base of the landslide parallel to


the dip direction of the sliding surface. Note that
spatial variability of the seismic motion has
been neglected, and the same acceleration
time history is imposed instantaneously along
the entire length of the sliding surface. Td is the
gravitational driving force acting on the
landslide mass

Td = g h sin

in which m is the cyclic shear resistance


mobilized along the shear band. A detailed
description of this term will be provided below.
Tf is the turbulent resisting force at the base of
the slide, represented by the quadratic Chezy
constitutive law

T f = T 2 sgn( )

(3)

in which is the turbulent (or hydraulic) friction


coefficient relating the viscous shear stress with
landslide velocity. Exploiting experimental
results from 1g models, Ancey et al. showed
that at small flow velocities (quasi-uniform flow
regime) is a linear function of the landslide
depth according to

in which g is the gravitational acceleration and


is the local bed slope. Tr is the resisting force
due to hysteretic (Coulomb) friction at the bed
influenced by bed curvature (Gray et al. 1999,
Iverson and Delinger 2001)

2
m
Tr = 1

g
x
cos

(5)

(4)

T = h

31

1 1

40 D g

(6)

in which Dg is the mean grain diameter. For


larger velocities, however, corresponding to
Froude numbers Fr > 1 (splashing flow regime),
assumes a constant depth-independent
value. Blagovechshenskiy et al. (2002), backanalysing measured field avalanche motion and
assuming that the hydraulic friction coefficient
is depth-independent, showed that is a
decreasing function of velocity with a range of
values from 0.01 to 0.10 for velocities between
3 m/s and 35 m/s. In general varies
significantly from avalanche to avalanche
(Hutter and Greve, 1993) and its contribution to
debris flow evolution can hardly be
distinguished from that of Coulomb friction.
Zwinger (2000) studied the Madlein avalanche
of 1984 and conjectured that reasonable
agreement with the observed deposition could
only be obtained when the sliding law was
Coulomb-type at low velocities and viscoustype at large velocities. Moreover, all the
aforementioned conjectures concern debris
flows with aspect ratios between 102 to 103
rather than landslides which are usually
characterized by aspect ratios of the order of 10
and thus exhibiting a completely different
rheological behaviour.
In Eqn(2), x is the average along the
depth of the sliding mass longitudinal normal
stress due to elongation or compression of the
soil mass in the x direction. The longitudinal
normal stress is assumed to be a combination
of a lithostatic (depthdependent) term and a
strain rate dependent term

x =

1
s g h cos d h
x
2

in which s , and d are the lithostatic and


viscous coefficients, respectively. is the
aspect ratio of the sliding mass, designated as

typical thichness [H c ]
typical length tangential to the bed [Lc ]

(8)
Pudasaini and Hutter (2003) suggest that
this ratio is of the order of 10-3 to 10-1 for typical
granular avalanches, with smaller values
corresponding to debris flow and larger ones to
landslides. Eqn(7) can be rewritten in the
following form

x =

1
K dyn g h cos
2

(9)

in which Kdyn is the dynamic lateral earth


pressure coefficient determined as

K dyn = s

2 d
g cos x

(10)

Kdyn varies in the following range of values

K a K dyn K p

(11)

Ka and Kp are the active and passive lateral


earth pressure coefficients, respectively,
derived from Coulomb failure theory and
determines the magnitude of the horizontal
normal stress at yielding, depending on whether
a soil column is expanding or contracting. For a
cohessionless (granular) soil, the following
expressions for K have been proposed by
Iverson and Delinger (2001):

(7)

1 + 1 cos 2 (1 + tan 2 )
1 ,
Kp = 2
cos 2

K =

2
2
K = 2 1 1 cos (1 + tan ) 1 ,
a
cos 2

in which is the internal friction angle of the soil


mixture, and the friction angle at the base of
the sliding mass. Similar relationships for the

0
x
(12)

>0
x

lateral earth pressure coefficient, based


however on different assumptions, have been
also proposed by Pudasaini and Hutter (2003).

32

1
d
=
1
du
uy

Eqn (12) is not valid for cohesive soils.


According to Rankines theory for lateral earth
pressure, negative values of K are also possible
when cohesion is taken into account.
The first term in the righthand side of Eqn
(7) governs the rheological behaviour of the
sliding mass. A typical range of values for the
lithostatic coefficient s is between 0.01 and 1,
with larger values corresponding to deformation
of fluidlike mixture. For the special case s = 0,
the sliding mass behaves as a rigid body and
Eqn (6) vanishes to the well known Newmark
sliding block model.
The second term on the righthand side of
Eqn (7) plays the role of a damping factor for
the surface waves emanating from the frontal
edge of the landslide and propagating along its
crest. These waves could be generated due to
either (i) elongationcompression of the soil
mass in the direction of flow, or (ii) abrupt
variations of bed inclination, or (iii) impact of the
avalanche front (shock wave) on a retaining
structure. A typical range of values for the
damping coefficient d is between 1 and 10,
with smaller values representative of fluidlike
mixture (less damping), and larger values
corresponding to solidlike mixture.

[b + (1 b ) sgn( )] }

(14)

in which uy is a parameter signaling yielding in


the soil (a rigid plastic behaviour is
approximated by assuming a very small value
of uy , say less than < 10-3 m). uy is defined as
the ratio of the ultimate shear strength y to the
shear modulus G of the soil, multiplied by the
shear band thickness db. However, as the shear
band thickness is considered to be zero in our
problem, uy can be calculated alternatively from
a direct or from a ring shear test. n and b are
dimensionless quantities that control the shape
of the hysteresis loop. Eqn (14) is obviously of a
hysteretic rather than a viscous type. Hence, its
solution is not frequency dependent.
Parameter n governs the sharpness of the
transition from the linear to the nonlinear range,
during initial virgin loading. It ranges from 0 to
, with elasticperfectly-plastic behaviour
practically achieved when n takes values
greater than 10. Values of n between 0.6 and 1
have been found to better fit experimental
results (Gerolymos and Gazetas, 2005).
Parameter
b
controls
the
shape
of
unloadingreloading curve. Its range of values
is between 0 and 1. When b = 0.5 the stiffness
upon loading reversal equals the initial tangent
stiffness, and the Masing criterion for
loadingunloadingreloading arises. For more
details on Eqn (14) and calibration of the
associated parameters, the reader is referred to
the recent publication of Gerolymos and
Gazetas (2005).
The shear strength y is given by Coulombs
friction law in conjunction with the hysteretic
BoucWen model (Gerolymos and Gazetas,
2005;
Gerolymos
and
Gazetas
2007;
Gerolymos et al., 2007):

Equations for Frictional Behaviour of the Shear


Band

A versatile onedimensional macroscopic


model is utilized to describe the shear stress
strain relationship inside the shear band. The
model is capable of reproducing an almost
endless variety of stressstrain forms,
monotonic as well as cyclic. Based on the
original proposal by Bouc (1971) and Wen
(1976), the model was extended by Gerolymos
& Gazetas (2005) and applied to cyclic
response of soils. A simple version of the model
is briefly outlined here.
The mobilized shear stress inside the shear
band is expressed as:

m = y

y = ( n p )

(15)

in which p is the excess pore water pressure,


is the mobilized friction coefficient, expressed in
terms of the friction angle at the base of the
sliding soil mass

(13)

in which y is the ultimate shear strength of soil,


and is a dimensionless hysteretic parameter
controlling the nonlinear response; it is
expressed with the following differential
equation:

= tan
and

33

(16)

n = g h cos

silt encountered at the head scarp of the


Terrano landslide was well weathered and soft.

(17)

is the total stress normal to the shear band.


THE HIGASHITAKEZAWA LANDSLIDE
Background

The main body of the landslide is indicated


in the plan of Fig 3 deduced from an air borne
laser scanning survey (Sassa et al., 2005)
carried out three days after the earthquake. A
crosssection of the landslide is also depicted
in Fig 3. The gentle slope inclination before the
head scarp reveals that the landslide was a
reactivation of a previous one. The landslide
involved a soil volume of about 1 200 000 m3
(Kokusho and Ishizawa, 2005). The maximum
dimensions in plan were about 300 m width and
250 m length (Kokusho and Ishizawa, 2005),
and the maximum thickness was about 40 m
(Sassa et al., 2005). The landslide mass moved
rapidly around 100 m, and hit the opposite bank
of Imokawa river (Sassa et al., 2005). A part of
the sliding mass spread across the road and hit
a school. From the head scarp of the landslide,
consisting of a rather impermeable stiff
siltstone, the inclination angle of the sliding
surface was estimated to be approximately 20o
(Sassa et al., 2005; Kokusho and Ishizawa,
2005).
A schematic geological section of the
landslide area is shown in Fig 3. The subsoil is
essentially constituted of a Neogene formation,
consisting of sandstone (the main body of the
landslide) underlain by siltstone. The terrace
along the river and below the toe of the
landslide consists of marine sand from the
Tertiary period (Sassa et al., 2005). The
groundwater flow over the siltstone layer, lead
Sassa et al. (2005) to assume the existence of
a thin silt layer between the sandstone and the
siltstone, due to weathering of the siltstone.
Although, this silt layer was not detected at the
head scarp, the assumption of Sassa et al.
(2005) was reinforced from field investigation of
the head scarp of the Terrano landslide located
in the vicinity of HigashiTakezawa and near
the Immokawa river. The Terrano landslide
(Sassa et al., 2005), was also triggered by
NiigataKen Chuetsu earthquake, and had the
same subsoil and groundwater conditions. The

Fig. 3. The Higashi Takezawa landslide: (a) plan


view, and (b) cross section (Sassa et al., 2005)

Water seepage observed on the head scarp


of the landslide three days after the earthquake
suggests that the water table was located well
above the sliding plane. No precipitation was
observed during a period of several days
preceding the earthquake, which occurred
during the rainy season.
The grain size distribution of the sand
involved in the sliding surface of the Higashi
Takezawa landslide is illustrated in Fig 4, along
with that of the Terrano silt which is considered
to be representative of the HigashiTakezawa
one. The strength properties of the soils under
consideration were obtained from consolidated
drained and undrained high speed ring shear
tests (Sassa et al., 2005). The undrained

34

friction angle of the sand was found to be 36.9o,


while the residual friction angle of the Terrano
silt was 23.9o. However, one peculiar aspect of
the HigashiTakezawa sand is its mechanical
instability due to grain crushing. The cyclic
loading test, resulting in an apparent friction
angle of 3.3o, indicated that the Higashi
Takezawa sand is susceptible to grain
crushinginduced liquefaction.

In the limit of undrained loading conditions,


which is a reasonable assumption when the
shear band is deformed at a large velocity
(rapid landslide), and assuming that advection
dominates upon diffusion, Eqn (18) reduces to:

dB p
p
p
+ =
n 0
t
x
dt

Fraction finer by weight

Eqn (19) instead of (18) will be used in all


subsequent analyses. The parameter controls
the ultimate value of the porewater pressure.
The larger the value of , the higher the
asymptotic value of the porewater pressure.

0.8
0.6
0.4

Equations for grain crushing

0.2
0
0.001

(19)

0.01

0.1

As already discussed, the breakage


potential Bp is a measure of the evolution of the
particle size distribution curve with loading, and
hence of the amount of grain crushing. We
assume that the evolution of Bp with time is
governed by the following equation (Gerolymos
and Gazetas 2007):

10

particle diameter D (mm)

Fig. 4. Grain size distribution of the Higashi


Takezawa sand (black line) and Terrano silt (gray
line), after Sassa et al. (2005)
Equations for grain crushinginduced pore
water pressure

dB p
dt

The mechanism of porewater pressure


generation due to particle breakage is assumed
to be governed by the following equation
(Gerolymos and Gazetas, 2007):

= (B pl B p )

(20)

in which is the coefficient of grain crushing.


Increasing values of correspond to increasing
rates of porewater pressure rise. Bpl is the final
(after loading) breakage potential as computed
at the current time of loading, given by:

B p
p
p

p
+
=
n 0
cv (B p )
t
x
z
z
t
(18)

B pl =

in which Bp is the current value of the breakage


potential; cv and are the coefficients of
consolidation and porepressurebreakage,
respectively. Note that cv is a function of Bp. In
fact, cv decreases with decreasing particle size
and thus with particle crushing evolution. It is
also important to notice the advection term in
the lefthand side of Eqn (18) which is usually
ignored in a conventional soil consolidation
analysis, as the developed flow velocities are in
general extremely small. However, this is not
valid for a rapidly evolved landslide in which
transport of the excess porewater pressure
along the shear band is a significant
mechanism of the sliding process, and
therefore can not be neglected.

B p0

1 + S nb

(21)

in which Bp0 is the initial (before loading) value


of Bp, defined as (Hardin, 1985). The definition
of Bp0 is schematically illustrated in Fig 5. The
breakage number, nb, is expressed according
to:
2

nb =

hc
+ 0.3
(1 + e0 ) n s

(22)

in which hc and ns is the crushing hardness and


shape number of the particle, respectively; and
e0 is the initial void ratio of the particles mixture.
For more details on the definitions of those

35

occurs only after the yield surface has been


reached, by contrast to conventional (mass)
liquefaction in which degradation of shear
resistance initiates below the yield surface,
when the phase transformation line has been
reached.

parameters the reader is referred to the work of


Hardin (1985). In Eqn (21), S is the stress
loading factor. In an undrained cyclic simple
shear test S is given by the following equation:
3

0.3

(23)

Shear resistance (MPa)

1 + e0 n m

S = 9
800 h 2 p a n

For a given shear stress time history, Eqns


(13), (14), (18), and (20) form a system of highly
nonlinear partial differential equations with four
unknowns: the excess porewater pressure p,
the breakage potential Bp, the hysteretic
parameter , and the displacement u.

0.2

0.1

(a)
0
0

0.1

Excess pore-water pressure ratio r u

25

oa
din
g
el
Be
for

50

ter
loa
din
g

Bp0

Silt

Af

Percent finer by weight (%)

Evolution of grain size


distribution with loading

75

0
0.01

0.2

0.3

0.4

0.5

Effective normal stress (MPa)

100

0.074 0.1

10
1
Particle diameter D (mm)

100

1
0.8
0.6
0.4

(b)

0.2
0
0

10

15

20

25

30

20

25

30

t (sec)

Fig. 5. Definition of the initial breakage potential Bp0,


after Hardin (1985)
Shear displacement (m)

Calibration of model parameters

Calibration of the parameters for shear band


behaviour is achieved through numerical
simulation of undrained cyclic ring shear tests
conducted by Sassa et al. (2005). The shape
number, crushing hardness, and initial void ratio
were assumed to be ns = 25, hc = 2.4, and e0 =
0.6, respectively, while the initial breakage
potential was calculated from Fig 4 to Bp0 =
0.34.
Detailed
information
on
the
aforementioned parameters is given by Hardin
(1985).
The experimental results are reproduced in
Figs 6 in the form of time history of the
developed shear displacement, and plot of the
shear resistance versus effective normal stress.
The breakage and porepressure breakage
coefficients correspond to the analysis are =
0.05 and = 25. Note that the proposed model
is capable of reproducing the brittle behaviour
of a soil undergoing grain crushinginduced
porepressure. That is, loss of shear resistance

0.8
0.6
0.4

(c)

0.2
0
0

10

15

t (sec)

Fig. 6. (a) Computed stress path (the loading is


plotted with gray line), and time histories of (b)
excess porewater pressure ratio, and (c) shear
displacement of the undrained cyclic ring shear test
of the HigashiTakezawa sand, computed with the
proposed model. The circles correspond to the
experimental data of Sassa et al. (2005)
ANALYSIS OF THE HIGASHITAKEZAWA
LANDSLIDE

With the developed model for seismic


triggering and evolution of graincrushing
induced landslides, we analyse the case of
HigashiTakezawa. The parameters are: s = 2

36

t / m3, K = 0.5, = 1, d = 5, T = 0.01h, n = 3, b


= 0.5, uy = 10-3 m, = 0.75, ns = 25, hc = 2.4, e0
= 0.6, Bp0 = 0.34, = 25, and = 0.05. The
seepage force is ignored, since the actual level
of the water table during the earthquake is not
known. The actual seismic excitation exerted on
the landslide cannot be known in detail, as it is
influenced by many parameters such as the
geology, topography, site conditions and
distance from the fault.

and the upper part of the siltstone is


assumed to have remained intact.
(b) The shear band formed within an assumed
thin silt layer atop the siltstone, but the sand
is not susceptible to grain crushing.
(c) The shear band formed within the sand
layer, the sand is susceptible to grain
crushing, and the upper part of the siltstone
is assumed to have remained intact.
The results of the analysis for cases (a) and
(b) are shown comparatively in Fig 7 in the form
of time histories of relative shear displacement.
The maximum computed displacement at the
end of shaking for case (a) is 0.65 m, which is
by far smaller than that of 3.4 m for case (b).
These values of displacement suggest that the
existence of a thin silt layer atop the siltstone is
more crucial for triggering the landslide.
However, none of them could explain the
observed rapid and large run-out distance of
the landslide. It is therefore reasonable to
assume that grain crushinginduced pore
pressures could be a major destabilizing factor
for the landslide.
The results of the analysis for case (c) are
presented in Fig 8 in terms of snapshots of the
landslide evolution, and distributions of velocity
(Fig 9), excess porewater pressure ratio ru
(Fig 10), and breakage potential Bp (Fig 11),
along the sliding surface. The following
observations are worthy of note regarding the
response of the sliding wedge:
At the early stages of the seismic motion,
excess pore waterpressure due to particle
crushing is generated at the head of the wedge
and propagates rapidly towards its toe. In the
following few seconds the excess porewater
pressure ratio rises up very quickly reaching
values larger than 0.9 along the entire length of
the sliding surface (t = 12.5 sec blue line). At
this time, sliding originates at the head of the
soil wedge, and landsliding begins. It is very
interesting that triggering occurs almost at the
end of seismic shaking, when the motion has
essentially subsided, and not during the strong
seismic shaking as one would expect. This
implies that grain crushinginduced pore
pressure is a cumulative process and thus
depends strongly on the history of loading.
After its initiation the landslide moves rapidly
towards the riverbed, developing velocities
between 5 m/s and 16 m/s. Velocities with
smaller values concentrate on the rear of the

15

Acceleration (m / s )

10
5
0
-5

(a)

-10
-15
0

10

15

20

t (sec)

Shear displacement (m)

(b)
0
0

10

15

20

t (sec)

Fig. 7. (a) Input acceleration time history (NIG019


EW 2004 - PGA = 1.3 g) at the base of the landslide,
and (b) computed time histories of relative shear
displacements, for sliding surface: (i) within the sand
layer(no grain crushing is considered, maxu = 0.65
m) (black line), and (ii) within an assumed thin silt
layer at the top of the siltstone ( maxu = 3.4 m) (gray
line)

Therefore, we apply as excitation the EW


component of the record from the nearest (to
the landslide) observation station NIG019 at
Ojiya (PGA = 1.3 g), around 10 km west of the
HigashiTakezawa landslide and WNW 7 km
from the epicenter of the main shock (Sassa et
al. 2005). Three scenarios are studied
regarding the potential location of the sliding
surface and the susceptibility of sand to grain
crushing:
(a) The shear band formed within the sand
layer (i.e., in the main body of the landslide),
the sand is not susceptible to grain crushing,

37

120
90
60
30

t=5s

0
120 0

50

100

150

200

250

300

350

400

100

150

200

250

300

350

400

100

150

200

250

300

350

400

100

150

200

250

300

350

400

100

150

200

250

300

350

400

100

150

200

250

300

350

400

90
60

t = 12.5 s

30
0
120

50

Elevation (m)

90
60

t = 19 s

30
0
120 0

50

90
60

t = 22.5 s

30
0
120 0

50

90
60

t = 26 s

30
0
120 0

50

90
60
30

t = 30 s

0
0

50

Horizontal distance (m)


Fig. 8 Snapshot of the computed landslide evolution, for the case of HigashiTakezawa. The school is
illustrated with the gray box

38

1.2

25
20
15
10
5
0

t=5s

0.9
0.6
0.3
0

100

150

200

250

300

350

400

50
25 0
t = 12.5 s
20
15
10
5
0

100

150

200

250

300

350

400

50

100

150

200

250

300

350

400

50
25 0
t = 19 s
20
15
10
5
0

100

150

200

250

300

350

400

50

100

150

200

250

300

350

400

50
25 0 t = 22.5
s
20
15
10
5
0

100

150

200

250

300

350

400

50

100

150

200

250

300

350

400

50
25 0
t = 26 s
20
15
10
5
0

100

150

200

250

300

350

400

50

100

150

200

250

300

350

400

50
25 0
t = 30 s
20
15
10
5
0

100

150

200

250

300

350

400

50

100

150

200

250

300

350

400

150

200

250

300

350

400

50
1.2 0 t = 12.5
s
0.9
0.6
0
0.9

t = 19 s

0.6
0.3
0
1.2 0
0.9

t = 22.5 s

0.6
0.3
0
1.2 0
0.9

t = 26 s

0.6
0.3
0
1.2 0
0.9

t = 30 s

0.6
0.3
0
0

Velocity (m / s)

Excess pore water pressure ratio ru

0.3
1.2 0

t=5s

50

100

Horizontal distance (m)

Horizontal distance (m)

Fig. 910 Snapshots of the computed excess porewater pressure ratio along the sliding surface (left), and
distribution of the landslide velocity, for the case of HigashiTakezawa. The school is illustrated with the gray
box

motion towards the riverbed ( 12.5 sec < t <


22.5 sec), separation of the frontal part from the
main body of the landslide (at t = 22.5 sec), and
deposition and deceleration ( t > 22.5 sec).

landslide, while those with larger values are


mostly at the front which essentially governs the
race of the entire landslide. At t = 22.5 sec the
sliding soil mass enters the riverbed while at
this time the frontal part of the landslide
detaches from the main body, spreads across
the river, hits the opposite bank with a velocity
of 25 m / s (at t = 26.5 sec), and finally reaches
the school at t = 30 sec. Following this frenetic
motion of the detached frontal part, the main
body of the landslide accumulates inside the
riverbed forming a natural reservoir which
decelerates the trailing part of the landslide.
The reduction in velocity begins at the rear and
progressively shifts to the front.
It is seen that the calculated sliding process
extended from the ruptured scrap in the source
zone to the deposition fan on the riverbed and
near the school, is consistent with the field
observation (Sassa et al., 2005). Clearly, there
are four major stages in the run-out process,
namely, triggering (at t 12.5 sec), accelerating

0.36

Bp

0.34
0.32
0.3
0

50

100

150
200
250
Horizontal distance : m

300

350

400

Fig 11. Distributions of the breakage potential Bp


along the sliding surface, at t = 5 sec (black line), t =
12.5 sec (blue line), t = 19 sec (golden line), t = 22.5
sec (gray line), t = 26 sec (green line), and t = 30 sec
(red line)

To get an insight into the mechanics behind


this disastrous response, Fig 11 plots the
evolution of particle breakage potential Bp.

39

in which st is the dynamic coefficient that


determines the magnitude of the gravitational
lateral soil pressure imposed to the retaining
structure. Bassanou (2000) studied the problem
experimentally and measured values of st
between 1 and 3 with smaller ones
corresponding to larger aspect ratios of the
retaining structure. hcr is the pile length covered
by the sliding soil mass, given by

Notice that Bp approaches a steady state value


of 0.30 at t > 15 seconds; this is larger than the
initial value of Bpl (computed to be 0.27 in
drained loading conditions), reflecting the
influence of the developed excess porewater
pressures. The slightly increasing breakage
potential at t > 15 seconds reveals that the
grain crushing process has been practically
terminated. The effective normal stress is not
adequate for further breakage. However, the
landslide is still accelerating due to the action of
gravity.

hcr = min(L, h )

in which L is the abovebedrock length of the


retaining structure.
Implementing equations (19) and (20) into a
beam on Winkler foundation model (Fig. 12)
and assuming that the beam axis is normal to
the basal surface, dynamic equilibrium of the
retaining wall gives the following equation

LANDSLIDEPILED WALL DYNAMIC


INTERACTION

The impact force exerted by a debris flow


against a retaining structure fixed at the
bedrock depends on the flow velocity, as well
as on the density of the flowing soil mass.
According to Armanini and Scotton (1993), the
impact force per unitary area of the structure
can be expressed as the sum of a viscous
(drag) component pd and a component due to
compressionelongation of the flowing mass pg

p = pd + p g

(27)

st
2M
+ ( st + C M ) Ast
+ pd p g = 0
2
t
z
(28)
st(z,t)

(24)

CM st(z,t)

For the computation of the first component,


several approaches have been developed
based on shallow water theory. Morison et al.
(1950) proposed the following formula for nonbreaking waves

stg(h-z)

1
p d = C d ( st ) 2 sgn( st )
2

(x,t)

(25)
x

0.5Cd

in which st is the velocity of the structural


member in the direction of flow, and Cd is the
drag coefficient depending on the Reynolds Re
and Froude Fr numbers of the flow, the
roughness, shape, and orientation of the object
with respect to flow direction. Representative
values of Cd for a cylindrical object (e.g. pile)
and an infinitely long flat plate oriented normal
to flow (e.g. retaining wall), are 1.2 and 2,
respectively.
The second component of the impact force
pg is given by

p g = st g (hcr z )

Plastic hinge

Rigid
bedrock

Fig 12: Schematic illustration of dynamic non-linear


interaction between a sliding soil mass and a
retaining piled wall. The piles are partially embedded
into the rigid bedrock.

in which Ast and st are the cross-sectional area


and mass density of the beam, respectively. CM
is the inertia coefficient representing the inertia
forces via an equivalent mass distribution along
the height of the retaining wall. CM varies
parabolically with depth obtaining its maximum
value at the top of the wall and vanishing to

(26)

40

zero at the interface with the bedrock. For a


cylindrical retaining structure (e.g. pile) with a
slenderness ratio L / d > 5 (d is the pile
diameter), the average value of CM over the
walls height is in the range of 0.81, with larger
values corresponding to larger slenderness
ratios.
In Eqn (28) M is the structural bending
moment which is assumed to be a nonlinear
hysteretic function of its curvature according to
the following equation (Gerolymos and
Gazetas, 2005).

M = st EI st + (1 st ) M y st

In the above equation, y is the value of wall


curvature at the initiation of yielding, and bst, nst
are dimensionless quantities that control the
shape of the hysteretic bending moment
curvature loop. Evidently, Eqn (30) is of the
same form as Eqn (14). In the special case of
ast = 0, My is equal to the ultimate (plastic)
bending moment of the wall Mp. For details on
the
calibration
of
the
aforementioned
parameters, the reader is referred to the work of
Gerolymos and Gazetas (2005).
The first term in the right hand side of Eqn
(28) represents the reaction force of the wall
normal to flow direction. The depthintegrated
wall reaction along the portion of the wall
covered by the flowing soil mass, is calculated
as

(29)

where EIst is the initial (elastic) bending stiffness


(also called flexural rigidity), st is a parameter
controlling the post yielding bending stiffness,
My is the value of bending moment that initiates
structural yielding in the wall, and is the wall
curvature. st is the hysteretic dimensionless
parameter that controls that controls the
nonlinear structural response of the retaining
wall. The latter is governed by the following
differential equation.

d st
1
=
1 st
y
d

nst

1
q =
hcr

hcr

2M
dz
z 2

(31)

Eqn (31) is inserted into Eqn (2) to model the


dynamic interaction of the retaining wall with the
landslide:

[bst + (1 bst ) sgn(& st )] }

(30)

(h x )
+ Td Tr T f
x

q (x x
j =1

1
(C d 2 + st g hcr )
2



) = h
+

x
t

(32)

sliding surface level, and assuming the


formation of a plastic hinge at this point, gives

in which (xxj) is the Croneckers delta


function, xj is the local horizontal coordinate of
the j retaining wall, and N is the total number of
the retaining walls distributed along the sliding
surface.
In the extreme case of a rigid structure (a
structure with infinite bending rigidity) Eqn (31)
vanishes to

q =

1
1
2
3
M = C d 2 hcr + st g hcr H ( M p M )
6
4

(34)
in which H(M) is the Heaviside step function,
suggesting that the mobilized bending moment
is equal to zero when the maximum bending
moment is reached. In reality, the maximum
bending moment is retained until to a specific
value of curvature, determined by the ductility
capacity of the wall (Priestley et al, 1996),
which in turn depends on the detailing of the
wall reinforcement. To determine the maximum

(33)

which was obtained from Eqns (25) and (26) by


neglecting the inertial force and setting st = 0.
Dynamic moment equilibrium with respect to the

41

wall covered by the landslide mass, Eqn (34) is


reformulated by setting M = Mp.

impact velocity the wall can sustain as a


function of both the plastic bending moment
and the abovebedrock length of the retaining

4M p
2 st

g hcr

2
3 Cd
C d hcr
max imp =

p g = st g (hcr z ) H ( z )

Ultimate Bending Moment Mp : MNm / m

808
7

(37)

269.4
133.3
65.7

32.1

15.4

7.2

606

3.1

404

1.0

202

failure

0
0

10

1
st g hcr3
6

15

20

25

(36)

In the right hand side of Eqn (36), the term


multiplied by the Heaviside step function H(-z)
is the lateral soil reaction acting at each point of
the embedded portion of the wall. It consists of
a viscous (first sub-term) and a hysteretic
(second sub-term) component, respectively.
Notice that negative values of z correspond to
points below the sliding surface. C is the
damping coefficient accounting for radiation of
waves emanating from the wall periphery and
propagating
through
the
soil
medium
towardsinfinity. Detailed information on the
calibration of C for a pile or an arbitrarily
shaped rigid caisson is provided by Gazetas
and Dobry (1984), and Gazetas and Tassoulas
(1987). py is the ultimate lateral soil reaction,
and s is an hysteretic dimensionless quantity
that determines the nonlinear response of the
soil, and is of the same form as Eqns (14) and
(30). Definition of this parameter is given in
Gerolymos and Gazetas (2005; 2006). For an
infinitely long retaining wall (plain strain
conditions) supporting a c soil, Rankines
theory suggests the following expression of py

Contours of max : m / s

Mp <

(35)

1
C d ( st ) 2 sgn( st ) H ( z ) + ( C st + p y s ) H ( z )
2

and

1000

1
st g hcr3
6

In the case of a compliant bedrock, Eqns


(25) and (26) for the impact force are
reformulated according to

Contours of the maximum impact velocity


against a given set of model parameters, as a
function of the local landslide thickness and
bending moment capacity of the wall, is plotted
in Fig 13.

pd =

, Mp

30

Landslide cross-sectional depth h : m

Fig 13: Contours of the maximum impact velocity


exerted on a wall founded in a rigid bedrock, as a
function of the local landslide thickness and bending
moment capacity of the wall, predicted from the
developed limit equilibrium method. The parameters
used [Eqn (35)], are: = 2 ton / m3, Cd = 1, and
st =1.

p y = 2 c tan(45 +

) + g z tan 2 (45 +

) (38)

in which c is the cohesion of the supporting soil.

42

velocity of the retaining structure are not zero.


Even though, assuming that the contribution of
the inertial force to the response of the wall is
insignificant and that the structural velocity st is
very small compared to the crosssectional
velocity of the landslide, the maximum impact
velocity can be computed by applying Broms
(1964) theory for ultimate lateral capacity. Even
if developed for piles, the applicability of Broms
method also holds for infinitely long retaining
walls. According to Broms, the lateral capacity
is reached when the moment from soil reactions
balances the ultimate bending moment at the
depth where the maximum soil resistance
develops (hinge point). The fundamental
assumption underlying the method is that the
structural movements are sufficient to fully
mobilize the plastic capacities everywhere, and
that the elastic deformations are ignored.
Applying Broms theory to our problem and
considering force and moment equilibrium with
respect to the hinge point, the following
expressions for the maximum impact velocity
imp and associated depth f of the hinge point
are obtained.

st(z,t)

CM st(z,t)

stg(h-z)
z

(x,t)
x

0.5Cd

Plastic hinge

py s

Flexible
bedrock

Fig 14: Schematic illustration of dynamic non-linear


interaction between a sliding soil mass and a
retaining piled wall. The piles are partially embedded
into the flexible bedrock.

When flexibility of bedrock is considered


(Fig. 14), Eqn (35) is no longer valid, as the
wall oscillates even if rigid. Inertial force and

Cd hcr

maximp =

p y g st hcr3 + p y hcr2 + 8 M p hcr ( py + g st hcr )


3

for
M p Mcr
M p < Mcr

for

factor larger than unity that accounts for soil


compliance. In other words, Mcr increases as
the ultimate lateral soil reaction py decreases,
while the maximum sustainable impact velocity
decreases. Eqns (40) and (41) assume that the
length of wall embedment Lemb is large enough
(Lemb > f ) to accommodate the formation of a
plastic hinge.
Fig. 15 plots the maximum impact velocity
and associated depth to fixity point for a given
set of model parameters, as a function of the
local landslide thickness and wall bending
moment capacity. The compliance of the
foundation soil corresponds to that of a very stiff
clay of undrained shear strength cu = 200 kPa.
Comparison between Figs 13 and 15 shows
that the flexibility of the foundation soil alters

and

f =

hcr
2

8M p
2 g st hcr
1+
+
2

3 py
p y hcr

1 (40)

In Eqn (39) Mcr is the minimum plastic bending


moment required for stability of the wall, given
by

M cr =

1
3 hcr
g st hcr3 1 +
g st
6
4 py

(39)

(41)

which is equal to that corresponding to a rigid


bedrock, given in Eqn (35), multiplied by a

43

discontinuity
arises
from
the
sudden
appearance of the retaining wall as is indicated
by the Croneckers delta in Eqn (32). To
overcome this problem, the above numerical
technique was used in conjunction with a
Kurganov
and
Tadmor
shockcapturing
scheme (Kurganov and Tadmor, 200) which
uses MUSCL reconstruction (Van Leer, 1979),
and the superbee algorithm (Roe, 1986) as a
TVD (Total Variation Diminishing) limiter
(Harten, 1983). It is also stressed that the
second term in the righthand side of Eqn (7)
provides physical viscocity to the problem that
damps the oscillations originating from
discontinuities and / or abrupt variations of the
basal topography, and thus contributes
beneficially to the numerical stability of the
solution.

dramatically the maximum impact velocity the


wall can sustain. Even for a relatively very stiff
soil, the maximum impact velocity could be less
than half of that corresponding to the rigid
foundation soil.

Ultimate Bending Moment Mp : MNm / m

Contours of max : m / s, and fixity depth f : m

100
22

19.8

258.6

80

139

17.6
20.9

74.5

10.8
13.2

39.7

60

15.4

5.4

= 2.44 m/s
11

40

8.8
6.6

20
0

failure

4.4
f = 2.2 m

10

15

NUMERICAL EXAMPLES
20

25

30

The capability of the model is investigated


through a number of numerical examples. The
problem under consideration is illustrated in Fig
16. The landslide material lies at rest on top of
a parabolically shaped sliding surface. The
length of the projection of the sliding surface on
the horizontal plane is 120 m, with local
inclination angles of = 31.6o at X = 0 m, and
= 0.21o at X = 120 m, with respect to the global
coordination system, and an average inclination
angle of avg = 6.82o. The surface of the
landslide material is a parabola (in section), has
a length of 30 m and a maximum thickness of 5
m at its centre (at X = 15 m). Initiation of the
landslide occurs when the basal friction
coefficient drops deliberately to = 0.05 (e.g.
due to material softening, corresponding to a
friction angle of 2.86o, that is much smaller than
the average inclination angle avg of the basal
surface), and the released soil material begins
to move down the inclined plane. Two cases
are analysed:
(a) the sliding process is free from the presence
of any retaining structure, and deposition
termination of the landslide occurs naturally.
(b) A retaining wall consisting of a row of piles
of d = 1 m in diameter and a centerto
center spacing of s = 1.5 m, with an above
bedrock height of L = 7 m and bending
moment capacity of Mp = 1 MNm, is
constructed at X = 60 m to decelerate the
sliding process. Flow of the soil mass
between the piles is allowed.

Landslide cross-sectional depth h : m

Fig 15: Contours of the maximum impact velocity


exerted on a wall founded in a compliant bedrock
and associated depth to fixity, as a function of the
local landslide thickness and bending moment
capacity of the wall, predicted from the developed
limit equilibrium method. The parameters used [Eqns
(39), (40)], are: = 2 ton / m3, Cd = 1 , st =1, and py
= 400 kPa.

The impact velocity given by Eqns (35) and


(39) should not be confused with that resulting
from collision of an individual boulder. The
collisional force of a single boulder can cause
severe local damage to the retaining structure,
and a cushion layer consisting of gravels or old
tires should be placed at its back for protection.
NUMERICAL FORMULATION

An explicit central finite difference technique


is used for the solution of Eqns (1) and (32) for
landslide motion, which are coupled with the
constitutive Eqns (14) for shear band
behaviour, and (28), (30) for non-linear bending
behaviour of the wall. The aforementioned
numerical scheme while providing accuracy of
secondorder, introduces spurious oscillations
into the solution where discontinuities or shocks
are present leading to ficticiously large
gradients of both the landslide thickness h and
the velocity . In our problem, such a

44

15

15

12

t=0s

12

15 0

20

40

60

80

100

120

15 0

t = 1.88 s

12

40

60

80

100

120

100

120

100

120

t = 1.88 s

15 0

20

40

60

80

100

120

15 0

t = 3.75 s

12

20

40

60

80

t = 3.75 s

12

15 0

20

40

60

12

80

100

120

Elevation : m

Elevation : m

20

12

t = 11.25 s

9
6
3
0

15

t=0s

15 0

20

40

60

12

80

t = 11.25 s

9
6
3
0

20

40

60

12

80

100

120

20

40

60

12

t = 16.88 s

15 0

100

120

t = 16.88 s

80

0
15 0

20

40

60

80

100

120

15 0

20

40

60

80

100

120

12

12
6

15 0

20

40

60

80

100

t = 22.50 s

t = 22.50 s

120

15

20

40

60

80

100

120

100

120

12

12

t = 30 s

t = 30 s

9
6

3
0

0
0

20

40

60

80

100

120

20

40

60

80

Horizontal distance : m

Horizontal distance : m

Fig 16: Snapshots of the landslide evolution for subcase a1 (cohesive soil: aspect ratio = 0.2).

Fig 17: Snapshots of the landslide evolution for subcase a2 (granular soil: aspect ratio = 0.8).

45

15

15
12

12

t=0s

9
6

wall

15 0

20

40

60

80

100

15 0

120

t = 1.88 s

12

20

40

60

80

100

120

15

t = 3.75 s

12

60

80

100

120

100

120

100

120

t = 1.88 s

20

40

60

80

t = 3.75 s

12

0
20

40

60

12

80

100

120

Elevation : m

15 0

Elevation : m

40

t = 11.25 s

9
6
3
0

15 0

20

40

60

12

80

t = 11.25 s

9
6
3
0

20

40

60

12

80

100

120

15 0

20

40

60

12

t = 16.88 s

80

100

120

t = 16.88 s

0
15 0

20

40

60

80

100

120

15 0

20

40

60

80

100

120

12

12

t = 22.50 s

t = 22.50 s

0
15

20

12

15

wall

15

t=0s

20

40

60

80

100

120

15 0

20

40

60

80

100

120

100

120

12

12

t = 30 s

t = 30 s

0
0

20

40

60

80

100

120

20

40

60

80

Horizontal distance : m

Horizontal distance : m

Fig 19: Snapshots of the landslide evolution for subcase b2 (granular soil: aspect ratio = 0.8). A row of
piles are located at X = 60 m.

Fig 18: Snapshots of the landslide evolution for subcase b1 (cohesive soil: aspect ratio = 0.2). A row of
piles are located at X = 60 m.

46

case a1, the flow remains expanding from its


initiation to its termination. The length of the
deposition zone amounts to 68 m, that is 1.5
times larger than in case a1 (cohesive soil),
covering the region between X = 38 m and X =
106 m. The resulting runout distance is now
76 m, that is 15 m larger than that of case a1.
The role of the retaining wall on the breaking
of the landslide movement (Figs 18 and 19) is
double. The existence of a piledwall at X = 60
m reduces both the runout distance of the
landslide and the extent of its deposition zone.
However, the contribution of the wall to the
termination of the landslide is significantly more
profound in case a2 (cohesive soil) than in case
b2 (granular soil). The runout distance and
associated length of the deposition zone
amount to 49 m and 36 m for case a2 and to 69
m and 64 m for case b2, respectively.

in both cases the bedrock is assumed to be


rigid, while in the second case the retaining wall
is assumed to behave as a rigidperfectly
plastic material. This means that q is
calculated directly from Eqn (33), and thus
differential Eqns (28) and (30) are simply
eliminated from the model. As the actual
problem is 2D in nature while the proposed
model is 1D, averaging the effect in 2D over the
distance between the piles is necessary. To this
end, Eqn (33) for the depthintegrated wall
reaction is modified as

q eq =

d
q
s

(42)

Eqn (42) suggests that the row of piles is


represented by an equivalent massive wall, the
width of which is equal to the pile diameter
multiplied by the ratio d / s. The actual force
and moments on the piles are then calculated
by dividing the computed ones by the
aforementioned ratio.
The values of the parameters used in the
analyses are: = 2 ton / m3, s = 0.5, = 0.2, d
= 5, = 0.08, st = 1.5, Cd = 1.5, b = 0.5, n = 3,
and uy = 10-3 m. Four sub-cases are
investigated regarding the effect of the aspect
ratio on the landslide evolution. Sub-cases a1
and b1 for = 0.2, representative for cohesive
soils, and subcases a2 and b2 for = 0.8,
representative for granular soils. The entire
computational area is divided into 120 cells,
resulting in a cell length of 1 m, while the time
step is taken equal to a very small value, t =
3.75 10-4 s, to ensure numerical stability.
The results of the analyses are plotted in
Figs 16 to 20. Fig 16 illustrates the simulated
sliding process for case a1 (cohesive soil: =
0.2). After its termination at t = 30 sec, the
landslide has a length of 45 m, that is 1.5 times
larger than its initial length of the sliding mass,
covering the region between X = 46 m and X =
91 m. Observe that at t < 17 sec the flow is
contracting, as the avalanche body has not yet
reached the gently sloping part of the sliding
surface. After a runout distance of 61 m, the
basal friction is large enough to bring the
avalanche to rest.
The corresponding results for case b1
(granular soil: = 0.8) are presented in Fig 17.
As the rheological component of the sliding soil
mass is now more intense, compared to that of

Impact moment at the pile base : kNmo

500
400
300
200
100
0
0

10

15

20

25

30

time : sec

Fig 20: Time histories of the impact moment


developed at the pile base for cases b1 (cohesive
soil: aspect ratio = 0.8) with black line, and b2
(granular soil: aspect ratio = 0.8) with grey line.

Fig 20 plots the time histories of the impact


moment developed at the pile base during the
sliding process, for cases a2 and b2. The figure
reveals that the frontal edge of the landslide
reaches the wall at t = 8.8 sec in case a2, later
than in case b2 where the wall is hit at t = 7.5
sec. The maximum impact moment is computed
to 431 kNm for case a2, and to 132 kNm for
case b2. Both values are significantly lower
than the capacity of the wall. The considerably
smaller bending moment in case b2 compared
to that in case a2, is attributed to the expanding
nature of the flowing process, which results in

47

significantly compared to that for a rigid


bedrock.
The capability of the model is demonstrated
through a set of numerical examples,
highlighting the role of a retaining wall on the
landslide evolution. It is shown that the
presence of the wall results in the reduction of
both the runout distance and the deposition
length of the landslide. The reduction becomes
more profound with increasing cohesiveness of
the landslide material. As the actual problem
studied in this paper is 2D in nature, validation
of the proposed 1D model through experiments
is strongly recommended.

low landslide thicknesses and thus in a small


impact force on the wall.
CONCLUSION

In this paper we developed a model for


seismic triggering, evolution and stopping of a
landslide with a retaining structure. The
evolution of the landslide is modeled via an
extended SavageHutter model coupled with
the MohrCoulomb sliding law for the frictional
deformation of the material within the shear
band.
The capability of the model is investigated
through prediction of the HigashiTakezawa
landslide. Three scenarios are analysed
regarding the location of the sliding surface and
the susceptibility of sand to grain crushing: (a)
shear band within the sand layer, but sand not
susceptible to grain crushing, (b) shear band
within an assumed thin silt layer at the top of
the siltstone, but sand not susceptible to grain
crushing, and (c) shear band within the sand
layer, but sand susceptible to grain crushing.
The residual displacement is calculated to be
0.65 m and 3.4 m for the first and second
scenario, respectively. The observed 100 m
displacement of the landslide, associated to a
shear velocity of 15 m / sec (20 sec after the
triggering), is only reproduced with the third
scenario, despite the residual friction angle of
the silt [scenario (b)] being 13 degrees smaller
than that of sand.
The developed depthintegrated model for
landslide kinematics is then appropriately
modified to implement the dynamic interaction
between a rapidly deformed landslide and a
retaining piledwall. The derived system of
differential equations is solved numerically
utilizing an explicit central finite difference
technique in conjunction with a shockcapturing
scheme to avoid spurious oscillations
originating from abrupt variations of the model
properties.
A limiting equilibrium approach to the
problem is also presented, resulting in design
diagrams of the maximum impact velocity the
wall experiences as a function of the local
landslide thickness and structural strength of
the wall. It is shown that bedrock compliance
influences greatly the maximum impact velocity.
Even for a relatively stiff bedrock, the maximum
impact velocity the wall can undertake reduces

ACKNOWLEDGMENTS

This paper is a partial result of the Project


PYTHAGORAS I / EPEAEK II (Operational
Programme for Educational and Vocational
Training II) [Title of the individual program:
Mathematical and experimental modeling of the
generation,
evolution
and
termination
mechanisms of catastrophic landslides].This
Project is co-funded by the European Social
Fund (75%) of the European Union and by
National Resources (25%) of the Greek Ministry
of Education.
REFERENCES
Armanini A, Scotton P (1993) On the dynamic
impact of a debris flow on structures. Proc. Of
the XXV IAHP Congress, Tokyo, 203210.
Baldwin J E, Donley H F, Howard T R (1987) On
debris flow/avalanche mitigation and control,
San Francisco Bay area, California. Reviews in
Engineering Geology. Geol. Soc. Of America,
VII: 223236.
Bassanou M (200) Landslides dynamics
Mathematical and experimental simulation of
debris flow. Ph. D. Thesis, National Technical
University, Athens, Greece (In Greeks).
Broms B B (1964) Lateral resistance of piles in
cohesive soils. J. soil mech. Fdns. Div., ASCE,
90: 2763.
Chen H, Lee C F (2002) Runout Analysis of Slurry
Flows with Bingham Model, J. of Geotechn. And
Geoenv. Eng., ASCE, Vol. 128(12): 10321042.
Fraccarollo L (1995) Un modello matematico
bidimensionale per la mappatura del riscio da
colata di detriti. GNDCI, Linea 1: 97105 (in
Italian).

48

Jan C D (1997) A study on the numerical modeling


of debris flow. Debris Flow Hazards Mitigation:
Mech., Pred. and Assessment, ASCE.

Gazetas G. and Dobry R (1984) Simple radiation


damping model for piles and footings. Journal of
Engineering Mechanics, ASCE, Vol. 110:
937956.

Jin M, Fread D L (1997) 1D routing of mud/debris


flow using NWS FLDWAV model. Debris Flow
Hazard Mitigation: Mech.,Pred. and Assessment,
ASCE.

Gazetas G, Tassoulas J L (1987) Horizontal


damping of arbitrarily shaped embedded
foundations.
Journal
of
Geotechnical
Engineering, ASCE, Vol. 113(5): 458475.

Kokusho T, Ishizawa T (2005) Energy approach to


slope failures and a case study during 2004
Niigataken Chuetsu earthquake. Proceedings
of
Geotechnical
Earthquake
Engineering
Satellite Conference Osaka, Japan, 10
September 2005, Performance Based Design in
Earthquake Geotechnical Engineering: Concepts
and Research, ISSMGE: 255-262.

Gerolymos N, Gazetas G (2005) Constitutive Model


for 1-D Cyclic Soil Behavior Applied to Seismic
Analysis of Layered Deposits. Soils and
Foundations 45(3): 147159.
Gerolymos N and Gazetas G (2007) A Model for
Grain Crushing Induced Landslides Application
to Nikawa, Kobe 1995. Soil Dynamics and
Earthquake
Engineering
(accepted
for
publication).

Kurganov A, Tadmor E (200) New highresolution


central schemes for nonlinear conservation laws
and convectiondiffusion equations. J. Comp.
Phys. 160: 214282.

Gerolymos N, Vardoulakis I and Gazetas G. (2007)


A thermo-poro-viscoplastic shear band model
for triggering and evolution of catastrophic
landslides. Soils and Foundations 47(1).

Laigle D, Coussot P (1997) Numerical modelling of


mudflows. J. of Hyd. Eng.

Gray J.M.N.T., Wieland M., Hutter K. (1999) Gravity


driven free surface flow of granular avalanches
over complex topography, Proc. R. Soc. London
A 455, 1841-1874.

Morison J R, OBrien M P, Johnson J W, Schaaf S A


(1950) The force exerted by surface waves on
piles. Petroleum Transaction, American Institute
of Mining Engineers 189: 149154.

Hadush S, Yashima A, Uzuoka R (2000)


Importance of viscous fluid characteristics in
liquefaction induced lateral spreading analysis.
Computers and Geotechnics 27: 199224.

OBrien J S, Julien P J, Fullerton W T (1993) Two


dimensional
water
flood
and
mudflow
simulation. J. of Hyd. Eng., Vol. 119(2).
Priestley M J N, Seible F, Calvi G M (1996) Seismic
design and retrofit of bridges. John Wiley and
Sons, Inc., New York

Harten H (1983) High resolution schemes for


hyperbolic conservative laws. J. Comp. Phys.
49: 357385.

Sassa K, Fukuoka H, Wang F, Wang G (2005)


Dynamic properties of earthquake-induced
large-scale rapid landslides within past landslide
masses. Landslides 2: 125-134.

Hungr O., Morgan G C, Van Dine D F and Lister D R


(1987) Debris flow defences in British
Columbia. Reviews in Engineering Geology
Geol. Soc. Of America, VII: 201222.

Pudasaini S. P., Hutter K. (2003) Rapid shear flows


of dry granular masses down curved and twisted
channels, J. Fluid Mech., 495, 193-208.

Hutter K, Greve R (1993) Twodimensional


similarity solutions for finitemass granular
avalanches with Coulomb and viscoustype
frictional resistance. J. Glac. 39: 357372.

Roe P L (1986) Characteristic based schemes for


theEuler equations. Ann. Rev. Fluid Mech. 18:
p337.

Hutter K, Koch T (1991) Motion of granular


avalanche in an exponentially curved chute:
experiments and theoretical predictions. Phil.
Trans. R. Soc. London, A 334: 93138.

Savage S B, Hutter K (1989) The motion of a finite


mass of granular material down a rough incline.
J. Fluid Mech. 199: 177215.

Iverson R M (1997) The physics of debris flows.


Rev. Geophys. 35: 245296.

Savage S B, Hutter K (1991) The dynamics of


avalanches of granular materials from initiation
to runout. Part I: Analysis, Acta Mechanica 86:
201223.

Iverson R. M., Delinger R. P. (2001) Flow of


variably fluidized granular masses across threedimensional terrain. 1. Coulomb mixture theory,
J. Geophys. Res. B, 106, 537-552.

Toro E F (1999) Riemann solvers and numerical


methods for fluid dynamics. SpringerVerlag,
Berlin Heidelberg.

49

Van Leer B (1979) Towards the ultimate


conservative difference scheme. A second order
sequel to Godunovs method.J. Comp. Phys.
32: 101136.

Tsukamoto Y, Ishihara K (2005) Residual strength


of soils involved in earthquakeinduced
landslides. Proceedings of Geotechnical
Earthquake Engineering Satellite Conference
Osaka,
Japan,
10
September
2005,
Performance Based Design in Earthquake
Geotechnical Engineering: Concepts and
Research, ISSMGE: 117-123.

Zwinger
T
(2000)
Dynamik
einer
Trockenschneelawine auf beliebig geformten
Berghngen. Ph. D. Thesis, Vienna University
of Technology (in German).

Uzuoka R, Yashima A, Kawakami T, Konrad J M


(1998) Fluid dynamics based prediction of
liquefaction
induced
lateral
spreading.
Computers and Geotechnics 22: 243282.

50

Interaction of Earthquake-Triggered Landslide with Foundation-Structure


Systems
R. Kourkoulis, F. Gelagoti, I. Anastassopoulos, G. Gazetas
National Technical University of Athens, Greece

Abstract
The paper studies the effects of earthquake induced landslides on structures founded on
the vicinity of slope crests. Plane-strain dynamic analyses are performed utilizing fully non
linear finite elements. The model is calibrated against published data to simulate the
observed strain-softening behavior of soil during a seismic event and under the action of
gravitational forces. The foundation is modeled as a flexural beam and the possibility of
sliding between the foundation and the underlying soil is considered through the use of
special gap elements. The influence of foundation type (shallow or piled), on the position of
the failure surface and the produced soil-displacements is explored parametrically. The
analysis shows that the use of mat foundation compared to that of isolated footings is
generally more suitable. Properly designed piled foundations can also significantly enhance
the seismic response of the structure.

INTRODUCTION

known that the resistance of soils whose


strength is characterized by different peak and
residual values, is progressively reducing with
increasing strain (Terzaghi and Peck, (1948),
Skempton (1964), Bjerrum (1967). The complex
mechanism of progressive failure of slopes
apparently cannot be modelled with simplified
limit equilibrium techniques and rather
necessitates advanced numerical methods
capable
to
describe
strain-localization
phenomena.
To simulate the progressive soil-failure and
the shear-zone development a popular
methodology is the use of Finite Elements
(Hoeg 1972, Lo and Lee 1973, Dounias et al
1988, Chen et al 1992, Modaressi et al 1995
,Potts et al 1997, Loukidis et al 2003,
Troncone, 2005 Pradel et al 2005).All the above
studies were performed under static conditions
and only the elements in the vicinity of the
expected failure surface were modelled so as to
obey a certain strain-softening law while the
rest of the soil model was assumed to behave
elastically.
The present study, utilizes a fully non-linear
finite element model with a strain-softening

The seismic bearing capacity of shallow


footings on the vicinity of sloping ground has
been thoroughly examined. Sarma and Chen,
(1995), Sarma (1999), Askari and Farzaneh,
(2003), Kumar and Rao (2003)), by means of a
pseudostatic limit equilibrium approach that
takes into account the inertia of soil mass,
concluded that the bearing capacity is minimal
when the footing is located at the edge of the
slope, and increases as the footing is carried
away from it until it ultimately converges to its
level-ground value.
Despite its broad acceptance, limit
equilibrium technique is inappropriate to
effectively simulate the soil-structure-interaction
effects that take place during earthquakes. First
of all it is inherently incapable to capture the
actual strain-development phenomena. The
method usually requires a pre-assumed sliding
surface that is not the natural product of
strength reduction due to strain accumulation.
Moreover the pseudostatic approach is not
applicable to soils that soften with increasing
number of cycles (Loukidis et. al (2003)), such
as those examined in the present study. It is

51

surface, while at the same time the


computational
time
was
exponentially
increased.

material law. The potential separation of the


foundation from the soil, due either to uplift or to
the downward slope movement during
landslide, is also taken into account. Since this
model can exhibit non-uniform straining, it
realistically captures the mechanism of
progressive slope failure and the effects of the
foundation on the position of the generated
failure surface. Because the sliding surface is
not pre-defined but is product of strain
softening, the model encapsulates both the
effect of the structure on the produced
displacement-field and on the position of the
failure surface itself. The soil is at this stage
assumed to be dry, neglecting the effect of pore
pressure build up due to cyclic loading.

SOIL MODELING AND CALIBRATION

A non-associated flow rule is adopted for the


soil. The pre-failure behavior and the loadingunloading of the soil obey the theory of
elasticity, while strain softening and the postfailure behavior are modeled with s MohrCoulmb failure criterion. The model parameters
are the cohesion, c, the friction angle and the
dilatancy angle . Strain softening is
incorporated into the finite element code
through a user defined subroutine that reduces
the strength parameters c and with increasing
plastic strain (Figure 1a). The material behavior
is calibrated against a viscous-plastic model for
the calculation of the strain and strength of
strain-softening cohesive soils developed by
Gerolymos et al., (2007), Based on several
experimental results (e.g. Lupini et al.,1981,
Bishop et al., 1971, Bromhead and Curtis,
1983, Skempton, 1985, Tika and Hutchinson,
1999) and utilizing an artificial neural network,
they proposed analytical relationships for the
calculation of the strength parameters of
cohesive
soils
depending
on
their
characteristics. The residual friction angle (Fig.
1b) has been found to be a function of the clay
content (CF ) and the clay activity (A):

FINITE ELEMENT MODEL

The analyses are performed in the finite


element code ABAQUS (2001) Plane strain
dynamic analyses are executed, utilizing a
fully non-linear elasto-plastic model. A 30 m
high, 23o steep slope is analysed. Two
different soil strata are considered. The top 40
m layer exhibits strain softening behaviour
according to the material law described below.
The bottom stratum is considered to be
elastic, while its thickness till the bedrock is 20
m
Quadrilateral 4-noded plane strain elements
are used for the representation of the soil while
the foundation slab is modeled with beam
elements. The interface between the foundation
slab and the underlying soil is modeled with
frictional elements which allow both slippage
and separation between the soil and the
foundation nodes. Both shallow and piled
foundations are examined. The piles are
modeled as beam elements; their diameter is
0.60 m, while their length extends to 30 meters.
This way the piles are founded within the stiff
soil layer below the level of the observed slope
failure surface. Free field conditions are applied
at the model boundaries. The optimum element
dimension that would ensure computational
efficiency without endangering the accuracy of
the simulation was parametrically investigated.
A size of L=0.5 m has been finally selected. It
was shown that the use of smaller elements
had no effect on the position of the failure

= P / C

(1)

where Ip the plasticity index.


Skempton (1985) suggests typical values of
displacements during various stages of the
ring shear test. The critical state friction angle
is calculated according to Mitchell (1976) as:
cv arcsin [ 0.6 0.14log(IP 5)]

(2)

Finally, the peak value of the friction angle


was found to be best calculated as follows:
tan P

tan cv

52

= 1 0.85 ln

2 ,
OCR

OCR > 2

(3)

PARAMETRIC ANALYSES

Infinite Strip

Scope of this study is a parametric


investigation of the effect of foundation type on
the failure surface and the generated soil
displacements for different levels of ground
shaking. Three types of structures are
investigated:
1) infinite strip of width 20 m and load q = 20,
40 and 80 kPa per unit area. The strip is
located at distances s=5 m, 8 m and 11 m from
the crest. The purpose of this study is to
examine whether and to what extend the load
and the location of the foundation affect the
position of the generated shear zone and the
resulting soil-displacement pattern.
2) a single storey structure of width 20 m and
load q=20 kPa per unit area founded on (a)
isolated footings and (b) rigid foundation slab.
Here the effect of foundation type is
investigated both in terms of stresses and
displacements on the structure.
3) A slab of width 20 m and load q=20 kPa per
unit area supported through piles. This set of
analyses aims to assess whether piled
foundation is a method of enhancing the
stability of structures in the vicinity of sliding
slopes.
The JMA accelerogram recorded during the
devastating Kobe 1995 earthquake (Fig 2)
scaled at PGAs of 0.5 g and 0.8 g is used as
the excitation motion of the model.

Position of the failure surface :


When a 20 m width rigid footing of q = 40 kPa
lies at 5 m distance from the slope crest the
picture of the generated plastic deformations is
notably changed. Fig 5a illustrates the new
failure surface when the slope is excited by
JMA accelerogram (PGA 0.8g). It is clear that
the footing causes the failure surface to deviate
from its free field position, as it imposes an
extra load on the slope crest which in turn
provokes failure of its underlying soil. As the
footing is taken further away from the crest at a
distance of 8 m (Fig. 5b), the failure surface
seems to be less affected until it finally returns
to its initial free field position once the footing
has been moved far away ( Fig. 5c). At that
distance, the influence of the footing on the
slope failure is minimal.
When the footing load is reduced to the half
(q = 20 kPa) the failure surface (Figure 6a) is
practically unaffected by the footings presence.
Conversely, in case the foundation load is large
(80 kPa), the shear zone is indeed influenced
by the footing: now the shape of the generated
failure surface is reminiscent of a bearing
capacitytype failure (Fig. 6c). As can be
deduced by the intensity of plastic deformations
generated underneath the footing (Figs 5 and
6), the bearing capacity of the ground increases
as the footing is moved away from the crest.

RESULTS

Footing Response :
Fig 7 plots the horizontal sliding of the soil
nodes underneath the two footing edges (nodes
1 and 2) along with the sliding of the footing
itself (Node 3). As expected, soil node 1 which
lies on the side of the sloping ground
systematically displaces the most, while soil
node 2 displaces less as it is less affected by
the slope movement. In case of the immense
JMA input motion scaled at 0.8 g., the
displacement of node 1 keeps increasing with
time an evidence of landslide. On the other
hand, as the footing detaches from the ground
once the friction force is exceeded, the footing
node 3 displaces less than the underlying soil
node 1. Note also that when the excitation
motion is the JMA record scaled at 0.5g the
landslide is not triggered (after the end of the
earthquake t > 27 sec the soil-displacements

We begin with the case when no structure is


founded on the top of the slope crest (FreeField case). Fig 3a plots the generated field
displacements when the input motion is the
JMA with PGA 0.5g, while Fig 3b shows the
same distribution when the peak ground
acceleration is 0.8g. The plastic strain
distribution is depicted in Fig 4. It is evident that
when the model is subjected to the JMA0.5g no
landslide is triggered (the maximum plastic
strain doesnt exceed 4 %), while the maximum
soil-displacement is 53 cm. On the contrary
when the excitation is the JMA0.8g the
formation of the shear zone is clear (40 %
plastic strain) and the maximum field
displacement after the application of the gravity
load is 1.8 m!

53

remain constant under the action of the gravity


load)

foundation slab decreases significantly the


developed moments compared to the use of
isolated footings. Greater are the differences at
the peripheral columns where the moments at
the top with the use of isolated footings are
double those developed with the mat
foundation.

One-Storey Building
The frame is considered to be elastic. Emphasis
is given on how the different types of foundation
affect the developed stresses on the structure
and the generated soil-displacements. It is
mentioned that in some cases the developed
stresses are greater than the actual resistance
of the cross-section. However, for comparison
reasons we keep the geometric characteristics
of the cross-section constant.

Displacements on the structure:


Fig 12 and 13 present the displacement time
history at particular points on the structure. Only
results for the JMA 0.8g excitation are depicted.
It is clear that the type of foundation controls
the magnitude of the differential settlement.
Note that when the foundation is a rigid slab the
differential settlement is overall smaller than the
one observed when isolated footings are used.
Moreover with the use of mat-foundation the
structure displaces as a rigid body which
results in general smaller deformations on the
frame.

Position of the failure surface :


Fig 8 displays the field plastic-strain distribution
when the excitation motion is the JMA
accelerogram scaled at 0.8g. In the upper
Figure the structure is supported on isolated
footings while in the bottom Figure the same
structure is founded on a rigid slab. It is clear
that the position and the shape of the shear
zone doesnt change with the foundation-type.

Piled Foundation
The piled foundations are only limited
examined as an alternative foundation method.
Our scope is to detect its potential benefits
compared to the shallow foundations. The piles
are considered to behave elastic. They go up to
depth 30 m where they reach the rock stratum
and their diameter is 60 cm.
The connection between the pile-node and
the soil-node is accomplished through properly
calibrated non-linear springs that follow
published p-y curves (Matlock, 1970). When
the spring deformation exceeds a certain value
(given by the empirical p-y curve) the spring
stiffness drops to zero. With this approach the
separation between pile and soil is achieved at
high values of soil displacement. It is believed
that with this numerical trick the actual 3-D
behavior of the pile-soil system is more
realistically described. Fig 14 depicts the typical
spring behavior during a seismic event that
precedes a landslide for two characteristic
springs (one at the top and one at the bottom of
the pile).
Figure 15 compares the final field
displacements when a shallow and a piled
foundation are used. Figure 16 presents the
horizontal displacement time histories of a
structure with (a) a shallow and (b) a piled
foundation respectively. It is noted that the
presence of piles improves the soil-

Field Displacements:
Fig 9 portrays the final soil-displacements when
the model is subjected to the JMA time-history
scaled at 0.8g. It is clear that the different
foundation type doesnt affect the resulting
displacement profile. The trends are reserved
when the input motion is the JMA0.5g.
Stresses on the structure:
Figures 10 and 11 depict the envelope of
maximum-bending-moments on the columns of
the frame when the input motion is the JMA
record scaled at 0.5g and 0.8g respectively.
The following are observed:
When the input is the JMA at 0.5g (an
excitation that doesnt trigger landslide), the
differences in the developed moments for the
two foundation-types are not significant. It is
believed that since this seismic time-history
doesnt evoke significant soil displacements,
the developed stresses are essentially the
result of the inertial response of the frame.
On the other hand, when the excitation is the
JMA 0.8g the developed soil displacements are
very high. This results further kinematic stress
on the frame due to the imposed displacements
of the landsliding soil stratum. In this case the
type of foundation is crucial. The use of rigid

54

Lo, K. Y. And Lee, C.E. (1973), Stress analysis and


slope stability in strain-softening materials.
Geotechnique 23, No1, 1 11

displacement pattern and decreases the peak


values. Most importantly the pile foundation
seems to relieve the supported structure which
deforms less.

Loukidis D., Bandini P. and Salgado R. (2003),


Stability of seismically loaded slopes using limit
analysis Geotechnique 53, No. 5, 463479

REFERENCES

Lupini J.F., Skinner A.E. and Vaughan P.R.


(1981),The drained residual strength of cohesive
soils. Geotechnique, 31(2), 181213

Anastasopoulos I. (2005), Behaviour of Foundations


over Surface Fault Rupture: Analysis of Case
Histories from the Izmit (1999) Earthquake,
Proceedings of the 16th International Conference
on Soil Mechanics & Earthquake Engineering,
Osaka, Japan, September 1216, 2005, pp.
26232626.

Mitchell J. K. (1976): Fundamentals of soil


behaviour, John Wiley and Sons, New York, 422.
Modaressi H., Faccioli E., Aubry D., Noret C. (1995),
Numerical modelling approaches for the analysis
of earthquake triggered landslides. Proceedings
of the Third International Conference on Recent
Advances
in
Geotechnical
Earthquake
Engineering and Soil Dynamics, St. Louis,
Missouri, II(INVLE.03), 833843

Anastasopoulos, . (2005), Fault RuptureSoil


FoundationStructure
Interaction,
Ph.D.
Dissertation, School of Civil Engineering,
National Technical University, Athens.
Askari, F. & Farzaneh, O. (2003). Upper-bound
solution for seismic bearing capacity of shallow
foundations near slopes Geotechnique 53, No. 8,
697702

Fardis N., Georgarakos P., Gazetas G.,


Anastasopoulos I.(2003) Sliding Isolation of
Structures: Effect of horizontal and vertical
acceleration. Proceedings of the fib SymposiumMay 6-8 Athens, Greece (in cd rom)

Bjerrum L. (1967) Progressive failure in slopes of


over-consolidated plastic clays and clay shales,
J. Soil Mech. Fdn Div. ASCE 93, 349

Potts D.M., Dounias G.T., and Vaughan P.R. (1990),


Finite element analysis of progressive failure of
Carsington Dam embankment. Geotechnique 40,
No 1, 79 101

Bromhead E.N. and Curtis R.D (1983): A


comparison of alternative methods of measuring
the residual strength of London Clay. Ground
Engineering, 16.

Potts D.M., Kovacevic N. And Vaughan P.R. (1997),


Delayed collapse of cut slopes in stiff clay.
Geotechnique 47, No. 5, 953982

Budhu, M., Al. Karni,Y.(1993) The seismic bearing


capacity of soils, Geotechnique, 43 (1), p.p. 181187

Potts, D.M. and Zdravcovic L. (1999), Finite element


analysis in geotechnical engineering: theory.
London: Thomas Telford

Chen. Morgenstern N. R. and Chan D. H. (1992)


Progressive failure of the Carsington Dam: a
numerical study. Can. Geotech. J. 29, 971988

Pradel D., Smith P.M., Stewart J.P. and Raad G.


(2005), Case History of Landslide Movement
during the Nothridge Earthquake JGGE, ASCE,
11, 1360-1369

Dounias G.T., Potts D.M. and Vaughan P.R. (1988)


Finite element analysis of progressive failure: two
case studies Comput. Geotech. 6, 155175.
Gerolymos N., Vardoulakis I. and Gazetas G. (2007)
A
thermoporovisco-plastic
shear
band
model for seismic triggering and evolution of
catastrophic landslides. Soils and Foundations
47 (1).

Sarma S.K. (1999), Seismic bearing capacity of


shallow strip footings adjacent to a slope.
Earthquake Geotechnical Engineering, Seco e
Pinto (ed), Balkema, Rotterdam, 309-313. Proc.
Second International Conference on Earthquake
Geotechnical Engineering, Lisbon, Portugal

Hoeg, K. (1972). Finite element analysis of strain


softening clay, J. Soil Mech. Fdns Div. Am. Sot.
Ciu. Engrs,
43-59.

Sarma S.K. and Kourkoulis R.S (2004) Investigation


into the prediction of Sliding Block Displacements
in Seismic Analysis of Earth Dams, Proceedings
13 WCEE, Vancouver B.C, (in cd-rom)

Larsson R., Runesson K. and Sture S. (1991), Finite


element
simulation
of
localized
plastic
deformation. Arch. Aplpl. Mechanics 61, 305
317.

Skempton A.W. (1985): Residual strength of clays in


landslides, folded strata and the laboratory,
Geotechnique, 35(1).
Terzaghi K. and Peck R. B.(1948) Soil Mechanics in
Engineering Practice,. New York: Wiley

55

Tika Th.E., and Hutchinson J.N. (1999), Ring shear


tests on soil from the Vaiont landslide slip
surface. Geotechnique, 49(1)
Troncone A. (2005), Numerical analysis of a
landslide in soils with strainsoftening behavior,
Geotechnique 55, No.8, 585596
(a)

(b)

Peak strength

Mobilized friction angle

Critical state

cv

Slow residual state


(at large displacements)

r,s
OC clay

r,d
Fast residual State
(at large velocities)

10

Slow residual friction angle r,s : deg

40
35

A = 0.75 - 1.2

30

A = 1.5 - 4.3

25

A = 0 - 0.6

20
15
10
5
0

400

Displacement : mm

20

40

60

80

Clay fraction CF : %

Fig 1 : (a) Typical behavior of strain-softening soils (b) Curve for the calculation of the slow residual angle as
a function of clay fraction and parameter A

0.8
0.6

a:g

0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
0

10

15

20

t : sec

Fig 2 : JMA recorded time-history, Kobe 1995, Japan

Fig 3 : Contours of horizontal displacement for the free field case when the model is subjected to the JMA
accelerogram scaled at (a) 0.5 g and (b) 0.8g.

56

Fig 4 : Contours of plastic deformations generated for the free field case when the model is excited by the
JMA recode scaled at (a) 0.5g and (b) 0.8g.

Fig 5 : Contours of plastic deformations generated for the case of a 20 m wide footing of q=40kPa lying at a
distance from the crest of (a) 5m, (b) 8 m, (c) 11 m, subjected to 0.8g JMA excitation.

57

Fig 6 : Contours of plastic deformations generated for the case of a 20 m wide footing lying at 8m from the
crest with load (a) q=20kPa (b) q= 40 kPa, (c) q = 80 kPa, subjected to the JMA excitation at 0.8g.

3
1

(a)

(b)

u (m )

u :m 1.60

u :m 0.50
0.40

u (m )

1.20

0.30
0.80

0.20
node 1
node 2
node 3

0.10
0.00
0

10

20

node 1
node 2
node 3

0.40
0.00
0

30

10

20

30

t:s

t:s

Fig 7: Horizontal displacements at nodes 1, 2 and 3 respectively when the input motion is (a) a JMA scaled at
0.5 g and (b) JMA scaled at 0.8 g.

58

(a)

(b)

Fig 8: Contours of plastic deformations in the case of a one-storey building at 8m from the crest with load
q=20kPa subjected to the JMA 0.8g, when the foundation is (a) mat-foundation and (b) isolated
footings.

umax = 2 m !

umax = 2 m

umax = 2 m !

umax = 2 m

Fig 9: Contours of horizontal field displacements in the case of a one-storey building at 8m from the crest with
load q=20kPa subjected to the JMA 0.8g, when the foundation is (a) mat-foundation and (b) isolated
footings.

59

Fig 10: Envelope of maximum bending-moments on a structure with (a) isolated footings and (b) matfoundation when the excitation is the JMA record at 0.5g.

Fig 11: Envelope of maximum bending-moments on a structure with (a) isolated footings and (b) matfoundation when the excitation is the JMA record at 0.8g.

60

u1 (m)

Point D

Point C

Point

Point

U 1(m)
u1 : m

1.2

Point

0.8

Point
Point C

0.6

Point D

0.4

0.2

0
0

10

15

20

25

30

35

t (sec)
: sec

-0.2
0.1

0
0

10

15

20

25

30

Point C

u2 (m)
Point

Uu2
2(m)
:m

-0.1

Point D

Point

35

Point D
Point C

-0.2

-0.3

Point



-0.4

Point

t (sec)

-0.5

t : sec

Fig 12: Displacement time-histories at points A, B, C and D on a structure with isolated footings when the
excitation is the JMA record at 0.8
1.2
1.2

Point A
Point C

1.01

U (m)

u:m

0.8
0.8

Point B

0.6
0.6
0.4
0.4

0.2
0.2

0
0

10

15

10

20

15

20

-0.2

: sec
t t(sec)
0.7
0.6
0.6

()
: degrees

0.5
0.4
0.4
0.3
0.2
0.2
0.1

0
0

-0.1

10

10

15

15

20

20

25

25

30

30

35

35

t (sec)

t : sec

Fig 13: Displacement time-histories at points A, B, C and D on a structure with mat-foundation when the
excitation is the JMA record at 0.8g.

61

P (KN)

P : kN

300
300
200
200
100
100
0
0.05
0.05

0.1
0.1

0.15
0.15

-100

-100

0.2
0.2

y:m

y (m)

P (KN)

P : kN

1010
0
-0.002
-0.0015
-0.001
-0.0005
-0.002
-0.0015
-0.001 -0.0005
-10
-10

0.0005
0.0005

0.001
0.001

y:m

y (m)

-20
-20

-30
-30

Fig 14: p-y curves for two characteristic springs

()

(a)

(b)
()

Fig 15: Final soil displacements when (a) a shallow and (b) a piled foundation are used and the excitation
motion is the JMA record scaled at 0.8g.

62

(a)
U (m)

(b)
1

0.8

uu = 0.9
mm
=0.9

0.8

found

u ==0.7
0.7 mm
ufound

0.6

u:m

0.6

u:m

usoil =0.95 m

0.4

0.2

0.4

0.2

0
0

10

15

20

25

30

35

-0.2

10

15

20

25

30

35

t (sec)

-0.2

t : sec

Fig 16: Horizontal Displacement time-histories on (a) shallow and (b) piled foundation when excited by the
JMA record scaled at 0.8g.

(deg)
1

: degrees

0.8
0.6
0.4

shallow foundation

0.2

piled foundation

0
0

10

20

30
t (s)
t : sec

Fig 17: Compare the time-histories of the tilt angle between a shallow and a piled foundation when the input
motion is the JMA record scaled at 0.8g.

63

Detrimental effects of urban tunnels on design seismic ground motions


G. Kouretzis1, G. Bouckovalas1, A. Sofianos1, P. Yiouta-Mitra1
1

National Technical University, Athens, Greece.

Abstract
This research deals with the issue of whether, and under what circumstances, the presence
of underground tunnels should be taken into account for the earthquake resistant design of
neighboring surface structures in urban areas. In order to investigate the effect of the
underground structure on surface seismic motion, a series of dynamic plane-strain
numerical analyses were conducted, considering a circular tunnel embedded in a
viscoelastic half-space, and a harmonic SV-wave excitation. The numerical methodology,
based on the Finite Difference Method, aims at quantifying the effect of the soil medium
characteristics, excitation frequency, tunnel diameter, depth of construction, and relative
flexibility of the lining compared to that of the surrounding soil. Conclusions include
preliminary criteria identifying the cases when the presence of an underground tunnel
should be considered in the design of a surface structure.

amplification of the free-field motion. The area


where this phenomenon is noticed is referred in
the literature as shadow zone.
- The presence of an underground tunnel
generates a parasitic vertical component of
surface motion, with significant amplitude.
The most important parameters, rising from
the analytical results presented by the aforementioned researchers to be influencing the
response at the ground surface, are:
- The ratio of the depth of structure axis
over the radius of structure, H/. For small
values of H/, not only the amplification due to
the underground cavity is larger, but also
seismic waves are trapped between the cavity
and the ground surface, thus producing more
complex response patterns.
- The dimensionless frequency n=2/,
where is the wavelength. Higher values of n
generally produce higher amplifications and
more complex response patterns.
- The dimensionless distance along the
ground surface from the projection of the
structure axis, x/, where is the radius of the
structure. The wave field affected from the

INTRODUCTION

The effect of the construction of underground structures such as Metro tunnels, on


aboveground buildings has been examined
thoroughly in the past, focusing mainly on the
resulting surface permanent settlements. Yet,
the effect that the presence of a massive
circular tunnel has on the seismic design of
nearby buildings is neglected in current
practice.
Nevertheless, many researchers (e.g. Lee
and Trifunac, 1979, Manoogian & Lee, 1996,
Manoogian, 1998, Lee & Karl, 1992, Barros &
Luco, 1993 and 1994, Lee et. al, 2001) have
investigated via analytical methods the effect
that the presence of a circular underground
structure, embedded in an elastic half-space,
has
on the
surface
ground
motion.
Summarizing their findings the following can be
said:
- Maximum amplification of the ground
surface motion, consisting of SV-waves, due to
the presence of an underground unlined circular
structure varies between 75% - 100%.
- The presence of a cavity can cause, under
certain conditions, intense and selective de-

64

Fig 1: Comparison of numerically computed normalized horizontal displacements on the ground surface, with
the results of Luco & De Barros (1994). The case shown here applies to =0.001, n=0.5, =1/3.
Table 1: Range of physical parameters corresponding to the numerical models.

Circular unlined tunnel

Dimensions
Shear wave velocity, Vs
Harmonic excitation
frequency, T

D = 520m

Circular lined tunnel

D = 520m, d = 0.040.17m
2001000 m/sec
0.05 1.0 sec

Code FLAC (Itasca, 1999). Following the


verification of the numerical model, a series of
parametric analyses are presented, aiming at
the definition of preliminary criteria identifying
the cases when the presence of an
underground tunnel should be considered in the
design of surface structures.

presence of the underground structure


attenuates at low rates, and is approximately
analogous to the inverse square root of the
distance from the tunnel axis. However,
analytical methods proposed in the literature
have produced results limited to x/<4.
- The relative stiffness of the tunnel lining,
compared to the surrounding soil, as stiffer
structures
appear
to
produce
lower
amplifications.
Existing analytical solutions provide a clear
overview of the problem, and identify the basic
parameters influencing the ground response.
Nevertheless, as they are based on complex
mathematics, they cannot be readily used for
parametric analyses to examine the effect of a
wide range of values of the aforementioned
parameters. For this reason, in the present work
the problem is treated numerically, with the aid
of the commercially available Finite Difference

NUMERICAL MODEL AND VERIFICATION

The numerical model is verified against the


analytical results of Luco and De Barros (1994),
who applied an indirect boundary integral
method based on two-dimensional Greens
functions to obtain the harmonic motion on the
ground surface. Their model consisted of an
infinitely long unlined cavity of circular crosssection with radius , embedded at a depth H,
oriented parallel to the free surface of the halfspace. The soil was modeled as a viscoelastic

65

The range for these parameters was based


on common characteristics of underground
tunnels and wave characteristics representing
high, as well as low frequency excitations.
Table 1 presents the range of the structure, soil
and motion characteristics examined in the
analyses. The combinations of these parameters results in a range of 0.051 for the
dimensionless
frequency
n,
while
the
dimensionless depth H/ obtains values of 17.
Obviously, since the depth of one radius would
apply to a tunnel emerging at the surface, it can
no longer be considered as underground. Thus,
it has been included only for reasons of
completeness.
The index of relative flexibility, J, was
calculated with the following equation (St. John
and Zahrah, 1987), considering a concrete
structure lining with Elining = 30 GPa and
lining = 0.2. Combined to the soil characteristics,
it yields two singular values, namely J=5 and
J=150, corresponding to a very stiff and a very
flexible tunnel, respectively.

medium, characterized by propagation velocities of P- and SV- waves equal to a and b


respectively, while the simplifying assumption
for hysteretic damping a=b= was also
adopted.
In this paper, the numerical model was
prepared along the above lines in order to
provide the means of verification. The
commercially available Finite Difference Code
FLAC (Itasca, 1999) was selected to perform
the analyses. The boundary conditions applied
at the artificial boundaries were a) lateral
dashpots to minimize wave reflections and
achieve free-field conditions, b) absorbing
boundaries at the bottom i.e. normal and shear
dashpots of coefficient c=CsVs (Kramer, 1996)
to represent the effect of radiation damping and
c) a stress boundary of amplitude xy=2CsVs at
the bottom to simulate the incoming harmonic
SV- wave. The factor of 2 in the above relation
corresponds to the fact that half of the input
energy is absorbed by the viscous boundary,
while in the above relations Cs is the velocity of
S-wave propagating through the soil medium,
is the soil mass density and Vs the excitation
particle velocity. Despite the existence of a
vertical plane of symmetry, the full model was
used due to the limitations of the numerical
code with respect to lateral dashpots. Hysteretic
damping is simulated in an approximate fashion
as Rayleigh damping according to the
guidelines of Itasca (1999) and Sofianos
(2003).
Figure 1 depicts a sample of the satisfactory
comparison with the Luco & De Barros
analytical model. In Figure 1, surface displacement amplitudes are normalized against the
amplitude of the free-field incident motion Us,
and they are illustrated against the normalized,
against the tunnel radius, distance from the
tunnel axis x/

J=

2
2Esoil (1 lining
)Rlining 3

Elining (1 + soil )t lining 3

(1)

Strong ground motion effects were


investigated for a distance of 18 radii from the
structure axis. Finally, a parameter that was
separately examined was the hysteretic
damping of the soil. Three values compatible
with the elastic model were selected, namely
=1%, 2% and 5%, corresponding to soft rockstiff soil formations. Results have shown no
practical difference between the three values,
therefore =2% was adopted in all analyses.
INDICATIVE RESULTS

A selection of diagrams is presented,


illustrating the normalized amplitudes of ground
surface displacements against the horizontal
distance from the structure axis of symmetry.
Normalization has been achieved by dividing
the results with the free-field horizontal
displacement amplitude, for horizontal but also
for vertical parasitic displacements, since there
is no free-field vertical component of motion to
compare with the latter.

PARAMETRIC ANALYSES

A series of parametric analyses were


conducted to quantify the effects of parameters
H/ (dimensionless depth), x/ (dimensionless
distance), n (dimensionless frequency) and J
(index of relative flexibility of the lining
compared to that of the surrounding soil), as
discussed in the introduction.

66

Fig. 2: Normalized horizontal displacement amplitudes for n=0.2, H/=2.

Fig. 3: Normalized horizontal displacement amplitudes for n=1.0, H/=2.

Fig. 4: Normalized vertical displacement amplitudes for n=0.4, H/=2.

further reduced to 15% (Fig. 6). It is therefore


confirmed that only for H/2 (Fig. 5) does any
strong motion aggravation arise, with maximum
amplification occurring at H/=1 and reaching
85%. Observations are somewhat different in
the case of the parasitic vertical component of
motion. At H/=1.5, maximum amplifications for
all frequencies are between 100-175 %.
Moreover, at a depth of three radii there is still a
70% amplification for n=0.6 (Fig. 7) and even at
four radii there is a 50% amplification for n=0.4.
The effect attenuates faster and drops at 15%
at a depth of seven radii.

Effect of the dimensionless frequency n

Low frequency values such as n=0.05 and 0.1,


that infer high values of wavelength as
compared to the tunnel dimensions, result in
waveforms practically unaffected by the
structure. For n higher than 0.2, the response
changes considerably. De-amplifications in the
shadow zone reach 55% and amplifications
30% (Fig. 2). For n=0.41.0 the shadow zone
becomes prominent (Fig. 3). Similar remarks
can be drawn for the vertical, parasitic,
component of motion. Higher frequencies result
in a vertical component that cannot be ignored
as in lower frequencies. Figure 4 depicts the
highest of these values at 90% of the free-field
horizontal component.

Effect of the relative flexibility of the tunnel J

For small depths, it can be observed from


Figures 8, 9 and 12 that decreasing values of J
tend to eradicate the effect of the presence of
the tunnel. At larger depths, the effect of the
tunnel lining is practically nullified.

Effect of the dimensionless depth H/

For depths larger than two radii, the


horizontal component amplification is no larger
than 25%. For depths larger than 7 radii it is

67

Fig. 5: Normalized horizontal displacement amplitudes for n=0.2, H/=1.5.

Fig. 6: Normalized horizontal displacement amplitudes for n=0.8, H/=7.

Fig. 7: Normalized horizontal and vertical displacement amplitudes for n=0.6, H/=3.

design of a nearby surface structure, all results


were evaluated via combined plots of the
maximum amplification values and of the
location of that maxima on the ground surface.
Figures 13 and 14 are only samples of such
plots, as the complete results are not shown for
brevitys sake. From Figure 13 it is drawn that
maximum amplification of the horizontal motion
reaches about 1.5 for a shallow (H/<2) unlined
tunnel. In Figure 14 it may be seen that for
small depths (H/<2), only the area close to the
tunnel is affected (0<x/<1.0). For 2<H/<4 the
affected area extends to 5 radii and for larger
depths it starts at 2 and extends to 7 radii.
Summarizing remarks from all relevant
diagrams, the following can be said for an
unlined tunnel:
- Ground response for excitations with
wavelengths larger than the tunnel diameter is
not affected by the presence of the tunnel.
- The presence of a tunnel results in

The last remark can be only cautiously


extended to all frequencies for the case of the
horizontal component of motion. As can be
seen from Figures 10 and 11, despite the large
depth of six radii, increasing frequencies result
in horizontal amplifications larger even than
those of the unlined tunnel. More specifically,
for J=5, when n=1.0 and H/=6.0, ANx fluctuates
from 0.45 to 1.38, while the same analyses for
an unlined tunnel resulted in amplifications of
0.54 to 1.17. With respect to the vertical
component however, the increase of the
structure stiffness always results in a decrease
of strong motion amplification, with the effect
becoming more prominent at smaller depths.
CONCLUSIONS

In order to extract preliminary criteria


identifying the cases when the presence of an
underground tunnel may be neglected in the

68

Fig. 8: Normalized horizontal displacement amplitudes for n=0.2, H/=2 for an unlined cavity (J=inf), very
flexible lining (J=150) and stiff lining (J=5).

Fig. 9: Normalized horizontal displacement amplitudes for n=0.2, H/=4 for an unlined cavity (J=inf), very
flexible lining (J=150) and stiff lining (J=5).

Fig. 10: Normalized horizontal displacement amplitudes for H/=6 n=0.4 for an unlined cavity (J=inf), very
flexible lining (J=150) and stiff lining (J=5).

Fig. 11: Normalized horizontal displacement amplitudes for H/=6 n=1.0 for an unlined cavity (J=inf), very
flexible lining (J=150) and stiff lining (J=5).

69

Fig. 12: Normalized vertical displacement amplitudes for n=0.4, H/=2 and 5 for an unlined cavity (J=inf), very
flexible lining (J=150) and stiff lining (J=5).

Fig. 13: Combined plot of maximum normalized horizontal displacement amplitudes due to the presence of an
unlined tunnel, for all depths and frequencies.

Fig. 14: Combined plot of the maximum amplification location, due to the presence of a flexible tunnel
(J=150), for all depths and frequencies.

70

Itasca Consulting Group Inc, (1999) FLAC, Fast


Lagrangian Analysis of Continua, Users
Manual, Mineapolis, MI, U.S.A.

amplifications that should be considered in the


design of a surface structure for H/ <3.0.
- The horizontal component of motion can
be amplified by 20% to 85% within a distance of
eleven radii from the tunnel axis (0<x/<11).
- The ground response is further
complicated by the appearance of a parasitic
vertical component of motion. Maximum vertical
displacement amplitudes occur within four radii
from the tunnel axis (0<x/<4), reaching values
of 1/3 to up to 3 times the value of horizontal
free-field component.
From the analyses for a lined tunnel, the
following remarks can be drawn:
- Stiffer linings generally attenuate the
effect of the tunnel presence on the ground
response.
- For n>0.8 and H/>6.0, stiffer tunnels
cause unexpected for such depths amplifications of about 30%.
- The horizontal component of motion can
be amplified from 20% up to 40%.
The vertical component of motion reaches
values from 1/3 to 1.5 times the value of the
horizontal free-field component.
- The value of the relative stiffness does
not alter the shadow zone, which therefore
coincides with that of the unlined tunnel.

Lee, V. W., (1977) "On the deformations near


circular underground cavity subjected to
incident plane SH-waves ", Proc. Application of
computer methods in engineering conference,
University of South California, L.A., pp.951-962.
Lee, V. W., Karl, J., (1993) "On the deformations
near a circular underground cavity subjected to
incident plane P waves", European Earthquake
Engineering 1, 29-39.
Lee, V.W., Karl,J., (1992) "Diffraction of SV-waves
by underground, circular, cylindrical cavities.",
Soil Dynamics and Earthquake Eng. 11,445456.
Lee, V.W., Trifunac, M.D., (1979) "Response of
tunnels to incident SH- waves.", J.eng.
mech.div. ASCE 105, 643-659.
Luco, J.E., De Barros F.C.P., (1994) "Dynamic
Displacements and stresses in the vicinity of a
cylindrical cavity embedded in a half- space.",
Earthquake Eng.& Structural Dynamics
23,321-340.
Manoogian ,M. E., "Surface motion above an
arbitrarily shaped tunnel due to elastic SH
waves." in Proc. Geotechnical Earthquake
Engineering and Soil Dynamics III.
Manoogian, M.E., Lee, V.W., (1996) "Diffraction of
SH-waves by subsurface inclusions of arbitrary
shape", J.eng. mech.div. ASCE 122,123-129.

ACKNOWLEDGEMENTS

The research was funded by Earthquake


Planning and Protection Organization, Greece.

Penzien.J., (2000) "Seismically induced racking of


tunnels linings", Earthquake Eng.& Structural
Dynamics 29,683-691.

REFERENCES
Kramer L. S., (1996) Geotechnical Earthquake
Engineering, Prentice Hall.

Sofianos, A.I., (2003) Effect of underground


structures on design response spectra, Final
Report (3), Earthquake Planning and
Protection Organization, Greece.

Datta, T.K., (1999) Seismic response of buried


pipelines: a state-of-the-art review, Nuclear
Engineering and Design, Vol. 192, Issues 2-3,
271-284.

St. John, M.C., Zahrah, T.F., (1987) "Aseismic


Design of Underground Structures", Tunneling
& Underground Space Technology 2(2),165197.

Davis, C.A. , Lee, V. W. , Bardet, J.P., (2001)


"Transverse response of underground cavities
and pipes to incident SV waves.", Earthquake
Eng.& Structural Dynamics 30, 383-410.

Vanzi,V., (2000) "Elastic and inelastic response of


tunnels under longitudinal earthquake excitation.",
J. Earthquake Engineering 4 (2),

De Barros F.C.P. & Luco, J.,E., (1993) "Diffraction of


obliquely incident waves by a cylindrical cavity
embedded in a layered viscoelastic halfspace", Soil Dynamics and Earthquake Eng.
12, 159-171.

71

Seismic Analysis of Near-Surface Tunnel Cross-Sections


Christos Vrettos
Technical University of Kaiserslautern, Germany

Abstract
Seismic loads exerted on tunnels in soft soils are controlled by the vibrational characteristics
of the surrounding soil and the flexibility of the tunnel lining. The analysis is performed by
the deformation method using imposed displacements. Available closed-form expressions
to calculate deformations and sectional forces are compared. By means of a finite element
model the accuracy of the decoupled analysis of the associated soil-structure-interaction
problem is assessed, and the influence of the tunnel embedment depth is investigated.
Finally, a 2D dynamic analysis of a shallow and of a deep tunnel in a layered soil deposit is
carried out to identify the relevant features of the seismic response.

INTRODUCTION

the tunnel lining. The present paper deals with


this loading case.

Tunnel structures are generally considered


less vulnerable to seismic actions than above
ground structures. In contrast to surface
structures, whose seismic response is governed
by inertia effects, the response of tunnels
embedded in the ground is primarily kinematic,
i.e. it is caused by the compatibility of the tunnel
deformation to that of the surrounding ground.
Therefore, soil-structure interaction effects are
of fundamental importance. The stiffness ratio
between tunnel structure and soil determines
the intensity of sectional forces. The analysis is
usually carried out by imposing appropriate
displacement patterns (deformation method).
Inertia effects are of secondary importance and
can be neglected. The available analysis
methods are presented by St. John and Zahrah
(1987), Wang (1993), Hashash et al. (2001),
and also by the author (Vrettos, 2005).
Impinging seismic waves cause in the
longitudinal direction i) axial deformations due
to alternating compression and tension, and ii)
bending due to curvature arising from the wave
components with particle motions perpendicular
to the tunnel axis. These cases are not
considered herein.
In the tunnel cross section vertically
propagating shear waves cause distortion of the
circular or rectangular cross sectional shape of

OVALING OF CIRCULAR TUNNELS

We first consider circular tunnel cross


sections subjected to shearing under plane
strain conditions. The equations given in the
following are based on the work by Wang
(1993) and Penzien (2000).
When the tunnel stiffness is equal to that of
the ground, the tunnel is modelled as a hole
filled with soil. The diametric strain ' ff in the
tunnel for shear strain J is
' d ff
d

J
2

(1)

where d ist he tunnel diameter.


For very low values of the tunnel stiffness
we obtain the other limiting case of a perforated
soil slice. The diametric strain 'dc is then

'd c
d

r 2 J (1  Q)

(2)

where Q is the Poissons ratio of the linearelastic soil medium. It becomes evident that the
relative stiffness between tunnel and soil
dominates the response. Two quantities are

72

used for this purpose: the compressibility and


the flexibility ratio, C and F, respectively:
C

E (1  Q "2 ) d
2 E " t " (1  Q)(1  2Q)
Q 2" ) d 3

E (1 
48 E "I " (1  Q)

K2

F [(1  2Q)  (1  2Q)C ]  (1  2Q)2 / 2  2


F [(3  2Q)  (1  2Q)C ]  C[2.5  8Q  6Q 2 ]  6  8Q
(10)

(3)

Expressions for diametric strain and bending


moment have not been derived yet, and it is
recommended to apply for the design the
values for full-slip conditions.
Solutions
for
sectional
forces
and
deformations due to soil-structure interaction
have also been derived by Penzien &
Wu (1998) and Penzien (2000). Assuming fullslip contact conditions the following closed form
solutions are obtained:

(4)

In the above equations E is the Youngs


modulus of the soil, and E" , I " , Q " and t " are the
Youngs modulus, the moment of inertia, the
Poissons ratio, and the thickness of the tunnel
lining.
Analytical solutions for the diametric strain,
the axial load, and the bending moment have
been derived by assuming full-slip contact
conditions between tunnel lining and soil.
Following Wang (1993) we obtain:

' d tunnel
' d ff

Nmax

2
K1F
3

n
' dtunnel

(5)

E d
1
J
K1
6
(1  Q) 2

1

(6)

Nmax

Mmax

Qmax

Mmax

E d
1
K1
J
6
(1  Q) 4

Rn
12(1  Q)
2F  5  6Q

Nmax

6E " I " R n

d 2 (1  Q 2" )
3E "I " R n

d (1  Q 2" )

d
J
2

(12)

12E "I " R n

d 2 (1  Q 2" )

(11)

(13)
J

(14)

4(1  Q)

(15)

Dn  1

(8)
with

Sliding along the contact surface is possible


only in tunnels embedded in very soft soils
under strong seismic actions. In most cases in
practice an intermediate condition between fullslip and bonded contact will develop. The
assumption of sliding yields conservative
estimates of sectional strain and bending
moments.
Under seismic shear loading maximum
values for axial forces in the tunnel are obtained
under the assumption of bonded contact:
E d
rK 2
J
(1  Q) 2

Rn

with R n defined as the racking ratio for fullslip conditions:

(7)

with
K1

R n ' d ff

Dn

12E "I " (5  6Q)

(16)

d 3G(1  Q 2" )

where G denotes the shear modulus of the


soil.
For the case of bonded contact the
corresponding solutions are obtained by
replacing R n by R in the above equations (11)
to (14), cf. Penzien (2000):
' d tunnel

(9)

Nmax

where

73

R ' d ff

12E "I " R

d 2 (1  Q 2" )

d
J
2

(17)
(18)

M max

Qmax

3E "I " R
Q 2" )

d (1 
12E "I " R
2

d (1 

Q 2" )

NUMERICAL INVESTIGATION ON
OVALING OF CIRCULAR TUNNELS

(19)
J

The numerical study is carried out by using


the finite element method code PLAXIS (2004).
Since modelling of full-slip is difficult with this
code a bonded contact is assumed here. The
equations presented above for a perforated soil
slice with/without tunnel are valid for a fullspace, i.e. the boundaries are located in a
sufficiently large distance from the structure.
The required minimum size of the FEM-mesh is
determined by comparing the numerical solution
with the analytical solution of a perforated soil
slice equation (2). For the soil assumed here
( Q 0.3) and J = 2.5210-3, we obtain for a
quadratic mesh of dimension equal to 15d an
accuracy in the value of diametric strain of
approx. 1.3%. The respective results for the
racking ratio and the sectional forces are:

(20)

where
R

4(1  Q)
D 1
24E "I " (3  4Q)
r

d 3G(1  Q 2" )

(21)
(22)

Comparison of the above equations is made


by using an example. The seismic shear strain
is estimated from the approximation by
Newmark (1967):
J

v
Cs

(23)

R 2.164,
Mmax 157.3 kNm/m,

where v is the maximum particle velocity of


the seismic excitation, Cs
G / U is the shear
wave velocity in the soil, and U is the density.
The following numerical values are selected
here: v = 0.63 m/s, Cs = 250 m/s, and
U = 1.92 Mg/m3. For the Poissons ratio we set
Q = 0.3.
For the tunnel lining the following values are
selected: d = 6 m, t " 0.3 m (cross sectional
0,3m2/m and moment of inertia
area A"
E"
I"
0.00225 m4/m),
24.80106 kPa,
Q " 0.20. It is further assumed that the tunnel
is located in a large depth so that the influence
of the free ground surface can be ignored.
For the case of full-slip equations (6) - (7)
and equations (12) - (13) yield identical results:
Nmax 62.94 kN/m und Mmax 188.81 kNm/m.
For the case of bonded contact equation (9)
by Wang (1993) yields an axial force
Nmax 1.045.8 kN/m while equation (18) by
Penzien (2000) yields a much lower value of
Nmax 124.4 kN/m. Bending moment and shear
force from equations (19) and (20) are
Mmax 186.95 kNm/m and Qmax 124.64 kN/m,
respectively. Racking ratio R according to Wang
is 2.58, while the equation by Penzien yields
2.55. The large discrepancy between equations
(9) and (18) has also been identified by
Hashash et al. (2005) and is investigated next.

Nmax
Qmax

1.040 kN/m,
108.1 kN/m

Based on this, the solution by Penzien


underestimates the axial force N by a factor of 8
and is thus not acceptable for use in design
practice. In contrast the solution by Wang
shows an excellent agreement with the
numerical solution. The values for the other
sectional forces show an accuracy of approx.
20%. Similar findings have been reported by
Hashash et al. (2005).
RACKING OF RECTANGULAR TUNNELS

Closed-form solutions for sectional forces


considering also the effects of soil-structure
interaction do not exist. A pseudo-coupled
procedure is usually applied: The distortion of
the structure is first estimated and subsequently
imposed on the structural system in order to
determine the sectional forces, cf. Wang (1993)
and Penzien (2000). This procedure is
summarized in the following.
The maximum relative deformation of the
tunnel lining due to the shear strain J induced
by the earthquake is:
' tunnel

J tunnel H

(24)

where H is the height of the rectangular


tunnel box. The relative free-field displacement

74

when the presence of the tunnel is ignored is


' ff J H .
Depending on the value of the relative
stiffness between tunnel and surrounding soil
' tunnel can be larger or smaller than ' ff . The
corresponding racking ratio is defined by:

' tunnel
' ff

Multiplication by R yields 'tunnel . This


displacement is finally applied to the structural
system of the tunnel, and the sectional forces
are determined.

(25)

The relative stiffness between tunnel and


soil is defined in terms of the flexibility ratio F:

GB
k" H

(26)
Fig 1:

where G is the strain-compatible shear


modulus of the soil, B is the width of the tunnel,
and k " is the horizontal stiffness of the
rectangular box. The latter is determined by
standard structural analysis methods on a
simple system consisting of the tunnel box
without the surrounding soil. It should be
noticed that the definition of the flexibility ratio is
different for circular and rectangular tunnels, cf.
equation (4) and equation (26), respectively.
A closed form solution for the racking ratio R
for deeply embedded tunnels has been derived
by Penzien (2000). It is determined from
equation (21), where D is given by the
following relationship:
D

ks

k " (3  4Q)
ks

G
H

Simple system to determine horizontal


stiffness of tunnel lining; W " k " / B .

The accuracy of the above procedure is


examined by comparison with a numerical
solution. A point that deserves attention is the
proper selection of Youngs modulus for the
tunnel under plane strain conditions. Finite
element codes like PLAXIS directly use the
Youngs modulus E and the Poissons ratio Q ,
while in beam analysis E has to be replaced by
E E /(1  Q 2" ) .
As an example we consider a rectangular
tunnel with dimensions B / H = 10 / 4 m, wall
thickness t" = 0.30 m, E" = 24.80106 kPa, and
Q " 0.20. For the soil we select G = 120 MPa
and Q = 0.3. The seismic shear strain is
assumed to act statically with J = 2.5210-3.
Considering the above comments regarding
the Youngs modulus E, the flexibility ratio F for
equation (26) can be calculated via an
expression given by Wang (1993) to F = 48.17.
PLAXIS yields for the system depicted in Fig 1
under unit shear stress loading and E A f a
maximum horizontal displacement equal to
1.6110-3 m, that corresponds to F = 48.30.
The racking ratio R according to
Penzien (2000) is then calculated with
D = 0.0373 to R = 2.70. For determining the
sectional forces in the lining the tunnel box
structure is subjected to a differential
displacement between roof and base equal to
' tunnel = 2.702.5210-34 = 27.2210-3 m.
In order to check the accuracy of this
pseudo-coupled procedure we compute the

(27)
(28)

The racking ratio is equal to 1 when the


stiffness of the tunnel lining k " is equal to that
of the soil k s . When k " becomes very small,
we obtain the limiting case of a perforated fullspace slice, i.e. R 4(1  Q) . It is evident, that
for very flexible structures the distortion of the
structure may be larger than that in the freefield, i.e. R ! 1.
Free-field displacement ' ff is determined
from a seismic site response analysis of the
particular soil profile or estimated by the
Newmark
approximation,
equation (23).

75

2D SEISMIC RESPONSE ANALYSIS

response of the entire tunnel-soil system under


shear loading using the code PLAXIS. The size
of the mesh is set equal to 10 times the tunnel
width and height, respectively. Distortion of the
tunnel is calculated from the lengths of the two
diagonals d1 and d 2 of the distorted cross
section:
' tunnel
J tunnel
(d 22  d12 ) /(4BH )
(29)
H

A dynamic 2D analysis by means of the


finite element method as usually carried out in
the frame of the detailed design of tunnel
projects is presented next.
We consider a layered soil deposit with a
thickness of 40 m subjected to a seismic motion
defined at outcropping rock. The acceleration
time history selected refers to the Loma Prieta
1989 earthquake recorded at Palo Alto SLAC
Lab, USGS Station 1601. The site conditions
correspond to rock. In the predominant
horizontal direction peak ground acceleration is
0.278 g that is scaled here to a peak value of
0.27 g.
First, a 1D free-field seismic site response
analysis is conducted using SHAKE. For the
variation of shear modulus and damping ratio
with shear strain we assume for all layers a
clayey soil of medium plasticity. The initial
small-strain shear moduli as well as the straincompatible ones are depicted in Fig 2. At the
interface between soil and bedrock a maximum
acceleration of 0.20 g is obtained from the
analysis while at the soil surface the maximum
acceleration is 0.45 g. The distribution of
maximum shear strain with depth is given in the
top graph of Fig 3.

From the numerical analysis we get J tunnel =


6.510-3 yielding a racking ratio R = 2.579.
Hence, the accuracy compared to the closedform solution by Penzien (2000) is 5 %.
INFLUENCE OF EMBEDMENT DEPTH IN
APPROXIMATE SOLUTION

The above closed form solutions are valid


for deep tunnels. The presence of a free
surface results in shear strain that varies with
depth. Furthermore, the stress distribution in
the vicinity of the tunnel depends on the
distance between tunnel roof and ground
surface. The effect of tunnel embedment depth
is investigated next by means of a numerical
analysis. We consider the rectangular tunnel
cross section used in the previous example and
vary the position of the tunnel relative to the
ground surface with cover thickness ranging
from 18 m to 2 m. The distance of the tunnel
from the bottom mesh boundary is kept
constant. The applied shear strain is set equal
to J = 2.52 10-3 m. The computed values of the
racking ratio R according to equation (25) are
summarized in Table 1.

Depth [m]

-10

Table 1: Variation of racking ratio with soil


cover thickness
tcover
[m]

18

12

2.58

2.52

2.44

2.26

2.08

-20

-30

It can be seen that the racking ratio varies


only weak with the soil cover thickness t cov er .
For a cover thickness equal to the tunnel height
the value of racking ratio is reduced by 13%
compared to the value of a deep tunnel. Similar
results are obtained for a much softer soil.
Hence, the solutions for deep sited tunnel
consists a practicable approximation.

-40
0

100

200

300

Shear modulus [MPa]

Fig 2:

76

Shear modulus profile in 1D analysis: The


dotted line is for the initial values, the solid
line for the strain compatible ones.

For the tunnel we select the single cell


structure of the previous examples with
dimensions height / width equal to 4 m / 10 m.
Two
positions
will
be
investigated
corresponding to soil cover thickness of 16 m
and 5 m, respectively.
The 2D seismic response analysis is carried
out using PLAXIS. The system width is set
equal to 100 m. The code works in the timedomain using frequency dependent Rayleigh
damping for the soil, thus requiring some
approximation to simulate the strain-compatible
hysteretic damping obtained from the 1D
analysis. For the shear modulus we use
constant values for the individual layers as
given in Fig 2. For the Poissons ratio we
assume Q = 0.3 for all layers. The seismic input
is defined at the bottom boundary of the finite
element mesh and corresponds to the time
history obtained by SHAKE at the soil/bedrock
interface with a maximum acceleration of
0.20 g.
The free-field analysis of this 2D system
yields at the surface a maximum acceleration of
0.3 g that is considerably lower than the 1D
value. The source of this discrepancy may be
found mainly in the 2D nature of the model, but
also in the Rayleigh damping approximation.
In the next step we analyze the system
including the tunnel structure. We monitor the
response at the upper right corner of the tunnel
structure, point C. The sectional forces are
evaluated at that point in time when the
maximum acceleration at point C occurs. For
the deep tunnel the maximum bending moment
is 565.2 kNm/m at an acceleration in point C
equal to 1.57 m/s2. For the shallower tunnel we
obtain a much lower value for the bending
moment equal to 283.3 kNm/m while the
acceleration at point C shows a much higher
value of 2.8 m/s2, cf. Fig 4.
Under 2D free-field conditions the
acceleration values at the depth level
corresponding to point C are: 1.72 m/s2 for the
deep and 2.45 m/s2 for the shallow tunnel,
respectively.
The fact that although the acceleration for
the deep tunnel is lower the associated bending
moment is larger is attributed to the larger shear
strain level along the height of the tunnel, cf.
Fig 3. This indicates that relative deformation is
governing the tunnel response.

Depth [m]

-10

-20

-30

-40
0

0.05

0.10

0.15

0.20

max. shear strain [%]

Depth [m]

-10

-20

-30

-40
0

0.05

0.10

0.15

0.20

shear strain [%]

Fig 3: Distribution of maximum shear strain with


depth. The top graph is from the 1D, the
bottom one from the 2D free field analysis
at the time of maximum acceleration at the
level of the shallow tunnel. The dotted lined
boxes indicate the tunnel position of the two
cases considered.

77

Hashash, Y.M.A., Hook, J.J., Schmidt, B., and Yao,


J.I.C. (2001) Seismic analysis and design of
underground
structures,
Tunnelling
and
Underground Space Technology, Vol. 16, pp.
247-29.
Newmark, N.M. (1967) Problems in wave
propagation in soil and rock, Proc. Int. Symp.
Wave Propgation and Dynamic Properties of
Earth Materials, New Mexico, Univ. of New
Mexico Press.
Penzien, J. (2000) Seismically induced racking of
tunnel linings, Earthquake Engng. Struct. Dyn.,
Vol. 29, pp. 683-691.
Penzien, J., and Wu, C.L. (1998) Stresses in linings
of bored tunnels, Earthquake Engng. Struct.
Dyn., Vol. 27, pp. 283300.
PLAXIS b.v. (2004): PLAXIS Finite Element Code for
Soil and Rock Analyses, Version 8, Delft,
Netherlands.
St. John, C. M. and Zahrah, T. F. (1987) Aseismic
Design of Underground Structures, Tunnelling
and Underground Space Technology, Vol. 2,
No. 2, pp. 165-197.
Fig 4:

Distribution of bending moments at the


point in time when maximum acceleration
occurs at the upper right connection
roof/wall.

Vrettos, C. (2005) Earthquake resistant design of


tunnels,
Rational
Tunnelling,
2nd
Summerschool, Innsbruck, D. Kolymbas &
A. Laudahn (Eds), pp. 261-283.
Wang, J.-N. (1993) Seismic Design of Tunnels: A
State-of-the-Art Approach, Monograph 7.,
Parsons, Brinckerhoff, Quade and Douglas Inc.,
New York.

CONCLUSIONS

Available closed form expressions for


computing sectional forces and deformations
due to ovaling/racking deformation of tunnels
have been compared with solutions obtained by
a finite element code, and some known
discrepancies for the condition of bonded
contact have been validated. The effects of the
free soil surface are assessed.
The difficulty in matching 1D and 2D free
field analyses is shown. The seismic response
of a single cell tunnel embedded in layered soil
is investigated for two embedment depths
demonstrating the relevance of kinematic soil
structure interaction and the appropriateness of
the deformation method for the design.
REFERENCES
Hashash, Y.M.A., Park, D, and Yao, J.I-C. (2005)
Ovaling deformations of circular tunnels under
seismic loading, an update on seismic design
and analysis of underground structures,
Tunneling and Underground Space Technology,
Vol. 20, pp. 435-441

78

Sliding of Rigid Block on Sloping Plane:


The Surprising Role of the Sequence of Long-Duration Pulses
E. Garini and G. Gazetas
National Technical University of Athens, Greece

Abstract
A numerical study is presented for a rigid block supported through a Coulomb friction
contact surface on an inclined plane subjected to parallel excitation. As excitation we utilize
idealized wavelets and near-fault seismic records strongly influenced by forward directivity
or fling-step effects (Northridge 1994, Kobe 1995, Kocaeli 1999, Chi-Chi 1999). For a given
level of peak acceleration the sliding displacement is found to be dependent on the following
parameters: the ratio ac1/aH of the critical sliding acceleration ac divided by the peak input
acceleration aH, the excitation frequency fH, the inclination angle of the plane, and, as
significantly, on the sign (+ or -), the sequence, and even the details of pulses of the
excitation time history. The parametric analysis shows that the displacements induced due
to sliding on inclined plane may exceed by far those induced on a horizontal plane, due to
the accumulation of slippage in one direction. Dynamic response of the block-base system
is found to be very sensitive to the unpredictable details of the excitation time history.

INTRODUCTION

acceleration aH (in g units). Moreover the rigid


block is supported through a Coulomb frictional
surface with a constant coefficient of friction
on the inclined plane.
The forces acting on the block are, as
illustrated in Figure 1b: (a) the weight, B, with its
two components, (b) the inertial force, F, due to
the blocks acceleration, (c) the perpendicular
reaction, Fk , of the plane and (d) the friction
force, T = Fk. Dynamic equilibrium of the block
results to the yielding downhill or uphill
acceleration. Respectivelly:

In earthquake geotechnical engineering the


analog of dynamic sliding of a block on an
inclined plane has been used to estimate the
response of earth dams, embankments, and
retaining walls during earthquakes. Introduced
in 1965, by Newmark the analog was further
utilized Seed and Martin (1966), Ambraseys
and Sarma (1967), Makdisi and Seed (1978)
who developed procedures for predicting the
permanent displacements of dams, applying
Newmarks sliding-block concept. Additional
studies of slip in relation to the strength of
ground shaking have been made by Richards
and Elms (1979) for evaluating the response of
shallow foundations and gravity retaining walls.
Also, Whitman and Lin (1983), Yegian et al
(1988), Gazetas and Uddin (1995), Kramer and
Smith (1997) used Newmarks model in
analyzing the stability of dams and
embankments.
The seismic slope stability problem is
analogous to the problem of a rigid block of
mass m, resting on an inclined plane which
forms an angle with the horizontal (Fig 1a).
The plane is subjected to seismic excitation
parallel to the base with peak ground

ac1 = cos sin


ac2 = cos + sin

(1)
(2)

Evidently, ac1 < ac2.


The acceleration ac1 at which the mass yields
downhill is called critical sliding acceleration.
Imposed bases acceleration in every moment
has a different value, aH(t), which can be
greater of lower than ac1. As long as the ground
acceleration, aH(t), does not exceed ac1 the
block does not slip. When aH(t) surpasses ac1
sliding takes place. The block can no longer
accelerate as fast as the plane and moves
relative to the base with a constant yielding
acceleration ac1. Because the block is assumed

79

cycle at the Ricker wavelet. As the ratio ac1/aH


decreases from 0.4 to 0.2, these two smaller
pulses cause significant sliding in the opposite
direction of the main cycle. For the other types
of idealized wavelets (T-Ricker, Sinus, Tsang)
such paradoxical behavior is not noticed (see
Figure 3).
Slippage is roughly inversely proportional to
ratio ac1/aH. By increasing ac1/aH, the velocity
gradient of the block is increasing and
consequently sliding occurs for a shorter period
with duration almost proportional to ac1/aH. The
explanation is conspicuous as can be seen
from Figure 4: In general, the resulting sliding is
equal to the area which is enclosed between
the velocity time histories of the block and the
base. The first sliding uphill is almost the same
for different values of ac1/aH (striped area in
velocity time-histories of Fig.4). However, at t =
1.7 sec the block stopped sliding uphill and
starts yielding at the opposite direction. That
makes the difference at the blocks response:
As illustrated in Fig. 4, when ac1/aH = 0.05
sliding downhill lasts 2 sec (starts at 1.7 sec
and finishes at 3.7 sec) whereas for ac1/aH = 0.1
yielding downhill lasts only 1 sec (from 1.7 sec
until 2.7 sec). As a result, the induced
displacement at the end of the sliding, which is
simply the shaded triangular area in Fig. 4, is
reduced to one half when the ratio ac1/aH is
doubled.

to be rigid, the acceleration throughout sliding is


constant. The movement continues until the
velocities of the block and the ground equalize.
Knowing the critical acceleration and the time
history
of
base
excitation,
permanent
displacements in every sliding period can be
calculated by a straightforward double
integration process.
Due to the transient nature of the
earthquake loading, the block may be subjected
to a number of acceleration pulses equal to, or
higher than its critical acceleration, which simply
produce some permanent deformation rather
than complete failure. Theoretically, sliding on
an inclined plane can happen in both directions:
uphill or downhill. However in practice, for
planes with inclination angle greater than 5o
sliding occurs only at downhill (where the mass
yields with acceleration ac1). Therefore, the
permanent deformation induced in every sliding
period, is added to the displacements of the
previous sliding periods.
ASYMMETRIC SLIDING INDUCED BY
IDEALIZED WAVELETS

Four types of idealized wavelets are used as


base excitation in this study: i) Ricker wavelets
of characteristic frequency: fH = 1, 2, and 4 Hz,
ii) One-cycle sinus pulse of frequency: fH = 1.5
and 3 Hz, iii) T-Ricker wavelet of frequency: fH
= 1 and 2 Hz, and iv) Tsang wavelet of
frequency: fH = 1 and 2 Hz. A number of
numerical analyses are performed in order to
understand the influence of each parameter in
asymmetric sliding. So, a wide range of values
is given to each parameter: ac1/aH = 0.05, 0.1 ,
0.2 , 0.4 , 0.6 and = 0o, 5o, 25o. Every
excitation time history is applied with its normal
and opposite sign (right and left in Figures).

Influence of the excitation frequency fH


Increasing the excitation frequency, leads to
dramatically decreasing slippage (with the ratio
ac1/aH and the angle remaining constant). The
explanation is straightforward: The higher the
frequency, the shorter the duration of block
acceleration relatively to the ground (Fig. 5).
Consequently, the velocities of the block and
the ground equalize more quickly and sliding
occurs for a shorter period, hence smaller
displacement. In Figure 6 is depicted the sliding
response for two different frequencies of Tsang
pulse in respect to yielding acceleration ratio
ac1/aH. The aforementioned trends are seemed
to be valid: the higher the frequency, the
smaller the slippage.
In addition, sliding is inversely proportional
to the square of the vibration frequency. Figure
6 shows that when frequency fH is doubled, the
induced slippage reduces by a factor of 4.

Influence of ac1/aH ratio


Evidently, increasing the ratio ac1/aH while
keeping the values of inclination angle and
peak ground acceleration aH constant, leads to
reduced slippage. Sliding displacements are
decreasing monotonically with increasing
friction coefficient, when > 0. However, when
= 0o (i.e. for sliding on horizontal plane) for a
Ricker excitation a paradoxical reduction of is
observed with decreasing ac1/aH (Fig. 2). This
phenomenon is caused by the presence of the
two smaller pulses before and after the major

80

The relative displacement d(t) results from the


integration of the differential of velocities
between the block and the base. In simple
cases such as in Fig. 6, this integral is given by
the area of a fundamental shape. Particularly in
Fig. 6, is the area or the almost rightangled
triangle whose one side is the yielding period
and the other one is the blocks velocity at the
beginning of the sliding. Notice that for fH = 1.5
Hz sliding lasts 4 sec (from 1.7 to 5.7 sec) while
for fH = 3 Hz sliding holds the half time (from 1.3
to 3.3 sec). Moreover, blocks velocity at the
beginning of the yielding is 2.2 m/s when fH =
1.5 Hz but it is reduced to the one half (1.1 m/s)
when fH = 3 Hz. Therefore by doubling the
excitation frequency, response increased by a
factor of 4.
Normalization of asymmetric response with
respect to the ratio of peak ground acceleration
divided by the square of peak ground velocity,
aH/VH2, converges the results to a single line; for
idealized pulses like T-Ricker and sinus, see
Figure 7 and 8 respectively. Observe that for all
excitation frequencies, fH, the normalized
slippage for any particular ratio ac1/aH is of the
same value.

Influence of excitation direction


Response of sliding block is strongly
depended on the direction (+ or -) of the
imposed excitation, as illustrated in Fig. 11, only
for the cases of asymmetric sliding ( 0).
A dramatic example is offered in Fig. 12. It is
presented the response of a rigid block sliding
on an inclined plane subjected to a single sinus
pulse. Time-histories on the left shows that the
first positive acceleration pulse of the base
induces downhill sliding, reaching a mere 0.6 m
in magnitude for 0.7 sec duration. The
immediately ensuing negative acceleration
pulse decelerates the body until sliding stops
without reversing uphill.
In contrast, when the direction of sinus pulse
reverted (right plots in Fig. 11) the sequence of
positive and negative cycles change. This time
the first acceleration pulse is the negative one,
causing as expected only a minor slippage
uphill, hardly moving the block from its rest
position. Then the positive acceleration pulse is
imposed leading to an almost unstoppable
downhill sliding of 4.2 m for 4 sec duration! It
appears that the vast difference between these
two cases is fully justified by the crucial
presence of the negative acceleration pulse at
the beginning or the end of excitation, because
this pulse acts like a brake in blocks sliding. A
practical conclusion: during earthquakes,
observed differences in behavior between
various
slopes,
retaining
walls,
and
asymmetrically-yielding structures (in general),
usually attributed to local site effects or different
structural characteristics, may only be the
unfortunate outcome of asymmetric sliding or
yielding,
arising
from
the
completely
unpredictable lack of symmetry of ground
shaking.

Influence of inclination angle


Generally, as the inclination of the base
increases, both the magnitude and the duration
of downward sliding also increase. As angle
increases, while keeping ac1/aH and fH constant,
sliding uphill becomes harder. It can be seen
from Figure 9 that for = 5o the block after the
first sliding downhill, starts at the moment t = 2
sec sliding uphill while for = 25o at the same
time (t = 2 sec) block is moving in absolute
contact with its baseno uphill sliding occurs.
Figure 10 depicts the acceleration, velocity
and slippage response of a block on either a 5o
inclined plane (left) or a 25o inclined plane
(right). The excitation is a Tsang pulse of 1 Hz
frequency. In case of the 5o slope not only
downward movement occurs; mild inclination
permits an uphill sliding of almost 0.3 m. On the
other hand, when the slope is steep ( = 25o)
yielding is always downhill leading to
substantially larger slippage.

ASYMMETRIC SLIDING INDUCED BY NEARFAULT GROUND MOTIONS

To verify the above trends obtained only


with idealized wave forms, analyses with actual
accelerograms are performed. Fortunately in
the last two decades, several strong
earthquakes have occurred providing a great
number of near-fault records, which are
influenced by forward rupture directivity and
fling-step effects (Somerville 2000, Abrahamson
2003). A selection of 15 horizontal components
of records is used as input shaking parallel to

81

the significance of the form of accelerogram is


portrayed in Figures 15, 16 and 17, which refer
to the north-south component of TCU068
record of the Chi-Chi 1999 earthquake. This
record comprised long period pulses of T 3
sec which reach acceleration values of only
0.20 g (compared with the much higher PGA
value of 0.42 g).
As can be seen from Figure 15, for a ratio
ac1/aH = 0.05, this excitation produces a sliding
displacement of 25 m! This is clearly due to the
mentioned long period pulse which starts at
about 7 sec. To further prove the importance of
this long duration pulse, a simplifying
approximation of the TCU068 n-s record is
made by clipping-off high accelerations and
keeping only the fundamental pulses of the
record. Figs 16-17 reveal that the resulting
slippage remains almost the same with that
produced with the actual record, especially in
case of the inverted accelerogram.

the base:

JMA 00, JMA 90 (Kobe 1995)


Fukiai (Kobe 1995)
Takatori 00, Takatori 90 (Kobe 1995)
Takarazuka 00, Takarazuka 90 (Cobe
1995)
No 4-140, 4-230, 6-140 (Imperial Valley
1979)
TCU068 NS, TCU068 EW, TCU102 NS
(Chi-Chi 1999)
Lucerne 275 (Landers 1992)
Sakarya EW (Kocaeli 1999)
Yarimca 60 (Kocaeli 1999)
Pacoima Dam 164 (Northridge 1994)
Santa Monica 90 (Northridge 1994)
UCLA 360 (Northridge 1994)
Rinaldi 228 (Northridge 1994)
Jensen 22 (Northridge 1994)
Newhall 360 (Northridge 1994)

For every record a series of numerical analyses


is performed and the influence of each
parameter in asymmetric sliding is assessed.
The following trends are noticed in the results of
all time-history analyses (not only those shown
here).

Influence of inclination angle


Figure 18 shows that when = 5o, the block
yields downhill as well as uphill (for ratio ac1/aH
= 0.05, one major uphill sliding occurred at t =
6.1 sec) while for larger values, block slides
exclusively downhill. Nonetheless, the induced
displacements are roughly the same. As a
consequence, few slidings happened uphill
even if the input motion includes multi-cycle
pulses such as in JMA, Fukiai, and Takatori
records.
Contrasting difference observed between
the symmetric ( = 0) and the asymmetric (
0) sliding response, as indicated by Figures 19
and 20. The key point is that in case of
horizontal plane, the block can slide both ways;
thus slippage induced in one direction during a
cycle can be inverted in next ones returning the
block to its initial position.

Influence of ac1/aH ratio


The observations made earlier with the
idealized wavelets, remain qualitatively valid in
case of actual records: as the ratio ac1/aH
increases, sliding displacement decreases.
However, with actual accelerograms slippage
does not always increase in inverse proportion
with ac1/aH (Fig. 13). This is explained by the
fact that, near-fault ground motions include a
large number of pulses. For instance, Figure 14
portrays the asymmetric response triggered by
the JMA record; an excitation of numerous
acceleration cycles. No uphill sliding occurred.
At each cycle a finite slippage is induced,
leading
to
successively
accumulating
displacement. However, almost all the records
include pulses in which sliding takes place
when ac1/aH has small values, but as the ratio
ac1/aH increases no sliding occurs.
Usually, a record is qualified by its peak
ground acceleration (PGA), presuming that the
magnitude of PGA is an important index of
destructiveness. However, this is not the rule for
elastoplastic systems, especially those involving
asymmetric sliding. An impressive evidence of

Influence of excitation direction


As obtained with idealized wavelets,
reversal of excitation direction may cause
significant magnification of sliding displacement.
Similar is the trend with actual accelerograms
(Figs 21-23).
The dominant direction of a record is of
importance because it determines if the
deleterious pulses of the record tend to move

82

the block uphill or downhill. Thus, Figure 21


portrays for two excitation directions (+ and -)
the response of a sliding block with ac1/aH = 0.1
on a plane of = 25, subjected to the Imperial
Valley No4 record. In the right handside of
Figure 21, notice that the distinctive long-period
pulse of T 3 sec which starts at 6.2 sec is
responsible for the slippage of 1.7 m. On the
contrary, when the record is reverted (see lefthandside of figure), this distinctive acceleration
pulse is the one who stops the block downhill
sliding, just at 0.9 m.
In fact, it is evident that all accelerograms
are recorded with a specific direction which
can not be changed. Hence which is the
purpose of applying a record with reverted
direction? Actually the direction of a record is
irreversible, but the direction of an inclined
plane is not. To appreciate the practical
significance of the change in sign of the
excitation, consider the case of the two steep
banks of a river; they have opposite slope
directions and in case of an earthquake they
subjected to the same input motion. This is
equal to the case of two identical slopes with
the same direction, which are excited by
opposite direction records. Even if they were
mathematically identical, their damage during
Chi-chi might be significantly different (see Fig.
22).

For an excitation, features such as the exact


sequence of cycles, the duration and magnitude
of acceleration pulses, and the velocity steps,
are all of major importance as they may play a
decisive role in the induced slippage,
sometimes equal important role as the
dominant frequency and the intensity of motion.
According to their form and details,
accelerograms can be grouped into one or
more of several categories:
Highfrequency records: Santa Monica,
UCLA, Lucerne, Sakarya

Records affected by forward or backward


directivity:
Fukikai,
JMA,
Takatori,
Takarazuka, TCU, Imperial Valley, Newhall,
Rinaldi, Jensen.

Considering the results of the above records


in asymmetric sliding, the following trends are
noticed:
Particularly disastrous are the records which
are characterized by a dominant velocity pulse
not only of long duration and large velocity step,
but mainly of monotonically increasing or
decreasing branch (Fig.24). By the term
monotonically increasing branch we mean that
the ratio of the velocity fluctuations amplitude
divided to the velocity step is small (Fig.24). A
necessity, to succeed this monotonic behavior
of the increasing and decreasing branch is
these variations to rise from distinctive
acceleration pulses; a forward rupture directivity
characteristic.
As can be seen in Figure 25, the velocity
pulse (of 3 sec duration and 1.5 m/s velocity
step) in the Yarimca record results in a slippage
of 4.5 m, generated by two dominant
successive acceleration pulses: the first of them
begins at the moment t 6 sec and finishes at t
7.5 sec, while the second one starts
immediately after the end of the first pulse and
finishes at t 10.5 sec. As a result, the induced
velocity pulse has a monotonically increasing
branch succeeded by a decreasing one.
On the other hand, even though the Sakarya
record includes a substantial velocity pulse of
4.5 sec duration and 0.9 m/s velocity step, it
causes a sliding displacement of only 1 m. This
may be explained by the fact that the velocity
pulse is built up by continuously fluctuating
acceleration spikes, each of which of high
magnitude.

Influence of excitation form

Records with fling-steps: Yarimca 060,


Sakarya, TCU 068 NS & EW, Lucerne

COMPARISON WITH AVAILABLE CHARTS

It is of great practical interest to compare the


results of our study with the relevant charts for
sliding displacement published by Makdisi &
Seed (1978), Sarma & Ambraseys (1967,
1978), and Yegian et al (1988). The
comparisons are displayed in Figures 26 28,
from which the following observations are
made:

Records with a single dominant velocity


pulse: Imperial Valley No 4-230 & 6-230,
Yarimca 060, Sakarya, TCU 068 NS & EW,
Lucerne 275, Rinaldi 228, Pacoima Dam
164.
Records with multiple dominant velocity
pulses: JMA, Takatori, ImperialValley No 4140, TCU 102 NS, Jensen 022.

83

1)

2)

3)

REFERENCES AND BIBLIOGRAPHY

Both in case of idealized wavelets and


actual records, the proposed curves by
Yegian et al (1988) are in very good
agreement with our results.
The curves of Makdisi & Seed (1978), and
Ambraseys & Sarma (1967, 1978)
generally lead to satisfactory results for all
idealized wavelets, but the sinus pulse.
With the latter excitation their charts
would
under-estimate
the
induced
slippage (computed here).
For most of the records (particularly the
strong near fault ground motions such as
TCU 068 NS, TCU 068 EW, TCU 102
NS, Fukiai, JMA, Takatori, and
Takarazuka) the curves of Makdisi &
Seed and Ambraseys & Sarma,
understandably, fail to predict even the
order of magnitude of the true sliding (Fig.
28).

Ambraseys, N.N., & Sarma, S.K. (1967) The


response of earth dams to strong earthquakes,
Gotechnique, Vol.17, pp 181 - 213.
Amirbekian, R.V. & Bolt, B.A. (1998) Spectral
comparison of vertical and horizontal seismic
strong ground motions in alluvial basins,
Earthquake Spectra, EERI, Vol. 14, No 4, pp
573- 595.
Baker, R. & Klein, Y. (2004) An integrated limiting
equilibrium approach for design of reinforced soil
retaining structures: Part I formulation,
Geotextiles and Geomembranes, Vol. 22, pp 151
177.
Baker, R. & Klein, Y. (2004) An integrated limiting
equilibrium approach for design of reinforced soil
retaining structures: Part II design examples,
Geotextiles and Geomembranes, Vol. 22, pp 119
150.
Bathe, K. J., & Mijailovich, S. (1988) Finite element
analysis of frictional contact problems, Journal
of Theoretical and Applied Mechanics, Vol. 7, pp
31 - 44.

CONCLUSIONS

Response in asymmetric sliding is


depending on several factors, which make the
extraction
of
specific
rules
difficult.
Nevertheless, the following general trends were
brought to light:
Peak ground acceleration does not have so
great importance in asymmetric sliding, as
do the frequency and direction of excitation.
The details, the sequence of pulses and the
form of the input motion have a dominant
influence on sliding response because of
the non-linearity of the problem.
On the basis of induced slippage, the most
devastating ground motions are these which
have a strong pulse character both in
acceleration and velocity. Such are the
records affected by forward directivity
effects.
Finally, comparison between the numerical
results of the present paper and literature,
results to the conclusion that curves of
Yegian et al describe even the extreme
reality accurately enough. On the other
hand, the proposed curves of Makdisi,
Seed,
Ambraseys
and
Sarma
underestimate, sometimes dramatically, the
slippage caused by near-fault ground
motions.

Bolt, B.A. (2004) Earthquakes, W.H. Freeman and


Company, Fifth edition, 41 Madison Avenue,
New York, NY 10010.
Bozorgnia, Y. & Bertero, V.V. (2004) Earthquake
Engineering: From Engineering Seismology to
Performance - Based Engineering, CRC Press,
2004.
Cai, Z. & Bathurst, R.J. (1996) Seismicinduced
permanent displacement of geosyntheticreinforced segmental retaining walls, Canadian
Geotechnical Journal, Vol. 33, pp 937 955.
Constantinou, M.C., & Gazetas, G. (1984)
Probabilistic seismic sliding deformations of
earth dams and slopes, Proceedings of
Specialty Conference on Probabilistic Methods
and Structural Reliability, ASCE, New York, pp
318 - 321.
Duncan, J.M. (1996) State of the art: Limit
equilibrium and finite element analysis of slopes,
Journal of Geotechnical Engineering, Vol. 122,
pp 577 596.
Elgamal, A.W.M., Scott, R.F., Succarieh, M.F., &
Yan, L. (1990) La Villita dam response during
five
earthquakes
including
permanent
deformation,
Journal
of
Geotechnical
Engineering, Vol. 116, pp 1443 - 1462.
Elms, D.G. (2000) Refinements to the Newmark
sliding block model, Proceedings of 12th World
Conference in Earthquake Engineering, pp 2132.

84

Gazetas, G. & Uddin, N. (1994) Permanent


deformation on pre-existing sliding surfaces in
dams, Journal of Geotechnical Engineering,
Vol.120, No.11, pp 2041 2061.

Michalowski, R. & You, L. (2000) Displacements of


reinforced slopes subjected to seismic loads,
Journal of Geotechnical and Geoenvironmental
Engineering, Vol. 126, No 8.

Junwu, D., Tong, M., Lee, G.C., Xiaorhai, Q. &


Wenting, B. (2004) Dynamic responses under
the excitation of pulse sequences, Earthquake
Engineering and Engineering Vibration, IEM and
MCEER, Vol. 3, No 2, pp 157 - 169.

Nadim. F., & Whitman, R.V. (1983) Seismically


induced movements of retaining walls, Journal
of Geotechnical Engineering Division, Vol. 109(7),
pp 915 - 931.
Newmark, N.M. (1965) Effect of earthquakes on
dams and embankments, Gotechnique, Vol. 15,
pp 139 -160.

Kramer, S. L. & Paulsen, S. (2002) A numerical


model for estimating seismic displacements of
reinforced steep slopes, M.S thesis, University
of Washington.

Richards, R., Elms, D.G. & Budhu, M. (1982)


Seismic bearing capacity and settlement of
foundations,
Journal
of
Geotechnical
Engineering Division.

Kramer, S. L. & Smith, M. (1997) Modified Newmark


model for seismic displacements of compliant
slopes,
Journal
of
Geotechnical
and
Geoenvironmental Engineering, Vol. 123, pp 635
644.

Richards, R. & Elms, D.G. (1979) Seismic


behaviour of gravity retaining walls, Journal of
Geotechnical Engineering Division.

Kramer, S. L. (1996) Geotechnical Earthquake


Engineering, Prentice Hall, Upper Saddle River,
NJ 07458.

Sarma, S.K. (1975) Seismic stability of earth dams


and embankments, Gotechnique, Vol. 25, No 4,
pp 743 - 761.

Lin, J.S., & Whitman, R.V. (1986) Earthquake


induced displacements of sliding block, Journal
of Geotechnical Engineering, Vol. 105, pp 1427 1434.

Seed, H.B., & Martin, G.R. (1966) The seismic


coefficient in earth dam design, Journal of the
Soil Mechanics and Foundation Division, ASCE,
Vol. 92, pp 25 - 58.

Lin, J.S. (1982) Probabilistic evaluation of the


seismically - induced permanent displacements
in earth dams, Research Report No R82-21,
Department of Civil Engineering, MIT.

Succarieh, M. (1990) Analysis of earthquake


induced large displacements in earth structures,
PhD thesis, Rensselaer Polytechnic Institute,
Troy, N.Y.

Ling, H. (2001) Recent applications of sliding block


theory to geotechnical design, Soil Dynamics
and Earthquake Engineering, Vol. 21, pp 189
197.

Wartman, J., Bray, J.D. & Seed, R.B. (2003)


Inclined plane studies of the Newmark sliding
block procedure, Journal of Geotechnical and
Geoenvironmental Engineering, Vol. 129, pp 673
684.

Ling, H., Leshchinsky, D. & Chou, N. (2001) Post


earthquake
investigation
on
several
geosynthetic- reinforced soil retaining walls and
slopes during the Ji-Ji earthquake in Taiwan,
Soil Dynamics and Earthquake engineering, Vol.
21, pp 297 313.

Westermo, B., & Udwadia, F. (1983) Periodic


response of a sliding oscillator system to
harmonic excitation, Earthquake Engineering
and Structural Dynamics, Vol. 11, pp 135 -146.
Yegian, M.K., Marciano E.A., & Ghahraman, V.G.
(1991)
Earthquake
induced
permanent
deformations: a probabilistic approach, Journal
of Geotechnical Engineering, Vol. 117, pp 35 50.

Makdisi, F.I. & Seed, H.B. (1978) Simplified


procedure for estimating dam and embankment
earthquake induced deformations, Journal of
Geotechnical Engineering Division, ASCE, Vol.
104, pp 849 867.

Yegian, M.K., Marciano E.A., & Ghahraman, V.G.


(1988) Integrated seismic risk analysis for earth
dams, Report No CE-88-15.

Makris, N. (1989) Analysis of harmonically excited


sliding systems, M.S thesis, State University of
New York, Buffalo, N.Y.

85

Fk

d(t)

d(t)

a(t)

a H(t)

(a)
Fig 1:

(b)

Schematic representation of Newmarks model used in the parametric study. (a) A


rigid block resting on an inclined sliding interface governed by Coulombs friction
law, (b) forces acting on a rigid block that tends to slide downwards when subjected
to upward excitation.

1.5

1.2
=0
=5

= 25

:m

0.9

0.6

0.3

0
0

0.1

0.2

0.3

0.4

0.5

0.6

ac1 / aH
Fig 2:

Sliding displacement according to acceleration ratio ac1 / aH in case of a Ricker


excitation with frequency fH = 1 Hz and peak acceleration aH = 10 m/s2. Notice the
difference between symmetric ( = 0o) and asymmetric sliding ( 0o).

86

1 .5

25

o
1.5
f = 1fH=
z

1 .2

f = 1fH=
z

1 Hz

f = 2f H=z 2

Hz

f = 4f H=z 4

Hz

H
H

0 .9

: m

1.2

0 .6

0 .3

0 .3

Hz

f = 4f H=z 4

Hz

H
H

0 .9

0 .6

1 Hz

f = 2f H=z 2

0
0

0 .1

0 .2

0 .3

0 .4

0 .5

0 .6

0 .1

0 .2

0 .3

a c1 / a

Fig 3:

0 .4

0 .5

0 .6

a c1 / a

Asymmetric sliding spectra illustrating the influence of critical acceleration ratio ac1 /
aH , for different excitation frequencies of a Ricker wavelet with maximum
acceleration aH = 10 m/s2. It is clear that by increasing the inclination of the plane the
slippage is also increasing.

Ground

Sliding block

o
12

-6

-6

acceleration : m / s

12

-1 2

-1 2

velocity : m / s

-0 .6

-0 .6

-1 .2

-1 .2

-1 .8

-1 .8

-2 .4

-2 .4

5
0 .8

0 .8

0 .4
d(t) : m

0 .4

0.76 m

0.40 m

-0 .4

-0 .4
0

t : s

Fig 4:

t : s

Effect of acceleration ratio to sliding response induced by a sinusoidal pulse of fH =


1.5 Hz. The acceleration ratio is ac1/ a H = 0.05 for the plots in the left handside and
0.1 for those in the right. The shaded areas in the velocity time histories represent the
downward (dotted) or upward (striped) sliding.

87

Ground
Sliding block

acceleration : m / s 2

Frequency: fH

= 1 Hz

Frequency:

12

12

velocity : m / s

0.5

1.5

2.5

3.5

4.5

0.5

2.5

2.5

1.5

1.5

0.5

0.5

1.5

2.5

3.5

1.5

2.5

3.5

4.5

0
0.5

1.5

2.5

3.5

0.5

4.5

4.5

1.86 m

1.5
d(t) : m

fH = 2 Hz

1.5

0.5

0.5

0.47 m

0.5

1.5

2.5

3.5

4.5

0.5

t:s
Fig 5:

1.5

2.5

3.5

4.5

t:s

Acceleration, velocity and displacement time-histories induced by a T-Ricker pulse of


two different frequencies (inclination angle = 5o and ac1/aH = 0.1 ). The single
dominant slip (dotted area) is the outcome of the one-sided acceleration pulse which
consequently generates a monotonically increasing velocity time history.

88

F re q u e n c y: f H = 1 .5 H z

F re q u e n c y: f H = 3 H z

12

12

-6

-6

acceleration : m / s

G ro u n d

-1 2

-1 2

velocity : m / s

S lid in g b lo c k

-0 .8

-0 .8

-1 .6

-1 .6

-2 .4

-2 .4
0

d(t) : m

4.25 m
4

1.10 m

t:s

Fig 6:

t:s

Acceleration, velocity and displacement time-histories for a sinusoidal excitation of


one cycle (inclination angle = 25o and acp1/aH = 0.05 ). The response reach the 4.25
m when the frequency of the pulse is fH = 1.5 Hz (left) whereas for the higher
frequency of 3 Hz (right) slippage falls to 1.10 m.

1g

25o
4.5

10

fH = 2 Hz

3.6

aH / VH2

fH = 1 Hz

: m

fH = 2 Hz

2.7

1.8

0.9

fH = 1 Hz

0
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Fig 7:

0.1

0.2

0.3

0.4

0.5

0.6

0.7

ac1 / a

ac1 / a

Dimensional (left) and normalized (right) sliding spectra for a T-Ricker wavelet in
case of a 25o inclined plane.

89

1g

1g

25

1.5

3.2

fH = 0.67 Hz

fH = 0.67 Hz
1.2

fH = 1.5 Hz

2.4

aH / VH2

1.6

0.8

fH = 3 Hz
0.9

0.6

0.3

0
0

0.1

0.2

0.3

0.4

0.5

0.6

0.1

0.2

ac1 / a

Fig 8:

= 5

0.5

0.6

= 25

10

acceleration : m / s

0.4

G ro u n d
5

G ro u n d
5

S lid in g b lo c k

-5

-5

-1 0

S lid in g b lo c k

-1 0
0

velocity : m / s

0.3

ac1 / a

Normalization effect of sliding spectra in case of a sinusoidal wavelet of one cycle.


Observe that dimensional slippage (left figure) is varying for every frequency fH,
however the normalized response (right plot) converge to a single line.

10

0 .5

0 .5

-0 .5

- 0 .5

-1

-1

-1 .5

- 1 .5
0

d(t) : m

: m

fH = 3 Hz

fH = 1.5 Hz

1 .5

1 .5

1 .2

1 .2

0 .9

0 .9

0 .6

1.47 m

0 .6

0.35 m

0 .3

0 .3
0

- 0 .3

- 0 .3
0

t : s

t : s

Fig 10: Effect of slope inclination to sliding response induced by a Tsang wavelet with fH = 1
Hz. The acceleration ratio is ac1/ a H = 0.05. The shaded areas in the velocity time
histories represent the downward (dotted) or upward (striped) sliding.

90

acceleration : m / s

Frequency: fH

= 1 Hz

Frequency:

10

10

-5

-5

Ground
Sliding block

-10

-10
0

velocity : m / s

fH = 2 Hz

1.2

1.2

0.6

0.6

-0.6

-0.6

-1.2

-1.2

0.4

0.4

0.38 m
0.2

0.10 m

d(t) : m

0.2

-0.2

-0.2

t:s

t:s

Fig 11: Response time-histories for Ricker excitation of frequency fH = 1 Hz (left) and fH = 2
Hz (right) in case of a plane with inclination = 5o and ac1/ a H = 0.1. The pulse is
imposed inverted to study sliding with respect to triggering input direction.

91

Ground
Sliding block

acceleration : m / s 2

25o
12

12

-6

-6
-12

-12

velocity : m / s

2.4

2.4

1.2

1.2

-1.2

-1.2

-2.4

-2.4
0

d(t) : m

4.5

4.5

3.6

3.6

2.7

2.7

1.8

4.25 m

1.8

0.62 m

0.9

0.9

0
0

t:s

t:s

Fig 12: By reversing the polarity of an one cycle sinusoidal pulse (fH = 1.5 Hz) the resulting
slippage, in case of an inclination angle = 25o and ac1/aH = 0.05, is increased seven
times (from 0.62 m to 4.26 m).

92

0.81 g
Ground
Sliding block
o

25

0.5

2.71 m

2
d(t) : m

velocity : m / s

1
-0.5
0

-1
0

10

12

14

10

12

14

0.5
0

0.85 m

1
-0.5
0

-1
0

10

12

14

10

12

14

0.5
2
d(t) : m

velocity : m / s

d(t) : m

velocity : m / s

-0.5

0.27 m
0

-1
0

10

12

14

t:s

10

12

t:s

Fig 13: Influence of acceleration ratio ac1/ a H on asymmetric sliding for the inverted JMAi
record (NS component and inclination angle = 25o). The two plots of the first top
line correspond to a ratio ac1/ a H = 0.05, whereas the plots of the middle and bottom
line correspond to critical acceleration ratios of 0.2 and 0.4, respectively.

93

14

Ground
Sliding block

25
6

acceleration : m / s

3
0
-3
-6
-9

10

12

14

10

12

14

velocity : m / s

0 .5
0
-0 .5
-1

d(t) : m

2.58 m

10

12

14

Tim e : sec

Fig 14:

Acceleration and velocity time-histories of a rigid block (black colored lines) and the input base
excitation (grey colored lines), as well as slippage time-history of the sliding block induced by
the JMA ground motion (NS component); in case of a slope with inclination angle =25o and
critical acceleration ratio ac1/aH = 0.05. The slippage of 2.58 m at the end of the motion is the
result of all the previously accumulated slides.

acceleration : m / s

4
2
0
-2

0.34 g

-4
0

12

15

18

21

12

15

18

21

velocity : m / s

2.97 m/s
2

-2
0

d(t) : m

12

8.81 m

12

15

18

21

T im e : s e c

Fig 15:

Asymmetric response time-histories for the TCU 068 NS ground motion ( = 25o and ac1/ a H =
0.05 ). The 8.8 m of total slippage is induced by a large sequence of remarkably long duration
acceleration pulses. Observe the outstanding ground velocity pulse of almost 6 sec duration
and 3 m/s magnitude.

94

Ground
Sliding block

25
4

-2

-2

-2

-4

-4

acceleration : m / s

velocity : m / s

12

18

24

12

18

24

-2

-2
0

d(t) : m

-4
0

12

18

20

20

15

15

8.8

12

18

24

13.6

10

12

18

24

12

18

24

16.0

0
0

24

18

15

10

12

20

10

-2
0

24

12

t : s

18

24

12

t : s

18

24

t : s

Fig 16: Simple approximation of the TCU 068- NS record through a series of rectangular
pulses (inclination angle = 25o and ac1/aH = 0.05 ). The left column plots show the
response to the detailed ground motion, while the middle and the right column display
both the simplified acceleration excitations that were used. Even though the first
major slippage of nearly 6.5 m (dotted line) is approximated by the simplified motions,
the cut off of the details leads to substantial increase in the final result.
4

-2

-2

-2

-4

-4

acceleration : m / s

velocity : m / s

18

27

-4
0

18

27

-2

-2

-2

-4
9

18

27

30

18

27

18

27

-4

-4

18

27
30

30

d(t) : m

23.3
20

25.3

10

27.1

20

10

10

0
0

18

t : s

27

20

18

t : s

27

18

t : s

Fig 17: Simple simulation of the inverted TCU 068-NS record through a series of rectangular
pulses (inclination angle = 25o and ac1/aH = 0.05 ). The final displacements in all
three cases (the actual record and the approximations) are in reasonable agreement.

95

27

Ground
Sliding block

acceleration : m / s 2

5o
10

10

-5

-5

velocity : m / s

-10

-10
2

10

12

14

1.5

1.5

0.5

0.5

-0.5

-0.5

-1

-1

10

12

14

10

12

14

-1.5

-1.5
0

d(t) : m

25

10

12

14

0.8

0.8

0.6

0.6

0.4

0.4

0.53 m

0.2

0.78 m

0.2
0

-0.2

-0.2

10

12

14

t:s

10

12

14

t:s

Fig 18: Effect of slope inclination to asymmetric sliding for the Fukiai ground motion. The
plots correspond to a ratio ac1/ a H = 0.2. In case of = 5o, the small inclination allows
an uphill sliding of the block (presented between the dotted lines in left handside
figures).

96

0.85 g

25

3.2

3.2

ac1/ a H = 0.05

d(t) : m

2.4
1.6

1.6

0.59

0.8

0.8

0
-0.8

-0.8
0

12

16

3.2

d(t) : m

12

16

3.2

ac1/ a H = 0.2

2.4

ac1/ a H = 0.2

2.4

1.6

1.6
0.31

0.8

0.78

0.8

-0.8

-0.8
0

12

16

3.2

12

16

3.2

ac1/ a H = 0.4

2.4
d(t) : m

3.13

2.4

ac1/ a H = 0.4

2.4

1.6

1.6
0.10

0.8
0

-0.8

-0.8
0

12

0.16

0.8

16

t:s

t:s

Fig 19: Influence of the symmetric (left) or asymmetric (right) nature of sliding to the
response induced by the Fukiai ground motion.

97

12

16

25

0.21 g

ac1/ a H = 0.05

d(t) : m

4
3
2

0.79

-0.1

-1

-1
0

12

16

20

12

16

20

ac1/ a H = 0.1

ac1/ a H = 0.1

4
d(t) : m

4.31

2.73

3
2

0.82

0.18

-1

-1
0

12

16

20

12

16

20

ac1/ a H = 0.2

4
d(t) : m

ac1/ a H = 0.05

ac1/ a H = 0.2

2
1

1.00

0.50

0.02

-1

-1
0

12

16

20

t:s

12

16

t:s

Fig 20: Symmetric (left) and asymmetric (right) sliding response to the Yarimca-060 record.
Observe the substantial larger slippage in case of the inclined plane, particularly for
small acceleration ratio ac1/ a H

98

20

Ground

acceleration : m / s 2

25

Sliding block

-2

-2

-4

-4

velocity : m / s

10

13

16

0.8

0.8

0.4

0.4

-0.4

-0.4

10

13

16

10

13

16

-0.8

-0.8
1

d(t) : m

10

13

16

1.5

1.5

0.93 m

1
0.5

1.70 m

1
0.5

0
1

10

13

16

t:s

10

13

t:s

Fig 21: Effect of excitation polarity on asymmetric sliding for the Imperial Valley No4 record
(230o component, inclination angle = 25o, ratio ac1/ a H = 0.1). Observe that by
reverting the triggering the subsequent two downward slides (dotted areas in left) are
replaced by a single major slide (dotted area in right).

99

16

acceleration : m / s 2

0.34 g
2

-2

-2

0.34 g

-4
0

10

-4
15

20

25

27

10

15

d(t) : m

18

25.30

18

8.81
9

0
0

10

15

20

25

10

15

20

25

27

27

ac1/ a H = 0.1
d(t) : m

25

27

ac1/ a H = 0.05

ac1/ a H = 0.1
18

18

5.46

12.90

0
0

10

15

20

25

27

10

15

20

25

27

ac1/ a H = 0.2
d(t) : m

20

ac1/ a H = 0.2

18

18

5.73

1.70
0

0
0

10

15

20

25

t:s

10

15

20

25

t:s

Fig 22: Sliding response of the TCU 068-NS record triggering a 25o inclined plane with the
peak ground acceleration of 0.34 g acting downwards (left handside) and upwards
(right handside).

100

Ground
Sliding block

acceleration : m / s 2

25
10

10

-5

-5

-10

-10

velocity : m / s

10

-1

-1

-2

10

10

-2
0

10
2.2

2.2

d(t) : m

2.00
1.04

1.1

1.1

0
0

10

t:s

t:s

Fig 23: Acceleration, velocity and displacement time-histories for the Rinaldi record (228o
component) when imposed parallel to the sliding interface (inclination angle = 25o
and ac1/aH = 0.1 ). Notice the asymmetric response of the block when the excitation is
inverted (plots in right); the well-shaped forward directivity pulse, shown between the
dotted lines, causes a major slippage of 2 m.

101

10

Ground
Sliding block

25

acceleration : m / s 2

5
2.5
0
-2.5
-5
0

12

15

18

21

12

15

18

21

1.2

velocity : m / s

Velocity Step
0.6

Fluctuation

0
-0.6
-1.2
0

d(t) : m

4
3
2

4.55

1
0
0

12

15

18

Time : sec

Fig 24: Acceleration and velocity time-histories of the rigid block and the input base
excitation, as well as slippage time-history of the sliding block triggered by the Jensen
record (22o component) for a slope of inclination =25o and critical acceleration ratio
ac1/aH = 0.05. The slippage of4.55 m is the outcome of three major sliding episodes
(dotted areas) induced by three subsequent velocity pulses resulting from the forward
directivity acceleration pulses (starting at 4 sec and ending at 12 sec).

102

21

Ground
Sliding block

25

-2

-2

acceleration : m / s

-4

-4

velocity : m / s

12

15

0.5

0.5

12

15

18

21

12

15

18

21

-0.5

-0.5
0

d(t) : m

12

15

4.31

1.01

1
0

0
0

12

15

t:s

Fig 26:

12

15

18

21

t:s

Acceleration, velocity and displacement time-histories for the Sakarya record (left)
and the Yarimca-060 component (right) of a block on an inclined plane with = 25o
and acceleration ratio acp1/aH = 0.05.

/ (aH TH )

0 .1

0 .0 1
Y e g ia n e t a l ( 1 9 8 8 )
S in u s o id a l

Sarma

tr ia n g u la r

(1976)

R ic k e r

0 .0 0 1

T - R ic k e r
s insinusoid
us
Tsang
0 .0 0 0 1
0

0 .1

0 .2

0 .3

0 .4

0 .5

0 .6

0 .7

0 .8

0 .9

a c1 / a

Fig 26:

Comparison of normalized slippage induced by idealized wavelets with the published


results of Yegian et al (1988) & Sarma (1979).

103

0.1

0.1

0.01

0.01

/ (aH TH )

JM A

TC U 068 - N S
0.001

Takatori

0.001

TC U 068 - E W

Takarazuka
Fukiai

TC U 102 - N S
0.0001

0.0001
0

0.1

0.2

0.3

0.4

0.5

0.6

0.1

0.2

0.3

0.4

0.5

0.6

a c1 / a H

/ (aH TH )

0.1

25

0.01

IV 4-140
0.001

IV 4-230

Y egian et al (1988)

IV 6-230

S inusoidal
Triangular

0.0001
0

0.1

0.2

0.3

0.4

0.5

Sarma
(1976)

0.6

a c1 / a H

: cm

Fig 27:

Comparison with the published results of Yegian et al (1988) & Sarma (1979) of the
normalized slippage induced by: Chi-chi records (top-left plot), Kobe records (topright), and Imperial Valley records (bottom-left). Each excitation imposed normally
and inverted.

1000

1000

100

100

10

10

y
JM A
Fukiai

IV 4-140
IV 4-230
IV 6-230

Takatori
Takarazuka
0.1

0.1
0

0.1

0.2

0.3

0.4

0.5

0.1

0.2

0.3

0.4

a c1 / a H
10000

: cm

1000

100

25

10

M = 6.5

T C U 068 - N S

T C U 068 - E W

M = 7.5

T C U 102 - N S
0.1
0

0.1

0.2

0.3

0.4

0.5

Makdisi & Seed


(1976)

Sarm a & Am braseys

a c1 / a H

Fig 28:

Comparison with the published results of Makdisi & Seed(1976) of the sliding
response induced by: Chi-chi records (top-left plot), Kobe records (top-right), and
Imperial Valley records (bottom-left). Each excitation imposed normally and inverted.

104

0.5

A Replacement to the MononobeOkabe Equations


by Stress Limit Analysis
G. Mylonakis, P. Kloukinas, C. Papantonopoulos
University of Patras, Greece

Abstract
A closed-form stress plasticity solution is presented for gravitational and earthquakeinduced earth pressures on retaining walls. The proposed solution is essentially an
approximate yield-line approach, based on the theory of discontinuous stress fields, and
takes into account the following parameters: (1) weight and friction angle of the soil material,
(2) wall inclination, (3) backfill inclination, (4) wall roughness, (5) surcharge at soil surface,
and (6) horizontal and vertical seismic acceleration. Both active and passive conditions are
considered by means of different inclinations of the stress characteristics in the backfill.
Results are presented in the form of dimensionless graphs and charts that elucidate the
salient features of the problem. Comparisons with established numerical solutions, such as
those of Chen and Sokolovskii, show satisfactory agreement (maximum error for active
pressures about 10%). It is shown that the solution does not perfectly satisfy equilibrium at
certain points in the medium, and hence cannot be classified in the context of limit analysis
theorems. Nevertheless, extensive comparisons with rigorous numerical results indicate that
the solution consistently overestimates active pressures and under-predicts the passive.
Accordingly, it can be viewed as an approximate lower-bound solution, than a mere
predictor of soil thrust. Compared to the Coulomb and Mononobe-Okabe equations, the
proposed solution is simpler, more accurate especially for passive pressures) and safe, as it
overestimates active pressures and underestimates the passive. Contrary to the
aforementioned solutions, the proposed solution is symmetric, as it can be expressed by a
single equation - describing both active and passive pressures - using appropriate signs for
friction angle and wall roughness.

INTRODUCTION

and need not be repeated herein [Kramer,


1996; Wood, 1973; Steedman & Zeng, 1990;
Veletsos & Younan, 1994; Theodorakopoulos et
al., 2001). Given their practical nature and
reasonable predictions of actual dynamic
pressures (e.g. Ebeling et al., 1992; Sheriff et.
al., 1982; Veletsos & Younan, 1994;
Theodorakopoulos et al., 2001), solutions of
this type are expected to continue being used
by engineers for a long time to come. This
expectation does not seem to diminish by the
advent
of
displacement-based
design
approaches, as the limit thrusts provided by the
classical methods can be used to predict the
threshold (yield) acceleration beyond which
permanent dynamic displacements start to
accumulate [Kramer, 1996; Pecker, 1995;
Richards & Elms, 1985; Whitman & Liao, 1985;

The classical equations of Coulomb [Coulomb,


1776; Heyman, 1972; Lambe, 1969; Clough &
Duncan, 1990; Chen, 1975) and MononobeOkabe (Okabe, 1924; Mononobe & Matsuo,
1929; Matsuo & Ohara, 1960; Seed & Whitman,
1970; Chen, 1975; Ebeling et al., 1992; Kramer
1996) are being widely used for determining
earth pressures due to gravitational and
earthquake loads, respectively. The MononobeOkabe solution treats earthquake loads as
pseudo-dynamic,
generated
by
uniform
acceleration in the backfill. The retained soil is
considered a perfectly plastic material, which
fails along a planar surface, thereby exerting a
limit thrust on the wall. The theoretical
limitations of such an approach are well known

105

can be criticized on the following important


aspects: (1) in the context of limit analysis, their
predictions are unsafe; (2) their accuracy (and
safety) diminishes in the case of passive
pressures on rough walls, (3) the mathematical
expressions are complicated and difficult to
verify 1 , (4) the distribution of tractions on the
wall are not predicted (typically assumed linear
with depth following Rankines solution), (5)
optimization of the failure mechanism is
required in the presence of multiple loads, to
determine a stationary (optimum) value of soil
thrust, and (6) in the context of limit-equilibrium
analysis, stress boundary conditions are not
satisfied, as the yield surface does not
generally emerge at the soil surface at the
required angles of 45o /2.
In light of the above arguments, it appears
that the development of a closed-form solution
of the stress type for assessing seismicallyinduced earth pressures would be desirable. It
will be shown that the proposed solution,
although approximate, is mathematically
simpler than the existing kinematic solutions,
offers
satisfactory
accuracy
(maximum
deviation for active pressures against rigorous
numerical solutions less than 10%), yields
results on the safe side, satisfies stress
boundary conditions, and predicts the point of
application of soil thrust. Last but least, the
solution is symmetric that is, it active and
passive pressures can be described by a single
formula, using appropriate signs for the friction
angle and wall roughness. Apart from its
intrinsic theoretical interest, the proposed
analysis can be used for the assessment and
improvement of other related methods.

Psarropoulos et. al., 2005).


Owing to the translational and staticallydetermined failure mechanisms employed, the
limit-equilibrium Mononobe-Okabe solutions
can be interpreted as kinematic solutions of limit
analysis (Collins, 1973). The latter solutions are
based on kinematically admissible failure
mechanisms in conjunction with a yield criterion
and a flow rule for the soil material, both of
which are enforced along pre-specified failure
surfaces (Chen, 1975; Pecker, 1995; Caquot &
Kerisel, 1948 ; Finn, 1967; Salencon, 1974).
Stresses outside the failure surfaces are not
examined and, thereby, equilibrium in the
medium is generally not satisfied. In the realm
of associative and convex materials, solutions
of this type are inherently unsafe that is, they
underestimate
active
pressures
and
overestimate the passive (Chen, 1975; Finn,
1967; Davis & Selvadurai, 2002; Salencon,
1974).
A second group of limit-analysis methods,
the stress solutions, make use of pertinent
stress fields that satisfy the equilibrium
equations and the stress boundary conditions,
without violating the failure criterion anywhere in
the medium (Davis & Selvadurai, 2002;
Atkinson, 1981; Parry, 1995). On the other
hand, the kinematics of the problem is not
examined and, therefore, compatibility of
deformations is generally not satisfied. For
convex materials, formulations of this type are
inherently safe that is, they overestimate active
pressures and underestimate the passive
(Chen, 1975; Davis & Selvadurai, 2002;
Atkinson, 1981). The best known such solution
is that of Rankine, the applicability of which is
severely restricted by the assumptions of
horizontal backfill, vertical wall and smooth soilwall interface. In addition, the solution may be
applied only if the surface surcharge is uniform
or non-existing. Owing to difficulties in deriving
pertinent stress fields for general geometries,
the vast majority of limit-analysis solutions in
geotechnical design are of the kinematic type
(Seed & Whitman, 1970; Ebeling et al., 1992;
Chen, 1975; Kramer 1996, Atkinson, 1981]. To
the best of the authors knowledge, no simple
closed-form solution of the stress type has been
derived for seismic earth pressures.
Notwithstanding the theoretical significance
and practical appeal of the Coulomb and
Mononobe-Okabe solutions, these formulations

METHODOLOGY

The problem under investigation is depicted


in Figure 1: a slope of dry cohesionless soil
retained by an inclined gravity wall, is subjected
to plane strain conditions under the combined
action of gravity (g) and seismic body forces (ah
x g) and (av x g) in the horizontal and vertical
direction, respectively. The problem parameters
are: height (H) and inclination () of the wall,
1

The story of a typographical error in the MononobeOkabe formula that appeared in a seminal article of the
early 1970s and subsequently propagated in a large
portion of the literature, is indicative of the difficulty in
checking the mathematics of these expressions (Davies et
al [41]).

106

inclination () of the backfill; roughness () of


the wall-soil interface; friction angle () and unit
weight () of the soil material, and surface
surcharge (q). Since backfills typically consist of
granular materials, cohesion in the soil and
cohesion at the soil-wall interface are not
studied here.

tan e =

inclined
backfill

+ e
cohesionless soil
( )

backfill, under zero rotation. Both assumptions


have important implications in the distribution of
earth pressures on the wall, as explained
below.
The resultant body force in the soil is acting
under an angle e from vertical towards the
backfill,

+a
+a

inclined wall,
roughness ()

Fig 1: The problem under consideration

ah
1 av

(1)

which minimizes passive resistance. In


accordance with the rest of the literature,
positive av is upward (downward ground
acceleration). However, its influence on earth
pressures, although included in the analysis, is
not studied numerically here, as it is usually
minor and often neglected in design (Ebeling et
al., 1992; Whitman & Liao, 1985).
In the absence of surcharge, the MononobeOkabe solution to the above problem is given
by the well-known formula (Kramer, 1996):

In addition, since the vibrational characteristics


of the soil are neglected, the seismic force is
assumed to be uniform in the backfill. Also, the
wall can translate away from, or towards to, the

2P
K E = E2 =
H

cos2 ( e + )

sin ( + ) sin ( e + )

cos e cos2 cos ( e + ) 1


cos ( e + ) cos ( )

e <

where PE denotes the limit seismic thrust on the


wall (units = F/L) and KE is the corresponding
earth pressure coefficient. In the above
representation (and hereafter), the upper sign
refers to active conditions (PE = PAE, KE = KAE),
and the lower sign to passive (PE = PPE, K =
KPE).
A drawback of the above equation lies in the
difficulty in interpreting the physical meaning especially signs - of the various terms (Davis &
Selvadurai, 2002, footnote in p4). As will be
shown below, the proposed solution is free of
this problem.
To prevent slope failure when inertial action
is pointing towards the wall, the seismic angle
e should not exceed the difference between
the friction angle and the slope inclination.
Therefore, the following constraint applies
(Ebeling et al., 1992):

(2)

(3)

A similar relation can be written for the case


where inertial action is pointing towards the
backfill, but it is of limited practical interest and
will not be discussed here.
To analyze the problem, the backfill is
divided into two main regions subjected to
different stress fields, as shown in Figure 2: the
first region (zone A) is located close to the soil
surface, whereas the second (zone B) close to
the wall. In both regions the soil is assumed to
be in a condition of impeding yielding under the
combined action of gravity and earthquake
body forces. The same assumption is adopted
for the soil-wall interface, which, however, is
subjected exclusively to contact stresses. A
transition zone between regions A and B is
introduced later on.
Fundamental to the proposed analysis is the
assumption that stresses close to the soil

107

which are valid for static conditions (h = v = 0)


and satisfy the stress boundary conditions at
the surface. Equations (4) suggest that the ratio
of shear to normal stresses is constant (tan) at
all depths, and that points at the same depth
are subjected to equal stresses. Note that due
to static determinacy and anti-symmetry, the
above relations are independent of material
properties and asymptotically exact at large
distances from the wall.
Considering the material to be in a condition
of impeding yielding, the Mohr circle of stresses
in region A is depicted in Figure 3. The different
locations of the stress point (, ) for active
and passive conditions and the different
inclinations of the major principal plane
(indicated by heavy lines) are apparent in the
graph.

surface can be well approximated by those in


an infinite slope, as shown in Figure 2. In this
region (A), the inclined soil element shown is
subjected to canceling actions along its vertical
sides. Thus equilibrium is achieved solely under
body forces and contact stress acting at its
bottom face.
Based
on
this
physically-motivated
hypothesis, the stresses and at the base
of the inclined element are determined from the
following expressions:

= z +
cos 2
cos

(4a)

= z +
sin cos
cos

(4b)

ZONE A

( , )

( , )

passive
case

unit length

active
case

rface
soil su

q
soil
surface

1
1

1A

1 +
SA

active

passive

ZONE B

(w , w )

ZONE A

passive

wall
plane

(w , w )

2+

passive

passive

1B

SB

active

(w , w )
active

wall
plane

(w , w )

Fig 3: Mohr circles of effective stresses and


inclination of the major principal planes in
zones A and B under gravitational loading.

active

ZONE B

From the geometry of Figure 3, the normal


stress is related to mean stress SA through
the proportionality relation

wall length
L = H / cos

Fig 2: Stress fields close to soil surface (Zone A)


and the wall (Zone B)

108

= S A [1 sin cos( 1 )]

transition in the orientation of principal planes in


the two zones, a logarithmic stress fan 2 is
adopted in this study, centered at the top of the
wall. In the interior of the fan, principal stresses
are gradually rotated by the angle separating
the major principal planes in the two regions, as
shown in Figure 4.

(5)

where 1 denotes the Caquot angle [23, 28]


given by

sin 1 =

sin
sin

(6)

For points in region B, it is assumed that


stresses are functions exclusively of the vertical
coordinate and obey the strength criterion of the
frictional soil-wall interface, as shown in Figure
2. Accordingly, at orientations inclined at an
angle from vertical,

w = w tan

AB

zone B

(7)

ACTIVE CONDITIONS

1
2
2
2 +
2

PASSIVE CONDITIONS

AB
z

(8)

zone B

where 2 is the corresponding Caquot angle


given by,

sin
sin 2 =
sin

zone A

2
2
2

where w and w are the normal and shear


tractions on the wall, at depth z. The above
equation is asymptotically exact for points in the
vicinity of the wall. The corresponding Mohr
circle of stresses is depicted in Figure 3. The
different signs of shear tractions for active and
passive conditions follow the directions shown
in Fig 2 (passive wall tractions pointing upward,
active tractions pointing downward), which
comply with the kinematics of the problem. This
is in contrast with the widespread view that
solutions based on equilibrium totally ignore the
displacement field (Papantonopoulos & Ladanyi
1973). From the geometry of Figure 3, the
normal traction w is related to the mean stress
SB through the expression

w = S B 1 sin cos ( 2 )

1 +
2

zone A

Fig 4: Rotation of major principal planes between


zones A and B for active and passive
conditions

(9)

In light of the foregoing, it becomes evident


that the orientation of stress characteristics in
the two regions is different and varies for active
and passive conditions. In addition, the mean
stresses SA and SB generally do not coincide,
which suggests that a Rankine-type solution
based on a single stress field is not possible.
To determine the separation of mean
stresses SA and SB and ensure a smooth

This additional condition is written as (Chen


1975):

S B = S A exp ( 2 tan )
2

(10)

This should not be confused with log-spiral shaped


failure surfaces used in kinematic solutions of related
problems.

109

The negative sign in the above equation


pertains to the case where SB < SA (e.g., active
case) and vice versa. The above equation is an
exact solution of the governing Ktter equations
for a weightless frictional material. For a
material with weight, the solution is only
approximate as Kotters equations are not
perfectly satisfied (Davis & Selvadurai, 2002;
Atkinson, 1981; Parry, 1995). In other words,
the log spiral fan accurately transmits stresses
applied at its boundaries, but handles only
approximately body forces imposed within its
volume. The error is expected to be small for
active conditions (which are of key importance
in design), because of the small opening angle
of the fan, and bigger for passive conditions. As
a result, the above solution cannot be
interpreted in the context of limit analysis
theorems. Nevertheless, it will be shown that
these violations are of minor importance from a
practical viewpoint.

is twice the angle separating the major principal


planes in zones A and B (Figure 4). The
convention regarding double signs in the above
equations is as before.
It is also straightforward to show that the
surcharge coefficient Kq is related to K through
the simple expression

K q = K

The total thrust on the wall due to surcharge


and gravity loading is obtained by the wellknown expression (Chen 1975):
(11)

which is reminiscent (though not equivalent) of


the bearing capacity equation of a strip surface
footing on cohesionless soil. In the above
equation, Kq and K denote the earth pressure
coefficients due to surcharge and self-weight,
respectively.
Combining Eqns (5), (8) and (10), and
integrating over the height of the wall, it is
straightforward to show that the earth pressure
coefficient K is given by [39]

K =

cos ( ) cos
cos cos 2

Solution including Earthquake Loading

Recognizing that earthquake action imposes


a resultant thrust in the backfill inclined by a
constant angle e from the vertical (Fig 1), it
becomes apparent that the pseudo-dynamic
problem does not differ fundamentally from the
corresponding static problem, as the former can
be obtained from the latter through a rotation of
the reference axes by the seismic angle e, as
shown in Fig 5. In other words, considering e
does not add an extra physical parameter to the
problem, but simply alters the values of the
other variables. This property of similarity was
apparently first employed by Briske (1927) and
later by Arango (Seed & Whitman, 1970;
Ebeling et al., 1992) in the analysis of related
problems.

(12)

1 sin cos ( 2 )

exp( 2 tan )
1 sin cos ( 1 )
where

2 = 2 ( 1 + ) + 2

(14)

which coincides with the kinematic solution of


Chen & Liu (1990), established using a
Coulomb mechanism. Note that for a horizontal
backfill ( = 0), coefficients Kq and K coincide
regardless of wall inclination and material
properties. Equation (14) represents an exact
solution for a weightless material with
surcharge. A simplified version of the above
solutions, restricted to the special case of a
vertical wall with horizontal backfill and no
surcharge (e = 0; = = 0; q =0), has been
derived by Lancelotta (2002). Another simplified
solution, which, however, contains some
algebraic mistakes (see application example at
the end of the paper) and is restricted to active
conditions and no surcharge, has been
presented by Powrie (1997).

Solution without Earthquake Loading

1
P = K q q H + K H 2
2

cos
cos ( )

(13)

110

strength parameters and are invariant to the


transformation.
In the light of the above developments, the
soil thrust including earthquake action can be
determined from the modified expression:

1
PE = K q* q* H * + K * * H *2
2

H*

in which parameters , , H, , and q have


been
replaced
by
their
transformed
counterparts. The symbols Kq* and K* denote
the surcharge and self-weight coefficients in the
modified geometry, respectively.
Substituting Eqns (15) through (19) in Eqn
(20) yields the modified earth pressure
coefficient

Fig 5: Similarity transformation based on a rotation


of the reference axes, for analyzing the
pseudo-dynamic seismic excitation as a
gravitational problem. Note the different wall
height, backfill slope, and wall inclination in
the modified geometry. Note also that the
rotation should be performed in the opposite
(clockwise) direction for passive pressures
(e<0)

KE = (1 av )

(15)

*= + e

(16)

H * = H cos ( + e ) / cos

(17)

* = (1 av ) / cos e

(18)

q * = q (1 av ) / cos e

(19)

cos ( ) cos( + e )
cos e cos cos2

(21)

1 sin cos ( 2 )

exp( 2 E tan )
*
1 sin cos 1 ( + e )

which encompasses seismic action and can be


used in the context of Eqn (11). In the above
equation,

Application of the concept to the present


analysis yields the following algebraic
transformations, according to the notation of
Figure 5:

*= + e

(20)

2 E = 2 ( 1* + ) + 2 e

(22)

is twice the revolution angle of principal


stresses in the two regions under seismic
conditions; 1* is equal to Arcsin[sin(+e)
/sin], following Eqns (6) and (15).
The seismic earth pressure coefficient KEq is
obtained as

K Eq = K E

The modification in and q is due to the


change in length of the corresponding vectors
(Fig 1) as a result of inertial action. To obtain
Eqn (19), it has been tacitly assumed that the
surcharge responds to the earthquake motion in
the same manner as the backfill and, thereby,
the transformed surcharge remains vertical.
Note that this is not an essential hypothesis -
just a convenient (reasonable) assumption from
an analysis viewpoint. Understandably, the

cos
cos ( )

(23)

which coincides with the static solution in Eqn


(14).
The horizontal component of soil thrust is
determined from the actual geometry, as in the
gravitational problem:

PEH = PE cos( )

111

(24)

predictions are in good agreement (largest


discrepancy about 10%), with the exception of
Coulombs
method
which
significantly
overestimates passive pressures. Moving from
the top to the bottom of each column, an
increase in KA values and a decrease in KP
values can be observed. This is easily
understood given the non-conservative nature
of the first two solutions (Coulomb, Chen), and
the conservative nature of the last two
(Sokolovskii 1965, proposed). This observation
does not hold for the zero extension line
solution of Habibagahi and Ghahramani (1977),
which cannot be classified in the context of limit
analysis theorems. Results for gravitational
active pressures on a rough inclined wall
obtained according to three different methods
as a function of the slope angle , are shown in
Fig 6.

Seismic Component of Soil Thrust

Following Seed & Whitman (1970), the


seismic component of soil thrust is defined from
the difference:

PE = PE P

(25)

which is mathematically valid, as the associated


vectors PE and P are coaxial. Nevertheless, the
physical meaning of PE is limited given that the
stress fields (and the corresponding failure
mechanisms) in the gravitational and seismic
problems are different. In addition, PE cannot
be interpreted in the context of limit analysis
theorems, as the difference of PE and P is
neither an upper nor a lower bound to the true
value.
MODEL VERIFICATION AND RESULTS

Presented in Table 1 are numerical results


for gravitational active and passive pressures
(KA, KP) from the present solution and
established solutions from the literature. The

Table I: Comparison of results for active and passive earth pressures predicted by various methods. The
results for = = 0 are identical for all methods. Note the decrease in KP values as we move from top to
bottom in each column, and the corresponding increase in KA values; = 0 (modified from Chen & Liu, 1990)
a. KA values

20

30

40

20

20

30

30

10

15

20

15

15

0.490

0.447

0.333

0.301

0.217

0.199

0.498

0.476

0.212

0.180

0.490

0.448

0.333

0.303

0.217

0.200

0.498

0.476

0.218

0.189

0.49

0.41

0.33

0.27

0.22

0.17

Slip line (Sokolovskii 1965)

0.490

0.450

0.330

0.300

0.220

0.200

0.521

0.487

0.229

0.206

Proposed Stress Limit


Analysis

0.490

0.451

0.333

0.305

0.217

0.201

0.531

0.485

0.237

0.217

2.04

2.64

3.00

4.98

4.60

11.77

2.27

3.162

5.34

12.91

2.04

2.58

3.00

4.70

4.60

10.07

2.27

3.160

5.09

8.92

2.04

2.55

3.00

4.65

4.60

9.95

Slip line (Sokolovskii 1965)

2.04

2.55

3.00

4.62

4.60

9.69

2.16

3.16

5.06

8.45

Proposed Stress Limit


Analysis

2.04

2.52

3.00

4.44

4.60

8.92

2.13

3.157

4.78

7.07

Coulomb
Kinematic Limit Analysis
(Chen & Liu 1990)
Zero extension (Habibagahi &
Ghahramani 1977)

b. KP values
Coulomb

Kinematic Limit Analysis


(Chen & Liu 1990)
Zero extension (Habibagahi &
Ghahramani 1977)

112

The performance of the proposed solution is


good (maximum deviation from Chens solution
about 10% - despite the high friction angle of
45o) and elucidates the accuracy of the
predictions. The performance of the simplified
solution of Caquot and Kerisel (1948) versus
that of Chen & Liu (1990) is as expected.
Corresponding
predictions
for
passive
pressures are given in Figure 7, for a wall with
negative backfill slope inclination, as a function
of the wall roughness . The agreement of the
various solutions, given the sensitivity of
passive pressure analyses, is very satisfactory.
Of particular interest are the predictions of
Sokolovskiis (1965) and Lee & Heringtons
(1972) methods, which, surprisingly, exceed
those of Chen for rough walls. This trend is
particularly pronounced for horizontal backfill
and values of above approximately 10o and
has been discussed by Chen & Liu.
Results for active seismic earth pressures
are given in Figure 8, referring to cases
examined in the seminal study of Seed &
Whitman (1970), for a reference friction angle of
35o. Naturally, active pressures increase with

0.6
1
K A = PA /( H2 )
2

Coefficient of Active Earth Pressure, K A

0.5

= 45o , = 2 / 3

PA
0.4

= 20o
0.3

Chen & Liu (1990)


Caquot & Kerisel (1948)
Proposed Stress Limit Analysis

0.2

= 0o

0.1

0.0

= 20o

10

15

20

Slope Angle of Backfill,

25
o

Fig 6: Comparison of results for active earth


pressures predicted by different methods
(modified from Chen 1975)

5
1
K P = PP /( H2 )
2

= 30 , = 20

Coefficient of Passive Earth Pressure, K P

PP
4

increasing levels of seismic acceleration and


slope inclination and decrease with increasing
friction angle and wall roughness. The
conservative nature of the proposed analysis
versus the Mononobe-Okabe (M-O) solution is
evident in the graphs. The trend is more
pronounced for high levels of horizontal seismic
coefficient (ah > 0.25), smooth walls, level
backfills, and high friction angles. Conversely,
the trend becomes weaker with steep backfills,
rough walls, and low friction angles.
A similar set of results is shown in Figure 9,
for a reference friction angle of 40o. The
following interesting observations can be made:
First: the predictions of the proposed analysis
are in good agreement with the results from the
kinematic analysis of Chen & Liu (1990), over a
wide range of material and geometric
parameters. Second, the present analysis is
conservative in all cases. Third, close to the
slope stability limit (Fig 9d), or for high
accelerations and large wall inclinations (Fig
9c), Chens predictions are less accurate than
those of the elementary M-O solution.

= 0o

= 10o

= 20o
1
Lee & Herington (1972)
Chen & Liu (1990)
Sokolovskii (1965)
Proposed Stress Limit Analysis

10

20

Angle of Wall Friction,

30

Fig 7: Comparison of results for passive earth


pressures predicted by different methods
(modified from Chen & Liu 1990)

113

interesting comparison is presented in Fig 10b:


average predictions from the two closed-form
solutions (M-O solution & proposed stress
solution) are plotted against the rigorous
numerical results of Chen & Liu (1990).
Evidently, in the range of most practical interest
(30o < < 40o), the discrepancies in the results
have been drastically reduced. This suggests
that the limit equilibrium (kinematic) M-O
solution and the proposed static solution
overestimate and underestimate, respectively,
passive resistances by the same amount in the
specific range of properties. Accordingly, this
averaging might be warranted for design
applications involving passive pressures.
Results for the earth pressure coefficient due
to surcharge KqE (Eq 23) are presented in
Figure 11, for both active and passive
conditions involving seismic action. The
agreement between the stress solution and the
numerical results of Chen & Liu (1990) is
excellent in the whole range of parameters
examined (except perhaps for active pressures,
where ah = 0.3). As expected, M-O solution
performs well for active pressures, but severely
overestimates the passive.

In the same extreme conditions, the


proposed solution becomes exceedingly
conservative, exceeding MO predictions by
about 35%. Note that whereas the M-O and the
proposed solution brake down in the slope
stability limit, Chens solution allows for
spurious mathematical predictions of active
thrust beyond the limit, as evident in Fig 9d.
Fourth,
with
the
exception
of
the
aforementioned extreme cases, Chens and MO predictions remain close over the whole
range
of
parameters
examined.
The
improvement in the predictions of the former
over the latter is marginal.
Results for seismic passive pressures
(resistances) are shown in Figure 10 for the
common case of a rough vertical wall with
horizontal backfill. Comparisons of the
proposed solution with results from the M-O
and Chens kinematic methods are provided on
the left graph (Fig 10a). The predictions of the
stress solutions are, understandably, lower than
those of Chen and Liu, whereas M-O
predictions are very high (i.e., unconservative)
especially for friction angles above 37
degrees. Given the sensitivity of passive
pressure analyses, the performance of the
proposed method is deemed satisfactory. An

114

0.7

0.7

0.6

= 35

= = 0o

0.6

0.5

= 35o

0.5

= 0o

= 0o

=/2

0.4

K E cos

Coefficient of Seismic Active Earth Pressure, K E

= = 0o

0.3

0.2

0.4

=/2
0.3

0.2

0.1

0.1
M - O Analysis
Proposed Stress Limit Analysis

0.0
0.0

0.1

0.2

0.3

0.4

M - O Analysis
Proposed Stress Limit Analysis

0.0
0.0

0.5

0.1

0.3

0.4

0.5

Horizontal Seismic Coefficient, ah

Horizontal Seismic Coefficient, ah

0.7

0.7

= = 0o

= = 0o
0.6

0.2

= 30

=/2

0.6

= 35 o ; = / 2
o

= 20

3 5o

4 0o

0.5

0.5

K E cos

K E cos

=0

0.4

0.3

0.2

0.4

0.3

0.2

0.1

0.1
M - O Analysis

M - O Analysis
Proposed Stress Limit Analysis

Proposed Stress Limit Analysis

0.0
0.0

0.1

0.2

0.3

0.4

0.5

Horizontal Seismic Coefficient, ah

0.0
0.0

0.1

0.2

0.3

0.4

Horizontal Seismic Coefficient, ah

Fig 8: Comparison of active seismic earth pressures predicted by the proposed


solution and from conventional M - O analysis, for different geometries, material
properties and acceleration levels ; av = 0. (modified from Seed & Whitman, 1970)

115

0.5

1
K AE = PAE /( H2 )
2
= = 0o ; = 2 / 3

(a)

0,6

(b)

PAE

0,5

0,5

0.30
0.20

0,4

30

35

40

M - O Analysis
Kinematic Limit Analysis (Chen & Liu 1990)
Proposed Stress Limit Analysis

0,1

45

0,0

0,1

0,2

0,3

0,4

Horizontal Seismic Coefficient, ah

Friction Angle, o

1,6

1,0

K AE

(d)

1
= PAE /( H2 )
2

Coefficient of Seismic Active Earth Pressure, K AE

(c)

= 40 ; ah = 0.20 ; = / 2
0,8

PAE
15o

0,4

= 0o
15o

0,2
M - O Analysis
Kinematic Limit Analysis (Chen & Liu 1990)
Proposed Stress Limit Analysis

-20

/ =1

0,2

0,1
25

Coefficient of Seismic Active Earth Pressure, K AE

M - O Analysis
Kinematic Limit Analysis (Chen & Liu 1990)
Proposed Stress Limit Analysis

0,0

0,4

ah = 0

0,2

0,6

PAE

0,3

0.10
0,3

1
K AE = PAE /( H2 )
2
= = 0o ; = 40o

0,6

K AE cos

Coefficient of Seismic Active Earth Pressure, K AE

0,7

-10

10

Slope Angle of Backfill,

1,4

1,0

slope
stability

= 40o ; = 0o ; = / 2

1,2

limit

PAE

0,8

=/2

0,6

/3
0o

0,4

0,2

0,0
0,0

20

1
K AE = PAE /( H2 )
2

M - O Analysis
Kinematic Limit Analysis (Chen & Liu 1990)
Proposed Stress Limit Analysis

0,1

0,2

0,3

Horizontal Seismic Coefficient, ah

Fig 9: Comparison of active seismic earth pressures predicted by different methods,


for different geometries, material properties, and acceleration levels; f = 40o,
av = 0. (modified from Chen & Liu, 1990)

116

0,4

25

25

Coefficient of Seismic Passive Earth Pressure, K PE

(a)

20

15

(b)

H PPE

H PPE

20

ah = 0
0.1

1
K PE = PPE /( H2 )
2

0.2
0.3

= 0o , = 0o
= 2 / 3

10

10

30

35

ah = 0
0.1
0.2
0.3

Kinematic Limit Analysis (Chen & Liu 1990)


Proposed Stress Limit Analysis

0
25

1
K PE = PPE /( H2 )
2

15

Mononobe - Okabe
(ah = 0)

Kinematic Limit Analysis (Chen & Liu 1990)


Average of M - O & Proposed Stress Limit Analysis

40

0
25

45

30

35

40

45

Angle of Internal Friction, o

Angle of Internal Friction, o

Fig 10: Comparison of results for passive seismic resistance on a rough wall
predicted by various methods. (Modified from Chen & Liu, 1990)

20

0,6

= = 0o

PAE

0,5

ah q
H

0.3
0.2

0,4

0.1
0,3

PPE

15

=2/3

K AE = PAE / q H

(b)

ah q

K PE = PPE / q H
q

0,7

(a)

10

ah = 0
0.1
0.2
0.3

= = 0o
=2/3

Mononobe - Okabe
(ah = 0)

ah = 0
5

0,2
Kinematic Limit Analysis (Chen & Liu 1990)
Proposed Stress Limit Analysis

0,1
25

30

35

Friction Angle,

40

Kinematic Limit Analysis (Chen & Liu 1990)


Proposed Stress Limit Analysis

45

0
25

30

35

Friction Angle,

40
o

Fig 11: Variation of KAEq and KPEq values with - angle for different acceleration levels.

117

45

Distribution of Earth Pressures on the Wall:


Analytical Findings

cos = 1 + tan 2 stands for the vectorial sum


of shear and normal tractions at the wall-soil
interface. Factor cos arises from the
equilibrium of the infinite slope in Eqn (4a).
Finally, cos( ) / cos2 is a geometric factor
arising from the integration of stresses along
the back of the wall, and is associated with the
inclination of the wall and backfill.
In light of
the above, the solution for gravitational
pressures can be expressed by the single
equation

Mention has already been made that in the


realm of pseudo-dynamic analysis, there is no
fundamental physical difference between
gravitational and seismic earth pressures.
Equations (4) indicate that stresses in the soil
vary linearly with depth (stress fan does not
alter this dependence), which implies that both
gravitational and seismic earth pressures vary
linearly along the back of wall. In the absence
of surcharge, the distribution becomes
proportional with depth, as in the Rankine
solution. Accordingly, the point of application of
seismic thrust is located at a height of H/3
above the base of the wall. It is well known from
experimental
observations
and
rigorous
numerical solutions, that this is not generally
true. The source of the difference lies in the
distribution of inertial forces in the soil mass
(which is often sinusoidal like - following the
time-varying natural mode shapes of the
deposit), as well as the various kinematic
boundary conditions (wall flexibility, foundation
compliance, presence of supports). Studying
the above factors lies beyond the scope of this
article, and like will be the subject of a future
publication. Some recent developments are
provided in the Master thesis of the second
author (Kloukinas, 2006) as well as in (Kramer,
1996;
Veletsos
&
Younan,
1994;
Theodorakopoulos et.al. 2005; Ostadan, 2001;
Paik & Salgado, 2003) .

K =

cos ( ) cos
cos cos2

1 sin cos ( 2 )

exp( 2 tan )
1 + sin cos[ 1 + ]

(26)

which is valid both for active conditions (using


positive values for and ) and passive
conditions (using negative values for and ). It
is straightforward to show that this property is
not valid for the Mononobe-Okabe solutions in
Eqn (4). The lack of symmetry in the limit
equilibrium solutions can be attributed to the
optimization
(i.e.,
maximization
and
minimization) involved in deriving the limit
thrusts. An application example elucidating the
simplicity of the solution is provided below.
Application Example

Active and passive pressures will be


computed for a gravity wall of height H = 5m,
inclination = 5o and roughness = 20o,
retaining an inclined cohesionless material with
= 30o, =18kN/m3 and = 15o, subjected to
earthquake accelerations ah = 0.2 and av = 0.
The static counterpart of the problem has been
presented by Powrie (1997).
The inclination of the resultant body force in
the backfill is obtained from Eqn (1):

DISCUSSION: SIMPLICITY & SYMMETRY

It is instructive to show that the proposed


solution can be derived essentially by
inspection. Indeed, basis of Equation (12) is the
familiar Rankine ratio (1 sin) / (1 sin). The
terms cos(2 ) and cos (1 ) in the
numerator and denominator reflect the fact that
stresses and w are not principal. Both terms
involve the same double signs as their
multipliers (i.e. sin and sin, respectively).
Angle and associated Caquot angle 1 have
to be in the denominator, as an increase in their
value must lead to an increase of active thrust.
The exponential term is easy to remember and
involves the same double sing ( ) as the
other terms in the numerator. With reference to
the
factors
outside
the
brackets,

e = Arctan(0.2) = 11.3o

(A-1)

The two Caquot angles are determined from


Eqns (6), (9) and (15) as

1* = sin 1[sin (15 + 11.3 ) / sin 30] = 62.4o

118

(A-2)

2 = sin [sin ( 20 ) / sin 30] = 43.2


1

PAH = 94.5 cos(5 + 20) = 85.6 kN / m


o

(A-3)
Note that according to Powrie (1997), the
value of the horizontal component is (Eqn 9.42,
p 333):

The angle separating the major principal


planes in Regions A and B is computed from
Eqn (22):

2 E = 43.2 ( 62.4 + 20 ) + 15 2 5 11.3


= 45.5o

1
0.395 18 52 (1 + tan 5 tan 20 )
2
= 91.7 kN / m

PAH =
(A-4)

which is in error as: (1) Ka, as determined from


Eqn 9.41 in Powries book, should be 0.385
not 0.395. (2) the correct sign in front of product
(tan x tan) is minus one. (3) Powries
equation does not encompass factor [cos( )
/ cos cos] arising from the integration of
stresses on the back of the wall.
For the passive case, the corresponding
parameters are:

Based on the above values, the earth


pressure coefficient is obtained from Eqn (21):

K AE =

cos ( 5 15) cos(15 + 11.3)


cos11.3cos20 cos2 5

1 sin 30 cos ( 43.2 20 )

1 + sin 30 cos 62.4 + (15 + 11.3)


exp(+ 45.5

180

(A-5)

e = Arctan(0.2) = 11.3o ,

tan 30) = 0.82

1* = sin 1[sin (15 11.3) / sin 30] = 7.4o ,


2 E = 43.2 + ( 7.4 + 20 ) + 15 2 5 + 11.3 = 86.9o ,

from which the overall active thrust on the wall


is easily determined (Eqn 11):

PAE =

1
0.82 18 52 = 185 kN / m
2

For
the
gravitational
corresponding parameters are

problem,

(A-8)

The passive earth pressure coefficient and


resistance are obtained from Eqns (21) and
(11):

(A-6)

K PE =

cos ( 5 15) cos(15 11.3)


cos11.3cos20 cos2 5

1 + sin 30 cos ( 43.2 + 20 )

1 sin 30 cos 7.41 (15 11.3)

the

1 = sin1[sin15 / sin 30] = 31.2o ,

exp(+ 2 E

2 = sin 1[sin ( 20 ) / sin 30] = 43.2o ,

180

(A-9)

tan 30) = 6.31

2 = 43.2 ( 31.2 + 20 ) + 15 2 5 = 3o ,

K A = 0.42 .

PPE =

Thus,

PA =

1
0.42 18 52 = 94.5 kN / m
2

1
6.3118 52 = 1420 kN / m
2

(A-10)

CONCLUSIONS

(A-7)

A stress plasticity solution was presented for


determining gravitational and earthquakeinduced earth pressures on gravity walls
retaining cohesionless soil. The proposed
solution incorporates idealized, yet realistic wall

The horizontal component of gravitational


soil thrust is determined from Eqn (24)

119

geometries and material properties. The


following are the main conclusions of the study:

appropriate signs for friction angle and wall


roughness.

1) The proposed solution is simpler than the


classical Coulomb and Mononobe-Okabe
equations. The main features of the
mathematical expressions, including signs, can
be deduced by physical reasoning, which is
hardly the case with the classical equations.

It should be emphasized that the verification


of the proposed solution was restricted to
analytical and not experimental results. Detailed
comparisons against experimental results
including distribution of earth pressures along
the wall, will be the subject of a future
publication.

2) Extensive comparisons with established


numerical solutions indicate that the proposed
solution is safe, as it overestimates active
pressures and under-predicts passive. This
makes the method appealing for practical
applications.

ACKNOWLEDGMENTS

The Authors are indebted to Professor


Dimitrios Atmatzidis for his constructive
criticism of the work.

3) For active pressures, the accuracy of the


solution is excellent (maximum observed
deviation from numerical data about 10%). The
largest deviations occur for high seismic
accelerations, high friction angles, steep
backfills, and negative wall inclinations.

REFERENCES
Anastasopoulos I., Gazetas G., Psarropoulos Pr.
(2003), Flexible Retaining Walls : Why They do
not often Fail in Strong Seismic Shaking, Proc.
Fib Int. Symposium on Concrete Structures in
Seismic Regions, Athens, May 2003.

4) For passive resistances, the predictions are


also satisfactory. However, the error is larger especially at high friction angles. Nevertheless,
the improvement over the M-O predictions is
dramatic. Taking the average between the
predictions of the M-O solution and the
proposed stress solution yields results which
are comparable to those obtained from rigorous
numerical solutions.

Atkinson, J. (1981), Foundations and slopes,


McGraw Hill, London.

5) The pseudo-dynamic seismic problem can


be deduced from the corresponding static
problem through a revolution of the reference
axes by the seismic angle e (Fig 5). This
similarity suggests that Coulomb and M-O
solutions are essentially equivalent.

Chen, W.F. (1975), Limit analysis and soil plasticity,


Developments in geotechnical engineering,
Elsevier, Amsterdam.

Briske, R. (1927), "Die Erdbebensicherheit von


Bauwerken", Die Bautechnik, Vol 5, 425- 430,
453-457, 547-555.
Caquot, A. and Kerisel, L. (1948), Trait de
mcanique des sols, Gauthier-Villars, Paris.

Chen, W.F., Liu, X.L. (1990), Limit analysis in soil


mechanics, Elsevier, Amsterdam.

6) Contrary to the overall gravitational-seismic


thrust PE, the purely seismic component PE =
PE P cannot be put in the context of a lower or
an upper bound. This holds even when PE and
P are upper or lower bounds.

Clough GW, Duncan JM. (1990), Earth pressures. In


Foundation Engineering Handbook, Fang HY
(ed.). Chapman & Hall:New York, pp.223-235.
Collins IL. (1973), A note on the interpretation of
Coulomb analysis of the thrust on a rough
retaining wall in terms of the limit theorems of
plasticity theory, Geotechnique, Vol. 24, No 1,
pp.106-108.

7) In the realm of the proposed model, the


distribution of earth pressures on the back of
the wall is linear with depth for both
gravitational and seismic conditions. This is not
coincidental given the similarity between the
gravitational and pseudo-dynamic problem.

Coulomb, C.A. (1776), "Essai sur une application


des regles de maximis et minimis a quelqes
problemes de stratique relatifs a l architecture".
Memoires de Mathematique et de Physique.
Presentes a l Academie Royale des Sciences;
Paris, 7, 343-382.

8) Contrary to the Mononobe-Okabe solutions,


the proposed solution is symmetric, as it can be
expressed by a single equation, describing both
active and passive pressures, by means of

120

Davis, R.O. and Selvadurai, A.P.S. (2002), Plasticity


and Geomechanics, Cambridge University Press,
Cambridge.

Lambe TW, Whitman RV. (1969), Soil Mechanics,


Wiley & Sons: NY.
Lancelotta R. (2002), Analytical solution of passive
earth pressure. Geotechnique, 52, 8, pp. 617619.

Davies TG, Richards R, Chen KH (1986). Passive


pressure during seismic loading, Journal of
Geotechnical Engineering, Vol. 112, No 4, pp.
479-483

Lee, I.K. and Herington, J.R. (1972), "A theoretical


study of the pressures acting on a rigid wall by a
sloping earth or rock fill", Geotechnique, 22, No
1, 1-26.

Ebeling, R.M., Morrison, E.E., Whitman, R.V., Liam


Finn, W.D. (1992), A Manual for Seismic Design
of Waterfront Retaining Strutures, US Army
Corps of Engineers, Tech. Report ITL-92-11.

Matsuo M, Ohara S. (1960), Lateral earth pressures


and stability of quay walls during earthquakes,
Proceedings, Second World Conference on
Earthquake Engineering, Tokyo, Japan.

Finn WD. Applications of Limit Plasticity in Soil


Mechanics. (1967), Journal of the Soil Mechanics
and Foundations Division, ASCE, Vol. 93 (SM5),
pp.101-120.

Mononobe, N., Matsuo, O. (1929), On the


determination
of
earth
pressure
during
earthquakes,
Proceeding
of
the
World
Engineering Congress, Tokyo, vol 9, 179-187.

Finn WD, Yogendrakumar M, Otsu H, Steedman


RS. (1989), Seismic response of a cantilever
retaining wall: centrifuge model test and dynamic
analysis. Proceedings of 4th International
Conference on Soil Dynamics and Earthquake
Engineering.
Computational
Mechanics
Publications: Southampton, pp.331-431.

Okabe, S. (1924), "General theory on earth pressure


and seismic stability of retaining walls and dams",
Journal of the Japanese Society of Civil
Engineers, 10, 6, 1277-1323.
Ostadan F. (2005), Seismic soil pressure for building
walls: an updated approach, Soil Dynamics and
Earthquake Engineering, 25, pp 785-793

Garini, E., Gazetas, G., and Gerolymos, N. (2006),


Asymmetric block sliding from idealized pulse
wavelets and near-fault ground motions,
Panhellenic Conference on Geotechnical &
Geoenvironmental Engineering, Xanthi, May 31
June 2 (in Greek).

Paik K and Salgado R. (2003), Estimation of Active


Earth Pressure Against Rigid Retaining Walls
Considering Arching Effects, Geote-chnique, 53,
No 7, pp 643-653.

Gazetas G., Psaropoulos P.., Anastasopoulos I.,


and Gerolymos N. (2004), Seismic Behaviour of
Flexible Retaining Systems Subjected to Short
Duration Moderately Strong Excitation, Soil
Dynamics & Earthquake Engineering, Vol. 24,
No. 7, pp. 537-550.

Papantonopoulos, C. and Ladanyi, B. (1973),


"Analyse de la Stabilitee des Talus Rocheux par
une Methode Generalisee de lEquilibre Limite",
Proceedings, 9th Canadian Rock Mechanics
Symposium, Montreal, 167-196 (in French).

Gazetas G., Gerolymos N., Anastasopoulos I.


(2005), Response of Three Athens Metro
Underground Structures in the 1999 Parnitha
Earthquake, Soil Dynamics & Earthquake
Engineering, Vol. 25, No. 7-10, pp. 617633.

Parry RHG. (1995), Mohr Circles, Stress Paths and


Geotechnics, E & FN Spon.
Pecker A. (1995), Seismic design of shallow
foundations. State-of-the-Art: 10th European
Conference on Earthquake Engineering, Duma
Ed. Balkema, pp 1001-1010.

Habibagahi K and Ghahramani A. (1977), Zero


extension theory of earth pressure. Journal of the
Geotechnical Engineering Division, ASCE;
105(GT7): pp.881-896.

Powrie W. (1997), Soil Mechanics: Concepts &


Applications, E & FN Spon, London.
Psarropoulos PN, Klonaris G, Gazetas G. (2005)
Seismic earth pressures on rigid and flexible
retaining walls, Soil Dynamics and Earthquake
Engineering; Vol. 24, p.p. 795809.

Heyman J. (1972), Coulombs Memoir on Statics; an


essay in the history of civil engineering,
Cambridge University Press
Kramer, S.L., (1996), Geotechnical Earthquake
Engineering, Prentice Hal.

Richards R, Elms DG. (1979), Seismic behaviour of


gravity retaining walls. Journal of the
Geotechnical Engineering Division, ASCE;
105(GT4): pp.449-464.

Kloukinas, P., (2006), "Gravitational and Seismic


Earth Pressures by Stress Limit Analysis", Ms
Thesis, University of Patras, Greece (in Greek
with extended English summary).

Salencon, J. (1974), Applications of the Theory of


Plasticity in Soil Mechanics. John Wiley.

121

Salencon, J. (1990), Introduction to the yield design


theory, European Journal of Mechanics A/Solids,
Vol. 9, No 5, pp. 477-500

Theodorakopoulos DD, Chassiakos AP, Beskos DE.


(2001), Dynamic pressures on rigid cantilever
walls retaining poroelastic soil media. Part II:
Second method of solution, Soil Dynamics &
Earthquake Engineering, 21, 4, p. 339-364.

Seed HB, Whitman, R.V. (1970), Design of earth


retaining
structures
for
dynamic
loads,
Proceedings of specialty conference on lateral
stresses in the ground and design of earth
retaining structures, ASCE, Ithaca, New York,
103-147.

Veletsos A.S. and Younan A.H. (1994). "Dynamic


soil pressures on rigid retaining walls",
Earthquake
Engineering
and
Structural
Dynamics, 23, pp. 275 301.

Sheriff MA, Ishibashi I, Lee CD. (1982), Earth


pressures against rigid retaining walls, Journal
of Geotechnical Engineering, ASCE ; 108: GT5,
pp. 679-696.

Whitman, R.V. and Liao, S. (1985), Seismic Design


of Gravity Retaining Walls, US Army Corps of
Engineers, Miscellaneous paper GL-85-1.
Wood, JH. (1973), Earthquake induced soil
pressures on structures. Doctoral Dissertation,
EERL 73-50, California Institute of Technology,
Pasadena, CA.

Sokolovskii, V.V. (1965), Statics of granular media,


Pergamon Press, New York.
Steedman RS, Zeng X. (1990), The influence of
phase on the calculation of pseudo-static earth
pressure on a retaining wall, Geotechnique, Vol.
40, pp.103-112
Terzaghi, K., (1943), Theoretical soil mechanics,
John Wiley & Sons Inc., New York.
Theodorakopoulos DD, Chassiakos AP, Beskos DE.
(2001), Dynamic pressures on rigid cantilever
walls retaining poroelastic soil media. Part I: First
method of solution, Soil Dynamics & Earthquake
Engineering, 21, 4, p. 315-338.

122

A Study on Ground Displacement due to Fault Slip


and its Influence to Underground Structure
Y. Adachi1, S. Yoshimura , T. Nakata
1
2

Hanshin Expressway Company Limited, Kyoto, Japan


Yachiyo Engineering Company Limited, Niigata, Japan

Abstract
The influence to shield tunnel structure due to ground slip, crack, and deformation
propagation in sedimentary layer caused by fault slip is studied. The base rock
displacement is estimated by empirical relations between the earthquake magnitude and the
base rock displacement. The ground displacement propagation is analyzed using linear and
non-linear finite element method with joint elements. As the result of this study, it is found
that the fault displacement does not reach to the surface where the base rock displacement
is relatively smaller than the thickness of the sedimentary layer. Therefore, in the case
where the tunnel existed in relatively shallow depth in deep sedimentary layer, critical
damage may not occur to the tunnel even though the tunnel suffers the ground
displacement.

between the experiment results and the site


survey results. Ramancharla and Megro (2001)
estimated the propagation of fault rupture in
deposit layer due to base layer slip by applied
element analysis. Konagai et al. (2001) also
simulated the propagation in deposit layer by
Lagrangian particle finite difference method.
There are some studies about the fault
rupture propagation in soil and its influence to
structures. Newmark et al (1975) proposed the
design method of buried pipeline to resist
against large fault displacement. Kennedy et al
(1977) estimated fault movement effects on
buried oil pipeline. Kiyomiya et al (2003)
calculated the fault movement effects on buried
sea bed pipeline by beam elements on soil
spring. Takada et al (2001) studied the relation
between the strain on buried pipeline and fault
rupture propagation and proposed the
estimation method of the maximum strain of the
pipeline. As described above, almost of the
researches were based on the ideal condition,
not the realistic condition.
The authors studied the shield tunnel
designed in the diluvium soil layer with crashed
planes
caused
by
fault
displacement
propagation by finite element analysis using

INTRODUCTION

A lot of social infrastructure such as bridges


and tunnels were suffered by the ground
displacement due to the fault slip of 1999
Kocaeli earthquake and Chi-Chi earthquake. It
is generally said very difficult to design resistant
structures against such ground displacement,
especially underground structures. However it is
very important to estimate the damage due to
such ground displacement and take possible
fail safe measures as much as possible.
There are a lot of studies about the fault
displacement propagation in deposit layer. Cole
et al. (1984) discussed influence zones in
alluvium over dip-slip faults by sand box
experiment. Bray et al. (1994) studied
earthquake fault rupture propagation through
soil due to reverse fault displacement by
nonlinear FEM analysis and sand box
experiment. Taniyama et al. (1998) estimated
the deformation in sandy deposit layer due to
reverse faulting by nonlinear FEM analysis and
sand box experiment. Onizuka et al. (1999)
studied the propagation of reverse faulting in
the sandy deposit layer by aluminum bar
experiment and showed the good agreement

123

600

550

500

450

400

350

300

100

250

200

150

100

50

-150

-200

-250

-300

-350

-400

METER

METER

METER

Crashed Planes
Shield Tunnel

100

Surface

Alluvium Layer

Diluvium Layer
-100

Base Line

-100

Base Rock

-200

-200

-300

-300

-400

-400

Fig. 1: Ground condition, location of crashed planes, and location of shield tunnel structure
Table 1: Estimated ground displacement

joint elements. This paper studies the difference


of the fault displacement propagation on linear
or non-linear soil properties and also discusses
the difference of the influence to the tunnel
structure.

Source

EQ Type

Ground

Matsuda

Inland (Japan)

Surface

Satoh et al

All (Japan)

Subface

69.2

Ave

Takemura et al

Inland (Japan)

56.5

Ave

SOILCONDITION AND SHIELD TUNNEL


Donald et al

The soil condition and the shield tunnel


structure studied in this paper are shown in Fig
1. Reflection test was performed in this field
and four crashed planes were found in the
deposit sediment layer indicated by ,,,
and in Fig. 1. There also found the fault
flexure lines at the ground surface related to
those crashed planes. The faults were
considered as reverse faults, and the return
period was estimated around 10,000 years, and
the maximum possible magnitude was
estimated M=6.5 by the area of the fault plane.
The planned tunnel was a shield tunnel with
composite segments. The diameter of the
tunnel was 10.6m. The aliment of the tunnel
was not so deep in the ground because of the
road tunnel.

Displ.(cm)
79.4

Ave

All (World)

Surface

55.4

Max

All (World)*

Surface

90.1

Max

All (World)

Surface

29.9

Ave

All (World)*

Surface

8.4

Ave

*: Only reverse faults are considered

displacement at the base layer.


Current studies use the statistical estimation
method for determining the base layer
displacement by analyzing the relation between
the displacement of the surface ground and the
magnitude of the earthquakes. These statistical
studies do not focused on the presence of the
deposit layer so that the obtained statistical
relation is not direct relation between the base
layer displacement and the earthquake
magnitude. But in this study, the statistical
estimation method can be assumed to estimate
the base layer displacement. Freeman et al
(2000) also took the same assumption and
studied about the assessment of the dam
against the fault displacement.
In this study, the works of Matsuda (1975)
which studied the fault displacement by
collecting active faults earthquake data
obtained in Japan, that of Sato et al (1989)
which studied about the active and plate
boundary earthquake in and around Japan, that
of Takemura et al (1998) which studied about
the in-crust earthquake, and that of Donald et al
(1994) which studied about the all of the

ESTIMATION OF FAULT DISPLACEMENT AT


BASE LAYER

First of all, the fault displacement at the


base layer is needed to be determined in order
to study the fault displacement propagation in
the deposit layer. Generally said, the base layer
displacement and the fault displacement are
different. Therefore the fault displacement used
for the strong motion analysis by fault rupture
modeling is not good for estimating the

124

Table 2: Soil properties of deposit layers


3

+40m

(kN/m ) VS(m/s) VP(m/s)


Alluvium
15.7
200
1,600
16.7
400
1,700
Diluvium

Crashed Plane()

18.6

Diluvium

Crashed Plane()

600

G(kN/m ) E(kN/m )
0.49 63,800 192,000
0.47 273,000 800,000
1,980,00
0.44 695,000
0

1,900

Shield Tunnel

Table 3: Joint element


Deposit Layers

m
Alluviu

Layer

m
Diluviu

Layer

Diluvium

as
Cr

es
lan
dP
he

C(kN/m )
0
0

Model 1
Model 2

Base Displacement

Layer

KnkN/m
1,000
1,000

Table 4: Physical properties of tunnel segment

yer
Base La

(kN/m )

Im

EkN/m

24.0

0.15

108.9

486,000

Equivalent
stiffness

Fig. 2: Modeling of ground and tunnel for FEM analysis

earthquake recorded in the world so far are


used
for
estimating
the
base
layer
displacement.
Table 1 shows the result of the base layer
displacement obtained by those studies above.
The target fault is considered as inland active
fault so that the study by the Matsuda (1975) is
seemed good to be applied in this study. By
using the Matsuda relation, 80cm of the base
layer displacement is computed for the target
fault. However, the Matsuda formula computes
the average value, so that the dispersion of
fault rupture phenomenon should be considered
from the engineering point of view. In this study,
10% of the dispersion is considered. So finally,
90cm of base layer displacement is used for
computation.
Fig 1 shows the four found crashed planes.
From the fault geology point of view, the most
distant crashed plane in the axis of fault incline
direction is the newest. But the cover depth of
surface layer is the deepest in the four planes.
Therefore, the oldest crashed plane where the
cover depth is the shallowest among the four is
employed in this study.

KrkN/m
1,000
1,000

35
20

Table 5: Study cases


Soil model
Linear
Linear
Non-linear
Non-linear

Case 1
Case 2
Case3
Case4

Crashed plane
Joint model 1
Joint model 1
Joint model 1
Joint model 1

n Normal stress

Tension

Tunnel stiffness
Equivalent Stiff.
Equivalent Stiff.
Equivalent Stiff.
Equivalent Stiff.

Shear stress
y=|n|tan+c

Sliding

Ks

Kn
Compression

Shear strain

Normal strain
Open

Sliding
-y=-|n|tan+c

(1) Normal direction


(2) Sliding direction
Fig. 3: Modeling of Crashed plane

is modeled by 3 layers, one alluvium, and two


diluvium layers. The perpendicular section in
which the shield tunnel is existed is used for the
two dimensional analysis.
As for the soil modeling, linear solid
elements are used for the three-dimensional
analysis and nonlinear solid model which
assumed the bilinear behavior that showed the
Mohr-Coulomb yield relation. The density , P
wave velocity Vp, and S wave velocity Vs are
computed from the result of geophysical logging.
The dynamic Poisson ratio , shear modulus G,
elastic modulus E are computed the estimated value
above mentioned.

ANALYSIS METHOD AND MODELING

Joint element which can express slide and


opening is employed for expressing the
geophysical property of the crashed plane. The
modeling of the joint element was shown in Fig.
3. Table 3 shows also the modeling of the joint
element. The cohesion is assumed c=0 for the
safety side computation. The internal friction
angle is assumed =35which is the standard
value for sand layer, and for=20which is

In order to study the fault rupture


propagation in deposit layer, three-dimensional
FEM analysis are performed with linear or nonlinear solid elements, joint elements with
standard value or with limited value.
Fig. 2 shows the FEM model used for the
three dimensional analysis. The analyzed area
is 720m long and 200m wide. The deposit layer

125

the reduced value which considers the repeated


fault slide.
For the shield tunnel, the tunnel is modeled
as a beam element which has effective stiffness
considering the ring joint stiffness.
As for the boundary condition, the base
layer is fixed. 90 cm of base layer displacement
with 45 incline angle is applied at the upper
(right) side of the base layer as shown in Fig. 1
and 2.
Table 5 shows the analytical cases
conducted in this study. All the cases, joint
elements are used for the crashed planes.
Case 1 is the case of three-dimensional
analysis with linear solid model. Case2 is the
two dimensional analysis with linear solid model.
Case 3 is the two dimensional analysis with
nonlinear solid model. Case 4 is the same as
case 3 except the internal friction angle
=20for the joint elements.
RESULT OF THE ANALYSES

Fig. 4 shows all the results of the


displacement of soil after being applied 90cm
offset at the base layer. Case1 shows the result
of the tunnel plane for the comparison. The
vertical direction is amplified 20 times of actual
scale. No evident gaps are observed along the
crashed plane for all the results even though all
the cases use joint elements for them.
Focused on the surface gap at the ground
surface indicated by the yarrows in Fig. 4,
2.1cm for Case 1, 2.1cm for Case 2, 0.0cm for
Case 3, and 0.0 cm for Case 4 are computed
so that the slide effects are so small compared
to the 90cm of the base offset. It can be
concluded that almost all the base offset are
absorbed in the soil in these cases. This agrees
with the condition that no evident fault line, only
flexure line, is observed at the field surface.
According to the previous studies, fault rupture
propagation can reach to the ground where the
base offset is over 4 to 5% of the cover
thickness of the deposit layer (Bray et al (1992)).
This case is the case where only 1% offset of
the deposit layer is applied to the deposit layer
so that the result obtained by this study agrees
with that of the previous ones.
Focused on the deformation mode of the
deposit layer, no difference to the Case1 and 2
was observed so that it can be said that there is
no clear difference to three and two

126

dimensional analyses. As for Case 1, the clear


difference is not observed for the deformation in
the tunnel plane and the side boundary plane
so that it can be said the tunnel stiffness is too
small compared to the soil.
There are clear difference of the soil
deformation modes compared to the results of
Case2, 3 and 4. As for Case2 where linear solid
model is used, only the upper side of the
deposit layer is displaced following with the
base offset. As for Case 3 and 4, not only upper
side but also lower side of the deposit layer
deform.
Fig. 5 shows the horizontal and vertical
displacement contour map of Case 2 and 3.
According to the result of Case 2, the lower side
of the deposit layer does not move in the
vertical direction but move in the horizontal
direction. The influenced area of Case 3 is
developed compared to that of Case 2. The
influence is concentrated to the joint element
where the linear solid model is used, but the
influenced area is widely developed where the
nonlinear solid model is used.
Fig. 6 shows the status of the joint elements,
Open, Slide, and Close. As for Case 2,
almost all elements of the crashed plane
show Slide except showing Open at the
bottom. The rest of the crashed plane show
Close so that it can be seen that the
displacement
is concentrated to just one
location. As for Case3 where the soil is
modeled by nonlinear elements so that the
lower deposit layer also deforms, the status of
the crashed planes are similar to those of Case
2. As for Case 4 where internal friction angle is
reduced, the crashed plane where the base
offset is applied and the adjacent plane show
Sliding.
Fig. 7 shows the relative displacement
between the upper and lower deposit layer
along the crashed plane of all the cases. As
for Case1 and 2, the influence of the offset of
the base layer reaches to about 30m upward
from the base which is almost 30 times of the
amount of the offset. The difference between
the two cases is pretty small. As for Case3 and
4 where the soil condition is modeled by
nonlinear elements, almost all the base offset is
absorbed with in 10m above the base. In these
cases, the base offset is absorbed in the lower
part of the deposit layer, not concentrated to the
joint elements.

(a) Case1

(b) Case2

(c) Case3

(d) Case4

Fig 4: Ground deformation (Red yarrows indicate the locations at the ground surface gaps)
0.

0.0

0.

0.0

-0.1

0.1

-0.1

0.1

-0.2

0.2

-0.2

0.2

-0.3

0.3

-0.3

0.3

-0.4

0.4

-0.4

0.4

-0.5

0.5

-0.5

0.5

-0.6

0.6

-0.6

0.6

-0.7

0.7

-0.7

0.7

-0.8

0.8

-0.8

0.8

-0.9

0.9

-0.9

0.9

-1.

1.0

Output Set: Interval 1, Step


Contour: X Translation

-1.

(a) Horizontal displacement contour map (Case2)

1.0Output Set: Interval

1, Step 100
Contour: X Translation

(b) Horizontal displacement contour map (Case3)Unit(m)

Unit(m)
1.

1.0

1.

1.0

0.9

0.9

0.9

0.9

0.8

0.8

0.8

0.8

0.7

0.7

0.7

0.7

0.6

0.6

0.6

0.6

0.5

0.5

0.5

0.5

0.4

0.4

0.4

0.4

0.3

0.3

0.3

0.3

0.2

0.2

0.2

0.2

0.1

0.1

0.1

0.1

0.

0.0

Output Set: Interval 1, Step


Contour: Y Translation

(c) Vertical displacement contour map (Case2)

Output Set: Interval 1, Step 100


Contour: Y Translation

Unit(m)

(d) Vertical displacement contour map (Case3))

Fig. 5: Horizontal and vertical displacement contour map

127

0.

0.0
Unit(m)

0
Case1
Case2
Case3
Case4

Depth from the Ground(m)

-10

Open 3.
Sliding2.
Close

Y
Z

1.

0.

-20

Tunnel location

-30
-40
-50
-60
-70
-80

(a) Case2

-90
0

20

40
60
80
Relative Displacement(cm)

100

Fig. 7: Relative displacement along the crashed


plane where the base offset is applied
0.7
0.6
3.

Vertical displacement (m)

Open 2.
Sliding
Close 1.

0.

(b) Case3

Location of
Crashed Planes

0.5
0.4
0.3

Case3

Case2

0.2
0.1
0
-0.1

200

400

600

Distance(m)
3.

Y
Z

Fig. 8: Vertical displacement along tunnel axis

Open 2.
Sliding1.
Close

consideration is needed for studying the cases


using nonlinear soil condition.
Fig. 8 shows the longitudinal displacement
of the tunnel of Case2 and 3. The lateral axis
shows the distance from the left edge of the
model. As for Case 2, the big displacement
jump occurs at the crashed plane where the
base offset is applied. But on the contrary, as
for Case 3, the displacement jump occurs at the
left area from the offset crashed plane.

0.

(c) case4
Fig. 6: Status of joint elements for crashed planes

The previous study by Bray et al (1992), shows


the fault rupture propagation can reach the
ground surface when the amount of the base
offset is over 3 to 6% of the depth of the deposit
layer. These research results show good
agreement to the result of Case 1 and 2 and not
to the result of Case 3 and 4. This is because
the previous studies assume the linear soil
property, but the Case3 and 4 assume
nonlinear soil behavior. Therefore more

INFLUENCE TO TUNNEL STRUCTURE

Using the ground displacement analyzed


above, the influence to the tunnel structure due
to the ground displacement caused by fault

128

maximum moment is observed at the left side of


the offset plane. These observations and
tendencies are similar to the vertical
displacement of the tunnel.
In any case. The relative displacement
along the offset plane is absorbed below the
tunnel position so that only a little influence can
be observed due to the relative displacement
caused by the fault activities.

4500
3500

Shear force (kN)

Case3

Location of
Crashed
Planes

2500
1500
500
-500

200

400

600

-1500

CONCLUSION

-2500

This paper studied about the influence to


shield tunnel structure due to ground slip, crack,
and deformation propagation in sedimentary
layer caused by fault activities. In order to study
the ground displacement effect, linear and nonlinear FEM analyses are performed using joint
elements for crashed planes observed by the
reflection test.
In the case where linear solid elements are
used
for
soil
modeling,
the
relative
displacement is concentrated to the crashed
plane which is modeled by the joint model. On
the contrary, in the case where non-linear solid
elements are used, the displacement due to the
base offset is widely absorbed in the soil.
In this study, the tunnel is existed relatively
very shallow in the ground and the deep deposit
layer is existed below the tunnel. Therefore, the
offset displacement is absorbed in the relatively
bottom part of the soil so that only very little
influence is observed to the tunnel.
Based on the result of this study, it can be
concluded that the influence to the tunnel
structure is very small even though the tunnel
crosses the fault line where the deposit layer is
thick, the tunnel is existed in the shallow depth.

Case2

Shear capacity

-3500
-4500

Distance(m)

Fig. 9: Shear force acted to the tunnel and the


strength
20000

Location of
Crashed
Planes

15000

Moment(kNm)

10000
5000
0
-5000

-10000
-15000

200

400

600

Case3

Moment capacity

Case2

-20000
Distance(m)

Fig. 10: Bending moment acted to the tunnel


and the strength

activities is studied by FEM analysis using


beam element with equivalent tunnel stiffness.
Fig. 9 and 10 show the results of Case 2 and 3,
respectively. The origin of the lateral axis is the
left edge of the modeled area.
Fig. 9 shows that the huge shear force is
observed around the crashed plane where the
base offset is applied. Not only Case2 but also
Case3 shows the same result. However the
shear capacity of the joint used for this segment
is big enough compared to the shear force.
Fig.10 show that the bending moment is
also big compared around the crashed plane
where the base offset is applied. As for Case2,
the maximum moment is observed at the right
side of the offset plane. On the contrary, the

REFERENCES
Adachi Y., Yoshimura S., and Nakata T. (2003)
Influence to underground structure by ground
displacement, Proc. of 12th Japan earthquake
engineering symposium, JEES (In Japanese)
Azuma H. and Kiyomiya O. (2001) Reliability
evaluation of sea bed pipeline suffered from
base offset caused by fault, Proc. of 2nd
improvement of seismic disaster prevention
symposium by failure process analysis, JSCE (In
Japanese)
Bray J.D., Seed R.B. and Seed H.B. (1994):
Analysis of earthquake fault rupture propagation

129

through cohesive soil, Journal of geotechnical


engineering, ASCE, 120,3, pp562-580

Sato Y., Abe K., Okada Y., Simazaki K., and Suzuki
Y. (1989) Relation between earthquake
magnitude and fault parameters, Fault
parameters handbook of Japan, Vol.2 2nd
chapter pp82-92, Kajima books

Cole, D. A.., Jr., and Lade, P. V. (1984) Influence


zones in alluvium over dip-slip faults, Journal of
geotechnical engineering, ASCE, Proc. Paper
18788, Vol. 110, No. GT5, pp599-615

Takada S., Nemat H., and Fukuda K. (2001)


Propose of simplified design method for buried
pipes crossing earthquake fault, Journal of
structural engineering, JSCE, No.668, -54,
pp187-194 (In Japanese)

Donald L., et al.(1994) New empirical relationships


among magnitude, rupture length, rupture width,
rupture area, and surface displacement, Bulletin
of the seismological society of America, Vol.84,
No.4, pp974-1002

Takemura M. (1998) Scaling rule of crust type


earthquake in Japan, Earthquake, Vol.2 No.51,
pp221-228, SSJ (In Japanese)

Freeman T., GILLON M., BERRYMAN K.,


MORIWAKI Y., SOMERVILLE P. and MEJIA L.
(2000) Matahina dam-fault surface displacement
design criteria, CD-ROM of the twelfth world
conference on earthquake engineering, Auckland
NZ

Tani K. and Ueda K. (1991) Location and shape of


discontinuous plane in sandy layer by base fault
displacement, Proc. of 26th geotechnical
engineers meeting,
pp1185-1188 JSCE (In
Japanese)

Irikura K. and Miyake H. (2001) Prediction of strong


motion based on scenario earthquake,
Earthquake, Vol. 110No. 6849-875, SSJ (In
Japanese)

Taniyama T. and Watanabe H. (1998) Deformation


propagation in sandy surface layer caused by
reverse fault activities, Journal of Geotechnical
Engineering, JSCE, 591/ -43, pp313-325 (In
Japanese)

Kennedy, R.P., Chow, A.W. and William, R.A. (1977)


Fault movement effects on buried oil pipeline,
Transportation engineering J., ASCE, Vol. 103,
No.TES, pp617-633
Konagai K. and Johanson J. (2001) Lagrangian
particles for modeling large soil deformations,
Proc. of a workshop on seismic fault-induced
failures, pp99-106
Matsuda T. (1975) Magnitude and return period of
the earthquakes caused by active faults,
Earthquake, Vol.2 No.28, pp269-283, SSJ (In
Japansese)
Newmark N. W. and Hall W. J. (1975) Pipeline
design to resist large fault displacement,
Proceedings of the U.S. National Conference on
Earthquake Engineering-1975, EERI, pp416-425
Onizuka N., Hakuno N., Iwashita K., and Suzuki T.
(1999) Stress and strain propagation due to
base displacement caused by reverse fault
activities, Journal of Applied Science, JSCE, 2,
pp533-542 (In Japanese)
Ramancharla P. K. and Megro K. (2001) Non-linear
numerical modeling of dip-slip faults for studying
ground surface deformation, Proc. of 2nd
improvement of seismic disaster prevention
symposium by failure process analysis, JSCE
Ramancharla P. K. and Megro K. (2001) Applied
simulation of non-linear behavior of dip-slip faults
for studying ground surface deformation, Proc.
of a workshop on seismic fault-induced failures,
JSPS, pp.109-114

130

Shallow Foundations and SoilStructure


Interaction

Study on Applicability of Seismic-isolation Foundation


to Railway Structures
X. Luo1, T. Miyamoto1, T. Imamura1
1

Railway Technical Research Institute, Japan

Abstract
Since seismic-isolation foundations generate relatively large displacements adversely
affecting the running safety of train, it is an important task to assess this influence while
design railway structures with seismic-isolation foundations. In this study, bridges and
viaducts are taken as the objects for assessment of influence due to different types of pile
head connections upon running safety of train during earthquakes. As a result, the
influence of seismic-isolation foundation upon running safety of train is well grasped through
structural dynamic analysis, vehicle running simulation and Running-Safety Assessment
(RSA) based on Spectral Intensity (SI).

INTRODUCTION

After the Hyogoken-Nunbu Earthquake of


Jan. 17, 1995, a new code, Seismic Design
Code for Railway Structures (the Railway Code,
in Japanese) (Railway Technical Research
Institute, 1999) was published. Some new
thought for seismic design have been adopted
by drawing the lesson of the earthquake.
According to the code, the ultimate bending
capacity of pile should be assured during the
intensive Level 2 (L2) earthquakes, which
makes the increase of cross-section and rebar
of the piles. To solve this problem and make an
economical
design,
seismic-isolation
foundations adopted in design is an effective
method.
Though the seismic-isolation
foundations can reduce the action by
earthquake vibration to a structure, on the other
hand, the energy absorption causes big
response displacement of the structure
adversely effecting the running safety of train.
Therefore, how to assess the degree of this
influence becomes a critical problem for design
of railway structures with seismic-isolation
foundations.
In this study, bridges and viaducts are taken
as the objects for assessment of the influence
due to different types of pile head connections
(rigid connection, semi-rigid connection and

131

pinned connection) upon running safety of train


during earthquakes. As a result, the influence
has been well grasped through structural
dynamic analysis, vehicle running simulation
and running-safety evaluation based on
Spectral Intensity (SI).
PUSHOVER ANALYSIS

To evaluate the seismic performance of the


structures with different pile head connections,
a static, non-linear procedure called pushover
analysis was applied to Shinkansen structures.
The structures with rigid pile head connection
are shown in Fig.1 (bridge, Case No.B1), and
Fig.2 (viaduct, Case No.V1). With regard to the
cross sections, the pile length for the bridge is
19.5m (Fig.1) and the pile length for the viaduct
is 25.5m (Fig.2). The surface ground is mainly
constituted by sandy soil and clay. Based on
these cross sections, the pile head connections
were modified with semi-rigid connection and
pinned connection through an optimization
design. As a result, the diameters of the piles
are decreased 10% to 15% compared with the
cases of rigid connection, which proves the
effects of seismic-isolation. The combination of
the structures and the pile head connections,
the decreased diameters are shown in Table 1.
According to the Railway Code (Railway

11700

2454

Seismic coefficient (kh)

Surface
layer

Sandy
soil

Sandy
soil

Clay

-20

3250
3250 1500
9500

-10

Clay

-20

Sandy
soil
Gravel

Cast-in-place RC pile (5 staggered


piles), =13001500

Depth (m)

Depth (m)

-10

21500

19500
1500

Surface
layer

7054

4600

2000

2500

Pulling-out side
pile head yielding

500

4500

2500

50

N Value

0.4

Response
displ. 360mm
Response displ. 435mm
Pulling-out side pile middle yielding

0.2

No.B2(Semi-rigid),Teq=0.765s

No.B3(Pinned),Teq=0.998s

500

Surface
layer

Clay

Clay

Depth (m)
Sandy
soil

-20

300

400

Response displ. 470.4mm


Pullingout side
pile head
yielding

0.6

-10

-10

200

500

Fig.3: Load-displacement relationship of bridge

Seismic coefficient (kh)

Surface
layer

Depth (m)

800

8000

900

100

Displacement at crest () (mm)

2000
3800

0.6

0.7

25500 1400 500


26900

800

0.8

0.0

11300

4500

Pier base yielding

Vs (m/s)

Reaching limit
of seismic
performance II

No.B1(Rigid),Teq=0.695s

Sandy
soil
Gravel

Fig.1: Schematic illustration of bridge with ground


condition

900

Response displ. 297.5mm

1.0

Sandy
soil

-20

0.5
0.4

Response displ. 430mm


Upper part
of column
yielding

Upper
part of
column
yielding

Reaching limit
of seismic
performance II

Response displ. 639.4mm


Pulling-out side pile
middle yielding
Upper part of column & Pulling-out
side pile head yielding

0.3
0.2

No.V3(Pinned),Teq=1.494s
No.V2(Semi-rigid),Teq=1.143s

0.1

No.V1(Rigid),Teq=1.081s

0.0

5400
Cast-in-place RC pile (one column
for one pile), =10001100

-30

Gravel

50

N Value

-30

Gravel

500

Vs (m/s)

Pile head
connection

No.B1
Rigid
No.B2 Bridge
Semi-rigid
No.B3
Pinned
No.V1
Rigid
Viaduct
No.V2
Semi-rigid
(Rigid frame)
No.V3
Pinned

200

300

400

500 600
Displacement at crest () (mm)

700

deformation characteristics of semi-rigid pile


head connection used in the analysis were
based on the reference (Kouda et al, 2005), in
which the concrete strength for the pile head is
equivalently increased due to raised restrain
effects of tense hoops distributed within the pile
head. For Case No.B3 and Case No.V3, the
deformation characteristics of pinned pile head
connection were based on a proposed model
(Tazoh et al, 2001), in which the loaddisplacement relationship depends on the
friction factor of metal.
In Fig.3, it is found that the yield seismic
coefficients of the cases with semi-rigid and
pinned pile head connections are about 10% to
20% lower than the case with rigid connection.
Also,
the
yield
seismic
coefficients
corresponding to the semi-rigid and pinned pile
head connections are larger than the seismic
coefficient 0.35 (G3 site classification) due to
response of Level 1 (L1) design earthquake
motion, which means the structures have

Table 1: Combination of structures and pile head


connections

Structure
type

100

Fig.4: Load-displacement relationship of viaduct

Fig.2 : Schematic illustration of viaduct with ground


condition

Case

Diameter
of pile (mm)
1500
1300
1300
1100
1000
1000

Technical Research Institute), the equivalent


natural periods of the structures should be
calculated by using the pushover analysis
method.
Given in Fig.3 and Fig.4 are the calculated
relationships between the seismic load and the
displacement at crest of the structures shown in
Table 1. For Case No.B2 and Case No.V2, the

132

Rigid pile head connection


Semi-rigid pile head
connection
Pined pile head connection

Response Velocity
Sv (mm/s)

Teq : Equivalent
natural period
of structure

SI =

Car body

Lateral stopper

Calculation
of limit SI

Wheel
Truck

Damper

Rail

2.5

S (h,T )dT
0.1 v
Period (sec)

0.1

70mm
Position for
judgment of
derailment

Spring

2.5

Calculation of
response SI

Spectral Intensity (SI) (mm)

Train running simulation

Response acceleration at
crest of bridge

10000
8000

Rigid

Semi-rigid

Pinned

Semi-rigid

Pinned

6000
4000

Rigid

2000
0
0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

Equivalent natural period (Teq) (sec)

Bedrock motion

Fig.5: Procedure for influence assessment of seismic-isolation foundation upon running safety of train

heavy casualty. Therefore, RSA of vehicle


must be conducted in design of railway
structures, which is a unique and complicated
problem in the field of railway dynamics.
With regard to the complicated response of
the vehicles, the index SI is proved suitable for
the RSA based on the concept of energy
balance between the input motions and the
response of vehicles (Luo, 2002, 2005). Given
in Fig.5 is a procedure for RSA with the index SI
against transverse vibration caused by
earthquakes.
Firstly,
the
response
accelerations at the crest of the structures were
calculated by inputting the bedrock motions as
shown in Table 2. In the calculation the soil-pile
interaction was taken into account. Secondly,
the response spectra of velocity were
calculated and the response SI values were
obtained by integrating the velocity from 0.1sec
to 2.5sec. Moreover, the running safety limits
of SI were calculated by vehicle running
simulations. Finally, the RSA of the vehicles
was carried out by comparing the response SI

enough capacity satisfying the seismic


performance I against L1 earthquake.
Compared with the equivalent natural period of
structure of Case No.B1 (rigid, Teq=0.695s), the
period Teq for Case No.B2 (semi-rigid)
increases about 10%, and that for Case No.B3
(pinned) increases about 40%.
The characteristics of the load-displacement
relationship of viaducts shown in Fig.4 are
similar to those of bridges shown in Fig.3.
However, the yield seismic coefficients of the
viaducts are about 20% to 30% smaller than the
corresponding cases of bridges because of
lower stiffness for the viaducts. Also, the yield
seismic coefficients corresponding to the semirigid and pinned pile head connections are
larger than the seismic coefficient 0.4 (G5 site
classification) due to response of L1 design
earthquake motion. Therefore, the structures
designed satisfy the seismic performance
criteria according to the Railway Code. Since
the soft surface ground and the rather low
structural stiffness, the equivalent natural
periods of the viaducts have increased about
50% compared to the bridges with same type of
pile head connection.

Table 2: Equivalent natural period and input


bedrock motions
Equivalent natural Design earthquake
Case
period (Teq )Sec motion on bedrock
No.B1
0.695
L1
No.B2
0.765
1.5 times of L1
No.B3
0.998
0.5 times of Spc.II
No.V1
1.081
Spc.I (L2)
No.V2
1.143
Spc.II (L2)
No.V3
1.494

ASSESSMENT OF RUNNING SAFETY OF TRAIN

Under intense ground shaking, running


railway vehicles might be rolled to induce a
large displacement resulting in derailment
and/or overturning which occasionally causes

133

Spectral Intensity (SI ) (mm)

10000

No.B2
8000

No.V1
No.B3

No.B1

in Fig.7 and Fig.8. The Figures show that the


running safety of the seismic-isolation
foundation can be assured against L1
earthquake, but there may be problems for
earthquakes over LI.

L1

No.V2

No.V3

Spc.II
(L2)
Spc.I
(L2)

6000

1.5 times
of L1
0.5 times
of Spc.II

4000

Limit SI stipulated in the Code

3000

CONCLUSIONS

0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6

The good cost performance of the non-rigid


pile
head connections
(seismic-isolation
foundations) has been proven and their RSA
can be conducted as the same as the common
rigid ones.

Equivalent natural period (Teq ) (sec)

Fig.6: Comparison of limit SI between those


obtained in this study and those stipulated in
the Displacement Limited Code

Spectral Intensity (SI) (mm)

with the limit SI.


Given in Fig.6 are the calculated results of
limit SI corresponding to the cases in Table 2.
Moreover, for comparison a line of limit SI for
common structures stipulated in a code called
Displacement
Limited
Code
(Railway
Technical Research Institute, 2006) was plotted
together.
This limit line was defined by
enveloping the limits (gray lines) for common
structures based on 11 representative
earthquake motions.
From Fig.6 it is
understood that even though the seismicisolation foundations generate quite large
displacement, the values of limit SI against
seismic vibration are higher than that for
common structures stipulated in the code.
Therefore, when designing foundations with
non-rigid pile head connections, the RSA can
be conducted in the same way as that for
common rigid ones. The results of RSA against
L1 and over L1 earthquake motions are shown

REFERENCES
Kouda, M., Hamada, Y., Sando, T., and Aoki, H.
(2005) Modeling Deformation Characteristics of
RC Pile Head with Tense Hoops, Proceedings
of the 60th Annual Meeting of Japan Society of
Civil Engineers, pp.961-962 (in Japanese)
Luo, X. (2002) A Code-Type Provision for Running
Safety Assessment of Train Undergone
Earthquake Motions, Proc., 12th European
Conference on Earthquake Engineering, CD
Version, Paper Reference 462, London, U.K.
Luo, X. (2005) Study on Methodology for Running
Safety Assessment of Trains in Seismic Design
of Railway Structures, Journal Soil Dynamics
and Earthquake Engineering; Vol. 25, No.2,
Elsevier Science Ltd., pp.79-91
Railway Technical Research Institute (1999)
Seismic Design Code for Railway Structures,
published by MARUZEN (in Japanese)
Railway Technical Research Institute (2006)
Displacement Limited Code for Railway
Structures, published by MARUZEN (in
Japanese)
Tazoh, T., Ohtsuki, A., Aoki, T., Mano, H., Isoda, K.,
Iwamoto, N., Ishihara, T., Ohkawa, M. (2001)
Developing a New Method of Pile Head
Connection for Decreasing Construction Cost
and
Increasing
Seismic
Performance,
Proceedings of the 26th JSCE Earthquake
Engineering
Symposium,
pp.881-884
(in
Japanese)

10000
8000

Limit SI stipulated in the Code

6000
4000
2000

No.B1 No.B2

No.V2
No.V3
No.V1
Response SI
No.B3

0
0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6

Equivalent natural period (Teq) (sec)


Spectral Intensity (SI) (mm)

Fig.7: RSA against L1 design earthquake motion


10000
8000
6000

Limit SI
stipulated in
the Code

Limit SI
stipulated in
the Code

4000
2000
0
0.5

1.0

Teq (sec)

1.5 0.5

1.0

1.5 0.5

Teq (sec)

Limit SI
stipulated in
the Code
1.0
1.5 0.5

Teq (sec)

Limit SI
stipulated in
the Code
1.0
1.5

Teq (sec)

(a) 1.5 times of L1, (b) 0.5 times of Spc.II, (c) Spc.I (L2), (d) Spc.II (L2)
Fig.8: RSA against earthquake motions over L1

134

A Study on Dynamic Soil-structure Interaction Effect Based on


Microtremor Measurement of Building and Surrounding Ground Surface
M. IIBA 1, M. Watakabe2, A. Fujii3, S. Koyama1, S. Sakai4 and M. Yasui2
1

Building Research Institute, Japan


2
Toda Corporation, Japan
3
Konoike Construction, Japan
4
Hazama Corporation, Japan

Abstract
Soil structure interaction (SSI) effects are investigated based on a microtremor measurement. The
instruments were set in a 7-storied residential building and on the ground to evaluate the sway and rocking
vibrations. The building constructed by HPC prefabricated method has a flamed structure in longitudinal
direction and a walled structure in transverse direction. Through transfer functions of buildings, predominant
frequencies under SSI model and those under based-fixed condition are calculated. The SSI effect is
remarkable in transverse direction due to predominant rocking effect. Through the random decrement
technique, the damping factor of buildings is obtained. It is founded that the damping factors are around 5 to
7% under the microtremor level.

INTRODUCTION

negligible (Iiba et al., 2002).


For
the
performance-based
design,
behaviors of buildings are clarified more
precisely based on data of earthquake motion
observation. But in general, the earthquake
motion observation is not popular, especially,
for purpose of the SSI phenomena.
A simplified method for incorporating SSI
effects and for calculating spring constants is
presented, to evaluate predominant frequency
of buildings with SSI. The SSI effects of
building are investigated based on a
microtremor measurement on a residential
building (Iiba et al., 2004). Through transfer
functions of the building, predominant
frequencies are calculated.
To investigate the applicability of the
simplified method, results by the simplified
method are compared with those by
microtremor measurement.

Phenomena that foundation displacements


occur due to base shear and overturning
moment is called soil structure interaction
(SSI). The SSI phenomena are the inertial soil
structure interaction. On the other hand, the
effect of SSI related to seismic input motions to
the buildings is called the kinematic soil
structure interaction. As a result of the
displacement of foundations due to the inertial
SSI, characteristics of superstructure are
changed as follows;
a) Elongation of natural period (compared
with base fixed condition)
b) Change of damping factor (compared with
base fixed condition)
The Building Standard Law in Japan and its
related enforcement and notices were revised
for the direction to the performance-based
design from 1998 (Midorikawa et al., 2000).
The calculation method of response and limit
strength was provided for checking structural
serviceability and safety of buildings. In the
calculation, SSI effects should be considered
when the interaction effect would be not

EVALUATION OF PREDOMINANT PERIOD OF


SSI SYSTEM

The
SSI
model
consisting
of
a
superstructure with a mass, a base and sway

135

and rocking springs is illustrated in Fig. 1.


Through an assumption that the influence of
foundation mass and moment of inertia at each
floor on response of SSI system is negligible
(Bielak, 1976), an external force acting to
springs of superstructure, sway and rocking is
only the inertial force of superstructure, as
shown in Fig. 2. In the case that the springs are
arranged in a series, an equivalent spring
constant (Ke) for the SSI system can be
obtained as follows (AIJ, 1996);
H2
1
1
1
(1)
=
+
+
Ke Kb K h K r
Where Kb, kh and Kr are spring constants for a
superstructure, sway motion and rocking
motion, respectively. Circular frequencies are
obtained by following equations;
K
K
Kr
2
(2)
b = b , h 2 = h , r 2 =
m
m
mH 2
Where m is the mass of superstructure, and b,
h
and
r are
circular
frequencies
corresponding to spring constants of each
motion. Based on the relationship of Te=2/e,
Tb=2 /b, Th=2 /h and Tr=2 /r, the
predominant period of the SSI system is
obtained as follows;
Te = Tb + Th + Tr
2

In the same way, an equivalent damping


factor of the SSI system (he) is estimated by
following equation.
3

hb =

(5b)

K r = K r + iK r ' = K r (1 + i 2hr ' ) = K r + icr (6b)


The
impedance
has
a
frequency
dependency. Considering the convenience to
incorporating into design, the value at rest
(frequency is zero) will be used to obtain spring
constant.
(7)
Spring constant: K = K ( = 0 )
To calculate static spring constants for
multiple layers of ground, a cone model is
applied, as drawn in Fig. 3(Wolf, 1994 and Iiba
et al., 2002). The spring constants (Khb, Krb) for

Kb
Cb
Ch

Kr

1 cb
2b m

For the sway and rocking motions, dynamic


impedances with a complex are expressed as
follows;
K h = K h + iK h ' = K h (1 + i 2hh ' ) = K h + ich (6a)

(3)

Kh

Cr

Imaginary Summit

Sway
and Rocking Model
Fig. 1: Sway-rocking model

Radius of Foundation :ro

Ground Surface
Shear Modulus

Building
Rocking Displ. Displ. ub

G1

uh

Z0
Z1

G2

Z2

ur
Sway Displ.

EVALUATION OF SPRING CONSTANTS FOR


SPREAD FOUNDATION

T
T
T
(4)
he = hb b + hh h + hr r
Te
Te
Te
Where hb, hh and hr are equivalent damping
factors. The hh and hr are calculated by
equation (5a). The hb is expressed by cb which
is the viscous damping coefficient of the
superstructure.

K '
(5a)
h = sin 0.5 tan 1
K

Inertial Force

Gi

Zi

Sum. of Displacement

Zn-2

= uh + ur + ub
=F/Kh + FH2/Kr + F/Kb
=F (1/Kh + H2/Kr + 1/Kb)

Gn-1
Gn

Zn-1
Engineering Bedrock

Fig. 3: Cone model and ground condition

Fig. 2: Force vs. displacement in simple SSI system

136

stiffness and Poissons ratio, and distance from


the cone summit for rocking motion of i-th layer
of ground. And rr0 is the equivalent radius for
rocking spring constant.
The soil property is expressed as follows;
(20)
Gi ' = Gi (1 + i 2hi )
Where hi is the damping factor of soil. The
Khb and Krb in Equations (8) and (9) is complex,
when using the complex shear modulus of soils
described
in
Equation
(20).
Dynamic
impedances are dependent on a type of
foundations, their dimension and properties of
soil. As to nonlinearity of the soil, there are two
kinds of nonlinearity. One is based on
nonlinearity of soil when seismic wave comes
up in the soil deposits. The shear stiffness and
damping factor are dependent on the shear
strain of the soil. The other is the nonlinearity
related to contact face between foundation or
pile and soil, so called, local nonlinearity. In the
calculation, only the former, that is, nonlinearity
of the soil property is considered.

sway and rocking motions at bottom for spread


foundation are expressed as follows;
K hb = h K1hb

(8)

K rb = r K1rb
1
h = n
1

i=1
hi
8G1rh 0
K1hb =
2 1

(9)
(10)

(11)

Z Z

hi hi 1
hi = i
(
G
Z
Z
1 h 0 hi Z hi 1 )

i = 1,2, , n 1

G Z

hn = n hn 1
G1 Z h 0
1
r = n
1

i =1
ri

(12a)
(12b)
(13)

K 1rb =

8 G1 rr 0
3 (1 1 )

E Z

(14)
4

Z Z

EVALUATION OF SPRING CONSTANTS FOR


PILE FOUNDATION

r 0 ri
ri = i ri 1
3
3
E
Z
Z
Z
1 r 0 ri 1 ri Z ri 1

i = 1,2, , n 1

As it is conformed that the sway spring


constant of a pile foundation is almost the same
as that of the spread foundation which has the
same dimension of plan configuration (Iiba et
al., 2002), the horizontal spring constant at pile
head is calculated by equation (8). On the other
hand, the rocking spring constant of the pile
foundation is remarkably larger than that of the
spread foundation.
The vertical spring of the pile consists of
springs of friction on pile surface and end
bearing at the pile tip. The spring constant of
friction per unit length (S) is estimated by
following equation (Randolf and Wroth, 1978).

(15a)

E Z
rn = n rn 1
E1 Z r 0
2 1 ,
9
2
Z h 0 = rh 0
Z r 0 = 1 1 rr 0
8
16
Ei = 2(1 + i )Gi

rh 0 = B D
rr 0 = 4

B D
3

(15b)
(16)
(17)
(18)

(19)

Where h is a modification factor for sway


and k1hb is the spring constant for rigid
foundation on semi-infinite uniform layer with
soil property of 1st layer. The 1 is a Poissons
ratio of ground under the foundation. Gi and Zhi
are the shear modulus of i-th layer of ground
and the distance of lower boundary from the
cone summit of i-th layer. And rh0, B and D are
an equivalent radius for the sway spring
constant, length and width of foundation. And r
is the modification factor for rocking and k1rb are
the rocking spring constant for rigid foundation
with semi-infinite uniform layer with soil
property of 1st layer. Ei, i and Zri are the elastic

Sv =

2Ge
log e (2rm / B)

(21)

rm = 2.5 L(1 e )
Ge =

(22)
n

1
1
Gi di e = i di

L i=1
L i=1

(23)

Where L and B= (2Ro) are pile length and


diameter, and Ge and e are an averaged shear
stiffness and Poissons ratio along the pile
length. The di is a height of i-th layer.
The spring constant at pile tip (kb) is

137

The building, which is constructed by HPC


prefabricated method, is composed of girders of
reinforced concrete (RC) structure, columns of
steel RC structure and floors and walls of precast RC boards. The building is supported by
the individual foundation with pre-stressed
concrete piles of 0.35m in diameter. Another
same building is constructed with distance of
about 70 m from the building.
The soil condition at the site is drawn in Fig.
5. The sandy and clayey layers are laminated.
The microtremors in the building and on the
ground surface are measured. Periods of
measurement are 600 or 500 s and 7 or 8 sets
of these periods are recorded. The interval of
records is 0.005 s.
Measuring points are shown in Fig. 4. Three
horizontal and two vertical sensors on the roof
and 1st floor, one horizontal one on the 4th
floor are set in the building. Three horizontal
sensors are installed on the ground surface
with 13m at both sides and 26m at one side far
from the building. The sensor arrangements are
the same in both buildings.
Types of sensors are velocity transducers
with servo type and those with moving-coil type
are used. The unit of measured data is velocity.
The vibration system with surrounding
grounds is assumed to be the sway and rocking
model shown in Fig. 6. Following systems are
considered, based on Fig.6 (Stewart and
Fenves, 1998).

estimated by following;
3 Gb Ro
(24)
kb =
8 1 b
Where Gb and b are the shear stiffness and
Poissons ratio of the engineering bedrock,
respectively. To combine two spring constants,
the vertical stiffness of pile is obtained (Masuda
et al., 1993).
E p Ap (1 e 2 L ) + k b (1 + e 2 L )
(25)
K v = E p Ap
E p Ap (1 + e 2 L ) + k b (1 e 2 L )
Where there is 2 (=Sv/EpAp), and Ep and Ap are
an elastic modulus and a cross-sectional area
of pile.
The rocking spring constant of a pile group
is expressed through the summation of all pile
(m). The rocking spring constants for x and y
axes are as follows.
m

i =1

i =1

K R x = K v y i 2 K R y = K v xi 2

(26)

MICROTREMOR MEASUREMENT

A plan at ordinary floor and a section in the


transverse direction of a building are drawn in
Fig. 4. The building, which has 7-storied
residential one with central corridor at
longitudinal direction, is constructed in a central
area of Tsukuba, Japan. There are three and
nine spans in transverse and longitudinal
directions, respectively. A ratio of transverse to
longitudinal width is 1:3.2 and ratios of height to
width in both directions are 1.47 and 0.64.

V alue

D epth
(m )
4,600

4,600

4,600

4,600

4,750

4,600

4,600

4,600

S oil
type

M ark

10

20

30

40

50

4,600

5,400

L o am
C lay

2,100

6.6

S and

5,400

11.9
13.0

10

C lay

15

20

RF

S and
25

7F

2,70

2,775

Plan

27.6
29.1

5F

34.0
35.4
36.1

2,70

Sensors

2,700

6F

Section

Vertical

30

C lay

35

S an d
C lay
w ith
S an d

3F

2,700 2,700

Horizontal

2,70

4F

C lay

S an d
w ith
C lay

40

41.5

2F

45

G ravel

1F

50.4

50

Fig. 5: Ground condition with soil types and


standard penetration values (N-values)

Fig. 4: Plan and section of building and


arrangement of sensors

138

System

Input

Output
(response)

SRB
ug
ug+uf +H+uB
RB
ug+uf
B
ug+uf +H
ug : Free Surface Ground Motion
uf: Input Loss due to Foundation
uf : Sway
uf: 1st Floor Motion Relative to Free Ground
(herein assumption to uf = uf + uf )
uB: Building Response Relative to 1st Floor
H: Equivalent Height
: Rocking angle at 1st Floor
1
2
3

u
u

H
uBB
H u

HH

uuu
f'
g +u

Surface Ground

Rigid Foundation
without Mass

uu
g

uuf' uuf''

Fig. 6: Displacement of ground surface, sway, rocking and building, and definition of transfer functions

Amplitude

13
12
11
10
9
8
7
6
5
4
3
2
1
0

-building
Transverse

2.87H z

3.25H z
4.14H z

A-building
Longitudinal

Amplitude

13
12
11
10
9
8
7
6
5
4
3
2
1
0

3.35H z
2.63H z

3.33H z

Frequency(Hz)
180

180
A-building
Transverse

135
90
45
0
-45

-90

-135

A-building
Longitudinal

135

Phase Delay (deg.)

Phase Delay (deg.)

Frequency(Hz)

90
45
0
-45

-90

-135

-180

-180
1

Frequency(Hz)
A-building
Transverse

1.0

Frequency(Hz)
A-building
Longitudinal

1.0

Coherence

0.8

Coherence

0.8
0.6

0.6

0.4

0.4

0.2

0.2

0.0

0.0
1

Frequency(Hz)

Frequency(Hz)

Fig. 7: Transfer functions and coherence of SRB-, RB- and B-systems

139

a) Sway and rocking model (SBR-system;


uT/ug)
b) Rocking model (RB-system; uT/(ug + uf) )
c) Building model with base fixed (B-system;
uT/(ug + uf + ur))
Where uT is a total response at top of
building including the ground (ug), sway (uf)
rocking (ur) and building responses (uB). ur is
product of the rocking angle at 1st floor and
height of building.
In case of focusing on the 1st mode, the
equivalent height of building should be used in
calculation of the transfer functions (Naito, et al.,
2003). In this analysis, the total height of
building is used.
Transfer functions and coherence of SRB-,
RB- and B-systems of the A-building are shown
in Fig. 7. The transfer functions are obtained as
follows;
Fourier spectral ratios with 80s in time
length and 40s in running time are calculated.
The number of running Fourier spectral ratios
are 84 and 99 in transverse and longitudinal
directions, respectively. After the average of the
running Fourier spectral ratios, the transfer
functions are obtained with the hanning
windows of 8 times.
The fundamental predominant frequencies
of SRB-, RB- and B-systems in transverse
directions are 2.87, 3.25 and 4.14Hz,
respectively. The reduction of predominant
frequencies occurs when including the sway
and
rocking
motions.
The
amplitude
(amplification factor) at predominant frequency
of SRB-system is the largest, while the
amplitude of B-system is lowest.
By the half power technique, the damping
factors of SRB-, RB- and B-systems are 5.37,
9.24 and 7.63%, respectively.
As to phase characteristics, the phase of
building top in SRB-, and B-systems at the
predominant frequency are faster (the same
result is shown in B-system of longitudinal
direction). Through the detail check for
phase characteristics in SRB-system, about
half of them are faster and the others are
delay. The phase characteristics are not
constant. The average phases are fast or
delay occasionally. In case of RB-system,
the delay cases are more than fast ones.
In longitudinal direction, the fundamental
predominant frequencies of SRB-, RB- and
B-systems are 2.63, 3.33 and 3.35Hz,

respectively. The amplitudes at predominant


frequencies of each system are almost similar.
The damping factors of SRB-, RB- and Bsystems are 8.69, 11.0 and 5.61%, respectively.
As to the coherences, RB- and B-systems
shows relatively large values. In the SRBsystem, since the coherence is less in
frequency range higher than the fundamental
predominant frequency, the input and output
are less corresponding to each other.
The predominant frequencies in various
systems in both directions are summarized in
Table 2. The sway and rocking frequencies are
calculated by using following equation (AlJ,
1996);
2
2
2
1 / f SRB = (1 / f B ) + (1 / f Sway ) + (1 / f Rocking ) (27)
The ratios of predominant frequencies in
SSI system (fSRB) to those at base-fixed
condition (fB) are 0.68-0.69 and 0.75-0.79 in
transverse
and
longitudinal
directions,
respectively. In the transverse direction, ratios
of sway and rocking are 0.22-0.23 and 0.300.31, respectively. In the longitudinal direction,
ratios of sway and rocking are 0.26-0.38 and
0.08-0.18, respectively.
VERIFICATION OF SIMPLIFIED METHOD

Table 3 presents properties of the building.


And equivalent values which contribute to 1st
mode vibration are shown. The spring
constants calculated by Equations 8, 9 and 26
are shown in Table 4. The results of
predominant frequency of the building with SSI
in the transverse direction are also presented in
the Table. After the group of 4 pre-stressed
piles is replaced to a cast-in place RC pile of
Table 2: Predominant frequencies and
ratios to frequency at B-system
Build
ing

A
Transverse
B
A
Longitudinal
B

140

Predominant frequency
of each system (Hz)
Sway

Rocking

4.14

6.12

5.24

(1.00)

(1.48)

(1.27)

3.19

4.11

5.83

5.04

(0.68)

(0.78)

(1.00)

(1.42)

(1.23)

2.63

3.33

3.35

4.29

30.5

(0.79)

(0.99)

(1.00)

(1.25)

(9.10)

2.76

3.21

3.70

5.39

6.47

(0.75)

(0.87)

(1.00)

(1.46)

(1.75)

SRB

RB

2.87

3.25

(0.69)

(0.79)

2.80

0.73m in diameter, the rocking spring is


calculated. The predominant frequency of
building (2.93Hz) is obtained by Equation 3, is
very similar to that in the microtremor
measurement (2.80-2.87Hz) shown in Table 2.
The displacement distribution of the building in
the analysis has a good agreement to that in
the measurement. It is proved that the
simplified method gives appropriate results.

the fundamental predominant frequencies are


calculated by RD (Random Decrement)
technique (Tamura, et al., 1993).
The waveforms of the top of building are
conducted with band-pass filter of 2-4Hz
remained. The waveforms of 5s, when the time
at peak values is adjusted to t=0, are
superposed along 600s (A-building) or 500s (BTable 4: Properties of soil and foundation, and
predominant frequency by simple analysis

DAMPING FACTOR IN SSI SYSTEM

To investigate the damping factor of building,


Table 3: Properties of building and its simple model
Items
Number of story
Dimensions of plan
Total height
Mass of unit area
Mass of each floor
Stiffness of each story
Natural frequency with base
fixed
Equivalent mass at first mode
Equivalent Stiffness at one
degree of freedom model
Equivalent height at first mode

Unit

Values

m
m
t/m2
t
MN/m
Hz

7
41.6 x 12.9
19.0
1.2
6.30x103
205
4.1

t
MN/m

3.60 x103
2.39 x103

15.2

Amplitude(cm/sec)

B-building Transverse
0.000

-0.002
0.5

1.0

1.5

2.0

Time(s)

Values
13.1m for sway spring
9.86m for rocking spring
2.50m
Unit mass
(t/ m3)
1.6

Shear wave
Velocity(m/s)
151

13.0

1.8

260

27.6

1.8

332

41.5

1.7

309

1.9

428

0.002

0.0

Items
Equivalent radius of
foundation
Embedment of
Foundation
Soil
Layer
Depth
Prop(m)
erty
1
6.6

Sway spring constant


Rocking spring constant
Predominant frequency
under SSI without piles
Predominant frequency
under SSI with rocking
fixed
Rocking spring constant
with RC piles (0.73m in
diameter)
Predominant frequency
under SSI with RC piles
Mode
Distribution
ratio
Analysis
Experiment

Fig. 8: Free vibration waveform by RD technique

5.39x103 MN/m
2.59x105 MNm/rad
2.17Hz (0.529: ratio to
base fixed condition)
3.41Hz (0.832: ratio to
base fixed condition)
1.07x106 MNm/rad

2.93Hz (0.714: ratio to


base fixed condition)
Sway Rocking Building
0.226
0.263
0.510
0.210.300.460.23
0.31
0.48

Table 5: Damping factor by RD technique


Direction

Transverse

Longitudinal

Set No.
ATM01
ATM02
ATM03
ATM04
ATM05
ATM06
Ave.
ATL01
ATL02
ATL03
ATL04
ATL05
ATL06
Ave.

A-Building
B-Building
Predominant Damping
Predominant Damping
Number of
Number of
Factor Set No.
Factor
Frequency
Frequency
Superposition
Superposition
(Hz)
(%)
(Hz)
(%)
1801
2.75
6.51
BTM01
1472
3.00
5.81
1796
2.95
5.97
BTM02
1478
2.75
5.04
1779
2.85
5.82
BTM03
1485
2.90
6.17
1771
2.85
5.73
BTM04
1460
2.75
5.04
1777
2.85
5.10
BTM05
1469
2.80
5.96
1796
2.90
6.67
BTM06
1468
2.80
5.32
1787
2.86
5.97
Ave.
1472
2.83
5.56
1785
2.80
5.57
BTL01
1472
3.00
7.48
1772
2.90
5.07
BTL02
1524
2.80
7.75
1792
2.80
7.40
BTL03
1519
2.80
7.45
1833
2.95
7.27
BTL04
1520
2.80
5.99
1833
2.85
7.51
BTL05
1505
2.90
6.16
1823
3.10
6.46
BTL06
1495
2.90
6.31
1806
2.90
6.55
Ave.
1515
2.87
6.86

141

Bielak, J. (1976) Modal analysis for buildingsoil interaction, Journals of Engineering


Mechanical Division, ASCE, Vol. 102, EM5,
pp.771-786
Iiba, M. et al. (2002) Introduction of Soil
Structure Interaction to Performance-based
Type of Provisions in Building Standard Law
of Japan, Proceedings of International
Conference on Advances and New
Challenges in Earthquake Engineering
Research, Harbin, part 2, pp.275-282
Iiba, M. et al. (2004) A Study on Dynamic SoilStructure Interaction Effect Based on
Microtremor Measurement of Building and
Surrounding Ground Surface, Proceeding
of Third US-Japan workshop on SSI, CDROM
Masuda A. et al. (1993) "Simplified Method for
Dynamic
Impedance
of
Pile-group,
Summaries of Technical Papers of Annual
Meeting, AIJ, 339-344 (in Japanese)
Midorikawa, M. et al. (2000) Development of
Seismic
Performance
Evaluation
Procedures in Building Code of Japan,
Proceedings of 12th World Conference of
Earthquake Engineering, Auckland, Paper
No. 2215
Naito, Y., Tanino, T., et al. (2003) System
Identification of Building Alone with the
Effect of Dynamic Soil Structure Interaction
Eliminated Journal of Structural and
Construction EngineeringNo.564pp.3946 (in Japanese)
Randolf M. and Wroth C.P. (1978) "Analysis of
Deformation of Vertically Loaded Piles,
Journals of Geotechnical Engineering
Division, ASCE, Vol.104, No.GT12, 14651488
Stewart, J. and Fenves, G. (1998) System
Identification for Evaluating Soil-structure
Interaction Effects in Buildings from Strong
Motion Recordings, Journals of Earthquake
Engineering and Structural Dynamics, 27,
pp.869-885
Tamura, Y., Sasaki, A. and Tsukagoshi, H.
(1993) Evaluation of Damping Ratios of
Randomly Excited Buildings Using the
Random Decrement Technique Journal
of Structural and Construction Engineering
No.454pp.29-38 (in Japanese)
Wolf J. (1994) Foundation Vibration Analysis
Using Simple Physical Models, PTR
Prentice Hall

building). Obtained waveforms are normalized


to the maximum amplitude at t=0. The
logarithmic damping factors are calculated by
applying the least square technique. An
example of free vibration waveform and
identified curve are drawn in Fig. 8. The
damping factors are summarized in Table 5.
The numbers of superposition are about 1500.
The damping factors in SSI in transverse and
longitudinal directions are 5.0-6.7 and 5.1-7.8%,
respectively. The average damping factors in
each direction are 5.77 and 6.71%. The
damping factor in longitudinal direction is larger
than that in transverse direction.
CONCLUSIONS

The SSI characteristics of building by


microtremor and the verification of simplified
method are summarized as follows;
1) For the performance-based design, a
simplified method incorporating SSI effects is
necessary to obtain precise response of
buildings
2) From microtremor measurement, the ratios
of predominant frequency of SSI systems to
that of building with base-fixed condition are
0.68 - 0.70, in the transverse direction due to
sway and rocking effects.
3) The predominant frequency of building
based on the simplified procedure is very
similar to that in the measurement. It is proved
that the simplified method is effective to
evaluate predominant frequency of SSI system.
4) From microtremor measurement, the
damping factors in SSI systems are evaluated
to be 5 to 7%, using the random decrement
technique.
ACKNOWLEDGEMENTS

The microtremor measurement is conducted


under the research theme entitled Seismic
observation of building and surrounding
ground in the Tsukuba Research Institute
Council. The authors express their sincere
thanks the members of the research group for
their help of microtremor measurements.
REFERENCES

Architectural Institute of Japan (1996) An


Introduction to Dynamic Soil-Structure
Interactionpp.17-19 (in Japanese)

142

Damage Index Evaluations of 3DOF System with Dynamic


Soil-Structure Interaction
K. Kawano1 , Y. Kimura2, Y.Nakamura3
1

Kagohisma University, Japan


Kagoshima University, Japan
3
Kagoshima University, Japan
2

Abstract
For the reliable performance-based design, it is necessary to evaluate the damage of
the structure. The seismic performance evaluations of dynamic soil-structure
interaction system by means of the damage index are examined. The soil-structure
interaction is represented with the sway-rocking model and the nonlinear
characteristics on the structure expressed with the tri-linear model. It is suggested
that it is very important to evaluate the relations of the dominated frequency between
the soil-structure interaction systems and the seismic input motions for the damage
evaluations of soil-structure interaction system due to severe seismic motions.

INTRODUCTION

significant effects on the seismic response


evaluation of the structures.
In the present study, the damage
evaluations by means of combining the ductility
ratio with the seismic response energy to the
SSI system are examined. The soil structure
interaction is represented with the sway-rocking
model (SR Model) and the nonlinear
characteristics on the structure expressed with
the tri-linear model. It is understood that the
ductility ratio and the seismic response energy
are closely related with the damage evaluation
of the structure. If strength characteristics of the
structure have uncertainty, it is expected to
cause the important effect on the damage
evaluations. In order to enhance the damage
evaluation, it is important to verify the
uncertainty effects on the nonlinear seismic
responses.

If some damages can be allowed in the


structural members, it provides very available
methods for the earthquake-proof designs
because of reducing the intensity of the input
forces. From such as background, a lot of
researchers have been carried out for the
performance-based designs. The strength
demand spectrum based on the ductility ratio is
one of the most useful methods which can be
treated with nonlinear effects on the structure
subjected to seismic forces (Iemura et al. 1998).
Recently, the strength demand spectrum based
on the damage evaluation of the structure has
been examined (Mikami et al. 1998). It has
been indicated that the total energy of the
seismic motion on a structure may be given by
one of the important factors on the damage
evaluations (Park et al. 1995).
Moreover, in order to perform the reliable
performance based design, it is important to
clarify for the damage evaluations not only the
dynamic characteristics of structure but also the
dynamic soil-structure interaction effects on the
nonlinear response situation. For the structure
with soil foundation system, it is supposed that
the dynamic soil-structure interaction (SSI) has

FORMULATION

Many researchers have examined for the


dynamic SSI and supposed that the SSI plays
important roles on the seismic response
evaluations. While the finite element method is
one of the most comprehensive methods, it
requires a lot of computing time and does not

143

seem to provide the relevant evaluation for a


preliminary design of the structure. Since the
dynamic responses of the superstructure
subjected to seismic forces mainly depend on
the
dominant
natural
frequency,
the
superstructure is represented with a singledegree-of-freedom (SDOF) system. The
nonlinear characteristics on the structure can
be expressed with the tri-linear model. The
response evaluation of the total system may be
represented with a simplified model as shown in
Figure 1. The governing equation of motion for
the nonlinear SSI system is expressed with

[M ]{&x&}+ [C ]{x&}+ [K (t )]{x} = {F }

k
H

kh

ch

c
VS1
J 0 VS2

m0

kr

cr

Fig.1: An idealized structure-soil-foundation model

(1)

The iterative procedure can be carried out using


the Newton Raphson method.
For damage evaluations of structure
subjected to seismic forces, it is necessary to
evaluate the estimation of energy response due
to seismic motions. Taking into accounts for the
energy on the SSI system, it can be expressed
with integration as follows:

{x&} [M ]{&x&}dt + {x&} [C ]{x&}dt +


t

{x&} [K (t )]{x}dt = {x&} {F }dt


t

(6)

Namely, this equation of motion energy can be


represented as follows:

EK + ED + EH = E

(7)

in which EK , E D , E H and E denote the


kinematic energy, the viscous damping energy,
the hysteretic energy and the total energy of
input seismic motion, respectively. The
hysteretic energy is caused by the restoring
force, and plays significant roles on the energy
evaluation with respect to the nonlinear
response due to seismic motions.
The damage evaluation is conveniently
conducted using the drift displacement and the
energy ratio of the hysteretic energy to the total
input seismic energy. The drift displacement is
depended upon the seismic motion properties
as well as the hysteretic characteristics and
structural properties. For the damage
evaluations by means of combining the ductility
ratio with the seismic response energy, the
damage index to the RC structure by Park et al

(2)

in which

~
~
[ K (t )]{x} = {F }

x1

m1

in which [M] and [C] denote the mass and


damping matrix, and [K (t )] denotes the
stiffness matrix on each time step including the
soil foundation system with linear properties. {F}
denotes vector for the input seismic motion.
The seismic motions corresponding to the
design spectra as shown in Figure 2 are near
field seismic motions recorded in Japan. { &x& },
{ x& } and { x } denote the acceleration responses,
the velocity responses and the displacement
responses, respectively. If the dynamic
response is carried out within linear region, the
governing equation can be solved with modal
analysis and spectral analysis. On the other
hand, if the response of the structure causes to
be nonlinear, it is hard to solve the equation in
the frequency domain. Thus, applying the
incremental method for the equation (1), this
equation can be solved with the increment
method because of the nonlinearity due to the
structure. The equation (1) can be expressed
with the incremental method as follows:

[ M ]{&x&} + [C ]{x&} + [ K (t )]{x} = {F }

x0 H

Zg

(3)

4
2
~
(4)
[ K (t )] = [ K (t )] + ( 2 )[ M ] + ( )[C ]
t
t
4
~
~
~
(5)
{F} = {F} + [M ]{( 2 ){x&}+ 2{&x&}} + [C ](2{x&})
t
Therefore, the increment of the responses can
be determined by solving the equation (3).

144

is applicable to assess the damage situation by


means of the maximum displacement and the
hysteretic energy as follows:

dE

Maximum Acceleration (m/sec**2)

D = 1M +
x1 u
Q y x1u

Kobens
Takans
Portns

30

(8)

in which x1M and x1u stand for the maximum


displacement and ultimate displacement,
respectively. The ultimate displacement can be
determined by the corresponding allowable
ductility ratio. The allowable ductility ratio is 5.0
in this present study. Qy denotes the yield force

25

20

15

10

0
0.1

Natural Period (sec)

of the structure and denotes the coefficient


depending on the characteristics of the
structural member (Fajfar 1992). The damage
index is practically related to the damage
situations by means of the damage assessment
of real structures experienced to seismic forces
as shown in Table 1. In this present study, the
strength demand spectra based on the damage
index from 0.1 to 1.0 are examined.

Fig.2: Acceleration response spectra

Table 1 :The relationship between Damage


Index and Damage level
Damage Index

Damage level

00.1

Slightly

0.10.2

Light

0.20.4

Moderate

0.41.0

Severe

1.0

Failure

RESULTS AND DISCUSSIONS


In order to clarify the damage evaluation of
the SSI system as shown in Figure 1, the
seismic response analysis with the SSI system
is carried out using typical seismic motions
corresponding to the design spectra as shown
in Figure 2. In this present study, the soilfoundation is supported by ground condition
which has two layers and the sway and rocking
spring
constant of soil-foundation can be
determined to the each ground condition
(Yamada et al. 1979). Table 2 shows the shear
wave velocity of each ground condition due to
the seismic motion as shown in Figure 2. The
structure has the viscous damping ratio 0.05,
and the nonlinearity of the structure expressed
with the tri-linear model. The seismic response
analysis is implemented with the Newmarks
method and the time increment is used to be
0.001 sec.
Figure 3 shows the comparison of natural
period of the superstructure with SSI system. It
is understood that the natural period of the SSI
system is longer than the superstructure in the
relatively high frequency range due to the
dynamic soil-structure effects.

Table 2: The shear wave velocity of each ground


condition for seismic motion
K obens

T ak a n s

P o rtn s

300

150

100

350

300

300

(Unit:m/sec)

On the other hand, there are few differences in


the low frequency range. The natural period of
SSI system is supposed to be mainly depended
on the superstructure. From this result, it is very
important to clarify the effects of dynamic soilstructure interaction for the damage evaluations
of the structure.

145

Demand Acceleration (m/sec**2)

Narural Period of the SSI system (sec)

35

Kobens
VS1=300m/sec,VS2=350m/sec
Takans
VS1=150m/sec,VS2=300m/sec
1

Portns
VS1=100m/sec,VS2=300m/sec

0.1
0.1

30
25
20
15
10
5
0

Natural period of the superstructure (sec)

0.1

Fig.3: Comparison the natural period of


superstructure with SSI system

Natural Period of the superstructure (sec)

Fig.4 : A strength demand spectra (Kobens)

demand spectra on damage


35

Demand Acceleration (m/sec**2)

Strength
index

SDOF
D=0.1
SSI system
D=0.1
D=0.2
D=0.4
D=0.6
D=1.0

Input seismic motion:Kobens


Damping ratio:0.05

The strength demand spectra are used to be


examined with the ductility ratio. However, the
damage evaluations can not be always carried
out using only the strength demand spectra
based on the ductility ratio. Therefore, it is
necessary to examine with other method of
assessment for the damage evaluations. The
strength demand spectra based on the damage
index from 0.1 to 1.0 are examined in this
present study.
Figure 4 and Figure 5 show the strength
demand spectra of the SSI system subjected to
the Kobens and Takans as shown in Figure 2,
respectively. The abscissa denotes the natural
period of the superstructure, and the ordinate
denotes
the
demand
maximum
input
acceleration. Each line corresponds to the
damage index from 0.1 to 1.0. The strength
demand spectrum of SDOF is shown by
comparison with the SSI system in each figure.
Comparing the demand acceleration of an input
seismic motion for the damage index, very
effective reduction of demand acceleration can
be caused by increase of the damage index.
For the input seismic motion, Kobens, the
strength demand spectra has a significant
difference between the SDOF and the SSI
system because of the dynamic SSI effects in
the relatively high frequency range. On the
other hand, since the strength demand spectra
due to the Takans yield relatively smaller
difference than the Kobens, it is understood that
the effects of dynamic soil-structure interaction

30

SDOF
D=0.1
SSI system
D=0.1
D=0.2
D=0.4
D=0.6
D=1.0

Input seismic motion:Takans


Damping ratio:0.05

25
20
15
10
5
0
0.1

Natural Period of the superstructure (sec)

Fig.5 : A strength demand spectra (Takans)

become smaller than the Kobens.


As previously mentioned, the strength demand
spectra are evaluated to the allowable ductility
ratio, 5.0. It is necessary to examine the
ductility ratio corresponding to the assigned
damage ratio for the SSI system. Figure 6
shows the maximum ductility ratio of SSI
system subjected to the Kobens. The abscissa
denotes the natural period of the superstructure,
and the ordinate denotes the maximum ductility
ratio of the SSI system. Each line corresponds
to the damage index from 0.2 to 1.0. For the
maximum ductility ratio due to the damage
index,1.0, it is understood that the maximum
ductility ratio has become under an allowable
ductility ratio for the natural period of the
superstructure from 0.2 sec to 2.0 sec. However,
for the maximum ductility ratio to the Kobens, it
is understood that the maximum ductility ratio
exceeds an allowable ductility ratio in the high

146

frequency range. For the maximum ductility


ratio to the Takans, the maximum ductility ratio
exceeds an allowable ductility ratio in the low
frequency range.
For the frequency region such as exceeding
the allowable ductility ratio, it is supposed to be
reexamined of the target damage index. A
constant damage index is obtained in case the
damage level of the structure is slight. Namely,
the damage index depends on the natural
period of the structure as well as the dynamic
characteristics of the input seismic motion.
However, if the severe damage level of
structure can be allowed, it is understood that
the damage index causes to some variations for
the natural period of the superstructure. The
damage index, which can be expressed in
Equation (8), is expressed with combining the
ductility ratio and the hysteretic energy of the
structure. When the damage level of the
structure is severe, the hysteretic energy also
increases, and it is indicated to be considerable
effects on the damage index. If the
performance- based design is applied to the
design of low damage level situation, the
maximum ductility ratio of the structure could be
evaluated with the damage spectra. However,
for the severe damage level of the structure, it
is necessary to carry out the damage
evaluations by means of combining the ductility
ratio with the hysteretic energy

7
6

Maximum ductility ratio

SSI system
D=0.2
D=0.4
D=0.6
D=1.0

Input seismic motion:Kobens


Damping ratio:0.05

5
4
3
2
1
0
0.1

Natural Period of the superstructure (sec)

Fig.6 : Maximum ductility ratio (Kobens)

7
6

Maximum ductility ratio

SSI system
D=0.2
D=0.4
D=0.6
D=1.0

Input seismic motion:Takans


Damping ratio:0.05

5
4
3
2
1
0
0.1

Natural Period of the superstructure (sec)

Fig.7 : Maximum ductility ratio (Takans)

Figure 8 shows the ratio of hysteretic energy


to total energy of SSI system subjected to
Kobens. The assigned damage index is
assumed to be 0.4 and 1.0. The abscissa
denotes the natural period of the superstructure,
and the ordinate denotes the ratio of hysteretic
energy to total energy. It is understood that the
ratio of energy gradually increased in the high
frequency range of superstructure by the
influence of the dynamic SSI. Comparing the
ratio of energy of SDOF with the SSI system, it
is understood that the ratio of hysteretic energy
lead to some differences under the natural
period of the superstructure, 0.5 sec. For the
SDOF system, the ratio of energy on the
damage index D=0.4 is about 60% and D=1.0 is
about 70%. For the SSI system, the ratio of
energy on the damage

147

100

Iuput seismic motion:Kobens


Damping ratio:0.05

90

SDOF
D=0.4
D=1.0
SSI system
D=0.4
D=1.0

80
70

EH/E (%)

Evaluations due to energy response

60
50
40
30
20
10
0
0.1

Natural Period of the superstructure(sec)

Fig.8 : Energy ratio due to the damageindex


(Kobens)

index D=0.4 is about 40% and is about 60% for


the damage index D=1.0. From these results, it
is understood that the SSI plays important roles
on the energy evaluations. In order to enhance
the estimation of the damage of the structure, it
is necessary to clarify the influence of the
dynamic SSI which is mainly depended on the
dynamic characteristics of the input seismic
motion and the natural period of the structure.

Drift displ. / Yield displ.

1.5

Input seismic motion:Portns


Damping ratio:0.05
D=0.2
D=0.4
D=0.6
D=1.0

1.0

0.5

Evaluations due to drift displacement


0.0
0.1

It is indicated that the drift displacement could


be a very important value from the viewpoint of
restoration at the early time of the structure
(Kawashima et al. 1994). The ratios of the drift
displacement to the yield displacement can be
represented with the drift displacement spectra
because it is evaluated with the demand
strength spectra by means of changing the yield
displacement in this present study. Figure 9
shows the drift displacement spectra to the
Portns. The abscissa denotes the natural
period of the superstructure, and the ordinate
denotes the ratio of drift displacement to yield
displacement. For damage index D=0.2, it is
noted that the ratio of the drift displacement to
the yield displacement indicates relatively small
value. It is noted that the ratio of the drift
displacement to the yield displacement
becomes approximately constant values in spite
of the natural period of the superstructure.
However, as the damage index increases, the
ratio leads to be very large fluctuation and it is
primarily depended on the natural period of the
superstructure. Moreover, it is understood that
the drift displacement yields large fluctuation for
not only the natural period of the superstructure
but also the damage index. Especially, the drift
displacement spectra corresponding to damage
index D=1.0 have very remarkable fluctuation.

Natural Period of the superstructure (sec)

Fig.9: Drift displacement spectra (Portns)

Demand Acceleration (m/sec**2)

15
Input seismic motion:Kobens
Damping ratio:0.05

D=0.4
D=0.6

12

0
0.1

Natural Period of the superstructure (sec)


Fig.10: A strength demand spectra with uncertainty (Kobens)

Demand Acceleration (m/sec**2)

15

Influence of the structural uncertainty


In general, the strength properties of the
structure are expec ted to have various
uncertain factors. Since the uncertainty could
yield the important contributions on the damage
evaluations, it is essential for the performancebased design of the structure to clarify the
uncertainty effects with respect to the nonlinear
response. The damage index shown in

Input seismic motion:Portns


Damping ratio:0.05

D=0.4
D=0.6

12

0
0.1

Natural Period of the superstructure (sec)


Fig.11 A strength demand spectra with uncertainty (Portns)

148

equation (8) is related to the yield displacement,


the maximum displacement, the expected
ultimate displacement and hysteretic energy.
Taking into account for the random process of
the input seismic motion and the structure
properties, it is important for the damage
evaluation of the structure to clarify the effects
of the uncertainties. In the present study, the
strength demand spectra with respect to the
uncertainty of the yield displacement are
examined. The coefficient of variation is
assumed to be 20%. Figure 10 shows the
strength demand spectra of SSI system with
uncertainty of yield displacement to Kobens.
Figure 11 similarly shows the strength demand
spectra of SSI system with uncertainty of yield
displacement to Portns. The abscissa denotes
the natural period of the superstructure, and the
ordinate denotes the demand acceleration. A
solid line and dotted line denote the strength
demand spectra due to the damage index
D=0.4 and D=0.6 in case of without uncertainty,
respectively. A dashed line and dashed-dotted
line denote the strength demand spectra to
uncertainty, respectively. Each response is
presented together with the one corresponding
to the mean value plus (or minus) the standard
deviation. Comparing the damage index, 0.4,
with the one 0.6, the demand acceleration is
mainly affected by extent of the uncertainty. If
the structure has the uncertainty of yield
displacement, it could be supposed to lead to
the severe damage level instead of the
moderate damage level as described by Table
1.From the results, it is important for the
damage evaluation to verify the influence of the
uncertainty.
.

to clarify the dynamic soil-interaction effects.


(2) The damage evaluation shows the different
tendency with respect to a maximum
response ductility ratio according to the
allowable
damage
level.
For
the
performance-based design, the light level of
damage can be evaluated by the maximum
ductility ratio of the structure. However, for
the severe damage level of the structure, it
is necessary to carry out using the damage
evaluations by means of combining the
ductility ratio with the hysteretic energy.
(3) The ratios of the drift displacement to the
yields displacement have a tendency
fluctuated by the natural period of the
superstructure. Increasing the damage
index, it leads to the fluctuation for the
natural period of the superstructure. Since
the drift displacement gradually increases
for the increment damage index, it is
important to carry out the available
evaluation for the drift displacement.
REFERENCES
Iemura,H., Igarashi,A. & Takahashi,Y.( 1998)
Ductility and Strength demand for near field
earthquake ground motion: Comparative study on
the Hyogo-ken Nanbu and the Northridge
earthquakes, Structural safety and Probability
pp.1705-1708
Mikami,T., Hirao,K., Sasada,S., Sawada,T. &
Nariyuki,Y.( 1998), A study on design spectra of
seismic intensity for level 2 earthquake, The 10th
Earthquake Engineering Symposium, pp.30613066 (in Japanese)
Park,Y.-J. & Ang,A.H.-S.(1995), Mechanistic
seismic damage model for reinforced concrete,
Journal of Structural Engineering, Vol.111, No.4,
pp.722-739
Yamada,Y, Takemiya,H.,Kawano,K,( 1979).
Random response analysis of a non-leanear
soil-suspension bridge pier, Earthq. Eng. Struct.
Dyn. Vol.7, pp.31-47
Kawashima,K., Macrae,G.A., Hoshikuma,J. &
Nagaya,K. (1994), Residual displacement
response spectrum and its application, Journal
of structural mechanics and earthquake
engineering, JSCE, No.501/ -29, pp.183-192 (in
Japanese)
Fajfar,P. (1992), Equivalent Ductility Factors,
Taking into Account Low-Cycle Fatigue,
Earthquake Engineering and Structural Dynamics,
Vol.21, No.10,pp.837-848

CONCLUSIONS
The seismic performance evaluations of SSI
system by means of the damage index are
examined. The main results are summarized as
follows:
(1) The damage index combined the maximum
displacement and the hysteretic energy can
be applied to estimate the strength demand
spectra for the SSI system. The damage
evaluation depends on the ground condition
and the dynamic characteristics of input
seismic motions. It is necessary for the
evaluations of the strength demand spectra

149

Seismic response of rigid structures stepping on nonlinear foundation


A. Palmeri1, N. Makris2
1
2

Department of Civil Engineering, University of Messina, Italy


Department of Civil Engineering, University of Patras, Greece

Abstract
The rocking motion of slender/rigid structures stepping on nonlinear yielding foundation is
examined. This work is the continuation of previous investigations on rocking structures
where the foundation behavior was restricted to linear viscoelastic. With yielding supporting
springs, the geometric nonlinearities from the dynamics of a rocking block combine with the
material nonlinearities of the foundation. This paper focuses in assessing the effects of the
geometric and material nonlinearities and identifies various trends of the dynamic response.
Selective results are presented.

a base undergoing horizontal motion were


presented by Housner (1963). Following this
seminal work, a large number of studies have
been presented to address the complex
dynamics of the free-standing block (Zhang &
Makris, 2001, and references provided therein),
while a handful of studies examined the rocking
response of structures stepping on a non-rigid
foundation. Psycharis and Jennings (1983)
presented an equivalent linear analysis of the
rocking response of a rigid block stepping on a
viscoelastic foundation, Yim and Chopra (1984)
presented a linear analysis of a flexible
structure supported on a base allowed to uplift,
while Gazetas (1999) documented the
permanent tilting of buildings due to foundation
failure. More recently, Palmeri and Makris
(2005) investigated in depth the rocking
response of a rigid block which steps on linear

INTRODUCTION

The possibility of tall structures to separate from


their foundation limits the earthquake generated
forces and moments, and in most cases leads
to more economical designs, assuming that
stability prevails. For instance, the piers of the
South Rangitikei Viaduct in New Zealand (Beck
& Skinner, 1974; Fig. 1 left), the Rio Vista
Bridge in Sacramento, California (Yashinsky &
Karshenas, 2003; Fig. 1 centre), and the North
Approach Viaduct of the Lions Gate Bridge in
Vancouver, British Columbia (Dowdell &
Hamersley, 2000; Fig. 1 right) are allowed to
rock on their foundation.
The analysis of the rocking response of a
rigid block on a monolithic foundation is well
known to the literature. Early studies on the
rocking response of a rigid block supported on

Fig. 1: Examples of bridges allowed to rock on their foundation.

150

R0

K,y
A

L
D

Fig. 2: Regimes of motion of a rigid block rocking on yielding foundation.

viscoelastic supports via a fully non-linear


analysis, which complements the linear one
presented by Psycharis and Jennings (1983).
In this paper, we present a study on the
rocking response of a rigid block which steps on
yielding supports. It is assumed that the
foundation system cannot take tension;
therefore, separation occurs when the upward
displacement of the heel of the structure is
greater than the static deflection of the support.
The developed methodology is applied to the
response analysis of a bridge peer that is
allowed to step, with the aim of investigating the
effects that a non-rigid foundations has on the
rocking response of tall/slender structures.

POSITION OF THE PROBLEM

We consider in this study the rigid block of mass


M shown in Fig. 2, in which R0 is the size and
is the slenderness. The nonlinear supports at the
points L and R are yielding springs with vertical
stiffness K = v2 M 2 and yielding deflection y .
The springs do not take tension, and the rigid
block can pivot about points L and R when it is
set to rocking. In Fig. 2, the horizontal dotdashed line is used for the position where the
supporting springs are unstressed.
When the motion is weak, then the body is in
continuous contact with both supports at pivot
points L and R. When the vertical uplift of the
heel of the body exceeds the static deflection of

151

the supporting springs, s , then separation


occurs: e.g., in the stage B of Fig. 2 the
separation is imminent, i.e. the supporting spring
at the heel (point R) is unstressed, while in the
stage C the body is supported only at its toe
(point L); the contact at the heel is successively
re-established in the stage E, where the support
is waiting in the unstressed configuration (see
stages C and D).
According to Fig. 2, the rigid block can
experience three regimes of motion: (i) fullcontact at both points L and R (stages A, B, and
E to G in Fig. 2); (ii) contact only at point L
(stages C and D, in which the block is tilting to
the left); and (iii) contact only at point R (stage H
and I, in which the block is tilting to the right).
There is a fourth regime of motion, when the
rigid body separates from both supports, which is
not the subject of this study: when this happens,
our analysis terminates.
It is worth noting that, given the elastoplastic
behaviour of the supporting springs, Fig. 2
shows the plastic deformations accumulated at
the pivot points (e.g., the transition between
stage C and D), which may determine a
permanent tilting of the block at the end of the
motion. As an example, in the stage E of Fig. 2
the elastic deformations at the pivot points L
and R are the same, as in initial stage A;
however, in the stage E the rigid block shows a
nonzero rotation due to the plastic deformation
cumulated at the pivot point L, while prior to the
stage A the supporting springs have not
experienced plastic deformations.

The loss of energy caused by the impacts is


governed in our analysis by the ratio of the
velocity of the imminent pivot point before the
impact to the velocity of the pivot point after the
impact at the support. This ratio defines the
coefficient of restitution, = v&La v&Lb = v&Ra v&Rb , which
has been introduced by Psycharis and Jennings
(1983). Given the abovementioned parameters,
the rotation of a rigid block rocking on a twosupport yielding foundation can be expressed by:
= f ( , , p, , v ,,excitation )

where = M R02 / I0 is the shape factor of the


rigid block, being I0 the moment of inertia of
the rigid block with respect to the centre of
mass; = y s is the bearing capacity index
of the foundation; and p is a frequency
parameter, which results from its s\ize, R0 , and
the intensity of the surrounding gravitational
field, g :
p=

g
+ 1 R0

(2)

Eq. (1) indicates that there are six


parameters of the rocking structure-foundation
system that influence its response other the
characteristics of the ground excitation.
For a bridge tower, the frequency parameter
can be as low as p = 0.5rad/s or even smaller;
whereas for a household refrigerator p = 2.0
rad/s . The slenderness of a 10 m tall classical
column can be as low as = 10 0.17 rad;

v
O

R0
R
L

v&&g (t )

(1)

u&&g (t )

Fig. 3: Schematic of a rigid block rocking on yielding foundation

152

F
Fy

F,v

n
K

Fig. 4: Rheological model (right) and hysteretic loop (centre) for the ideal elastic-perfectly plastic behaviour.
Hysteretic loop of the Bouc-Wen model (left).

the bearing capacity index, = y s , are in


the range 0 1 and > 1.

whereas, the slenderness of a three-storey


building which is 8 m wide is approximately
= 36 0.62 rad. The vertical frequency, v ,
can originate from the vertical stiffness of an
elastoplastic device or the stiffness of a
conventional foundation, depending on the
configuration of the rocking structure.
Accordingly, practical values of interest for the
ratio = v p are in the range 10 < < 100.
Of course, there may be special cases where
the ratio lies outside this range. Finally, the
values of the coefficient of restitution, , and

EQUATIONS OF MOTION

Since the rigid block is not allowed to slide, the


system has only two degrees of freedom: the
vertical displacement of the centre of mass, v ,
and the rotation of the block, (see Fig. 3).
Using the two degrees of freedom, the equations
of motion in the full-contact regime can be
derived in the form:

&& g
&&
v + 2 ( zL + zR ) = g v g (t )
(3)

&&g (t )
u
1

+
2
2
2
1 + cos ( ) && +
p {[ sin( ) + cos( ) ] zL [ sin( ) cos( ) ] zR } = (1 + m ) p cos( )


2
g

where u&&g (t ) and v&&g (t ) are the time histories of


the horizontal and of the vertical components of
the ground acceleration, respectively; while zL
and zR are the dimensionless hysteretic
variables associated with the yielding springs at
the pivot points L and R, respectively, which are
proportional with the corresponding reaction
forces:

FL = Fy zL ; FR = Fy zR

(4)

in which Fy = K y is the yielding force of the


supporting springs. When the rigid block
separates from the support on the right, the
associated hysteretic variables goes to zero,
zR = 0, and the equations of motion are:

&& g
v + 2 zL = g v&&g (t )

u&& (t )
1 + cos2 ( ) && + 1 + p 2 [ sin( ) + cos( ) ] z = (1 + ) p 2 [cos( ) sin( ) ] g
L

2
g

(5)

In the same way, when the rigid block


separates from the support on the left, zL = 0,
and the equations of motion are:
&& g
v + 2 zR = g v&&g (t )

u&& (t )
1 + cos2 ( ) && + 1 + p 2 [ sin( ) cos( ) ] z = (1 + ) p 2 [cos( ) + sin( ) ] g
R

2
g

153

(6)

Each of Eqs. (3), (5) and (6) rules the


dynamic equilibrium of the rigid block in one of
the three regimes of motion depicted in Fig. 2.
These equations, however, have to be coupled
with those ruling the time variation of the
hysteretic variables zL and zR , whose values
depends in principle on the whole time histories
of the vertical displacements of the pivot points
L and R, respectively:

effectively approximated by the Bouc-Wen


model (Bouc, 1967; Wen, 1976), which is
described by the equations:
F (t ) = K y z(t )
z&(t ) =

where n is a dimensionless parameter that


controls the rate of the transition between the
elastic and the perfect plastic phases (Fig. 4
right). In the following, it is always assumed
n = 5.
Taking into account Eqs. (7), the state
equations ruling the hysteretic variables zL and
zR when the rigid block is in contact with the
supporting springs can be derived in the form:

v L = v + R0 sin( ) ; v R = v R0 sin( ) (7)


Among the rheological models available in
the literature for the hysteresis of yielding
springs, the simplest one is the elastic-perfectly
plastic model (Fig. 4 left and centre), made of
an elastic spring with stiffness K in series with
a Coulomb element with sliding force Fy = K y .
The ideal elastic- perfectly plastic model can be

2 p2
sin( )
z&L =
v& +

g
1+

2 p2
sin( )
&
z
=
v&
R
g
1+

(8)
v&
n 1
n
1 0.5 sign(v& ) z z 0.5 z

g sin( ) &
& 1 0.5 sign v& +
zL
(1 + ) p 2

g sin( ) &
& 1 0.5 sign v&
zR
(1 + ) p 2

n 1

zL 0.5 zL

n 1

zR 0.5 zR

(9)

separates from the left spring zL = 0 and


z&L = 0.

When the block separates from the right


spring, the corresponding hysteretic variables
takes the constant value zR = 0, and then
z&R = 0. In the same way, when the block

Table 1: Information pertinent to the strong motions selected in this study.


Earthquake
(Date)

Station
(Component)

D
km

PGD
m

PGV
m/s

PGA
g

Loma Prieta,
California
(18-Oct-1989)

16 LGCP
(000)

6.9

6.1

0.412

0.948

0.563

Erzikan, Turkey
(13-Mar-1992)

95 Erzikan
(NS)

6.9

2.0

0.273

0.839

0.515

7.1

8.5

0.410

1.274

1.497

6.7

6.4

0.327

0.130

0.843

6.7

7.1

0.288

1.661

0.838

6.9

0.3

0.358

0.127

0.611

Cape Mendocino,
89005 Cape
California
Mendocino
(25-Apr-1992)
(000)
Northridge,
24514 Sylmar Olive
California
View Med FF
(17-Jan-1994)
(360)
Northridge,
77 Rinaldi
California
Receiving Station
(17-Jan-1994)
(228)
Kobe, Japan
(16-Jan-1995)

0 Takatori
(000)

NS: North-South.

154

1992 Erzican
PGA= 0.515 g

0.12

max

1995 Kobe
PGA= 0.611 g

0.12

0.08

0.08

0.04

0.04

0.12

max

1989 Loma Prieta


PGA= 0.563 g

1994 Northridge, Rinaldi


PGA= 0.838 g

1994 Northridge, Sylmar


PGA= 0.843 g

1992 Cape Mendocino


PGA= 1.497 g

0.12

0.08

0.08

0.04

0.04

: Rigid foundation
Case No. 1 (+): =40, =2.5
Case No. 3 (): =80, =2.5

Case No. 2 (,): =40, =3.5


Case No. 4 (*): =80, =3.5

Fig. 6: Maximum rotation values of the bridge tower when subjected to the strong motions listed in Table 1.

can see that for the first three ground motions a


foundation with larger values of stiffness and/or
yielding deflection produces larger responses; in
these cases, moreover, the model with rigid base
(solid lines) gives conservative estimations of the
peak rotation. Remarkably, in two among the last
three ground motions the opposite happens.
Nevertheless, what is important to note is that large
variations in the mechanical properties of the
foundation have just a small effect on the value of
the peak rotation. Furthermore, the tower, although
very slender ( = 0.13 rad), experiences always
small rotations because it has such large size
R0 = 60.1m).
( p = 0.36 rad/s,
and
More
precisely: i) the maximum peak rotation of the
stepping bridge under analysis (max = 0.114 ) is
induced by the 1989 Loma Prieta earthquake
(PGA = 0.563 g ) when the base is assumed rigid;
and ii), contrary to what one would intuitively
expect,
the
minimum
peak
rotation
(max = 0.0191 ) is induced by the 1992 Cape
endocino earthquake, which exhibits the maximum
peak ground acceleration (PGA = 1.497 g ), also
in this case when the base is assumed rigid.

NUMERICAL APPLICATIONS

The proposed nonlinear formulation is used to


estimate the seismic-induced response of a
stepping bridge about 60 m tall to six strong
ground motions chronologically listed in Table 1.
Frequency parameter, angle of slenderness and
shape factor of the tower are assumed to be
p = 0.38 rad/s, = 0.13 rad ( 7.4) and = 3,
respectively. Moreover, the coefficient of restitution
is = 0.05, while four combinations of parameters
and are considered: 1) = 40 and = 2.5;
2) = 40 and = 3.5 (larger yielding deflection);
3) = 80 (stiffer foundation) and = 2.5; and 4)
= 80 and = 3.5. The peak rotations computed
with these combinations of the foundation
parameters are also compared with that one
computed with a monolithic base.
Fig. 6 plots the peak normalized rotations of the
tower of interest when subjected to the six ground
motions listed in Table 1, including also the vertical
component. They are ordered with increasing peak
ground acceleration, from PGA = 0.515 g (1992
Erzikan) to 1.497 g (1992 Cape Mendocino). One

155

= 40, = 2.5
0

&
p

= 40, = 3.5

10 15 20 25 0

10 15 20 25 0

= 80, = 2.5
5

= 80, = 3.5

10 15 20 25 0

10 15 20 25

0.12

0.12

0.06

0.06

-0.06

-0.06

-0.12

-0.12

0.6

0.6

0.3

0.3

-0.3

-0.3

-0.6

E
M

-0.6

-1
-1

u&&g (t )
g

0.6

0.6

0.3

0.3

-0.3

-0.3

-0.6

-0.6
0

10 15 20 25 0

t [s]

10 15 20 25 0

10 15 20 25 0

t [s]

t [s]

10 15 20 25

t [s]

-4 -3 -2 -1 0 -3 -2 -1 0 -4 -3 -2 -1 0 -3 -2 -1 0 -8 -6 -4 -2 0 -6 -4 -2 0 -8 -6 -4 -2 0 -6 -4 -2 0
0

F -0.25
Fy -0.5

-0.5

-0.75
-1

-0.25

-0.75

-1

-4 -3 -2 -1 0 -3 -2 -1 0 -4 -3 -2 -1 0 -3 -2 -1 0 -8 -6 -4 -2 0 -6 -4 -2 0 -8 -6 -4 -2 0 -6 -4 -2 0

v y

v y

v y

v y

Fig. 7: Time histories of rotation , angular velocity & and total energy E (top graphs) of a bridge
pier with parameters = 3, = 0.13 rad and p = 0.38 rad/s rocking on a yielding foundation
with coefficient of restitution = 0.05 and four combinations of parameters and , when
subjected to the 1989 Loma Prieta earthquake. Hysteretic loops v F (bottom graphs) of the
yielding springs..

156

It is worth noting that, although is some cases


the peak rotations computed for different
configurations of the foundation are very close
each other, the time histories of the rocking
response may be quite different. In Fig. 7, for
instance, normalized rotation (first graph from
the top), angular velocity & (p ) (second graph)
and total energy E M (third graph, with values in
J/kg) of the tower of interest when subjected to
the 1989 Loma Prieta earthquake are compared.
When the foundation is rigid (solid lines), the peak
rotation takes the larger value, while the
oscillations looks very regular. In the case of
yielding foundation, on the contrary, the peak
rotations take slightly smaller values, while the
rocking responses prove to be very sensitive to
stiffness and load bearing capacity of the
foundation. For instance, the hysteretic loops of the
nonlinear supporting sprigs (first graphs from the
bottom) show that = 80 and = 2.5 is the case
where the inelastic demand is larger (third column).

ACKNOWLEDGMENTS

We thank the European Social Fund (ESF),


Operational Program for Educational and
Vocational Training II (EPEAEK II), and
particularly the Program PYTHAGORAS II, for
funding the above work.
REFERENCES

Beck, J.L., and Skinner, R.I. (1974). The seismic


response of a reinforced concrete bridge pier
designed to step, Earthquake Engineering and
Structural Dynamics, Vol. 2, pp. 343-58.
Bouc, R. (1967). Forced vibration of mechanical
systems with hysteresis, Proc. 4th Conf. on
Nonlinear Oscillation.
Dowdell, D.J., and Hamersley, B.A. (2000). Lions
Gate Bridge North Approach Seismic retrofit,
Proc. 3rd International Conference STESSA 2000.
Gazetas, G. (1999). Overturning and settlement in
Adapazari during the 1999 Izmit (Turkey)
Earthquake, Proc. 1st International Conf. on the
Kocaeli Earthquake.

CONCLUSIONS

In this paper, the rocking response of


slender/rigid structures stepping on nonlinear
yielding foundation is investigated. The response
analysis of this study complements previous
investigations where the foundation behaviour is
restricted to linear viscoelastic. A novel set of
equations of motion is presented, where two
hysteretic variables are included, aimed to account
for the plastic deformations cumulated in the
supports. In the proposed model the geometric
nonlinearities arising from the dynamics of a
rocking block combine with the material
nonlinearities of the foundation. As a result, the
response of the rocking block to a given
accelerogram depends on its size, shape and
slenderness, together with the stiffness and the
yielding deflection of the foundation, and with the
amount of energy dissipated during the impacts.
The numerical applications show that a simple
model with rigid base may give accurate
estimations of the peak rotation, while more refined
models are required when the inelastic demand in
the foundation is of importance. Some
counterintuitive results are also emphasized: for
instance, an increase in the stiffness and/or in the
strength of the foundation may indifferently
produce larger or smaller rotations.

Housner, G.W. (1963). The behaviour of inverted


pendulum structures during earthquakes, Bulletin
of the Seismological Society of America, Vol. 53, pp.
404-17.
Palmeri, A., and Makris, N. (2005). Response analysis
of rigid structures rocking on viscoelastic
foundation, Report No. EEAM 2005-02,
Department of Civil Engineering, University of
Patras, Greece.
Psycharis, I.N., and Jennings, P.C. (1983) Rocking of
slender rigid bodies allowed to uplift, Earthquake
Engineering and Structural Dynamics, Vol. 11, pp.
57-76.
Wen, Y.K. (1976). Method for random vibration of
hysteretic systems, Journal of Engineering
Mechanics Division ASCE, Vol. 102, pp. 249-63.
Yashinsky, M., and Karshenas, M.J. (2003).
Fundamental of seismic protection for bridges,
EERI Monograph MNO-9, Earthquake Engineering
Research Institute, Oakland, California.
Yim, C.K., Chopra, A.K., and Penzien, J. (1980)
Rocking response of rigid blocks to earthquakes,
Earthquake Engineering and Structural Dynamics,
Vol. 8, pp. 565-87.
Zhang, J., and Makris, N. (2001) Rocking response
and overturning of free-standing blocks under
cycloidal pulses, ASCE Journal of Engineering
Mechanics, Vol. 127, pp. 473-83.

157

Numerical simulations of shaking table experiments on a shallow


foundation test model at PWRI, Japan
R. Paolucci1, M. Shirato2 , M.T. Yilmaz3
1

Politecnico di Milano, ITALY


Public Works Research Institute, Tsukuba, JAPAN
3
Middle East Technical University, Ankara, TURKEY
2

Abstract
The set of experiments carried out at the Public Works Research Institute, Tsukuba, Japan,
is a unique opportunity to analyze the seismic behaviour of shallow foundations under
seismic loading of various levels of intensity up to complete failure. The experimental set up
and the different phases of the experiments are described in more detail in a companion
paper for this Workshop (Shirato et al., 2007). In this contribution the experiments are
simulated using a simplified approach consisting of a single degree-of-freedom structure
founded on a compliant foundation with 3 degrees-of-freedom (vertical, horizontal, rocking).
The nonlinear behaviour of the soil-foundation system is allowed by numerical modelling
using a nonlinear elasto-perfectly plastic macro-element, defined by a suitable yield surface
and plastic flow rule.
The results of the numerical analyses are very satisfactory in terms of the accurate
simulations of the time history of overturning moments, driving the nonlinear behaviour of
the foundation. The simulation of the accumulated settlements and rotations was much
more difficult: satisfactory results were obtained by using a suitable stiffness degradation
rule as a function of the total plastic rotation accumulated during shaking. These results
throw light on the debated issue of the foundation ductile behaviour during strong seismic
shaking and on its possible role in the evolution of performance-based design concepts
including nonlinear soil-structure interaction.

INTRODUCTION

design, there is now an increasing awareness


of the effects of the interaction between the
foundation and the superstructure and its role
on the overall seismic capacity of the system
(see e.g. ATC-40, 1996; Martin and Lam, 2000;
Pecker 2006). This awareness is conflicting with
the lack of reliable methods for predicting
foundation yielding under strong earthquake
shaking and for putting it into proper
consideration in the framework of such
approaches.
For this purpose, nonlinear dynamic finite
element simulations of large numerical models,
including the superstructure, the foundation and
the surrounding soil, are probably not the best
tool, since they require the use of sophisticated
soil constitutive laws and very large
computational
times
to
perform
a
comprehensive set of parametric analyses. To
overcome such limitations, but preserving at the

In the framework of seismic design


according to capacity principles, it is generally
recognized that any damage to the foundation
should be avoided. This means that the
nonlinear capacity of the system is exploited at
the superstructure level alone, typically
permitting the energy dissipation at suitably
selected points through formation of plastic
hinges or insertion of isolation/dissipation
devices. This requirement is partly supported by
budget considerations, due to the time and
economic costs related to the post-earthquake
inspection and verification of the foundation
system, but it is also justified based on the lack
of well established and calibrated methods to
study the post-yielding behavior of soilfoundation systems under strong seismic loads.
With the ever increasing interest towards
performance based approaches for seismic

158

defined according to Nova and Montrasio


(1991) and a plastic flow rule according to
Cremer et al. (2002). Further details on the
numerical procedure are reported in Paolucci et
al. (2007).

same time the essential features of the dynamic


soil-structure interaction problem, the macroelement concept has become more and more
popular (Nova and Montrasio, 1991; Paolucci,
1997; Cremer et al., 2001 and 2002; Le Pape
and Sieffert, 2001). It basically consists of
modelling the soil-foundation system as a single
non-linear macro-element, having 3 degrees of
freedom (dof) for in-plane analyses, defined by
a suitable failure surface and plastic flow rule.
However,
although
the
macro-element
approach is quite promising, it has not been
supported so far by adequate experimental
evidence, at least for seismic applications.
Indeed, few experimental results are
available on the nonlinear soil-shallow
foundation interaction under dynamic seismic
loads, including, among the most relevant ones,
the quasi-static large-scale tests performed at
the Joint Research Center at Ispra, Italy (Negro
et al., 2000; Faccioli et al., 2001) and at the
Public Works Research Institute (PWRI, 2005)
at Tsukuba, Japan, and the fully dynamic tests
in centrifuge (Zeng and Steedman 1998, Gajan
et al., 2005) and in shaking table (Maugeri et al.
2000).
In this work we make reference to the
companion paper of Shirato et al. (2007)
presented at this Workshop, which summarizes
the experimental setup and main results of a
comprehensive set of shaking table tests on the
performance of a model shallow foundation
resting at the surface of a laminar box filled with
dry sand, and excited by real accelerograms of
various levels of amplitude. The purpose of this
contribution is to show that, after a suitable
improvement of the method proposed by
Paolucci (1997), it is possible to capture with a
satisfactory agreement many of the details of
the experimental response, even during the
highly nonlinear phases of foundation behavior.

x0

x1

m1
k1,c1

k0,c0
m0
kr,cr

kv,cv

xg
yg

Fig 1: The 4 degrees-of-freedom model used for


dynamic nonlinear soil-structure interaction
analyses

A further improvement of our results was


obtained by taking into account that during the
strongly nonlinear phase of the excitation the
instantaneous foundation-soil contact area
decreases, owing to successive cycles of
foundation rotations. This was implemented by
a simple degradation rule for the foundation
stiffness parameters (k0, kv and kr, see Fig. 1),
described by the following equation:

B = B (1-D)

(1)

where B is the foundation width and D is a


degradation parameter defined as follows:

( )

D p =

COMPUTATIONAL PROCEDURE

D1
1 + 1 D2 p

(2)

where D1 and D2 are model parameters related


to the ultimate D value and to the degradation
speed, respectively, while p is the cumulated
plastic foundation rotation at a specified instant
of time. The best agreement with observations
has been obtained using D1 = 0.75 and D2 =
5000/rad. Details on this model and on its
implementation in the numerical procedure are
given in Paolucci et al. (2007).

The computational procedure is based on the


simplified approach proposed by Paolucci
(1997), consisting of a two-dimensional soilstructure interaction model having 4 dof (Figure
7), one to account for the horizontal motion of
the superstructure, assumed to behave in the
linear-elastic range, while the soil-foundation
system is represented by a 3 dof (horizontal,
rocking and vertical motion) elasto-perfectly
plastic macro-element, with yielding function

159

NUMERICAL SIMULATIONS RESULTS

without
degradation
is
already
rather
satisfactory, the improvement obtained by the
stiffness degradation rule (1) and (2) leads in
many cases to capture some important details
of the nonlinear response, including the period
elongation following the major yielding phases
of the model response. Note that, while in the
calculations with the non-degrading model the
damping value was kept constant to the 5%
value estimated from the small-amplitude
vibrations, in the degrading model we raised
such value to 10% (15% for Case 1-2), in order
to limit the amplitude of fluctuations of the
computed response, especially in terms of
rotation, and improve the agreement with the
observations. The higher frequency excitation
for Case 1-2, and the corresponding shortduration yielding phases, can explain the higher
damping ratio required to fit the observations.
The parameter set used for the analyses is
reported in Table 1, while the reader is referred
to Paolucci et al. (2007) for further details on
their selection.
When considering foundation rotation (Fig. 4),
the agreement is again quite satisfactory,
especially if we consider the loading magnitude
and the large permanent effects observed on
the test model. Again, the degrading model is
suitable to capture some important details of
the observed time histories, except for Case 1-4
for which the numerical calculation diverges.

We have studied the response of the model to


the four earthquake input motions considered in
the sequence of shaking table experiments
(Fig. 2), that will be denoted in the following as
Case 1-2, 1-4, 1-5 and 2-2, according to the
sequence of tests.

Fig 2: Earthquake records used as base excitations


for shaking table tests: (a) 1993 Hokkaido
Nansei Oki Earthquake Schichihou Bridge
record, (b) and (c) 1995 Kobe Earthquake
JMA record, (d) 80% scaled 1995 Kobe JMA
record

A sample of results is shown in Fig.3 in terms of


the normalized overturning moment M/VB. We
have omitted Case 1-5 for sake of brevity,
mainly because it was run just after Case 1-4,
without any intermediate preparation work of
the foundation soil, so that the dynamic soil
properties are not the same as in the other
experiments. The horizontal displacements will
also be omitted in the discussion, since they
generally result in negligible permanent
deformations. All plots in Fig. 3 show the
normalized overturning moment observed (thick
lines) and simulated (thin lines) within the 10 s
long most severe phase of the excitations.
Within this phase, the foundation behaviour is
highly nonlinear, and frequently reaches a
threshold eccentricity ratio that in our
experiments turns out to be slightly higher than
0.4. For all cases, we have compared both
results with and without stiffness degradation,
to show that, although the simplest approach

Table 1: Elastic parameters used for numerical


simulations
k0
N/m
3.0107
cv
Nms/
rad
1.8104

160

kr
Nm/ra
d
2.0106

kv
N/m

k1
N/m

c0
Ns/m

3.8107

3.0109

1.6104

cr
Nms
/rad
2.0103

c1
Ns/m

m1
kg

m0
kg

h
m

J
kgm2

0.

664

161

0.54

1.5102

Fig 3: Time histories of the normalized overturning moment observed (thick lines) and simulated (thin lines)
either with or without stiffness degradation. The small picture at the bottom shows a zoom of the
observed vs. simulated comparison during the pre-yielding phase of the excitation. From Paolucci et al.
(2007)

161

Fig 4: The same as Fig. 3, in terms of foundation rotation time history. From Paolucci et al. (2007)

Fig 5: The same as Fig. 3, in terms of foundation vertical settlement. From Paolucci et al. (2007)

162

Faccioli E, Paolucci R, Vivero G. (2001)


Investigation of Seismic soil footing interaction
By large scale cyclic tests and analytical
models., 4th Int. Conf. on Recent Advances in
Geotechnical Earthquake Engineering and Soil
Dynamics, Special Presentation Lecture SPL-05,
San Diego.

The main limitation of our numerical


calculations is clearly the vertical settlement
prediction, as shown in Fig. 5: our elastoperfectly plastic macro-element model is not
presently able to simulate the accumulation of
large vertical settlements throughout the
excitation. However, when the magnitude of
settlements and the number of loading cycles is
limited as in Case 2-2, the prediction of the final
vertical displacement is relatively satisfactory,
although details of the time history are missed.
Comparing rotations and vertical displacements
in Figures 4 and 5, it seems that there is a
coupling of these degrees of freedom in the
nonlinear range, that is not suitably accounted
for in our simplified model.

Gajan S, Kutter B, Phalen J, Hutchinson TC, Martin


GR. (2005) Centrifuge modeling of loaddeformation behaviour of rocking shallow
foundations. Soil Dynamics and Earthquake
Engineering, Vol. 25: pp. 773-783.
Le Pape Y, Sieffert JP. (2001). Application of
thermodynamics to the global modelling of
shallow foundations on frictional material.
International Journal for Numerical and Analytical
Methods in Geomechanics, Vol. 25, pp. 13771408.

CONCLUSIONS

Martin GR, Lam IP (2000). Earthquake resistant


design of foundations - Retrofit of existing
foundations; Proc. GeoEng2000, Melbourne,
2000; pp. 1025-1047.

The numerical method used in this study,


based on the macro-element concept to
account for in a simplified way the nonlinear
dynamic
interaction between
soil
and
foundation, has been proven rather successful
to simulate the results of the sequence of
shaking table experiments on a shallow
foundation test model carried out at the Public
Works Research Institute, Tsukuba, Japan. The
agreement
with
observations
is
quite
satisfactory for the rotational degree of
freedom, for which the introduction of a simple
stiffness degradation rule allowed us to capture
also some important details of the foundation
response. Still some progress should be made
as regards the prediction of vertical settlements,
for which a more sophisticated constitutive
model is required.

Maugeri M, Musumeci G, Novit D, Taylor CA.


(2000). Shaking table test of failure of a shallow
foundation subjected to an eccentric load. Soil
Dynamics and Earthquake Engineering, Vol. 20:
pp. 435-444.
Negro P, Paolucci R, Pedretti S, Faccioli E. (2000).
Large-scale
soil-structure
interaction
experiments on sand under cyclic loading, Proc.
12th World Conference on Earthquake
Engineering, Auckland, NewZealand, paper #
1191.
Paolucci R. (1997). Simplified evaluation of
earthquake induced permanent displacements of
shallow foundations Journal of Earthquake
Engineering, Vol. 1, pp. 563-579.
Paolucci R. Shirato M, Yilmaz MT (2007). Seismic
behaviour of shallow foundations: shaking table
experiments vs. numerical modelling. Submitted
for publication to Earthquake Engineering &
Structural Dynamics.

REFERENCES
ATC-40. Seismic evaluation and retrofit of concrete
buildings. Chapter 10: Foundation effects.
Technical Rep. SSC 96-01, Seismic Safety
Commission, State of California, 1996.

Pecker A (2006). Enhanced seismic design of


shallow foundations: example of the Rion
Antirion bridge. 4th Athenian Lecture on
Geotechnical Engineering, Athens, Greece,
January 2006.

Cremer C, Pecker A, Davenne L. (2001) Cyclic


macro-element for soil-structure interaction:
material
and
geometrical
non-linearities.
International Journal for Numerical and Analytical
Methods in Geomechanics,Vol. 25: 1257-1284.

Shirato M, Nakatani S, Fukui J, Paolucci R (2007)


Large-scale model tests on shallow foundations
subjected to earthquake loads. Proc. 2nd
Japan-Greece Workshop on Seismic design,
observation and retrofit of foundations, 3-4 April
2007, Tokyo.

Cremer C, Pecker A, Davenne L. Modelling of


nonlinear dynamic behaviour of a shallow strip
foundation with macro-element. Journal of
Earthquake Engineering, 2002; 6: 175-212.

163

Zeng X, Steedman RS. (1998) Bearing capacity


failure of shallow foundations in earthquakes,
Gotechnique, Vol. 48, pp. 235-256.

164

Analytical Modelling of Footings under Large Overturning Moment


M. Apostolou and G. Gazetas
National Technical University of Athens, Greece

Abstract
Non linear features of the rocking response of tall structures founded on shallow
foundations are investigated. To this extent a macroscopic modelling of the soil
foundation system is developed, capable of representing the large-displacement domain
of the response. Analytical equations for the monotonic load-displacement relationship
are extracted incorporating both geometric and material nonlinearities. Such analytical
backbone curves may be implemented in dynamic SSI analysis through the concept of
nonlinear macro-element to represent the near-field soil-foundation system. The
limitations of conventional Winkler-based modelling are also highlighted under strong
overturning moments.

INTRODUCTION

The use of overstrength factors, is


necessitated by the so-called capacity
design principle, under which plastic hinging
is allowed only in the super-structural
elements not in the below-ground (and thus
un-inspectable)
foundation
and
soil.
Therefore, structural yielding of the footing
and mobilisation of bearing capacity
mechanisms is not allowed. Only a limited
amount of sliding deformation and uplifting at
the foundationsoil interface is allowed.
However, there is a growing awareness in the
profession of the need to consider soil
foundation inelasticity, in analysis and
perhaps even in design [see: Pecker (1998),
Martin & Lam (2000), and Allotey & Naggar
(2003)]. This need has emerged from:

The conventional approach to foundation


design introduces factors of safety against
sliding and exceedance of ultimate capacity,
in a way similar to the traditional static design.
This approach involves two consecutive steps
of structural and foundation analysis:

Dynamic analysis of the structure is


performed in which the soil is modeled as
an elastic medium, represented by
suitable translational and rotational
springs (and, sometimes, with the
associated dashpots).

The dynamic forces and moments


transmitted onto the foundation are
derived from the results of such analyses
along with considerations for inelastic
structural response (e.g. by reducing the
moments in columns through the
behaviour [ductility] factor q). The
foundations are then designed in such a
way that these transmitted horizontal
forces
and
overturning
moments,
increased by overstrength factors, would
not induce sliding or bearing capacity
failure.

165

The large (often huge) acceleration (and


velocity) levels recorded in several
earthquakes which are associated with
even larger elastic spectral accelerations
(of the order of 2g). Enormous ductility
demands would be imposed to structures
by such accelerations if soil and
foundation yielding did not effectively
take place to limit the transmitted
accelerations.

interface and the passivetype resistance


often enjoyed by embedded foundations.

In seismically retrofitting a building or a


bridge, allowing for soil and foundation
yielding is the only rational alternative.
Because increasing the structural capacity
of some elements would imply that the
forces transmitted onto the foundation be
increased, to the point that it would not be
technically or economically feasible to
undertake them elastically. Thus, new
retrofit design guidelines (e.g. FEMA 356)
explicitly permit inelastic deformations in
the foundation.
Even with new structures, it has been
recognised that with improved analysis
methods we need to better evaluate
performance in terms of levels of damage.
For the superstructure, performance
based
design
or
equivalently
displacementbased design have been
used for a number of years, with inelastic
push-over analyses becoming almost
routine in seismic design practice. It is
logical to extend the inelastic analysis to
the supporting foundation and soil.

Sliding at the soil-foundation


interface

Foundation uplifting from the


supporting soil

NEW DESIGN PHILOSOPHY: PLASTIC


HINGING IN SHALLOW FOUNDATIONS

Bearing capacity type of


soil failure

Excluding structural yielding in the isolated


footing or the foundation beam, three types of
nonlinearity (shown schematically in Fig 1)
can take place and modify the overall
structurefoundation response:

Fig 1: Plastic hinge approach at soilfoundation


interface.

Separation and uplifting of the foundation


from the soil. This would happen when the
seismic overturning moment tends to produce
net tensile stresses at the edges of the
foundation (Fig 1b). The ensuing rocking
oscillations in which uplifting takes place
involve primarily geometric nonlinearities, if
the soil is competent enough. There is no
detriment to the vertical load carrying capacity
and the consequences in terms of induced
vertical settlements may be minor. Moreover,
in many cases, footing uplifting is beneficial
for the response of the superstructure, as it
helps reduce the ductility demands on
columns. Housner (1963), Pauley & Priestley
(1992), and many others have reported that
the satisfactory response of some slender
structures in strong shaking can only be
attributed to foundation rocking. Deliberately
designing a bridge foundation to uplift in

Sliding at the soilfoundation interface.


This would happen whenever the transmitted
horizontal force exceeds the frictional
resistance (see Fig 1a). As pointed out by
Newmark (1965), thanks to the oscillatory
nature of earthquake shaking, only short
periods of exceedance usually exist in each
one direction; hence, sliding is not associated
with failure, but with permanent irreversible
deformations. The designer must only ensure
that the magnitude of such deformations
would not be structurally or operationally
detrimental. Although this philosophy has
been applied to the design of earth dams and
gravity
retaining
walls,
its
practical
significance for foundations might be
somewhat limited in view of the large values
of the coefficient of friction at soilfooting

166

analyses in which inelastic action in the soil is


considered or is ignored. With inelastic action
(including uplifting) the shear wall sheds
some of its load onto the columns of the
frame, which must then be properly
reinforced; the opposite is true when linear
soilfoundation behaviour is assumed. Thus,
computing the consequences of plastic
hinging in shallow foundation analysis may
be a necessity. The interplay between uplifting
and mobilization of bearing capacity
mechanisms is governed primarily by the
following factors:

rocking has been proposed as an effective


seismic isolation method by Kawashima &
coworkers (2005). Moreover, even with very
slender and relatively rigid structures, uplifting
would not lead to overturning except in rather
extreme cases of little concern to the engineer
(Makris & Roussos 2000, Gerolymos et al
2005). In soft and moderately-soft soils much
of what was said above is still valid, but
inelastic action in the soil is now unavoidable
under the supporting edge of the uplifting
footing in rocking. At the extreme, inelastic
deformations in the soil take the form of
mobilization of failure mechanisms, as
discussed below.

Mobilization of bearing capacity failure


mechanisms in the supporting soil. Such
inelastic action under seismic loading would
always be accompanied with uplifting of the
foundation (Fig 1c). In static geotechnical
analysis large factors of safety are introduced
to ensure that bearing capacity modes of
failure are not even approached. In
conventional seismic analysis, such as in the
EC8 Part 5 bearing capacity is avoided
thanks to an overstrength factor of about
1.40. The oscillatory nature of seismic
shaking, however, allows the mobilisation (for
a short period of time!) of the maximum soil
resistance along a continuous (failure)
surface. No collapse or overturning failure
occurs, as the applied (causative) moment
quickly reverses, and a similar bearingcapacity failure mechanism may develop
under the other edge of the foundation. The
problem again reduces to computing the
inelastic deformations, which in this case
means permanent rotation. The designer must
ensure that its consequences are not
detrimental.

the vertical foundation load N in


comparison with the ultimate vertical
capacity Nu, expressed through the ratio
= /u
the height, h, of the mass center of gravity
from the base compared with the
foundation dimensions (width B, length L)
the intensity, frequency content and
sequence of pulses of the seismic
excitation.

CONVENTIONAL WINKLER MODELLING

In common SSI analysis procedures,


Winkler modelling of soil has often been found
a simple and efficient tool because of its ability
to incorporate different nonlinear aspects of
the rocking behaviour at relatively low levels
of computational cost. For example, notension springs can capture uplifting effects at
the soil-foundation interface whereas soil
yielding can be represented with elastic
perfectly plastic springs. In the context of a
conventional Winkler model the following
postulations are often encountered:

The concept of allowing mobilization of


bearing capacity mechanisms in foundation
design may represent a major change in
foundation design philosophy (FEMA 1997,
Pecker 1998). However, for analysis of the
ultimate response of a structurefoundation
system to extreme earthquake shaking,
accounting for such a possibility is necessary.
Martin & Lam (2000) illustrate with an
example of a hypothetical structure containing
a shear wall connected with a frame how
dramatically different are the results of

167

A unique constant spring modulus kv is


adopted: for any type of loading
(symmetric or antisymmetric), independent
of the distance from the mid-point x .
Correspondingly, a purely vertical or
moment loading results in a uniform or
triangular distribution of contact pressures
along the foundation.
The rotation pole of the structural system
remains fixed at the foundation mid-point
even after uplift or soil yielding initiates.
During a clockwise (counter-clockwise)
rotation, uplift onsets when the vertical

displacement at the left edge (right edge)


of the foundation becomes zero.
Second order ( P ) effects are ignored,
even in the domain of large lateral
displacements of tall superstructures.

Elastic soil

For purely elastic soil behaviour,


nonlinearity of the momentrotation response
emerges once uplifting occurs and leads to an
ultimate capacity of Mu = Nb . During fullcontact conditions (state 1), the moment with
reference to the centre of the footing can be
easily extracted by integrating distributed
moments over the foundation:

A schematic of a rigid footing resting on


Winkler supports is shown in Fig 2. Bartlet
(1979) used a conventional soil model of
distributed elasticperfectly plastic springs to
study the rocking response of a footing on
clay. His considerations on the different states
of moment-rotation response have been
implemented by the FEMA guidelines
(273/274 documents) in modelling shear walls
as portrayed in the graph of Fig 3. The
following considerations of the Mtheta
relationship are presented in this plot:
(a) the extreme case of elastic soil
conditions represented with the path 1-2-3-5a,
(b) the case of a soil with limited strength
where uplift occurs before yield (path 1-2-3-45b-6), and
(c) the case of a soil with limited strength
where yield precedes the uplifting (path 1-2c3c-4c-5c-6).

M1 =

2
kv x d x =

(=

uplift

b N /(2kv b ) ) together with Eq 1:

uplift =

N
Nb
and Muplift =
= M2
2
2kv b
3

(2)

Upon the onset of uplifting, the moment


rotation curve enters a softening mode (state
3 and 5a). The moment at this level is derived
with integration over the remaining in contact
part of the footing 2 as follows (Siddharthan
et al., 1992):

M3,5 a =

wbo

x2 d x

b 2

kv

pole of
rotation

2b

1 2N
= Nb 1
3 k b 2
v

2N
= Muplift 3

kv b 2

(a)

(1)

The rocking rotation and moment at


incipient uplift (state 2) can be derived by the
uplifting
criterion
w ( b ) = 0

N=mg

p(x)

2kv b 3

po

po (uplift initiation)

(3)

This curve reaches a maximum value of


max M = Nb which represents the ultimate
moment capacity of the foundation for elastic
soil response when P effects are ignored.

po

+
pu po

(b)

(yield initiation)

Elastoplastic soil

Fig 2: Conventional beam-on-winkler-foundation


model: (a) configuration of the model and
(b) superposition of soil contact pressures
at full-contact state.

In the general case of soil with limited


strength, an ultimate value of the spring
reaction pu = Nu / 2b is implemented with Nu

168

The equation of the ultimate moment can


be also obtained from moment equilibrium at
limit state 6 (see Fig 3). After some algebraic
manipulations, Eq 7 yields to the analytical
solution of the failure curve in the N M
plane:

being the foundation bearing capacity under


central vertical load. It is revealed that the
inverse of the safety factor under purely
vertical loading,

= N / Nu = ( FSv )

(4)

N
M = Nb 1

Nu

has a significant effect on the rocking


behaviour (Allotey et al., 2003). In particular,
two separate modes of the moment-rotation
curve can be distinguished dependent upon
the value of : an uplift prevailing state
corresponding to a lightly-loaded foundation
( < 0.5 ) and a soil-yield prevailing state in

(8)

or in nondimensional form
=
m

n
(1 n )
2

= M / Nu B ( B = 2b ) are
n = N / Nu = and m
respectively the nondimensional vertical load
and moment.

case of a heavily-loaded foundation ( > 0.5 ) .


The former follows the path 1-2-3-5b-6 in the
graph of Fig 3 whereas the latter is
represented with the path 1-2c-3c-4c-5c-6. As
shown in Fig 2b uplift initiates before soil yield
if po < pu po or po < 0.5 pu ( < 0.5 ) . The
rocking rotation and moment at uplift onset
are given by Eq 2. Similarly if po > 0.5 pu

(b)
(a)

5a
4

( > 0.5 )

soil yield occurs first. In this case


the yield criterion becomes (Siddharthan et
al., 1992):

5b

5c

(c)
3c

4c

2c

q
N
y = u
kv b 2kv b 2

(5)

and

My =

2qu b 2 Nb

= M 2c
3
3

Allotey et al., (2003) derived the analytical


relationship during concurrent uplift and yield
(state 5b or 5c):

M5 b,c = Nb

pu 3
N2

2 pu 24 ( kv )2

N2
2 pu

3c Yield prior to uplift

2 Elastic at uplift

5c Yield after uplift

3 Elastic with uplift

6 Limit state

(a) Elastic soil behaviour

(6)

(c) Elastoplastic soil behaviour


(yield initiates before uplift)

Fig 3: Rocking of a shear wall on strip footing: the


different states of the M-theta curve under
monotonic loading (Bartlet, 1979, reprinted
in FEMA274 document).

Regardless the value of the ultimate


moment capacity of the foundation derived
from Eq 6 ( for ) is:
M6 = Nb

1 Elastic prior to uplift

The analytical expressions of the

interaction curves in the dimensionless n m


plane for foundation failure (overturn) and

(7)

169

the optimum foundation behaviour in terms of


the moment ultimate capacity is obtained:

incipient uplift and soil yield are summarized


in Table 1 together with their values
at = 0, 0.5, and 1. These curves are also
plotted in Fig 4. The loading plane bounded
by the failure curve is partitioned in regions of
linear and nonlinear response. The nonlinear
area is subdivided in smaller parts depending
on whether uplift or soil yield or both are
encountered. It is worthy of note that a
perfectly symmetric response is achieved
about the vertical axis at = 0.5 . At this point

1
Nu b
4

max Mu = 0.125Nu B =

(9)

In addition, at = 0.5 the maximum range


of linearly elastic response is achieved before
nonlinear conditions due to uplift or soil yield
are engaged, with a threshold moment
of 0.083 Nu B = ( 2 / 3 ) max M .

plane for foundation


Table 1 Analytical expressions of the interaction curves in the dimensionless n m
failure (overturn) and incipient uplift and soil yield.
Interaction curve
Vertical load
factor:
=

(1)
Foundation
uplift:

N
Nu

uplift =
m

Muplift

y =
m

Nu B

0.5

My

Mu
Nu B

=
m

Nu B

rigid =
m

Mu ,rigid
Nu B

(3 4 )
6
1
(1 )
6

6
1
(1 )( 4 1)
6

0.5

(4)
Overturn on
rigid soil:

(3)
Failure
(overturn):

(2)
Soil yield:

(1 )

=0

= 0.5

1
12

1
12

1
8

=1

1
8
1
2

0.25
(1)
(4)

Mu : Nu B

0.2
0.15

(3)
(2)

(d)

0.1

(1)
(3)

(b)

0.05

(c)

((2)
4)

(a)
0
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

N : Nu
(a)

Elastic,
full-contact

(b)

Elastic,
uplift

(c)

Yield,
full-contact

(d)

Yield,
uplift

Fig 4: Interaction curves in the normalized NM plane for bearing capacity failure on rigid or deformable soil.
Decomposition of uplifting and soil-yielding mechanisms.

170

onsets whenever the subgrade reaction due to


moment loading at a corner point pm reaches
the initial vertical reaction po . The different
states of the foundation response in terms of a
large-displacement analysis procedure are
presented in Fig 5.

MACROSCOPIC FOUNDATION MODELLING


Elastic soil

For a vertical load factor << 1 (i.e. for a


lightly loaded foundation or for very stiff soil
conditions) foundation rocking is associated
with large levels of uplift and small soil
deformations. Under such circumstances the
assumption of elastic soil behaviour is
generally a reasonable approximation in
analyzing the rocking response. Nonlinear
behaviour under moment loading is then
attributed primarily to the reduction of the
footing contact area (geometric nonlinearity).
In slender structures geometrical nonlinearity
is even more amplified at large rotations due
to the accompanying lateral movement of the
mass centre ( P effect).
In light of an elastic approach for the soil
medium, concentration of contact pressures
occurs in the vicinity of corner points which
cannot be captured by the conventional
Winkler modelling. The increase in local soil
stiffness with the distance from the footing
midpoint is higher under antisymmetric
loading which corresponds to larger values of
rocking stiffness K m compared to the vertical
stiffness Kv . It is therefore evident that a
more efficient and precise soil model should
be incorporated in the nonlinear analysis of
footings subjected to severe overturning
moments. Neglecting the effects of soil
nonlinearity, such a rigorous macroscopic
modelling of rocking behaviour should be in
agreement with: (a) the classical elastic
medium solutions of the soil-foundation
stiffnesses during full-contact conditions (i.e.
Gazetas, 1991), and (b) the limiting case of an
uplifting foundation on a rigid soil.
We consider a rigid, strip footing of width
B = 2b resting on the surface of a
homogeneous half-space. Initially the footing
is subjected only to a vertical loading N .
Then a gradually increasing horizontal force is
applied at the mass center of the
superstructure (located at height h ) leading to
an overturning moment M = Nh cos about
the foundation center. It is postulated that
tensile forces cannot be undertaken by the
soil-foundation interface. In this way uplift

N
M ( < Muplift )

wbo

2b

(a) Full-contact phase


N
Uplift condition :
pz(-b) = 0

M ( = Muplift )

wbo
= uplift
2b

(b) Incipient uplift


Mass
center

N
M ( > Muplift )

wbo

2b

(c) Uplifting phase

Fig 5: Rocking of a rigid strip footing on elastic soil:


different states of response.

Full-contact phase: For sufficiently low


levels of moment, the confinement due to
vertical loading ensures a full-contact
condition at the interface. At this state, the
strip is rocking about its midpoint and
therefore any vertical displacement w at a
distance x from the center comprises (a) a
vertical component w bo and (b) a rocking
component x sin :

w b ( x ) = w bo + x sin

(10)

Moreover, the contact pressure to the footing


at a distance x can be determined with the

171

superposition of the symmetric (vertical) and


the antisymmetric (moment) components of
loading ( pv and pm respectively):

pz ( x, ) = pv ( x ) + pm ( x, )

where Kv , K m are the (global) soil-foundation


stiffnesses for each loading case. These
stiffnesses may be calculated with closedform solutions of the literature. For example
for a rigid strip on elastic half-space it is
Kv = 0.73 G /(1 ) and K m = G b 2 / 2(1 )
(Gazetas, 1991). Recalling Eqs 13 the
subgrade moduli ratio at a distance x from
the center is:

(11)

It is remarked that in contrast to the


conventional winkler modelling, the sensitivity
of soil response to the loading conditions is
now incorporated by adopting two separate
spring constants for vertical and moment
loading.
The problem of a rigid strip on a
homogeneous isotropic medium can be
analyzed in a variety of ways. These include
Greens function techniques, complex variable
methods and integral transform methods.
Within the assumption of a smooth interface,
closed-form solutions can be derived for
simple loading cases. Sadowsky (1928) and
Muskhelishvilli (1953) developed solutions for
the contact pressures in symmetric or antisymmetric loading:
pv ( x ) =

kv ( x ) K v b 2
=
km ( x ) 2K m

which is independent of the distance x .


The sensitivity of the subgrade modulus to
the distance from the midpoint and the loading
conditions is reflected in the p w curves,
calculated at the center and at the edges of
the foundation with two-dimensional finite
element analysis and plotted in Fig 6. The
vertical load which is initially applied to the
footing induces a uniform settlement w bo . The
horizontal (displacementcontrolled) loading
imposed at the level of the mass center
invokes an additional loading in the right part
of the footing (see curve 3) with stiffer
subgrade modulus as well as unloading of the
left part (see curve 1). At this state of
response, symmetric points of the footing with
respect to the vertical central axis have equal
subgrade moduli (i.e. km1 = k m 3 ).

b 1 x 2 / b2

and

(12)
pm ( x, ) =

2M x

b3 1 x 2 / b 2

Evidently, the contact pressure for each


loading case theoretically approaches infinity
at the edges of the footing. In reality the
maximum value of the contact pressure is
bounded due to the finite soil strength and
redistribution of stresses. The subgrade
modulus for vertical and moment loading at a
distance x from the foundation mid-point is
derived after dividing Eqs 12 by w bo and
x sin respectively:

kv ( x ) =

4
km3*

pz / pvo

(3)

(1), (3)

(1)

kv1*, kv3*
kv2*
0

0.25

km1*

(2)
km2*

0.5

0.75

1.25

1.5

w b / w bo

Fig 6: p w curves at the center (2) and at the


edges of the foundation (1, 3) during a
push-over loading test.

2K m

(2)

b 1 x 2 / b2

(1)
2
1

(13)

km ( x ) =

(3)

Kv

and

(14)

1 x 2 / b2

172

The resultant moment of the contact


pressures with reference to the center can be
computed with integration over the foundation:

geometry of the supporting medium and


induces uplift to initiate at a vertical
displacement w ( b ) larger than zero. This is
evident in the numerical analysis depicted in
Fig 6 where uplift onsets before the
normalized vertical displacement of the
unloading (left) edge w bo / w b becomes zero.
Taking into account that at marginal uplift the
vertical displacement of the footing pivot point
is w ( b ) = w uplift the following expression for

M = M p,v + M p,m
b

k ( x )w
v

bo

x cos d x

(15)

x 2 cos sin d x

the rotation at incipient uplift is derived


according to Fig 6:

where M p,v are M p,m the moments associated


with the symmetric and antisymmetric loading
respectively. Setting the transformation
x = b sin ( d x = b cos d ) , these moment
components yield:
b

Mf ,v =

Nx
1 x 2 / b2

Nb cos

/2

sin uplift =

cos d x
(16a)

uplift

sin d = 0

/ 2

Mf ,m =

K m x 2 sin 2

1 x 2 / b2

K m sin 2

w bo kv (b )
b km (b )

(18)

Substitution of Eq 14 into Eq 18 derives the


following uplifting criterion for the rocking
angle:
Mrigid
Nb
=
2K m
2K m

(19a)

Mrigid
Nb
=
2
2

(19b)

and therefore

and
b

w bo w uplift

Muplift

dx

/2

1 cos 2
d

2
/ 2

According to the conventional model, the


uplifting criterion of Eq 17 is satisfied when
w ( b ) = 0 meaning that at the uplift onset the
footing edge reaches its initial position. On the
contrary, from the proposed model uplifting
occurs before the footing edge returns to its
initial position. For a homogeneous soil profile
the arising restitution ratio is depending upon
the footing geometry and the presence of
shallow bedrock.
Moreover, in terms of the conventional
model the critical angle for incipient uplift is
or
uplift N / Kv b
uplift = sin1 (w bo / b )

(16b)

= K m
Through Eqs 16 it is verified that the
symmetric part of loading does not contribute
to the resultant moment whereas a linear
moment-rotation relationship is established
under antisymmetric loading.
Uplift initiation: For a clockwise rotation,
the footing marginally lifts off the supporting
soil when the contact pressure at the left edge
counter-balances the initial reaction of vertical
loading:

pz ( b ) = pv ( b ) pm ( b ) = 0

= Nb / 3K m and therefore the uplifting moment


Evidently,
the
is Muplift = N b / 3 = M rigid / 3 .
conventional Winkler model underestimates
the moment at incipient uplift by a factor of 1.5
in comparison to the exact two-dimensional
solution.

(17)

In contrast to conventional Winkler-based


approach, in the proposed model moment
loading is associated with stiffer subgrade
moduli. This is a result of the two-dimensional

Uplifting phase: Once uplifting occurs the


foundation area remaining in contact with the

173

ground is gradually decreasing (see Fig 5c).


As a result the rocking response enters a non
linear regime even under purely elastic soil
conditions. During the uplift mode, the
moment of the foundation with respect to its
midpoint is the sum of (a) the moment due to
the contact pressures ( M p,v and M p,m ) and (b)

wb

w b = w b + b (1 ) ln

(20)

pm ( x, ) x cos d x

b 2

N h sin

1 2 2

(24)

2M

pm ( , ) =

1 2 2

By virtue of Eqs 21 and 24 the moment due to


purely vertical loading for the uplifting footing
yields:

(21)

The contact pressures of the effective


footing on the basis of elastic soil behaviour
are calculated with the hypothesis of the
incipient uplift:
The distribution of subgrade reactions
along the footing due to combined vertical and
moment loading once uplifting occurs is the
same with that of a fictitious footing of
width 2 , which under the same combined
loading is being at incipient uplift.
A consequence of this hypothesis is that
the center of the effective footing at each step
is the instantaneous rotation pole of the
footing. This can be expressed by the
following incremental equation:

w b = b (1 )

(23b)

pv ( ) =

We define the distance of a point from


the effective footing (the part of the footing
remaining in contact with the ground)
midpoint. In this case:
x = +b

The latter equation provides the analytical


relationship between the vertical displacement
and the rocking angle of the footing over
elastic soil.
The hypothesis of the incipient uplift
allows for the superposition principle to be
applied for the analytical calculation of the
contact pressures developed along the
effective footing:

pv ( x ) x cos d x +

b 2

(23a)

or

M = M p,v + M p,m + MP
b

b (1 )

uplift

w bo

the moment ensued by the lateral movement


of the superstructure ( P effects). In this
way the foundation moment becomes:

wb =

Mf ,v =

N ( + b )

1 2 / 2

/2

N cos

cos d

( sin + b ) d

(25)

/2

= N cos ( b )
This denotes that the moment Mf ,v is the
product of the vertical reaction resultant times
the distance from the effective footing center
to the footing midpoint. The moment of the
foundation due to purely moment loading is:

Mf ,m =

(22)

K m,eff sin 2 ( + b )

3 1 2 / 2

where is the effective-to-initial footing width


ratio ( / b ) .
Now, the vertical displacement of the
foundation w b can be derived by integration:

K m,eff sin 2

174

/2

2
b
sin + sin
d

/2

K m,eff sin 2

K m,eff

(26)

The
above
equation
verifies
the
expression adopted for the subgrade stiffness
due to moment loading.
Finally, the total overturning moment of
the uplifting foundation yields:
M = K m,eff + N ( b ) cos Nh sin

limiting case that the footing is supported by


rigid soil it yields = 0 for any > 0 and
therefore Eq 29 leads to the well-known
momentrotation relationship:
M = Nb cos Nh sin = NR sin ( c )

(27)

While the analytical formulation of the


foundation moment was based on a strip
footing it can also be extrapolated to any
rectangular spread foundation where the
direction of horizontal loading runs parallel to
a normal axis of the foundation cross-section.

However, the rocking stiffness of the effective


footing is:
K m,eff =

G 2
2
= Km 2
2 (1 )
b

(28)

1 : rocking at incipient uplift

By substituting Eq 28 into Eq 27 we finally get:


M = Km

b2

(31)

2 : rocking with uplift

2b sinuplift = 2 sin

uplift

+ N ( b ) cos Nh sin (29a

/ b uplift /

2b sinuplift

2 sin

or
M = K m 2 + Nb (1 ) cos Nh sin

(a)

(29b
)
N = 500 kN,

Finite element analysis of rocking


response reveals that the width of the
effective footing is inversely proportional to
the rocking angle as shown in Fig 7b. In this
way the effective footing ratio may be
calculated with the following equation:

=
=

= 1

uplift

uplift

(30a)

> uplift

(30b)

2b = 2 m,

h=5m

E = 100 MPa - Analytical


-//-

0.8

FE

= /b

E = 20 MPa - Analytical
-//-

0.6

FE

0.4

0.2

0
0

0.02

0.04

0.06

0.08

0.1

: rad

(b)

A geometrical explanation of the relationship


between the effective width and the rocking
angle is shown in Fig 7a.
Eqs 29 and 30 provide the analytical
expression of the foundation moment as a
non linear function of the rocking rotation.
Denotably the above analytical procedure can
be utilized in the computation of the moment:
(a) in the linear domain where full-contact
conditions are encountered, (b) in the largeamplitude region where the gradually
amplifying P
effects dominate the
response, and (c) at near-overturning
conditions where c = tan1(b / h ) . In the

Fig 7: Comparison of the numerical curves of the


- relationship with the analytical
prediction.

A comparison of the analytical calculation


of the foundation moment and vertical
displacement to the results of the finite
element analysis is shown in Fig 8 for a strip
footing ( 2b = 2 m, h = 5 m, N = 500 kN) on
elastic soil layer over rigid bedrock
( E1 = 20 and E2 = 100 MPa ) . An excellent
agreement between analytical and numerical

175

method is achieved throughout the range of


the rocking rotation.
N = 500 kN,

2b = 2 m,

500

h=5m

500

M : kNm

E = 100 MPa
400

400

300

300

200

200

100

100

E = 20 MPa

0
0

0.05

0.1

0.15

0.2

0.05

: rad
0.2

0.2

0.15

0.2

E = 20 MPa

0.15

0.15

0.1

0.1

0.05

0.05

-0.05

-0.05
0

0.05

0.1

0.15

0.2

0.05

0.1

: rad

: rad

500

500

E = 100 MPa

M : kNm

0.15

0.2

E = 100 MPa

wb : m

0.1

: rad

E = 20 MPa

400

400

300

300

200

200

100

100

0
-0.05

0.05

0.1

0.15

0.2

0
-0.05

wb: m

Analytical curves
FE curves

0.05

0.1

0.15

0.2

wb: m

M ( ) = K m 2 + Nb (1 ) cos Nh sin
w b ( ) = w bo + b (1 ) sin + ln (1 )

Rigid soil boundary

uplift / ,
1 ,

> uplift
uplift

Fig 8: Analytical curves of a rigid strip footing on elastic soil (Eqs 22, 29, and 31) and comparison with twodimensional finite element results.

176

During the uplift regime the foundation


moment may be expressed explicitly as a
function of the effective footing ratio by
substituting the rocking angle given by Eq 30b
to Eq 29b:

Nh
M = Muplift 2 2

Km

and the moment at this point:

Nh
Mu = Nb 1

K m

or

(32a)

2u
Mu = Nb 1
tan c

or without P effects

M = Muplift ( 2 )

u
Nb 1 2
c

(35b)

We notice that when u 0 the ultimate


moment approaches the rigid soil limiting
value ( Nb ), while for u c / 2 the moment

(32b)

The latter equation is that presented by


Crmer et al (2002) extracted empirically from
the results of a parametric numerical study.
In the foregoing it was pointed out that
once uplifting initiates the foundation stiffness
enters into a softening fashion which bounds
the overturning moment to an ultimate value.
This upper limit corresponds to the moment
capacity of the foundation. It is well known
that for the extreme case of an infinitely rigid
soil this ultimate moment equals to the vertical
load times the half-width of the footing. In
reality though, soil deformations in the vicinity
of the base edge are inevitable even when
dealing with very stiff soils. Thus, soil
compliance shifts the axis of the resultant
vertical reaction towards the base centre,
reducing the moment capacity of the
foundation. For elastic soil conditions, the
moment Mu is the local maximum of the
function M = M ( ) as defined in Eqs 29.
Hence it can be calculated by the
condition d M / d = 0 . For small values of the
rocking angles (compared to the critical angle
c ) it is sin and cos 1 . Therefore, the
rotation at which the ultimate moment of the
foundation occurs can be computed with
derivation of Eq 29 with respect to :
dM
N 2b2
=
N h
d
4K m 2

(35a)

Mu tends to zero. This means that the locus

of points (u , Mu ) tracks onto the median of


the angle defined by the ordinate and the
rigid M-theta softening line. That conclusion
is validated through the results of finite
element analysis as depicted in Fig 9. The
exact location of an ultimate point (u , Mu ) at
this locus depends on the vertical load, the
height of the mass-point and the soilfoundation rocking stiffness.
600

E = 100 MPa

500

E = 20 MPa
M : kNm

400

E = 5 MPa
300
200
100
0
0

0.05

0.1

0.15

0.2

: rad

Fig 9: Monotonic M curves for strip footing on


elastic soil (5, 20, and 100 MPa) calculated
with finite element analysis and the arising
locus of the critical equilibrium points.
( N = 500 kN, 2b = 2 m, h = 5 m ).

(33)
Inelastic soil

Therefore the angle that satisfies the equality


d M / d = 0 is:

u =

Nb 2
Nh
= tan c
4K m h
4K m

The simplification of elastic soil behaviour


allows for a tractable analytical calculation of
the uplifting response. Excluding some cases
of lightly loaded foundations ( < 0.2 0.3 ) , in
reality yielding zones of the supporting soil

(34)

177

emanating from the area underneath the


foundation edges are rather inevitable. This
may lead to a visible non linear fashion of
the momentrotation relationship even during
full-contact conditions. Moreover, once uplift
initiates,
soil
material
nonlinearity
counterbalances the uplifting displacements
and results into substantially non-linear
foundation behaviour.
We consider again a rigid, strip footing of
width B = 2b resting on the surface of a
homogeneous half-space. Initially the footing
is subjected only to a vertical loading N . The
vertical
displacement
of
the
footing
(settlement) w bo is now a non linear function
of the applied load. Two analytical curves are
most suitable to describe the backbone N w
curve according to the exponential and the
hyperbolic law (Eqs 36a and b respectively):

1200
1000

N: kN

800
600
400
200
0
0

0.08

0.1

0.12

0.1

0.12

1200

800
600
400

(36b)

200
0
0

where = K v / Nu . In Fig 10 the analytical


N w curves are plotted in comparison with
those calculated from finite element analysis
of a strip footing ( 2b = 2 m, h = 5 m ) on
homogeneous
soil
layer
over
rigid
An
bedrock ( E = 100 MPa, su = 100 kPa ) .
excellent agreement is achieved for both
models with the numerical results. According
to the afore-discussed models, the settlement
at the end of this loading phase may be
calculated as follows:
=

0.06

HYPERBOLIC
MODEL

(36a)

ln (1 N / Nu )

0.04

1000

1 e w b

wb
N=
1
1
+
w
Kv Nu b

wb =

0.02

wb: m

N: kN

N = Kv

EXPONENTIAL
MODEL

ln (1 )

Kv / Nu
N
1
1
wb = u
=
Kv 1 Nu / N
(1 1/ )

0.02

0.04

0.06

0.08

wb: m

Fig 10: Comparison of the exponential and the


hyperbolic model with the finite element
analysis in the calculation of the N w
curve during monotonic vertical loading.

At the second loading stage, a gradually


increasing horizontal force is applied at the
level of the superstructure mass (located at
height h ) leading to an overturning moment
M = Nh cos about the foundation center. As
in the elastic case, uplift onsets whenever the
subgrade reaction due to moment loading at a
corner point pm reaches the initial vertical
reaction po .

(37a)
(37b)

Full-contact phase: For sufficiently low


levels of moment, the confinement due to
vertical loading ensures a full-contact
condition at the interface. In contrast to the
elastic case, the strip is rocking about a point
(pole of rotation) which is not fixed at the
footing midpoint but shifts towards its
unloading edge. This leftward (for a clockwise

The analytical calculation of the settlement


with the exponential model has been
successfully evaluated by Nova and
Montrasio (1991) through experimental N w
curves on sand.

178

rotation) movement of the instantaneous pole


is attributed to the plastification of the
supporting soil underneath the loading edge
of the footing. The larger the structural weight
is, the more rapidly the pole moves towards
the unloading edge. In the limiting case of
1 the footing tends immediately to rotate
about its unloading edge even under a very
small overturning moment. In this case the
vertical displacement (settlement) becomes:
w b = b sin

k ( b)
v
w vo

pu 1 e pu
= pu

where kv and km are the elastic stiffnesses


as determined above. Eq 40 may reduce to
the following expression of the stiffness ratio:
w vo w uplift
kv ( b )
=
k m ( b )
w vo

(38)

On the other hand, for elastic soil


behaviour ( = 0 ) the displacement w b prior
to uplift initiation is zero.
The M curve at the full-contact phase
may be calculated with the exponential or the
hyperbolic model as follows (Eqs 39a and 39b
respectively):
M = Km

M=

400
300

(39a)

200
100

1
1
+

K m Mu

(41)

EXPONENTIAL
MODEL

M: kNm

1 e

k ( b)
m

(w bo wuplift )
p
1 e u
(40

0
0

(39b)

0.01

0.02

0.03

0.04

0.03

0.04

: rad

where = K m / Mu and Mu is the ultimate


moment of the foundation if uplift is prevented.
The afore-mentioned analytical models
are evaluated in the calculation of the M
relationship through finite element analysis for
a strip footing ( N = 500 kN, 2b = 2 m, h = 5 m )
on homogeneous soil layer over rigid bedrock
( E = 100 MPa, su = 100 kPa ) as plotted in Fig
11. Both models seem to capture the basic
features of the backbone curve. Nevertheless,
a slightly closer fit is achieved with the
exponential model throughout the loading
sequence.

HYPERBOLIC
MODEL

400

M: kNm

300
200
100
0
0

0.01

0.02

: rad

Fig 11: Comparison of the exponential and the


hyperbolic model with the finite element
analysis in the calculation of the M
curve during monotonic horizontal loading
at the level of the superstructure mass
center.

Uplift initiation: As in the elastic case,


the footing marginally lifts off the supporting
soil when the uplifting criterion described with
Eq 17 is satisfied. By applying the exponential
law to the p w curve of the unloading edge,
the following uplifting criterion is obtained:

From Eq 41 it is derived that the critical


rocking angle at marginal uplift is the same
obtained with the elastic approach:

179

uplift

Nb
2K m

The components of the moment


associated with the eccentricity of the
resultant reaction N ( M p,v ) and the P

(42)

effects ( MP ) are the same with those of the


elastic case.
For the analytical calculation of the
moment component associated with the
antisymmetric part of the external loading
( M p,m ) calculation of soil reactions due to that

The uplifting moment is then derived by


substituting Eq 42 to Eq 39a:
Muplift

Nb

2 Mu
= Mu 1 e

(43)

Evidently, the uplifting moment is not


anymore a linear function of the vertical load
N but exhibits a softening behaviour as N
increases. As the failure curve in the N M
plane is described with a parabola, it is
expected that a threshold value of N exists
where the uplift and failure curves intersect
and beyond which no uplifting occurs.

loading and integration along the effective


footing is required. For inelastic soil conditions
however, this method leads to complex
integral expressions of the resultant moment
even when simple analytical p w curves are
employed (for example the exponential law).
More than that, the shift of the rotation pole
towards the edge of the footing, further
complicates the decoupling of the developed
soil
reactions
into
symmetric
and
antisymmetric components. An alternative
approximate method to calculate the moment
component M p,m may arise through the

Uplifting phase: Once uplifting occurs,


the foundation area remaining in contact with
the ground is gradually decreasing. However
the magnitude of uplifting (i.e. expressed with
the effective footing ratio = / b ) is limited
with the increase in the structural weight
(expressed with the load factor ).
The vertical displacement w b in the
limiting cases of = 0 and = 1 is given by
Eqs 23b and 38 respectively. From these two
states of response, linear interpolation may
provide the vertical displacement w b of a load
factor as follows:
w b w b + b (1 ) ln

extrapolation of the linear M correlation


extracted for elastic soil conditions.
This linear trend of the M correlation
is confirmed by the numerical results plotted
in Fig 12b.
In addition, the effective footing ratio
must satisfy: (a) the limiting case of elastic soil
( = 0 ) and (b) the limit state condition = .
The exponential law of the effective footing
ratio with respect to the rocking angle
which satisfies both limitations is given by the
following equation:

(44a
)

or
w b w b + b (1 ) ln

(44b
)

(46)

Numerical validation of the latter equation


through nonlinear finite element analysis of
rocking response is presented in Fig 12c.
Eventually, the moment of the foundation
for inelastic soil conditions becomes:

During the uplift mode, the moment of the


foundation with respect to its midpoint is the
sum of (a) the moment due to the contact
pressures ( M p,v and M p,m ) and (b) the


M = Km
+ Nb (1 ) cos Nh sin (47
1
)

moment ensued by the lateral movement of


the superstructure ( P effects):
M = M p,v + M p,m + MP

uplift (1 ) +

(45)

180

Eqs 46 and 47 comprise the analytical


expression of the foundation moment as a
non linear function of the rocking rotation for
the general case of inelastic soil.

(50b)

Mu = Nb (1 )(1 )

(51)

in dimensionless form:
Mu

=
(1 )(1 )
2
Nu B
n
= (1 ) (1 n )
m
2

(52a)
(52b)

Eqs 52 provide the analytical expression


of the failure curve in the nondimensional

plane when P
effects are
n m
considered. Remark that for = 0 (i.e. no
P effects are considered) the failure curve
reduces to that calculated with the
conventional Winkler model (Eq 8).

CONCLUSIONS

Non linear features of the rocking


response of tall structures founded on shallow
foundations are investigated. Principally,
based on conventional Winkler modelling,
interaction curves in the nondimensional
N M plane were calculated for failure
(overturn) as well as incipient uplift and soil
yield. A perfectly symmetric response is
achieved about the vertical axis at = 0.5 . At
this point the optimum foundation behaviour in
terms of the moment ultimate capacity is
obtained.
To
highlight
the
induced
nonlinearities in the large-displacement
domain, a macroscopic modelling of the soil
foundation system was developed. In this
respect
analytical
equations
for
the

(48)

The rocking angle u at which the ultimate


moment is attained is therefore:
Nh
4K m

2u
u
Mu = Nb 1
Nb 1 2
c
tan c

After some algebraic manipulations, the


ultimate moment may be expressed as a
function of the nonlinear parameters
= Nh / K m
and , which represent
respectively geometrical and soil material
nonlinearities.

Interaction in the N M plane: As


shown in the foregoing discussion, the M
backbone curve follows a softening fashion
due to geometric and soil material
nonlinearities. As a result the moment
developed by the foundation is bounded by an
ultimate value Mu . In the optimum case of a
rigid supporting soil the foundation undertakes
the maximum possible moment which
corresponds to the vertical load N times the
half-width of the footing b . In common
geotechnical applications however soil
compliance and plastification of the supporting
soil result in substantial reduction of the
ultimate moment. Moreover, further decrease
in Mu is expected in tall structures due to
P effects. In the general case the ultimate
moment of the foundation may be derived
analytically as the local maximum of Eq 47.
To this extent derivation of Eq 47 yields:

u = tan c

(50a)

or

A comparison of the analytical calculation


of the foundation moment and vertical
displacement to the results of the finite
element analysis is shown in Fig 13 for a strip
footing ( 2b = 2 m, h = 5 m, N = 500 kN) on
inelastic soil layer over rigid bedrock
( E = 100 MPa and su = 100 kPa ) .
A close agreement of the analytical results to
those obtained from numerical analysis is
attained throughout the range of the rocking
rotation.

dM
N 2b2
=
N h =0
d
4K m 2

Nh
Mu = Nb 1

K m

(49)

Eventually, the ultimate moment is:

181

to uplift, (b) M correlation, and (c) the effective


width 2 with respect to the rocking angle and
comparison with the analytical prediction.

overturning moment and the vertical


displacement were extracted as a function of
the rocking angle (a) for elastic soil, and (b)
for inelastic soil. From the latter case,
analytical equation of the failure curve in the
nondimensional N M plane was obtained,
incorporating both geometric and material
nonlinearities.

E = 100 MPa
400

M : kNm

N = 500 kN,

500

E = 100 MPa,

2b = 2 m,

0.5

h = 5 m,

300
200
100

120

0
0

gap force: kN

100

0.05

0.1

0.15

0.2

: rad

80

xp
0.2

60

E = 100 MPa

40

0.15

20
-1

-0.5

0.5

wb : m

0
1

distance from axis: m

0.1
0.05

(a)
0

400

-0.05
0

M: kNm

300

0.02

0.04

0.06

0.08

0.1

: rad

200

500

Muplift

E = 100 MPa

100

400

0.5

0.6

0.7

0.8

0.9

M : kNm

0
1

300
200

(b)
100

2.00

0
-0.016

-0.014

2: m

-0.012

-0.01

-0.008

wb: m

1.50


M = Km
+ Nb(1 )cos Nh sin
1

1.00
Analytical curves
FE curves

0.50

Rigid soil boundary

0.00
0

0.005

0.01

0.015

w b w b + b (1 ) ln
uplift + ( uplift )

0.02

: rad

1,

> uplift

uplift

(c)

Fig 13: Analytical curves of a rigid strip footing on


inelastic soil and comparison with twodimensional finite element results.

Fig 12: Nonlinear rocking response of a strip


footing on inelastic soil, calculated with twodimensional finite element analysis: (a) distribution
of soil reactions in characteristic increments prior

182

Crmer, C., Pecker, A. and Davenne, L. [2002].


Modeling of nonlinear dynamic behaviour of a
shallow strip foundation with macro-element,
J. Earth. Engng, 6(2).

ACKNOWLEDGEMENTS

The major part of the present study was


funded by the Greek General Secretariat for
Research and Technology through the
research project X-SOILS. Also part of this
work was conducted in terms of the EU
research project QUAKER, funded through
the EU Fifth Framework Programme:
Environment, Energy, and Sustainable
Development, Research and Technological
Development Activity of Generic Nature: the
Fight against Natural and Technological
Hazards (contract number: EVG1-CT-200200064).

FEMA 356 [2000]. Prestandard and commentary


for the seismic rehabilitation of buildings,
FEMA, Washington, D.C.
Gajan S, Kutter B, Phalen J, Hutchinson T and
Martin G. [2005]. Centrifuge modelling of loaddeformation behaviour of rocking shallow
foundations, Soil Dyn. & Earth. Engrg 25, 773783.
Gazetas, G. [1991]. Foundations vibrations,
Foundation Engineering Handbook ch 15 (H.Y. Fang, van Nostrand Reinhold, NY).
Gazetas G., Apostolou M. and Anastasopoulos I.
[2003], Seismic Uplifting of Foundations on
Soft Soil, with Examples from Adapazari (Izmit
1999, Earthquake), BGA International Conf. on
Foundations Innovations, Observations,
Design & Practice in the University of Dundee,
Scotland, September 25, pp. 37-50.

REFERENCES

Allotey, N. and El Naggar, M. H. [2003]. Analytical


moment-rotation curves for rigid foundations
based on a Winkler model, Soil Dyn. & Earth.
Engrg 23, 367-381.

Gazetas, G. and Apostolou, M. [2004]. Nonlinear


soilfoundation interaction: Uplifting and soil
yielding, Proc. 3rd Joint US-Japan Workshop
on SSI, Menlo Park CA.

Apostolou M., Gazetas G., Makris N.,


Anastasopoulos I. [2003], Rocking of
Foundations under Strong Seismic Excitation.
Proc. Fib International Symposium on Concrete
Structures in Seismic Regions, Athens, May
2003.

Gazetas G., Apostolou M., and Anastasopoulos I.


[2004], Seismic Bearing Capacity and Uplifting
Foundations : Adapazari 1999, Proc. 5th Int.
Conf. on Case Histories in Geotechnical
Engineering, New York, April 13-17.

Apostolou, M., Thorel, L., Gazetas, G., Garnier, J.


and Rault, G. [2007]. Physical and numerical
modelling of soil-footing-structure under
lateral cyclic loading, Proc. 4th ICEGE,
Thessaloniki, Greece.

Gazetas, G. and Apostolou, M. [2006]. Numerical


modelling of shallow foundations, Quaker
program, Final report on Topic B2.

ATC 40 [1996]. The seismic evaluation and retrofit


of concrete buildings, Applied Technology
Council. Vols. 1 and II, Palo Alto, CA.

Gerolymos, N., Apostolou, M., and Gazetas, G.


[2007], Neural network analysis of overturning
response under nearfault type excitation,
Earthquake Engineering and Engineering
Vibration, Vol. 4, No. 2, pp. 213-228.

Bartlett, P. [1979]. Foundation rocking on a clay


soil, M.E. Thesis, Report No. 154, School of
Eng., Univ. of Auckland, 144 pp.

Hibbitt,
Karlsson
and
Sorensen
[2002].
Abaqus/Standard Users Manual, Version 6.3,
Hibbitt Providence, Rhode Island.

Bransby, F. and Randolph, M. [1997]. Shallow


foundations subjected to combined loadings,
Proc. ISOPE.

Houlsby G.T. and Puzrin A.M. [1999]. The bearing


capacity of strip footing on clay under
combined loading, Proc. Royal Society, 455A:
893-916.

Butterfield, R. and Ticof, J. [1979]. The use of


physical models in design, Discussion. Proc.
7th Eur. Conference Soil Mechanics, Brighton,
4, 259-261.

Martin, G. and Lam, I. [2000]. Earthquake


resistant design of foundations Retrofit of
existing
foundations,
Proc.
GeoEng
Conference, Melbourne.

Butterfield, R. and Gottardi, G. [1994]. A complete


three-dimensional failure envelope for shallow
footings on sand, Gotechnique 44, No. 1,
181-184.

183

Nova, R. and Montrasio, L. [1991]. Settlements of


shallow foundations on sand, Gotechnique
41 (2), 243-256.
Pecker, A. [1998] Capacity design principles for
shallow foundations in seismic areas, Keynote
lecture, Proc. XIth Europ. Conf. on Earthquake
Engrg. Eds. Bisch, Labb, Pecker Balkema.
Priestley, M. J. N., Evison, R. J. and Carr, A. J.
[1978]. Seismic response of structures free to
rock on their foundations, Bull. New Zealand
Nat. Soc. Earth. Engrg. 11(3), 141-150.
Sadowsky,
M.
[1928].
Zweidimensionale
probleme der Elastizittstheorie, Zeit. Angew.
Math. Mech. 8(2), 107-121.
Siddharthan, R., Ara, S., Norris, G. [1992]. Simple
rigid plastic model for seismic tilting of rigid
walls, J. Struct. Engng, ASCE 118(2), 469487.
Somerville, P. G. [1998]. The amplitude and
duration effects of rupture directivity in near
fault ground motions, Geot. Earth. Engrg. Soil
Dyn. III ASCE, eds. P. Dakoulas, M. K. Yegian
and R. D. Holtz.
Quaker project website:
www.dundee.ac.uk/civileng/quaker/index.htm.

184

Natural Period and Effective Damping


of Simple Structures on Footings and Piles
A. Maravas, G. Mylonakis and D. L. Karabalis
University of Patras, Greece

Abstract
Novel analytical solutions are presented for single degree-of-freedom (SDOF) oscillators founded on footings
and piles on compliant ground. First, exact formulas for the fundamental natural period of the above
structures, encompassing the frequency dependence of the various impedance terms, are derived. Second,
closed-form solutions for the corresponding damping coefficients are derived. It is shown that the common
approximation of neglecting higher-order terms involving products of damping coefficients is unnecessary and
potentially inaccurate for highly-damped soil-structure systems. Third, the influence of foundation mass on the
period and damping of the system is incorporated. To address the issue of coupled swaying-rocking
oscillations at the pile head, the reference system is translated to the depth below the pile head where the
resultant soil reaction to the pile is applied, to ensure a diagonal foundation impedance matrix. Fourth, the
amounts of radiation damping generated from a single pile and a surface footing are compared. To this end, a
new concept of statically and geometrically equivalent SSI systems is introduced. It is shown that a structure
founded on a pile may generate twice the amount of radiation damping produced by a similar structure on a
spread footing. Results are provided in ready-to-use graphs and charts that elucidate the salient features of
the problem and can be directly implemented in design. The paper complements and extends the seminal
studies in the subject by Parmelee, Veletsos, Bielak and co-workers.

(1985), (2) to present a novel solution for


determining the fundamental natural period and
effective damping of a simple oscillator
supported on a surface foundation, (3) to
extend the solution to encompass pile
foundations, (4) to present results for typical
structural systems on compliant ground, and (5)
to compare radiation damping generated from a
surface foundation and a pile foundation.

INTRODUCTION

Knowledge in the subject of dynamic SoilStructure Interaction (SSI) has been derived
mainly from studies of structures on mat
foundations, during the last forty years. The
seismic response of structures on pile
foundations has received considerably less
research attention. More importantly, the results
of these efforts have not yet lead to established
design methods and/or code provisions, such
as the simple methods developed for structures
on surface foundations (NEHRP-03, EC-8).
Therefore much is yet to be learned on the
subject before a comprehensive understanding
is developed on the role of basic problem
parameters on the seismic response of pilesupported systems.
The goals of this article are: (1) to review
available methods on the subject with emphasis
on the design-oriented solutions by Veletsos
and co-workers (1974, 1975, 1977) and Wolf

STRUCTURE ON SURFACE FOOTING

The classical approach for elastodynamic


analysis of soil-structure interaction aims at
replacing the actual structure by an equivalent
simple oscillator supported on a set of
frequency-dependent springs and dashpots
accounting for the stiffness and damping of the
soil medium. This model has been adopted by
several researchers, including Parmelee (1967),
Veletsos et al (1974, 1975, 1977), Jennings &
Bielak (1973), Wolf (1985) and, more recently,

185

Aviles et al (1996, 1998). A brief overview of


available methods leading to closed-form
solutions is presented below. Based on these
procedures,
a
novel,
accurate
and
straightforward scheme for analyzing the
problem is presented.

( ) =

K x = ax K ,

The system studied is shown in Figure 1. It


involves a simple oscillator on flexible base
representing a single storey structure, or a multi
storey structure after a pertinent reduction of its
degrees-of-freedom (e.g., considering that the
mass is concentrated at the point where the
resultant inertial force acts).
The structure is described by its stiffness k,
mass m, height h, and damping ratio , which
may be either viscous or linearly hysteretic. The
foundation consists of a rigid surface circular
footing of radius r resting on a homogeneous,
linearly elastic, isotropic halfspace described by
a shear modulus Gs, mass density s, Poissons
ratio s, and hysteretic damping ratio s.
Foundation stiffness is modeled by frequencydependent springs Kx and K representing
stiffness in translational and rocking oscillations,
respectively. Following Veletsos & Nair (1974),
to ensure uniform units in all stiffness terms, K
is expressed by a translational vertical spring
acting at distance r from the center of the
footing. Damping is modeled by a pair of
dashpots Cx, C, attached in parallel to the
springs, representing energy loss due to
hysteretic action and wave radiation in the soil
medium. In the present formulation, the
influence of foundation embedment on the
response is neglected.
The dynamic impedance K * ( ) along any
degree of freedom of the system is defined
according to the formula

K ( ) = K + i C = K (1 + 2 i )

(2)

For the model in Figure 1a, the foundation


springs and dashpots can be expressed by the
formulas proposed by Veletsos & Meek (1974):

Classical solutions

Im( K * ) C
=
2 Re( K * ) 2 K

Cx = x

Kr
,
Vs

K = a K
C =

(3)

Kr
Vs

(4)

where Vs denotes the propagation velocity of


distortional waves in the halfspace and K is the
static horizontal stiffness of the foundation
defined by

K=

8
Gs r
2 s

(5)

x, , x and are dimensionless factors that


depend on Poissons ratio for the halfspace
material, and the dimensionless frequency
a0 =

(6)

Vs

Under seismic excitation, the system


deflects as shown in Figure 2. The translation of
the mass relative to ground is composed of
three parts: (1) horizontal translation due to
swaying motion of footing ux, (2) horizontal
translation due to rocking motion of footing, u
and (3) horizontal deflection of column, uc.
Based on these definitions, the impedance of
the system is defined as:

j*
K

(1)

P
i 1 + 2ii
=K
u x + u h + uc

(7)

i and i denote the apparent stiffness


where K
and damping coefficients at elevation h.
The response of the soil-structure system
depends on the mechanical properties of the
foundation, the soil, the superstructure and the
characteristics of the excitation. These are
summarized in the following dimensionless
parameters (Veletsos et al 1974, 1975, 1977):
(i) The wave parameter

in which K is the real part of the impedance,


C is the corresponding imaginary part,
is the cyclic excitation frequency, and i
(= 1 ) the imaginary unity. is an energy
loss parameter, analogous (yet not
identical) to the viscous damping coefficient
of a simple oscillator.

186

Vs
fc h

Solution by Wolf (1985)

(8)

The system considered by Wolf is identical


to that shown in Figures 1 and 2. The main
difference with the Veletsos approach is that
frequency-independent moduli defined by the
values ax = 1, x = 0.575 , a = 0.15, = 0.15 , are
adopted for the foundation, and that the
response of the system is evaluated by directly
solving a set of three simultaneous governing
equations of the system.
The properties of the replacement oscillator
in this solution are given by

where fc = k / m / 2 denotes the natural


frequency of the fixed base structure.
(ii) The relative mass density for the structure
and the soil

m
s hr 2

(9)

(iii) The damping ratio of the structure for fixed


base conditions.
(iv) The slenderness ratio (h/r).
(v) The Poissons ratio s of the soil.
(vi) The hysteretic damping ratio s of the soil.

 =

The aim of these solutions is to connect the


properties of the soil-structure system ( Ti , i )
with the properties of the fixed base structure
(, ), so that the influence of soil-structure
interaction on the dynamic behavior of the
structure can be elucidated. This connection is
expressed by the following pair of equations
(Veletsos 1977):
2

( )

( )

i
i = j0 + T T

cyclic natural frequencies of the system under


rocking oscillations of the base (superstructure
assumed rigid), swaying oscillations of the base,
and
oscillations
of
the
superstructure
(foundation assumed rigid), respectively. Note
the simpler form of Eqn (14) as compared to
Eqn (12).
Notwithstanding the theoretical significance
and practical appeal of the above methods,
they both can be criticized on the following
important aspects:

(11)

(a) Both methods neglect products of damping


ratios (i x j) as negligible higher order
terms. This approximation is questionable
for highly-damped SSI systems.
(b) The effective damping in the Veletsos
approach arises from an approximate
procedure leading to an expression
containing imaginary terms (Eqn 12). This
limits significantly its suitability for practical
applications.
(c) Structural damping in the Veletsos solution
is strictly of viscous nature.
(d) Frequency dependence of foundation
springs and dashpots in the Wolf approach
is neglected.

radiation damping of the footing. The latter is


given by (Veletsos and Nair, 1975)

( )

4 Ti
2 3 T

(14)

x = K x / m , c = k / m define the uncoupled

where represents the damping of the structure


(assumed to be of viscous nature) and j the

j0 =

(13)

In the above equations, = K r 2 / mh 2 ,

1 a h
1+3 s x (10)
2s a r

k kh 2
+

K x K


i 2
i 2
i 2
i 2

+
+


x
s

c
c
x

Solution by Veletsos and co-workers (1974, 1975,


1977)

k Kh
2
Ti =T 1+ 1+ x =T 1+ s
Kx K
2 ax 2 h
r

i = 2 1+

(2 s )x r 2
3(1 s )

(12)
+
ax (ax + ia0 x ) h a (a + ia0 )

The method is based on setting the resonant


period and peak pseudo-acceleration of the
actual infinite-degree-of-freedom elastodynamic
system equal to that of an equivalent simple
oscillator. More discussion is given below.

187

(e) Structural damping in the Wolf solution is


strictly of hysteretic nature.
(f)
Both solutions employ rather complex
procedures involving either equivalence
of responses of different dynamic systems
(Veletsos), or solutions, of simultaneous
linear equations (Wolf).
(g) In both solutions, foundation mass and
rotational inertia are neglected.

2
2

i2
1 + 4 i
1 + 4 i
i 2 = 1 + 4

+
+
2
2
2
2
2
2
x 1 + 4 x
c (1 + 4 )
1 + 4

(18)

The above equations are exact solutions to the


problem, in the sense that no approximations
apart from those typically involved in
numerically
evaluating
the
foundation
impedances are involved. Given the frequencydependnet nature of foundation impedances, an
iterative procedure is generally required to
implement these formulas. Note that omitting
the products i2, the above solutions duly reduce
to those in Eqns (14) and (13), respectively.

Proposed exact procedure

In this section a simple exact solution to the


problem shown in Figures 1 and 2 is presented.
The solution contains no approximations in the
derivation of the fundamental natural period and
effective damping of the system. Furthermore,
the
exact
frequency-varying
foundation
impedances may be employed.
Mention has already been made that the
total horizontal deflection of the system can be
decomposed as sum of the three modular
displacements shown in Fig 2, i.e.,

Influence of foundation mass

he influence of foundation mass and inertia


(mf, If) on dynamic behavior of the system
presented in Figures 1 and 2 can be
incorporated by using modified frequencydependent springs Kx and K according to:

(15)

K x = Kx mf 2

(19)

This implies that the associated compliances


can be viewed as complex springs attached in
parallel and, thereby, the dynamic impedance
of the system can be expressed through the
summation rule

K = K f 2

(20)

ut = uc + u x + u

where Kx and K are given by Eqn (3). This


modification does not alter the form of the
solution in Eqs (17) and (18), yet introduces an
additional dimensionless parameter, , which
expresses the ratio of footing mass to
superstructure mass
m
= f
(21)
m

1
1
1 h
1
= * + * + *
*
i
Kx
K r k
K

(16)

in which the associated impedances are


complex valued (to account for phase
differences in the response of the springs) and
frequency
dependent.
Substituting
each
complex impedance term in Eq. (16) by its
representation according to Eq. (1) yields the
exact damping and natural frequency of the
system as (Maravas 2006)

i =

x 2 1 + 4 x 2

x 1 + 4 x
2

2 1 + 4 2

1 + 4
2

which varies typically between 0 and 1. More


discussion is given below.
Parametric analysis and comparisons with
classical methods

Comparative graphs for the variation of the


i / ) and the effective
ratio Ti / T (inverse of
c
damping of the system i , versus the

c 2 (1 + 4 2 )
1
c (1 + 4 2 )

(17)

slenderness ratio h/r, are presented in Figure 3.


Using the proposed exact procedure, the
influence of the relative mass ratio and the
material hysteretic damping ratio s on the
period and effective damping of the soilstructure system is presented in Figure 4. For

188

the parametric analysis, values of factors x, ,


x and correspond to a Poissons ratio of
0.45.
It is evident from Figures 3(a,b) that the
results from the method of Veletsos are in
relative agreement with those obtained by the
proposed exact procedure. Also in Figures
3(c,d), we observe that results from the method
of Wolf are quite different from those of the
proposed procedure. This is mainly due to the
assumption of frequency-independent springs
and dashpots adopted by Wolf.
The above observation is justified by the
fact that for low values of ratio 1/ (i.e., low
values of c) the results of the two methods are
virtually identical. It is also apparent that the
results obtained by Wolf loose accuracy with
decreasing h/r.
Figures 4(a,b) show that the variation in
relative mass ratio affects significantly the
period and damping of the soil-structure system.
More specifically, increasing leads to more
flexible systems and higher values of damping
ratio. The hysteretic damping ratio of soil s,
does not affect much the system period,
especially for tall structures (h/r=5), as shown in
Figure 4(c). On the other hand, it affects
considerably the effective damping of the
system, as shown in Figure 4(d).
Figures 5(a,b) show that the variation of
mass ratio affects the natural period and
damping of soil-structure systems with low
values of h/r. On the contrary, it does not affect
systems with high values of the slenderness
ratio, as shown in Figures 5(c,d).
STRUCTURE ON PILE FOUNDATION

The case of a structure on a single pile


foundation is investigated next. The problem is
treated using a variance of the method for
spread footings presented in the previous
section. The amounts of energy radiated by a
pile- or a footing-supported structure are
compared via by the concept of statically
equivalent SSI systems, introduced in this work.
Soil-pile-structure system and method of
analyses

The system studied is shown in Figure 6. It


consists of the linear elastic SDOF structure
described in previous section, founded on a
single flexible, circular solid pile of Youngs

189

modulus Ep, diameter d, mass per unit length


mp, and length L, which is considered to be
greater than the pile effective length Le.
Accordingly, the pile can be considered
infinitely long. The soil is considered as a
linearly
elastic
homogeneous
isotropic
halfspace, as described above.
Soil-pile system can be represented by
three dynamic impedances K xx* , K rr * , and K xr * ,
corresponding to swaying, rocking, and crossswaying-rocking of the pile head, respectively.
In this study, analytical expressions for the
dynamic impedances are used, as derived by
Novak, (1974) and Mylonakis (1995).
K xx = 4 E p I p 3 , K xr = 2 E p I p 2 , K rr = 2 E p I p
*

k x m p 2 + i cx
=

4E p I p

(22)

1/ 4

(23)

where is a wave number parameter, Ip is the


moment of inertia of the pile cross section, and
kx and cx are the moduli of distributed springs
and dashpots along the pile. The latter are
given by (Gazetas & Dobry 1984):

k x = Es ,

cx = 6a -1/4
op sVs d + 2

s kx

(24)

where aop=2ao (where the footing radius r in Eq.


(6) is replaced by the pile diameter d) and is
the Winkler factor, given as function of pile-soil
stiffness ratio Ep/Es (Dobry et al, 1982)

Ep

Es

= 1.67

0.053

(25)

Under horizontal loading the soil reacts in


the manner shown in Figure 6(b). The resultant
of the distributed reactions is applied at depth e
below the pile head. Since the reference
system is anchored at the pile head, a cross
i xr is
swaying-rocking impedance term K
necessary for modeling the compliance of the
foundation. This term is not compatible with the
analysis of the spread footing presented earlier.
In order to overcome this problem, the
reference system can be translated to a depth
e, where the total soil reaction is applied. In this
manner, the cross impedance K xr * vanishes

and the impedance matrix of the pile becomes


diagonal, as shown in Figure 6(c). This
transformation is approximate, as it implies a
rigid pile between depths z = 0 and z = e.
However, this introduces little error, since e is
usually small (up to a few pile diameters)
compared to the overall pile length. The
transformed impedances K xxe* and K rre* are
given by the expressions
*

K xxe = K xx , K rre = K rr 2 K xr e + K xx e 2 , K xre = 0

Results of parametric analyses

The response of the soil-pile-structure


system depends on the properties of the pile,
the supporting soil, the superstructure and the
excitation. These properties are included in
above dimensionless parameters, as in the
case of the structure supported on surface
footing.
In Figure 7, results obtained with the use of
the proposed exact procedure are presented.
Specifically, Figure 7(a) presents the influence
of the slenderness ratio (h/d) on the natural
period of the interacting system. Evidently, the
behavior is similar to the behavior of the soilfooting-structure system presented earlier. The
same observation holds for the effective
damping, as shown in Figure 7(b).
The influence of pile-soil stiffness ratio Ep/Es,
on the properties of the system is significant, as
shown in Figures 7(c,d). For relatively flexible
piles (low values of Ep/Es), the soil-pile-structure
system is more flexible and dissipates larger
amounts of energy.

(26)

where
*

K
1
e = xr * =
2
K xx

(27)

is the aforementioned eccentricity.


The transformation shown in Figure 6(c)
allows usage of the exact procedure developed
for the analysis of structures on surface
footings. Thus, the natural frequencies i of the
soil-pile-structure system required by Eqs. (17)
and (18) are computed by Maravas (2006) as

K xx
K rr
, =
m
m( h + e) 2

x =

Comparison of two SSI systems

In this section, the concept of statically and


geometrically equivalent interacting systems is
introduced, in an effort to compare the damping
of systems on pile or surface footing
foundations. To achieve this, the lateral
stiffness of the two SSI systems should be
comparable at the elevation of mass m.
Accordingly, the two systems are geometrically
equivalent, i.e., h/d=h/2r, if the ratio of Youngs
moduli for the soil (Maravas 2006):

(28)

where
K xx = ( 4 E p I p )

1/ 4

K rr =

( k m 2 ) 2 + ( c ) 2
p
x
x

3/8

3
cos (29)
4

1/ 8
2
3/ 4
1
2
1
4 E p I p ) ( k x m p 2 ) + (cx ) cos
(

4
4

cx

k x m p 2

= Arc tan

(30)
1/ 4

(f)
Es
1 3
=

( p)
2 16
Es

(31)

= tan
2
4

(2 vs )(1 + vs ) 1 + 4 ( h / d )

1/ 4
1/ 4

16 E p
h

1 + 1 + 2
( p)
d

Es

(33)

In the above equation, symbol f denotes


footing and symbol p pile. Because a pile is
much stiffer than a footing, the supporting soil
of the footing must have a modulus of
(f)
elasticity Es quite larger than the modulus of

The corresponding damping ratios i are given


by the expressions (Maravas 2006)
x = tan ,
2
4

1/ 4

Ep
( p )
Es

(32)

( p)

elasticity of the soil around a pile Es .


Using the proposed exact procedure, the
effective damping of the two SSI systems is
compared for different ratios h/d and Ep/Es, as
shown in Figures 8a and 8b. Furthermore,

in which is defined by Eq. (31).

190

comparative graphs of radiation damping for the


two foundation types are presented in Figures
8c and 8d. Evidently, a structure on a pile may
experience as much as 3 times the amount of
damping generated from the same structure on
a spread footing (Figure 8a). This observation is
justified in view of the tow-dimensional nature of
wave propagation around a pile, which results
to much higher energy dissipation.

REFERENCES
Aviles, J. & Perez-Rocha, L.E.(1996) Evaluation of
interaction effects on the system period and the
system damping due to foundation embedment
and layer depth Soil Dynamics & Earthquake
Engineering. 15(11), 27.
Aviles, J. & Perez-Rocha, L. E (1998). Effects of
foundation embedment during building-soil
interaction, Earthquake Engineering & Structural
Dynamics, 27(12), 1523-1540.

CONCLUSIONS

Dobry, R., Vicente, E., O'Rourke, M. J., & Roesset,


J. M (1982). "Horizontal stiffness and damping of
single piles", J. Geotech. Engng Div., ASCE,
108(3), 439-459.

A novel analytical procedure for determining


the dynamic characteristics of simple structures
founded on surface footings and piles was
presented. Using the proposed methodology,
the influence of common assumptions on the
computation of the mechanical properties of
such systems was elucidated. Results were
provided in ready-to-use graphs and charts that
elucidate the salient features of the problem
and can be directly implemented in design. By
introducing the concept of statically and
geometrically equivalent SSI systems, the
amounts of radiation damping generated from a
single pile and a footing were compared. The
main conclusions of the study are:
(1) The proposed solution is simpler, more
accurate, and more general than the classical
methods by Veletsos, Bielak and co-workers.
(2) The common approximation of
neglecting
higher-order
terms
involving
products of damping coefficients may be
inaccurate for highly-damped SSI systems.
(3)
The
foundation
mass
affects
considerably dynamic SSI response of systems
with low values of slenderness ratio h/r.
(4) The proposed analysis can easily
incorporate
embedded
foundations,
by
translating the reference system to the depth
below the surface where the resultant soil
reaction is applied. This ensures a diagonal
foundation impedance matrix and greatly
simplifies calculations.
(5) A structure founded on a pile may
generate 100% more radiation damping than a
similar structure on a spread footing. The
difference gets more pronounced with highfrequency, squatty structures on stiff soil.

Eurocode EC-8 (1990) Structures in seismic


regions, Part 5: Foundations, retaining structures,
and
geotechnical
aspects,
Brussels:
Commission of the European Communities
Gazetas, G and Dobry, R. (1984) Simple Radiation
Damping Model for Piles and Footings, Journal
of Engineering Mechanics, Vol. 110, No. 6, June,
pp. 937-956
Gazetas G., Anastasopoulos I., Gerolymos N.,
Mylonakis G., & Syngros C. (2005), The
Collapse of the Hanshin Expressway (Fukae)
Bridge, Kobe 1995 : SoilFoundationStructure
Interaction, Reconstruction, Seismic Isolation,
Entwicklungen
in
der
Bodenmechanik,
Bodendynamik und Geotechnik, Festschrift zum
60. Geburstag von Univ.-Professor Dr.-Ing.habil.
Stavros A. Savidis, Frabk Rackwitz, Springer,
pp. 93-120.
Gerolymos, N., (1997), Effective period and damping
of bridge piers on pile groups, Diploma Thesis,
School of Civil Engineering, National Technical
University, Athens (in Greek).
Gerolymos, N., Gazetas, G. & Mylonakis, G.
Fundamental natural period and effective
damping of pile-supported bridge piers, Proc.,
11th European Conference on Earthquake
Engineering, Balkema, Rotterdam, 1998
Jennings, P.C. & Bielak, J.(1973) "Dynamics of
building-soil interaction", Bulletin of the
Seismological Society of America, 63(1) 9-48.
Maravas, A. (2006). Discrete model for dynamic
soil-structure interaction of structures on rigid
surface or pile foundations, MS Thesis,
University of Patras, Greece (in Greek).
Mylonakis, G. (1995) Contributions to the static and
seismic analysis of pile-supported bridge piers,
Ph.D. Dissertation, SUNY-Buffalo.
NEHRP (1997) Recommended Provisions for
Seismic Regulations for New Buildings and other

191

Veletsos, A.S. & Meek, J.W. (1974). Dynamic


behavior of building-foundation systems, Earthq,
Engng & Struct, Dyn,, 3(2), 121-138.

Structures, Building Seismic Safety Council,


Washington, D.C.
Novak, M. (1974) Dynamic stiffness and damping of
piles, Canadian Geotech. Journal, 11, 574-591.

Veletsos, A.S. & Nair, V.V. (1975) Seismic


interaction
of
structures
on
hysteretic
foundations, Journal of Structural Engineering,
ASCE, 101(1), 109-129.

Parmelee, R. (1967). Building-foundation interaction


effects, Journal of Engineering Mechanics
Division, ASCE, 93(EM2), 131-152.

Veletsos, A. S. (1977) Dynamics of StructureFoundation Systems, in: Hall, W. J. (ed.),


Structural & Geotechnical Mech., Prentice-Hall.

Sextos, A, Kappos. A, Pitilakis, K. Inelastic dynamic


analysis of RC bridges accounting for spatial
variability of ground motion, site effects and soilstructure interaction phenomena. Part 2:
Parametric study, Earthquake Engineering and
Structural Dynamics 32 (4), pp. 629-652

Wolf, J. P. (1985). Dynamic


Interaction, Prentice Hall.

Stewart, JP, Fenves, GL, Seed, RB. (1999).


Seismic Soil-Structure Interaction in Buildings I:
Analytical Methods, Journal of Geotechnical and
Geoenvironmental Engineering, ASCE, 125(1).

192

Soil-Structure

FIGURES

u
m

C
r

(b)

(a)

Fig 1: (a) Structure idealized by a stick model, (b) Reduced single degree-of-freedom model

ux

Cx

Kx

Fig 2:. Deflection diagram for soil-structure system

193

uc

h/r=5

2.2

h/r=3

0.4

0.2

2.0

0.1

1.8

1.6

h/r=1

0.3

h/r=1

Exact procedure
Veletsos (1977)

h/r=2

0.05
0.04

0.03

1.4

0.02

1.2

0.01

Exact procedure
- - - - - Veletsos (1977)
h/r=5

1/

1.0
0.1

(a)

0.2

0.3

0.4

0.5

= 0.05, s = 0.45, = 0.15, s = 0

1.0

0.005
0.02

0.6

(b)

h/r=1

0.05

0.10

0.25

0.50

1/

0.75 1.00

= 0.05, s = 0.45, = 0.15, s = 0


0.5

h/r=0.33

0.8

0.4

h/r=5

c 0.6

h/r=2

h/r=1

0.3

0.4

0.2

Exact procedure
Wolf (1985)

Exact procedure
Wolf (1985)

h/r=0.33

0.2

0.0
0.1

(c)

h/r=2

0.1

h/r=5

1/

0.0

1/

0.1

(d) = 0.02, s = 0.45, = 0.15, s = 0.05


Fig 3:. Comparison of proposed exact solution with those of Veletsos (1977) and Wolf (1985)

= 0.02, s = 0.45, = 0.15, s = 0.05

3.5

0.7

3.0

2.5

=0
0.10
0.20
0.30
0.60
1

2.0

1.5

1.0

0.5
0.2

(a)

0.4

0.6

0.8

1.0

1/

= 0.02, s = 0.45, h/r = 1, s = 0.05

0.01
0.1

(b)

3.0
0.1

1/

0.5

h/r=1

= 0.02, s = 0.45, h/r = 1, s = 0.05

h/r=5

3.5

=0
0.10
0.20
0.30
0.60
1

0.1

2.5

s = 0
0.05
0.20

si

2.0

s = 0
0.05
0.20

1.5

1.0
0.2

0.4

0.6

0.8

0.01

1/

1.0

0.1

1/

(d) = 0.02, s = 0.45, h/r = 5, = 0.15


(c) = 0.02, s = 0.45, = 0.15
Fig: 4. Parametric results using the proposed exact procedure. (a) Period of SSI system as function of , (b)
Effective damping of SSI system as function of , (c) Period of SSI system as function of soil damping
ratio s, (d) Effective damping of SSI system as function of s

194

2.5

1.0
0.9
0.8
0.7

2.0

0.6
0.5
0.4

1.5

=0
0.20
0.50
0.80
1

1.0
0.2

(a)

0.4

0.6

0.8

1.0

=0
0.20
0.50
0.80
1

0.3
0.2
0.1

1/

0.0
0.2

= 0.02, s = 0.45, h/r = 0.33, s = 0.05

(b)

2.5

0.4

0.6

0.8

1.0

1/

= 0.02, s = 0.45, h/r = 0.33, s = 0.05

0.14

0.12

2.0

0.10

0.08

=0
0.20
0.50
0.80
1

1.5

0.06

=0
0.20
0.50
0.80
1

0.04

0.02

1.0

(c)

1/
0.4
0.6
0.8
1.0
= 0.02, s = 0.45, h/r = 2, = 0.15, s =0.05

0.00

0.2

(d)

1/
0.2
0.4
0.6
0.8
1.0
= 0.02, s = 0.45, h/r = 2, = 0.15, s= 0.05

Fig 5:. Influence of foundation mass on the dynamic behavior of SDOF systems

i xx
K

i xr
K
i rr
K

I=

ix
K

i
K

Ep I p
d

(a)

(c)

(b)

Fig 6:. (a) Model of pile-supported-structure, (b) Distribution of soil reactions due to horizontal loading, (c)
Reduced model with two dynamic impedances

195

3.0

0.30

0.25

h/d = 1
5
10

2.5

0.20

2.0

0.15

0.10

1.5

h/d = 1
5
10

0.05

1.0
0.2

0.4

0.6

0.8

1.0

1/

0.00
0.2

(a) = 0.02, s = 0.45, = 0.15, s = 0.05, p/s = 1.40, Ep/Es = 100

0.6

0.8

1.0

1/

(b) = 0.02, s = 0.45, = 0.15, s = 0.05, p/s = 1.40, Ep/Es = 100

2.5

0.12

Ep/Es = 100
1000
10000

0.10

2.0

0.4

0.08

i
1.5

0.06

0.04

Ep/Es = 100
1000
10000

0.02

1.0
0.2

0.4

0.6

0.8

1.0

1/

0.00
0.2

(c) = 0.02, s = 0.45, = 0.15, s = 0.05, p/s = 1.40, h/d = 5

0.4

0.6

0.8

1.0

1/

(d) = 0.02, s = 0.45, = 0.15, s = 0.05, p/s = 1.40, h/d = 5

Fig 7:. (a) System period as function of h/d, (b) System damping as function of h/d. (c) System period as
function of Ep/Es, (d) System damping as function of Ep/Es
0.12

0.28
0.24

0.10

Pile
Footing

0.20

Pile
Footing

0.08

0.16

0.06
0.12

0.04

0.08
0.02

0.04
0.00
0.1

0.2

0.3

0.4

(f)

(a) = 0.02, s = 0.45, h/d = 1, s = 0.05, p/s = 1.40, Ep/Es = 1000

0.00
0.02

0.08

0.12

0.16

(f)

(b) = 0.02, s = 0.45, h/d = 5, s = 0.05, p/s = 1.40, Ep/Es = 1000

(p)

1.0

0.04

1
(1/)
f

0.20

0.8
0.15

0.6

Pile
Footing

Pile
Footing
0.10

0.4
(f)

0.2

0.05

0.0
0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

0.00

(a0)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

(a0)

(c) s = 0.45, s = 0.05, p/s = 1.40, Ep/Es = 1000


(d) s = 0.45, s = 0.05, p/s = 1.40, Ep/Es = 1000
Fig 8: (a, b) Effective damping of a structure founded on pile and footing, (c) Radiation damping of a pile and
footing due to translational oscillation, (d) Radiation damping of a pile and footing due to rocking
oscillation

196

Interaction of Shallow and Deep Foundations


with a Rupturing Normal Fault
I. Anastasopoulos and G. Gazetas
National Technical University of Athens, Greece

Abstract
The paper studies the response and distress of shallow and deep foundations to a seismic
fault rupture emerging directly underneath. The developed numerical methodology has been
calibrated with centrifuge experiments. The outlined parametric results provide valuable
insight to the respective soilfoundation interplay, and could explain qualitatively the
observed behaviour in a number of case histories from recent earthquakes. It is shown that
rigid mat or box foundations may divert the rupture path and, if properly designed, survive
the rupture with only some unavoidable rotation. The role of piles is not nearly as clear, and
the paper highlights a possibly detrimental effect.

INTRODUCTION

field studies, centrifugal experiments, and


numerical / analytical modeling. Specifically:
Field studies of documented case
histories motivated our investigation and
offered material for calibration of the
theoretical methods and analyses,
Carefully
controlled
centrifugal
experiments helped in developing an
improved understanding of mechanisms
and in acquiring a reliable experimental
data base for validating the theoretical
simulations, and
Theoretical
methods
(analytical
or
numerical) calibrated against the above
field and experimental data offered
additional insight into the nature of the
interaction,
and
were
utilised
in
developing parametric results and design
aids.
This paper summarises some of the key
findings of these theoretical studies, which were
later supplemented with further analyses
pertaining to pile and caisson foundations.

Numerous cases of devastating effects of


earthquake surface fault rupture on structures
were observed in the 1999 earthquakes of
Kocaeli, Duzce, and Chi-chi. However,
examples of satisfactory, even spectacular,
performance of a variety of structures also
emerged. In some cases the foundation and
structure were quite strong and thus either
forced the rupture to deviate or withstood the
tectonic movements with some rigid-body
rotation and translation but without damage. In
other cases structures were quite ductile and
deformed without failing. Thus, the not-so-new
idea that a structure can be designed to survive
with minimal damage a surface fault rupture reemerged. The work presented here was
motivated by the need to develop deep
quantitative understanding of the interaction
between a rupturing normal fault and a variety
of foundation types. Much of the reported
research was conducted within the framework
of
a
Europian-Union
research
project
(QUAKER) with the participation of the
research teams from the University of Dundee,
Godynamique
et
Structure,
Studio
Geotechnico Italiano, and National Technical
University of Athens.
The study was based on an integrated
approach, comprising three interrelated steps:

FAULTRUPTURE PROPAGATION AND


INTERACTION WITH FOUNDATIONS

ITS

Statement of the Problem


It has long been recognised (Duncan &
Lefebvre, 1973; Bray 1990) that a strong

197

structure founded on/in soil can resist


successfully the loading induced by a rupturing
seismic fault. In the Kocaeli, Dzce-Bolu, and
Chi-Chi earthquakes of 1999 numerous
structures (single-storey and multi-storey
buildings, bankers, bridge piers, retaining
structures, electricity pylons, dams, tunnels)
were located directly above the propagation
path of the rupturing (normal, strike-slip,
reverse) faults. Some of these structures
exhibited a remarkably good behaviour. This
observation had a strong motivating influence
for our research effort.
For it became
immediately clear that the strict prohibition: Do
not build in the immediate vicinity of active
faults, which the prevailing seismic codes
invariably imposed, was unduly restrictive (and
in many cases meaningless).
Indeed, along the ground surface in the
freefield, ruptures are neither continuous, nor
do they follow precisely the surface outcrop of
pre-existing faults (Ambraseys & Jackson,
1984).
In addition to several geologic factors that
contribute to such behaviour, significant
appears to be the role of a soil deposit that
happens to overlie the rock base through which
the rupture propagates. If, where, and how
large will the dislocation emerge on the ground
surface (i.e. the fault will outcrop) depends not
only on the style and magnitude of the fault
rupture, but also on the geometric and material
characteristics of the overlying soils. Field
observations and analytical and experimental
research findings [Bray et al, 1994a, b; Cole &
Lade, 1984; Lade et al, 1984; Lazarte & Bray,
1995] show that deep and loose soil deposits
may even mask a small-size fault rupture which
occurs at their base; whereas by contrast with a
cohesive deposit of small thickness, a large
offset in the base rock will likely cause a distinct
fault scarp of nearly the same displacement
magnitude. One important finding of the above
studies is that the rupture path in the soil is not
a simple extension of the plane of the fault in
the base rock: phenomena such as diffraction
and bifurcation affect the direction of the
rupture path, and make its outcropping location
and offset magnitude difficult to predict.
Our interest here is not on the propagation
of a rupture within the soil, but on how a
structure sitting on top of the fault breakout
behaves. It turns out that a fascinating interplay

takes place between the propagating fault


rupture, the deforming soil, the differentially
displacing foundation, and the supported
structure. Two different phenomena take place.
First, the presence of the structure modifies the
rupture path. Depending on the rigidity of the
foundation and the weight of the structure, even
complete diversion of the fault path before it
outcrops may take place. Obviously, the
damage to a given structure depends not only
on its location with respect to the fault outcrop
in the free-field, but also on whether and by
how much such a diversion may occur.
Second, the loads transmitted from the
foundation on to the soil tend to compress the
asperities and smoothen the anomalies of
the ground surface that are produced around
the fault breakout in the freefield, i.e. when the
structure is not present. Thus, depending on
the relative rigidity (bending and axial) of the
foundation with respect to the soil, as well as on
the magnitude of the structural load, the
foundation and the structure will experience
differential displacements and rotation different
from those of the free-field ground surface.
This phenomenon, given the name FaultRuptureSoilFoundationStructure Interaction
[FRSFSI] by Anastasopoulos & Gazetas
(2007), is briefly elucidated in the sequel for
shallow and deep foundations.
Numerical Analysis and Results: Shallow
Foundations
The problem studied here is illustrated in Fig. 1.
We consider a uniform soil deposit of thickness
H at the base of which a normal fault, dipping at
an angle (measured from the horizontal),
produces
downward
displacement
(dislocation, offset) of vertical amplitude h.
The analysis is conducted in two steps. First,
fault rupture propagation through soil is
analysed in the free field, ignoring the presence
of the structure. Then, a strip foundation of
width B carrying a uniformly distributed load q
or a multistory framestructure is placed on top
of the freefield fault outcrop at a specified
distance s (measured from its corner), and the
analysis of deformation of the soilstructure
system due to the same base dislocation h is
performed. The analyses are conducted under
2-D plane-strain conditions evidently a
simplification, in view of the finite dimensions of

198

B
q

x
y
s

Hanging wall

free-field
rupture path

H H

Footwall

Figure 1: Configuration of the soilfoundation system subjected to a normal fault dislocation at the base rock.

detachment of the foundation from the bearing


soil (i.e. gap formation beneath the foundation).
The interface shear properties follow Coulombs
friction law, allowing for slippage. Both
detachment and slippage are important
phenomena for a realistic foundation model.
A typical result elucidating the interplay
between loose (Dr = 45%) soil, rupture path,
and a perfectly rigid foundation carrying a 4storey structure is given in Fig. 2. A base rock
dislocation of 2 m (5% of the soil thickness) is
imposed.
The
structure
is
placed
symmetricallystraddling the free-field fault
breakout (i.e. the foundation is placed with its
middle coinciding with the location where the
fault would outcrop in the free field). Yet, a
distinct rupture path (with high concentration of
plastic shearing deformation and a resulting
conspicuous surface scarp) is observed only in
the freefield. The presence of the structure
with its rigid foundation causes the rupture path
to bifurcate at about the middle of the soil layer.
The resulting two branches outcrop outside the
left and the right corner of the foundation,
respectively. The soil deformations around
these branches are far smaller and diffuse than
in the freefield, and the respective surface
scarps are much milder. Thanks to the
substantial weight of the structure and the
flexibility of the ground, the structure settles and
rotates as a rigid body. The foundation does
not experience any loss of contact with the
ground ; apparently, the foundation pressure is
large enough to eliminate any likely asperities
of the ground surface.
As a result of such behaviour, the structure
and its foundation do not experience any

a real structure in the direction parallel to the


fault. The relative location of outcropping is
varied parametrically through the distance S.
Comparing
soil
and
groundsurface
deformations in the two steps gives a first
picture of the significance of SFSI.
Among several alternatives that were
explored, the FE model shown in Fig. 2
produced results in excellent accord with
several centrifugal experiments conducted at
the University of Dundee for both steps of the
analysis (Anastasopoulos et al 2007 a, b). A
parametric investigation revealed the need for a
long (B = 4H) and very refined mesh (element
size of 0.5 m 1.0 m) along with a suitable slipline tracing algorithm in the region of soil
rupture and foundation loading. An elastoplastic
constitutive model with the Mohr-Coulomb
failure criterion and isotropic strain softening
was adopted and encoded in the ABAQUS
finite element environment. Similar models
have been successfully employed in modeling
the failure of embankments and cut slopes
(Potts et al, 1987, 1990). Modelling strain
softening was shown to be necessary; it was
introduced by suitably reducing the mobilised
friction angle mob and the mobilised dilation
angle mob with increasing plastic octahedral
shear strain. With all the above features, the FE
formulation is capable of predicting realistically
the effect of large deformations with the
creation and propagation of shear bands.
The foundation, modeled with linear elastic
beam elements, is positioned on top of the soil
model and connected to it through special
contact elements. The latter are rigid in
compression
but
tensionless,
allowing

199

L = 4H
free-field
fault outcrop

Foot wall

(a)

(b)

Hanging wall

Foundation

Figure 2:

Figure 2. Finite element discretisation and the two steps of the analysis: (a) fault rupture
propagation in the free-field, and (b) interplay between the outcropping fault rupture and the
structure (termed Fault RuptureSoilFoundationStructure Interaction, FR-SFSI).

substantial distress, while their rotation and


settlement could perhaps be acceptable.

and kinematic () characteristics of the soil


along the depth.

The main factors influencing FR-SFSI are:

The style of faulting (normal, thrust, strikeslip), the angle of dip and the
offset (dislocation) at the basement rock.
The total thickness (H) of the overlying soil
deposit, and the stiffness (G), strength (, c)

200

The type of the foundation system (for


example, isolated footings, mat foundation,
box-type foundation, piles, caissons).

The flexural and axial rigidity of the


foundation system (thickness of mat
foundation cross-section and length of tie
beams, etc.)

The load of the superstructure and the


foundation.

The stiffness of the superstructure (cross


section of structural members, spacing of
columns, presence or not of shear walls).

he location s from the foundation corner to


the free-field outcrop.

dislocation h/H = 5% would cause only a minor


diversion of the rupture path, easily noticeable
only in the loose soil (about 2 m towards the
hanging wall, to the left of the foundation). The
differential settlement is higher on loose sand.
The main difference between the two soils is in
the uplifting of the foundation. In dense sand
the building loses contact at both sides, uL 5
m, uR 3 m ; with only its central part
maintaining contact over a width bC 12 m. On
the other hand, in loose sand the building uplifts
only at the left side, uL 3 m.
The foundation distress is about 50% higher
in the dense sand, as a result of the creation of
a wider cantilever, whereas in the loose sand
the greater compression of the scarp is
beneficial. However, the differential settlement
and (rigid-body) rotation of the foundation is
three-time higher on the loose sand.

However, a detailed investigation of the role


of all the above parameters is beyond the
scope of this chapter. Reference is made to
Anastasopoulos (2005) and Anastasopoulos &
Gazetas (2007, b) for such a parameter study.
Here we only outline a few characteristic results
pertaining to a 20 m wide rigid mat foundation,
transmitting 30 kPa uniform pressure. The soil
layer is either loose (Dr 45%) or dense (Dr
80%) sand of total thickness H = 40 m. Three
locations of the foundation with respect to the
free-field outcrop are considered: s = 4 m,
10 m, and 16 m, i.e. near the left edge, in the
middle, and near the right edge of the
foundation, respectively.
Figs 35 portray the response of the
soilfoundation system for each location S and
each of the two soil densities, for a
parametrically variable ratio of base dislocation
over layer thickness:

.
Shown in each figure are the deformed
mesh, the distribution of plastic strains, the
diversion of the rupture D, the vertical
displacement profile y, the distortion angle ,
and the contact pressures p along the soilfoundation interface. In all cases the results are
compared with the corresponding free-field
results, to visualize the effects of FR-SFSI. The
contact stresses are compared to their initial
distribution (i.e., for h/H = 0, before the bedrock
displacement is applied) to reveal which parts of
the structure are losing contact with the bearing
soil, and hence foundation uplifting takes place.
The left part of the building that uplifts will be
denoted as uL, the right uR, and uC if the uplifting
takes place around the centre. In similar
fashion, the part of the foundation that
maintains contact will be denoted as bL, bR and
bC, if it is located at the left side, the right side,
or the middle, respectively.
The following trends are worthy of note:

(2) For the fault emerging (without SFSI) in


the middle of the foundation (s = 10 m, or s/B =
0.50) the rupture path is diverted, becomes very
diffuse, and bifurcates for h/H = 5%. The left
branch (which is a secondary one) diverts by
about 3 m to 4 m towards the hanging wall. In
both cases of loose and dense sand a fault
scarp develops beneath the building. The
foundation maintains always contact at its left
edge and at its middle part. In dense sand,
moving from left to right, there is first a small
part of the building bL 2 m that is in contact,
followed by an uplifted portion, uL 4 m, then
the middle part that remains in contact, bR 10
m, and finally the far most right part of the
foundation that uplifts uR 4 m. Although the
situation is qualitatively similar with loose sand,
uplifting is much less extensive (uR 1 m). The
effective width of the foundation, i.e. in contact
with the soil, is now: bL = bC 19 m. And the
differential settlement is about 2 m in both
cases.
(3) For the fault emerging (without SFSI)
near the right edge of the foundation (s = 16 m,
or s/B = 0.80), we see again diversion,
diffusion, and a minor bifurcation of the rupture
path for h/H = 5%. The right branch (which is
the most significant) diverts slightly towards the
footwall by about 24 m, while the (barely
noticeable) left branch is diverted towards the
hanging wall ( 3 m). But a clear significant
difference is noted between the loose and
dense sand cases:

(1) For the fault emerging (without SFSI)


near the left edge (s = 4m, or s/B = 0.20), this
lightly loaded foundation for a relative base

201

Dense Sand

Loose Sand
(i)

Dense
Sand
B = 20
m

(i)

Loose
Sand
B = 20
m

y = 21 c m

y = 61 c m

q = 30 kPa

q = 30 kPa

Free Field
Free Field

foundation
building

foundation
building

0
h/H = 1 %

h/H = 1 %

h/H = 3 %

-1

(ii)

-2
10

15

20

25

h/H = 5 %

(ii)
0

30

10

15

20

25

30

foundation
building

-50

-50

pv (kPa)

pv (kPa)

h/H = 4 %

-2

foundation
building

h/H = 0 %

-100

-150

h/H = 0 %

-100

h/H = 5 %

-150
h/H = 5 %

(iii)

(iii)

-200

-200
0

10

80

15

20

25

30

10

80

foundation
building

60

60

40

40

(%)

(%)

-1

-1.5

h/H = 5 %

h/H = 3 %

free field

h/H = 4 %

-1.5

h/H = 2 %

-0.5

h/H = 2 %

y (m)

y (m)

-0.5

20

15

20

25

30

foundation
building

free field

20
0

(iv)

-20
0

10

15

20

25

(iv)

-20
0

30

Di stance (m)

10

15

20

25

30

Distance (m)

Figure 3: FR-SFSI analysis of rigid B = 20 m foundation subjected to q = 30 kPa surcharge load. Fault
rupture in the free field emerging at s = 4 m : (i) Deformed mesh and plastic strain, (ii) Vertical
displacement at the surface, (iii) contact pressure p, and (iv) distortion angle . The results of the
FR-SFSI analysis (red lines) are compared with the Free-field results (blue lines) for h/H = 1
to 5%.

202

Dense Sand

Loose Sand
(i)

B = 20
m
Dense
Sand

(i)

Loose
B = Sand
20 m

y = 218 c m

q = 30 kPa

y = 191 c m

q = 30 kPa

Free Field
Free Field

foundation
building

foundation
building

h/H = 1 %

h/H = 1 %

-0.5

h/H = 2 %

y (m)

y (m)

-0.5

h/H = 3 %

-1

h/H = 4 %

-1.5

h/H = 2 %
h/H = 3 %

-1

h/H = 4 %

-1.5

h/H = 5 %

h/H = 5 %

free field

-2

-2

(ii)
0

10

15

20

25

30

building
foundation

10

15

20

25

30

foundation
building

-50

pv (kPa)

pv (kPa)

-50
h/H = 0 %

-100

-150

-100

h/H = 0 %

h/H = 5 %

-150
h/H = 5 %

(iii)

-200
0

10

15

20

25

(iii)

-200

30

80

10

80

foundation
building
60

15

20

25

30

foundation
building

60

40

40

free field

(%)

(%)

(ii)

-2.5

-2.5

20

20

-20

-20

(iv)

-40
0

10

15

20

25

(iv)

-40
0

30

Di stance (m)

10

15

20

25

30

Di stance (m)

Figure 4: FR-SFSI analysis of rigid B = 20 m foundation subjected to q = 30 kPa surcharge load. Fault
rupture in the free field emerging at s = 10 m : (i) Deformed mesh and plastic strain, (ii) Vertical
displacement at the surface, (iii) contact pressure p, and (iv) distortion angle . The results of the
FR-SFSI analysis (red lines) are compared with the Free-field results (blue lines) for h/H = 1
to 5%.

203

Dense Sand

Loose Sand

B = 20 m

(i)

(i)

B = 20 m

y = 150 cm

y = 82 cm

q = 30 kPa

q = 30 kPa

Free Field
Free Field

foundation
building

foundation
building

0
h/H = 1 %

h/H = 1 %

-0.5

h/H = 2 %

-1

y (m)

y (m)

-0.5

h/H = 3 %
h/H = 4 %

-1.5

h/H = 4 %

-2

(ii)

-2.5

(ii)
-2.5

10

15

20

25

30

foundation
building

10

15

20

25

30

foundation
building

h/H = 0 %

h/H = 0 %

-100

pv (kPa)

-100

-200

h/H = 5 %

-200

-300

-300
h/H = 5 %

(iii)

-400
0

10

15

20

25

(iii)

-400
0

30

10

15

20

25

30

120

120

foundation
building

foundation
building
80

(%)

80

(%)

h/H = 3 %

h/H = 5 %

-2

pv (kPa)

-1

-1.5

free field

h/H = 5 %

h/H = 2 %

free field

40

40

(iv)

-40
0

10

15

20

25

(iv)

-40
0

30

Di stance (m)

10

15

20

25

30

Di stance (m)

Figure 5: FR-SFSI analysis of rigid B = 20 m foundation subjected to q = 30 kPa surcharge load. Fault
rupture in the free field emerging at s = 16 m : (i) Deformed mesh and plastic strain, (ii) Vertical
displacement at the surface, (iii) contact pressure p, and (iv) distortion angle . The results of the
FR-SFSI analysis (red lines) are compared with the Free-field results (blue lines) for h/H = 1
to 5%.

204

The fault scarp that is formed near the


right edge of the foundation is
conspicuous only with loose sand.
On dense sand, the middle part of the
foundation loses contact with the bearing
soil, uC 11 m, while the left and right part
of it remain in contact, bL 2 m and bR
7 m.
On loose sand, the response is quite
favourable: not only is the dislocation
diverted by more than 4 m and outcrops
beyond the right edge of the structure, but
full contact is maintained over the whole
length of the soilfoundation interface.
The distress of the foundation is thus
significantly less with loose than with
dense sand. Also smaller on loose sand
is the (rigid-body) rotation of the
foundation.
Such a good response of a building on loose
soil on the hanging wall is reminiscent of
several success stories from the Kocaeli 1999
earthquake, especially of the building in
Denizerler across the entrance from the Ford
factory, near Glck (see Anastasopoulos et
Gazetas 2007).
(4) Although not shown here, the effect of an
increase in the transmitted load from 30 kPa to
60 kPa is quite beneficial on loose sand, but
almost negligible on dense sand. The most
significant benefits are the decrease of
foundation rotation (and of structure tilting) and
the elimination of a large part of uplifting. As a
consequence, the survival of a heavy
structure on top of a major fault rupture in loose
soil seems quite possible, in qualitative accord
with numerous such success stories in several
earthquakes.

tied to the different blocks of the fault may


indeed be vulnerable. An interesting analogy
has been brought to our attention by Professor
J. Bray (2005): deep-rooted trees being torn
apart by a fault rupturing directly underneath,
apparently as a result of their roots being pulled
in opposite directions.
Two typical foundation systems are examined
here in order to highlight the interaction
between a deep foundation and an emerging
fault rupture:
a 3x3 capped pile group
a square rigid embedded foundation
(caisson)
A 3-D finite element model was developed for
each case, using eight-noded elements, and
employing the same soil and interface
constitutive models as described in the
preceding section. Fig 6 presents a plane
section of the complete model. In both cases
the soil deposit consisted of dense sand, of
total thickness H = 20 m. Needless to say, the
choise of this limited depth was motivated solely
by the desire for the smallest possible size of
this 3D model.

Numerical Analysis
Foundations

and

Results:

Piles
The piles are of length Lp = 15 m, diameter
dp = 1 m and are spaced 4 m apart (from axis
to axis). Their cap is 10 m x 10 m in plan and
2.5 m thick, and carries a structural vertical load
of 10 MN. A rigid connection is assumed
between cap and piles (fixed-head piles). Only
ideally elastic pile behaviour is considered at
the present time, although the necessity for
accounting for pile inelasticity will become
apparent (if a realistic assessment of the
response of the system to large fault offsets is
needed).
Aiming at giving a first picture of the
possible straining to be experienced by the
piles, Fig. 7 portrays the deformed finiteelement mesh with the distribution of plastic
shear strains. Four positions of the pile group
with respect to the outcropping fault in the
freefield are examined: S = 1, 5, 9, and 13
meters, where S is measured from the edge of
the pile cap (which lies 1 m to the left of the
nearby pile axis). Then Fig. 8 presents detailed
results (deformations and internal forces) for the
case of S = 5 m, only. Several trends are worth
noting in these figures:

Deep

Whereas piles are used for protecting structures


by helping to keep total and differential
settlements small, their role in supporting
structures straddling seismic faults is far from
clear. Scant (perhaps only circumstantial)
evidence from recent earthquakes has
implicated the piles in some structural damage
see for example the analysis of the damage
of the pile-supported Attaturk Stadium in
Denizerler during the Kocaeli Earthquake
(Anastasopoulos & Gazetas, 2007). Systems

205

a
Axis of symmetry

Free field rupture path

h
b

a
Axis of symmetry

Free field rupture path

Figure 6: Pile group and caisson foundations in the path of a rupturing fault. Cross section aa of the 3D finite
discretisation

larger than the imposed base offset, .


The pile group, however, remains almost
intact: there is no displacement or rotation
of the pile and only the piles in the front
row experience some (rather minor)
distress (bending moments of the order of
300 kNm).
(2) For s = 5 m, the fault would have emerged
at the center of the foundation in the

(1) For s = 1 m, when the fault emerges near


the left edge of the pile group (and the
group is therefore almost all in the
footwall), a slight diversion of the rupture
path to the left takes place. A very distinct
scarp is formed immediately next to the
piles. The scarp forms a slope of about the
same angle, a, as that of the triggering
basement rupture, and is appreciably

206

(3) For s = 9 m, the fault would have emerged


near the right edge of the group in the free
field. The piles with their presence and
transmitted loads diffuse the rupture,
thereby suffering unequal settlements and
non-uniform large displacements. As a
result,
the
rotation
and
lateral
displacement of the cap and the bending
moments in the piles attain very large
(unacceptable) values.

freefield. The presence of the axially


loaded piles makes the rupture path: (i) to
partly divert to the left and emerge just at
the edge of the front of piles, and (ii) to
become diffuse in the region between the
piles. Substantial rotation and horizontal
displacement of the pile cap take place.
The front row of piles is being pulled
outward and downward by the dropping
hanging wall of the fault; as a result very
large bending moments would develop at
the pilehead, in excess of 12 MNm for a
dislocation of 2 m. The middle row of piles
would experience much less distress, but
the last row and especially the corner piles
would
develop
substantial
bending
moments (almost 6 MNm for a dislocation
of 2 m), as a result of being pushed near
their middle. Notice the completely
different pattern of bending moments with
depth between front-row and back-row
piles in Fig. 7 : whereas for pile 1 (front
row) the maximum is at the top, for pile 6
(corner pile) the maximum appears at 10
m depth. Evidently, such large bending
moments, especially in the front row,
exceed the maximum conceivable capacity
of a well-reinforced 1 m diameter pile,
implying structural failure at least of
conventional-type piles.

(4) Finally, for s = 13 m (freefield fault outcrop


3 m beyond the pile-cap), there is a slight
diversion of the rupture to the right with a
simultaneous slight diffusion of the plastic
shear strains. Only the last row of piles is
stressed significantly (max M 4MNm for
h = 2 m). There is apparently no rotation of
the pile cap, but a downward and outward
displacement are unavoidable; such
displacements might have a detrimental
effect on a framed structure, one column of
which is supported on the studied piledfoundation.
In conclusion, it appears that the response of
piled foundations may be less favourable than
that of rigid mat foundations. However, two
significant limitations of the performed analyses
on which these conclusions are partly based
must be noted here:

s=1 m

s=5 m

s=9 m

s = 13 m

Figure 7: Deformed mesh of the soilpilecap system with the concentration of plastic octahedral strains, for
different positions (s = 1 13 m) of the emerging fault rupture.

207

vertical component of dislocation h = 2 m.


By contrast, recall that the corresponding
piled foundation had developed a rotation
and horizontal displacement, while its front
row of piles had been substantially
distressed.
Notice also that a secondary rupture has
begun to form, propagating at an angle of
about 30o to the left of main rupture. It is
about to reach the ground surface for h = 2
m, and the associated graben between the
two normal ruptures is (barely) visible in the
scale of the figure.

perfect (bonded) contact was assumed


between piles and soil
the piles were modelled as a perfectly
elastic material.
As a result of the first assumption, the forces
upon the piles by the outward and downward
moving hanging wall are exaggerated. Soil
sliding around the piles would reduce the
magnitude of such drag forces, thereby
leading to smaller pile distress and smaller cap
rotation / displacement. Regarding the second
assumption, note that the large bending
moments in the piles would not of course
materialize in reality, since their ultimate
structural capacity can not be exceeded.
Prediction of the consequences of the
unavoidable redistribution of loads among the
piles, and among piles and raft, can not be
made reliably with the results presented above
for purely elastic piles.

(3) For s = 9 m, the rupture is diffused the


caisson rotates substantially to the left (8o
for h = 2 m), and an active state of stress
develops on the back side of the caisson.
Clearly this behaviour is not so favourable;
for instance it would cause distress in a
framed structure one column of which is
supported on such a caisson. But by
contrast to the piled foundation, the
capability of the caisson to transmit the
vertical load would be hardly affected.

Rigid Caisson
The caisson is 10 m x 10 m in plan and also 15
m in depth. It carries 10 MN vertical load. Only
fully bonded contact between the caisson and
the soil is considered --- an idealization that is
likely to lead to a conservative assessment of
the caisson displacement / rotation.
Dominant role in the response of a given
caisson to fault rupturing underneath plays its
position with respect to the freefield rupture
outcropping. Again four such positions are
considered: 1, 5, 9, and 13 meters. Figs 8-9
portray the deformed mesh with the distribution
of plastic octahedral shear strains for each
value of s. Fig. 8 gives the plane section
(along the axis) while Fig. 9 depicts a 3-D view
(of half the model). The following conclusions
are drawn:

(4) For s = 13 m, the rupture path hits the


base corner of the caisson and defracts to
the right, emerging at the ground surface at
a distance of S = 18 m, i.e., 5 m to the right
of the freefield outcrop. The caisson
essentially follows the movement of the
hanging wall, thereby experiencing an
appreciable rotation of about 3o for h = 2 m.
In conclusion, it appears that the response of
deep embedded foundations (caissons) would
in most cases be quite satisfactory, especially if
structural provisions are taken to accommodate
their unavoidable rotation at large fault offsets.
Once again, one of the limitations in our
modelling, namely the assumption of a
perfectly-bonded
interface,
may
have
exaggerated the lateral displacement/rotation of
the caisson.

(1) For s = 1 m, the fault emerges to the left of


the caisson, diverted slightly, and forms a
distinct scarp similar to that in the case of
the piled foundation. The caisson does not
experience any measurable rotation or
displacement.

ACKNOWLEDGEMENTS
This work formed part of the EU research
project QUAKER, funded through the EU Fifth
Framework Programme: Environment, Energy,
and Sustainable Development, Research and
Technological Development Activity of Generic
Nature: the Fight against Natural and
Technological Hazards (contract number:

(2) More significant is the diversion of the


rupture path in case of s = 5 m; the fault
now emerges vertically along the side of the
caisson. The latter hardly feels the rupture,
experiencing a rotation of merely 1o for a

208

aa

bb

pile cap

pile cap

0
h=0.2 m

h=0.2 m
h=0.4 m
h=0.8 m

-1
h=1.2 m

-1.5

h=0.8 m

-1
h=1.2 m

-1.5

h=1.6 m

-2

h=0.4 m

-0.5

y (m)

y (m)

-0.5

h=1.6 m

-2

h=2 m

x (m)

-2.5
-10

10

h=2 m

x (m)

-2.5
-10

20

10

20

16

16

Pile 2

Pile 1

12

Pile 5

M (MNm)

M (MNm)

12

-5

0.5

1.5

h (m)

M (MNm)

M (MNm)

10

15

-5
0

static

y (m)

y (m)

0.5

h (m)

1.5

10

15

Pile 6
static
h=0.1

h=0.1

12

h=0.8

16

Pile 4

0
0

Pile 1

12

Pile 6

Pile 3

h=2.0

16

h=0.8
h=2.0

Figure 8: Detailed results (vertical displacement y of the ground surface, largest bending moment in the
piles, and distribution of bending moments in piles 1 and 6 for a 3 x 3 capped pile group, for a fault
rupturing position s = 5 m.

209

s=1 m

s=5 m

s=9 m

s = 13 m

s=1 m

s=5 m

s=9 m

s = 13 m

(a)

(b)

Figure 9: Deformed mesh of the caissonsoil system with the concentration of plastic octahedral strains, for
different positions (s = 1 13 m) of the emerging fault rupture: (a) Section aa ; (b) 3-D view.

REFERENCES AND BIBLIOGRAPHY

EVG1-CT-2002-00064). The research on pile


and caisson foundations built on top of a
rupturing normal fault was part of a project
funded by the Greek Railway Organization
OSE.

Anastasopoulos, . (2005), Fault RuptureSoil


FoundationStructure
Interaction,
Ph.D.
Dissertation, School of Civil Engineering,
National Technical University, Athens, pp.570.

210

Anastasopoulos I., Gazetas G., Bransby M.F.,


Davies M.C.R., and El Nahas A.(2007), "Normal
Fault Rupture Interaction with Strip Foundations",
Journal of Geotechnical and Geoenvironmental
Engineering, Vol. 133

Journal of the Soil Mechanics and Foundation


Division, ASCE, Vol. 99, pp. 1153-1163.
Lade, P.V., Cole, D.A., Jr., and Cummings, D.
(1984), Multiple Failure Surfaces Over Dip-Slip
Faults, Journal of Geotechnical Engineering,
ASCE, Vol. 110, No. 5, pp. 616-627.

Anastasopoulos I., & Gazetas G. (2007),


"Foundation-Structure Systems over a Rupturing
Normal Fault : Part I. Observations after the
Kocaeli
1999
Earthquake",
Bulletin
of
Earthquake Engineering, Vol. 5, No. 3

Lazarte, C.A., and Bray, J.D. (1995), Observed


Surface Breakage due to Strike-Slip Faulting,
Third International Conference on Recent
Advances in Geotechnical Engineering and Soil
Dynamics, Vol. 2, pp. 635640.

Anastasopoulos I., & Gazetas G. (2007), "Behaviour


of Structure-Foundation Systems over a
Rupturing Normal Fault : Part II. Analysis of the
Kocaeli Case Histories", Bulletin of Earthquake
Engineering, Vol. 5, No. 3

Potts, D. M., Dounias, G. T. & Vaughan, P. R. V.


(1987), Finite element analysis of the direct
shear box test, Gotechnique, Vol. 37, No. 1, pp.
1123.

Anastasopoulos I., Gazetas G., Bransby M.F.,


Davies M.C.R., and El Nahas A.(2007), "Fault
Rupture Propagation through Sand : Finite
Element Analysis and Validation through
Centrifuge
Experiments",
Journal
of
Geotechnical
and
Geoenvironmental
Engineering , Vol. 133

Potts, D. M., Dounias, G. T. & Vaughan, P. R. (1990),


Finite element analysis of progressive failure of
Carsington Embankment, Gotechnique, Vol. 40,
No. 1, pp. 79101.
Potts, D. M., Kovacevic, N. & Vaughan, P. R. (1997),
Delayed collapse of cut slopes in stiff clay,
Gotechnique, Vol. 47, No. 5, pp. 953982.

Ambraseys, N. & Jackson, J., (1984), Seismic


Movements, Ground Movements and their
Effects on Structures, P.B. Attewell and R.K.
Taylor Editors, Surrey University Press, pp. 353380.
Bray, J.D. (1990), The effects of tectonic movements
on stresses and deformations in earth
embankments, Ph.D. Dissertation, University of
California, Berkeley.
Bray, J.D. (2001), Developing mitigation measures
for the hazards associated with earthquake
surface fault rupture, Workshop on Seismic
Fault-Induced Failures: Possible Remedies for
Damage to Urban Facilities, Japan Society for
the Promotion of Science, University of Tokyo,
pp. 5579.
Bray, J.D., Seed, R.B., Cluff, L.S., and Seed, H.B.
(1994), Earthquake Fault Rupture Propagation
through
Soil,
Journal
of
Geotechnical
Engineering, ASCE, Vol. 120, No.3, March, pp.
543-561.
Bray, J.D., Seed, R.B., and Seed, H.B. (1994b),
Analysis
of
Earthquake
Fault
Rupture
Propagation through Cohesive Soil, Journal of
Geotechnical Engineering, ASCE, Vol. 120, No.3,
March, pp. 562-580.
Cole, D.A. Jr., and Lade, P.V. (1984), Influence
Zones in Alluvium Over Dip-Slip Faults, Journal
of Geotechnical Engineering, ASCE, Vol. 110,
No. 5, pp. 599-615.
Duncan, J.M., and Lefebvre, G. (1973), Earth
pressures on Structures Due to Fault Movement,

211

The Role of the Soil Stiffness on the Dynamic Impedance Functions


D. Pitilakis1, D. Clouteau2, A. Modaressi2
1

Ecole Centrale Paris, France (currently in Aristotle University of Thessaloniki, Greece)


2
Ecole Centrale Paris, France

Abstract
This paper provides an insight in the role of the soil stiffness on the estimation of the
dynamic impedance functions. The softening of the soil under strong ground shaking is not
taken into account in the dynamic impedance functions, under the assumption of linear
elastic soil behavior. Nevertheless, the resulting shear wave velocity reduction due to
nonlinear soil behavior may have important effects on the amplitude and shape of the
dynamic stiffness and radiation damping coefficients. A simple parametric analysis is
performed for a typical footing resting on a halfspace soil profile and subjected to a scaled
ground motion. The dynamic impedance coefficient are estimated with an equivalent linear
procedure and compared to the linear elastic case. The dynamic stiffness coefficient is
found to decrease in amplitude and become frequency dependent, depending on the initial
soil shear wave velocity. The radiation damping is found to be unaffected by the nonlinear
soil behavior.

INTRODUCTION

normalizes the excitation frequency with a


characteristic dimension B of the foundation
and the shear wave velocity Vs of the
supporting soil.
Geometrical and material linearity is implicit
in equation (1). The static stiffness of the
footing depends on the initial shear modulus G
of the supporting soil and the Poisson's ratio ,
while the dynamic stiffness and damping
coefficients depend on the excitation frequency.
The hysteretic material damping depends on
the initial linear material properties of the soil.
Besides,
the
dimensionless
frequency
parameter 0 depends on the shear wave
velocity of the linear soil. For convenience and
to avoid any misconceptions, the shear wave
velocity of the supporting soil will be assumed
herein as the shear wave velocity Vs,30 of the
upper 30m of soil, according to the Eurocode 8.
As opposed to the idealization of the linear
soil behavior, the dynamic material properties of
the soil are not constant when the soil behaves

The importance, as well as the general


notion of the foundation dynamic impedance
function, is well established in engineering
practice, during the past three decades
(Veletsos 1971, Veletsos 1973, Gazetas 1983,
Dobry1986a, Dobry19886b, Gazetas1991a,
Gazetas 1991b, Sieffert 1992, Pecker 1997,
Mylonakis 2006). The general equation,
extracted from Gazetas 1983,
S = K ( k(,) + 0 c(,) ) (1+2 )

(1)

provides the dynamic impedance function S in


terms of the static stiffness K, the dynamic
stiffness and damping coefficients k and c
respectively, and the hysteretic material
damping ratio of an equivalent SDOF
oscillator. The dimensionless parameter
0 = B/Vs

(2)

212

in a nonlinear way. The shear modulus G and,


consequently, the shear wave velocity Vs,30
change in time with varying earthquake
excitation amplitude and frequency content.
Thus, equation (1) cannot be directly
implemented in the nonlinear soil behavior
case, since the abovementioned parameters
vary. As a result, the foundation dynamic
impedance functions have to be calculated in
terms of the modified soil-foundation system
properties. For that reason, an equivalent linear
approach can be implemented in the calculation
of the foundation dynamic impedance functions,
using equivalent linear shear modulus and
damping characteristics.
An equivalent linear procedure is used
herein to calculate the dynamic impedance
functions of a foundation supported on a soil
profile having equivalent linear properties
(Pitilakis D., 2006). The soil-foundation system
is subjected to an earthquake ground motion
consisting of body waves, propagating only on
the vertical direction. The effect of the different
initial shear wave velocity Vs,30 on the dynamic
foundation impedance functions is clearly
demonstrated.
The computation of the dynamic foundation
response is performed assuming that there is
no structure founded on the footing. Moreover,
the foundation is assumed usually massless in
a typical calculation of the dynamic impedance
functions. Therefore, no secondary nonlinear
effects are expected to appear in the soil, as
there is no diffracted wave field, created by the
foundation-structure vibration and emanating
away from the foundation to infinity. There
would still be created a diffracted wave field due
to kinematic interaction but it vanishes in the
present case for vertical incidence and surface
foundations. Assuming that the secondary
nonlinearities induced in the soil are
approximately zero, the dynamic impedance
functions estimated by the equivalent linear
procedure can be used for the computation of
the SSI with an equivalent linear soil behavior.

SYSTEM IDENTIFICATION

In a linear soil-foundation system, factors


that influence the dynamic foundation
impedances are identified by several authors.
Among them are the foundation shape and
flexibility, the embedment ratio, the excitation
frequency, the type of soil profile along with the
depth and stiffness of the bedrock, the soil
properties (shear modulus, Poisson's ratio,
hysteretic damping ratio), the soil anisotropy,
inhomogeneity and nonlinearity.
In the nonlinear soil--foundation system,
however, parameters such as the intensity and
frequency content of the signal may modify
significantly the response from one case to
another, by exciting the nonlinear response of
the soil. In contrast with the linear soil case, the
nonlinear soil behavior tends to decrease the
stiffness of the system and to increase the
energy dissipation by viscous and hysteretic
mechanisms in the soil. Thus, additional
parameters enter in the estimation of the
foundation dynamic impedance functions.
In order to reveal the effects of the nonlinear
soil behavior, and more specifically of the
different initial soil shear wave velocity, on the
foundation dynamic impedance functions, a
series of parametric analyses is performed. A
rigid
massless
footing
is
placed
a
homogeneous halfspace soil profile and is
subjected to a harmonic motion. Up to this point
the procedure is similar to the largely exploited
procedure, assuming linear soil behavior,
presented in numerous studies. The nonlinear
soil behavior is approximated by shear modulus
reduction and damping curves for a typical
clayey soil materials, notably clay with IP0.
Furthermore, the soil--foundation system is
subjected to an earthquake record and the
response is obtained in the light of the
substructure technique with equivalent linear
soil properties.
Due to the inherent complexity of such an
approach to the equivalent linear dynamic
impedance functions, several assumptions
have to be made for sake of simplicity and

213

different sizes as well as different types of


elements. More specifically, configurations were
tested using 99 quadrilateral elements, 321
quadrilateral elements, 241 triangular elements,
and the chosen configuration of 334
quadrilateral and 4 triangular elements. The
real and imaginary parts of the estimated
dynamic impedances were compared against
the reference solution proposed by Sieffert
(1992). The presented configuration (334
quadrilateral and 4 triangular elements with an
average size of 0.5m) is chosen based on the
accuracy and the time efficiency of the
achieved solution.
For a rigid, circular, massless footing resting
on a perfectly elastic halfspace, the static
stiffness of the circular disk in the horizontal,
vertical, rocking and torsional modes are
calculated according to Veletsos (1973),

comprehension.
Accordingly,
the
soilfoundation system properties are chosen so as
to cover the wider possible range of cases with
the minimum number of interfering parameters.

Foundation Identification

The foundation consists of an infinitely rigid,


massless circular footing of diameter d=10m,
resting on the free soil surface. Any arbitrary
shaped footing can be approximated by an
equivalent circular one, by equating the contact
surfaces for the three translational degrees of
freedom and the area moments of inertia for the
three rotational components.
The radius of the circular foundation, r=5m,
was chosen sufficiently large in order to
account for the kinematic interaction, along with
the inertial interaction, for a vast range of shear
wave velocities and frequencies. The circular
footing is meshed in 334 quadrilateral and 4
triangular elements with an average size of
0.5m, as shown in Fig 1.

Kx = 8Gr / (2-)
Kz = 4Gr / (1-)
K = 8Gr3 /3(1-)
Kt = 16Gr3 / 3

(3)
(4)
(5)
(6)

where G and are the initial shear modulus and


Poisson's ratio respectively for the soil,
assuming linear conditions, and r is the radius
of the footing. The static stiffness of the disk
expresses the force which is necessary to
produce a unit displacement or rotation, for the
two translational and the rotational modes of
vibration respectively.
In the vibration of a footing under loading,
the amplitude of the dynamic response in each
mode depends on the zone of influence, which
is the depth to which the normal stresses
extend in the soil. Therefore, for a circular
footing resting on a homogeneous halfspace,
the vertical normal stresses induced along the
centerline of the foundation extend to a depth of
five radii. For a horizontally loaded footing, the
horizontal stresses induced in the soil vanish at
depth larger than two radii, while for moment
and torsional loading the stresses practically
exist down to a depth of 1.25 and 0.75 radii
respectively (Gazetas 1983).

Fig 1: Rigid massless circular footing used in the


parametric analyses

Different mesh configurations were tested,


using larger or smaller number of elements,

214

naturally, in the case of a nonlinear soil


behavior and thus, an initial hysteretic damping
ratio is set at 2% for linear soil conditions. The
Poisson's ratio is set at 1/3, a typical value for
unsaturated soil.

Soil Profile Identification

The footing is placed on the free surface of


a homogeneous soil profile, as shown in Fig 2.
For convenience, the thickness of the soil
stratum is chosen equal to H=30m. For the
homogeneous halfspace shown in Fig 2, five
different shear wave velocities Vs,30 are
assigned, notably 100m/s, 180m/s, 250m/s,
350m/s, 500m/s, classifying the soil profile to
category types C and B according to the
Eurocode 8.

Fig 3: Shear modulus reduction and damping curves


for the clay IP0 soil material
Fig 2: Homogeneous halfspace soil profile
supporting the circular footing
Ground Motion Identification

In order to approximate the nonlinear


behavior of the soil with an equivalent linear
approach, a clay with plasticity index IP0,
according to Vucetic (1991), is used to describe
the clayey soil material (Fig 3). This curve is
widely used in engineering practice and
describes a typical clayey soil material.
A discretization of the soil profile in soil
layers is performed, in order to calculate the
shear strain level at different depths and vary
the shear modulus and the damping
accordingly. Therefore, the soil profile of 30m is
divided into five layers, having thickness from
top to bottom of 1m, 2m, 4m, 8m and 15m. The
soil profile in all the analyses has a typical unit
weight of 2000kg/m3. Moreover, even though in
most of the available literature on the
foundation impedance functions zero material
damping is assumed in the soil, it is confirmed
that even at small strain levels the soil behaves
in a hysteretic way (Dobry 1986a, Dobry
1986b). The latter is more pronounced,

In order to promote the nonlinear soil


behavior, the soil--foundation system is
subjected to the strong earthquake record of
Aegion, 1995 Aegion, Greece, chosen
according
to
the
European
tectonic
environment, with varying amplitude and
frequency content. The earthquake record is
scaled to maximum acceleration amplitude of
0.01g, 0.10g, 0.20g, 0.30g and 0.50g, in order
to cover a wide range of ground motion
amplitudes. However, one must keep in mind
that scaling an accelerogram might affect and
indeed
modify
its
frequency
content.
Nonetheless, in the parametric analyses
conducted, the frequency content of the input
ground motion is not as important parameter for
the nonlinear soil behavior as is the excitation
amplitude. For acceleration amplitude 0.01g,
the soil is expected to behave in a linear way.

215

An attempt is made to estimate such


dimensionless graphs for the equivalent
dynamic stiffness and dashpot coefficients.
Then, knowing the dynamic impedance
coefficients of the footing, the response of the
superstructure can be calculated from standard
rigid body dynamics.
Four components of the circular footing
vibration are examined, notably the horizontal,
vertical, rocking and torsional. The real part of
the dynamic impedance in equation (1), namely
the dynamic stiffness coefficients are noted by
kx , kz , k and kt respectively, while the
imaginary part of equation (1) expresses the
radiation damping coefficients, which are
denoted by cx, cz, cr and ct, respectively. Both
the real and imaginary parts of the impedances
are always normalized by the linear static
stiffness for the corresponding mode of
vibration, presented in equations (3) to (6). For
the real part of the dynamic impedance, the
dynamic stiffness coefficient, the dimensionless
frequency parameter, notably 0 = B/Vs,30,LIN in
a linear analysis, is enriched by a correction
factor of Vs,30,LIN/ Vs,30,EQL in order to include the
effect of the soil softening due to nonlinear
behavior. The correction factor Vs,30,LIN/ Vs,30,EQL
is indeed the ratio of the initial, linear shear
wave velocity assigned to the soil profile over
the shear wave velocity calculated for the soil
profile having equivalent linear characteristics.
Attention should be made, as will be seen, in
reading the charts, as the equivalent linear
shear wave velocity differs in each soil case,
depending on the properties of the soil and the
dynamic characteristics of the input signal.
From a practical point of view, multiplying the
dimensionless 0 parameter by Vs,30,LIN/ Vs,30,EQL
moves the dynamic stiffness coefficient
impedance curve to the right, i.e. to the higher
frequency range. Nevertheless, this horizontal
translation to the higher frequency range is the
counterpart of the shifting of the dynamic
stiffness coefficient to the lower frequency
range, due to the soil softening. The result is
the canceling of the shifting to the lower and to
the higher frequency range, that is the peak
response will appear at the same position as it
appears in the linear case.

Fig 4: Aegion, 1995 Aegion, Greece earthquake


record (PGA=4.92m/s2)

Thus, using the earthquake record with


scaled amplitudes covers sufficiently a wide
range of earthquake scenarios with scattered
amplitudes.

PARAMETRIC ANALYSES

A parametric analysis is performed in order


to investigate the effects of the nonlinear
behavior of the soil on the dynamic foundation
impedance functions. The analyses of soil-foundation systems comprise a halfspace soil
profile of a typical clay, five initial shear wave
velocities for the profile (100m/s,180m/s,
250m/s, 350m/s, 500m/s), and an earthquake
record scaled to five different amplitudes.
In the simple linear case the foundation
dynamic impedance function are affected
primarily by the parameters mentioned in a
previous section. Thus, it is immediately
understood that the complexity of the problem
increases in the nonlinear soil case as the
number of the parameters that influence the soil
response increase.
As the typical procedure for calculating the
dynamic impedances of a footing suggests
(equation (1)), the static stiffness is calculated
from equations (3) to (6) and the dynamic
stiffness and damping coefficients are
estimated from normalized graphs and charts.

216

Contrary to the dynamic stiffness coefficient


k, the imaginary part of the dynamic
impedance, namely the radiation dashpot
coefficient c times the dimensionless parameter
0, 0 c, is plotted against 0 without any
normalization. The imaginary part is plotted in
its integrity (0 c), instead of c, in order to avoid
divergence at low frequencies. For sake of
simplicity, the imaginary part of the dynamic
impedance 0 c, will be referred to herein simply
as radiation dashpot coefficient.

frequency range the magnitude of the dynamic


impedance stiffness increases with increasing
frequency, attains a peak and then decreases
following the trend of the linear case, showing a
certain
frequency
dependence.
This
contradictory behavior in the low frequency
range (dominated by the response of the
deeper soil layers to low frequency, large wave
length, pulses) is caused by an inversion of the
shear wave velocity with depth. For an
earthquake amplitude of 0.30g, the shear wave
velocity decreases with increasing depth. At a
depth of 11m, the shear wave velocity is 77m/s,
while it is 171m/s at a depth of 0.5m under the
foundation. Similarly, the shear modulus
reduces down to 10% of the maximum linear
elastic value Gmax. Thus, the low frequency
response is dominated by pulses propagating at
significantly lower shear wave velocities than
the initial linear case, resulting in a stiffness
coefficient significantly decreased in amplitude
from the linear case. On the other hand, in the
high frequency range the response is
dominated by the surface waves which are
allowed to be created. These waves propagate
in a higher shear wave velocity, with a shorter
wavelength, than the low frequency deeper
body waves, leading to a increase in the
dynamic stiffness of the footing in the higher
frequency range. In conclusion, the increase of
the dynamic stiffness coefficient with increasing
frequency is due to the decrease of the shear
wave velocity with increasing depth, after the
equivalent linear analysis. In the case of Fig 5,
the decrease of the stiffness in the high
frequency range, after the peak, is caused by
an eventual increase of the shear wave
velocity, computed to be 83m/s at a depth of
22.5m (from 77m/s at 11m depth).

Effect of Different Soil Shear Wave Velocity

The
circular
footing
rests
on
a
homogeneous halfspace. Assuming that the soil
is a clay with plasticity index IP0 and initial
shear wave velocity Vs,30,LIN=180m/s, suggests
a soil type B according to the Eurocode 8.
When the soil--foundation system is subjected
to the Aegion earthquake record the horizontal,
vertical, rocking and torsional dynamic
impedances are shown in Fig 5 to Fig 8
respectively. The impedances are plotted for
the linear case and for five different earthquake
amplitudes, as shown in the legend of the
graphs.
The horizontal dynamic stiffness coefficient
decreases from the linear case in amplitude
with increasing level of excitation amplitude (Fig
5). The fluctuations in the stiffness coefficient
curves are the apparent result of resonances
that occur in the soil. The initially homogeneous
halfspace soil profile behaves in a nonlinear
way and interfaces are formed in the soil
between layers with different impedance ratios.
Consequently, waves emanating from the
vibrating foundation are reflected on those
interfaces and propagate back towards the
footing. The result of this propagation is the
increase of the foundation motion in some
frequencies, close to the resonance frequencies
of the newly formed inhomogeneous soil. The
short and flat peaks imply no significant
impedance ratio between the formed soil layers,
whereas they appear in the resonant
frequencies of the soil.
Contrary to the linear case, in the lower

217

Fig 5: Horizontal dynamic stiffness and radiation dashpot coefficients for a homogeneous soil profile, with
clay IP0 soil, initial shear wave velocity Vs,30,LIN=180m/s, subjected to the Aegion earthquake record

Fig 6: Vertical dynamic stiffness and radiation dashpot coefficients for a homogeneous soil profile, with clay
IP0 soil, initial shear wave velocity Vs,30,LIN=180m/s, subjected to the Aegion earthquake record

218

Fig 7: Rocking dynamic stiffness and radiation dashpot coefficients for a homogeneous soil profile, with clay
IP0 soil, initial shear wave velocity Vs,30,LIN=180m/s, subjected to the Aegion earthquake record

Fig 8: Torsional dynamic stiffness and radiation dashpot coefficients for a homogeneous soil profile, with clay
IP0 soil, initial shear wave velocity Vs,30,LIN=180m/s, subjected to the Aegion earthquake record

219

The imaginary part of the dynamic


impedance, plotted in Fig 5, is the combined
effect of the radiation viscous damping and the
hysteretic material damping. Assuming a purely
elastic soil, if there were no hysteretic damping
in the equivalent linear analyses, the values of
the imaginary part of the dynamic impedance
would be zero, implying that the radiation
dashpot would be zero. On that account, the
non zero value of the radiation dashpot for zero
frequency, even for the linear case, denotes the
presence of the hysteretic damping (equal to
2% in the linear analyses). In the linear case,
the imaginary part increases at a constant rate
with increasing frequency, implying that the
radiation damping is practically independent of
the frequency. In the equivalent linear case, the
magnitude of the imaginary part of the
impedance increases from the linear case with
increasing excitation amplitude. This increase is
attributed primarily to the hysteretic material
damping of the soil, which indeed increases in
the equivalent linear soil with increasing level of
excitation. For an earthquake amplitude of
0.30g, the hysteretic damping increases up to
17% in the deeper soil layers. The radiation
viscous damping in the horizontal mode
increases with increasing frequency at the
same rate as in the linear case. This explains
the fact that the radiation dashpot coefficients
are transposed in parallel from the linear case,
apparently unaffected by the soil nonlinearity.
For the nonlinear case, the larger the excitation
amplitude, the larger the parallel transpose of
the curve to higher values.
In the low frequency range, however, the
hysteretic damping in the equivalent linear
analyses does not increase from the linear
case, causing the radiation dashpot coefficient
curves for all excitation amplitudes to be similar
in amplitude with the linear case. This arises
from the fact that in the low frequency range no
surface waves are created. The response of the
soil is dominated by the wave fields having
longer wave lengths, which create resonance
phenomena at larger depths. As in the
equivalent linear analysis the halfspace soil
profile is discretized into soil layers with refined
thicknesses closer to the surface, the increase

of the hysteretic material damping is taken into


account mainly in the upper soil layers and not
in the deeper. Thereby, the predominant low
frequency waves in the deeper soil layers do
not influence the equivalent linear soil
response, resulting in a minor increase of the
hysteretic damping from the linear case.
The effects of the soil softening due to
nonlinear behavior are more pronounced in the
vertical stiffness coefficient in Fig 6. This is
expected because of the deeper zone of
influence of the vertical normal stresses in the
soil, caused by the vertical loading of the
footing. The magnitude of kz decreases with
increasing excitation amplitude, reaching a
reduction of 80% for the case of the excitation
amplitude 0.50g. Furthermore, the nonlinear
soil is not homogeneous any more, as seen
from the peaks and valleys for values of the
dimensionless frequency parameter 0 c
Vs,30,LIN/ Vs,30,EQL less than 1. The location of the
resonant valleys coincides with the resonant
frequencies of the soil.
Concerning the vertical radiation dashpot
coefficient, the increase in the hysteretic
damping with increasing excitation amplitude, in
the medium to higher frequency range, is
apparent in the vertical mode as well. The low
frequency range is dominated by the effects of
the deeper soil layers, in which the increase in
hysteretic damping is not accounted in the
equivalent linear analysis. The radiation
damping coefficient in the vertical mode is
slightly affected by the increase of amplitude,
increasing with frequency with a bit larger rate
than in the linear case.
Similar trends are observed in the rocking
(Fig 7) and torsional (Fig 8) vibration modes
equally. The stiffness decreases from the linear
case with increasing excitation amplitude, while
it tends to become frequency independent with
increasing level of excitation amplitude. The
hysteretic damping increases with increasing
excitation amplitude while the radiation
damping is unaffected by that increase. The
radiation damping for the nonlinear cases
increases with frequency at the same rate as in
the linear case. The small, compared to the
translational modes, radiation coefficient values

220

attested by researchers for the rocking and


torsional modes, are confirmed for the nonlinear
case as well. Yet, the radiation damping
coefficients are at least 50% lower than they
are for the translational components.
Assuming a higher initial shear wave
velocity of the soil Vs,30,LIN=350m/s, the dynamic
response of the footing resting on the stiffer soil
is shown in Fig 9 for the horizontal mode, in Fig
10 for the vertical mode, in Fig 11 for the
rocking mode and in Fig 12 for the torsional
mode.
The stiffer soil causes the dynamic stiffness
coefficient decrease less from the linear case
than for the softer soil case (Vs,30,LIN=180m/s).
Evidently, the softer soil exhibits larger
deformations and the nonlinearities are more
pronounced. For the soil with Vs,30,LIN=350m/s,
the dynamic stiffness coefficients increase
constantly with increasing frequency in the
horizontal mode, in contrast with the linear
case. This is due to the inversion of the shear
wave velocity, that is the decrease of the
velocity of the body waves with increasing
depth. As opposed to the previous soil case
with Vs,30,LIN=180m/s, the dynamic stiffness
coefficient increases with frequency in the
whole frequency range, due to the constant
decrease of the shear wave velocity with depth.
The shear wave velocity decreases to 178m/s
at a depth of 22.5m. Some flat undulations
appear due to the soil inhomogeneity caused by
the nonlinear behavior and the discretization of
the soil profile. In the vertical mode, the peaks
and valleys at a dimensionless frequency 0 c
Vs,30,LIN/ Vs,30,EQL are again more pronounced
than in the horizontal mode, suggesting
stronger resonance phenomena to occur in the
soil due to the deeper zone of vertical influence.
These undulations are quite flat due to the
existence of hysteretic damping in the soil.
The radiation damping coefficients for the
horizontal and vertical modes for the stiffer soil
resemble to the ones of the softer soil, in
magnitude and in increasing rate with
increasing frequency.

For the stiffer soil the rocking and torsional


dynamic stiffnesses decrease from the linear
case less than they decrease in the softer soil
case. Nevertheless, they tend also to become
independent of frequency in the whole
frequency range. On the other hand, in the
stiffer soil the radiation dashpot coefficients for
the rocking and torsional modes attain values
less than 50% of the ones attained for the softer
soil (Vs,30,LIN=180m/s). Especially for the
torsional mode, the dashpot coefficient is
practically independent of the exciting
frequency.

CONCLUSIONS

A parametric analysis was conducted to


demonstrate the effect of the soil shear wave
velocity on the dynamic impedance functions,
when nonlinear soil behavior is accounted. The
nonlinear soil behavior was approximated by an
equivalent linear procedure and the dynamic
impedance functions of a typical footing resting
on a typical soil profile were calculated and
compared with the linear elastic soil case. The
principal findings can be summarized as
follows:
z The dynamic response of the footing for
the nonlinear case depends on more
parameters than in the linear case. The
complexity of the linear problem is
augmented in the nonlinear case by the
influence of the shear wave velocity of
the profile (it decreases in the nonlinear
case), the soil material (characterized by
the shear modulus reduction and
damping curves) and the excitation
amplitude and frequency content.
z The
dynamic stiffness coefficient
decreases from the linear case with
increasing excitation amplitude and with
decreasing initial shear wave velocity

221

Fig 9: Horizontal dynamic stiffness and radiation dashpot coefficients for a homogeneous soil profile, with
clay IP0 soil, initial shear wave velocity Vs,30,LIN=350m/s, subjected to the Aegion earthquake record

Fig 10: Vertical dynamic stiffness and radiation dashpot coefficients for a homogeneous soil profile, with clay
IP0 soil, initial shear wave velocity Vs,30,LIN=350m/s, subjected to the Aegion earthquake record

222

Fig 11: Rocking dynamic stiffness and radiation dashpot coefficients for a homogeneous soil profile, with clay
IP0 soil, initial shear wave velocity Vs,30,LIN=350m/s, subjected to the Aegion earthquake record

Fig 12: Torsional dynamic stiffness and radiation dashpot coefficients for a homogeneous soil profile, with
clay IP0 soil, initial shear wave velocity Vs,30,LIN=350m/s, subjected to the Aegion earthquake record

223

Journal of Geotechnical Engineering Division


ASCE, Vol. 112, No. 2, pp. 109-135

The dynamic stiffness coefficient for the


initially homogeneous soil profile may
increase,
decrease,
or
become
practically constant with increasing
frequency. The response of the stiffness
coefficient depends on the modification
of the soil shear wave velocity due to
the equivalent linear soil behavior. The
shear wave velocity may indeed
decrease significantly with increasing
depth, contrary to the traditional
increase, or stiffening, with increasing
depth. Therefore, an increase of the
stiffness coefficient with increasing
frequency may be as well expected,
since this inversion of the shear wave
velocity
may
be
encountered.
Nevertheless, this phenomenon is
produced only when there is a
significant decrease of the shear wave
velocity in the deeper soil layers, due to
the equivalent linear soil behavior and
its inherent deficiencies.
The dynamic stiffness coefficient
decreases with decreasing shear
modulus of the soil for a fixed level of
shearing in the soil.
In
the
nonlinear
soil
profile
inhomogeneous soil layers are formed
and,
consequently,
resonant
frequencies of the soil appear as
fluctuations of the dynamic stiffness
coefficient at those frequencies. These
undulations are of larger amplitude for
the vertical and horizontal vibration
modes, while they do not appear in the
rocking and torsional modes. The larger
fluctuations for the vertical mode are
attributed to the deeper zone of
influence of the normal stresses induced
by the response of the footing to vertical
loading.

Dobry R., Gazetas G., Stokoe II K.H. (1986)


Dynamic response of arbitrarily shaped
foundations,
Journal
of
Geotechnical
Engineering Division ASCE, Vol. 112, No. 2,
pp. 136-154
Gazetas G. (1983) Analysis of machine foundation
vibrations: state of the art. Soil Dynamics and
Earthquake Engineering, Vol. 2, No. 1, pp. 2-42
Gazetas G. (1991) Formulas and charts for
impedances of surface and embedded
foundations,
Journal
of
Geotechnical
Engineering Division -- ASCE, Vol. 117, No. 9,
pp. 1363-1381
Gazetas G. and Stokoe K.H. (1991) Free vibration
of embedded foundations: Theory versus
experiment,
Journal
of
Geotechnical
Engineering Division -- ASCE, Vol. 117, No. 9,
pp. 1382-1401
Mylonakis G., Nikolaou S., Gazetas G. (2006)
Footings under seismic loading: Analysis and
design issues with emphasis on bridge
foundations, Soil Dynamics and Earthquake
Engineering, Vol. 26, No. 9, pp. 824-853
Pecker A. (1997) Analytical formulae for the seismic
bearing capacity of shallow strip foundations in
Seismic Behavior of Ground and Geotechnical
Structures, Balkema, Rotterdam, pp. 261-268
Pitilakis D. (2006) Soil-structure interaction modeling
using equivalent linear soil behavior in the
substructure method, Ph.D. Thesis presented in
Ecole Centrale Paris, France
Sieffert J.-G., and Cevaer F. (1992) Manuel de
fonction d'impedance, Ouest Editions/AFPS,
Nantes, France
Veletsos A.S., Wei Y.T. (1971) Lateral and rocking
vibration of footings, Journal of Soils Mechanics
and Foundation Division ASCE, Vol. 97, No.
SM9, pp. 1127-1248
Veletsos A.S., Verbic B. (1973) Vibration of
viscoelastic
foundations,
Earthquake
Engineering and Structural Dynamics ASCE,
Vol. 2, pp. 87-102

REFERENCES

Vucetic M. (1994) Cyclic threshold shear strains in


soils, Journal of Geotechnical Engineering
Division ASCE, Vol. 120, No. 12, pp. 22082228

CEN (2002) Eurocode 8, Comite Europeen de


Normalisation, Prenorme ENV 1997-1
Dobry R., and Gazetas G. (1986) Dynamic
response of arbitrarily shaped foundations,

224

Fundamental Period of Sdof Systems Including Soil-Structure Interaction


and Soil Improvement
E. Kirtas1, K. Trevlopoulos1 , E. Rovithis1, K. Pitilakis1
1

Department of Civil Engineering, Aristotle University of Thessaloniki, Greece

Abstract
The effect of soil-structure interaction on the fundamental period of a structural system is
recognized by several modern seismic codes, whereas simple straightforward relationships
are provided to estimate the modified period of the building. In the present paper an effort
takes place to estimate the fundamental period of structures with surface foundation
including soil-structure interaction phenomena, utilizing 2D plane-strain numerical
simulations with FE codes. Three characteristic fixed base periods of single degree of
freedom systems are considered during the numerical investigation, while the influence of
parameters such as the soil category, superstructure mass and height values is highlighted.
Comparison with theoretical relationships and seismic code recommendations reveal the
accuracy of the employed numerical calculations, whereas a discussion takes place
regarding the observed variation of the calculated effective period values. Further
investigation regarding the effective period modification due to subsoil stiffening
interventions, highlights the necessity to consider the enhanced soil properties during the
study of the systems dynamic response.

INTRODUCTION

provide a comprehensive procedure to


incorporate interaction phenomena into seismic
design of buildings. The employed approach
involves modification of the dynamic properties
of the structure and evaluation of the response
to the prescribed free-field motion (Jennings
and Bielak, 1973). The
expected
consequences of SSI would include an increase
in the fundamental natural period of the
structure and a change (usually increase) of the
effective damping. The effective structural
period (including interaction) is calculated using
the relationship first proposed by Veletsos and
Meek (1974):

The
mechanisms
that
soil-structure
interaction (SSI) affects the dynamic response
of structures during an earthquake event have
been highlighted by several studies over the
last few decades. Soil deformation under
seismic motion is modified by the foundation
stiffness at the first stage of interaction
(kinematic part), whereas structural oscillation
imposes additional horizontal and rotational
deformations on the foundation creating
outgoing waves (inertial part), constituting
together a rather complicated phenomenon.
Depending on the foundation shape and
formation, the soil stiffness as well as the
structural dynamic characteristics, soil-structure
interaction may possess a paramount role in
the systems seismic performance. Both
induced seismic motion and structural dynamic
response may be altered during the seismic
event, changing dramatically the behaviour
especially in the case of structures with stiff
foundations on soft soil formations.
FEMA 450 regulations (BSSC, 2003)

TSSI = T 1 + k

1 + k h2
K y
K r

(1)

where T is the fixed-base period, Ky and K the


spring stiffness values to lateral and rocking
motions calculated, k and h the stiffness and
effective height of the fixed-base structure.
Seismic base shear reduction compared to the
fixed-base structure is then calculated, using an

225

appropriate expression that considers also a


seismic response coefficient based on the
effective period determined above.
Eurocode 8, Part 5 (CEN 2002) instructions
refer to the consideration of soil-structure
interaction phenomena in the cases that SSI
effects could be detrimental, such as:
a) structures where P- (2nd order) effects
play a significant role
b) structures with massive or deep-seated
foundations, such as large bridge piers,
offshore caissons, and silos
c) slender tall structures, such as towers
and chimneys
d) structures supported on very soft soils,
with average shear wave velocity Vs,max less
than 100m/s
According to EC8, dynamic soil-structure
interaction should also take into account the
non-linear soil behavior and the radiation
damping, resulting in the majority of usual
building structures to beneficial seismic
response and reduced structural bending
moments and shear forces. Nevertheless,
specific instructions on the quantitative
consideration of SSI effects are not provided in
the context of EC8.
The Greek Seismic Code (EAK 2000) on the
other hand does not mention directly soilstructure interaction issues, apart from a
requirement to take properly into account soil
compliancy at the foundation level that may
alter the dynamic response of the structure.
Yet, none of the above seismic codes
mentions the effects of soil properties
enhancement on the SSI mechanisms during
the dynamic response of the modified system.
Indeed, the fundamental period of structures,
bridge piers or other surface constructions is
expected to change in the presence of typical
soil interventions such as soil replacement and
compaction, stone columns, jet grouting,
compaction piles etc (Pitilakis et 2005, Kirtas et
al 2007).
In the first part of the present paper 2D plain
strain calculations of the effective (SSI)
fundamental period of sdof structures are
validated by comparing with corresponding
calculations determined from seismic code
recommendations or other explicit expressions
and theoretical solutions proposed in the
bibliography. In a second part the influence of
soil improvement in the modification of structure

fundamental frequency is revealed from


validated numerical simulations, whereas the
modification of the dynamic characteristics as
well as input motion alteration due to subsoil
stiffening (Fig 1) are calculated and discussed
accordingly.

Reference
System

Modified
area
G=10G

Fig 1: Application of soil stiffness enhancement


below foundation.
CALCULATION OF EFFECTIVE PERIOD VALUES

The soil-structure system employed during


the validation procedure consists of a sdof
structure with surface strip foundation (Fig 2).
The rigid foundation is considered bonded with
the soil: sliding and foundation uplift have not
been taken into consideration during the
theoretical and numerical calculations of the
SSI period. The crucial parameters that affect
significantly the intensity of the soil-structure
interaction phenomena, and subsequently the
modification of the structural dynamic
characteristics compared to the fixed-base
case, are properly accounted for during the
investigation as presented in the following
paragraphs.
Structural mass

The superstructure mass is directly related


to the development of soil-structure interaction
phenomena. Increased values of structural
mass result in enhancement of the foundation
rocking motion, due to structural oscillations
during the dynamic response of the system.
is
Normalized
structural
mass
mnorm
implemented in order to quantify the influence
of the superstructure mass to the obtained
effective structural period TSSI, according to the
expression (Wolf, 1985):
m
(2)
mnorm = str 3

226

where mstr is the superstructure mass, is the


soil density and the characteristic dimension
of the foundation, equal to half-width B of the
strip foundation employed in this paper.

for several combinations of footing geometries


and soil profiles have been recently discussed
in Mylonakis et al. (2006), focusing primarily to
bridge
foundation
cases,
whereas
dimensionless charts that provide effective
periods and damping values of soil-structure
systems have been proposed from Aviles and
Suarez (2002).
Despite the fact that numerical calculations
concern only one specific case of foundation
geometry, it is quite interesting to review the
variation
between
different
theoretical
expressions for foundation dynamic stiffness
values. Therefore, during the theoretical
calculation of stiffness parameters in the
present
study,
several
foundation-soil
combinations are examined, as presented in
Table 1 and Fig 3. The provided expressions
are adequate to estimate the effective structural
period based on Equation 1 that concern SSI
behaviour in 2 dimensions considering
horizontal and rocking motion. Details of the
approach adopted in each specific case are
summarized in the following paragraphs.

Structural height

Structural height is also strongly related to


the rocking motion of the foundation due to soilstructure interaction, increasing the oscillation
of the superstructure and enhancing effective
period modification. Normalized structural
height is considered herein to study structures
of different height, according to the expression
(Wolf, 1985):
hnorm =

hstr
B

(3)

Foundation soil compliancy

Depending on the ground type, foundation


soil compliancy may differ between different soil
categories, affecting significantly the dynamic
properties of the SSI system. In this paper three
soil types with different properties are
examined, each indicative of the soil categories
B, C and D as defined by EC8. The considered
shear wave velocities Vs are equal to 400m/s,
200m/s and 100m/s for soil types B, C and D
respectively. This correlation is useful from the
engineering point of view, during the
comparative evaluation of the obtained TSSI
values with respect to the examined soil
categories.

(a) Square foundation over uniform half-space


The expression used for the dynamic
stiffness calculation of square foundation
geometry over uniform half-space can be found
in Gazetas (1991) and Gazetas (1997). The
dynamic stiffness coefficient of the horizontal
response mode is usually considered equal to 1
for square footings, whereas the coefficient of
the rocking response mode is frequency
dependent. The dimensions of the square
footing are considered equal to the width of the
strip foundation of Fig 2.

Theoretical calculation of effective period values

Several theoretical relationships have been


proposed to estimate the soil stiffness
parameters of Equation 1. Static stiffness
values are usually first determined, and the
dynamic stiffness is then obtained by utilizing a
dynamic coefficient that is proposed for each
degree of freedom at the soil-foundation
interface. This dynamic coefficient depends on
the geometry, stiffness and embedment of the
foundation body and is usually frequency
dependent. Numerous studies in the last few
decades have investigated several cases of
foundation geometries including strip, circular,
square, rectangular or in general arbitrary
foundation shapes. Most of the studies concern
homogeneous half-space consideration of the
foundation subsoil conditions, yet layered soil
profiles over bedrock or uniform half-space
have also been studied. Analytical expressions

(b) Arbitrary mat foundation over uniform halfspace


The effective period of arbitrary mat
foundation over uniform half-space is calculated
directly according to the expression of Table 1,
proposed in FEMA 450 Commentary (BSSC,
2003). The effective weight and height of the
structure are equal to the total W and h for sdof
systems where the whole mass is concentrated
in one specific level. The value of the
foundation characteristic radius taken into
consideration during the stiffness calculations
varies, depending on the response mode as
presented in Table 1.

227

mstr
hstr
2B
Shear wave velocity VS
Shear modulus G

Density
Damping

Fig 2: Investigated soil-structure system


Table 1: Theoretical expressions for dynamic stiffness calculation
Foundationsoil
properties

Response
Mode

(a)
Square foundation
over uniform halfspace

Horizontal

Static Stiffness

Ky =

Kr =

Rocking

(b)
Arbitrary mat
foundation over
uniform half-space

Dynamic Coefficient

L
proposed
k y = k y ,0 ,

B
diagram

9GB
2v

3.6GB
1 v

effective period TSSI = T 1 +

25 r h
1.12 r h 2
1 +
2
2

Vs T
rm3

Horizontal

8Gr 1 r
Ky =
1 +
2 v 2 H

Rocking

8Gr 3 1 r
m
Kr =
1 + m
3 (1 v ) 6 H

(d)
Strip foundation on
soil layer over
bedrock

Horizontal
Rocking

2L

Gazetas 1991,
Gazetas 1997

k rx 1 0.200

(c)
Arbitrary foundation
on soil layer over
bedrock

Ky

Reference

FEMA 450
(BSSC 2003)

y 1

(Commentary FEMA 450)


H
proposed
k y = k y ,0 ,

B
diagram

2G
B
1 + 2
2v
H

Kr
GB2
B

1 + 0.2
2L 2 (1 v )
H

k r 1 0.200

FEMA 450
(BSSC 2003),
Elsabee et al. (1977),
Kausel and Roesset (1975)

Gazetas and Roesset


(1976), Gazetas (1983),
Mylonakis et al. (2006)

In the expressions of the Table:

(a) 0 =
, 2B is the footing width and G is the soil shear modulus
VS
(b) =

W
, A o is the foundation area, h and W are the height and weight of the sdof structure, dynamic rocking coefficient
A0 h

(FEMA 450), r =

A0
, rm =

4 I0
and Io is the footing moment of inertia about axis vertical to the rocking motion

(c) H is the soil layer thickness, whereas stiffness values are valid for r

(a)

< 0.5 , where for K y , r = r and for K r , r = rm

(c)

(b)

Fig 3: Foundation geometries and soil profiles examined

228

(d)

structural period. Simulation efficiency of the


codes has been previously verified in wave
propagation and site effects investigation, as
well as in selected physical centrifuge
experiments of SSI systems (Pitilakis et al.
2005, Kirtas et al. 2006a).
In order to identify the effective structural
period TSSI, the results of numerical time-history
analysis are employed. An earthquake
recording is imposed at the base of the model
(bedrock level) and seismic waves propagate
throughout the soil deposit to the surface and
the structural base, triggering therefore soilstructure interaction phenomena. The response
period of the structure in each soil-structure
combination is then identified, using the Fourier
transform of the corresponding response timehistories at the base and the top of the structure.
Indeed the top-to-base ratio of the response
histories depicted in the frequency domain
(transfer function), can be utilized for the
determination of the resonance period at the
point of maximum motion amplification. This
effective period varies for different soil types,
given the same structural fixed-base T, subject
to different soil compliances that affect the
extent of the developing interaction phenomena.
Using the Fourier ratio instead of plain
modal analysis to identify the dynamic
characteristics of the soil-structure system can
provide additional information apart from the
effective period value. An indicative example in
the case of the structure with fixed-base period
equal to 0.2s is presented in Fig 4. The
resulting effective period of the structure can be
identified quite easily from the diagram in the
case of soil categories B and C. On the other
hand, an effective period value is also
determined when soil type D is examined, by
selecting the period value at the peak of the
Fourier ratio. Nevertheless, the frequency
content of the response in this particular
combination is not very clear, since the ratio
has almost uniform value for a significant
frequency range. Additional information is
therefore provided from the Fourier ratio,
highlighting
the
enhanced
interaction
phenomena in the case of the soft soil deposit
and questioning the use of a single effective
period value to characterize uniquely the overall
structural response. This phenomenon is also
reviewed in Aviles and Suarez (2002) for
systems involving short structures.

(c) Arbitrary foundation on soil layer over


bedrock
The relationships employed to calculate
dynamic stiffness of arbitrary foundation
geometry on soil layer over bedrock can be
found in FEMA 450 (BSSC, 2003), based on
the work of Elsabee et al. (1977) and Kausel
and Roesset (1975). The same expressions
with the addition of proper coefficients can be
also employed for embedded foundations on
soil stratum over bedrock.
(d) Strip foundation on soil layer over bedrock
Dynamic stiffness in the case of strip
foundation lying on soil layer over bedrock was
calculated using the expressions that can be
found in Mylonakis et al. (2006), based on the
work of Gazetas and Roesset (1976) and
Gazetas (1983). The strip foundation on soil
stratum over bedrock investigated during the
numerical simulation is more closely related to
the specific theoretical calculation, presenting
therefore a particular interest when reviewing
the results. The dynamic coefficient for
horizontal movement presents fluctuations with
frequency, depending on the H/B ratio and the
examined frequency range. In this paper, with
the acceptance of a small error, the horizontal
dynamic coefficient is assumed equal to 1 for all
calculations. Taking into consideration the
examined H/B ratio (>4), the dynamic
coefficient values calculated here are rather
overestimated as can be observed in the
corresponding diagram in Mylonakis et al.
(2006).
Numerical calculation of effective period values

Numerical plain strain calculation of the


effective period for the structure with the strip
footing is based on the model schematically
depicted in Fig 2. The value of L/H ratio is
selected equal to 4, to avoid any undesired
reflected waves emanating from the side
boundaries, whereas H/B ratio is equal to 4.6.
Soil behavior is considered linear elastic,
characterized by the shear modulus G and
density that determine the shear wave
velocity within the soil deposit as:
G
(4)
VS =

General purposed FE codes ADINA (2005)


and ANSYS (2000) are employed for the
numerical determination of the effective

229

16

Soil type B

12

mnorm=0.5
mnorm=1
mnorm=2

Structural Transfer Function

0
0.1

0.2

0.3

0.4

12

0.5

0.6

0.7

0.8

0.9

1.0

0.7

0.8

0.9

1.0

0.7

0.8

0.9

1.0

Soil type C

0
0.1

0.2

0.3

0.4

12

0.5

0.6

Soil type D

0
0.1

0.2

0.3

0.4

0.5

0.6

T (sec)

Fig 4: Top-to-base Fourier ratios (transfer functions) for structure with T=0.2s (fixed base)
Table 2: Theoretical and numerical analysis results of effective period (hnorm=1)
Soil-mnorm case

Fixed
base
structure
T=0.2s

Theoretical

Numerical

Fixed
base
structure
T=0.4s

Theoretical

Numerical

Fixed
base
structure
T=0.6s

Theoretical

Numerical

B-0.5

B-1.0

B-2.0

C-0.5

C-1.0

C-2.0

D-0.5

D-1.0

D-2.0

(a)

0.21

0.219

0.236

0.239

0.271

0.325

0.338

0.425

0.555

(b)

0.208

0.216

0.231

0.232

0.259

0.306

0.311

0.389

0.507

(c)

0.208

0.216

0.23

0.231

0.259

0.305

0.311

0.388

0.505

(d)

0.213

0.225

0.247

0.249

0.289

0.354

0.366

0.469

0.622

ADINA

0.206

0.21

0.217

0.213

0.251

0.266

0.268

0.39

0.506

ANSYS

0.217

0.226

0.245

0.24

0.253

0.281

0.336

0.347

0.488

(a)

0.405

0.41

0.419

0.42

0.438

0.473

0.478

0.543

0.65

(b)

0.404

0.408

0.416

0.416

0.432

0.461

0.463

0.518

0.612

(c)

0.404

0.408

0.416

0.416

0.431

0.461

0.463

0.517

0.611

(d)

0.406

0.413

0.425

0.426

0.449

0.494

0.499

0.578

0.709

ADINA

0.402

0.406

0.41

0.41

0.418

0.431

0.422

0.506

0.532

ANSYS

0.406

0.41

0.422

0.422

0.44

0.471

0.482

0.506

0.546

(a)

0.603

0.606

0.613

0.613

0.625

0.65

0.652

0.7

0.786

(b)

0.603

0.605

0.611

0.611

0.622

0.643

0.643

0.683

0.757

(c)

0.603

0.605

0.611

0.611

0.621

0.642

0.643

0.683

0.755

(d)

0.604

0.608

0.617

0.617

0.633

0.665

0.668

0.729

0.835

ADINA

0.602

0.602

0.611

0.611

0.611

0.621

0.63

0.661

0.719

ANSYS

0.611

0.611

0.611

0.611

0.621

0.63

0.621

0.63

0.64

230

uniform half-space and arbitrary foundation


lying on soil layer over bedrock respectively.
Both expressions, proposed by FEMA 450
(2003), produce almost identical results,
indicating that the examined H/B ratio is rather
large to interfere to the calculation of the
effective period value. On the other hand, the
observed difference between (a) and (b) is quite
large considering that they refer to the same
subsoil conditions and equivalent foundation
properties (i.e. square and arbitrary footing
shape). Furthermore, it is evident from Table 1
that when the thickness of the soil layer
underneath the foundation is small compared to
the characteristic footing dimension, the
impedance factor is increased compared to the
case of uniform halfspace (which occurs when
H/B approaches infinity). This is not depicted in
Fig 5 where the foundation case (d) constantly
presents larger values of effective period (i.e.
smaller overall stiffness) compared to any other
case, even though the corresponding stiffness
parameters where already overestimated during
the calculations since the horizontal dynamic
coefficient was assumed equal to 1 as
mentioned previously.
It is therefore deduced from all the above
observations, that in order to make an accurate
estimation of the effective period it is not
compulsory to consider in great detail the
particular characteristics of the involved subsoil
conditions and foundation properties, at least
for cases similar to the ones examined here. A
rough description of the soil-structure system in
terms of subsoil geometry and foundation
shape, along with the proper determination of
the soil type and structural mass, height and
fixed base period, seem to provide a
satisfactory effective period value. The variation
in the case of the utilized theoretical
relationships lies in relatively small values, not
exceeding 15-20% between the different
calculations, verifying the coherence amongst
the aforementioned expressions. Taking into
consideration the inherent difficulties in the
prediction of the dynamic characteristics due to
soil-structure interaction, a more detailed
description of the systems dynamic response
than simply the estimation of TSSI may require
the thorough procedures proposed in the
corresponding references.
It is evident in all examined structural
periods that the numerical results tend to

EVALUATION OF THE RESULTS AND


VALIDATION OF NUMERICAL SIMULATIONS

Validation of effective period numerical


calculations is presented in the next few
paragraphs, based on the comparison of
numerical analyses using sdof structures on
rigid strip foundations, towards calculations that
consider theoretical expressions given in the
bibliography. The examination of the several
soil typestructural mass combinations for
each structural fixed-base period provides the
results of Table 2. In Fig 5 a comprehensive
diagram of the various investigated cases is
illustrated, presenting the variation of the
effective period for each combination. In the
same diagram both theoretical and numerical
results are presented, allowing for a
comparative evaluation of the corresponding
calculations. It is important to mention here that
increased values of structural mass or lower soil
categories are directly related to increasing
interaction effects and larger effective period
values. Nevertheless, when combinations of the
parameters soil type-structural mass are
examined (X axis in Fig 5 diagram), it is not
possible to predict in advance the case that
presents more significant period modifications.
The
sequence
of
the
soil-structure
combinations examined in the X axis of the
specific diagram is therefore indicative, since for
example C-2 case (soil type-normalized mass)
could result in larger effective period than D-0.5
combination that follows.
Natural period modification due to soilstructure interaction effects is more pronounced
in the case of lower period (stiffer) structures
with increased normalized mass. Indeed, the
structure of fixed base period equal to 0.2s
approaches effective period values of the order
of 0.5s to 0.6s, in the extreme case of very soft
subsoil conditions and large superstructure
mass that result in the development of
significant interaction effects. The modification
of the natural period in the specific structure is
equal to 150-200% compared to the fixed base
value. On the other hand, maximum effective
period alteration is reduced when more flexible
structures are examined, presenting increase of
40-75% and 17-40% in the cases of 0.4s and
0.6s structures respectively.
It is quite interesting to comment on the
calculated effective periods of cases (b) and (c)
that refer to arbitrary mat foundation over

231

Structural Period T=0.2s (fixed base)

1.00
0.90
0.80

T SSI (sec)

0.70
0.60
0.50

(a) Square on halfspace


(b) Mat found. arbitrary on halfspace
(c) Flexible arbitrary on soil layer
(d) Strip on soil layer
Numerical Analysis ADINA
Numerical Analysis ANSYS

0.40
0.30
0.20
0.10
0.00
B-0.5

B-1

0.80

T SSI (sec)

0.70
0.60
0.50

C-0.5
C-1
C-2
Soil Type - m norm

D-0.5

D-1

D-2

D-1

D-2

D-1

D-2

Structural Period T=0.4s (fixed base)

1.00
0.90

B-2

(a) Square on halfspace


(b) Mat found. arbitrary on halfspace
(c) Flexible arbitrary on soil layer
(d) Strip on soil layer
Numerical Analysis ADINA
Numerical Analysis ANSYS

0.40
0.30
0.20
0.10
0.00
B-0.5

B-1

B-2

C-0.5
C-1
C-2
Soil Type - m norm

D-0.5

Structural Period T=0.6s (fixed base)

1.00
0.90
0.80

T SSI (sec)

0.70
0.60
0.50
0.40
0.30
0.20
0.10

(a) Square on halfspace


(b) Mat found. arbitrary on halfspace
(c) Flexible arbitrary on soil layer
(d) Strip on soil layer
Numerical Analysis ADINA
Numerical Analysis ANSYS

0.00
B-0.5

B-1

B-2

C-0.5
C-1
C-2
Soil Type - m norm

D-0.5

Fig 5: Effective period variation with soil type-normalized mass combinations (hnorm=1)

232

period
modification
due
to
subsoil
interventions of limited scale in the foundation
area. Indeed several interventions result in
subsoil stiffness increase in order to enhance
soil strength and reduce settlements. The
indirect consequence of similar mitigation
methods would be the modification of the soilstructure systems dynamic properties,
altering simultaneously the input motion and
the seismic response in the case of an
earthquake event (Kirtas et al., 2006b). Yet,
as mentioned previously, seismic codes do
not require a proper consideration of the
imposed
foundation
soil
properties
modification on the input motion and the
dynamic characteristics of the soil-structure
system.
The specific intervention case examined
here, refers to a soil shear modulus increase
by a factor of 10, resulting in a much stiffer
soil formation compared to the initial
conditions at a specific area below foundation
(Fig 1). Therefore, soil category D with a
reference shear modulus value equal to 18
GPa, after the intervention is assumed to
obtain a G value equal to 180 GPa,
corresponding to shear wave velocity of 316
m/s. Such an increase implemented in soil
type B would result in a mixture that can no
longer be characterized as soil, since the
modified shear wave velocity is larger than
1000 m/s resembling properties of soft rock
formations. Thus, the theoretical predictions in
the diagrams of Fig 8 referring to soil type B
(marked as B* to denote enhanced soil
stiffness), should not be taken into
consideration since the modified soil
properties are outside the valid range of the
corresponding relationships.
Moreover, during the evaluation of the
effective period, an inherent deviation is
expected between the theoretical predictions
and the numerical simulations due to the
partial application of the soil stiffness
enhancement. Indeed, theoretical predictions
utilizing the modified soil stiffness would
actually concern a general mitigation in the
surface layer over the whole area beneath the
foundation, providing therefore erroneously
smaller effective period estimates in the
present case that only a limited area is
modified.

slightly underestimate, compared to the


existing analytical expressions, the effective
structural
period
when
soil-structure
interaction is considered. Yet, from a
preliminary evaluation of the obtained results,
both employed numerical codes seem to
follow closely the trend of the TSSI variation
with different soil-structure combinations.
Effective periods calculated with ANSYS are
closer to the theoretical solutions in the case
of 0.2s and 0.4s structures, whereas in the
case of the 0.6s structure it is ADINA
calculations that fit better the theoretical
results. In general, effective period values
calculated with ADINA appear to form a lower
bound to the examined approaches,
presenting a constant deviation compared to
theoretically calculated values. On the other
hand, ANSYS results do not present a uniform
pattern. In certain cases it coincides with the
analytical results whereas it is failing to follow
the trend of the various analytical approaches
in the specific structure of T=0.6s.
The deviation of the numerical simulation
results compared to cases (c) and (d) of the
corresponding analytical calculations is
presented in the diagrams of Fig 6. The
aforementioned conclusion regarding the
efficiency of each FE code is verified from the
specific diagram. The maximum deviation of
the numerical calculations compared to the
analytical solutions is slightly over 20%,
revealing an overall satisfactory performance
of ANSYS and ADINA on TSSI determination.
When higher structures are investigated,
the interaction effects become more
significant resulting in increased effective
period with structural height as presented in
Fig 7 for the specific structure of 0.4s fixed
base period and normalized structural mass
equal to 1. Numerical investigation with
ADINA code revealed an even better
agreement between theoretical predictions
and numerical simulations. Indeed numerical
results lie within the range of the analytical
calculations, indicating an efficient modelling
of the interaction phenomena in terms of
structural period modification.
APPLICATION OF SUBSOIL STIFFENING
INTERVENTION

An interesting question within the context


of the present paper concerns the effective

233

Numerical Calculations deviation from


foundation case (c)
100
75

Numerical Calculations deviation from


foundation case (d)
100

Tstr=0.2s (fixed-base)

50
25

25

0
-25

0
-25

-50
-75

-50

-100

-100

-75

Tstr=0.4s (fixed-base)

Deviation (%)

75
50
25
0
-25
-50
-75
-100

Tstr=0.4s (fixed-base)

75
50
25
0
-25
-50
-75
-100

Tstr=0.6s (fixed-base)

75

Tstr=0.6s (fixed-base)

75

50
25

50

0
-25

0
-25

25

-50
-75

-50

ADINA Deviation (%)


ANSYS Deviation (%)

-100

ADINA Deviation (%)


ANSYS Deviation (%)

-75
-100

B-0.5

B-1

B-2 C-0.5 C-1

C-2 D-0.5 D-1

D-2

B-0.5

B-1

B-2 C-0.5 C-1

C-2 D-0.5 D-1

Soil type - mnorm

Soil type - mnorm

Fig 6: Deviation of numerical calculations from theoretical approaches (c) (diagram on the left)
and (d) (diagram on the right)

Structural Period T=0.4s (fixed base)


1.00
(a) Square on halfspace
0.90
0.80

(b) Mat found. arbitrary on halfspace


(c) Flexible arbitrary on soil layer

0.70
(d) Strip on soil layer
T SSI (sec)

Deviation (%)

Tstr=0.2s (fixed-base)

75
50

0.60

Numerical Analysis ADINA

0.50
0.40
0.30
0.20
0.10
0.00
B-1

B-2

B-3

C-1
C-2
C-3
Soil Type - h norm

D-1

D-2

Fig 7: Effective period variation with soil type-normalized height combinations

234

D-3

D-2

Structural Period T=0.2s (fixed base)

0.60
0.55
0.50

TSSI (sec)

0.45
0.40
0.35
0.30

*(a) Square on halfspace


*(b) Mat found. arbitrary on halfspace
*(c) Flexible arbitrary on soil layer
*(d) Strip on soil layer
*Numerical Analysis ADINA
*Numerical Analysis ANSYS
Numerical Analysis ADINA
Numerical Analysis ANSYS

0.25
0.20
0.15
0.10
B*-0.5

B*-1

B*-2

C*-0.5
C*-1
C*-2
Soil Type - mnorm

D*-0.5

D*-1

D*-2

D*-1

D*-2

D*-1

D*-2

Structural Period T=0.4s (fixed base)

0.80
0.75
0.70

TSSI (sec)

0.65
0.60
0.55
0.50
0.45
0.40
0.35
0.30
B*-0.5

B*-1

B*-2

C*-0.5
C*-1
C*-2
Soil Type - mnorm

D*-0.5

Structural Period T=0.6s (fixed base)

1.00
0.95
0.90

TSSI (sec)

0.85
0.80
0.75
0.70
0.65
0.60
0.55
0.50
B*-0.5

B*-1

B*-2

C*-0.5
C*-1
C*-2
Soil Type - mnorm

D*-0.5

Fig 8: Effective period variation with soil type-normalized mass combinations (* to denote upgraded soil)
(stiffened soil in black numerical calculation of unmodified soil cases are in grey colour)

235

constantly underestimate the effective period


values, due to the inherent consideration of an
overall application of the soil properties
enhancement. Therefore, in similar intervention
cases, numerical investigation is deemed more
appropriate even for a rough estimation of the
TSSI values.

Observation of the effective period variation in


Fig 8 verifies the above prediction. Theoretical
relationships result systematically in smaller
effective period values compared to the
numerically obtained results. The inadequacy of
the analytical relationships to predict the TSSI
value without the proper consideration of the
limited intervention extent is therefore evident,
especially when soft soil conditions and large
superstructure mass characterize the initial soilstructure case. The relative deviation between
ADINA and ANSYS is similar to the one
observed in the reference soil-structure
systems, yet this time forming the upper bound
of the illustrated values. Comparison of
numerical results (grey and black lines in Fig 8)
is indicative of the significant intervention effect,
leading the structural response to frequencies
closer to the fundamental frequency of the fixed
base structure.

REFERENCES
ADINA (2005) "Automatic Dynamic Incremental
Nonlinear Analysis. Theory and Modelling Guide",
ADINA R&D, Inc.
ANSYS (2000) "ANSYS Users manual. Version 8.1",
SAS IP Inc., Houston, USA
Aviles J., Suarez M. (2002) "Effective Periods and
Dampings of Building-Foundation Systems
Including Seismic Wave Effects", Engineering
Structures, Vol. 24, 553-562
Building Seismic Safety Council (2003), "FEMA 450
- NHRP Recommended Provisions for Seismic
Regulations for New Buildings and Other
Structures", Federal Emergency Management
Agency, Washington D.C.

CONCLUSIONS

The effective structural period in several


soil-structure cases has been estimated using
well-known available theoretical expressions
and numerical simulations with FE codes
ADINA and ANSYS. Subsoil conditions refer to
soil categories B, C and D according to EC8,
whereas three structural fixed base periods as
well as three different superstructure masses
have been implemented during the investigation
to include a wide range of representative soilstructure cases. Validation of the effective
period calculations from the numerical
simulations that was based on theoretical
solutions, revealed a satisfactory approximation
of the structural period modification due to soilstructure interaction. Application of the same
comparison in soil-structure cases of locally
mitigated subsoil area beneath the foundation,
indicates the induced reduction of the effective
period due to similar interventions of enhanced
soil stiffness. Numerical investigation revealed
that the imposed alteration of the systems
dynamic characteristics, due to the employed
mitigation, may result in a significant
modification of the seismic response, especially
in cases of stiff structures founded on soft soil
conditions. On the other hand, the inadequacy
of the analytical relationships to predict the TSSI
value without the proper consideration of the
limited
intervention
extent
is
evident.
Calculations based on analytical relationships

CEN (2002) prEN 19985: Eurocode 8: Design of


Structures for Earthquake Resistance, Part 5:
Foundations,
Retaining
Structures
and
Geotechnical Aspects, European Committee for
Standardisation, Brussels
Elsabee F., Kausel I. and Roesset J.M. (1977)
"Dynamic Stiffness of Embedded Foundations",
Proceedings of the ASCE Second Annual
Engineering Mechanics Division Specialty
Conference, pp. 40-43
Gazetas G. (1983) "Analysis of Machine Foundation
Vibrations: State of the Art", Soil Dynamics and
Earthquake Engineering, Vol. 2, No. 1, pp. 2-42
Gazetas G. (1997) "Dynamic Soil-Structure
Interaction During Earthquakes", Proceedings of
the Advanced Study Course on Seismic Risk
(SERINA), ITSAK, Thessaloniki, Greece
Gazetas G. (1991) "Formulas and Charts for
Impedances of Surface and Embedded
Foundations",
Journal
of
Geotechnical
Engineering Division, ASCE, Vol. 117, No. 9, pp.
1363-1381
Gazetas G. and Roesset JM. (1976) "Forced
vibrations of strip footings on layered soils",
Proceedings of the Specialty Conference on
Methods of Structural Analysis, Dynamic SoilStructure Interaction Session, ASCE, Vol. 1, pp.
115-131

236

Jennings P.C., Bielak J. (1973) "Dynamics of


Building-Soil Interaction", Bulletin of the
Seismological Society of America, Vol. 63, No. 1,
pp. 9-48
Kausel E. and Roesset JM. (1975) "Dynamic
Stiffness of Circular Foundations", Journal of the
Engineering Mechanics Devision, ASCE, Vol.
101, No. 6, pp. 771-785
Kirtas E., Rovithis E., Pitilakis K. and Sextos A.
(2006a) "Numerical Investigation of Potential
Foundation Intervention as a Means for
Mitigating Seismic Risk", Proceedings of the 8th
U.S. National Conference on Earthquake
Engineering, San Francisco, California
Kirtas E., Rovithis E. and Pitilakis K. (2006b)
"Numerical Investigation of Subsoil Interventions
Towards Structural Seismic Risk Mitigation",
Proceedings of the First European Conference
on Earthquake Engineering and Seismology,
Geneva, Switzerland
Kirtas E., Trevlopoulos K., Rovithis E., Pitilakis K.
(2007), "Discussion on the Fundamental Period
of Sdof Systems Including Soil-Structure
Interaction", Accepted for presentation in the 4th
International
Conference
on
Earthquake
Geotechnical Engineering, Thessaloniki, Greece
Ministry of Public Works (2000) "Greek Seismic
Code, EAK 2000", Athens (in Greek)
Mylonakis G., Nikolaou S. and Gazetas G. (2006)
"Footings Under Seismic Loading: Analysis and
Design Issues with Emphasis on Bridge
Foundations", Soil Dynamics and Earthquake
Engineering, Vol. 26, No. 9, pp. 824-853
Pitilakis K., Kirtas E. and Rovithis E. (2005) "Is it
Possible to Improve the Seismic Structural
Behaviour with Intervention to Subsoil and
Foundation Conditions?" Proceedings of the 1st
Greece-Japan Workshop: Seismic Design,
Observation and Retrofit of Foundations, 185202, Athens, Greece
Veletsos AS, Meek JW. (1974) "Dynamic Behaviour
of Building-Foundation Systems", Earthquake
Engineering and Structural Dynamics, Vol. 3, No.
2, pp. 121-138
Wolf JP. (1985) Dynamic soil-structure interaction,
Prentice-Hall, Englewood Cliffs, NJ

237

Analytical Study on Seismic Design of Footings


Subjected to Earthquake Loading
K. Kosa1, T. Ando2, Y. Adachi3, M. Shirato4
1

Kyushu Institute of Technology, Japan


2
Hanshin Expressway Co., Japan
3
Hanshin Expressway Co., Japan
4
Public Works Research Institute, Japan

Abstract
The upper part of the footing on the loading side suffers tension damage when a bridge pier
is subjected to earthquake loading. To find the mechanism of this damage, a seismic
loading test was conducted and the obtained results were evaluated using two-dimensional
elasto-plastic finite element method (FEM) analysis. It was found that the upper part of the
footing on the loading side is damaged by the bending-preceding type shear, and that this
damage occurs after the reinforcement in the upper footing has reached yielding due to a
horizontal shear force. It was also found that this shear damage usually does not result in a
brittle fracture owing to the confinement effect provided by the main reinforcement in the
lower part of the footing.

INTRODUCTION

conducted a seismic behavior simulation test


and evaluated the obtained results using twodimensional elasto-plastic FEM analysis.

When a bridge pier with a pile foundation is


subjected to earthquake loading, both the upper
and lower parts of the footing (hereinafter
referred to as upper and lower footings) on the
loading side are tensioned due to an inertia
force occurring on the pier. The behavior of the
lower footing under this loading has been
studied well1). On the other hand, the behavior
of the upper footing which should be designed
against tensile force from the column and piles,
has been experimented little and its damage
mechanism is still largely unclarified.
After the Hyogo-ken Nambu Earthquake
(Mw=6.9) in 1995, the horizontal seismic
coefficient in the Japanese Specification for
Highway Bridges was increased to twice the
conventional coefficient. Several studies also
reported that the footing shape is predominantly
determined by the resistance from the upper
footing on the loading side2). Therefore, for the
effective seismic design of footings, it is
important
to
understand
the
damage
mechanism of the upper footing under a tensile
force. As one such effort, the authors

EXPERIMENTAL PROGRAM

Specimens and experimental method


Specimens were constructed to be a
monolithic type having a pier, a footing, and
piles. They were constructed to 1/3 and 1/2
scales of an actual bridge which sustained C
rank damage (slight damage) in the Hyogo-ken
Nambu Earthquake. Fig. 1 shows the structure
of specimens. Table 1 shows their attributes.
The effective depth of footing was the same for
the specimens of the same scale, but their
shear span ratio (a/d) was changed to 0.39,
0.75, or 1.0. The reinforcement ratio on the
upper and lower footings of Specimen No. 3
was increased to 1.5 times that of Specimen No.
1. Load was applied by two axial loading
manner. Namely, a horizontal load equivalent to
a seismic force was applied to the upper part of
the column, while a vertical load of 1.6 N/mm2
equivalent to the reaction force of the

238

Table 1: Attributes of specimen


PC steel bar

No.1

Vertical jack

a/d
Width (mm)
Depth (mm)
Footing Reinforce Upper
ment ratio
Lower
(%)
Spec. for
reinforcement

200

Horizontal jack
Reaction
wall

Spesimen

a=300

340

640

Loading
direction

2000
3150

550

0.266

0.320

0.75
1830

1990 Spec.

No.3
0.75
1830
0.120

No.4
1/2
0.39
2500
750
0.075

0.408
0.144
1.5 x 1990
1990 Spec.
Spec.

a=300

340

500 450

640

0.080

No.2
1/3
1.00
2030
450
0.108

Scale

340

Unit:mm

300 550 300


340
1150
1830

Upper face of
specimen

Fig. 1: Structure of specimen

Fig. 2: Damage to the upper footing

Damage to the footing


1) Damage to the upper footing
In each specimen, dominant damage
occurred in the upper footing on the loading
side, as shown in Fig. 2. The damage
developed in the following manner.
Cracking started at the intersection of the
upper footing and the column bottom. They
expanded to the right and left sides and to the
load application side.
The main reinforcement in the upper footing
sustained a yield strain at the foot of the column
on the loading side. Almost concurrently, the
main reinforcement in the lower footing on the
compression side also sustained a yield strain.
The load reached its maximum and cracks
at the corner of the column bottom started to
propagate towards the corner of the footing.
After reaching the maximum load, the loaddisplacement curve began to decrease
gradually. Then, the upper footing on the
loading side began to rise, and a corn-shaped
failure linking the corner of the column bottom
and the corner of the footing was clearly found
(the black area in Fig. 2).

Horizontal load (kN)

superstructure was being applied to the column


top as a uniform load.

Fig. 3: Horizontal load-displacement


relationship at the loading position

relationship at the horizontal load application


position. After the main reinforcement in the
upper footing yielded and the maximum
strength is reached, the load began to decrease
gradually. In the case of Specimen No. 2 with a
large shear span ratio, the behavior was similar
up to the maximum strength, but a punching
shear failure occurred in the footing during the
decrease of load.

2) Load-displacement relationship at the


horizontal load application position
Fig. 3 shows the load-displacement

239

3) Damage to the cross section of footing


A core was taken from the cross section of
the footing in Specimen No. 4. Fig. 4 shows the
damage sustained. A diagonal crack due to the
tensile force from the column and piles is seen.
As the upper footing on the loading side rose,
this tensile shear crack appears to be a
dominant failure cause of this specimen.

Main reinforcement in the column

Column

Shear damage due to


tensile force

4) Footing damage vs. ductility


Table 2 shows the strength and ductility of
each specimen. From the damage observed, it
is considered that the load decrease was
caused by the tensile shear damage of the
footing on the loading side. If the strength
reduction up to yielding of main reinforcement
in the upper footing is taken as the ductility of
the footing, a ductility of about 2 can be
expected.

Pile

Fig. 4: Damage to the footing (No. 4)


Table 2: Strength and ductility of specimens
Hor. Load
(kN)
Yield strength
Hor. displacement
(mm)
Hor. load
(kN)
Maximum
strength
Hor. displacement
(mm)
Hor. load
(kN)
100% x yield
strength
Hor. displacement
(mm)
max/y
Ductility
100%/y

ANALYTICAL PROGRAM

Analysis method
For the analysis, two-dimensional elastoplastic FEM analysis was used. The width of the
footing was made to b (width of column) + d
(effective depth of footing) = 940 mm from the
range where the main reinforcement in the
upper footing reached yielding4).
As the loading conditions, a monotonic
load was applied to the upper side of the
column by the displacement control method
while a load equivalent to the dead weight of
the superstructure was being applied to the
column top. As the boundary condition, the
column bottom was completely fixed.

No.1

No.2

No.3

No.4

Py

245

338

342

814

22

21

25

25

Pmax

273

370

367

1032

max

38

33

40

60

P100%

245

338

342

814

100%

54

45

60

150

1.7

2.1

1.6

2.4

2.5

2.3

2.4

6.0

Table 3: Characteristics of concrete and


reinforcement
Compressive
strength
(N/mm2)
28.028

Column
Footing, pile

Material model
As the elements, plane stress elements
and linear elements were used for the concrete
and
reinforcement,
respectively.
The
reinforcement and concrete were assumed as
completely bonded. Table 3 shows the
materials characteristics of the concrete and
reinforcement. As the failure criteria, the
Drucker-Prager criterion was used for the
compressive side of the concrete and the
maximum principal stress criterion for the
tensile side5). Fig. 5 shows the stress strain
model of the concrete5). In the compression
increase range, a quadratic curve was used up
to the compressive strength, and after this
strength the stress was made to decrease

Reinforcement

Tensile
strength
(N/mm2)
2.107

Modulus of
elasticity
(N/mm2)
1.35104

25.284
1.637
1.24104
Modulus of
Yield strength Yield strain
elasticity
()
(N/mm2)
(N/mm2)
5
345
1640
2.010

Poissons
ratio
0.2
0.2
Tensile
strength
(N/mm2)
490

ft
cu

4cu

t
cr

= f c(

c
0 . 002

)( 2

c
0 . 002

0.75GF/(ftL)
5GF/(ftL)

0.2fc
)

fc
Fig. 5: Stress-strain of concrete

240


u
y

Es
100
Es

Fig. 6: Shear reduction coefficient


Fig. 7: Stress-strain of reinforcement

linearly. In the tensile range, the stress was


assumed to increase linearly up to the
maximum tensile stress (ft), and after that a 1/4
model which took softening and failure energy
into account was used. The failure energy of
the concrete Gf was assumed as 0.1N/mm. The
diagonal line of an element was used as the
equivalent length (leq) of an element which is
required for analysis. Fig. 6 shows the
relationship between the shear stress transfer
coefficient and the strain after the occurrence of
cracks.6) As shown in Fig. 7, the model for the
stress-strain relationship of reinforcement had a
yield plateau range of 8.5 times the yield strain
after reaching the reinforcement yielding, and
then took strain hardening into account.

Fig. 8: Load-displacement relationship at the


horizontal load application position

ANALYTICAL RESULTS AND DISCUSSIONS


In the analysis, all specimens showed a
behavior similar to an experimental behavior
until around the maximum load when damage
occurred to the footing on the loading side.
Taking Specimen No. 1 as the representative,
this process is reviewed in detail. Here, the
tensile strain is given as positive and the
compressive strain as negative.
Behavior up to flexural yielding
1) Horizontal load-displacement relationship
Fig. 8 shows the load-displacement relationship
at the horizontal load application position. In the
experiment, the horizontal displacement
increased after the reinforcement in the upper
footing yielded, and the load began to decrease
after reaching the maximum strength. In the
analysis, the horizontal displacement also
increased after the yielding of reinforcement,
but the load decrease was not reproduced due

Fig. 9: Distribution of maximum principal strain in the


footing on the loading side

241

Column
Upper
reinforcement
Assumed plane for
shear damage

4000
Lower
reinforcement

2000

Maximum principal strain ()

Fig. 11: Distribution of strain in the upper footing on


the loading side

Fig. 10: Cracks under the maximum load

1000
0

Pile

(When : 14 mm displacement)

Fig. 13: Distribution of maximum principal strain in


the footing on the loading side

Fig. 12: Distribution of maximum principal strain in


the footing on the loading side

3) Strain of main reinforcement in the upper


footing on the loading side
to divergence. This is probably due to a small
amount of reinforcement in the upper footing
and the effect of two-dimensional modeling.

Fig. 11 shows the distribution of strain in the


main reinforcement in the upper footing. Good
agreement was seen between the analytical
and experimental results concerning the
distribution of strains and the occurrence of a
large strain at the outer rim of the column
bottom on the loading side as the load
increased.

2) Maximum principal strain in the upper


footing on the loading side
Fig. 9 shows the maximum principal strain
in the upper footing on the loading side when
the displacement was 9 mm. Fig. 10 describes
the distribution of cracks under the maximum
load in the experiment. In the experiment, the
dominant cracks occurred in the upper footing
and flexural cracking started at the front side of
the column. In the analysis, a maximum
principal strain in the horizontal direction also
occurred in the upper footing on the loading
side. It is also presumed that flexural cracks
occurred at the front side of the column.

Behavior after flexural yielding


1) Shear damage to the footing on the
loading side
Fig. 12 shows the distribution of maximum
principal strain in the footing on the loading side
when the displacement was 14 mm. In the
experiment, diagonal cracking penetrating from
the upper face to the lower face of the footing

242

Horizontal force in the column


bottom-pile head range
Horizontal shear force
Horizontal force at the pile head
Horizontal force at the pile head

Strain in the lower reinforcement ()

400
300
200
100
0
-100
0

Fig. 14: Distribution of main reinforcement in the


lower footing on the loading side

Horizontal shear force

5
10
15
Displacement at the loading position (mm)

20

Fig. 16: Relationship between horizontal shear force


and loading position

Horizontal force at
the column bottom

Vertical force by
column reinforcement

Vertical force by
column reinforcement

Horizontal force at
the pile head

Cross section for check


of horizontal shear force

Vertical force
at the pile head

Fig. 15: Schematic of shear forces acting on the footing

was confirmed, as seen in Fig. 10. In the


analysis, a maximum principal strain in the
diagonal direction which was different from the
flexural cracking was caused at the shear span
position indicated by the circle in the figure.
Because this crack was considered as a shear
crack from its extension in the diagonal
direction, a shear damage cross section which
was perpendicular to the strain was assumed.
Fig. 13 shows the relationship between the
maximum principal strain occurring on the
assumed shear damage cross section and the
displacement at the loading position. The
maximum principal strain here is the mean
value of principal strains which occurred in the
element crossing the assumed shear damage
cross section. Strain reaching the maximum
tensile stress did not occur until yielding of

upper reinforcement when the footing was


considered to be taking flexural behavior. But,
after the yielding of upper reinforcement, strain
increased significantly.
Fig. 14 shows the distribution of strain in
the reinforcement in the lower footing on the
loading side. While the load was small,
compressive
strain
occurred
in
the
reinforcement in the lower footing. But, after the
yielding of reinforcement in the upper footing,
tensile strain also occurred in the reinforcement
in the lower footing in both the experiment and
analysis.
2) Shear force acting on the footing
Fig. 15 shows a schematic of the shear
forces acting on the footing. Dominating forces
contributing to the shear damage of the footing
were the shear force in the horizontal direction

243

Horizontal shear force (kN)

due to acting inertia force and the shear force in


the vertical direction due to pullout action of the
column and piles. From the fact that the shear
damage shown in Fig. 12 occurred after the
yielding of reinforcement in the upper footing,
and then the tension force also occurred in the
reinforcement in the lower footing, it is
considered that the force in the direction the
main reinforcement is affected, namely, the
shear force in the horizontal direction, is
predominant.
3) Horizontal shear force acting on the
footing on the loading side
Fig. 16 shows the relationship between the
horizontal shear force and displacement at the
loading position. Here, the acting horizontal
shear force was calculated by subtracting the
horizontal force acting on the pile head from the
horizontal force acting on the column bottom.
With the increase of horizontal displacement,
acting horizontal shear force increased
monotonically without showing an abrupt
increase or decrease.
Fig. 17 shows the relationship between the
ratio of horizontal shear force carried by the
concrete
and
reinforcement
and
the
displacement at the loading position. The shear
force carried by the reinforcement was
calculated from the stress acting on the
reinforcement which crosses the cross section
for the horizontal shear force check shown in
Fig. 15. The shear force carried by the concrete
was obtained by subtracting the shear force
carried by the main reinforcement from the
acting shear force.
In the elastic state, most of the horizontal
force was carried by the concrete. But, this ratio
began to decrease from around a displacement
of 7 mm when cracking appeared on the upper
footing, and instead the ratio carried by the
reinforcement
increased.
Therefore,
the
behavior in this range is considered to be close
to a flexural behavior.
In the case of general flexural behavior, a
major compression force acts on the
reinforcement in the lower footing after the
reinforcement in the upper footing has reached
yielding. But, here a tension force was
generated, and the ratio of shear force carried
by the reinforcement in the lower footing
increased and the ratio of shear force carried by

Fig. 17: Horizontal shear force carried

Fig. 18: Schematic of shear damage to the footing

the concrete decreased. At this time, shear


damage was probably caused in the footing.
After this damage, the main reinforcement in
the lower footing resisted the horizontal shear
force.
4) Mechanism of shear damage to the
footing
Fig. 18 shows a schematic of shear damage
to the footing due to a horizontal shear force. It
is considered that shear damage to the footing
is caused by the horizontal shear force acting
on the column and piles. When this force acts,
the upper footing becomes tensioned due to the
rotational behavior of the column. But, after the
reinforcement in the upper footing has yielded,
only the concrete must resist the shear force
and soon suffers shear damage at the middle of
the footing. Even though shear damage has

244

REFERENCES

occurred, a brittle fracture will not be caused


because the main reinforcement in the lower
footing still resists.

1) Shirato, M., et al. (2001) Study on the Shear


Strength Equation for Deep Beam-Footing
Connection,
Journal
of
Structural
Engineering (Japan), Vol. 47A, pp. 13151325
2) Fujii, Y., et al. (1999) Analysis of Shear
Effect on the Footing under New Highway
Specifications, Proc. of JCI, Vol. 21, pp.
1231-1236
3) Kosa, K., et al. (2000) Experimental Study
on the Behavior of Footing under Seismic
Loading with a Focus on the Shear Span
Ratio, Journal of Structural Engineering
(Japan), Vol. 46A, pp 1405-1412
4) Road Association of Japan (1996) Specifications
for
Highway
Bridges,
IV
Substructure
5) Chen, W.F. (translated by Shikibe, Kawazumi,
and Adachi) (1985) Plastic Analysis of
Concrete Structures
6) Rots, J.G. (1988) Computational Modeling of
Concrete Fractures Dissertation, Delft
University of Technology

CONCLUSIONS
The following conclusions were drawn from
the experiment and two-dimensional plastoelastic FEM analysis performed on the footing
subjected to earthquake loading.
(1) Shear damage to the upper footing on the
loading side is often preceded by the flexural
damage, because the ratio of the main
reinforcement in the upper footing is small.
(2) Shear damage to the footing occurs after
the main reinforcement in the upper footing
has reached yielding.
(3) Damage due to horizontal force does not
result in a brittle fracture owing to the
confinement effect provided by the main
reinforcement in the lower footing.

245

Development of the sheet-pile foundation;


a series of seismic proving tests
S. Higuchi1, H. Nishioka2, K. Tanaka1, M. Koda2, J. Hirao3
1

Technical Research Institute of OBAYASHI Co., Japan


2
Railway Technical Research Institute, Japan
3
Civil Engineering Technology Dept., OBAYASHI Co., Japan

Abstract
ABSTRACT: The authors have proposed the sheet-pile foundation as a new foundation type.
This foundation has advantages as follows; 1) Wider applicability to various soil conditions
than the shallow foundations, 2) Limited impact to the environment with construction work,
3) More economical than the pile foundations in terms of cost. In this paper, investigations
of the fundamental characteristics of the sheet-pile foundation during the earthquake for the
development of the design method of this structure are presented. A series of static loading
tests (laboratory and full-scale in the field) and centrifuge tests were carried out. Results
show the sheet-pile foundation has excellent performance against seismic force.

INTRODUCTION

Recently, development of construction


methods for densely populated urban area is
emphasized in Japan. For example, in order to
ease traffic congestion, railroads are re-laid on
viaducts. For this project, structures are usually
constructed very close to existing structures,
and the space allowed for construction work is
limited. In addition, it is required to reduce costs,
as well as minimizing the impact to the
environment, such as noise, vibration and
disposals from construction work.
Sheet-pile Foundation (SPF, hereafter),
which combines the footing and sheet-piles,
proposed as a new foundation form (Koda et al.
2003, Nishioka et al. 2004) is one solution.
Because of the confinement of the ground is
increased by the sheet-piles, both bearing
capacity and horizontal resistance of the SPF
are improved compared to those of the shallow
foundation. Therefore, the applicability became
wider than that of the shallow foundations. For
example, SPF can be adopted on the loose
sandy ground to which the pile foundation has
been usually applied. The construction cost of
SPF is almost the same as that of the shallow
foundation and more competitive than that of

246

the pile foundation. On the other hand, since


the pile work is not necessary, it can avoid
various disadvantages of pile foundation, such
as noise, vibration and the disposal of surplus
soil. Fig 1 shows an outline of the sheet-pile
foundation com-pared with the shallow
foundation and the pile foundation.
In this paper, a series of static loading tests
(laboratory and full-scale in the field) and
centrifuge tests carried out for the purpose to
evaluate the performance of the sheet-pile
foundation are presented and discussed.
Shallow
foundation

Sheet-Pile
foundation

Pile
foundation

Sheet-Piles

Dense Sand
(SPT N-value >30)

Loose Sand
(10< N-value >30)

Fig 1: Outline of the sheet-pile foundation.

HORIZONTAL LOADING TESTS

block of 100 mm in width B, and placed on the


model ground surface. Model sheet-pile was
made of phosphor bronze plates with 0.2 mm
thickness, and they were pressed to concavoconvex form shown in Fig 2. The length L that
was installed into the ground was 100 mm
(L/B=1.0) or 50 mm (L/B=0.5). The value of L
shown in Eq.1 of the model sheet-piles was the
same grade as that of prototype.
(1)
L = 4 k h D 4 EI L

1) Outline of model ground and foundations


A model ground was prepared in a rigid
container with dry sand. The model was a twodimensional in plane strain condition. The sand
containers side walls were made of transparent
acrylic plates to allow observation of the
deformation of the ground. In order to reduce
friction between the acrylic plate and sand,
rubber membranes were pasted on acrylic plate
with grease. Target points were marked on the
rubber membrane for measuring displacement
of the ground by an image processing system.
Table 1 summarizes the conditions of the
model ground. The relative density Dr of the
model ground was controlled by the height of
the sand hopper to 90% or 60%.
The model footing was made of aluminum

where : Characteristic value of pile (1/m), kh:


Coefficient of horizontal subgrade reaction
(kN/m3), D: Width of sheet-piles (m), EI:
Flexural rigidity of sheet-pile (kNm2), L: Length
of sheet-pile (m). Table 2 summarizes
specifications of the model sheet-piles.
2) Test procedures
In order to simulate the effects of inertia
force by the earthquake, the horizontal loading
tests were conducted. The horizontal loads
were applied to the bridge pier top.
An outline of the tests is shown in Fig 3.
The horizontal displacement was applied with a
screw jack statically and cyclically at a height of
230mm from the footing model bottom that was
corresponding to the bridge pier top. The
vertical load was applied to the pier top at 1.2
kN by an air cylinder, which was about 10% of
the bearing capacity of the shallow foundation
previously tested on the dense ground model.

Table 1: Conditions of modeled ground


Ground size
(W H D)
Material of ground
Dry unit weight d
Lubricated layer

2000 mm 580 mm
600 mm
Dry Toyoura sand
d = 16.2 kN/m3 (Dr = 90%)
d = 15.1 kN/m3 (Dr = 60%)
Rubber membrane (t=0.2mm)
with Grease (10m)

Horizontal
Displacement

Vertical Load
V = 1.2 kN

Fig 2: Picture of model sheet-pile


Table 2: Specifications of model sheet-piles

Thickness
Height of concavoconvex form
Width of Footing B
Width of
sheet-piles D
Young's modulus
E
Geometrical moment
of inertia I
Coefficient of
horizontal subgrade
reaction kh
Length L
L

Prototype
TYPE IV

Modeled
sheet-pile

15.5 mm

0.2 mm

340 mm

1.5 mm

4.8 m

100 mm

4.8 m
200 kN/mm
( Steel )

B=100mm

596 mm
2

1.2810-3 m4

D = 596 mm

h=230mm

M odel ground
Dry Toyoura sand

L=100mm
2

110 kN/mm
( Phosphor bronze )

Fig 3: Outline of horizontal reciprocal loading test.

42.9 mm4

78,600 kN/m3
( Sand N = 30)

45,700 kN/m3
( Dr=90% )

4.8 m (L/B=1.0)
2.4 m (L/B=0.5)
3.74 (L/B=1.0)

100 mm (L/B=1.0)
50 mm (L/B=0.5)
3.42 (L/B=1.0)

Sheet-Pile model

Table 3: Cases of horizontal loading tests

247

Case

Density of ground

Foundation Form

HD1
HL1
HL2
HL3

Dr=90% (Dense)

Shallow foundation
Shallow foundation
Sheet-pile foundation L/B=0.5
Sheet-pile foundation L/B=1.0

Dr =60% (Medium
dense)

0.4

of more than 10% of footing width, the SPF on


the same ground model (Case-HL2, HL3) has
only small settlements, which are as same as
that of shallow foundation on the dense ground
(Case-HD1). Therefore, it is clear that the
sheet-piles restrained the settlement.
Photo 1 shows the deformation of the
ground in two cases (Case-HL1 and HL2),
computed by the image processing system
previously described. The lines in the figures
show the locus of each target point until the
horizontal displacement at the top of pier
reaches 20mm. The deformation of the ground
was observed in a large area around the model
footing. In the case of the shallow foundation
(Case-HL1), the ground failed like a circular slip
Displacement (mm)
0

case-HL1 (SF)
case-HL3 (SPF L/B=1.0)

0.3

10

15

20

2
4
6
8
10

0.2

Load (kN)

Settlement (mm)

3) Test results
The relations between the horizontal load P
and displacement at the top of pier are shown
in Fig 4 and Fig 5. Fig 4 shows hysteresis
curves, and Fig 5 shows skeleton curves, which
connected the turning points on each loading
cycles together. The SPF (Case-HL2, HL3) has
higher horizontal resistance than that of shallow
foundations (Case-HD1, HL1). Since the loop of
the hysteric curve of the SPF is larger than that
of the shallow foundation, it is clear that the
hysteric damping of the SPF is larger than
shallow foundations. In addition, the residual
horizontal displacement of the SPF was almost
negligible after the experiment.
The settlement characteristic is another
important function of the railway structure. Fig
6 shows settlements of footings when the
horizontal displacement reached to the peak in
each cycle. Settlements became larger as the
increase of horizontal displacement in all cases.
Although the shallow foundation on the medium
dense ground (Case-HL1) had large settlement

case-HD1 (SF)
case-HL1 (SF)
case-HL2 (SPF L/B=0.5)
case-HL3 (SPF L/B=1.0)

12

Fig 6: Skeleton curves of settlements

0.1
0.0

-0.1
-0.2
-0.3
-25

-20

-15

-10

-5

10

15

20

25

Displacement (mm)

Fig 4: Hysteresis curves of P- relationship


a) Case-HL1: Shallow foundation

0.35
0.30

Load (kN)

0.25
0.20
0.15
case-HD1 (SF)
case-HL1 (SF)
case-HL2 (SPF L/B=0.5)
case-HL3 (SPF L/B=1.0)

0.10
0.05
0.00
0

10

15

20

b) Case-HL2: Sheet-pile foundation (L/B=1.0)

Displacement (mm)

Photo 1: Displacement computed by the Image


Processing System when horizontal displacement
reaches 20mm. (Dr=60%)

Fig 5: Skeleton curves of P- relationship

248

8. Sheet-pile length and ground density are


chosen as parameters of this experiment. Table
4 summarizes the conditions of the model and
ground. Shake table tests were carried out
utilizing the input motions as shown in Table 5.
2) Test results
Distribution of the vertical self load between
the sheet-piles and the footing bottom after the
centrifuge acceleration process are summarized
in Table 6. It is seen that larger self weight is
shared by the sheet-piles on the cases of either
the sheet-pile length become longer or the
ground became denser.
Predominant period of the ground-structure
system analyzed by the transfer functions (A/D
in Fig 7) during a small event (white noise
motion; Peak Acceleration=0.02g in prototype)
are summarized in Table 7. This shows the
initial stiffness of the SPF is larger than that of
the shallow foundations, and mainly influenced
by the ground condition.
Skelton curves, which consists of relations

CENTRIFUGE TESTS

1) Outline of the centrifuge models


Because of the stress dependent physical
properties of ground materials, centrifuge shake
table tests were conducted to simulate actual
characteristics of the SPF during earthquake
event. Centrifuge tests were carried out under a
25g centrifugal acceleration, therefore, the
model structures were scaled as 1/25, as
shown in Fig 7. A 9m high viaduct with 5m
square footing was selected as the prototype
(Photo 2). Sheet-pile was also scaled in this
experiment except connections as shown in Fig

Rigid Container

Strain Guages
SP40+pier2=42
Dry Sand

200 (5m)

250
11@16=176

Unit: mm

1900 (47.5m)

200
198
11

200
D

Fig 7(a): Centrifuge setup


(Two models were installed
in one model ground)

Shake Direction

100 (2.5m)

Earth Pres. Cell

300 200 (5m)

Disp. Transducer 2

11

Sheet-pile type III


t=13mm

400

Sheet-pile

t=0.5mm
10

15

50

Accelerometer

Shake Direction

400 (10m)

600 (15m)

in the limited area shallower than 50mm on


both sides of the footing. It is seen that the
deformation of the ground around the footing
spreads outward direction. On the other hand,
in the case of the SPF (Case-HL2), the
horizontal deformation of the ground is
restrained by the sheet-piles.

s=1/25
16

Fig 8: Sheet-pile model

Front/Back
Sheet-pile

Filled with
epoxy resin

Footing

Side Sheet-pile

Photo 2: Model structure and ground

Fig 7 (b): Plan view of the footing

249

ground. y of the SPF with L/B=1.0 is 0.6g in


this case, and this is twice as large as that of
the shallow foundation, of which the y is 0.3g,
as proportional to the dense ground cases.
Time histories of the earth pressure cells
installed at the bottom surface of the footings
during the severe earthquake input motion
(Bedrock PA=0.45g) event are shown in Fig 10.
From this figure, it is confirmed that relatively
large area under the footing is separated from
the ground surface during the earthquake event,
and the separation was recovered after the
earthquake.
Fig 11 shows the stress time histories of the
sheet-pile installed perpendicular to the shake
direction during the shake event mentioned
above. Both axial and bending stress
components are far smaller than its yield limit

between the horizontal response acceleration


(g) and the response displacement of the pier
top (mm), on the shallow foundation and the
sheet-pile foundations (L/B=1.0 and 0.5) at the
dense ground are shown in Fig 9. These
curves are found with sinusoidal input motion
tests. Yield acceleration y of the SPF with
L/B=1.0 is 0.8g, and this is twice as large as
that of the shallow foundation on the same
ground, of which y is 0.4g. y of the SPF with
L/B=0.5 is also improved as 0.6g. This means
with installing the sheet-piles in the ground,
seismic performance of the shallow foundation
can be largely improved, and the performance
is affected by the length of the sheet-piles.
Yield accelerations y are also compared in
Table 8, in the case of the medium dense

Table 4: Summary of test conditions


Case
Foundation type
Ground condition

1-1
Shallow
foundation

1-2

1-3

SPF

SPF

2-1
2-2
Shallow
SPF
foundation
Medium denseDr=60%,
Vs=180m/s
1.0B

Dense Dr=90% : Vs=200m/s

Sheet-pile length
Note : B is the footing width

1.0B

0.5B

Table 5: Summary of input motions for the shake table tests


Item
Sinusoidal motion
Earthuake motion

Summary
3 different frequencies (f=1.2, 2.4 and
3.6 Hz with 25 cycles)
A standard seismic motion, L2-spectral
1, prepared for the raylway structures
by RTRI

Purpose
Investigating the basic dynamic
characteristics of the SPF
Investigating the performance of the
SPF during the earthquake event

Table 6: Vertical self load distribution after the centrifuge acceleration process
Case
Foundation type
Ground
Sheet-pile
Vertical pressure
(kPa)
Vertical load
distribution
(Footing : SP)

1-1
SF
-

1-2
SPF
Dense
1.0B

1-3
SPF
0.5B

2-1
2-2
SF
SPF
Medium dense
1.0B
-

300

160

200

300

130

1 : 0.87

1 : 0.5

1 : 1.3

Note : SF is the shallow foundation; SP is the sheet-pile

Table 7: Predominant period of ground-structure system measured at the small shake event
Case

1-1

Foundation type
Ground
Sheet-pile
Predominant Period (s)

SF
0.66

1-2
SPF
Dense
1.0B
0.50

250

1-3

2-1

SPF

SF
SPF
Medium dense
1.0B
0.68
0.55

0.5B
0.51

2-2

MSP: Resistant moment against the rotation of


the footing by sheet-piles
Ms: In-plane bending moment by the side sheetpiles
NspC: Axial force of the sheet-pile at the front
side of the footing (Compression side)
NspT: Axial force of the sheet-pile at the back
side of the footing (Tension side)
MoutC: Bending moment of the front side sheetpile
MoutT: Bending moment of the back side sheetpile
B: Width of the footing
On the other hand, the rotation resistance
from the ground reaction under the footing
against the rocking motion is estimated by
assuming all vertical components can be

(240MPa) during the severe earthquake event.


This means the sheet-pile is not a critical
member on the seismic design of the SPF.
3) Bearing mechanism of the SPF
As it was seen, seismic performance of the
SPF is largely improved compared with the
shallow foundation. Therefore, the bearing
mechanism of the SPF against the seismic
action will be discussed in this section.
Bearing mechanism against the rocking
motion of the SPF is investigated utilizing the
force equilibrium around the footing defined as
eq.2, as illustrated in Fig 12.

MSP = Ms + (Nsp Nsp ) B / 2 + (Mout + Mout )


T

(2)

Earth Pres. (kPa)

1
Yield Acc. L/B=1.0
0.8
0.6 Yield Acc. of L/B=0.5
0.4
0.2
0
Yield Acc. of
-0.2
shallow foundation
-0.4
Shallow foundation
-0.6
SPF L/B=1.0
-0.8
SPF L/B=0.5
-1
-400
-200
0
200
400
Disp. (mm)

300
100
0
-100
0

50

Axial stress
Bending stress

50
0
-50
-100
0

2-2
SPF
1.0B
0.6

10

20
30
Time (sec)

40

Fig 11: Stress time histories of sheet-pile


mg

Mg

MoutC

NspT

40

100

Ms
MoutT

20
30
Time (sec)

Stress monitoring location

Table 11: Yield accelerations


(Medium dense ground)
2-1
SF
0.3

10

Fig 10: Earth pressure time histories

Fig 9: Load-displacement skelton curves of


tested foundations (Dense ground)

Case
Foundation type
Sheet-pile length
y (g)

Center
South 1/4

200

Stress (MPa)

Response Acc. (g)

Here,

NiS

NspC

Ng
b

NiT

Fig 12: Force equilibrium around the footing in


terms of rotation from the sheet-piles

251

NiC

Fig 13: Schematic drawing of the rotation


resistance from the ground reaction

50

supported by the vertical resistance of the front


side sheet-pile as its axial force, as shown in
Fig 13. Therefore, total vertical force from the
ground reaction Ng and the resistant moment by
the ground Mg can be calculated as follows.
(3)
N g = mg N i

assumption was suitable.


Since the calculation assumption of the
bearing mechanism is reasonable, share of
each reaction component against rotation can
be analyzed. According to the calculation
example of Fig 14, about 1/3 of the resistant
moment is shared by Mg (Ground reaction
component), and 2/3 is shared by MSP (Sheetpile reaction component).
Detailed analysis on the sheet-pile reaction
component is carried out as shown in Fig 15. At
the time step shown in the Fig 15(a), it is found
that 86% of the resistant moment is shared by
the axial resistance of the front/back sheet-piles,
and only 5% is shared by the bending
resistance of the sheet-piles. Share of each
component is illustrated in Fig 15(b).

Here,
mg: Mass of the superstructure x Centrifugal
acceleration
N i : Sum of the vertical force of sheet-piles
(4)

Here,
b: B/2
Therefore, total resistant moment against the
rotation around the footing is defined as eq.5.
M RSPF = M SP + M g
(5)

FULL-SCALE FIELD TEST

a) Outline of full-scale models


Aiming at practical use, the full-scale field
tests were conducted in Kawagoe-City, Japan.
The test setup and the models are shown in
Photo 3. The surface diluvial clay (Kanto loam),
with a thickness of 5m, is laid on the gravel
layer. These models have a 3.6m square
footing, 6m high pier. The sheet-piles length is
3.6m, the same as the width of the footing.
Therefore, the tips of the sheet-piles were not
installed into the gravel layer.

Rotation moment (kNm)

Case 1-2:L/B=1.0
Sin 1.2Hz, 0.4g

M
2
1

M (N)
out
M (S)

0.5
0.0

out

T=4.026sec
4.01

4.02
4.03
Time (sec)

4.04

4.05

Fig 15(a): Time histories of each resistant


moment component of MSP

SP

M +M
SP

1.0

Case 1-2:L/B=1.0
Sin 1.2Hz, 0.4g

-0.5
4.00

4
3

1.5

MoutT+ MoutC

Rotation moment (kNm)

2.0

* B/2

Resistant moment is calculated by eq.2


utilizing the forces measured by strain gauges
on the sheet-piles. The resistant moment by the
ground Mg is calculated at limited time section,
as illustrated in Fig 14. Calculation results show
there is good matching between MRSPF and MA.
Therefore, it is confirmed that the calculation

MSP

Time histories of the action moment


calculated from the inertia force of the structure
and the resistant moment defined as eq.5 are
shown in Fig 14. Action moment around the
footing center MA is calculated as follows.
M A = miili
(6)
Here,
mi: mass of the superstructure, pier and footing
i: Measured response acceleration
li: Arm from the center of the footing

(NSPT-NSPC)

M g = Ng b

Axial resistance by the front/back sheet-piles

Bending resistance
resistance
66of the front/back sheet-piles

-1

Rotation resistance by the side sheet-piles

-2

-4
3.7

0.09 0.05

Resistant moment from the ground Mg

-3

0.86

is calculated throughout this section only.


3.8

3.9

4.0
4.1
Time (sec)

4.2

4.3
0.0

0.2

0.4

0.6

0.8

1.0

Fig 15(b): Share of each resistant moment


component of MSP

Fig 14: Time histories of the action moment and


the resistant moment

252

condition.
3 Sheet-pile is not a critical member on the
seismic design of the SPF.

b) Horizontal static loading test


The horizontal static loading tests were
conducted by pulling the tops of two models to
each other by a hydraulic jack. At first, the
shallow foundation was pulled by the sheet-pile
foundation. Next the sheet-pile foundation was
pulled by the shallow foundation reinforced by
the ground anchor.
The P- relation of each case is shown in
Fig 16, together with the photos of the last
deformation of each model. Fig 16 shows that
the ratio of the horizontal resistance of the
sheet-pile foundation against that of the shallow
foundation was about four. This ratio is larger
than that obtained by the laboratory test.

Based on the above test results, a guideline


of design and construction of the sheet-pile
foundation was published by the Railway
Technical Research Institute, Japan, and
several actual projects were realized so far.
Because of its excellent performance and cost
competitiveness, application of the sheet-pile
foundation may increase in future.
REFERENCES
Koda, M., et al. (2003) The Proposal of Sheet-Pile
Foundation Combining a Footing with SheetPiles (in Japanese), TSU-CHI-TO-KISO, Vol. 51,
No. 11, pp.8-10
Nishioka, H., et al. (2004) A Series of Static Loading
Tests of Modeled Sheet-Pile Foundation
Combining Footing with Sheet-Piles on Sand,
Proceeding
of
15th
Southeast
Asian
Geotechnical Society Conference, pp.199-204,
Bangkok, Thailand.

CONCLUSION

Followings were found from a series of tests.


1 The seismic performances of the sheet-pile
foundation are greatly improved compared with
those of the shallow foundation.
2 Performance of the sheet-pile foundation is
influenced by both sheet-pile length and ground
Hydraulic Jack

Shallow
foundation

Sheet-Pile
foundation

Photo 3: Test setup and full-scale models

Horizontal load P (kN)

1000
800
600
Shallow Foundation
400

Sheet-Pile Foundation

200
0
0

100

200

300

400

500

Horizontal displacement (mm)

Fig 16: P- relationship and pictures of the models after loading

253

Rocking Seismic Isolation of Bridges Supported by Spread Foundations


K. Kawashima1, T. Nagai2, D. Sakellaraki2
1

Tokyo Institute of Technology, Japan


Former Graduate Student, Tokyo Institute of Technology, Japan

Abstract
This paper presents an analysis on the effectiveness of rocking seismic isolation of bridges
supported by spread foundations. Separations of footing from and contacts of footing with the
underlying ground which occur during an extreme ground motion result in mitigation of bridge
response. Separations and contacts of footing must have occurred in past earthquakes although
their effect was not rigorously included in seismic design. The effect of rocking seismic isolation is
presented for a 10 m tall standard bridge supported by spread foundations under three directional
excitation. It is shown that the effect of rocking seismic isolation is significant in reducing the plastic
deformation of columns at the plastic hinge regions although this increases deck and columns
response displacement.

INTRODUCTION

Direct or spread foundations are widely used


to support bridges where soil condition is stable.
They are designed so that performance for
sliding, settlement and overturning is assured.
Since overturning is generally critical in spread
foundations, it is important to clarify the safety
for overturning. Rocking response of spread
foundations provides a unique structural
response of bridges. Although spread
foundations have been conservatively designed
so that uplift from the underlying ground can be
minimum for preventing overturning, it is often
observed
from
the
post
earthquake
investigation that cracks along footings
occurred on the ground surface. This obviously
shows that rocking response occurred in spread
foundations during past earthquakes. Positive
use of rocking response of spread foundations
results in the isolation effect on bridges
response. The rocking response of spread
foundations increases as size of footings
decreases however it could result in overturning
of bridges under an extreme ground motion.
Consequently careful evaluation on size of
spread foundation is required.

254

Rocking response of structures has been


investigated by many researchers (Housner
1963, Ishiyama 1982, Kawashima et al. 1989,
Priestley et al. 1996). Housner studied rocking
response of a rigid block on a rigid base
(Housner 1963). Inelastic rocking response of
large rigid foundations was investigated by
Kawashima et al., and it was found that rigid
foundations did not overturn even if they were
subjected to ground motions with peak
accelerations
much
larger
than
static
accelerations in the static analysis (Kawashima
et al. 1989, 1991, 1994). This fact was taken
into account in the seismic design of HonshuShikoku Bridges, including the world longest
Akashi Straight Bridge and Kurushima Straight
Bridge.
More recently, Ciampoli et al. presented an
importance of rocking response and bridge
response interaction (Ciampoli et al., 1995).
Priestley et al. presented a contribution of
rocking response of a footing to a total deck
displacement
(Priestley
et
al.
1996).
Kawashima and Hosoiri (2003) and Kawashima
et al (2005) showed seismic rocking isolation
effect on standard bridges. Mergos and
Kawashima (2005) showed three directional
excitation effect on seismic rocking isolation of
a bridge (Mergos and Kawashima, 2005).

bottom of footing; FL and FLa : demand and


capacity of sliding force, respectively; e and ea :
eccentricity resulting from lateral force and its
allowable value, respectively; M B and V :
moment resulted from lateral force and vertical
force resulted from dead weight of deck,
column,
footing
and
overburden
soil,
respectively; and A : footing base area. It is
noted that ea is generally one third of
foundation width.
However, as rocking response increases,
footing starts to uplift and separate from the
ground at an edge as shown in Fig 1. Static
settlement of a spread foundation, v FS , with a
width l resulted from a static dead load of deck,
column and footing, V , may be written as

Sakellaraki and Kawashima verified seismic


rocking isolation effect based on shake table
test (Sakellaraki and Kawashima, 2006).
This paper shows the rocking isolation effect
of a bridge supported by spread foundations,
and clarifies the effect of bilateral excitation on
seismic rocking isolation.
ROCKING OF SPREAD FOUNDATION

In static seismic design, spread foundations


are sized so that the seismic performance can
be assured for based on bearing capacity of the
ground, sliding and overturning of foundation as

SB =

a A

< S Ba ; S S =

e=

MB
< ea
V

FL
< S Sa ;
FLa

v FS =

V
k sv l

(2)

When foundation uplifts with an angle F


from the resting position, an edge of
foundation starts to separate from the
underlying ground as upward displacement of
footing at edge, l FS / 2 , becomes larger
than initial settlement v FS~. Rotational spring
stiffness of foundation, K F , may then be
written as

(1)

in which, S B and S S , S Ba and S Sa : safety


factors for bearing capacity and sliding,
respectively, and their allowable safety factors;
a : vertical stress capacity of the ground at the

l F
l/2
2
< v FS
l / 2 k sv ( x) x dx
~
2
K F =
l FS
Xl / 2 k sv ( x) x 2 dx
v FS
2

(a)

(3)

in which X represents distance from the


center of footing to a point where footing uplifts
to the original level of the ground.
To
represent
rocking
soil-foundation
interaction, an analytical model as shown in Fig
2 (a) is widely used for a bridge supported by
spread foundations. Rotational spring stiffness,
K F , idealizes the rocking soil-foundation
interaction. Assuming that subgrade reaction of
soils per unit area, k sv , is linear at the entire
response displacement range of footing,
rotational stiffness of foundation K F may be
obtained as

(b)

K F = l /l 2/ 2 k sv ( x) x 2 dx

(4)

in which x is distance from the center of


footing.
Because the idealization by Fig 2 (a) with a
linear rotational stiffness K F only represents
linear response of a spread foundation, an

(c)
Fig 1: Uplift of spread foundation from the
underlying ground; (a) Static equilibrium, (b)
Uplift at left edge and (c) Uplifted equilibrium

255

separation as shown in Fig 3. Consequently,


the restoring force of the i-th spring is written as
k (v Fi v FSi )
fVi = SVi
0

v Fi < v FSi
(5)
v Fi v FSi

in which k SVi : stiffness of the i-th soil spring,


vFSi : initial settlement of footing due to dead
weight of bridge, and v Fi : relative
displacement at the i-th soil spring between the
footing and the ground which is defined as

v Fi = v Fi vGi

(a)

(6)

where v Fi and vGi are vertical response


displacement of footing and the ground,
respectively, at the i-th soil spring.
Although it is not shown in this paper, the
underlying ground may yield due to rocking of a
spread foundation when the underlying soil
does not have sufficient strength. The
saturation of the restoring force of the
underlying ground can be incorporated in the
idealization by Fig 3 (Kawashima et al 1991,
Kawashima and Hosoiri 2003).
It is known that sliding and uplift of a rocking
body develops a complex interaction between
sliding, rocking and jumping (Ishiyama 1982).
Sliding is however restrained in the following
analysis because sliding is not critical in spread
foundations embedded in the ground. The
analytical model shown in Fig 2 (b) can be
easily extended to three-dimensional model.

(b)

Fig 2: Idealization of spread foundation; (a)


without separation, and (b) with separation of the
footing from underlying ground

Fig 3: Idealization of restoring force of underlying


ground by soil spring

analytical model as shown in Fig 2 (b) is used


here to take account of uplift and separation of
a spread foundation from the underlying ground.
It is assumed in the model that the i-th soil
spring has nonlinear restring force with a
stiffness of k SVi in compression and 0 in

TARGET BRIDGE
A bridge which is analyzed here is presented
in Fig 4. It is a 200m long five span continuous
bridge supported by two abutments and four
reinforced concrete columns. Abutments and

Fig 4: Target bridge

256

design code (Japan Road Association 2002).


Since the bridge and the soil condition are
almost uniform along the bridge axis, a
structural system consisting of a column, a
foundation and a tributary deck is analyzed
here. The seismic response in the longitudinal,
transverse and vertical directions is analyzed.
The spread foundation determined from static
seismic
design
assuming
0.2g
lateral
acceleration is 2 m thick, 6.5 m long in the
longitudinal direction and 7 m wide in the
transverse direction. Overturning is the most
critical requirement in sizing foundations.
A foundation, a column and a part of deck are
idealized by a three-dimensional discrete
analytical model as shown in Fig 5. The footing
was idealized by a rigid grid consisting of beam
elements. Four sides of footing in the
longitudinal and transverse directions are
referred hereinafter A- and B-sides and C- and
D-sides, respectively (refer to Fig 5). The soilfoundation interaction was idealized by
nonlinear soil springs as shown in Fig 3. Soil
spring stiffness was determined from subgrade
reaction of soils based on the design practice
(Japan Road Association 2002). Column at the
plastic hinge was idealized by 3D fiber elements,

Fig 5:
Idealization of spread foundationunderlying ground-column-deck

Acceleration (m/s2)

columns rest on six spread foundations.


Superstructure is consisting of a plate girder
deck supported by five fixed steel bearings.
Spread foundations are embedded in sandy
soils with the bottom of foundations resting on
gravels with N-value of the standard penetration
test over 50. Since the fundamental natural
period of the ground is less than 0.2 s, this site
is designated as a stiff site (ground group I)
according to the Japanese highway bridge
9.8
0
-9.8

Acceleration (m/s )

(a)
9.8
0
-9.8

Acceleration (m/s )

(b)
9.8
0
-9.8
0

10

15
Time (s)

20

25

30

(c)
Fig 6: Ground motion for analysis; (a) NS, (b) EW and (c) UD components of JMA Kobe Observatory
record during the 1995 Kobe, Japan earthquake

257

Acceleration (m/s )

20
0
-20

Acceleration (m/s )

(a)
20
0
-20

Displacement (m)

(b)
0.3
0
-0.3

Displacement (m)

(c)
0.1
0
-0.1

10

15
Time (s)

20

25

30

(d)
Fig 7: Response accelerations and displacements at the deck under unilateral excitation when uplift of the
footing is not taken into account; (a) longitudinal acceleration, (b) vertical acceleration, (c) longitudinal
displacement, and (d) vertical displacement

Displacement (m)

0.15
A-side
B-side
0

-0.15

(a)
Restoring Force (kN)

1500
A-side
B-side
0

-1500
0

10

15
Time (s)

20

25

30

(b)
Fig 8: Response of the footing at A and B sides under unilateral excitation when uplift of the footing is not
taken into account; (a) vertical displacement of the footing, and (b) reaction restoring force of the soil spring

258

10000

500

500

5000

0
-500

-1000
-0.05
0
0.05
0.1
Relative vertical displacement (m)

Moment (kNm)

1000
Restoring force (kN)

Restoring force (kN)

1000

-5000

-500

-1000
-0.05
0
0.05
0.1
Relative vertical displacement (m)

(a)

-10000
-0.01

0
Curvature (1/m)

0.01

(b)

Fig 9: Restoring force of a soil spring under unilateral excitation


when uplift of the footing is not taken into account; (a) A-side,
and (b) B-side

and linear beam elements with the cracked


stiffness elsewhere. Flexure strength and
ductility capacity of the column were determined
based on an empirical stress and strain relation
of confined concrete (Hoshikuma et al 1997,
Sakai et al. 2000) and reinforcing bars
(Menegotto et al. 1973). Foundation and deck
were assumed rigid and their masses were
lumped at their gravity centers.
Because effect of overburden soil is
important in analysis, it was assumed that
weight of the overburden soil was included in
the evaluation of static settlement v FS in terms
of V in (1) and (2), but it was disregarded in
evaluation of the inertia force.
Three-dimensional bridge response under
bilateral and vertical excitation was computed.
As shown in Fig 6, NS, EW and vertical
components of ground accelerations measured
at Kobe Observatory of Japan Meteorological
Agency (JMA Kobe record) during the 1995
Kobe, Japan earthquake was imposed to the
bridge in the longitudinal, transverse and
vertical directions, respectively. Response
under unilateral and vertical excitation was also
computed for clarifying effect of unilateral
excitation. NS and UD components were
imposed to the bridge in the longitudinal and
vertical directions, respectively, in this
evaluation. Rayleigh damping was assumed to
represent energy dissipation (Clough and
Penzien 1993). Damping ratio was assumed
0.05 for the first and second modes.

Fig 10: Moment vs. curvature


hysteresis of column at the
Plastic hinge

For clarifying effect of three directional


excitation, response of the bridge under
unilateral and vertical excitation was first
evaluated. Fig 7 shows deck accelerations and
displacements when the bridge was subjected
to NS and US components in the longitudinal
and vertical directions, respectively. Soilfoundation interaction was idealized by a linear
rotational soil spring (refer to Fig 2(a)) without
taking uplift and separation of the footing from
the underlying ground into account in analysis.
The peak deck acceleration and displacement
in the longitudinal direction are 13.1 m/s2 and
0.219 m, respectively.
Fig 8 shows relative vertical displacement of
footing vFi and restoring force of a soil spring
fVi at both sides in the longitudinal direction (Aand B-sides, refer to Fig 5). The peak vertical
displacement of footing resulted from rocking
response was 15.2 mm and 22.9 mm at A-and
B-sides, respectively. Fig 9 shows restoring
force fVi vs. relative vertical displacement v Fi
hysteresis of the same soil springs. The peak
compression and tension of a soil spring is 903
kN and 472 kN, respectively, at A-side. They
result in compression and tension stress of 2.54
MPa and 1.33MPa, respectively, in the
underlying ground at A-side.
As a consequence, column shows moment
vs. curvature hysteresis at the plastic hinge in
the longitudinal direction as shown in Fig 10.
Column undergoes significant plastic range with
the peak curvature of 9.4x10-3/m.
On the other hand, Fig 11 shows deck
accelerations and displacements of the bridge

SEISMIC RESPONSE UNDER UNILATERAL


AND VERTICAL EXCITATION

259

Acceleration (m/s )

15
0
-15

Acceleration (m/s )

(a)
20
0
-20

Displacement (m)

(b)
0.3
0
-0.3

Displacement (m)

(c)
0.1
0
-0.1

(d)
Fig 11: Response accelerations and displacements at the top of column under unilateral excitation
when uplift of the footing is taken into account; (a) longitudinal acceleration, (b) vertical
acceleration, (c) longitudinal displacement, and (d) vertical displacement

Displacement (m)

0.15
A-side
B-side
0

-0.15

(a)
Restoring force (kN)

1500
A-side
B-side
0

-1500
0

10

15
Time (s)

20

25

30

(b)
Fig 12: Response of the footing at A and B sides under unilateral excitation when uplift of the footing
is taken into account; (a) vertical displacement of the footing, and (b) reaction restoring force of the
soil spring

260

750
0
-750

-1500
-0.05
0
0.05
0.1
Relative vertical displacement (m)

(a)

1500

4000

750

2000
Moment (kNm)

Restoring force (kN)

Restoring force (kN)

1500

0
-750

-2000

-1500
-0.05
0
0.05
0.1
Relative vertical displacement (m)

-4000
-0.001

0
Curvature (1/m)

0.001

(b)

Fig 13: Restoring force of a soil spring under unilateral excitation


when uplift of the footing is taken into account; (a) A-side, and
(b) B-side

Fig 14: Moment vs. curvature


hysteresis of column at the plastic
hinge

springs at A- and B-sides. The peak


compression of a soil spring is 725 kN at A-side,
which corresponds to compression stress of
2.04 MPa. No tension stress is induced in the
soil springs. The compression stress of 2.04
MPa is 80 % of the stress in the underlying
ground by disregarding uplift and separation of
footing from the underlying ground.
Fig 14 shows moment vs. curvature
hysteresis of column at the plastic hinge in the
longitudinal direction. The plastic deformation of
column is limited and the peak curvature is
7.63x10-4/m. This is only 8% the curvature
computed by disregarding uplift and separation
of footing from the underlying ground. Rocking
seismic isolation effect is thus significant for
mitigating damage of column at the plastic
hinge.

computed by idealizing nonlinear soilfoundation interaction by nonlinear soil springs


(refer to Fig 2 (b)) taking uplift and separation of
footing from the underlying ground into account.
The bridge was subjected to NS and US
components of JMA Kobe ground motion in the
longitudinal and vertical directions, respectively.
The peak deck acceleration and displacement
in the longitudinal direction are 6.69 m/s2 and
0.253 m, respectively, which are 49% smaller
and 16% larger, respectively, than those
computed by disregarding uplift and separation
of footing from the underlying grounds. It is
obvious that the significant decrease of the
peak deck acceleration results from rocking
seismic isolation. It should be assured that the
slight increase of deck displacement does not
result in any problem for the seismic
performance of bridge.
Fig 12 shows vertical relative displacement
of footing vFi and restoring force of a soil
spring fVi at A- and B-sides in the longitudinal
direction. Footing rocked and uplifted 110 mm
and 56 mm at A- and B-sides, respectively,
from an initial settlement due to dead weight
vFS (refer to (2)) of 2.48 mm. It is important to
note that spread foundation which was
designed in accordance with static design
assuming 0.2 g lateral acceleration by (1) uplift
56-110 mm under a near-field ground motion
recorded during the 1995 Kobe earthquake.
This means that rocking isolation must have
occurred in past significant earthquakes
although this effect was not considered in
design.
Fig 13 shows restoring force fVi vs. relative
vertical displacement vFi hysteresis of soil

SEISMIC RESPONSE UNDER THREE


DIRECTIONAL EXCITATION
Fig 15 shows deck responses under three
directional excitation evaluated by taking uplift
and separation of footing from the underlying
ground into account. The peak deck
acceleration and displacement are 7.16 m/s2
and 0.263, respectively, in the longitudinal
direction and 7.50 m/s2 and 0.194 m,
respectively, in the transverse direction.
Because the peak deck acceleration and
displacement in the longitudinal direction are
6.69 m/s2 and 0.253 m, respectively, under
unilateral and vertical excitation, the peak deck
acceleration and displacement increase by 7%
and 4%, respectively, under three directional
excitation. Fig 16 shows relative vertical
displacement of footing v Fi and restoring

261

Displacement (m)

0.3
0
-0.3

Displacement (m)

(a)
0.3
0
-0.3

Displacement (m)

(b)
0.1
0
-0.1
0

10

15
Time (s)

20

25

30

Displacement (m)

(c)
Fig 15: Response displacements at deck under three directional excitation when uplift of the footing is taken
into account; (a) longitudinal displacement, (b) transverse displacement, and (c) vertical displacement
0.15
A-side
B-side

-0.15

Displacement (m)

(a)
0.15

C-side
D-side

-0.15

Restoring force (kN)

(b)
1000
A-side
B-side

0
-1000

Restoring force (kN)

(c)
1000
C-side
D-side

0
-1000 0

10

15
Time (s)

20

25

30

(d)
Fig 16: Response of footing under three directional excitation when uplift of footing is taken into account; (a)
vertical footing displacement at A and B-sides, (b) vertical footing displacement at C and D-sides, (c) soil
spring reaction restoring force at A and B-sides, and (d) soil spring restoring force at C and D-sides

262

1000
Restoring force (kN)

Restoring force (kN)

1000
500
0
-500
-1000
-1500
-0.1

0.1

500
0
-500
-1000
-1500
-0.1
0
0.1
0.2
Relative vertical displacement (m)

0.2

Relative vertical displacement (m)

(a)

(b)

5000

5000

2500

2500

Moment (kNm)

Moment (kNm)

Fig 17: Restoring force of a soil spring under three directional excitation when
uplift of the footing is taken into account; (a) AC-corner, and (b) BC-corner

0
-2500
-5000
-0.001

0
Curvature (1/m)

0
-2500
-5000
-0.001

0.001

(a) Longitudinal direction

0
Curvature (1/m)

0.001

(b) Transverse direction

Fig 18: Moment vs. curvature hysteresis of column at the plastic hinge
under three directional excitation when uplift of the footing is taken into account

force of a soil spring fVi at A- and B-sides in


the longitudinal direction and C- and D-sides in
the transverse direction. The peak relative uplift
v Fi is 127 mm and 79 mm at A- and B-sides,
respectively. Because v Fi at A- and B-sides
under unilateral and vertical excitation is 110
mm and 56 mm, respectively, v Fi under three
directional excitation is 15% and 41% larger.
Fig. 17 shows the restoring force fVi vs.
relative vertical displacement v Fi hysteresis of
the soil springs at the corner of A- and C-sides
(AC corner) and the corner of B- and C-sides
(BC corner). The peak soil compression is 647
KN and 1107kN at the AC and BC corners,
respectively, which results in the peak soil
stress of 1.8 MPa and 6.23 MPa, respectively.
Consequently, the compression stress induced
in the underlying soil at the corners under three
directional excitation is nearly 1.8-2.2 times the
soil stress under unilateral and vertical
excitation. Protection of soils at the corners is

needed depending on bearing capacity of the


ground.
Fig 18 shows moment vs. curvature
hysteresis of column at the plastic hinge in the
longitudinal and transverse directions. The peak
curvature is 8.24x10-4/m and 2.0x10-4/m in the
longitudinal
and
transverse
directions,
respectively. Since the strength of column is
much higher in the transverse direction than the
longitudinal direction, the curvature of column in
the transverse direction is limited. Because the
peak curvature in the longitudinal direction is
7.63x10-4/m under unilateral and vertical
excitation, it increases by 7% under three
directional excitation.
CONCLUSIONS
Effectiveness of the rocking seismic isolation
of spread foundations on the seismic

263

Clough R.W. and Penzien J. (1993) Dynamics of


Structures, 2nd edn, McGraw Hill, New York.
Hoshikuma, J. Kawashima, K. Nagaya, K. and A. W.
Taylor (1997) Stress-Strain Model for Confined
Reinforced Concrete in Bridge Piers, Journal of
Structural Engineering, ASCE, 123(5), pp. 624633.
Housner, G. W. (1963) The Behaviour of Inverted
Pendulum Structures during Earthquakes,
Bulletin Seismological Society of America, 53,
pp. 404-417.
Ishiyama, Y. (1982) Motion of Rigid Bodies and
Criteria for Overturning by Earthquake
Excitations, Earthquake Engineering and
Structural Dynamics, 10, pp. 635-650.
Japan
Road
Association
(2002)
Design
Specifications of Highway Bridges, Maruzen,
Tokyo.
Kawashima, K. and Unjoh, S. (1989) Rocking
Response of a Rigid Foundation subjected to
Seismic Excitation, Civil Engineering Journal,
32 (10), pp. 60-66 (in Japanese).
Kawashima, K. and Unjoh, S (1991) Overturning of
Rigid Foundation Resting on Ground with
Insufficient Yield Strength, Civil Engineering
Journal, 33(3), pp. 54-59 (in Japanese).
Kawashima, K., Unjoh, S. and Mukai, H. (1994)
Inelastic Rocking of Direct Foundation during
an Earthquake, Civil Engineering Journal, 36(7),
pp. 50-55 (in Japanese).
Kawashima, K. and Hosoiri, K. (2003) Rocking
Response of Bridge Columns on Direct
Foundations, Proc. fib-Symposium, Concrete
Structures in Seismic Region, Paper No. 118
(CD-ROM), Athens, Greece.
Kawashima, K., Watanabe, G., Sakeraraki, D. and
Nagai, T. (2005) Rocking Isolation of Bridge
Foundations, 9th World Seminar on Seismic
Isolation, Energy Dissipation and Active
Vibration Control of Structures, pp. 609-630,
Kobe, Japan
Mergos, P.E. and Kawashima, K. (2005) Rocking
Isolation of a Typical Bridge Pier on Spread
Foundation, Journal of Earthquake Engineering,
9(2), pp. 395-414.
Menegotto, M.and Pinto, P.E. (1973) Method of
Analysis for Cyclically Loaded R.C. Plane
Frames including Changes in Geometry and
Non-Elastic Behavior of Elements under
Combined Normal Force and Bending, Proc.
IABSE Symposium on Resistance and Ultimate
Deformability of Structures Acted on by Well
Defined Repeated Loads, pp. 15-22.
Priestley, N. M. J., Seible, F. and Calvi, G. M. (1996)
Seismic Design and Retrofit of Bridges, John
Wiley & Sons, New York.
Sakai, J. and Kawashima, K. (2006) Unloading and
Reloading Stress-Strain Model for Confined
Concrete, Journal of Structural Engineering,
ASCE, 132(1), pp. 112-122.

performance of bridges was studied based on


nonlinear analysis for a 10 m tall standard
bridge. Analytical model which takes account of
uplift and separation of footing from the
underlying ground was verified by a shake table
test. Based on the results presented herein, the
following conclusions may be deduced on the
effect of rocking seismic isolation:
1) If separation of footing from the underlying
ground due to rocking response occurs, the
plastic deformation of column at the plastic
hinge significantly decreases as a result of
softening of moment vs. rotation hysteresis of
footing. As a consequence, inelastic rocking of
footing results in an isolation effect on the
response of bridge. However because the
isolation effect results in an increase of
response displacement of bridge, this must be
properly considered in design.
2) Spread foundation designed in accordance
with static seismic analysis assuming 0.2g
lateral static acceleration and the working stress
design approach may rock separating from the
underlying ground under near-field ground
motions. Uplift at the edge of footing is 80-130
mm in a 10 m tall standard bridge under JMA
Kobe ground motion in the 1995 Kobe, Japan
earthquake. This means that rocking seismic
isolation must have occurred in past significant
earthquakes although this effect was not
considered in design. Positive use of the
rocking seismic isolation brings benefit in the
seismic design of bridges.
3) Uplift and separation of footing from the
underlying ground can be analyzed using
nonlinear soil spring idealization as shown in
Figs 2 (b) and 3.
4) Three directional excitation results in an
increase of bridge response acceleration and
displacement. In particular, stress induced in
the underlying ground at the corners
significantly increases under three directional
excitation. Underlying ground at the corners
needs to be protected, if necessary, for use of
rocking seismic isolation.
REFERENCES
Ciampoli, M. and Pinto, P. E. (1995) Effects of SoilStructure Interaction on Inelastic Seismic
Response of Bridge Piers, Journal of Structural
Engineering, ASCE, 121 (5): 806-814.

264

Sakellaraki, D. and Kawashima, K. (2006)


Effectiveness of Seismic Rocking Isolation of
Bridges based on Shake Table Test, 1st

European
Conference
on
Earthquake
Engineering and Seismology, Paper No. 364, pp.
1-10, Geneva, Switzerland

265

266

Deep Foundations

Effects of Limit State Performance of Steel Bearings on a Bridge Upper


Structure
K. Ohtomo1, Y. Sato1, M. Sakai1
1

Central Research Institute of Electric Power Industry, Japan

Abstract
The present study discusses the effect of bearing support performance on a bridge
superstructure during a strong ground motion. A hybrid seismic response experiment for a
steel bearing is conducted for this purpose. Slip-type hysteresis is identified for bridge axis
and transverse direction excitations arising from contact between pin and other members in
the shoe. The degree of superstructure response is also affected by such hysteresis,
particularly in bridge axis direction.

INTRODUCTION

42000
M

42000
M

16435

MMovable Shoe

16665

The effects of seismic limit state


performance of bearings on a whole bridge
system during strong ground motions have
been greatly concerned from past earthquake
damages. Although rubber bearings or seismic
isolation devices have replaced steel bearings
since the 1995 Hyogo-ken Nanbu earthquake,
they are still in use in existing bridges. This
indicates that the effects of steel bearing failure
on earthquake resistance of a bridge are a
crucial issue.
There are several physical models (Otsuka
et al, 1997; Yabe et al, 1997; Iemura et al,
1998) for steel bearing required in developing
dynamic seismic response analysis model of a
bridge system. However, a systematic nonlinear
model for steel bearing performance is yet to be
established and geometric behavior modeling
should be also discussed as well as material
strength in developing a nonlinear model.
The objective of the present study is then to
experimentally discuss the effect of near limit
state in regard to fixed steel bearing on seismic
performance of a total bridge system. For this
purpose, a hybrid seismic experiment is
performed to examine earthquake response
nature of a total bridge system associated with
nonlinear behavior of steel bearings.

75000

Fixed Shoe

unitmm

Fig 1: Target bridge


Table 1: Properties of superstructure of the target
bridge
Mass per Unit Lengtht/m
2

EIkNm

19.0
7

Vertical

1.9510

Transverse

1.9510

GJkNm

6
5

7.73910

TARGET BRIDGE AND ITS MODEL

Fig. 1 shows a target bridge consisting of


160m long three-span continuous decks
supported by four rectangular reinforced
concrete piers with one fixed bearing and three
movable bearing conditions. The superstructure
is assumed to have 3,000t in mass. As far as a

267

Table 2: Properties of superstructure of reinforced concrete column


1st Point

2nd Point

M1(kNm) 1(1/m) M2(kNm)


Fixed Shoe Column
Movable Shoe Column

-5

-5

Bridge Axis 1.5210 3.9010


Transverse 1.1510 3.9010

3rd Ponit

2(1/m)

1.3410 6.5510
1.0110 6.5510

M3(kNm) 3(1/m)
5

-3

-3

1.7210 3.8110
1.3010 3.8110

Bridge Axis,
4
-5
4
4
5
-3
1.0010 3.7010
Transverse 1.2710 4.9010 7.0610 7.5610

Table 3: Restraining of relative displacement between superstructure and substructure


Shoe Condition

Bridge Axis

Transverse

Vertical

Around
Bridge Axis

Around
Transverse

Around
Vertical

Fixed Support

Restrained

Restrained

Restrained

Restrained

Free

Free

Movable Support

Free

Restrained

Restrained

Restrained

Free

Free

HYBRID SEISMIC RESPONSE EXPERIMENT

bearing seismic performance is concerned,


major bridge damage due to poor bearing
performance
during
an
earthquake
is
predominant on three continuous girder bridges
among
other
bridges
(JSCE,
1996).
Furthermore,
such
bridge
structure
is
susceptible to sustain localized damage on the
piers that support fixed steel bearings. Then,
the three continuous girder bridge well satisfies
the present study purpose.
Three dimensional finite element seismic
response analysis (hereafter, 3DFEM analysis)
models of the target bridge were constructed
respectively for bridge axis and transverse to
bridge axis directions to facilitate numerical
analysis and experimental works in the present
study The superstructure properties (mass per
unit length, flexural rigidity EI and torsional
rigidity GJ) are tabulated in Table 1. Reinforced
concrete columns corresponding to both fixed
and movable bearings were idealized as a
beam element with a tri-linear moment M and
curvature relationship and their parameters
are listed in Table 2. Mass density of the
columns was set to be 2.4 t/m3.
The fixed and movable bearings were
modeled as just nodal in the 3DFEM analysis
and their freedoms are summarized in Table 3.
Although soil springs are often used to express
dynamic soil and foundation interaction in a
dynamic analysis, the current study discards
this option due to the limitation on degrees of
freedoms that can be handled in the hybrid
experiment control.

In fact, a hybrid experiment scheme


employed in the present study is a real-time
hybrid experiment that has an actuator
response delay compensation function using a
digital signal processor (DSP) instead of a
classic pseudo dynamic experiment procedure.
Hereafter, the term real-time is dropped unless
confusion arises.
Bearing
The present study concerns the earthquake
damage only on a bearing body. The pin-type
bearing specimen prepared in the hybrid
experiment is schematically illustrated in Fig. 2.
The pin is a steel cylinder whose middle part
diameter
is
intentionally
narrowed
for
assembling purpose with upper and lower
shoes. The metal material was SS400, which is
recommended in the Handbook of Bearing
Support for Highway Bridge (Japan Road
Association, 1991).
To design the pin profile, demand and
capacity balance on the transverse direction
excitation associated with the applied force F
was considered (Japan Road Association,
1991; Abe et al, 2004) as illustrated in Fig. 3.
The outer diameter D and the narrowed part
diameter d of the steel cylinder can be
determined by the tensile strength capacity of
the narrowed part and the shear strength
capacity at the shear key characterized by
thickness t in the shoe (Abe et al, 2001). Finally,
40mm and 20mm for the outer diameter and the

268

210

Pin
400

400

Pin

14.25

Transverse
Lower Shoe
Fig 2: Configuration and dimension of bearing specimen

Bridge axis

Acceleration(m/s )

8
4
0
-4
-8
0

112

Unitmm

8.25

14

Upper Shoe

10

105 105

4
25
25

400
20
40

Upper Shoe

400

10
Time(s)

15

20

Fig 4: Takatori ground motion (EW) component

Fig 3: Shear key structure of the pin

25
h0.05
2

Acceleration(m/s )

narrowed part diameters respectively were


determined so that the cylinder would sustain
breakout under the maximum load capacity of
the hydraulic actuator used in the hybrid
experiment, i.e., 500kN.
Numerical Model Development
An EW component of the observed ground
motion at Takatori-eki station of JR West during
the 1995 Hyogo-ken Nanbu earthquake
(Hereafter, Takatori record for simplicity) was
used for the dynamic analysis throughout this
study. In reality, first twenty seconds of duration
of Takatori record is used for analysis purpose.
The acceleration time history and the
acceleration response spectrum (h=5%; h
represents damping ratio) are presented in Fig.
4 and Fig. 5, respectively.
A nonlinear dynamic analysis was
implemented to assess earthquake response of
the target bridge. The equation of motion was
solved by numerical direct time integration
scheme, or Newmarks method ( =0.25) was
employed with time increment of 0.0007812s.
Rayleigh type damping was used to express
viscous damping constants involved in the
equation of motion. Modal damping ratio of 2%

20
15
10
5
0
0

2
Period(s)

Fig 5: Acceleration response spectrum

is assumed to be valid on the first and the third


natural periods for both bridge axis and
transverse directions.
The 3DFEM analysis model as constructed
previously for the target bridge was then
reduced to a two-degrees-of-freedom model
(hereafter, 2DOF model) for bridge axis as well
as transverse to bridge axis directions as
presented in Fig. 6. In order to develop the
2DOF model, the analytical correlation between
the 2DOF and the 3DFEM model analyses was
examined
focusing
on
the
maximum
displacement response at the column top under
Takatori record excitation.

269

Superstructure

Superstructure

Mass
Steel Bearing
Vertical Actuator (1000kN)

Stiffness
(Experiment)
Steel Bearing

Pier Mass
Horizontal Actuator (500kN)

Pier

Reaction Wall

Pier
Stiffness

Steel Bearing
Support Beam for Bearing

Strong Floor

Fig 6: 2DOF model for the hybrid experiment

Fig 7: Experiment setup (bridge axis direction)


Table 4: Dynamic properties of 2DOF model
Bridge Axis

Transverse

15.728

7.478

Superstructure
Mass(t)

To assess seismic response of the target


bridge, a similitude rule must be applied. The
similitude rule widely used among many hybrid
seismic response experiments for a bridge
structure is employed as presented in Table 5.
The scale factor was determined on the ground
that dimensions of the model bearing can be
scaled down from a so-called 300t class
bearing whose narrowed part diameter size is
equal to 60mm by referring one (Hanshin
Expressway Company Limited, 1989) of the
practical design examples. As a result, the
scale factor was determined to be 3.2. In fact,
five bearings are assumed to exist on the pier.

1.187

Pier Mass(t)
Pier Stiffness(kN/m)

1.4610

9.29104

Damping Ratio

0.2086

0.2825

Natural FrequencyHz

1.140

1.269

Natural Periods

0.8773

0.7879

Table 5: Similitude rule for the hybrid experiment


Quantity

Dimension

Scale Factor

Length

Mass

S3

Time

Velocity
Acceleration

-1

-2

1/S

LT

LT

Force

MLT-2

S2

Rigidity

MT-2

Experiment Setup
The schematic view of the hybrid
experiment setup is depicted in Fig. 7.
Photographs of transverse and bridge axis
directions setups are also shown Fig. 8 and Fig.
9, respectively. The horizontal and vertical
loads were applied to the bearing specimen by
500kN and 1,000kN actuators, respectively.
The 500kN actuator played a role on response
displacement loading from the computed
superstructure response, while the 1,000kN
actuator added the dead weight of the
superstructure.
The 500kN loading was applied to the
bearing specimen as a computed response
under Takatori record for bridge axis and
transverse directions. In other words, for an
example, the pin axis was placed perpendicular
to the 500kN actuator axis for representing a
bridge axis excitation.

The superstructure mass, the column mass


and the column stiffness were determined so
that the both displacement responses fall into
an acceptable tolerance. The dynamic
properties of the established 2DOF are
tabulated in Table 4. As far as the column
stiffness is discussed, values of equivalent
stiffness and damping ratio for the column
correspond to nonlinear load and displacement
relationship listed in Table 2 were identified
using an equivalent linear model approach.

270

Reaction Force (kN)

600
400

6.66m/s

8.00m/s

200
0
-200
-400
-600
-8

-4

0
4
Displacement (mm)

Fig 10: Hysteresis behavior of bearing (bridge axis)

Fig 8: Entire view of transverse direction setup

Upper Shoe

Pin

Lower Shoe

Fig 11: Translation movement around the pin

friction force between the upper and lower


shoes. The positive and negative reaction
forces correspond to the slip stays almost at
constant values regardless of the magnitude of
PGA. These absolute values are, however,
unequal between positive and negative
movement. This may depend on intact
inclination of the bearing seating.
The cause of relatively large stiffness as
observed in Fig. 10 can be explained as
follows: When the shoe rotates around the pin
axis, a slight translation of the pin occurs due to
existing gap as illustrated in Fig. 11. After the
translation reaches the pin-support surface
between upper and lower shoes, like a
mechanical key appears, resulting in essential
stiffness of bearing body.
Fig. 12 also presents load and displacement
hysteresis under 6.66m/s2 and 8.00m/s2 in PGA
excitations,
in
which
rotation-induced
displacement is extracted. Then, slip length is
obtained as about 0.5mm. In addition, the
difference between the pin diameter and the
pin-support diameter is 0.5mm, which indicates
that the maximum translation should be equal
to 0.5mm and meets the slip length as shown in
Fig. 12. This observation demonstrates the

Fig 9: Local view of axis direction setup


RESULTS AND DISCUSSIONS

Excitation Cases
Peak ground acceleration (PGA) value
equal to 6.66m/s2 in Takatori record was
adjusted like 1.11m/s2, 3.33m/s2, 6.66m/s2 and
8.00m/s2. Then, four sets of excitation cases
were applied for the bridge axis direction.
According to the similar manner with the case of
bridge axis direction, 1.11m/s2, 3.33m/s2,
6.66m/s2, 10.00m/s2 and 20.00m/s2 in PGA
were applied for the bridge transverse direction.
Hysteresis Characteristics of Bearing
Behavior
Fig. 10 shows the load and displacement
relationship under 6.66m/s2 and 8.00m/s2 in
PGA excitation cases for the bridge axis
direction. A slip-type hysteresis loop is clearly
observed in the respective case. The slip is
probably due to rotation of the bearing around
the pin axis, which simultaneously occurs when
the applied load exceed the maximum static

271

above-mentioned discussion.

Reaction Force (kN)

600
400

6.66m/s

8.00m/s

200
0
-200
-400

Fig 14: Breakout of the pin

-600
-2

-1

0
1
Displacement (mm)

Fig. 14 presents the pin breakout occurred


during the 20.00m/s2 excitation. One can
observe that the failure surface deviates 45
degrees from the pin axis. However, this seems
to be still brought by a tensile force. The tensile
strength of the narrowed part of the pin is
calculated about 160kN, this value virtually
becomes about 200kN because of additional
static friction force of 40kN. On the other hand,
the breakout load is identified as 247kN as
shown in Fig. 13, leading a twenty percent
increase to the above-mentioned value.

Fig 12: Hysteresis behavior of bearing using


corrected displacement (bridge axis)
300
Reaction Force (kN)

Estimated breakout load


200
2

6.66m/s
2
20.00m/s

100
0
-100

Estimated breakout load


-200

Effects on the Superstructure


Experimental results discussed here are
enlarged to a real size and correlated with the
2DOF analysis results. In this respect, the
2DOF model excludes the relative displacement
of the bearing to investigate clearly the effect of
the bearing performance.
The superstructure displacement time
histories for the bridge axis direction are plotted
in Fig. 15. The duration patterns are similar to
each other, especially peak response are
almost identical in the duration of 10s and more.
On the other hand, the magnitude of the peak
response in the experiment result is relatively
larger than those of in the analysis result. Thus,
the effect of the nonlinear hysteresis
characteristics of the bearing on the
superstructure response is observed. The
maximum displacement obtained by the
experiment is 18.4mm or about 14% larger than
that of the 2DOF analysis result. This suggests
that essential response of a bridge
superstructure needs to be estimated taking the
bearing hysteresis into account.
The slip-type hysteresis as shown in Fig. 10
undoubtedly contributes to such a response. As
a result, relatively larger peak responses occur

-300
-8

-6

-4
-2
Displacement (mm)

Fig 13: Hysteresis behavior of bearing (transverse)

Fig. 13 indicates load and displacement


relationship under 6.66m/s2 and 20.00m/s2 in
PGA excitations for the bridge transverse
direction, in which significant slip-type
hysteresis loops are also obtained. Note that
pin breakout at the narrowed part occurred
during the 20.00m/s2 excitation; subsequently
the hybrid experiment was terminated due to
the limiting values of displacement and force of
the 500kN actuator. The degree of the reaction
force at which the slip occurs is roughly
constant under respective excitation, indicating
that the pin moves along with the gap around
the shear key (See Fig. 3) of the bearing when
the applied force becomes larger than the static
friction force. The magnitude of the slip
movement is further intensified arising from the
enlargement of the narrowed part brought by
cyclic stress-induced plastic deformation.

272

reinforced concrete piers and a superstructure


are numerically modeled and assigned to the
numerical part and a steel bearing is dealt with
the experimental part. From the hybrid
experiment and additional numerical analysis
results, the following conclusions can be
addressed:
1) Load and displacement characteristics is
identified as a slip-type hysteresis for both
bridge axis and transverse to bridge axis
directions. In bridge axis direction, rotation of
the upper and lower shoes around the pin
contributed to develop such a specific loop. On
the other hand, the pin movement in the gap
distance around the shear key plays a role on
the slip-type hysteresis characteristic for
transverse direction. The magnitude of the slip
is further intensified under accumulated cyclic
loading.
2) The slip-type hysteresis indeed affected
the superstructure response. The maximum
horizontal displacement at the superstructure
for bridge axis direction and transverse direction
to bridge axis in the hybrid experiment are
larger than those of the 2DOF analysis that
uses a prescribed condition at the bearing to
represent the bearing behavior. Namely, 14%
and 4% increases for bridge axis direction and
transverse direction to bridge axis, respectively
are recognized. In particular, the 14% increase
may have a serious consequence on the design
of the gap allowance between adjacent decks.

in negative value in the reflection of negatively


biased hysteresis loop.
0.2
Displacement(m)

Experiment
2DOF

0.1
0
-0.1
-0.2

10
Time(s)

15

20

Fig 15: Superstructure displacement time histories


in bridge axis under 6.66m/s2 excitation
0.1
Displacement(m)

Experiment
2DOF

0.05
0
-0.05
-0.1

10
Time(s)

15

20

Fig 16: Superstructure displacement time histories


in bridge transverse under 6.66m/s2
excitation

ACKNOWLEDGEMENT

Fig. 16 compares the superstructure


displacement time histories for the bridge
transverse direction obtained by the hybrid
experiment and the 2DOF analysis. Peak
responses as well as duration characteristics
are
almost
identical.
The
maximum
displacement is 98.2mm and 94.2mm for the
experiment and the 2DOF analysis, respectively.
This comparison indicates that the experimental
maximum response is about 4% larger than that
of the 2DOF analysis and therefore the effect of
the
slip-induced
displacement
on
the
superstructure is insignificant.
COCLUDING REMARKS

The present study examined the effect of


the bearing performance associated with limit
state on a superstructure based on the hybrid
seismic response experiment in which

The present study was granted by NIED


(National Research Institute for Earth Science
and Disaster Prevention) in connection with FY
2005 US-Japan corroboration of experimental
studies on seismic performance of bridge
structures by utilizing Three-Dimension FullScale Earthquake Testing Facility (E-Defense)
program.
REFRENCES
Abe, M., Yanagino, K., Fujino, Y. and Hashimoto, S.
(2001) "Damage Analysis of Three-Span
Continuous Girder Bridges in 1995 Hyogo-ken
Nanbu Earthquake (in Japanese) ", Journal of
Structural
Mechanics
and
Earthquake
Engineering, No.668/I-54, pp.83-101
Abe, M., Yoshida, J., Fujino, Y., Morishige, Y, Uno S.

273

and Usami, S. (2004) "Experimental Investigation


of Ultimate Behavior of Metal Bridge Bearing
under Seismic Loading (in Japanese) ", Journal
of Structural Mechanics and Earthquake
Engineering, No.773/I-69, pp.63-78

JSCE Editorial Committee for the Report on the


Hanshin-Awaji Earthquake Disaster, (1996)
Report on the Hanshin-Awaji Earthquake
Disaster, Damage to Civil Engineering Structures,
Bridge Structures (in Japanese), pp.59-68

Hanshin Expressway Company Limited (1989)


Standard Design Drawings of Steel Bearing for
Steel Girder (in Japanese)

Otsuka, H., Kanda, M., Suzuki, N. and Kawakami, M.


(1997) "Dynamic Analysis Concerned with
Rotational Displacement of Skewed Bridges
Caused by Horizontal Ground Motion (in
Japanese) ", Journal of Structural Mechanics and
Earthquake Engineering, No.570/I-40, pp.315324

Iemura, H., Miyamoto, A. and Takahashi, Y. (1998)


"Influence of Failure of Steel Bearings on
Damage Modes of Bridges under Strong
Earthquake Motion (in Japanese) ", Journal of
Structural Engineering, Vol.44A, pp.659-666

Yabe, M., Takemura, H. and Kawashima, K. (1997)


"Effects of Impacts between Two Adjacent Decks
in a Straight Bridge and a Deck and Abutment in
a Skewed Bridge (in Japanese) ", Journal of
Structural Engineering, Vol.43A, pp.781-791

Japan Road Association, (1991) Handbook of


Bearing Support for Highway Bridge (in
Japanese)

274

Large-scale model tests of shallow foundations subjected to earthquake


loads
M. Shirato1, T. Kouno1, S. Nakatani1, R. Paolucci2
1

Public Works Research Institute, Japan


2
Politecnico di Milano, Italy

Abstract
We conducted a series of 1G large-scale shake table tests and cyclic eccentric loading tests
of a shallow foundation model. The experimental parameters were the difference in loading
methods (i.e. dynamic and static), input seismic motions (i.e. intensity and number of cycles), soil densities (i.e. dense and medium dense), and the ratio of horizontal and overturning moment loads. The result provided benchmark data sets for the development of numerical models for the response of a shallow foundation subjected to a large earthquake. The
experimental data set contained the acceleration and displacement of the soil and those of
the foundation model. The dataset also contained the distribution of reaction normal and
shear forces at the foundation base. The experimental results suggest that the coupling effect among vertical, horizontal, and overturning loads, the uplift of shallow foundation and a
sophisticated hardening rule should be considered to separately predict the evolution of the
residual displacement.

soil is caused by an increase in the eccentricity


of vertical loads especially when a shallow
foundation is rested on sturdy ground. It results
in a nonlinear relationship between a moment
and rotation of footing. The Specifications for
Highway Bridges has considered this kind of
force reduction in the structural design of shallow foundation footings for large earthquakes. A
nonlinear rocking load-displacement curve is
calculated using a Beam-on-Winkler Foundation
model. Then, the energy conservation rule can
be used to estimate the intensity and position of
the reaction force normal to the foundation base.
Recent case histories also back up the empirical engineering judgment that current shallow foundations have a sufficient safety margin
even against large earthquakes. The damage to
highway bridge shallow foundations such as
excessive settlement or inclination was not
caused in recent large earthquakes in Japan.
However, our understanding of the actual
behavior of shallow foundations during large
earthquakes is not sufficient enough to model
shallow foundations as macroscopic structural
elements in the seismic design calculation of

INTRODUCTION

Shallow foundations are considered to have


larger safety margins than pile foundations
against large earthquakes, because design
norms request that shallow foundations are directly rested on sturdy bearing layers. As for the
Specifications for Highway Bridges that is the
highway bridge design norm in Japan, further
empirical regulations for design of the normal
situation and small-to-mid scale earthquake
situation are required. For example, the choice
of bearing layer, the limitation of maximum soil
reaction stress intensity, and the limitation of
the degree of partial uplift are specified to prevent the foundations from an excessive settlement from the viewpoint of long-term serviceability. After all, such additional safety margins
can prevent bearing failure, excessive settlement, and inclination even during large earthquakes.
In addition, seismic loads to shallow foundations are considered to be reduced during large
earthquakes. It is expected that a progressive
reduction in contact area between footing and

275

structures. In design, we are unable to estimate


the plastic displacement caused by the cyclic
large force to the soil beneath the footing. Much
of the current design just owes the past experience.
The behavior of a shallow foundation subjected to a combination of vertical load (V),
horizontal load (H), and moment (M) has been
extensively investigated. As for the ultimate
bearing capacity of shallow foundations to
combined loading, a failure locus concept in the
V-H-M space has been developed (Nova and
Montrasio 1991, Butterfield and Gottardi 1994,
Houlsby and Martin 1993). As for the displacement of the foundation up to the ultimate state,
the macro-element approach has been developed (Nova and Montrasio 1991, Gottardi and
Butterfield 1995, and others). A work-hardening
plasticity is applied to the evolution of the yield
locus in the V-H-M space. As a combined load
increases, the yield locus expands and finally
becomes the failure locus in the V-H-M space.
Therefore, the foundation response can be obtained within the context of plasticity with a
relevant flow rule. This approach is much simpler than, for example, other sophisticated finite
element approaches and easy to implement
into the computation of dynamic soil-structure
interactions (Paolucci 1997, Cremer et al. 2001,
Okamura and Matsuo 2002, di Prisco et al.
2002).
However, the current macro-element modeling is basically based on monotonic loading experiments and the investigation of the macroscopic foundation behavior for cyclic combined
loading is nascent. Although, recently, some
experimental findings have shown the seismic
behavior of shallow foundations (e.g. Haya and
Nishimura 1998, Negro et al. 2000, Faccioli et
al. 2001, Gajan et al. 2005), available experimental datasets are still limited.
This paper presents the results of large
scale 1G tests of model pier footings on sand
that were subjected to shake table loading, cyclic lateral loading, and concentric vertical loading. This paper first introduces the testing
methods and then represents major findings
that should be useful for the establishment of
macro-element modeling for earthquake loading.
Note that a numerical simulation for the present
experiment is tackled in a companion paper
(Paolucci et al. 2007).

276

SHAKE TABLE EXPERIMENTAL PROGRAM

The experiments were conducted at the


Large-scale Shake Table Facility in the Public
Works Research Institute, Tsukuba, Japan. The
details in the experiment are reported in a
PWRI report (Fukui et al. 2007b).
Test apparatus

Fig. 1 shows photos of the experiment setup. The size of the shake table was 8 m 8 m
in plan. A laminar shear box having internal dimensions 4 m 4 m in plan and 2.2 m high was
placed on the table. It consisted of 10 layered
frames. The shake table was rocked only in the
North-South horizontal direction.
A dry dense sand deposit was made of Toyoura sand up to a height of 2.1 m in the laminar
shear box and compacted in layers so that a
satisfactory homogeneous soil condition was
obtained. The soil relative density, Dr, was 80%,
and the mass density, , was 1.60 103 kg/m3.
CD triaxial compression tests revealed that the
internal friction angle, , is 42.1. Undrained cyclic triaxial compression laboratory tests determined a small strain elastic modulus E0:
E0 = 34.92 'c0.4436 103 kN/m2

(1)

in which the unit of confining stress, 'c, is


kN/m2. 33 accelerometers were embedded in
the sand deposit.
A model pier footing was located on the center of the sand deposit surface. Fig. 2 shows a
schematic diagram of the model pier footing.
The model comprised of three main structural
components: a top steel rack, a short steel Ibeam column, and a footing. Fig. 2 also shows
the mass, mG, structural moment of inertia
about the gravity center, J, and height from the
footing base to the gravity center, hG, for each
structural component. The total weight of the
structural model was 8385 N. The total height of
the model was 0.753 m. The height of the center of gravity was 0.420 m from the base of the
footing. A steel rack at the top was 5227 N
heavy, including steel plates gathered on the
rack. A short steel beam having an I crosssection and end diaphragms connected the
steel rack and the footing. The I-beam connection was much stiffer than the soil-foundation
system and is considered rigid. The footing
shape was a 0.5 m sided square block. 11 bi-

earthquake are greatly affected by both intensity and duration (or the number of cycles) of
the earthquake, and, therefore, two types of
ground motion are involved in design. The Type
I seismic motion is ground motion associated
with the interplate-type earthquake having a
magnitude of approximately 8 and is generated
at plate boundaries in the ocean. The Type II
seismic motion is ground motion associated
with the inland-strike-type earthquake having
magnitudes of approximately 7 and is caused
by faults located at short distances from bridge
sites. The peak amplitudes of Type I seismic
motions are smaller than those of Type II seismic motions, but Type I seismic motions have
longer durations. Type II motions have high intensities but short durations. The shake table
was also excited by a sweep wave to check the
basic vibration properties at different stages.
The shake table motions were captured with
laser displacement transducers and accelerometers.

directional load cells were attached as the base


of the footing, so that the distribution of normal
and shear reactions to the base of the footing
was captured. The long side of the load cell had
the same length as the foundation side and it
was perpendicular to the excited direction. The
contact surfaces of the load cells to the soil deposit were rough, being covered with sand paper. 25 accelerometers were also attached to
capture horizontal and vertical accelerations of
the model pier footing.
The static safety factor in terms of the bearing capacity is considered to be 29, based on a
monotonic vertical loading experiment result
that will be mentioned later. It is much larger
than the required safety factors of the Japanese
highway design norm, a safety factor of 3 for
the normal design situation considering both
dead and live loads and 2 for the small-to-midscale earthquake design situation considering
dead loads and the effect of the earthquakeinduced inclination and eccentricity of load on
the bearing capacity. However, as mentioned in
INTRODUCTION, the Japanese design norm
also requires the limitations of the degree of
partial uplift and the intensity of subgrade reaction stress for both normal and small-to-midscale earthquake design situations. As a result
that we examined several past design results of
highway bridge shallow foundations, the ratios
of the bearing capacities to the dead loads
tended to be in the range of 9 to 24. The ratio of
the height of the center of gravity of the model
pier footing to the footing length was 0.84, and
it is also included in typical ratios in past design
case histories of highway bridge pier footings
(Fukui et al. 1999). Therefore, the model pier
footing had similar mechanical properties to
those of typical design cases.
Fig. 3 shows the two earthquake records
that rocked the shake table. A motion recorded
at Shichiho Bridge, Hokkaido, Japan, during the
1993 Hokkaido Nansei-Oki Earthquake (MW =
7.8) and the N-S component recorded at JMA
(Japan Metrological Agency) Kobe during the
1995 Hyogo-ken Nanbu (Kobe) Earthquake (MW
= 6.9) were adopted. The Shichiho Bridge motion is consistent with the Type I seismic motions and the Kobe motion is a very representative of the Type II seismic motions in the Japanese Specifications for Highway Bridges (JRA
2002). The characteristics of inelastic behavior
of both soils and structural members during an

Test series

We had two test series, Case S1 and Case


S2, where S denotes the shake table experiment. The sand deposit was separately prepared for each Case. Each case had several
excitation phases with different input motions.
Test series are tabulated in Table 1. Because
the system was gigantic, earthquake motions
were not perfectly reproduced on the shake table. Case S1 was originally planned for Case
S1-2, i.e. Type I seismic motion case, and Case
S2 was planned for Case S2-2, i.e. Type II
seismic motion case. However, after Case S1-3,
we rehearsed for the Case S2-2 experiment,
applying the Kobe motion several times with
different experimental conditions as test runs.
Although we were originally thinking that the
data was going to be dumped, the test excitation phases exhibited unique behavior, and we
ended up dealing with them as formal excitation
cases S1-4 and S1-5 and analyzing the data.
In Case S2-2, the Kobe motion was applied
but the acceleration amplitude was 80% of the
original record in time domain to prevent from
overturning failure and obtain a dynamic shallow foundation motion until the end of the excitation. Although a 10 mm embedment (4% of
the footing height) was involved, we presumed
we can disregard the embedment effect on the
overall foundation behavior.

277

Test apparatus
Data processing

Fig. 5 shows a schematic diagram and a


photo of the experimental setup. In a deep test
pit having internal dimensions 4 m 4 m in plan,
a dry sand deposit was made of Toyoura sand
to a height of 2 m with an average relative soil
density of Dr = 80% (soil density = 1.60 103
kg/m3) or 60% ( = 1.60 103 kg/m3). The deposit was compacted in layers so that homogeneous soil conditions were achieved. At a relative density Dr of 60%, CD triaxial compression
tests revealed that the internal friction angle, ,
was 39.9 and undrained cyclic triaxial compression laboratory tests determined a small
strain elastic modulus E0:

As shown in Fig. 4, the measured forces


and displacements of the footing are expressed
in terms of the displacements and resultant
forces at the base center of the footing, where v
is the settlement, u is the sliding, is the rotation, V is the vertical force, H is the horizontal
force, and M is the moment. The positive direction of the sliding, u, in Fig. 4 coincides with the
North direction in Fig. 2. The horizontal displacement of the model pier footing relative to
the soil deposit surface displacement was taken
and will be shown hereafter. Because the settlement of the soil deposit was little, the settlement will be shown using the absolute displacement hereafter.
In the following results, accelerations were
obtained with accelerometers. Displacements of
the model pier footing and soil deposit surface
were captured via an image analysis of digital
VCR records. The VCR image analysis was
able to capture more accurately the long-period
part of displacement, including permanent settlement and rotation. The double-integration of
acceleration together with a proper high-pass
filtering process also can provide displacement
records. Comparisons of the displacement records obtained with the VCR image analysis
and the double-integration of acceleration
showed that the phase changes and amplitudes
agree well with each other in the main excitation part but the double-integration was unable
to recover the long-period part reliably, including permanent settlement and rotation because
of the high-pass filtering process.
The resultant forces at the base center point
of the footing were obtained with the bidirectional load cells. Accordingly, the P-Delta
effect, the additional moment caused by selfweight of the structure moving through a lateral
displacement, is automatically included in the
estimated moment, M.

E0 = 24.68 'c0.4776 103 kN/m2

(2)

in which the unit of confining stress, 'c, is


kN/m2.
Fig. 6 shows a schematic diagram of the
model