3rd Greece Japan 2009 Proceedings

Editors
GeorgeGazetas
YozoGoto
TakashiTazoh
Hellenic Society for Earthquake Engineering & Laboratory of Soil Mechanics NTUA
Earthquake Engineering Committee of the J apanese Society of Civil Engineers
Proceedingsofthe
3
rd
GreeceJapanWorkshop
Santorini2223September2009
Seismic
Design, Observation, Retrofit
of Foundations
Special Theme :
Seismic Protection of Cultural Heritage

3
rd
GreeceJ apan Workshop : Seismic Design, Observation, and Retrofit of
Foundations.
Special Theme : Seismic Protection of Cultural Heritage

Proceedings

Organized by:

the Hellenic Society for Earthquake Engineering and
the Laboratory of Soil Mechanics of the National Technical University of Athens

with the
Earthquake Engineering Committee of the J apanese Society of Civil Engineering

Sponsored by :
ATTIKO METRO, NEOTEK

2009 Laboratory of Soil Mechanics, National Technical University of Athens
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.

Scientific & Organizing Committee

Co-Chairmen :

G. Gazetas (NTUA, Greece)
Y. Goto (ERI, J apan)
T. Tazoh (Shimizu Co., J apan)

Members :
N. Yoshida
M. Saitoh
K. Ohtomo
F. Miura
M. Sato
H. Kimata
K. Pitilakis
G. Boukovalas
P. Dakoulas
N. Gerolymos
I. Anastasopoulos

Local Organizing Team

M. Georgiopoulou
Th. Zafeirakos
D. Fyselia

Preface

The 3
rd
GreeceJapan Workshop on Seismic Design of Foundations follows the very successful
previousWorkshops(WS):inAthens(October2005)andTokyo(April2007).Acommonfeature
of these WS was the active participation of leading researcherspractitioners (mostly from
Japan) and academics (mostly from Greece). The topics of these WS covered a wide range of
foundation problems, as well as a few topics of more general interest in earthquake
geotechnics.TheGreekauthorshadmostlyemphasizednumericalandanalyticalsolutions;the
Japanese researchers utilised largescale experimental testing along with numerical
simulations, mainly in support of innovative design concepts. As a result, valuable cross
fertilisation of methods, topics, and ideas has been developing between our two engineering
communities.
WhilemanyofthecharacteristicsofthefirsttwoWShavebeenpreserved,this3
rd
WSinitiates
adoubledeparture:
Thetopicsbroadentoencompassearthquakeandgeotechnicalproblemsnotstrictlyrelated
to foundations, including seismic retrofitting and analysis of bridges, as well analysis of
dams,slopes,retainingsystems.
A Special Theme Session on the Seismic protection of cultural heritage is established as a
significantpartoftheWS.Thetopicsofthissessionrangefromseismicanalysisofancient
monuments,protectionofartifacts,restorationofearthquakedamagedhistoricstructures,
to policy aspects of restoration and prioritisation. Seven of the related lectures have been
submittedinwritingtoappearintheseProceedings.
As always, in addition to Japanese and Greek authors, there are articles and/or oral
presentationsfromleadingengineersfromFrance,USA,U.K.,India,China,Algeria,Austria.
WehopethatthereadersoftheProceedingswillbesatisfiedwith(ordespiteof)suchabroad
varietyofsubjects.
G.Gazetas
1
,Y.Goto
2
,T.Tazoh
3
September2009

1
NationalTechnicalUniversityofAthens(NTUA),Greece
2
EarthquakeResearchInstitute,TheUniversityofTokyo,Japan
3
InstituteofTechnology,ShimizuCorporation,Japan

Proceedings:ListofPapers
1.PileFoundations,Liquefaction
Astudyonthebehaviorofrakedpilesinseismicallyliquefiablesoils
S.Bhattacharya,T.Tazoh,J.Jang,M.Sato........................................................................................1
KinematicResponseofBatterPileFoundation:CentrifugeTests
T.Tazoh,M.Sato,J.Jang,Y.Taji,G.Gazetas.....................................................................................20
KinematicandInertialBehaviorofBatterPiles
A.Giannakou,N.Gerolymos,G.Gazetas...........................................................................................36
SeismicResponseofBridgePileColumns
V.A.Drosos,N.Gerolymos,G.Gazetas........................................................................................51
EfficiencyofEqualEnergyAssumptionforEvaluatingDuctilityFactorsofaPile
S.Mori..............................................................................................................................................71
SinkageofaPileFoundationduringtheNiigatakenChuetsuokiEarthquakein2007
Y.Goto.................................................................................................................................................82
LargescaleShakeTableTestsonLateralSpreadingofSheetpileQuayWallandPileFoundation
M.Sato,K.Tabata...............................................................................................................................89
EDefenseShakingTableTestonLiquefactionInducedLateralSpreadingofLargeScaleModelGround
withQuayWallandPileSupportedStructure
K.Tabata,M.Sato...............................................................................................................................96
Designchartsforsinglepilesunderlateralspreadingofliquefiedsoil
A.Valsamis,G.Bouckovalas,E.Drakopoulos....................................................................................104
PilesinLiquefactionInducedSoilFlowbehindQuayWall:ASimplePhysicalMethodVersusCentrifuge
Experiments
P.Tasiopoulou,N.Gerolymos,T.Tazoh,G.Gazetas.........................................................................114
FiniteElementAnalysisofPileSoilInteractionSystembyOverlayingMeshMethod
A.Ohta,F.Miura,Y.Ono,J.Kiyono....................................................................................................128
UniquedescriptionofliquefactionbehaviorofToyourasandswithdifferentdensities
F.Zhang,Y.Jin.....................................................................................................................................140
2.EarthquakeObservations,GroundMotions
ImpactofModerateEarthquakesinPostBhujEra:CaseStudyofSikkim2006andDurgapur2008
Earthquakes,India
S.C.Dutta,P.Mukhopadhyay,S.Bhattacharya.........................................................................154
StatisticalEvaluationofEmbedmentEffectonDamagetoRCBuildingStructuresduringthe1995
HyogokenNanbuEarthquake
A.Mikami,Y.Nariyuki,T.Matsuda...............................................................................................171
GlobalIncreaseofNaturalDisastersandInternationalCooperationforDisasterMitigation
MasanoriHamada,XuWu..............................................................................................................177
TsunamiDamageStudiesandConstructionoftheMemorialPolesinBandaAceh
H.Iemura,M.H.Pradono,M.Sugimoto.......................................................................................186
TheAlgerianExperienceofCoveringEarthquakeDamagesandAppliedTechniquesof
Reinforcement
A.Moulay..........................................................................................................................................193
NumericalAnalysisofNearFieldAsymmetricVerticalMotion
T.Tobita,S.Iai,T.Iwata..................................................................................................................206
EffectofPreYieldingElasticityonSlidingTriggeredbyNearFaultMotionsModeledasIdealized
Wavelets
E.Garini,G.Gazetas,N.Gerolymos..............................................................................................219
TheIrrationalityofCurrentSeismicCodeSpectraforSoftSoils:ProposedRemedy
A.Ziotopoulou,G.Gazetas.............................................................................................................229
3.SoilStructureInteraction
CalculationofseismicresponseofbuildingbasedonpushoveranalysisofSSImodel
M.Iiba,Y.Umemura,O.Kurimoto,T.Akita,M.Teshigawara,K.Watanabe..238
AftershockEffectsonDamageEvaluationsforSoilStructureInteractionSystem
K.Kawano,Y.Kimura.258
SoilStructureInteractionProblemofaSchoolBuildingBasedonEarthquakeRecordsand
DynamicSubstructureMethod
M.Nakamura,Y.Kitamura,J.Suzumura,K.Hanada...........................267
EffectofFoundationSoilInterventionstotheSeismicResponseofMdofStructures
K.Pitilakis,E.Kirtas,E.Rovithis.....276
Shakingtabletestonseismicbehaviorofgravestomesimilartostonepillarswithandwithout
reinforcement
S.Miwa,A.Furukawa,J.Kiyono...............................287
MechanicalRepresentationofDynamicStiffnessofSoilFoundationSystems
M.Saitoh.....................................................................................................296
NonlinearDynamicAnalysisofPartiallySupportedBeamColumnsonNonlinearElastic
FoundationIncludingShearDeformationEffect
E.J.Sapountzakis,A.E.Kampitsis............................302
Seismicriskoftheundergroundstructureconsideringtheuncertaintyofseismicaction
S.Nakamura,S.Nishiyama,T.Matsumoto,Y.Miyagawa..........................................330
Topographicirregularitiesandsoilfoundationstructureinteraction
D.Pitilakis.............................................335
Numericalmodellingofshallowfoundationundercyclicoverturningmomentandvalidation
throughcentrifugeexperiments
M.Apostolou.............................................344
4.Bridges:Analysis,Design,andRetrofit
RetrofittechnologiesoflongspanbridgesinHanshinExpressway
Y.Adachi,H.Kanaji,T.Nishioka......................................................................................355
PileFoundationsinImprovedSoilforHighwayBridgefacingLiquefactionInducedSoilFlow
N.Gerolymos,G.Gazetas.................................................................................................362
MethodologyforDesignAnalysisofBridgesagainstanEmergingFaultRupture
I.Anastasopoulos,R.Kourkoulis,V.Drosos,T.Georgarakos,G.Gazetas..375
TowardsaReversalofSeismicCapacityDesign.PartA:AnalysisofBridgePierFoundation
System
I.Anastasopoulos,M.Loli,N.Gerolymos,M.Apostolou,G.Gazetas...............................393
TowardsaReversalofSeismicCapacityDesign:PartB.ShakingTableTestingofBridgePier
FoundationSystem
I.Anastasopoulos,T.Georgarakos,V.Drosos,S.Giannakos,G.Gazetas...407
SimplifiedMethodfortheDesignofRaftFoundationsagainstaDirectHitbyThrustFaulting
I.Anastasopoulos....419
PreliminarySFSIStudiesfortheMessinaBridgeFoundations
E.Stavropoulou,I.Anastasopoulos,G.Gazetas.....438
5.Dams,Slopes,RetainingWalls
FieldObservationsofWenchuanEarthquakeofMay122008
H.Wang..449
LateralandLongitudinalSeismicVibrationsofConcreteFaceRockfillDams
P.Dakoulas....470
DynamicCrackPropagationAnalysisofConcreteGravityDamswithJointedRockFoundation
H.Kimata,Y.Fujita,H.Horii,M.Yazdani.................................486
EngineeringaspectsofsiteandtopographyeffectsatAegion,Greece
O.J.Ktenidou,D.Raptakis,K.Pitilakis,F.J.ChvezGarca.....495
PilesforStabilisingSeismicallyPrecariousSlopes.PartA:DevelopmentandValidation
R.Kourkoulis,F.Gelagoti,I.Anastasopoulos,G.Gazetas......506
PilesforStabilisingSeismicallyPrecariousSlopes.PartB:ParametricAnalysisandDesignCharts
R.Kourkoulis,F.Gelagoti,I.Anastasopoulos,G.Gazetas......520
Onseismicresponseofretainingstructures
N.Sitar,L.AlAtik.........................................................................534
TheNikawa(1995)andHigashiTakezawa(2004):ModellingandNumericalAnalysis
N.Gerolymos......545
EffectofhystereticdampingandstiffnessatunloadingonresponseofgroundduringEarthquake
N.Yoshida......573
6.Monuments
Structuralrestorationprocesstoprotectaworldheritagemonumentafteranearthquake
disaster:thecaseoftheKatholikonofDafniMonasteryinAtticaGreece
A.MiltiadouFezans...584
OntheDynamicBehaviorofaLightweightIsolatorforMuseumArtifacts
V.Koumousis......605
LessonslearntfromtheImpactofSikkim2006EarthquakeonHeritageStructures
P.Mukhopadhyay,S.C.Dutta,S.Bhattacharya.....615
SeismicResponseAnalysisofUndergroundMonumentalStructures.TheCatacombsofKom
ElShoqafa,Alexandria,Egypt.
K.Pitilakis,S.Hemeda.....629
Estimationoftheseismicresponseofhistoricalstructuresandevaluationofinterventions
I.N.Psycharis,H.P.Mouzakis,A.Miltiadou,P.Pavlopoulou,I.M.Taflampas644
InPlaneResponseofoldMasonryunderSeismicLoads
T.Zimmermann,A.Strauss....658
ThereconstructionofthenortheastcornerofthetempleofZeusatNemea,Greece
N.Makris..665
DynamicAnalysisofRuptureProcessofMasonryBuildingsusingNewSimulationMethod
FurukawaA.,KiyonoJ.,TokiK....678
1.
PileFoundations,Liquefaction
A study on the behavior of raked piles in seismically liquefiable

soils

S. Bhattacharya
University of Bristol (U.K)
T. Tazoh & J.Jang
Institute of Technology, Shimizu Corporation, Japan
M. Sato
National Research Institute for Earth Science & Disaster Prevention, Japan

1 INTRODUCTION
ABSTRACT: Raked piles are often used in seismic areas to support structures, see for example
ports, wharfs or bridges. However, there is confusion on the performance of raked pile i.e.
whether or not it is beneficial. This paper focuses on two aspects: (a) Critically analyzing the
pattern of observed failures during the past earthquakes; (b) Reporting a series of well-planned
centrifuge tests to understand the behavior of raked piles. Broad conclusions on the behavior of
raked piles will be drawn.

Pile foundations are invariably used in seismic areas particularly in soft soil zones (liquefiable
zones or soft clay) to support structures such as medium to tall buildings or medium to large
span bridges. These foundations continue to collapse or get severely damaged following most
major earthquakes causing great concern to the earthquake geotechnical engineering commu-
nity, see for example the collapse of structures in the aftermath of 1995 Kobe earthquake, 1999
Taiwan and Koceli earthquake or the 2001 Bhuj earthquake. While vertical piles are common,
there are cases where raked (inclined) piles have been used to support structures. Typically,
raked piles are used to support marine (port/wharfs/offshore platforms) structures or bridge
foundations where large lateral loads are expected.
While broad aspects of the various failure mechanisms of vertical pile foundations are un-
derstood, there is little consensus on the predominant/dominant mechanism of failure. The main
failure mechanisms for vertical pile foundations are:
(a) Bending of piles due to inertial loads at the pile head or kinematic loads from soil flow
and/or wave propagation. Details of calculating inertial loading can be found in most codes such
EC8. On the other hand, methods for calculating kinematic loads on piles can be found in JRA
(2002). Mylonakis (2001) discusses simplified methods to calculate kinematic bending mo-
ments in piles in layered soils due to wave propagation effects.
(b) Buckling of piles due to loss of restraint owing to soil liquefaction. Details can be found
in Bhattacharya et al (2004, 2005).
(c) Shear loading on the pile. This can be problematic in hollow piles.
(d) Dynamic failure due to the frequency effects of the earthquake due to the change in natu-
ral frequency of the structure-pile-soil system, see Bhattacharya et al (2009).
On the other hand, there is a confusion regarding the use of raked or inclined piles in seismic
areas mainly due to the mixed performances during the past earthquakes. Many codes of prac-
tice prohibit the use of raked piles in seismic areas; see commentary on the codes by Giannakou
et al (2007), Gerolymos et al (2008), Escoffier et al (2008), AFPS (1990). This paper has there-
fore two aims:
(a) Collating and critically analyzing the performance of raked piles during the past earth-
quakes.
(b) Report a series of centrifuge tests in order to understand some aspects of the behavior.
Proceedings, 3rd Greece-Japan Workshop: Seismic Design, Observation, and Retrofit of Foundations. Santorini 2009
1
2 REPORTED CASE STUDIES ON RAKED PILES

Study of damages of structures supported on raked piles during the past earthquakes showed
mixed observations i.e. some foundations performed superbly while others collapsed. For ex-
ample, raked piles remained undamaged in Landing Bridge in the aftermath of 1987 Edgecumbe
earthquake and also in the port structures in Kandla Port (India) following the 2001 Bhuj earth-
quake. Severe damages to raked piles were observed in Rio Banano Bridge during the 1991
Costa Rica earthquake, Wharf structure in the Port of Oakland during the 1989 Loma Prieta
earthquake). Table 1 lists some of the performances.

Table 1: Case studies of raked pile
Case history and reference Details Remarks
Performance of landing bridge
during 1987 Edgecumbe earth-
quake , Berrill et al (2001)
The piles are 406mm square PSC piles at 1:6 rake (an-
gle of batter is

). The pile passes through 4m of liq-
uefied soil. Both piers and abutments are supported by
raked piles.
5 . 9
Good per-
formance
Port of Oakland after the 1989
Loma Prieta earthquake, see
Seed et al (1991).
Raked piles were used in conjunction with vertical
piles. The piles passed through bay mud and then were
founded on dense sand. In the 7
th
street terminal (pile
dimension is 406.4mm square PSC section). At Piers
27 and 29, the section is 508 mm square PSC section.
Poor per-
formance
Rio-Banano bridge during the
1991 Costa Rica earthquake, see
EERI (1991) see Figure 1 for
the section of the Bridge.
Abutments were supported on two rows of 360mm
square driven precast piles. The front row had a 1:5
rake (angle of batter is

) and the abutment rotated. 3 . 11
Poor per-
formance
Rio-Bananito bridge during the
1991 Costa Rica earthquake,
EERI (1991)
Piles were installed in rake of 1:5 and 1:10, see Figure
2.
Poor per-
formance
Inclined pile supported quay
wall following the 1995 Kobe
earthquake, see Kastranta et al
(1998)
Inclined piles supported Maya wharf . Good per-
formance
Berths of Kandla Port during the
2001 Bhuj earthquake (India)
Hollow RC pile of 0.51m and average length of 17.7m Good per-
formance

Figure 1(a): Example of raked piles (Rio-Banano Bridge), following EERI (1991)

2

Figure 1(b): Example of Rio-Bananito Bridge using raked piles, following EERI (1991)
3 ANALYTICAL METHOD TO COMPARE STIFFNESS OF VERTICAL PILES AND
RAKED PILES

One of the reasons in using raked piles is the enhanced lateral stiffness. This section aims to ob-
tain a comparative study of vertical and lateral stiffness. Figure 3 shows two pile foundations
one having a group of 4 vertical piles and the other a group of 4 raked piles. In both the cases,
the length of the pile is L having bending rigidity of EI. The raked piles are defined by an angle
of batter (| ) which is the angle that the pile forms with the vertical. It is assumed that the pile is
fixed at some level below the liquefiable zone. This point of fixity depends on the relative pile-
soil stiffness. Typical values of fixity lies between 5 and 10 times the diameter of the pile. In
this section, simple analytical expression is computed for finding the lateral stiffness of piled
foundations for these two cases.

3.1 Lateral stiffness of a group of vertical piles
The boundary condition is assumed as free to translate at the pile head and free to rotate. The
lateral stiffness of a group of 4 piles is given by:
3
12
4
L
EI
K = (1) where
EI = Bending Rigidity of the pile
L = Length of the pile
The expression (1) can be found in most standard text books.
3
|
u
Plane of fixity in non-
liquefiable hard layer
VERTICAL PILE RAKED PILE

Figure 2: Vertical and raked piles
3.2 Lateral stiffness of the group of raked piles
The lateral stiffness of a raked pile is given by Equation 2.
(
+ =
(
+ = u u u u
2 2
2
3
2
3
2
sin cos
12
12
4 sin
12
cos 4
I
AL
L
EI
L
EI
L
EA
K
raked
(2) where
EA = Axial stiffness of the pile where A = Cross sectional area of the pile
u = Angle made by the pile with the horizontal as shown in the Figure 3

If it is noted that the second moment of area I is given by:
2
min
Ar I = (3) where
r
min
= Minimum radius of gyration

Combining equations 2 and 3 we get:
(
+
|
|
.
|
\
|
=
(
+ = 1 cos 1
12
12
4 sin cos
12
12
4
2
2
3
2 2
2
min
2
3
u
u u
L
EI
Ar
AL
L
EI
K
raked
(4) where
min
r
L
= known as Slenderness ratio of the pile in the likely unsupported zone.
Bhattacharya et al (2004) analyzed 14 case studies of pile performance in seismic zones. The
analysis showed that piles that survived the earthquake has slenderness ratio less than 50. Com-
paring equation 1 and 4 we define an improvement factor Z
i
as:
1 cos 1
12
2
2
+
|
|
.
|
\
|
= = u
K
K
Z
raked
i
(5)
4
Equation 5 suggests that raked piles are always stiffer that vertical piles. Equation 5 can be re-
casted using angle of batter (| ) as Equation 6.
1 sin 1
12
2
2
+
|
|
.
|
\
|
= = |
K
K
Z
raked
i
(6)
Equation 6 is plotted in Figure 3. The factor Z
i
represents the ratio of the stiffness of a raked
pile with respect to the vertical piles and can be coined as Improvement Factor. In the Figure
Z
i
is plotted against the angle of batter for various slenderness ra io ( t ).

0 2 4 6 8 10 12 14 16 18 20
1
10
20
30
40
50
60
70
80
90
100
Angle of Batter (|)
I
m
p
r
o
v
e
m
e
n
t

F
a
c
t
o
r

(
Z

i
)

=40
=60
=80
=100

Figure 3: Normalized stiffness ratio of a raked pile plotted against the angle of batter

The next section describes a series of centrifuge tests that were carried out at Institute of Tech-
nology (Shimizu Corporation) to investigate various aspects of raked piles and vertical piles.
4 CENTRIFUGE TESTS
4.1 Description of the test
A series of centrifuge tests has been carried out in Institute of Technology (Shimizu Corpora-
tion) to study the various aspects of raked piles. Two types of pile group were tested as shown
in Figure 4. Table 2 summarizes the different centrifuge tests carried out. However, this paper
concentrates on one particular test (Case 32). In some of the tests, a small scaled superstructure
was used: see Figure 5 for the photograph of the superstructure and Figures 6 and 7 for sche-
matic diagram of the pile-supported structures. The input motion applied at the base of the
model is given by Figure 8 and the FFT of the signal is given by Figure 9.

5
Table 2: Schedule of the tests

ID of the test Details of the test Remarks
SERIES -1
Case 11
Two vertical pile groups near a quay wall were tested.
In one case the quay wall collapsed and in the other the
wall did not collapse. The piled foundations are placed
200mm from quay wall.
SERIES -1
Case 12
SERIES-1
Case 13
In the three tests of SERIES -1, the
main aim was to study the effect of
quay wall collapse on the piled
foundations. The distance between
the quay wall and the piled founda-
tions were varied. All the piled
foundations had a superstructure.
SERIES-2
Case 21
Two vertical pile groups were tested. In one case, there
was a superstructure and in the other there is no super-
structure.
SERIES-2
Case 21
Two vertical pile groups were tested. In one case, the
foundation was 50mm from the quay wall and in the
other the foundation was 100mm from quay wall.
In SERIES -2 tests, the effect of in-
ertia on the superstructure was in-
vestigated
SERIES-3
Case 31
Two pile groups were tested. One had a vertical pile
and the other had a raked pile. The angle of batter is
. The distance from the quay wall is 200mm.
10
SERIES-3
Case 32
10
SERIES-3
Case 33
10
SERIES-3
Case 34
5
In SERIES -3 tests, comparison is
made between vertical piles and
raked piled.

Figure 4: Photograph of the vertical pile and raked piles used in the centrifuge tests

6

Figure 5: Centrifuge setup of the raked piles

Liquefiable
soil
ACC-Footing
ACC-Base AG(0)
ACC-Structure
A-PP2
A-PP1
100mm
130mm
30mm
588.75gm
392.5gm
80mm
60mm
Quay wall
Columns consists of 4
nos of 6mmx2mmx60
mm steel section
A-4
A-2
A-1
A-5
A-6
A-7
A-9
A-10

Figure 6: Schematic diagram of the vertical pile showing the instrumentation. [{A-i denotes Strain
Gauges where i varies from 1 to 10 except 3 and 8}, {ACC is accelerometer}, {PP denotes Pore Pres-
sure}].

7
Liquefiable
soil
ACC-Footing
ACC-Base AG(0)
ACC-Structure
B-PP2
A-PP1
100mm
130mm
30mm
588.75gm
392.5gm
80mm
60mm
Quay wall
Columns consists of 4
nos of 6mmx2mmx60
mm steel section
B-4
B-6
B-7
B-9
B-10

Figure 7: Schematic diagram of the vertical pile showing the instrumentation [{A-i denotes Strain Gauges
where i varies from 1 to 10 except 3 and 8}, {ACC is accelerometer}, {PP denotes Pore Pressure}].

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-10
-8
-6
-4
-2
0
2
4
6
8
10
Time(s)
B
e
d
r
o
c
k

a
c
c
e
l
e
r
a
t
i
o
n

(
g
)

AG(0)

Figure 8: Input motion used in Test Case 32
8
0 30 60 90 120 150 180 210 240
-20
0
20
40
60
80
100
Frequency (Hz)
L
o
g

a
m
p
l
i
t
u
d
e

(
d
B
)

o
f

i
n
p
u
t

Figure 9: Input motion used in the tests
4.2 Measured responses from the tests
Figure 10 shows the time history of the measured responses for the vertical pile. The input ac-
celeration, the pore pressure responses and the acceleration of the footing and structure are plot-
ted. The responses indicate that as the soil liquefies i.e. at about 0.15sec and after the responses
of the footing and structures alter. The instrumentation layout can be found in Figure 6. Figure
11 shows the similar data for the raked piles and the corresponding instrumentation. Figure 12
compares the acceleration of structure for two types of foundations. Figure 13 shows the corre-
sponding acceleration for footing for two types of foundations.
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
-10
0
10
A
c
c
e
l
e
r
a
t
i
o
n

(
g
)

Vertical pile-Structure-AS1
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
-10
0
10
A
c
c
e
l
e
r
a
t
i
o
n

(
g
)

Vertical pile-Footing-AS1
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
0
20
40
45
P
o
r
e

p
r
e
s
s
u
r
e

(
k
P
a
)

A-PP1
A-PP2
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
-10
0
10
Time(s)
B
a
s
e

a
c
c
e
l
e
r
a
t
i
o
n

(
g
)

AG(0)

Figure 10: Time history of results for vertical piles

9
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
-20
0
20
A
c
c
e
l
e
r
a
t
i
o
n

(
g
)

Raked pile-Structure-AS1
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
-20
0
20
A
c
c
e
l
e
r
a
t
i
o
n

(
g
)

Raked pile-Footing-AS1
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
0
20
40
P
o
r
e

p
r
e
s
s
u
r
e

(
k
P
a
)

B-PP1
B-PP2
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
-20
0
20
Time(s)
B
a
s
e

a
c
c
e
l
e
r
a
t
i
o
n

(
g
)

AG(0)

Figure 11: Time history of results for raked piles

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55
-30
-20
-10
0
10
20
30
Time (s)
A
c
c
e
l
e
r
a
t
i
o
n

(
g
)

Input
Structure-vertical piles
Structure-raked piles

Figure 12: Time history of acceleration for structures for two types of foundations

10
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55
-10
-8
-6
-4
-2
0
2
4
6
8
10
Time (s)
A
c
c
e
l
e
r
a
t
i
o
n

(
g
)

Input
Footing-vertical piles
Footing-raked piles

Figure 13: Time history of acceleration of footing for two types of foundations

4.3 Frequency response analysis of the foundations
Before we analyse the experimental data, it is necessary to estimate various parameters. Table 3
characteristics of the pile.

Table 3: Characteristics of the pile
Material Steel
Pile outside diameter (d
o
) 10mm
Wall thickness (t) 0.2mm
Second Moment of Area (I) 73.95mm
4

Radius of gyration ( mm
A
I
r 46 . 3
min
= = )
3.46mm
Pile length (L) 270mm
Youngs Modulus (E) 210GPa
EI (Bending rigidity of the pile) 15.5310
6
Nmm
2

4.3.1 Stiffness of vertical piles

Based on Figure 6, the boundary condition of the vertical pile is assumed as free to translate at
the pile head and free to rotate. The length of unsupported length is given by the summation of
top non-liquefiable zone (as the zone moves and 60mm), the liquefiable zone (130mm) and
50mm (depth of fixity). Therefore L is about 240mm. The lateral stiffness of a group of 4 piles
is given by equation 1 is given by equation 7.
mm
N
mm
Nmm
L
EI
K 92 . 53
240
10 53 . 15 48 12
4
3 3
2 6
3
=

= = (7)
The section of the column is (2mm x 6mm) and 60mm high. This can be considered to be
very rigid compared to the pile. As a first approximation, the time period of the structure can be
11
computed by considering the stiffness of the pile and the mass of the footing, columns and
structure. The total mass of the structure is given by equation 8.

kg gm gm gm gm M 1 85 . 10003 6 . 22 75 . 588 5 . 392
I
= = + + = (

8)
he frequency is given by: T
Hz
m kg
N
f
n
9 . 36
10 1
92 . 53
2
1
3
=
=

t
(9)
Figures 14 and 15 shows the frequency domain analysis of the acceleration data of the foot-
ing and structure. This is essentially, plotting the acceleration time history of the footing data
and structure in the normalized frequency domain. The analysis indicates that there is a peak at
about 36 Hz which indicates that this is natural frequency of the system.

0 10 20 30 40 50 60 70 80 90 100
0
0.5
1
1.5
2
2.5
3
3.5
4
Frequency (Hz)
D
y
n
a
m
i
c

a
m
p
l
i
f
i
c
a
t
i
o
n

f
a
c
t
o
r

(
F
o
o
t
i
n
g
)

Figure 14: Frequency response function of e footing for the vertical pile foundation. th
12
0 10 20 30 40 50 60 70 80 90 100
0
0.5
1
1.5
2
2.5
3
3.5
4
Frequency (Hz)
D
y
n
a
m
i
c

a
m
p
l
i
f
i
c
a
t
i
o
n

f
a
c
t
o
r

(
S
t
r
u
c
t
u
r
e
)

Figure 15: Frequency response function of the structure for the vertical pile foundation

4.3.2 Stiffness of raked piles

The angle of batter of the raked pile is . Based on Figure 7, and equation 6 we may obtain
the stiffness of the raked pile. The slenderness ratio of the pile required is given by equation 10
10
36 . 69
46 . 3
240
min
= = =
r
L
(10)
The improvement factor is given by equation 11.
13 1 ) 10 sin 400 ( 1 sin 1
12
2 2
2
~ + = +
|
|
.
|
\
|
= = |
K
K
Z
raked
i
(11)
The stiffness of the raked pile foundation is given by equation 12.

mm
N
mm
N
K
raked
710 92 . 53 13 = = (12)
As a first approximation, the frequency of the raked pile foundation supported structure is
given by:
Hz
m kg
N
f
n
134
10 1
710
2
1
3
=
=

t

Figures 16 and 17 plots the frequency response function of the acceleration data of the foot-
ing and structure. The results indicate that there is a peak at about 82Hz (Figure 17).
13
0 20 40 60 80 100 120 140 160 180 200
0
0.2
0.4
0.6
0.8
1
1.2
1.4
Frequency (Hz)
D
y
n
a
m
i
c

a
m
p
l
i
f
i
c
a
t
i
o
n

f
a
c
t
o
r

(
F
o
o
t
i
n
g
)

Raked pile

Figure 16: Dynamic amplification factor measured from the structure for the raked piles (Figure 7).

0 20 40 60 80 100 120 140 160 180 200
0
0.5
1
1.5
Frequency (Hz)
D
y
n
a
m
i
c

a
m
p
l
i
f
i
c
a
t
i
o
n

f
a
c
t
o
r

(
S
t
r
u
c
t
u
r
e
)

Raked pile

Figure 17: Dynamic amplification factor measured from the structure for the raked piles (Figure 7).

14
4.4 Bending moment at the front and back side of the vertical pile
Strain measurements were taken at various locations of the vertical pile (see Figure 6 for the in-
strumentation layout). Figure 18 shows the bending strain at various locations of the pile near to
the quay wall. Figure 19 shows the bending strain at various locations in the back side. The tests
indicate that there are large moments at the interface possible due to kinematic interactions.
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-2000
-1500
-1000
-500
0
500
1000
Time (s)
B
e
n
d
i
n
g

s
t
r
a
i
n
-
v
e
r
t
i
c
a
l

p
i
l
e

(

)

A1(1d)
A2(6d)
A4(19d)
A5(27d)

Figure 18: Bending strain along the vertical pile in the front of the quay wall

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-2000
-1500
-1000
-500
0
500
1000
1500
Time (s)
B
e
n
d
i
n
g

s
t
r
a
i
n
-
v
e
r
t
i
c
a
l

p
i
l
e

(

)

A6(1d)
A7(6d)
A9(19d)
A10(27d)

Figure 19: Bending strain along the vertical pile in the back pile

Figures 20 and 21 shows the bending strain for the raked piles (see Figure 7 for instrumenta-
tion layout). Figure 22 compares the bending strain for the raked pile and the vertical pile which
is possible due to inertia. Figure 23 shows the kinematic bending moment at the interface for
vertical pile and raked pile.

15
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-1500
-1000
-500
0
500
1000
1500
Time (s)
B
e
n
d
i
n
g

s
t
r
a
i
n
-
r
a
k
e
d

p
i
l
e

(

)

B1(1d)
B2(6d)
B4(19d)
B5(27d)

Figure 20: Bending strain along the raked pile in the front of the quay wall

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-1600
-1200
-800
-400
0
400
800
1200
1600
Time (s)
B
e
n
d
i
n
g

s
t
r
a
i
n
-
r
a
k
e
d

p
i
l
e

(

)

B6(1d)
B7(6d)
B9(19d)
B10(27d)

Figure 21: Bending strain along the vertical pile in the back pile

16
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-600
-400
-200
0
200
400
600
800
1000
1200
Time (s)
I
n
e
r
t
i
a
l

b
e
n
d
i
n
g

s
t
r
a
i
n

(

)

Vertical Pile (A1)
Raked Pile (B1)

Figure 22: Comparison with vertical pile and raked pile

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-600
-400
-200
0
200
400
600
800
1000
1200
Time (s)
K
i
n
e
m
a
t
i
c

b
e
n
d
i
n
g

s
t
r
a
i
n

a
t

t
h
e

i
n
t
e
r
f
a
c
e

(

)

Vertical Pile (A4)
Raked Pile (B4)

Figure 23: Vertical pile and raked pile

4.5 Bending strain in the front and back pile
Figures 24 and 25 shows the bending strain for the front pile and the back pile.
17
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-500
-400
-300
-200
-100
0
100
200
300
400
Time (s)
B
e
n
d
i
n
g

s
t
r
a
i
n
-
v
e
r
t
i
c
a
l

p
i
l
e

(

)

(Front pile A4)
(Back pile A9)

Figure 24: Strain measurements taken at the interface (A4 and A9) of the raked pile configuration.

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
-600
-400
-200
0
200
400
600
800
1000
1200
Time (s)
B
e
n
d
i
n
g

s
t
r
a
i
n
-
r
a
k
e
d

p
i
l
e

(

)

(Front pile B4)
(Back pile B9)

Figure 25: Strain measurements taken at the interface (B4 and B9) of the raked pile configuration.

5 CONCLUSIONS
Raked piles are often used in seismic areas to support structures. However, confusion still exists
on their performance i.e. whether or not they are beneficial. This paper presents a study in this
regard. Following conclusions are drawn:
1. Raked piles are always stiffer than the vertical piles. The stiffness increase depends
on the angle of batter and also on the slenderness ratio of the pile. An expression of
this increase i.e. the ratio of lateral stiffness of raked piles to the lateral stiffness of
the vertical pile denoted by Improvement factor is derived. This enhanced stiffness
affects the performance in various ways and is explained in the next two points.
2. As the stiffness of a raked pile system is comparatively high, the frequency of the
structure-foundation system is also high. Therefore, the raked pile foundation is dy-
namically sensitive and the performance depends on the type of earthquake.
3. Due to relatively high horizontal stiffness of the raked pile foundation, the horizontal
displacement of the pile cap or the structure is comparatively less making its per-
18
formance satisfactory from serviceability point of view. This further reduces the P-
delta effect of the superstructure.
4. Due to this high stiffness, raked piles attract significant bending moments near the
pile head. As a result, proper connection details are necessary.
5. Raked pile requires careful design consideration and can prove beneficial to most en-
gineering structures.
6 REFERENCES
AFPS (1990), Recommendations AFPS 90, 3 Volumes, Association franaise de gnie parasismique,
France
Bhattacharya, S., S.Adhikari and N.A.Alexander (2009): A simplified method for unified buckling and
free vibration analysis of pile-supported structures in liquefiable soils, Journal of Soil Dynamics and
Earthquake Engineering, August 2009 Issue
Bhattacharya, S., Bolton, M.D. and Madabhushi, S.P.G. (2005): A reconsideration of the safety of the
piled bridge foundations in liquefiable soils, Soils and Foundations, Volume 45, August 2005 issue,
No 4, pp 13-26
Bhattacharya, S., Madabhushi, S.P.G., and Bolton, M.D. (2004): An alternative mechanism of pile fail
ure in liquefiable deposits during earthquakes, Geotechnique 54, April issue, No.3, pp 203-213.
Berril J. B., Christensen S.A., Keenan R. P., Okada W. and Pettinga J. R. (2001), Case study of lateral
spreading forces on a piled foundation, Geotechnique, Vol. 51, No 6, pp. 501-517
EERI (1991), Costa Rica Earthquake of April 22, 1991, Reconnaissance Report, Earthquake Spectra,
Vol. 7, Supplement B.
Giannakou, A., Gerolymos,N. and Gazetas,G. (2007): Influence of Batter Piles on the Seismic Response
of Pile Groups, 2
nd
Japan Greece workshop on seismic design, observation and retrofit of foundations,
3
rd
-4
th
April 2007, Tokyo.
Gerelymos, N., Giannakou, A., Anastasopoulos, I., and Gazetas, G. (2008): Evidence of beneficial role of
inclined piles: Observation and summary of numerical analysis, Bulletin of Earthquake Engineering,
Vol 6, pp 705-722.
Escoffer,S., Chazelas, J-L., and Garnier, J. (2008): Centrifuge modeling of raked piles, Bulletin of Earth-
quake Engineering, Vol 6, pp 689-704.
Kastranta G. & Gazetas G, Tazoh T. (1998), Performance of three quay walls in Maya Wharf: Kobe
1995, Proceedings of the 11th European Conference on Earthquake Engineering
JRA (1996,1980,1972): Japanese Road Association, Specification for Highway Bridges, Part V, Seismic
Design.
Mylonakis G., 2001, Simplified model for seismic pile bending at soil layer interfaces, Soils and
Foundations, 41, 47-58.

19
Kinematic Response of Batter Pile Foundation: Centrifuge
Tests

T. Tazoh
M. Sato
National Research Institute for Earth Science & Disaster Prevention, Japan
J. Jang & Y. Taji
G. Gazetas
National Technical University of Athens, Greece

ABSTRACT: We carried out centrifuge tests to clarify the seismic behavior of batter-pile
foundations. A vertical-pile foundation and a batter-pile foundation without the presence of a
superstructure were installed parallel to each other in a soil container filled with dry sand, and
were excited simultaneously. Through a comparison of the acceleration and displacement
response of the footing, as well as the axial and bending strain of the piles for the two pile
foundations, the kinematic response of the seismic behavior of the batter-pile foundation was
experimentally investigated.

1 INTRODUCTION

The lateral stiffness of a pile foundation can be increased by adopting batter piles, which is why
they are commonly used in landing piers that are subject to large lateral forces. However, batter
piles are seldom used for buildings or civil engineering structures even in the case of large
lateral forces. The reasons are as follows:
1) When soil settlement occurs, not only the safety of the pile foundation but also that of the
structure as a whole system may be threatened by settlement-induced vertical loads acting
on the batter piles.
2) During an earthquake, the piles in a batter-pile foundation may be subject to excessive
axial compression and pullout forces, which are not generated in a vertical-pile foundation.
3) The strength of concrete piles is reduced by decreasing the compressive force acting on the
piles due to rocking motions induced by the adopted batter piles.
4) Since infinite lateral ground planes cannot be assumed for batter piles, they cannot be
expected to have the same horizontal subgrade reaction as that of vertical piles.
5) In urban areas, the use of batter piles is constrained by the boundary lines of adjacent land.

The 1995 Great Hanshin Earthquake in Japan has increased the demand for pile foundations
with high seismic performance, as well as lower cost and easier construction. Batter piles can be
used with little additional expense, no special design, and hardly any difficulty in construction.
Therefore, the seismic behavior of batter piles has recently attracted much research interest, as
has research and development related to easy and accurate methods of installing batter piles
(Gerolymos, N., et. al., 2008, Giannakou, A., et. al., 2007, Poulos., N., 2006).

In this study, we carried out centrifuge shaking table tests to clarify the seismic behavior of
batter-pile foundations. A vertical-pile foundation and a batter-pile foundation were installed
20
parallel to each other in a soil container filled with dry sand, and were excited simultaneously
(Tazoh, T., et. al., 2005, Tazoh, T., et. al., 2007). As our objective was to investigate the
fundamental characteristics of the seismic behavior of batter piles, none of the pile-foundation
models had a superstructure. This study focused on the kinematic interaction of batter piles
(Fan, K., et. al., 1991, Mylonakis, G., et. al., 1997, Mylonakis, G., 2001, Nikolaou, S., et. al.,
2001, Sica, S., et., al., 2007, Tazoh, T., et. al., 1987). Through a comparison of the acceleration
and displacement response of the footing, as well as the axial and bending strain of the piles for
the two pile foundations, the kinematic nature of the seismic behavior of the batter-pile
foundation was experimentally studied.

2 CENTRIFUGE TESTS

The most direct and effective way to quantitatively and qualitatively investigate the seismic
behavior of batter piles is to compare the seismic behavior between a vertical-pile foundation
and a batter-pile foundation under the same input motions. Each test for each model must be
carried out under nearly identical conditions with respect to input motions, soil conditions, and
soil behavior. Note, however, that it is impossible to achieve complete similarity between
shaking table tests due to the difficulty of reproducing the input motion and nonlinear behavior
of the soil.
Therefore, a vertical-pile foundation and a batter-pile foundation without the presence of a
superstructure were installed parallel to each other in a soil container, as shown in Figure 1, and
were excited simultaneously.
A laminar box was used as the soil container to allow shear deformation of the soil deposit as
in the free field. Actually, installing two models that behave differently in a laminar box is not
an appropriate testing method because the behavior of the models might influence each other.
However, considering the inconsistency of the input motion and the difficulty of reproducing
the soil conditions and nonlinearity, we believe that this method is more reasonable than
individually testing the vertical-pile foundation and batter-pile foundation separately.
The interior of the soil container is 805 mm in length, 474 mm in width, and 324 mm in
height. All tests were conducted at centrifugal acceleration of 30 g on a 1/30-scale model.
Table 1 shows the scaling ratios of the models.
The vertical-pile foundation and the batter-pile foundation each have four piles, and the pile
heads and pile tips are rigidly connected to the footing and the base of the soil container,
respectively. The batter piles are identically inclined at a 10 angle. The soil deposit is a
uniform layer consisting of dry silica sand No. 7 (Mean particle diameter D
50
= 0.15 mm; Soil
density
s
= 2.635 g/cm
3
; Maximum dry density
max
= 1.539 g/cm
3
; Minimum dry density
min
= 1.206 g/cm
3
). Thickness and relative density of the soil deposit is 300 mm (prototype:
9 m) and Dr = 60%, respectively.
Figure 2 shows the grain size accumulation curve of silica sand No. 7. Table 2 shows the
materials and size of the experimental model used in the tests and Photograph 1 shows the test
model. Sixty-two monitoring channels in total were installed, with the sensors comprising
seventeen accelerometers, five non-contact displacement meters, and forty strain gauges
(Table 3). The test was conducted a total of nine times, varying the input motion and maximum
acceleration as shown in Table 4.
While the purpose of this study was to clarify the kinematic interaction of the batter piles,
consideration must also be given to effects from the mass of the footing (made of steel, size:
355 cm). The inertial interaction caused by the inertial force of the footing might be included
in the results, which consequently may not represent the perfect kinematic interaction.

3 KINEMATIC NATURE OF SEISMIC BEHAVIOR OF BATTER PILE

Figure 3 shows the frequency transfer function calculated by the acceleration records between
the soil surface and the input motion of the sweep test. The predominant frequency of the
21
ground is 3.5 Hz in the case of maximum acceleration of input motion at 5 Gal. The
predominant frequencies are 3.23.3 Hz and 3.03.1 Hz, and also the peak acceleration
amplification factors decrease corresponding to the increase in maximum acceleration of the
input motion to 15 Gal and 30 Gal.
Figure 4 shows the frequency transfer function between the ground surface and input motion
obtained from El Centro record excitation. The predominant frequency of the ground is 3.4 Hz
in the case of maximum acceleration of input motion at 50 Gal. The predominant frequencies
are 2.82.9 Hz and 2.42.5 Hz, and also the peak acceleration amplification factors decrease
according to the increase in maximum acceleration of the input motion to 100 Gal and 200 Gal.
These phenomena were obviously produced by the nonlinearity of the soil.

Figure 5 shows the relationship between horizontal displacement and rotational angle of the
footing based on the data from sinusoidal excitation of 3.5 Hz, in order to investigate the
rotational characteristics of the footing of the vertical-pile foundation and the batter-pile
foundation. The rotational angle is calculated by dividing the difference in the vertical
displacement based on the data of the accelerometers installed at both sides of the footing by the
distance between the two accelerometers.
The fact that there is no phase difference between the sway and the rocking motion indicates
that the response of the footing to motion to the right is counterclockwise rotation, as shown in
Figure 7. There is no phase difference between the sway and the rocking motion of the
vertical-pile foundation; on the other hand, anti-phase behavior can be seen in the data for the
batter-pile foundation.
Figure 6 shows the data obtained from El Centro record excitation at the maximum
acceleration of 200 Gal. The same trend as seen in the case of sinusoidal excitation can also be
found in Figure 6. The phenomena of the opposite phase between the sway and the rocking
motions of the vertical-pile foundation and the batter-pile foundation can be found in all of the
other test data. From Figures 5 and 6, it can also be seen that the rotation angles of the
batter-pile foundation are almost two times larger than those of the vertical-pile foundation.

Figure 8 shows the maximum-value distribution of the bending and axial strains of the piles
in the vertical-pile foundation (pile-VA1) and the batter-pile foundation (pile-BA1) obtained
from sinusoidal excitation of 3.5 Hz. The frequency of 3.5 Hz closely corresponds to that of the
predominant frequency of the ground as shown in Figure 3. The largest values were obtained at
the pile heads, and the bending and axial strains of the batter-pile foundation are larger than
those of the vertical-pile foundation in all cases, as shown in Figure 8.
Figure 9 shows the maximum-value distribution of the bending and axial strains of the piles
in the vertical-pile foundation (pile-VA1) and the batter-pile foundation (pile-BA1) obtained
from El Centro record excitation. The largest values were obtained at the pile heads, and the
bending and axial strains of the batter-pile foundation are larger than those of the vertical-pile
foundation, likely due to the sinusoidal excitation.

Figures 10 and 11 show the maximum values for acceleration of the footings and the ground
surface, and the bending and axial strains at the pile heads corresponding to the increments in
maximum acceleration of the input motion. From the figures, it can be seen that the maximum
acceleration of the footing of the vertical-pile foundation is larger than that of the batter-pile
foundation and that both the bending and axial pile strain of the batter-pile foundation are larger
than those of the vertical-pile foundation in both the sinusoidal and El Centro record excitation.

4 ASEISMICITY OF BATTER PILE

Figures 12 and 13 compare the frequency transfer functions of the horizontal acceleration of the
footing and input motion between the vertical-pile foundation and the batter-pile foundation
obtained from sweep test and El Centro record excitation. The difference between the frequency
transfer functions of the two pile foundations represents the aseismicity of the batter-pile
22
foundation. From these figures, it can be elucidated that the batter-pile foundation has a certain
level of aseismicity in all of the frequency ranges.
Figures 14 and 15 compare the frequency transfer functions of the bending and axial strains
of the piles and input motion between the vertical-pile foundation and the batter-pile foundation
calculated using the data from the sweep test and El Centro record excitation. From these
figures, it can be seen that the strain of the batter piles is larger than that of the vertical piles.
Therefore, it is considered that the compensation for the aseismicity of batter piles seeks large
cross-sectional efficiency for the batter piles.

5 CONCLUSIONS
The main conclusions of the study are as follows:

1) The response of the footing of the vertical-pile foundation to motion to the right is
counterclockwise rotation. On the other hand, that of the batter-pile foundation is rotation
in the opposite direction to that of the vertical-pile foundation.
2) Bending and axial strains attain the largest values at the pile heads in both the vertical-pile
foundation and batter-pile foundation.
3) Improved aseismicity by adopting batter piles can be gained in almost all frequency ranges.
4) Bending and axial strains of the batter-pile foundation are larger than those of the
vertical-pile foundation. In other words, the compensation for the aseismicity of batter piles
seeks large cross-sectional efficiency for the batter piles.

6 ACNOLEDGEMENT
We would like to express our sincerest gratitude to Mr. Katsumi Yoshinari for his invaluable
help to carry out the centrifuge tests.

7 REFERENCES
Fan, K., Gazetas, G., Kaynia, A., Kausel, E., & Ahmad, S. 1991. Kinematic seismic response of single
piles and pile groups, Journal of Geotechnical Engineering, 117, 18601879.
Gerolymos, N., Giannakou, A., Anastasopoulos, I. & Gazetas, G. 2008. Evidence of beneficial role of
inclined piles: observations and summary of numerical analyses, Bulletin of Earthquake Engineering:
Vol. 6, No. 4, 705722: Springer.
Giannakou, A., Gerolymos, N., & Gazetas, G. 2007. Kinematic response of groups with inclined piles,
Proceedings of the 4
th
International Conference on Earthquake and Geotechnical Engineering
(ICEGE), Thessaloniki, Greece, on CD-Rom.
Mylonakis, G., Nikolau, S., & Gazetas, G. 1997. Soil-pile bridge seismic interaction: kinematic and
inertial effects. Part I: Soft soil, Earthquake Engineering and Structural Dynamics, 26, 337359.
Mylonakis, G. 2001. Simplified model for seismic pile bending at soil layer interfaces, Soils and
Foundations, 41, 4, 4758.
Nikolaou, S., Mylonakis, G., Gazetas, G., & Tazoh, T. 2001. Kinematic pile bending during earthquakes:
analysis and field measurements, Geotechnique, 51, 5, 425440.
Poulos, N. 2006. Raked piles---virtues and drawbacks, Journal of Geotech Geoenviron Engineering,
132(6), 795803. Doi:10.1061/(ASCE) 10900241(2006)132:6(795).
Sica, S., Mylonakis, G., & Simonelli, A. L., 2007. Kinematic bending of piles: Analysis vs. code
provisions, Proceedings of the 4
th
International Conference on Earthquake and Geotechnical
Engineering (ICEGE), Thessaloniki, Greece, on CD-Rom.
Tazoh, T., Shimizu, K., & Wakahara, T. 1987. Seismic observations and analysis of grouped piles.
23
Geotechnical Special Publication No. 11, Dynamic Response of Pile Foundations, ASCE, 120.
Tazoh, T., Sato, M., & Gazetas, G. 2005. Centrifuge tests on pile-foundation structure systems affected
by liquefaction-induced flow due to quay-wall collapse, Proceedings of the 1
st
Greece-Japan
Workshop on Seismic Design, Observation and Retrofit of Foundations, Athens, Greece, 79106.
Tazoh, T., Sato, M., Jang, J., & Gazetas, G. 2007. Centrifuge tests on remedial measure using batter piles
against liquefaction-induced soil flow after quay wall failure, Proceedings of the 2
nd
Greece-Japan
Workshop on Seismic Design, Observation and Retrofit of Foundations, Tokyo, Japan, 431-439.
24

Figure 1 Longitudinal sections and plan of the 1/30-scale centrifuge model
(scale unit: mm, for the prototype dimensions: multiply by 30. A vertical-pile foundation and a
batter-pile foundation without the presence of a superstructure were set parallel to each other in a soil
container which was filled with dry sand, and were excited simultaneously.)

Photograph 1 Testing Model
(The pile foundations have four piles.)
807
4
7
5
3
2
0
2
3
7
.
5
3
0
0
2
0
3
2
0
3
0
0
2
0
2
3
7
.
5
330
D-VF-X
D-BF-X
A-G-X1
10 10
A-G-X55
A-G-Z1
A-VF-Z
330
D-T-X1
A-G-X2
A-G-X3
A-G-X4
A-VF+Z
A-VF-X
A-BF-X
A-BF+Z A-BF-Z
A-G-X6 A-G-X7
A-G-X8 A-G-X9
S-VA130 S-VA230
S-VA2260
S-VA2190
S-VA2130
S-VA270
S-VA1260
S-VA1190
S-VA1130
S-VA170
S-BA230 S-BA130
S-BA1260
S-BA1190
S-BA1130
S-BA170
S-BA2260
S-BA2190
S-BA2130
S-BA270
D-G-X
A-G-X1
A-G-X55 A-G-Z1
D-G-Z
D-VF-X
D-BF-X
A-G-X6
A-G-X7
A-VF+Z A-VF-X
A-BS-X
A-BF-Z A-BF+Z
D-T-X1
D-G-Z
D-G-X
A-G-X4
X
Z
X
Z
X
Y
25 50 125 130
3
0
0
1
1
8
EV-P
EP-B
Accelerometer
Strain gauge
Noncontact displacement meter
Earth pressure meter

Vertical-pile f. model
Batter-pile f. model
Vertical-pile f. model
Batter- pile f. model
25

Table 1 Scaling ratios of the testing model
Item Symbol Unit
Centrifuge
model
Prototype Scale
Sand
stratum
Depth H m 0.3 9 1/N
Density
t
KN/m
3
19.8 19.8 1
Structure
Width W m 0.05 1.5 1/N
Height H m 0.04 1.2 1/N
Mass M kg 0.785 21,195 1/N
3

Footing
Width W m 0.05 1.5 1/N
Height H m 0.03 0.9 1/N
Mass M kg 0.58875 15,896 1/N
3

Column
Width L m 0.006 0.18 1/N
Width
(shaking direction)
W m 0.004 0.12 1/N
Moment
of inertia of area
I m
4
3.20E-11 2.59E-05 1/N
4

Length L m 0.06 1.8 1/N
Pile
Diameter D m 0.01 0.3 1/N
Thickness t m 0.0002 0.006 1/N
Young's modulus E MN/m
2
2.06E+05 2.06E+05 1
Area A m
2
6.16E-06 5.54E-03 1/N
2

Moment
of inertia of area
I m
4
7.40E-11 5.99E-05 1/N
4

Normal stiffness EA MN 1.27E+00 1.14E+03 1/N
2

Bending stiffness EI MN-cm
2
1.52E-09 1.23E-03 1/N
4

Acceleration
Centrifuge g g 30 1 N
Earthquake Gal 6000 200 N
Other
parameters
Displacement m 1 30 1/N
Force F N 1 900 1/N
2

Stress kPa 1 1 1
Strain 1x10
-6
1x10
-6
1
Time t s 1 30 1/N
Frequency f Hz 30 1 N

Figure 2 Grain size accumulation curve of silica sand No. 7
(Mean particle diameter D
50
= 0.15 mm, Soil density
s
= 2.635 g/cm
3
,
Maximum dry density
max
= 1.539 g/cm
3
, Minimum dray density
min
= 1.206 g/cm
3
)
0
20
40
60
80
100
0.01 0.1 1
P
e
r
c
e
n
t

f
i
n
e
r

b
y

w
e
i
g
h
t

(
%
)
Grain size (mm)
D
50
=0.147mm
U
c
=1.63
G
s
=2.635
F
c
=5.2%
dmax
=1.539
dmin
=1.206

26

Table 2 Materials and dimensions of the test model
Parts Material & size
Laminar box 805mm, 475mm, 324mmInner size: length, width, depth
Soil deposit Dry sand: Silica No.7Dr 60Thickness: 300 mm
Vertical pile
Stainless steel
No. of piles422, Inclination angle: 0
Length: 270 mmDiameter:10 mmThickness: 0.2 mm
Batter pile
Stainless steel
No. of piles422, Inclination angle: 10
Length: 274 mmDiameter: 10 mmThickness: 0.2 mm
Footing
Steel
Thickness: 30 mmPlan size: 50 mm50 mm

Table 3 Installed sensors
(62 monitoring channels were installed, with the sensors comprising 17 accelerometers,
5 non-contact displacement meters, and 40 strain gauges.)
Transducer Location Direction Number Subtotal Total
Accelerometer
Batter pile
X 2
17
62
Z 2
Vertical pile
X 2
Z 2
Ground X 6
Base X 1
Table control X 1
Centrifugal acc. Z 1
Non-contact
displacement
meter
Batter pile X 1
5
Vertical pile X 1
Ground
X 1
Z 1
Base X 1
Strain gauge
Batter pile
Pile-BA1 10
40
Pile-BA2 10
Vertical pile
Pile-VA1 10
Pile-VA2 10

27

Table 4 Test cases
Input Motion
Freq. of input
motion (Hz)
Max. acc. of input
motion (Gal)
Test case No.
Sweep test motion 1.7-10 Hz
5 1-1
15 1-2
30 1-3
Sinusoidal
excitation
3.5 Hz
50 2-1
100 2-2
200 2-3
El Centro record
El Centro record
N-S component
50 3-1
100 3-2
200 3-3

Figure 3 Frequency transfer function of the ground surface obtained from sweep test
(5 Gal, 15 Gal, 30 Gal)

0
5
10
15
0 2 4 6 8 10
Sweep 5 Gal
Sweep 15 Gal
Sweep 30 Gal
A
m
p
l
i
f
i
c
a
t
i
o
n
Frequency (Hz)
Ground surface

-180
-90
0
90
180
0 2 4 6 8 10
Sweep 5 Gal
Sweep 15 Gal
Sweep 30 Gal
P
h
a
s
e

(
d
e
g
r
e
e
)
Frequency (Hz)
Ground surface
28

Figure 4 Frequency transfer function of the ground surface obtained from the El Centro record excitations
(50 Gal, 100 Gal, 200 Gal)

Figure 5 Comparisons of horizontal displacement and rotational angle of the footings between the
vertical-pile foundation and the batter-pile foundation (Sinusoidal excitation: 3.5 Hz, 200 Gal)

-180
-90
0
90
180
0 2 4 6 8 10
El Centro 50 Gal
El Centro 100 Gal
El Centro 200 Gal
P
h
a
s
e

(
d
e
g
r
e
e
)
Frequency (Hz)
Ground surface

-15
0
15
-0.8
0
0.8
Footing displacement
Footing rotation
D
i
s
p
.

(
m
m
)
R
o
t
a
t
i
o
n

(
d
e
g
.
)
Vertical pile
-10
0
10
-1.5
0
1.5
0 2 4 6 8 10
Footing displacement
Footing rotation
D
i
s
p
.

(
m
m
)
R
o
t
a
t
i
o
n

(
d
e
g
.
)
Time(sec)
Batter pile

0
10
20
30
0 2 4 6 8 10
El Centro 50 Gal
El Centro 100 Gal
El Centro 200 Gal
A
m
p
l
i
f
i
c
a
t
i
o
n
Frequency (Hz)
Ground surface
29

Figure 6 Comparisons of horizontal displacement and rotational angle of the footings between the
vertical-pile foundation and the batter-pile foundation (El Centro record: 200 Gal)

Figure 7 Kinematic responses of footings

(a) 50 Gal (b) 100 Gal (c) 200 Gal

Figure 8.1 Bending strain distributions of the vertical-pile foundation (pile-VA1) and the batter-pile
foundation (pile-BA1) obtained from the sinusoidal excitation of 3.5 Hz (50 Gal, 100 Gal, 200 Gal)

-30
0
30
-1
0
1
Footing displacement Footing rotation
D
i
s
p
.

(
m
m
)
R
o
t
a
t
i
o
n

(
d
e
g
.
)
Vertical pile
-30
0
30
-2
0
2
0 2 4 6 8 10
Footing displacement Footing rotation
D
i
s
p
.

(
m
m
)
R
o
t
a
t
i
o
n

(
d
e
g
.
)
Time(sec)
Batter pile
()
()
A-V_F-Z() A-V_F-Z()
A-V_F-X
()
A-B_F-Z () A-B_F-Z()
A-B_F-X
()

0 50 100
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)
0 100 200
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)
0 200 400
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)
30

(a) 50 Gal (b) 100 Gal (c) 200 Gal

Figure 8.2 Axial strain distributions of the vertical-pile foundation (pile-VA1) and the batter-pile
foundation (pile-BA1) obtained from the sinusoidal excitation of 3.5 Hz (50 Gal, 100 Gal, 200 Gal)

(a) 50 Gal (b) 100 Gal (c) 200 Gal

Figure 9.1 Bending strain distributions of the vertical-pile foundation (pile-VA1) and the batter-pile
foundation (pile-BA1) obtained from El Centro record (50 Gal, 100 Gal, 200 Gal)

0 25 50
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)
0 50 100
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)
0 75 150
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)

0 50 100
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)
0 150 300
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)
0 200 400
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)
31

(a) 50 Gal (b) 100 Gal (c) 200 Gal

Figure 9.2 Axial strains distributions of the vertical-pile foundation (pile-VA1) and the batter-pile
foundation (pile-BA1) obtained from El Centro record (50 Gal, 100 Gal, 200 Gal)

Figure 10 Maximum values of the accelerations of the footings and the ground surfaces, and the
bending and axial strains at the pile-heads (Sinusoidal excitation: 3.5 Hz)

0 25 50
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)
0 50 100
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)
0 75 150
-9
-6
-3
0
Vertical pile
Batter pile
Bending strain (x10
-6
)
D
e
p
t
h

o
f

p
i
l
e

h
e
a
d

(
m
)

32

Figure 11 Maximum values of the accelerations of the footings and the ground surfaces, and the
bending and axial strains at the pile-heads (El Centro record)

(a) 5 Gal (b) 15 Gal (c) 30 Gal

Figure 12 Aseismicity of the batter-pile foundation: Comparison of the frequency transfer function
between the horizontal acceleration and input motion of the footing in the vertical-pile foundation and the
batter-pile foundation obtained from sweep tests (5 Gal, 15 Gal, 30 Gal)

0
2.5
5
7.5
10
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n
Frequency (Hz)
Footing
(Sweep 5 Gal)
0
2.5
5
7.5
10
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n
Frequency (Hz)
Footing
(Sweep 15 Gal)
0
2.5
5
7.5
10
0 2 4 6 8 10
Vertical-pile
Batter-pile
A
m
p
l
i
f
i
c
a
t
i
o
n
Frequency (Hz)
Footing
(Sweep 30 Gal)
33

(a) 50 Gal (b) 100 Gal (c) 200 Gal

Figure 13 Aseismicity of the batter-pile foundation: Comparison of the frequency transfer function
between the horizontal acceleration and input motion of the footing in the vertical-pile foundation and the
batter-pile foundation obtained from El Centro record (50 Gal, 100 Gal, 200 Gal)

(a) Bending strains

(b) Axial strains

Figure 14 Comparisons of the frequency transfer functions of the bending and axial strains of the piles
and input motion between the vertical-pile foundation and the batter-pile foundation (sweep tests)

0
2
4
6
8
10
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n
Frequency (Hz)
Footing
(El Centro 50 Gal)
0
2
4
6
8
10
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n
Frequency (Hz)
Footing
(El Centro 100 Gal)
0
2
4
6
8
10
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n
Frequency (Hz)
Footing
(El Centro 200 Gal)

0
1
2
3
4
5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
Sweep 5 Gal
Pile-A1 head
(Bending strain)
0
1
2
3
4
5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
Sweep 15 Gal
Pile-A1 head
(Bending strain)
0
1
2
3
4
5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
Sweep 30 Gal
Pile-A1 head
(Bending strain)

0
0.5
1
1.5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
Sweep 5 Gal
Pile-A1 head
(Axial strain)
0
0.5
1
1.5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
Sweep 15 Gal
Pile-A1 head
(Axial strain)
0
0.5
1
1.5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
Sweep 30 Gal
Pile-A1 head
(Axial strain)
34

(a) Bending strains

(b) Axial strains

Figure 15 Comparisons of the frequency transfer functions of the bending and axial strains of the piles
and input motion between the vertical-pile foundation and the batter-pile foundation (El Centro record)

0
1
2
3
4
5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
El Centro 50 Gal
Pile-A1 head
(Bending strain)
0
1
2
3
4
5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
El Centro 100 Gal
Pile-A1 head
(Bending strain)
0
1
2
3
4
5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
El Centro 200 Gal
Pile-A1 head
(Bending strain)

0
0.5
1
1.5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
El Centro 50 Gal
Pile-A1 head
(Axial strain)
0
0.5
1
1.5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
El Centro 100 Gal
Pile-A1 head
(Axial strain)
0
0.5
1
1.5
0 2 4 6 8 10
Vertical pile
Batter pile
A
m
p
l
i
f
i
c
a
t
i
o
n

(
m
i
c
r
o
/
G
a
l
)
Frequency (Hz)
El Centro 200 Gal
Pile-A1 head
(Axial strain)
35

1 INTRODUCTION
Batter piles have been used for a long time to resist large lateral loads from winds, water waves,
soil pressures, and impacts. Their distinct advantage over vertical piles is that they transmit the
applied loads partly in axial compression, rather than only through shear and bending. Thus,
batter piles offer larger stiffness and bearing capacity than same-diameter-and-depth vertical
piles a superiority of particular importance when the near-surface soils are soft and/or the
lateral load is large.
Despite these advantages, they do not enjoy a good reputation for seismic resistance.
Following the poor performance of batter piles in a series of earthquakes, the seismic behaviour
of inclined piles has been considered detrimental, and many codes require that such piles be
avoided. For instance, the French Seismic Code (AFPS 90) states flatly that Inclined piles
should not be used to resist seismic loads. The seismic Eurocode EC8 / Part 5, dealing with
geotechnics and foundations, is a little less restrictive, stating: It is recommended that no in-
clined piles be used for transmitting lateral loads to the soil. If, in any case, such piles are used,
they must be designed to carry safely axial as well as bending loading.
Case histories that have recently confirmed the potential for unsatisfactory performance of
improperlydesigned batter piles include the wharf in the Port of Oakland in the 1989 Loma
Kinematic and Inertial Behavior of Raked Piles
A. Giannakou
1
, N. Gerolymos
2
, G. Gazetas
2

1
Fugro West, Oakland, California, USA
2
ABSTRACT: Several aspects of the seismic response of groups containing non-vertical piles are
studied, including the lateral pile head stiffnesses, the kinematic pile deformation, and the
inertial soil-pile-structure response. A key goal is to explore the conditions under which the
presence of batter piles is beneficial, indifferent, or detrimental. Parametric analyses are carried
out using 3D finite element modeling, assuming elastic behavior of soil, piles, and superstruc-
ture. The model is first utilized to obtain the lateral stiffnesses of single batter piles and to show
that its results converge to the available solutions from the literature. Then, real accelerograms
covering a broad range of frequency characteristics are employed as base excitation of simple
fixed-head 2-pile group configurations, embedded in homogeneous, inhomogeneous, and
layered soil profiles, while supporting very tall or very short structures. Five pile inclinations are
considered while the corresponding verticalpile group results serve as reference. It is found
that in purely kinematic seismic loading, batter piles tend to confirm their negative reputation,
as had also been found recently for a group subjected to static horizontal ground deformation.
However, the total (kinematic plus inertial) response of structural systems founded on groups
of batter piles offers many reasons for optimism. Batter piles may indeed be beneficial (or
detrimental) depending on, among other parameters, the relative size of the overturming
moment versus the shear force transmitted onto them from the superstructure.
36
Prieta earthquake (M
s
= 7.1), the Port of Los Angeles in the 1994 Northridge earthquake (M
s
=
6.8), and the Rio Banano and the Rio Vizcaya Bridges in the 1991 Costa Rica earthquake (M
s
=
7.5). The bad reputation of batter piles has been reinforced by these incidents. The culmination
was the following statement in the ASCE monograph on Seismic Design of Port and Harbor
Facilities : The use of batter piles in ports is typically not encouraged because of their poor
seismic performance during past earthquakes. [Technical Council on Lifeline Earthquake En-
gineering 2001, Kavazanjian, 2006].
However, a more thorough investigation on the causes of these failures, showed that the in-
adequate reinforcement in the top of the piles and also the improper connection of piles to their
caps were the culprits of the observed damage (Mitchell et al, 1991; Priestley et al, 1991) a
result of the early isostatic method of analysis which assumed that batter piles transmit only
axial load.
But what if batter piles were properly designed to resist the developed moment and shear
loads at their head ? And if, furthermore, they were designed to posses sufficient ductility at the
head and the connection to the cap ? Would their seismic behavior still remain poor ?
A goal of this paper is to give at least a partial answer to these questions. Indeed, in recent
years, evidence has been accumulating that well-designed batter piles may not only have a satis-
factory performance themselves, but may also be beneficial for the structure they support. Re-
cent research on the seismic response of batter piles and micropiles (Guin, 1997; Lam and Mar-
tin, 1987; Sadek and Shahrour, 2004 and 2006) has shown that the seismic response of a
structure may improve in many respects when supported by inclined piles. Moreover, case his-
tories referring to the Maya Warf in the Kobe 1995 earthquake and the Landing Road Bridge in
the Edgecumbe New Zealand 1987 earthquake have highlighted the potential help provided by
inclined piles (Berril et al 2001, Gazetas & Mylonakis 1998).
As a result of the improved understanding of the source of the observed poor performance,
batter piles in recent years seem to have been re-established in their traditional role of with-
standing large horizontal loads applied to deep foundations (as pointed out in an enlightening
professional article by Kavazanjian, 2006). The piers for the new San Francisco Bay Bridge
East Span present a characteristic example of this trust in batter piles to carry huge lateral seis-
mic loads in very soft soil.
Presently, research on the seismic response of batter piles has been rather limited (Juran et al,
2001; Sadek and Shahrour, 2004 and 2006; Okawa et al, 2005; Poulos, 2006 ; Deng et al, 2007 ;
and Ravazi et al, 2007). Aiming at filling part of this gap, we study several aspects of the
seismic response of batter piles through parametric 3D analyses employing the FE method. Only
the idealized case of linear viscoelastic soil response is treated here the shortcomings of line-
arity will be explored in a forthcoming companion paper, but the conclusions drawn in this
study remain qualitatively valid even in the presence of soil nonlinearities.
The first part of the paper outlines the numerical model and shows its consistency with avail-
able analytical results.
2 PROBLEM DEFINITION AND FINITE ELEMENT MODELING
The seismic behavior of symmetric 2 x 1 group configurations with piles battered at various
angles is investigated using ABAQUS. Batter angles commonly encountered in practice are
considered, such as 5
o
15
o
, in addition to the less usual cases of 20
o
and 25
o
. The vertical fixed-
head pile group is used as a reference for delineating the role of pile inclination. Figure 1 depicts
the finite element meshes of the six configurations. All piles are of Youngs modulus E
P
= 30
GPa, diameter d = 1 m, and the depth to their tip is L = 15 m. The center-to-center distance, s,
between the piles at pilehead elevation is three pile diameters (S = 3d). The piles are rigidly
connected to a perfectlyrigid pile cap. The massandcolumn superstructure is modeled as a
single degree of freedom oscillator. The concentrated superstructure mass M
str
is such that the
load per pile in each configuration is 1.0 MN typical of actual pile designs. In all cases
studied, the fundamental period of the superstructure is 0.44 sec and of the soil 0.29 sec.

37
= 15
= 20
= 25
= 0
= 5
= 10
= 15
= 20
= 25
= 0
= 5
= 10
s1
s2
s1
s2
3.5 d
I II III IV
s
z / d
T
s
= 0.29 s
Soil Profiles
11.5 d
3d
L

=

1
5

m
d = 1 m

s
z
s1
s2
s1
s2
3.5 d
I II III IV
s
z / d
T
s
= 0.29 s
Soil Profiles
11.5 d
3d
L

=

1
5

m
d = 1 m

s

Figure 1.The six studied pile group configurations and the corresponding 3-D finite element models.

Four idealized soil profiles are considered : (a) a homogenous, (b) a nonhomogenous Gib-
son soil, (c) a two-layer profile with a bottom stiffer layer, and (d) a two-layer profile with a
top stiffer layer [crust] (Figure 2).
Both pile and soil are linear visco-elastic. Soil is modeled with eight-noded brick elements
while the piles are represented with a series of beam elements rigidly linked to the peripheral
nodes in order to properly model the pile geometry. The performed mesh sensitivity study
revealed that an element dimension of 0.5 m (i.e., one pile radius) leads to nearly accurate
results. Appropriate kinematic constraints are imposed to the lateral edges of the model, allow-
ing it to move as the free field, while elementary transmitting boundaries (V dashpots) ab-
sorb much of the wave energy emitted from the oscillating piles.
Three real acceleration time histories, covering a wide range of frequencies, are used as base
excitation: (i) The record of the 2003 Lefkada M
s
6.4 earthquake: PGA = 0.42 g, dominant pe-
riod range T
p
0.20.65 sec, [Gazetas et al., 2005]; (ii) the rock outcrop motion of the 1995 Ae-
gion M
s
6.2 earthquake : PGA = 0.39 g, dominant period range T
p
0.140.6 sec [Gazetas,
1996]; and (iii) the JMA record of the 1995 M
JMA
= 7.2 Kobe earthquake PGA = 0.83 g, T
p

0.251.0 sec [Architectural Institute of Japan, 1995].

Figure 2.Problem geometry and the four soil profiles.
38
3 STATIC STIFFNESSES, CONVERGENCE TO PUBLISHED SOLUTIONS
As a starting point, the static stiffnesses of a single batter pile are computed and contrasted with
available solutions for vertical (Poulos & Davis 1980, Gazetas 1991) and batter (Poulos 1980)
piles in Figures 3 and 4. For a homogeneous soil stratum of thickness H, which exceeds the
depth L to the pile tip (Soil Model I), our numerical (FE) results are compared with those com-
puted : (a) on the basis of Poulos (1980) and Poulos & Daviss (1980) approximation for angles
of batter = 0
and 25
o
, and (b) the approximate closed-form expressions of Gazetas (1991) for
a vertical pile. The latter were developed using the Blaney & Roesset (1976) innovative
dynamic finite-element which incorporates perfect transmitting boundaries. Figure 3 depicts the
variation of the normalized lateral static stiffnesses, K
HH
, K
RR
, K
HR
, as functions of the ratio
E
p
/E
s
of the pile and soil Youngs moduli, and the angle of batter as a parameter. We draw the
following conclusions :
The rocking K
RR
and, especially, the cross-coupled horizontalrocking K
HR
stiffnesses are
essentially independent of : vertical and battered piles have nearly identical response, as
expected by Poulos 1980 simplification (Poulos & Madhav 1971, Poulos & Davis 1980).
Therefore the closed-form expressions developed for vertical piles (Gazetas 1991),
( )
0 75
3
0 15
RR s p s
K E d E E
.
. (1)
( )
0 50
2
0 22
HR s p s
K E d E E
.
. (2)

are in excellent accord with both the FE results of this study and Poulos approximation.
The above expressions can therefore be used even with batter piles.
The horizontal stiffness K
HH
exhibits a small dependence on . Both the FE analysis and
Poulos approximation show that stiffness increases by about 30% on average when
increases from 0
o
to 25
o
. In percentage, the difference declines with increasing E
P
/E
s
ratio.
The results of the Poulos (1980) and Gazetas (1991) approximations only slightly underes-
timate the finite results. The following simple expression :
( ) ( )
2
0.21 tan
2
1.08 1 4 tan +
HH s p s
K E d E E

(3)
has been developed by fitting the FE results for batter piles. It will suffice in practical
applications for any angle .
For the linearly-inhomogeneous stratum (Gibson Soil) with Youngs modulus of the form :
( ) ( ) =
s
E z E z d (4)
in which apparently E
S
= is the modulus at one-diameter depth, similar conclusions can be
drawn from Figure 4 :
The rocking and cross swaying-rocking stiffnesses, K
RR
and K
HR
, are again practically
unaffected by the inclination of the pile, while they are sensitive to the E
P
/E
S
ratio. The
Poulos (1980) and Gazetas (1991) approximations are again in excellent accord with the
present FE results. Thus the expressions developed for vertical piles
( )
0 80
3
0 15
RR s p s
K E d E E
.
. (5)
( )
0 60
2
0 17
HR s p s
K E d E E
.
. (6)

provide very good estimates for all values of . (he reader should notice that E
S
, the
modulus at depth z = d, in the above Eqns has a different meaning from the constant E
S

modulus of Eqns 13.)

39
0
0
0
0
0
0
E
d
L

L
(z)
E
s
100 1000 10000 100000
E
p
/ E
s
= 0 [Poulos, 1980]
= 25
o
[Poulos, 1980]
= 0 [Gazetas, 1991]
= 0 [FE Analysis]
= 25
o
[FE Analysis]
12
10
8
6
4
5
0
200
160
120
80
40
0
30
25
20
15
10
5
0
K
H
H

/

E
s
d
K
R
R

/

E
s
d
3
K
H
R

/

E
s
d
2
=25
o
=0
o
all
0
0
0
0
0
0
E
d
L

L
(z)
E
s
E
d
L

L
(z)
E
s
100 1000 10000 100000
E
p
/ E
s
= 0 [Poulos, 1980]
= 25
o
[Poulos, 1980]
= 0 [Gazetas, 1991]
= 0 [FE Analysis]
= 25
o
[FE Analysis]
12
10
8
6
4
5
0
200
160
120
80
40
0
30
25
20
15
10
5
0
K
H
H

/

E
s
d
K
R
R

/

E
s
d
3
K
H
R

/

E
s
d
2
=25
o
=0
o
all

Figure 3. Normalized static stiffnesses of single pile : (a) swaying, (b) rocking, and (c) cross swaying
rocking for batter angle = 0
o
and 25
o
as a function of pile-soil stiffness ratio E
p
/ E
s
(L/d = 15, homoge-
nous soil). Comparison with solution for vertical and inclined piles from the literature.

40
100 1000 10000 100000
E
p
/ E
s
= 0 [Poulos, 1980]
= 25
o
[Poulos, 1980]
= 0 [Gazetas, 1991]
= 0 [FE Analysis]
= 25
o
[FE Analysis]
E
d
L

L
(z)
E
s
z/d
60
50
40
30
20
10
0
300
250
200
150
100
50
0
54
45
36
27
18
9
0
=25
o
=0
o
all
K
H
H

/

E
s
d
K
R
R

/

E
s
d
3
K
H
R

/

E
s
d
2
100 1000 10000 100000
E
p
/ E
s
= 0 [Poulos, 1980]
= 25
o
[Poulos, 1980]
= 0 [Gazetas, 1991]
= 0 [FE Analysis]
= 25
o
[FE Analysis]
E
d
L

L
(z)
E
s
z/d
E
d
L

L
(z)
E
s
z/d
60
50
40
30
20
10
0
300
250
200
150
100
50
0
54
45
36
27
18
9
0
=25
o
=0
o
all
K
H
H

/

E
s
d
K
R
R

/

E
s
d
3
K
H
R

/

E
s
d
2

Figure 4. Normalized static stiffnesses of single pile : (a) swaying, (b) rocking, and (c) cross swaying
rocking for batter angles = 0
o
and 25
o
as a function of pile-soil stiffness ratio E
p
/ E
s
(L/d = 15, inho-
mogenous Gibson soil). Comparison with solution for vertical and inclined piles from the literature.

41
On the contrary, the horizontal stiffness K
HH
is very sensitive to the batter angle . ncreasing
25
approximately doubles the stiffnesses according to our FE results, or triples them

according to Poulos approximation for all values of the E
P
/E
S
ratio. We note that Poulos
(1980) solution is based on an (additional) approximation, necessary to handle the soil
inhomogeneity while still utilizing Midlins solution for a homogeneous halfspace (Poulos
1979) . This may be the cause of some inaccuracy which shows up with large pile
inclinations. The FE results for batter piles have been fitted with the expression :
( ) ( )
( )
2
0.35 1 0.5tan
0.60 1 tan
+
+
HH s p s
K E d E E

(7)
It applies to flexible piles (L/d >10) ; for = 0 it reduces to the expression of Gazetas (1991).
Note that all the above expressions (Eqns 13 , 57 ) apply only for floating piles. If the pile
bears on a rigid base, the effect of becomes more prominent, noticeable even for K
RR
and K
HR
.
It is also worth noting that the pile slenderness ratio, L/d, plays a more significant role with
batter piles of large inclination than with vertical piles, because the axial stiffness which affects
K
HH
of batter piles is more sensitive to pile length ; while by contrast the stiffness K
HH
of a
vertical pile is essentially unaffected by L (at least for flexible pile ; e.g., Randolph, 1979).
4 KINEMATIC RESPONSE OF INCLINED PILE GROUPS
The kinematic response of vertical piles has been thoroughly studied by several researchers
including Flores-Barrones & Whitman (1982), Kaynia & Kausel (1982), Dobry & O Rourke
(1983), Harada et al (1981), Gazetas (1984), Fan et al (1991), Kavvadas & Gazetas (1993),
Bentley and El Naggar (2000), Nikolaou et al (2001), Takewaki & Kishida (2005). However,
little attention has been paid to the kinematic response of groups containing inclined piles.
Among the few exceptions: Sadek & Shahrour (2006) studied the seismic response of inclined
micropiles subjected to a sinusoidal motion at the eigenfrequency of the soil profile, and showed
that for kinematic loading a group of four symmetrically inclined micropiles exhibits lower
values of lateral acceleration at the cap level and larger values of internal forces in the piles
compared to a group of vertical micropiles. Deng et al (2007) performed kinematic analysis for
a large pile group containing inclined piles and found that kinematic loading can have a major
impact on the magnitude of the maximum axial force that develops in the batter piles. In their
study, such piles developed 5 to 8 times greater axial forces than the vertical piles.
The maximum values of the bending moments that develop for all motions and all soil pro-
files is summarized in Figure 5. Several trends are worthy of note :
One advantage of groups with batter piles is the reduction of the lateral displacement at the
pile cap. Evidently, the incompatibility between freefield and inclinedpile displacement
profiles becomes more pronounced as the inclination increases. A profound effect of in-
creased angle of batter is the increase and, more importantly, the change in direction of the
pilehead rotation. In other words, for the soil and pile group moving to the right, the cap
rotates counter-clockwise.
It is interesting to note that the above conclusions are similar to those recently presented by
Poulos (2006). Applying to a group of 6 rigidly-capped piles a triangular horizontal
ground displacement (with the maximum at the surface) he also found that the group
rotation is profoundly affected by the rake angle, and that for 15
o
angle of batter the
group rotation is more than four times greater than, and in the opposite direction to that
for vertical piles. (Poulos 2006, p. 799).
The bending moments and axial forces that develop in fixed-head pile groups containing
batter piles are larger than in the vertical group, in the case of the Gibson soil. Kinematic
interaction has a major effect on the maximum seismic load of batter piles. While the hori-
zontal motion of the soil during the passage of seismic waves tends to cause mainly lateral
motion of vertical piles (and thus they develop mainly shear and moment), inclined piles
experience significant axial forces as well.
These effects are largest in the homogeneous soil profile. However, in profiles where a stiff
soil layer is present (i.e., III and IV) the maximum bending moment that develops does not
vary significantly with pile inclination (Figures 5c and 5d).
42
0
o
5
o
10
o
15
o
20
o
25
o
0
o
5
o
10
o
15
o
20
o
25
o
m
a
x

M
k

/

m
a
x

M
k

v
e
r
t
i
c
a
l

g
r
o
u
p
3.5
3
2.5
2
1.5
1
0.5
0
3.5
3
2.5
2
1.5
1
0.5
0
3.5
3
2.5
2
1.5
1
0.5
0
3.5
3
2.5
2
1.5
1
0.5
0
(a) (b)
(c) (d)
0
o
5
o
10
o
15
o
20
o
25
o
0
o
5
o
10
o
15
o
20
o
25
o
m
a
x

M
k

/

m
a
x

M
k

v
e
r
t
i
c
a
l

g
r
o
u
p
3.5
3
2.5
2
1.5
1
0.5
0
3.5
3
2.5
2
1.5
1
0.5
0
3.5
3
2.5
2
1.5
1
0.5
0
3.5
3
2.5
2
1.5
1
0.5
0
0
o
5
o
10
o
15
o
20
o
25
o
0
o
5
o
10
o
15
o
20
o
25
o
0
o
5
o
10
o
15
o
20
o
25
o
0
o
5
o
10
o
15
o
20
o
25
o
m
a
x

M
k

/

m
a
x

M
k

v
e
r
t
i
c
a
l

g
r
o
u
p
3.5
3
2.5
2
1.5
1
0.5
0
3.5
3
2.5
2
1.5
1
0.5
0
3.5
3
2.5
2
1.5
1
0.5
0
3.5
3
2.5
2
1.5
1
0.5
0
(a) (b)
(c) (d)

Figure 5. Kinematic response of rigidlycapped 2pile group: normalized peak maximum bending mo-
ment along the pile. Normalization with respect to the response of the group of vertical piles (maximum,
minimum, and average values from the three accelerograms [Lefkada, Aegion, JMA]). The results refer to
the configuration and piles shown in grey at the bottom of the figure. (T
soil
= 0.29 s)

In order to understand these differences among the developing moments in the four soil pro-
files, we need to consider the two sources of kinematic straining of batter piles. First and most
obvious cause is the existence of an abrupt change in stiffness between two successive soil
layers (profiles III and IV). In this case the largest bending moment is generated at or near the
interface of the two layers as is well known from the studies on vertical piles (Dobry and
ORourke, 1983; Nikolaou and Gazetas, 1997), and hence it is practically independent of pile
inclination. (Note that this source of kinematic straining of deep foundations is recognized in
some recent codes [e.g., EC8].
The second, and perhaps more important, source of kinematic straining is the constraint im-
posed by the rigid pile cap. In this case the maximum kinematic bending moment develops at or
near the pile head (and will be later added to the inertial bending moment generated by the
oscillation of the superstructure), in contrast to the aforementioned case where the maximum
bending moment is generally developed at greater depths. In the case of the homogenous soil,
the E
P
/E
S
ratio near the soil surface is relatively small ; therefore the increase of batter angle
leads to significantly larger values of bending moment due to the increase in lateral stiffness. In
case of Gibson soil, however, the E
P
/E
S
ratio near the soil surface is very large, and the
additional stiffness provided by the pile inclination leads to smaller increase in bending
moments (in the order of 1.5 times the bending moment in the vertical pile compared to 3 in the
case of the homogenous soil).
5 SOILPILESTRUCTURE INTERACTION
The influence of the (super)structure on the seismic response of groups with batter piles is con-
sidered in this section. Specifically, to illustrate the effect of structural height, two one-degree-
of-freedom oscillators are studied, modelling: (i) a tall slender structure (H
st
= 12 m) whose cru-
43
H
str
= 12 m
H
str
= 1 m
0
o
5
o
10
o
15
o
20
o
25
o
3
2.5
2
1.5
1
0.5
0
m
a
x
M
k
+
i
m
a
x
M
k
+
i
,

f
o
r
v
e
r
t
i
c
a
l

g
r
o
u
p
H
str
= 12 m
H
str
= 1 m
0
o
5
o
10
o
15
o
20
o
25
o
3
2.5
2
1.5
1
0.5
0
m
a
x
M
k
+
i
m
a
x
M
k
+
i
,

f
o
r
v
e
r
t
i
c
a
l

g
r
o
u
p
H
str
= 12 m
H
str
= 1 m
H
str
= 12 m
H
str
= 1 m
0
o
5
o
10
o
15
o
20
o
25
o
0
o
5
o
10
o
15
o
20
o
25
o
3
2.5
2
1.5
1
0.5
0
m
a
x
M
k
+
i
m
a
x
M
k
+
i
,

f
o
r
v
e
r
t
i
c
a
l

g
r
o
u
p
cial loading is the overturning moment, an example being a tall bridge pier, and (ii) a short
structure (H
str
= 1 m) whose crucial loading is the shear force. The fundamental fixed-base pe-
riod of the two structures is kept constant, 0.44 sec. The groups are embedded in the non-
homogenous Gibson-type profile (II).
Pile distress and the role of batter are highly dependent on the type of crucial loading (mo-
ment versus shear). Figures 6 and 7 summarize in dimensionless form of the maximum bending
moment and axial force that develop in the piles of the two structures, for all earthquake excita-
tions. Note that groups with batter piles lead to small lateral displacement in earthquake
shaking. Horizontal displacements decrease invariably for both structures as the inclination
increases. However, groups with inclined piles develop larger cap rotations regardless of the
type of structure they support !
The bending moment experienced by batter piles supporting a tall structure is larger than by
vertical piles (Figure 6). The bending moment increases monotonically with pile rake. This
conclusion is valid even for large fixed-base fundamental structural periods where the inertial
effect is limited. Proper reinforcement of the pile-cap connection is necessary for undertaking
safely this bending moment and for securing adequate inelastic deformation in an unpredictably
large earthquake motion. On the contrary, groups with inclined piles supporting a short structure
develop smaller bending moments than the vertical group. This observation is in agreement with
other published numerical and experimental studies (Sadek and Shahrour, 2004; Okawa et al.,
2005; Sadek and Shahrour, 2006), and is also compatible with the prevailing engineering
perception about the role of batter piles, as will be explained in the sequel. However, when the
contribution to bending moment from inertial loading of the structure is small (i.e. flexible
structures with very large fundamental periods), the contribution to bending moment from the
kinematic deformation of the pile prevails and batter piles may suffer larger bending moments
than vertical piles.

Figure 6. Total (kinematic + inertial) response: normalized peak maximum bending moment along the
pile supporting a short (H
str
= 1 m) and a tall (H
str
= 12 m) structure. Normalization with respect to the re-
sponse of the group of vertical piles (maximum, minimum, and average values from the three accelero-
grams [Lefkada, Aegion, JMA]). The results refer to the configuration and piles shown (in grey) at the
bottom of the figure. (Gibson soil, T
s
= 0.29 s, E
p
/ E
s
= 1000, T
str
= 0.44 s)

44
0
o
5
o
10
o
15
o
20
o
25
o
2.5
2
1.5
1
0.5
0
H
str
= 1 m
H
str
= 12 m
m
a
x
N
k
+
i
m
a
x
N
k
+
i
,

f
o
r
v
e
r
t
i
c
a
l

g
r
o
u
p
0
o
5
o
10
o
15
o
20
o
25
o
0
o
5
o
10
o
15
o
20
o
25
o
2.5
2
1.5
1
0.5
0
H
str
= 1 m
H
str
= 12 m
H
str
= 1 m
H
str
= 12 m
m
a
x
N
k
+
i
m
a
x
N
k
+
i
,

f
o
r
v
e
r
t
i
c
a
l

g
r
o
u
p

Figure 7. Total (kinematic + inertial) response: normalized peak maximum axial force along the pile sup-
porting a short (H
str
str
= 12 m) structure. Normalization with respect to the response of
the group of vertical piles (maximum, minimum, and average values from the three accelerograms [Le-
fkada, Aegion, JMA]). The results refer to the configuration and piles shown (in grey) at the bottom of the
figure. (Gibson soil, T
s
= 0.29 s, E
p
/ E
s
= 1000, T
str
= 0.44 s).

Perhaps surprisingly, groups of inclined piles supporting a tall structure attract smaller axial
forces than those of the group with vertical piles (Figure 7) ! This must be attributed to the
disproportionately large overturning moment resisted mainly by axial reactions of the vertical
piles. However, batter piles embedded in relatively stiff soils (small E
P
/E
S
ratio) and supporting
structures with large (fixed-base) fundamental periods may develop larger axial forces than the
vertical piles owing to the larger contribution of the kinematic soil-pile interplay to their seismic
response. In stark contrast, in the case of a short structure, inclined piles develop larger axial
forces than the vertical piles. This maximum, however, occurs at great depths (in the order of 10
pile diameters) and is mainly due to the kinematic interaction of the pile with the soil.
To better understand the differences in the distress of batter piles supporting a tall and a short
structure, we compare the snapshots of displacement of a vertical and a 25
o
batter pile group,
supporting either a short (H
ostr
ostr
= 12 m) structure. Figure 8 portrays the
comparison. Notice that in the case of the batter group supporting a short structure (Figure 8a)
the displacementrotation of the pile cap is out-of-phase with the displacement of the structure.
This is not the case with the tall structure : cap rotation and structure are nearly in-phase.
Figures 9 and 10 attempt to illustrate schematically the mechanisms through which the iner-
tial forces of the superstructure are undertaken. Our goal is to convince that the presented re-
sults are explainable (hence reasonable), and at the same time to develop a deeper understanding
of the problem mechanics.
With the short structure, where the inertial shear force dominates, the vertical group develops
primarily a pair of shear forces, and only secondarily bending moments due to the rotation fixity
at the cap (Figure 9a). Axial forces at the vertical piles are negligible.
By contrast, a horizontal shear force on the cap of a batter pile group results in the develop-
ment of both shear and axial forces in each pile. In fact, since the lateral pile deformation far ex-
ceeds the deformation due to the axial forces, the arrows of the shearforce vectors define quali-
tatively the rotation of the cap (Figure 9b).

45
(a)
(b)
(a)
(b)
Q
Q / 2
N 0
1
Q
1

2
Q
2
Q
(a)
(b)
Q
Q / 2
N 0
1
Q
1

2
Q
2
Q
(a)
(b)

Figure 8. Exaggerated snapshots of the deformed shapes of groups with vertical piles (left-hand side) and
inclined (right-hand side) supporting: (a) short (H
str
= 1 m) and (b) tall (H
str
= 12 m) superstructures in
Gibson soil

Figure 9. Mechanisms for undertaking the inertial forces of a short superstructure for : (a) vertical, and
(b) inclined pile groups. The vectors indicate the forces imposed from the on the piles. The dashed lines
correspond to the location of the cap if the axial displacements of the piles are completely ignored.
46
M
q
f
M
q
f
M
q
f
With the tall structure, where the overturning moment dominates, the vertical piles are sub-
jected to a pair of axial forces that undertake most of this load (Figure 10a). Secondarily, head
moments develop due to pile fixity to the cap. In stark contrast, batter piles may undertake this
large moment mainly largely by flexure [bending] (Figure 10b). As a result, substantial cap
rotations take place. Indeed, the capped batter piles are now rotationally more flexible than the
capped vertical piles! To elucidate why, imagine two piles inclined at very large (certainly un-
realistic) angles (Figure 11) with a large overturning moment applied on the cap : replacing
conceptually the soil with Winkler springs, one can realize that the piles will unavoidably bend,
activating these springs alternately in tension and compression, near the cap, q. The (remaining)
portion of the load, undertaken by the framing action of the piles activates the friction, f, at the
pile-soil interface, would be very small (equal to zero in the extreme [unrealistic, of course] case
of two piles inclined at 90
o
).

Figure 11. Mechanisms for undertaking the inertial forces of a short superstructure for : (a) vertical, and
(b) inclined pile groups. The vectors indicate the forces imposed from the on the piles. The dashed lines
correspond to the location of the cap if the axial displacements of the piles are completely ignored.

Having analysed the mechanics of the pile distress, let us examine the structural distress. Di-
mensionless diagrams for the drift of the superstructure are presented in Figure 12. The
horizontal drift when the structure is supported on a group of batter piles is generally smaller
than with the exclusivelyvertical pilegroup . This reduction of structural distress is
appreciable only with the tall structure, and is attributed to the observed simultaneous
occurrence of smaller horizontal cap displacement, and larger cap rotation with the group of
batter piles. Increased pile batter relates to smaller structural distress and smaller horizontal
displacement of the mass, at the cost of larger cap rotation.
6 CONCLUSIONS
This paper has tried to contribute towards a better understanding of the seismic behavior of bat-
ter piles, which may under certain circumstances be beneficial rather than detrimental, for both
the structure they support and the piles themselves. Admittedly, the linear approximation of the
SSI phenomena is not without shortcomings. Phenomena created by strong nonlinearities of the
soil, such as permanent soil deformations due to extensive soil plastification, residual bending
moments on the piles etc, cannot be captured by linear (or equivalent linear) FE analyses. None-
theless, valuable insight is gained into the description, understanding, and explanation of the be-
havior of batter piles. The main conclusions are as follows :
1) The purely kinematic response of batter piles tends to confirm their negative reputation :
the parametric analyses show that they experience larger bending moments than vertical
piles. Moreover, batter piles exhibit significantly larger axial forces than vertical piles for
all four idealized profiles, due to exclusively horizontal shaking of the soil. In fact both of
these internal forces increase as the angle increases. This conclusion is in full accord with
the conclusion of a recent study by Poulos (2006) who imposed on the piles a lateral static
ground displacement linearly decreasing with depth, and thus not very different in shape
from the first mode free-field displacements that is the main source of kinematic pile
straining in our study.
47
0
o
5
o
10
o
15
o
20
o
25
o
1.4
1.2
1
0.8
0.6
0.4
0.2
0
H
str
= 1 m
H
str
= 12 m m
a
x
U
s
t
r
m
a
x
U
s
t
r
,

f
o
r
v
e
r
t
i
c
a
l

g
r
o
u
p
0
o
5
o
10
o
15
o
20
o
25
o
1.4
1.2
1
0.8
0.6
0.4
0.2
0
H
str
= 1 m
H
str
= 12 m m
a
x
U
s
t
r
m
a
x
U
s
t
r
,

f
o
r
v
e
r
t
i
c
a
l

g
r
o
u
p
0
o
5
o
10
o
15
o
20
o
25
o
1.4
1.2
1
0.8
0.6
0.4
0.2
0
H
str
= 1 m
H
str
= 12 m m
a
x
U
s
t
r
m
a
x
U
s
t
r
,

f
o
r
v
e
r
t
i
c
a
l

g
r
o
u
p
m
a
x
U
s
t
r
m
a
x
U
s
t
r
,

f
o
r
v
e
r
t
i
c
a
l

g
r
o
u
p

Figure 12. Total (kinematic + inertial) response: normalized peak maximum horizontal drift of a short
(H
str
str
= 12 m) structure. Normalization with respect to the response of the group of
vertical piles (maximum, minimum, and average values from the three accelerograms [Lefkada, Aegion,
JMA]). (Gibson soil, T
soil
= 0.29 s, E
p
/ E
s
= 1000, T
str
= 0.44 s).

2) However, the total (kinematic plus inertial) response of structural systems founded on
groups of batter piles offers many reasons for optimism. It has been shown that the role of
batter may be quite beneficial or detrimental depending on the relation between shear force
and overturning moment. Specifically, a tall (slender) structure and a short (squatty) struc-
ture have been selected, as two extreme cases : large moment and small shear charac-
terizes the tall structure; large shear and small moment the short one.
3) For the batter piles supporting a tall slender structure we conclude that :
(a) Configurations with batter piles undergo smaller horizontal displacements than the
vertical group, but at the same time they develop larger cap rotations, often outof
phase with lateral displacements.
(b) The bending moment in batter piles is larger than in vertical piles. In fact, the bending
moment increases as the pile rake increases. Proper reinforcement of the pile-cap
connection is necessary for undertaking safely this bending moment and securing
adequate inelastic deformation in case of an unpredictably large (exceeding the
design) earthquake motion.
(c) Perhaps surprisingly, the symmetric group of batter piles attracts smaller axial forces
than a group of exclusively vertical piles! This is attributed to the disproportionately
large share of the overturning moment resisted by axial reactions in the vertical piles --
- not the case with batter piles which undertake this moment partially with flexure
(bending).
(d) The lateral distortion (and drift) of the structure on batter piles is significantly smaller
than of on vertical piles.
4) For batter piles supporting a short squatty structure :
(a) Embedded in Gibson soil they develop smaller bending moments than vertical piles,
within the range of the considered excitations.
(b) Now batter piles sustain larger axial forces than the vertical group, for two reasons : (i)
kinematic loading, which constitutes an important component of the total loading,
produces larger head moment in the batter piles ; (ii) the inertial loading induces
mostly a dynamic shear force which, while being resisted by lateral loads in vertical
piles, it loads axially (and laterally) the batter piles.
(c) The horizontal drift of the superstructure is less sensitive to pile batter.
48
ACKNOWLEDGEMENT
This work formed part of the EU research project QUAKER which is 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, under contract number : EVG1-CT-2002-00064. We also thank Evan-
gelia Garini for her kind contribution in the preparation of the figures.
REFERENCES
AFPS (1990). Association franaise de Gnie Parasismique. Recommandations AFPS 90. Presses des
Ponts et Chausses, Paris.
Architectural Institute of Japan (1995). Strong motion records of the Hyogo-ken Nanbu Earthquake (in
Japanese).
Banerjee P.K. and Driscoll R.M.C. (1975). A program for the analysis of pile groups of any geometry
subjected to any loading conditions, Department of the Environment, London, HECB/B/7/.
Banerjee P.K. and Driscoll R.M.C. (1976). Three dimensional analysis of raked pile groups, Proceed-
ings of Institution Civil Engineers, Part 2, Vol. 61, 653-671.
Banerjee P.K. and Davies T.G. (1977). Analysis of pile groups embedded in Gibson soil, Proceedings of
9
th
International Conference Soil Mechanics, Tokyo.
Banerjee P.K. and Davies T.G. (1978). The behavior of axially and laterally loaded piles embedded in
non-homogeneous soils, Geotechnique, Vol. 28 (3), 309-326.
Banerjee P.K. and Davies T.G. (1980). Analysis of some reported case histories of laterally loaded pile
groups, Numerical Methdods in Offshore piling. Institution of Civil Engineers, ICE, London, 101-
108.
Bentley K.J. and El Naggar H. M. (2000). Numerical analysis of kinematic response of single piles,
Canadian Geotechnical Journal , Vol. 37, 1368-1382.
Berrill J. B., Chris Tensen S. A., Keenan R. R., Okada W., & Pettiga R. J (2001), Case study of lateral
spreading forces on a piled foundation, Gotechnique, Vol. 51(6), 501-517.
Deng N, Kulesza R., and Ostadan Farhang. (2007). "Seismic soil-pile group interaction analysis of a bat-
tered pile group", 4
th
International Conference on Earthquake Geotechnical Engineering, Thessalo-
niki, June 25-28.
Dobry R. & O Rourke M. J. (1983). Discussion on "Seismic response of end-bearing piles" by Flores-
Berrones R. & Whitman R. V., Journal of Geotechnical Engineering Division, ASCE, Vol. 109.
EC8 (2000). Structures in seismic regions, Part 5: Foundations, retaining structures, and geotechnical
aspects. Seismic Eurocode.
Fan K., Gazetas G., Kaynia A., Kausel E. and Ahmad S. (1991). "Kinematic seismic response of single
piles and pile groups," Journal of Geotechnical Engineering, ASCE, Vol. 117, 1860-1879
Flores-Berrones R. & Whitman R. V. (1982). "Seismic response of end-bearing piles," Journal of Geo-
technical Engineering Division, ASCE, Vol. 108, 554-569.
Gazetas G. (1984). "Seismic response of end-bearing piles Soil Dynamics and Earthquake Engineering,
Vol. 3, 8993.
Gazetas G. (1991). Foundation Vibrations, Foundation Engineering Handbook 2
nd
Edition, H.Y. Fang
(Ed), Kluwer/Springer Publishers, Chapter 15, 553-5938993.
Gazetas G. and Mylonakis G. (1998). "Seismic soil-structure interaction: new evidence and emerging is-
sues," Geotechnical Earthquake Engineering and Soil Dynamics III, ASCE, Vol. 2, 1119-1174.
Gazetas, G., Anastasopoulos I., and Dakoulas P. (2005). Failure of harbor quaywall in the Lefkada 2003
earthquake, Performance Based Design in Earthquake Geotechnical Engineering : Concepts and Re-
search, Proceedings of Geotechnical Earthquake Engineering Satellite Conference, Osaka, 62-69.
Giannakou A. (2007). Seismic behavior of inclined piles. Ph.D Thesis, National Technical University of
Athens.
Guin J. (1997). Advances in soil-pile-structure interaction and non-linear pile behavior, PhD thesis, State
University of New York at Buffalo.
Harada T., Kubo K., and Katayama T. (1981). Dynamic soil-structure interaction by continuum formula-
tion method, Report, Institute of Industrial Science, University of Tokyo, Vol. 29 (5).
Juran I., Benslimane A., and Hanna S (2001). Engineering analysis of dynamic behavior of micropile
systems, Transportation Research Record, No. 1772, 91-106
Kavazanjian E. (2006). A drivenpile andvantage: Batter piles, Piledriver, Q4, 21-25
Kavvadas M. and Gazetas G. (1993). "Kinematic seismic response of free-head piles in layered soil,"
Geotechnique, Vol. 43, 207-222
49
Kaynia A. M. & Kausel E. (1982). Dynamic stiffness and seismic response of pile groups, Research re-
port R82-03, Massachusetts Institute of Technology.
Lam, P. I and Martin, G.R. (1986). Seismic design of highway bridge foundations, Vol. II, Design
procedures and guidelines. Report No. FHWA/RD-86/102, Federal Highway Administration, Vir-
ginia.
Mitchell, D., Tinawi, R., and Sexsmith, R.G. (1991). Performance of bridges in the 1989 Loma Prieta
earthquake : Lessons for Canadian designers. Canadian Journal of Civil Engineering, Vol. 18 (4),
711-734.
Nikolaou S., Mylonakis G., Gazetas G., and Tazoh T. (2001). Kinematic pile bending during earth-
quakes: analysis and field measurements, Geotechnique Vol. 51, 425-440.
Okawa K., Kamei H., Zhang F., and Kimura M. (2005). Seismic performance of group-pile foundation
with inclined steel piles, Proceedings of the 1
st
GreeceJapan Workshop on Seismic Design, Obser-
vation, and Retrofit of Foundations, Athens, 53-60
Poulos H.G. and Madhav M.R. (1971). Analysis of the movement of battered piles, Proceedings of the
1
st
Australia-New Zealand Conference on Geomechanics, Institution of Australian Engineers, Mel-
bourne, 1, 268-275.
Poulos H.G. (1974). Analysis of pile groups subjected to vertical and horizontal loads, Australian Ge-
omechanics Journal, G4, No. 1, 26-32.
Poulos H.G. (1979). Settlement of single piles in non-homogeneous soil, Journal of GeotechnicalEngi-
neering Division, ASCE, Vol. 105(5), 627-641.
Poulos H.G. & Davis E.H. (1980). Pile Foundation Analysis and Design. John Wiley & Sons.
Poulos H.G. (1980). An approach for the analysis of offshore pile groups, Numerical Methods in
Offshore Piling, Institution of civil Engineers. ICE, London, 119-126.
Poulos H.G. (1999). Approximate computer analysis of pile groups subjected to loads and ground
movements. International Journal Numerical Analytical Methods in Geomechanics, Vol. 23, 1021-
1041,
Poulos H. G. (2006). Raked pilesVirtues and drawbacks Journal of Geotechnical and Geoenviron-
mental Engineering, ASCE, Vol. 132, 795-803.
Priestley, N., Singh, J., Youd, T., and Rollins, K. (1991). Costa Rica Earthquake of April 22, 1991
Reconnaissance Report, Pub. 91-02, Earthquake Eng. Research Inst., 59-91.
Randolph M.F. (1981). The response of flexible piles to lateral loading Geotechnique, Vol. 31, 247-
259.
Ravazi S.A, Fahker A., and Mirghaderi S.R. (2007). An insight into the bad reputation of batter piles in
seismic performance of wharves, 4
th
International Conference on Earthquake Geotechnical Engi-
neering, Thessaloniki, K. Pitilakis editor, June 25-28, 2007
Sadek M. & Shahrour I. (2004). "Three-dimensional finite element analysis of the seismic behavior of in-
clined micropiles", Soil Dynamics and Earthquake Engineering, Vol. 24, 473-485.
Sadek M. & Shahrour I. (2006). "Influence of the head and tip connection on the seismic performance of
micropiles," Soil Dynamics and Earthquake Engineering, Vol. 26, 461-468.
Takewaki I. & Kishida A. (2005). "Efficient analysis of pile-group effect on seismic stiffness and strength
design of buildings," Soil Dynamics and Earthquake Engineering, Vol. 25, 355-367
50

1 INTRODUCTION
Current seismic design of bridge structures is based on a presumed ductile response. A capacity
design methodology ensures that regions of inelastic deformation are carefully detailed to pro-
vide adequate structural ductility, without transforming the structure into a mechanism. Brittle
failure modes are suppressed by providing a higher level of strength compared to the corre-
sponding to ductile failure modes. For most bridges, the foundation system may be strategically
designed to remain structurally elastic while the pier is detailed for inelastic deformation and
energy dissipation. Essentially-elastic response of the foundation is usually ensured by increas-
ing the strength of the foundation above that of the bridge pier base so that plastic hinging oc-
curs in the pier instead of the foundation.
The concept of ductility design for foundation elements is still new in earthquake engineering
practice. The potential development of a plastic hinge in the pile is forbidden in existing regula-
tions, codes and specifications. The main reasons are: (i) the location of plastic hinges is not ap-
proachable for post-seismic inspection and repair, (ii) the high cost associated with repair of a
severely damaged foundation, and (iii) failure due to yielding in the pile prior to exceeding soil
capacity is an undesirable failure mechanism, by contrast to that in which soil capacity is mobi-
lized first.
However, several case-histories (especially from the Kobe 1995 earthquake) have shown that:
(a) pile yielding under strong shaking cannot be avoided, especially for piles embedded in soft
soils; and (b) pile integrity checking after an earthquake is a cumbersome, yet feasible task. Fur-
thermore, there are structures where plastic hinging cannot be avoided in members of the foun-
dation during a severe earthquake. A good example of such structure is the pile-column (also
Seismic Response of Bridge Pile-Columns
V.A. Drosos, N. Gerolymos & G. Gazetas
ABSTRACT: While seismic codes do not allow plastic deformation of piles, the Kobe earth-
quake has shown that limited structural yielding and cracking of piles may not be always detri-
mental. As a first attempt to investigate the consequences of pile yielding in the response of a
pilecolumn supported bridge structure, this paper explores the soilpilebridge pier interaction
to seismic loading, with emphasis on structural nonlinearity. The pile-soil interaction is modeled
through distributed nonlinear Winkler-type springs and dashpots. Numerical analysis is per-
formed with a constitutive model (Gerolymos and Gazetas, 2005a; 2005b; 2006a) materialized
in the OpenSees finite element code (Mazzoni et al. 2005) which can simulate: the nonlinear
behaviour of both pile and soil ; the possible separation and gapping between pile and soil; ra-
diation damping ; loss of stiffness and strength in pile and soil. The model is applied to the
analysis of pile-column supported bridge structures, focusing on the influence of soil compli-
ance, intensity of seismic excitation, pile diameter, aboveground height of the pile, and above
or below ground development of plastic hinge, on key performance measures of the pier as is:
the displacement (global) and curvature (local) ductility demands and the maximum drift ratio.
It is shown that kinematic expressions for performance measure parameters may lead to errone-
ous results when soilstructure interaction is considered.
51
known in the American practice as extended pile-shaft), where the column is continued below
the ground level as a pile of the same or somewhat larger diameter. Obviously, the design of
such foundation requires careful consideration of the flexural strength and ductility capacity of
the pile.
An advantage of supporting a column bent on drilled pile is the cost savings associated with the
construction of large cast-in-drilled-hole (CIDH) piles instead of multiple piles of smaller di-
ameter, which must later be integrated into a structural unit using a pile cap. Another advantage
of such a design is that localized damage that could otherwise develop at the columnpile cap
joint is avoided by the pile-column combination, since there is no structural distinction between
the pile and the column other than the presence of a construction joint at the pilecolumn inter-
face.
While the design of piled footings favors forcing plastic hinging into the superstructure, with
the piles remaining elastic, pile-columns can be designed with overall ductile performance in
mind. In case of a single pile-column, formation of a plastic hinge in the pile shaft is the only
mechanism by which ductile performance can be attained. A pile-column bent may first tend to
plastify at the column-beam joint, but the full flexural capacity of the system can only be ob-
tained through the formation of a secondary plastic hinge, below ground surface (at least
slightly below). Bending moment distribution varies with height, but diminishes after attaining a
maximum bending moment below the ground level. A typical depth for maximum bending
moment, and possibly the location of the plastic hinge, ranges from one to three or four pile di-
ameters below ground surface, depending on the above-ground height and soil stiffness.
Damage below ground in the form of cracking or spalling of concrete, fracture of transverse re-
inforcement, or buckling of longitudinal reinforcement is generally difficult to assess after an
earthquake. This, coupled with the potential high cost of repair, resulted in the current use of a
design displacement ductility factor that is smaller than that of columns in order to limit the
amount of yielding in the pile below the ground level. For example, in the United States, ATC-
32 (ATC 1996) prescribes a displacement ductility factor of 3 for pile-columns compared to a
displacement ductility factor of 4 for well-confined fixed-base reinforced concrete columns. A
similar approach of prescribing higher lateral strength for piles has been adopted for seismic de-
sign of highway bridges in New Zealand. For plastic hinging that may develop at a depth less
than 2 m below the ground level, but not below the mean water level, the design displacement
ductility factor is limited to no more than 4. For plastic hinging at a depth greater than 2 m be-
low the ground level or below the mean water level, the design displacement ductility factor is
reduced to no more than 3 (Chapman 1995, Park 1998).
In this paper, a parametric investigation of the nonlinear inelastic response of pile-column
bridge systems is conducted, and the influence of pile inelastic behavior and soilstructure in-
teraction on structure ductility demand is identified. The role of various key parameters are ex-
amined, such as: (a) soil compliance, (b) above-ground height of the column shaft, (c) pile di-
ameter, (d) intensity of the input seismic motion, and (e) location of the plastic hinge, on
characteristic performance measures of the soil-structure system response, such as: the dis-
placement (global),
, and curvature (local),
, ductility demands and the maximum drift ra-

tio max. It is shown that: (a) neglecting the consideration of the soil-structure interaction ef-
fects may lead to unconservative estimates of the actual seismic demand, and (b) the
development of a plastic hinge along the pile (for instance for cases that the pile is designed
with inferior or equal strength compared to that of the pier) is beneficial for the pier response.
2 THE PROBLEM EQUATIONS AND PARAMETERS
2.1 Definition of the problem
The studied problem is sketched in Figure 1: a pile-column embedded in clay or sand deposit,
monolithically connected to the bridge deck is excited by a seismic motion. It is assumed that
the transverse response of the bridge structure may be characterized by the response of a single
bent, as would be the case for a regular bridge with coherent ground shaking applied to all
bents.
The height of the pier H is given parametrically the values of 5 and 10 m, so that a typical urban
bridge and a rather short viaduct, in respect, are examined. The diameter b of the pile-column
52
above-ground takes values of 1.5 and 3.0 m. However, to investigate the influence of the plastic
hinge position on the system response, two more cases are examined: the below-ground pile-
column diameter d is increased by 33 % relatively to the above-ground diameter b. So, for pile
diameters d = 1.5, 2.0, 3.0, and 4.0 m, pier diameter equals to b = 1.5, 1.5, 3.0, and 3.0 m, re-
spectively. For sake of simplicity, the term diameter will refer from this point on, to the below-
ground diameter d. The embedment length of the pile L is considered in every case equal to 30
m. In total, a set of four structural configurations are analyzed.
The mass of the deck is calculated so that the fundamental period of the fixed-base pier would
be T = 0.3 sec for all cases studied. This restriction for the fixed-base period leads to a mass of
45 Mg for the pile diameter of d = 1.5 m, and 720 Mg for that of d = 3.0 m. The nonlinear be-
havior of the pile-column is characterized through the predefined momentcurvature relations
illustrated in Figure 2. These curves have been obtained with the BWGG model (Gerolymos &
Gazetas 2005b), discussed in the sequel, for n = 1, initial stiffness equal to the uncracked flex-
ural stiffness EI of the pile-column, and ultimate strength equal to the conventionally calculated
moment at the ground surface considering that a critical acceleration of 0.2 g is applied on the
deck mass. In the case of the variable-diameter piers, the bending moment capacity of the pile
cross-sections is calculated to be proportional to the square power of the cross-section diameter
d2, which is a reasonable assumption for a given detailing of reinforcement. In that way, the po-
tential development of a plastic is forced to occur in the above-ground portion of the pile-
column.
It is noted that the objective of the parametric study described herein is to investigate the seis-
mic response of the system in the inelastic regime and not to design the structure. Therefore, (a)
we are mainly concerned about achieving equivalence of the studied systems in the framework
of nonlinear response analysis without considering soil-structure interaction effects, rather than
about reinforcement details that correspond to the utilized momentcurvature curves. And (b)
the critical acceleration was scaled to 0.2 g, to ensure that the system will enter the inelastic re-
gime under the used seismic excitation.

Figure 1. The problem investigated and the two types of presumed soil deposits
2.2 Constitutive equations and numerical modeling
The developed BWGG model is a versatile one-dimensional actionreaction relationship, capa-
ble of reproducing an almost endless variety of stressstrain or forcedisplacement or moment
rotation relations, monotonic as well as cyclic. It is being applied here to model the monotonic
and cyclic response of piles, expressing both the py and momentcurvature relationships. A
simple version of the model is outlined below. More details can be found in Gerolymos & Ga-
zetas (2005a, 2005b, 2006a, 2006b), although the model utilized here is a slightly im-
proved/simplified version of the model in the latter reference.
53

Figure 2. Predefined momentcurvature relations used in the analyses

The lateral soil reaction against a deflecting pile is expressed as the sum of an elastic and an ine-
lastic component according to:
( ) 1
x s s s y s
p k y p = + (1)
where s is a dimensionless inelastic soil parameter expressed in the following differential
form:
( )
{ }
0
1
x
n
s s s
s s s s
d h
b g sign dy
dy y

= +

(2)
where p
x
is the resultant (in the direction of loading) of the normal and shear stresses along the
perimeter of a pile segment of unit length and it includes both in-phase and out-of-phase
components ; the latter reflects radiation and hysteretic damping in the soil. y is the pile deflec-
tion at the location of the spring; k
s
is a reference spring stiffness ; s is a parameter governing
the post yielding stiffness ; p
y
is a characteristic value of the soil reaction related to the initiation
of significant inelasticity (yielding) ; y
0
is a characteristic value of pile deflection related to the
initiation of yielding in soil reaction. n
s
, b
s
and g
s
, are dimensionless quantities that control the
shape of the hysteretic soil reaction pile deflection loop, and
s
, r
s
and h
s
are strain hardening
parameters for stiffness decay, strength degradation, and pinching behavior, respectively ; where
c
s
is the damping coefficient at small amplitude motions, and c
sd
is a viscoplastic parameter
which controls the coupling of soil and soilpile interface nonlinearity with radiation damping
The reader is referred to Gerolymos and Gazetas (2005 a,b) for more details.
The inelastic behavior of the pile is similarly expressed in terms of a strength-of-materials-type
bending momentpile curvature relation, which includes an elastic and an inelastic component:

p y p p p p
M a
z
y
I E M ) 1 (
2
2
+
=
(3)
where E
p
I
p
is the initial (elastic) bending stiffness (also called flexural rigidity),
p
is a parame-
ter controlling the post yielding bending stiffness, My is the value of bending moment that initi-
ates structural yielding in the pile, and
p
is the hysteretic dimensionless parameter which con-
trols the nonlinear structural response of the pile. The latter is governed by
54
( )
{ }
0
1
p
n
p p p
p p p p
d h
b g sign d
d

= +

(4)
where is the pile curvature, and b
p
, g
p
, n
p
,
p
, r
p
, and h
p
, are dimensionless quantities that con-
trol the shape of the hysteretic bending momentcurvature loop in the same manner as n
s
, b
s
, g
s
,
p
, r
p
, and h
p
, control the shape of the lateral soil reactiondeflection loop.
0
is the value of pile
curvature at initiation of yielding in the pile.
Evidently, Equations 3 4 are of the same form as Equations 1 2, except that no viscous term
(radiation damping) is included in the structural pile response.
The seismic response of the soilpilestructure system is investigated herein via a beam-on-
nonlinear-Winkler-foundation (BNWF) finite element model developed in OpenSees (Fig. 3).
The pile-column is discretized into nonlinear beam elements with length 0.5 to 1.0 m, whose
bending behavior is governed by the macroscopic constitutive BWGG model. The mass of the
deck is simulated as a concentrated mass at the top node of the pile-column, whereas the dis-
tributed mass of the extended pile is simulated by lumped masses on beam-element nodes.
The near-field soilpile interface is simulated with nonlinear py spring elements, the behavior
of which is described also by the BWGG model. Model parameters were appropriately cali-
brated to match the py curves of Reese et al. (1974) and Matlock (1970). The free extremities
of the soil springs were excited by the acceleration time histories obtained at each depth from
the free-field seismic response analysis (Banerjee et al. 1987).
Although the developed finite element model has the capability to reproduce higher order phe-
nomena (e.g. P effects), such phenomena were ignored, considering that their strong depend-
ence on the mass of the structure and the geometry would obscure the role of other parameters
(e.g. structural inelasticity and soil compliance).

Figure 3. Schematic illustration of the model used for the analyses
2.3 Soil parameters
The influence of near-field soil compliance on the seismic response of the soilpilestructure
system is investigated parametrically considering four different homogeneous soil profiles (Fig.
1): (a) sand with friction angle = 30
o
, (b) sand with friction angle = 40
o
, (c) clay with
undrained shear strength S
u
= 40 kPa, and (d) clay with undrained shear strength S
u
= 200 kPa.
The small-amplitude stiffness k (= p
y
/ y
0
) was obtained from the available beam-on-dynamic-
Winkler-Foundation solutions (e.g. Gazetas & Dobry 1984, Makris & Gazetas 1992) in terms of
the Youngs modulus of the soil.
For piles in cohesive soils the ultimate soil reaction per unit length of pile can be approximated
by the well known expression
d S P
u y 1
=
(5)
where S
u
is the soil undrained shear strength, and
1
varies from 9 to 12, depending on the fric-
tion ratio f
s
/ S
u
at the pilesoil interface. A value of
1
= 9 is often used for a soft clay, while
1

55
= 11 is more appropriate for a stiff clay. At shallow depths, the plane strain assumption of Equa-
tion 8 is inappropriate because of the nonzero vertical deformation of the soil during lateral
motion of the pile. The following formulation has been proposed for P
y
near the surface
(Matlock 1970)
( )
J
S
d
d
z d S
d
z
J
S
P
u
s
u
u
v
y
+
<
+ + =
,
2 1
,
2
,
(6)
where
'
v
is the vertical effective stress, and
s
the effective specific weight of the soil, and
2

and J are a dimensionless quantities. Broms (1964) proposed a value of
2
= 2, whereas Matlock
(1970) used
2
=3. Matlock (1970) stated that the value of J was determined experimentally to
be 0.5 for a soft clay and about 0.25 for a medium clay, whereas Reese (1975) suggested a value
of J = 2.83 for every type of clay. For piles embedded in cohesionless soils, Broms (1964) pro-
posed an analytical expression for the ultimate soil reaction :
z d P
s y

+ =

2
45 tan 3
2 ,

(7)
where is the angle of friction. Equation 7 is very often preferred in practice among other more
rigorous expressions for its simplicity and compatibility with experimental results.
For the description of the nonlinear behavior of the near-field soil the well-known py relations
of Reese et al. (1974) and Matlock (1970) are used for sand and clay, respectively.
2.4 Soil Profiles, Seismic Excitations, and Site Response Analysis
The influence of soil amplification on the seismic response of the soilpilestructure system is
not examined, mainly for two reasons: (a) a thorough investigation of seismic ground response
is out of scope of this paper, and (b) the unavoidable differences in free-field motions from the
soil response analysis of the four different soil profiles, would complicate the comprehension of
the related phenomena. Therefore, a single soil profile was selected for ground response analy-
sis (Fig. 4): a category C profile, according to NEHRP (1994). Bedrock was assumed to be at 50
m depth.

Figure 4. Shear wave velocity distribution of the adopted soil deposit used for the wave propagation
analysis

The influence of shaking on the seismic response is investigated by selecting three real accelera-
tion records as seismic excitations:
the record from Aegion earthquake (1995),
the record from Lefkada earthquake (2003), and
the JMA record from Kobe earthquake (1995).
56
The first two records are representative strong motions of the seismic environment of Greece,
with one and many cycles, respectively. JMA record is used to investigate the dynamic response
of the soilpilestructure system to a quite unfavorable incident. The dominant periods of the
acceleration time histories for the aforementioned three earthquake records range from 0.2 to
0.8 s, resulting in a fixed base fundamental period ratio (designated as the fixed base fundamen-
tal period of the superstructure divided by the predominant period of the free-field surface ac-
celeration time history) which ranges from 0.66 to 2.67. This is a wide range of values which
ensures generalization of the results presented herein. Near-fault effects such as rupture
directivity and fling (Gerolymos et al. 2007) are also captured by the utilized accelerograms.
All the records were first scaled to a PGA of 0.5 g and 0.8 g at the ground surface; then through
deconvolution analyses conducted with SHAKE (Schnabel et al. 1972), the bedrock motion as
well as the motion at various depths along the pile, were estimated. The ground motion profiles
obtained from SHAKE analyses are then used as input motion in the developed BNWF model.
The acceleration time histories at the surface and the corresponding elastic response spectra
scaled to a SA = 0.8 g (T = 0 s) for 5 % damping, are presented in Figure 5.
It should be stated here in that from a seismological point of view, simply scaling an accelera-
tion time history to a large PGA value for representing the severity of an earthquake might not
be always correct. It is well known from the literature that high peak ground accelerations are
usually accompanied by a large number of predominant cycles. Obviously, this is not the case
for Aegion record which can be satisfactorily approximated by a single sinusoidal pulse.

Figure 5. Real acceleration time histories used as seismic excitation, after scaling to a peak ground accel-
eration of a
g
= 0.5 and 0.8 g, and corresponding ( = 5 %) response spectra scaled to S
a
(T = 0 s) = 0.8 g.
2.5 Analysis Methodology and Performance Measure Parameters
Besides the fundamental response amounts (acceleration, displacement, moments, etc.) that de-
scribe the behavior of a structure under dynamic loading, other important seismic performance
measures are the local and global ductility demand
and
, and the maximum drift ratio

max
.
The local (curvature) ductility demand
is defined as the maximum curvature

max
imposed on
the structure by an earthquake, divided by the yield curvature
y
, which is a property of the pile-
column cross-section.
y
max
=
(8)
57
For bridge structures supported on extended piles, the local ductility demand imposed on the
pile shaft might govern the design of the system, because damage to the pile (such as spalling of
cover concrete, crack widths, potential for buckling or fracture of longitudinal reinforcement) is
related to the local curvature ductility.
The following procedure is followed for the assessment of local curvature ductility demand in
the analyses conducted. The momentcurvature curve of each pile-column cross-section is ap-
proximated by a bilinear elasticperfectly plastic relation, in which the first (linear) section is
defined as the secant stiffness through the first-yield point
f y
(yielding of first longitudinal rein-
forcement bar) and the second section by the tangent line on the post-yielding section of the ac-
tual momentcurvature curve. The intersection of these two lines defines the cross-section yield
curvature
y
(Fig. 6).

Figure 6. Definition of yield curvature of the soil-pile-structure system

Similarly, the global (displacement) ductility demand
is the ratio of the maximum displace-

ment of the system u
max
, imposed by an earthquake, to the yield displacement u
y
, which is a
soilpilestructure system property.
y
u
u
max
=
(9)
The yield displacement u
y
is assessed through static nonlinear analyses (push-over analyses) ac-
cording to the following procedure:
At the center of mass of the superstructure, a horizontal force is gradually applied. The maxi-
mum displacement and the curvature along the pile-column are continuously monitored. The
displacement measured, when the pile curvature reaches the first-yield point
f y
, is defined as
the first-yield displacement u
f y
. Then, similarly to the procedure followed for the determination
of yield curvature, the loaddisplacement curve is approximated by an equivalent bilinear elas-
ticperfectly plastic curve, in which the first (linear) section is defined as the secant stiffness
through the first-yield point u
f y
and the second section by the tangent line on the post-yielding
section of the loaddisplacement curve. The intersection of these two lines defines the yield dis-
placement u
y
(Fig. 7).

Figure 7. Definition of yield displacement of the soil-pile-structure system
58
It has to be noticed, that for the estimation of pile curvature, we did not use the FEM original
curvature results as these showed mesh sensitivity. Instead, plastic rotation results which are
mesh insensitive, were used and divided by the plastic hinge length L
p
to derive pile curvature.
The length of plastic hinge L
p
for the pile-columns was estimated according to Budek et al.
(2000) approximation:
H d L
p
+ = 06 . 0
(10)
where d is the pile diameter and H the above-ground height. Similar expressions, based however
on different assumptions, have also been provided in Caltrans (1986,1990), Dowrick (1987),
Priestley et al. (1996), Chai (2002), and Chai & Hutchinson (2002).
The drift ratio is defined as the maximum displacement of the deck imposed by an earthquake
relative to pier base displacement divided by the height of the pier:

H
u u
base pier deck
=
max max
(11)
3 ANALYSIS RESULTS AND DISCUSSION
In this paragraph, typical results of the nonlinear analyses are presented in terms of acceleration
time-histories; peak bending moment, curvature and displacement distributions.
Results of the seismic response of the examined structural systems are presented in Figures 8
26. The acceleration time histories calculated for a pile-column of diameter d = 3 m and height
H = 5 m are presented in Figure 8a for every soil profile examined. The response of the deck is
quite smaller in case of soft clay. On the contrary, stiff foundation soils lead to increased re-
sponse of deck. An exception to this is the increased structural (deck) response for the case of
loose sand. A possible explanation is that the lateral confinement providing by the soil might be
considerable even for small values of internal friction angles of the soil, thus stiffening the re-
sponse of the pile.

Figure 8. Acceleration time histories of the deck for free-head pile-columns with above-ground height H
= 5 m. Left diagram: pile with diameter b = d = 3.0 m. Right diagram: pile founded in soft clay (excitation
at ground surface: JMA, Kobe 1995 a
g
= 0.5 g)
59
Similarly, differences are observed in Figure 8b, where the acceleration time histories of a 5 m
high pier embedded in soft clay are illustrated for different pile diameters. Smaller pile-column
diameter leads to higher deck acceleration. The small pile-column diameters (d = 1.5, 2.0 m)
correspond to a deck mass which is 16 times smaller than that for the larger pile-column diame-
ters (d = 3.0, 4.0 m). This substantial difference in deck masses is responsible for an also large
discrepancy in the fundamental natural periods of the pierfoundationsoil system (effective pe-
riods). Indeed, as can be hardly seen in the response acceleration time histories after the input
motion has subsided, at t = 15 s (free response), the effective period of the small-diameter
bridge columns is approximately 1.0 s while that of the large-diameter bridge columns is about
2.0 s. It is therefore shown that the effective period of the pier systems increases with increasing
mass of deck, despite that the fixed-base period is held constant. A similar trend in the response
of bridge-piers supported either by a single pile or by a group of piles has been also shown in
Gerolymos et al. (1998).
Furthermore, a small decrease in peak acceleration values is observed in constant-diameter ex-
tended pile (d = 1.5, 3.0 m) compared to the response of variable-diameter systems (d = 2.0, 4.0
m). The response of the constant-diameter systems is associated with more intense pile and soil
inelasticity compared to the response of variable-diameter systems. The ample mobilization of
soil plastification and structural yielding mechanisms in the case of constant-diameter piers re-
sults in an increase of both hysteretic and radiation damping, which in turn slows down the deck
response.
The distributions of maximum displacements with depth for several cases are presented in Fig-
ure 9 for the JMA record as seismic excitation at ground surface. The influence of soil in this
distribution is depicted in Figure 9a: the softer the soil, the larger the maximum displacement
within the soil. The increased compliance of the soft soils implies significant deflection of the
pile. In Figure 9b, the role of pile diameter on the system response is illustrated. Larger pile-
diameter piers (d = 3.0 and 4.0 m) are associated with larger effective periods compared to those
of smaller pile-diameter (d = 1.5 and 2.0 m). Given the response spectra of the JMA accelera-
tion time history, this means that the larger pile-diameter systems exhibit larger displacements
and smaller response accelerations compared to the piers with smaller pile diameters (see also
Fig. 8).

Figure 9. Maximum displacement distributions for pile-columns with above-ground height H = 10 m. Left
diagram: pile with diameter b = d = 3.0 m. Right diagram: pile founded in soft clay (excitation at ground
surface: JMA, Kobe 1995 a
g
= 0.5 g)

The distributions of maximum bending moments with depth for several cases are presented in
Figures 1011. The influence of soil type in this distribution is depicted in Figure 10. Like in
displacement distributions, soft soils result to increased pile effective lengths. With a maximum
60
moment of 1200 kNm (yield moment), the depth where this value becomes half increases from
4 m, in case of stiff clay and dense sand, to 6 m for loose sand, and to 10 m in case of soft clay.
It is interesting to observe in Figure 12 that the maximum bending moments among the four
cases are almost equal, indirectly revealing the formation of plastic hinges at certain locations of
the piles. Given the severity of the ground motion (the JMA record scaled to a peak ground ac-
celeration of 0.8g), mobilization of the full bending moment resistance of the piles is unavoid-
able irrespectively of the type of foundation soil.
As shown in Figure 11, the increase of pile diameter results to shifting of the maximum-moment
depth to greater depths. It is noticed, that the maximum-moment depth does not always coincide
with plastic hinge position, due to difference in pile and pier diameters.

Figure 10. Maximum bending moment distributions for free-head pile-columns with above-ground height
H = 10 m and diameters b = d = 1.5 m (excitation at ground surface: JMA, Kobe 1995 a
g
= 0.8 g)

Figure 11. Maximum bending moments distributions for free-head pile-columns with above-ground
height H = 10 m founded in stiff clay (excitation at ground surface: JMA, Kobe 1995 a
g
= 0.8 g)
61
The position of plastic hinge is easily assessed via distributions of pile curvature with depth. In
Figure 12a, the distributions of curvature are presented for a 3-m diameter pile-column with
above-ground height H = 10 m, embedded in different soils. It is observed that the plastic hinge
is developed within 2 diameters below ground surface. The amplitude of curvature increases
significantly in stiffer soils. The effect of pile diameter is shown in Figure 12b. In case of pile-
columns with pier of smaller diameter than that of the pile (d = 2.0, 4.0 m), the pier is highly
stressed and the plastic hinge is formed at its base. Below the ground surface, curvature values
decrease rapidly. On the other hand, constant-diameter pile-columns (d = 1.5, 3.0 m) may de-
velop plastic hinge below surface. In every case, however, plastic rotations are distributed in
greater length with consequent decrease of the maximum curvature.

Figure 12. Maximum curvature distributions for free-head pile-columns with above-ground height H = 10
m. Left diagram: pile with diameter d = 3.0 m (excitation at ground surface: JMA, Kobe 1995 a
g
= 0.5
g). Right diagram: pile founded in soft clay (excitation at ground surface: JMA, Kobe 1995 a
g
= 0.8 g)

In Figure 13, the correlation of the local curvature ductility demand to the global displacement
ductility demand is presented. All the analyses resulted to nonlinear behavior of the extended
pile shaft (
> 1) are depicted categorized according to the foundation soil. The mean ratio (

1) / (
1) equals to 5.4 for soft clay, 3.4 for loose sand, 2.6 for dense sand, and 2.7 for stiff
clay. Similar results have been also obtained by Hutchinson et al. (2004). At first sight, it seems
that founding pile-columns in soft soils is unfavorable: for a given earthquake imposed global
displacement ductility, the local curvature ductility demand is higher than the one corresponds
to stiffer soils. This impression, as will be revealed later on, may be deceptive.

Figure 13. Correlation of local and global ductility demands for different soil types
62
A similar trend appears in Figure 14 where analyses results have been categorized according to
the potential location of plastic hinge. For constant-diameter pile-columns the plastic hinge is
likely developed below the ground surface (on pile) whereas for variable-diameter pile
columns, plastic hinges are developed at the base of pier. The average ratio (
1) / (
1)
takes a value of 3.5 for plastic hinge on the pile, and 2.7 for plastic hinge on the pier. The results
discourage the inelastic design of pile, however, the picture is yet to be cleared.
In the next figures (Figs. 15 16), analyses results have been grouped according to pier diame-
ter and shaking intensity, in respect. A slight predominance of the larger pier (d = 3.0 m) is ob-
served as the average value of (
1) / (
1) ratio is 3.3 instead of 3.7 in case of smaller pier

(d = 1.5 m). As expected, the shaking amplitude does not affect the value of ductility demand
ratio.

Figure 14. Correlation of local and global ductility demands for different plastic hinge locations

Figure 15. Correlation of local and global ductility demands for different diameters

Figure 16. Correlation of local and global ductility demands for different seismic motion amplitudes
63
The influence of above-ground height H on the correlation between local and global ductility
demand is illustrated in Figure 17. Shorter piers exhibit greater local curvature ductility demand
for a given displacement ductility level.

Figure 17. Correlation of local and global ductility demands for different above-ground heights

In Figure 18, the correlation of local curvature ductility demand to the maximum drift ratio is
presented for all the soil profiles examined. For a given maximum drift ratio, the required curva-
ture ductility is greater for stiffer soils. The depth of the plastic hinge location increases with
decreasing soil stiffness resulting in larger rigid body displacement, which however is not asso-
ciated with strain in the pier. An inversion in the trend observed earlier is evident.
The same trend is observed in Figure 19, where the effect of plastic hinge location is examined:
for a given maximum drift ratio, the required curvature ductility is greater when the pier is plas-
ticized. Indeed, the rigid body motion component of the displacement, which increases with in-
creasing depth of plastic hinge location, does not produce any structural damage and hence does
not affect the ductility demand on the pier.
The pile-column diameter and the amplitude of ground motion only slightly affect the
/
max

ratio (Figs. 20 and 21). The influence of above-ground height H on the correlation between local
ductility demand and maximum drift ratio is illustrated in Figure 22. Taller piers exhibit greater
local curvature ductility demand for a given drift ratio. Indeed, for a given drift ratio the differ-
ential horizontal displacement between deck and pier base decreases with decreasing above-
ground height of the pier, leading to smaller pier distress and thus to smaller ductility demand.

Figure 18. Correlation of local ductility demand and maximum drift ratio for different soil types
64

Figure 19. Correlation of local ductility demand and maximum drift ratio for different plastic hinge loca-
tions

Figure 20. Correlation of local ductility demand and maximum drift ratio for different diameters

Figure 21. Correlation of local ductility demand and maximum drift ratio for different seismic motion
amplitudes
65

Figure 22. Correlation of local ductility demand and maximum drift ratio for different above-ground
heights

In Figures 2326, the mean and peak values of the factors
, and
max
are illustrated for vari-
ous parameters examined. It is clearly observed that the mean and maximum values of both

and
factors are lower for soft soils and plasticized piles. This phenomenon discredits the trend
appeared in Figures 13 14 and reveals the beneficial influence of soil compliance and pile ine-
lasticity on the response of the structure examined. The apparent paradox stems from the fact
that kinematic expressions do not distinguish between capacity and demand, as also stated in
Mylonakis et al. (2000). For example, according to Figure 13, for a given displacement ductility
demand the curvature ductility capacity of a pile-column embedded in soft soil needs to be lar-
ger than that of a pile-column embedded in stiff soil. However, this does not mean that for a
given seismic excitation both pile-columns would exhibit the same displacement ductility.

Figure 23. Variation of local curvature ductility (
) demand for different parameters examined

66

Figure 24. Variation of global curvature ductility (
) demand for different parameters examined

Although the ratio (
1) / (
1) may take higher values for soft soils, the absolute values of
are small and so are the values of
. The maximum drift ratio

max
seems to stay insensitive
to parameters like soil stiffness and location of plastic hinges (Fig. 25). On the contrary, it de-
pends strongly on the intensity of the seismic excitation. Increase of the above-ground height, as
shown in Figure 26, causes increase in the mean values of
and decrease in the mean value of
max
, whereas mean

value slightly increases, if not remain constant.
4 CONJECTURES
From the analysis of the results of the exploratory parametric analyses conducted herein, the fol-
lowing conclusions could be drawn:
For a given global (displacement) ductility demand
(Mu),
the local (curvature) ductility demand
increases for increased soil compliance.

the potential formation of plastic hinge below ground surface also increases the local (curva-
ture) ductility demand
(M).
the curvature ductility demand slightly decreases with increasing pile diameter.
the curvature ductility demand increases in case of column-piles with smaller above-ground
height ratios (d / H).
The opposite trends for the local ductility demand
are observed, when the maximum drift ra-

tio
max
is kept constant.
However, the conclusions above do not reveal the true nature of the problem and the following
remarks should be considered:
67

Figure 25. Variation of maximum drift ratio (max : %) for different parameters examined

Figure 26. Variation of performance measures for different above-ground heights
68

For a given earthquake, the global displacement ductility demand
decreases as the soil

compliance increases. Thus, while (
1) / (
1) ratio has a higher value for a soft soil,

the small
demand may refrain the local ductility demand
at levels lower than what cor-

responds to a stiffer soil.
The same comment holds for the location of plastic hinge. The potential of plastic hinge de-
velopment on the pile (i.e. below ground surface) reduces
demand, with consequent reduc-

tion of local ductility demand.
Most of the available relations for the performance measures in literature are functions of struc-
ture geometry and reinforcement details only. However, from the results presented in this paper,
the need for modification of these expressions in order to include soil-compliance and pile-
plastification effects on structure dynamic response is demonstrated. Some very early, improved
correlations are proposed herein.

Nevertheless, it has to be noted that ductility capacity required in a structure does not always co-
incides with ductility demand which depends on the characteristics of the seismic loading and
inelasticity of soil-pile-structure system. Thus, a structure with higher required ductility capacity
may experience lower developed ductility than another structure with lower ductility capacity
requirements. The actual ductility demands of a structure can be assessed accurately exclu-
sively within the framework of a nonlinear dynamic analysis, in which the influence of soil
properties and excitation characteristics are parametrically investigated.
REFERENCES
ATC-32 Applied Technology Council 1996. Improved seismic design criteria for California bridges:
provisional recommendations, Redwood City, California
Banerjee, S., Stanton, J. F., and Hawkins, N. M. 1987. Seismic performance of precast concrete bridge
piles, Journal of Structural Engineering, 113(2): 381396
Broms, B. 1964a. Lateral resistance of piles in cohesive soils, Journal of Soil Mechanics and Foundation
Division, ASCE, 90(3): 27-63
Broms, B. 1964b. Lateral resistance of piles in cohesionless soils, Journal of Soil Mechanics and Founda-
tion Division, ASCE, 90(3): 123-156
Budek, A.M., Priestley, M. J. N., and Benzoni, G. 2000. Inelastic seismic response of bridge drilled-shaft
RC pile/columns, Journal of Geotechnical and Geoenvironmental Engineering, 126(4): 510517
Caltrans 1986. Bridge Design Specifications manual, Caltrans, Sacramento, California
Caltrans 1990. Bridge Design Specifications / Seismic Design References, Caltrans, Sacramento, Califor-
nia
Chai, Y.H. 2002. Flexural strength and ductility of extended pile-shafts. I: Analytical model, Journal of
Structural Engineering, 128(5): 586594
Chai, Y.H., and Hutchinson, T.C. 2002. Flexural strength and ductility of extended pileshafts. II: ex-
perimental study, Journal of Geotechnical and Geoenvironmental Engineering, 128(5): 595602
Chapman, H. E. 1995. Earthquake resistant bridges and associated highway structures: current New Zea-
land practice, Proc., National Seismic Conference on Bridges and Highways, San Diego, California
Dowrick, D. J. 1987. Earthquake resistant design, 2nd edition, Wiley-Interscience, New York
Gerolymos N., Gazetas G., Mylonakis G. 1998. Fundamental Period and Effective Damping of Pile
Supported Bridge Piers, Eleventh European Conference on Earthquake Engineering, Paris, in CD
Rom.
Gerolymos N., Apostolou M., Gazetas G. 2005. Neural network analysis of the overturning response un-
der nearfault type excitation, Earthquake Engineering and Engineering Vibration, 4(2): 213-228.
Gerolymos N., Gazetas G. 2005a. Constitutive model for 1D cyclic soil behavior applied to seismic
analysis of layered deposits, Soils and Foundations, 45(3): 147-159.
Gerolymos N., Gazetas G. 2005b. Phenomenological model applied to inelastic response of soilpile in-
teraction systems, Soils and Foundations, Vol 45(4): 119-132.
Gerolymos N., Gazetas G. 2006a. Development of Winkler Model for Static and Dynamic Response of
Caisson Foundations with Soil and Interface Nonlinearities, Soil Dynamics and Earthquake Engineer-
ing, 26(5): 363-376.
Gerolymos N., Gazetas G. 2006b. Static and Dynamic Response of Massive Caisson Foundations with
Soil and Interface NonlinearitiesValidation and Results, Soil Dynamics and Earthquake Engineering,
26(5): 377-394.
69
Hutchinson, T.C., Boulanger, R.W., Chai, Y.H., and Idriss, I. M. 2004. Inelastic seismic response of ex-
tended pile shaft supported bridge structures, Report PEER 2002/14, Pacific Earthquake Engineering
Research Center, University of California, Berkeley
Mazzoni, S., McKenna, F., and Fenves, G.L. 2005. OpenSees command language manual, The Regents
of the University of California, 375 p.
Matlock, H. 1970. Correlations for design of laterally loaded piles in soft clay, Proc., 2nd Annual Off-
shore Technology Conference, Houston, Texas. 577-594.
Mylonakis G., Gazetas G., Gerolymos N., and Anastasopoulos I. 2000. Detrimental Role of Soil-Structure
Interaction and the Collapse of the 18-Pier Fukae Bridge in Kobe, Recent Advances in Applied Me-
chanics, Honorary Volume for Professor A.N.Kounadis, N.T.U.A., 2000, 145-159.
NEHRP 1994. Recommended provisions for seismic regulations of new buildings: Part 1, provisions.
FEMA 222A, National Earthquake Hazard Reduction Program, Federal Emergency Management
Agency, Washington, D.C.
Park, R. 1998. New Zealand practice on the design of bridges for earthquake resistance, Proc., 1st Struc-
tural Engineers World Congress, San Francisco, Elsevier Science
Priestley, M. J. N., Seible, F., and Calvi, G. M. 1996. Seismic design and retrofit of bridges, Wiley-
Interscience, New York
Reese, L.C., Cox, W.R., and Koop, F.D. 1974. Analysis of laterally loaded piles in sand. Proc., 6th An-
nual Offshore Technology Conference, Houston, Texas, 473485
Schnabel, P. B., J. Lysmer and H. B. Seed 1972. SHAKE- A Computer Program for Earthquake Response
Analysis of Horizontally Layered Sites, EERC 72-12, Univ. of Calif., Berkeley.

70
1 INTRODUCTION
Performance-based design approach has been promoted for design codes for a structure to be
newly constructed. Seismic evaluation for existing structures is expected to adopt performance-
based approach as well. Considering that the amount of existing structures is much greater than
that of newly constructed ones, it needs easier and more cost-effective measures, for an exam-
ple, like an evaluation chart.
Seismic actions on a pile foundation are classified to inertial forces on a structure and to a lat-
eral soil displacement acting on piles. In particular, the soil displacement has been newly intro-
duced into some of revised design codes as a seismic action for designing a pile foundation. It is
considered not to have been taken into account in the design of existing structures. Thus, the
seismic evaluation considering the effect of lateral soil displacements on existing pile founda-
tions is anticipated to be conducted early.
Aiming at development of a simplified method for the seismic evaluation of existing pile
foundations (Mori, Suga & Akaishi, 2009), validity of equal energy assumption (EEA) for esti-
mation of ductility factor response of a pile is studied and discussed in this paper through com-
parison between results of fully nonlinear dynamic response analyses and evaluation results
based on EEA.
2 EQUAL ENERGY ASSUMPTION
2.1 Concept of equal energy assumption
Equal energy assumption was originally proposed for the estimation of nonlinear dynamic re-
sponse of a single-degree-of-freedom (SDOF) system by Newmark & Rosenblueth (1971). On
the basis of this concept, it is assumed that the maximum response of strain energy of a nonlin-
ear inelastic pile of a structural system is about equal to the maximum response of strain energy
Efficiency of Equal Energy Assumption for Evaluating Ductility
Factors of a Pile

S. Mori
Ehime University, Matsuyama, Japan

ABSTRACT: This paper discusses the applicability of a concept of equal energy assumption
(EEA) to the estimation of plastic deformation of a pile subjected to seismic displacements of
soils on the basis of an earthquake response analysis using an elastic pile model in place of a
nonlinear pile model. The earthquake response analyses in this study adopt a lumped mass and
nonlinear element model for a structure-pile-soil coupled system. It is concluded that the EEA is
fairly applicable to and efficient for rough evaluation of ductility factors of a pile on the basis of
results showing that the average ratios of the predominant pile curvatures estimated by the EEA
application to that by nonlinear analysis vary from 50 to 120 % in the applied cases.

71
of a linear elastic pile the system. The maximum ductility factor as an index of plastic response
characteristics can be estimated through an elastic analysis based on the equal energy assump-
tion (EEA). In this assumption, rearrangement of stresses in the system due to the plastic behav-
ior of the pile is neglected. This neglect must produce a certain degree of discrepancy against
true response of the nonlinear pile of the system (Mori & Hirata, 2002a). This paper discusses
the discrepancy through comparison between pile responses estimated based on EEA and ana-
lyzed by nonlinear response analyses, from the engineering point of view.
2.2 EEA application to a RC pile with tri-linear model
Figure 1 shows a schematic diagram of EEA application to a relation between bending moment
M and curvature of a reinforced concrete (RC) pile in the case of a tri-linear model. The tri-
linear model in the M- relation has three line segments and two folding points and an ultimate
point. M
c
and
c
are the moment and curvature at first cracking, respectively. M
y
and
y
are
those at first yielding. M
u
and
c
are those at ultimate resisting. Flexural rigidity EI that is the
slope of a line or a line segment in the M- plane is herein donated K with a suffix. K
1
is M
c
/
c
,
K
2
is the slope of the line segment between the points (M
c
,
c
) and (M
y
,
y
), and K
3
is the slope
of the line segment between the points (M
y
,
y
) and (M
u
,
u
). If you obtain an elastic response of
pile curvature
e
by an elastic earthquake response analysis, you calculate an elastic strain en-
ergy response E
e
. Comparing E
e
with the energies consumed up to the cracking, yielding and ul-
timate states that are E
c
, E
y
, and E
u
, respectively, you select one of the equations as shown be-
low. E
e
, E
c
, E
y
, and E
u
are calculated according to the equations shown below. Thus, based on
the EEA, you can calculate the response of curvature
eq
that is approximately equivalent to the
maximum response by a nonlinear analysis according to Equations 1 to 3. (Mori & Hirata,
2002a)
e eq
= (
c e
E E < ) (1)
+ + =
e
c
c c c eq
E
K
M
K M M
K
4
1
1
2
2
2
2
(
y e c
E E E < < )(2)
+ + + =
e
c y
c
y y y eq
E
K
M M
K
M
K M M
K
4
1
2
2 2
1
2
3
2
3
(
u e y
E E E < < )(3)
Figure 1. Equal energy assumption in a bending moment curvature relation in the case of tri-linear model

M
(
e
,M
e
)
(
p
,M
p
)
K
1
K
2
K
3
M
u
u
M
y
(
eq
,M
eq
)
M
(
e
,M
e
)
(
p
,M
p
)
K
1
K
2
K
3
M
u
u
M
y
(
eq
,M
eq
)
Curvature,
M
o
m
e
n
t
,

M
72
where
2 /
e e e
M E = ,
2 /
c c c
M E = ,
2 / ) )( (
c y c y c y
M M E E + + = , and
2 / ) )( (
y u y u y u
M M E E + + =
3 METHOD AND MODEL OF ANALYSIS
3.1 Description of object for study
The object of this study is an actual bridge pier recently constructed at a soft soil site shown in
Figure 2a. The pier is T-shaped and its column has a rectangular section having 3.0 m width and
3.5 m length. The mass of the pier is 336 ton and the design horizontally inertial force according
to the JRA (Japan Road Association) design specification (JRA,1996) is 11.76MN (1200 tf).
The footing of the pier was constructed 4 m below the ground surface and is supported by 12
cast-in-place reinforced concrete end-bearing piles of lengths of 30 m. The pile arrangement is
in a form of 4 rows of 3 piles, and Figure 2a shows the elevation view in a longitudinal direc-
tion. The piles were placed through a 10 m thick soft clayey soil layer characterized by the av-
eraged SPT N-value of 2 underlain by a 14 m thick medium-hard clayey soil layer characterized
by that of 8 in the bearing stratum of sandy soil with SPT N-values of more than 50 with 2 m
embedded lengths. Therefore, stress concentrations along the depth of the piles raised by seis-
mic soil displacements are anticipated to be generated at or around boundaries of certain two
soil layers in between which ratios of rigidity show high contrast. Such boundaries would be lo-
cated at relative depths of 10 m and 24 to 28 m to pile heads. Additionally, the depth of pile
heads is also another point of stress concentration due mainly to horizontal inertial forces from
the superstructure, the pier and the footing and to the seismic soil displacements.

Figure 2. Object of study and its schematic model of analysis

Fill
N=4
Clayey soil
N=2
Clayey soil
N=8
Sandy soil
N=16
Sandy soil N=50
10 m
3.5 m
RC piles
Bridge pier
Fill
N=4
Clayey soil
N=2
Clayey soil
N=8
Sandy soil
N=16
Sandy soil N=50
10 m
3.5 m
RC piles
Bridge pier
1
2
.
5

m
3
0

m
3
0

m
1
0

m
4

m
2

m
4

mFill
N=4
Clayey soil
N=2
Clayey soil
N=8
Sandy soil
N=16
Sandy soil N=50
10 m
3.5 m
RC piles
Bridge pier
Fill
N=4
Clayey soil
N=2
Clayey soil
N=8
Sandy soil
N=16
Sandy soil N=50
10 m
3.5 m
RC piles
Bridge pier
1
2
.
5

m
3
0

m
3
0

m
1
0

m
4

m
2

m
4

m
Pile-structure system
Free field soil
system
Pairs of spring and dashpot
for interaction
Superstructure mass
Input earthquake motion
Pile-structure system
Free field soil
system
Pairs of spring and dashpot
for interaction
Superstructure mass
Input earthquake motion
(a) Soil-pile-structure for study (b) Schematic model of analysis
73
3.2 Method of analysis
The earthquake response analyses in this study is carried out in time domain adopting a lumped
mass and nonlinear element model for a structure-pile-soil coupled system proposed by Penzien
et al. (1964). The schematic diagram of the model shown in Figure 2b illustrates that the model
consists of both a structure-pile system and a soil system representing a free field, and that the
springs and dashpots between the two systems transmit the displacements of the free field to the
structure-pile system. Such springs and dashpots, therefore, is hereafter called interaction
springs. The method proposed by Penzien et al. has two features for considering a pile-soil in-
teraction; firstly, a set of vertically distributed interaction springs realizes a spatial or kine-
matic interaction effect, and secondly, newly introduced effective masses realize an inertial in-
teraction effect and a dissipation damping effect at the same time. The effective mass at a node
corresponding to a unit thickness of a soil layer is evaluated to be equivalent with regard to an
integral of kinematic energy over soil elements surrounding a pile in the soil layer. The inertial
interaction is incorporated into the equation of motion as a product of the effective mass and the
relative displacement of a pile element to a soil element in the free field. It is here necessary to
be noticed that the rigidity of a vertical column of the effective masses is not taken into account.
(Mori, 2000)
Eventually in this study, earthquake response analyses are carried out by two kinds of mod-
els; one model has a nonlinear pile, and the other has an elastic linear pile. In the two models,
shear springs connecting lumped masses in the column of the soil system are fully nonlinear, but
the interaction springs are elastic linear. The condition that the interaction springs are elastic is
expected to be conservative for evaluating the response of pile moment and curvature in deeper
portions according to the previous research results obtained by Mori & Hirata (2002b).
3.3 Models of analyses
The structure-pile system is modeled to have 43 nodes with lumped masses that are basically ar-
ranged in an equal interval of 1 m. The soil system is also modeled to be arranged in an equal
interval of 1 m, accordingly it has 34 nodes with lumped masses. The boundary conditions of
the model are as follows; the end of the deepest pile element is pined to the base for the model,
and a rotation spring, which represents the resistance of a spatially-distributed-pile group, is
fixed to the base. The rotation spring has a constant calculated by only considering axial rigidity
of piles and neglecting resistance of surrounding soils. Further, it is modeled as an elastic spring
fixed to the base, because that the axial force in the most outer pile is expected not to reach its
ultimate bearing capacity. The interaction springs are evaluated according to Sugimura (1972).
The input earthquake motions for the model are defined as a total motion in the base. The
damping for the model is numerically Rayleigh damping, whose parameters and are deter-
mined for making damping ratios at the 1st and 2nd eigenfrequencies of the soil system be 2 %.
3.4 Constitutive model of soils and its parameters
The nonlinear model of shear springs for soil elements in the column for the free field adopts
the model proposed by Ramberg-Osgood (abbreviated as R-O model.) Determination of pa-
rameters of the R-O model is basically using soil properties adopted for the design of the bridge
is shown in Table 1. Shear wave velocity, Vs for each layer is determined by Equations 4 and
5 empirically based on SPT N-value, N according to JRA design specification (JRA, 1996). Ri-
gidity of soil at a strain of 10
-4
%, G
0
is determined by the Equation 6 using density of soil,
theoretically.

3 / 1
100N V
s
= for fine-grained soils (4)
3 / 1
80N V
s
= for sandy soils (5)
2
0 s
V G = (6)

74
Table 1. Soil properties for design and this study
Rigidity of soil, G and damping constant, h are dependant on a shear strain in soil element, .
For specifying such dependency as a target, Equations 7 and 8 proposed by Imazu & Fukutake
(1986) through compilation of related literatures. These equations can be applied up to a strain
of 0.5 %. Parameters a, b, c, and d in these equations are provided in cases of sand, clay, and
gravel for the average.
( )
b
a G G + = 1 1
0
(7)
d
c h = (8)
In order to fit the target dependency curves, G/G
0
- and h-, the two parameters of R-O
model. One of the two parameters is reference strain,
0.5
, which is strain of an element when the
rigidity of the element decreased to half of rigidity at a strain of 10
-4
%, and the other is damp-
ing ratio when the strain of element is infinity, h
max
. Eventually, the parameters adopted in this
study are shown in Table 2, and G/G
0
degradation curves of the model and target for sands, as
an example, is shown in Figure 3. Using the two parameters, the loading curve, or skeleton
curve in the plane of shear stress, and shear strain is expressed in Equation 9, where sub-
parameters and (These and are along the same notations as the two parameters for
Rayleigh damping, but these have different meanings and different values.) shown in Equations
10 and 11.
( )
+ = 1
0
G
(9)
max
max
2
2
h
h
= (10)
=
0 5 . 0
2
G
(11)

Table 2. Parameters of Ramberg-Osgood model for analysis

No. Soil type
Thickness
of layer
Averaged
SPT N-
Unit mass Cohesion
Shear wave
velocity
Rigidity
(m) (ton/m
3
) (kN/m
2
) (m/s) (kN/m
2
)
1 Fill 4 4 1.8 0 127 29028
2 Clayey soil 10 2 1.7 19.6 126 26989
3 Clayey soil 14 8 1.8 78.4 200 72001
4 Sandy soil 4 16 1.9 0 202 77528
5 Sandy soil >= 2 50 2.0 0 295 174048
Note
Sandy soil layer of No.5 was adopted as the bearing stratum for the design.
Reference strain
*1
Maximum damping ratio
*2
0.5
(%) h
max
(%)
Sand 0.0607 24.9
Clay 0.1406 14.5
Gravel 0.0348 14.4
Soil type
*1 Strain of an element when the rigidity of the element
*2 Damping ratio when the strain of element is infinity
decreased to half of that at a strain of 10
-4
%
75

Figure 3. G/G0 degradation curves of the model and target for sands
3.5 Constitutive model of RC piles
Properties of the RC pile are shown in Table 3. A tri-linear model is adopted as a nonlinear
model for the moment and curvature relationship. The tri-linear model adopted in this study fol-
lows Takeda model defining a loading path along a curve consisting of three line segments with
two bents, an unloading path directing a past maximum response point, and a reloading along
the unloading path toward the unloading point. Characteristic points of the tri-linear model for
RC piles are shown in Table 4. Using the points in this table, the skeleton curves of the RC piles
are shown in Figure 4. On the other hand, an elastic linear model with an initial flexural stiff-
ness is adopted in the model to be applied EEA to.

Table 3. Properties of pile

Table 4. Characteristic points of tri-linear model for RC piles

Depth (m)
c
M
c

y
M
y

u
M
u
Condition
(m
-1
)
(kNm)
(m
-1
)
(kNm)
(m
-1
)
(kNm)
Cracking 2.3*10
-4
675 2.3*10
-4
675 2.3*10
-4
648
Yielding 2.4*10
-3
2199 2.4*10
-3
2225 2.2*10
-3
1556
Ultimate 2.1*10
-2
3190 1.2*10
-2
3190 1.4*10
-2
2105
0 - 2.4 m 2.4 - 10 m 10 - 30 m

0
0.2
0.4
0.6
0.8
1
1.2
0.0001 0.001 0.01 0.1 1
)
G
/
G
0

R-O
Imazu-Fukutake (Target)
Ramberg-Osgood model
Shear strain, (%)

0
0.2
0.4
0.6
0.8
1
1.2
0.0001 0.001 0.01 0.1 1
)
G
/
G
0

R-O
Imazu-Fukutake (Target)
Ramberg-Osgood model
Shear strain, (%)
Item
Type
Diameter
Length
Strength of concrete
Young's modulous
Specification of Re-bar
Depth from pile head major bar stirrup
0 to 2.4 m 28D29 D16 ctc150
2.4 to 10 m 28D29 D16 ctc300
10 to 30 m 14D29 D16 ctc300
24 N/mm
2
2.5*104 N/mm
2
SD295A
Specification
Cast-in-place reinforced concrete pile
1.2 m
30.0 m
76
Figure 4. Skeleton curves of the RC piles by tri-linear model
3.6 Input earthquake motions
Two types of design earthquake motion waves are defined by and provided by JRA specifica-
tions. Type I is classified as a typical strong ground motion recorded at site a large distance
away during a large earthquake occurred in a subduction zone. On the other hand, Type II is
classified to a typical severely strong ground motions recorded at a site within a small distance
during a moderate-sized earthquake occurred in an inland seismic region. Input earthquake mo-
tion waves adopted in this study are show in Table 5. KPI wave, which is the NS component of
accelerogram of earthquake motion recorded at a depth of 83 m of Kobe Port Island (KPI) site
during Hyogoken-Nambu earthquake, is shown in Figure 5 as an example.

Table 5. Input earthquake motions

Figure 5. Acceleration time history of KPI wave for an input earthquake motion

0
500
1000
1500
2000
2500
3000
3500
0 0.005 0.01 0.015 0.02 0.025
Curvature (m
-1
)
M
o
m
e
n
t

(
k
N
m
)
0 - 2.4 m
2.4 - 10 m
10 - 30 m
Depth (m)
Cracking
Ultimate
Yielding
No. Type Earthquake motion wave Earthquake PGA (cm/s
2
)
1 I Kaihoku Bridge (Longi.) 1978 Miyagiken-Oki 319
2 I Itajima Bridge (Longt.) 1968 Tokachi-Oki 363
3 II Kobe Port Island (NS) 1995 Hyogoken-Nambu 679
4 II JR Takatori Station (NS) 1995 Hyogoken-Nambu 687
5 II Osaka Gas Fukiai (N27W) 1995 Hyogoken-Nambu 736
6 II JMA Kobe Obs. (NS) 1995 Hyogoken-Nambu 812
Note
Type I: Typical strong ground motion recorded at site a large distance away during a large
earthquake occurred in a subduction zone
Type II: Typical strong ground motions recorded at a site within a small distance during a
moderate-sized earthquake occurred in an inland seismic region
Types of design earthquake motion waves are defined by and provided by JRA.

-800
-400
0
400
800
0 5 10 15 20 25 30
Time (s)
A
c
c
.

(
c
m
/
s
2
)
KPI-83 m, NS
Abs.Max 679cm/s
2
, 8.06
77
4 RESULTS AND DISCUSSIONS
4.1 Effects of nonlinearity of interaction springs
In this study, the interaction springs are, for simplicity, elastic linear in both models with the lin-
ear pile and with the nonlinear pile. However, effects of nonlinearity of the interaction springs
on the response of pile need to be discussed, for an example, on the basis of the results of a pre-
vious research by Mori & Hirata (2002b), prior to discussion on effectiveness of EEA for esti-
mation of nonlinear pile curvature. In their research, the object and basic model of theirs are the
same of this research, but the difference is additional adoption of a nonlinear model of interac-
tion springs, which used R-O model adjusted to characteristics of interaction springs. Due to
space limitation, the detail of determination of parameters for this R-O model is omitted here.
Input earthquake motion wave is only KPI wave, which has a severe intensity. Therefore, actual
inputs, however, are four differently scaled waves; 0.1, 0.2 and 0.5 times and no scaled or origi-
nal for studying the progress of degree of nonlinear response.
The observations on difference of responses of reactions and deformations of the interaction
springs under different magnitudes of input motions in terms of comparison between the linear
and nonlinear interaction springs will be basis for coming discussion. As the magnitude of input
motion from 0.1 times to 1.0 times scale, nonlinear responses of the interaction springs appear
firstly at and around the pile head, secondly at depths of 28 m and 32 m corresponding to
boundaries of high contrast of impedance, and finally appear at a depth of 14 m corresponding
to boundaries between soft and hard clayey soil layers. This suggests that an inertial force from
superstructure is most influential to the pile head, and that soil displacements are more influen-
tial to deep portions having higher contrast of impedance of soils at a boundary.
Distributions of maximum moment of pile in cases of differently scaled input motions are
shown in Figure 6a to Figure 6d for comparing linear elastic and nonlinear models of interaction
springs. We cannot find any obvious difference in the case of 0.1 times scaling in spite of ap-
pearance of clearly nonlinear responses of the interaction springs and the piles at and around
pile head. Even in the case of 0.2 times scaling, differences are very slight except depths of 28
m and 32 m showing clear predominance of the linear spring case. Taking into account the re-
sponse of the nonlinear spring case under cracking moment while the response of the linear case
go beyond it, we can understand a mechanism that nonlinear behavior of the interaction spring
reduce flexural responses of pile at deeper layer boundaries with relatively high impedance.
Quantifying the effect of such mechanism, a new index, nonlinear effect ratio (NER) with re-
gard to interaction spring is defined as the ratio of response of nonlinear case to that of linear
case. Inverse of this index means degree of overestimating flexural response of pile by elastic
interaction springs.

Figure 6. Distributions of maximum moment of pile in cases of differently scaled input motions for com-
paring linear elastic and nonlinear models of interaction springs

0
5
10
15
20
25
30
35
(
m
)
N
4
2
8
16
50
0
5
10
15
20
25
30
35
(
m
)
N
4
2
8
16
50
Interaction springs
Nonlinear
Linear

(c)
0.5
0 2000 4000
(kNm)
M
c
M
y
M
u
(d)
1.0
0 2000 4000
(kNm)
M
c
M
y
M
u
(b)
0.2
0 1000 2000 3000
(kNm)
M
c
M
y
(a)
0.1
0 500 1000 1500
(kNm)
M
c
Moment (kNm) Moment (kNm) Moment (kNm) Moment (kNm)
(a) 0.1 times scaling (b) 0.2 times scaling (c) 0.5 times scaling (d) No scaling
Soil profile
and SPT N
value
0
D
e
p
t
h

(
m
)
0
5
10
15
20
25
30
35
(
m
)
N
4
2
8
16
50
0
5
10
15
20
25
30
35
(
m
)
N
4
2
8
16
50
Interaction springs
Nonlinear
Linear

(c)
0.5
0 2000 4000
(kNm)
M
c
M
y
M
u
(d)
1.0
0 2000 4000
(kNm)
M
c
M
y
M
u
(b)
0.2
0 1000 2000 3000
(kNm)
M
c
M
y
(a)
0.1
0 500 1000 1500
(kNm)
M
c
Moment (kNm) Moment (kNm) Moment (kNm) Moment (kNm)
(a) 0.1 times scaling (b) 0.2 times scaling (c) 0.5 times scaling (d) No scaling
Soil profile
and SPT N
value
0
D
e
p
t
h

(
m
)
78
In the case of 0.2 times scaled input, NERs of moment varies from 0.9 to 0.95 at most of pre-
dominantly responding portions except a depth of 32 m showing a NER of 0.6. In the case of
0.5 times scaled input, NERs of moment varies from 0.85 to 1.03 at predominantly responding
portions except a depth of 32 m showing a NER of 0.45. This indicates that a model with linear
elastic interaction springs overestimate flexural responses of pile at deeper layer boundaries
with relatively high impedance, especially a boundary of the upper end of a bearing layer.
Furthermore, in the case of no scaled input, a NER of moment is 1.25 at pile head but it is
0.95, 0.75 and 0.3 at depths of 15, 28 and 32 m, respectively. Observing the tendency in change
of NER distributions, we can understand that the increase of NER at pile head is induced from
deepening a point of resultant reaction of the interaction springs. On the other hand, the de-
crease of NERs at deeper potions can be understood that the deformation of nonlinear pile due
to larger shear strain of soil becomes more and more plastic and that the growth of moment of it
is reduced as well.
4.2 Influences of nonlinearity in flexural rigidity of pile and effects of EEA application
In the previous section, we understand about modeling of interaction springs that the linear elas-
tic model tends to overestimate the moment of pile at deeper portions rather than the nonlinear
(or tri-linear) model when the entire model has a nonlinear soil system and nonlinear pile ele-
ments. This means linear elastic modeling of interaction springs leads us conservative or safer
judgment. The results to be discussed hereafter are according to the analyses described in Chap-
ters 2 and 3 of this paper.
For discussing applicability of EEA to estimation of nonlinear flexural responses of a pile,
distributions of the responses are studied. The distributions of maximum moment of pile along
the depth in the case of Kaihoku wave input is shown in Figure 7a as an example. We can find
that there is much difference in the distributions between nonlinear and linear elastic models in
terms of magnitude and shape. Further discussions focus the depths of 4 m, 15 m, and 32 m. The
depth of 4 m corresponds to pile heads and that of 32 m to the upper end of the bearing stratum.
We should reconsider the tendencies that a lateral force on a pile head produces concentration of
moment there, and on the other hand, lateral soil displacements acting on the pile along the
depth produces concentrations of moment at boundaries with high contrast of impedance such
as the upper end of a bearing stratum. In addition, previously defined new index, nonlinearity
effect ratio (NER) is also applied with regard to a pile element as the ratio of response of
nonlinear case to that of linear case. On the basis of Figure 7a, NERs of moment at depths of 4
m, 15 m and 32 m are 3.6, 3.6 and 6.1, respectively. On the other hand, NERs of moment with
EEA application at the same depths are 0.60, 0.80 and 0.63, respectively. Therefore, EEA appli-
cation remarkably can revise estimation of moment response because of NER becoming closer
to one.
(a) Moment for Kaihoku (b) Curvature for Kaihoku (c) Curvature for KPI
Figure 7. Distributions of maximum flexural response of pile in cases of Kaihoku & KPI waves

100 1000 10000 100000
(kNm)

Mc My Mu

0.0001 0.001 0.01 0.1
(m
-1
)
u

0.0001 0.001 0.01 0.1
(m
-1
)
u
0
5
10
15
20
25
30
35
(
m
)
N
4
2
8
16
50
0
5
10
15
20
25
30
35
(
m
)
N
4
2
8
16
50
Linear
Nonlinear
EEA estimation
Moment (kNm) Curvature (m
-1
) Curvature (m
-1
)
D
e
p
t
h

(
m
)
SPT N

100 1000 10000 100000
(kNm)

Mc My Mu

0.0001 0.001 0.01 0.1
(m
-1
)
u

0.0001 0.001 0.01 0.1
(m
-1
)
u
0
5
10
15
20
25
30
35
(
m
)
N
4
2
8
16
50
0
5
10
15
20
25
30
35
(
m
)
N
4
2
8
16
50
Linear
Nonlinear
EEA estimation
Moment (kNm) Curvature (m
-1
) Curvature (m
-1
)
D
e
p
t
h

(
m
)
SPT N
79
Hereafter, the curvature of pile, which is of importance in performance-based approach, is
discussed. The distributions of maximum curvature of pile along the depth in the case of Kai-
hoku wave input is shown in Figure 7b. We find that there is also much difference in the distri-
butions between nonlinear and linear elastic models in terms of magnitude and shape, while
EEA application revise a distribution shape of linear elastic response much better than moment.
NERs of curvature at depths of 4 m, 15 m and 32 m are 0.16, 0.39 and 0.11, respectively. On the
other hand, NERs of curvature with EEA application at the same depths are 0.56, 1.02 and 0.66,
respectively. Moreover the situation in the depth of 28 m is very similar to that in the depth of
32 m. Therefore, EEA application considerably can revise estimation of curvature response in
this case. In particular, it is understood that EEA is practically very effective to estimation of
plastic response of pile curvature in deeper portions. For confirming this understanding, let us
see other example. The distributions of maximum curvature of pile along the depth in the case
of KPI wave input is shown in Figure 7c. The shape of each models and the tendencies of
nonlinear effects that can be evaluated by NER are almost the same as described previously.
Thus, the understanding can be confirmed.
4.3 Influences of nonlinearity in flexural rigidity of pile and effects of EEA application
In the previous section, influences of nonlinearity in flexural rigidity of pile was discussed on
the basis of distribution of flexural responses of pile, and accordingly effects of EEA application
for estimation of nonlinear curvature of pile were confirmed, especially in its deeper portions.
Then, the applicability of EEA is discussed through comparison among predominant curvature
responses at three specific depths in cases of various types of input waves. Figure 8 shows
variation of NERs of curvature at depths of 4 m, 15 m and 32 m subjected to all six input mo-
tions. For linear models of pile, NERs vary approximately from 0.1 to 0.4. On the other hand,
NERs vary approximately from 0.5 to 1.2 for EEA application. In particular, the average of
NERs in deeper portions is roughly 0.8. Considering the degree of the effect of overestimate by
using linear interaction springs, which was discussed in 4.1, I can conclude that EEA applica-
tion to an entire model with linear elastic pile elements and linear elastic interaction springs is
practically very efficient for evaluating plastic curvature of pile subjected to severe earthquake
motions.

Figure 8. Variation of nonlinearity effect ratio (NER) of curvature at depths of 4 m, 15 m and 32 m sub-
jected to six input motions
5 CONCLUSIONS
The applicability of equal energy assumption (EEA) to linear elastic pile response for estimation
of ductility factor response of a nonlinear pile was studied on the basis of fully nonlinear dy-
namic response analyses on an actual bridge pier. It is concluded as follows;

0.01
0.1
1
10
/

4 m

15 m

32 m
/

4 m

15 m

32 m
1 2 3 4 5 6
EEA application/Nonlinear
4 m
15 m
32 m
Linear/Nonliner
4 m
15 m
32 m
Number of input motion
R
a
t
i
o

o
f

c
u
r
v
a
t
u
r
e
s

0.01
0.1
1
10
/

4 m

15 m

32 m
/

4 m

15 m

32 m
1 2 3 4 5 6
EEA application/Nonlinear
4 m
15 m
32 m
Linear/Nonliner
4 m
15 m
32 m
Number of input motion
R
a
t
i
o

o
f

c
u
r
v
a
t
u
r
e
s
80

(1) On the basis of fully nonlinear response analyses of soil-pile structure system subjected to
four differently scaled input waves for studying the progress of degree of nonlinear behavior in
each element of the system, the nonlinear behavior of the interaction spring reduces flexural re-
sponses of pile at deeper layer boundaries with relatively high contrast of soil impedance. De-
fining a nonlinear effect ratio (NER) as the ratio of response of nonlinear case to that of linear
case with regard to interaction spring for quantifying the effect of such mechanism, it was found
that NERs at pile head and deeper potions become lesser than one as the input wave increases.
This means elastic modeling of interaction spring overestimates flexural response of pile rather
than nonlinear modeling of it.

(2) EEA application to linear elastic pile modeling considerably can revise the estimation of
nonlinear flexural responses of pile such as moment and especially curvature in terms of pre-
dominant response in their distribution along the depth. In particular, it is understood that EEA
is practically very effective to estimation of plastic response of pile curvature in deeper portions.
NERs of curvature of pile subjected to all six input motions vary approximately from 0.1 to 0.4
for a linear elastic pile, on the other hand, vary approximately from 0.5 to 1.2 for EEA applica-
tion. In particular, the average of NERs in deeper portions for EEA application is roughly 0.8.

(3) EEA application to an entire model with linear elastic pile elements and linear elastic inter-
action springs is practically very efficient for evaluating plastic curvature of pile subjected to
severe earthquake motions.
REFERENCES
Imazu, M. & Fukutake, K. 1986. Dynamic deformation properties of gravels. Proc. 21st Ann. Conf. Jap.
Soc. Soil Mech. Found. Eng., JSSMFE: 509-512. (in Japanese)
Japan Road Association (JRA) 1996. Design Specification of Highway Bridge, Part V Seismic Design:
Maruzen. (in Japanese)
Mori, S. 2000. Proposal of spring-mass model for pile-foundation structure and its application to really
damaged structures, Journal of Applied Mechanics, JSCE, 3: 609-620. (in Japanese with English ab-
stract)
Mori, S. & Hirata, A. 2002a. Evaluation of nonlinear seismic response of pile by constant energy rule.
Proc. 37th Ann. Conf. Jap. Geotech. Soc., JGS, 1495-1496. (in Japanese).
Mori, S. & Hirata, A. 2002b. Effects of nonlinearity of members on seismic response of pile-foundation
structure. Journal of Structural Engineering, JSCE 48A: 469-478. (in Japanese with English abstracts)
Mori, S, Suga, K. & Akaishi, T. 2009. Evaluation chart of existing pile foundation against seismic soil
displacements. Performance-Based Design in Earthquake Geotechnical Engineering, Proc. Intern.
Works. Kokusho, Tsukamoto & Yoshimine (eds) , London: Taylor & Francis Group: 667-674.
Newmark, N.M. & Rosenblueth, E. 1971. Fundamentals of Earthquake Engineering, Prentice-Hall Inc.
Penzien, J., Scheffey, C.F. & Parmelee, R.A. 1964. Seismic Analysis of Bridges on Long Piles, Journal
of the Eng. Mech. Div., Proc. of ASCE 90(EM3): 223-254.
Sugimura Y. 1972. Research on Seismic Vibration Characteristics of Foundations Supported by Long
Piles in Soft Soils, Dissertation to Waseda University. (in Japanese)

81
Sinkage of a Pile Foundation during the Niigata-ken Chuetsu-oki
Earthquake in 2007

Y. Goto
Earthquake Research Institute, the University of Tokyo, Bunkyo-ku, Tokyo, Japan
1 INSTRUCTION
ABSTRACT: A new building supported by pile foundation in the Kashiwazaki sewage treat-
ment plant sank by a hundred and several tens millimeter during the strong shaking (more than
X of EMS98) of the Niigataken Chuetsu-oki Earthquake in 2007. The building was designed
according to the new seismic design code revised after the 1995 Kobe Earthquake but sank
more than the adjacent old building of mat foundation did. This paper reports an overview of the
damage, analyses the mechanism of the sinking and discusses some issues about the earthquake
proof design of foundations against high level of seismic load.

1.1 General overview of the earthquake
Kashiwazaki-city, one of the middle scale cities on
the northern-central shore of Japan main island, hav-
ing ninety thousand population, was hit by a very
strong shaking more than X in EMS98 scale due to
the earthquake of July 16, 2007 having Magnitude
6.8. The world largest scale nuclear power plant is
located in Kashiwazaki-city. Very many professionals
focused on the effect of so strong shaking to the nu-
clear power plant and considerable lessons have been
extracted. However, there are some useful lessons ex-
tracted from damages of ordinary structures.
Epicenter
Aftershock
Area
Nuclear
Power Plant
The author made damage reconnaissance of a sew-
age-treatment plant of Kashiwazaki-city focusing his
interest on a pile foundation. Sewage-treatment plant
is often located on a shore or a river side, where
ground condition is commonly poor. So it is unavoid-
ably easy to suffer earthquake damages.
1.2 Kashiwazaki-city sewage-treatment plant
The Kashiwazaki sewage-treatment plant (purifica-
tion capacity is 40,700 kiloliter a day) is located on
the mouth of Sabaisi River also (Fig. 1). It is about
15 km south-southwest of the epicenter.
Kashiwazaki
City Office
Railway
Station
Sewage
Treatment
Plant
2km
200km
Figure 1. Location of Epicenter and
Kashiwazaki-city
(detached to Google
)
82

Although there was not any seismometer in the plant, the shaking intensity is estimated to have
exceeded X (EMS98) from an interpolation of seismometer records around the site, namely,
JMA Intensity 6-upper (about EMS98 X) at the city office of Kashiwazaki and Kashiwazaki
railway station and JMA Intensity 7 in the nuclear power plant site (maximum among the many
seismometers in that site).
Main building
Purification facilities
Digester tanks
Sabaisi River
50 m
Main sludge-treatment building
Sludge Inspecttion
Building
Dense sand dune
Gas facility
Figure 3. Total layout of Kashi-
wazaki sewage treatment plant
Figure 2. Typical boring logs
Figure 2 is a typical borehole exploration data of the site. The surface layer is loose buried
sandy soil of 3 meters depth overlying on loose sand sediment layer of 4 meters. These layers
are followed by dense sand dune layer. The depth of bedrock of this area is supposed to be sev-
eral hundred meters.
1.3 Damage overview of the sewage-treatment plant
Figure 3 shows total layout of the plant. Although some sections of the sewage line in Kashi-
wazaki-city were heavily damaged (mainly by liquefaction), influent wastewater to the plant
was increased 1.3 times because of inflow of underground water, and drift sand increased con-
siderably.
The function to flow the sewage through the plant and to run it out was not damaged. Main fa-
cilities for sewage-treatment were not damaged too, because they were series of massive rein-
forced concrete pools based on improved sand layer, where sand compaction pile method was
used for the improvement. Power outage was only one minute and purification function of the
plant was recovered soon after the earthquake.
While, the sludge treatment function of the plant went down because of some damages on the
treatment facilities, such as digester tanks, sludge squeeze transportation line, and gas facilities
for heating sludge. These damages did not affect the purification function of sewage directly. A
temporary sludge transportation line was equipped within 5 days, and the sludge was stored un-
digested, then dewatered and carried out from the plant.
83
2 DAMAGE OF SLUDGE INSPECTION BUILDING
2.1 Profile of the Sludge Inspection Building
As mentioned above, main facilities for purification were not damaged, but a large vibrational
and residual displacement of an attached building, which had been designed according to the
new seismic design code revised after the 1995 Kobe Earthquake, made trouble. The building,
called as Sludge Inspection Building, had pile foundation and was constructed on a side of main
sludge-treatment building which had deep mat foundation based on dense sand layer. Figure 4 is
a side view of the main building and the Sludge Inspection Building. The latter is reinforced
concrete building having three stories and had been supported by fifty-one PHC piles (shear-
strength-improved type) as shown in Figure 5. The pile was eight meters in length and 700mm
in diameter and supported 900kN static load.

Sludge Inspection
Building
Main Sludge
Treatment Building
Figure 4. Side view of Main Sludge Treatment Building (left) and Sludge Inspection Building (right)

Future
extension
Figure 5. Profile of pile and plan of pile layout
Pile top
reinforcement
Footing
Concrete
filled in
4
5
0
0

4
5
0
0

5
6
0
0

5
4
0
0

6000 6000
5
4
0
0

PHC pile
shear strength improved type
84
2.2 Sinking and relative displacement
Due to the earthquake, the Sludge Inspection Building settled down 90mm and displaced later-
ally 200mm relative to the main building (Fig. 6 & 7). As the large interrrspaces, more than one
hundred millimeter, were left between the sides of the base slab and the ground, relative instan-
taneous displacement between the two buildings during the earthquake was supposed to be lar-
ger than 200 mm. This large gap was one of the causes of the breakage of the sludge squeeze
transportation line (Fig. 8). The main sludge-treatment building, itself, was estimated to have
sunken down several tens millimeter by a leveling after the earthquake. Accordingly, the Sludge
Inspection Building sank down one hundred and several ten millimeter absolutely.
2.3 Damage of the piles
Figures 9 & 10 show the upper ends of the piles, which were cracked by bending clearly. How-
ever, this type of failure is unexplainable for the
sinking more than several ten millimeters unless
the extent of the failure causes detaching of
concrete from the pile. The used piles were
typical shear-strength-improved type which was
developed after the 1995 Kobe earthquake. Al-
though there were many evidences of liquefac-
tion of the surface layer around the building,
large scale lateral spreading or flowing were not
observed in the site. So shear failure of the piles
seemed hard to occur at underground. Finally, it
is concluded that the piles sank down from their
tips.
Figure 6. Right side is pile foundation

Figure 8. Breakage of sludge transportation line
(Kume 2007)
Figure 7. Left side is pile foundation

Figure 10. Crack at a pile top Figure 9. Crack at a pile top
85
2.4 Design and construction method of the piles
The applied design formula for estimating bearing capacity of the piles was commonly used
one;

(1)

here, RaL ; long term allowable bearing
capacity (kN)
R l =
3
1
( N A 1.5 L ) a
_
p + f
; end bearing capacity factor

10
D
L
90 ; = 25 (L 5m)
90 <
D
L
110 ; = 25
4
1
(
D
L
- 90)

L ; length of pile
D ; external diameter of pile
; average standard penetration number value of ground between pile tip - 4D depth
and pile tip +1D depth.
Ap ; cross section area of pile defined by outer diameter
Lf ; length (in meter) of pile where surface friction can be countable.
N
_
; perimeter length (in meter) of pile

Allowable bearing capacity for short period load such as earthquake shall be double of
long term bearing capacity.

According to the formula, the allowable long term bearing capacity of one pile was calculated as
1040 kN, and 2080 kN for the short term bearing capacity. So, the cause of the sinking seemed
not to be in design calculation as far as the formula (1) is reliable.
2.5 Piling method
The piling method of the piles was one of the typical inner-boring, foot-protection and tip-
broadening methods. The process is shown in Figure 11. Diameter of the tip-broadening part
was 900mm, while that of the piles was 700mm. The broadening was elongated one meter from
the tip of the piles. The void of the elongated part was filled by cement milk (w/c; 60%).

Drilling
and pile
insert
Broadening Pile insert Drill clawback Cement milk filling
Figure 11. Piling method
86
2.6 Scenario of the sinking
Fine sand
Middle sand
Sandy silt
Silty clay
50
Broaden part level
SPT number
Figure 12. Boring log and pile tip level
Dese dune sand layer
Pile
Pile tip level
Figure 12 shows the boring log and the setting depth of
the piles comparatively. The tip of the broadening part
almost reaches the bottom boundary of the dense dune
sand layer. So the authors scenario is;
(1) Surface sandy soil was liquefied due to the strong
earthquake, and the skin friction between the piles and
the soil disappeared.
(2) The piles became supported by the broadening parts
only.
(3) Pore water pressure in the dense dune sand layer and
the soft sandy soil layer just below the dune sand rose
up, and the bearing capacity of these layers decreased.
(4) The tips of the broadening part punched the dense
dune sand through and sank.
3 CONSIDERATION FROM A VIEW POINT OF
DESIGN PHILOSOPHY
3.1 Structural safety was reserved but performance was lowered
The sinking was uniform and the superstructure of the Sludge Inspection Building did not suffer
any significant structural damage. As the structural safety was reserved against such strong
shaking, the seismic design applied to the building was judged to be adequate. However there
remains one problem, that is, the sludge transportation line was broken due to the large relative
displacement against adjacent building.
It is a common knowledge of geotechnical engineers that same type of foundation should be ap-
plied to buildings which need to minimize relative displacement each other. But, in this plant,
pile foundation was applied to a new building even if it would be in contact with a previous
building which had been built on deep mat foundation.
The following circumstances were supposed to be to this issue;
(1) As the new building needed no underground floor, a deep mat foundation was not good
choice because of its high cost.
(2) While, a shallow mat foundation, even if a soil improvement was combined, was not appli-
cable because the sliding and the over-turning stabilities could not be satisfied against the
high design seismic force (0.6G of level 2) which had been adapted after the Kobe Earth-
quake.
(3) Consequently, a pile foundation was adopted. There seems to have been no other ways, even
if the design engineer noticed the large displacement of pile foundation might occur during
strong earthquake,
3.2 Discrepancy between high design earthquake force and design calculation method
The discrepancy between the high design earthquake force (0.6G) and the design calculation
method concerning sliding and over-turning stabilities should be considered much more. The
stabilities are usually evaluated by the classic plastic-equilibriums theory of inertia force to fric-
tion, bearing capacity of sub-soil and geometric stability. But, the classic theory is adequate to
inertia force around 0.2G and loses its balance when applied to so high inertia force as 0.6G.
Low possibility but extremely high earthquake load such as 0.5-1.0G has become accepted to be
rational from the experiences of recent destructive earthquakes and researches.
Allowable residual displacement concept based on the deformation estimation must be im-
proved and the effort to apply it in the earthquake proof design for foundations is expected.
87
4 RESTORATION WORK

A series of careful inspection and back-analyses revealed that all heads of the piles had been
cracked, but at the same time, revealed that most of the pile cap footings and the under floor
beams had not been damaged. Then, small holes were drilled through the base floor, and ce-
ment-mixed soil columns were created under the whole area of the footings using a high pres-
sured jet mixed method.
5 CONCLUSION

(1) A new building supported by pile foundation sank by a hundred and several tens millimeter
due to a strong shaking of X or more in EMS98 scale during the 2007 Niigata-ken Chuetsu-
oki Earthquake.
(2) The cause of sinking was supposed to be the pile tips' punching through a supporting layer.
(3) Structural safety of the new building was reserved but its performance was lost due to the
sinking and the large lateral displacement relative to an adjacent building which was based
on deep mat foundation.
(4) The choice of pile as the foundation was rational as far as the new design code, which had
been innovated after the 1995 Kobe Earthquake, was concerned.
(5) Problem is in the discrepancy between the high design earthquake force (0.6G) and the de-
sign calculation method, which automatically excludes use of shallow mat foundation.
(6) Challenge we are facing is to improve an allowable residual displacement concept based on
deformation estimation and to apply it in earthquake proof design for foundations.

ACKNOWLEDGEMENT

The author would like to express sincere thanks to Mr. Atsushi Nagaoka, Head of Kashiwazaki
Sewage Treatment Plant, for his limitless cooperation to this survey.

REFERENCES

JSCE, JGS, JAEE, AIJ and SSJ 2007. Disaster Reconnaissance on the 2007 Niigata-ken Chuetsu-oki
Earthquake; Handout for Debrief Session (in Japanese)
Kume E. 2007. Quick Report on Sewage Treatment Plants and Facilities due to the 2007 Niigata-ken
Chuetsu-oki Earthquake, Quarterly journal Mizusumasi 2007 summer, 20-24 Japan Sewage Works
Agency Japan Sewage Works Agency (in Japanese)

88
1 INTRODUCTIONS
Soil liquefaction brought severe damage to pile foundations of road bridges and buildings dur-
ing the 1995 Kobe Earthquake, liquefaction and lateral spreading of the liquefied sand in Kobe
port area caused extensive damage to port facilities. Seaward displacements of the sheet pile
quay walls due to lateral spreading of the liquefied soil extended over several meters, as a result,
translation and inclination occurred to neighboring pile foundations.
Sheet pile quay walls have been widely used in metropolitan areas where many structures and
bridges have been constructed using pile foundations. In order to mitigate the damage to these
structures resulting from probable future large earthquakes, it is therefore necessary to develop
appropriate countermeasures. Consequently, it is important to understand the mechanisms re-
garding seismically induced ground deformation behind sheet pile quay walls and to evaluate
their effects on neighboring pile foundations.
Recently, at National Research Institute for Earth Science and Disaster Prevention in Japan
(NIED), a series of large scale shake table tests were conducted to study the seismic response of
sheet pile wall system and the liquefaction and deformation characteristics of the saturated sand
backfill, as well as the response of the neighboring pile foundations. In these experiments, both
the liquefaction and post liquefaction stages were modeled and studied. The largest laminar box
in the world was employed in order to obtain nearly full-scale testing results to ascertain the
mechanisms of lateral ground flow of the liquefied soil behind sheet pile quay walls and to eva-
luate the effects of the liquefied earth pressure acting on pile foundations both during the ground
shaking and post liquefaction stage.
2 LARGE-SCALE TESTS
2.1 Apparatus
A large-scale laminar box (6.0m x12.0m x 3.5m) which is the largest in the world to this date,
and a large shake table (15m x 14.5m) in National Research Institute for Earth Science and Dis-
Large-scale Shake Table Tests on Lateral Spreading of Sheet-pile
Quay Wall and Pile Foundation

M. Sato & K. Tabata
National Research Institute for Earth Science and Disaster Prevention, Tsukuba, Japan
ABSTRACT: Shaking table tests of two large-scale models with liquefiable deposit on lateral
spreading under earthquake motions were performed to reveal the failure mechanism of sheet-
pile-type quay wall and pile foundation behind the wall. These tests simulated the lateral
spreading behavior of the deposit with slow rate of ground deformation, which is assumed to
occur even after earthquake shaking. The test results also indicated that deposits residual de-
formation induced by cyclic motions due to an earthquake and structures inertial force influ-
ence largely the deformations of the quay wall and pile foundation.

89
aster Prevention (NIED) were used to perform the experiments. Photo 1 shows the shake table
and the assembled laminar box. The laminar box is designed to slide to maximum one meter in
horizontal direction. The each layer of the box can move independently regarding its upper and
lower layers. A rubber membrane is placed to provide a waterproof inside the box. External
frames with horizontal rollers ensure safe and accurate movement of the layers, while the inter-
nal stoppers are used as limiting measures to stop excessive movement of the individual layers.
2.2 Test cases and input motions
Two tests were executed. The depth of the deposit in CASE-1 was 4.5m, while that in CASE-2
was 5.0m. The experimental conditions of two tests are almost same except these depths and
pile materials. Reinforced concrete piles were used in CASE-1, while steel piles in CASE-2.
Figure 1 shows the test model in a large-scale laminar box and locations of transducer in CASE-
2. The model and the transducer locations were almost same in CASE-1.

Five cycles of sinusoidal wave were used as the input motion. Figure 2 shows the time histories
of the input accelerations and velocities in CASE-1 and CASE-2. The maximum acceleration
450 gal and the maximum velocity 17 cm/s with the frequency 4 Hz were used in CASE-1,

Photo 1 Large-scale laminar container fixed on the shake table in NIED
Section
East
Acc.West1
Acc.Structure
PWP-West1
Disp.Structure
Disp.Sheet-pile
Accelerometer
Pore Pressure meter
Displacement
Bending Strain
Acc.Input
11, 600mm
5, 000mm
500mm
800mm
3, 700mm
West
PWP-West2 Acc.West2
Acc.East1
Acc.East2
Acc.East3
PWP-East1
PWP-East2
PWP-East3
Pile-A

Figure 1 Schematic illustration of the shake table test model (CASE-2)
90
while 350 gal and 37 cm/s with 1.5 Hz were used in CASE-2. The maximum velocity in
CASE-2 was about twice of CASE-1. In consequence, the table input energy of Case-2 was
about four times of CASE-1.
2.3 Soil material
Clean sand from Kasumigaura area in Ibaraki prefecture in Japan was sampled and used in this
experiment. Index tests were performed on this sample, and Figure 3 shows the grain size distri-
bution and physical properties of the material.
2.4 Sample preparation method
Four reinforced concrete piles in CASE-1 were installed in the center of the laminar box by pin
connection to the base. Diameter of the piles was 15 cm and their length was 4.5 m. Center to
center space between the piles was 0.9 m and a steel top cap provided a rigid connection on top
of the piles. Similarly, four steel piles in CASE-2 were installed. A sheet pile in CASE-1 and
CASE-2 was installed in the east side of the piles with 30 cm space from the laminar box wall.
This sheet pile was used to provide lateral soil pressure on the liquefied soil to keep it moving
toward the water side. The box was partially filled with water, and dry sand was pluviated in the
water. As the hydraulic sediment was filled in the box, the water level was also increased. Based
on the past experience, it was proved that this method gives a uniform and relatively loose satu-
rated sample (Dr= 45%). In the next step, the sheet pile installed, and the filling procedure con-
tinued until the soil level in east and west side of the sheet pile reached to 4.0 and 3.2 m in
CASE-1, respectively. The water level at this stage was 0.5 m from ground surface. Then, in the
east side, a layer of unsaturated soil was placed to raise the ground level to 4.5 m. Soil, piles and
sheet piles were heavily instrumented with pore water pressure and displacement transducers, as
well as accelerometers and strain gauges. Instruments were fixed on a thin net and installed in
the box, before sand pluviation. Total number of 256 channels was used for data acquisition of
-500
0
500
A
c
c
.

(
g
a
l
)
(a)Input Acc.
-50
0
50
0 1 2 3 4 5 6 7 8
V
e
l
.

(
c
m
/
s
)
Time (s)
(b)Input Velocity
-500
0
500
A
c
c
.

(
g
a
l
)
(a)Input Acc.
-50
0
50
0 2 4 6 8 10 12 14
V
e
l
.

(
c
m
/
s
)
Time (s)
(b)Input Velocity

(CASE-1Sine 4.0Hz, a
max
=450Gal) (CASE-2Sine 1.5Hz, a
max
=350Gal)
Figure 2 Measured input motion time histories on shake table

0
20
40
60
80
100
0.01 0.1 1 10
P
e
r
c
e
n
t

f
i
n
e
r

b
y

w
e
i
g
h
t

(
%
)
Grain size (mm)
Kasumigaura Sand
D50 (mm) 0.31
Uc 3.0
Gs 2.718
Fc (%) 5.4
emax 0.961
emin 0.570

Figure 3 Grain size distribution curve and physical
properties of Kasumigaura sand

Photo 2 Structure and sheet-pile model on large-scale
test
91
outputs of the instruments. Then, heavy plates were mounted and fixed on top cap until its
weight reached to about 10 tons, as a model for a massive super structure.
3 TEST RESULTS
3.1 Acceleration responses and excess pore water pressures of sand deposits
Figure 4 shows the time histories of response accelerations observed at the backyard of the pile
foundation and sea-side in front of the quay wall. Similarly, Figure 5 shows the time histories of
excess pore water pressures. The comparison of these time histories of the accelerations and
excess pore water pressures in Figures 4 and 5 indicates that soil liquefaction occurred after 2 or
3 cycles and then vibrations did not transfer to upper layers due to soil liquefaction. The test re-
sults reveal that after a few pulses of shaking, excess pore water pressure in loose backyard in-
creased.
3.2 Displacements of sheet-piles and structures
The large displacement caused by soil liquefaction was measured on the top of the pile-group
structure and sheet pile. The displacement time histories in CASE-1 and CASE-2 are shown in
Figure 6. The lateral displacements of the sheet pile in these cases cyclically accumulated dur-
ing the shaking and slowly grew up after the shaking. On the other hand, the lateral displace-
ments of the structure accumulated during the shaking, while these did not grow up remarkably
after the shaking.
The displacement of the top of the sheet pile at the end of the shaking (about t=5 second) was
about 25 mm in CASE-1. The figure indicates that at t=50 second the lateral displacement in-
creased only about 5 mm, and after this moment a rapid change in rate of lateral displacement
was observed. The flow continued until t=200 second and the maximum displacement reached
to about 100 mm at the top of the sheet pile. In CASE-2, the displacement of the top of the sheet
-500
-250
0
250
500
East-1(GL-1140mm)
East-2(GL-2400mm)
A
c
c
.

(
G
a
l
)
(a) Acc. Ground East (backyard)
-500
-250
0
250
500
West-1(GL-1100mm)
West-2(GL-2360mm)
0 1 2 3 4 5 6 7 8
A
c
c
.
(
G
a
l
)
(b) Acc. Ground West (seaside)
Time (s)
-500
-250
0
250
500
East-1(GL-1220mm)
East-2(GL-2480mm)
A
c
c
.

(
G
a
l
)
(a) Acc. Ground East (backyard)
-500
-250
0
250
500
West-1(GL-1180mm)
West-2(GL-2440mm)
0 2 4 6 8 10 12 14
A
c
c
.

(
G
a
l
)
(b) Acc. Ground West (seaside)
Time (s)

max
max
=350Gal)
Figure 4 Acceleration time history measured in ground of backyard and seaside

-10
0
10
20
30
East-1(GL-1140mm)
East-2(GL-2400mm) P
W
P

(
k
P
a
)
(a) PWP-East
(backyard)
-10
0
10
20
30
West-1(GL-1100mm)
West-2(GL-2360mm)
0 1 2 3 4 5 6 7 8
P
W
P
(
k
P
a
)
(b) PWP_West
(seaside)
Time (s)
-10
0
10
20
30
East-1(GL-1220mm)
East-2(GL-2480mm)
P
W
P

(
k
P
a
)
(a) PWP-East
(backyard)
-10
0
10
20
30
West-1(GL-1180mm)
West-2(GL-2440mm)
0 2 4 6 8 10 12 14
P
W
P

(
k
P
a
)(b) PWP-West
(seaside)
Time (s)

max
max
=350Gal)
Figure 5 Excess pore water pressure generation and dissipation time histories in ground of backyard and seaside
92
pile at the end of the shaking (about t=9 second) was about 300 mm. The flow continued until
t=180 second and the maximum displacement reached to about 440 mm.
Based on the results, it can be clearly said that these tests were able to reproduce the phenome-
non of post-liquefaction behavior of the liquefied backfill sand that slowly flew toward seaside.
The displacements of the sheet pile and structure dividing with during and after the shaking
show in Table 1. The displacements of the sheet pile and structure in CASE-1 and CASE-2 were
almost same after the shaking, but these were large different during the shaking. The reason is
that these displacements depended on the input motion energy and the liquefaction response of
the sand deposit.
The influence on the displacement of the structure during the shaking was larger than that after
the shaking because the ratio of the displacement of the structure and sheet pile quay wall was
0.80 during the shaking and 0.20 after the shaking in CASE-1, and 0.47 during the shaking and
0.19 after the shaking in CASE-2.
These facts indicate that, when predicting displacement, it is not sufficient to take into account
the liquefaction-induced large ground deformation only after the shaking.
3.3 Bending Strains of piles
Strain gauges were attached on ten different height levels of the piles to measure the bending
strain time history during and after the shaking. Figure 7 shows bending strain time histories on
each depth of the sheet pile in CASE-1 and CASE-2. The distributions of bending strain about
the altitude shows in Figure 8.
The bending strain of pile top in CASE-1 was too small to release bending moment by the struc-
ture rocking. In addition, the large bending strain concentrated at the local point of GL-3000
mm to break out cracks of the concrete pile. A local point of concentration of the bending strain
of the steel pile in CASE-2 could not be found, but the value at around GL-3000 mm was large.
This depth was near the bottom of the sheet pile because the stress of the pile became large in
the depth to cause ground deformation due to lateral spreading.
In CASE-1, the distributions of bending strain at t=2.39 and t=2.58 seconds during the shaking
was small due to rotation of the pile top. The distribution at t=3.33 second during the shaking
showed peak at GL-3,000 mm, and that bending strain caused by ground deformation due to lat-
eral spreading. When the test model ground was demolished after the test, the damage of piles
-50
0
50
100
150
Sheet-pile
Structure
1 10 100 1000
D
i
s
p
.
(
m
m
)
Time (s)
-250
0
250
500
/ /
Sheet-pile
Structure
1 10 100 1000
D
i
s
p
.

(
m
m
)
Time (s)

max
max
=350Gal)
Figure 6 Displacement time histories on top of the sheet pile and structure

Table 1 Displacements of sheet-pile and structure for during shaking and after shaking
During shaking
(CASE-1:25sec)
(CASE-2:29sec)
After shaking
(CASE-1:5650sec)
(CASE-2:91000sec)
CASE-1
Sine4.0Hz,
a
max
=450Gal
Sheet-pile 25mm 75mm
Structure 20mm 15mm
Structure/Sheet-pile 0.80 0.20
CASE-2
Sine 1.5Hz,
a
max
=350Gal
Sheet-pile 300mm 85mm
Structure 140mm 15mm
Structure/Sheet-pile 0.47 0.19
93
was investigated. As the results, several large cracks on the reinforced concrete pile caused near
GL-3000 mm and remained large bend.
The distribution shape of bending strain in CASE-2 is smooth because of the steel pile. The dis-
tributions at t=3.11 and t=4.49 seconds during the shaking are small similarly CASE-1, but
show large value near bottom of the sheet pile. The distributions at t=50 and t=665 seconds after
the shaking show a steady shape except near the pile tip.
4 CONCLUDINS
Two large-scale shake table tests were conducted to study the seismic response of sheet pile
wall system and the liquefaction and deformation characteristics of saturated sand backfill, as
well as the response of its neighboring pile foundations.
The test results revealed that after a few pulses of shake, excess pore water pressure in saturated
and relatively loose backfill increased, and the consequent loss of effective stress resulted in lat-
eral spread of the liquefied sand.
Post-liquefaction behavior of the liquefied sand was quite remarkable. The results clearly dem-
onstrate that the tests can reproduce the phenomenon of the post-liquefaction behavior of the li-
quefied backfill sand that slowly flows toward seaside.
-200
0
200
400
S
t
r
a
i
n

(
x
1
0
-
6
)
(a) Bending strain (Pile-A)
GL-250mm
-200
0
200
400
S
t
r
a
i
n

(
x
1
0
-
6
)
(b) Bending strain (Pile-A)
GL-500mm
-300
0
300
600
S
t
r
a
i
n

(
x
1
0
-
6
)
(c) Bending strain (Pile-A)
GL-1250mm
-400
0
400
800
S
t
r
a
i
n

(
x
1
0
-
6
)
(d) Bending strain (Pile-A)
GL-2000mm
1200
600
0
-600 S
t
r
a
i
n

(
x
1
0
-
6
)
(a) Bending Strain (Pile-A)
GL-380mm
1200
600
0
-600
S
t
r
a
i
n

(
x
1
0
-
6
)
(b) Bending Strain (Pile-A)
GL-1220mm
1200
600
0
-600
S
t
r
a
i
n

(
x
1
0
-
6
)
(c) Bending Strain (Pile-A)
GL-2060mm
1200
600
0
-600 S
t
r
a
i
n

(
x
1
0
-
6
)
(d) Bending Strain (Pile-A)
GL-2900mm
-800
0
800
1600
1 10 100 1000
S
t
r
a
i
n

(
x
1
0
-
6
)
(e) Bending strain (Pile-A)
GL-3000mm
Time (s)
1600
800
0
-800
1 10 100 1000
S
t
r
a
i
n

(
x
1
0
-
6
)
(e) Bending Strain (Pile-A)
GL-3320mm
Time (s)

max
max
=350Gal)
Figure 7 Bending strain time histories on pile-A

-1000 0 1000 2000
-5000
-4000
-3000
-2000
-1000
0
t=2.39(s)
t=2.58(s)
t=3.33(s)
t=5.00(s)
t=50. 0(s)
t=665(s)
Bending strain (x10
-6
)
D
e
p
t
h

(
m
m
)

-2000 -1000 0 1000
-5000
-4000
-3000
-2000
-1000
0
t=3.13(s)
t=3.50(s)
t=3.85(s)
t=4.94(s)
t=50. 0(s)
t=665(s)
Bending strain (x10
-6
)
D
e
p
t
h

(
m
m
)

max
max
=350Gal)
Figure 8 Bending strain distributions on pile-A
94
The influence of during the shaking on the displacement of structure is larger than that after the
shaking. The displacements of the sheet pile and structure is dominated in the input motion
energy and the liquefaction response of sand deposit.
It is important to note that this study is an on-going research process and more data analyses,
shake table tests, centrifuge tests and numerical analyses are being conducted and the results
will be published.
5 REFERENCES
Editorial Committee for the Report on the Hanshin-Awaji Earthquake Disaster, 1997, Report on the Han-
shin-Awaji Earthquake Disaster, Damage to Civil Engineering Structure, Port and Coastal Facilities,
River Management Facilities and Sediment and Slope Movement Control Facilities, JSCE
Inatomi, T., et al., 1997, Technical Note of Damage to Port and Port-related Facilities by the 1995 Hyo-
goken- nanbu Earthquake, The Port and Harbor Research Institute, Ministry of Transport, Japan
Hanshin Expressway Public Corporation, 1996, Investigation on the Seismic Damages of Bridge Founda-
tion in the Reclaimed Land
Sato, M., Ogasawara, M., Tazoh, T., 2001, Reproduction of lateral ground displacements and lateral-flow
earth pressures acting on pile foundations using centrifuge modeling, Fourth International Conference
on Recent Advance in Geotechnical Earthquake Engineering and Sail Dynamics, San Diego, USA,
6pp.(CD-Rom), Paper No.9.33.
Tazoh, T., Sato, M., Gazetas, G., 2005, Centrifuge tests on pile-foundation-structure systems affected by
liquefaction-induced flow due to quay-wall collapse, Proceedings of the 1
st
Greece-Japan Workshop
on Seismic Design, Observation and Retrofit of Foundations, pp.79-106.
Tokimatsu, K.., Mizuno H., Kakurai M., 1996, Building Damage Associated with Geotechnical Prob-
lems, Special Issue on the 1995 Hyogoken-nanbu Earthquake, Soils and Foundations, p.p. 219-234.
Yasuda, S., Ishihara, K., Morimoto, I., Orense, R, Ikeda, M., and Tamura, S., 2000, Large-scale shakingt-
able tests on pile foundations in liquefied ground, Proceedings, 12th World Conference on Earth-
quake Engineering, Auckland, New Zealand, Paper No. 1474, 8pp.

95
1 INTRODUCTION
Many large earthquakes have caused severe damage to various structures. Especially in the
1995 Hyogoken-Nambu earthquake, a lot of pile-supported structures behind quay walls in port
areas collapsed due to widespread liquefaction and its resulting lateral spreading of ground.
Such structures are necessary as facilities for rescue supplies to disaster areas as well as eco-
nomical distribution systems during rehabilitation process after the disaster. Therefore, under-
standing the lateral spreading behavior of liquefied ground and its influence so as to mitigate
earthquake disaster for port structures is one of the very important problems in geotechnical
earthquake engineering discipline. In spite of its importance, reproduction of a lateral spread-
ing phenomenon in the model ground on a centrifuge apparatus or shaking table is almost im-
possible because mainly of the small model size, so that the influences of lateral spreading on
structures still have not been completely revealed. In other words, it is necessary to conduct
shaking tests of a large-scale model ground with a structure subjected to actual magnitude
earthquake motions to understand the behavior and influences of lateral spreading.
For this reason, the authors carried out a series of tests of a large-scale model ground with a
quay wall and group-pile-supported structure on lateral spreading due to liquefaction by the E-
Defense shaking table (MEXT 2007). The objective of the testing series is reproduction of li-
quefaction-induced lateral spreading to observe the phenomena of the model ground and struc-
tures in detail. This experimental study involves revealing the failure mechanism of a quay
wall and pile-supported structure behind the wall. The paper here describes the results of one
of the tests of the model ground with a caisson-type quay wall and pile-supported structure, and
explains their behaviors.
E-Defense Shaking Table Test on Liquefaction-Induced Lateral
Spreading of Large-Scale Model Ground with Quay Wall and
Pile-Supported Structure
K. Tabata & M. Sato
National Research Institute for Earth Science and Disaster Prevention, Tsukuba, Japan
ABSTRACT: A shaking table test of a large-scale model was performed at the E-Defense
three-dimensional shaking table facility in order to observe the behavior of lateral spreading of
liquefiable ground and to evaluate the mechanism of its influence on the failure of structures.
The model ground was prepared in a large rectangular container with a caisson-type quay wall,
pile-supported structure and about nine hundred sensors. To this model, horizontal and vertic-
al input motions based on one of the 1995 Hyogoken-Nambu earthquake records were applied.
The motions induced liquefaction of the ground and horizontal displacement of the caisson to
waterside, causing deformation of the structure due to the piles bent and collapsed. Observa-
tion of the test explains contributions of structural inertia and ground deformation to the beha-
vior of the caisson and structure.
96
2 OUTLINE OF THE TEST
2.1 E-Defense shaking table
The test was carried out by the E-Defense shaking table. E-Defense is the name of a full-scale
three-dimensional earthquake testing facility operated by the National Research Institute for
Earth Science and Disaster Prevention (Ohtani et al. 2003). After the start of its operation in
2005, a lot of shaking tests on soil-structure interaction and collapse behavior of various struc-
tures have been successfully performed (Kajiwara et al. 2007, Tabata & Kajiwara 2009). The
key of the E-Defense is a shaking table, which is the worlds largest three-dimensional table 20
m long and 15 m wide as shown in Figure 1. The remarkable feature of the table presented in
Table 1 is the capability to reproduce ground motions recorded in the 1995 Kobe earthquake for
a 12-MN structure by 10 horizontal and 14 vertical actuators.

Table
14 vertical
actuators
10 horizontal
actuators
2
0
m
lo
n
g
15m
w
ide

Figure 1. Illustration and photo of the E-Defense shaking table.

Table 1. Specifications of the E-Defense shaking table.
Horizontal (x and y) Vertical (z)
acceleration 900gal 1500gal
velocity 2m/s 0.7m/s
displacement 1m 0.5m
allowable moment 150MNm 40MNm
* at the maximum load
Table size
Loading capacity
Maximum performance*
20m x 15m
12MN

2.2 Specimen of the model ground
Figure 2 and Photo 1 show the specimen of the model prepared in a rectangular container 16 m
long, 4 m wide and 5 m high. The specimen was a liquefiable ground with a caisson-type quay
wall and structure supported by a 3-by-2 pile group. The liquefiable ground was made of Al-
bany silica sand compacted to 60-percent relative density and saturated by de-aired water before
testing. The properties and indices of the sand are presented in Table 2, and its deformation
characteristics are described by Yasuda et al. (2006). The grain size distribution of the sand is
also shown in Figure 3, indicating that the distribution is similar to that of Toyoura sand. The
deposit was divided by the caisson into 2.5-meter-thick waterside and 4.5-meter-thick land-
side deposits. The water table was 0.5 m below the landside ground surface, i.e. 1.5 m above
the waterside ground surface. In the landside deposit, the pile-supported structure was in-
stalled behind the caisson, consisting of six hollow steel piles, footing and the weight modeling
a superstructure. The six piles were aligned with three parallel to the caisson in two rows and
fixed to the footing and pinned at the containers bottom. The 10-ton footing penetrated into
the landside deposit to a depth of 0.5 m. The weight was placed on the footing, inducing iner-
tial force by a 12-ton weight.

97
1
6
0
0
WATER TABLE
4
0
0
0
5
0
0
2
0
0
0
7
5
0
16000
4
5
0
0
4
0
0
0
1300 4400 9300
1
7
5
0
CAISSON
MOUND
4
0
0
1600
PILE
LAYOUT
1000
2
0
0
0
9
6
0
9
6
0
CAISSON
1600
3550
600 250 800 600
Dr = 60%
Dr = 90%
SUPER-
STRUCTURE
D F
D F
PILE
FOOTING
UNIT: mm
(a) Plan view
(b) Section view
PILE A2
PILE B2
WATERSIDE
WATERSIDE
LANDSIDE
LANDSIDE
y
z
y z

Figure 2. Illustrations of the specimen and the locations of the measurement points D and F: (a) Plan
and (b) section views.

Photo 1. Specimen on the table before shaking.

Table 2. Properties and indices of Albany silica sand.

Figure 3. Grain size distributions of Albany silica sand and Toyoura sand.
98
2.3 Measurement installation
To achieve the objective , measurement sensors listed in Table 3 were installed with the speci-
men. The table presents a total of 836 sensors that were mounted to observe the behavior in
detail. Additionally, a unique three-dimensional displacement measuring system was em-
ployed to investigate the motions of ground surface and structures (Tokuyama et al. 2007). In
this system, displacement is determined by digital video cameras that observed reflective, spher-
ical markers attached to an object.

Table 3. Sensors installed with the specimen.
Type of sensor

G
r
o
u
n
d

C
a
i
s
s
o
n

P
i
l
e
s

S
u
p
e
r
s
t
r
u
c
t
u
r
e

a
n
d

f
o
o
t
i
n
g

C
o
n
t
a
i
n
e
r

T
o
t
a
l
Strain gauge 234 40 274
Accelerometer 89 8 22 24 24 167
Veolcity transducer 3 3 6
Displacement transducer 11 14 11 36
Earth pressure transducer 17 104 16 137
Water pressure transducer 119 7 72 198
Load cell 18 18

2.4 Testing program
The specimen set on the table was shaken under two-dimensional, horizontal and vertical mo-
tions based on the north-south and up-down motions recorded at the JR Takatori station during
the 1995 Kobe earthquake. In the test, the north-south component was applied to the speci-
mens long direction, and the up-down component to the vertical direction. The peak table ac-
celerations were approximately 6.0 and 1.7 m/s
2
in the horizontal and vertical directions respec-
tively, and the shaking duration was about 42 seconds. Figure 4a shows the acceleration time
histories of the target input motion and actual, observed table motion from zero (data acquisition
start time) to 30 seconds. Because the nature of the specimen was significantly changed due to
liquefaction caused by strong motions, the tables control system could hardly follow such
changes and reproduce table motions that were identical to the target. However, as shown in
Figure 4b, both shapes of Fourier spectrum are very similar, especially in the domain of lower
frequency.

Figure 4. Comparison between the target input motion signaled to the table and actual, observed table
motion: (a) Acceleration time histories and (b) acceleration Fourier amplitude spectra.
99
3 TEST RESULTS
3.1 Observation of the specimen after test completion
Photo 2 shows the specimen before and after the test, and Figure 5 illustrates the section of the
specimen that demonstrates its change due to shaking. The caisson overturned toward the wa-
terside (left of the specimen in the photo) with the horizontal displacement at the top of about 2
m and 25-degree decline, accompanying its mound with horizontal deformation and small set-
tlement. Three waterside piles bent at almost the same level of the mound. This, in turn,
caused horizontal deformation of the pile-supported structure toward the caisson with 47-degree
decline of the weight and footing.
As shown in Figure 5, horizontal displacement of the landside ground surface ranged up to 2
m and decreased with the distance from the caisson. Such trend is similar to some cases ob-
served in the 1995 Kobe earthquake (Ishihara et al. 1996). The liquefaction and resulting lat-
eral spreading also caused settlement of the landside deposit of about 20 cm, while relatively
small settlement was observed at the waterside deposit.

Photo 2. Caisson, pile-supported structure and their surroundings (a) before shaking and (b) after test
completion.

Dr = 60%
Dr = 90%
WATERSIDE LANDSIDE
y
z

Figure 5. Section of the change of the specimen due to shaking. Gray and black colors indicate the
structures and measurement points before and after the test respectively.
3.2 Excess pore water pressure change during and after shaking
Figure 6 shows the changes of excess pore water pressure, u, at four different levels (0.6, 1.7,
2.9 and 3.7 m from the landside surface) under the measurement point F. As shown in the fig-
ure, u build-up occurred at all levels when the excitation started, and then reached their over-
burden pressure in 7 to 10 seconds after the start. In consequence, it can be said that all layers
of the deposit were liquefied due to the applied input motions. Following this u increasing
process, u dissipation started from the bottom layer to the surface. This observation implies
that liquefied deposit became dense from the bottom to the surface during the dissipation
process of u. In addition, u began to dissipate when its pressure level reached that of the
lower layer.
(a) (b)
100

Figure 6. Excess pore water pressure changes at different levels under the point F.
3.3 Behavior of the caisson during shaking
Just after start of the shaking, the caisson moved to the waterside and landside alternately a few
times, and finally overturned toward the waterside at 10.6 seconds after the test start. To eva-
luate the effects of its inertial force and backfill earth pressure change on the caissons behavior,
the time histories of the displacements and effective earth pressure changes are introduced as
following. Figure 7 shows the vertical displacements, z, of the waterside and landside tops of
the caisson, presenting decline of the caisson began at 6 seconds. Figure 8 shows the horizon-
tal displacements, y, of the caisson, the measurement point D on its backfill, and the shaking ta-
ble. In the figure, a positive y value means the displacement to the waterside. As shown in
the figure, the horizontal displacement of the caisson is always larger than that of the point D
with almost opposite phase of the table. Additionally, the effective earth pressure change on
the caissons landside, p', at different levels from the surface shown in Figure 9 is constantly
negative from 5 seconds to 10.6 seconds at which the caisson overturned toward the waterside.
These facts suggest that the behavior of the caisson during shaking was principally dominated
by its inertial force rather than the p' change of the backfill. Indeed, many cracks on the
ground surface and gaps between the caisson and its backfill are observed in Photo 3, which is
the close view of the caisson and its surroundings including the footing and the point D around
at 8 seconds. Therefore, it can be said that the influence of caissons inertial forces is more
dominant to the stability of the caisson under earthquake motions than that of the p' change of
its backfill.

Figure 7. Vertical displacements determined at the waterside and landside tops of the caisson.

Figure 8. Horizontal displacements of the caisson, point D and shaking table.
101

Figure 9. Effective earth pressure changes on the backfill side of the caisson.

Waterside Landside
Caisson
Point D Footing

Photo 3. Close view of the caisson, backfill surface and footing during shaking.
3.4 Behavior of the pile-supported structure during shaking
Figure 10 shows horizontal displacements of the caisson, weight of the pile-supported structure
and their surroundings including the measurement points D and F. As already explained, the
caisson overturned toward the waterside at 10.6 seconds after the test started. Until the caisson
overturned, the horizontal displacement of the point D on the backfill of the caisson was always
smaller than that of the caisson, while the displacement of the weight was smaller than those of
the caisson and point D and almost same behavior of the point F on the ground behind the pile-
supported structure. It is also observed that these displacements change with nearly same
phase. Just after the caisson overturned at 10.6 seconds, the displacement of the weight
reached the exactly same of the point D and increased, and finally declined to the waterside.
In this process, the displacement of the point F was smaller than those of the weight and point
D. This observation implies that one of the triggers to decline the pile-supported structure can
be the large deformation of the ground between its footing and the caisson due to the caisson
overturned.

Figure 10. Horizontal displacements of the caisson, weight, and points D and F.
102
4 CONCLUSIONS
In order to investigate the behavior of liquefaction-induced lateral spreading and the mechanism
of its influence on structures, a shaking table test of a large-scale model ground with a caisson-
type quay wall and group-pile-supported structure assumed as a situation of port areas was con-
ducted at the E-Defense shaking table facility. In the test, about nine hundred sensors moni-
tored the behavior in detail and the three-dimensional displacement measuring system was em-
ployed to observe large displacements.
The specimen of the model ground was shaken under two dimensional, horizontal and vertic-
al motions based on one of the 1995 Kobe earthquake records. Such motions induced lique-
faction in all layers of the deposit, causing overturn of the caisson toward the waterside and ho-
rizontal deformation of the landside ground. According to the displacement and effective earth
pressure change of the caisson, the overturn was mainly caused by its inertial force, not the in-
fluence of its backfill ground deformation. After the caisson overturned and the following de-
formation of its backfill ground caused, the pile-supported structure declined to the waterside.
It can be considered that this phenomenon was triggered by large deformation of the ground be-
hind the caisson. Note that these observations and considerations are based on the model test-
ing results, meaning that the observed area is still limited compared to the actual field. Hence,
to apply the knowledge from this study to practice, other research procedures, such as parame-
tric testing of small models, field investigations and computational analyses, as well as more
large-scale model testing are needed.
ACKNOWLEDGEMENT
The testing program described here is a part of the Special project for mitigation of earthquake
disaster in urban areas funded by the Japanese Ministry of Education, Culture, Sports, Science
and Technology. The motions applied to the shaking table were based on the record obtained
at the JR Takatori station during the 1995 Kobe earthquake that is originally from the Railway
Technical Research Institute in Japan. These supports and contributions are gratefully ac-
knowledged.
REFERENCES
Ishihara, K., Yasuda, S. & Nagase, H. 1996. Soil characteristics and ground damage. Special issue, Soils
and Foundations: 109-118.
Kajiwara, K., Sato, M. & Nakashima, M. 2006. Shaking table and activities at E-Defense. Proc. 1st Eu-
ropean Conference on Earthquake Engineering and Seismology: Paper no.733.
Ministry of Education, Culture, Sports, Science and Technology (MEXT). 2007. Summary Report, Spe-
cial project for mitigation of earthquake disaster in urban areas: 66-75 (in Japanese).
Ohtani, K., Kagawa, N., Katayama, T. & Shibata, H. 2003. Construction of E-Defense (3-D full-scale
earthquake testing facility). Proc. 2nd Intern. Symp. New Technologies for Urban Safety of Mega Ci-
ties in Asia: 69-76.
Tabata, K. & Kajiwara, K. 2009. Experimental research on behavior of various structures under earth-
quake motions at E-Defense. Proc. 2nd China-Japan Science Forum, Beijing, China: 205-206.
Tokuyama, H., Tabata, K., Nakazawa, H. & Sato, M. 2007. Applicability of three-dimensional displace-
ment measurement to model ground in large-scale shaking table testing at E-Defense. Proc. 42nd An-
nual Symp. Japanese Geotechnical Society: 1643-1644 (in Japanese).
Yasuda, S., Saito, S. & Suzuki, S. 2006. Effect of confining pressure on liquefaction behavior of sand.
Proc. 61st Annual Symp. Japanese Society of Civil Engr.: 547-548 (in Japanese).
103
Design charts for single piles under lateral spreading of liquefied
soil

A. Valsamis
Dr Civil Engineer, Civil Engineering School, National Technical University of Athens, Greece
G. Bouckovalas
Professor, Civil Engineering School, National Technical University of Athens, Greece
E. Drakopoulos
Civil Engineer Msc, National Technical University of Athens, Greece
1 INTRODUCTION
ABSTRACT: Multi-variable design charts and relationships are presented for the preliminary
computation of maximum pile displacement and bending moment in the case of earthquake-
induced liquefaction and lateral spreading of the natural ground. The charts were initially devel-
oped on the basis of results from a large number of parametric analyses which were performed
with the pseudo static (P-y) method, and consequently calibrated with the aid of 2-D and 3-D
dynamic analyses of the liquefied soil and the pile. Three different combinations of pile and
soil conditions were considered, which are commonly encountered in practice.

One of the most damaging effects of earthquake-induced soil liquefaction is the lateral spread-
ing of soils, where large areas of ground move laterally to lengths ranging from some centime-
ters to a few meters. This phenomenon may occur in the case of even small surface inclination
(e.g. 24%) or small topographic irregularities (e.g. 23m) such as those near river and lake
banks.
In such cases, the kinematic interaction of single piles and pile groups with the lateral
spreading ground may induce significant additional residual horizontal loads and bending mo-
ments to the pile, which cannot be predicted by common design methods for superstructure
loading.

2 PSEUDO-STATIC PREDICTION METHODS
The detailed analysis of piles against lateral spreading is a rather complicated soilstructure in-
teraction problem which, strictly speaking, requires a sophisticated numerical simulation, well
beyond the limits of common applications. Thus, for simplified computations, a number of
pseudo-static methodologies have been developed, where the loads or displacements applied by
the laterally spreading ground are being estimated independently and subsequently applied as
external loads to the pile. Existing pseudo-static methodologies may be divided in two catego-
ries:
a) The P-y method, which relies upon the substitution of the ground with Winkler
type springs that are governed by a non-linear load-displacement (P-y) law. Ac-
cording to this methodology an independent estimation of the ground displacement
is made and the resulting displacements are applied to the base of the springs in or-
der to evaluate the pile deflection and the corresponding shear forces and moments
(e.g. Tokimatsu 1999, Boulanger et al 2003).
b) The limit equilibrium method, which is based on a pseudo-static estimation of the ul-
timate pressure that the laterally spreading ground applies to the pile. Pile displace-
ments and bending moments can be consequently evaluated (e.g. JRA 1996, Dobry
et al 2003) from beam theory.
104
Recently, Ashford & Juirnarongrit (2004) concluded that the P-y method is the most reliable
method of the two, after comparing the two most commonly used limit equilibrium methods
(JRA, 1996 and Dobry et al., 2003) with a simple P-y method that used the curves proposed
from Reese et al. (1974) for sands, degraded with a factor = 0.1 in order to take into account
the soil liquefaction. Bhattacharya et al. (2003) also concluded that the limit equilibrium method
of JRA (1996) is systematically non-conservative. Thus, on the ground of these independent
findings, the P-y method has been chosen to derive the design charts in this paper.
More specifically, the method proposed by Branderberg (2002) has been selected, according
to which the P-y curves of (1995) for the non-liquefied sands should be used, after being
degraded with a loading factor . This factor represents the effect of liquefaction on the me-
chanical characteristics (soil strength and deformation) of the natural soil and can be computed
according to Table 1, in terms of the corrected blow count of the Standard Penetration Test
(N
1
)
60-CS
.
The aforementioned methodology has been chosen among seven (7) compatible methodolo-
gies (Ishihara & Cubrinovski, 1998, Cubrinovski et al., 2006, Rollins et al., 2005 & 2007, To-
kimatsu, 1999, High Pressure Gas Safety Institute of Japan, 2000, Railway Technical Research
Institute of Japan, 1999, and Matlock, 1970) following an extensive evaluation through com-
parison to three centrifuge experiments (Abdoun 1998) and one large shaking table experiment
(Cubrinovski et al. 2004).

Table 1. Proposed degradation factors after Branderberg (2000)

(N
1
)
60-CS

<8 0 to 0.1
8-16 0.1 to 0.2
16-24 0.2 to 0.3
>24 0.3 to 0.5

3 PARAMETRIC ANALYSES
The numerical analyses have been performed with the help of the finite elements program
NASTRAN (MacNeal-Schwendler Corp. 1994). Simulation of the liquefied soil layers was
based on the P-y methodology outlined in the previous paragraph. The non-liquefied soil layers
have been simulated with the P-y curves proposed by API (1995, 2002) without the use of a
degradation factor. It should be mentioned that, as long as the non-liquefiable base layer does
not fail, the exact P-y curve used for its simulation does not affect significantly the results, as its
stiffness is almost 100 times larger than that of the liquefied soil above it.
Based on a previous study of the lateral spreading phenomenon (Valsamis et al., 2007), the
variation with depth of the lateral displacements of the liquefied soil was assumed as a quarter
sine, with the maximum displacement developing near the top of the layer and zero displace-
ment at the bottom of the layer. On the other hand, the displacement of the non-liquefied soil
layers was assumed to remain constant with the depth.
In total, one hundred sixty two (162) parametric analyses have been performed, concerning
three different combinations of piles and ground layer profile (Figure 1):
o A 2-layered soil profile, where there is a surface liquefied soil layer which spreads lat-
erally, while the pile rests inside a non-liquefiable bottom layer. This is a common case
near river or lake banks, where there are loose surface alluvial deposits.
o A 2-layered soil profile, same as above, where the pile head is fixed in position, due to
superstructure constraints. This case is often encountered in piles supporting bridges or
other large structures, where the superstructure restrains the pile head from moving.
o A 3-layered soil profile, similar with the 2-layered profile with the addition of a non-
liquefiable soil crust. This is the most common case in small distances from water
fronts, where the water table is some meters below the soil surface, leaving a non-
liquefiable surface soil crust.

105
This categorization is justified on the grounds that any possible constraints on the free pile
head displacement and rotation, either due to the non-liquefiable soil crust or a superstructure,
proved to be among the the most important factors controlling the pile response. Note that, Ishi-
hara & Cubrinovski (1998), Brandenberg (2002), Rollins et al. (2005) have also reached similar
conclusions for the effect of the pile head constraint enforced by a non-liquefiable soil crust.
Sixty six (66) parametric analyses have been performed for the 2-layerd case and the follow-
ing pile & soil input parameters:
o Relative Density Dr=35 90 % (degradation factor = 0.05 0.4 and friction angle
= 32
o
42
)
o Thickness of liquefied soil layer
liq
= 6 to 10m
o Elasticity Modulus for the pile ( = 30 to 210 GPa)
o Pile diameter (D = 0.15m to 0.6m),
o Pile stiffness = 16 to 1336 Mm2, and
o Ground surface displaments D
h
= 0.125m to 1.20m.

For the 2-layered geometry with fixed pile head, forty-six (46) analyses have been performed
with the same range of parameters. For the 3-layer geometry fifty (50) numerical analyses have
been performed. These analyses cover the above mentioned range of parameters in combination
with soil crust thickness crust = 1 to 4m.

Figure 1. Static models for the (a) 2-layered, (b) 3-layered and (c) fixed pile head cases

4 DESIGN CHARTS
Pile design against lateral spreading must assure that, following the seismic excitation:
a) no structural failure of the pile has occurred (no development of plastic hinges at any
depth), and
b) no performance failure of the superstructure should be encountered due to excessive
superstructure displacements.
To check against these criteria, both the maximum developing moment and the maximum dis-
placement of the pile head are needed.
The depth of the maximum bending moment is in general variable. For the cases considered
in this article (Figure 1), it is known before-hand that maximum moments develop at the inter-
face between the liquefied soil layer and the non-liquefied base layer. Similarly, it is known be-
fore hand that the maximum pile displacement develop at the pile head, for the 2- and the 3-
layered cases, and near the mid-depth of the liquefiable soil layer for the fixed pile head case.
For these reasons, the statistical analysis of the parametric analyses results has been focused
upon the magnitude of those two design parameters and not upon the respective location along
the pile.
It should also be mentioned that the statistical processing was not blind, e.g. based only on
some algorithm that minimizes the error of the empirical predictions. On the contrary, a general
form of the prediction relations was initially obtained based on analytical solutions of the static
q
q
()
( )
P
()
(a) (b) (c)
106
models presented in Figures 1a, 1b and 1c and subsequently the statistical processing was used
to calibrate the general relations against the results of the parametric analyses.
Figures 2a and 2b present the proposed design charts for the maximum pile displacement and
the associated bending moment for the 2-layered soil profile. Alternatively, the pile head dis-
placement D
pile
(m) may be computed from Figure 2a and the respective maximum bending
moment
max
(kN/m) may be subsequently estimated as:
2
max
2 . 2
liq
pile
H
EID
M = (1)
where Hliq(m) is the liquefied soil layer thickness and EI(kN/m2) is the pile stiffness.
Observe that the relation in Figure 2a is strongly non-linear. This is due to the fact that the
Winkler springs representing the soil are elasto-plastic and thus, after a certain soil displace-
ment, the loads due to the lateral movement of the soil remain constant. This elastoplastic re-
sponse of the soil springs is the main reason why the derivation of a simple analytical expres-
sion for the pile displacement was not possible. Moreover, note that the correlations of Figure
2a are not dimensionless and thus they should always be used in conjunction with the interna-
tional system unit SI (kN, m).

0.01 0.1 1 10 100
(D
h*
EI)/(H
l i q
6
D
2
)
0.01
0.1
1
10
100
(
D
p
i
l
e
*
E
I
)
/
(
H
l
i
q
6
D
2
)
b=0,4 +D=0,3m
b=0,4 +D=0,15m
b=0,2 +D=0,3m
b=0,2 +D=0,15m
b=0,1 +D=0,6m
b=0,1 +D=0,3m
b=0,05 +D=0,6m
b=0,05 +D=0,3m
(a)
200 400 600 800 1000
(D
pi l e*
EI)/H
l i q
2
0
500
1000
1500
2000
M
m
a
x
b=0.05 & D=0.3m
b=0.1 & D=0.3m
b=0.2 & D=0.3m
b=0.,4 & D=0.3m
b=0.4 & D=0.15m
b=0.2 & D=0.15m
b=0.1 & D=0.6m
b=0.05 & D=0.6m
(b)
Dr
= 50%
Dr
= 65%
D
r
= 85%
Dr
= 40%

Figure 2. Design charts (a) for the maximum pile displacement and (b) for the maximum developing
bending moment of the pile, for the 2-layered soil profile

The design charts for the 3-layered soil profiles are shown in Figures 3a and 3b. Observe that
the pile head displacement follows systematically the non-liquefied soil crust displacement. This
observation has been also confirmed from centrifuge experiments (Abdoun, 1999) which show
that pile head displacements are only slightly larger than soil surface displacements. In this case,
it was possible to develop simplified analytical relations, both for the pile head displacement
and the developing bending moments, namely:
h pile
D D = 22 . 1 (2)
65 . 0
2
max
18
|
|
.
|
\
|
=
liq
pile
H
EID
M (3)
107
0.01 0.1 1 10
D
h
0.01
0.1
1
10
D
p
i
l
e
b=0,4 +D=0,3m
b=0,4 +D=0,15m
b=0,2 +D=0,3m
b=0,2 +D=0,15m
b=0,1 +D=0,6m
b=0,1 +D=0,3m
b=0,05 +D=0,6m
b=0,05 +D=0,3m
(a)
1000 2000 3000 4000
(D
pi l e*
EI)/H
l i q
2
0
1000
2000
3000
4000
M
m
a
x
b=0,05 +D=0,3m
b=0,1 +D=0,3m
b=0,2 +D=0,3m
b=0,4 +D=0,3m
b=0,4 +D=0,15m
b=0,2 +D=0,15m
b=0,1 +D=0,6m
b=0,05 +D=0,6m
(b)

Figure 3. Design charts (a) for the maximum pile displacement and (b) for the maximum developing
bending moment on the pile, for the 3-layered soil profile

Finally, for the fixed pile head case, the design chart for the maximum pile displacement and
bending moment, are being presented in Figures 4a and 4b respectively.
In this case also, it was possible to phrase analytical relations for the estimation of the above
mentioned design parameters:
( )
2
2 2
3 . 0 12
EI
D
D H D
h liq pile
|
= (4)
2
max
18
liq
pile
H
EID
M = (5)
where is the degradation factor for the soil strength due to the liquefaction which can be taken
from Table 1.

0.0001 0.001 0.01 0.1 1 10
(D
h
)
0.3
H
liq
12
(D/EI)
2
0.001
0.01
0.1
1
D
p
i
l
e

(
m
)
b=0,4 +D=0,15m
b=0,3 +D=0,3m
b=0,2 +D=0,3m
b=0,2 +D=0,15m
b=0,1 +D=0,6m
b=0,1 +D=0,3m
b=0,05 +D=0,6m
b=0,05 +D=0,3m
(a)
20 40 60 80 100
EID
p
/H
liq
2
400
800
1200
1600
2000
M

(
k
N
m
)
b=0,4 +D=0,15m
b=0,3 +D=0,3m
b=0,2 +D=0,3m
b=0,2 +D=0,15m
b=0,1 +D=0,6m
b=0,1 +D=0,3m
b=0,05 +D=0,6m
b=0,05 +D=0,3m
(b)

Figure 4. Design charts for the (a) maximum developed pile displacement and (b) the maximum develop-
ing bending moment on the pile, for the fixed pile head case

108
5 COMPARISON WITH DYNAMIC NUMERICAL ANALYSES RESULTS
The proposed design charts have been evaluated with the help of 2-Dimension and 3-Dimension
dynamic numerical analyses. The numerical analyses were performed with the aid of the non-
linear Finite Difference method combined with an effective stress constitutive model which can
simulate the static and the dynamic response of cohesionless soils, including liquefaction. In
brief, this is a bounding surface model, with a vanished elastic region, which was built based
upon the Critical State Theory framework (Papadimitriou et al., 2001, Andrianopoulos et al.
2006a, 2006b and 2007). One of its main characteristics is that the monotonic and cyclic re-
sponse of soils is described using a single set parameters which is soil-specific, but does not de-
pend on the initial stress and density conditions.
The 2-Dimension simulation of a clearly 3-Dimensioned problem with the help of a plane-
strain algorithm, has been achieved with the help of the special pile elements, incorporated in
the finite difference program FLAC-2D (Itasca, 2005). Those elements, are in fact beam ele-
ments which have the ability to connect to the FLAC grid that simulates the surrounding soil
with the help of special Winkler springs, whose maximum strength depends on the mean effec-
tive stress of the surrounding soil. Those elements can simulate in a simple way the liquefac-
tion-induced degradation of the soil characteristic, since their strength depend on the effective
strength and thus can be used for the simulation of piles inside laterally spreading soil (Valsamis
2008). Figure 5a shows a typical 2-Dimension grid used for the simulation of the problem. Sixty
seven (67) such numerical analyses have been performed with the following range of parame-
ters:
o Relative Density of liquefiable soil layer Dr = 35 to 85%
o Thickness of liquefiable soil layer liq = 2 to 8m
o Thickness of non-liquefiable soil crust Hcrust = 0 to 6 m
o Pile stiffness EI = 500 to 80000 kPam4
o Pile diameter = 0.3 to 1.0m
o Ground surface displacement Dh = 0.2 to 1.4m.
o 2-layered and 3-layered soil profiles have been simulated, both with free and fixed
pile head.
Moreover, thirty (30) 3-dimension numerical analyses have been performed, in which the
ground has been simulated with the same constitutive model. The pile in this case was simulated
with a simple elastic model. Between the soil grid and the pile, contact elements have been used
with friction angle equal to 2/3 x (Itasca 1997), which allows the relative movement of the
liquefied ground around the pile. Figure 5b shows a typical grid used for the 3-D simulation of
the problem. 3D numerical analysis has the advantage of simulating directly the problem and
thus not having any predefined parameters such as the coupling springs parameters as with the
case of the 2D numerical analysis. This increases our confidence that the 3D numerical simula-
tion will be able to simulate accurately a large range of parametric lateral spreading problems.
The 3-D analyses cover the following range of parameters:
o Relative Density of liquefiable soil layer Dr = 40 to 85%
o Thickness of liquefiable soil layer liq = 6 to 8m
o Pile stiffness EI = 4 to 16500 GPam4
o Pile diameter = 0.3 to 2.0m
o Ground surface displacement Dh = 0.1 to 0.3m.
Those analyses concern exclusively 2-layered soil profiles, which is the geometry that has the
biggest difficulties in the simulation, since the developed moments and displacements are exclu-
sively due to the forces of the liquefied soil.

109

P
ult
= f (p')
P
y
K
Constitutive
model based
on CS Theory
Pile simulated
as a beam
following an
elastic law
(a)

Constitutive
model based
on CS Theory
Pile simulated
with a simple
elastic model

Figure 5. Typical grids (a) 2-dimension and (b) 3-dimension numerical simulation of single pile under
lateral spreading
(b)

The results of both 2-D and 3-D dynamic numerical analyses, agree in general with the design
charts which were developed by the P-y methodology and have already been presented (Figures
2, 3 and 4). Differences were only encountered concerning the design chart for the evaluation of
the pile displacement in the case of the 2-layered soil geometry with free pile head (Figure 2a).
For this case, Figure 6a presents the results of the 2-D and 3-D numerical analyses together with
the previously presented design chart. As it can be seen, the numerical analyses gave systemati-
cally smaller displacements and thus the design chart presented in Figure 6b should be pre-
ferred.

110
0.01 0.1 1 10 100
(D
h*
EI)/(H
l i q
6
D
2
)
0.01
0.1
1
10
100
(
D
p
i
l
e
*
E
I
)
/
(
H
l
i
q
6
D
2
)
FLAC 2D Dr=45%
FLAC 3D Dr=45%
FLAC 3D Dr=65%
FLAC 3D Dr=85%
0.01 0.1 1 10 100
(D
h*
EI)/(H
l i q
6
D
2
)
0.01
0.1
1
10
100
(
D
p
i
l
e
*
E
I
)
/
(
H
l
i
q
6
D
2
)
FLAC 2D Dr=45%
FLAC 3D Dr=45%
FLAC 3D Dr=65%
FLAC 3D Dr=85%
D
r
= 50%
Dr
= 65%
D
r
= 85%
D
r
= 40%
D
r

=

4
0
%
D
r

=

6
5
%
D
r

=

8
5
%

(a) (b)
Figure 6. (a) Comparison between the 2-D and 3-D predictions for the 2-layered soil profile and the design
chart from the P-y analyses (b) Final design chart for the ground surface displacement in 2-layered soil pro-
files
6 CONCLUSIONS
In the previous paragraphs, diagrams and relations were presented for the approximate evalua-
tion of the maximum displacement and bending moment of the pile due to liquefaction-induced
lateral spreading. The charts concern three different combinations of pile and ground conditions,
often encountered in practice. The proposed design charts and relations should be used with the
following limitations:
(a) They were derived pseudo-statically, taking only into account the final displacement of
the ground, at the end of shaking. Any effects of the superstructure inertia are ignored.
(b) The expected free-field maximum ground surface displacement should be computed in-
dependently, based on the (many) available empirical relations which are published in the litera-
ture (e.g. Hamada, 1999, Youd et al, 2002, Valsamis, 2009).
(c) All the above mentioned charts and relations, and more specifically those concerning
the 2-layered soil profile case, should be applied only when the soil has the capability to flow
freely around the pile under investigation. In all other cases (e.g. small distance between the
piles, sheet-pile wall, etc) they may lead to non-conservative predictions of the pile displace-
ment and bending moment.
(d) It has been assumed that the pile has been adequately embedded to the non-liquefiable
base soil layer so as to guarantee fixed bottom conditions during lateral ground spreading. When
the pile has not been driven adequately to the bottom (non-liquefiable) soil layer, there is the
possibility of pile extortion or significant pile base rotation which results to larger displacements
for the pile head and smaller developing moments.
7 REFERENCES
Abdoun T. H. (1999) Modeling of seismically induced lateral spreading of multi-layered soil and its ef-
fect on pile foundations, PHD Thesis, Rensselaer Polytechnic Institute, Troy, New York
Andrianopoulos, K.I. (2006), Numerical modeling of static and dynamic behavior of elastoplastic soils,
Doctorate Thesis, Department of Geotechnical Engineering, School of Civil Engineering, National
Technical University of Athens (in Greek).
Andrianopoulos, K.I., Papadimitriou, A.G. and Bouckovalas, G.D. (2006), Implementation of a bound-
ing surface model for seismic response of sands, Proceedings of the 4th International FLAC Sympo-
sium on Numerical Modeling in Geomechanics, Madrid, Spain
111
Andrianopoulos, K.I., Papadimitriou, A.G. and Bouckovalas, G.D. (2007), Use of a new bounding sur-
face model for the analysis of earthquake-induced liquefaction phenomena, paper no 1443, Proceed-
ings of 4th International Conference on Earthquake Geotechnical Engineering.
API (1995), Recommended practice for planning, designing and constructing fixed offshore platform,
Washington, DC: American Petroleum Institute.
API (2002), Recommended practice for planning, designing and constructing fixed offshore platform,
Washington, DC: American Petroleum Institute.
Ashford S. A. & Juirnarongrit T. (2004), Evaluation of force based and displacement based analyses for
responses of single piles to lateral spreading, 11
th
International conference on Soil dynamics &
earthquake engineering, 3
rd
International conference on earthquake geotechnical engineering, 7-9
January 2004, Berkeley
Bhattacharya S. (2003), Pile instability during earthquake liquefaction, PHD Thesis, University of
Cambridge, UK.
Boulanger R.W., Kutter B.L., Brandenberg S.J., Singh P. and Chang D. (2003), Pile foundations in liq-
uefied and lateral spreading ground during earthquakes: Centrifuge experiments and analyses Report
No. UCD/CGM-03/01, Univ. of California at Davis.
Boulanger R.W., Wilson D.W., Kutter B.L. and Abghari, A. (1997), "Soil-pile-superstructure interaction
in liquefiable sand", Transportation Research Record No. 1569, TRB, NRC, National Academy Press,
55-64
Brandenberg S.J. (2002), Behavior of Pile Foundations in Liquefied and Laterally Spreading Ground,
PHD Thesis, University of California, Davis
Cubrinovski M, Kokusho T. & Ishihara K. (2004), Interpretation from Large-Scale Shake Table Tests
on Piles subjected to Spreading of Liquefied Soils, 11
th
International conference on Soil dynamics &
earthquake engineering, 3
rd
International conference on earthquake geotechnical engineering, 7-9
January 2004, Berkeley
Cubrinovsky M. , T. Kokusho and K. Ishihara (2006), Interpretation from large scale shake table tests
on piles undergoing lateral spreading in liquefied soils Soil Dynamics and Earthquake engineering,
vol.26
Dobry, R., Abdoun, T., O Rourke T.D., Goh S.H. (2003), Single piles in lateral spreads: Field Bending
Moment Evaluation, ASCE Journal of Geotechnical and Geoenvironmental Engineering, Vol. 129,
No. 10, October, pp. 879889
Hamada M. (1999), Similitude law for liquefied-ground flow, Proceedings of the 7th U.S.-Japan
Workshop on Earthquake Resistant design of lifeline facilities and countermeasures against soil lique-
faction, pp. 191-205.
High Pressure Gas Safety Institute of Japan (2000), Design method of foundation for Level 2 earthquake
motion, (In Japanese)
Ishihara K. & Cubrinovski M. (1998), Soil-pile interaction in liquefied deposits undergoing lateral
spreading, XI Danube-European Conference, Croatia, May 1998
Itasca (2005), FLAC version 5.0: Fast Langrangian Analysis of Continua, Itasca Consulting Group,
Minneapolis, Minnesota.
Itasca (1997), FLAC3D version 2.0: Fast Langrangian Analysis of Continua in 3 Dimentions, Itasca
Consulting Group, Minneapolis, Minnesota.
Japan Road Association (1996), "Specifications for highway bridges", Part V Seismic Design
Karamitros, D.K. (2009), Development of a numerical algorithm for the dynamic elastoplastic analysis
of geotechnical structures in two (2) and three (3) dimensions, PHD thesis , N.T.U.A.
The MacNeal-Schwendler Corporation (1994), MSC/NASTRAN for Windows: Reference Manual
Matlock, H. (1970). Correlations of design of laterally loaded piles in soft clay. Proc. Offshore Tech-
nology Conference, Houston, TX, Vol 1, No.1204, pp. 577-594.
Papadimitriou A., Bouckovalas G. and Dafalias Y. (2001), A plasticity model for sand under small and
large cyclic strains, Journal of Geotechnical and Geoenviromental Engineering, Vol.127, No. 11
Railway Technical Research Institute (1999), Earthquake resistant design code for railway structures,
Maruzen Co. (in Japanese)
Reese L.C. and Van Impe W. F. (2001), "Single piles and pile groups under lateral loading", A.A.
Balkema/Rotterdam/Brookfield, Book p.p.463.
Rollins K.M, Gerber T.M., Lane J.D. and Ashford S.A. (2005), "Lateral resistance of a full-scale pile
group in liquefied sand", ASCE Journal of Geotechnical and Geoenviromental Engineering, Vol 131,
No. 1, January, pp. 115-125.
Rollins K.M., Bowles S., Brown D. & Ashford S. (2007), Lateral load testing of large drilled shafts af-
ter blast-induced liquefaction, 4
th
International Conference on Earthquake Geotechnical Engineering,
Paper no 1141, June 25-28, Thessaloniki, Greece
112
Valsamis A., Bouckovalas G. & Dimitriadi V., (2007), Numerical evaluation of lateral spreading dis-
placements in layered soils, 4
th
International Conference on Earthquake Geotechnical Engineering,
Thessaloniki, Greece, June 25-28
Valsamis (2008), Numerical simulation of single pile response under liquefaction-induced lateral
spreading, Doctorate Thesis, Department of Geotechnical Engineering, School of Civil Engineering,
National Technical University of Athens.
Youd L. T., Hansen M. C. and Bartlett F. S. (2002), "Revised multilinear regression equations for predic-
tion of lateral spread displacement", Journal of Geotechnical and Geoenvironmental Engineering, Vol.
128, No. 12, December 1, pp 1007-1017.
Youd L. T., Idriss I. M., Andrus R.D., Arango I., Castro G., Christian J.T., Dobry R., Finn W.D.L.,
Harder L. F. jr, Hynes M. E., Ishihara K., Koester J. P., Liao S.S.C., Marcuson W.F. III, Martin G.R.,
Mitchell J. K., Moriwaki Y., Power M.S., Robertson P.K., Seed R. B. and Stokoe K.H. II (2001),
Liquefaction resistance of soils: summary report from the 1996 NCEER and 1998 NCEER/NSF
Workshops on evaluation of liquefaction resistance of soils, Journal of Geotechnical and Geoenvi-
ronmental Engineering, Vol. 127, No. 10, October, pp 817-833.

113
1 INTRODUCTION
Pile foundations may sustain significant damage due to soil liquefaction, especially when soil
flow occurs. The problem became intensely apparent during the 1995 Great Kobe Earthquake.
Extensive soil flow was triggered along river and sea sides, causing seaward displacements to
sheet pile quay-walls in the Kobe port area accompanied by considerable translation of the
neighboring pile foundations (Ishihara, 1997, Tokimatsu et al., 1997, Yasuda et al., 1996), illus-
trated in Figure 1.

soilflow soilflow

Figure 1. Soil liquefaction along sea sides may cause significant seaward displacement to quay walls af-
fecting the neighboring pile foundations.

Numerous numerical and experimental researches has been performed during the last decade
in order to figure out the mechanism of soil-pile interaction under soil flow conditions. The
moving soil mass provides the driving force to the pile and displaces the pile a certain amount
Piles in LiquefactionInduced Soil Flow behind QuayWall: A
Simple Physical Method versus Centrifuge Experiments

P. Tasiopoulou
Research Ast., University of California, Davis, USA
N. Gerolymos
Lecturer, National Technical University, Athens, Greece
T. Tazoh
Director, Institute of Technology, Shimizu Corporation, Tokyo, Japan
G. Gazetas
Professor, National Technical University, Athens, Greece
ABSTRACT: The paper presents a new physically simplified methodology for computing dis-
placements and internal forces on piles under conditions of lateral spreading. The results com-
pare well with results from centrifuge tests. To this end, 2D effective stress dynamic analysis of
a cross-section of the wall-soil system without the presence of the piles is combined with an also
2D quasi-static analysis of a horizontal slice of the system with the group of piles.
114
depending on the relative stiffnesses between the pile and the liquefied soil (Boulanger et al.,
2003), as depicted in Figure 2. Thus, the magnitude of the soil movement, the lateral load of the
surficial non-liquefiable soil layer and the stiffness degradation in the liquefied zone are the key
parameters that need to be taken into account when evaluating the pile response due to soil flow
(Cubrinovsky et al., 2004).

Beforeliquefaction Afterliquefaction
flow
Beforeliquefaction Afterliquefaction
flow
K

Figure 2. The moving soil mass provides the driving force to the pile and displaces the pile a certain
amount depending on the relative stiffnesses between the pile, K, and the liquefied soil.

In engineering practice, several methods have been formulated based on these understand-
ings, either for designing purposes or prediction of field performance. In general, the methods
can be classified into three categories: (a) the force methods, including the Japanese Road Asso-
ciate Method (JRA, 1996), the limit equilibrium method (Dobry and Abdoun, 2000) and the vis-
cous fluid method (Hamada 2002, Yasuda 2002), (b) the displacement methods, or else known
as pseudo-static beam on nonlinear Winkler foundation method according to which the free
field soil displacement is imposed to the pile through p-y springs (Boulanger et al., 2003), (c)
the hybrid forcedisplacement methods, which are a combination of the first two (Cubrinovsky
and Ishihara, 2004).
The aforementioned methods are mainly single pile analyses dependent on the soil profile and
the geometry of the problem. Moreover, lots of assumptions are required regarding the stiffness
degradation in the liquefied layer and soil-pile interaction issues, such as the direction of the
load exerted on the pile by the upper non-liquefiable layer. Inevitably, considerable uncertainty
is hidden behind all methods of post liquefaction analysis (Finn and Thavaraj, 2001).
In this study, we present a new physically simplified methodology, appropriate for every soil
profile and type of pile configuration (single piles, pile groups). This methodology, described
below, falls into the displacementmethod category, but avoids the associated empirical selec-
tion of stiffnessreduction factors and does not involve the use of p-y curves. Continuously, the
method is applied to two different centrifuge experiments (Tazoh et al., 2005, Sato et al., 2001),
reproducing the test results with satisfying accuracy for engineering purposes.
2 CENTRIFUGE EXPERIMENTS
Several series of dynamic centrifuge experiments were conducted at the Institute of Technol-
ogy, Shimizu Corporation, in Japan, in order to evaluate the damage of pilefoundation systems
triggered by liquefactioninduced soil flow after quay-wall collapse. One of these, Test Case
CD, presented by Tazoh et al. (2005), explores the effect of the superstructure on the response
115
of a 2x2 pile group under soil flow conditions, portrayed in Figure 3. A partition was placed at
the center of the laminar box behind the sheet pile quay-wall in order to separate the two pile
foundation models: the one with a superstructure (C side) and the other without a superstructure
(D side), as depicted in Figure 3.

0 m
1.8 m
6.0 m
8.1 m
9 m
0.6 m
Dry SilicaSandNo8
Dr=50%
Liquefiable SilicaSandNo8
Dr=50%
Toyoura Sand
Dr=90%
SilicasandNo3
Sheetpile
Quaywall
24m
3.6 m
6m 3m Superstructure
Side D
with superstructure
Side C
no superstructure
Partition
Floating
Quaywall
0 m
1.8 m
6.0 m
8.1 m
9 m
0.6 m
Dry SilicaSandNo8
Dr=50%
Liquefiable SilicaSandNo8
Dr=50%
Toyoura Sand
Dr=90%
SilicasandNo3
Sheetpile
Quaywall
24m
3.6 m
6m 3m Superstructure
Side D
with superstructure
Side C
no superstructure
Partition
Floating
Quaywall

Figure 3. Geometry and soil properties of the centrifuge model of test Case CD, side D into the laminar
box, in prototype scale (Tazoh et al., 2005). A partition behind the quay-wall separates the two pile-
foundation models: the one with a superstructure (C side) and the other without a superstructure (D side).

The input motion at the base of the laminar box is shown in Figure 4 along with the recorded
time histories of excess pore water pressure in the liquefied layer at the almost free field. Liq-
uefaction starts at around 4 sec and 5 sec at the depths of z = 3 m and z = 5 m, respectively. Fig-
ure 5 depicts the horizontal displacements of the footings and the quay-wall of the Case CD.
Evidently, the horizontal displacement of the C-side footing (with no superstructure) is larger
than that of the D-side footing, (with a superstructure), but the difference is not significant. The
effect of the inertial force of the superstructure can be identified during the excitation period;
however it can not be recognized during the liquefaction-induced soil flow some time after the
end of excitation. Figure 6 illustrates the distribution of the maximum bending strains along the
pile at the end of the shaking with and without the influence of the superstructure. It is evident
that the effect of the inertial force of the superstructure on the final pile strains is not significant.
Thus, the main load on the pile is the kinematic one coming from the soil-flow.

116
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
A:g
A:g
time(sec)
Tazohetal.(2005)
Satoetal.(2001)
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
A:g
A:g
time(sec)
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
A:g
A:g
time(sec)
Tazohetal.(2005)
Satoetal.(2001)
40
20
0
40
z=3m
0 5 400 15 10 800 1200

20
0
40
60
20
t(sec)
z=5m
E
x
c
e
s
s
P
o
r
e
P
r
e
s
s
u
r
e
(
k
P
a
)
(a) (b)
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
A:g
A:g
time(sec)
Tazohetal.(2005)
Satoetal.(2001)
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
A:g
A:g
time(sec)
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
A:g
A:g
time(sec)
Tazohetal.(2005)
Satoetal.(2001)
40
20
0
40
z=3m
0 5 400 15 10 800 1200

20
0
40
60
20
t(sec)
z=5m
E
x
c
e
s
s
P
o
r
e
P
r
e
s
s
u
r
e
(
k
P
a
)
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
A:g
A:g
time(sec)
Tazohetal.(2005)
Satoetal.(2001)
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
A:g
A:g
time(sec)
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
0.6
0.4
0.2
0
0.2
0.4
0.6
0 5 10 15
A:g
A:g
time(sec)
Tazohetal.(2005)
Satoetal.(2001)
40
20
0
40
z=3m
0 5 400 15 10 800 1200

20
0
40
60
20
t(sec)
z=5m
E
x
c
e
s
s
P
o
r
e
P
r
e
s
s
u
r
e
(
k
P
a
)
(a) (b)

Figure 4. (a) Input accelerograms recorded at the base of the centrifuge models of the two experiments
(Tazoh et al., 2005 and Sato et al., 2001) in prototype scale, and (b) Indicative time histories of excess
pore water pressures recorded in the free field (away from the pile foundation) at the depths 3 m and 5
m below the ground surface

0 5 10 15 400 800
0.8m
0.1 m
0.15 m
Quaywall
With superstructure
No supersructure
t(sec)
D
i
s
p
l
a
c
e
m
e
n
t

Figure 5. Time histories of displacements of the quay wall and the footing with and without the super-
structure (Tazoh et al., 2005).

Similar centrifuge results with the same soil profile were described by Sato et al. (2001) indi-
cating the seismic performance of a 2x8 pile group, situated 3 m behind the floating sheet pile
quay-wall. Two different centrifuge models were designed: one with (Case 2) and one without
the pile group (Case 1). Figure 4 shows the input wave recorded at the base of the model of
Case 1, which caused excessive pore water pressure generation in the loose saturated sand layer
and maximum seaward displacement of the quay-wall of 0.8 m during the shaking. In Case 2,
the existence of the pile group limited the quay-wall movement to 0.45 m approximately. The
footing of the pile foundation sustained even smaller displacement, about 6 cm. Long after the
end of shaking, when considerable dissipation of pore water pressure has occurred the quay-
wall reached a displacement of 1.15 m, while the footing displacement increased by a mere 2
cm. The trend of the time-dependent results is practically identical with that of the results of
Test Case CD.
3 A NEW PHYSICALLY SIMPLIFIED METHODOLOGY
To begin with, the soil response without the piles needs to be identified. On these grounds, a
2D effective stress numerical analysis of the soil profile including the quay-wall hereafter
117
called free field is performed using the code FLAC, as depicted in Figure 7. Thus, the free
field soil response behind the quay-wall is obtained in terms of:
the distribution of the horizontal soil movement with depth,
the depth and the thickness of the liquefied zone,
the strength degradation in the liquefied layer and
the shear strain distribution with depth.

0
1
2
3
4
5
6
7
8
9
0.7 0.5 0.3 0.1 0.1 0.3 0.5
bendingstrains
(10
3
)
d
e
p
t
h
(
m
)
Cside(NOsuperstructure)
Dside(WITHsuperstructure)

Figure 6. Distributions of the maximum bending strains along the piles at the end of shaking with and
without a superstructure. The influence of the superstructure in terms of bending distress of the piles is
not as significant.
Although the dynamic numerical analysis provides the required results as a function of time,
we are primarily interested in the final values after the end of the shaking, which are also the
maximum ones due to the accumulative nature of the liquefactioninduced soil flow. Obviously
our methodology is not restricted to this code.
In the next step, a horizontal slice in the middle of the liquefied zone is isolated, including the
piles and the quay-wall, as demonstrated in Figure 8. Our purpose is to perform an elastic plane
strain analysis of this horizontal slice by imposing pseudo-statically a unit uniform displacement
at the quay-wall boundary, illustrated in Figure 9, so as to estimate the ratio of the pile dis-
placement to the soil displacement in the free-field (away enough of the piles), named ratio
and depicted in the same figure. This ratio of displacements represents the soil-pile interaction
due to soil flow, in quantitative terms.
The numerical model of the horizontal slice consists of the liquefied soil with uniform prop-
erties (shear modulus, G
l
) surrounding the pile sections, as shown in Figure 9. So far, the pile
sections are simulated as rigid bodies into the liquefied soil, G
l
. In order to provide the required
horizontal resistance to the pile sections against the moving soil mass, an out of plane horizontal
spring, K, is connected to each pile section.
118
Sheetpile
Quaywall
A: g
Sheetpile
Quaywall
Sheetpile
Quaywall
A: g
Sheetpile
Quaywall

Figure 7. 2D Finite-difference mesh of a numerical model of the free field without the piles, in FLAC,
before and after shaking.

G
l
G
l
HorizontalSlice
Largedisplacement
androtation
ofquaywall

G
l
G
l
HorizontalSlice
Largedisplacement
androtation
ofquaywall

Figure 8. A horizontal slice in the middle of the liquefied zone, G
l
, is isolated, including the pile and the
quay-wall sections.

In retrospect, a realistic numerical simulation of the horizontal slice at the middle of the liq-
uefied zone requires the appropriate calibration of the horizontal stiffness of each pile section,
K, and of the shear modulus of the liquefied soil, G
l
.

Horizontal stiffness of each pile section, K

Every single pile of the pile group is simulated as a vertical beam element with suitable
boundary conditions. The rotation at the top depends on the pile cap (Mokwa and Duncan,
119
2003), the number of piles of the foundation and the stiffness of the surrounding surficial non-
liquefiable soil. All these features tend to restrain the rotation at the pile-head.

G
l
Unit
displacement
Pilecapconstraint
pilesection
Q
u
a
y
w
a
l
l
liquefiedsoil
G
l
Unit
displacement
Pilecapconstraint
pilesection
Q
u
a
y
w
a
l
l
liquefiedsoil
Freefielddisplacement
Disp.1
Pilegroupdisplacement
Disp.2
liquefiedsoil
Ratio =
Disp.2
Disp.1
G
l
Unit
displacement
Pilecapconstraint
pilesection
Q
u
a
y
w
a
l
l
liquefiedsoil
G
l
Unit
displacement
Pilecapconstraint
pilesection
Q
u
a
y
w
a
l
l
liquefiedsoil
Disp.1
Disp.2
liquefiedsoil
Ratio =
Disp.2
Disp.1
Ratio =
Disp.2
Disp.1
Ratio =
Disp.2
Disp.1
G
l
Unit
displacement
Pilecapconstraint
pilesection
Q
u
a
y
w
a
l
l
liquefiedsoil
G
l
Unit
displacement
Pilecapconstraint
pilesection
Q
u
a
y
w
a
l
l
liquefiedsoil
Disp.1
Disp.2
liquefiedsoil
Ratio =
Disp.2
Disp.1
G
l
Unit
displacement
Pilecapconstraint
pilesection
Q
u
a
y
w
a
l
l
liquefiedsoil
G
l
Unit
displacement
Pilecapconstraint
pilesection
Q
u
a
y
w
a
l
l
liquefiedsoil
Disp.1
Disp.2
liquefiedsoil
Ratio =
Disp.2
Disp.1
Ratio =
Disp.2
Disp.1
Ratio =
Disp.2
Disp.1

Figure 9. (Top) Numerical model of the horizontal slice. The pile sections sustain the same horizontal
displacement due to the pilecap constraint. (Bottom) Definition of the ratio .

The active length of the beam element depends on the depth to fixity, below the liquefied
layer. Therefore, every single pile is simulated as a vertical beam fixed at the bottom with a cer-
tain degree of rotational freedom at the top depending on the aforementioned kinematic con-
straints.
Based on this simulation, the pile section of the numerical model is just a section of the beam
element at a characteristic depth at the middle of the liquefied zone. Thus, the horizontal stiff-
ness, K, of each pile section, is defined as the point load exerted on the beam, in order to cause a
unit displacement of the beam at the characteristic depth, as depicted in Figure 10.

120

0.0
0.2
0.4
0.6
0.8
1.0
0.001 0.01 0.1 1 10 100 1000
K/EI
z
/
L
0.0
0.2
0.4
0.6
0.8
1.0
0.001 0.01 0.1 1 10 100 1000
K/EI
z
/
L
0.0
0.2
0.4
0.6
0.8
1.0
0.001 0.01 0.1 1 10 100 1000
K/EI
z
/
L
0.0
0.2
0.4
0.6
0.8
1.0
0.001 0.01 0.1 1 10 100 1000
K/EI
z
/
L

Figure 10. Dimensionless horizontal stiffness along the pile for free to fixed boundary conditions at the
pile head.

0.0
0.2
0.4
0.6
0.8
1.0
0.1 1 10 100
K/ G
l
r
a
t
i
o
"
"
1
0.8
0.6
0.4
0.2
0
0.1 1 10 100
singlepile
2x2pilegroup
2x8pilegroup
0.0
0.2
0.4
0.6
0.8
1.0
0.1 1 10 100
K/ G
l
r
a
t
i
o
"
"
1
0.8
0.6
0.4
0.2
0
0.1 1 10 100
singlepile
2x2pilegroup
2x8pilegroup

Figure 11. The ratio as a function of the relative stiffnesses between the pile and the liquefied soil for
three different pile configurations: single pile, 2x2 and 2x8 pile group, obtained from several parametric
numerical analyses of the horizontal slice.

Shear modulus of the liquefied soil, G
l

In the framework of an elastic analysis of the horizontal slice, an equivalent linear shear
modulus of the liquefied soil can be determined as:
0
1
liq
H
res
l liq
res
G dz
H

(1)
H
liq
is the thickness of the liquefied zone, is the residual shear stress and is the maximum
shear strain after the end of shaking, obtained from the numerical analysis of the free field. That
is how the stiffness degradation of the liquefied soil is taking into account.
The elastic pseudo-static analysis of the horizontal is conducted for two different pile con-
figurations: (a) a 2x2 pile group (Tazoh et al., 2005) and (b) a 2x8 pile group (Sato et al., 2003).

121
0
0.2
0.4
0.6
0.8
1
0 0.5 1
w/ ( pL EI w
max
)
z
/
L
uniformload
invertedtriangleload
invertedtrapezoidload
0
0.2
0.4
0.6
0.8
1
0 0.5 1
w/ ( pL EI w
max
)
z
/
L
uniformload
invertedtriangleload
invertedtrapezoidload

Figure 12. Deformation shape of a pile with no rotation at the top and the bottom, determined for three
different load distributions: uniform, inverted triangle shaped and inverted trapezoid one.

The ratio , is obtained as function of the relative stiffness between the pile and the lique-
fied soil, K/G
l
, shown in Figure 11.
When the relative stiffness tends to zero (K/G
l
0), the ratio tends to unity, which
means that the pile sections move just like the soil, as a rigid body. On the contrary, when the
relative stiffness tends to infinity (K/G
l
), the ratio tends to zero. This is due to the fact
that the soil has practically zero shear strength (G
l
0) and flows around the pile without exert-
ing any significant load on them. Moreover, Figure 11 indicates that increasing the number of
piles, the resistance of the foundation to the moving soil mass becomes stronger. It is worth
mentioned at this point that the numerical modeling of the pile-group section into the liquefied
soil is based on the assumption that the piles sustain the same horizontal displacement due to the
pile-cap constraint.
In conclusion, as long as the relative stiffness, K/G
l
, is determined, the ratio can be esti-
mated. Eventually, multiplying the ratio with the free-field soil displacement in the middle
of the liquefied zone, at the position where the piles would be present behind the quay-wall, we
can calculate the pile displacement at the same characteristic depth.
In the last step of the methodology, the whole pile deformation and mainly the pile displace-
ment at the top, remains to be evaluated. The deformation shape of each pile, simulated as a
beam element, is defined primarily by its boundary conditions and secondarily by the load dis-
tribution along it, as depicted in Figure 12. A potential load distribution, is imposed on the pile,
so as to obtain its deformation shape as a function of the unknown load value, p.
Until now, we have just estimated the pile displacement in the middle of the liquefied zone.
Using this known displacement, we can calibrate the shape function of the pile with depth and
finally, estimate the unknown load value, p. Continuously, the distribution of the pile dis-
placements with depth, as long as the bending moments along the pile, can be determined by
imposing the already fully known load distribution on the beam-pile.

122
0
1
2
3
4
5
6
7
8
9
15 10 5 0 5 10 15 20
shearstress(kPa)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8 10 12 14 16
shearstrain(10
2
)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
15 10 5 0 5 10 15 20
shearstress(kPa)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8 10 12 14 16
shearstrain(10
2
)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
freefielddisplacement(m)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
15 10 5 0 5 10 15 20
shearstress(kPa)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8 10 12 14 16
shearstrain(10
2
)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
15 10 5 0 5 10 15 20
shearstress(kPa)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8 10 12 14 16
shearstrain(10
2
)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
(i)
(ii)
0
1
2
3
4
5
6
7
8
9
15 10 5 0 5 10 15 20
shearstress(kPa)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8 10 12 14 16
shearstrain(10
2
)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
15 10 5 0 5 10 15 20
shearstress(kPa)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8 10 12 14 16
shearstrain(10
2
)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
15 10 5 0 5 10 15 20
shearstress(kPa)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8 10 12 14 16
shearstrain(10
2
)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
15 10 5 0 5 10 15 20
shearstress(kPa)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8 10 12 14 16
shearstrain(10
2
)
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
0
1
2
3
4
5
6
7
8
9
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
d
e
p
t
h
(
m
)
Tazohetal.(2005)
Satoetal.(2001)
(i)
(ii)

Figure 13. Distributions of horizontal soil displacements, shear strains and stresses, obtained from the
numerical analysis of the two models: (i) Tazoh et al., 2005 (Case CD) and (ii) Sato et al. (2001) without
the piles.

0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0 2 4 6 8 10 12
time(sec)
Q
u
a
y
w
a
l
l
d
i
p
l
a
c
e
m
e
n
t
a
t
t
o
p
(
m
)
Tazohetal.(2005)
Satoetal.(2005)
(i)
(ii)

0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
0 2 4 6 8 10 12
time(sec)
Q
u
a
y
w
a
l
l
d
i
p
l
a
c
e
m
e
n
t
a
t
t
o
p
(
m
)
Tazohetal.(2005)
Satoetal.(2005)
(i)
(ii)

Figure 14. Horizontal quay wall displacements, obtained from the numerical analysis of the two models:
Tazoh et al., 2005 (Case CD) and Sato et al. (2001) without the piles.
4 COMPARISON WITH RESULTS FROM CENTRIFUGE TESTS
The results obtained from the numerical analysis of the free field including the quay-wall in
the first step for the two centrifuge experiments (Tazoh et al., 2005 and Sato et al., 2001) are il-
lustrated in Figures 13 to 16. The only differences between the two models, both the centrifuge
and the numerical ones are the input wave motions, shown in Figure 4 and the distance of the
quay wall from the boundaries parallel to it, depicted in Figure 3. According to the excess pore
pressure time histories, demonstrated in Figures 15 and 16, the liquefaction seems more exten-
sive in case of Tazoh et al. model. This is one of the reasons why the quay-wall displacement of
this model is larger than the one of the Sato et al. model, as portrayed in Figure 14.
In the second step, every single pile of the pile groups is simulated as a beam fixed both at
the top and the bottom, following the assumption that the pile cap does not allow any rotation at
the pile heads. The horizontal stiffnesses, K, are calculated at the characteristic depths depicted
in Figure 17 which coincide with the middle of the liquefied layer. The values required to esti-
123
mate the pile displacement at the characteristic depth are given in table 1. Once the relative
stiffness is estimated for each model, the ratio is determined, choosing the 2x2 pile-group
curve in case of the Tazoh et al. model and the 2x8 pile-group curve for the Sato et al. model.

0.0
0.2
0.4
0.6
0.8
1.0
0 2 4 6 8 10 12
time(sec)
E
x
c
e
s
s
p
o
r
e
p
r
e
s
s
u
r
e
r
a
t
i
o
z=4m
z=3.5m
z=5.5m
z=6m
0.0
0.2
0.4
0.6
0.8
1.0
0 2 4 6 8 10 12
time(sec)
E
x
c
e
s
s
p
o
r
e
p
r
e
s
s
u
r
e
r
a
t
i
o
z=4m
z=3.5m
z=5.5m
z=6m

Figure 15. Time histories of excess pore pressure ratio at depths of 4, 3.5, 5.5, 6 m below the ground sur-
face, obtained from the numerical analysis of the Sato et al. model (2001) without the piles.

0.0
0.2
0.4
0.6
0.8
1.0
0 2 4 6 8 10 12 14
time(sec)
E
x
c
e
s
s
p
o
r
e
p
r
e
s
s
u
r
e
r
a
t
i
o
z=4m
z=3.5m
z=5.5m
z=6m
0.0
0.2
0.4
0.6
0.8
1.0
0 2 4 6 8 10 12 14
time(sec)
E
x
c
e
s
s
p
o
r
e
p
r
e
s
s
u
r
e
r
a
t
i
o
z=4m
z=3.5m
z=5.5m
z=6m

Figure 16. Time histories of excess pore pressure ratio at depths of 4, 3.5, 5.5, 6 m below the ground sur-
face, obtained from the numerical analysis of the Tazoh et al. model (2005) without the piles.

In the last step, five different potential load distributions are imposed on the pile-beam. A
range of estimated values regarding the displacements and the bending moments along each pile
of the pile group are illustrated in Figure 18 for the two models. The range of pile displacements
is not very sensitive to the shape of the load distribution whereas the bending moments vary
significantly, especially on the top. However, the maximum bending moment for both of the
two models was recorded at the pile tip, during the centrifuge experiments.
5 CONCLUDING REMARKS
In this paper, a new simple physically-motivated methodology is proposed for the evaluation
of pile response due to liquefactioninduced soil flow. The main characteristics of this method-
ology verified above are:
It avoids the associated empirical selection of stiffnessreduction factors and does not in-
volve the use of p-y curves.
124
1.3 m
L=7.5m 9 m
Liquefiedlayer
2 m
6 m
3 m
P
1.3 m
L=7.5m L=7.5m 9 m
Liquefiedlayer
2 m
6 m
3 m
P
EI=17000kNm
2
0.6 m
L=8.0m 9 m
Liquefiedlayer
2 m
6 m
3.5 m
P
0.6 m
L=8.0m L=8.0m 9 m
Liquefiedlayer
2 m
6 m
3.5 m
P
EI=6480kNm
2
1.3 m
L=7.5m 9 m
Liquefiedlayer
2 m
6 m
3 m
P
1.3 m
L=7.5m L=7.5m 9 m
Liquefiedlayer
2 m
6 m
3 m
P
EI=17000kNm
2
0.6 m
L=8.0m 9 m
Liquefiedlayer
2 m
6 m
3.5 m
P
0.6 m
L=8.0m L=8.0m 9 m
Liquefiedlayer
2 m
6 m
3.5 m
P
EI=6480kNm
2
1.3 m
L=7.5m 9 m
Liquefiedlayer
2 m
6 m
3 m
P
1.3 m
L=7.5m L=7.5m 9 m
Liquefiedlayer
2 m
6 m
3 m
P
EI=17000kNm
2
0.6 m
L=8.0m 9 m
Liquefiedlayer
2 m
6 m
3.5 m
P
0.6 m
L=8.0m L=8.0m 9 m
Liquefiedlayer
2 m
6 m
3.5 m
P
EI=6480kNm
2
1.3 m
L=7.5m 9 m
Liquefiedlayer
2 m
6 m
3 m
P
1.3 m
L=7.5m L=7.5m 9 m
Liquefiedlayer
2 m
6 m
3 m
P
EI=17000kNm
2
0.6 m
L=8.0m 9 m
Liquefiedlayer
2 m
6 m
3.5 m
P
0.6 m
L=8.0m L=8.0m 9 m
Liquefiedlayer
2 m
6 m
3.5 m
P
EI=6480kNm
2

Figure 17. Each pile of the 2x8 pile group of the Sato et al. (2001)model (Top) and of the 2x2 pile group
of the Tazoh et al. (2005) model (Case CD -- Bottom), is simulated as a vertical beam element with no ro-
tation at the top and the bottom. The horizontal stiffness, of the pile is estimated at a characteristic depth
of 3 m (3.5 m for the Tazoh et al model) from its top, which coincides with the middle of the liquefied
layer.

Table 1. Table of the values needed to determine the ratio and the pile displacements at the middle of
the liquefied layer for the two models:

depth
(m)
G
l
(kPa)
K
(kN/m)
K /G
l
ffdisp
(m)
ratio
""
piledisp.
(m)
3.50 90 370 4.1 0.5 0.152 0.076
depth
(m)
G
l
(kPa)
K
(kN/m)
K /G
l
ffdisp
(m)
ratio
""
piledisp.
(m)
3.00 200 1020 5.1 0.42 0.105 0.044
Tazohetal.
(2005)
Satoetal.
(2001)

It introduces an elastic pseudo static numerical analysis of a horizontal slice into the lique-
fied layer including the piles, in order to estimate the soil-pile interaction under soil flow
conditions in quantitative terms. This interaction is determined as a function of the relative
stiffness between the pile and the liquefied soil. The pile stiffness is assumed to remain lin-
ear-elastic throughout the analysis.
It can be applied to any type of soil profile and pile configuration (single pile or pile
group).
125
The effectiveness of the new methodology in combination with suitable engineering judg-
ment and reasonable assumptions can provide sufficient accuracy for designing and evalu-
ating purposes.

0
1
2
3
4
5
6
7
0 0.02 0.04 0.06 0.08
pilediplacement(m)
d
e
p
t
h
(
m
)
Centrifugeexperiment
Rangeofpredictedvalues
0
1
2
3
4
5
6
7
0 0.02 0.04 0.06 0.08
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
0 0.02 0.04 0.06 0.08
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
600 400 200 0 200 400 600
bendingstrains()
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
600 400 200 0 200 400 600
bendingstrains()
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
8
0 0.05 0.1 0.15
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
8
0 0.05 0.1 0.15
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
8
0 0.05 0.1 0.15
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
8
1500 1000 500 0 500 1000 1500
bendingstrains()
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
0 0.02 0.04 0.06 0.08
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
0 0.02 0.04 0.06 0.08
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
0 0.02 0.04 0.06 0.08
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
600 400 200 0 200 400 600
bendingstrains()
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
600 400 200 0 200 400 600
bendingstrains()
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
8
0 0.05 0.1 0.15
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
8
0 0.05 0.1 0.15
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
8
0 0.05 0.1 0.15
pilediplacement(m)
d
e
p
t
h
(
m
)
0
1
2
3
4
5
6
7
8
1500 1000 500 0 500 1000 1500
bendingstrains()
d
e
p
t
h
(
m
)

Figure 18. Calculated range of distributions of (a) the horizontal displacement, and (b) the bending strain,
in comparison with the centrifuge test results by Sato et al., 2001 (Top), and Tazoh et al., 2005 (Bottom).
ACKNOWLEDGMENTS
This work forms part of an EU 7
th
Framework research project funded through the European
Research Council (ERC) Programme Ideas, Support for Frontier Research Advanced Grant,
under Contract number ERC-2008-AdG 228254-DARE.

126
REFERENCES
Boulanger R. W., Kutter B. L., Brandenberg S. J., Singh P., and Chang P., (2003), Pile Foundations in
Liquefied and Laterally Spreading Ground during Earthquakes: Centrifuge Experiments and Analyses,
Report No. UCD/CGM-03/01, Department of Civil and Environmental Engineering, University of
California at Davis.
Cubrinovsky M., and Ishihara K., (2004), Simplified Method for Analysis of Piles undergoing Lateral
Spreading of Liquefied Soils, Soils and Foundations, Vol. 44, No. 5, pp. 119-133.
Cubrinovsky M., Kokusho T., and Ishihara K., (2004), Interpretation from large-scale shake table tests
on piles subjected to spreading of liquefied soils, Proceedings of 11
th
Int. Conf. Soil Dynamics and
Earthq. Engrg. / 3
rd
Int. Conf. Earthq. Geotech. Engrg., Berkeley, USA, Vol. 2, pp. 463-470.
Dobry R., and Abdoun T. H., (1998), Post-Triggering Response of Liquefied Soil in The Free Field and
Near Foundations, State-of-the-art paper, Proc. ASCE 1998 Specialty Conference on Geotechnical
Earthquake Engineering and Soil Dynamics (P. Dakoulas, M. Yegian and R. D. Holtz, eds.), Univer-
sity of Washington, Seattle, Washington, August 3-6, Vol. 1, pp. 270-300.
Dobry R., and Abdoun T. H., (2000), Recent studies on seismic centrifuge modeling of liquefaction and
its effect on deep foundations, Proceedings of the fourth International Conference on Recent Ad-
vances in Geotechnical Earthquake Engineering and Soil Dynamics , San Diego.
Finn Liam W. D., and Thavaraj T., (2001), Deep Foundations in liquefiable soils: Case Histories , cen-
trifuge tests and methods of analysis , Proceedings of the fourth International Conference on Recent
Advances in Geotechnical Earthquake Engineering and Soil Dynamics, San Diego.
Ishihara K., (1997), Geotechnical aspects of the 1995 Kobe Earthquake, Proceedings of 14
th
Int. Conf.
SMFE, Terzaghi Oration, Hamburg.
Mokwa R. L., Duncan J. M (2003), Rotational restraint of pile caps during lateral loading, Journal of
Geotechnical and Geoenvironmental Engineering, ASCE, 129(9), pp. 829-837.
Ramos R., Abdoun T. H., and Dobry, R., (2000), Effect of Lateral Stiffness of Superstructure on Bend-
ing Moments of Pile Foundation Due to Liquefaction-induced Lateral Spreading, Proc. 12th World
Conf. on Earthquake Engineering, Auckland, New Zealand, Jan. 30 - Feb. 4, 8 pages.
Sato M., Tazoh T. and Ogasawara M. (2001), Reproduction of Lateral Ground Displacement and Lateral
Flow Earth Pressure Acting on Pile Foundations using Centrifuge Modeling, Proceedings of the
Fourth International Conference on Recent Advances in Geotechnical Earthquake Engineering and
Soil Dynamics, San Diego.
Tazoh T., Gazetas G. (1996), Pile Foundations Subjected to Large Ground Deformations: Lessons from
Kobe and Research Needs, Proceedings of the Eleventh World Conference on Earthquake Engineer-
ing, Acapulco, Mexico, paper 2081.
Tazoh T., Sato M., and Gazetas G., (2005), Centrifuge Tests on PileFoundationStructure Systems Af-
fected by LiquefactionInduced Soil Flow after Quay Wall Failure, Proceedings of the 1
st
Greece
Japan Workshop: Seismic Design, Observation, and Retrofit of Foundations, Athens, Oct. 11-12 pp.
79-106.
Tokimatsu K., Oh-oka H., Samoto Y., Nakazawa A., and Asaka Y., (1997), Failure and Deformation
Modes of Piles Caused by Liquefaction-Induced Lateral Spreading in 1995 Hyogo-ken Nambu Earth-
quake, Proceedings of the Third Kansai International Forum on Comparative Geotechnical Engi-
neering, (KIG-Forum 97), pp. 239-248.
Yasuda S., Ishihara K., Harada K., and Namura H., (1996), Area of Ground Flow Occurred behind
Quaywalls due to Liquefaction, Proceedings of the Third Kansai International Forum on Compara-
tive Geotechnical Engineering, (KIG-Forum 97), pp. 85-93.
127
FINITE ELEMENT ANALYSIS OF PILE-SOIL INTERACTION
SYSTEM BY OVERLAYING MESH METHOD

A. Ohta & F. Miura
The graduate school of science and engineering, Yamaguchi University

Y. Ono & J. Kiyono
The graduate school of engineering, Kyoto University

1 INTRODUCTION
The Overlaying mesh method (OMM) is an analytical approach that overlaps two or more
independent different-size-mesh models. In the OMM, detailed mesh model is used in elected area
under consideration, with coarser mesh model else where, in order to optimize calculation effort. In
the previous study, different-size-models are used to express a complex area with different material
constants, but same type elements, such as plane strain two dimensional elements are used. In this
research, we propose a new application method of the overlaying mesh method using different
elements such as beam elements and solid elements. We analyzed two types of pile foundation
models using OMM, and proved that the proposed method is valid.
2 THEORY OF THE PROPOSED METHOD
2.1 Derivation of the fundamental equations for the OMM
In the OMM, two or more different-sized-mesh models are used, one is for modeling the wide area,
which we call Global area, the other/others is/are used to model detailed area(s), which we call Local
area, where we want to know the detailed behavior. In the soil-structure interaction problem, for
example, the former is used to model the ground which widely extends, and the latter is used to
model the structure of which shape is complex.
Let designate the Global area as
G
, the Local area as
L
and the boundary between these areas as
GL
. The image of the relationship of them is illustrated in Figure1.

Figure 1. Superimposition of global and local areas

Displacement fields are independently defined in each
G
and
L
, i.e., u
i
G
and u
i
L
, respectively.
The actual displacement u
i
in
L
is defined as the sum of u
i
G
and u
i
L
, while u
i
is equal to u
i
G
outside the
L
. Namely, the displacement u
i
is defined as the following equations.

128
L L
i
G
i i
u u u + = in
(1)

L G G
i i
u u = in
(2)
To satisfy the continuity of the displacement at the boundary
GL
, the following condition is
needed.
on 0
GL
=
L
i
u
(3)
Displacements u
i
G
and u
i
L
in
G
and
L
are expressed by using shape function matrices N
G
and N
L

and nodal displacement vectors u
i
G
and u
i
L
as follows.
G
j
G
ij
G
i
u N u =
(4)
L
j
L
ij
L
i
u N u =
(5)
By partially differentiating Eq.(1) and using above equations, we obtain strain
ij
as,
L
ij
G
ij ij
+ = (6)
In which
G
k
G
ijk
G
ij
u B =
(7)
L
k
L
ijk
L
ij
u B =
(8)
By using the principle of virtual work, we can obtain the next equation.

+ = d t u d b u d D
i i i i kl ijkl ij
(9)
Where,
ij
,

ij
,
i
u , b
i
, t
i
, D
ijkl
are virtual strain, strain, virtual displacement, body force, surface
traction and constitutive tensor, respectively. The left side of the equation stands for the virtual
work due to the internal strains and the right side represents the virtual work done by the external
forces. By substituting Eq.s (1), (6), (7) and (8) into Eq. (9), we can obtain the following equations.

+ + + =
+ +
d t u u d b u u
d D
i
L
i
G
i i
L
i
G
i
L
ij
G
ij ijkl
L
ij
G
ij
) ( ) (
) ( ) (

(10)

+ + + =
+ +
d t u N u N d b u N u N
d u B u B D u B u B
i
L
i
L
ij
G
i
G
ij i
L
i
L
ij
G
i
G
ij
L
m
L
ijm
G
m
G
ijm ijkl
L
m
L
ijm
G
m
G
ijm
) ( ) (
) ( ) (

(11)
129
By rewriting the above equations in the matrix form, we obtain the following equation.
L
G
L
G
L LG
GL G
f
f
u
u
K K
K K
(12)
Where :

+ =
+ =
=
=
=
=
d t N d b N f
d t N d b N f
d B D B K
d B D B K
d B D B K
d B D B K
i
L
i i
L
i
L
i
G
i i
G
i
G
L L
kl ijkl
L
ij
L
L G
kl ijkl
L
ij
LG
L L
kl ijkl
G
ij
GL
G G
kl ijkl
G
ij
G
L
L
L
G
(13)

In which K
G
and f
G
are stiffness matrix and external force vector for the global area
G
, and K
L
and f
L

are stiffness matrix and external force vector for the local area
L
, respectively.
2.2 Linking the beam element and the plane strain solid element
According to the previous work
1)
, linkage matrices between global and local plane strain elements,
K
GL
and K
LG
, are obtained from Eq.(13). Linkage matrices between plane strain elements and beam
elements, however, cannot be obtained in the same manner, because the strains are different between
the beam element and the solid element. It is, therefore, necessary to develop a new method to link
them.
The global nodal displacement at the same position as that of the local node, u
G
can be obtained
by using the global shape function N
G
and global nodal displacements
G
k u as Eq. (14).
G
k
G
kl
G
l
u N u = (14)
Global strain at arbitrary point,
G
, can be obtained from Eq.(7), and also obtained using other
element if the point is included inside the element and the coordinate of the nodal points of the
element. Therefore, global strain can be obtained by using u
G
and local shape function B
L
.
G
l
L
ijl
G
k
G
ijk
G
ij
u B
u B
=
=
(15)
Using Eq. (14), we can obtain the following relationships.
G
kl
L
ijl
G
ijk
G
k
G
ijk
G
l
L
ijl
G
k
G
kl
L
ijl
N B B
u B
u B u N B
=
=
=
(16)
130

Therefore, K
LG
can be obtained as follows.
[ ]
[ ][ ]
G L
G
mn
L L
klm ijkl
L
ij
L G
mn
L
klm ijkl
L
ij
L G
kl ijkl
L
ij
LG
N K
N d B D B
d N B D B
d B D B K
L
L
L
=
=
=
=
(17)

In the same manner, K
GL
is expressed in the following way.
[ ] [ ] [ ]
L
T
G LG
K N K =
(18)

2.3 Constitution of the local mesh
Figure 2 shows the total system which includes global model and local model. The local model
contains beam elements of which area is designated by
C
. The local area modeled by solid elements
is expressed by
B
and the global area by
A
. It is assumed that the areas
A
and
C
are not in
contact. The constants of elasticity in the areas
A
and
B
are the same and expressed as D
1
ijkl
and in
the area
C
, D
1
ijkl
in the global model and D
L
ijkl
in the local model.

As for the boundaries, the boundary between
A
and
B
is designated by
AB
, in the same
manner, the boundary between
B
and
C
is designated by
BC
. The boundary is divided into
A
,
B

and
C
according to the division of the areas
A
,
B
and
C
, respectively.
With the definitions above, K
G
, K
L
and K
GL
are obtained as follows.
[ ] [ ] [ ][ ]
+ +
=
C B A
d B D B K
G
kl ijkl
T
G
ij
G 1
(19)
[ ] [ ] [ ][ ] [ ] [ ][ ]

+ =
C B
d B D B d B D B K
L
kl
L
ijkl
T
L
ij
L
kl ijkl
T
L
ij
L 1
(20)
131
[ ] [ ] [ ][ ] [ ] [ ][ ]

+ =
C B
d B D B d B D B K
L
kl
L
ijkl
T
G
ij
L
kl ijkl
T
G
ij
GL 1
(21)
Eq. (11) can be written in the tensor form as;

+ + + =
+ +
+
d t u d t u d b u d b u
d D d D
d D d D
i
L
i i
G
i i
L
i i
G
i
L L
kl ijkl
L
ij
L G
kl ijkl
L
ij
L L
kl ijkl
G
ij
G
kl ijkl
G
ij
L
L L
L

(22)
The displacements can be written in the following equation, in which symbols G, L, A, B and C stand
for Global, Local and areas A, B and C.
+ = +
+ = +
=
=
in
in
in
C LC
i
GC
i
L
i
G
i
B LB
i
GB
i
L
i
G
i
A GA
i
G
i
i
u u u u
u u u u
u u
u
(23)
As for the global displacement concerning the virtual displacement
G
i
u , and strain
G
ij
, we can
obtain the following equation.

+ + +
+ + =
+ +
+ +
d t u d t u d t u
d b u d b u d b u
d D d D
d D d D d D
i
GC
i i
GB
i i
GA
i
i
GC
i i
GB
i i
GA
i
LC
kl
L GC
ij
LB
kl
GB
ij
GC
kl
GC
ij
GB
kl
GB
ij
GA
kl
GA
ij
C B A
C
ijkl
B
ijkl
A
ijkl
B
ijkl
A
ijkl

1
1 1 1
(24)
By partially integrating the left part of Eq.(24) using the Greens formula, the following equation is
obtained.
{ } { }
{ }
{ } { }
{ }
{ } 0 ) ( ) (
) (
) ( ) (
) (
) (
1 1
1 1
1 1
1
, ,
1
, ,
1
,
1
= + + +
+ +
+ + + +
+ + +
+ + +

d u n D D D
d u n D D
d u t n D D d u t n D
d u t n D d u b D D
d u b D d u b D
GBC
i
C
j
LC
kl
L
ijkl
GC
kl ijkl
LB
kl
GB
kl ijkl
GAB
i
B
j
LB
kl
GB
kl ijkl
GA
kl ijkl
GC
i i
C
j
LC
kl
L
ijkl
GC
kl ijkl
GB
i i
B
j
LB
kl
GB
kl ijkl
GA
i i
A
j
GA
kl ijkl
GC
i i
LC
l kl
L
ijkl
GC
l kl ijkl
GB
i i
LB
l kl
GB
l kl ijkl
GA
i i
GA
l kl ijkl
BC
AB
C B
A C
B A

(25)
As the global displacements u
i
G
is continuous in area , the following relations can exist.
AB GAB
i
GB
i
GA
i
u u u = = on
(26)
BC GBC
i
GC
i
GB
i
u u u = = on
(27)
132
On the other hand, as for the local displacement concerning the virtual displacement
L
i
u , and
strain
L
ij
, we can obtain the following equation.

+ + + =
+ +
+
C B C B
C B
C B
d t u d t u d b u d b u
d D d D
d D d D
i
LC
i i
LB
i i
LC
i i
LB
i
LC
kl
L
ijkl
LC
ij
LB
kl ijkl
LB
ij
GC
kl
L
ijkl
LC
ij
GB
kl ijkl
LB
ij

1
1
(28)
In the same manner as in the global area, Eq. (28) can be written as,
{ }
{ }
{ } { }
{ }
{ } 0 ) ( ) (
) (
) ( ) (
) (
1 1
1 1
1 1
, ,
1
, ,
1
= + + +
+ +
+ + + +
+ +
+ +
d u n D D D
d u n D D
d u t n D D d u t n D
d u b D D
d u b D
GBC
i
C
j
LC
kl
L
ijkl
GC
kl ijkl
LB
kl
GB
kl ijkl
GAB
i
B
j
LB
kl
GB
kl ijkl
GA
kl ijkl
GC
i i
C
j
LC
kl
L
ijkl
GC
kl ijkl
GB
i i
B
j
LB
kl
GB
kl ijkl
GC
i i
LC
l kl
L
ijkl
GC
l kl ijkl
GB
i i
LB
l kl
GB
l kl ijkl
BC
AB
C B
C
B

(29)
And,
AB
on 0 =
LB
i
u
(30)
In the Eq.s (25) and (29), as the virtual displacements are arbitrary, we obtain the following
equations.
A
i
GA
l kl ijkl
b D = + in 0
,
1
(31)
B
i
LB
l kl
GB
l kl ijkl
b D = + + in 0 ) (
, ,
1

(32)
C
i
LC
l kl
L
ijkl
GC
l kl ijkl
b D D = + + in 0
, ,
1

(33)
C
i
LC
l kl
L
ijkl
GC
l kl
L
ijkl
b D D = + + in 0
, ,

(34)
A
i
A
j
GA
kl ijkl
t n D = on 0
1
(35)
B
i
B
j
LB
kl
GB
kl ijkl
t n D = + on 0 ) (
1

(36)
C
i
C
j
LC
kl
L
ijkl
GC
kl ijkl
t n D D = + on 0 ) (
1

(37)
133
C
i
C
j
LC
kl
L
ijkl
GC
kl
L
ijkl
t n D D = + on 0 ) (
(38)
{ }
AB B
j
LB
kl
GB
kl ijkl
GA
kl ijkl
n D D = + on 0 ) (
1 1

(39)
{ }
BC C
j
LC
kl
L
ijkl
GC
kl ijkl
LB
kl
GB
kl ijkl
n D D D = + + on 0 ) ( ) (
1 1

(40)
{ }
BC C
j
LC
kl
L
ijkl
GC
kl
L
ijkl
LB
kl
GB
kl ijkl
n D D D = + + on 0 ) ( ) (
1

(41)
By subtracting Eq. (34) from Eq. (33), Eq. (38) from Eq. (37), Eq. (41) from Eq. (40), we obtain
Eq.s (42), (43), (44), respectively.
C GC
l kl
L
ijkl ijkl
D D = in 0 ) (
,
1
(42)
C C
j
GC
kl
L
ijkl ijkl
n D D = on 0 ) (
1
(43)
BC C
j
GC
kl
L
ijkl ijkl
n D D = on 0 ) (
1
(44)
From Eq.s (43) and (44), equilibrium of stress is independently satisfied within the global model
on the boundaries
BC
and
C
, and normal stress outward direction is 0. From Eq.s (42), (43), (44) we
can get the next relationship.
C GC
kl
L
ijkl ijkl
D D = in 0 ) (
1
(45)
This means that the stress of beam elements due to global model is 0 on the boundary of area
C
.
And from Eq. (45)
C GC
kl
= in 0
(46)
Eq.s (33) and (34) become
C
i
LC
l kl
L
ijkl
b D = + in 0
,
(47)
In the same way, Eq. (40) and (41) become
{ }
BC B
j
LC
kl
L
ijkl
LB
kl
GB
kl ijkl
n D D = + on 0 ) (
1

(48)
This means that the stresses due to displacements in local model within area
C
on the
boundary
BC
, equilibrium to those within area
B
. Stresses in the beam elements, therefore, can be
expressed only by the local model and obtained only by the stiffness of the beam elements.

134

Figure 5 Finite element mesh using the OMM
Figure 3 Vertical pile model
Figure 4 Ordinal finite element mesh of vertical
135
3 ANALITICAL EXAMPLES
3.1 Vertical pile model
The vertical pile-footing-ground model used in this analysis is illustrated in Figure 3. Ground and
piles and footing are assumed to be elastic materials and material constants of ground and footing are
the same for the simplicity. Youngs modulus, sectional area, and moment inertia of the section of
the pile are 206GPa, 0.009628m
2
, and 7.86x10
-5
m
4
, respectively. Shear wave velocity, unit weight
and Poissons ratio are 300m/s, 16.67kN/m
3
, 0.3, respectively. Finite element models of the model
are shown in Figures 4 and 5. Figure 4 is the ordinal finite element model and Figure 5 is overlaying
mesh model.
Numerical analysis results are compared in Figures 6, 7 and 8. Horizontal placements of the
beam elements are illustrated in Figure 6, vertical displacements in Figure 7 and rotational angles in
Figure 8. In these figures, normal means the results from the ordinal element model, and OMM
from the overlaying mesh model.

The difference of the horizontal response displacements is about 0.1mm, and this is very small
compared with the maximum response of the system of 2.9mm, in the vertical direction, the
difference is about 0.03mm, while the maximum response is about 13.5mm. The distributions of the
response displacements in the total system are shown in Figures 9, 10, 11 and 12. From these figures
the results are almost same in two models. This means the validity of the proposed method.
Figure 6. Comparison of horizontal displacements Figure 7. Comparison of vertical displacements
Figure 8. Comparison of rotational angle pile
136

Figure 9. Distribution of horizontal displacement
from normal finite element mesh.
Figure 10. Distribution of horizontal displacement
from finite element mesh with OMM.
Figure 12. Distribution of vertical displacement
from finite element mesh with OMM.
Figure 11. Distribution of vertical displacement
from normal finite element mesh
Figure 13. Battered pile model.
Figure 14. Normal finite element mesh of battered
piles.

Figure 15. Finite element mesh of battered piles with OMM

137

3.2 Battered pile model
Figure 13 shows the battered pile-ground-footing model. The material constants are the same as
those of the vertical pile models. Figures 14 and 15 are ordinal mesh model and OMM, respectively.
The mesh of the former model is very complicated to express the battered piles, on the other hand, the
mesh is very simple for the latter model as shown in Figure 15. Same global mesh as in the vertical
pile is used for the OMM.
The comparisons of horizontal and vertical displacements and rotational angle are made in Figures
16, 17 and 18. The differences between these two models are little larger than those from the vertical
piles model, especially for rotational angle. One of the possible reasons is the irregularity of the
ordinal mesh model. To generate the mesh shown in Figure 14, we employed automatic mesh
generation software, thats why the mesh is not symmetry about the centerline. We need to seek and
examine other cause of the difference of the results.
The comparisons of the distributions of horizontal and vertical displacements in the global system
are shown in Figures 19 and 20. As can be seen in Figures 16 and 17, the difference of the responses
can be observed at the top and the tip of the piles.
The CPU time to analyze models shown in Figures 14 and 15 are almost same in both cases. To
generate the OMM is very easy, because we just put the battered piles models on the global model
(ground model). This is a typical advantage of using OMM.

Figure 16. Comparison of horizontal displacements Figure 17. Comparison of vertical displacements

Figure 18. Comparison of rotational angle of pile

138

4 CONCLISIONS
We derived the OMM in application of the soil-structure interaction system. Then we examine the
validity of the method. For vertical pile model, we could get good agreement between the ordinal
model and the OMM, but in the analysis of battered pile model, the difference is little larger than
those for vertical pile models. We need to examine the reason and establish the analysis method for
the soil-structure interaction problem, and more we need to extend the method to three dimensional
problem in which the advantage of the method will be remarkable.
REFERENCES
T. Belytchko, J. Fish and A. Bayliss (1990), The spectral overlay on finite elements for problems with high
gradients, Computer methods in applied mechanics and engineering, Vol.81 pp.71-89.
Figure 19. Distribution of horizontal displacement from ordinal finite element mesh.
Figure 20. Distribution of horizontal displacement from finite element mesh with OMM.
139
1 INTRODUCTION
Many researches have been done on the liquefaction of sand experimentally, empirically, and
mathematically. Cyclic mobility due to soil skeleton dilation at large shear strain excursions is
one of typical mechanical behaviors of sand during liquefaction. A large number of computa-
tional models have been, and continue to be developed for simulation of nonlinear soil re-
sponse, and the dilatancy effects. Research related to testing methods and modeling of the cy-
clic mobility of sand can be found in many publications, e.g., the work by Oka et al. (1992);
LIQCA program by Yashima et al. (1991). In recent years, research on constitutive model for
soils has been developing very quickly. For instance, the concept of subloading proposed by
Hashiguchi and Ueno (1977), and the concept of superloading proposed by Asaoka et al.
(1998), make it possible not only to describe remolded soils (Roscoe et al., 1963 and Schofield
and Wroth 1968), but also naturally deposited soils in which overconsolidation, structure and
anisotropy of soils play a very important rule in determining the mechanical behaviors of soil.
Asaoka et al. (2002) proposed an elasto-plastic constitutive model based on the evolution rules
which describe the collapse of the soil skeleton structure (the concept of superloading), the loss
in overconsolidation (the concept of subloading), and the development of anisotropy during
shearing. In their paper, the importance of stress-induced anisotropy is introduced.
In liquefaction analysis, a constitutive model that can properly describe stress-strain relations of
sand during liquefaction is most important. Liquefaction analysis is a solution of boundary
value problem (BVP). If the stress and strain path of sand cannot be expressed rightly in ele-
mental level, the solution of a BVP is impossible or has no meaning. However, almost constitu-
tive models can describe the stress-strain relations obtained from different values of material
parameters for sands with variant densities. This is unreasonable in a liquefaction process be-
cause the density of ground is not constant and will increase due to flow out of the pore water
from ground.
Zhang et al. (2007) proposed a new constitutive model for soils that can describe liquefaction
behaviors of sands with different densities. The model provides an approach for describing the
Unique description of liquefaction behavior of Toyoura sands
with different densities

F. Zhang & Y. Jin
Nagoya Institute of Technology, Nagoya, Japan
ABSTRACT: In this paper, based on the model proposed by Zhang et al.(2007), a minor modi-
fication for the evolution equation of overconsolidation is carried out at first and then a unique
description of the overall mechanical behaviors of Toyoura Standard Sand (TSS), a typical
clean sand, is conducted, in which the eight parameters for describing TSS is kept constant no
matter what kind of loadings or drainage conditions may be. In the theoretical simulation, based
on conventional drained triaxial compression tests and undrained triaxial cyclic loading tests,
the material parameters of TSS are determined. The capability of the model to describe
uniquely the overall behaviors of the sand under different drainage conditions and different
loadings with one set of fixed parameters is verified.
140
stress-induced anisotropy with a new evolution rule for changes in overconsolidation, by which
the mechanical behavior of soils subjected to cyclic loading under undrained conditions, includ-
ing the cyclic mobility of medium dense sand, can be uniquely described. Moreover, different
soils can be explained efficiently with the concepts of overconsolidation, structure and stress-
induced anisotropy. In his paper, the proposed constitutive model is used to describe the overall
behaviors of Toyoura Standard Sand with different densities subjected to various kinds of load-
ing under different drainage conditions with a unique set of parameters. The material parame-
ters of the sand are determined based on conventional triaxial tests. By comparing the test re-
sults with predicted ones, it is possible to confirm the performance of the constitutive model.
In this study, based on conventional drained triaxial compression tests and undrained triaxial
cyclic loading tests, material parameters of Toyoura Standard sand is determined at first and
then other tests under different loadings and drainage conditions are simulated by the constitu-
tive model. The purpose of the research is to verify if the model can describe uniquely the over-
all behaviors of the sand under different drainage conditions and different loadings with one set
of fixed parameters.
2 MODIFICATION OF ORIGINAL CONSTITUTIVE MODEL
In the original model (Zhang et al. 2007), apart from the concepts of subloading (Hashiguchi
and Ueno, 1977) and superloading (Asaoka et al., 1998), a new evolution rule for the develop-
ment of overconsolidation was proposed. Here we use the word of development instead of
loss just want to emphasize that during plastic loading the degree of overconsolidation some-
time may even increase, not always the case in which overconsolidation only develops in elastic
unloading process. In the original model, the changing rate of overconsolidation is assumed to
be controlled by two factors, namely, the plastic component of stretching that was employed as
the only factor in the work by Asaoka et al. (2002), and the increment in anisotropy, in other
words,

MD M
p
R f
R JU

= +
o
&
D

(1)
in which,
o
is proportional to the norm of the plastic component of stretching

p
s
D
, and U is
given by the following relation as:

'
2
'
0
( ) ln
D
m p
U R
p
=
(
'
0
p =98.0kPa, reference stress) (2)
The problem arisen in Eq.(2), however, is that when a confining stress becomes very high, then
the value of (p/p
0
)
2
will be even much larger, which may result in a very quick development of
overconsolidation in large confining stress. Obviously this is not a true behavior in reality. To
solve this problem, a new function is proposed in the paper as shown in the following relation:

' ' 2
0
' ' 2
0
( / )
( ) ln
D ( / ) 1
p p m
U R
p p
=
+
(
'
0
p
=98.0kPa, reference stress) (3)
in which, the function (y=x
2
/(x
2
+1)) satisfying the requirement that when x is very small, y=x
2

while if x is very large, y will be equal to 1. This modification can then solve the problem per-
fectly without affecting the performance of the model in describing the cyclic mobility. A brief
description of the model can be found in Appendix.
3 SIMULATION OF TOYOURA SAND WITH DIFFERENT CONDITIONS
3.1 Undrained triaxial cyclic loading tests and its simulation
In this section, evolution parameter a which controls the degradation of structure when soil is
subjected to shearing or compression is reconsidered from the results of undrained cyclic triax-
ial test. The material used in experiment is standard Toyoura sand, whose particle size is ex-
tremely uniform and widely used in geotechnical experiments in Japan. Three different ampli-
tudes of shear stress ratio (q/2
0
) of 0.15, 0.20 and 0.25 are tested. The experimental conditions
141
are listed as Table 1. Cyclic loading with sine wave is applied with 0.01Hz under confining
stress 98kPa.
Figure 1 shows experiment results of cyclic triaxial test under undrained condition with differ-
ent amplitude of cyclic loading in which p (=
1
+2
3
) is mean effective stress, q is stress dif-
ference (=
1
3
,
1
and
3
are the maximum and minimum principal effective stresses). In
the case of small amplitude of cyclic loading, it is seen that it needs more cyclic loading num-
bers before cyclic mobility take places than those of large amplitude of cyclic loading. To de-
termine the evolution parameter, element simulations are conducted under the same condition
with experiment.
Figure 2 shows theoretical results with different amplitude of cyclic loading. The tendency of
predicted effective stress paths and stress-strain relations are well coincident with test ones,
while the cyclic number necessary for causing cyclic mobility is less than test results.
A unique set of material parameters of Toyoura sand are listed in Table 2. Eight parameters are
involved in the constitutive model, among which five parameters, M, N,
%
, % , and are the
same as in the Cam-clay model obtained from laboratory tests. The values of the parameters ex-
cept a and m are the same as the those in the works by Ye et al. (2007).
From the parametric study shown in Figure 2, the evolution parameter a which controls the deg-
radation of structure is decided to be 0.5. Degradation parameter of overconsolidation state m
has already been determined by the works of Sago et al. (2008). The value of parameter b
r
is
determined by Ye et al. (2007) that is based on the performance of the soil influenced by the
development of the stress-induced anisotropy when the soil is subjected to shearing or com-
pression. On the other hand, it is difficult to determine the initial conditions of the state parame-
ters for soils because the values are not only dependent on present stress state but also on its
history. The initial conditions of the state parameters for the sand are listed in Table 3. Initial
anisotropy
0
is determined to be `zero` , which means the soil is assumed as isotropic at the
beginning if the specimen is not disturbed. Structure R
*
which influenced by the process of the
soils in its deposition, is quickly collapse during shearing and never recover. Therefore initial
degree of structure R
0
*
is assumed to be a relatively large value R
0
*
=0.75 for the medium dense
sand. While the initial degree of overconsolidation 1/R
0
is determined to be 70 after considering
the experiment results.

Table 1. Experiment conditions of sand with different amplitude of cyclic loading
Amplitude of shear stress ratio (q/2
0
) 0.15 0.2 0.25
Initial void ratio e
0
0.748 (Dr=0.62) 0.753 (Dr=0.61) 0.77 (Dr=0.56)
Initial mean effective stress p (kPa) 98 98 98
Cyclic loading frequency (Hz) 0.01 0.01 0.01

Note: e
max
=0.97; e
min
=0.61

-60
-30
0
30
60
-5 -2.5 0 2.5 5
Axial strain a()
D
e
v
i
a
to
r

s
t
r
e
s
s

q

(
k
P
a
)
-60
-30
0
30
60
0 50 100 150
Mean effective stress p`(kPa)
D
e
v
i
a
t
o
r
s
tr
e
s
s

q

(
k
P
a
)

-60
-30
0
30
60
-5 -2.5 0 2.5 5
Axial strain a()
D
e
v
i
a
to
r

s
t
r
e
s
s

q

(
k
P
a
)
-60
-30
0
30
60
0 50 100 150
D
e
v
ia
t
o
r

s
t
r
e
s
s

q

(
k
P
a
)

-60
-30
0
30
60
-5 -2.5 0 2.5 5
Axial strain a()
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
k
P
a
)
-60
-30
0
30
60
0 50 100 150
D
e
v
ia
t
o
r

s
t
r
e
s
s

q

(
k
P
a
)

(a) q/2
0
=0.15 (b) q/2
0
=0.20 (c) q/2
0
=0.25
Figure 1. Experiment results of cyclic triaxial test under undrained condition with different
amplitude of cyclic loading
142
Table 2. Material parameters of Toyoura Sand
Compression index 0.05
Swelling index 0.0064
Critical state parameter 1.30
Reference void ratio N (p=98 kPa on N.C.L.) 0.87
Poissons ratio 0.30
Degradation parameter of overconsolidation state m 0.01
Degradation parameter of structure a 0.5
Evolution parameter of anisotropy b
r
1.5

-60
-30
0
30
60
0 50 100 150
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
-60
-30
0
30
60
-5 -2.5 0 2.5 5
Axial strain a()
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)

-60
-30
0
30
60
0 50 100 150
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
-60
-30
0
30
60
-5 -2.5 0 2.5 5
Axial strain a()
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)

-60
-30
0
30
60
0 50 100 150
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
-60
-30
0
30
60
-5 -2.5 0 2.5 5
Axial strain a()
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)

(a) q/2
0
=0.15 (b) q/2
0
=0.20 (c) q/2
0
=0.25
Figure 2. Analysis results with different amplitude of cyclic loading

Table 3. Initial conditions of sand with different amplitude of cyclic loading
Amplitude of shear stress ratio(q/2
0
) 0.15 0.2 0.25
0
0.81 0.81 0.81
Initial mean effective stress p (kPa) 98 98 98
Initial degree of structure R
0
*
0.75 0.75 0.75
Initial degree of overconsolidation 1/R
0
70 70 70
Initial anisotropy
0
0.0 0.0 0.0

3.2 Drained triaxial compression test and its simulation
Triaxial compression test and element simulation under constant mean principal stress are con-
ducted to confirm the performance of the constitutive model. The test results on Toyoura sand
are reported from the works by Nakai et al. (2004). Loose specimen was prepared by depositing
the saturated sand slowly in de-aired water using a funnel with an opening of 3mm. Dense
specimen was prepared by pouring the saturated sand into the mold in several layers and com-
pacting each layer with a rod whose diameter was 6mm. The initial void ratio of the loose sand
is e
0
=0.851, and the dense sand is e
0
=0.666.
It is, however, very different to identify the reference void ratio N in a small confining stress
condition. Therefore, by extension of e-ln p relation to the small stress range, the reference void
ratio N of Toyoura sand is determined to be 0.87. Material parameters used in the analysis are
the same as the Table 2, and the initial values of the state parameters determined from the ele-
ment simulation are listed in Table 4. Initial degree of structure is determined from the fact that
well remolded sand has extremely low structure. Figure 3 shows comparison between experi-
ment and analysis results obtained from triaxial compression test sands with different densities.
The dots and the lines in the figures represent the relations between principal stress ratio q/p,
deviator strain
d
and volumetric strain
v
. In the Figure 3(a), theoretical result of loose sand
represents well the test result. The theoretical result for dense sand as shown in Figure 3(b) is
reproduced relatively well before peak strength while the residual stress and strain relations are
143
discrepancy slight from test results. On the whole, however, the constitutive model can describe
the test results qualitatively and quantitatively.

Table 4. Initial conditions of sand with different densities
Loose sand Dense sand
0
0.78 0.69
Initial mean effective stress p (kPa) 196 196
0
*
0.99 0.99
0
4.0 30.0
Initial anisotropy
0
0.0 0.0

-1
-0.5
0
0.5
1
1.5
2
0 5 10 15
Axial strain
a
(%)
P
r
i
n
c
i
p
a
l

s
t
r
e
s
s

r
a
t
i
o

q
/
p
0
-10
-5
0
5
V
o
l
u
m
e
t
r
i
c

s
t
r
a
i
n

v

(
%
)

0
Experimental result
Analysis result

-1
-0.5
0
0.5
1
1.5
2
0 5 10 15
Axial strain a (%)
P
r
i
n
c
i
p
a
l

s
t
r
e
s
s

r
a
t
i
o

q
/
p
0
-10
-5
0
5
V
o
l
u
m
e
t
r
i
c

s
t
r
a
i
n

v

(
%
)

0
Experimental result
Analysis result

(a) Loose sand (b) Dense sand
Figure 3. Experiment and analysis results obtained from triaxial compression test

3.3 Numerical prediction of sands with different densities
The behavior of sand is known as being dependent on its density. In order to verify the influ-
ence of the initial density on the behavior of sand, a series of calculations with different initial
densities are conducted at the same confining pressure.
It is known that loose sand will undergo large volume change and get denser under repeated
perturbation. Firstly, loose sand subjected to repeated triaxial compaction under constant lateral
pressure is simulated under drained condition to explain the phenomenon of densification of
loose sand. Similar to the work by Nakai K. (2005), in the simulation, the density of specimen
is adjusted with a small vibrating compact with amplitude of 2.3kPa on the sand under a small
confining pressure 10kPa. After the compaction, these sands with different densities are isot-
ropically consolidated to the prescribed confining pressure of 294kPa. The set of sands with
eight different densities are prepared by different numbers of vibration compaction, as shown in
Figure 4. All these sands have the same five parameters which are listed in Tables 2 and 5. It
means that these eight sands are made from the same sand material. The initial sand is supposed
to be extremely loose condition with a very large void ratio which is normally consolidated and
highly structured without stress-induced anisotropy. Table 6 presents initial conditions of sands
obtained from compaction procedure. By using these material parameters, the sands with dif-
ferent densities in the laboratory tests under both drained and undrained conditions subjected to
monotones and cyclic loading, are calculated systematically.

Table 5. Reference conditions of sand before vibrating compaction
0
1.19
Initial mean effective stress p (kPa) 10.0
0
*
0.1
0
1.0
Initial anisotropy
0
0.0

144
0.5
0.6
0.7
0.8
0.9
1
1.1
1 10 100 1000
Mean effective stress (kPa)
V
o
i
d

r
a
t
i
o

e
294
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]

Figure 4. Set of sands with different densities prepared from loose sand by vibration compaction and iso-
tropic compression

Table 6. Initial conditions of sands obtained from compression procedure
No.
Vibration
number
n
Initial
void ratio
e
0
Initial degree of
overconsolidation
OCR
Initial degree of
structure
R
0
*
Initial
anisotropy

0
[1] 2 0.903 1.30 0.103 6.44E-06
[2] 15 0.842 4.77 0.113 5.28E-06
[3] 30 0.785 15.99 0.124 7.30E-06
[4] 35 0.771 21.70 0.128 1.01E-05
[5] 40 0.758 28.14 0.132 1.70E-05
[6] 50 0.737 42.96 0.139 8.13E-05
[7] 70 0.707 78.88 0.152 1.75E-03
[8] 120 0.660 208.47 0.168 2.64E-02

3.3.1 Mechanical behaviors of sand subjected to undrained/drained cyclic loading
In the simulation, confining stress of the specimen is 294kPa and amplitude of the cyclic load-
ing in shear stress ratio (q/
m
) is 0.12.
Figure 5 shows the stress paths, stress-strain relations with different densities subjected to cy-
clic triaxial test under undrained condition. It is clear from the figures that very loose sands pro-
ducing a large failure strain along the way towards the zero effective stress state before cyclic
mobility has a chance to occur, as shown in Figures 5 (a) to (d). For medium dense sand, how-
ever, cyclic mobility occurs and the strain which results in limited deformation gradually gets
larger, as shown in Figures 5 (e) to (g). On the other hand, dense sand having a small strain
never shows cyclic mobility, as depicted in Figure 5 (h). The above results mean that the me-
chanical behavior of sand subjected to cyclic loading without drainage, can be uniquely and
properly described by the constitutive model no matter what density it may have.
Figure 6 displays the result of representative loose sand (No.2) subjected to cyclic loading un-
der drained condition. Other simulate conditions are the same as aforementioned undrained test.
As the cyclic loading goes on, the loose sand is compacted and undergoes a large volumetric
stain, which is the same as the results reported by Nakai K.et al. (2004), in which the phenom-
ena of consolidation and liquefaction of sand were just dependent on drainage condition. That
is, if the compaction conducted under drained condition, then the consolidation of sand will oc-
cur; while under undrained condition, the liquefaction will occur.

145
-100
-50
0
50
100
0 50 100 150 200 250 300
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
[1]
-100
-50
0
50
100
-8 -4 0 4 8
Shear strain a()
D
e
v
ia
to
r s
tre
s
s
q
(
k
P
a
)
[1]
-100
-50
0
50
100
0 50 100 150 200 250 300
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
[2]
-100
-50
0
50
100
-8 -4 0 4 8
Shear strain a()
D
e
v
ia
to
r s
tre
s
s
q
(
k
P
a
)
[2]
-100
-50
0
50
100
0 50 100 150 200 250 300
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
[3]
-100
-50
0
50
100
-8 -4 0 4 8
Shear strain a()
D
e
v
ia
to
r s
tre
s
s
q
(
k
P
a
)
[3]
-100
-50
0
50
100
0 50 100 150 200 250 300
D
e
v
ia
to
r
s
tre
s
s
q
(
k
P
a
)
[4]
-100
-50
0
50
100
-8 -4 0 4 8
Shear strain a()
D
e
v
ia
to
r s
tre
s
s
q
(
k
P
a
)
[4]

(a) No.1 (b)No.2 (c)No.3 (d) No.4

-100
-50
0
50
100
0 50 100 150 200 250 300
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
[5]
-100
-50
0
50
100
-8 -4 0 4 8
Shear strain a()
D
e
v
ia
to
r s
tre
s
s
q
(
k
P
a
)
[5]
-100
-50
0
50
100
0 50 100 150 200 250 300
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
[6]
-100
-50
0
50
100
-8 -4 0 4 8
Shear strain a()
D
e
v
ia
to
r s
tre
s
s
q
(
k
P
a
)
[6]
-100
-50
0
50
100
0 50 100 150 200 250 300
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
[7]
-100
-50
0
50
100
-8 -4 0 4 8
Shear strain a()
D
e
v
ia
to
r s
tre
s
s
q
(
k
P
a
)
[7]
-100
-50
0
50
100
0 50 100 150 200 250 300
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
[8]
-100
-50
0
50
100
-8 -4 0 4 8
Shear strain a()
D
e
v
ia
to
r
s
tr
e
s
s
q
(
k
P
a
)
[8]

(e) No.5 (f) No.6 (g) No.7 (h) No.8
Figure 5. Stress paths, stress-strain relations of the sand specimens with different densities subjected to
cyclic triaxial test under undrained condition

-100
-50
0
50
100
0 100 200 300 400
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
k
P
a
)

0.6
0.7
0.8
0.9
0 100 200 300 400
V
o
i
d

r
a
t
i
o

e
e=0.189

(a) (b)
Figure 6. Stress paths, stress-strain relations of the loose sand specimen
(No.2) subjected to cyclic triaxial test under drained condition

3.3.2 Mechanical behaviors of sand under undrained/drained compression
The relationships between stress, strain and void ratio under undrained/drained triaxial com-
pression tests on different densities of sand are discussed in this section. The parameters of the
sand used in the calculation are the same as in Table 2 and Table 6.
Under undrained conditions, three different types of typical stress-strain relations can be ob-
served in the simulation, as shown in Figure 7. For loose sands marked with [1] to [2], the sands
exhibit peak strength at first in small strain level, then collapse and flow rapidly to the original,
showing a typical strain hardening and strain softening behavior. For medium dense sands
marked with [3] to [7], stiffness of the sands decrease abruptly in small strain but strain harden-
ing continues within an intermediate range of shear strains and finally followed by a limited
strain softening, in which a relative higher strength can be approached in spite of a limited
strain softening. Dense sand marked with [8], however, only shows strain hardening and large
shear strength can be reached. Figure 7(c) depicts the traces of the states of all sands during
146
shearing in e-ln p curves, showing that all sands change towards the C.S.L. horizontally at criti-
cal states.
Figure 8 displays the simulated results of stress, strain and dilation of the sands under drained
condition. It is observed that all sands approached to the same values of ultimate shear stress
and the same void ratio, irrespective of different initial densities at the beginning of shearing. It
is also known that initial dense sands show typical strain hardening - strain softening and dila-
tion while initial loose soils only show strain hardening along with monotonic compression.
The simulated facts that the behaviors of sands subjected to shearing under drained/undrained
condition are largely dependent on their initial densities, are already well-known in laboratory
tests of sands and it is not necessary to give any comparison between test and simulation.

0
150
300
450
600
0 150 300 450 600
D
e
v
i
a
to
r
s
t
r
e
s
s

q

(
k
P
a
)
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]

0
150
300
450
600
0 2 4 6 8 10
Shear strain a()
D
e
v
i
a
to
r
s
t
r
e
s
s

q
(
k
P
a
)
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]

0.6
0.7
0.8
0.9
1
1 10 100 1000 10000
V
o
id

r
a
ti
o
e
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
294
start point
end point

(a) (b) (c)
Figure 7. Stress paths, stress-strain relations of the sand specimens with different densities subjected to
triaxial compression test under undrained condition

0
300
600
900
1200
1500
0 2 4 6 8 10
Shear strain a()
D
e
v
i
a
to
r

s
t
r
e
s
s

q
(
k
P
a
)
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]

-6
-4
-2
0
2
4
6
8
10
0 2 4 6 8 10
Shear strain a()
V
o
lu
m
e
tr
i
c
s
t
r
a
i
n
v
(
)
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]

0.5
0.6
0.7
0.8
0.9
1
1.1
0 300 600 900 1200 1500
V
o
i
d
r
a
t
io

e
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
start point
end point

(a) (b) (c)
Figure 8. Stress-strain-dilatancy relations of the sand specimens with different densities subjected to triax-
ial compression test under drained condition

3.4 Confining-stress dependency of sand in undrained monotonic loading tests
Verdugo and Ishihara (1996) reported their experimental results of TSS, in which undrained
triaxial compression tests on sands with the same void ratio but different confining pressures
were conducted under very high confining pressures (up to 3MPa). The test results given in
Figure 9 show that under the same void ratio, if a confining stress is large, the sand behaves like
a loose sand, while if the confining stress is small, the sand behaves like a dense sand. Such a
phenomenon is called as confining-stress dependency of sand, originally defined in the re-
search by Ishihara (1993). Nakai (2005) also reported the same phenomenon in his tests on sil-
ica sands.
The parameters of the sand used in the simulation are the same as Table 2. The initial values of
the void ratios are set to be equal to 0.78, 0.70 and 0.65 respectively and are listed in Table 7
with other initial conditions of the sands. Figure 10 shows that the simulated results on the
whole, coincide well with the test results quantitatively and qualitatively. It is also known that
the mechanical behavior of sands with the same density but different confining stresses can also
be reproduced uniquely with one set of the same material parameters in all different conditions.
147
0
1
2
3
4
0 5 10 15 20 25
Axial strain a (%)
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)
'0=2.0 MPa, e=0.908
'0=1.0 MPa, e=0.906
'0=0.1 MPa, e=0.908
Toyoura sand
( e=0.907, Dr=18.5% )
0
1
2
3
4
0 1 2 3 4
Effective mean stress p' (MPa)
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)
Toyoura sand
( e=0.907, Dr=18.5% )

0
1
2
3
4
0 5 10 15 20 25
Axial strain
a
()
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)
0
1
2
3
4
0 1 2 3 4
Mean effective stress p'(MPa)
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)
[3]

(a) Loose sand (e=0.907, Dr=18.5%) (a) Loose sand (e=0.78)
0
1
2
3
4
0 1 2 3 4
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)
Toyoura sand
( e=0.833, Dr=37.9%)
0
1
2
3
4
0 5 10 15 20 25
Axial strain a (%)
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)
Toyoura sand
( e=0.833, Dr=37.9% )
'0=3.0 MPa
'0=2.0 MPa
'0=1.0 MPa
'0=0.1 MPa

0
1
2
3
4
0 5 10 15 20 25
Axial strain a (%)
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)
0
1
2
3
4
0 1 2 3 4
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)

(b) Medium dense sand (e=0.833, Dr=37.9%) (b) Medium dense sand (e=0.70)
0
1
2
3
4
0 1 2 3 4
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)
Toyoura sand
( e=0.735, Dr=63.7%)
0
1
2
3
4
0 5 10 15 20 25
Axial strain a (%)
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)
Toyoura sand
( e=0.735, Dr=63.7% )
'0=3.0 MPa
'0=2.0 MPa
'0=1.0 MPa
'0=0.1 MPa

0
1
2
3
4
0 5 10 15 20 25
Axial strain a (%)
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)
0
1
2
3
4
0 1 2 3 4
D
e
v
i
a
t
o
r

s
t
r
e
s
s

q

(
M
P
a
)

(c) Dense sand (e=0.735, Dr=63.7%) (c) Dense sand (e=0.65)
Figure 9. Test results of TSS Figure 10. Simulation of the test results in Fig.9
(Verdugo and Ishihara, 1996)

Table 7 Initial conditions of TSS samples before undrained triaxial compression
No.
Confining
stress
p' (MPa)
Initial
void ratio
e
0
Initial degree of
overconsolidation
OCR
Initial degree of
structure
R
0
*
Initial
anisotropy

0
0.1 0.776 40.0 0.21 0.0
1.0 0.781 3.0 0.20 0.0
2.0 0.781 1.6 0.15 0.0
0.1 0.698 111.0 0.45 0.0
1.0 0.698 12.0 0.30 0.0
2.0 0.703 9.0 0.16 0.0
3.0 0.709 5.3 0.15 0.0
0.1 0.651 270.0 0.55 0.0
1.0 0.651 35.0 0.30 0.0
2.0 0.651 16.0 0.30 0.0
3.0 0.651 12.0 0.25 0.0

3.5 Dense sand subjected to drained cyclic loading
Finally, behaviors of dense sand subjected to drained cyclic loading under constant-mean-
effective-stress constant mean effective stress are simulated. The confining pressure of the sand
is 196 kPa and cyclic loading condition is that the mean effective stress is kept constant and a
maximum principal stress ratio (
1
/
3
) is loaded to 4. Figure 11 shows the test results by Hi-
nokio (2000), in which the stress-strain curves are plotted in terms of effective stress ratio
1
/
3
, a dimensionless normalized stress. The volumetric strain shows dilatancy at the very be-
148
ginning under cyclic loading and then turns to compression until it reaches a steady state at
which the compression almost stops, as shown in Figure 11 (b). For deviatory stress-strain rela-
tion, at the beginning, it shows a relatively large loop, as the cyclic loading number increases,
however, the stiffness of the sand grows up and the stress-strain relation comes into an almost
fixed loop as shown Figure 11(c). In the simulation, the parameters of the sand are the same as
those listed in Table 2 and initial conditions of dense sand are shown in Table 8. In determining
the initial conditions of the sand, it is assumed that the sand is well-remolded one with ex-
tremely low structure and relatively high OCR. As can be seen in Figure 12, the overall charac-
teristics of the sand predicted by the theory, for instance, the changes in dilatancy and stress-
strain relations, agree qualitatively well with the test results, but showing a slight over-
estimation of volume strain. It should be emphasized here that in the simulation, volumetric
compression also stopped automatically after certain cycles of loadings, which is completely
coincides with the reality. The reason why the model can describe this behavior is quite simple.
Taking a look at Figures 12(d) and (e), in which the changes of overconsolidation and stress-
induced anisotropy are plotted, it is easy to find out that during plastic loading the degree of
overconsolidation sometime may even increase, not always the case in which overconsolidation
only develops in elastic unloading process. One of the most important features of the model is
that the changing rate of overconsolidation is assumed to be controlled by two factors, namely,
the plastic component of stretching and the increment in anisotropy. The physical meaning of
the assumption is clear because stress-induced anisotropy is usually dependent on the roundness
of soil particles and their orientation of deposition. Therefore a sand with strong anisotropy will
have a stronger resistance against volumetric change than those of weak one in which soil parti-
ciples deposit rather randomly. It is clear from the figures that during cyclic shearing, overcon-
solidation gets higher and higher, in other words, the density is getting higher, resulting in the
difficulty to further compression.

(a) (b) (c)
Figure 11. Test results of dense sand in drained cyclic loading (after Hinokio, 2000)

Table 8. Initial conditions of sand before drained cyclic loading
0
0.661
Confining stress p (kPa) 196.0
0
*
0.99
0
55.0
Initial anisotropy
0
0.0

149
-4
-1.5
1
3.5
6
0 150 300
S
t
r
e
s
s

r
a
t
i
o

1
/
3
comp.
ext.
196
a/r=5.4
r/a=5.4
3.5
6

-2
0
2
4
6
8
-4 -1.5 1 3.5 6
Stress ratio
1
/
3
V
o
l
u
m
e
t
r
i
c

s
t
r
a
i
n

v

(
%
)

0
comp. ext.
6 3.5

-4
-1.5
1
3.5
6
-0.5 0 0.5 1 1.5 2
Deviator strain
d
(%)
S
t
r
e
s
s

r
a
t
i
o

1
/
3
comp.
ext.
3.5
6

(a) (b) (c)
0
300
600
900
1200
-4 -1.5 1 3.5 6
Stress ratio
1
/
3
O
C
R
comp. ext.
6 3.5

0
0.25
0.5
0.75
1
-4 -1.5 1 3.5 6
Stress ratio
1
/
3
D
e
g
r
e
e

o
f

a
n
i
s
o
t
r
o
p
y

comp. ext.
6 3.5

(d) (e)
Figure 12. Simulated stress paths, stress ratio-strain relations of dense sand in drained cyclic loading under
constant mean effective stress
4 SUMMARY
In this study, based on conventional drained triaxial compression tests and undrained triaxial
cyclic loading tests, material parameters of Toyoura Standard sand is determined and other tests
under different loadings and drainage conditions are simulated by the constitutive model. The
capability of the model to describe uniquely the overall behaviors of the sand under different
drainage conditions and different loadings with one set of fixed parameters, is verified and the
following conclusions can be given.
1) The evolution parameter a, the parameter controlled the degradation of structure, is deter-
mined to be 0.5 from the parametric study based on the undrained cyclic triaxial test. Mean-
while, from the parametric study based on the drained triaxial compression test, reference
void ratio N is determined to be 0.87.
From the comparison between experiment and theoretical results obtained from undrained
triaxial cyclic loading tests and drained triaxial compression tests, it is known that the con-
stitutive model can describe the test results relatively well.
2) The behavior of sand is known to be largely dependent on its density. For this reason, a se-
ries of simulations on sands with different densities originally compacted from the same
sand are conducted with one set of material parameters. The results reveal the fact that the
mechanical behaviors of sand subjected to cyclic loading under drained/undrained condition
can be uniquely and properly described by the constitutive model no matter what density it
may have. Furthermore, it is concluded theoretically that if the compaction conducted under
drained condition, then the consolidation of sand will occur; while under undrained condi-
tion, the liquefaction will occur.
3) Under undrained triaxial compression test, loose sands exhibit peak strength at first in
small strain level, then collapse and flow rapidly to the original, showing a typical strain
hardening and strain softening behavior. For medium dense sands, stiffness of the sands de-
crease abruptly in small strain but strain hardening continues within an intermediate range of
shear strains and finally followed by a limited strain softening, in which a relative higher
strength can be approached in spite of a limited strain softening. Dense sand, however, only
150
shows strain hardening and large shear strength can be reached. In the drained condition, it is
observed that all sands approached to the same values of ultimate shear stress and the same
void ratio, irrespective of different initial densities at the beginning of shearing.
4) Confining-stress dependency of sand, a typical behavior of sands with the same density but
different confining stresses, can also be reproduced with the set of fixed material parameters.
5) Dense sand subjected to drained cyclic loading under constant mean effective stress is also
simulated. The overall characteristics of the sand is predicted well by the theory, for in-
stance, the changes in dilatancy and stress-strain relations are qualitatively the same as the
test results with a slight over-estimation of volume strain.
It cannot say that the model can perfectly describe the colorful behaviors of TSS, but that the
model can give a unified description of TSS to a rather satisfactory level if considering the fact
that only eight material parameters are employed in the model. Meanwhile, if the influence of
intermediate principal stress can be properly taken into consideration like the t
ij
concept (Nakai
and Mihara, 1984), the model would be much better.
REFERENCES
Asaoka, A., Nakano, M. and Noda. T. (1998): Super loading yield surface concept for the saturated struc-
tured soils, Proc. of the Fourth European Conference on Numerical Methods in Geotechnical Engi-
neering-NUMGE98, 232-242.
Asaoka, A., Noda, T., Yamada, E., Kaneda, K. and Nakano, M. (2002): An elasto-plastic description of
two distinct volume change mechanisms of soils, Soils and Foundations, Vol.42, No.5, 47-57.
Hashiguchi, K. and Ueno, M. (1977): Elastoplastic constitutive laws of granular material, Constitutive
Equations of Soils, Pro. 9th Int. Conf. Soil Mech. Found. Engrg., Spec. Ses. 9, Murayama, S. and
Schofield, A. N. (eds.), Tokyo, JSSMFE, 73-82.
Hinokio, M. (2000): Deformation characteristic of sand subjected to monotonic and cyclic loadings and
its application to bearing capacity problem, Doctoral Dissertation, Nagoya Institute of Technology,
33-38 (in Japanese).
Ishihara, K. (1993): Liquefaction and flow failure during earthquake, The 33
rd
Rankine Lecture, Geotech-
nique.
Nakai, K. (2005): An elasto-plastic constitutive modeling of soils based on the evolution laws describing
decay of soil skeleton structure, loss of overconsolidation and development of anisotropy , Doctoral
Dissertation, Nagoya University, 115-118 (in Japanese).
Nakai, K., Nakano, M., Noda, T. and Asaoka, A. (2004): Description of Compaction and Liquefac-
tion behavior of sand based on evolution of soil skeleton structure, Proc. 2
nd
International Workshop
on New Frontiers in Computational Geotechnics, Fortaleza, Brazil, Zhang et al. (eds), 135-144.
Nakai, T. and Mihara, Y. (1984): A new mechanical quantity for soils and its application to elastoplastic
constitutive models, Soils and Foundations, Vol.24, No.2, 82-94.
Nakai T. and Hinokio M. (2004): A simple elastoplastic model for normally and over consolidated soils
with unified material parameters, Soils and Foundations, Vol.44, No.2, 53-70.
Oka, F., Yashima, A., Kato, M. and Sekiguchi, K. (1992): A constitutive model for sand based on the non-
linear kinematic hardening rule and its application, Proc. 10th World Conf. Earthquake Engineering,
Madrid, Vol.5, Balkema, 2529-2534.
Roscoe, K. H., Schofield, A. N. and Thurairajah, A. (1963): Yielding of clays in states wetter than critical,
Geotechnique, Vol.13, No.3, 250-255.
Sago, H. (2008). The evaluation of earthquake resistant capability on self-standing steel sheet pile, Master
Dissertation, Nagoya Institute University (in Japanese)
Schofield, A. N. and Wroth, C. P. (1968): Critical State Soil Mechanics, McGraw-Hill.
Verdugo, R. and Ishihara, K. (1996): The steady state of sandy soils, Soils and Foundations, Vol.36, No.2,
81-91.
Yashima, A., Oka, F., Shibata, T., and Uzuoka, R. (1991): Liquefaction analysis by LIQCA, Proceedings
of JGS Conference on Liquefaction of Ground and its Counter measure, 165-174 (in Japanese).
Ye, B., Ye, G., Zhang, F. and Yashima, A. (2007): Experiment and numerical simulation of repeated liq-
uefaction-consolidation of sand, Soils and Foundations, Vol.47, No.3, 547-558.
Zhang, F., Ye, B., Noda, T., Nakano, M. and Nakai, K. (2007): Explanation of Cyclic Mobility of Soils:
Approach by Stress-Induced Anisotropy, Soils and Foundations, Vol.47, No.4, 635-648.

151
APPENDIX: BRIEF DESCRIPTION OF CONSTITUTIVE MODEL
The model (Zhang et al., 2007) is proposed based on the concepts of subloading (Hashiguchi and
Ueno, 1977) and superloading (Asaoka et al., 1998). Here a brief description of the model is given.
The similarity ratio of the superloading surface to normal yield surface R
*
and the similarity ratio of the
superloading surface to subloading surface R are given as,
'
* *
' ' '
, 0 1
p q q q
R R and
q p p p
= = < =
% % %
%
A-(1)
'
' ' ' '
, 0 1,
p q q q q
R R and
q p p p p
= = < = =
%
%
A-(2)
where,
' ' '
( , ), ( , ) and ( , ) p q p q p q % % represent the present stress state, the corresponding normally con-
solidated stress state and the structured stress state at p-q plane, respectively. The normally yield surface is
given in the following form as:
' 2 2 2
'
' 2 2
0 0
0
M
( , , ) tr MDln MDln tr 0
M
t t
p p
p
f p J d J d
p

+
+ = + + =

% %
% %
%
D D A-(3)
where,
* *
= % , and the other variables involved in Equations, (1), (2) and (3) are defined as:
' ' '

3 / 2, , tr / 3
' '
p p , p
= = + = S/ S T I T = , = = , = = , = = , = A-(4)
2/ 3 = , 2/ 3 = A-(5)
where, S SS S is the deviatory stress tensor; is the anisotropic stress tensor, and
'
T is the Cauchy effec-
tive stress tensor and is assumed to be positive in tension. In Equation A-(3), J is the Jacobean determina-
tion of deformation gradient tensor F and can be expressed as:
0 0
1
det
1
v e
J
v e
+
= = =
+
F A-(6)
where v and v
0
are the specific volume at the current time (t) and the specific value at the reference
time (t=0). D is the dilatancy parameter which can be expressed by ,
%
% the compression and the swell-
ing index, respectively, as follows:
0 0
D
M(1 ) Mv e

= =
+
% %
% %
A-(7)
D
p
denotes the plastic component of stretching D which is assumed to be positive in tension, and is re-
lated to the plastic volumetric strain rate in the following form under the condition that the compressive of
the volumetric strain is supposed to be positive:
0
tr
t
p p
v
J d =
D
A-(8)
By substituting Equations A-(1) and A-(2) into Equations A-(3), subloading yield surface can also be
written as:
' *
0
' 2 2 2
' 2 2
0
0
( , , ) MDln MDln tr
M
MDln MDln MDln MDln tr 0
M
t
p
t
p
f p R R J d
p
R R J d
p

+ +
+
= + + + =
D
D
A-(9)
An associated flow rule is employed in the present model, namely,
'
/
p
f = D T A-(10)
The consistency equation for the subloading yield surface can then be given as:
'
MD MD tr 0
p
f f R R
J
R R

+ + + =

o
o & &
T D
T
'
A-(11)
where,
o
T
'
and
o
are the Green-Naghdi rates of stress tensor T
'
and anisotropic stress tensor , re-
spectively. is material spin tensor. It is easy to obtain the following relation:
152
2 2 2
2 2 2
' ' ' 2 2 2 '
1 ( ) M
MD( /(M )) MD
(M )
f
p p p p

= + + =
+
A-(12)
From Equation A-(12), it is clear that the C.S.L., defined by the condition in which
'
/ 0 f p = , always
satisfies the relation =M, implying that the C.S.L., as the threshold between plastic compression and
plastic expansion, does not move with the changes in the anisotropy.
Evolution rule for the anisotropic stress tensor is defined as:
(M ) 2 / 3
D
p
r s
J
b =
o

D
A-(13)
where, <M provides a natural physical limitation on the development of anisotropy automatically., it is
known from the evolution Equation A-(13) that development of anisotropy will stop at the state
when = .
Evolution rule for degree of structure R*, which is the same as in the work by Asaoka et al. (2002), is
adopted:
2 / 3
p
s
R JU

=
&
D , (1 ) / D U aR R

= , ( 0 1 R
< ) A-(14)
The changing rate of overconsolidation is assumed to be controlled by two factors, namely, the plastic
component of stretching and the incremental anisotropy as,
MD M
p
R f
R JU

= +
o
&
D
A-(15)
In which,
o
is proportional to the norm of the plastic component of stretching
p
s
D and U is given
by the following relation as:
' ' 2
0
' ' 2
0
( / )
( ) ln
D ( / ) 1
p p m
U R
p p
=
+
A-(16)
where
'
0
p =98.0 kPa is reference stress.
If the stretching is divided into elastic and plastic components, and the elastic components follow
' '
'
, ,
e e p
f
= = + =
o o
T ED D D D T ED E
T
A-(17)
The positive valuable ( =) can be rewritten as :
'
2 2
s ' ' 2 2 *2 '
MD
(M )
(M )
f
f f
J
p

=

+
+
ED
E
T T
T A-(18)
where

' ' 2
2 2 *2 2 2 2 0
s ' ' 2
0
*2 2
* *
2 2 *2 2 2
( / ) Mln 1
M M { } 6 (M )
( / ) 1 3
6M (M ) (2M 3 )
2 M(1 ) (1 )
M (M )(M )
r
p p m R
R p p
b
a R

= +
+

+
+

A-(19)
The loading criteria are given below:

0
0
0
loading
neutral
unloading
>
<
A-(20)
In most cases, the denominator is positive, therefore, 0 > is equivalent to the following relation:

'
0
f
>
ED
T
A-(21)
153
2.
EarthquakeObservations,GroundMotions
Impact of Moderate Earthquakes in Post-Bhuj Era: Case Study of

Sikkim 2006 and Durgapur 2008 Earthquakes, India

S. C. Dutta
Professor of Civil Engineering, School of Infrastructure, Indian Institute of Technology, Bhubaneswar,
Orissa, India (formerly Professor, Department of Civil Engineering, Bengal Engineering and Science
University, Shibpur, Howrah, West Bengal, India)
P. Mukhopadhyay
Assistant Professor, Department of Architecture, Town & Regional Planning, Bengal Engineering and
Science University, Shibpur, Howrah, West Bengal, India
S. Bhattacharya
Lecturer in Dynamics Engineering, Department of Civil Engineering, University of Bristol, Queens
Building, Bristol, United Kingdom
ABSTRACT: The 2001 Bhuj Earthquake was an eye opener to the Indian earthquake research-
ers as it established the fact that devastation of earthquakes was not necessarily restricted to in-
ter-tectonic regions, but may occur in intra-tectonic regions as well. All these motivated the au-
thors to study and compare the impact of moderate earthquakes in hilly areas as well as in plain
lands and to closely scrutinize the local building and construction practices from the view point
of their adequacy in seismic resistance. The present paper reports one such endeavour where the
impact of two moderate earthquakes of magnitudes 5.7 and 4.0, occurring respectively at Sik-
kim in 2006 and Durgapur in 2008, are compared. Though the earthquakes were of modest na-
ture, yet their impacts were extensive. Reconnaissance based damage surveys show that the ef-
fect was more pronounced because of the presence of a number of non-engineered or semi-
engineered, one or two-storied dwellings, whose walls were mostly made of brick masonry and
mud, sometimes using wood or bamboo as framing elements. Development of vertical and in-
clined shear cracks at the corner of wall junctions, and that of horizontal shear cracks above the
plinth and near the roofs, were common features of damages at both the hilly areas of Sikkim
and plain lands of Durgapur. In addition to these, inclined slopes, inadequate foundations and
possibility of pounding between adjacent buildings are additional factors of vulnerability ob-
served during the damage survey at Sikkim. The analysis of the damage profiles and suggested
remedial measures indicate that simple guidelines, adequately worked out in the socio-economic
context of a locality and country, if strictly implemented, may be the best viable solution with-
out changing the nature of habitats.
1 INTRODUCTION
It was generally conceived by Indian earthquake researchers before 2001 that earthquakes are
restricted only to inter-tectonic zones. Accordingly vast territories of India were considered as
free from seismicity and were depicted as Zone I in the Map of India showing seismic zones of
India by the Bureau of Indian Standards (BIS). But the experience of the 2001 Bhuj Earthquake
was an eye opener as it established the fact that devastation of earthquakes may occur in intra-
tectonic regions as well. This led to a major change of the BIS map of India (IS 1893: 1984 up-
dated to IS 1893 Part 1: 2002) showing four instead of previous five seismic zones, where the
entire Zone I was abolished.
India has been visited by low to moderate magnitude earthquakes with almost regular interval
of about two to three years. On the other hand, earthquake of very severe magnitude is relatively
rare. Data from Table 1, showing list of earthquakes occurring in India, confirms this. Again
though the impact of such medium to low magnitude earthquakes may not be severe on the
framed buildings, one cannot ignore their impact on the so called non-engineered or semi-
engineered structures of India and other developing economies. In fact, the people affected by
the damage or collapse of such structures are generally from the economically weaker sections
of the society who cannot pay for consulting structural engineers and architects! All these con-
154
texts motivated the authors to study and compare the impact of frequent yet moderate earth-
quakes. The present paper reports such endeavours where the impact of two moderate earth-
quakes of magnitudes 5.7 and 4.0, occurring respectively at Sikkim in 2006 and Durgapur in
2008, are compared. The experience of damage survey of these two earthquakes show that how
by implementing very simple measures, specific to local nature of habitat, the damage could
have been reduced. These two earthquakes are really a revelation to find that building con-
sciousness and attitude towards implementation of adequate socio-economy friendly technology
can play a major role in minimizing the severity of damage, casualties and economic loss during
such low to medium magnitude earthquakes.

Table 1. Major Indian earthquakes.
_____________________________________________________________________________________________________
Date Location Magnitude Date Location Magnitud
_____________________________________________________________________________________________________
e
16 Jun. 1819 Kutch, Gujarat 8.0 21 Jul. 1965 Anjar, Gujarat 7.0
10 Jan. 1869 Cachar, Assam 7.5 10 Dec. 1967 Koyna, Maharashtra 6.5
30 May 1885 Sopor, Jammu & Kashmir 7.0 19 Jan. 1975 Kinnaur, Himachal Pradesh 6.2
14 Jul. 1885 North-West of Dhaka 7.0 06 Aug. 1988 Manipur-Myanmar border 6.6
12 Jun. 1897 Shillong plateau 8.7 21 Aug. 1988 Bihar-Nepal border 6.4
04 Apr. 1905 Kangra, Himachal Pradesh 8.0 20 Oct. 1991 Uttarkashi, Uttar Pradesh 6.6
08 Jul. 1918 Srimangal, Assam 7.6 30 Sep. 1993 Latur, Maharashtra 6.3
02 Jul. 1930 Dhubri, Assam 7.1 22 May 1997 Jabalpur, Madhya Pradesh 6.0
15 Jan. 1934 Bihar-Nepal border 8.3 29 Mar. 1999 Chamoli, Uttar Pradesh 6.8
31 May 1935 Quetta 7.6 26 Jan. 2001 Bhuj, Gujarat 6.9
26 Jun. 1941 Andaman Islands 8.1 14 Sep. 2002 North Andaman Islands 6.5
23 Oct. 1943 Assam 7.2 08 Oct. 2005 North Kashmir 7.6
15 Aug. 1950 Arunachal Pradesh 8.5 14 Feb. 2006 Sikkim 5
_____________________________________________________________________________________________________
.7
2 DESCRIPTION OF EARTHQUAKES
2.1 Sikkim Earthquake
An earthquake of magnitude 5.7 in the Richter scale violently rocked Sikkim on 14th February
2006, which though did not cause major casualties, but had resulted in considerable structural
damage. Sikkim, one of the hilly states of India, situated in the Eastern Himalayas, is wedged
between Nepal in the west, Bhutan in the east and China in the north and northeast. The inten-
sity of the earthquake as experienced by various parts of Gangtok city, capital of Sikkim, as
well as its surrounding region in East and South Sikkim, was in the range of V to VI as per the
MSK scale. On the other hand, this area comes under the earthquake Zone IV where a severe
seismic intensity is generally expected. In this context, the present earthquake can be treated as
a warning only, in view of the widespread building construction undertaken in different hilly
centres of Sikkim, many of which are not properly designed. As per the Indian Metrological
Department (IMD), the earthquake, which had its epicentre at Chungthang, 75 km north of
Gangtok, lasted 22 seconds and caused extensive damage to property. However, United Nations
Development Programme, Bhubaneswar reports the same earthquake to have the epicentre at 25
km west-north of Gangtok. Such an earthquake has also another specialty of being followed by
a large number of aftershocks, and in this particular case, tremors from about 10 such earth-
quake aftershocks were reported by the IMD.
Two Indian Army soldiers were killed when the vehicle they were travelling in was struck by
a rock fall at Sherathang near the border outpost at Nathu La in Sikkim. Two people were in-
jured in east Sikkim by this earthquake. A child received minor scratches when a flower pot fell
on him during the earthquake at Ranipool while a man suffered minor head injuries after being
hit by a boulder along the Indira Bypass. In addition, 8 people were discharged after being ad-
ministered first aid at the Tashi Namgyal Memorial Hospital in Gangtok. The people in Gangtok
felt very severe shaking. The furniture was displaced from their positions. Books and other
loose objects were thrown from shelves. Hanging lights and fans started swinging. On the other
hand, New Jalpaiguri which was about 120 km away from Gangtok seems to have experienced
almost an intensity of V as understood from the awakening of sleeping persons and displace-
155
ments of small unstable objects. So, less change in intensity over a large distance indicates geo-
logical strata has a better transmissibility of the seismic waves and increases the concern about
the seismic vulnerability of not only Sikkim, but also the surrounding plane land.
The Department of Land Revenue and Disaster Management, Government of Sikkim, in its
Revised Memorandum for Restoration Works of Earthquake Damages of 14.02.2006, Sikkim,
has claimed that more than 500 buildings have been damaged by the earthquake. In view of the
same, the Government of Sikkim has sought to the Government of India a Central Assistance of
Rs. 301.62 million.
2.2 Durgapur Earthquake
A mild earthquake was felt in the Gangajalghati-Raniganj-Mejia region of the state of West
Bengal, India, as well as in the adjacent districts of the state of Jharkhand, India, in the morning
of 6th February, 2008. As per IMD, the tremor was felt at around 11:39 hrs. IST, and had a
magnitude of 4.3 in the Richter scale, lasting for 10-15 seconds. The epicentre of the quake was
reported to be near Andal in Bardhaman district of West Bengal. The main earthquake was fol-
lowed by three mild aftershocks. Another aftershock was felt at Bankura, Barjora, Gangajal-
ghati, Indpur, Mejia, Raniganj and Onda as well as in other parts of the districts of Bankura,
Burdwan and West Midnapore in the evening at 17:12 hrs. IST, which was reported to last for 3
seconds.
Panic-stricken people rushed out of their houses towards open fields as doors and windows of
buildings rattled. School authorities at many places declared holiday fearing any eventuality. An
elderly woman was killed at Raniganj and at least 50 people injured in this earthquake, includ-
ing two persons who were struck by falling debris, one each in Bishnupur and Gangajalghati in
Bankura district. 12 students at the Durvalpur High School at Mejia and 12 students at Patrahat
Girl's School in Onda fell unconscious as a result of the tremor. Reports from other districts said
that the quake was severe as several buildings developed cracks, including eight railway signal-
ling cabins on the Delhi-Howrah mainline. Over two dozen mud houses were collapsed at Gan-
gajalghati as did a few abandoned buildings at Searsole near Raniganj. A few mobile telecom
towers collapsed and production at the Mejia Thermal Power Station stopped.
3 OBSERVATION OF DAMAGES
3.1 Sikkim Earthquake
The damage survey of the Sikkim Earthquake was conducted from 26th March to 1st April,
2006. A large number of government and private buildings in the Gangtok city and in the sur-
rounding regions have undergone both structural as well as non-structural damages. Damages
suffered by purely framed structures were limited primarily to non-structural ones. However, the
infill walls of some framed structures, which were constructed with poor masonry, suffered
badly due to the seismic shocks; these included government buildings as well. In the midst of
the different engineered structures, some traditional buildings, built with combination of wood
and bamboo, were observed to have suffered less. Damage profiles of some of the significant
buildings are described below.
3.1.1 Legislative Assembly, Gangtok:
The assembly house is a reinforced concrete frame-building in hexagonal shape, whose main
hall consists of eight large sized columns arranged in circular pattern. The columns are having
overhanging beams at the top supporting the domical roof. The building is situated on a hill
slope with perimeter-columns having isolated foundations at different levels. The building ex-
perienced damage earlier and was repaired. However, such cracks are observed to be aggravated
in this occasion. These cracks are generally in the vicinity of the perimeter and confined in the
infill wall. The floor of the building on the rear side has experienced a settlement by about a few
inches. Views of some cracks in roof supporting beams due to such settlements are presented in
the photographs of Figure 1. The observation shows that intensity of the earthquake was not
adequate to cause damage to the framed building.

156
(a) Horizontal cracks (b) Vertical cracks

Figure 1. Shear cracks in the beams supporting the roof of the Legislative Assembly at Gangtok, Sikkim.

3.1.2 Tashling Secretarial building, Gangtok:
This building is a five-storey reinforced concrete framed structure. The infill walls are made of
chisel dressed stone masonry at ground floor. The infill walls at the upper floors consist of ma-
sonry made with mortar blocks. These masonry walls have two layers with a hollow portion in
between, probably for the purpose of insulation. The primary nature of the damage is of non-
structural type. The infill wall has been separated from the actual frame of the building. In addi-
tion to that, the infill walls have undergone considerable shear cracking (Fig. 2a) accompanied
by local failure at places as shown in Figure 2b.

(a) Shear crack on the walls (b) Failure of hollow walls

Figure 2. Damaged walls of the Tashling secretariat building at Gangtok, Sikkim.

Again, considerable pounding observed at the separation joint indicates that the relative
movement of one side of the building is higher than the other. The damage of the infill wall
should be retrofitted by providing reinforcements to the extent possible apart from the patch re-
pairing through mortar grouting. On the other hand, separation joint should be made clear
enough by removing any mortar or stiff structural material.
3.1.3 Police headquarter building, Gangtok:
This three-storied reinforced concrete framed building has partly open ground storey provided
for parking (Fig. 3a), while the other portion is having basement. The portion adjacent to the
basement has undergone considerable damage. Primarily, the infill walls on one side have been
damaged perhaps due to excess deformation on that side resulting from lack of symmetry. The
nature and reasons of the damages are very similar to those observed in earlier cases as may be
157
understood from Figure 3b. Partly open ground storey has not only made this storey weaker
than the other storeys, it has also introduced asymmetry endangering the structure in the event
of an earthquake.

(a) Partially open ground storey of the police head (b) Horizontal cracks in infill walls adjacent to
quarter building the basement

Figure 3. Police headuarter building at Gangtok, Sikkim.

3.1.4 Office Building of the Geological Survey of India (GSI), Deorali:
The GSI office is housed in a four-storied building at Deorali. The building has a regular beam-
column grid at the front side; whereas, some of these frames are not continuous at the backside.
This has introduced irregularity in plan as well as in elevation of the building, which causes se-
vere damage at the backside. Figure 4 illustrates one such damage, which comprises of damage
of beam-column joints leading to distortion of the frame, which in turn, has caused crushing of
the brick walls at many places. Attempts have been made to repair the cracks by mortar. How-
ever, such repairing did not involve any attempt for structural strengthening. The partially open
ground storey behaving as soft storey further acts as reasons of vulnerability of the building.

Figure 4. Damaged infill walls at the backside of
the office building of the GSI at Deorali, Sikkim.

158
3.1.5 Government Secondary School, Sichey:
The out-of-plane rotation suffered by the walls of the Government Secondary School building at
Sichey, East Sikkim, has caused the separation of the wall from the main structural frames and
the shear cracking has caused further damage to them. Figure 5 illustrates the case where a view
is taken from inside of one of the devastated class rooms. The provision of vertical reinforce-
ments and horizontal reinforced concrete bands could have solved the problems. The nature as
well as reason of the non-structural damages in this school building is similar to all other struc-
tures.

Figure 5. Damaged infill wall showing out-of-plane
rotation of the school building at Sichey, Sikkim.

3.1.6 A private building at Deorali, National Highway 31A (NH-31A):
Generally, the private buildings are designed with less financial investment and that is why the
damages in the private buildings are found to be much more than government ones. While the
damage in most of the government buildings were limited to the non-structural elements, both
structural and non-structural elements are found to be considerably damaged in private or semi-
private buildings.
The severely damaged private building, under discussion, is situated beside the NH-31A. It
has four storeys at its rear side, while three storeys are at the front; thus one storey of the build-
ing is along the hill slope (Fig. 6a). The overall asymmetry as well as irregular placement of
columns, which are alike features of many buildings at hilly regions, are the prime structural de-
ficiencies. In fact, Figure 6a further shows that the damage is primarily concentrated at the rear
side of the building, which seems to lack from adequate framing action. The fact that the col-
umns are severely damaged, especially at the rear portion along with the infill walls, is clear
from Figure 6b. The figure further points out that the columns are found to be devoid of ade-
quate lateral ties; this has caused the concrete to easily bulge out from the core.
In some places, the infill walls have been very severely damaged exposing the mortar blocks
(Fig. 7). This points out to the fact that infill walls are not only made by mortar blocks of inade-
quate strength; there is also a lack of adequate cementing material due to adoption of poor con-
struction practices. Figure 7 further shows clear gaps between the mortar blocks. Such inade-
quate quality has been one of the major reasons of failure of many such infill walls.
Furthermore, early failure of these walls might have lead to transfer of more lateral load to the
columns leading to severe damage of the same.

159
(a) Side view of the damaged private building (b) Damaged rear columns and infill walls

Figure 6. Damaged RC framed private building of Mr. Sharma at Deorali, Sikkim beside NH-31A.

Figure 7. Poor construction practice of use of
very less mortar is exposed in infill walls of the
damaged building at Deorali, Sikkim.

3.1.7 Better performance of few non-engineered buildings:
It was observed at Gangtok that many-a-times buildings constructed with wood, bamboo and
other traditional non-engineered materials, sustained the aftermath of earthquakes rather better
than those built with so called engineered materials. The point is best demonstrated through
Figures 6a and 8, in which one can find an undamaged private residence built with wood and
bamboo, plastered with mud, situated at the rear side of the damaged private building at Deorali,
NH-31A. The light weight of these materials may explain the reason of the survival. However,
this warrants further investigation.

160

Figure 8. Close-up of the undamaged residence
built with traditional materials, viewed from
within the damaged building at Deorali, Sikkim.

3.2 Durgapur Earthquake
The damage survey was conducted from 29th February to 2nd March, 2008. As per the recom-
mendations of Census of India - 2001, the buildings covered during the damage survey can be
broadly classified into three categories, viz. Permanent houses, Semi-permanent houses, and,
Temporary houses. This classification is based on the type of material used in the construction
of roof and wall. Here, permanent houses are defined as such houses, the walls and roof of
which are made of permanent materials; whereas, houses in which both walls and roofs are
made of materials, which have to be replaced frequently are defined as temporary houses. On
the other hand, houses, in which either the wall or the roof is made of permanent material, and
the other is made of temporary material are called as Semi-permanent houses. It may be men-
tioned that the houses surveyed in the urban and suburban areas were either of the permanent
or semi-permanent category; but those surveyed in the villages belonged to either of the semi-
permanent or temporary types. Damage profiles of buildings belonging to each category are
described below.
3.2.1 Permanent houses:
The permanent houses surveyed are mostly single or double storied structures, whose rein-
forced concrete floor / roof slabs are supported by reinforced concrete beams, which in turn rest
on load bearing brick masonry. The moderate 2008 Durgapur Earthquake produced cracks pre-
dominantly along the junctions of different structural components, like those between beams
and walls, between walls and roof slabs, between walls and walls etc. Figure 9 shows one such
example where one can see development of vertical shear cracks at the corner of the room from
roof to floor. Horizontal shear cracks have also developed at the junction of ceiling with walls.
In some other earthquake affected buildings, shear cracks occurred at the window sills, and at
the junctions between door-window frames and walls. However, these cracks were minor in na-
ture. Further, it was observed that the wall and roof junctions of some of the earthquake affected
buildings was clamped with wood or iron angles; and eventually those roofs were protected
from possible sliding and overturning.

161

Figure 9. Minor vertical and horizontal shear cracks
produced at the junctions of ceiling with walls in a
single-storied building at DPL colony, West Bengal.

3.2.2 Semi-permanent houses:
The walls of all the semi-permanent houses were made of load bearing brick masonry, one of
the most popular permanent building construction materials in the tropics. The temporary roofs,
however, varied between burnt clay tiles and asbestos cement sheets supported on bamboo
and/or wooden roof structures. The cracks occurring in these buildings were, obviously, more
severe than those surveyed in the permanent ones. Figure 10 depicts a typical semi-permanent
house surveyed at Village Mejia in West Bengal, India.

Figure 10. A typical semi-permanent house at
Village Mejia, West Bengal.

The photographs at Figure 11 illustrate the different types of failures that had occurred in
these semi-permanent houses due to the 2008 Durgapur Earthquake. Whereas Figure 11a dem-
onstrates failure of part of a wall, which has suffered out-of-plane rotation, Figure 11b demon-
strates the failure of two mutually perpendicular walls forming a T-junction. Absence of proper
toothing between two mutually perpendicular walls has been the main reason behind the second
failure. Vertical cracks appear in the external brick masonry of another single-storied building
shown at Figure 11c, which has occurred probably due to absence of reinforced concrete beam
at the door-lintel level. The photograph at Figure 11d shows the example of a horizontal crack,
which has further resulted into failure of mortar and brick masonry.
162

(a) Part of a wall failing due to out-of-plane rotation (b) Failure at the T-junction between two walls
nry walls probably due to (d) Failure of brick and mortar masonry
houses damaged during the 2008 Durgapur Earth-
uake, surveyed at Village Mejia in West Bengal.

o perform
etter by offering more resistance to earthquake forces in the form of framing action.

(c) Separation of maso
absence of RC frame

igure 11. Failure profiles of different semi-permanent F
q

3.2.3 Temporary houses:
The temporary houses, which were analysed during the damage survey used mud as the princi-
pal walling material, and thatched roof or burnt clay tiles on bamboo roof-structure as the major
roofing material. The photographs at Figure 12 show complete failure of one of the gable-end
walls of a two-pitched damaged mud house at Village Birvanpur of West Bengal due to the
2008 Durgapur Earthquake. Figure 12b depicts the overturning of the wall due to out-of-plane
rotation, illustrating the separation of two mutually perpendicular walls due to absence of any
significant tying in between them.
The photograph of the temporary house shown in Figure 13, surveyed in the same village, il-
lustrates the fact that though it suffered from out-of-plane rotation, it did not collapse like the
one observed in Figure 12. This is probably due to the fact that the gable-end wall of this house
was not constructed with only mud; rather it was made of brick in combination with mud. The
superior strength of this gable-end wall prevented the adjacent damaged mud wall from collaps-
ing.
The photograph at Figure 14 shows the failure of the walls of another temporary house in the
same village probably due to absence of proper framing. Such walls can be made t
b

163
(a) Front view of the collapsing wall, put at rest by (b) Side view of the collapsing wall, showing
a number of supports separation of adjacent walls from corner

Figure 12. Overturning of the gable-end of a mud house due to out-of-plane rotation during the 2008
Durgapur Earthquake at Village Birvanpur, West Bengal.

Figure 13. Brick wall in combination with mud
preventing adjacent wall from collapse during
2008 Durgapur Earthquake at Village Birvanpur,
West Bengal.

Figure 14. Failure of a mud wall during the 2008
Durgapur Earthquake due to absence of proper
raming, surveyed at Village Birvanpur, West Bengal. f

164
4 ANALYSIS OF DAMAGES AND REMEDIAL MEASURES
Observation of the nature of vulnerability of building structures, suffered during the two moder-
ate earthquakes, reveals some common features and understanding. The similarity in the range
of intensity of the two earthquakes in the suffered localities and in the nature of the construction
materials are reasons behind the likeness of the damage profiles, which therefore warrant com-
mon remedial measures. On the other hand, gross difference in the terrain category of the two
places demands separate particular approach in the remedial measures for the hilly region.
These two aspects are discussed separately in the following paragraphs.
4.1 General analysis and recommendations
The general nature of vulnerability of the buildings in both Sikkim and Gangajalghati-Raniganj-
Mejia regions can be attributed to poor quality of in-fill walls. In most of the buildings, the ma-
sonry made by mortar block is used. The material used as cementing mortar seems to be of very
poor quality and at the same time the mortar blocks seem not to have adequate strength. Infill
panel made by this kind of masonry is performing very poorly, undergoing diagonal shear
cracks and overturning. Furthermore, the strength may be poorer due to use of hollow mortar
blocks. These different inadequacies may be avoided by undertaking the following methodolo-
gies.
4.1.1 Out-of-plane collapse
The infill walls are very frequently found to get separated from the rest of the frames because of
out-of-plane rotations. These tendencies may be checked by providing adequate horizontal rein-
forced concrete bands and vertical reinforcements in places as shown in Figure 15, including the
boundary of the openings. Out-of-plane collapse in masonry system may further be arrested if
the walls in mutually orthogonal directions are properly bonded with each other. To this end, in
addition to vertical reinforcements, alternating toothed joint in walls at corners and T-junctions
as shown typically in Figure 16 may be adopted.

Figure 15. Vertical and horizontal reinforced Figure 16. Detail at corners and joints of structures
concrete bands in masonry building structures. with masonry walls.
(Source: Murty, 2005a) (Sources: IS 4326: 1993, IAEE & NICEE: 2004)

165
4.1.2 Arresting shear cracks
It has been observed that the horizontal, vertical and diagonal shear cracks of the earthquake af-
fected structures are generally originating either from the corners of openings or from junctions
of different structural elements viz. those between walls and walls, walls and floor slabs etc.
Shear cracks originating from junction of structural elements may be taken care by the horizon-
tal reinforced concrete bands and vertical reinforcements in masonry, as suggested above. The
possibility of generation of shear cracks from the corner of openings in the walls may be re-
duced significantly by adopting appropriate structural configuration around them. Some impor-
tant guidelines suggested by the Indian standards (IS 4326: 1993) and literature jointly pub-
lished by the International Association for Earthquake Engineering, Tokyo and National
Information Center of Earthquake Engineering, India (IAEE & NICEE: 2004) are pictorially
presented in Figure 17.

(a) Opening in major walls (b) Masonry strengthening around openings

Figure 17. Recommendations for arresting shear cracks in and around openings of masonry structures
((Sources: IS 4326: 1993, IAEE & NICEE: 2004)

4.1.3 Steel reinforcement details
The effect of moderate earthquakes were though less damaging for framed structures, yet few
simple guidelines regarding detailing of steel reinforcement needs to be followed to minimise
the possibility of damage during earthquakes of higher intensity. It was observed during the
damage surveys that the steel detailing of the reinforced concrete beams and columns were not
adequate. The photograph at Figure 18 was taken on-site at Gangtok, illustrating one aspect of
the issue, where not even the stirrup detailing was properly done. It is understood that stirrups
help in three ways; namely, they
(a) carry the shear force and thereby resist diagonal shear cracks;
(b) protect the concrete from bulging outwards due to flexure; and,
(c) prevent the buckling of the compressed longitudinal bars owing to flexure.
In moderate to severe seismic zones, the Indian code for ductile detailing of reinforced con-
crete structures subjected to seismic forces (IS 13920:1993) demands among other requirements
related to stirrups that their both ends should be bent into a 135 hook and extended at least 10
times its diameter beyond the hook to ensure that they do not open out during an earthquake
(Figure 19). Adherence to this simple prescription can be made without any difficulty through
appropriate understanding of the issue and proper supervision.
166

Figure 18. Inadequate stirrup detailing practice Figure 19. Illustration of one of the provisions of
observed at Gangtok, Sikkim. IS 13920-1993 for stirrups (Source: Murty, 2005b).

4.2 Suggestions for Hilly Regions
4.2.1 Inadequate gripping of foundations
The foundations of many buildings are just resting on hard rock as shown in Figure 20. No fur-
ther excavation within the rock bed is made because the surface of the hard rock layer usually
has sufficient bearing capacity. Generally, filling up by soil is observed up to about 1.0-1.5 m
above the rock level. This soil layer is generally retained by constructing masonry wall at the
edge of the hill bed. These masonry walls are made of stone masonry using local stones with
adequate chiselling. The slope of such retaining wall is maintained as one horizontal: four
vertical. Such kind of foundations may not be adequate to provide resistance against sliding
and overturning. To stop sliding, shear keys may be provided. Overturning may be avoided by
provision of friction piles, as shown in Figure 21.

Figure 20. Columns with inadequate Figure 21. Suggestions for provision of friction piles for
foundations simply resting on rock bed, buildings on rock bed at hilly regions to avoid sliding and
at Darjeeling, Eastern Himalayas. overturning.

4.2.2 Vulnerable rear side columns on sloped site
Most of the buildings of hilly regions are situated on sloped site. The road-facing sides of such
buildings have lesser number of storeys involving stiffer frames, while the frames at the back-
sides are longer and are comparatively more flexible. These frames have higher number of sto-
167
reys. Mostly the lower panels of the backside frames do not have infill walls (refer Figure 22);
thus they become flexible and weak in nature due to less strength and stiffness. These backside
frames may be strengthened by provision of bracings or strong lintel beams. In this context, it is
worthwhile to mention that hill side slopes, liable to slide during earthquake, are suggested to be
avoided and only stable slopes should be chosen as site for building construction. Also it would
be preferable to build separate blocks of a building on separate terraces than to build one large
building block with different footings at different terraces of varying elevations. Further, a site
subject to the danger of rock falls has to be avoided.

Figure 22. Buildings with vulnerable backside
frames without infill having exposed foundation
without any gripping at Lohapul near Rangpo,
Sikkim.

4.2.3 Pounding of buildings
Pressure of urbanisation, lack of new land within existing cities, coupled with inadequate im-
plementation of the provisions of the Municipal Building Rules for side open spaces of build-
ings have created such a situation in hilly regions of the developing countries that many-a-time
two buildings are constructed on two adjacent plots, keeping practically no space in between
them. The situation is illustrated through the photographs at Figure 23. Oscillations during the
2006 Sikkim Earthquake did cause such adjacent structures to pound against each other causing
damage to each of them. Strict adherence of municipal building rules can be the only solution to
such problems.

168

Figure 23. Pounding of adjacent building structures at Gangtok, Sikkim.

The damage survey and analysis presented in this paper leads to the following important lessons
to be learnt by the engineering community.
1. A moderate or low magnitude earthquake, which may occur with a greater frequency, can
cause severe damage to so-called non-engineered or semi-engineered structures which are very
large in numbers and provide shelter to most of the people in the world.
2. The damage survey shows that inadequate corner joints between walls, absence of lintel
band and vertical reinforcements, and asymmetry and irregularity in configurations are some of
the very predominant causes of failure. The vulnerability in hilly regions further increases due
to inclined slopes and inadequate foundations. The measures for strengthening such buildings
are available and highlighted in the present paper.
3. The earthquake-resistant techniques for attributing better protection during seismic actions
for such structures are available but may be needed to be tuned a little bit in the particular con-
text of a locality and socio-economic background. However, due to lack of adequate attention
towards such structures, any stringent design, supervision and construction guidelines are either
not made or not implemented.
4. The lightweight structures with wooden or bamboo frame can be viable alternative to ma-
sonry structures depending on local, social, cultural, economic and other backgrounds of a par-
ticular place as they are found to perform safely. Hence, careful attempt for implementing such
structures as common peoples building should be made.
5. In the context of availability of technical know-how, the major task is to build up adequate
consciousness in the level of municipal framework and individual householders. In fact, the
biggest challenge is to make the common people understand about the vulnerability of the shel-
ter they are leaving, even during frequent low magnitude earthquake and make them aware to
take adequate technical consultancy for design, construction and/or retrofitting of the same. The
technical community and governmental organization may take the help of the public media in an
adequate manner for the same; and should immediately take up this task to avoid recurrence of
such damage frequently.
At the end, the authors feel that such recurrence of damages in moderate to low magnitude
earthquake occur in many places of the world including the recent earthquake in Italy at
LAquila, which has clearly pointed out the general negligence of civilizations in terms of re-
search and its implementation towards the popular traditional masonry and other non-
engineered structures. Scientific and technical community has to shift their focus more towards
169
the need of human civilization than carrying utopian research for satisfying their quest only.
REFERENCES
Bureau of Indian Standards. 1986. Indian Standard 1893: 1984 (reaffirmed 1998), Criteria for earth-
quake resistant design of structures (fourth revision). New Delhi: BIS.
Bureau of Indian Standards. 2002. Indian Standard 1893 (Part 1): 2002, Criteria for earthquake resis-
tant design of structures, Part 1 General provisions and buildings (fifth revision). New Delhi: BIS.
Bureau of Indian Standards. 2002. Indian Standard 4326: 1993 (reaffirmed 1998), Earthquake resistant
design and construction of buildings Code of practice (second revision). New Delhi: BIS.
Bureau of Indian Standards. 2002. Indian Standard 13920: 1993 (reaffirmed 1998), Ductile detailing of
reinforced concrete structures subjected to seismic forces. New Delhi: BIS.
International Association for Earthquake Engineering & National Information Center of Earthquake En-
gineering. 2004. Guidelines for Earthquake Resistant non-engineered construction (English Edition).
Tokyo: IAEE & Kanpur: NICEE.
Murty, C.V.R. 2005a. Why is vertical reinforcement required in masonry buildings? In Earthquake Tips
Learning Earthquake Design and Construction: 29-30. Kanpur: NICEE.
Murty, C.V.R. 2005b. How do columns in RC buildings resist earthquakes? In Earthquake Tips Learn-
ing Earthquake Design and Construction: 37-38. Kanpur: NICEE.
Registrar General & Census Commissioner, India. 2001. Census of India, 2001. New Delhi: Government
of India.
170
Statistical Evaluation of Embedment Effect on Damage to RC
Building Structures during the 1995 Hyogoken-Nanbu Earth-
quake

A. Mikami & Y. Nariyuki
The University of Tokushima, Tokushima, Japan
T. Matsuda
Oriental Consultants Co., Ltd., Tokyo, Japan
1 INTRODUCTION
ABSTRACT: This study statistically evaluates the effect of embedment due to presence of
basements on damage mitigation of reinforced concrete (RC) building structures based on the
building damage database during the 1995 Hyogoken-Nanbu Earthquake. A multivariate analy-
sis (Hayashis quantification II) is used to investigate factors (items) that have a strong effect on
the degree of damage to RC buildings. The results indicate that embedment has a remarkable ef-
fect on reducing damage; degree of which is similar to the construction year effect (due to major
revision of Japanese building standard law in 1981). This finding may support more positive
adoption of embedment effect in seismic design practice.

Variation of earthquake motions from the free-field to the foundation is caused by differences of
stiffness and mass of the foundation from the soil. Soil-structure interaction (SSI) associated
with the difference in stiffness is regarded as kinematic soil-structure interaction, which is con-
veniently represented by the transfer function in the frequency domain as the ratio of massless
foundation motion to free-field surface ground motion.
Transfer functions for an embedded cylinder were theoretically calculated by Elsabee et al.
(Elsabee & Morray, 1977) for soil resting on rock, and by Day (Day, 1977) for halfspace soil.
Their theoretical works indicate that embedment has a marked influence on kinematic soil-
structure interaction.
However, it was not until quite recently that the effect was incorporated into the building de-
sign guideline of the Architectural Institute of Japan (AIJ, 2004). Hence, buildings constructed
before the 1995 Hyogoken-Nanbu Earthquake might have some amount of extra strength, as-
suming that kinematic SSI effect was not taken into account in their design, and this fact may
have resulted in reducing damage to such buildings.
As a consequence, in this study, effect of embedment due to the presence of basement on
damage reduction of building structures is statistically investigated by using a multivariate
analysis based on the damage database of RC public building structures during the 1995
Hyogoken-Nanbu Earthquake.

2 DATABASE
2.1 Database of damage to RC buildings
The AIJ carried out investigation into damage to RC/SRC building structures during the 1995
Hyogoken-Nanbu Earthquake, and the data was compiled into a report published in 1997 (AIJ,
1997). The dataset includes construction year, number of stories (story stands for number of floors
above the ground) and number of basement levels, for both damaged and undamaged buildings.
171
Table 1. Database used for the analysis
No. Damage level Const. year
No. of
Stories
Basement
Levels
PGV(cm/s)
1 4 1992 8 1 120-140
2 2 1980 9 2 -50
3 4 1980 10 1 -50
4 3 1993 10 2 50-80
5 5 1964 7 2 120-140
6 6 1957 8 1 120-140
7 4 1977 8 1 140-
8 3 1993 7 2 140-
9 2 1991 8 1 140-
10 1 1975 3 2 -50
11 5 1975 12 0 140-
12 2 1987 3 1 120-140
13 2 1991 8 0 50-80
14 1 1979 8 1 -50
15 3 1966 9 2 -50
16 1 1988 6 1 -50
17 5 1971 8 3 120-140
18 2 1972 7 2 -50
19 1 1990 4 3 120-140
20 3 1972 4 1 120-140
21 2 1985 3 1 -50
22 2 1992 10 1 -50
23 3 1964 13 2 -50
24 3 1970 3 2 -50
25 3 1958 5 2 -50
26 4 1965 5 1 120-140
27 3 1977 11 0 90-110
28 3 1973 9 2 50-80
29 2 1970 13 2 90-110
30 2 1990 14 4 90-110
31 3 1966 13 2 90-110
32 4 1981 9 3 -50
33 3 1972 8 3 50-80
34 3 1981 5 2 90-110
35 4 1965 6 1 120-140
68 4 1921 2 1 50-80
69 4 1964 3 0 50-80
70 4 1969 5 0 90-110
71 4 1992 2 1 140-
72 4 1987 2 0 140-

This study utilizes the data of public building structures that have an adequate number of both em-
bedded and non-embedded structures to investigate the effect of embedment. As a representative
index of expressing strong ground motion, peak ground velocity (PGV) distribution in the Kobe
area estimated by Hayashi et al. (Hayashi, 1997) was added by the authors to the database. Dis-
carding some data collected from obviously liquefied area (such as Port Island), 72 sets of building
172
Table 2 Damage levels
Damage level Description
Level 1 No damage
Level 2 Slight damage
Level 3 Minor damage
Level 4 Medium damage
Level 5 Serious damage
Level 6 Collapse

5
6
4
3
2
Level 1
Level 2
Level 3
Level 4
Level 5
Level 6

Before 1981
1

After 1982
5 & 6
4
3
2
1
Level 1
Level 2
Level 3
Level 4
Level 5
Level 6

Figure 1. Effect of construction year on damage ratio of buildings
data were extracted from the original database as shown in Table 1. Damage to building structures
is classified into 6 levels (from level 1 to level 6) as shown in Table 2. Missing data from the
original database was gathered by our telephone inquiry.
2.2 Effect of different construction years
Major revision was made to the Japanese Building standard law in 1981, thus, it is recognized
173
that construction year is the key factor that influences the degree of damage to building struc-
tures. Figure 1 compares the ratio of each damage level of buildings constructed before 1981
and after 1982. It is obvious that buildings constructed after 1982 show less degree of damage
level compared with those constructed before 1981.

3 STATISTICAL EVALUATION OF EMBEDMENT EFFECT BY A MULTIVARIATE
ANALYSIS
3.1 Method
Effect of embedment on damage reduction of buildings is examined using a multivariate analy-
sis. The problem to be solved in this study is to ascertain factors (items) that have strong influ-
ence on the degree of damage to buildings. Hayashis Quantification II (Hayashi, 1952) is util-
ized in this study. This method is basically a multivariate discrimination analysis, however, it
can deal with a discrimination problem even when variables are given as qualitative data by in-
corporating 0-1 dummy variables. The method is so well known in Japan that some computer
program packages include Hayashis quantification.
Suppose that there are K outside criteria (response variables), R items (each item has c
R
cate-
gories). To deal with qualitative data, 0-1 dummy variables are introduced.
i
becomes 1 only
when a sample ( -th sample of i-th group) fall into a sub-category of each item (k category of j
item).

=
0
1
) ( jk
i
(1)

To discriminate individuals into several groups, sample scores are computed by the following
linear equation.

= =
=
R
j
c
k
i jk i
j
jk a Y
1 1
) (

(2)
where
jk
is unknown coefficients. As already known in the analysis of variance, total sum of
squared differences between scores on
i
and the grand mean is partitioned into sum of
squared differences between group means and the grand mean (between groups variability), and
sum of squared difference between individual scores and their respective groups means (within-
group variability).
a
Y

= = = = =
+ =
K
i
n
i i
K
i
i i
K
i
n
i
i i
Y Y Y Y n Y Y
1 1
2
1
2
1 1
2
) ( ) ( ) (

(3)

where Y and
i
Y are mean values of total sample scores (grand mean) and samples scores
within subgroups. Correlation ratio is calculated using the following equation as the ratio the
sum of squared differences between group means and the grand mean to total sum of squared
differences between groups means and the grand mean. Unknown coefficients are deter-
mined so that the correlation ratio becomes the maximum.
jk
a

max
) (
) (
1 1
2
1
2
2
= =
=
K
i
n
i
K
i
i i
i
Y Y
Y Y n

(4)

174
Table 3 Item categories
Item Description Category
before 1971 1
1972-1981 2 Construction years
after 1982 3
1 5 1 Stories above
ground 6 or more 2
0 1
B1 2
Basement
levels
B2 or more 3
less than 80 1
PGV(cm/s)
90 or more 2
Level 1 3 1
Damage level
Level 4 6 2

Implementing some mathematical manipulation, the problem attributes to generalized
eigenvalue proble
Table 4 Results
Item Category
Number
of items
Category score Range
Partial Cor-
relation Co-
efficient
before 1971 29 -0.6630
1972-1981 23 0.0725
Construction
years
after 1982 20 0.8779
1.5409 0.3426
1 5 28 -0.0293 Stories above
ground 6 44 0.0186
0.0479 0.0130
0 16 -0.5180
B1 32 -0.4176
Basement
levels
B2 or more 24 0.9021
1.4201 0.3359
less than 80 31 0.3918
PGV(cm/s)
90 or more 41 -0.2963
0.6881 0.1911
Level 1-3 39 0.4629
Damage level
Level 4-6 33 -0.5470
Partial correlation coefficient = 0.2568

m.
3.2 Categories for the Quantification II
The original datasets were re-categorized for the analysis by quantification II as shown in Table
3. Dummy variables were, then, introduced based on the new category to quantify the qualita-
tive data.
3.3 Results
Results calculated by the quantification II are shown in Table 4. Looking at the category scores
of the outside variable (damage level), it is recognized that higher damage level has a lower
category score. Hence, category scores of each item have an inverse relationship (i.e. larger
score contributes to reduce damage). From the values of the range and the partial correlation co-
efficient, it is apparent that items of construction year and presence of adequate basements have
about the same value as also shown in Figure 2. This indicates that these items have an influ-
ence on damage to the same degree. Paying attention to the category scores of each item, we can
understand that buildings that have two or more basement levels show less damage degree in
175
0.0 0.5 1.0 1.5 2.0
PGV
basement levels
stories above ground
construction years
Item Range

Figure 2 Item range
addition to buildings constructed after 1982. Thus, these categories contribute to lessen damage
to building structures.

This study statistically investigated the effect of embedment due to presence of basement on
damage mitigation of reinforced concrete building structures during the 1995 Hyogoken-Nanbu
Earthquake. The results indicate that the effect of embedment has a remarkable influence on re-
ducing damage to buildings when there exists sufficient embedment (two or more basement lev-
els). The degree of reducing damage by embedment is similar to the effect of construction year
due to the major revision of the Japanese building standard law in 1981. Although this result
may support more positive adoption of embedment effect in design practice in Japan, the fol-
lowing two points should be noted: (1) The number of data used in the analysis was only 72
which may not be adequate. (2) The result reflects only a single earthquake event (1995
Hyogoken-Nanbu Earthquake). Therefore, verification of our findings using some other earth-
quakes and their damage data is needed in the future.

ACKNOWLEDGMENTS
The authors would like to express their gratitude to Kansai Geo-informatics Network (KG-NET)
for providing soil information in Kobe area. The authors would also like to thank Mr. Tsutomu
Seo, Ms. Rie Yoshioka and Ms. Sakiko Tsujino for their support to establish the database.

REFERENCES
Architectural Institute of Japan. 2004. Recommendations for Loads on Buildings: Architectural Institute
of Japan (in Japanese).
Architectural Institute of Japan. 1997. Report of damage to reinforced-concrete building structures dur-
ing the 1995 Hyogoken-Nanbu Earthquake: Architectural Institute of Japan (in Japanese).
Day, S.M. 1977. Finite element analysis of seismic scattering problems, Ph.D. Dissertation, University of
California, San Diego.
Elsabee, F. & Morray, J.P. 1977. Dynamic behavior of embedded foundations, MIT report, R77-3.
Hayashi, Y., Miyakoshi, J. & Tamura, K. 1997. Study on the distribution of peak ground velocity based
on building damage during the 1995 Hyogoken-Nanbu Earthquake, J. Struct. Constr. Eng., 502: 61-
68. (in Japanese)
Hayashi, C. 1952. On the prediction of phenomena from qualitative data on the quantification of qualita-
tive data from the mathematico-statistical point of view, Ann. Inst. Statist. Math. 3, 69-98.
176
Global Increase of Natural Disasters and
International Cooperation for Disaster Mitigation

Masanori Hamada
Professor, Dept. of Civil and Environment Engineering, Waseda University, Tokyo, Japan
Xu Wu
Dr. Eng., Technical Center, OYO Corporation, Saitama, Japan
1 INTRODUCTION
ABSTRACT: This paper describes the current situation of global increase of natural disasters
such as earthquake, tsunami, flood and storm. The change of natural environments is increasing
disasters, and the vulnerability of the society proves serious menace to safe and secure life of
mankind. This paper touches upon the basic concept of the policy against natural disasters under
change of natural and social environments. This document also introduces the international co-
operation by Japan Society of Civil Engineering and a non-profit organization, Engineers with-
out Borders, Japan for reduction of natural disasters in the world.

Natural disasters such as earthquake, tsunami, storm and flood have been increasing in the re-
cent years in the world. There are two main factors underlying the increase of natural disasters.
Those are the natural environmental change and the social vulnerability.
The natural environmental change such as global warming, heat island phenomena in mega
cities, the decrease of the forest, desertification and erosion of rivers, are resulting in extremely
heavy rains and snows, huge typhoons and hurricanes, abnormally high temperature, and high
tidal waves. The social environment is also changing and it is becoming fragile against natural
disasters. Those are highly congested urban areas, depopulation of rural areas, human habitation
on disaster-prone lands, lack of cooperation and communication among recent urban societies,
and insufficient infrastructures for the disaster mitigation.
This paper briefly reviews the recent natural disasters in the world and the associated sub-
jects, and describes the basic concept of the policy for the reduction of future natural disasters in
future. Furthermore, the authors introduce internationally cooperative activities for the disaster
mitigation and the restoration of affected societies.
2 NATURAL DESASTERS AND CLIMATE CHANGE IN THE WORLD
2.1 Recent Natural Disasters in the World
During recent few years, the disastrous earthquakes and tsunamis have attacked the Asian coun-
tries. The 2008 Wenchuan earthquake caused an unprecedented damage to the vast and
mountainous areas of the Sichuan Province of China. The 2004 Sumatra earthquake and conse-
quent tsunami killed more than 200,000 people in the areas around the Indian Ocean (JSCE 2005).
In 2005, about 70,000 people were killed in Pakistan (JSCE and AIJ 2005), and in 2006, a
disastrous earthquake attacked the Java Island, Indonesia.

177

N
U
M
B
E
R

O
F

E
V
E
N
T
S
Asia America
Europe Others
7
6
5
4
3
2
1
0
8
7
6
5
4
3
2
1
0
8
1946 51 56 61 66 71 76 81 86 91 96 2001

50 55 60 65 70 75 80 85 90 95 2000 05
Figure 1 Damaging Earthquakes and Tsunamis in
the World in the Last 60 Years (Events
with more than 1000 casualties)

Figure 1 shows the number of earthquakes and tsunamis with more than 1,000 deaths in each
five years period during the last 60 years (1946~2005) in the world. The number of events has
drastically increased in the last two decades.
The number of earthquakes with magnitudes more than 7.0 and 6.0 in the world during the
last 60 years is shown in Figure 2. On the contrary of the increase of the earthquake and tsunami
disasters, the number of occurrences of earthquakes with magnitudes more than 7.0 has been
decreasing during the last 60 years. The number of the earthquakes with more than 6.0 slightly
increased during the last decade, but was not consistent with the rapid increase of the number of
the earthquake and tsunami disasters. This suggests that the reason of the increase of the earth-
quake and tsunami disasters is the increase of the vulnerability of our human societies.

0
20
40
60
80
100
120
140
160
180
0
100
200
300
400
500
600
700
800
900
1946
50
51
55
56
60
61
65
66
70
71
75
76
80
81
85
86
90
91
95
96
2000
01
05
N
u
m
b
e
r

o
f

E
v
e
n
t
s
M
o
r
e

t
h
a
n

M
7
.
0
N
u
m
b
e
r

o
f

E
v
e
n
t
s
M
o
r
e

t
h
a
n

M
6
.
0
More than M7.0

More than M6.0
0
20
40
60
80
100
120
140
160
180
0
20
40
60
80
100
120
140
160
180
0
100
200
300
400
500
600
700
800
900
0
100
200
300
400
500
600
700
800
900
1946
50
51
55
56
60
61
65
66
70
71
75
76
80
81
85
86
90
91
95
96
2000
01
05
1946
50
1946
50
51
55
51
55
56
60
56
60
61
65
61
65
66
70
66
70
71
75
71
75
76
80
76
80
81
85
81
85
86
90
86
90
91
95
91
95
96
2000
96
2000
01
05
01
05
N
u
m
b
e
r

o
f

E
v
e
n
t
s
M
o
r
e

t
h
a
n

M
7
.
0
N
u
m
b
e
r

o
f

E
v
e
n
t
s
M
o
r
e

t
h
a
n

M
6
.
0
More than M7.0

More than M6.0
More than M7.0
More than M6.0

Figure 2 Number of Earthquakes with
Magnitudes More than 7 and
More than 6
Figure 2 Number of Earthquakes with Magnitudes
More Than 7 and 6 in the Last 60 Years.

Storm and flood disasters also have suffered the people in the world in recent years. Most of
the storm and flood disasters were concentrated in the Asian region and Central America. In
2008, more than one hundred and thirty thousand people were killed by a huge cyclone in
Myanmar, and another large cyclone caused an extensive damage to Bangladesh in 2007. Hurri-
cane Katrina attacked Louisiana in United States and killed more than 5,000 people around the
Caribbean Sea in 2005.
Figure 3 shows the number of storm and flood disasters with more than one thousands deaths
in each five years period during the last 60 years (1946~2005). The storm and flood disasters in
the world have also increased during the last two decades, and the disasters in Asia are domi-
nant.

178

N
u
m
b
e
r

o
f

E
v
e
n
t
s
1946 51 56 61 66 71 76 81 86 91 96 2001

50 55 60 65 70 75 80 85 90 95 2000 05
14
12
10
8
6
4
2
0
14
12
10
8
6
4
2
0
Asia America
Europe Others

Figure 4 Damaging Storms and Floods in the
World during the Last Half Century
(E

vents with more than 1000 casualties)
Figure 3 Number of Storm and Flood Disasters in the
World during the Last 60 Years (Events with more than
1000 deaths)

There may be two main reasons for the increase of the flood and storm disasters. One is the
increase of the vulnerability of the human societies against disasters, which is particularly re-
markable in the Asian countries. Another reason may be the global climate change. Figure 4 il-
lustrates the change of air temperature in the world during the last century. The air temperature
has drastically risen in the northern hemisphere, and this trend is predicted to continue during
the 21 century. The seawater temperature is also rising due to the rise of air temperature. The
rise of seawater temperature can be considered to be one of causes of the occurrence of huge ty-
phoons and hurricanes, and high tidal waves.

-2.5 -1.5 -0.5 0.5 1.5 2.5 -2.5 -1.5 -0.5 0.5 1.5 2.5 -2.5 -1.5 -0.5 0.5 1.5 2.5

Figure 4 Change of Air Temperature during Last Cen-
tury in the World (JMA 2005).

NumberofEvents:59 NumberofDeath:1,027,011
Asia
71
42
America
20
11
Europe
5
3
Others 5
3
Asia
89
915,453
America
6
62,392
4 42,300 Europe
Others
1
6,866

Figure 5 Regional Ratio of Natural Disasters and
Death (1986~2007)

179
Figure 5 shows the regional ratio of the number of natural disasters and deaths in the world
(1986~2007). Including earthquake, tsunami, storm and flood, the natural disasters which killed
more than 1000 people occurred 59 times during last two decades, and 42 events were in the
Asian region. About one million people lost their lives due to the natural disasters since 1986,
and the 89% of them, about 9 hundred thousands were in the Asian region. This statistics shows
that the natural disaster mitigation in the Asian region is one of urgent and major subjects to
create safe and secure world.
2.2 Recent Natural Disasters in Japan
The epicenters of main earthquake disasters occurred in Japan after the 1995 Hygoken Nanbu
(Kobe) earthquake are shown in Figure 6. The total 6,547 lives were lost, including the Kobe
earthquake.

2003.9.26 Tokachi-oki EQ
(Death: 1, Missing: 1, Injured: 849
2008.7.24 Miyagiken Enan-hokubuEQ
(Death:1, Injured: 211
2005.3.20 Fukuokaken Seiho-oki EQ
Death:1, Injured:1087
1995.1.17 Hygoken NanbuEQ
Death: 6,434, Missing: 3, Injured: 43,792
2001.3.24 Geiyo EQ
Death: 2, Injured: 288
1997.5.13 KagoshimakenSatsuma-chihoEQ
Injured: 74
2008.6.14 Iwate-Miyagi Nairiku EQ
(Death: 13, Missing: 10, Injured: 451
2004.10.23 Niigata Chuetsu EQ
(Death: 65, Injured: 4805)
2007.7.16 Niigata Chuetsu-oki EQ
(Death:15, Injured:2345 )
2007.3.25 Noto Hanto EQ
(Death: 1, Injured: 356)

Figure 6 Recent Earthquake Disasters in Japan
(1995~2008, the Number of Death and Missing: 6,547)

A large number of infrastructures was destroyed.
(a) The 1995 Hygoken Nanbu earthquake
Shinkansen was derailed (JSCE 2004).
(b) The 2004 Niigata Chuetsu earthquake

A huge landslide was triggered.
(d) The 2008 Iwate-Miyagi Nairiku earthquake
A nuclear power plant was fired.
(c) The 2007 Niigata Chuetsu-oki earthquake

Figure 7 Typical Disasters in the Recent Earthquakes in Japan.

Some typical disasters occurred in the resent earthquakes in Japan are shown in Figure 7. We
have learned new lessons whenever earthquakes occurred.
From the 1995 Kobe earthquake, we learned lack of earthquake resistance of concrete
bridges, subway structures and buildings against the strong earthquake ground motions in the
near field of the earthquake fault. A large man-made island reclaimed from the sea extensively
was liquefied resulting in severe damage to structures such as storage tanks, and quay walls.
The 2004 Niigata Chuetsu earthquake taught us new lessons, one of which was the derailment
of the high speed train, Shinkansen. This event raised serious discussions on the safety of the
high speed train against earthquakes in the vicinity of railway line. Another lesson from this
180
earthquake was an overlap of different natural disasters. Three days before the earthquake a
huge typhoon attacked the same area with heavy rains. The slopes in the mountainous area and
embankments were saturated. The earthquake ground motion caused huge landslides and em-
bankment failures.
The 2007 Niigata Chuetsu-oki earthquake severely damaged a nuclear power station. This
was the first occurrence of earthquake damage to nuclear facility in the world. The 2008 Iwate-
Miyagi Nairiku earthquake hit the mountainous areas in the northern part of Japan. A huge
landslide was triggered and resulted in serious damages to infrastructures.
Besides the earthquake disasters the flood and storm disasters also frequently occur in Japan.
Flood and storm disasters with casualties meaningfully occurred twice a year, and about 500
lives were lost during the last decade. The recent ones are shown in Figure 8.
In 2006, a huge mudflow, which was caused by a heavy rain, attacked residential area in Na-
gano Prefecture, and killed 12 people. In 2004, Itsukushima shrine, one of national treasures
was severely damaged by high tidal wave and strong wind by a large-scale typhoon. In 2005, a
downpour with 100 mm rain per hour flooded a wide area of the downtown of Tokyo. This was
considered to be caused by the heat island phenomena in highly urbanized mega cities.

The number of death
and missing12
The number of deat

h and missing45
The number of death and missing31
(a) 2006 Mud Flow by
Heavy Rain (Nagano)
(b) 2004 Itsuku
which was
Typhoon N
(c) 1999 Tidal Wave by
Typhoon No.18
(Kumamoto)
(d) 2005 Fl
Rain (T

Figure 10 Recent Flood and Storm Disaster in Japan
shima-shrine
Destroyed by
o.18 (Hiroshima)
ood by Heavy
okyo)
Figure 8 Recent Flood and Storm Disasters in Japan.

3 BASIC STRATEGY OF THE POLICY AGAINST FUTURE DISASTERS UNDER THE
CHANGE OF NATURAL AND SOCIAL ENBIRONMENTS
In the following the authors introduce basic strategy of the policy against future disasters under
the change of natural and social environments.
As shown in Figure 9 the change of natural environments such as global warning, heat Island
phenomena in urbanized area, deforestation, desertification and erosion of river and seashore is
increasing natural disasters. Those are extremely heavy rains and snows, huge typhoons, hurri-
canes and cyclones, drought, abnormally high temperature and high tidal waves.
In addition to the change of natural environments, our social environments are also changing,
becoming fragile against natural disasters. Those are too congested urban areas, depopulation of
rural area, human habitations on fragile ground, lack of cooperation and communication among
the recent urban societies, budget deficit of central and rural governments, and finally poverty.
The poverty is the most important factor for the increase of the natural disasters in the Asian de-
181
veloping countries. The poverty is expanding the disaster, and the disaster is worsening the pov-
erty.
The characteristics of natural disaster are changing due to the change of natural and social
environments. The key point for the measures against future disasters is how to prepare unex-
pected natural phenomena and against external forces largely exceeding the design level, in an-
other word, how to prepare against natural disasters with a huge scale, but comparatively low
probability of occurrence.

ChangeofNatural Environment
GlobalWarming
HeatIslandofUrbanizedArea
Deforestation
Desertification
ErosionbyRiverandSeashore
ChangeofSocialEnvironment
HeavyRainandSnow
HugeTyphoon,Hurricaneand
Cyclone
Drought
HighTemperature
HighWaveandStormSurge
ChangeofCharacteristics
ofNaturalDisasters
Measuresagainstunexpected
naturalphenomenaandexternal
forceslargelyexceedingdesignlevel
Implementationofbothsoftware
andhardwaremeasures
CongestionofUrbanArea
DepopulationofRuralArea
HumanHabitationonFragileGround
LackofCooperationand
CommunicationamongSocieties
BudgetDeficitofCentralandRural
Governments
Poverty andLowGovernance
1

Figure 9 Change of Natural and Social Environments
and Increase of Natural Disasters.

Figure 10 illustrates a basic strategy for the measures against huge natural disasters with low
probability of occurrence and moderate-sized disasters with medium probability. That is a com-
bination of hardware measures and software measures. Hardware measures mean, for an exam-
ple, reinforcement of dikes against floods, and the soft ware measures are evacuation systems
during flood and the education of people. Moderate-sized disasters with medium probability are
prevented mainly by hardware measures. However, against huge disasters with low probability,
the disasters are reduced by both hardware and software measures.
The problem is how to determine the rational level of the investment for disaster mitigation.
One of methods to judge the rational level of the investment is the comparison of the risk with
the cost of the measures. The risk is estimated as the product of the total loss with the probabil-
ity. For the estimation of the total loss, we have to take into consideration, various factors, not
only loss of human lives and properties, but also probable national power decline resulting from
the disaster, ruining of national landscape and furthermore, psychological damage to the people.
And the consensus among the people is essential to determine the rational level of disaster miti-
gation measures.
Figure 11 shows the disaster management cycle. The disaster management should cover these
three stages of management cycle.
NaturalDisasterMitigationbyCombination
ofHardwareandSoftwareMeasures
H
a
r
d
w
a
r
e

M
e
a
s
u
r
e
s
E
x
t
e
r
n
a
l

F
o
r
c
e
s
ModeratesDisasterswith
MediumProbability
HugeDisasterswith
LowProbability

Reinforcement
of Hardware
Measures
Software
Measures
2
Figure 10 Basic Strategy of the Policy against Future
Disasters under the Changes of Natural and Social Envi-
ronments.
182
First stage is the preparedness before disaster, which involves planning of national land-use,
development of social systems, reinforcement of residential houses, building and infrastructures,
stock of goods and materials and education of people. The second stage is emergency response
immediately after disasters such as rescue operation, fire fighting, care of refugees and medical
treatment. The third stage is re-development of damaged areas, reconstruction of infrastructures,
house and buildings and industrial facilities, transfer of experience of disaster.
The disaster management on these three stages is carried out by various organizations such as
the central government, prefectural and municipal governments, non-government organizations
regional communities, private companies and residents.

Disaster
National Land-
use Plan
Development of
Social Systems
Reinforcement
of Structures and
Facilities
Stock of Goods
and Materials
Education
Rescue Operation
Fire Fighting
Relief of Refugees
Medical Treatment
Redevelopment of the Damaged Areas
Reconstruction of infrastructures,
Houses and Buildings, and Industrial
Facilities
Transfer of Disaster Experiences
P
r
e
p
a
r
e
d
n
e
s
s
b
e
f
o
r
e
D
i
s
a
s
t
e
r
s
E
m
e
r
g
e
n
c
y
R
e
s
p
o
n
s
e
RestorationandReconstruction

Figure 11 Cycle of Disaster Management (Cabinet Of-
fice, Government of Japan 2006).
4 INTERNATIONAL COOPERATION FOR NATURAL DISASTER MITIGATION BY
JAPAN SOCIETY OF CIVIL ENGINEERS AND ENGINEERS WITHOUT BORDERS,
JAPAN

As mentioned previously, most of the natural disasters concentrated in the Asian region in the
recent years. Therefore, reduction of natural disasters in this area is one of our primary subjects.
The Japan Society of Civil Engineers (JSCE), which was established in 1914 with a mission that
it should contribute to the advancement of scientific culture and the creation of a safe and secure
society by developing technology and knowledge in civil engineering field. Therefore, the
members of JSCE have been continuously promoting the transfer of technologies and experi-
ences for natural disaster mitigation, technical assistance for restoration works for the damaged
infrastructures and societies, and education and training of people for future disaster mitigation
in the world, particularly in the Asian countries.
The Engineering without Borders, Japan (EWBJ), a non-profit organization was organized by
members of JSCE and the Architectural Institute of Japan (AIJ) with purposes to promote inter-
national activities for natural disaster mitigation, with supports governmental institutions and
construction industries of Japan. JSCE and EWBJ have promoted various kind activities for
natural disaster mitigation in the Asian region, which are introduced below.

4.1 Technical Assistances for Restoration and Reconstruction
After the 2004 and 2005 Sumatra earthquakes, JSCE and EWBJ jointly dispatched their teams
to the counties around the Indian Ocean for technical assistances for restoration of the affected
areas and reconstruction of damaged infrastructures. In northern Sumatra, they proposed a strat-
egy for reconstruction of the road along the west coast of Sumatra, which was washed away and
many bridges were destroyed by the tsunami.
183

Steeltrussbridgeswerewashedaway 2004 Sumatra Tsunami 2005 Sumatra Earthquake

Seminar and Workshop after the 2008 Wenchuan Earthquake

Figure 12 Technical Assistances for Restoration and
Reconstruction by JSCE and EWBJ.

In Nias island, where located in the off coast of the west of Sumatra was attacked by the sec-
ond earthquake in 2005. The dispatched members by JSCE and EWBJ proposed a plan of re-
construction of bridges and buildings damaged by soil liquefaction.
After the 2008 Wenchuan earthquake, JSCE and its associated other engineering and science
societies of Japan dispatched their members for the discussions on the restoration methods of
damaged buildings, bridges, tunnels and slopes with the Chinese counterparts. They introduced
some examples of restoration of structures damaged by Japanese earthquakes including the 1995
Kobe earthquake.

4.2 Educational Activities
Student members of JSCE and EWBJ have continued educational program for pupils of primary
and middle schools in Indonesia after the tsunami tragedy. They are teaching children the
mechanism of the occurrence of earthquake and tsunami, and how to save lives from future nat-
ural disasters.
Members of JSCE constructed one hundred monuments in Banda Ache, the tops of which
show the height of the tsunami at each location. The epitaph of the monuments describes the
tragedy and the mourning for the dead. These monuments have been used for the education of
the children.
In Chengdu, Sichuan of China, JSCE and its associated other engineering and science socie-
ties started a series of lectures on seismology and earthquake engineering for the students and
young engineers. They are lecturing the system of the earthquake prediction, estimation of
ground motion, earthquake resistant design of structures, and strategies of mid-and long term
restoration of the affected areas

Series of Lecture on Seismology and Earthquake Engineering
Education Program in Indonesia Monuments of Tsunami

Figure 13 Educational Actives by JSCE and EWBJ.
184
4.3 Transfer of Technology
Figure 14 shows the tsunami warning system which was proposed by JSCE and EWBJ members
to respond the request from the North Sumatra provincial government.
EWBJ members are also installing seismographs in Padang area of West Sumatra to develop
the earthquake and tsunami warning system, and to survey the characteristics of the earthquake
ground motion in the areas, under cooperation with the researchers in Indonesia.
In Bangladesh EWBJ made a manual for the construction of shelters against future cyclones,
and has been training regional engineers by using this manual under financial aid of Ministry of
Foreign Affairs, Japan. It is planned that 10 thousand shelters will be constructed in the coastal
areas of Bangladesh.

Satellit e
Seismic st at ion
Cent ral disaster
miti gati on center
Broadcast
center
Televi sion Radio
Sat ell ite
Mosque

Figure 14 Proposal of Regional Tsunami Warning Sys-
tem for North Sumatra Provincial Government.
5 CONCLUSIONS
In recent years, global societies especially in the Asian region have experienced many nature
disasters such as earthquake, tsunami, flood and storm. These nature disasters produced great
human suffering and economic loss.
Due to the natural environmental change and the social vulnerability, the increasing risk asso-
ciated with such disasters seems continue to grow, unless effective measures are developed and
implemented to mitigate the disasters.
Over the last few decades Japan has gathered many invaluable experiences and knowledge in
both the technological and societal fields which can be shared in the natural disasters mitigation.
REFERENCES
[1] Japan Society of Civil Engineer (2005), The Damage Induced by Sumatra Earthquake and Associ-
ated Tsunami of December 26, 2004.
[2] Japan Society of Civil Engineers and Architectural Institute of Japan (2005), A quick Report on
Kashmir Earthquake.
[3] Japan Meteorological Agency (2005), A Report on Abnormal Weather 2005.
[4] Japan Society of Civil Engineers (2004), General Investigation about the Damage of Social Infra-
structure Systems due to the 2004 Niigata Earthquake / Result of Investigation and Urgent Proposals.
[5] Cabinet Office, Government of Japan (2006), White Paper on Disaster Prevention 2006.

185
Tsunami Damage Studies and Construction of the Memorial
Poles in Banda Aceh

Hirokazu Iemura
Kinki Politchnic College, Kishiwada, Osaka,Japan
Mulyo Harris Pradono
Agency for the Assessment and Application of Technology, Jakarta, Indonesia
Megumi Sugimoto
Graduate School of Global Environmental Studies, Kyoto University, Kyoto, Japan

ABSTRACT: During a survey led by the first author in Banda Aceh in Indonesia and surround-
ing area after the great Sumatra Earthquake, questionnaires have been distributed to the people
affected by the earthquake and tsunami. One important result of the questionnaires shows that
even if people had started running away just after the big earthquake, the percentage of ex-
pected survivors would have been less s than 100%. The practical implication from this result is
that education, socialization, escape structures, warning systems, and wave resisting structures
are among important factors for people to be safer against future tsunami attacks. Moreover, for
learning and educational purposes, the memorial poles showing the height of the tsunami attack
were constructed.
1 INTRODUCTION
Huge earthquake and tsunami on December 26th, 2004 has caused a great number of casualties
around Indian Ocean rim countries especially in Banda Aceh city and Aceh Besar Sub-province,
in Sumatra Island, Indonesia. Lessons from this huge disaster shall be learnt by locals and peo-
ple all around the world.
In a quick response to the disaster, a Japanese group of researchers led by the first author de-
parted to Banda Aceh and surrounding areas in attempt to study the lessons by the huge earth-
quake and tsunami. The authors also conducted tsunami questionnaires during their first visit to
Banda Aceh and surroundings area (Iemura, et al, 2006).
2 IMPORTANT RESULTS FROM THE QUESTIONNAIRES
The important results from the questionnaires are that the local peoples knowledge on tsunami
was very low. The question was Did you know that tsunami would come after a big earth-
quake? Figure 1 shows the result.

Figure 1. Respondents Knowledge on Tsunami

Kne w if tsunami would c ome ?
No
97%
Yes
3%
186
Other important result is that, according to the respondents, even if people had started running
away just after the big earthquake, the percentage of expected survivors would have been less
than 100% (Figure 2). The practical implication is that education, socialization, escape routes,
escape structures, warning system, wave resisting structures are among important factors for
people to be safer against future earthquake and tsunami attacks.

Figure 2. Percentage of Survivors in Banda Aceh City (Number in Parenthesis Shows Percentage of Sur-
vivors if People Would Have Escaped Immediately after the Earthquake)
3 TSUNAMI HEIGHT MEMORIAL POLES AND DISASTER EDUCATION
Since tsunami is a rare event, the future generations may forget the disaster easily. People need
encouraging and reminding words that would be written on lasting memorial structures. One of
the ideas prompted by the first author to local people is to build poles with the height of tsunami
run up throughout the affected areas in the city.
The poles have many important purposes: (1) to encourage people to be prepared for the next
one, (2) to keep the memory of tsunami attack, (3) to educate next generation the important les-
sons from the tsunami, (4) to mourn the passed away people and to restore and reconstruct
Banda Aceh from the disaster, (5) to keep accurate data of tsunami height for future planning,
(6) to be escaping sign with the tsunami height, (7) to encourage local people to live with hope
and ease under tsunami risk, and (8) to be a symbol of Banda Aceh as the tsunami-attacked city.
The idea was supported by local people and the Embassy of Japan in the Republic of Indone-
sia. The Embassy made available a grant up to US $ 91,411 (US $ Ninety one thousand, four
hundred-eleven) by March 31, 2006. It contributes to the execution of the construction by the
Yayasan Umi Abasiah. The project entitled The Project for Supporting Education of Tsunami
Disaster Prevention in Nanggroe Aceh Darussalam.
The first author also provides technical assistance. The pole structure should be strong and
yet reasonably economical and feasible in the material availability point of view. Hybrid struc-
ture consisting of masonry and reinforced concrete was finally adopted as the material. The first
drawing for designing the poles is shown in Figure 3.

187

Figure 3 First Drawing of the Pole for Design Purpose

Figure 3. First Drawing of the Pole for Design Purpose

The poles would be built along the evacuation routes. The purpose is for providing evacuees
with escaping sign with height, so that the evacuees escape toward the lower poles. After con-
sidering many aspects, including limitation of the land for construction, the realized location of
the poles is as shown in Figure 4. The height of the poles is the inundation of tsunami water,
when the tsunami wave was relatively flat and around ten minutes later, the tsunami water re-
turned back to the sea.

Figure 4. Locations of 85 Tsunami Height Memorial Poles in Banda Aceh and Aceh Besar

3
1
17
7
50
51
52
63 61
59
46
13
14
12
11
15
5
4
2
10
68
8
9
18
19
21
44
23
25
1.5 km
North
T TS SU UN NA AM MI I H HE EI IG GH HT T M ME EM MO OR RI IA AL L P PO OL LE ES S
B Ba an nd da a A Ac ce eh h a an nd d A Ac ce eh h B Be es sa ar r, , I In nd do on ne es si i a a
Malaka Strait

Yayasan Umi Abasiah
Banda Aceh,
Indonesia
in collaboration with
Embassy of Japan
in Indonesia
and
Kyoto University
Japan
B BA AN ND DA A A AC CE EH H C CI IT TY Y
41
48
47
49
54
56
57
58
60
45
40
62
16 28
29
6
31
30
37
65
27
26
20
66
22
24
39

Indian
Ocean Jakarta
Singapore
Malaysia
Sumatera
Indonesia
Epicenter, 26 Dec. 2004
Simelue Is.
Nias Is.
Padang
Medan
Banda Aceh
A AC CE EH H B BE ES SA AR R
D DI IS ST TR RI IC CT T
B BA AN ND DA A
A AC CE EH H C CI IT TY Y
8 85 5
8 84 4
8 83 3
8 81 1
8 82 2
8 80 0
7 79 9
7 78 8
7 77 7
7 76 6
7 75 5
7 73 3
7 72 2
7 71 1
6 69 9
7 70 0
6 68 8
7 74 4
32
36
35
33
3
4
67 38
42
43
53 55
64
188

Figure 5. The Firstly
Built Tsunami Pole
(February 2006)

Pole No.: 01,
Height: 3.15 m, Dis-
tance from Shore:
2.80 km, Location:
Gampong Lamdin-
gin, Kecamatan
Kuta Alam, Banda
Aceh.
The placard in front of
the pole shows the in-
undation height, dis-
tance from shore, time
of tsunami arrival,
time of earthquake, lo-
cation name, words of
wisdom, developer,
and donor
Figure 6. The Lowest Tsu-
nami Pole
Pole No.: 28, Height: 0.90 m,
Distance from Shore: 4.30 km,
Location: Dinas Pertanian dan
Tanaman Pangan, Banda Aceh.

Figure 7. Tsunami Pole at a School

Pole No.: 15, Height: 1.52 m, Distance from Shore:
3.90 km, Location: SD Negeri 28, Kp. Keuramat,
Banda Aceh

Figure 8. The Tallest Tsunami Pole
Pole No.: 69, Height: 9.00 m (w),
Distance from Shore: 0.50 km, Lo-
cation: Masjid Lam Tengoh, Peukan
Bada, Banda Aceh.
The wave height was justified from
witnesses, because there was no in-
undation mark left by the tsunami.

Figure 9. Tsunami Pole at a School

Pole No.: 49, Height: 3.80 m, Distance from Shore: 3.10 km,
Location: SDN 2, Punge Jurong, Banda Aceh

189
Table 1. List of 85 Tsunami Height Memorial Poles in Banda Aceh and Aceh Besar
1. 3.15 m 2.80 km GPG.
LAMDINGIN KEC.
KUTA ALAM
23. 1.75 m 3.50 km Asrama
Mahasiswa Unsyiah,
Darussalam
45. 2.30 m 3.20 km
Kantor Pengadilan
Negeri Banda Aceh
67. 1.30 m 3.30 km RUMAH
BP. ALAMSYAH UMAR
JALAN SYIAH KUALA
2. 2.90 m 3.00 km GPG.
LAMBARO SKEP
KEC. KUTA ALAM
24. 1.45 m 3.60 km Kantor
Rektorat Iain Ar-Raniry,
Darussalam
46. 2.00 m 3.50 km
Taman Kanak2
YKA Taman Sari
68. 5.70 m 1.50 km MASJID
BAITUL MAGHFIRAH
PEUKAN BADA
3. 1.40 m 3.75 km MAN 1
25. 2.00 m 3.40 km MAN 3,
RUKOH
47. 3.40 m 3.20 km
SMA NEGERI 1
BLANG PADANG
69. 9.00 m (w) 0.50 km
MASJID LAM TENGOH
PEUKAN BADA
4. 1.95 m 3.40 km SMP
NEGERI 2
26. 1.00 m 4.00 km SMP
NEGERI 18
48. 1.80 m 3.40 km
SMPN 17 BLANG
PADANG
70. 7.00 m 0.40 km MASJID
INDRA PURWA
LAMGURON
5. 1.39 m 3.55 km SD
NEGERI 25
27. 1.80 m 3.85 km SMP
NEGERI 6
49. 3.80 m 3.10 km SDN
2 PUNGE JURONG
71. 2.50 m 2.50 km
MEUNASAH TANJONG
KEC.LHOKNGA
6. 1.84 m 3.70 km Masjid
Agung Al Makmur
Bandar Baru
28. 0.90 m 4.30 km DINAS
PERTANIAN
TANAMAN PANGAN
50. 4.50 m 2.00 km
Rumah Ny. Zulkifli
Narukaya Blang Oi
72. 5.50 m 2.50 km SDN
KAJHU
KEC.BAITUSSALAM
7. 2.60 m 3.60 km SMK
NEGERI 2/STM
NEGERI
29. 1.60 m 3.80 km SMK
NEGERI 3
51. 5.80 m 1.90 km
Masjid Syech Abdul
Rauf Blang Oi
73. 7.00 m 2.20 km Tanah
Widari Dawam Dawood
Sp, Cot Paya
8. 2.60 m 3.40 km SD
NEGERI 80, PRADA
30. 2.45 m 3.50 km Kantor
Dekranas Taman Ratu
Safiatuddin
52. 7.00 m 0.50 km
Rumah Bp. Bachtiar
Zakaria Deah Baro
74. 3.50 m 2.70 km SMAN-1
BAITUSSALAM
9. 2.60 m 3.50 km Kanwil
Kehakiman Dan
HAM
31. 2.65 m 3.40 km KAN-
TOR BAWASDA
53. 3.90 m 1.80 km
SMPN 11
LAMJABAT
75. 5.10 m 1.50 km MASJID
LAMBADA LHOK
BAITUSSALAM
10. 1.55 m 4.10 km
DINAS KOPERASI
DAN UKM
32. 1.00 m 3.85 km
Direktorat Politeknik
Kesehatan
54. 3.70 m 2.10 km SDN
95 GAMPONG
BARO
76. 4.60 m 2.00 km
GAMPONG LABUI
BAITUSSALAM
11. 3.20 m 3.50 km SMA
NEGERI 2
33. 1.80 m 3.75 km
Politekkes NAD Jurusan
Keperawatan
55. 2.20 m 2.90 km Mas-
jid Baitul Muqarra-
bin Punge Blang Cut
77. 4.00 m 1.50 km
GAMPONG LAMNGA
JALAN KRUENG RAYA
12. 2.70 m 3.70 km SD
NEGERI 20, POCUT
BAREN
34. 2.00 m 3.40 km SD 35
LAMPRIET
56. 2.20 m 2.90 km SDN
18 PUNGE BLANG
CUT
78. 3.40 m 1.30 km
MEUNASAH NEUHEUN
MASJID RAYA
13. 3.50 m 3.60 km SMP
NEGERI 9,
PEUNAYONG
35. 1.80 m 3.35 km MASJID
JAMIK SILANG
RUKOH
57. 3.40 m 2.70 km
Universitas Iskandar
Muda, Surien
79. 2.20 m 0.40 km
MEUNASAH DURONG
MASJID RAYA
14. 2.57 m 3.90 km
KANTOR BKPMD
36. 1.20 m 3.80 km SMP
NEGERI 8
DARUSSALAM
58. 2.30 m 3.00 km SDN
97 LAMTEUMEN
TIMUR
80. 3.30 m 1.00 km Meunasah
Paya Kameng Masjid Raya
15. 1.52 m 3.90 km SD
NEGERI 28, KP.
KEURAMAT
37. 3.40 m 3.40 km SDN 27
GAMPONG MULIA
59. 2.00 m 3.10 km SDN
93 LAMTEUMEN
TIMUR
81. 3.40 m 0.50 km MASJID
KRUENG RAYA
16. 0.89 m 3.90 km
DARUL ULUM,
YPUI
38. 3.50 m 3.35 km MIN
Merduati Jalan
Malahayati, GP.MULIA
60. 1.40 m 3.30 km
PGSD FKIP
Unsyiah Goheng
82. 3.20 m 0.80 km
GAMPONG MEUNASAH
KULAM
17. 1.52 m 3.70 km SD
KARTIKA XIX-I,
LAMPRIET
39. 4.60 m 3.00 km MASJID
AL MUKARRAMAH
GP. MULIA
61. 2.00 m 3.30 km MIN
TELADAN
LAMTEUMEN
83. 3.20 m 0.80 km
GAMPONG MEUNASAH
MON
18. 1.91 m 3.40 km Badan
Perpustakaan
Wilayah
40. 4.50 m 2.50 km Masjid
Tgk. Dianjong
PEULANGGAHAN
62. 1.00 m 3.50 km
Rumah Zakaria
Ismail Lamteumen
84. 2.50 m 0.30 km MASJID
LAMREH KRUENG
RAYA
19. 1.80 m 3.80 km 41. 7.00 m 1.80 km SMPN 63. 2.35 m 3.30 km Biro 85. 3.10 m 0.50 km
190
MASJID
LAMGUGOB
12 GAMPONG JAWA Logistik Polda NAD
Lamteumen
Pasantren/Dayah Al
Mahfuzhah Krueng Raya
20. 1.40 m 3.80 km MIS
LAMGUGOB
42. 6.00 m 2.00 km SDN 6
KEUDAH
64. 3.80 m 2.30 km
SMPN 15 LAMPOH
DAYA
21. 1.00 m 3.90 km
STIES/AMBA
43. 8.00 m (w) 1.80 km SDN
8/38 MERDUATI
65. 3.70 m 2.70 km SDN
61 JEULINGKE
22. 1.30 m 3.70 km SD
NEGERI 69,
DARUSSALAM
44. 2.70 m 2.70 km SD
MUHAMMADIYAH
LAMPASEH
66. 3.20 m 2.70 km SDN
106 RUKOH
Notes:
- Data in [m] is height and
data in [km] is distance from
shore
- Poles 43 & 69 show wave
height. Others show inundation
height
- red in Banda Aceh, blue in
Aceh Besar

Figure 10. Tsunami Height Memorial Poles as part of Education Tools on Hazard Mitigation

In May 2007, all of the 85 poles were constructed throughout Banda Aceh city and surround-
ing areas. Some photos are shown in Figures 6 to 10 and the list of the poles is shown in Table
1.
The tsunami height memorial poles are also education tools for hazard preparedness. One ex-
ample is by showing the students the record of the data mentioned in the poles and why they
were built. The on going survey now is to map the three dimension coordinate of the poles. This
data is invaluable for future tsunami disaster mitigation.
The tsunami height memorial poles will not only provide hazard data dissemination but also
education for the present and next generations. The poles are expected to last long and convey
the messages to every generation.
4 RECOMMENDATION FOR THE RECONSTRUCTION
Other than the tsunami poles and disaster education above (Not forget but understand) above,
institute or center for earthquake and tsunami research, tsunami and earthquake museum
(monuments, facts, data, education materials, and so on), international collaborations among re-
search institution, and tsunami and earthquake safe structural design (technologies and codes),
shall be realized.
191
5 CONCLUSION
1) From the questionnaires about the tsunami damage in Banda Aceh, it was found that local
people did not know the tsunami behavior, even by words.
2) Even if people had started to escape just after the earthquake ground motion, some people
could not have survived because of long evacuation distance.
3) The tsunami memorial poles were constructed with many purposes for future safety of
Banda Aceh city, with financial aid from Japanese people and understanding and support of
local people.
ACKNOWLEDGEMENT
The authors would like to express their sincere gratitude to the Embassy of Japan in the Re-
public of Indonesia, to the Japan Society for the Promotion of Science, to the Ministry of Edu-
cation, Culture, Sports, Science, and Technology, and to the people of Nanggroe Aceh Darussa-
lam, Indonesia, for their strong interest and support.
REFERENCES
Iemura, H., Takahashi, Y., Pradono, M. H., Sukamdo, P., and Kurniawan, R. (2006) Earthquake and
Tsunami Questionnaires in Banda Aceh and Surrounding Areas, Disaster Prevention and Manage-
ment, Emerald, Vol. 15, No. 1, pp. 21-30.
Agusssalim bin Husen, Thantawi Jauhari, Hirokazu Iemura, Muryo Harris Pradono, Megumi Sugimoto
and Takashi Furuwatari (2007) Tsunami Height Memorial Poles in Banda Aceh for disaster Mitiga-
tion and Education Proc. APRU/AEARU Research Symposium, Jakarta, Indonesia, June, 2007.

192
The Algerian Experience of Covering Earthquake Damages and
Applied Techniques of Reinforcement

Moulay Ali
President General Manager CTC OUEST
Civil Engineer, DEA INSA LYON, PGD IISEE-BRI TOKYO

ABSTRACT: Algeria is located on the northern edge of the African plate, which is converging
with the European plate, with a shortening rate of about 48 mm/yr. Northern Algeria is a
highly seismic area, as evidenced by the historical (13651992) seismicity. During the last two
decades, northern Algeria experienced several destructive moderate-to-strong earth-quakes.
Since 1980, El Asnam earthquake (M
s
7.3), which claimed over 2700 lives and destroyed about
60,000 housings, many moderate, but destructive, earthquakes occurred, such as the Constantine
October 27, 1985 (M
s
5.7), Chenoua October 29, 1989 (M
s
6.0), Mascara August18, 1994
(M
s
5.6), Algiers September 4, 1996 (M
s
5.6), Ain Temouchent December 22, 1999 (M
s
5.6), and
Beni Ourtilane November 10, 2000 (M
s
5.5) earthquakes. On May 21, 2003, the Algiers
Boumerdes region was struck by a magnitude 6.8 (M
w
) earthquake, which caused considerable
damages and claimed over 2300 lives.
The paradox is that this region was trough all ages a convergence point of migration of
populations and a Carrefour of cultural exchange and commercial established currents
especially along the Mediterranean coast.
The historical heritage of many types of ancient constructions and significant cultural
monuments and archaeological sites is remaining as a witness of the melting common history of
generations of populations.
This heritage is in danger. Its exposure to strong earthquake events beside the weight of time
and the carelessness of users are worsening the situation, the assessment is traduced in
catastrophic results. The reserved potential for restoration of existing patrimony is remaining
weak and insignificant. Many cases of collapse occurred and many valorous monuments and
sites did extinguish or disappeared embedded under inappropriate extension works and
anarchical use.
Because of a rather high growth of the population, important developing construction
programs are engaged mainly in this northern part of the country where the climate is mild and
the whole part of populations is concentrated. Huge construction programs are planned to face
the continuously increasing needs for all purposes of use (housing, education, sanitary,
administrations...)
The challenge is to protect both of existing historical sites and representative classified
monuments and buildings and new construction programs from potential natural risks and to
promote a policy of maintenance and restoration of old buildings and specially to keep alive the
historical symbols of the past civilisation and the stamps of historical occupancies (ROMAN
EMPIRE ,OTOMAN EMPIRE, SPANISH, etc)

193
1 CHARACTERISTICS OF THE ALGERIAN SEISMICITY

The characteristics of the Algerian seismicity are based upon the information recorded in the
historical seismicity, the seismo-tectonic studies, the seismology, the paleo-seismology, seismic
hazard. Today we possess important indications on the seismic activity in Northern Algeria. The
Earthquake monitoring is conduced trough a seismological network implemented with a
countrywide accelerometer reseal.
This activity is essentially focused in the Northern area of the country, although with much
sporadic way, some micro quakes are recorded in the sub-Saharan side in the Saharan party. In
the Northern area, the Tellian region is the most active one.

The established zoning shows that the regions of El Chelif, Oran and Algiers are the most
threatened areas since in those zones the most important earthquakes occurred.
This seismicity is linked to the complex tectonic movements of the convergence of the
African plate in the south towards the Eurasian plate in the north.
The seismic zoning of the Algerian territory shows that the Tellian strip, notably in its coastal
fringe, is subject to the highest degree of seismic hazard.
The lack of strong ground motion data was significantly experienced when elaborating the
first Algerian aseismic building code in 1976. Revisions were gradually introduced after mean
earthquakes in 1981, 1983, 1988, 1990, 1999 and 2003.
Actually it becomes urgent to set a new policy of urban planning and occupancy of this
important area considering its exposure to the earthquake risks.
The developing construction programs should concern more and more the middle region of
the high plateaux and the large Sahara in the south to fix more and more population
considering the total inexistence of earthquake risk there and the rare presence of consistant
monuments. ( projects of the two main new cities of Hassi Messaoud and Bouzeghoul)
However, the existing constructions of the most important cities and urban areas mainly
concentrated in the northern part of Algeria requires a continuous effort of maintenance and
protection of the established historical patrimony, beside the cumulated economic
infrastructures.

Figure 1. Algeria Erthquaque Hazard Map
194
Figure 2. Algiers, Casbah in XVII Century
2 STATUS OF THE EXISTING FRAME IN ALGERIA

The evolution of the housing sector in Algeria remains obviously marked by the history of
the country. The actual stock of houses is consisting of more than 6.5 million units. More than
25 % of existing constructions are threatened by an advanced decomposition process.
The identified construction typologies are closely related to the construction periods.
In such a way the main cities of the northern part are still keeping particular styles of
construction as a heritage of all the historical development of construction technologies.
Thus, the frame heritage is presented as a stratification of construction techniques.

TYPOLOGY OF EXISTANT CONSTRUCTIONS

Figure 3. Typical street and houses arrangement in Medina
2.1 Category A:

Encloses the remaining rural and urban structures of houses and districts built before 1920
(old medina), including chiefly constructions made of adobe, clay, random rocks, with wooden
stiffener: (Casbah of Algiers, Oran, Constantine, Dellys, Cherchel etc)
On the urban scale they are organized in:
BLOCKS: varying number of houses, individualized by communication streets which
surround them. This makes each block reacts independently from the others during the shakings.
THE HOUSES: are semi-detached, overlapping and leaning against each other forming a
compact homogeneous unit.

195
Figure 4. Covered built Circulation Figure 5. Masonry wood reinforced walls

THE SABATS: Number of streets are covered by galleries on top of which the houses extend
and thus creating roofed passageways.
They can be flat with wooden logs incorporated or vaults built out of stones or bricks.
Bearing stone walls are linked one to another by alternate crossing of wood logs
(consolidating the angles) every 50 cm in height, wood logs of about 2 m long.
Partition brick walls are connected to main walls by wood logs.
On a structural scale, floors are built by superposition of two layers of wood logs inserted in
all width of bearing walls (rigid diaphragm).

Corbellings for frontal balcony are set overhanging and supported by wood logs forming an
angle with bearing-walls.
Wooden framing is found around the openings.

2.2 Category B:

Buildings with load-bearing walls: carved rocks, stone and vaulted metallic flours.
This type of construction represents 90% of old urban construction frames (colonial texture)
which constitute the main central heart of the towns in major important cities like Algiers, Oran,
Constantine, Annaba. Generally, the predominant type of building is raised on 01 to 05 floors.
This class is in a rather altered state, with an average of age reaching the century.

2.3 Category C:

Early in the sixties, new constructions were set on the existing urban occupancy mainly made
of column-beam reinforced concrete frames which represent the major part of individual private
houses and public buildings in the new urbanized areas around old centers of the cities. A
massive exude of population from rural areas to the urban sites chiefly for economical trend and
employment The main reason

A great part of this class of constructions suffered hardly under destructive earthquakes
occurred in main old towns ( El Asnam OCTOBER 10,1980 (M
s
7.3), Constantine October 27,
1985 (M
s
5.7), Mascara August18, 1994 (M
s
5.6), Ain Temouchent December 22, 1999 (M
s
5.6),
ALGIERSBoumerdes May 21, 2003(M
s
6.8).
Although many of them seem moderate, the poor quality of construction and bad soil
conditions were the main reasons of massive destructions and wide collapse.
The cases of Ain Temouchent city extended in a river bed or the surrounding districts of
Algiers built in an anarchical way with a very poor quality of materials remain typically
representative of the fragility and the extreme vulnerability of this urban texture category.
The behavior of this class of construction under earthquake has been traducing the lack of
the seismic code and the inexistence of quality control procedures.

196
2.4 Category D:

This category consists of rigid panels concrete construction systems. This technology has
been applied intensively after El Asnam earthquake in 1980 and the standard revision
recommendations consequently introduced.

2.5 Category E:

New designed important structures are nowadays more and more engineered with regards to
advanced earthquake resisting methods and various construction processes.
From these categories, those of A & B are usually called old constructions
This heritage is in danger. Both of the weight of time and the carelessness of users are
worsening the situation, the assessment gave catastrophic results.
The requalification of this old heritage should inevitably goes through a multi dimensional
scientific step (Urban, technical, social and economical). The urban fabric includes a certain
number of buildings that represents a culture and the marks of countrys History.
The in-depth study of the old constructions as an architectural & urban creation holds the
marks of a particular civilization and a specific transformation; the old districts are
sentimentally more attractive than the new or modern ones, and more pleasant in appearance.
This leads back to sociological and cultural referents.
The consciousness of the critical status of the old constructions showed up the importance of
saving it. This requires a global strategy of intervention and specific means given the extreme
fragility of the traditional urban fabrics.
The status of these constructions and the high density of population give a particular
vulnerability to the Algerian great metropolis.
Indeed, 30% of the population lived in urban areas in 1966, 62% in 1998 and according to
the latest RGPH results; more than 80% in 2008. The towns of more than 100 000 inhabitants
jump from 3 in 1962 to 32 in 1987 and more than 60 nowadays.
10 have more than 200 000 inhabitants. The urban areas which are the result of conurbations
and the densification of the urban networks become larger and larger, while the general
population of Algeria had tripled during that period. The urban population from its side was
multiplied by 10.
The public authorities are interested by the old constructions; this interest is materialized
today by lots of rehabilitations conducted over buildings and notably over historical sites and
monuments, we cite some instances regarding that in what follows:

3 EXPERTS RENOVATION EXPERIENCES CASES

3.1 Rais Palace (Bastion 23) ALGIERS
Figure 6. Bastion 23 General plan distribution
197
Figure 7. General view before intervention
Summary sheet of building revamping:
Country : Algeria Written by: BERDJANI Mustapha
Monument : Ras Palace (Bastion 23)
Identification:Description of the original building
Revamping goals: To restore Ras Palace (Bastion23) and to create the centre of arts and
culture in order to promote arts & culture for a large public.
Address: 23, AMARA Rachid Boulevard - Bab El Oued Algiers.
Location of the Building: Included in the saved perimeter of Algiers Casbah (World
classification in 1992)
Built environment: Weak density
Protection for Building: Yes, since 1909 (carried on by the
Law n67-281 of 20/12/1967)
Owner: National agency of archeology (Ministry of Culture)
Contractor: SCI-MBM Contractors (Italy)
Main companies: SCI-MBM Contractors (Italy), Company of creation of cultural heritage
(ERPC)/ Algerian architects.
Date of renovation: Novembre1987- End 1994
Overall cost of the operation (Taxes included): Twelve million of dollars (12.000.000 $)

Financing: Equipment budget financed by the government.
Documentation source: Study performed by (Yapi-Merkezi) Turkish engineering office.
Archeological research (Algerian team) written data, photographic and iconographic.
Bastion 23 is an architectural set composed of palaces and houses of Moorish type. It was
part of the urban fabric of the ancient town, which was surrounding a set of districts that
constituted El-Djazair town at the Ottoman time between the XVI XIX century.
Walls: Mixed bearing walls, made of bricks and stone masonry.
Floor: Traditional floor made of thuyas log with marble and baked clay covering.
Waterproof quality: Sand covering, lime and baked clay mortar.
Outer covering: Coating used: sand, traditional lime.
Openings and front elements: Faades with discrete elements toward outside.
Water supply: Rain water collecting system by canalization made of baked clay as well as the
presence of three wells.
Purification system: Peripheral gutter system.

198
Figure 8. Site works view
Figure 10. Architectural improvement of the main side Figure 9. Internal decoration of
Diagnosis conducted on constructive elements:
Walls and columns: Analysis of the structure and materials.
Floors: Analysis of the structure and materials.
Roofing: Analysis of the impairment status of the traditional waterproof quality.
Coverings: Analysis and localization of the altered areas.

Renovation works

Walls:Reinforcement of the existent walls
in order to consolidate and stabilize the structure and the rebuilt of the collapsed walls.

Floor:
Reconstruction of floors at 100%, with the re-use of bearing elements that was in good
conditions.

Waterproof quality:
New layers have been applied Pose (sand, cement, baked clay)
Coverings:
General cleaning of the impaired coatings. A Treatment of all deep and superficial cracks has
been conducted. Marble, baked clay and schist have been used.
Paving:
Lying of paving on outside areas.
Equipments:
Renovation of different networks (electricity, gas, AEP, phone network, purification system)
Outer arrangement:
The outer area was arranged as a parking with a covering made of cobblestone Based on the
overall analysis of all main characteristics of the various components, a multidisciplinary
199
Figure 11. Protection retaining wall
Figure 12. General view after completion
Turkish engineering team (Yapi Merkezi), has elaborated a synthesis and an assessment of the
building.
Restoration, renovation and arrangement of Monuments surrounding areas. Excavation
works had been conducted in order to determine the exact plan of the Monument, the
reinforcement of walls, reconstruction of collapsed parts (faade walls etc), the treatment of
faades and architectural corrections after restoration and consolidation of the entire original
structure, renovation of different networks (electricity, gas, telephone network, purification
system), mending the pavement of the soil surface.

Evaluation of the results
Integration of the building: after nearly seven years of restoration, Bastion 23 became the
urban or scenery environment: it was the first experience of retrieving a historical and presented
to the public.
Compliance with the initial program: The program that has been applied in this project is
regarded as a public one. Its conversion to a museum and to an Arts & culture centre contributes
to the promotion of culture and to keeps alive the memory of Algiers.
Compliance of usage: Air conditioning, fire security and all the necessary comfort.
Social balance: Employment source, a place for meetings and cultural exchange.
Bastion 23 plays a socio-cultural & economic role.
Maintenance of the building: The Arts & Culture center is permanently kept under a
continuous and global maintenance program.

200
Figures 13, 14. Walls cracks repairing and reinforcement
Figure 15. Reinforcement of poof dome support and stone walls
3.2 Abdellah Ibn Salem Mosque restoration

The construction of Abdellah Ibn Salem mosque goes back to the colonial period and its
inauguration dates from the year 1908.
Abdellah Ibn Salem Mosque is one of the historical sites in Oran city thanks to its arab-
moorish architectural style.
The mosque was subject to time vagary and the lack of maintenance, it showed important
degradations and this lead the public authority to decide its rehabilitation.
The bearing structure is composed of bearing stone walls with vaulted veined bricks by IPN
steel profiles. The prayer room which occupies the central part is protected by a false ceiling in
half circle made of lathing reeds with plaster covering surmounted by a double side tiles roofing
which lays over mixed steel-wood framework.
The two minarets are placed over the two angles of the main faade are made by rock blocks
of large seize linked by a hydraulic lime mortar.
The restoration works consist of:

201
- The preparation of cracked and broken walls.
- The rehabilitation of damaged vaulted roof.
- The rehabilitation of impaired wooden framework elements.
Figure 16. Internal view of the buildings Figure 17. Waterproofing coating
protection
Figure 18. Original view of Arenas Figure 19. Before renovation

3.3 Orans Arenas rehabilitation

The arenas have been built between 1910 and 1911.In the 40s, Oran was still more Spanish
than Arab. The city was founded by the Andalousians in the year 903; the second

Spanish occupation does back to 1774 and after the French conquest, the city kept without
doubt the Spanish urban stamp.
In the 50s, a big majority of the inhabitants of ORAN were from Spanish origin. Some are
political refugees, but most of them were staying there since many generations
The best world toreadors came to perform in these large arenas which could contains
thousands of persons. They confronted the Toros of the most famous Spanish Cowboy boots
or herd of bulls.
This is not surprising at all, especially with the presence of the Eckmhl site arenas found in
the Southern working class area in ORAN, the tradition of bullfighting is still felt there
This imposing architecture which was abandoned and wrongly used had to wait for the
heritage to retrieve its place in the cultural priorities, and then Orans arenas were
202
programmed for the rehabilitation of the heritage sites of the local council program, while
waiting to be indexed in the list of classified sites.
Since then, local concerned authority started to look over this site and planed reinforcement
and renovating operation

STRUCTURAL DESCRIPTION OF THE ARENAS

The arenas have a circular form, composed of a hybrid structure made of stones, which are in
staggered rows, and reinforced concrete for the extension added part.
It is founded on a channeled linear feet. The link between the different stones is ensured by a
lime link mortar.
The brace of the external and internal surrounding walls of the arenas in the radial
direction is ensured by thick masonry shear walls in form of a multi span stone arc that lies over
stone poles.
Discharging arches built out of stones or bricks are set in a circular arrangement of a thick
framework, to allow transfer of horizontal loads to the ground.
Construction is erected as a compact dynamic block.
IPN steel profile tie beams are arranged at different highs to ensure the rigidity of stone poles.
The hollow block roofing lies directly over a rack which at its turn lies directly over the brace
rock walls. Circular openings are laid out at the level of the surrounding walls of the arenas in
order to ventilate the place and avoid the condensation of the air saturated by water on the walls
(humidity). The access to the terraces for the public is ensured by stairs of stone masonry.
The hollow block roofing lies directly over a rack which at its turn lies directly over the brace
rock walls.

Figure 20. Supporting walls and arch distribution

203

Figure 21.22.23 Repairing works (concrete Jacketing)

4 CONCLUSIONS
Two main reasons has been developed as an explanation of omitting to maintain the heritage
patrimony and monuments in acceptable environmental conditions and stable configuration.
The first is the great demand of housing and construction programs concentrated in the
existing metropolis in the north putting all the potential on this aim.
The second cause of deterioration is due to age effect and the weight of time added to
destructive sporadic earthquakes increasing the secular collapse process.
A new appearing reason is an economical tendency, thus the price of land inside urban areas
has grown to incredible values in main cities due to the high density of occupation of the soil.
This is actually encouraging the squatting of the latest free area of land after total collapse of
existing constructions and ancient monuments.

204
REFERENCES
Architectural project of renovation of BASTION-23 from Ministry of Culture by Turkish engineering
office : Yapi-Merkezi
Project undertaken with the help of The European Union, Euromed HERITAGE and The International
Cooperation Agency of Spain (AECI).
Architectural design of Ibn Salem Mosque by IBDAA Architects Oran 2007
Design architectural group of BARCO Architects Oran for Arenas Project 2008
CTC OUEST diagnostic and expertise works on the three Exposed cases 2003 2006 2008
CTC reinforcement methods and applicable renovation techniques 2006
205

1
1 INTRODUCTION

Nonlinear dynamic response is one of the distinguished characteristics of the ground under
strong shaking. The nonlinearity has been reported and investigated by many researchers for
more than 40 years. The investigation has been mainly concerned with the following two sub-
jects;
(1) Shift of predominant frequency of acceleration towards lower frequency [e.g., Idriss and
Seed (1968), Idriss (1990)].
(2) High spikes on horizontal acceleration time histories due to cyclic mobility in relatively
dense sand deposit [e.g., 1987 Superstition Hills earthquake (Holzer, et al. 1989), 1993 Ku-
shiroki, Japan, earthquake (Iai, et al. 1995), 1994 Northridge earthquake (Bardet and Davis
1996), 1995 Hyogoken Nanbu, Japan, earthquake (Iwasaki and Tai 1996), 2001 Nisqually
earthquake (Frankel, et al. 2002), 2004 Niigataken Chuetsu, Japan, earthquake, 2007 Niiga-
taken Chutsu-oki, Japan, earthquake].
The phase-shift of predominant frequency (1) is caused by inelastic response of the ground
material against large ground strains. This nonlinearity can be indirectly observed by comparing
the frequency transfer functions between small and large earthquakes as a shift of the peak fre-
quency of large earthquakes towards lower frequency. In recent studies, to derive such a transfer
functions, time histories obtained from seismometers forming the vertical array have been inten-
sively employed, with which simultaneous recordings of the base at certain depth and surface
motions are achieved. Mechanism of the nonlinearity has been investigated through laboratory
experiments for soils under cyclic loadings, and it has been understood that large cyclic shear
strain amplitude causes degradation of soil stiffness. Because of this type of stiffness degrada-
tion predominant frequency of the ground is shifted toward lower ones.
High-spikes on horizontal acceleration records (2) are typically found in locations where the
ground material constitutes dense saturated sand. Results of stress controlled undrained cyclic
Numerical Analysis of Near-Field Asymmetric Vertical Motion

T. Tobita, S. Iai & T. Iwata
Disaster Prevention Research Institute, Kyoto University, Kyoto, Japan
ABSTRACT: An unprecedented vertical surface acceleration of nearly four times of gravity was
recorded during the 2008 Iwate-Miyagi Inland, Japan, earthquake (Mw 6.9). The motion was
recorded at the KiK-net, IWTH25 station, located 3 km southwest of the epicenter. The station
is equipped with three-component accelerometers, installed at both the free-surface and the bot-
tom of a 260-m borehole. Wave form of the vertical acceleration shows clear asymmetric form
with large amplitude in upward direction. To study this recently discovered nonlinear behavior
of the surface ground motion, numerical analysis with the finite element method has been con-
ducted with parameters derived from the borehole data at the station. The analysis successfully
simulates the asymmetric vertical motion. The analysis indicates that the asymmetric motion
may be characterized by the existence of lower bound of negative acceleration and high positive
pulses caused by the compression stress of the disturbed surface ground material.

206

2
triaxial tests for dense sands show increase of volumetric strain due to stress dilatancy at large
strains. This increase of volumetric strain causes rapid reduction of the excess pore water pres-
sure, and results in the rapid recovery of shear strength. When this phenomenon occurs in real,
the rapid increase of shear stress may be observed as spiky response on acceleration records
(Ishihara 1993). When the spikes appear, other frequency components are attenuated due to
build up of excess pore water pressure reaching to liquefaction state. These spikes may give
overestimation of the maximum acceleration amplitude and duration of shaking (Bonilla, et al.
2005).
In addition to the above mentioned two types of nonlinear response of the ground, the 3rd
nonlinearity under strong ground motion has been recently discovered as,
(3) Asymmetric from of the surface vertical acceleration amplitude (Aoi, et al. 2008).
In the 2008 Iwate-Miyagi Inland, Japan, earthquake, very large acceleration amplitude which
exceeded 40 m/s
2
(3 component combined) was recorded at the KiK-net, IWTH25 station (Fig.
1 and 2). The site is located about 3 km southwest of the epicenter on the hanging wall of the
seismic fault. The earthquake was caused by the inland reverse fault with 30 km strike and 20
km in depth. At IWTH25 station, 3 components seismometers are installed at the surface and
G.L. -260 m. The site is located in volcanic region and locally it is on the river sediments under-
lain by igneous rocks, such as tuff. The recorded wave form of the surface vertical acceleration
has large amplitude only in positive direction. The maximum amplitude of vertical acceleration
is 4 times larger than acceleration of gravity and 2 times larger than its horizontal components.
Aoi, et al. (2008) reported and qualitatively explained the mechanism of this phenomenon by
the analogy of bouncing a matter on a trampoline and called the Trampoline effect.
The asymmetric form may be attributed to physical characteristics of granular media, which
shows asymmetric response against normal compression and extension force. That is, granular
media, such as dry sands, has less resistance against tension force. Objective of the present
study is to simulate this asymmetric response of the surface vertical acceleration and give physi-
cal background to explain this phenomenon.

Figure 1 Location of the epicenter of the 2008 Iwate-Miyagi Inland, Japan, earthquake and IWTH25,
KiK-net station.

207

3

Figure 2 Measured time histories and Fourier amplitudes of surface acceleration at IWTH25 after the
2008 Iwate-Miyagi Inland, Japan, earthquake. Asymmetric amplitude appears on the UD record.
2 ANALYTICAL MODEL

Numerous analytical models which simulates the nonlinear dynamic ground motion have
been proposed. Equivalent-linear method [SHAKE (Schnabel, et al. 1972)EERA (Bardet, et
al. 2000)] has been widely used for its simple configuration and precise results for low to inter-
mediate levels of ground shaking. To directly simulate nonlinear stress-strain response of soils,
the H-D model (Hardin and Drnevich 1972) and the R-O model (Ramberg and Osgood 1943)
have been widely used. By combining these nonlinear constitutive models with Masings law
(Masing 1926), hysteresis response of soils can be simulated [e.g., NERA (Bardet and Tobita
2001)]. Iwan (1967) and Mroz (1967) proposed a model for soil nonlinearlity and hysteretic be-
havior by using multiple springs and sliders.
By taking advantage of plasticity theory started in the field of metal engineering, constitutive
models based on the theory of plasticity have been formulated and widely used in the frame
work of finite element analysis (Roscoe, et al. 1963, Roscoe and Burland 1968, Matsuoka 1974,
Dafalias and Popov 1975, Sekiguchi and Ohta 1977, Mroz, et al. 1978, Hashiguchi 1980,
Towhata and Ishihara 1985, Nakai 1986). These models generally require many parameters,
which are more or less derived from experimental results, to simulate more realistic and compli-
cated behavior of soils.
To simulate soil liquefaction, dynamic solid and fluid coupling behavior has to be incorpo-
rated into constitutive modeling. Constitutive models based on the effective stress concept have
been proposed for such a material (Martin, et al. 1975, Prevost 1985, Iai, et al. 1992, Oka, et al.
1999). In these models, by applying dependency on the effective mean stress to shear behavior
of soils, liquefaction as well as dilatant behavior can be successfully simulated.
In this study, the multiple simple shear mechanism proposed by Towhata and Ishihara (1985)
and Iai, et al. (1992) is implemented.

2.1 Multiple simple shear mechanism
The finite element code called FLIP (Iai, et al. 1992) is implemented for nonlinear site re-
sponse analysis. Total stress analysis is conducted, i.e., no excess pore water pressure buildup
during shaking is assumed. The code utilizes the multiple simple shear mechanism as nonlinear
208

4
constitutive relation (Towhata and Ishihara 1985). In this model, contact forces between sand
particles are idealized by evenly distributed multiple springs whose property is characterized by
the hyperbolic type (Fig. 3). The model automatically accommodates the principal stress rota-
tion which plays an important roll in the cyclic behavior of anisotropically consolidated sands.
In what follows, core of the modeling is briefly introduced. Details are found in Ozutsumi
(2003). Let us consider the following vectors of stresses and strains,
{ } { } ' ' ' '
T
x y xy o o o t = (1)
{ } { }
T
x y xy c c c = (2)
where ' x o and ' y o are the normal stress, x c and y c are normal strain, 'xy t and xy
are shear stress and shear strain, respectively. Incremental form of constitutive relationship can
be written as follows,
{ } { }{ } { }
/ /
'
T
L U L U
d R n n d o c = (3)
where the vector { }
/ L U
n specifies direction of stress increment, the scalar
/ L U
R defines mag-
nitude of stress increment per unit strain increment along the direction { } n , vector { } n gives
the direction of strains. The subscripts L/U indicate the components are different in the direction
of loading L and unloading U. Iai et al. (1992) postulated that the incremental constitutive
relation, Eq. (3), is given by I+1 separate mechanisms for i=0 to I in associated form, i.e.,
{ } { }
/ L U
n n = , as
{ } { }{ } { }
( ) ( ) ( )
/
0
'
I
T
i i i
L U
i
d R n n d o c
=
=
(4)
To specify the volumetric behavior, the first term of Eq. (4), i=0, is independently defined so
that Eq. (4) becomes,
{ } { }{ } { } { }{ } { }
(0) (0) ( ) ( ) ( )
/
1
'
I
T T
i i i
p L U
i
d K n n d d R n n d o c c c
=
= +
(5)
where K is rebound modulus and
{ }
p
dc is an additional volumetric strain increment due to di-
latancy given by
{ } { }
/ 2 / 2 0
p p p
d d d c c c = (6)
Direction vectors are given as
{ } { }
(0)
1 1 0
T
n = (7)
{ } { }
( )
cos -cos sin
T
i
i i i
n u u u = (for i = 1 to I) (8)
where i u is the angle of the ith spring from the horizontal axis, and given as,
( 1) i i u u = A (for i = 1 to I) in which / I u t A = .
Rewriting Eq. (5) yields,
{ } { }{ } { }{ } { } { }{ } { }
{ } { }{ } { }
(0) (0) ( ) ( ) ( ) (0) (0)
/
1
(0) (0)
'
=[D]
I
T T T
i i i
L U p
i
T
p
d K n n R n n d K n n d
d K n n d
o c c
c c
=
(
= +
(

(9)
in which matrix [D] is,
{ }{ } { }{ }
(0) (0) ( ) ( ) ( )
/
1
[ ]
I
T T
i i i
L U
i
D K n n R n n
=
= +
(for i = 1 to I) (10)
Rebound modulus K is expressed as
0
0
'
'
m
K K
o
o
| |
=
|
\ .
(11)
209

5
where K
0
is the rebound modulus at a reference confining stress
0
' o , ' o is an effective stress,
and m is a parameter which correlates K and stress ratio
0
'/ ' o o . In the frame work of the mod-
eling, tangent shear modulus is given as,
( )
/
( )
i i
L U i i
dF
R d
d

u
= A (12)
in which the function ( ) i F is a hyperbolic function which is equivalent to the spring force
due to the displacement of a multi-spring in one radian. By assuming Masings law for cyclic
response and defining damping of spring, the shape of function ( ) i F is determined (see
Ozutsumi (2003) for detail).
In each shear mechanism, loading and unloading conditions can be defined by the sign of,
{ } { } ( )
( )
cos sin
T
i
i x y i xy i
d n d d d d c c c u u = = + (13)
where
{ } { } ( )
( )
cos sin
T
i
i x y i xy i
n c c c u u = = + is the virtual shear strain of the mechanism
i= 1 to I. Each tangent modulus depends on the current state and history of each virtual simple
shear strain
i
.
In matrix form, Eq. (10) can be written as,
| | 1 2 3
1 1 0 1 1 0 0 0 1 0 0 0
1 1 0 1 1 0 0 0 1 0 0 0
0 0 0 0 0 0 1 1 0 0 0 1
D K G G G
| | | | | | | |
| | | |
= + + +
| | | |
| | | |
\ . \ . \ . \ .
(14)
where
2
1
1
( )
cos
I
i
i
i
dF
G
d

u u
=
= A
(15)
2
1
( )
cos sin
I
i
i i
i
dF
G
d

u u u
=
= A
(16)
2
3
1
( )
sin
I
i
i
dF
G
d

u u
=
= A
(17)
In the finite element code, FLIP, if the volumetric strain is positive, judged by the following
relation,
( ) 0
x y p
c c c + > (Tension) (18)
then, all the stress components are set to be zero, i.e., ' ' ' 0
x y xy
o o t = = = . As it will be shown
in the next section, this is the property that simulates asymmetric behavior of granular media
with no cohesion.

Figure 3 Schematic view of the multiple simple shear mechanism
210

6
2.2 Model behavior under normal compression and tension stress
Behavior of the multiple simple shear model is investigated by applying vertical compression
and tension forces to a single element (Fig. 4). Model parameters for this particular study are,
shear modulus=3.210
5
(kPa), Poissons ratio=0.33, density=1.8 (t/m
3
), and friction an-
gle=35. Computation was carried out under drained condition. Firstly the element is isotropi-
cally consolidated with stress o
0
= 49 (kPa) with nodal constraints shown in Fig. 4(a). Then, it is
stretched and compressed vertically by enforcing cyclic normal force to nodes 1 and 4 (Fig. 4b).
Here, the horizontal displacement of nodes 1 and 4 are constrained (Fig. 4b). Analytical time
step of At=10
-4
was employed. Figure 5(a) depicts time histories of mean stress,
o
m
=(o
xx
+o
yy
)/2, and volumetric strain, c
v
=c
xx
+c
yy
, on the left and right vertical axes, respec-
tively. For both strains and stresses, tension is taken as positive. As shown in Fig. 5(a), when the
mean stress is zero, tensile strain abruptly increases, that is, the element is vertically stretched.
Stress strain curve in Fig. 5(b) gives another view of this property. Namely, the curve starts at
the point indicated as 1 in Fig. 5 where mean stress is -49 kPa, then gradually stretched and
once the mean stress reached to zero at 2, volumetric strain abruptly increases up to about
0.3%. In the compression side, in the given range of enforced vertical stress, the stress-strain re-
lationship is within an elastic range following the line whose slope is defined by the rebound
modulus of K=605 MPa.

Figure 4 Illustration for application of isotropic confining pressure (a) and cyclic axial stress (b) to a sin-
gle element.

Figure 5 Time histories of (a) mean stress and volumetric strain, and (b) mean stress versus volumetric
strain curve.

211

7
2.3 Model behavior against vertical input acceleration

To simulate the asymmetric vertical ground motion with simple model, 1D column of 10 m
deep with 20 elements are implemented. Each mesh is a square with 0.5 m at the side. Dis-
placement degree of freedom of nodes at the bottom mesh is fixed in both horizontal and verti-
cal directions. A pair of nodes at the same height was given the same displacement degree of
freedom to simulate half-space ground motion. Analysis is carried out by assuming dry condi-
tion. Sinusoidal input acceleration with 10 Hz and the amplitude of 19.6 m/s
2
(=2 g) is given at
the base of the model. In this study, only the vertical motion is given. Computational time step
of At=0.001 (s) was employed.
Figure 6(a) is a section of the time histories from 0.5 to 0.8 (s) of volumetric strain (c
v
=c
x
+c
y
),
input and surface accelerations. In the same manner, Fig. 6(b) shows mean stress
[o
m
=(o
x
+o
y
)/2] of the surface element, in which vertical scale of input and surface accelerations
is magnified 8 times for clarity. Here, tension is taken as positive. In Fig. 6(a), tensile strain can
be observed while tensile stress is not generated [Fig. 6(b)] as it was shown in the previous sec-
tion. As shown in Fig. 6(b), surface acceleration start to abruptly increase when the input verti-
cal motion becomes lower than 9.8 m/s
2
. This is because, at this moment, the surface element is
free falling and compression force is generated in an element, and reaction to this compression
force may create acceleration pulse in the vertical upward direction. Again from Fig. 6(b), lower
bound of surface acceleration is limited by the minimum amplitude of vertical input accelera-
tion. Thus, the asymmetric motion of the vertical acceleration can be characterized by the exis-
tence of lower bound of negative acceleration and positive pulse in the vertical acceleration due
to compression of soils.
Figure 7 compares mean stress and volumetric relationship of the surface element for various
amplitudes of sinusoidal vertical input accelerations. As shown in Fig. 7(a), when the vertical
input acceleration is small, response is in the compression side and follows the line whose slope
is defined by the initial elastic rebound modulus. As amplitude of input acceleration increases,
the curve starts to show nonlinearity. If the amplitude of input vertical acceleration exceeded 9.8
m/s
2
, tensile strain appears while mean stress is kept to be zero.

Figure 6 Time histories of (a) volumetric strain, vertical input and vertical surface acceleration, (b) mean
stress, vertical input and vertical surface acceleration of the surface element. Scale of acceleration in (b)
is magnified 8 times of that of (a).

212

8

Figure 7 Mean stress versus volumetric strain relationship of the surface element for sinusoidal input ac-
celeration with various amplitudes: (a) 4.9 m/s
2
, (b) 9.8 m/s
2
and (c) 19.6 m/s
2
.
3 NUMERICAL ANALAYSIS
3.1 Model parameters for site response analysis of KiK-net, IWTH25 station
Physical model parameters, such as, shear modulus, poissons ratios, and densities, are esti-
mated from the PS velocity profile obtained at the KiK-net, IWTH25 station. Table 1 summa-
rizes model parameter implemented in the following analysis. Based on the velocity profile,
model ground is divided into 7 layers from the surface to G.L. -242 m (Fig. 8). From P and S
wave velocities, shear modulus and poissons ratios of each layer can be estimated. Borehole
profile deeper than 120 m shows that the tuff whose S wave velocity exceeds 1,300 m/s is pre-
dominant. Above it, mud stones and terrace deposit constitute the layers. Densities of material
are given as follows; 2.2 t/m
3
for surface cover soil, 2.3 t/m
3
for sediments, and tuff and mud
stones are set uniformly 2.6 t/m
3
(Toyota 2009).
In the analysis, the ground is modeled with single column elements (133 elements) whose
width is 1 m, and height varies from 0.2 m (layer 1) to 5.43 m (layer 7). With this model
ground, frequency component of the maximum 50 Hz can be properly simulated unless being
attenuated.

Table 1 Model parameter of the ground profile at the KiK-net, IWTH25 station

213

9
0
80
160
240
0 4500
Velocity (m/s)
D
e
p
t
h

(
m
)
P
S
Layer 1
Layer 2
Layer 3
Layer 4
Layer 5
Layer 6
Layer 7

Figure 8 P and S wave velocity profile at IWTH25 station (NIED 2009) and assumed layering system for
numerical analysis.
3.2 Site response analysis with small input motion
Site response analysis with small input motion is conducted to examine the model parameter
defined above (Table 1). The small input motion used in the analysis (M
w
=6.8, Depth 108 km
on 24/7/2008) was recorded at the same site, IWTH25, where large acceleration was measured
during the 2008 Iwate-Miyagi Inland, Japan, earthquake. In the analysis, both the horizontal
(NS) and vertical motion (UD) were simultaneously input at the base of the model. The meas-
ured maximum accelerations of NS and UD components at the base are, respectively, 1.2 m/s
2

and 0.83 m/s
2
and at the surface, 1.1 m/s
2
and 0.71 m/s
2
, respectively. Time step of the analysis
was At=0.005 (s).
Computed time history of the horizontal motion (NS) in Fig. 9 is in good agreement with the
recorded motion in terms of both maximum amplitude and the shape of envelope, while the
maximum amplitude of vertical motion is over estimated about 40%. As shown in Fig. 9, in the
case of small input motion, amplitude of both measured and computed vertical motions are
symmetric.
Transfer functions of both horizontal and vertical motions in Fig. 10 show significant agree-
ment below 10 Hz. However, for response above 10 Hz is under estimated. This may be because
high frequency components in the input acceleration were too small to be transmitted to the sur-
face without attenuation. In reality, there are many sources of high frequencies in the ground,
such as inclined boundaries between soil stratum and/or heterogeneity, which may be properly
included in two-dimensional analysis.

Figure 9 Measured and computed surface acceleration (NS and UD components) for small input motion.
214

10

Figure 10 Transfer functions of measured and computed acceleration for small input motion.
3.3 Site response analysis for the 2008 Iwate-Miyagi Inland, Japan, earthquake
Site response analysis is conducted to simulate asymmetric vertical motion observed in the
2008 Iwate-Miyagi Inland, Japan, earthquake. Recorded input motion of both horizontal (NS)
and vertical (UD) components at the base of the KiK-net array at IWTH25 was input at the base
of the model ground with time step of 0.01 s which was the same time step as measured accel-
eration. As shown in Fig. 11, computed surface acceleration has similar asymmetric shape in the
vertical component as measured one, i.e., positive spikes and bounded negative amplitude. The
maximum amplitude of measured vertical acceleration was 33 m/s
2
, while the one computed
overestimates about double to be 62 m/s
2
. Negative amplitudes of surface vertical acceleration
show, on average, significant agreements between observation and analysis, and it is close to the
acceleration due to gravity. This might indicate the temporal free-fall of the ground.
Transfer function of computed vertical acceleration [Fig. 12(b)] shows a peak at 5 Hz which
is not seen in the observation. Cause of this is under investigation. Except the peak, transfer
function is over all in good agreement. Higher frequency components are properly simulated
because input acceleration might be large enough for higher frequency components to be trans-
mitted to the surface. Transfer functions of horizontal motions also show in good agreement
[Fig. 12(a)]. However, there are no clear peaks in the transfer function of computed horizontal
motion.

Figure 11 Measured and computed surface acceleration (NS and UD components) for the 2008 Iwate-
Miyagi Inland earthquake.
215

11

Figure 12 Transfer functions of measured and computed acceleration for the 2008 Iwate-Miyagi Inland
earthquake. (a) NS and (b) UD components.

Figure 13 compares measured and computed horizontal and vertical acceleration. Computed ho-
rizontal acceleration [Fig. 13(a)] shows no clear peaks when vertical acceleration is negative.
When the vertical acceleration is negative, elements near the surface is free-falling, and at this
state confining pressure may be very small, and shear waves cannot be transmitted in such a soft
material. While as shown in Fig. 13(b), this is not clear in the measured acceleration. One possi-
ble reason may be that in reality there are sources of higher frequencies as mentioned earlier.
Computed time histories from 4.5 to 5 s of volumetric strain [Fig. 14(a)], mean stress [Fig.
14(b)] with computed vertical surface acceleration, and mean stress and volumetric strain rela-
tionship [Fig. 14(c)] are plotted to see the model behavior of the surface element. When the sur-
face acceleration is negative, tensile volumetric strain indicated by number 3 to 5 in Fig. 14(a)
appears at zero mean stress. When the surface element is under compression indicated by num-
ber 6 and 7 in Figure 14, positive spikes appear on the surface vertical acceleration. In the same
manner as Chapter 2, mean stress and volumetric strain relation [Fig. 14(c)] follows a quadratic
curve which is a volumetric stress and strain relationship indicated by Eq. (11).

Figure 13 Computed (a) and measured (b) time histories of NS and UD components of surface accelera-
tion with magnified time scale for 3 to 8 seconds of Fig. 9.
216

12

Figure 14 Computed time histories from 4.5 to 5.0 (s); (a) volumetric strain, (b) mean stress. In (a) and
(b), computed UD components of surface acceleration is plotted as a reference. (c) Mean stress versus
volumetric strain relationship.
4 CONCLUSIONS
During the 2008 Iwate-Miyagi Inland, Japan, earthquake, very large acceleration record
which exceeded 40 m/s
2
(3 components combined) was observed at the vertical array site, KiK-
net, IWTH25 station. The surface vertical acceleration record had asymmetric amplitude. In the
present study, analytical studies were carried out for this newly discovered nonlinear site re-
sponse associated with the vertical strong ground motion.
Firstly, model behavior against normal compression and tension was examined by using sin-
gle element. Then response against vertical motion was investigated by using a simple 1D col-
umn mesh with 10 m depth and sinusoidal input motion. The analysis indicated that the mecha-
nism of the asymmetry on vertical surface acceleration might be characterized by the existence
of lower bound of negative acceleration and positive pulse due to compression of soils.
Then, observed ground motion was simulated with model ground of 1D column whose model
parameters were derived from the borehole data at the vertical array, KiK-net, IWTH25 site.
The ground model constituted with 7 layers from the surface to G.L. -246 m which was assumed
to be the base of IWTH25 station. With the base records of NS and UD components as input
motions, 1D site response analysis was conducted. Vertical acceleration at the surface clearly
showed asymmetric form as can be seen in the observed acceleration.

ACKNOWLEDGEMENT
Strong motion data used in this study was provided by the digital strong-motion seismograph
network, KiK-net, National Research Institute for Earth Science and Disaster Prevention
(NIED), Japan.

REFERENCES
Aoi, S., Kunugi, T. and Fujiwara, H. 2008. Trampoline effect in extreme ground motion. Science 322: 727-730.
Bardet, J.P. and Davis, C.A. 1996. Performance of San Fernando dams during the 1994 Northridge Earthquake. Journal of Geotech-
nical Engineering, ASCE 122 (7): 554-564.
Bardet, J.P., Ichii, K. and Lin, C.H. 2000. EERA: A computer program for Equivalent-linear Earthquake site Response Analyses of
layered soil deposits. A report to Department of Civil Engineering, University of Southern California,
http://geoinfo.usc.edu/gees.
Bardet, J.P. and Tobita, T. 2001. NERA: A computer program for Nonlinear Earthquake site Response Analysis of Layered Soil De-
posits. Department of Civil Engineering, University of Southern California, http://geoinfo.usc.edu/gees.
Bonilla, L.F., Archuleta, R.J. and Lavallee, D. 2005. Hysteretic and dilatant behavior of cohesionless soils and their effects on
nonlinear site response: Field data observations and modeling. Bulletin of the Seismological Society of America 95 (6): 2373-
2395.
Dafalias, Y.F. and Popov, E. 1975. A model of nonlinearity hardening materials for complex loading. Acta Mechanica 21 (3): 173-
192.
Frankel, A.D., Carver, D.L. and Williams, R.A. 2002. Nonlinear and linear site response and basin effects in Seattle for the M 6.8
Nisqually, Washington earthquake. Bulletin of the Seismological Society of America 92 2090-2109.
217

13
Hardin, B.O. and Drnevich, V.P. 1972. Shear modulus and damping in soils: design equations and curves. J. Soil Mech. Found. Div.
1: 2-89.
Hashiguchi, K. 1980. Constitutive equation of elastoplastic materials with elasto-plastic transition. Journal of Applied Mechanics,
ASME 102(2): 266-272.
Holzer, T. L., Youd, T. L. and Hanks, T. C. 1989. Dynamic liquefaction during the 1987 Superstition Hills, California, earthquake.
Science 244: 56-59.
Iai, S., Matsunaga, Y. and Kameoka, T. 1992. Strain space plasticity model for cyclic mobility. Soils and Foundations 32(2): 1- 15.
Iai, S., Morita, T., Kameoka, T., Matsunaga, Y. and Abiko, K. 1995. Response of a dense sand deposit during 1993 Kushiro-oki
earthquake. Soils and Foundations 35 (1): 115-131.
Idriss, I.M. and Seed, H.B. 1968. Analysis of ground motions during the 1957 San Francisco earthquake. Bulletin of the Seismologi-
cal Society of America 58: 2013-2032.
Idriss, I.M. 1990. Response of soft soil sites during earthquakes. Proceedings of the H. Bolton Seed Memorial Symposium 2 J. M.
Duncan (Editor), BiTech Publisher, Vancouver: 273-289.
Ishihara, K. 1993. Liquefaction and flow failure during earthquakes. The 33rd Rankine lecture, Geotechnique 43 (3): 351 - 415.
Iwan, W.D. 1967. On a class of models for the yielding behavior of continuous and composite systems. Journal of Applied Mechan-
ics, ASME 34: 612-617.
Iwasaki, Y. and Tai, M. 1996. Strong motion records at Kobe Port Island. Special Issue of Soils and Foundations, Japanese Geo-
technical Society 29-40.
Martin, G.R., Finn, W.D.L. and Seed, H.B. 1975. Fundamentals of liquefaction under cyclic loading. Journal of Geotechnical Engi-
neering, ASCE 101 (5): 423-438.
Masing, G. 1926. Eiganspannungen und verfestigung beim messing. Proceedings of Second International Conference of Applied
Mechanics: 322-335.
Matsuoka, H. 1974. Stress-strain relationship of sands based on the mobilized plane. Soils and Foundations 14 (2): 47-61.
Mroz, Z. 1967. On the description of anisotropic workhardening. Journal of Mechanics and Physics of solids 15: 163-175.
Mroz, Z., Norris, V.A. and Zienkiewicz, O.C. 1978. An anisotropic hardening model for soils and its application to cyclic loading.
International Journal for Numerical and Analytical Methods in Geomechanics 2: 203-221.
Nakai, T. 1986. A generalized elastoplastic constitutive model for clay in a three-dimensional stresses. Soils and Foundations 26 (3):
81-89.
NIED 2009. Digital Strong-Motion Seismograph Network, Kik-net. http://www.k-net.bosai.go.jp/k-net/.
Oka, F., Yashima, A., Tateishi, Y., Taguchi, Y. and Yamashita, S. 1999. A cyclic elasto-plastic constitutive model for sand consider-
ing a plastic-strain dependence of the shear modulus. Geotechnique 49 (5): 661-680.
Ozutsumi, O. 2003. Development of a numerical method for earthquake damage estimation of soil and structure systems on liquefi-
able ground. Doctor's dissertation submitted to the Department of Civil and Earth Resources Engineering, Kyoto University (in
Japanese)
Prevost, J.H. 1985. A simple plasticity theory for frictional cohesionless soils. Soil Dynamics and Earthquake Engineering 4 (1): 9-
17.
Ramberg, W. and Osgood, W.R. 1943. Description of stress-strain curve by three parameters. Tech. Note 902, National Advisory
committee for Aeronautics, Washington, D.C.
Roscoe, K.H., Schofield, A.N. and Thurairajah, A. 1963. Yielding of clay in states wetter than critical. Geotechnique 13 (3): 211-
240.
Roscoe, K. H. and Burland, J. B. 1968. On the generalised stress-strain behaviour of 'wet' clay. Engineering Plasticity, Cambridge
Press 535-609.
Schnabel, P.B., Lysmer, J. and Seed, H.B. 1972. SHAKE-A computer program for earthquake response analysis of horizontally lay-
ered sites. Earthquake Engineering Research Center, College of Engineering, University of California, Berkeley Report No.
EERC 72-12.
Sekiguchi, H. and Ohta, H. 1977. Induced anisotropy and time dependency in clays. Proc. 9th ICSMFE, Speciality Session 9 229-
238.
Towhata, I. and Ishihara, K. 1985. Modelling soil behaviour under principal stress axes rotation. Proceedings of 5th International
Conference on Numerical Methods in Geomechanics, Nagoya 1: 523-530.
Toyota, H. 2009. Personal communication.

218
Effect of Pre-Yielding Elasticity on Sliding Triggered by
Near-Fault Motions Modeled as Idealized Wavelets
E. Garini, G. Gazetas, N. Gerolymos
Soil Mechanics Laboratory, National Technical University of Athens, Greece

ABSTRACT: The influence of elastic pre-yielding on the response of a mass resting on an
inclined plane is investigated in this paper. The ultimate shearing capacity of the interface
obeys Coulombs friction law. The slope is subjected to near-fault triggering by two types of
idealized wavelets: (i) a Ricker wavelet, representative of forward directivity affected motions,
containing strong long-period acceleration pulses, and (ii) an one-cycle sinusoidal wavelet,
representative of fling-affected motions, containing an one-sided velocity pulse with an ensuing
permanent displacement. The asymmetric sliding response is analyzed and the effect of a
number of parameters is explored. They include: the critical acceleration ratio, a
C
/a
H
, the
excitation frequency, f
o
, the changing polarity of excitation, and the magnitude of elastic
pre-yielding displacement, dy.
1 INTRODUCTION
Several applications in geotechnical earthquake engineering require an understanding of the
dynamic sliding response of a block of mass m supported on seismically vibrating base through
an asymmetric frictional contact. n his 1965 seminal Rankine Lecture, Newmark proposed that
the seismic performance of earth dams and embankments be evaluated in terms of permanent
deformations which occur whenever the inertia forces on a potential slide mass are large enough
to overcome the frictional resistance at the failure surface. He proposed the analog of a rigid
block on inclined plane for a simple way of analytically obtaining approximate estimates of
these deformations. Newmarks analog has seen numerous applications and extensions: seismic
deformation analysis of earth dams and embankments, displacements associated with landslides,
seismic deformation of landfills with geosynthetic liners, seismic settlement of surface
foundations, movements of wedges in rock slopes, and even potential sliding of concrete gravity
dams. The extension of the analog by Richards & Elms (1979)

to gravity retaining walls has
met worldwide acceptance, and has found its way into seismic codes of practice. Several other
generalized applications have also appeared.
A numerical study has been recently presented by the authors (Garini et al 2007, Gazetas et
al 2009) for a rigid block supported through a rigid-plastic frictional contact surface on an
inclined plane, and subjected to slope-parallel excitation. The latter was described with near-
fault seismic records strongly influenced by forward-directivity or fling-step effects. Our
study had consistently and repeatedly revealed a profound sensitivity of both maximum and
residual slippage: (i) on the sequence and even the details of the pulses contained in the
excitation, and (ii) on the polarity of shaking. A few of the findings contradicted some of the
prevailing beliefs that have emanated from statistical correlation studies in literature.
However, all these finding were based on the extreme assumption of a perfectly rigid-plastic
interface. Since in most realistic systems some pre-sliding elasticity is unavoidable, this paper
investigates an elastic-plastic support interface. Fig.1 illustrates the problem studied herein.

219

(a)
A
(t)
D
(t)
m
=
f(D
)
T(t)
D(t)
mg(cos sin)
mg(cos+sin)
T(t)
D(t)
mg(cos sin)
mg(cos+sin)
dy

b) (c)
(a)
=
f(D
)
=
f(D
)
D
(t)
D
(t)

Figure 1. (a) The problem studied in the paper sliding on inclined plane undergoing excitation parallel
to the slope, (b) ideally rigid-plastic behavior of the interface as studied by Garini et al (2007), and (c)
elastic-perfectly plastic sliding response studied here.
2 DIRECTIVITY AND FLING IN NEAR-FAULT MOTIONS

Earlier studies of Newmark-sliding were based on records available in the late 1970s and
1980s. Very few of those motions were near-fault records from largemagnitude (M > 6.5)
events. Today, however, such near-fault records are known to often contain either long-period
high-amplitude acceleration pulses, or large residual displacements the outcome,
respectively, either of the coherent arrival of seismic waves when the fault rupture propagates
towards the site, or of tectonic permanent displacement (offset) of the earth in the proximity of
the seismogenic fault rupture. The terms forward-rupture directivity and fling step have
been given to the two phenomena (Singh, 1988; Somerville et al, 1996; Abrahamson, 2000; and
Bolt, 2004).
Fig. 2 illustrates in idealized form some fundamental characteristics of these two types of
near-fault motions. For strikeslip earthquakes, the signature of forward rupture directivity
appears in direction normal to the fault; whereas, the fling step is significant in the parallel
component of motion in close proximity to the fault, especially if the latter emerges on the
surface with a large static offset. The two phenomena (and directivity in particular) have been
the subject of seismological (theoretical and instrumental) as well as earthquake engineering
research.
Two idealized motions (wavelets) are used as excitation in this paper to represent in a
simple way typical directivity and fling affected ground motions. They are the one-cycle
sinusoidal and the Ricker wavelets: the former modeling a typical fling affected motion, and the
latter a directivity affected motion. Four characteristic frequencies are utilized for each of these
motions as follows:
Richer: f
o
= 0.35 Hz, 0.75 Hz, 1.5 Hz, and 3,5 Hz
Sinusoidal: f
o
= 0.57 Hz, 1.25 Hz, 2.5 Hz, 6.67 Hz
(
A
(t)
m
A
(t)
m
T(t)
D(t)
mg(cos sin)
mg(cos+sin)
T(t)
D(t)
mg(cos sin)
mg(cos+sin)
T(t)
D(t)
mg(cos sin)
mg(cos+sin)
dy
T(t)
D(t)
mg(cos sin)
mg(cos+sin)
dy
b) (c) (
220
The response spectra of corresponding pairs of these motions are approximately matched to give
almost the same maximum spectral acceleration at nearly the same dominant period. For
instance, Fig.3 illustrates the spectral matching of two of the pairs.

0
2
4
6
3 5 7 9 11
0
2
4
6
3 5 7 9 11
-0.7
0
0.7
1.4
3 5 7 9 11
-3
0
3
3 5 7 9 11
-10
-5
0
5
10
3 5 7 9 11
D
residual
= 0
Fault
strike
Directivity pulse (Normal component
of Displacement)
Fling step (Parallel component
of Displacement)
Site
(t) :
m/s
2
V(t) :
Fault
strike
of Displacement)
Site

Figure 2. Explanatory sketch of the forward-directivity and fling-step phenomena as reflected in the
displacement records; and examples of simple wavelets bearing the signature of the two effects.
3 RESULTS: THE EFFECT OF PRE-YIELDING ELASTICITY ON SLIDING
Fig.4 depicts the asymmetric sliding response in terms of acceleration, velocity, and
displacement time-histories for a mass of m = 1 Mgr on a slope of inclination = 25
o
. The
system is subjected to a Ricker wavelet of peak acceleration 1 g and central frequency 1.5 Hz.
The right hand-side portrays the time-histories corresponding to a rigid-plastic interface with
critical ratio of a
C
/a
H
= 0.1. Notice that the mass is moving in unison with the base as long as
the critical acceleration a
C
is not exceeded. Whenever base acceleration exceeds the critical
m/s
D(t) :
Fling Step
General Example:
(One-cycle Sinus pulse )
t : s
m
Forward Directivity
General Example:
(Ricker wavelet)
t : s
V 5.6
-10
-5
0
5
10
3 5 7 9 11
V 5.3 m/s
D
maxl
= 1.31 m
D
residual
= 4.9 m
0
2
4
6
3 5 7 9 11
0
2
4
6
3 5 7 9 11
-0.7
0
0.7
1.4
3 5 7 9 11
-3
0
3
3 5 7 9 11
-10
-5
0
5
10
3 5 7 9 11
D
residual
= 0
of Displacement)
Fault
strike
of Displacement)
Site
of Displacement)
(t) :
m/s
2
V(t) :
s m/
D(t) :
Fling Step
General Example:
Fling Step
General Example:
t : s
m
Forward Directivity
General Example:
(Ricker wavelet)
Forward Directivity
General Example:
(Ricker wavelet)
t : s
V 5.6
-10
-5
0
5
10
3 5 7 9 11
V 5.3 m/s
D
maxl
= 1.31 m
D
residual
= 4.9 m
221
acceleration, the block slides either downward (usually) or upward (rarely). As a result, a
residual yielding displacement of 0.36 m occurs.
The left hand-side of Fig.4 illustrates the elastic-plastic response of the same mass-base
system. Here, the block can displace elastically up to 0.1 m before yielding. As a consequence,
a phase difference between input excitation and induced mass response occurs. Notice this
phase shift in Fig.4 in the acceleration and velocity time-histories (see arrows). It is emphasized
that pre-yielding displacement exists in both the upward and downward direction. Thus, even
though no uphill sliding happens after 1 sec in Fig.4, an elastic upward displacement does take
place reversal of accumulated slippage. In addition, the block does not rest after the last
sliding period but continues uphill and downhill elastic oscillations after the end of triggering.
The force-displacement hysteresis loop in Fig.5 shows the two yielding events and the
accompanying elastic branches.

0
1
2
3
0 1 2 3 4
Ricker, f
o
= 0.35 Hz, 1.50 Hz
One-cycle Sinus, f
o
= 0.57 Hz, 1.25 Hz
T : s
S
A
: g
One-cycle Sinus
Ricker
1 g
1 g
1 g
0.6 g
0
1
2
3
0 1 2 3 4
Ricker, f
o
= 0.35 Hz, 1.50 Hz
One-cycle Sinus, f
o
= 0.57 Hz, 1.25 Hz
T : s
S
A
: g
One-cycle Sinus
Ricker
One-cycle Sinus
Ricker
1 g
1 g
1 g
0.6 g
1 g
0.6 g

Figure 3. The two idealized time histories used as base excitation, with their response acceleration spectra
S
A
:g T. (Peak ground acceleration: 1 g).
4 COMPILATION OF RESULTS
All our numerical results with Ricker wavelets as base excitations are compiled in Figs 6 and 7.
Fig.6 depicts the permanent slippage with respect to the critical acceleration ratio, a
C
/a
H
, for
four values of pre-yielding deformation, dy, and four excitation frequencies, f
o
. As expected,
when the acceleration ratio, a
C
/a
H
, increases the induced slippage decreases. This general trend
is independent of the existence or not of the pre-yielding displacement. Notice that the
existence of pre-yielding elasticity may lead to larger or smaller permanent displacements,
depending on f
o
and a
C
/a
H
.
Furthermore, in Fig.7 observe the influence of frequency on sliding displacement. For
frequency, f
o
= 0.35 Hz, as the elastic region dy increases the slippage D also increases. The
222
same is valid for f
o
= 0.75 Hz. However, when the frequency, f
o
, increases up to 1.50 Hz the
response changes. Observe that while slippage becomes greater for pre-yielding region between
0 and 0.01, for larger values of dy the sliding response decreases, in the case of 1.50 Hz. The
behavior is more complicated when the frequency takes the value of 3.50 Hz.

Figure 4. Acceleration, velocity, and displacement time histories for a maximum elastic deformation of
dy = 0.1 m are presented at the left, and for dy = 0 are presented at the right (excitation: Ricker wavelet of
frequency f
o
= 1.5 Hz).
5 EFFECT OF CHANGING POLARITY
The next two Figures (Figs 8 and 9) address a most astonishing effect: that of the reversal in
polarity (i.e., change from + to direction in which the excitation is applied). [This is the same
as having two identical slopes, one opposite to the other (across the street so to speak),
subjected to the same excitation, as sketched at the top of Figs. 8 and 9.]
A few researchers and only in recent years (Kramer & Lindwall, 2004; Fardis et al, 2003)
appear to have published on the importance of the polarity of shaking. This has much to do

0
0.1
0.2
0.3
0.4
0 1 2 3
-10
-5
0
5
10
0 1 2 3
0.14 m
2
nd
slide 0.29 m
1
st
slide
t : s
:
s
2
s
m/
V :
m/
D :
m
Ground
Sliding block
25
o
C
/
= 0.1
dy = 0.1 m
t : s
dy = 0 m
-10
-5
0
5
10
0 1 2 3
-0.8
-0.4
0
0.4
0.8
0 1 2 3
0.36 m
-0.8
-0.4
0
0.4
0.8
0 1 2 3
0
0.1
0.2
0.3
0.4
0 1 2 3
0
0.1
0.2
0.3
0.4
0 1 2 3
-10
-5
0
5
10
0 1 2 3
0.14 m 0.14 m
2
nd
slide 0.29 m
1
st
slide
t : s
:
s
2
s
m/
V :
m/
D :
m
Ground
Sliding block
25
o
25
o
C
/
= 0.1
C
/
= 0.1
dy = 0.1 m dy = 0.1 m
t : s
dy = 0 m dy = 0 m
-10
-5
0
5
10
0 1 2 3
-0.8
-0.4
0
0.4
0.8
0 1 2 3
-0.8
-0.4
0
0.4
0.8
0 1 2 3
0
0.1
0.2
0.3
0.4
0 1 2 3
0.36 m
223

T : kN
-4
-3
-2
-1
0
1
2
0 0.1 0.2 0.3 0.4 0.5
1
st
slide 2
nd
slide
0.08
Initial elastic deformation induced
by the m g sin component
D : m
T : kN
-4
-3
-2
-1
0
1
2
0 0.1 0.2 0.3 0.4 0.5
1
st
slide 2
nd
slide
0.08
Initial elastic deformation induced
by the m g sin component
D : m

Figure 5. Force-displacement response for the case of an elasto-plastic sliding system with dy = 0.1 m,
= 25
o
, a
C
/a
H
= 0.1 and a Ricker excitation of 1.5 Hz frequency. Notice that yielding occurs only in one
direction, as it was expected.

0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
dy = 0
dy = 0.01 m
dy = 0.05 m
dy = 0.1 m
0
3
6
9
12
15
18
21
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
C
/
H
S
l
i
p
p
a
g
e

:

m
1 g
0.6 g
S
l
i
p
p
a
g
e

:

m
C
/
H
25
o
dy = 0
dy = 0.01 m
dy = 0.05 m
dy = 0.1 m
dy = 0
dy = 0.01 m
dy = 0.05 m
dy = 0.1 m
dy = 0
dy = 0.01 m
dy = 0.05 m
dy = 0.1 m
Ricker 0.35 Hz Ricker 0.75 Hz
m
=
1
M
g
r
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
dy = 0
dy = 0.01 m
dy = 0.05 m
dy = 0.1 m
0
3
6
9
12
15
18
21
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
C
/
H
S
l
i
p
p
a
g
e

:

m
1 g
0.6 g
1 g
0.6 g
S
l
i
p
p
a
g
e

:

m
C
/
H
25
o
25
o
25
o
dy = 0
dy = 0.01 m
dy = 0.05 m
dy = 0.1 m
dy = 0
dy = 0.01 m
dy = 0.05 m
dy = 0.1 m
dy = 0
dy = 0.01 m
dy = 0.05 m
dy = 0.1 m
Ricker 0.35 Hz Ricker 0.35 Hz Ricker 0.75 Hz Ricker 0.75 Hz
m
=
1
M
g
r

Figure 6. Influence of maximum elastic deformation, dy, on asymmetric sliding response triggered by
Ricker wavelets of maximum acceleration 1 g and of different frequencies.

224

S
l
i
p
p
a
g
e

:

m
S
l
i
p
p
a
g
e

:

m
dy : m dy : m
0
3
6
9
12
15
18
21
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05
0
1
2
3
4
5
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05
0
0.03
0.06
0.09
0.12
0.15
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05
C
/
H
0.05
0.1
0.02
1 g
0.6 g
25
o
S
l
i
p
p
a
g
e

:

m
S
l
i
p
p
a
g
e

:

m
dy : m dy : m
0
3
6
9
12
15
18
21
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05 0.05
0
1
2
3
4
5
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05
0
1
2
3
4
5
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05
0
0.2
0.4
0.6
0.8
1
1.2
1.4
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05
0
0.03
0.06
0.09
0.12
0.15
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05
0
0.03
0.06
0.09
0.12
0.15
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1 0.05
C
/
H
0.05
0.1
0.02
C
/
H
0.05
0.1
0.02
1 g
0.6 g
1 g
0.6 g
25
o
25
o
25
o
Figure 7. Effect of characteristic frequency, f
o
, and of
C
/
ratio on maximum slippage with respect to

elastic deformation, dy (Excitation: single Ricker wavelet).

with the asymmetry of recorded motions, which is what accentuates the importance of polarity.
It is mainly the nearfault strong motions which are highly asymmetric due to the contained
directivity and fling pulses. But few such motions had been recorded worldwide twenty
years ago. Now a large number has become available.
The sliding analysis of Fig. 8 is simple but most revealing. For a steep slope ( = 25
) and a
yield acceleration ratio a
C
/a
H
= 0.2, we notice the following : When the first sinusoidal
acceleration half-pulse is downward [as in Fig. 8(a) on the right] the block remains almost
attached to the base. Only a mere 0.18 m uphill displacement takes place (see the upward
slippage region enclosed by the dotted lines, starting at 1 sec and ending at 1.8 sec). Even in
this small 0.18 m deformation, the yielding part is particularly smaller than the elastic. The
subsequent, second (and last), upward half-pulse acceleration of the base initiates an uninhibited
downslope slippage of the block, which lasts for a long time after the excitation has terminated
t 2.4 sec on t 5.2 sec. The result is a huge 7.70 m.
In stark contrast, when the first sinusoidal acceleration half-pulse of the base is upward [as in
Fig. 8(b) at the left] the block starts sliding downslope almost immediately. But it soon comes
to a stop after about 1.4 seconds, as the upward base motion decelerates and then reverses. The
resulting residual slip is only 3.14 m, almost 2.5 times smaller than the 7.70 m produced with
the reverse motion!
This effect of reversing the polarity of shaking is of profound importance, especially with
fling type motions (as the sinus pulse studied above). It may not however be as dramatic with
225
directivity affected motions if they contain several competing cycles of pulses, as seen in
Fig. 9 with the Ricker wavelet.

-12
-10
-8
-6
-4
-2
0
2
4
-1 0 1 2 3 4 5 6 7 8
0
1
2
3
4
5
6
7
8
0 1 2 3 4 5 6
-12
-10
-8
-6
-4
-2
0
2
4
-1 0 1 2 3 4 5 6 7 8
A
(
t
)

:

m
/
s
2
t : s
-15
-10
-5
0
5
10
15
0 1 2 3 4 5 6
D
(
t
)

:

m
T

:

k
N
D : m
25
o
25
o
+
-15
-10
-5
0
5
10
15
0 1 2 3 4 5 6
-1
0
1
2
3
4
5
6
7
8
0 1 2 3 4 5 6
t : s
D : m
Ground
Sliding block
C
/
= 0.2
3.14 m
7.70 m
(b) (a)
-12
-10
-8
-6
-4
-2
0
2
4
-1 0 1 2 3 4 5 6 7 8
0
1
2
3
4
5
6
7
8
0 1 2 3 4 5 6
-12
-10
-8
-6
-4
-2
0
2
4
-1 0 1 2 3 4 5 6 7 8
A
(
t
)

:

m
/
s
2
t : s
-15
-10
-5
0
5
10
15
0 1 2 3 4 5 6
D
(
t
)

:

m
T

:

k
N
D : m
25
o
25
o
+
25
o
25
o
+
-15
-10
-5
0
5
10
15
0 1 2 3 4 5 6
-1
0
1
2
3
4
5
6
7
8
0 1 2 3 4 5 6
t : s
D : m
Ground
Sliding block
C
/
= 0.2
C
/
= 0.2
3.14 m
7.70 m
(b) (a)

Figure 8. Acceleration and displacement time histories for a stiff elasto-plastic system with yield
displacement dy = 0.05 m. The third row of figures illustrates the forcedisplacement response.
(Excitation: one-cycle Sinus of 0.57 Hz frequency).
6 CONCLUSIONS
The solutions portrayed graphically in the paper (through the acceleration and velocity time
histories of the base and mass) are easy to understand, offering considerable insight into the
dynamics of asymmetric sliding when elastic pre-yielding is taken into account. The effects
of elastoplastic yielding on the final accumulated displacement are: (a) the time shift in the
226
sliding time-histories of the mass, and (b) the increase of uphill deformation for accelerations
smaller than the critical because of elasticity.
Forwarddirectivity and flingstep affected motions, containing severe acceleration pulses
and/or large velocity steps, may have an unpredictably-detrimental effect on residual slip,
especially for small values of the critical acceleration. The unpredictability of asymmetric
response arises from the sensitivity of the sliding on the sequence, duration, and details of
motion.
Changing the polarity of excitation (i.e., applying it in the + and then in the direction) has a
significant effect on the accumulated slippage.

25
o
25
o
+
0
9
18
27
36
45
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
C
/
H
+

Figure 9. The impact of alternating excitation polarity on the induced yielding for both the Ricker and
one-cycle sinus pulses. Three different values of elastic deformation, dy, are presented.
7 ACKNOWLEDGEMENT
The research presented in this paper was financially supported by the Secretariat for Research
and Technology of Greece, under the auspices of PENED Programme with Contract number
03ED278. We thank Dr. Ioannis Anastasopoulos for his thoughtful help in programming
aspects of this study.
REFERENCES
Abrahamson, N. A. (2000), Effects of rupture directivity on probabilistic seismic hazard analysis, Pro-
ceedings, 6
th
International Conference on Seismic Zonation, Palm Springs, California, Earthquake En-
gineering Research Institute.
Bolt, B.A. (2004), Earthquakes, W. H. Freeman & Co, Fifth Edition, New York.
Boore, M. D. (2001), Effect of baseline corrections on displacements and response spectra for several Re-
cordings of the 1999 Chi-Chi, Taiwan, Earthquake, Bulletin of the Seismological Society of America,
Vol. 91, No. 5, pp. 1199-1211.
Bray, J.D., and Rathje, E.M. (1998), Earthquake-Induced Displacements of Solid-Waste Landfills, Jour-
nal of Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 124, pp. 242-253.
Constantinou, M. C., and Gazetas, G. (1987), Probabilistic Seismic Sliding Deformations of Earth Dams
and Slopes, Proceedings of the Specialty Conference on Probabilistic Mechanics and Structural Reli-
ability, ASCE, Berkeley, pp. 318-321.
S
l
i
p
p
a
g
e

:

m
C
/
H

C
/
H
f
o
= 0.57 Hz,
f
o
= 0.57 Hz,
f
o
= 0.35 Hz,
f
o
= 0.35 Hz,
One-Cycle Sinus
Ricker
}
}
+
+
dy = 0.01 m dy = 0.05 m
dy = 0.1 m
25
o
25
o
+
25
o
25
o
+
0
9
18
27
36
45
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
C
/
H

C
/
H

C
/
H
S
l
i
p
p
a
g
e

:

m
0
9
18
27
36
45
0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8
C
/
H

C
/
H

C
/
H
f
o
= 0.57 Hz,
f
o
= 0.57 Hz,
f
o
= 0.35 Hz,
f
o
= 0.35 Hz,
One-Cycle Sinus
Ricker
}
}
+
S
l
i
p
p
a
g
e

:

m
+
f
o
= 0.57 Hz,
f
o
= 0.57 Hz,
f
o
= 0.35 Hz,
f
o
= 0.35 Hz,
One-Cycle Sinus
Ricker
}
}
++
++
dy = 0.05 m dy = 0.05 m dy = 0.01 m dy = 0.01 m

dy = 0.1 m dy = 0.1 m
227
Fardis, N., Georgarakos, P., Gazetas, G., and Anastasopoulos, I. (2003), Sliding Isolation of Structures:
Effect of Horizontal and Vertical Acceleration, Proceedings of the FIB International Symposium on
Concrete Structures in Seismic Regions, Athens, Greece (in CD-Rom).
Garini E. and Gazetas G. (2007), Sliding of Rigid Block on Sloping Plane: The Surprising Role of the
Sequence of Long-Duration Pulses, Proceedings of 2
nd
Japan-Greece Workshop on Seismic Design,
Observation, and Retrofit of Foundations, Tokyo, Japan, pp. 79-104.
Gazetas, G, Garini, E., and Anastasopoulos, I. (2009),

Effects of NearFault Ground Shaking on Sliding
Systems, Journal of Geotechnical and Geoenviromental Engineering, ASCE (in press).
Gazetas, G. (1996), Soil Dynamics and Earthquake Engineering: Case Histories. Symeon Publishers,
Athens, Greece; Chapter 5: The Aegion 1995 Earthquake, pp. 269-317.
Gazetas, G., and Uddin, N. (1994), Permanent deformation on pre-existing sliding surfaces in dams,
Journal of Geotechnical Engineering, ASCE, Vol.120, No.11, pp. 2041 2061.
Makdisi, F.I., and Seed, H.B. (1978), Simplified Procedure For Estimating Dam And Embankment Earth-
quake Induced Deformations, Journal of Geotechnical Engineering Division, ASCE, Vol. 104, 849
867.
Makris, N., and Psychogios T. (2006), Dimensional response analysis of yielding structures with first
mode dominated response, Earthquake Engineering Structural Dynamics, Vol.35, pp. 1203-1224.
Makris, N., and Roussos, Y.S. (2000), Rocking Response of Rigid Blocks Under Near-Source Ground
Motions, Gotechnique, 50, No. 3, pp. 243-262.
Newmark, N. M. (1965), Effects of Earthquakes on Dams and Embankments, Gotechnique, Vol. 15, No.
2, pp. 139-160.
Richards, R., and Elms, D.G. (1979), Seismic Behaviour of Gravity Retaining Walls, Journal of Geotech-
nical Engineering Division, ASCE, Vol. 105, pp. 449-464.
Singh, J.P. (1988), Earthquake Ground Motions: Implications for Designing Structures and Reconciling
Structural Damage, Earthquake Spectra, Vol. 1, pp. 239-270.
Somerville, P.G., Saikia, C., Wald, D., and Graves, R. (1996), Implications of the Northridge earthquake
for strong ground motions from thrust faults,
Bulletin of the Seismological Society of America, Vol. 86, pp. 115 - 125.
Yegian, M. K., Harb, J. N., and Kadakal, U. (1998), Dynamic Response Analysis Procedure for Landfills
and Geosynthetic Liners, Journal of Geotechnical and Geoenvironmental Engineering, ASCE, Vol.
124 (10), pp. 1027-1033.

228
1. THE PROBLEM: CODE SPECTRA CONTRADICT REALITY
It is well known how important soil effects are on the intensity and frequency content of ground
motions. In civil engineering practice these effects are often computed theoretically (wave
propagation analysis assuming equivalent-linear or nonlinear soil behavior). Yet, the seismic
codes have universally faced the problem of soil amplification in a purely empirical and (un-
avoidably) oversimplified way :
The soil deposits were classified in a few broad categories, each of which encompasses a
wide range of soil layer stiffnesses and thickness down to bedrock.
The response spectra S
a
(T) from numerous world wide accelerograms recorded on top of
soils belonging to each category, were statistically processed. The shape of the design
spectrum for the particular soil category was based on the average of the normalized
spectrum, S
a
(T)/A, for each period T, after some conservative smoothening.
The design spectra that have thus resulted share a crucial characteristic : the more flexible a
soil deposit (i.e. the smaller its stiffness and/or the larger its thickness), the flatter the design
spectrum.
Yet, reality has repeatedly shown the opposite trend ! Numerous records in soft soils have
produced response spectra of a sharp rather than flat shape, with well defined peaks around the
site fundamental period. Fig. 1 highlights the discrepancy between seismic codes and reality.
The Irrationality of Current Seismic Code Spectra for Soft Soils :
Proposed Remedy

A. Ziotopoulou
#
, G. Gazetas
Soil Mechanics Laboratory, National Technical University of Athens, Greece

#
presently, University of California at Davis
ABSTRACT: Seismic codes have universally adopted smooth design acceleration spectra, on
the basis of averaging of a large number of elastic response spectra of actual recordings. Such
spectra have, for each soil category, an essentially constant acceleration plateau, S
a
, usually
equal to 2.5A, followed by a descending acceleration branch. The period range of the constant-
acceleration plateau is larger for softer soils. However, such a flat shape of spectra has little re-
semblance to an actual soil-amplified spectrum. The unrealistic shape stems basically from the
fact that the spectra of motions recorded on soft soils belonging to one soil category attain their
maxima at different well-separated periods ; thereby, averaging them eliminates their peaks and
leads to a (spurious) flat spectrum. Through an extensive analytical parametric study we demon-
strate that by normalizing the period of the Spectra by the predominant period of motion, and
then averaging, results in a bi-normalized spectrum (S
a
/A : T/T
p
)which has a sharp peak at T/T
p

= 1. It is found out that this spectrum has a peak of S
a
/A 3.75 (rather than 2.5), for a narrow
range of normalized periods. The effect of such a spectrum especially on SSI studies may be
drastically different from the effect of a (conventional) code spectrum

229
The consequences of such a disparity,, especially on SSI systems may be significantly detrimen-
tal.

2. WHY THIS DISCREPANCY ?
As illustrated in the sketch of Fig. 2, the culprit is the averaging of dissimilar response spectra ;
its accomplish : the very broad range of stiffness and thickness of each soil category. A range
of natural periods in the ratio of 1 to 4 is quite possible within one single category, say category
D (according to NEHRP). The actual seismic motions in a number of (soft) soil profiles be-
longing to category D but with so vastly different fundamental periods are likely to have re-
sponse spectra with sharp peaks at well-separated periods. Thus, at the period for which one
spectrum has a peak the spectra on sites with different periods are likely to have very small val-
ues. Hence, by averaging all these different values we simply annihilate the real sharp peaks.
In other words, spuriously and against safety, we disregard (or rather depress) the reso-
nance between soil deposit and excitation!
The topic has already been brought to light by Mylonakis & Gazetas (2000) and Gazetas
(2006), in an attempt to reevaluate the importance of soil-structure interaction (SSI). More re-
cently, Xu & Xie (2004) along similar lines developed a unique average bi-normalized spectrum
for 206 strong-motion records of the Chi-Chi (1999) earthquake. Each and every individual ac-
celeration response spectrum was doubly normalized : the ordinate, S
a
, with respect to the peak
ground acceleration, A; the abscissa, T, with respect to the predominant period T
p
of the spec-
trum. The average of the individual S
a
/A : T/T
p
spectra exhibited indeed a sharp peak, at T/T
p
= 1, with a maximum value of the order of 4, rather than the 2.5 of the code spectra. The practi-
cal indirect conclusion from the above studies was that the increase of the period of a structure-
soil system with decreasing soil stiffness would not necessarily lead to reduced intensity of
shaking, as presently implied by the code spectra.

3. SUMMARY OF THE ANALYTICAL (REMEDIAL) STUDY
In contrast with the purely empirical method with which the Code Spectra have been developed,
we follow an analytical methodology which comprises the following steps :
For a particular soil category (for example C according to EC8, or D according to NEHRP)
we construct a number of idealized generic soil profiles having the following characteris-
tic parameters:
velocity :
30 , S
V = 180 m/s, 260 m/s, 360 m/s
30 , S
V = average shear wave velocity from the
ground surface down to a depth of 30 m
distribution of V
s
with depth : uniform, trapezoidal, with-crust (see Fig. 3)
depth to rock : H = 30 m and 60 m.
rock to soil wave velocity ratio : V
S, ROCK
/ V
S,30
= 1.5 and 5
Seven accelerograms recorded on rock are utilized as (rock-outcrop) excitation after be-
ing scaled (up or down) to achieve peak ground acceleration : A = 0.20 g, 0.40 g, 0.60 g.
The names and earthquakes of these records are :
Stone Canyon Reservoir, Northridge 1994
Aegion-Rock, Aegion 1995
Sakarya, Izmit 1999
Dayhook, Tabas 1978
Gilroy-1, Loma Prieta 1989
Lucerne, Landers 1992
Superstition Mountain, Imperial Valley 1979.
Exciting all the aforesaid soil profiles with each record in all possible combinations results
in 1009 cases. They are analysed, first, with the well-known equivalent-linear method of
Schnabel et al, 1972(SHAKE) and, second, with the inelastic method (NL-DYAS) introduced
by Gerolymos & Gazetas, 2005.
230

Fig. 1: The discrepancy between a Code Design Spectrum typical for soft soils and the response
spectra of two actual soil amplified motions.

Fig. 2: Sketch illustrating the derivation of code spectra from the average (for each specific period)
of the S
a
/A values of all recorded spectra. The three individual idealized spectra are from possible
motions in three soft soil profiles, all belonging to the same Soil Class (Category), and all bearing the
effects of resonance but at different periods. The resulting spectrum spuriously suppresses the
soilexcitation resonance.
T ( sec )
0
1
4
5
6
0 0.5 1.5 2.5 1 2
Mexico (1985)
SCT
Kobe (1995)
Takatori
CODE
Spectrum
S
a
/ A
2
3
Damping= 5%
T ( sec )
0
1
4
5
6
0 0.5 1.5 2.5 1 2
Mexico (1985)
SCT
Kobe (1995)
Takatori
CODE
Spectrum
S
a
/ A
2
3
Damping= 5%
T
1
2
3
4
Response Spectra of Possible
Motions on 3 Soil Profiles
2.5
S
a
/ A
T
1
2
3
4
Response Spectra of Possible
Motions on 3 Soil Profiles
2.5
S
a
/ A
231

The response spectra of the ground surface motions resulting from each of the 2x1009
analyses are utilized in two different ways :

(a) We normalize only the spectral accelerations by dividing with the corresponding peak
ground acceleration, S
a
/A --- the established conventional normalization used for deriv-
ing the current design spectra (S
a
/A :T).
(b) We normalize both the spectral acceleration, S
a
/A, and the period, T, by dividing it with
the predominant period T
p
of the ground surface motion. We call the plot S
a
/A : T/T
p
Bi-
Normalized Spectrum (BNS).

The average for each period T of the 1009 simply normalized spectral values (type (a)) give
a mean response spectrum (S
a
/A : T) which is expected to be quite similar with the current
code spectrum for this soil category.
The average for each period ratio T/T
p
of the 1009 doubly normalized spectra (type (b))
give a mean response spectrum (S
a
/A : T/T
p
) which is expected to differ both in shape and in
amplitude from the conventional spectrum.
4. RESULTS: TOWARDS A MORE RATIONAL SPECTRUM
All the 1009 response spectra obtained with the equivalent-linear soil response analyses and
simply or doubly normalized as afore-explained, are portrayed in Fig. 5(a) and 5(b) respec-
tively. Their average response spectra, after some conservative smoothening could serve as
the design spectra. The following conclusions emerge from the two figures :
(a) Regarding the conventionally derived spectrum as anticipated, its shape is indeed quite
similar with the smooth shape of the code spectrum for this soil category : a nearly constant
ordinate, approaching (from below) S
a
/A 2.5, for the range of periods from 0.15 sec to
0.60 sec, approximately. (Of course, if more excitations had been employed , and additional
and more realistic soil profiles had been considered, the period range of nearly constant S
a

would have likely increased, and the spectrum would have been even smoother.)
(b) Regarding the Bi-Normalized Spectrum its shape is vastly different from the conven-
tional spectrum : a sharp peak at T/T
p
1 dominates. Its maximum value, max (S
a
/A),
reaches 3.75, i.e. it is 50% greater than the peak value of the conventional spectrum.
Evidently, the (true or pseudo) resonance between soil and excitation is well preserved only in
the bi-normalized spectrum. The conventional Spectrum does not reflect the physics of the
problem, while being unsafe for many structures (with T T
p
) and leading to erroneous conclu-
sions on the possible effects of soil-structure interaction.

5. THE UNIQUENESS OF THE BI-NORMALIZED SPECTRUM
Several interesting attributes of the Bi-Normalized Spectrum (BNS) have been demonstrated
analytically by Ziotopoulou and Gazetas (2009). Specifically,:
The BNS is hardly influenced by soil category, i.e., it is practically the same for all soil
categories! The same conclusion was drawn by Xu & Xie (2004) for the strong records of
the Chi-Chi (1999) earthquake. (Of course, T
p
may change significantly from soil to soil,
decreasing with soil stiffness ; and moreover, it is often affected by the nature of seismic
excitation. Its estimation is a totally different ball game.)
232
The BNS is only marginally influenced by the nature of the performed wave propagation
analysis : equivalent-linear and truly nonlinear analyses differ appreciably only in the
low-period range (T/T
p
< 0.5), not in the basic shape of the spectrum.
The BNS is only marginally influenced by the nature of seismic excitation. (Of course,
again, the above argument does not extend to T
p
which is partially controlled by the
dominant excitation periods.)
Indicative of the uniqueness of this BNS is Fig. 6, which reveals the practical independence of
BNS from the soil category contrary to the behavior of the conventional spectrum.

Fig. 3: The three types of generic soil profiles used in our parametric investigation

Fig. 4 Historic rock accelerograms used as excitation in our analyses (shown here scaled to
0.40 g).

z

(

m

)
V
s
V
S,30
V
S,30
V
S,30 V
s
V
s
V
S ( z = H )
V
R
V
R
V
R
z

(

m

)
V
s
V
S,30
V
S,30
V
S,30 V
s
V
s
V
S ( z = H )
V
R
V
R
V
R
Aegion Sakarya
Dayhook
Superstition Mtn.
Gilroy Stone Canyon
Lucerne
0.4
0.4
0
(g)
Aegion Sakarya
Dayhook
Superstition Mtn.
Gilroy Stone Canyon
Lucerne
0.4
0.4
0
(g)
233

Fig. 5: Compilation of response spectra of ground surface motions from all the equivalent-linear
analyses. (a) Conventionally normalized spectra ; (b) Bi-Normalized spectra. The thick curves are
the mean response spectra.

7. CONCLUSION
One unique Bi-Normalized Spectrum (BNS), for all soil categories and most likely seismic exci-
tations, emerged from the comprehensive set of wave-propagation analyses reported in this arti-
cle. This unique spectrum is sketched in Fig. 7 and is approximated with the following algebraic
expressions :

S
a
/A = exp (1.35 [T / T
p
]

) for T/T
p
< 1

S
a
/A = 3.75 ( T/T
p
)
1.2
for T/T
p
1

The potential benefits from adopting this simple spectrum have been highlighted in the article.
However, the imprecise definition of T
p
and the difficulty in predicting it in reality remain the
main obstacles in adopting it at present.

T (sec)
H=60m
BiNormalizedSpectra
1
S
a
/A
T / T
P
Bi-Normalized Spectra
S
a
/A
3.75
T (sec)
H=60m
BiNormalizedSpectra
1
S
a
/A
T / T
P
Bi-Normalized Spectra
S
a
/A
3.75
234

Fig. 6: Comparison of the mean spectra of Figure 5 (Soil: V
S,30
= 180 360 m/s) , with the mean
spectra for a much softer soil (V
S,30
= 100 m/s). The differences of the conventionally normalized
spectrum (top) almost disappear in the normalized spectra (bottom)

T ( sec )
C
C
C
S
a
/ A
S
a
/ A
T / T
p
T ( sec )
C
C
C
S
a
/ A
S
a
/ A
T / T
p
S
a
/ A
T /
S
a
/ A
T /
235
Fig. 7: Mean Bi-Normalized Spectra (BNS) from the equivalent linear wave propagation (SHAKE)
and from the inelastic wave propagation (NL-DYAS) studies, and the idealized smooth spectrum
proposed for design. The algebraic expressions for two branches of this spectrum are given in the text.

ACKNOWLEDGMENTS
th
Research Councils (ERC) Programme Ideas, Support for Frontier Research Advanced
Grant, under Contract number ERC-2008-AdG 228254-DARE.
REFERENCES and RELATED BIBLIOGRAPHY
Afra H. and Pecker A. (2001), Calculation of free field response spectrum of a non-homogeneous soil
deposit from bed rock response spectrum, Soil Dynamics and Earthquake Engineering, 22, 157-165.
Bachman R and Bonneville D (2000), The seismic provisions of the 1997 Uniform Building Code, Earth-
quake Spectra, 16(1), pp. 85-100.
Bouckovalas G. D., Papadimitriou A.G., Karamitros D. (2006), Compatibility of EC8 ground types and
site effects with 1-D wave propagation theory, Geotechnical Evaluation and Application of EC8, 53-
67.
Borcherdt RD (1994), Estimates of site-dependent response spectra for design (methodology and justifi-
cation), Earthquake Spectra 10 (4): 617-53.
Building Seismic Safety Council (BSSC), The 2003 NEHRP recommended provisions for new buildings
and other structures. Part 1: Provisions, Federal Emergency Agency (FEMA 450), Washington D.C.
Dobry R, Borcherdt RD, Crouse CB, Idriss IM, Joyner WB, Martin GR, Power MS, Rinne EE and Seed
RB (2000), New site coefficients and site classification system used in recent building seismic code
provisions, Earthquake Spectra , 16(1), 41-67.
Drosos V, Gerolymos N and Gazetas G (2007), Calibration and verification of nonlinear wave propaga-
tion method, Proceedings of the 4
th
Int. Conference on Earthquake Geotechnical Engineering, K. Piti-
lakis, editor, Paper No. 1594.
EC-8 (2001), Part 1 General Rules, seismic actions and rules for buildings, Eurocode 8, Draft 4 (prEN
1998), European Committee for Standardization.
Gazetas G (1982), Vibrational characteristics of soil deposits with variable wave velocity, Numerical and
Analytical Methods in Geomechanics, 1982, 6 (1), 1-20.
Gazetas G (2006), Seismic design of foundations and soil-structure interaction, 1
st
European Conference
on Earthquake Engineering and Seismology, Geneva, Switzerland, 3-8 September.
Gazetas G and Bianchini G (1979), Field evaluation of body and surface-wave soil-amplification theo-
ries, Proceedings of the 2
nd
U.S. National Conference on Earthquake Engineering, EERI, 603-612.
Gerolymos N and Gazetas G (2005), Constitutive model for 1-D cyclic soil behavior applied to seismic
analysis of layered deposits, Soils and Foundations, JGS, 147-159..
Hartzell S, Bonilla LF, and Williams RA (2004), Prediction of nonlinear soil effects, Bull. Seism. Soc.
Am. 94, 1609-1629.
Holmes WT (2000), The 1997 NEHRP Provisions for seismic regulations for new buildings and other
structures, Earthquake Spectra, 16 (1), pp. 101-114.
Housner GW (1959), Behavior of structures during Earthquakes, Proceedings. ASCE, 85, October.
Idriss IM (1990), Response of soft soils during earthquakes, Proceedings of H. Bolton Seed memorial
symposium, Vol. 2, J. Michael Duncan (ed.), Bitech, Vancouver.
Mylonakis G and Gazetas G (2000), Seismic soil-structure interaction: beneficial or detrimental?, Journal
of Earthquake Engineering, 4 (3), 277-301.
NEHRP (2000), Recommended provisions for seismic regulations for new buildings and other structures:
Part 1, provisions, FEMA 368, Building Seismic Safety Council (BSSC), Washington D.C.
Pitilakis K. (2004), Site effects, recent advances in earthquake geotechnical engineering and microzona-
tion, A. Ansal editor, Kluwer Acacemic Publishers, 139-193.
Pitilakis K., Gazepis C., Anastasiadis A. (2006), Design response spectra and soil classification for seis-
mic code provisions. Geotechnical Evaluation and Application of EC8, G. Bouckovalas editor, 37-52.
236
Rathje EM, Abrahamson NA and Bray JD (1998), Simplified frequency content estimates of earthquake
ground motions, Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 124(2) 150-
159..
Rodriguez-Marek A, Bray JD and Abrahamson NA (1999), Task 3: Characterization of site response
general site categories, Pacific Earthquake Engineering Research Center (PEER) Report 1999/03.
Schnabel PB, Lysmer J and Seed HB (1972), SHAKE: A computer program for earthquake response
analysis of horizontally layered sites, Rep. EERC 72-12, University of California, Berkeley.
Schnabel PB, Seed HB and Lysmer J (1972), Modification of seismological records for effects of local
soil conditions, Bulletin of the Seismological Society of America, 62, 1649-1664.
Seed HB, Ugas C and Lysmer J (1976), Site-dependent spectra for earthquake-resistant design, Bulletin
of the Seismological Society of America, 66, 221243.
Somerville P (1998), Emerging Art: Earthquake ground motion, geotechnical special publication, 75
Geotechnical Eng. & Soil Dynamics III, ASCE, Dakoulas P, Yegian M and Holtz RD editors.
Tazoh T., Sato M., Shimizu K. and Koyama K. (1988), Nonlinear seismic response analysis of horizon-
tally layered soil deposits, Proceedings of Ninth World Conference on Earthquake Engineering, To-
kyo-Kyoto, II, 495-501.
Tazoh T., Shimizu K., Sato M., Hirosse T. and Kouama K., (1988), Identification of time-varying pre-
dominant period and damping constant in nonlinear seismic response of soft soil deposit, Proceedings
of Ninth World Conference on Earthquake Engineering, Tokyo-Kyoto, II, 483-488.
Towhata, I. (1996), Seismic wave propagation in elastic soil with continuous variation of shear modulus
in the vertical direction, Soils and Foundations, JGS, 36(1), 61-72.
Travasarou Th. and Gazetas G., (2004), On the linear seismic response of soils with modulus varying as a
power of depth the Maliakos marine clay, Soils and Foundations, JGS, 44(5), 85-93.
Kramer SL, Geotechnical Earthquake Engineering, Prentice Hall, 1996.
Li S. and Xie L (2007), Effect of hanging wall and forward directivity in 1999 Chi-Chi earthquake on
inelastic displacement response of structures., Earthquake Engineering and Engineering Vibration,
6(1) , 77-84.
Stewart JP, Liu AH, Choi Y and Baturai MB (2001), Amplification factors for spectral acceleration in ac-
tive regions, Pacific Earthquake Engineering Research Center (PEER) Report 2001/10.
Vucetic M and Dobry R (1991), Effect of soil plasticity on cyclic response, J. Geotech. Eng. ASCE 117,
89-107.
Xu L and Xie L (2004), Bi-Normalized response spectral characteristics of the 1999 Chi-Chi Earthquake,
Earthquake Engineering and Engineering Vibration , 3 (2), 47-155

237
3.
SoilStructureInteraction
1 INTRODUCTION
The Building Standard Law in Japan and its related enforcement and notices were revised for
the direction to the performance-based design in 1998. The calculation method of response and
limit strength was provided for checking structural serviceability and safety of buildings (Mido-
rikawa et al. 2000, Kuramoto et al. 2000, Kuramoto et al. 2002). The soil structure interaction
(SSI) effects should be considered when the effect will be not negligible. The appropriate me-
thod for the structural safety of buildings during severe earthquake needs to be proposed.
A method for incorporating SSI and calculating the equivalent period and damping effect is
presented. To obtain relationships between force and displacement of the building with a pile
foundation and surrounding soil, a pushover analysis is conducted (Watanabe et al. 2004). To
get the damping effect of the building with SSI, a radiation damping through piles is considered
(AIJ 2006). Based on these results, the building, pile foundation and surrounding soil are re-
placed with an equivalent single degree of freedom (ESDOF) model. The spectrum-based
method is applied to earthquake responses of a residence building with a span in short direction.
The results by the spectrum-based method are compared with those by the time history analy-
sis of the SSI model. Through comparison, the applicability of the method is discussed.
Calculation of seismic response of building based on pushover
analysis of SSI model

M. Iiba
Building Research Institute, Tsukuba, Japan
Y. Umemura
Ando Corporation, Minato-ku, Tokyo, Japan
O. Kurimoto
Obayashi Corporation, Kiyose, Tokyo, Japan
T. Akita
Chiba University, Inage-ku, Chiba, Japan
M. Teshigawara
Nagoya University, Chikusa-ku, Nagoya, Japan
K. Watanabe
Urban Renaissance Agency, Shinjyuku-ku, Tokyo, Japan

ABSTRACT: The soil structure interaction (SSI) effect should be considered when the SSI ef-
fect would be not negligible. The appropriate method for incorporating the SSI effect to res-
ponses of structure to evaluate structural safety needs to be proposed. To obtain relationships
between force and displacement of a building with a pile foundation and surrounding soil, a pu-
shover analysis is conducted. To evaluate a damping effect of the building with SSI, a radiation
damping through piles is considered. The building, pile foundation and soil system are replaced
with an equivalent single degree of freedom system and responses of the system are calculated
based on the acceleration spectrum-based method. The responses of the building are compared
with those by the time history analysis. The responses almost agree with those by time history
analysis and the proposed method is demonstrated to have enough accuracy for a practical use.
238
2 EVALUATION PROCEDURE
The evaluation procedure involves the application of the equivalent linearization technique us-
ing the ESDOF system and the response spectrum analysis (Kuramoto et al. 2000), as shown in
Figure 1. The building is replaced with the ESDOF system with equivalent period and damping
factor. This is based on the result of the nonlinear pushover analysis with the distribution in
proportion to the first mode of the building vibration (Kuramoto et al. 2002).

Transition Curve
Demand Spectra
Capacity Spectrum
for Inelastic (h
eq
=7%, 10%,15%)
for Elastic (h
eq
=5%)
T
5%
T
15%
S
d
S
a
h
eq
15%
Performance Point (Response)
h
eq
at Performance Point
5%
T
7%
T
10%
7%
10%
Limit Value

Figure 1. Illustration of seismic response evaluation procedure based on response spectrum.

The flow for response evaluation of buildings during earthquake is shown in Figure 2. The
procedure consists of several steps as follows;
a) The ARS at ground surface is set. The acceleration (S
a
) and displacement response spectra
(S
d
) are drawn up (Demand Spectrum in Fig. 1).
b) The nonlinear pushover analysis for the model with a building, a foundation and a surround-
ing soil as shown in Figure 3. (Watanabe et al. 2004)
c) The force-displacement relationship of the ESDOF system (Capacity Spectrum in Fig. 1). is
established based on the results by the nonlinear pushover analysis.
d) Equivalent damping coefficient based on dissipation energy is calculated.
e) Based on the equivalent period and equivalent damping factor, the response of the ESDOF
system is obtained (Performance Point in Fig. 1).

The earthquake motion is given as the ARS at the outcropped engineering bedrock whose
shear wave velocity is about 400 m/s. The time histories of earthquake are fitted to the ARS
with random phases. By using equivalent linear analysis (Program SHAKE), the time history
responses on the ground surface whose average ARSs are calculated. A distribution of external
horizontal force to the superstructure is obtained through the time history analysis of the SSI
model. In the transverse direction, the distribution of horizontal force is the combination of a
uniform and a triangular distribution through height (Umemura et al. 2004).
The building is replaced with an ESDOF system as shown in Figure 4 (Kuramoto et al. 2002).
The force-displacement relationship of the ESDOF system is given by equations (1) and (2),
when the force corresponds to the base shear (
B 1
Q ), and its displacement (
1
) corresponds to
the displacement at an equivalent height (
e
H ) where the modal participation function is equal to
1.0 ( { } 0 . 1
1 1
= u )
239
a 1
1
B 1
S M Q = (1)
d 1 1
S = (2)
where, M 1 = equivalent mass corresponding to the 1st mode, and
a 1
S and
d 1
S = accelera-
tion response and displacement response for the 1st mode, respectively.

START
Building, ground and earthquake motion, etc. conditions
Response acceleration spectrum
(considering amplification
properties of surface ground)
Response of each story
Pushover analysis of building,
foundation and soil model
ESDOF system (force
vs. displacement,
equivalent period
Damping coefficient and equivalent
damping factor of ESDOF
Response of ESDOF (shear force and displacement)
END

Figure 2. Flow for response evaluation.

Column
Girder
Footing
girder
Shear wall
Horizontal
Soil spring
Seismic
horizontal
force
Pile end bearing spring
Pile
Pile
peripheral
friction
spring

Figure 3. Overall model for pushover analysis.

240

K
inel
1
SDOF System
Q
B
Structure Model
1
Q
B
H
e
Force Vector and
Displaced Mode
h
e
: equivalent height, Q
B
: base shear force
1
: horizontal Displacement at equivalent height

Figure 4. Multi-story model and ESDOF system of building
3 RELATIONSHIP BETWEEN LOAD AND DEFLECTION FOR PUSHOVER ANALYSIS
3.1 Modeling of members of superstructure
Column members are modeled as a rigid beam element with a rotational spring at the column
capital and column base. Beam members are also modeled as the rigid beam element with the
rotational spring in both ends, as shown in Figure 5. In these cases, the bending moment (M)
and rotating angle () make the model with a tri-linear type curve.
As far as the shear walls are concerned, as shown in Figure 6, beams and columns are able to
bear only axial tension only (beam ends, column capitals and bases are supported by hinges).
Wall member has axial and shear springs with the rotational springs at top and base. In other
words, wall member is modeled as virtual columns where the wall members can bear bending
and shearing forces.

McCracking moment
MyYielding moment
K0Elastic stiffness
KcStiffness after cracking
KyStiffness after yielding
Mc
M
My
Kc
K0
Ky

Figure 5. Model of beams.

241
Nonlinear shear
spring
Nonlinear rotary spring
Rigid beam
Nonlinear axial spring

Figure 6. Model of shear walls.

3.2 Model of piles
Pile members are modeled, as shown in Figure 7. The relationship between the bending moment
(M) and curvature () (M - curve) is set to a model as a tri-linear type skeleton curve. In the
modeling, the pile head axis force of superstructure required horizontal yield strength (at equiv-
alent to required Ds) was used. Then based on the M- curve calculated from the cross-section
equilibrium, the member characteristics of body structure was set up so as to obtain the expres-
sion: (Area -1) = (Area -2). Moreover, the relationship of the concrete stress-strain is specified
by using the bi-linear type skeleton curve, where the material strength of reinforcement is speci-
fied 1.1 times larger than the design standard strength, according to e function method. More-
over, the dissecting length of pile members is standardized as 1/2 of the pile diameter. In addi-
tion, the curvature distribution in the dissecting zone is specified constant. It was assumed that
the pile and pile cap is jointed rigidly.

M

Mc
MyMu
c y u
McCracking moment
cCurvature of cracking
MyYielding moment
yCurvature of yielding
MuUltimate moment
uCurvature of ultimate
Muu
E1I1
E2I2
Mcc
Cross-section Analysis
Area-2
Area-1

Figure 7. Model of piles (M- relation).

3.3 Model of sugrade reactions
The superstructure is modeled to a three-dimensional frame, and columns and girders are re-
placed with beam elements and bearing walls are replaced with the column beam. Spring mod-
els for horizontal subgrade reaction, skin friction and end bearing are set based on the recom-
242
mendation by Architectural Institute of Japan (AIJ 2001, Watanabe et al. 2004). The models of
subgrade reactions are drawn in Figure 8. The pile group effect is not included in the analysis.

a) Horizontal subgrade reaction
b) End bearing reaction
c) Skin friction
R
F

10mm(c)20(s)
0.8R
F
3mm(c)5(s)
cClayersSandy
R
u

1/2R
u

0.1D

0.6mm 10mm
P
max

A (=2)BC (=1/3)k
ho

y
P
A (=2)BC (=1)k
ho

A (=2)BC (=4)k
ho

k
h
= y
-1/2
k
ho

Figure 8. Springs of soil reactions to piles.

Each spring is expressed as follows;

a) Horizontal subgrade reaction
Horizontal subgrade reaction is expressed as follows;
k
h
=ABCk
ho
(3)
where, k
h
= horizontal subgrade reaction per unit area; A, B and C = factors for dependency of
creep at static loading test, of strain at soil response and of soil nonlinearity, respectively; k
ho
=
horizontal subgrade reaction at displacement of 10mm (= 80E
o
d
-3/4
); where E
o
= modulus of
deformation of ground; and d = diameter of piles.
The value of P
max
is a plastic subgrade reaction and is expressed as follows;
P
max
= 3.0 K
p
z d for sandy soil
= 9c (z 2.5d) for clayer soil (4)
=2(1+1.4z/d) cz < 2.5d
where K
p
= coefficient of passive earth pressure; = unit weight; z = depth; and c = value of co-
hesion.

b) End bearing reaction
The relationship between force and displacement at pile tip is expressed as follows;
243
n
p pu
p p
p pu
p p p
A R
A R
A R
A R d S
+ =
/
/
) 1 (
/
/
1 . 0
/
(5)
where, = 0.23 (sandy soil) and 0.12 (gravel); and n = 2.70 (sandy soil) and 3.31 (gravel). And
S
p
= displacement at pile tip; Rp = vertical load; Ap = cross sectional area of pile; and R
pu
= ulti-
mate end bearing capacity. The value of R
pu
is expressed as follows;
R
pu
100N
ave
A
p
(6)
where N
ave
= average N value obtained from the standard penetration test.

c) Skin friction along pile
The ultimate skin friction of pile (R
F
) is as follows;
R
F
= 10/3NsLs for sandy soil (7a)
R
F
= 0.5q
u
Lc for clayer soil (7b)
where Ns = N value for sandy soil layers ( 30); Ls = thickness of sandy soil; q
u
= unconfined
compressive strength ( 200kN/m
2
); Lc = thickness of clayer soil; and = perimeter of pile.
4 EVALUATION OF EQUIVALENT DAMPING FACTOR
4.1 Evaluation of equivalent damping coefficient
As the contribution of damping effects by SSI, following items are mentioned;
i) Hysteretic damping due to soil nonlinearity at contact points between piles or embedment and
soil
ii) Material damping of soil due to nonlinearity with shear strain dependency during S-wave
propagation in the ground
iii) Radiation damping through pile or embedment to ground
In the study, only radiation damping is considered (AIJ 2006). A horizontal dashpot (viscous
damper) is set in parallel to the springs, as shown in Figure 9.

BWidth of Pile
D
u0Horizontal Displacement at Pile
Pile Tip
Ground Surface
C i
K i
uiHorizontal Displacement at Pile

Figure 9. Soil spring and dashpot in horizontal motion.

244
The damping force in the perpendicular direction to pile axis is evaluated as follows (Iiba et
al. 2007);
i Hij Hij
u C F & = (8)
ij j Li i Hij
D d V C = (9)
where C
Hij
= viscous damping coefficient; u
i
= response displacement of all piles at i-th layer in
horizontal direction;
i
= density of soils at i-th layer; V
Li
= Lysmers velocity of soils at i-th
layer; d
j
= diameter of j-th pile; and D
ij
= length of division along j-th pile. When the dis-
placement at all pile head is set to be u
0
and the equivalent viscous damping coefficient at pile
head for j-th pile is to be C
H0j
, the damping force at j-th pile head is obtained as follows;
( )
= =
j i
i Hij j H j H
u C u C F & &
0 0 0
(10)
The total equivalent viscous damping coefficient of all piles (C
sw
) is shown.
) (
0 0 0
u C u C
j H sw
& & = (11)
The C
sw
is expressed by the damping coefficient of each dashpots (C
H0i
) as follows;
0
/ ) ( u u C C
i Hij
j i
sw
=
(12)
In case of rotational motion, the radiation damping model is illustrated in Figure 10. The ver-
tical displacement at i-th layer and j-th pile is set to be v
ij
. The equivalent viscous damping coef-
ficient at pile head is expressed as follows;
=
j i
j ij Vij ro
L v C C
0
/ ) (
(13)
ij j i S i Vij
D d V C = (14)
where
0
= average rotational angle at pile head; L
j
= distance from center of rotation; C
Vij
=
equivalent viscous damping coefficient at pile; and Vs
i
= S-wave velocity at i-th layer.

LjDistance to Pile
Center of Rotation
0Average Rotational Angle at Pile Head
iAverage Rotational Angle at Pile
vijVertical Displacement at Pile

Figure 10. Soil spring and dashpot in rocking motion.
245
4.2 Equivalent damping factor of buildings with SSI system
The equivalent viscous damping factor (he) of the total system is expressed as follows;
3 3 3
=
eq
b
b
eq
ro
ro
eq
sw
sw e
T
T
h
T
T
h
T
T
h h (15)
where T
b
, T
sw
, T
ro
and T
eq
= predominant periods for superstructure, sway, rocking and whole
system in the ESDOF; and the h
b
, h
sw
and h
ro
= equivalent viscous damping factors of super-
structure, sway and rocking system corresponding to predominant period of each mode.
The damping factors are obtained using following equations;
03 . 0
1
1
1
+
b
h
(16)
M
C
h
sw
sw
sw
1
2
= ,
2
1
2
e ro
ro
ro
MH
C
h
= (17)
where
1
= value dependent on structural system of superstructure (0.25 for general RC struc-
ture); = the plastic ratio (ratio of story drift to yield drift) of superstructure; and
sw
and
ro
=
predominant circular frequencies for sway and rocking motions.
4.3 Reduction factor for Acceleration response spectrum
To get the acceleration and displacement responses according to the damping factor of system, a
following reduction factor (F
h
) for the ARS is used;
e
h
h
F
10 1
5 . 1
+
=
(18)
5 BUILDING-PILE-SOIL MODEL FOR RESPONSE CALCULATION
5.1 Outline of Models
Three buildings with 5, 8 and 14-story and one soil ground are selected. Elevations of 8 stories
building are drawn in Figure 11. The short span (transverse) direction of the residential building
is analyzed, which consists of continuous bearing walls through the height in the direction. The
response of building in transverse direction is much influenced by sway and rocking motions of
the foundation. The height, mass distribution and other properties of the buildings are presented
in Tables 1, 2. The soil properties of ground are summarized in Table 3 and Figure 12. The soil
deposits have 0.73 s in predominant period at small shear strain region.

a) Longitudinal direction b) Transverse direction

Figure 11. Elevations of 8 stories building.
246

Table 1. Dimensions and properties of buildings.
_______________________________________________________________________
Structure Floor 5-story Floor 8-story 14-story
Height Mass Height Mass Height Mass
________________________________________________
m t m t m t
_______________________________________________________________________
Superstructure R 42.7 764
14 39.9 674
13 37.2 682
12 34.4 687
11 31.7 703
10 28.9 711
9 23.7 765 26.1 728
8 20.9 667 23.3 737
7 18.2 680 20.5 748
6 15.3 900 15.4 683 17.7 755
5 12.6 686 12.7 696 14.9 775
4 9.80 681 9.90 706 12.1 787
3 7.05 680 7.15 708 9.20 795
2 4.30 700 4.40 715 6.35 799
Foundation 1 1.10 805 1.00 1038 2.00 1405
Total 4452 6658 11749
Effective Value 9.32 4314 14.1 6094 27.5 9690
at first mode*
Embedment 1.30 1.40 3.20
______________________________________________________________________
* To include the depth of embedment

Table 2. Dimension of piles.
________________________
Story Diameter Tip depth
m m
_______________________
5 1.3-1.5 32
8 1.6-1.8 32
14 2.2 32
________________________

Table 3. Soil properties of ground model.
_________________________________________
Depth Soil type Density Vp Vs
_________________________________________
m t/m
3
m/s m/s
_________________________________________
14 clay 1.6 500 130
20 Fine sand with silt 1.8 450 150
28 Fine sand 1.8 575 200
>28 Gravel 2.0 1000 400
_________________________________________

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.00
0.20
0.40
0.60
0.80
1.00
1.0E-05 1.0E-04 1.0E-03 1.0E-02 1.0E-01
h
G
/
G
0
Shear Strain
Clayey Soil
Sandy Soil

Figure 12. Nonlinear characteristic of soil.
247
5.2 Earthquake motion
The earthquake motion is given as the ARS at the outcropped engineering bedrock as drawn in
Figure 13. The time history responses on the ground surface whose average ARS is drawn in the
same Figure, are calculated. Table 4 summarized the equivalent S-wave velocities and damping
factors which are obtained by the equivalent linear analysis.

0
5
10
0.1 1.0 10.0
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
/
s
)
Period (s)
Ground Surface
Outcropped Engineering Bedrock

Figure 13. ARS of engineering bedrock and ground surface.

Table 4. Equivalent S-wave velocity and damping factor.
___________________________________________________
Depth Initial Values Values during earthquake
Vs h Vs h
___________________________________________________
m m/s m/s
___________________________________________________
2.33 130 0.02 123 0.023
4.66 130 0.02 106 0.071
6.99 130 0.02 92.2 0.104
9.32 130 0.02 83.9 0.122
11.7 130 0.02 77.7 0.134
14.0 130 0.02 73.0 0.143
17.0 150 0.02 58.8 0.193
20.0 150 0.02 55.6 0.196
24.0 200 0.02 97.8 0.173
28.0 200 0.02 89.3 0.182
>28.0 400 0.02 400 0.02
___________________________________________________

5.3 Load and deflection relationship of buildings
Figure 14 presents the relationships between shear force and drift at each story of 8-story build-
ing through the nonlinear pushover analysis. The acceleration-displacement relationship of the
ESDOF system is obtained as shown in Figure 15. In the figure, the displacements of super-
structure, sway motion, rocking motion and total at the equivalent height of building are drawn.
The equivalent period of the system at any horizontal displacement, as shown in Figure 16, is
calculated using a following equation.
) / (
2
1 1
1
B
eq
Q
M
T = (19)
The equivalent period gradually increases with displacement due to the nonlinear characteris-
tics of structural members and soil reaction.

248
0
10000
20000
30000
40000
50000
0 5 10 15
S
t
o
r
y

S
h
e
a
r

F
o
r
c
e

(
k
N
)
Story Drift (cm)
1F
2F
3F
4F
5F
6F
7F
8F
Pile Head

Figure 14. Shear force vs. drift at each story (8F building).

0
200
400
600
800
1000
0 5 10 15 20 25
S
p
e
c
t
r
a
l

A
c
c
e
l
e
r
a
t
i
o
n

(
c
m
/
s
e
c
2
)
Spectral Displacement (cm)
Building Rocking
Sway
1
S
a
-
1
S
d

Figure 15. Relationship between
1
S
a
and
1
S
d
(8F building).

0.0
0.5
1.0
1.5
0 5 10 15 20
E
q
u
i
v
a
l
e
n
t

P
e
r
i
o
d

(
s
)

Figure 16. Equivalent period with displacement (8F building).

249
5.4 Equivalent damping coefficient
Based on horizontal and vertical displacements of piles based on pushover analysis and the
equivalent s-wave velocities of soil deposits (Table 4), the equivalent viscous damping coeffi-
cients at pile head are obtained as three lines drawn in Figure 17. The dependency of displace-
ment or rotational angle at pile head on the damping coefficients is not so remarkable. The
damping coefficients tend to increase with horizontal displacement or rotational angle.

a) Sway motion
0.0E+00
1.0E+04
2.0E+04
3.0E+04
4.0E+04
5.0E+04
6.0E+04
0 1 2 3 4 5 6 7 8 9 10
V
i
s
c
o
u
s

C
o
e
f
f
i
c
i
e
n
t

(
k
N
s
/
m
)
Displacement (cm)
5F2c
8F2c
14F2c
5F2c(TLE,h=2%)
8F2c(TLE,h=2%)
14F2c(TLE,h=2%)

b) Rocking motion
0.0E+00
2.0E+06
4.0E+06
6.0E+06
8.0E+06
1.0E+07
1.2E+07
1.4E+07
1.6E+07
0.000 0.002 0.004 0.006 0.008 0.010
V
i
s
c
o
u
s

C
o
e
f
f
i
c
i
e
n
t

(
K
N
m
s
/
r
a
d
)
Rotation Angle (rad)
5F2c
8F2c
14F2c
5F2c(TLE, h=2%)
8F2c(TLE, h=2%)
14F2c(TLE, h=2%)

Figure17. Equivalent viscous damping coefficients.
5.5 Bending moment and curvature relationship of piles
The pile arrangement of 8-story building is shown in Figure 18. Three different diameters of
piles are used in the building. The relationships between bending moment and curvature for two
piles are drawn in Figure 19. The characteristics of pile at position (X4, Y1) are calculated on
condition that the pile has an additional pulling force due to an overturning moment during
earthquake. On the other hand, the characteristics of pile at position (X4, Y2) are calculated un-
der the additional compressive force due to the overturning moment.

250
Y2
Y1
X1 X2 X3 X4 X5 X6 X7
Diameter of pile
: 1600 : 1700 : 1800

Figure18. Pile arrangement of the foundation (8F building).

0
2000
4000
6000
8000
0 0.001 0.002 0.003 0.004 0.005 0.006
B
e
n
d
i
n
g

M
o
m
e
n
t

(
k
N
m
)
Curvature (1/m)
PilepositionX4,Y1

0
2000
4000
6000
8000
10000
12000
14000
0 0.001 0.002 0.003 0.004 0.005 0.006
B
e
n
d
i
n
g

M
o
m
e
n
t

(
k
N
m
)
Curvature (1/m)
PilepositionX4,Y2

Figure19. Pile arrangement of the foundation (8F building).
6 RESPONSE OF BUILDINGS
Based on the equivalent period and equivalent damping factor, the response of the ESDOF sys-
tem is obtained through Figure 1. The cross points between demand and capacity spectra give
the maximum acceleration and displacement of the system. Figure 20 shows the responses of 8-
and 14-story buildings under the acceleration responses at ground surface and the force-
displacement relationship of the system with SSI.
Table 5 summarizes the maximum acceleration at equivalent height, the maximum shear
force at both 1
st
story and pile head and the maximum displacement of sway, rocking and build-
ing at building top. The maximum response accelerations of buildings are about 4 m/s. The
shear forces at pile head is 10-20% larger than those at 1
st
story. For 5-story building, the ratio
of sway to total is 59% and the effect of sway on the total displacement is remarkable. For 14-
story building, the effect of rocking is remarkable. In case of 8-story building, the effects of
sway and rocking are similar.
251

a) 8-story building
0
200
400
600
800
1000
0 5 10 15 20 25 30
S
p
e
c
t
r
a
l

A
c
c
e
l
a
r
a
t
i
o
n

(
c
m
/
s
e
c 2
)
Sa-Sd Spectrum (h=5%)
Sa-Sd Spectrum (h=14.7%)
Capacity Specturm
Te=0.675s
Response

b) 14- story building
0
200
400
600
800
1000
0 5 10 15 20 25 30 35 40
S
p
e
c
t
r
a
l

A
c
c
e
l
a
r
a
t
i
o
n

(
c
m
/
s
e
c
2
)
Sa-Sd Spectrum (h=5%)
Sa-Sd Spectrum (h=12.3%)
Capacity Specturm
Te=1.49s
Response

Figure 20. Response of ESDOF system.

Table 5. Maximum response of buildings.
a) Acceleration and shear force
_______________________________________________
Story Acceleration shear force
Pile head 1
st
story
_______________________________________________
m/s
2
kN kN
_______________________________________________
5 3.94 1.70E+04 1.37E+04
8 3.87 2.36E+04 2.07E+04
14 4.77 4.65E+04 4.22E+04
_______________________________________________
b) Displacement (sway, rocking and building)
_____________________________________________
Story Displacement
Total Sway Rocking Building
_____________________________________________
cm cm cm cm
_____________________________________________
5 2.62 1.54 0.87 0.21
8 5.85 2.18 2.90 0.77
14 36.0 5.94 25.6 14.45
_____________________________________________

The equivalent periods and damping factors of buildings are summarized in Table 6. The
equivalent period is calculated as follows;
252
2 2 2 2
b ro sw e
T T T T + + = (20)
a total e
S u T
1
/ 2 = (21)
where u
total
= the displacement at equivalent height. The mode whose period is the largest
among the periods is predominant in the SSI system. For 5-story building, the sway mode is
predominant. For 14-story building, the mode of rocking is remarkable.

Table 6. Equivalent periods and damping factors of buildings.
a) Equivalent periods
_________________________________________
Story T
b
T
sw
T
ro
SSI(T
e
)
_________________________________________
s s s s
__________________________________________
5 0.12 0.39 0.23 0.47
8 0.24 0.47 0.42 0.68
14 0.61 0.70 1.17 1.49
__________________________________________
b) Equivalent damping factors
_______________________________________________________
Story Building(h
b
) Sway(h
sw
) Rocking(h
ro
) SSI(h
e
)
_______________________________________________________
% % % %
_______________________________________________________
5 3.46 23.5 30.6 17.3
8 4.32 25.6 24.0 14.7
14 4.07 31.0 18.5 12.3
_______________________________________________________

The maximum responses of bending moment of piles at positions (X4, Y1) and (X4, Y2) are
plotted in Figure 19. The distributions of bending moment at these piles are drawn in Figure 21.

0
5
10
15
20
25
30
-4000 -2000 0 2000 4000
D
e
p
t
h

(
m
)
Bending moment (kNm)
(X4,Y1)

0
5
10
15
20
25
30
-6000 -3000 0 3000 6000
D
e
p
t
h

(
m
)
Bending moment (kNm)
(X4,Y2)

Figure 21. responses of bending moment of piles.

7 DISCUSSION ON PILE STIFFNESS, DAMPING COEFFICIENT AND EARTHQUAKE
RESPONSE
7.1 Equivalent stiffness of pile head and viscous damping coefficient
The equivalent stiffness of pile and viscous damping coefficients during earthquake by pushov-
er analysis are compared with those by the thin layer element (TLE) analysis. In the TLE analy-
sis, the damping factor of soil deposits is set to be 2%. Because that the material damping of soil
253
deposit due to nonlinearity is not considered in this study. The dynamic impedance of the pile
foundation for horizontal and rocking motions are obtained, based on the linear approximation.
The dynamic impedances of pile foundation of 8-story building are drawn in Figure 22. As the
impedance functions have a frequency dependency. Because the spring and dashpot values are
assumed to be constant in the following time history analysis, the real parts of impedances at
frequency of 0.1Hz are adopted as the spring constant, and the minimum values of the damping
coefficients are selected.

a) Sway motion
0.0E+00
2.0E+06
4.0E+06
6.0E+06
8.0E+06
1.0E+07
1.2E+07
0 1 2 3 4 5
I
m
p
e
d
a
n
c
e

(
k
N
/
m
)
Frequency (Hz)
Real Part
Imaginary Part
Spring Constant at f=0.1 Hz
Damping Coefficient(3.11*E+04 kNs/m)

b) Rocking motion
0.0E+00
2.0E+08
4.0E+08
6.0E+08
8.0E+08
1.0E+09
1.2E+09
0 1 2 3 4 5
I
m
p
e
d
a
n
c
e

(
k
N
m
/
r
a
d
)
Frequency (Hz)
Real Part
Imaginary Part
Spring Constant at f=0.1 Hz
Damping Coefficient(3.09*E+06 kNms/rad)

Figure 22. Impedance functions of foundation of 8F building.

The equivalent spring constants and viscous damping coefficients are compared with the re-
sults by the TLE analysis in Table 7. And the equivalent viscous damping coefficients by the
TLE analysis are plotted in Figure 17. The sway spring values by the pushover analysis are a
good agreement to those by the TLE analysis. The difference is increased with height of build-
ing. The rocking spring values by pushover analysis are lower than the TLE results. The reason
is the difference of the characteristics of spring models for end bearing and skin friction of piles.
The spring values by the pushover analysis decrease with the relative displacement between pile
and soil. On the other hand, the properties of soil are treated to be equivalent linear in the TLE
method. While the horizontal damping coefficients by the radiation effect have a good agree-
ment to those by TLE results, the rotational damping coefficients are overestimated and about
three times those by TLE results.
In order to investigate the effect of spring values for end bearing, the springs which are al-
most the same as those in the TLE analysis are set at the pile tip. The equivalent viscous damp-
254
ing coefficients at pile head are drawn in Figure 23. By using the same amount of spring at the
pile tip, the rotational damping coefficients are about two times those by TLE results.

Table 7. Comparison of spring values and viscous damping coefficients.
a) Sway motion
_____________________________________________________
Story Pushover analysis Thin layer element
Spring Damping Spring Damping
_____________________________________________________
kN/m kNs/m kN/m kNs/m
_____________________________________________________
5 1.12E+06 3.24E+04 1.51E+06 2.81E+04
8 1.11E+06 4.16E+04 1.65E+06 3.11E+04
14 7.83E+05 5.41E+04 1.91E+06 3.68E+04
_____________________________________________________
b) Rocking motion
__________________________________________________________
Story Pushover analysis Thin layer element
Spring Damping Spring Damping
__________________________________________________________
kNm/rad kNms/rad kNm/rad kNms/rad
__________________________________________________________
5 2.82E+08 7.78E+06 5.37E+08 2.31E+06
8 2.80E+08 1.04E+07 6.97E+08 3.09E+06
14 2.11E+08 1.45E+07 1.04E+09 4.70E+06
__________________________________________________________

0.0E+00
2.0E+06
4.0E+06
6.0E+06
8.0E+06
1.0E+07
1.2E+07
0.000 0.001 0.002 0.003 0.004
V
i
s
c
o
u
s

C
o
e
f
f
i
c
i
e
n
t

(
K
N
m
s
/
r
a
d
)
Rotation Angle (rad)
5F2c
8F2c
14F2c
5F2c(TLE, h=2%)
8F2c(TLE, h=2%)
14F2c(TLE, h=2%)

Figure 23. Equivalent viscous damping coefficients for rocking motion as to another model.

7.2 Earthquake response of building
In order to investigate the amount of the responses by spectrum-based method, the time his-
tory analysis on multi degree of freedom model of 5, 8 and 14-story buildings is conducted. The
relationship between shear force and story drift of superstructures is a tri-linear skeleton curve
based on the results by pushover analysis with base fixed condition. A standard rule under
unloading and reloading processes is used as the hysteretic characteristics. The damping factor
of superstructure is 3% in proportional to the initial stiffness of the story. In order to analyze the
SSI model, the spring and dashpot are added to the superstructure (SR model). The spring and
dashpot values are assumed to be constant in the time history analysis, as before mentioned. As
the spring constant, the real parts of impedances at frequency of 0.1Hz are adopted and the min-
imum values of the damping coefficients are selected (Fig. 22).
The distributions of maximum shear force and maximum displacement for 8-story building
by the spectrum-based method are compared with those by the time history analysis, as drawn in
Figure 24. While the maximum shear force by the spectrum-based method is a little smaller than
those by the time history analysis, the displacement is larger. The reason why the displacement
is larger is that the rocking (vertical) spring values by pushover analysis are lower than the TLE
255
results (Table 7). Since the sway and rocking spring value of 8-story building by pushover anal-
ysis is two third and one forth times those by TLE method respectively, the displacement of
sway and rocking motions is remarkable in whole displacement.
The reason why the story shear force is smaller is that the response of building does not in-
clude the effect of higher vibration modes than 1
st
mode. From Table 1, the effective mass of
building for 1
st
mode is 0.92 times total mass. The effect of higher modes on the response needs
to be considered.

0
1
2
3
4
5
6
7
8
9
0 10000 20000 30000
S
t
o
r
y
Story Shear Force (kN)
Proposed
Method
Time
History
0
1
2
3
4
5
6
7
8
9
0 5 10
S
t
o
r
y
Displacement (cm)
Proposed
Method
Time
History

Figure 24. Story shear force and displacement distribution with height.
Concluding remarks in the responses of buildings with SSI through the spectrum-based method
are summarized as follows.
1) The equivalent period of SSI system is estimated based on the pushover analysis of the build-
ing, foundation and surrounding soil model.
2) The equivalent viscous damping factor is estimated based on the radiation damping effect
through piles.
3) The sway spring values by the pushover analysis are a good agreement to those by the TLE
analysis. The difference is a little increased with height of building. The rocking spring values
by pushover analysis are lower than those by the TLE analysis.
4) The horizontal damping coefficients by radiation effect based on the pushover analysis have
a good agreement to those by TLE analysis. The rotational damping coefficients are about three
times those by TLE analysis. In case where the springs of pile tip are set to be almost the same
as those used in the TLE analysis, the difference of equivalent viscous damping coefficients be-
comes small.
5) The maximum shear forces by the spectrum-based method are a little smaller than those by
the time history analysis, while the displacements are larger. Because of low spring values,
sway and rocking displacements are more remarkable in the results by pushover analysis. As to
the smaller story shear force, the effect of higher vibration modes on the response of building
needs to be considered.
ACKNOWLEDGEMENTS
The authors express their sincere thanks to members of the committee Response evaluation
method and its example using spectrum-based method based on pushover analysis of building,
256
foundation and soil model, for their suggestion related to the model of pushover analysis and
the method of spectrum-based method.
REFERENCESS
Architectural Institute of Japan, 2001. Recommendation for the design of building Foundations (in Japa-
nese)
Architectural Institute of Japan, 2006. Seismic response analysis and design of buildings considering dy-
namic soil-structure interaction (in Japanese)
Iiba, M., Umemura, Y., Kurimoto, O. & Inoue, Y. 2007. A study on seismic response of building based
on response spectrum using pushover analysis of combined super- and sub-structures, Structural field,
Summaries of technical papers of annual meeting, Architectural Institute of Japan, 99-106 (in Japa-
nese)
Kuramoto, H., Teshigawara, M., Okuzono, T., Koshika, N., Takayama, M. & Hori, T. 2000. Predicting
the earthquake response of buildings using equivalent single degree of freedom system, Proceedings
of 12th World Conference on Earthquake Engineering, Paper No. 1039
Kuramoto, H., Midorikawa, M. Teshigawara, M., Ueda, T. & Yoshimura, M. 2002. An overview of per-
formance-based seismic code of buildings in Japan, Advances in Mechanics of Structures and Mate-
rials, 333-338, A.A. Balkema Publishers
Midorikawa, M., Hiraishi, H., Ohkawa, I., Iiba, M., Teshigawara, M. & Isoda, H. 2000. Development of
seismic performance evaluation procedures in building code of Japan, Proceedings of 12th World
Conference on Earthquake Engineering, Paper No. 2215
Umemura, Y., Kurimoto, O., Iiba, M., Watanabe, K., Kuramoto, H. & Tamura, M. 2004. A study on the
seismic load distribution along vertical direction considering soil-building interaction, Structural field,
Summaries of technical papers of annual meeting, Architectural Institute of Japan, 217-220 (in Japa-
nese)
Watanabe, K., Umemura, Y., Teshigawara, M., iiba, M. & Tamura M. 2004. A study on seismic response
evaluation by monotonic analysis of building and its foundation, Proceedings of 13th World Confe-
rence on Earthquake Engineering, Paper No. 1939
257
1 INTRODUCTION
For more available method on the seismic designs, it is very important to develop application of
the performance based design method. The strength demand spectrum based on the ductility fac-
tor is one of the most useful methods, which can be treated with nonlinear effects on the struc-
ture subjected to seismic forces (Iemura et al., 1998). However, on the performance evaluation
due to the strength demand spectrum, it is not always clear how to evaluate the damage on the
structures with respect to the hysteretic energy and input seismic motion energy. The strength
demand spectrum based on the damage evaluation of the structure has also been examined (Mi-
kami et al., 1998; Kimura et al., 2007). It has been indicated that one of the important factors on
the damage evaluations could be given by the total energy of the seismic motion on the structure
(Park et al., 1995). However, these researches have been carried out in the point of view on the
influence of the main shock.
It is presumed that the earthquake with severe intensity is so considerable effect due to after-
shocks that it is important to examine effects on the damage evaluation. Since the damage of
structure may be closely related to the earthquake input energy, it is essential to evaluate accu-
mulated energy on the structure due to aftershocks. For the evaluation of the damage of struc-
ture by the aftershock, it is important to clarify the accumulated damage on the performance
based design due to earthquake. The appropriate estimation of the deterioration due to after-
shock plays the important roles on evaluation of accumulated damages on the nonlinear re-
sponse situation due to aftershocks.
In this present study, effects on the aftershocks are examined with the damage index for the
soil structure interaction system (SSI), which is modeled with the sway rocking model. It is in-
dicated that the damage index is closely related with accumulated damage estimation of the
structure due to aftershocks. It is suggested that an increase of seismic intensity ratio of after-
shock to main shock leads to an increase of accumulated damage of the structure. Therefore, it
Aftershock Effects on Damage Evaluations for Soil Structure
Interaction System

K. Kawano & Y. Kimura
Kagohisma University, Japan
ABSTRACT: An available evaluation on aftershocks plays important roles on the performance
based design method to seismic motions. In this present study, it is examined about the uncer-
tainty effects on the damage index, which is evaluated with combining the ductility factor with
the hysteretic energy for the severe damage level of the structure for the soil-structure interac-
tion system. While there are a little bit influences of the damage index due to after shocks, it is
shown that the uncertainty on the maximum seismic motion plays important roles on the dam-
age index evaluation. For the performance based design to the seismic motion, the evaluation of
the uncertainty to the damage situation becomes so important that it can be carried out with the
fragility on the assigned damage level.

258
is shown that the fragility on the assigned damage index leads to the available estimation on
damages due to aftershocks.
2 FORMULATION
Many researchers have examined for the dynamic SSI and supposed that the SSI plays impor-
tant roles on the seismic response evaluations. Taking into account the dynamic response repre-
sented with a few dominate vibration mode, the superstructure can be expressed with a single-
degree-of-freedom (SDOF) system. The nonlinear characteristics on the structure can be ex-
pressed with the trilinear 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 nonlin-
ear SSI system is expressed with

[ ]{ } { } [ ( )]{ } { } M x C x K t x F + + = (1)

in which [M] and [C] denote the mass and damping matrix, and [K(t)] denotes the stiffness ma-
trix on each time step including the soil foundation system with linear properties. {F} denotes
vector for the input seismic motion. { } x , { } x and { } x denote the acceleration responses, the ve-
locity responses and the displacement responses, respectively. If the dynamic response is carried
out within linear region, the governing equation of motion 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.

Figure 1. An idealized structure-soil-foundation model.

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 + + = (2)
in which

[ ( )]{ } { } K t x F = (3)

259
2
4 2
[ ( )] [ ( )] ( )[ ] [ ]

K t K t M C
t t
= + + (4)

2
4
{ } { } { }{( ){ } 2{ }} [ ](2{ })
F F M x x C x
t
= + + + (5)

Therefore, the increment of the responses can be determined by solving the equation (3).The
iterative procedure can be carried out using the Newton Raphson method. The damage evalua-
tion 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, 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 is applicable to assess the damage situation by
means of the maximum displacement and the hysteretic energy as follows:

1
1 1
M
u y u
x
D dE
x Q x
= +

(6)
in which x
1M
and x
1u
stand for the maximum displacement and ultimate displacement, respec-
tively. The ultimate displacement can be determined by the corresponding allowable ductility
ratio. The allowable ductility ratio is 5.0 in this present study. Q
y
denotes the yield force 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 exam-
ined.

Table 1. Relationship between Damage Index and Damage level.

3 RESULTS AND DISCUSSIONS
The accumulated damage estimation due to aftershocks is examined using typical seismic mo-
tions in Japan. In the present study, the damage evaluation is carried out with the three seismic
motions of near field type, Kobe-ns component, Taka-ns component and Port-ns component as
shown in Figure 2, which are obtained by Kobe earthquake (1995). All the seismic motions
260
have a dominant frequency of relatively narrow range, for example, for Kobe ns from 0.3 sec to
0.8 sec.
In order to clarify the damage evaluation of the SSI system as shown in Figure 1, the seismic
response analysis of the SSI system is carried out using typical seismic motions corresponding
to the design spectra as shown in Figure 2. The soil foundation is supported by ground condition
with two layers, and the sway and rocking spring constant of soil foundation can be determined
by the each ground condition. Figure 3 shows the relation between the natural period and
ground condition of each seismic motion site. The natural period of SSI system is supposed to
be mainly depended on the superstructure.

Figure 2. Acceleration response.

Figure 3. Natural period due to each ground condition.

4 SEISMIC RESPONSES WITH AFTERSHOCK
For the evaluation on the influence of aftershock, it is necessary to simulate earthquake ground
motions including the aftershock. In this present study, it is assumed that substantial aftershock
would be occurred after the main shock. Assumption is made by the maximum acceleration of
aftershock, which yields the intensity of seismic motion from 10% to 70 % of main shock and
the same characteristic of motion.

261
Figure 4 shows the time history of displacement response due to simulated earthquake ground
motions of the input intensity of 50% to Kobe-ns. The two lines denotes the displacement re-
sponse due to the damage index D=0.1 and 0.6, respectively. The natural period of the structure
is 0.5 sec. If the damage index is 0.6, the response leads to severe nonlinear response and to drift
of the displacement. Figure 5 similarly shows the time history of the displacement due to the
same seismic motion including aftershock. The natural period is 1.0 sec. The two responses are
denoted to the damage index D=0.1 and 0.6, respectively. The displacement response becomes
larger responses than the natural period 0.5 sec because of the dominant frequency of the input
motion. It is understood that the dynamic response due to aftershocks is affected for the damage
index 0.6 and the natural period of the structure.

Figure 4. Time history of displacement responses due to aftershock.

Figure 5. Time history of displacement responses due to aftershock.

Figure 6 shows the relation between restoring force and displacement to Kobe-ns. Each line
denotes responses due to the main shocks and aftershock. The abscissa denotes the ratio of the
displacement to the yield one and the ordinate does the ratio of the restoring force to yield one.
The natural period of the superstructure is 0.5 sec and the damage index 0.6. It is understood
that the maximum nonlinear response is caused by the main shock of input seismic motion and
aftershock plays important roles on increase of the hysteretic energy.

262

Figure 6. Response of restoring force.
5 DAMAGE INDEX DUE TO AFTERSHOCK
In order to evaluate the influence of aftershock, the seismic response analysis is carried out with
the simulated earthquake ground motion. Figures 7 and 8 show the damage index with respect to
the influence of the aftershock, which is assumed to be the input intensity of 50%, to Taka-ns,
and Kobe-ns, respectively.
The effects on the aftershock are estimated with the difference from the corresponding target
damage index. When the damage level of the structure is slight such as 0.2, a constant damage
index is obtained in spite of the natural period of the structure. It is understood that there are a
little bit influences on the damage index due to aftershock. However, if the severe damage level
is allowed in structure, it is indicated that the influence of aftershock leads to somewhat increase
of the damage index in comparison with the damage index due to only the main shock. More-
over, it is understood that the damage index is considerably influenced by the natural period of
the structure. Therefore, the aftershock plays the important roles on the damage evaluation for
the performance based evaluation of the structure.

Figure 7. Damage index with aftershock (Taka ns).

263

Figure 8. Damage index with aftershock (Kobe ns).

6 ENERGY RESPONSE DUE TO AFTERSHOCK
As previous mentioned, it is indicated that the hysteretic energy plays important roles on the
damage index evaluation. The damage index due to the aftershock is essentially influenced by
the hysteretic energy. The damage evaluation is so important roles that it could be demonstrated
by combining the ductility factor with the energy ratio of the hysteretic energy to the total input
energy. Figure 9 shows the energy responses to Taka-ns with the influence of the aftershock.
The natural period of the structure is 1.0 sec and the target damage index is 0.6. Each line corre-
sponds to the damping energy, the hysteretic energy, and the input seismic energy, respectively.
It is noted that the hysteretic energy of main shock comes out suddenly sharp increment be-
tween 5.0 sec and 10.0 sec in accordance with increase of energy of the input seismic motion. It
is known that increase of the hysteretic energy after about 60.0 sec is brought about the influ-
ence of the aftershock. Moreover, the increment of hysteretic energy aftershock indicates rela-
tively smaller than the main shock. However, the hysteretic energy due to the aftershock is sup-
posed to extend as input intensity of the seismic motion increases. Therefore, in order to
evaluate the influence of aftershock, it is very important to clarify the increment of hysteretic
energy due to the aftershock.
7 FRAGILITY ON ASSIGNED DAMAGE INDEX
A few of uncertainty are inherently involved in the input seismic motion and the strength
characteristics of the structure. For the available estimation of the damage index, it is so signifi-
cant roles that it is essential to clarify the effects on the uncertainty for the damage index
evaluation. Moreover, it is suggested that the uncertainty effects with respect to the damage
evaluation play important contributions on the damage evaluations for the performance based
design of the structure (Kimura et al, 2007). The damage index is evaluated with the yield dis-
placement, the maximum displacement response, the expected ultimate displacement and hys-
teretic energy. Taking into accounts the random property of the input seismic motion, the effects
of the uncertainty are so important that it should be involved with the damage evaluation of the
structure. In the present study, it is examined about the effect on the damage index due to the
uncertainty of the input maximum acceleration with the Monte Carlo simulation. The maximum
input acceleration of seismic motion is assumed to have a lognormal distribution. The coeffi-
cient of variation is 10%.
264

Figure 9. Time history of energy response (Taka ns).

The target design damage index is assumed to be 0.6 for the simulation. The probability of
excess is evaluated for the assigned damage index such as 0.2, and 0.4. The damage index 0.2
corresponds to the slight damage situation and 0.4 to the moderate damage situation as shown in
Table 1.
Figure 10 shows the fragility on the assigned damage index of the SSI system. The input
seismic motion is Kobe-ns and the natural period 0.5sec. The abscissa denotes the mean value of
the maximum acceleration of the input seismic motion. When the assigned damage index is 0.2,
the excess is occurred from about 3.2m/s
2
, and the damage index simulated becomes above
the 0.2 for the maximum acceleration above about 5.0m/
s2
. For the damage index assigned
0.4, the excess is similarly appeared above the maximum acceleration 5.0m/s
2

and the prob-
ability of excess above 0.4 becomes almost 100% for the maximum acceleration about
7.5m/s
2
. The differences between the damage index 0.2 and 0.4 are so considerable that it is
important to figure out the probability of excess for the target design damage index.
8 CONCLUSIONS
The effects on damage evaluation due to aftershocks of SSI system are examined. The main re-
sults are summarized as follows:
(1) The damage index combined the maximum displacement with the hysteretic energy is
useful to evaluate the damage situation for the SSI system. The ground condition and the dy-
namic characteristics of input seismic motion play significant roles on evaluations of the dam-
age index.
(2) While the performance based design of the light level of damage can be evaluated by the
maximum ductility factor of the structure, it is available for the damage evaluations of the SSI
system to carry out combining the ductility factor with the hysteretic energy to the severe dam-
age level of the structure.
(3) For the performance based design to the seismic motion, it is significant to clarify the un-
certainty to the damage situation as demonstrated the fragility on the damage level assigned to
SSI system.

265

Figure 10. Fragility on assigned damage index (Kobe ns).
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, Struc-
tural safety and Probability, pp.1705-1708
Kawashima,K., Macrae,G.A., Hoshikuma,J. & Nagaya,K.(1994.), Residual displacement response spec-
trum and its application, Journal of structural mechanics and earthquake engineering, JSCE, No.501/
-29, pp.183-192 (in Japanese)
Kimura,Y. Kawano,K. & Nakamura,Y( 2007), Damage evaluation on nonlinear SSI systems due to
forces, Applications of statistics and probability in civil engineeringKanda,Takada&Furuta (eds), pp
1-8
Mikami,T., Hirao,K., Sasada,S., Sawada,T. & Nariyuki,Y.( 1998), A study on design spectra of seismic
intensity for level 2 earthquake, The 10
th
Earthquake Engineering Symposium, pp.3061-3066 (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

266
Soil-Structure Interaction Problem of a School Building Based on
Earthquake Records and Dynamic Substructure Method

M. Nakamura, Y. Kitamura, J. Suzumura & K. Hanada
Department of civil engineering, College of science and technology, Nihon university, Tokyo, Japan

1 INTRODUCTION
ABSTRACT: The importance of dynamic interaction has been often pointed out for common
low- and medium-rise buildings however effects of dynamic interaction have not been actively
discussed at the stage of structural design. One of the reasons is that observation records for dis-
cussing soil-building interaction have not been accumulated sufficiently. College of science and
technology, Nihon university, has established the earthquake observation network for the
ground and school buildings in Funabashi Campus. The authors discuss inertial and kinematic
interaction effects for a SRC school building that has five stories above the ground and one
story under the ground for which earthquake observations are performed. Inertial interaction ef-
fects on the subject building's behaviors during an earthquake, which was recorded over 100gal
on the surface ground, are small, and such behaviors almost greatly depends on conditions of
the ground around the embedment of the building.

Earthquake resistance of socially important structures such as nuclear facilities has been evalu-
ated by earthquake observation, vibration tests and theoretical analyses with dynamic interac-
tion considered conventionally (Hanada 1987). On the other hand, the importance of dynamic
interaction has been often pointed out for common low- and medium-rise buildings however ef-
fects of dynamic interaction have not been actively discussed at the stage of structural design
(Tobita et al. 2009). One of the reasons is that observation records for discussing soil-building
interaction have not been accumulated sufficiently (Tobita et al. 2009).
College of science and technology, Nihon university, has established the earthquake observation
network for the ground and school buildings in Funabashi Campus as major subjects for under-
standing input earthquake motions to structures and their behaviors during earthquakes (Naka-
mura et al. 2006). Therefore, in this paper, the authors focus on a school building, which is a
typical low- and medium-rise building, and discuss inertial and kinematic interactions. The soil
and building is modeled with lumped mass system, respectively. The dynamic substructure
method is applied to connect the soil and building with interaction springs. For the interaction
springs, solutions of the circular foundation by the theory of elastic half-space are applied.
Moreover, inputs in the case of with and without excavation forces are discussed.
2 TARGET SITE AND OBSERVATION RECORDS
2.1 Subject structure
The subject structure is Building 14 of a steel framed reinforced concrete (SRC) structure lo-
cated around the center of the campus shown in Figure 1. Figures 2 and Photo 1 show outlines
and an overview of the building, respectively. The north and east wings of Building 14 are in L-
267
type form and neighbor each other. In the foundation of both wings, cast-in-situ concrete piles
are placed at 52 spots (27 for the north wing and 25 for the east wing) with a bearing stratum of
GL-26m. Superstructures of both buildings are connected through expansion joints. The subject
for this study is the north wing, which has five stories above the ground (steel framed structure)
and one story under the ground (reinforced concrete structure). We focus only on the records
obtained in the lengthwise direction (X-direction) of X5-Y8 in RFL. The X-direction almost
corresponds to the EW-direction. The east wings behaviors are excluded because observation
has been just started therefore both quantities and quality of the data are not sufficient. More-
over, the behavior of north wings Y-direction is also excluded because it is presumed to be
greatly affected by the east wing. Some in-situ soil investigation results indicated the building
region is almost on horizontally layered soil locally (Adachi et al 1995).

Building 14
(The north wing)
Point-D
N
0m 50m 100m
Blg.8
Blg.4
Blg.13
Blg.9
Blg.12
Blg.11
Blg.10
Blg.6
Blg.5
Blg.3
Blg.2
Blg.1
Blg.7
Sports
hall
Library

Figure 1. Campus layout.

unit:mm

RFL
B1FL
5FL
4FL
3FL
2FL
1FL GL
X1 X2 X3 X4 X5 X6 X7 X8 X9
2
2
,
4
0
0
6
,
4
5
0

(1) Vertical section view.

(2) Horizontal section view.
60,600
X1 X2 X3 X4 X5 X6 X7 X8 X9
Y9
Y8
Y7A
Y7
Y6
Y5
1
9
,
6
5
0

268

(3) Pile location.
X1 X2 X3 X4 X5 X6 X7 X8 X9
Y9
Y8
Y7A
Y7
Y6
Y5
Y5A
Figure 2. Subject building.

The north wing

The east wing

Photo 1. A view of subject building.
2.2 Free field
Figure 3 shows in-situ soil investigation results conducted at Point-D shown in Figure 1. Inves-
tigation depth was from GL-46.5m to GL0m. From GL-46.5m to GL-20m, the S-wave velocity
(Vs) propagation varies in the range of 300-450m/s and it suddenly falls from GL-15m to the
ground surface. As for soil properties, the ground surface is Kanto loam, the soil down to GL-
20m is clay and loose sand, and those deeper are dense sand layers containing silt at intervals.
N-value exceeds 50 in the region of GL-26m, which is a bearing stratum of Building 14. The
vertical array observations are carried out at Point-D. In this study, ground surface behaviors of
Point-D are assumed free field behaviors.

0
10
20
30
40
Vs (m/s) Damping
Density (t/m
3
)
Soil investigation
Lumped mass model
Multiple reflection model

seismograph

Figure 3. The results of in-situ soil investigation at Point-D.

269
2.3 Observation result
The transfer functions estimated by observed records to be referred as an index of numerical
model creation are described. Transfer functions are estimated by Hv estimation (Bendat et al.
2000) for ten records of small earthquakes. Figure 4 shows epicenter distributions of those small
earthquakes. Epicenters of those small earthquakes are seen around Kanto and their epicentral
distances are below 60km. Figure 5 shows acceleration response spectrum at GL0m of Point-D.
Dominant frequencies are seen around 5.5Hz in each record. Figure 6 shows transfer functions
of the free field system (GL0/GL-46.5m) and soil-building system (RFL/GL-46.5m).

M5
M4

Figure 4. Epicenter distributions.

0.01
0.1
1
10
100
1000
1 1

0
Frequency (Hz)
0.2

Figure 5. Acceleration response spectrum.

0
10
20
30
40
50
60
0 2 4 6 8 10
Frequency (Hz)
0
5
10
15
20
25
30
35
0 2 4 6 8 1

0
Frequency (Hz)

(1)GL0/GL-46.5m. (2)RFL/GL-46.5m.
Figure 6. Transfer functions.
270
3 NUMERICAL MODEL
3.1 Interaction model
Figure 7 shows numerical interaction models. The soil and building were modeled with lumped
mass system. The dynamic substructure method is applied to connect the soil and building with
interaction springs for the purpose of examining inertial interaction effects. Rocking behavior is
ignored because the number of pile foundations placed in the subject building is large and only
the lengthwise behavior is focused on.
Model-A is the case without excavation force and, Model-B is that with excavation force. Re-
sponses and excavation force is calculated from the earthquake responses obtained from a one-
dimensional multiple reflection model. They are inputted to multiple points at the same time as
shown in Figure 8.

(1) Model-A. (2) Model-B.
Figure 7. Interaction model.

Figure 8. Considering excavation force.
271
3.2 Ground model
First, a lumped mass model of the soil is examined. A depth range of the modeling is from
GL0m to GL-46.5m. Vs distributions were obtained by revising the in-situ soil investigation re-
sults. The values used for this model are shown in Figure 3. Moreover, in order to assure that
the soil behaves as a free soil without being affected by building's behaviors, a total mass of the
soil was determined 1,000 times as large as the total mass of the subject building.
Next, a one-dimensional multiple reflection model is examined to evaluate excavation forces.
The soil specifications used for Figure 3 are already shown. Same as the lumped mass modeling,
we revised the Vs distribution so that the dominant frequency of observed values can corre-
spond to the model.
Transfer functions of GL0/GL-46.5m obtained from the lumped mass model and multiple re-
flection model are shown in Figure 9 together with observed values.
Over 7Hz for the multiple reflection model, differences occur from observed value however we
focus on effects of excavation forces here therefore we presume that these differences do not af-
fect the following discussions.

0
5
10
15
20
25
0 2 4 6 8 10
Observed
Lumped mass model
Multiple reflection model
Frequency(Hz)

Figure 9. Transfer functions (GL0/GL-46.5m).
3.3 Interaction springs
A foundation part of the subject building is rectangle. However, the authors predicted that there
would be no significant differences between the spring value, which is a detailed explanation for
the subject building, and approximate solutions. Therefore, in this study, we used a simple equa-
tion to substitute the embedded rectangular foundation for a circular foundation with a radius of
r
0
whose area is equivalent to the rectangular foundation as shown in Figure 10 (1).
For the interaction springs between the bottom face of the building's embedment and the soil,
we applied equations derived from Tajimi's vibration admittance theory (Tajimi 1968). As
shown in Figure 10 (3), the authors discuss the case that rigid distribution (RD) is assumed as
reaction distribution of the soil in the case that a circular rigid plate on an anti-infinite soil is ex-
cited horizontally on the interaction spring at the bottom face, and in the case of flat distribution
(FD). Novak's interaction springs are used as the interaction springs between the side face of the
structure's embedment and the soil. For Model-B with excavation force considered, dynamic
soil spring is approximated to a static spring with an assumption that Novaks interaction spring
on the building side are assumed uniform.

(1) Substitution of foundation. (2)Side, bottom, soil. (3)Circular rigid plate horizontally excited.
Figure 10. Substitution of foundation and circular rigid plate horizontally excited.
272
3.3.1 Effects of structures stiffness
Effects of the buildings stiffness on the dynamic characteristics are discussed by comparing
transfer functions of the numerical model with observed values. Table 1 shows values of mass
and stiffness in the case that the subject building is substituted to lumped mass system. The
mass was calculated from design drawings. For the stiffness, the values calculated from the de-
sign drawings were multiplied by 1.6 times so as to correspond it to the first dominant fre-
quency of transfer functions obtained from the observed values. Figure 11 shows transfer func-
tions of RFL/GL-46.5m obtained with the parameters shown in Table 1. It was reported that
stiffness often do not correspond with small earthquake and/or microtremor observation results
when the specifications in the design drawings were used. A similar tendency was confirmed in
this study. This is because of the effects of so-called secondary members such as window
glasses, window frames and the walls between rooms, the authors presume.
Figure 11 also shows results estimated with RD and FD for reaction distribution of the bottom
face. As mentioned later, the behaviors do not greatly vary in either case. In other words, dy-
namic characteristics of the transfer function expressing roofs / foundations of the building de-
pend only on the buildings stiffness. We assume that the buildings stiffness behavior was
properly evaluated hence proceed to the following discussion.

Table 1. Parameters of the subject building.

Mass (t) Stiffness (kN/m)
m
SR
950.9
m
S5
824.9 k
S5
461744
m
S4
830.9 k
S4
669680
m
S3
840.7 k
S3
949616
m
S2
1020.0 k
S2
1031408
m
s1
2266.8 k
S1
1182112
m
s0
3456.2 k
S0
61573400

0
10
20
30
40
50
60
0 2 4 6 8 10
Observed
RD
FD
Frequency (Hz)

Figure 11. Transfer functions (RFL/GL-46.5m)

3.3.2 Effects of interaction spring
Comparing the transfer functions of the numerical model with the observed values, we focus on
the effects of differences in reaction distributions of the bottom face spring given to dynamic
characteristics. We have discussed the case of RD for a bottom face spring of the building and
that of FD, same as the transfer functions in Figure11. There are no significant differences be-
tween those two cases. The subject soil is comparatively soft and has a rigid foundation on it
therefore we use RD for the following study.
Moreover, great variations are not seen even after changing the spring on the bottom face of the
building therefore we cannot expect dramatic changes from a discussion on pile models.
273
4 EARTHQUAKE RESPONSE ANALYSIS
4.1 Input earthquake motion
An earthquake response analysis is performed and interaction effects are duscussed. EW-
direction record observed at GL-46.5m of Point-D shown in Figure 12 is used as an input earth-
quake motion which was recorded over 100gal on the ground surface. The earthquake wave in
Figure 12 is inputted into the bottom of the each model and was performed an earthquake re-
sponse analysis in time domain.

-50
-25
0
25
50
0 10 20 30 40 50 60 70 80
Time (s)

Figure 12. Input earthquake motion.
4.2 Comparison of acceleration response spectrum
Figure 13 shows absolute acceleration response spectrum obtained from the top of the models as
a result of the earthquake response analysis. The figure also shows observed values in the build-
ing RFL. The peak frequencies almost agreed with the observed values in Model-A however the
response amplitudes were much lower than the observed values. On the other hand, both the
peak frequencies and the response amplitude agreed with observed values in Model-B.

(1) Observed value. (2) Without excavation force. (3) With excavation force.
(Model-A) (Model-B)
Figure 13. Absolute acceleration response spectrum at RFL.
5 CONCLUSION
In this study, the authors have discussed inertial and kinematic interaction effects for a SRC
school building that has five stories above the ground and one story under the ground for which
earthquake observations are performed. In conclusion, it has been revealed that inertial interac-
tion effect on the building's behavior during an earthquake is small and only those of input
should be considered. The followings are the conclusion obtained from this study.
1) Effects of interaction springs on the subject building's behavior during small earthquake is
small and such behavior almost depends on the buildings stiffness.
2) The subject building's behavior during an earthquake which was recorded over 100gal on
the surface ground is greatly affected by conditions of the soil around the embedment of the
building and we simulated building responses almost successfully by considering excavation
force.
0
200
400
600
800
1000
1200
1 10
h=0.05
h=0.20
h=0.40
A
c
c
e
l
e
r
a
t
i
o
n

r
e
s
p
o
n
s
e
(
g
a
l
)
Frequency(Hz)
0
200
400
600
800
1000
1200
1 10
h=0.05
h=0.20
h=0.40
A
c
c
e
l
e
r
a
t
i
o
n

r
e
s
p
o
n
s
e
(
g
a
l
)
Frequency(Hz)
0
200
400
600
800
1000
1200
1 10
h=0.05
h=0.20
h=0.40
A
c
c
e
l
e
r
a
t
i
o
n

r
e
s
p
o
n
s
e
(
g
a
l
)
Frequency(Hz)
274
REFERENCES
Hanada, Kazufumi 1987. The modal Identification and System Identification for the Soil-Structure Sys-
tem. The report of Central Research Institute of Electric Power Industry UO5: (in Japanese)
Tobita, J., Fukuwa, N., Kojima, H. & Hamada, E. 2009. Development of Soil-Structure Earthquake Re-
sponse Observation for Evaluation of Dynamic Characteristics of Buildings, Journal of JAEE 9(3):
37-56. (in Japanese)
Nakamura, Masataka (The earthquake motion and earthquake-resistant structure research group of CST
Nihon university) 2006. Earthquake observation system in Funabashi area, the state of the data man-
agement: Journal of the research institute of science and technology Nihon university 112: (in Japa-
nese)
Adachi, H. (The earthquake motion and earthquake-resistant structure research group of CST Nihon uni-
versity) 1995. Study on earthquake motion characteristics based on earthquake observation system in
Narashino campus, Special report of the research institute of science and technology Nihon university
19: (in Japanese)
Bendat, J.S. & Piersol, A.G. 2000. Random data analysis and measurement procedure. New York: Wiley
Interscience.
Tajimi, H., Soil-structure interaction. 1968. Earthquake engineering. Tokyo: Shokoku-sha. (in Japanese)

275

ABSTRACT: Soil interventions applied locally at the foundation level of structures result inevi-
tably in the modification of the foundation soil properties. The modified stiffness of the soil be-
low the structures affects in some extent both seismic wave propagation and soil-structure inter-
action mechanisms that may finally lead to a substantially different structural seismic response.
Such interventions may include either conventional soil mitigation techniques, oriented for up-
grading soil strength, or innovative solutions aiming at modifying beneficially the dynamic
properties of the oscillating system. Previous investigations have verified the beneficial or det-
rimental contribution of several soil foundation interventions to the seismic response of single
degree of freedom (SDOF) structures. This paper extends the study to multi-degree of freedom
(MDOF) structures, emphasizing on frame-type structures with seismic response dominated by
the fundamental oscillation mode. The numerical investigation includes finite element modeling
of the soil-structure system, utilizing validated numerical procedures and codes. The modifica-
tion of MDOF seismic response resembles the one obtained for SDOF structures, revealing the
existence of specific frequency ranges where significant reduction of structural response can be
achieved.
1 INTRODUCTION
Foundation soil interventions related to the increase of soil strength and stiffness are tradition-
ally employed to improve soil bearing capacity and reduce settlements under static and seismic
loading. However, recent studies have alternatively examined soil interventions considering the
modified seismic wave field imposed on the foundation, aiming at a possible upgrade of the
structural dynamic response. Evidently, any interference to the foundation soil properties may
modify the dynamic characteristics of the soil-structure system and thus affect soil-structure in-
teraction mechanisms, leading to a different seismic behavior of the superstructure. Along these
lines, an extended investigation has been presented by Yegian & Catan (2004) and Yegian &
Kadakal (2004), utilizing a smooth synthetic liner placed within the soil deposit. This innovative
type of intervention aims at dissipating seismic energy through slip deformations along the liner
interface. The soil-structure system enclosed by the low friction liner is practically isolated from
the underlying soil during ground shaking, leading to substantially reduced amplitude of the
seismic motion that excites the superstructure. A similar concept, of interposing an artificial soil
layer between the superstructure and the foundation soil, has been examined by Doudoumis et
al. (2002). The low shearing resistance of the artificial layer reduces the inertia forces imposed
on the superstructure, thus resulting in beneficial structural response compared to the non-
slipping foundation case, especially for low-rise buildings.
In the field of train and machine induced vibrations, numerous studies investigate the use of
open or in-filled trenches with flexible or stiff materials in order to modify the wave propaga-
tion pattern in the soil media [Kattis et al. 1999, Andersen and Nielsen 2005]. Despite the dif-
ferent nature of this specific problem that mostly involves surface waves, the common trend to
explore artificial interference to the wave propagation pattern before approaching the structures
location is obvious.
Recently, the effect of foundation soil interventions on the seismic response of structures has
been investigated in the case of single-degree-of-freedom (SDOF) structures (Kirtas & Pitilakis,
2009). Several subsoil interventions, including commonly applied as well as innovative tech-
niques, have been investigated, revealing a beneficial or detrimental role on the seismic behav-
Effect of Foundation Soil Interventions to the Seismic Response
of Mdof Structures

K. Pitilakis, E. Kirtas & E. Rovithis
Department of Civil Engineering, Aristotle University of Thessaloniki, Greece
276
ior of the superstructure. In the present study the investigation is extended considering multi-
degree-of-freedom (MDOF) structures. Two-dimensional frame-type structures are selected,
characterized by the dominant role of the fundamental oscillation mode. The analysis is per-
formed in the time-domain utilizing validated numerical models, whereas results are evaluated
both in the time and frequency domain and compared to the corresponding SDOF cases.
2 SUBSOIL INTERVENTIONS EFFECT IN THE CASE OF SDOF STRUCTURES
The foundation soil intervention techniques employed during the investigation of the response
of MDOF structures, have been previously examined in the case of SDOF structures (Kirtas &
Pitilakis, 2009) as schematically presented in Figure 1. A SDOF structure with surface rigid
foundation resting on a homogeneous soil deposit serves as the benchmark soil-structure system
(Figure 1a), whereas the soil interventions applied in the foundation area comprise of:
- Soil-stiffening (Figure 1b): increase of the foundation soil stiffness is expected to mod-
ify both seismic waves propagating through the foundation soil and dynamic properties
of the system, resulting in a different soil-structure interaction mechanism
- Stiff diaphragms (Figure 1c): construction of stiff diaphragms upgrades the foundation
subsoil strength and modifies the dynamic properties of the system
- Soft horizontal layer (Figure 1d): inclusion of a horizontal layer of small shear resis-
tance could serve as a seismic wave filter reducing the amplitude of the seismic motion
transmitted through the soft layer
- Soft diaphragms (Figure 1e): flexible diaphragms placed vertically next to the founda-
tion may partially isolate the soil-structure area and shift the system response to higher
oscillation periods
- Soft caisson (Figure 1f): the combination of flexible diaphragms and soft horizontal
layer would form an isolated soil-structure area, altering systems dynamic properties
and input motion characteristics
The numerical investigation of the interventions effect on the seismic response of single-
degree-of-freedom (SDOF) structures has been performed utilizing a validated numerical
scheme, as presented in Kirtas et al. (2009) and Kirtas & Pitilakis (2009). In order to evaluate
the imposed response modification, the notion of a response ratio was introduced, as the ratio
of the response Fourier spectra in the modified case to the corresponding response in the bench-
mark system, obtained in selected locations of the system. Thus, values of the response ratio be-
low unity are indicative of mitigation efficiency of the structural response in the examined fre-
quencies. On the other hand, response ratios exceeding unity demonstrate a possible detrimental
effect on the systems seismic performance.

(a) (b) (c)
(e) (d) (f)
2B
hstr
mstr
D D
D
D
D

Figure 1. Examined intervention techniques.
Indicative results of the interventions effect in the case of a medium rise building are depicted
277
in Figure 2. It is mentioned that the fundamental period of the examined structure is equal to
0.4s considering fixed-base conditions. The results are presented in terms of superstructure ac-
celeration ratio, e.g. the ratio of the acceleration Fourier spectrum at the top of the structure, ob-
tained from the analysis of the system after the intervention, to the corresponding value consid-
ering the benchmark soil-structure case. This particular response ratio is characteristic of the
overall structural response when SDOF structures are examined. Numerical analysis results pre-
sented in the plots of Figure 2 correspond to foundation soil types B and C of EC8, whereas
both structural mass and height have been properly taken into account during the 2D plain strain
modeling procedure (Kirtas, 2007).
It is observed that soil stiffening and stiff diaphragms solutions produce a response ratio
above or close to unity at the period range of interest, regardless of the soil category, corre-
sponding to either detrimental or no effect on the structural seismic response. Moreover, the soft
horizontal layer technique results in a superstructure acceleration ratio very close to unity, with-
out substantially modifying the superstructures dynamic behavior. This reveals that the pre-
intervention state is more or less retained.
Soft diaphragms and soft caisson on the other hand present quite a different effect. A wide
frequency range is observed, near the fundamental period of the structure, where the ratio is
constantly below unity, indicating a possible mitigation of the structural seismic response. Es-
pecially in the case of the soft caisson, the beneficial contribution is extended for periods higher
than 1.0s, since the seismic isolation of the soil-structure system leads the fundamental period to
large values. These conclusions obtained from the observation of the superstructures response
ratio were further verified in the time domain, utilizing the respective acceleration time-histories
(Kirtas & Pitilakis, 2009).

0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1. 0
T ( sec)
0. 00
1. 00
2. 00
3. 00
4. 00
5. 00
0. 00
1. 00
2. 00
3. 00
4. 00
5. 00
6. 00
S
u
p
e
r
s
t
r
u
c
t
u
r
e

A
c
c
.

R
a
t
i
o
Soi l C ( Tst r =0. 4s)
Soi l St i f f eni ng
St i f f Di aphr agms
Sof t Layer
Sof t Di aphr agms
Sof t Cai sson
Soi l B ( Tst r =0. 4s)

Figure 2. Indicative intervention results for a T=0.4s structure (fixed base value).

Concluding the investigation of subsoil interventions on SDOF structures, it was observed
278
that subsoil stiffening may lead to unfavorable seismic behavior, increasing the seismic loading
sustained by the superstructure. Notwithstanding soil strength enhancement and excessive set-
tlement reduction, this particular method may have a detrimental effect on the superstructures
seismic response. The performance of stiff diaphragms, although slightly different from the soil
stiffening technique, didnt prove efficient in terms of reducing structural seismic response. The
imposed effect of the intervention is rather negative, especially when stiff structures are founded
on soft soil deposits. This possible detrimental effect of interventions similar to (b) and (c) of
Figure 1 is not normally considered in engineering practice.
On the other hand, the innovative approach of low-stiffness elements introduced in the foun-
dation soil seems to provide an interesting alternative in several soil-structure cases, albeit the
placement of the soft horizontal layer alone resulted inadequate to provide a seismic wave filter
for the upwards propagating seismic motion. It is noted though, that the construction of the layer
directly under the foundation was not examined since it is related to superstructure isolation
techniques. The partial isolation of the soil-structure system, achieved using vertical soft dia-
phragms, introduces a promising solution towards the reduction of the structural seismic re-
sponse. The mitigation effect is quite significant, especially with increasing structural mass.
Most importantly, the observed beneficial behavior is pronounced when resonance phenomena
are generated in the benchmark case, leading to an efficient reduction of the large acceleration
levels imposed on the superstructure. Finally, the combination of intervention techniques (d)
and (e) of Figure 1 forms a soft caisson that isolates a well-defined soil-structure area. This type
of intervention shifts structural response to significantly higher periods, beyond the frequency
range of common earthquakes, resulting thus in a beneficial effect on the structural dynamic be-
havior. However, an increase of the soil deformations and structural displacements should be
expected and taken into consideration in design procedures. It should be noted that the use of
reduced stiffness elements in the foundation soil raises several issues, including the develop-
ment of new materials and aging effects, which are still in the stage of research. Nevertheless,
the numerical examination of such interventions is important since it may lead to a different
perspective towards mitigating structural seismic response.
3 EXAMINED CASES CONSIDERING MDOF STRUCTURES
Given the effect of subsoil interventions on the seismic response of SDOF structures, it was
considered interesting to extend the investigation to MDOF structures. For this reason, two-
dimensional multi-storey frame-type structures have been employed, reproducing more closely
the majority of existing building types compared to simplified SDOF systems.
The numerical model and methodology employed have been thoroughly described in Kirtas
et al. (2009) and Kirtas & Pitilakis (2009), following the same principles with a soil-structure
investigation scheme that was validated using physical experiments in centrifuge apparatus. The
analysis takes place in the time-domain whereas the results are evaluated both in time and fre-
quency domain. An important point of the evaluation stage, as has been demonstrated in the
previous studies, is that the use of the response ratio eliminates the effect of the soil deposit
depth and of the particular characteristics of the utilized input motions. Therefore, any conclu-
sions derived from the analysis can be generalized to excitations of different frequency content
in various soil deposit depths.
Three different frame-type structures are utilized to investigate the effect of soil interventions
depicted in Figure 1, corresponding to three different fundamental period values equal to 0.2s,
0.4s and 0.6s for fixed-base conditions.
The structural mass is taken into account normalized by a soil mass quantity as (Wolf 1985):
str
norm 3
m
m
B
=
(1)
where m
str
=superstructure mass, =soil density and B=characteristic foundation dimension
(half the foundation width for strip foundation type).
Structural height is also normalized using the characteristic foundation dimension, according
to the expression:
279
str
norm
h
h
B
= (2)
where h
str
=superstructure height. For all examined cases considering MDOF structures,
m
norm
=2 and h
norm
=1 values have been selected during the numerical investigation.
In order to compare the imposed effect amongst the examined soil-intervention-structure
cases, it was deemed necessary to establish an equivalence between the SDOF and MDOF soil-
structure interaction systems. The relation between the dynamic characteristics of MDOF struc-
tures and equivalent SDOF oscillators is given in the following expressions (Wolf, 1985):
2
j j
j
SDOF 2
j j
j
m
m
m
| |
|
\ .
=

(3)
j j j
j
SDOF
j j
j
mh
h
m
=

(4)
where m, h and are the mass, height and modal storey displacement (1st mode) of storey j
of the MDOF structure.
The correlation between MDOF and SDOF structures is presented in Table 1, indicatively for
a structure with T=0.2s under fixed-base conditions. The equivalent frame-type MDOF structure
is schematically presented in Figure 3. The mass ratio participating in the fundamental mode of
response for all the examined MDOF structures lies between 83-88%.
The equivalence between the SDOF and MDOF soil-structure systems is verified using the
transfer function of the structure, i.e. the structural top to base acceleration Fourier ratio. The
comparison of these transfer functions is presented in Figure 4, pertaining to foundation soil
conditions of type C according to EC8. The similarity of the transfer functions is obvious and
verifies the equivalence of the soil-structure systems for all three examined fundamental period
values. The effect of higher modes is minimal as can be observed in the respective plots at lower
period values.

Table 1. Correlation between MDOF and SDOF structure (indicative for T=0.2s).
MDOF structure
SDOF structure
Storey (j)
Modal dis-
placement
j
Height z
j
Mass m
j

1
st
0.0230 3m 44t
2
nd
0.0575 6m 44t

2
j j
j
2
j j
j
m
m 146.1t
m
| |
|
\ .
= =

3
rd
0.0861 9m 44t

4
th
0.1038 12m 44t

j j j
j
j j
j
mh
h 9.0m
m
= =

280

Figure 3. Equivalent MDOF structure for T=0.2s SDOF oscillator (fixed base value).

0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1. 0
T ( sec)
0
5
10
15
20
25
30
0
5
10
15
20
25
30
35
0
5
10
15
20
25
30
0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1. 0
0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1. 0
St r uct ur e Tst r =0. 2s
S
t
r
u
c
t
u
r
a
l

T
r
a
n
s
f
e
r

F
u
n
c
t
i
o
n
Mdof
Sdof

Figure 4. Transfer functions of MDOF and SDOF structures founded on soil type C conditions.

281
4 EVALUATION OF INTERVENTION RESULTS
Direct comparisons of the subsoil interventions effect on MDOF and equivalent SDOF struc-
tural response under simplified harmonic motions are presented in Figure 5 (Kirtas et al. 2006).
Both base and superstructure acceleration ratios exhibit almost identical variation for the exam-
ined soil-structure-intervention configurations in the frequency domain. As it has already been
noted in the case of SDOF structures, increase of the foundation soil stiffness presents a rather
detrimental effect, depending on the particular characteristics of each soil-structure system. On
the other hand, soft diaphragms or soft caisson techniques seem to mitigate structural response
within the frequency range of interest. The soft horizontal layer case was not examined since
preliminary findings pointed out minimal effect on the overall response.

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
T (sec)
0
0.5
1
1.5
2
0
0.5
1
1.5
2
0
0.5
1
1.5
2
0
0.5
1
1.5
2
Sdof structure
Frame structure
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
T (sec)
0
1
2
3
0
1
2
3
0
1
2
3
0
1
2
3
Sdof structure
Frame structure
Base Acc. Ratio
Superstructure Acc. Ratio

Figure 5. Comparison of intervention effect between MDOF and SDOF structures (Kirtas et al. 2006).

Having established the effect of the interventions on MDOF structures for harmonic motions,
soft diaphragm and soft caisson techniques have been further examined under real earthquake
recordings applied at the base of the soil profile.
The similarity of the soft diaphragm intervention effect between MDOF and SDOF struc-
tures, in terms of base and superstructure accelerations, is obvious in Figures 6, 7. The plots
shown in these Figures correspond to soil type C and a typical application depth equal to 5B/3.
The differences observed in the low period range of the superstructure acceleration ratio (Figure
7) should be attributed to the second response mode of the MDOF structures, which cannot be
captured by SDOF oscillators. This is also verified from the comparison of acceleration ratios
computed at the base of the structures (Figure 6), which are almost identical, indicating that the
variation of the superstructure response ratios of Figure7 is exclusively attributed to the higher
modes of the MDOF building. Yet, due to the small participation ratio of the higher modes of
the multi-storey structures, the response in the time domain remains practically unaffected. In-
deed, as shown in Figure 8, the effect of soft diaphragms on computed accelerations at the top
of the structure is very much the same for both SDOF and MDOF cases, modifying similarly the
response of the benchmark soil-structure system.

282
0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1. 0
T ( sec)
0. 00
1. 00
2. 00
3. 00
4. 00
0. 00
1. 00
2. 00
3. 00
4. 00
0. 00
1. 00
2. 00
3. 00
4. 00
5. 00
B
a
s
e

A
c
c
.

R
a
t
i
o
Sdof st r uct ur e
Mdof st r uct ur e
T=0. 2s ( Soi l C, D=5B/ 3, mnor m=2)
T=0. 4s ( Soi l C, D=5B/ 3, mnor m=2)
T=0. 6s ( Soi l C, D=5B/ 3, mnor m=2)

Figure 6. Soft Diaphragms. Base acceleration ratio results for SDOF and MDOF structures.

S
u
p
e
r
s
t
r
u
c
t
u
r
e

A
c
c
.

R
a
t
i
o
0. 1 0. 2 0. 3 0. 4 0. 5 0. 6 0. 7 0. 8 0. 9 1. 0
T ( sec)
0. 00
2. 00
4. 00
6. 00
0. 00
1. 00
2. 00
3. 00
4. 00
0. 00
1. 00
2. 00
3. 00
4. 00
5. 00
Sdof st r uct ur e
Mdof st r uct ur e
T=0. 2s ( Soi l C, D=5B/ 3, mnor m=2)
T=0. 4s ( Soi l C, D=5B/ 3, mnor m=2)
T=0. 6s ( Soi l C, D=5B/ 3, mnor m=2)

Figure 7. Soft Diaphragms. Superstructure acceleration ratio results for SDOF and MDOF structures.
283

S
u
p
e
r
s
t
r
u
c
t
u
r
e

A
c
c
.

t
i
m
e
-
h
i
s
t
o
r
i
e
s

(
m
/
s
)
0 1 2 3 4 5 6
t ( sec)
- 10. 0
- 8. 0
- 6. 0
- 4. 0
- 2. 0
0. 0
2. 0
4. 0
6. 0
8. 0
- 10. 0
- 8. 0
- 6. 0
- 4. 0
- 2. 0
0. 0
2. 0
4. 0
6. 0
8. 0
10. 0
0 1 2 3 4 5 6
Benchmar k
I nt er vent i on
Mdof ( Soi l C, mnor m=2)
Sdof ( Soi l C, mnor m=2)

Figure 8. Soft Diaphragms. Acceleration time-histories of MDOD and SDOF structures (Tstr=0.4s).

Substantial intervention effect on mitigating the seismic response of MDOF structures was
also observed during the investigation of the soft caisson technique. The superstructure accel-
eration ratios between SDOF and MDOF structures are compared in Figure 9, revealing an al-
most identical pattern. Structural response in the time domain (Figure 10) is obviously reduced,
verifying further the efficiency of the soft caisson intervention method.
In general, the effect of both soft diaphragms and soft caisson interventions on the seismic re-
sponse of MDOF structures follows the same pattern that was previously obtained for equiva-
lent SDOF structures. It is therefore reasonable to extrapolate several of the conclusions ob-
tained during the study of SDOF structures, to equivalent multi-storey structures with dynamic
properties similar to those adopted herein.
It is fair to note that the present study concerns only 2D frame-type MDOF structures, charac-
terized by the large participation ratio of the fundamental response mode. However, even under
these restrictions, the practical importance of this investigation stems from the fact that several
interventions employing flexible construction materials may prove a promising alternative to
reduce the seismic response of multi-storey structures. On the other hand, it is equally important
to focus on a few disconcerting observations, regarding the unfavorable modification of the
seismic response of MDOF structures observed during the installation of intervention tech-
niques that increase the soil stiffness in the foundation area.

284
S
u
p
e
r
s
t
r
u
c
t
u
r
e

A
c
c
.

R
a
t
i
o
0. 2 0. 4 0. 6 0. 8 1. 0 1. 2 1. 4 1. 6 1. 8 2. 0
T ( sec)
0. 00
2. 00
4. 00
6. 00
0. 00
2. 00
4. 00
6. 00
0. 00
2. 00
4. 00
6. 00
8. 00
Sdof
Mdof
T=0. 2s ( Soi l C, mnor m=2)

Figure 9. Soft Caisson. Superstructure acceleration ratio results for SDOF and MDOF structures.

S
u
p
e
r
s
t
r
u
c
t
u
r
e

A
c
c
.

t
i
m
e
-
h
i
s
t
o
r
i
e
s

(
m
/
s
)
0 1 2 3 4 5 6 7 8 9 10 11 12
t ( sec)
- 10. 0
- 8. 0
- 6. 0
- 4. 0
- 2. 0
0. 0
2. 0
4. 0
6. 0
8. 0
- 10. 0
- 8. 0
- 6. 0
- 4. 0
- 2. 0
0. 0
2. 0
4. 0
6. 0
8. 0
10. 0
0 1 2 3 4 5 6 7 8 9 10 11 12
Benchmar k
I nt er vent i on
Mdof ( Soi l C, mnor m=2)
Sdof ( Soi l C, mnor m=2)

Figure 10. Soft Caisson. Acceleration time-histories of MDOF and SDOF structures (Tstr=0.2s).

285
5 CONCLUSIONS
Several subsoil interventions were examined towards mitigating structural seismic response, in-
cluding commonly applied methods in design procedures as well as innovative techniques re-
lated to flexible vertical diaphragms and/or soft horizontal layers of low shear resistance. Hav-
ing established the effect of these interventions on SDOF oscillators the investigation was
extended to frame-type MDOF structures.
The numerical results demonstrated almost identical effect of the examined subsoil interventions
between MDOF and equivalent SDOF structures, both for harmonic and real earthquake mo-
tions. Notwithstanding certain restrictions regarding the dynamic characteristics of the exam-
ined Mdof structures, subsoil intervention techniques introducing the partial or complete isola-
tion of the soil-structure system may provide improved seismic behavior in a wide frequency
range. On the other hand, interventions related to soil stiffening, traditionally employed to up-
grade structural stability under static loading, could lead to an unfavorable modification of the
soil-structure interaction mechanism, resulting in higher seismic forces imposed on the super-
structure.

6 REFERENCES
Yegian, M. & Catan, M. 2004. Soil isolation for seismic protection using a smooth synthetic liner. Jour-
nal of Geotechnical and Geoenvironmental Engineering, ASCE, 130(11): 1131-1139
Yegian, M. % Kadakal, U. 2004. Foundation isolation for seismic protection using a smooth synthetic
liner. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 130(11): 1121-1130
Doudoumis, I., Papadopoulos, P. & Papaliangas, T. 2002. A Low-Cost Base-Isolation System on Artifi-
cial Soil Layers with Low Shearing Resistance. Proceedings of the 12 European Conference on
Earthquake Engineering. London, England
Kattis, S.E., Polyzos, D. & Beskos, D.E. 1999. Modelling of Pile Wave Barriers by Effective Trenches
and their Screening Effectiveness. Soil Dynamics and Earthquake Engineering, 18(1): 1-10
Andersen, L. & Nielsen, S.R.K. 2005. Reduction of Ground Vibration by Means of Barriers or Soil Im-
provement Along a Railway Track. Soil Dynamics and Earthquake Engineering, 25(7-10): 701-716
Kirtas, E., Rovithis, E. & Pitilakis, K. 2009. Subsoil Interventions Effect on Structural Seismic Response.
Part I: Validation of Numerical Simulations. Journal of Earthquake Engineering, 13(2): 155-169
Kirtas, E. & Pitilakis, K. 2009. Subsoil Interventions Effect on Structural Seismic Response. Part II: Pa-
rametric Investigation. Journal of Earthquake Engineering, 13(3): 328-344
Kirtas, E. 2007. Numerical Investigation of Structural Seismic Behavior Enhancement Utilizing Subsoil
Interventions. PhD Thesis (in Greek with English abstract). Department of Civil Engineering, Aris-
totle University of Thessaloniki
Kirtas, E., Rovithis, E., Pitilakis, K. & Sextos A. 2006. Numerical Investigation of Potential Foundation
Intervention as a Means for Mitigating Seismic Risk. Proceedings of the 8th U.S. National Confer-
ence on Earthquake Engineering. San Francisco, California
Wolf, J.P. 1985. Dynamic soil-structure interaction. Prentice-Hall, Englewood Cliffs, NJ

286
1. INTRODUCTION

There are many cultural heritage in the world. Stone structures like stone pillar in Greece are
one of the cultural heritages. As for such structures, most of them are not sufficient for earth-
quake safety. In Japan, gravestone has simillar shape with an stone pillar. In this paper, earth-
quake reinforcement for gravestone is discucced as an example of reinforcement of stone struc-
ture which is a type of the cultural heritage. Gravestones are prone to be overturned easily
because they have a simple structure of being only piled up without any reinforcement. During
the 1995 Hyogoken-Nambu Earthquake, many gravestones were overturned and damaged. After
the earthquake, because of the increase of the awareness for disaster mitigation, various kinds of
reinforcement measures for preventing from overturning of gravestone have been developed and
applied in practice, such as gluing with a bonding agent between gravestones, adhesion with an
adhesive agent, interlocking with steel bars in the holes of gravestones. However, there remain
many un-reinforced gravestones, mainly built before the 1995 Hyogoken-Nambu Earthquake.
Dynamic behavior of a gravestone during an earthquake is very complicated; it is thought that
the shaking table test on full-scale gravestone models is one of the reliable ways to study the
dynamic behavior and effectiveness of seismic retrofitting measure for gravestones. In this re-
search, three-dimensional shaking table tests on full scale gravestone models were conducted in
order to investigate the behavior of gravestones during strong motion and the effect of rein-
forcement measure ([1],[2],[3]Furukawa et al., 2006, 2008a, 2008b, [4]Kiyono et al., 2007,
[5]Miwa et al., 2007). At first, Japanese type and Western type gravestone without reinforce-
Shaking table test on seismic behavior of gravestome similar to
stone pillars with and without reinforcement
S. Miwa
a
, A. Furukawa
b
& J. Kiyono
c

a
Director, Research Institute of Technology, Tobishima Corporation, Chiba, JAPAN
b
Assistant Professor, Dept. of Civil Engineering, Kyushu University, Fukuoka, JAPAN
c
Associate Professor, Dept. of Urban Management, Kyoto University, Kyoto, JAPAN
ABSTRACT: In Japan, there are many gravestons which are similar shaped to the stone pil-
lars consturucted in the ancient age, these structures are usually cultural heritages. Gravestone
are easy to be damaged by earthquakes. Recently, many types of reinforcement have been de-
veloped to prevent gravestone from overturning. 3-dimensional shaking table tests of full-scale
gravestone models were conducted to investigate the seismic behavior and overturning mechan-
ism of the gravestones and the effect of reinforcement. Full scale Japanese-type and European-
type gravestone models without any reinforcement and gravestones with bonding, adhesive and
interlocking reinforcement were tested. Effectiveness of reinforcement measures was discussed
based on the pictures and acceleration measurements. The conclusions obtained from the expe-
riment are summarized as follows, 1) The main reason for damage of Japanese-type gravestones
without reinforcement during the earthquake is overturning by rocking vibration, 2) The main
reason for damage of European-type gravestones without reinforcement during the earthquake is
dropping from base stone caused by large displacement of sliding, 3) Japanese-type gravestones
were overturned by strong motion in compliance with JMA seismic intensity of 6+. European-
type gravestones were overturned by strong motion in compliance with JMA seismic intensity
of 7, 4) Reinforcement measures by gluing, adhesion or interlocking of gravestones are effective
for mitigating the seismic disaster to gravestones.

Keyword:stone heritage, gravestone, shaking table test, overturning, reinforcement

287
ment were tested to clarify the characteristics of normal gravestone with seismic intensity of 6-,
6+ and 7 in compliance with JMA Intensity scale. Next, gravestones with reinforcement by
gluing, adhesion or interlocking with steel bar were tested to study the effectiveness of the rein-
forcement measures.

2. MODELS OF GRAVESTONES

2.1 Gravestone without reinforcement
2.1.1 Japanese type gravestones

Figure 1 shows the models of gravestone. Table 1 shows the size and mass for each part of
the models. The type of gravestone used widely in Tokyo area was selected as the typical Japa-
nese gravestone. The first, second, third and fourth stones from the top are known as "Saoishi",
"Jodai", "Chudai" and "Shibadai", respectively. "Saoishi", "Jodai" and "Chudai" were installed
in such a way the center of the gravity was in a straight line with a vertical axis of symmetry.

2.1.2 Western type gravestones

In this type of gravestone, the top stone is known as "Saoishi", followed by "Jodai" and bot-
tom as "Shibadai" respectively. The stones were installed with a vertical axis of symmetry.

2.2 Gravestones with reinforcement by gluing and adhesion

Both bonding agent and adhesive agent are commonly used for joining separated objects or
materials. Bonding agent has low initial liquid viscosity.After joining, the liquid hardens
gradually and yields high bonding strength. On the other hand, adhesive agent is a semi solid
with initially high viscosity and low elasticity modulus. It's characteristics dose not change after
joining. In other words, the process of hardening is not essential for adhesive agent.
Gluing measure is to join the stones by elastic bonding agent. This method is widely used as
retrofitting measure for gravestones. Elastic bonding agent like denatured silicone has proper
material strength comparing with the strength of stone. Adhesive measure is to join the stones
by adhesive agent, like butyl gum with shock and energy absorption characteristics. It is to be
mentioned that both measures have issues pertinent to degradation of joining agent (bonding
agent or adhesive agent).

2.3 Gravestones with reinforcement by interlocking using steel bar

The interlocking reinforcement measure using steel bar is carried out by drilling holes at the
bottom of the top stone "Saoishi" and at the surface of the second stone "Jodai", followed by
inserting a steel bar, before joining the two stones with the steel bar. One steel bar is inserted
without bonding with mortar or bonding agent. Moreover, there is no reinforcement between
"Jodai" and "Chudai" as well as between "Chudai" and "Shibadai". The interlocking reinforce-
ment measure has been widely used. The seismic behavior of three types of gravestones with
different interlocking reinforcement measures, namely one with a long and thick steel bar, one
with a short and thin steel bar, one with two short and thin steel bars, were compared. For Mod-
el 6, the diameter and depth of hole were 21 mm and 210 mm, respectively. Besides, the diame-
ter and length of the steel bar were 18 mm and 400 mm, respectively . For Model 7 and Model
8, the diameter and depth of hole were 12 mm and 80 mm, respectively. Furthermore, the di-
ameter and length of the steel bar were 10 mm and 150 mm, respectively. There were two steel
bars in Model 8, and the distance between the two holes was approximately 115 mm. The loca-
tions of the holes for the steel bar on the bottom surface of 'Saoishi' are shown in Figure 2.

288

Figure 1 Gravestone models for Shaking table test

Table 1 Gravestone Models for Shaking Table Test (unit: mm)
Mo
-
del
Type Rein-
forced meas-
ure
Stee
l bar
size
Steel
bar
number
Saoi-
shi
Jodai Chu-
dai
Shi-
badai
Mi-
zubachi
1 Japa-
nese

No rein-
forcement
- - W242
D242
H636

Mass
108kg
W424
D424
H273

Mass
142kg
W606
D606
H303

Mass
322kg
W848
D848
H152

Mass
315kg
W484
D152
H272

Mass
46kg
2 Gluing - -
3 Adhesion - -
4 Interlock-
ing
d18
L400
1

5 Interlock-
ing
d10
L150
1

6 Interlock-
ing
d10
L150
2
7 Western

No rein-
forcement
- - W606
D303
H485
Mass
211.1kg
W758
D455
H212
Mass
197.8kg
-

W848
D848
H152
Mass
315kg
-

8 Gluing - -
9 Interlock-
ing
d18
L400
2
d: diameter, L: length, W: width, D: depth, H: height

c) Japanese type gravestone
with reinforcement by adhe-
sion measure
a) Japanese type gravestone
without reinforcement
e) Western type gravestone
without reinforcement
b) Japanese type gravestone
with reinforcement by
gluing measure
d) Japanese type gravestone
with reinforcement by Inter-
locking using steel bar
f) Western type gravestone with reinforce-
ment by Interlocking using steel bar
e) Western type gravestone with
reinforcement by gluing measure
289

a) model 4, 5 b) model 6

Figure 2 Location of the holes for the steel bar on the bottom surface of Saoishi

Figure 3 Acceleration response spectra and waveforms of Input motion for the test

Table 2 Maximum acceleration for each seismic intensity (m/s
2
)
JMA Inten-
sity
x y
6-
5.53 4.84 2.79
6+
8.94 8.18 4.68
7
1.170 1.049 7.33

Table 3 Result of shaking table test
Model Type
Reinforced
measure
Steel
bar size
Steel bar
number 6- 6+ 7
1
Japanese

No rein-
forcement
- - B D D
2 Gluing - - A A A
3 Adhesion - - A A A
4 Interlocking
d18
L400
1
B B E
5 Interlocking
d10
L150
1
B C E
6 Interlocking
d10
L150
2 B C E
7
Western

No rein-
forcement
- - B B D
8

Gluing - - A A A
9

Interlocking
d18
L400
2 B B B
A: No Damage, B: Displaced, C: Displaced, Mizubachi was overturned,
D: Saoishi was overturned or dropped, E: Saoishi was overturned with Jodai
290
2.4 Input motion

In order to investigate the vibration characteristics of gravestones which have different natu-
ral frequencies, artificial strong motion in horizontal directions were simulated by using ampli-
tude of the response spectra in compliance with the seismic design of highway bridges ([6]Japan
Road Association, 2002), which has flat amplitude in major frequencies, and phase characteris-
tics of the earthquake record obtained by JMA Kobe during the 1995 Hyogoken-Nambu Earth-
quake. Besides, vertical motion was simulated by adopting the spectra data for evaluating earth-
quake reistance of communication equipments ([7]NTT, 1998) and the phase characteristics of
the earthquake record obtained by JMA Kobe during the 1995 Hyogoken-Nambu Earthquake.
Acceleration response spectra and waveforms are shown in Figure 3. Strong motion in com-
pliance with JMA Intensity 6-, 6+ and 7 were simulated by using these artificial strong motions,
by multiplying with the same proportionality constant in three directions. Maximum values of
these strong motions are shown in Table 2. These strong motions are known as "6-", "6+" and
"7" hereafter in this paper.

3. RESULTS OF SHAKING TABLE TESTS

3.1 Summary

The behaviors of gravestones were recorded in the form of animations by digital cameras.
Table 3 shows the results of the shaking table test. Model 1, with no reinforcement, slid during
"6-", and overturned during "6+". The models reinforced with gluing agent or adhesion agent
did not overturn even at "7". The model reinforced with interlocking measure restrained from
overturning during "6+", but the top stone "Saoishi" and the second stone "Jodai" were over-
turned as one unit during "7".

3.2 Behaviors of gravestone during strong motion

3.2.1 Japanese type gravestone without reinforcement

Figure 4 shows the condition of the Japanese type gravestone model without reinforcement
during the shaking table tests. Typical behavior during "6-" was rocking of "Saoishi" and dis-
placements in association with rocking. "Saoishi" was not overturned, but was displaced to the
edge of "Jodai". "Jodai" and "chudai" were found rotated slightly. During "6+", "Saoishi" was
rocked hard, and was displaced to the edge of "Jodai". The model was finally overturned after
22 second from the start of shaking. During "7", "Saoishi" was rocked very intensely and was
overturned after 21 second from the start of shaking. For this model, the rocking of "Saoishi",
which has high height to width ratio, was predominant. As a result, the rocking caused overturn-
ing of the stone.

3.2.2 Western type gravestone without reinforcement

Figure 5 shows the condition of the Western type gravestone model without reinforcement
during the shaking table tests. "Saoishi" and "Jodai" rotated in horizontal plane during '6-'. The
stones rotated largely during "6+". "Saoishi" slid and dropped after 26 second from the start of
shaking. As for the Western type gravestone without reinforcement, typical behavior was ob-
served in the form of large movements of "Saoishi" by slide, which eventually caused dropping.
Because the height to width ratio of the Western type gravestone was smaller than that of the
Japanese type gravestone, rocking was not predominant and did not result in overturning of this
type of gravestone.

3.2.3 Gluing measure

Figure 6 shows the condition of the gravestone models with reinforcement by gluing during
the shaking table tests. Gluing measure exhibited its effect on the seismic behaviors of graves-
291
tone models. For the Japanese type gravestone models, movement did not occur even at '7', al-
though the side stone, known as "Mizubachi" moved or was overturned because there was no
bonding applied. As for the Western type gravestone models, movement was not observed even
at "7"for all parts. This confirmed the effectiveness of the gluing measure.

(a) before test (b) after 6- (a) before test (b) after 6-

(c) during 6+ (d) after 6+ (c) after 6+

(e) during 7 (15senonds) (f) after 7 (d) during 7 (25second) (e) after 7

(a) before test (b) after 6- (c) after 6+ (d) after 7
(1) Japanese Type

(2) Western Type
Figure 6 Test results of gravestones with reinforcement by gluing

Figure 4 Test results of Japanese type
gravestones without reinforcement
Figure 5 Test results of Western type
gravestones without reinforcement
292
3.2.4 Adhesive measure

Figure 7 shows the condition of the gravestone model with reinforcement by adhesion during
the shaking table tests. Adhesive measure also yielded effect its effect on the seismic behaviors
of the gravestone models. Gravestone parts including "Mizubachi", which was adhered in this
case, did not recorded any movement even at "7",

Figure 7 Test results of gravestones with reinforcement by adhesion

(a) during 6- (b) after 6- (c) during 6+ (d) after 6+ (e) during 7 (f) after 7
(1) Model 4 (1 long steel bar)
(2) Model 5 (1 short steel bar)

(3) Model 5 (2 short steel bar)
Figure 8 Test results of Interlocking measure using steel bar
(Japanese type gravestone models)
293
3.2.5 Interlocking measure using steel bar (Japanese type gravestone models)

Figure 8 shows the condition of the Japanese type gravestone model with reinforcement by
interlocking measure using steel bar during the shaking table tests. Steel bar was used for join-
ing "Saoishi" and "Jodai" in order to prevent rocking, sliding and overturning of "Saoishi".
Through joining, the two stones moved as one unit, and therefore, the center of gravity became
lower than the case in which "Saoishi" was able to move independently. Rocking was prevented
by low center of gravity. Large friction was expected because of the increase in weight by join-
ing the two stones,. Moreover, it was also expected that this kind of joint is able to prevent slid-
ing. However, this joint was found not effcective at "7". During "7", "Saoishi" interlocked with
"Jodai" by a steel bar was overturned as a unit. This suggested the necessity to reinforce be-
tween "Jodai" and "Chudai", and between "Chudai" and "Shibadai" at the same time.

It was found that steel bars with a larger diameter and a longer length as in Model 6 re-
strained rotational movement, which did not happen for steel bars with a smaller diameter and a
shorter length. Relative movement between "Saoishi" and "Jodai" was predominant in Models 7
and 8, both of which small steel bars were used. Relative movement should be reduced because
such a motion causes a more considerable damageto the gravestones. It is reckoned that steel
bars of large diameters and long lengths, e.g. that used in Model 6, should be used for this rein-
forced measure, concerning the simplicity in execution.

3.2.6 Interlocking measure using steel bar (Western type gravestone models)
Figure 9 shows the condition of the Western type gravestone model with reinforcement by in-
terlocking measure using steel bar during the shaking table tests. "Saoisi" interlocked with "Jo-
dai" by using a steel bar slid on "Shibadai" during all seismic intensities. During "7", "Saoisi"
and "Jodai" moved as one unit to the edge of "Shibadai". It is supposed that these stones would
be overturned in the same way as the Japanese type gravestone models, if the duration of strong
motion becomes slightly longer or the intensity of the motion becomes slightly larger. There-
fore, it is necessary to apply reinforce measure between "Jodai" and ''Shibadai".

Figure 9 Test results of Interlocking measure using steel bar
(Western type gravestone models)

4. CONCULUSIONS

The conclusions obtained from this research are summarized as follows:
1) The main reason for damage of Japanese type gravestones without reinforcement during the
earthquake is overturning of "Saoishi" by rocking vibration.
2) The main reason for damage of European type gravestones without reinforcement during the
earthquake is dropping from base stone caused by large displacement of sliding.
3) Japanese type gravestones were overturned by strong motion in compliance with JMA seis-
mic intensity of 6+. European type gravestones were overturned by strong motion in compliance
with JMA seismic intensity of 7.
4) Reinforcement measures by gluing, adhesion of gravestones are effective for mitigating the
seismic disaster to gravestones. Both gluing and adhesion measures have issues pertaining to
degradation of joining agent
294
5) Interlocking measure using steel bars is effective for strong motion up to '6+'. However,
"Saoishi" interlocked with "Jodai" by a steel bar was overturned as one unit during '7'. It is
therefore necessary to reinforce between "Jodai" and "Chudai", and between "Chudai" and
"Shibadai" at the same time.
6) It is reckoned that steel bars with larger diameters and longer lengths like that applied in
Model 6 should be used for this reinforced measure, concerning the simplicity in execution.

ACKNOWLEDGEMENT
A part of this research was supported by Grant-in-Aid for Scientific Research 'KAKENHI'
(19560490).
The Author would like to highly appreciate the member of Japan Stone Association for the
cooperation and support on this research. We also highly appreciate Prof. Hattori of Nihon Uni-
versity for providing the results of damage investigation ob gravestones during earthquakes.
We would like to express our gratitude to Dr. Hwakian Chai of Tobishima Corporation for his
kind suggestion of writing this paper.

REFERENCES

[1]Furukawa, A., Kiyono, J., Miwa, S. and Momiyama, T. (2006). Verification of Reinforce-
ment Measures for Gravestones Based on Shaking Table Test Using Full Scale Models, Journal
of Social Safety Science, No.8, 31-39, (in Japanese).
[2]Furukawa, A., Miwa, S. and Kiyono, J. (2008). Simple Anti-Earthquake Measure for Exhi-
bited Gravestones at Stone Stores, Journal of Social Safety Science, No.10, (in printing, in Japa-
nese).
[3]Furukawa, A., Miwa, S. and Kiyono, J. (2008). Effect of size and installation method of steel
bar on seismic performance of gravestones based on shaking table test and numerical test using
full scale models, Journal of Applied Mechanics, Japan Society of Civil Engineering, Vol.11,
(in printing, in Japanese).
[4]Kiyono, J. Miwa, S. and Furukawa, A. (2007). Experimental Study on Seismic Performance
of Gravestones with Interlocking Reinforcement and Seismic Isolation System, Journal of
Earthquake Engineering, Japan Society of Civil Engineering, Vol.29, 412-419, (in Japanese).
[5]Miwa, S., Kiyono, J., Furukawa, A. and Momiyama, T. (2007). Experimental study on seis-
mic performance of gravestones with bonding and adhesive reinforcement, Journal of Structur-
al Engineering, Japan Society of Civil Engineering, Vol.53A, 329-338, (in Japanese).
[6]Japan Road Association (2002). Seismic Design Specifications of Highway Bridges, Japan
Road Association.
[7]NTT (1998). Test of Seismic Performance for Communication Equipment, NTT.

295
Mechanical Representation of Dynamic Stiffness of Soil-
Foundation Systems

M. Saitoh
Graduate School of Science and Engineering, Saitama University, Japan

ABSTRACT: This paper presents an application example of a newly proposed model that can
simulate frequency-dependent oscillations in impedance functions of soil-foundation systems by
using a so called gyromass. A gyromass is an element defined as a unit system that generates
a reaction force due to the relative acceleration of the nodes between which the gyromass is
placed. This study also introduce a so-called soil-foundation emulator. This emulator is cre-
ated as a real mechanical system with springs, dampers, and mechanical gyromass elements
based on the proposed model. This emulator is to be used as a virtual soil-foundation system
supporting a superstructure excited by an experimental shaking table when the real soil-
foundation system is difficult to construct below the superstructure on the shaking table.

1 INTRODUCTION
In general, soil-foundation systems show various frequency-dependent impedance characteris-
tics, such as rapid oscillations or cut-off frequencies, depending on the types of foundations, soil
profiles, and the direction of excitations, etc. In contrast, many recent studies of structural dy-
namics have focused on the inelastic behavior of structural systems because the methodology of
performance-based seismic design, which has been applied to seismic codes and guidelines in
many countries, allows modeling of the inelastic behavior of structural systems. The frequency
dependency of soil-foundation impedance characteristics is usually considered in a numerical
method performed in the frequency domain, whereas the nonlinearity of superstructures is in the
time domain because the inelastic behavior of materials and structural members strongly de-
pends on the stress path being integrated stepwise. In actual, the effect of the frequency depend-
ency of the impedance functions upon the nonlinear responses of superstructures has not yet
been investigated, even qualitatively, because numerical techniques taking account of both fre-
quency dependency and nonlinearity are generally complicated and still immature. The most
powerful and acceptable tool in practice is a mechanical representation approximating the im-
pedance functions by frequency-independent spring, dashpot, and mass elements [e.g., Meek &
Veletsos (1974), Wolf & Somaini (1986), Nogami & Konagai (1986, 1988), Wolf (1994), Wolf
& Song (2002)]. The reason for this is that the frequency-independent mechanical elements can
be applied directly to conventional structural analysis. This mechanically represented system is
generally called a Lumped Parameter Model. In recent years, various types of lumped pa-
rameter models have been proposed for soil reaction, shallow foundations, embedded founda-
tions, etc. Although increasing the number of elements and degrees of freedom results in more
accurate fitting to the ideal impedance functions, in practical applications, simple models having
a small number of elements and degrees of freedom are used. In general, however, such simple
models show a moderate variation with frequency, and thus do not express the frequency-
dependent impedance functions with sufficient accuracy.
Saitoh (2007) proposed a model by which various types of frequency-dependent impedance
functions can be simulated using a mechanical element called a "gyromass". The gyromass is
frequency-independent and is defined as a unit system that generates a reaction force due to the
relative acceleration of the nodes between which the gyromass is placed. The advantage of the
use of this element is that a rapid change in the impedance functions with frequency can easily
be expressed by the element. In the present study, the impedance functions of a 33 piles em-
296
bedded in a homogeneous soil medium are simulated by the model for assessing the precision of
the model.
Moreover, it is noteworthy that this lumped parameter model can realistically be constructed
by using real mechanical elements as a so called soil-foundation emulator. The purpose of this
construction is that the soil-foundation emulator can represent a soil-foundation system on a
shaking table in experimental researches. In general, a structural system is constructed on a
shaking table for investigating its dynamic behavior when subjected to earthquake waves. How-
ever, only a superstructure is usually constructed because the soil-foundation system is difficult
to create due to the restriction of the space. It is known that the lack of a soil-foundation system
may change the natural periods and damping of the systems. The present study introduces the
status of this emulator under construction as an innovative research project supported by the
Ministry of Education, Culture, Sports, Science and Technology, Japan.
2 SOIL-PILE-SUPERSTRUCTURE SYSTEM STUDIED
One of the most impressive examples is the application to impedance functions of pile groups.
The reason for this is that impedance functions of pile groups in general show complicated os-
cillations with frequency. These oscillations usually occur due to in-phase and out-of-phase
pile-to-pile interactions, as explained in Dobry and Gazetas (1988). It is known that simulating
impedance functions of pile groups by using spring-dashpot-mass models is quite difficult since
some of these parameters often take negative values, which is considered to be unrealistic, and
in general, unstable in numerical calculations. Here, a 3x3 pile group is taken as an application
example using the published results of a rigorous solution (Kaynia and Kausel (1982)).
2.1 Introduction of Type II model
In the following simulation, one of the proposed models, called the Type II Model, is used.
Details of this model are described in Saitoh (2007); therefore, an explanation of the fundamen-
tals of this model is omitted here. Figure 1 shows that the Type II model consists of two types of
unit systems: one is the base system, and the other is the core system composed of a spring
and a unit having a gyro-mass
k
m and a dashpot arranged in parallel. The base system and
the core system are arranged in parallel. As described in Saitoh (2007), an additional core sys-
tem having different coefficients can be arranged in parallel with the Type II model in order to
enhance the precision of the simulation, if needed. The impedance functions of the Type II
model having multiple core systems shown in Figure 1 can be expressed by the following for-
mula:
c

( )
( ) | |
( ) ( )
(
0 0
1
2
0
2
2
2
0
0
2
0 0
1
2
0
2
2
2
0
2
0
2 2
0
2
0
0
1 1
1
1 a u
a a
a i a
a a
a a a
K a F
N
i
i i
i i
N
i
i i
i i i i
(
(
+
+
+
+
+
+ =

= =

|

|
)
,
(1)

where the coefficient K is equivalent to the static stiffness (
0
0 = a );
i
,
i
, and
i
| are, re-
spectively, the dimensionless coefficients of the damper and the gyro-mass, and the stiffness ra-
tio of the -th core system shown in Figure 1; and is the total number of core systems. i N
2.2 Effect of the number of core systems on simulation
Figure 2 shows a comparison of a horizontal impedance function of a 3 3 pile group (stiff-
ness ratio of pile to soil
1000 =
s p
E E
, slenderness ratio 15 = d H , Poissons ratio 4 . 0 = v ,
material damping of soil
s
, and distance ratio
05 . 0 = | 10 = d S , where is the axis-to-axis
distance of the piles and is the diameter of the piles) simulated by the Type II model with the
published results. In this comparison, the effect of the number of core systems upon the preci-
S
d
297
sion of the simulation is assessed. The properties used here are summarized in Table 1. Figure 2
indicates that the precision of the simulation gradually increases as the number of core systems
increases: in particular, the oscillations n the high frequency region appearing in the rigorous
impedance functions shows good agreement with the simulated results by using the Type II
model that consists of three core systems.

Figure 1. Type II Model.

3 SOIL-FOUNDATION EMULATOR
A project so called Soil-Foundation Emulator Project has been conducted since 2008, sup-
ported by the ministry of Education, Culture, Sports, Science and Technology, Japan as shown
in Figure 3. The role of this emulator in earthquake engineering is to enhance the precision of
the prediction of damage to structures subjected to earthquake waves in experimental researches
where shaking tables are used. In general, only a superstructure is constructed on a shaking ta-
ble: the soil-foundation system that supports the superstructure is difficult to construct due to
Table 1. Dimensionless coefficients of simple models simulating
impedance functions of a 3x3 pile group.

coefficient Base Core-1 Core-2 Core-3
Re.(
( )
s d
K K 9
9
)
0.56 0.56 0.56 0.56
0
7.80 0.16 0.16 0.00
0
0.00 1.60 1.60 1.60
1
| 2.00 1.67 1.65
1
3.90 4.20 4.20
1
7.00 8.50 9.50
2
| 1.20 0.25
2
0.50 0.45
2
0.90 1.80
3
| 1.25
3
0.40
3
0.80

1
f
f
u
2
f
K
C
M
2
u
11
u
1
m
1
c
1
k
12
u
2
m
2
c
2
k
13
u
3
m
3
c
3
k

298
the restriction of the space on the table. This soil-foundation emulator can represent the soil-
foundation system in a small space. The emulator is constructed based on the lumped parameter
model explained above. Therefore, this emulator is a passive-controlled soil-foundation system.
The merit of this passive-controlled system is that the system maintains the constant status of a
target impedance function while the superstructure shows strong nonlinearity in its structural
members. In general, active-controlled systems often show convergence problems when the
nonlinearity occurs in the superstructure. This emulator consists of springs, dampers, and gy-
romass elements. It is noted that an extremely large mass is necessary for expressing the oscil-
lation in impedance functions in general so that lamped parameter models using general mass
(not gyromass) are very difficult to be constructed as a real mechanical system on a shaking ta-
ble due to a limited space. In contrast, the model proposed in this study consists of no general
mass at all; only gyromass elements are used. As described below, a mechanically-created gy-
romass element in the present study is a markedly compact system owing to a specific mechani-
cal technique.

0.0 0.2 0.4 0.6 0.8 1.0
0
1
2
3
4
5
Real Part
Type II Model (Base)
Rigorous Solution
Imaginary Part
K
d
(
9
)
/
9
K
s
(
1
)
Dimensionless Frequency a
0
(=ea/c
s
)
0.0 0.2 0.4 0.6 0.8 1.0
0
1
2
3
4
5
Real Part
Type II Model (Core1)
Rigorous Solution
Imaginary Part
K
d
(
9
)
/
9
K
s
(
1
)
0
(=ea/c
s
)

0.0 0.2 0.4 0.6 0.8 1.0
0
1
2
3
4
5
Real Part
Rigorous Solution
Imaginary Part
K
d
(
9
)
/
9
K
s
(
1
)
0
(=ea/c
s
)
0.0 0.2 0.4 0.6 0.8 1.0
0
1
2
3
4
5
Real Part
Rigorous Solution
Imaginary Part
K
d
(
9
)
/
9
K
s
(
1
)
0
(=ea/c
s
)

Figure 2. Normalized horizontal impedance functions of a 33 pile group simulated by using Type II
model, showing (a) base system with no core system, (b) single core system, (c) double core systems, and
(d) triple core systems. Comparison with the rigorous solution of Kaynia and Kausel (1982). (Imaginary
part of the impedance is divided by the dimensionless frequency ;
0
a
( ) 9
d
K is the dynamic impedance of
a 9-pile group; is the static stiffness of a single pile.)
( ) 1
s
K
(a) (b)
(c) (d)
Base
Core1
Core2
Core3
Base
Core1
Core2
Base
Core1
Base
299
3.1 Mechanical System of gyromass elements
The mechanical analogy of the gyromass element is shown in Figure 4 (a). The analogy consists
of a rotational disk and a rod attached to the disk with strong friction, gears for instance. It is as-
sumed that the mass of the rod is negligible. The disk rotates with rotational acceleration u

as an external force is given to the rod. The relative acceleration of the rod with respect
to the fixed node at the right hand side is geometrically related to the rotational acceleration .
Consequently, the following relation between the external force and the relative acceleration
can be obtained:
F u
u

F
u
u m F = (2)

Figure 3. Overview of Soil-Foundation Emulator Project
(a) Soil-Foundation-
Superstructure System

(b) Soil-Foundation-
Superstructure Simulator
(c) Soil-Foundation Emulator (under construction)

where

2
r
J
m . (3)

Here, r is the distance from the center of the disk to the point where the rod is attached; is
the moment of inertia of the disk; and
J
m is the equivalent mass generated by the rotation of
the disk. Thus, the reaction force at the left hand side of the rod is identical to the product of the
equivalent mass m and the relative acceleration . Accordingly, the equivalent mass u m is
termed gyro-mass in order to be distinguished from ordinary mass. In order to increase the gy-
romass much larger value than the mass of the disk itself, the following conventional techniques
can be used in mechanical engineering: (1) increasing the radius of gyration of the disk by con-
centrating the mass at the edge of the disk; and (2) increasing the rotational acceleration of the
300
disk by combinations of disk gears. These techniques make the ordinary mass of the disk negli-
gible when compared with the gyromass in the system. Figure 4 (b) shows the real mechanical
system of the gyromass element, where the second technique is used in the system.
3.2 Future Planning
The total system has been under construction and will be accomplished in the end of 2009. The
results of this project will be presented in the next opportunity.

Figure 4. Mechanical analogy of gyromass element and its creation by using mechanical elements. (a)
Mechanical analogy, and (b) Mechanical system of gyromass element

rod

F
u
r
J
u
(a)
Gears
Rotational mass
(b)
4 REFERENCES
Meek, J. W. & Veletsos, A. S. 1974. Simple models for foundations in lateral and rocking motion. Proc.,
5th World Conf. on Earthquake Engrg., 2: 2610-2613.
Wolf, J. P. & Somaini, D. R. 1986. Approximate dynamic model of embedded foundation in time do-
main. Earthquake Engrg. & Struct. Dyn. 14: 83-703.
Nogami, T. & Nonagai, K. 1986. Time domain axial response of dynamically loaded single piles. J.
Engrg. Mech., ASCE 112(11): 1241-1252.
Nogami, T. & Konagai, K. 1988. Time domain flexural response of dynamically loaded single piles. J.
Engrg. Mech., ASCE 114(9): 1512-1525.
Wolf, J. P. 1994. Foundation vibration analysis using simple physical models. Eaglewood Cliffs, NJ,
Prentice-Hall.
Wolf, J. P. & Song, C. 2002. Some cornerstones of dynamic soil-structure interaction. Engrg. Struct. 24:
13-28.
Saitoh, M. 2007. Simple model of frequency-dependent impedance functions in soil-structure interaction
using frequency-independent elementsJournal of Engineering Mechanics, ASCE 133(10): 1101-
1114.
Kaynia, A., & Kausel, E. 1982. Dynamic stiffness and seismic response of pile groups. Research Report,
R82-03, Dept. of Civil Engrg., MIT, Cambridge, Mass.

ACKNOWLEDGEMENTS
This paper is based on the work supported by the ministry of Education, Culture, Sports, Sci-
ence and Technology, Japan, under Grant-in-Aid for Young Scientists (A) No. 20686031. Any
opinions, findings, conclusions or recommendations expressed in this material are those of the
author and do not necessarily reflect the views of the sponsors.
301
Nonlinear Dynamic Analysis of Partially Supported Beam-
Columns on Nonlinear Elastic Foundation Including Shear
Deformation Effect

E.J. Sapountzakis & A.E. Kampitsis
School of Civil Engineering, National Technical University of Athens,
Zografou Campus, GR-157 80 Athens, Greece

ABSTRACT: In this paper, a boundary element method is developed for the nonlinear dynamic
analysis of beam-columns of arbitrary doubly symmetric simply or multiply connected constant
cross section, partially supported on tensionless Winkler foundation, undergoing moderate large
deflections under general boundary conditions, taking into account the effects of shear
deformation and rotary inertia. The beam-column is subjected to the combined action of
arbitrarily distributed or concentrated transverse loading and bending moments in both
directions as well as to axial loading. To account for shear deformations, the concept of shear
deformation coefficients is used. Five boundary value problems are formulated with respect to
the transverse displacements, to the axial displacement and to two stress functions and solved
using the Analog Equation Method, a BEM based method. Application of the boundary element
technique yields a nonlinear coupled system of equations of motion. The solution of this system
is accomplished iteratively by employing the average acceleration method in combination with
the modified Newton Raphson method. The evaluation of the shear deformation coefficients is
accomplished from the aforementioned stress functions using only boundary integration. The
proposed model takes into account the coupling effects of bending and shear deformations along
the member as well as the shear forces along the span induced by the applied axial loading.
Numerical examples are worked out to illustrate the efficiency, wherever possible the accuracy
and the range of applications of the developed method.

1 INTRODUCTION
The study of nonlinear effects on the dynamic analysis of structural elements is essential in civil
engineering applications, wherein weight saving is of paramount importance. This non-linearity
results from retaining the square of the slope in the straindisplacement relations (intermediate
non-linear theory), avoiding in this way the inaccuracies arising from a linearized second
order analysis. Thus, the aforementioned study takes into account the influence of the action of
axial, lateral forces and end moments on the deformed shape of the structural element.
Besides, the admission of tensile stresses across the interface separating the beam from the
foundation is not realistic. When there is no bonding between beam and subgrade, regions of no
contact develop beneath the beam. These regions are unknown and the change of the transverse
displacement sign provides the condition for the determination of the contact region. Moreover,
due to the intensive use of materials having relatively high transverse shear modulus and the
need for beam members with high natural frequencies the error incurred from the ignorance of
the effect of shear deformation may be substantial, particularly in the case of heavy lateral
loading. The Timoshenko beam theory, which includes shear deformation and rotary inertia
302
effects has an extended range of applications as it allows treatment of deep beam (depth is large
relative to length), short and thin-webbed beams and beams where higher modes are excited.
When the beam-column deflections of the structure are small, a wide range of linear analysis
tools, such as modal analysis, can be used, and some analytical results are possible. During the
past few years, the linear dynamic analysis of beams on elastic foundation has received a good
amount of attention in the literature with pioneer the work of Hetenyi (1966) who studied the
elementary Bernoulli-Euler beams on elastic Winkler foundation. Rades (1972) presented the
steady-state response of a finite rigid beam resting on a foundation defined by one inertial and
three elastic parameters in the assumption of a permanent and smooth contact between beam and
foundation considering only uncoupled modes. Wang & Stephens (1977) studied the natural
vibrations of a Timoshenko beam on a Pasternak-type foundation showing the effects of rotary
inertia, shear deformation and foundation constants of the beam employing general analytic
solutions for simple cases of boundary conditions. De Rosa (1995) and El-Mously (1999)
derived explicit formulae for the fundamental natural frequencies of finite Timoshenko-beams
mounted on finite Pasternak foundation. Explicit expressions for the natural frequencies and the
associated amplitude ratios of a double beam system and the analytical solution of its critical
buckling were also derived by Zhang et. al. (2008). Moreover, semi-analytical closed form
solutions (Coskun 2003) and numerical methods such as the method of power series expansion
of displacement components employing Hamilton's principle (Matsunaga 1999), the Galerkin
method (Coskun 2000), the differential quadrature element method (Chen 2002), a combination
between the state space and the differential quadrature methods (Chen et. al. 2004), double
Fourier transforms (Kim & Cho 2006) and the finite element technique (Yokoyama 1988,
Karamanlidis & Prakash 1988, Arboleda-Monsalve et. al. 2007) have also been used for the
vibration and buckling analysis of beam-columns on one- or two parameter linear or nonlinear
(tensionless) elastic foundations taking into account or ignoring shear deformation effect.
As the deflections become larger, the induced geometric nonlinearities result in effects that
are not observed in linear systems. Contrary to the good amount of attention in the literature
concerning the linear dynamic analysis of beam-columns supported on elastic foundation, very
little work has been done on the corresponding nonlinear problem, such as the nonlinear free
vibration analysis of multispan beams on elastic supports presented by Lewandowski (1989)
employing the dynamic finite element method, neglecting the horizontally and rotary inertia
forces and considering the beams as distributed mass systems.
In this paper, a boundary element method is developed for the nonlinear dynamic analysis of
beam-columns of arbitrary doubly symmetric simply or multiply connected constant cross
section, partially supported on tensionless Winkler foundation, undergoing moderate large
deflections under general boundary conditions, taking into account the effects of shear
deformation and rotary inertia. The beam-column is subjected to the combined action of
arbitrarily distributed or concentrated transverse loading and bending moments in both
directions as well as to axial loading. To account for shear deformations, the concept of shear
deformation coefficients is used. Five boundary value problems are formulated with respect to
the transverse displacements, to the axial displacement and to two stress functions and solved
using the Analog Equation Method, (Katsikadelis 2002), a BEM based method. Application of
the boundary element technique yields a nonlinear coupled system of equations of motion. The
solution of this system is accomplished iteratively by employing the average acceleration
method in combination with the modified Newton Raphson method, (Chang 2004, Isaacson &
Keller 1966). The evaluation of the shear deformation coefficients is accomplished from the
aforementioned stress functions using only boundary integration. The proposed model takes into
account the coupling effects of bending and shear deformations along the member as well as the
shear forces along the span induced by the applied axial loading. The essential features and
novel aspects of the present formulation compared with previous ones are summarized as
follows.
i. Shear deformation effect and rotary inertia are taken into account on the nonlinear
dynamic analysis of beam-columns subjected to arbitrary loading (distributed or
303
concentrated transverse loading and bending moments in both directions, as well as axial
loading).
ii. The homogeneous linear half-space is approximated by a tensionless Winkler foundation.
iii. The beam-column is supported by the most general nonlinear boundary conditions
including elastic support or restrain, while its cross section is an arbitrary doubly
symmetric one.
iv. The proposed model takes into account the coupling effects of bending and shear
deformations along the member as well as shear forces along the span induced by the
applied axial loading.
v. The shear deformation coefficients are evaluated using an energy approach, instead of
Timoshenkos (Timoshenko & Goodier 1984) and Cowpers (1966) definitions, for
which several authors (Schramm et. al.1994), (Schramm et. al. 1997) have pointed out
that one obtains unsatisfactory results or definitions given by other researchers, (Stephen
1980), (Hutchinson 2001), for which these factors take negative values.
vi. The effect of the materials Poisson ratio is taken into account.
vii. The proposed method employs a BEM approach (requiring boundary discretization)
resulting in line or parabolic elements instead of area elements of the FEM solutions
(requiring the whole cross section to be discretized into triangular or quadrilateral area
elements), while a small number of line elements are required to achieve high accuracy.
Numerical examples are worked out to illustrate the efficiency, wherever possible the accuracy
and the range of applications of the developed method.
2 STATEMENT OF THE PROBLEM
Let us consider a prismatic beam-column of length (Fig.1), of constant arbitrary doubly
symmetric cross-section of area . The homogeneous isotropic and linearly elastic material of
the beam-column cross-section, with modulus of elasticity
L
A
E , shear modulus G and Poissons
ratio occupies the two dimensional multiply connected region v O of the plane and is
bounded by the boundary curves, which are piecewise smooth, i.e. they may
have a finite number of corners. In Fig.1b is the principal bending coordinate system
through the cross sections centroid. The beam-column is partially supported on a tensionless
homogeneous elastic soil with
y,z
(
j
j 1,2,..., I = ) K
Cyz
x
k , and the moduli of subgrade reaction for the x, y, z
directions, respectively (Winkler spring stiffness). Taking into account the unbonded contact
between beam and subgrade, the interaction pressure at the interface is compressive and can be
represented for the horizontal and vertical directions by the following relations
y
k
z
k
( )
sx u x
p U x k u = (1a)
( )
sy v y
p U x k v = (1b)
( )
sz w z
p U x k w = (1c)
where ( )
i
U x is the unit step function defined as
( )
i
0 if i
U x i u,v,w
1 if i 0
< 0
= =
>
(2)
The beam is subjected to the combined action of the arbitrarily distributed or concentrated time
dependent axial loading
x x
p p ( x,t ) = , transverse loading ( )
y y
p p x,t = , ( )
z z
p p x,t = acting
in the and directions, respectively and bending moments y z ( )
y y
m m = x,t
z z
m m = ,
along and axes, respectively (Fig. 1a).
( ) x,t
y z

304

p
x
(x,t)
k
z
m
z
(x,t)
z,w
x,u
l
total
p
z
(x,t)
Timoshenko beam-column
l
supported

(a)
n
t
s
y,v
z,w
1
CS
()

(C: Center of gravity
S: Shear center)
(b)
Figure 1. x-z plane of a prismatic beam-column in axial - flexural loading (a) with an arbitrary doubly
symmetric cross-section occupying the two dimensional region O (b).

Under the action of the aforementioned loading, the displacement field of the beam taking
into account shear deformation effect is given as
( ) ( ) ( ) ( )
z y
u x, y, z,t u x,t y x,t z x,t u u = + (3a)
( ) ( ) v x,t v x,t = (3b)
( ) ( ) w x,t w x,t = (3c)
where u , v , w are the axial and transverse beam displacement components with respect to the
system of axes; Cyz ( ) u x,t , ( ) v x,t , ( ) w x,t are the corresponding components of the centroid
and C ( )
y
x,t u , ( )
z
x,t u are the angles of rotation due to bending of the cross-section with
respect to its centroid.
Employing the strain-displacement relations of the three - dimensional elasticity for
moderate displacements (Ramm & Hofmann 1995, Rothert & Gensichen 1987), the following
strain components can be easily obtained
2 2
xx
u 1 v w
x 2 x x
c
(
c c c | | | |
= + +
| |
c c c
\ . \ . (

( (4a)
xz
w u v v w w
x z x z x z
c c c c c c |
= + + +
c c c c c c
\ .
|
|
(4b)
xy
v u v v w w
x y x y x y
| | c c c c c c
= + + +
c c c c c c
\ .
|
(4c)
yy zz yz
0 c c = = = (4d)
305
where it has been assumed that for moderate displacements
( )
2
u u
x x
c c
<<
c c
,
( )( ) ( ) ( )
u u u u
x z x z
c c c c
<< +
c c c c
,
( )
( )
( )
( )
u u u u
x y x
c c c c
<< +
y c c c c
. Substitu-ting the
displacement components (3) to the strain-displacement relations (4), the strain components can
be written as
( )
( )
2 2
xx y z
1
x, y, z,t u z y v w
2
c u u
' '
' ' = + + + ' (5a)
xy
v
z
u ' = (5b)
xz
w
y
u ' = + (5c)
where
xy
,
xz
are the additional angles of rotation of the cross-section due to shear
deformation . Considering strains to be small, employing the second Piola Kirchhoff
stress tensor and assuming an isotropic and homogeneous material, the stress
components are defined in terms of the strain ones as
xx xx
xy xy
xz x
S E 0 0
S 0 G 0
0 0 G
S
c

(

(
=
`
(

(

) z
)
(6)
or employing eqns. (5) as
(
2 2
xx y z
1
S E u z y v w
2
u u
' '
' ' = + + +

)
(
'
(
(7a)
( )
xy
S G v
z
u ' = (7b)
( )
xz
S G w
z
u ' = + (7c)
On the basis of Hamiltons principle, the variations of the Lagrangian equation defined as
( )
2
1
t
ext
t
U K W dt 0 o =
}
(8)
and expressed as a function of the stress resultants acting on the cross section of the beam in the
deformed state provide the governing equations and the boundary conditions of the beam
subjected to nonlinear vibrations. In eqn.(8) ( ) o denotes variation of quantities, while U , ,
are the strain energy, the kinetic energy and the external load work, respectively given as
K
ext
W
( )
xx xx xy xy xz xz
V
U S S S d o oc o o = + +
}
V (9a)
( )
2 2 2
V
1
K u v w
2
o o o o = + +
}

dV (9b)
( )
ext x u x y v y z w z y y z z
L
W p u U k u u p v U k v v p w U k w w m m o o o o o o o ou o = + + + +
}
dx u
(9c)
Moreover, the stress resultants of the beam are given as
xx
N S d
O
O =
}
(10a)
y xx
S zd
O
M =
}
O
O
(10b)
z xx
S yd
O
M =
}
(10c)
y
y xy
A
Q S dO =
}
(10d)
306
z
z xz
A
Q S dO =
}
(10e)
Substituting the expressions of the stress components (7) into equations (10), the stress
resultants are obtained as
(
2 2
1
N EA u v w
2
' ' ' = + +

)
(
(
y
(11a)
y y
M EI u
'
= (11b)
z z z
M EI u
'
= (11c)
y y
Q GA
xy
= (11d)
z z
Q GA
xz
= (11e)
where A is the cross section area, , the moments of inertia with respect to the principle
bending axes given as
y
I
z
I
A d
O
O =
}
(12)
2
y
I z d
O
O =
}
(13a)
2
z
I y d
O
O =
}
(13b)
and , are its shear rigidities of the Timoshenkos beam theory, where
y
GA
z
GA
z z
z
1
A A
a
k = = A
y y
y
1
A A
a
k = = A (14a,b)
are the shear areas with respect to , axes, respectively with y z
y
k ,
z
k the shear correction
factors and , the shear deformation coefficients. Substituting the stress components given
in eqns.(7) and the strain resultants given in equations (5) to the strain energy variation
y
a
z
a
int
oE
(eqn.9a) and employing equation (8) the equilibrium equations of the beam are derived as
( )
u x x
EA u w w v v u U k u p A '' ' '' ' '' + + + + = (15a)
( )
( ) y z v y y
Nv v GA v U k v A u
'
'
' '' + + = p (15b)
( )
z z z z y z z
EI I GA v u u u
''
' + =
m (15c)
( )
( ) z y w z z
Nw w GA w U k w A u
'
'
' '' + + + = p (15d)
( )
y y y y z y y
EI I GA w u u u
''
' + + + =
m (15e)
Combining equations (15b,c) and (15d,e), the following differential equations with respect to u ,
, are derived v w
( )
u x x
EA u w w v v u U k u p A '' ' '' ' '' + + + + = (16a)

( ) ( )
( )
( )
( )
( )
2
2
y
z z
z z
2 2
y y
z z z z
y y y v y y y z
y y y y
Ea Nv
EI I v
EI v I 1 v Nv Nv A v
G GA GA
x t
EI I EI I
k v k v" k v U p p p m
GA GA GA GA
A

| |
'
' | | c
c
''' ' |
'''' ' ' + + +
|
|
c c
\ .
\ .
| |
'' '
+ + = + |
|
\ .

(16b)
307
( ) ( )
( )
( )
( )
( )
2
2
y y
z
y y
2 2
z z
y y y y
z z z w z z z y
z z z z
EI I Nw
Ea w
EI w"" I 1 w Nw Nw Aw
G GA GA
x t
EI I EI I
k w k w" k w U p p p m
GA GA GA GA
A

| |
'
' c
c | |
''' ' |
' ' + + +
|
|
c c \ .
\ .
| |
'' '
+ + = + +
|
\ .
(16c)
Eqns. (16) constitute the governing differential equations of a Timoshenko beam-column,
partially supported on a tensionless Winkler foundation, subjected to nonlinear vibrations due to
the combined action of time dependent axial and transverse loading. These equations are also
subjected to the pertinent boundary conditions of the problem, which are given as
( ) ( )
1 2
a u x,t N x,t
3
o o + = (17)
( ) ( )
1 2 y
v x,t V x,t
3
| | | + = ( ) ( )
1 z 2 z 3
x,t x,t | u | M | + = (18a,b)
( ) ( )
1 2 z
w x,t V x,t
3
+ = ( ) ( )
1 y 2 y 3
x,t x,t u M + = (19a,b)
at the beam ends x 0,l = , together with the initial conditions
( ) ( )
0
u x,0 u x = ( ) ( )
0
u x,0 u x =
(20a,b)
( ) ( )
0
v x,0 v x = ( ) ( )
0
v x,0 v x =
(21a,b)
( ) ( )
0
w x,0 w x = ( ) ( )
0
w x,0 w x =

(22a,b)
where ( )
0
u x , ( )
0
v x , ( )
0
w x , ( )
0
u x
, ( )
0
v x
and ( )
0
w x
are prescribed functions. In eqns. (18),

(19) , and
y
V
z
V
z
M ,
y
M are the reactions and bending moments with respect to , ,
respectively, which together with the angles of rotation due to bending
y z
y
u ,
z
u are given by the
following relations
( )
z
y z y v y
y
EI v
V Nv EI v p U k v' Nv I
GA x
z z
A
c (
''
'
' ''' ' = + +
u
(
(23a)
( )
y
z y z w z
z
EI
w
V Nw EI w p U k w' Nw A I
GA x
y y
u =
c (
''
'
' ''' ' +
(
(23b)
( )
z
z z y v y
y
EI
M EI v p U k v Nv Av
GA

'
'' ' = + +

(
(
(23c)
( )
y
y y z w z
z
EI
M EI w p U k w Nw Aw
GA

'
'' ' = +

(
(
(23d)
( ) ( )
y
y z w z y y y
2 2
z
z
EI
w 1
A p U k w' Nw EI w I GA w
x GA
G A
u u
c | |
''
'
z
' ''' ' = + + +
|
c
\ .
(23e)
( ) ( )
z
z y v y z z z
2 2
y
y
EI v 1
y
p U k v' Nv A EI v I GA v
x GA
G A
u
c | |
''
'
u ' ''' ' = + + +
|
c
\ .
(23f)
Finally,
k k k k k k
, , , , , o o | | ( k 1 ) are functions specified at the beam ends ,2,3 = x 0,l = .
Eqs. (17)-(19) describe the most general nonlinear boundary conditions associated with the
problem at hand and can include elastic support or restraint. It is apparent that all types of the
conventional boundary conditions (clamped, simply supported, free or guided edge) can be
derived from these equations by specifying appropriately these functions (e.g. for a clamped
308
edge it is
1 1 1
1 o | = = = ,
1
1 1
1 o | = = = ,
2 3
2 3 2 3 2 3 2
o o | | o o | = = = = = = = = =
3 2 3
0 = | = = ).
The solution of the initial boundary value problem given from eqns (16), subjected to the
boundary conditions (17)-(19) and the initial conditions (20)-(22) which represents the
nonlinear flexural dynamic analysis of a Timoshenko beam-column, partially supported on a
tensionless Winkler foundation, presumes the evaluation of the shear deformation coefficients
, , corresponding to the principal coordinate system . These coefficients are
established equating the approximate formula of the shear strain energy per unit length (Stephen
1980)
y
a
z
a Cyz
2
2
y y
z z
appr.
a Q
a Q
U
2AG 2AG
= + (24)
with the exact one given from
( ) ( )
2
2
xz xy
exact
U
2G
O
t t
dO
+
=
}
(25)
and are obtained as [28]
( ) ( )
y
2
y
1 A
a
O
d O O
k
A
( ( = =

}
V V e e O (26a)
( ) ( )
z
2
z
1 A
a
O
u u
k
A
( ( = =

}
V V d d dO (26b)
where ( ) ( )
xz xy
j
j
, t t are the transverse (direct) shear stress components,
( ) ( ) ( ) y c c + c c V
y z
i i
A
z is a symbolic vector with the unit vectors along and axes,
respectively, is given from
,
y z
i i y z
( )
y z
2 1 A v = + (27)
v is the Poisson ratio of the cross section material, and d are vectors defined as e
2 2
y
y z
I
2
v
| |
= +
|
|
\ .
y
e i
y
I yz v
z
i (28a)
2 2
z z
y z
I yz I
2
v v
| |
\ .
y
d i
|
|
z
i (28b)
and ( ) y, z O , ( ) y, z u are stress functions, which are evaluated from the solution of the
following Neumann type boundary value problems Sapountzakis, & Mokos (2005)
2
y
2I y O V = in O (29a)
n
O c
=
c
n e on
K 1
j
j 1
I I
+
=
=

(29b)
2
z
2I z u V = in O (30a)
n
u c
=
c
n d on
K 1
j
j 1
I I
+
=
=

(30b)
309
where is the outward normal vector to the boundary n I . In the case of negligible shear
deformations . It is also worth here noting that the boundary conditions (29b), (30b)
have been derived from the physical consideration that the traction vector in the direction of the
normal vector vanishes on the free surface of the beam.
z y
a a 0 = =
n
3 INTEGRAL REPRESENTATIONS NUMERICAL SOLUTION
According to the precedent analysis, the nonlinear flexural dynamic analysis of Timoshenko
beam-columns, partially supported on a tensionless Winkler foundation, undergoing moderate
large deflections reduces in establishing the displacement components ( ) u x,t and ( ) v x,t ,
( ) w x,t having continuous derivatives up to the second order and up to the fourth order with
respect to x , respectively, and also having derivatives up to the second order with respect to t
(ignoring the inertia terms of the fourth order (Thomson 1981). Moreover, these displacement
components must satisfy the coupled governing differential equations (16) inside the beam, the
boundary conditions (17)-(19) at the beam ends x 0,l = and the initial conditions (20)-(22).
Eqns (16) are solved using the Analog Equation Method (Katsikadelis 2002) as it is developed
for hyperbolic differential equations in Sapountzakis & Katsikadelis (2000).
3.1 For the transverse displacements v, w.
Let ( ) v x,t , ( ) w x,t
(
be the sought solution of the aforementioned boundary value problem.
Setting as ) ( )
2
u x,t = v x,t , ( ) ( )
3
u x,t w x,t = and differentiating these functions four times
with respect to x yields
( )
4
i
i
4
u
q x,t
x
c
=
c
( ) i 2,3 = (31)
Eqns. (31) are quasi-static, that is the time variable appears as a parameter. They indicate that
the solution of eqns. (16b), (16c) can be established by solving eqns. (31) under the same
boundary conditions (18)-(19), provided that the fictitious load distributions ( )
i
q x,t
are first established. These distributions can be determined using BEM as follows.
( ) i 2,3 =
Following the procedure presented in (Thomson 1981) and employing the constant element
assumption for the load distributions along the L internal beam elements (as the numerical
implementation becomes very simple and the obtained results are of high accuracy), the integral
representations of the displacement components
i
q
i
u ( ) i 2,3 = and their first derivatives with
respect to x when applied for the beam ends ( ), together with the boundary conditions (18)-
(19) are employed to express the unknown boundary quantities
0,l
( )
i
u ,t , , ( )
i x
u , ,t , , ( )
i xx
u , ,t ,
and ( )
i xx
u , ,t
x
, ( 0,l , = ) in terms of as
i
q
11 13 14 2 xxx 3
21 22 23 2 xx 3
2
31 32 33 34 3 2 x
41 42 43 4 2
0 0 ,
0 0 ,
, 0
0 0
(
(

(
= +
` ` `
(

(

) )
D D D u
D D D u
q
E E E E F u
E E E F u

)
(32a)

310
11 13 14 3 xxx 3
21 22 23 3 xx 3
3
31 32 33 34 3 x 3
41 42 43 3 4
0 ,
0 , 0
, 0
0 0
(
(

(
= +
` ` `
(

(

) )
G G G u
G G G u
q
E E E E u F
E E E u F
0

)
(32b)
where , , , , , , , , , , , are known
square matrices including the values of the functions
11
D
13
D
14
D
21
D
22
D
23
D
11
G
13
G
14
G
21
G
22
G
23
G 2 2
j j j
, , ,
j
| | ( j 1,2 = ) of eqns (18)-(19);
,
3
3
, ,
3 3
are known column matrices including the boundary values of the
functions
2 1
3 3
, ,
3
,
3
| | of eqns (18)-(19); , (
jk
E j 3,4 = , k 1,2,3,4 = ) are square 2 known
coefficient matrices and
2
j
F ( ) j 3 = ,4 are 2 L rectangular known matrices originating from
the integration of kernels on the axis of the beam. Moreover,
( ) ( ) {
T
i i i
u 0,t u l ,t = u
}
(33a)
( ) ( )
T
i i
i x
u 0,t u l ,t
,
x x
c c
=

c c

)
u

`
(33b)
( ) ( )
T
2 2
i i
i xx
2 2
u 0,t u l ,t
,
x x

c c

=

c c

)
u
`
(33c)
( ) ( )
T
3 3
i i
i xxx
3 3
u 0,t u l ,t
,
x x

c c

=

c c

)
u
`
(33d)
are vectors including the two unknown boundary values of the respective boundary quantities
and
{ }
T
i i i
i 1 2 L
q q ... q = q ( ) i 2,3 = is the vector including the unknown nodal values of the
fictitious load.
L
Discretization of the integral representations of the displacement components
i
u ( ) i 2,3 =
and their derivatives with respect to x , after elimination of the boundary quantities employing
eqns. (32), gives
i i i
= + u Tq t
i
i = 2,3 (34a)
i ,x ix i ix
= + u T q t i = 2,3 (34b)
i ,xx ixx i ixx
= + u T q t i = 2,3 (34c)
i ,xxx ixxx i ixxx
= + u T q t i = 2,3 (34d)
i xxxx i
, = u q i = 2,3 (34e)
where , , , , are vectors including the values of and their
derivatives at the nodal points, , , , are known L
i
u
i x
, u
i xx
, u
L
i xxx
, u
i xxxx
, u
i
T
( )
i
u x,t
ix
T
ixx
T
ixxx
T L matrices and , ,
, are known matrices.
i
t
ix
t
ixx
t
ixxx
t L 1
In the conventional BEM, the load vectors are known and eqns (34) are used to evaluate
and their derivatives at the nodal points. This, however, can not be done here since
are unknown. For this purpose, additional equations are derived, which permit the
establishment of . These equations result by applying eqns (16b), (16c) to the collocation
points, which after ignoring the inertia terms of the fourth order arising from coupling of shear
deformations and rotary inertia Thomson (1981), lead to the formulation of the following set of
simultaneous equations
i
q
( )
i
u x,t
i
q
2L
L
2L
i
q L
311
2 2 2 2 2 2 2
+ + = M q S q K q f (35a)
3 3 3 3 3 3 3
+ + = M q S q K q f (35b)
where the , , , , ,
2
M
3
M
2
S
3
S
2
K
3
K L L matrices and the ,
2
f
3
f L 1 vectors are given
as
(
y
z
2 2 z 2xx x 2x 2xx y 2
y
Ea
I
A 1
G GA
I
| |
= + +
|
\ .
M T T N T NT K T
)
(36a)
(
z
2 xt 2x t
y
2 I
GA
)
2xx
= + S N T N T (36b)
( )
( )
z
2 z x 2x 2xx xtt 2x tt 2xx y 2
y
z
xxx 2x xx 2xx x 2xxx y 2xx
y
I
GA
EI
3 3
GA
= + +
+ + + +
K EI N T NT N T N T K T
N T N T N T N K T
+
(36c)
( ) ( )
z z
2 y y 2 y,xx y 2xx y,tt z,x x 2x 2xx
y y
z z
xxx 2x xx 2xx x 2xxx xtt 2x tt 2xx
y y
EI I
GA GA
EI I
3 3
GA GA
( = + + +

+ + + +
f p K t p K t p m N t Nt
N t N t N t N t N t
(36d)
(
y
z
3 3 y 3xx x 3x 3xx z 3
z
I
Ea
A 1
G GA
I
| |
= + +
|
\ .
M T T N T NT K T
)
(36e)
(
y
3 xt 3x t
z
2 I
GA
= + S N T N )
3xx
T (36f)
( )
( )
y
3 y x 3x 3xx xtt 3x tt 3xx z 3
z
y
xxx 3x xx 3xx x 3xxx z 3xx
z
I
GA
EI
3 3
GA
= + +
+ + + +
K EI N T NT N T N T K T
N T N T N T N K T
+
(36g)
( ) ( )
y y
3 z z 3 z,xx z 3xx z,tt y,x x 3x 3xx
z z
y y
xxx 3x xx 3xx x 3xxx xtt 3x tt 3xx
z z
EI I
GA GA
EI I
3 3
GA GA
( = + + + +

+ + + +
f p K t p K t p m N t Nt
N t N t N t N t N t
(36h)
where , N
km
N ( k,m x,t = ) are diagonal matrices containing the values of the axial force
and its derivatives with respect to k and m parameters at the nodal points, is the unit matrix,
, are diagonal matrices including the values of the corresponding quantities at
the aforementioned points, while , , , , , , and are
L L
y
p
L
z,xx
I
y
EI
z
EI L L
y,xx
p
y,tt
p
z
p p
z,tt
p
y,x
m
z,x
m L 1
vectors containing the values of the external loading and its derivatives at these points. Finally
y
K ,
z
K are diagonal matrices whose diagonal elements are given as
( ) ( ) ( )
y y
i
i ,i i
k U = K
v
(37a)
( ) ( ) ( )
z z
i i
i ,i
k U = K
w
(37b)
where
( )
y
i
k , ( )
z
i
k , ( )
v
i
U , are the values of the corresponding moduli of subgrade or
the unit step function at the i-th nodal point.
( )
w
i
U
312

3.2 For the axial displacement u.
Let ( )
1
u u x,t = be the sought solution of the boundary value problem described by eqns (16a)
and (17). Differentiating this function two times yields
(
2
1
x
2
u
q x,t )
x
c
=
c
(38)
Eqn (38) indicates that the solution of the original problem can be obtained as the axial
displacement of a beam with unit axial rigidity subjected to an axial fictitious load ( )
x
q x,t
under the same boundary conditions. The fictitious load is unknown. Following the same
procedure as in 3.1, the discretized counterpart of the integral representations of the
displacement component and its first derivative with respect to
1
u x when applied to all nodal
points in the interior of the beam yields
1 1 1
= + u T q t
1
(39a)
1,x 1x 1 1x
= + u T q t (39b)
where , are known matrices, similar with those mentioned before for the
displacements . Application of eqn. (16a) to the collocation points, after employing
eqns. (34), (39) leads to the formulation of the following system of equations with respect to
, and fictitious load vectors
1
T
1x
T
3
q
L L
2 3
u , u L
L
1
q
2
q
( ) ( ) ( )
( ) ( )
x 1 1 2xx 2 2xx 2x 2 2x
dg.
3xx 3 3xx 3x 3 3x x 1
dg.
( = + +

( + + +

1 1 x
EA K T q AT q p EA T q t T q t
EA T q t T q t K t

(40)
where , are diagonal matrices including the values of the corresponding
quantities at the L nodal points and
EA A L L
x
K is a diagonal matrix similar with those of equations (37)
whose diagonal elements are given as
( ) ( ) ( )
x x u
i
i ,i
k U = K
i
(41)
Moreover, substituting eqns. (34), (40) in eqn. (11a) the discretized counterpart of the axial
force at the neutral axis of the beam is given as
( ) ( ) ( )
( ) ( )
1x 1 1x 2x 2 2x 2x 2 2x
dg
3x 3 3x 3x 3 3x
dg
1
2
1
2
( = + + + +

( + + +

N EA T q t EA T q t T q t
EA T q t T q t
+
(42)
Eqns. (35a), (35b), (40) and (42) constitute a nonlinear coupled system of equations with respect
to , , and quantities. The solution of this system is accomplished iteratively by
employing the average acceleration method in combination with the modified Newton Raphson
method (Chang 2004, Isaacson & Keller 1966).
1
q
2
q
3
q N
3.3 For the stress functions ( ) y, z O and ( ) y, z u
The evaluation of the stress functions ( ) y, z O and ( ) y, z u is accomplished using BEM as this
is presented in Sapountzakis and Mokos (2005).
313
Moreover, since the nonlinear the nonlinear flexural dynamic problem of Timoshenko beam-
columns is solved by the BEM, the domain integrals for the evaluation of the area, the bending
moments of inertia (eqns.12) and the shear deformation coefficients (eqns.26) have to be
converted to boundary line integrals, in order to maintain the pure boundary character of the
method. This can be achieved using integration by parts, the Gauss theorem and the Green
identity. Thus, the moments, the product of inertia and the cross section area can be written as
( )
2
y y
I yz n
I
=
}
ds
ds
(43a)
( )
2
z z
I zy n
I
=
}
(43b)
( )
y z
1
A yn zn
2
I
= +
}
ds (43c)
while the shear deformation coefficients and are obtained from the relations
y
a
z
a
( )
2 2
y y y yy ed
2
A 1
a 4v 2 I I v I I I
4
O
A
|
= + +
\ .
e O
|
|
(44a)
( )
2 2
z z z z ed
2
A 1
a 4v 2 I I v I I I
4
u
A
|
= + +
\ .
d u
|
|
(44b)
where
( )
e
I
O
I
O =
}
n e ds (45a)
( )
d
I
u
I
u =
}
n d ds (45b)
4 4 2 3
ed z y z
2
I y zn z yn y z n
3
I
|
= + +
\ .
}
ds
|
|
(45c)
( ) ( )
4
y yy z y
1
I 2I y zn 3 n y y
6
O
I
O
2
ds
(
= +

}
n e (45d)
( ) ( )
4
z zz y z
1
I 2I z yn 3 n z z
6
u
I
u
2
ds
(
= +

}
n d (45g)
4 NUMERICAL EXAMPLES
On the basis of the analytical and numerical procedures presented in the previous sections, a
computer program has been written and representative examples have been studied to
demonstrate the efficiency, wherever possible the accuracy and the range of applications of the
developed method. In all the examples treated the results have been obtained using
nodal points along the beam-column and a time step of
L 41 =
t 1.0 sec A = .
4.1 Example 1
For comparison reasons, the linear dynamic analysis of a simply supported uniform beam-
column of length l 6 ( .096m = E 24.82GPa = , , v 0
3
3387 g / m K = .3 = , )
resting on a homogeneous (either bilateral or unilateral) elastic foundation with modulus of
subgrade reaction , as this is shown in Fig.2 is examined. The free
vibrations case of this example was analyzed by Timoshenko et al. (1974). Lai et al. (1992),
Thambiratnam and Zhuge (1996) and Friswell et al. (2007) while the forced vibrations one by
Calim (2009). The beam-column is subjected to a triangular impulsive load of amplitude
5 4
I 143.9 10 m
=
2
5 MN / m
z
k 16.5 =
l / 2
P 100 kN = at its midpoint, as this is shown in Fig.2.
314
In Table1, the evaluated first five natural frequencies of the beam-column resting on the
bilateral elastic foundation are presented as compared with those obtained from the literature. In
Figs.3-5 the time history of the transverse displacement ( ) w l / 2 at the beam-columns
midpoint, of the bending moment ( )
y
M l / 2 at the same point and of the shear force ( )
z
Q l at
the right supported end, respectively are presented either for the bilateral or the unilateral elastic
foundation model and compared with those obtained from a complementary functions method
and a FEM solution demonstrating the accuracy of the results of the proposed method.
Moreover, in Table2 the extreme values of the displacement ( ) w l / 2 and of the soil reaction
sz
p ( l / 2 ) at the beam-columns midpoint are also presented for both cases of bilateral and
unilateral soil reaction.

P(t)
Beam-Column
k
z

x
z
(a)
Time (sec)
0.0125 0.025
100
P(t) kN

(b)
Figure 2. Prismatic beam-column on elastic foundation (a) subjected to a triangular impulsive load (b).

Table1. First five natural frequencies (Hz) of the simply supported beam of example 1.

Modes
Timosh.et
al.(1974)
Lai et
al.(1992)
Thambiratnam &
Zhuge(1996)
Friswell et
al.(2007)
Calim
(2009)
ANSYS
Present
Study
1 32.9032 32.9049 32.9033 32.8980 32.8633 32.8624 32.7946
2 56.8135 56.8220 56.8193 56.8080 56.5972 56.5891 56.5476
3 112.908 111.973 111.961 111.900 110.759 110.739 110.722
4 - - - 193.760 189.939 189.901 189.489
5 - - - - 222.078 222.043 222.077
Table2. Extreme values of the displacement ( ) w l / 2 and the foundation reaction

of the
beam-column of example 1.
sz
p ( l / 2 )

Winkler Winkler Tensionless
( ) w l/2 (m)
sz
p (l/2) (kN) ( ) w l/2 (m)
sz
p (l/2) (kN)
max 2.6310
-3
43.6 2.6310
-3
43.6
min -2.5010
-3
-41.5 -6.8010
-3
0.00

4.2 Example 2
In order to illustrate the importance of the nonlinear analysis and the influence of the shear
deformation effect in flexural vibrations, a clamped beam-column of length , having a
hollow rectangular cross section (
l 4.90m =
E 210 GPa = , v 0.3 = , ) resting on a
homogeneous (either bilateral or unilateral) elastic foundation of stiffness , as this is shown
in Fig.6 is examined.
/
3
m
z
k
7.85 tn =
315

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Time (sec)
-0.0012
-0.001
-0.0008
-0.0006
-0.0004
-0.0002
0
0.0002
0.0004
0.0006
0.0008
0.001
0.0012
N
o
n
d
i
m
e
n
s
i
o
n
a
l

D
i
s
p
l
a
c
e
m
e
n
t

w
(
l
/
2
)
Present study, Winkler
Present study, Winkler Tensionless
Calim (2009)
ANSYS
Figure 3. Time history of the transverse displacement ( ) w l / 2 at the midpoint of the beam-column of
example 1.

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.
Ti
1
me (sec)
-0.012
-0.01
-0.008
-0.006
-0.004
-0.002
0
0.002
0.004
0.006
0.008
0.01
0.012
N
o
n
d
i
m
e
n
s
i
o
n
a
l

M
o
m
e
n
t

M
y
(
l
/
2
)
Calim (2009)
ANSYS

Figure 4. Time history of the bending moment ( )
y
M l / 2 at the midpoint of the beam-column of
example 1.

316
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.
Ti
1
me (sec)
-0.04
-0.03
-0.02
-0.01
0
0.01
0.02
0.03
0.04
N
o
n
d
i
m
e
n
s
i
o
n
a
l

S
h
e
a
r

F
o
r
c
e

Q
z
(
l
)
Calim (2009)
ANSYS

Figure 5. Time history of the shear force ( )
z
Q l at the right supported end of the beam-column of
example 1.

k
z

h
=
1
4
c
m

t=4mm
y
z
x
p
z
=350kN/m
l=4.9m
a
z
= 3.664
a
y
= 1.766
b=23cm
y
z
My = 200kNm
l/2=2.45m

Figure 6. Clamped beam of hollow rectangular cross section subjected to the suddenly applied
concentrated bending moment
y
M

and uniformly distributed load .
z
p

In Figs.7, 8 the deflection along the beam-column subjected to a suddenly applied
consecrated bending moment
w
=
y
M 200kNm at its midpoint is presented at the time instant

for various values of the stiffness for the cases of bilateral and unilateral
soil reaction, respectively. The influence of both the foundation stiffness parameter and the
unilateral character of the soil reaction are easily verified. Moreover, in Fig.9 the time history of
the central transverse deflection
2
t 1.6 10 sec
=
z
k
z
k
( ) w l / 2 and in Table 3 the maximum central deflection
( )
max
w m and the period ( )
z
T sec

of the first cycle of motion of the beam-column additionally

317
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Beam-Column Length (m)
5
-0.008
-0.006
-0.004
-0.002
0
0.002
0.004
0.006
0.008
D
i
s
p
l
a
c
e
m
e
n
t

w

(
m
)
kz =6.45 kPa
kz =64.5 kPa
kz =645 kPa
kz =64500 kPa

Figure 7. Deflection along the beam-column of example 2, for various stiffness

values of the
bilateral Winkler springs.
w
z
k

kz =6.45 kPa
kz =64.5 kPa
kz =645 kPa
kz =64500 kPa
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Beam-Column Length (m)
5
-0.01
-0.005
0
0.005
0.01
0.015
0.02
0.025
D
i
s
p
l
a
c
e
m
e
n
t

w

(
m
)

Figure 8. Deflection along the beam-column of example 2, for various stiffness

values of the
unilateral Winkler springs.
w
z
k

318
Table 3. Maximum central deflection ( )
max
w m and period ( )
z
T sec

of the first cycle of motion of the
clamped beam of example 2.
Without Shear Deformation With Shear Deformation
Linear Analysis
Nonlinear
Analysis
Linear Analysis Nonlinear Analysis
max
w
0.3729 0.2572 0.3914 0.2688
z
T

0.01890 0.01482 0.01973 0.01607

subjected to a uniformly distributed load /
z
p 350 kN m = (Fig.6) is presented for a unilateral
subgrade model with
z
, performing either a linear or a nonlinear analysis and
taking into account or ignoring both shear deformation effect and rotary inertia. From the
obtained results, the discrepancy between the linear and the nonlinear analysis is not negligible
and should not be ignored, while the significant influence of the shear deformation effect
increasing both central transverse displacement and the obtained period of the first cycle of
motion is remarked in both linear and nonlinear analysis.
k 645.0 kPa =
4.3 Example 3
To demonstrate the range of applications of the proposed method, a partially embedded pile in a
homogeneous elastic Winker foundation with spring stiffness , of total
length ( ,
/
2
z y
k k 85.0 N m = = M
l 15.0m =
free
l 6.20 = m m
embed
l 8.80 = ), of circular cross section of diameter
( D 1. = 0m E 29 = GPa A 0.7 = , ,
2
5m v 0.2 8 = , ,
4
49m
y z
a a 1,172
y z
I I 0.0 = = = = ), as this is
shown in Fig.10 has been studied. According to its boundary conditions, the pile end at the
elastic foundation is clamped, while the other end is free according to its displacements and
blocked according to its rotations. The pile is subjected to a suddenly applied concentrated axial
load ,
x
P 0,t ) 1 ( = .0MN ( ) t 0 > .0 and to a uniformly distributed transverse load
( ) /
y
p t 500kN = m , ( ) t 0.0 > acting to the free part of the length of the pile.
In Figs.11, 12 the time history of the head displacement of the pile and the deflection
along the pile at the time instant are presented, respectively performing either a
linear or a nonlinear analysis, taking into account or ignoring both rotary inertia and shear
deformation effect. Moreover, in Table 4 the maximum value of the head displacement
top
v v
2
t 7.0 10 sec
=
( )
top
max
v and the period of the first-cycle of motion are presented for the aforementioned
cases. Moreover, the examined pile additionally to the aforementioned loading is also subjected
to a uniformly distributed load
y
T
/
z
p 100 kN m = at its free length and to a suddenly applied
concentrated load ( )
z
P t 50 6 = kN ( , ) 0.0 t 0.15 s s acting at its top, as this is shown in Fig.10.
In Fig.13 the time history of the head displacement of the pile performing either a linear or
a nonlinear analysis is presented taking into account or ignoring shear deformation effect.
Finally, in Table 5 the maximum values of the head displacements
top
w
( )
top
max
v ,
( )
top
max
w and the
periods

of the first cycle are presented for the same cases of analysis. At this point
it is worth noting that the small discrepancy of the head deflections
y z
T , ( T s )
( )
top
v
y,
max
z
between the
Tables 4, 5 is due to the coupling effect of the transverse displacements in directions in the
nonlinear analysis.

319
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04
Time (sec)
-0.4
-0.35
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
D
i
s
p
l
a
c
e
m
e
n
t

w
(
l
/
2
)

m
Nonlinear Analysis with Shear Deformation
Nonlinear Analysis without Shear Deformation
Linear Analysis with Shear Deformation
Linear Analysis without Shear Deformation
e

Figure 9. Time history of the central deflection ( ) w l / 2
45.0 kPa
of the beam-column of example 2, for a
unilateral subgrade model with
z
k 6 = .

p
y
=
5
0
0
k
N
/
m

p
z
=
1
0
0
k
N
/
m

y
P
x
=1.0 MN
P
z
=650kN
l = 6.20 m
l = 8.80 m
(a)
a
y
=a
z
=1.172
P
x

p
y

p
z

650
P
z
(kN)
0.0
Time (sec) 0.15
Time (sec) 0.0
1.0 MN
100 kN/m
500 kN/m
A =0.785m
2

z
D=1.0m
(b)
Figure 10. Pile of circular cross section in axial - flexural loading subjected to rectangular impulsive
concentrated load , concentrated axial load
z
P
x
P and to uniformly distributed loading
, .
y
p
z
p

320
0 0.05 0.1 0.15 0.2 0.25 0.3 0.
Time (sec)
35
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0.22
0.24
D
i
s
p
l
a
c
e
m
e
n
t

v

a
t

p
i
l
e
-
h
e
a
d

(
m
)
e

Figure 11. Time history of the transverse displacement of the head of the pile of example 3.
top
v

Table 4. Maximum head deflection and period
( )
top
max
v ( )
y
T sec

of the first cycle of motion of the pile
of example 3.
Linear Analysis
Nonlinear
Analysis
Linear Analysis
Nonlinear
Analysis
( )
top
max
v
0.2699 0.2321 0.2699 0.2353
y
T

0.1105 0.1180 0.1111 0.1201

Table 5. Maximum head displacements
( )
top
max
v ,
( )
top
max
w and periods ( )
y z
T , sec T

of the first cycle of
the examined pile of example 3.
Linear Analysis
Nonlinear
Analysis
Linear Analysis
Nonlinear
Analysis
( )
top
max
v
0.2054 0.2320 0.2699 0.2351
( )
top
max
w

0.0992 0.1109 0.1002 0.1111
y
T

0.1105 0.1172 0.1111 0.1192
z
T

0.1133 0.1189 0.1143 0.1215
321

0 2 4 6 8 10 12 14
Depth (m)
16
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0.22
D
i
s
p
l
a
c
e
m
e
n
t

(
m
)

Figure 12. Displacement v at time instant along the pile of example 3.
-2
t = 7.010

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Time (sec)
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0.1
0.12
D
i
s
p
l
a
c
e
n
e
n
t

w

a
t

p
i
l
e
-

h
e
a
d

(
m
)
e

top
w
322

4.4 Example 4
In this example, a fully embedded in stiff cohesive soil with non constant stiffness free head pile
of length l 8 of a hollow circular cross section ( .0 m = E 210 GPa = , , ),
as this is shown in Fig.14 is examined. The pile is subjected to a concentrated axial
/
3
7.85 tn m = v 0.3 =
x
P t ) 500 kN ( = ( , ) t 0.0 > and transverse ( ) ( )
z
P t 750 cos t kN e = loading acting at the free
pile head, where / 9rad sec 614.32 e = is the first natural frequency of the pile-soil system.

l = 8.0 m
x
0.8 m
K
z,y
(MPa)
100 0 200
(a)
y
z
P
z
(KN)
Time (sec)
0
0.0
500
P
x
(kN)
Time (sec) 0.0
a
z
=2.226
a
y
=2.226
D = 80cm
15 mm
(b)
Figure 14. Hollow circular pile in axial-flexural loading embedded in non constant stiffness soil,
subjected to concentrated axial
x
P and transverse loading at pile-head.
z
P

In Figs.15, 16 the time history of the pile head displacement
top
w the deflection w along
the pile at the time instant t 4 1 = re presented, respectively, performing either a linear
or a nonlinear analysis and taking into account or ignoring both rotary inertia and shear
deformation effect. The discrepancy between linear and nonlinear analysis in the resonance case
is remarkable and is justified from the varying with time stiffness of the pile-soil system in the
case of a nonlinear analysis. Moreover, to demonstrate this discrepancy, in Fig.17 the time
history of the head displacement of the pile is presented for the time interval 0.0 - 0.05sec.
and
sec a
2
0

top
w
4.5 Example 5
As the final application a clamped-pinned HEA320 beam-column of length l 6
(
.5 m =
E 210 GPa = , , ) resting on a constant stiffness soil of , as
this is shown in Fig.18, is considered. The beam-column is subjected to a uniformly distributed
axial loading
v 0.3 =
x
/
3
7.85 tn m =
z
k 1 = .2MPa
p t ) 500 kN ( = ( , ) t 0 > .0 and to a transverse concentrated moving load
( ) ( )
z
P t 10 si = n t M e N with constant velocity of / v 65m s = , where / rad sec 100 e = .

323

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
Time (sec)
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
D
i
s
p
l
a
c
e
m
e
n
t
a
t
p
i
l
e
h
e
a
d
(
m
)
Linear Analysis with Shear Deformation /10 (m)
e

Figure 15. Time history of the transverse displacement of the head of the pile of example 4 (for
graphic purposes displacements coming from linear analysis are divided by 10).
top
w

0 1 2 3 4 5 6 7
Depth (m)
8
-0.05
-0.045
-0.04
-0.035
-0.03
-0.025
-0.02
-0.015
-0.01
-0.005
0
0.005
D
i
s
p
l
a
c
e
m
e
n
t

(
m
)
Liinear Analysis without Shear Deformation
e

Figure 16. Displacement w at time instant sec along the pile of example 4.
-2
t = 4.010
324

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045 0.05
Time (sec)
-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0.1
D
i
s
p
l
a
c
e
m
e
n
t

w

a
t

p
i
l
e
-
h
e
a
d

(
m
)
e

top
w

P
z
=10.0sin(t) MN
500kN/m
65m/sec
k
z
= 1.2MPa
y
z
x
l=6.5m
a
z
= 4.5118
a
y
= 1.4752
HEA320

Figure 18. Beam-column in axial-flexural loading on constant stiffness soil subjected to uniformly
distributed axial load
x
p and to concentrated transverse moving load .
z
P

In Fig.19 the time history of the central transverse deflection ( ) w l / 2 of the beam-column,
performing either a linear or a nonlinear analysis, taking into account or ignoring both rotary
inertia and shear deformation effect is presented. In Fig.20 the displacement at the time
instant along the beam-column either for conventional Winkler or for tensionless
Winkler soil are presented performing a nonlinear analysis and taking into account or ignoring
shear deformation effect. Finally, in Table 6 the deflection extreme values ,

and in
Fig.21 the effect of the frequency
w
sec t 0.05 =
max
w
min
w
e of the concentrated moving load to the maximum
deflection are presented for all of the aforementioned cases.
max
w

325
Table 6. Extreme values of the displacement ( )
2
w 10 m
of the beam-column of example 5.

Winkler
Linear Analysis
Nonlinear
Analysis
max
w
6.21 8.43 6.45 8.58
min
w

-7.87 -9.15 -8.17 -9.26
Tensionless Winkler
Linear Analysis
Nonlinear
Analysis
max
w
5.92 6.84 6.67 6.98
min
w

-10.18 -10.96 -11.46 12.34

0 0.02 0.04 0.06 0.08 0.1 0.12
Time (sec)
-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0.1
D
i
s
p
l
a
c
e
m
e
n
t

w
(
l
/
2
)

(
m
)
e

Figure 19. Time history of the central deflection ( ) w l / 2 of the beam-column of example 5.
A boundary element method is developed for the nonlinear dynamic analysis of beam-columns
of arbitrary doubly symmetric simply or multiply connected constant cross section, partially
supported on tensionless Winkler foundation, undergoing moderate large deflections under
general boundary conditions, taking into account the effects of shear deformation and rotary
inertia. The beam-column is subjected to the combined action of arbitrarily distributed or
concentrated transverse loading and bending moments in both directions as well as to axial
loading. The proposed model takes into account the coupling effects of bending and shear
deformations along the member as well as the shear forces along the span induced by the
applied axial loading. The main conclusions that can be drawn from this investigation are
326

0 1 2 3 4 5 6
Beam-Column Lenght (m)
7
-0.12
-0.1
-0.08
-0.06
-0.04
-0.02
0
D
i
s
p
l
a
c
e
m
e
n
t

w

(
m
)
Bernouli Beam-Column on Winkler Springs
Timoshenko Beam-Column on Winkler Springs
Bernouli Beam-Column on Tensionless Winkler Springs
Timohenko Beam-Column on Tensionless Winkler Springs
e

Figure 20. Displacements w at time instant t = along the beam - column of example 5. 0.05sec

0 50 100 150 200 250 300 350 400 450 500
Frequency (rad/sec)
0
0.1
0.2
0.3
0.4
0.5
0.6
w
m
a
x
e

Figure 21. Frequency effect to the maximum deflection of the beam - column of example 5.
327

a. The numerical technique presented in this investigation is well suited for computer aided
analysis for beams of arbitrary simply or multiply connected doubly symmetric cross section.
b. In some cases, the effect of shear deformation is significant, especially for low beam
slenderness values, increasing both the maximum transverse displacements and the
calculated periods of the first cycle of motion.
c. The discrepancy between the results of the linear and the nonlinear analysis is remarkable.
d. The effect of both the tensionless character and the stiffness of the soil is important.
e. The discrepancy in the response of a beam-column in the resonance case, performing a linear
or a nonlinear analysis is remarkable.
f. The coupling effect of the transverse displacements in both directions in the nonlinear
analysis influences these displacements.
6 ACKNOWLEDGEMENT
This work forms part of an EU 7th Framework research project funded through the European
Research Councils (ERC) Programme Ideas, Support for Frontier Research Advanced Grant,
under Contract number ERC-2008-AdG 228254-DARE.
7 REFERENCES
ANSYS Swanson Analysis System, Inc., 201 Johnson Road, Houston, PA 15342-1300, USA.
Arboleda-Monsalve, L. G., Zapata-Medina, D. G. & Aristizabal-Ochoa, J. D. 2007. Timoshenko Beam-
Column with Generalized End Conditions on Elastic Foundation: Dynamic-Stiffness Matrix and Load
Vector. Journal of Sound and Vibration 310: 10571079.
Calim, F. F. 2009. Dynamic analysis of beams on viscoelastic foundation. European Journal of
Mechanics A/Solids 28: 469-476.
Chang, S. Y. 2004. Studies of Newmark Method for Solving Nonlinear Systems: (I) Basic Analysis.
Journal of the Chinese Institute of Engineer, 27(5): 651-662.
Chen, C. N. 2002. DQEM Vibration Analyses of Non-Prismatic Shear Deformable Beams Resting on
Elastic Foundations. Journal of Sound and Vibration 255(5): 989-999.
Chen, W. Q., Lu, C. F. & Bian, Z. G. 2004. A Mixed Method for Bending and Free Vibration of Beams
Resting on a Pasternak Elastic Foundation. Applied Mathematical Modeling 28: 877890.
Coskun, I. 2000. Non-Linear Vibrations of a Beam Resting on a tensionless Winkler Foundation. Journal
of Sound and Vibration 236(3): 401-411.
Coskun, I. 2003. The Response of a Finite Beam on a Tensionless Pasternak Foundation Subjected to a
Harmonic Load. European Journal of Mechanics A/Solids 212: 151161.
Cowper, G.R.1966. The Shear Coefficient in Timoshenkos Beam Theory. Journal of Applied Mechanics
ASME 33(2): 335-340.
De Rosa, M.A. 1995. Free Vibrations of Timoshenko Beams on Two-Parameter Elastic Foundation.
Computers & Structure, 57(1): 151-156.
El-Mously M. 1999. Fundamental Frequencies of Timoshenko Beams Mounted on Pasternak Foundation.
Journal of Sound and Vibration 228(2): 452-457.
Friswell, M.I., Adhikari, S. & Lei, Y. 2007. Vibration analysis of beams with non-local foundation using
the finite element method. International Journal for Numerical Methods in Engineering 71(11):
13651386.
Hetenyi, M. 1966. Beams and Plates on Elastic Foundations and Related Problems. Applied Mechanics
Reviews 19: 95-102.
Hutchinson, J.R. 2001. Shear Coefficients for Timoshenko Beam Theory. ASME Journal of Applied
Mechanics 68: 8792.
Isaacson, E. & Keller, H.B. 1966. Analysis of Numerical Methods. New York: John Wiley and Sons.
Karamanlidis, D. & Prakash, V. 1988. Buckling and Vibration Analysis of Flexible Beams Resting on an
Elastic Half-Space. Earthquake Engineering and Structural Dynamics, 16: 1103-1114.
328
Katsikadelis, J.T. 2002. The Analog Equation Method. A Boundary-Only Integral Equation Method for
Nonlinear Static and Dynamic Problems in General Bodies. Theoretical and Applied Mechanics 27:
13-38.
Kim, S.-M. & Cho, Y.-H. 2006. Vibration and Dynamic Buckling of Shear Beam-Columns on Elastic
Foundation Under Moving Harmonic Loads. International Journal of Solids and Structures 43: 393
412.
Lai, Y.C., Ting, B.Y., Lee, W.S.& Becker, W.R. 1992. Dynamic response of beams on elastic
foundation. International Journal for Structural Engineering 118: 853-858.
Lewandowski, R. 1989. Nonlinear Free Vibrations of Multispan Beams on Elastic Supports. Computers
& Structures 32(2): 305-312.
Matsunaga, H. 1999. Vibration and Buckling of Deep Beam-Columns on Two-Parameters Elastic
Foundation. Journal of Sound and Vibration 228(2): 359-376.
Rades, M. 1972. Dynamic Analysis of an Inertial Foundation Model. Int. J. Solids Structures 8: 1353-
1372.
Ramm E., Hofmann T.J. 1995. Stabtragwerke, Berlin: Der Ingenieurbau, Ed.G. Mehlhorn, Band
Baustatik/Baudynamik, Ernst & Sohn.
Rothert H., Gensichen V. 1987. Nichtlineare Stabstatik. Berlin: SpringerVerlag.
Sapountzakis E.J. & Katsikadelis J.T. 2000. Analysis of Plates Reinforced with Beams. Computational
Mechanics 26: 66-74.
Sapountzakis, E.J. & Mokos, V.G. 2005. A BEM Solution to Transverse Shear Loading of Beams.
Computational Mechanic 36: 384-397.
Schramm, U., Kitis, L., Kang, W. & Pilkey, W. D. 1994. On the Shear Deformation Coefficient in Beam
Theory. Finite Elements in Analysis and Design 16: 141162.
Schramm, U., Rubenchik, V. & Pilkey, W. D. 1997. Beam Stiffness Matrix Based on the Elasticity
Equations. International Journal for Numerical Methods in Engineering 40: 211232.
Stephen, N.G. 1980. Timoshenkos Shear Coefficient from a Beam Subjected to Gravity Loading. ASME
Journal of Applied Mechanics 47: 121127.
Thambiratnam, D. & Zhuge, Y. 1996. Free vibration analysis of beams on elastic foundation. Computer
& Structures 60(6): 971-980.
Thomson, W. T.1981. Theory of Vibration with Applications. EnglewoodCliffs Prentice Hall.
Timoshenko, S., Young, D.H. & Weaver, W. 1974. Vibration problems in Engineering. New York:
fourth ed Wiley.
Timoshenko, S.P. & Goodier, J.N. 1984. Theory of Elasticity. New York.3
rd
edn: McGraw-Hill.
Wang, T.M. & Stephens, J. E. 1977. Natural Frequencies of Timoshenko Beams on Pasternak
Foundation. Journal of Sound and Vibration 51(2): 149-155.
Yokoyama, T. 1988. Parametric Instability of Timoshenko Beams Resting on an Elastic Foundation.
Computers & Structures 28(2): 207-216.
Zhang, Y. Q., Lu, Y., Wanga, S. L. & Liu, X. 2008. Vibration and buckling of a double-beam system
under compressive axial loading. Journal of Sound and Vibration 318: 341352.

329
1 INTRODUCTION
The legend that an underground structure was safer than the structure above the ground col-
lapsed on 1995 Hyogoken Nanbu earthquake. It is important to secure the aseismic reliability of
an underground structure with the frame similar to the structure above the ground. The execut-
ing time for nonlinear response analysis by use of the method is not short though the finite ele-
ment method has been often used in the seismic design of an underground structure. Because
the aseismic reliability is evaluated based on the variation of the response to by analyzing a lot
of cases to consider the uncertainty of the seismic ground motion and the material properties of
ground, the application of the finite element method is difficult on practical use.
The objective of this report is to develop the simple method to evaluate the damage probabili-
ty of underground structure accurately. The method to take into account of the uncertainty in
term of the seismic action and material properties of ground based on FEM is described. The
method to evaluate the probability distribution of the seismic responses obtained by using the
finite element method proposed by Tsutumi et al is combined with the idea to consider the un-
certainty of the seismic action according to the ground motion simply.
2 METHOD TO EVALUATE REFERENCE EARTHQUAKE GROUND MOTION
CONSIDERING UNCERTAINTY
The objective point to evaluate fragility curve is Kobe. The curve is evaluated by using haz-
ard information exhibited on web site as seismic hazard station (JSHIS). Hence, PGV (peak
ground velocity) is used as the intensity of earthquake ground motion in the fragility curve as
well as the previous exhibited hazard information. 20 records among the seismic observed re-
cords in the past earthquakes were taken as reference earthquake ground motions at each PGV
by considering the source property of the records which are similar to the source property with
the high influence to the object site and the maximum amplitude of the records were adjusted to
Seismic risk of the underground structure considering the uncer-
tainty of seismic action

S. Nakamura
Dept. of Civil Engineering, College of Engineering, Nihon University , Koriyama, Japan
S. Nishiyama
Nikken Sekkei Civil Enginnering, Osaka, Japan
T. Matsumoto
Newjec, Osaka, Japan
Y. Miyagawa
Central Research Institute of Electric Power Industry ,Chiba, Japan
ABSTRACT: In order to evaluate the earthquake risk of an underground structure, the method
for combining a dynamic analysis with FOSM considering the uncertainty of the spatial dis-
tribution of soil properties has been proposed. The dynamic response analysis such as a FEM
has generally multiple degrees of freedom. Therefore, the method has difficulty to apply the risk
evaluation considering the uncertainty such as the seismic ground motion because the huge
numbers of analysis are necessary. Here, the method which is able to consider the uncertainty of
the seismic ground motion and the uncertainty of a structural material as major uncertainties in
simply is proposed to evaluate the risk of an underground structure.

330
PGV. If the fragility curve according to an underground structure is evaluated by use of monte
carlo method for considering the uncertainty of the earthquake ground motion and the uncertain-
ty of material properties in ground based on FE analysis, the FE analyses for huge number of FE
models have to be carried out. It is difficult to obtain the fragility curve within the practical
time. Therefore, the median of the dynamic response of structure is evaluated by the analysis for
FE model with the mean material properties of soil against one reference earthquake ground
motion that has an average characteristic among 20 waves. The influence of the spatial variance
of the ground physical properties is considered indirectly.
The method to evaluate the reference earthquake motion with the average characteristic
among the 20 waves is described. First of all, 1D nonlinear earthquake response analyses for
ground are carried out for the models considering not only the uncertainty of reference earth-
quake motion but also the uncertainty of the material properties in ground. The relative dis-
placement between the depth at upper floor of the structure and the depth at the lower floor is
noticed among the analytical results because dynamic response of an underground structure is
strongly affected by the deformation characteristics of ground. After converting those values in-
to the naturalized logarithm, the relationship between PGV and the median is calculated. Next,
1D nonlinear earthquake response analyses for ground are carried out for the models with mean
material properties in ground considering only the uncertainty of reference earthquake motion.
A corresponding result to the median according to the relative displacements obtained by the
former analyses is chosen among 20 results obtained by the latter analyses. The reference earth-
quake motion used as a input motion for the analytical model of the result is considered as the
motion which have the average characteristics among 20 waves at each PGV.
Next, the ground model for the analysis considering the uncertainty of material properties in
ground is described as follows. The ground structure according to S wave velocity is shown in
Figure.1. The engineering bedrock is specified at the lower interface in the layer with shear
wave velocity of 450m/sec under the depth of 17.2m. Moreover, the uncertainty of material
properties in ground is considered by the spatial distribution according to the initial shear
modulus. The stochastic characteristic is assumed to be the normal distribution. Among the
characterized values, the mean value is specified to be the value obtained by shear wave veloc-
ity. 1000 ground structural models of shear wave velocity are generated based on the assump-
tion of probabilistic distribution according to shear wave velocity. An example among 1000
models is shown in Figure.1. When 1000 ground structure models arrange with 20 reference
earthquake ground motions, 50 ground structure models allocate to a reference earthquake
ground motion.
3 STRUCTURAL MODEL AND DEFINITION OF THE LIMIT STATE
The subway station among the underground structures is taken into account as the analytical
model. The structural dimension is modeled as the same with Daikai subway station which was
20
15
10
5
0
D
e
p
t
h
(
m
)
300 200 100 0
Shear wave Velocity(m/s)
Variable Example
Mean Value

Figure.1 Example model for the distribution of shear wave velocity with depth
331
suffered by Hyogoken-Nanbu earthquake in 1995. However, section performance such as the
arrangement of steel bar is designed against the design input motion called as level 2 earthquake
motion specified on the aseismic design standard for the railway structure revised after the
earthquake. On the standard, the failure mode at the ultimate state is specified to be flexural
failure due to sufficient arrangement for shear strengthening. The ground structure in terms of
shear velocity is shown in Figure.1. 2 dimensional FE model is used to consider the dynamic
interaction between the structure and ground. The nonlinear seismic response analysis consider-
ing nonlinearity of soil and structural member is carried out in the time-domain. A example of
the analytical model is shown in Figure.2.
The damage degree generated in the column is noticed as the limit state of the structural
system because the column among the structural members in the subway station governs the
seismic performance according to the collapse of the structure in comparison of the floor plate
and the sidewall. When the deformation characteristics of the structural member is defined by
the relationship between the bending moment (M) and curvature (!), the damage level of the
structural system specified by the relationship between the external load acted to the frame
model of structure and the relative displacement between the upper floor and the lower floor is
defined as shown in Figure.3. The three different limit state are assumed to be caused at the
three different points y, m and n indicated on the relationship and are correspond with damage
level 2, damage level 3 and damage level 4 respectively. The uncertainty of the value at each
limit state is mainly considered by the uncertainty of material properties of structural member
because the value at limit state is obtained by the push over analysis considering flexural defor-
mation characteristics as the member properties. The value at limit state is named as the limit
value.
When the push over analysis is carried out to obtain the limit value, the corner at member
joint is modeled to be rigid zone and the nonlinear characteristics for the flexural deformation
between the bending moment and curvature specified by the railway standard is used. Moreo-
ver, after the earth pressure etc acted to the structure as permanent load, the concentric load acts
to the upper floor as the seismic action during the earthquake. As a concrete strength, three cas-
es are set to be 26.9, 30.3 and 33.7 N/mm
2
respectively. As a result, each relationship between
the applied load and the obtained displacement is shown in Figure.4.

Figure.2 Analytical 2D FE model

Figure.3 Damaged level of the member shown in relationship
between Load and displacement

Damage level
332
In any concrete strength, the damage state at first appears in the column. When the displace-
ment reaches to the values between 97mm and 102mm, the damage at level 4 is caused.
Moreover, it is found that the concrete strength doesnt influence the limit value.
4 DAMAGE PROBABILITY OF STRUCTURE
The damage probability is obtained by the following equation.

!
f P (
B PGV ) ="(#ln
R M (
B PGV )
C M
/
R
2
$ +
C
2
$ ) (1)

hence,
!
R M (
B PGV ) represents the median among the response relative displacement between
upper floor and lower floor. The value obtained by nonlinear 2 dimensional FE analysis is used
as the median of the stochastic characteristics according to the response relative displacement.
Furthermore,
!
c M
and
!
B PGV
represent the median of the limit values obtained by push over
analysis on the limit state, peak ground velocity at the engineering bedrock respectively.
!
R
"
and
!
C
" represent the Logarithm standard deviation for the response values and the limit values
respectively. Here, the statistical characteristics of the response and the limit value is assumed
to be expressed by lognormal distribution, and
!
" in equation represents the standard normal
probability distribution.
Based on the method proposed by Tsutsumi (2007), the variation of the response values is
expressed by two kinds of variation. One is the relative displacement between the depth at upper
floor in the ground and the depth at lower floor in the ground. And the others is the relative dis-
placement in the structure mentioned above. The latter variation is ignored because the former
variation is larger than the latter variation. As for the variation of limit value, an epistemic

Figure.4 Relationship between load applied to whole structure model and displacement
at the representative position

1.0
0.8
0.6
0.4
0.2
0.0
C
o
n
d
i
t
i
o
n
a
l

D
a
m
a
g
e
d

P
r
a
b
a
b
i
l
i
t
y
5 6 7 8 9
10
2 3 4 5 6 7 8 9
100
2 3 4 5
PGV at base(cm/s)
Calculated Damaged Probability
Fitted Fragility curve

Figure.5 Fragility curve

Displacement
Damage
Level3(n)
Damage
Level2(m)
Damage
Level1(y)
A
p
p
l
i
e
d

L
o
a
d

P

333
uncertainty was not considered because the values are calculated analytically by the over push
analysis, and an aleatory uncertainty was also disregarded because Tsutsumi et al (2007) de-
scribed that the influence on the variation is small. Here, the limit state is assumed to be a state
at damage level 4, and the relation between PGV at engineering bedrock and the conditional
damage probability corresponding to PGV is shown in Figure.5. The conditional probability of
the damage per year can obtain by multiplying the annual occurrence frequency for each PGV.
The annual damage probability can be calculated by summing the conditional probability of the
damage per year, and becomes 5.35"10
-4
. The reliability index converted from the probability
becomes 3.27.
The damaged probability of the underground structure at KOBE was evaluated by the com-
bined methods to evaluate simply the probability distributions of the seismic responses taken in-
to account of the uncertainty of the material properties of ground by using the finite element
method proposed by Tsutsumi H, (2007) by considering the uncertainty of the reference earth-
quake ground motion based on exhibited seismic hazard information. The same structural di-
mension of subway station with Daikai station collapsed during Hyogoken Nanbu earthquake in
1995 is considered as the underground structure to evaluate the annual damage probability.
The analytical structure model is designed by the revised railway seismic design standard after
the earthquake. As a result, It is found that the annual damage probability of the subway station
becomes 5.35"10
-4
.
References
Tsutsumi H. , Ebisawa K., Nakamura S. 2007. A Proposal of the Method to evaluate the failure probabili-
ty for an underground structure in nuclear poer plants Vol63, No.4, pp. 704-715, J. S. C. E. (A) (in
Japanese)

334
1 INTRODUCTION
The aggravation of the seismic motion in the vicinity of topographic irregularities has been stud-
ied in the past as it !as validated against e"perimental analytical and numerical results# $ list of
these studies can be found in $ssima%i & 'ausel ()**+,# In the most recent years the combined
effects of topographic features and soil-foundation-structure interaction (-.-I, have been the fo-
cus of various studies ($ssima%i & 'ausel )**+ -te!art & -holtis )**/ 'im & -te!art
)**0, trying to illustrate the effect of the modification of the incident !ave field due to t!o-di-
mensional phenomena caused by the topographic aggravation on the seismic response of a soil-
foundation-structure system#
The present study is meant to be a continuation of the afore mentioned studies focusing on the
combined effects of the -.-I and topography on the dynamic response of simplified structures
resting on the free surface of the soil close to the topographic irregularity# Through a series of
parametric analyses using the finite element method the additive effects of the initial soil stiff-
ness of the topographic geometry (inclination and height, and of the parameters affecting tradi-
tionally the -.-I phenomenon (relative soil to structure stiffness slenderness structural to foun-
dation mass, are elucidated on the dynamic response of a simple soil-foundation-structure system#
) NU12RIC$3 1OD23
)#1 Geometry
The numerical model used in the analyses is sho!n in .igure 1# It consists of appro"imately 4/**
plane 5-node finite elements the si6e of !hich !as selected based on the dynamic characteristics of
the input motion# In the regions of interest i#e# in the vicinity of the crest at the foot of the slope and
around the foundation of the structure the finite element mesh is refined# $bsorbing elements are
placed at the sides and bottom of the soil model to ensure the correct !ave propagation and to avoid
spurious reflections into the soil#
The full length of the soil model is /**m !hile the total height varies depending on the chosen
slope height# .or the t!o slope heights of )*m and /*m chosen in the analyses the total soil profile
model height at its thic% side is /*m and 7*m respectively# The slope inclination !as chosen to be
Topographic irregularities and soil 8 foundation - structure inter-
action
D# 9itila%is
Aristotle University of Thessaloniki, Greece
$:-TR$CT; This study focuses on the influence on the seismic response of a soil-foundation-
structure system; i, of the geometry of a nearby topographic irregularity (inclination of slope
height, ii, of the dynamic soil characteristics (shear modulus stratification depth of bedroc%,
iii, of the dynamic characteristics of the simplified superstructure (stiffness slenderness mass,#
.rom a series of parametric analyses using the finite element method useful conclusions are e"-
tracted on the degree of influence of the aforementioned parameters on the seismic response of the
complete soil-foundation-structure system and conse<uently on the seismic design of the founda-
tions#
335
0* and 5/ degrees# The centerline of the structure is founded at small distance from the crest !here
pea% topographic aggravation is determined according to $ssima%i et al# ()**/,#
.igure 1# Numerical model used in the finite element analyses#
)#) Soil
.or the soil model five different configurations of a uniform homogeneous material !ere used#
The soil density and 9oisson=s ratio !ere %ept to 17**%g>m
0
and *#00 respectively !hile the shear
!ave velocity !as chosen to be 1**m>s )**m>s 0**m>s 5**m>s and 4**m>s resulting in elastic
shear moduli ranging bet!een 1719a and 4/*19a#
The resonant fre<uency of the soil profile of depth /*m at free-field conditions is *#/?6 1?6
1#/?6 )?6 and 0?6 respectively for the five shear !ave velocities mentioned above !hile for
the deeper soil profile of 7*m the soil resonant fre<uency attains the values of *#01?6 *#4)?6
*#@0?6 1#)/?6 and 1#7+?6 respectively#
1oreover according to Aeli et al# (1@77, the topographic aggravation is e"pected at fre<uen-
cies calculated by the simplistic formula of H/
0
=0.2 !here H is the slope height and
0
is the
dominant propagating !avelength# This !ould produce for the slope height of the )*m and the
above mentioned soil shear !ave velocities (1**m>s to 4**m>s, e"pected topographic resonant
fre<uencies at 1?6 )?6 0?6 5?6 and 4?6 respectively# $ccordingly for the slope height /*m
the e"pected resonant fre<uencies of the topographic aggravation !ould be at *#5?6 *#7?6
1#)?6 1#4?6 and )#5?6 respectively#
)#0 Structure
The first structure used in the analyses is a single-degree-of-freedom system consisting of a mass
of 1**t clumped at a free height of 1*m over the foundation center of gravity# The properties of
the hypothetical structure bearing the mass produce a resonant fi"ed-base fre<uency of 0#)?6#
The foundation is a s<uare !ith side dimension /m# This results in a slenderness ratio 5#
The second structure differs from the first in the total height the latter being /m resulting in a
slenderness ratio of )# The resonant fre<uency of the second shorter structure is appro"imately
e<ual to the one of the tall at 0#1?6#
0 9$R$12TRIC $N$3B-2-
-everal sets of parametric analyses !ere performed in this study in order to identify and evaluate
the combined effects of the topographic irregularities and the soil-foundation-structure interaction
on the dynamic response of a structure# Aiven the comple"ity of the )D problem of a soil-founda-
tion-structure system resting close to the crest of a cliff the parametric analyses focused on the
effect of the soil stiffness the geometry of the topographic irregularity and the structural charac-
teristics affecting the interaction !ith the soil# .ollo!ing are presented some of the basic findings
in the form of fre<uency response functions (.R., or acceleration time histories#
Three consecutive linear elastic analyses !ere performed to obtain the .R. of the soil-founda-
tion-structure system; The first set of analyses included only the soil profile !ithout topographic
irregularities or structure in order to obtain the free-field response# This !as necessary because
as it !ill be sho!n ne"t the free-field response to vertically propagating shear !aves -C should
be clean of parasitic hori6ontal and mainly vertical components caused either by the structural
vibration or by the topographic aggravation and )D phenomena# Ideally a control point in the fi-
nite element model at the free soil surface a!ay from the structure or the slope !ould do ho!ev-
er this !as found to be not ade<uate for the needs of this study# The second set of analyses com-
336
prised of the soil-foundation-structure system !ithout the topographic irregularity !hile the third
set of analyses included the topographic aggravation as !ell#
0#1 ffect of the initial soil stiffness
Initially the effect of the soil stiffness is studied# The slope inclination and height is 5/ degrees
and )*m respectively# The tall structure !ith the height of 1*m is founded on the free soil surface
close to the crest# Due to the slope height and inclination the soil shear !ave velocity of 1**m>s
!as e"cluded from the results as it !ould probably cause slope stability problems# $ccording to
Celetsos & 1ee% (1@+5, the principal parameters affecting the dynamic response of a soil-foun-
dation-structure system are; i, the !ave parameter ! = "s/f
0
h !here f
0
is the fi"ed-based struc-
tural fre<uency and h the structure=s height and ii, the slenderness ratio h/r !here r is the char-
acteristic dimension of the foundation# In this case the slenderness ratio is 5 !hile the !ave pa-
rameter e"pressing the relative stiffness bet!een the soil and the structure ranges from 4#)/ for
CsD)**m>s to 17#+/ for CsD4**m>s#
In .igure ) is sho!n the .R. of the soil-foundation-structure system bet!een the top of the
structure and the free-field for the soil shear !ave velocities of )**m>s 0**m>s 5**ms> and
4**m>s# The fi"ed-base resonant fre<uency of the tall structure is at 0#)?6# Clearly the -.-I ef-
fects are more pronounced for the softer soil profile sho!n by a larger shifting of the system res-
onant fre<uency to lo!er fre<uencies# .urthermore the response of the complete system including
topography is more comple" as it resonates in more than one fre<uency# Ehile the -.-I re-
sponse is governed mainly by a single resonant fre<uency each time at the fundamental fre<uency
of the soil-foundation-structure system the response of the system including the topographic ir-
regularity resonates at at least three fre<uencies notably the resonant fre<uency of the system the
resonant fre<uency of the topographic aggravation and a third parasitic fre<uency#
.igure )# .re<uency response function bet!een the top of the structure and the free-field for linear elas-
tic soil response slope height )*m and 5/ degrees inclination for four different soil shear !ave veloci-
ties (cloc%!ise from top left; )**m>s 0**m>s 4**m>s and 5**m>s,# The fi"ed-base response (solid line,
is compared !ith the -.-I response (dashed line, and the -.-I and topography response (dashed-dotted
line,#
337
.igure 0# .re<uency response function bet!een the top of the structure and the foundation (left, and the
top of the structure and the elastic bedroc% for linear elastic soil response (right, for slope height )*m
and 5/ degrees inclination and for shear !ave velocity of )**m>s# The fi"ed-base response (solid line, is
compared !ith the -.-I response (dashed line, and the -.-I and topography response (dashed-dotted
line,#
To clarify the above mentioned results ne"t the case of the soil shear !ave velocity of )**m>s
is isolated# The .R. bet!een the top of the structure and the center of the foundation (.ig# 0 left,
isolates the response of the soil-foundation-structure system !hich naturally coincides for the
t!o different cases (!ith and !ithout topography effects, at appro"imately 1#5?6 shifting do!n
from 0#)?6 for fi"ed-base condition# Ehen loo%ing the .R. bet!een the top of the structure and
the bedroc% (.ig# 0 right, ho!ever the resonant fre<uency of the topographic aggravation and
the parasitic component of the motion are clearly sho!n at 1#1?6 and at 1#+?6 respectively#
These resonances are not met in the case of solely the -.-I !here the resonant fre<uency of the
soil profile and of the complete system are found at 1?6 and 1#5?6 respectively# 1oreover the
topographic resonant fre<uency is in general found at lo!er fre<uencies than the ones predicted
by Aeli et al#(1@77,#
Ehen the system is subFected to the $egion 1@@/ $egion Areece earth<ua%e record (9A$
*#5@g, the acceleration response time history at the top of the structure is presented in .igure 5#
$s !itnessed from the .R. as !ell the effects of the -.-I are more pronounced from the effects
of the topography for the stiffer soil material !hile for the softer soil material the topographic ef-
fects control mostly the response# Conse<uently the acceleration response at the top of the struc-
ture for the softer soil is found to be larger for the case !here the topography is included !hile
for the stiffer soil it is dominated by the -.-I phenomenon#
The .R. for the vertical component of the motion of the system is sho!n in .igure / !here
the response of the fi"ed-base and the -.-I is practically nil at least in the range of fre<uencies
of interest up to /?6 !hile for the case !here the topography is included the system responds in
as many as four resonant fre<uencies# The vertical acceleration response at the top of the struc-
ture elucidates the presence of this vertical component of the vibration caused by the trapping of
the incident and the diffracted !aves !ithin the structure forcing the structure to intense roc%ing#
338
.igure 5# $cceleration response at the top of the structure !hen subFected to the $egion 1@@/ earth-
<ua%e record (9A$ *#5@g, for linear elastic soil response slope height )*m and 5/ degrees inclination
for four different soil shear !ave velocities (cloc%!ise from top left; )**m>s 0**m>s 4**m>s and
5**m>s,# The -.-I response (dashed line, is compared !ith the -.-I and topography response (dashed-
dotted line,#
.igure /# .re<uency response function of the vertical component bet!een the top of the structure and the
free-field (left, and vertical acceleration response at the top of the structure (right, !hen subFected to the
$egion 1@@/ earth<ua%e record (9A$ *#5@g, for linear elastic soil response slope height )*m and 5/
degrees inclination and soil shear !ave velocity )**m>s# The fi"ed-base response (solid line, is com-
pared !ith the -.-I response (dashed line, and the -.-I and topography response (dashed-dotted line,#
339
0#) ffect of the #eometry
0#)#1 Hei#ht of slo$e
Ne"t the effect of the topographic geometry on the system response is designated# In the results
presented above the slope height !as )*m# Ehen the slope height increases from )*m to /*m
the .R. of the system are sho!n belo! in .igure 4 for the four different soil shear !ave veloci-
ties#
In this case the -.-I response for all four soil stiffness resembles the response for the -.-I
presented for a )*m slope height in .igure )# The only difference is that in the case of the slope of
/*m the total soil profile height is 7*m !hich in turn affects the resonant fre<uency of the sys-
tem# Eorthy to note ho!ever is the fact that for the larger slope height the influence of the to-
pography is more pronounced for all four soil stiffness than it is for solely the -.-I case# This
might be e"pected as the topographic aggravation in this case is much more severe causing in-
tense t!o-dimensional phenomena#
.igure 4# .re<uency response function bet!een the top of the structure and the free-field for linear elas-
tic soil response slope height /*m and 5/ degrees inclination for four different soil shear !ave veloci-
ties (cloc%!ise from top left; )**m>s 0**m>s 4**m>s and 5**m>s,# The fi"ed-base response (solid line,
is compared !ith the -.-I response (dashed line, and the -.-I and topography response (dashed-dotted
line,#
The acceleration response at the top of the structure !hen the system is subFected to the $e-
gion 1@@/ $egion Areece earth<ua%e record (9A$ *#5@g, is presented in .igure +# .ollo!ing
the trend in the .R. the amplification of the motion due to topographic effects is larger than only
for -.-I for soft and for stiff soil as !ell# 1oreover the richer fre<uency content of the acceler-
ation response in the case !ith the topographic irregularity is attributed to the topographic reso-
nant fre<uency and the parasitic components of the motion that are caused in the vicinity of the
soil-foundation-structure system#
340
.igure +# $cceleration response at the top of the structure !hen subFected to the $egion 1@@/ earth-
<ua%e record (9A$ *#5@g, for linear elastic soil response slope height /*m and 5/ degrees inclination
for four different soil shear !ave velocities (cloc%!ise from top left; )**m>s 0**m>s 4**m>s and
5**m>s,# The -.-I response (dashed line, is compared !ith the -.-I and topography response (dashed-
dotted line,#
0#)#) ffect of the slo$e inclination
Ne"t the slope inclination effect on the system response is investigated# The soil shear !ave ve-
locity is )**m>s in order to promote -.-I effects# The slope inclination is chosen to be 0* and 5/
degrees# .igure 7 sho!s the .R. bet!een the top of the structure and the free-field for slope
height )*m (.ig# 7 left, and /*m (.ig# 7 right,# The main effect of the slope inclination on the
system response seems to be that for larger angles of the slope the response presents more para-
sitic components of motion# Naturally this topographic aggravation is more pronounced for larg-
er slope heights#
341
.igure 7# .re<uency response function bet!een the top of the structure and the free-field for slope incli-
nation 0* and 5/ degrees for soil shear !ave velocity )**m>s and slope height )*m (left, and /*m
(right,# The fi"ed-base response (solid line, is compared !ith the -.-I and topography response for
slope inclination 0*degrees (dashed line, and for slope inclination 5/ degrees (dashed-dotted line,#
0#0 ffect of the structure slen%erness
.inally the effect of the slenderness of the structure is e"amined# To this end the response of the
shorter system (h>rD), is compared !ith the response of the taller structure (h>rD5,# The slope
height and inclination is )*m and 5/ degrees respectively and the soil shear !ave velocity is
)**m>s#
.igure @ sho!s the response for the soil-foundation-structure system alone (.ig#@ left, and for
the complete system including the topographic irregularity (.ig#@ right,# Concerning the case ne-
glecting the topographic effect the response naturally follo!s the e"pected -.-I trend !ith larg-
er system softening for the tall slender structure# On the other hand the amplification of the mo-
tion due to topographic effects happens to be larger for the shorter structure possibly because of
resonance of the system !ith the resonant fre<uency of the topographic aggravation at appro"i-
mately )?6#
.igure @# .re<uency response function bet!een the top of the structure and the free-field for slope height
and inclination )*m and 5/ degrees respectively soil shear !ave velocity )**m>s and -.-I !ithout to-
pography effects (left, and !ith topography effects (right,# The fi"ed-base response (solid line, is com-
pared !ith the response of the shorter structure (h>rD), (dashed line, and !ith the taller structure
(h>rD5, (dashed-dotted line,#
342
5 CONC3U-ION
The combined effects of -.-I and topographic aggravation !ere sho!n by means of parametric
analyses# Due to the comple"ity of the phenomenon the parameters and characteristics entering
the finite element model !ere chosen so as to elucidate the effects on the structural response of
the soil dynamic characteristics of the geometry of the topography and of the characteristics of
the foundation-structure system# Thus for a structure founded close to the verte" of a cliff the
principal conclusions can be summari6ed as follo!s; i, The larger the slope height and inclina-
tion the larger the influence of the topography on the response of the system# This is due to the
development of vertical vibration modes caused by t!o-dimensional phenomena e"aggerated in
the vicinity of the structure because of the interaction !ith the structure# ii, The incident motion
can be amplified even for stiffer soil due to the topographic irregularity# iii, The %no! effects of
the soil-foundation-structure interaction (system softening increased damping due to radiation,
are still met in the analyses even though more pronounced in some cases due to t!o-dimensional
phenomena near the crest of the cliff#
$C'NOE32DA12NT-
The author !ould li%e to ac%no!ledge the contribution of his students that !or%ed on the para-
metric analyses#
R2.2R2NC2-
$ssima%i D# & 'ausel 2# )**+# 1odified Topographic $mplification .actors for a single-faced slope
due to %inematic soil-structure interaction# &ournal of Geotechnical an% Geoenvironmental n#i'
neerin# 100(11,; 1515-1501#
$ssima%i D# 'ausel 2# & Aa6etas A# )**/# Eave propagation and soil-structure interaction on a cliff
crest during the 1@@@ $thens earth<ua%e# Soil (ynamics an% arth)uake n#ineerin# )/(+-1*,;
/108/)+#
Aeli 3# :ard 9#-B# & Gullien :# 1@77# The effect of topography on earth<ua%e ground motion; $ re-
vie! and ne! results# *ulletin of Seismolo#ical Society of America +7; 5)840#
'im -# & -te!art G# 9# )**0#'inematic soil-structure interaction from strong motion recordings# &our'
nal of Geotechnical an% Geoenvironmental n#ineerin# 1)@(5,; 0)0800/#
-te!art G#9#& -holtis -#2# )**/# Case study of strong ground motion variations across cut slope# Soil
(ynamics an% arth)uake n#ineerin# )/; /0@-/5/#
Celetsos $# & 1ee% G#E# 1@+5# Dynamic behavior of building-foundation systems# arth)uake n#i'
neerin# an% Structural (ynamics 0; 1)1-107#
343
1. INTRODUCTION
Non linear soil-structure interaction of a shallow foundation subjected to ground shaking has not
yet received much attention in the geotechnical research and practice. Earthquake analysis of
soil-foundation systems is often based on the assumption of linear or equivalent-linear soil be-
haviour and full-contact conditions between foundation and soil. However, post-earthquake ob-
servations of shallow foundations over the last decades have demonstrated that nonlinear behav-
iour and permanent displacements is rather inevitable during strong seismic shaking. In
particular, for slender structural systems where rocking is the prevailing mode of response,
nonlinear effects emerge from:
the negligible tensile capacity of the soil-foundation interface during swaying-
rocking motion, which results in uplifting of the foundation as well as in developing
of second order (P-) effects (geometrical nonlinearity),
the plastification of the supporting soil, especially underneath the foundation edges,
generated by the concentration of vertical stresses and amplified by the cyclic re-
sponse of the superstructure (material nonlinearity).
These two mechanisms may lead to severe maximum and permanent foundation displace-
ments (rotation and settlement) which cannot be captured by conventional, linear or equivalent
linear computational methods. A series of centrifuge tests have been performed at LCPC in the
framework of the EU research project QUAKER to highlight the nonlinear aspects of the cy-
clic, dynamic response of shallow foundation. Then, non-linear finite element analysis was util-
ised (a) to simulate the experimental tests and (b) to carry out a fundamental sensitivity study in
view of the dynamic failure envelope in the N-M plane, the dynamic (transient) foundation up-
lift, and the permanent vertical displacement (settlement) accumulated due to cyclic loading.

Numerical modelling of shallow foundation under cyclic
overturning moment and validation through centrifuge
experiments
M. Apostolou
National Technical University, Athens
ABSTRACT: Rocking of stiff structures on shallow foundation has been found to enter into the
nonlinear regime even under moderate seismic shaking. Post-earthquake observations of shal-
low foundations on stiff soils have revealed that cyclic inertial moment is often associated with
large levels of transient uplift about the foundation edges. Several cases of soft soils have shown
that permanent rigid-body displacements of the building (settlement, tilting) may occur, even if
no prominent structural failure takes place. In the framework of the QUAKER research project
a series of centrifuge tests have been performed at LCPC to investigate a building with a slen-
derness ratio of resting on clay, under monotonic and cyclic loading. Ultimate capacity and per-
manent deformation of the foundation have been recorded and discussed. Nonlinear finite ele-
ment modeling of the experimental tests highlights the effects of the problem parameters on the
foundation response. It is shown that cyclic loading may increase the moment capacity of a
foundation when compared to the static case.
344
2. CENTRIFUGE TESTS
In this series of experiments, a 100x100 mm
2
footing (80x80 mm
2
in few cases) over soft
saturated clay is submitted to (a) monotonic, displacement-controlled vertical loading to failure
(determination of bearing capacity), (b) monotonic horizontal loading to failure under constant
dead weight, and (c) cyclic horizontal loading under self weight (displacement-controlled) up to
0.4 mm (0.4 m in prototype dimensions) and under a driving frequency of 0.10 Hz - 0.16 Hz.
The tests were performed under a centrifugal acceleration of 100 g meaning that a scale of 1/100
should be applied to derive the prototype model. Horizontal loading is applied to the structural
center of gravity at a height of 100 mm above the foundation level leading to a height-to-width
ratio of two. Two values of the structural dead weight were chosen to investigate the influence
of the vertical load on the rocking response. Principally, a building with a dead weight of 1284 t
(for test Tub3 - T07) or 1370 t is implemented, corresponding to a heavily-loaded foundation
(M1). A building with a dead weight of 580 t is also used corresponding to a lightly-loaded
foundation (M2). The former gives a vertical loading ratio of = N/N
u
= 0.6 whereas for the lat-
ter it is = N/N
u
= 0.26. Soil material used in the experiments is saturated kaolin Speswhite
clay. The total depth of the soil sample inside the container is 263 mm corresponding to a soil
stratum width of 248 mm. Each container has been prepared by consolidation under stress in lab
with three or four successive layers of clay. Cone Penetrometer tests were performed at 1 g be-
fore consolidation in the centrifuge, and also in-flight after reconsolidation and just before load-
ing the structure. Shear vane tests were also carried out at several points to estimate s
u
. Profiles
of s
u
with depth were determined using CPT and well-established correlation between q
c
and s
u

measured in-flight on Speswhite clay (Garnier, 2001):
/ 18.5
c u
q s = (1)
Geometry and instrumentation of the model utilised at the centrifuge tests are shown in Fig-
ure 1.

Figure 1. Loading device (left) and deformed position of model M2 after loading (right)

The list of the loading tests performed is presented in chronological order in Table 1 together
with the values of the ultimate capacity.

Synopsis of test results: Prior to lateral loading, preliminary displacement-controlled tests
have been performed to estimate the vertical bearing capacity of the foundation for the two
structural configurations (tests T01 to T06). Due to the log-type shape of the vertical load-
settlement curve no clear failure point could be identified. To overcome this, two conventional
failure criteria were established for settlement level of 4.5 mm and 10 mm. On the other hand,
horizontal load-displacement monotonic curve after initial yielding tends to a horizontal line de-
termining the lateral load capacity of the foundation. Ultimate loads in horizontal and vertical
loading direction are presented in Table 1.
LVDTs
PPTs
Load cell
Load cell
Servo-jack
Building model
LVDTs
Loading
direction
LVDTs
PPTs
Load cell
Load cell
Servo-jack
Building model
LVDTs
Loading
direction
345
Table 1. Loading program of the centrifuge tests and ultimate load
Tub n
o
Test Foundation (mm x mm) Loading sequences
Ultimate load
(MN)
Tub 1 T01
8 x 8 Vertical static (DC) 10

T02 10 x 10
Vertical static (DC) (can-
celled)
-
Tub 2 T03
8 x 8 Vertical static (DC) 16.5
T04
10 x 10 Vertical static (DC) 24
Tub 3 T05
10 x 10 Vertical static (DC) -
T06
10 x 10 Vertical static (LC) 22
T07
10 x 10 - Building M1 Horizontal static (DC) 1.4
Tub 4 T08

T09 10 x 10 - Building M1
Horizontal static (DC)
Horizontal cyclic (DC)
1.3
2.0
T10
10 x 10 - Building M1
-
Tub 5 T10
10 x 10 - Building M1
Horizontal cyclic (LC)
-
T11
10 x 10 - Building M2 Horizontal cyclic (LC) -
Tub 6 T12

T13
10 x 10 - Building M1 (+sand
layer)
Horizontal cyclic (LC) -
Tub 7 T14
10 x 10 - Building M1 Horizontal static (DC) -
T15 10 x 10 - Building M2 (+sand
layer)
Horizontal cyclic (LC) -

3. NUMERICAL SIMULATION OF CENTRIFUGE TESTS
A series of two-dimensional finite element analysis was performed to simulate the centrifuge
experiments. The prototype model has been implemented in the numerical study so that all di-
mensions at the centrifuge model have been properly scaled up. A lumped-mass structure with a
square footing (10x10m
2
) is considered to represent the building. The mass point located at a
height of 10 m above the foundation level is connected to the foundation with a (rigid) beam
element so that no flexural deformation of the superstructure is permitted. Horizontal loading is
applied at this level. Rigid beam elements have been also utilised to prevent foundation mat de-
formations. The rigid boundary at the bottom is placed at a depth of 25 m below the foundation
level. Nonlinear soil behaviour is described with a nonlinear constitutive model which incorpo-
rates the von Mises yield criterion combined with an isotropic and kinematic hardening model
in the post-yield domain. This model is most suitable for the analysis of the dynamic behaviour
of cohesive soils under undrained conditions. Some of the analyses were repeated by utilizing
the elasticperfectly plastic M-C model. Linear undrained strength profiles were estimated from
in-flight CPT results in association with Equation (1) as presented in Table 2. These profiles
have been utilised in the two-dimensional finite element analysis. The favorable effect of verti-
cal loading to soil strength underneath the foundation has been taken into account by increasing
the values of su at surface, up to s
u
(B/4). Due to lack of experimental data for the soil stiffness,
Youngs modulus at low deformations is considered as a linear function of the undrained shear
strength. Different formulae are implemented to estimate the soil stiffness during horizontal and
vertical direction.
A typical comparison of the numerical and the experimental vertical load-settlement curve
(tub3 test) is presented in Figure 2. Upper and lower bound distributions of su from tub3 (pro-
files b and c respectively) have been utilized in the numerical simulation. Both numerically
computed backbone curves capture the initial stiffness and the hardening behaviour (after the
yield onset), of the centrifuge test. The large-displacement response and the ultimate bearing
capacity calculated with profile b however are much closer to the centrifuge results. Similar
346
trends for the vertical backbone curve are extracted from the simulation of the other tests. The
experimental load-displacement curve under monotonic horizontal loading (tub3 test) is pre-
sented in Figure 3. Both profiles b and c of s
u
were used for the numerical interpretation. In this
case however, the soil underneath the foundation has been strengthened due to the gravitational
preloading (12.6 MN). Hence, an increase of the undrained strength is considered as shown in
Figure 3. A correlation of E = 1200s
u
was adopted for the elastic soil modulus. An excellent
agreement between the experimental and the numerical results is achieved when the profile b is
considered. It is also uncovered from the numerical analysis that the foundation response is
marginally influenced by the contact conditions at the interface (rough or smooth). On the con-
trary, the ultimate horizontal load merely reaches up to 300 kN when the profile c is used; a
value which is far less than the centrifuge result.
The foregoing analysis procedure was repeated with profile b, for different values of the ini-
tial load N varying from near zero to the ultimate value N
u
. The horizontal load was obtained (a)
at incipient yield and (b) at u/B = 0.1. The derived failure envelope in the N-Q space (plotted in
Figure 4) can be approximated by a parabola with a local maximum at near the half of N
u
. This
maximum value of the shear force reaches merely 1.7 MN which is significantly lower than
As
uo
= 2.9 MN. The difference between the two values is attributed to the interaction in Q-M
space in the former case. Furthermore, for values of N close to N
u
the lateral load at u/B = 0.1
exceeds increasingly the yield load, due to a hardening effect.

Table 2. Linear distributions of the undrained strength with depth based on the CPT results at 100g. These
values of s
u
have been implemented in the finite element analysis.
Test
Undrained shear strength
(kPa)
Notes
Tub1 T01 12.5 3.24 = +
u
s z
Tub2 T03/T04 8.9 1.74 = +
u
s z
T06 18.5 4.24
u
s z = + Profile b
Tub3
T07 7.6 2.64
u
s z = + Profile c
Tub4 T09 6.4 3.65
u
s z = +
Tub6 T012 5.6 5.87 = +
u
s z

Figure 2. Monotonic vertical load-settlement curve calculated with centrifuge experiment (tub3, testT06)
and comparison with the two-dimensional FE analysis.

0
5
10
15
20
25
0 20 40 60 80 100
s
u
: kPa
z
:

m
18.5+4.24z
7.6+2.64z
0
5
10
15
20
25
30
0 0.2 0.4 0.6 0.8 1
w: m
N
:

M
N
2-d FE analysis (Profile c - rough interface)
-//- (Profile b - rough interface)
Tub3 results
347

Figure 3. Monotonic lateral load-displacement curve from centrifuge experiment (tub3, testT07) com-
pared with the two-dimensional FE analysis

Figure 4. Monotonic failure envelope in the N-Q space calculated with two-dimensional FE analysis
(tub3). For each level of vertical load, the ultimate horizontal force is calculated at incipient yield and at
u/B = 0.1

Cyclic loading: Cyclic horizontal loading at the level of the gravity centre, under a constant
vertical load was also performed in centrifuge. Typical results of the tub4 test (T09) are pre-
sented in Figure 5. Initially, a monotonic loading is applied to the structure until a prototype dis-
placement of 0.2 m is obtained. Then the building is removed automatically to its initial posi-
tion. The second (dynamic) loading phase is subdivided in three displacement-controlled cyclic
sequences: (a) 10 cycles at 0.1 Hz with an amplitude of 0.4 mm (0.04 m in prototype dimen-
sions), (b) 10 cycles at 0.16 Hz with an amplitude of 0.4 mm, and (c) 9 cycles at 0.16 Hz with
an amplitude of 0.4 mm. After cyclic loading, a monotonic loading is applied to re-calculate the
ultimate horizontal force. In this step the maximum force has increased from 1.2 MN to 2.1 MN
which is attributed to the preceding cyclic loading. The interpretation of tub4 cyclic test is also
presented in Figure 5. Finite element analysis captures both the initial and the residual (in-
creased) shear force capacity of the foundation. Also dynamic numerical analysis provides the
0
5
10
15
20
25
0 20 40 60 80 100
s
u
: kPa
z
:

m
18.5+4.24z
7.6+2.64z
0
0.5
1
1.5
2
2.5
3
0 0.1 0.2 0.3 0.4 0.5 0.6
u
cm
: m
Q
:

M
N
2-D FE analysis (Profile b - rough interface)
-//- (Profile b - smooth interface)
Tub3 results
0
500
1000
1500
2000
2500
3000
0 5000 10000 15000 20000 25000
N : kN
Q

:

k
N
at u/B = 0.1
at yield point
348
same maximum force with the experimental value (1.8 MN). In the numerical loops however an
isotropic behaviour is revealed in the loading and unloading directions.

Figure 5. Backbone Q-u curves before and after cyclic loading (tub4) from the centrifuge (left) and nu-
merical simulation (right).

4. FUNDAMENTAL SENSITIVITY STUDY
Nonlinear behaviour of shallow foundations subjected to earthquake loading is investigated
next through a parametric finite element study.
Dynamic failure envelope N-M: A plethora of analytical studies for the calculation of inter-
action curves in the generalized N-M loading space has been presented by many researchers up
to recently. They are all limited to the study of the problem through a static treatment (e.g. upper
and lower bound solutions). However, dynamic rocking behaviour of the footing with many
significant cycles may lead to substantially higher levels of the moment capacity especially for
large values of the vertical load as observed in recent experiments including the QUAKER cen-
trifuge tests (see Figure 6).

Figure 6. Failure envelope points in the N-M plane from recent experiments (left) as presented in Gajan et
al. (2005) and centrifuge results from the QUAKER project (right)

(1) : Initial static loading
(4) : Final static loading (after dynamic loading*)
-1.5
-1
-0.5
0
0.5
1
1.5
2
2.5
0 0.2 0.4 0.6 0.8 1
u
cm
: m
Q
:

M
N
Start of dynamic
loading
End of dynamic loading** -
Start of monotonic reloading
* First 10 cycles (0.1Hz, amplitude 0.04 m), 10 cycles (0.16Hz, 0.04 m), and last 9 cycles (0.16Hz, 0.04 m)
** 6 cycles, 0.1Hz, amplitude 0.04 m
**
* Centrifuge experiments (cyclic loading) at Davis, California.
** Limit-state pseudostatic analysis.
*
*
*
N [MN]
349
A parametric finite element analysis has been performed in this study to elucidate the way
dynamic loading affects the failure envelope of a shallow foundation. Initially, a pseudostatic
approach is considered. Butterfield and Gottardi (1994) first proposed that the shape of the static
failure envelope in the N-M plane for a strip footing on sand can be approximated by a parabola
with a local maximum at N = N
u
/2. Houlsby and Puzrin (1999) utilised the theorems of limit
analysis to derive upper and lower bounds of the failure envelope for a cohesive (undrained)
soil. The simplest analytical treatment to derive the failure envelope in the N-M plane is to rep-
resent the supporting soil with distributed elasticperfectly plastic springs (beam-on-Winkler-
foundation). A zero tensile capacity is attributed to the springs so that separation of the founda-
tion from soil is permitted. After yield onset at an individual spring, a perfectly plastic behav-
iour is considered with p = p
u
. The failure envelope in the N-M plane for the Winkler model is
presented in Figure 7. In the same graph the failure envelope calculated with a non-linear push-
over numerical analysis is plotted. An excellent agreement is achieved between the analytical
and the numerical solution. Remarkably, for values of N close to N
u
a small swelling of the
numerical curve is observed, attributed to the passive forces developed behind the footing edge.

Figure 7. The simplified Winkler foundation model (top left) and the failure envelopes from static and
earthquake analysis.

A time-domain finite element analysis series to calculate failure envelopes under dynamic
loading was also performed. In this case a s-dof uplifting structure is assumed (2b = 1 m, h = 5
m) supported by a soil layer with s
u
= 50 kPa. The seismic bedrock is merely at the depth of -5
m so that any filtration of the excitation frequency content through soil is prevented (T
s
< 0.1
sec). The footing however, can deliberately undergo rocking oscillations as if it was supported
by a half-space soil medium (K
R,H=5m
= 1.04K
R,H-S
). A Ricker pulse-type excitation is applied at
the seismic bedrock with a predominant period of 0.33 sec, 0.67 sec, and 1.33 sec (Ricker
nominal frequency f
R
of 2.0, 1.0, and 0.5 respectively). It is considered that these values cover
the period range of a typical near-fault pulse-type motion. Three levels of ground shaking have
been implemented in the analysis; a weak, a moderate and a strong shaking level (PGA: 0.2 g,
0.4 g, and 0.6 g correspondingly). For each loading case the dynamic moment capacity of the
foundation is calculated for different values of the gravitational load N and failure envelopes in
the N-M loading space are derived. Initially, the moment capacity of the footing is calculated
for the short-period excitation (0.33 sec) as plotted in Figure 7. In the same graph the linear
0
50
100
150
200
250
300
0 100 200 300 400 500 600
N: kN
M
:

k
N
m
FE - Ricker1.0 - 0.20g
FE - Static
Winkler soil model:
M = Nb(1-)
Rigid soil: M = Nb
0
50
100
150
200
250
300
0 100 200 300 400 500 600
N: kN
M
:

k
N
m
FE - Static
Winkler soil model:
M = Nb(1-)
Rigid soil: M = Nb
w
bo
2 2
N
p
u
Q
N
Overturning moment: Q h
Restoring moment: N (b - )
M
p
u
2 = N or
=
350
rigid soil failure envelope is also presented. For the moderate and strong shaking level, a dy-
namic over-strength is developed which is enhanced with the increase of N. Counter-
intuitively, this beneficial dynamic behaviour is highly amplified close to the limiting value of
= 1 where the static M
u
approaches zero. On the other hand, for low levels of N the ultimate dy-
namic response can be successfully predicted by the static approach or even by the rigid soil
linear curve (0). Only for the weak level of ground shaking the dynamic failure envelope is
bounded by the static curve almost throughout the range of N. This occurs because for low lev-
els of shaking the dynamic capacity of the foundation may not be reached. Two more Ricker
wavelets with a long-duration pulse are also utilised in the study with the above-mentioned
shaking levels. The resulting failure envelopes are plotted in the graphs of Figure 7. What is
more interesting now is that the dynamic over-strength not only is even more amplified but it
also approaches the rigid soil moment capacity, especially for the case of T
E
= 1.33 sec.

Foundation uplift: During strong seismic shaking, rocking motion of a shallow foundation is
often associated with large uplift from the supporting soil. The uplifting level is higher in case
of a lightly-loaded foundation or a slender structural system. A simplified estimation for the up-
lifting level at static overturning conditions can be derived from the schematic of Figure 8. The
supporting soil has been substituted with elastoplastic distributed springs (see Figure 7).

Figure 8. The numerically computed effective width under monotonic loading (a) at the increment of
maximum moment (M
u
), (b) at failure (M = 0), and comparison with the elastoplastic Winkler model. B =
2b = 2 m, s
u
= 50 kPa.

Equilibrium of forces acting on the foundation at limit state provides that the part of the foot-
ing remaining in contact with soil is 2 = N / p
u
or = / b = . A finite element validation of
this approximation comes from the static results portrayed in the top-left graph of Figure 8.
Moreover, the earthquake-induced uplift of a shallow foundation is investigated for a Ricker-
type excitation. In the other three graphs, the effective width 2 at M = M
u
is calculated from
the above-discussed numerical study of the dynamic moment capacity. Depending on the fre-
0
0.5
1
1.5
2
0 100 200 300 400 500 600
N: kN
2
:

m
Winkler foundation ( = )
at M = Mu
at failure (M = 0)
0
0.5
1
1.5
2
0 100 200 300 400 500 600
N: kN
2
:

m
FE - Static
0
0.5
1
1.5
2
0 100 200 300 400 500 600
N: kN
2
:

m
FE - Static
0
0.5
1
1.5
2
0 100 200 300 400 500 600
N: kN
2
:

m
FE - Static
351
quency content of ground shaking, the dynamic interaction curve in the 2N plane may be lo-
cated above or below the static linear trend = . The difference of the dynamic from the static
failure envelope is enlarged by the intensity of ground shaking. Hence, for a strong shaking
(PGA = 0.6 g) the dynamically-induced uplift may be substantially lower (T
E
= 0.33 sec) or
higher (T
E
= 1.33 sec) than the static prediction. On the other hand, for a weak excitation (PGA
= 0.2 g) equation = can practically describe the uplifting level.

Permanent cyclic settlement: It has been found that cyclic rocking motion of the foundation
may lead to significant accumulation of permanent settlement (Gajan et al, 2004 among others).
This cyclic vertical displacement w
res
which is added to the initial (static) settlement w
o
is
primarily sensitive to the number of cycles, the footing width, the safety factor , and the fre-
quency content of ground shaking. A parametric study was conducted to find estimates of the
additional settlement as a function of the rotation amplitude
max
. It has been shown that the cy-
clic settlement is very sensitive to the level of vertical loading whereas for limited levels of up-
lift, linear trends in the w
res
-
max
relationship may be extracted. All the numerical results are
summarized in the normalized diagram of Figure 9. Similar correlation between the cyclic set-
tlement and the rocking amplitude has been recently presented in the literature extracted from a
large data base of centrifuge test (Gajan, et al, 2005) as portrayed in the same figure.

Half amplitude of cyclic rotation, max [rad]

Figure 9. Normalized cyclic (additional) settlements of the foundation from (a) the present numerical
study (top), (b) centrifuge experiments by Gajan et. al, 2004 (bottom).
0
0.01
0.02
0.03
0.04
0 0.04 0.08 0.12 0.16
max
: rad
w
r
e
s
:
2
b
= 0.15 0.25
= 0.35 0.45
= 0.55 0.65 = 0.75 0.85
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
0.01
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
N
o
r
m
a
l
i
z
e
d

S
e
t
t
l
e
m
e
n
t

p
e
r

C
y
c
l
e
,

U
V
/
c
y
c
l
e
FSv = 2.0 (Weissing)
FSv = 3.4 (SSG02)
FSv = 3.4 (SSG02)
FSv = 4.0 (SSG03)
FSv = 4.1 (KRR02)
FSv = 5.3 (SSG02~Dynamic)
FSv = 6.8 (SSG02)
FSv = 8.2 (SSG03)
FSv = 9.6 (SSG02)
Dotted lines:
FS = 4 or less
Solid lines:
FS = 5 or more
Clay?, FS=2.8
0
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
0.01
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07
N
o
r
m
a
l
i
z
e
d

S
e
t
t
l
e
m
e
n
t

p
e
r

C
y
c
l
e
,

U
V
/
c
y
c
l
e
FSv = 2.0 (Weissing)
FSv = 3.4 (SSG02)
FSv = 3.4 (SSG02)
FSv = 4.0 (SSG03)
FSv = 4.1 (KRR02)
FSv = 6.8 (SSG02)
FSv = 8.2 (SSG03)
FSv = 9.6 (SSG02)
Dotted lines:
FS = 4 or less
Solid lines:
FS = 5 or more
Clay?, FS=2.8
352

5. CONCLUDING REMARKS
The most important findings drawn from the present study are as follows:
Two-dimensional finite element analysis provides in general a very good interpreta-
tion of the centrifuge results under dynamic and cyclic loading conditions. A satisfac-
tory agreement between plane-strain and experimental tests is also achieved.
After cyclic loading, a slight increase of the horizontal capacity of the foundation is
observed at the centrifuge and captured in the numerical simulation.
Dynamic failure envelopes in the N-M plane may reach considerably higher levels
than the corresponding static curves.
Under monotonic loading the part of the footing remaining in contact with soil (effec-
tive width: 2) when the moment capacity of the foundation is mobilised can be es-
timated by the simple relationship = 2 = . This formula can be easily drawn by
the footing on elastic-perfectly plastic Winkler springs model. Even under dynamic-
cyclic loading conditions the effective width can be approximately calculated through
this equation; however the frequency and the amplitude of the seismic excitation may
be significantly different from the linear trend under certain circumstances.

ACKNOWLEDGEMENTS
The writers gratefully acknowledge the financial support received from the European Commis-
sion in terms of the research project Fault-Rupture and Strong Shaking Effects on the Safety of
Composite Foundations and Pipeline Systems (QUAKER), contract number: EVG1-CT-2002-
00064. The authors acknowledge the help the LCPC-Nantes group of, especially of Jacpue Gar-
nier, Sandra Escoffier, Jean Louis Chazelas.

REFERENCES
Anastasopoulos I., Gazetas G., Loli M., Apostolou M., Gerolymos N. (2009), Soil failure can be
used for seismic protection of structures, Bulletin of Earthquake Engineering (available
online).
Apostolou M, Gazetas G & Garini E. Seismic response of slender rigid structures with founda-
tion uplifting, Soil Dynamics and Earthquake Engineering, 2007, No. 7, 642-654.
Apostolou M, Thorel L, Gazetas G, Garnier J, and Rault G. Physical and numerical modelling
of soil-footing-structure under lateral cyclic loading, Proc. 4
th
ICEGE, 2007, Thessaloniki,
Greece.
Bartlett P. (1976), Foundation rocking on a clay soil, M.E. Thesis, Report No. 154, School of
Eng., Univ. of Auckland, 144 pp, 1976.
Housner G. The behaviour of inverted pendulum structures during earthquakes Bulletin of the
Seismological Society of America, 53, No. 2, 403-417, 1963.
Hibbitt, Karlsson, & Sorensen. Abaqus 6.5/Standard users manual, Hibbitt Providence,
Rhode Island, 2005.
Houlsby G. and Purzin A. The bearing capacity of strip footing on clay under combined load-
ing, Proc. Royal Society, 455A, 893-916, 1999.
Butterfield R and Gottardi G. A complete three-dimensional failure envelope for shallow foot-
ings on sand, Gotechnique 44, No. 1, 181-184, 1994.
Gazetas G and Apostolou M. Nonlinear soil-structure interaction: Foundation uplifting and soil
yielding, 3
rd
UJNR Workshop on Soil-Structure-Interaction, Menlo Park, California, 2004.
Gazetas G, Apostolou M and Anastasopoulos J. Seismic uplifting of foundations on soft soil,
with examples of Adapazari (Izmit 1999 earthquake), International Conference on Founda-
tions, 37-49 Dundee, Scotland, 2003. Gajan S, Kutter B, Phalen J, Hutchinson T and Martin
353
G. Centrifuge modelling of load-deformation behaviour of rocking shallow foundations,
Soil Dynamics and Earthquake Engineering 25, 773-783, 2005.
Garnier J. Physical models in geotechnics: state-of-the-art and recent advances, 1st Coulomb
Conference, Paris, 2001.
Rault G, Thorel L and Garnier J. Non linearity of soil-footing interaction, Quaker program,
Topic B2, Progress Report No. 1, 2005.
354
4.
Bridges:Analysis,Design,andRetrofit
Company
Logo
Yukio ADACHI, Hanshin Expressway Engineering
Hidesada KANAJI, Hanshin Expressway
Tsutomu NISHIOKA, Hanshin Expressway
Retrofit technologies of
long-span bridges in Hanshin Expressway
Company
Logo
Expressway network in Osaka area
Kyoto
Osaka
Kobe
Company
Logo
Long-span bridges in Osaka bay
Wangan route
Higashi Kobe Bridge
885m, 1992
Minato Bridge
980m, 1974
Yamatogawa Bridge
653m, 1982
14 Bridges
Rokko Island Bridge
217m, 1992
Company
Logo
Lessons from 1995 Kobe EQ
Rokko Island Bridge
355
Company
Logo
Higashi-Kobe Bridge
Company
Logo
Nishinomiya-ko Bridge
Company
Logo
Strategy of retrofit
Consider source and site specific
ground motions
Use latest structural analysis
Use latest isolation and damping
enhancement technologies
Company
Logo
Strategy of retrofit
Strengthen detail parts of
bearings and etc.
-Especially connection portions.
Provide fail safe structure to keep
bridge function
356
Company
Logo
Example: Minato bridge
Type Cantilever Truss
Span 980 m
Weight 45000 ton
Traffic per day 89,400 vehicles
Completion 1974
|Long span Truss}
No.1: Quebec Bridge
No.2: Forth Bridge
No.3: Minato Bridge
Company
Logo
Example: Minato bridge
Hinge Hinge
[Elevation]
[Plan]
[Tower]
Suspended Span
Cantilever Span Floor Deck
Sway Bracing
Lateral Bracing
Chord member
Company
Logo
Site specific ground motion
10
100
1000
10000
0.1 1 10
Period (sec)
A
c
c
e
l
e
r
a
t
i
o
n

(
g
a
l
)
Asperity-a Asperity-b
Asperity-c Mode#1
Mode#2 Original
10
100
1000
10000
0.1 1 10
Period (sec)
A
c
c
e
l
e
r
a
t
i
o
n

(
g
a
l
)
Asperity-a Asperity-b
Asperity-c Mode#1
Mode#2 Original
Longitudinal direction Transverse direction
Effective in period shift Isolation Not effective in period shift Damping enhancement
Company
Logo
Damage control design
Main truss
Sub
members
Elastic
Inelastic
Floor
Isolation
Buckling
Restrained
Brace
357
Company
Logo
Floor isolation
Rubber Spring Sliding Bearing
Cross Beam
of Truss
Floor Deck
Company
Logo
Buckling Restrained Brace
Lower Lateral Braces
Tower Sway Braces
Hollow Steel
Restraining steel
Core Steel (LY225)
Company
Logo
Buckling Restrained Brace
Effect by Floor Isolation for Longitudinal Dir.
Effect by inelastic brace for Transverse Dir.
Red: Buckling, Blue: Tensile yield, Green: Elastic
Stress Ratio
a
1<R R<-1 -1R1
Before
After
After
Before
Company
Logo
Cable Installation at Hinges
Cable installation
Cable
Existing eye bar
Hinge
358
Company
Logo
Restrainers
Base
Restrainer
Cable
Company
Logo
Steel adhesion
Steel Plate
Plate
addition
Company
Logo
Struts
Strut addition
Company
Logo
Example: Tempozan bridge
Type Cable stayed
Span 640 m
Weight 22,000 ton
Completion 1990
359
Company
Logo
Strengthen existed bearings
Fixed bearing
Original Design: Flexible design using high tower
Fixed bearing
Excessive displacement
limiting structure
Excessive displacement
limiting structure
Fixed bearing
Company
Logo
Steel panel damper
Steel panel damper
Steel panel damper
Conc. filling
Company
Logo
Strengthen of side piers
Side piers
Restrainer cable
Rubber
+
Stopper
Company
Logo
Example: Higashi Kobe bridge
Type Cable stayed
Span 885 m
Weight 25,549 ton
Completion 1995
360
Company
Logo
'P187)
(c)O(`I019I1I
(e)^`51
(d)^`5`/
(a)``I (b) |0g
Company
Logo
|j'.1.';I
jj'.1@)g
16.7n 16.7n
!`
`
`
]/|
4,700
2
,5
0
0

]|

/]|
1)
1,200
J5`/0}|
Company
Logo
Summary
Lessons from the 1995 Kobe
earthquake were very important to
conduct seismic retrofit of long-span
bridges.
Latest technologies, isolation and
damping enhancement, reborn the
long-span bridges.
Strengthen of the detailed parts,
especially connection parts, are
necessary to reborn.
361
1 INTRODUCTION
Nestos bridge is among the 12 major highway bridges of Egnatia Odos. Located at the North-
eastern part of Greece, at the borders of KavalaXanthi provinces (Fig 1), has a total length of
460m. The broader area where the bridge is located is almost flat with a light bilateral inclina-
tion ranging from 0.5% to 1.5% towards Nestos riverbed. The bridge consists of two independ-
ent branches, each with 11 piers and 2 abutments, and spans of 38 m. A photo of the bridge at
near completion stage is given in Fig 2.
The general geology of the broad area consists of Pleistocenic sediments (gravel, coarse sand
and moraine) and semi-consolidated pleistocenic sediments (cemented gravel and sands).
At the study area where the bridge is founded, soil materials are distinguished as follows
(Klimis et al., 2004):
- River deposits of river Nestos consisting of sands, silty and clayey silty sands, gravels
and occasionally pebbles of gneiss or marble.
- Alluvial deposits essentially of sandy nature, with fluctuating percentage of clays, silts
and gravels. Occasionally, thin layers of gravels and clays / silts are encountered. Ac-
cording to geotechnical investigation programs carried out for this area, alluvial depos-
its are of considerable thickness ranging from 12 to almost 45m. Those layers are rela-
tively loose and considerably uneven, and in some cases an increased percentage of
organics has been detected.
The bedrock consists mainly of biotitic gneiss locally interpolated by amphibolites and mar-
bles encountered at depth which varies from borehole to borehole from 15 m to 45 m. the
ground water level was founded at a depth ranging from 1.5 m to 6.0m below ground surface.
An extensive geotechnical survey showed the existence of a 10 m to 20 m thick soil layer sus-
ceptible to mass liquefaction.
Pile Foundations in Improved Soil for Highway Bridge facing
LiquefactionInduced Soil Flow

N. Gerolymos
G. Gazetas
Professor, National Technical University, Athens, Greece
ABSTRACT: The recently constructed 460 m long Nestos bridge (Egnatia Odos), includes 11
piers and 2 abutments with spans of about 38 m. It is founded on fluvial deposits containing po-
tentially liquefiable sand and silty sand for a total thickness of 5 m to 20 m. The objective of
this paper is to analyse the seismic response of the pile foundation and to evaluate the effective-
ness of ground improvement with stone columns. Three solutions are parametrically investi-
gated: 35 m long friction piles of d = 1.2 m diameter (1) without soil improvement; (2) with lim-
ited soil improvement (with dimensions in plan equal to that of the pile-cap); and (3) extended
improvement (with dimensions in plan equal to about of a span). The effective stress dynamic
analysis is carried out numerically, considering full coupling of the poromechanical soil prop-
erties.
362
According to the official study it had been scheduled the bridge to be constructed by the can-
tilever method with a total length of 952 m and a central span of 120m. The piers would have
been supported by group of piles 54 m in length and 1.5 m in diameter (Fig 3). Despite the com-
puted liquefaction of the foundation soil, the piles had been designed to behave satisfactorily
under the loading of the liquefied soil. However, at the hazard of liquefaction the risk of exten-
sive lateral spreading of the liquefied soil was quite large to necessitate the use of mitigation
measures. The proposed countermeasure was the local improvement of the foundation soil of
each pile group (limited to plan dimensions equal to that of the pile cap) through the construc-
tion of 0.8 m in diameter gravel columns at center-to-center distances of 2.25 m.

Nestos Bridge Nestos Bridge

Fig. 1. The Nestos bridge is located at the Northeastern part of Greece, at the borders of KavalaXanthi
provinces

Taking into consideration that the construction of 54 m long piles in loose sandy soil below
ground water level is not an easy task, the foundation solution proposed in the official study was
withdrawn. Instead, the use of 30 m to 35 m long friction piles 1.2 m in diameter was suggested
(Fig 4). Given the increased vulnerability of friction piles to liquefactioninduced displacements
and to settlements due to soil consolidation (in the post-liquefaction stage) compared to that of
end-bearing piles, the design of soil improvement proposed in the official study was also re-
vised.
The goal of this paper is to study the effectiveness of three foundation solutions to seismic
loading with respect to the plan dimensions of the gravel columns grid: (1) without soil im-
provement, (2) with limited soil improvement (with dimensions in plan equal to that of the pile-
cap), and (3) extended improvement (with dimensions in plan equal to about of a span). The
analysis is carried out with the finite difference code FLAC, considering full coupling of the
poromechanical soil properties.
A simple constitutive soil model is used (Byrne, 1991), with the advantage against more so-
phisticated models that only one parameterthe SPT valueis needed for the analysis, a pa-
rameter which is included in almost every geotechnical program. A brief description of the
model is given in the sequel.
2 GEOTECHNICAL PARAMETERS AND SEISMIC EXCITATION
Evaluating the findings of an extensive geotechnical program which included SPT, CPT, and
crosshole tests, we formed idealized profiles of the shear wave velocity (Fig 5) which is a cru-
cial step in a soil-dynamic response analysis.

363

Fig 2. Aerial photo of Nestos bridge at near completion stage (Egnatia Odos, August 2008)

The calculation was based on the following two-step methodology: First, the initial (from the
crosshole tests) shear wave velocity profile is utilized as the input soil profile for an equivalent
linear type of analysis with the computer code SHAKE (Schnabel, Lysmer and Seed, 1972).
Then, the shear modulus and damping ratio distributions from the last iteration of the SHAKE
analysis, serve as the input soil profile in the analysis with FLAC.
The contribution of gravel columns to the increase of the shear wave velocity is estimated
from the following empirical correlation (Rollins et al, 1998):
( ) ( )
0.17 0.12
60
115
s
V N o' = (1)
In which N
60
is the SPT value, and o' is the effective vertical stress.
Four recorded acceleration time histories were selected and scaled down to a peak ground ac-
celeration of 0.26g to describe the excitation at rock outcrop. Among them, the Kalamata record
was found to give the most unfavorable results.
The permeability coefficient is taken equal to k = 4 10
-6
m/s for the silty-sands, and equal to k
= 10
-2
m/s for the stone columns. The improved soil mass (the gravel columns and the native
soil) is considered as a soil block of homogenized properties with an equivalent permeability
coefficient of k
eq
10
-3
m/s. This was calculated by equating the time periods for radial (gravel
columns and native soil---actual case) and vertical (equivalent soil with homogenized proper-
ties) drainage, for a consolidation percentage which corresponds to a duration time that does not
exceed the duration of strong motion. That is about 6 sec for Kalamata record.
3 METHOD OF ANALYSIS
The seismic analysis is performed using the finite difference code FLAC2D (plain strain analy-
sis) applying the effective stress method and a simple elasto-plastic constitutive soil model to
describe the cyclic stressstrain relationship of soil behaviour.
364
The aforementioned constitutive model includes (a) a Mohr-Coulomb failure criterion, (b) a
non-associative flow rule, and (c) a simple hardening law (Byrne, 1991) for excess pore-water
pressure build-up. According to this law the relation of the increment of volume decrease,
vd
,
to the cyclic shear strain amplitude, , where is presumed to be the engineering shear strain,
is given by the following empirical expression:

Gravel
columns
3
5

m
Silty
Sand
Rock
Liquefiable
zone
?
M4 M6 M6 M7
38 m
2
0

m
5
4

m
Gravel
columns
3
5

m
Silty
Sand
Rock
Liquefiable
zone
?
M4 M6 M6 M7
38 m
2
0

m
5
4

m
M5
Gravel
columns
3
5

m
Silty
Sand
Rock
Liquefiable
zone
?
M4 M6 M6 M7
38 m
2
0

m
5
4

m
Gravel
columns
3
5

m
Silty
Sand
Rock
Liquefiable
zone
?
M4 M6 M6 M7
38 m
2
0

m
5
4

m
M5

Fig 3. Sketch of Nestos bridge foundation (Piers M4 to M7). Official Solution (2003)

3
5

m
Liquefiable
zone
?
M4 M6 M6 M7
38 m
2
0

m
Silty
Sand
Rock
Gravel
columns
3
5

m
Liquefiable
zone
?
M4 M6 M6 M7
38 m
2
0

m
Silty
Sand
Rock
Gravel
columns

M5
3
5

m
Liquefiable
zone
?
M4 M6 M6 M7
38 m
2
0

m
Silty
Sand
Rock
Gravel
columns
3
5

m
Liquefiable
zone
?
M4 M6 M6 M7
38 m
2
0

m
Silty
Sand
Rock
Gravel
columns

M5

Fig 4. Sketch of Nestos bridge foundation (Piers M4 to M7). Final Solution (2007)

1 2
exp
vd
vd
C C
c
c
| |
A =
|
\ .
(2)
where C
1
and C
2
are constants. In many cases, C
2
= 0.4/C
1
, so Eq. (2) involves only one inde-
pendent constant; however, both C
1
and C
2
have been retained for generality. In addition, a third
parameter, C
3
, sets the threshold shear strain (i.e., the limiting shear strain amplitude below
which volumetric strain is not produced). The constant C
1
represents the tendency of the soil
material for contraction. Consequently, a high value of C
1
corresponding to a saturated soil
365
material, increases the possibility of liquefaction. On the other hand, the constant C
2
represents
the tendency of the soil material for no further volume change, as the accumulated volumetric
deformation increases.
N
SPT
s 20
N
SPT
~ 14
N
SPT
~20
PGA
rock
= 0.26 g
PGA
free field
= 0.40 g
Improved with
stone columns soil
profile
Initial Soil Profile
N
SPT
= 25
N
SPT
= 25
N
SPT
= 20
0
10
20
30
40
0 500 1000
Vs : m/s
z

:

m
N
SPT
~ 16
N
SPT
s 20
N
SPT
~ 14
N
SPT
~20
PGA
rock
= 0.26 g
PGA
free field
= 0.40 g
Improved with
stone columns soil
profile
Initial Soil Profile
N
SPT
= 25
N
SPT
= 25
N
SPT
= 20
0
10
20
30
40
0 500 1000
Vs : m/s
z

:

m
N
SPT
~ 16

Fig 5. Idealized geotechnical section used in our analyses

4.50 m
2
.
2
5

m
2.25 m
4.50 m
11 m
4.50 m
11 m
13.5 m
3
1
.
5

m
4.50 m
2.25 m
2.25 m
4
.
5
0

m
1
1
m
4
.
5
0

m
1
0
.
2
5

m
1
1
m
31.5 m
4
7
m
4.50 m
2
.
2
5

m
2.25 m
4.50 m
11 m
4.50 m
11 m
13.5 m
3
1
.
5

m
4.50 m
2
.
2
5

m
2.25 m
4.50 m
11 m
4.50 m
11 m
13.5 m
3
1
.
5

m
4.50 m
2.25 m
2.25 m
4
.
5
0

m
1
1
m
4
.
5
0

m
1
0
.
2
5

m
1
1
m
31.5 m
4
7
m
4.50 m
2.25 m
2.25 m
4
.
5
0

m
1
1
m
4
.
5
0

m
1
0
.
2
5

m
1
1
m
31.5 m
4
7
m

Fig 6. Cases II and III: Limited and extended soil improvement with gravel columns, respectively

FLAC contains a built-in constitutive model (named the Finn model) that incorporates Eq.
(9) into the standard Mohr-Coulomb plasticity model. As it stands, the model captures the basic
mechanisms that can lead to liquefaction in sand. In addition to the usual parameters (friction,
366
moduli, etc.), the model needs the three constants for Eq. (2). For Eq. (2), Byrne (1991) notes
that the constant, C
1
, can be derived from relative densities, D
r
, as follows:

-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
Rock outcrop
motion
Free field motion at ground surface
t : s
A

:

m

/

s
2
A

:

m

/

s
2
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
Rock outcrop motion
Free field motion at ground surface level
t : s
A

:

m

/

s
2
A

:

m

/

s
2
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
Rock outcrop
motion
t : s
A

:

m

/

s
2
A

:

m

/

s
2
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
Rock outcrop
motion
t : s
A

:

m

/

s
2
A

:

m

/

s
2
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
Rock outcrop motion
t : s
A

:

m

/

s
2
A

:

m

/

s
2
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
Rock outcrop motion
t : s
A

:

m

/

s
2
A

:

m

/

s
2

Fig 7. Computed acceleration time histories at ground surface for Case I (without soil improvement), and
Case II (limited soil improvement) with the Kalamata record scaled at 0.26 g as the seismic excitation at
the base of the finite difference models (for piers M6 and M7).

( )
2.5
1
7600
r
C D

= (3)
Furthermore, using an empirical relation between Dr and normalized standard penetration test
values, (N
1
)
60
,
( )
0.5
1
60
15
r
D N = (4)
then,
( )
1.25
1 1
60
8.7 C N

= (5)
C2 is then calculated from C2 = 0.4/C
1
in this case.
The corrected SPT value (N
1
)
60
is expressed as a function of the uncorrected SPT value N,
according to the following empirical relationship (Seed and Idriss, 1982; Liao and Whitman,
1986):
( ) ( )
1
60
N PI
N a b C N c = + (6)
where

0
a
N
v
p
C
o
=
'
(7)
is the correction factor for initial effective overburden depth. a and b are correction factors
due to the effect of the fine percentage FC:
367
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
Rock outcrop motion
t : s
A

:

m

/

s
2
A

:

m

/

s
2
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
-0.4
-0.2
0
0.2
0.4
0 4 8 12 16
Rock outcrop motion
t : s
A

:

m

/

s
2
A

:

m

/

s
2

Fig 8. Computed acceleration time histories at ground surface for Case III (extended soil improvement)
with the Kalamata record scaled at 0.26 g as the seismic excitation at the base of the finite difference
model (for piers M6 and m7).

( )
( )
2
1.5
0, 5%
exp 1.76 190/ , 5% 35%
5.0, 35%
1.0, 5%
1.0, 5%
0.99 /1000 , 5% 35%
1.2, 35%
a FC
a FC FC
a FC
b FC
b FC
b FC FC
b FC
= s
(
= < <

= >
= s
= s
(
= + < <

= >
(8)
and
PI
c the correction factor due to plasticity effect:

( )
1.0 10
1 0.022 10 10
PI
if PI
c
PI if PI
<
=

+ >
(9)
Pilesoil interaction is modeled through special composite elements (beam supported on
Winkler springs designated as pile element in FLAC element library). Each beam element is con-
sidered to behave elastically while the forcedisplacement relationship of the springs follows an
368
elasticrigidly plastic behaviour. In this way the 3-dimensional interplay of the pile with the soil
grid is efficiently captured in an equivalent 2-dimensional approach.

0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
4 m
12 m
16 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
4 m
12 m
16 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
4 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
4 m
16 m
20 m
26 m
36 m
2 m
16 m
26 m
35 m
48 m
B
B
AA
BB
2 m
16 m
26 m
35 m
48 m

0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
AA
BB
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
4 m
12 m
16 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
4 m
12 m
16 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
4 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
4 m
16 m
20 m
26 m
36 m
2 m
16 m
26 m
35 m
48 m
B
B
AA
BB
2 m
16 m
26 m
35 m
48 m

0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
AA
BB
2 m
16 m
26 m
35 m
48 m

0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
AA
BB

Fig 9. Computed excess pore water pressure ratio time histories at selected depths for Case I (without soil
improvement) and Case II (limited soil improvement) with the Kalamata record scaled at 0.26 g as the
seismic excitation at the base of the finite difference models (for piers M6 and M7).

The spring stiffness of the elastic branch is appropriately calibrated for the kinematic re-
sponseprior to soil liquefactionof the pile to match the actual one in 3-dimensions. The ul-
timate spring reaction governs the lateral spreading of the liquefied soil around the piles in an
equivalent plain-strain approach. Details on the calibration methodology of FLACs pile ele-
ments are provided in Tasiopoulou (2007).
4 RESULTS AND DISCUSSION
The calibration methodology for Byrnes (1991) constitutive model, described in the previ-
ous chapter, was implemented as a fish subroutine in code FLAC and applied to the analysis of
three foundation scenarios with respect to the plan dimensions of the improved soil area (a grid
of stone columns with center-to-center distances of 2.25 m, 20 m in length and 0.8 m in diame-
ter):
- A group of friction piles 35 m in length and 1.2 m in diameter, without soil improvement.
- A group of friction piles 35 m in length and 1.2 m in diameter, with limited soil im-
provement (31.5 m x 13.5 m in plan dimensions), see Fig 6.
- A group of friction piles 35 m in length and 1.2 m in diameter, with extended soil im-
provement (47 m x 31.5 m in plan dimensions), see Fig 6.
369
2 m
16 m
26 m
35 m
48 m
B
B
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
12 m 16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
12 m 16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
AA
BB
2 m
16 m
26 m
35 m
48 m
B
B
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
12 m 16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
r
u
8 m
12 m 16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
0
0.2
0.4
0.6
0.8
1
0 4 8 12 16
t : s
r
u
8 m
4 m
12 m
16 m
20 m
26 m
36 m
AA
BB

Fig 10. Computed excess porewater pressure ratio time histories at selected depths for Case III (ex-
tended soil improvement) with the Kalamata record scaled at 0.26 g as the seismic excitation at the base
of the finite difference model (for piers M6 and M7).

To investigate the effect of the rather abrupt inclination angle of the basal topography on the
response of piles under conditions of soil liquefaction, the seismic behaviour of the aforemen-
tioned foundation configurations was also checked against a possible bedrock inclination of 3
o
.
So we analysed six in total cases.
The results of the analyses are presented in Figs 7 to 12 in the form of time histories of (a)
horizontal acceleration at ground surface level, (b) excess pore-water pressure ratio at selected
depths, (c) seismically-induced settlement at the pile cap and the free-field surface, (d) seismic
bending moments at the pile-to-cap connection and at interfaces of abrupt change in stiffness
between any two successive soil layers. The results depicted in those figures correspond to piers
M6 and M7 with the Kalamata record scaled at 0.26 g as the seismic excitation at the base of the
model, and to flat (0
o
inclination angle) basal topography.
The following trends are worthy of note:

370
-30
-20
-10
0
0 4 8 12 16 20
t : s
w

:

c
m
: 12 cm
: 21.1 cm
-30
-20
-10
0
0 4 8 12 16 20
t : s
w

:

c
m
: 12 cm
: 21.1 cm
18 m
8 m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
: 2.5 MNm
: 0.97 m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
: 2.5 MNm
: 0.97 m
-20
-15
-10
-5
0
0 4 8 12 16 20
t : s
: 7.1 cm
: 16.5 cm
w

:

c
m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
:
1.5 MNm
:
0.65 m
18 m
8 m
Interface: 0.97 MNm
Pile head: 2.5 MNm
Pile head: 1.5 MNm
Interface: 0.65 MNm
Free field: 16.5 cm
Pile cap: 7.1 cm Pile cap: 12 cm
Free field: 21.5 cm
Consolidation
Consolidation
Interface: 0.97 MNm
Pile head: 2.5 MNm
Free field: 21.5 cm
Interface: 0.97 MNm
Pile head: 2.5 MNm
Pile cap: 12 cm
Free field: 21.5 cm
Interface: 0.97 MNm
Pile head: 2.5 MNm
Consolidation
Pile cap: 12 cm
Free field: 21.5 cm
Interface: 0.97 MNm
Pile head: 2.5 MNm
-30
-20
-10
0
0 4 8 12 16 20
t : s
w

:

c
m
: 12 cm
: 21.1 cm
-30
-20
-10
0
0 4 8 12 16 20
t : s
w

:

c
m
: 12 cm
: 21.1 cm
18 m
8 m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
: 2.5 MNm
: 0.97 m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
: 2.5 MNm
: 0.97 m
-30
-20
-10
0
0 4 8 12 16 20
t : s
w

:

c
m
: 12 cm
: 21.1 cm
-30
-20
-10
0
0 4 8 12 16 20
t : s
w

:

c
m
: 12 cm
: 21.1 cm
18 m
8 m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
: 2.5 MNm
: 0.97 m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
: 2.5 MNm
: 0.97 m
-20
-15
-10
-5
0
0 4 8 12 16 20
t : s
: 7.1 cm
: 16.5 cm
w

:

c
m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
:
1.5 MNm
:
0.65 m
18 m
8 m
-20
-15
-10
-5
0
0 4 8 12 16 20
t : s
: 7.1 cm
: 16.5 cm
w

:

c
m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
:
1.5 MNm
:
0.65 m
18 m
8 m
Interface: 0.97 MNm
Pile head: 2.5 MNm
Pile head: 1.5 MNm
Interface: 0.65 MNm
Free field: 16.5 cm
Pile cap: 7.1 cm Pile cap: 12 cm
Free field: 21.5 cm
Consolidation
Consolidation
Interface: 0.97 MNm
Pile head: 2.5 MNm
Free field: 21.5 cm
Interface: 0.97 MNm
Pile head: 2.5 MNm
Pile cap: 12 cm
Free field: 21.5 cm
Interface: 0.97 MNm
Pile head: 2.5 MNm
Consolidation
Pile cap: 12 cm
Free field: 21.5 cm
Interface: 0.97 MNm
Pile head: 2.5 MNm

Fig 11. Computed seismic bending moment and settlement time histories for Case I (without soil im-
provement) and Case II (limited soil improvement) with the Kalamata record scaled at 0.26 g as the seis-
mic excitation at the base of the model.

- In the first case (without soil improvement): excessive soil liquefaction, maximum seis-
mic bending moment at pile-to-cap connection which exceeds the 2.5 MNm, and a resid-
ual settlement (at the end of the consolidation phase) of the pile cap equal to 12 cm.
- In the second case (limited soil improvement): also excessive soil liquefaction. Local soil
improvement is insufficient to mitigate the risk of liquefaction. The maximum seismic
bending moment at the pile-to-cap connection and residual settlement at the cap are 1.5
MNm and 7 cm, respectively. The beneficial influence of the limited soil improvement is
evident, but its contribution to ameliorate the seismic response of the piles is not ade-
quate.
- Finally, in the third case (extended soil improvement): marginal liquefaction with rapid
dissipation of excess pore water pressures during the seismic shaking. The maximum
seismically induced bending moment at pile heads is merely 0.9 MNm while the residual
settlement of the pile cap is less than 3.5 cm.
- The triggering of soil liquefaction is clearly reflected in the ground surface acceleration
time history. Lowamplitude accelerations corresponding to low frequency cycles for the
first and second case, by stark contrast to the third case where we observe higher in am-
plitude accelerations accompanied by higher in frequency cycles.
To shed light on the influence of bedrock inclination on the foundation response, Figs 13 and
14 summarize the results in terms of maximum seismic bending moments and residual dis-
placements, respectively, for flat and inclined at 3
o
basal topography. In fact, the modelling of
bedrock inclination is achieved by projecting the vector of gravitational acceleration to both axis
(the horizontal x and the vertical y). The following trends are interesting to observe:

371
-30
-20
-10
0
0 4 8 12 16 20
t : s
w

:

c
m
: 3.5 cm
: 26.1 cm
-30
-20
-10
0
0 4 8 12 16 20
t : s
w

:

c
m
: 3.5 cm
: 26.1 cm
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
: 0.9 MNm
: 0.92 m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
: 0.9 MNm
: 0.92 m
18 m
Consolidation
Pile cap: 3.5 cm
Free field: 26.1 cm
Interface: 0.92 MNm
Pile head: 0.9 MNm
-30
-20
-10
0
0 4 8 12 16 20
t : s
w

:

c
m
: 3.5 cm
: 26.1 cm
-30
-20
-10
0
0 4 8 12 16 20
t : s
w

:

c
m
: 3.5 cm
: 26.1 cm
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
: 0.9 MNm
: 0.92 m
-3
-2
-1
0
1
2
3
0 4 8 12 16
t : s
m
: 0.9 MNm
: 0.92 m
18 m 18 m
Consolidation
Pile cap: 3.5 cm
Free field: 26.1 cm
Interface: 0.92 MNm
Pile head: 0.9 MNm

Fig 12. Computed seismic bending moment and settlement time histories for Case III ( (extended soil im-
provement) with the Kalamata record scaled at 0.26 g as the seismic excitation at the base of the model.
0
1
2
3
Without improvement Limited improvement Extended improvement
2.5
1.5
0.9
2.0
1.0
1.9
0.65
:

M
N
m
0
1
2
3
Without improvement Limited improvement Extended improvement
2.5
1.5
0.9
2.0
1.0
1.9
0.65
:

M
N
m

Fig 13. Computed maximum seismic bending moment for the three cases studied: at pile-to-cap connec-
tion for flat bedrock (yellow shaded area---circles) and inclined at 3
o
bedrock (green shaded area---
squares), and at interfaces between two successive soil layers with abrupt change in stiffness (light blue
shaded area---triangles).
372
Extended
improvement
0
5
10
15
20
25
23
19
10
u
:

c
m
without
improvement
limited
improvement
Horizontal displacement
Extended
improvement
0
5
10
15
20
25
23
19
10
u
:

c
m
without
improvement
limited
improvement
Horizontal displacement

Fig 14. Computed maximum seismically induced horizontal displacements for the three cases studied, for
inclined at 3
o
bedrock.
- Bedrock inclination does not affect the maximum seismically developed bending moment
in the third case (extended soil improvement). As soil liquefaction is only marginal, the
inclined bedrock does not lead to lateral spreading. On the contrary, the maximum bend-
ing moment is about the same for the first two cases (without and with limited soil im-
provement), revealing that improving the soil properties just beneath the pier foundations
when the risk of soil liquefaction characterize the broader area of the bridge, is not an
adequate countermeasure.
- In the first case (without soil improvement), bedrock inclination seems to alleviate the
structural distress of piles. The computed maximum bending moment is 2 MNm com-
pared to 2.5 MNm for flat bedrock. In the case of inclined bedrock the liquefied soil
flows around the piles (lateral spreading) decreasing their distress at the cost, however, of
larger displacements. For flat bedrock, lateral spreading is not a consequence. The role of
soil is now two-fold. Not only it imposes load on the piles but also supports them. The
limited flow of the liquefied soil around the piles increases the distress on piles.
- The maximum kinematic bending moment (at interfaces between any two consecutive
soil layers of large stiffness contrast) is the minimum one for the second case compared to
that of the other two cases. Perhaps surprisingly, the extended soil improvement deterio-
rates the response giving a 0.9 MNm maximum kinematic bending moment compared to
0.6 MNm and 1.0 MNm for the second and first case, respectively. This is attributed to
that the improved soil forms an artificial interface of large stiffness contrast with the un-
derlying native soil layer, which gives rise to large seismicallyinduced strain and
thereby curvature on the pile.
- The effectiveness of the extended soil improvement is also reflected in Fig 14 for inclined
bedrock. Not only the residual (postliquefaction) settlement but also the maximum hori-
zontal displacement appears to be significantly smaller than those for the other two cases
(without and with limited soil improvement).
ACKNOWLEDGMENTS
The authors would like to express their gratitude to Engineers Stamatis and Kyriakos Statho-
poulos from DOMI, Panagiotis Panetsos and Helen Sakoubenta from Egnatia Odos, and
John Ritsos and Eftihis Keramidas from EDAFOMICHANIKI, for their fruitful cooperation dur-
ing the project.
373
REFERENCES AND BIBLIOGRAPHY
Berrill J., & Yasuda S., (2002), Liquefaction and Piled Foundations: Some Issues, Journal of Earth-
quake Engineering, Special Issue 1, Vol. 6, pp. 141.
Boulanger R. W., Kutter B. L., Brandenberg S. J., Singh P., Chang P. (2003) Pile foundations in lique-
fied and laterally spreading ground during earthquakes: Centrifuge experiments and analyses, Report
No. UCD/CGM-03/01, University of California at Davis
Byrne, P. (1991), A Cyclic Shear-Volume Coupling and Pore-Pressure Model for Sand, Proc. Second
International Conference on Recent Advances in Geotechnical Earthquake Engineering and Soil Dy-
namics, St. Louis, Missouri, Paper No. 1.24, pp. 47-55.
Byrne Peter M., Park Sung-Sink, Beaty Michael, Sharp Michael, Gonzalez Lenart and Abdoun Tarek
(2004) Numerical modeling of liquefaction and comparison with centrifuge tests, Proc. 13th World
Conference on Earthquake Engineering, Vancouver, B.C., paper 3387
Dobry R., & Abdoun T., (2001), Recent Studies on Seismic Centrifuge Modeling of Liquefaction and
its Effect on Deep Foundations, Proc. 4
th
International Conference on Recent Advances in Geotech-
nical Earthquake Engineering and Soil Dynamics (S. Prakash, ed.), San Diego, CA, pp. 2631.
Dobry R., & Abdoun T., (2001), PostTriggering Response of Liquefied Sand in the Free Field and
Near Foundations, Proc. 4
th
International Conference on Recent Advances in Geotechnical Earth-
quake Engineering and Soil Dynamics (S. Prakash, ed.), San Diego, CA.
Finn Liam W. D., Lee Kwok W., Martin Geoffrey R. (1977) An effectivestress model for liquefaction,
Journal of the Geotechnical Engineering Division, Vol. 103 NO.GT6, June
Garini E. (2005) Single pile response to liquefaction induced lateralspreading, Master thesis, Depart-
ment of Civil, Structural and Environmental Engineering, State University of New York at Buffalo
Ishihara K., & Cubrinovski M., (1998), Problems Associated with Liquefaction and Lateral Spreading
during Earthquakes, Soil Dynamics III, ASCE, Specialty Geotechnical Conference, Seattle, Vol. 1
pp. 301312
Japanese Geotechnical Society. (1996), Special Issue on Geotechnical Aspects of the January 17 1995
YyogokenNambu Earthquake, published by the Soils and Foundations.
Klimis N., Anastasiadis A., Gazetas G., Apostolou M. (2004) Liquefaction Risk Assessment and Design
of Pile Foundations for Highway Bridge, Proc. 13th World Conference on Earthquake Engineering,
Vancouver, Canada, paper No 2973.
Okamura M., Abdoun T.H., Dobry R., Sharp M.K., & Taboada V.M. (2001), Effects of Sand Perne-
ability and Weak Aftershock on Earthquake Induced Lateral Spreading , Soils & Foundations, Vol.
41, pp. 6378.
Papadimitriou A. G., Moutsopoulou M. E., Bouckovalas G. D. (2007), Numerical Analysis of Gravel
Drain Performance in Liquefiable Soils, Proceedings of the 2nd Japan-Greece Workshop on Seismic
Design, Observation, and Retrofit of Foundations, Tokyo, Japan, pp. 467-478.
Rollins K. M, Diehl N. B., Weaver T. J. (1998), Implications of V
s
BPT (N
1
)
60
Correlations for lique-
faction assessment in gravels, Soil Dynamics III, ASCE, Specialty Geotechnical Conference, Seattle
Vol. 1, pp. 506-517.
Seed H. B. and Idriss I.M. (1982), Ground Motions and soil Liquefaction during Earthquakes, Mono-
graph, Earthquake Engineering Research Institute, Berkeley, California.
Tasiopoulou P. (2007), SoilPile Interaction due to LiquefactionInduced Soil Flow, after Large Dis-
placement of QuayWall, Diploma Thesis, NTUA, Athens (in English).
Tazoh T., Ohtsuki A., Fuchimoto M., Nanjo A., et al (2000), Analysis of the Damage to the the Pile
Foundation of a Highway Bridge caused by Soil Liquefaction and its Lateral Spread Due to the 1995
Great Hanshin Earthquake, Proc. 12
th
World Conference Earthquake Engineering, New Zealand.
Tokimatsu K. & Asaka Y. (1998), Effects of LiquefactionInduced Ground Displacements on Pile Per-
formance in the 1995 HyogokenNambu Earthquake, Soils and Foundations, Special Issue, pp. 163-
177.

374
1 INTRODUCTION
In a large magnitude earthquake, structures on top of a fault outcrop may undergo significant
differential movements that could lead to failure. Seismic codes have thus prohibited construc-
tion in the immediate vicinity of seismically active faults. But for long facilities and structures
such as water channels, tunnels, pipelines, embankments, and long bridges crossing several geo-
logic formations, such a prohibition has often been impossible to respect. Moreover, past and
recent earthquakes have revealed that survival of structures on top of a fault is not impossible
even when fault displacements are large (of the order of meters).
In fact, the three 1999 earthquakes in Turkey (Kocaeli and Dzce-Bolu) and Taiwan (Chi-
Chi), provided many examples of satisfactory structural performance (Youd et al., 2000; Erdik,
2001; Bray, 2001; Ulusay et al., 2002), serving as an actual confirmation of the older belief that
structures can be designed against large tectonic displacements (Duncan & Lefebvre, 1973; Nic-
cum et al., 1976; Youd, 1989; Berrill, 1983).
Nevertheless, several bridges were damaged due to surface faulting in the 1999 Chi-Chi
earthquake (Kawashima, 2001; Pamuk et al., 2005). One such example is illustrated in Figure 1,
referring to a prestressed concrete bridge, the Bei-Fung viaduct in Fung-Yan City (photos
adapted from Hwang, 2000). The Chelungpu thrust fault crossed its south abutment with an up-
thrust of about 7 m. The result: collapse of two spans of the bridge due to differential displace-
ment between the piers.
Bridge failures, but also successes, were also reported after the 1999 Turkey earthquakes
(Ulusay et al., 2002; Pamuk et al., 2005). One such case is the failure of the 100 m Arifiye
Overpass, near Adapazari. Consisting of four simply supported pre-stressed concrete spans, it
was crossed by the 2 m offsetting fault: all spans fell off due to unseating.

Methodology for Design Analysis of Bridges against an Emerg-
ing Fault Rupture

I. Anastasopoulos, R. Kourkoulis, V. Drosos, T. Georgarakos, G. Gazetas

ABSTRACT: This paper presents a methodology for design of bridges against faulting. The
problem is decoupled in two steps, the first dealing with the response of a single pier and its
foundation, and the second analyzing a detailed model of the superstructure subjected to differ-
ential displacements of Step 1. A parametric study investigates typical models of viaduct and
overpass bridges, founded on piles or caissons. Fixedhead piled foundations are shown to be
rather vulnerable to tectonic deformation. Endbearing piles in particular are unable to survive
bedrock offsets exceeding few centimeters. Floating piles perform better, and if combined with
hinged piletocap connections, they could survive much larger offsets. Soil resilience is bene-
ficial in reducing pile distress. Caisson foundations are almost invariably successful. Statically
indeterminate superstructures are quite vulnerable, while staticallydeterminate are insensitive.
An actual application of the developed method is presented for a major bridge, demonstrating
the feasibility of design against tectonic deformation.

375
F
a
u
l
t

r
u
p
t
u
r
e
F
a
u
l
t

r
u
p
t
u
r
e

Figure 1. Collapse of two spans of the Bei-Fung bridge due to tectonic dislocation in the 1999 Chi-Chi
(Taiwan) earthquake.

Evidently, such failures are the result of a seismic design which did not systematically study
the consequences of fault imposed deformations. To bridge the apparent gap in our understand-
ing, recent research efforts combining field studies, centrifuge model testing, and numerical
modeling (Anastasopoulos & Gazetas 2007; Bransby et al. 2008 ; Faccioli et al. 2008 ; Anas-
tasopoulos et al. 2007; 2008) have culminated in the development of a validated methodology
for analysis and design of foundationstructure systems against surface fault rupture. It was
shown that foundation continuity and stiffness are critical for the survival of buildings. How-
ever, for bridges such continuity is meaningless: bridges are founded on separate supports.
While a building on a continuous and stiff foundation may convert the imposed dislocation to
rigid-body rotation without being substantially distressed, a bridge cannot avoid the differential
displacement between its supports (piers). As attested by the previously discussed case histories,
such differential displacement may cause structural failure or deck fall, depending on the type of
the superstructure.
The basic goal of this paper is to develop a fundamental methodology for bridge design
against large tectonic deformation. Since this work was part of a research project in Greece, em-
phasis is placed on normal faulting (the dominant mode in Greece).
2 PROBLEM DEFINITION AND ANALYSIS METHODOLOGY
The problem investigated herein and the employed analysis methodology is illustrated sche-
matically in Figure 2. The analysis of the bridgefoundation system subjected to faulting
induced deformation is conducted in two steps, in which the interaction between rupture, soil,
foundation, and superstructure is rationally taken into account. Specifically :
In Step 1 (local level), we analyze the response of a single bridge pier subjected to fault rup-
ture deformation. A detailed model is employed for the aforementioned fault rupture soil
foundationstructure interaction (FR-SFSI), with the superstructure modeled in a simplified
manner: the pier, of height H
p
and stiffness EI
p
, is included in the model; the bridge deck is re-
placed by equivalent lateral and rotational springs, K
x
and K
, respectively. For the case of a

continuous deck monolithically connected to piers, K
x
represents the axial stiffness of the deck
and K
the bending stiffness of the pier-deck connection. Correspondingly, for a seismically iso-
lated bridge, K
x
and K
represent the lateral and rotational stiffness of the (elastomeric) bearings.

The output of this step is dual: (i) it provides information regarding the distress of the founda-
tion system (e.g. the internal forces in piles, in case of a piled foundation) ; (ii) it provides the
necessary input for the second step: the horizontal and vertical displacements x and y and the
rotation at the base of the pier.
In Step 2 (global level), the detailed model of the superstructure is subjected to the computed
x , y , and from Step 1.
376
Fault offset
x
ff
y
ff
x
y
x, y, = 0
x, y, = 0
Detailed model of superstructure
x
,
H
h
H
Hanging wall
Detailed model
of foundation
Simplified model
of superstructure
p
footwall
x

H
p
y
Step 1
Step 2
Fault offset Fault offset
x
ff
y
ff
x
y
x, y, = 0
x, y, = 0
x
ff
y
ff
x
ff
y
ff
y
ff
x
y y
x, y, = 0
x, y, = 0
Detailed model of superstructure
x
,
H
h
HH
Hanging wall
Detailed model
of foundation
Simplified model
of superstructure
p
footwall
x

H
p
y
Step 1
Step 2

Figure 2. Problem definition and analysis methodology. The analysis of the soilstructure system sub-
jected to faulting-induced deformation is conducted in two steps. In Step 1, we analyze the response of a
single bridge pier subjected to fault rupture deformation. A detailed model is employed to model fault
rupture soilfoundationstructure interaction (FR-SFSI), with the superstructure modeled in a simplified
manner. In Step 2, the detailed model of the superstructure is subjected to the computed displacements
and rotations of Step 1.
377
350 m
70 m
50 m
(a)
(b)
(c)
d
= 100 GNm
2
P1
P2
P3
P4
P5
P6
A1 A2
P1
P2
P3
P4
P5
P6
A1 A2
P1
P2
P3
P4
P5
P6 A1 A2
T = 3.0 sec
T = 3.0 sec
T = 1.0 sec
m
d
= 5000 Mg
d
= 100 GNm
2
m
d
= 5000 Mg
d
= 100 GNm
2
m
d
= 5000 Mg
350 m
70 m
50 m
(a)
(b)
(c)
d
= 100 GNm
2
P1
P2
P3
P4
P5
P6
A1 A2
P1
P2
P3
P4
P5
P6
A1 A2
P1
P2
P3
P4
P5
P6 A1 A2
T = 3.0 sec
T = 3.0 sec
T = 1.0 sec
m
d
= 5000 Mg
d
= 100 GNm
2
m
d
= 5000 Mg
d
= 100 GNm
2
m
d
= 5000 Mg
After a worldwide overview, bridge systems were categorized according to their geometric
characteristics, the typology of their superstructure, and their foundation. Based on this, and
aiming to render the results of this research as general as possible, two generalized bridge types
were selected : (i) a typical 350 m long viaduct bridge (Figure 3), and (ii) a typical 75 m long 3-
span overpass bridge (Figure 4). For each bridge type, alternative superstructure typologies
were investigated.
As shown in Figure 3, five different alternatives were investigated for the typical viaduct:
(a) a 7-span viaduct, with continuous deck monolithically connected to piers;
(b) the same system, but the deck supported through elastomeric bearings;
(c) 7 simply supported decks on elastomeric bearings;
The deck is a box section of sectional stiffness EI
d
= 100 GNm
2
, adequate for standard con-
struction of the 50 m spans. The sectional stiffness EI
p
of the piers varies with their height H
p
,
so that the member stiffness of each pier K
p
is kept constant : a commonly used rule in practice.
The value of K
p
was computed based on the mass of the deck m
d
, so that the dominant period T
of each system corresponds to realistic values : T = 1.0 sec for alternative (a) ; T = 1.5 sec for
alternative (b). For the seismically isolated alternatives (b and c), the stiffness of the elastomeric
bearings was computed so that T = 3.0 sec.
Two alternatives were selected for the typical overpass bridge (Figure 4) :
(a) a three-span continuous deck monolithically connected to piers; and
(b) three simply supported decks on elastomeric bearings.
As in the previous case, the two alternatives were selected to correspond to realistic bridges.
Since the span is smaller (25 m), a smaller box section of EI
d
= 60 GNm2 was selected. The
cross-sectional stiffness of piers was set to EI
p
= 20 GNm
2
, so that (for m
d
= 800 Mg) the non-
isolated system would yield T = 0.5 sec in the longitudinal direction (a typical value). For both
alternatives, the stiffness of the elastomeric bearings was selected so as to achieve T = 2.5 sec.

Figure 3. Parametrically investigated 7-span viaduct bridges : (a) continuous deck monolithically con-
nected to piers ; (b) same system, but deck supported through elastomeric bearings ; (c) 7 simply sup-
ported spans on elastomeric bearings.
378
8 m
75 m
25 m
(a)
(b)
d
= 60 GNm
2
m
d
= 800 Mg
T = 2.5 sec
T = 2.5 sec
p
= 20 GNm
2
m
d
= 800 Mg
d
= 60 GNm
2
m
d
= 800 Mg
8 m
75 m
25 m
(a)
(b)
d
= 60 GNm
2
m
d
= 800 Mg
T = 2.5 sec
T = 2.5 sec
p
= 20 GNm
2
m
d
= 800 Mg
d
= 60 GNm
2
m
d
= 800 Mg

Figure 4. Parametrically investigated typical 3-span overpass bridges : (a) continuous deck supported on
elastomeric bearings ; (b) 3 simply supported spans on elastomeric bearings.

As depicted in Figures 3 and 4, for each bridge type different scenarios were investigated
with respect to the location of fault outcropping : the first set (in grey) assumes fault rupture
emergence between two consecutive piers; the second (in black) refers to the case of the disloca-
tion taking place at the location of a pier. In the first case, the input to the detailed bridge model
(of Step 2) only includes the displacements x and y, and the interaction analysis of Step 1 is
redundant. In the second case (rupturing at the location of a pier), the FRSFSI analysis is man-
datory to compute x and y (which are affected substantially by the presence of the pier foun-
dation), and the input to the Step 2 model also includes the rotation at the base of the pier,
which is equally (if not even more) important to x and y, especially for the case of tall piers.
At the foundation level, an adequate number of local rupture location scenarios were parametri-
cally investigated. For the input to the subsequent bridge superstructure analysis (Step 2), the
worst-case local rupture location scenario was employed.
Since the response of the bridge system largely depends on the response of its foundation to
the imposed tectonic dislocation, a number of typical foundation systems were parametrically
investigated in the first analysis step. As for the superstructure, realistic foundation systems
were selected for analysis, corresponding to the parametrically analyzed bridge types (Figures 3
and 4). The idealized foundation types of Figure 5 were selected for analysis :
(i) a small 2 4, d = 1.0 m, L = 15 m pile group, suitable for the overpass bridge ;
(ii) a large 3 3, d = 1.5 m, L = 15 m pile group, suitable for the 7-span viaduct bridge ;
(iii) a small 5 m x 5 m x 10 m caisson foundation, suitable for the 7-span viaduct) ; and
(iv) a large 10 m 10 m 15 m caisson foundation, also suitable for the 7-span viaduct.
In all cases, different scenarios were investigated with respect to the soil conditions, with
Layer 1 ranging from idealized loose to dense sand (Anastasopoulos et al., 2007), and Layer 2
(for the piled foundations) ranging from dense sand to rock-type material. Thus, both floating
(with the material of Layer 2 being the same with Layer 1) and end-bearing piles (with the ma-
terial of Layer 2 being substantially stiffer than Layer 1) were investigated.

379
s
d = 1.0 m L = 15 m
H = 18 m
11 m
5 m
3d
3d
h
Layer 1
Layer 2
4d
Layer 2
s
d = 1.5 m L = 15 m
H = 20 m
h
Layer 1
3d
14 m
14 m
2.5d
2.5d
s
H = 20 m
5 m
5 m
h
Layer 1
10 m
s
H = 20 m
10 m
10 m
h
Layer 1
15 m
(a) (b)
(c) (d)
s
d = 1.0 m L = 15 m
H = 18 m
11 m
5 m
3d
3d
h
Layer 1
Layer 2
4d
Layer 2
s
d = 1.5 m L = 15 m
H = 20 m
h
Layer 1
3d
14 m
14 m
2.5d
2.5d
s
H = 20 m
5 m
5 m
5 m
5 m
h
Layer 1
10 m
s
H = 20 m
10 m
10 m
h
Layer 1
15 m
(a) (b)
(c) (d)

Figure 5. Parametrically investigated foundation systems : (a) 2 x 4, d = 1.0 m, L = 15 m pile group (suit-
able for the overpass bridge) ; (b) 3 x 3, d = 1.5 m, L = 15 m pile group (suitable for the 7-span viaduct) ;
(c) small 5 m x 5 m x 10 m caisson foundation (suitable for the 7-span viaduct) ; and (d) large 10 m x 10
m x 15 m caisson foundation (also suitable for the 7-span viaduct, for worse soil conditions).

The following section discusses the finite element (FE) analysis method employed for
FRSFSI analysis at the local pier-foundation level (Step 1). Then, the key findings concerning
the response of piled and caisson foundations are discussed, followed by the main results of the
global analysis of the superstructure (Step 2).
3 FINITE ELEMENT MODELLING
The analysis is conducted in 3-D, in order to model the response of piled and caisson founda-
tions realistically. Utilizing the FE code ABAQUS, the soil is modeled with hexahedral (8-
node) brick-type elements of dimension d
FE
= 1 m to achieve a reasonably refined mesh, as
documented in detail in Anastasopoulos et al. (2007). In the case of caisson foundations, the
mesh is made sparser far from the area of interest. The caisson is also modeled with brick-type
elements, assumed linear elastic : E = 25 GPa (concrete). Following the results of an initial sen-
sitivity study, the total width of the model was set to B = 3H. Although this is less than the B =
4H recommendation of Bray (1990; 1994a; 1994b), the sensitivity analysis showed that the re-
sults of interest are hardly affected, while the computational time was reduced substantially.
Hence, it was accepted as a reasonable compromise.
380
The superstructure is taken into account, as described previously : the pier is modeled with
beam elements ; the deck with appropriate grounded springs. In all cases, half of the foundation
system is analyzed, taking advantage of problem symmetry.
In the case of piled foundations, the mesh is refined further close at the area of the piles
(d
FE
0.25 m) to capture their geometry. Dummy (i.e. of zero mass and stiffness) brick-type
elements are used to model the geometry of the piles. The piles are actually modeled with beam
elements, rigidly connected to the peripheral nodes of the corresponding dummy elements. This
way, soil-to-pile interaction is modeled realistically : contact is attained on the actual periphery
of the pile and the actual pile tip area. The piles are connected to a rigid pile cap, which is mod-
eled with hexahedral brick-type elements. Both the piles and the pile cap are assumed liner elas-
tic, with E = 25 GPa (concrete).
The analysis is performed in two steps. First, fault rupture propagation through soil is ana-
lyzed in the free field, ignoring the presence of the foundation (pile group or caisson). Then,
knowing the location of fault rupture emergence, the foundation is positioned so that the unper-
turbed rupture outcrops at distance s from its hanging wall (left) edge. In the case of the small
2 4 pile group, five scenarios were parametrically investigated with respect to the location of
fault outcropping : s = 2, 5, 7, 8, 9, 10, and 11 m. Observe that although the width of the pile-
cap is 5 m, due to the non-vertical propagation of the rupture larger values of s are required to
cross the pile group at all possible locations. Similarly, for the large 3 3 pile group, s = 3, 7,
11, 15, 16, and 20 m ; for the small caisson, s = 2, 4, and 8 m ; and for the large caisson,
s = 1, 5, 9, 13, 14, and 18 m.
4 SOIL CONSTITUTIVE MODELLING AND CALIBRATION
Soil behavior is modeled with an elastoplastic constitutive model with a Mohr-Coulomb failure
criterion and isotropic strain softening, encoded in ABAQUS through a user subroutine. Strain
softening is introduced by reducing the mobilised friction angle
mob
and the mobilised dilation
angle
mob
with the increase of plastic octahedral shear strain. Soil behaviour before yielding is
modeled as linear elastic. Model parameters are calibrated through direct shear test results, and
an approximate scaling method is employed to take account of scale effects. Two idealized soil
materials and a simplified rock-type material are utilized in the analysis :
Loose Sand :
p
= 32
,
res
= 30
,
p
= 3
, and
y
= 0.030.
Dense Sand :
p
= 45
,
res
= 30
,
p
= 18
, and
y
= 0.015.
Rock-type material :
p
= 37
,
res
= 25
,
p
= 15
, and
y
= 0.002.
where :
y
is a parameter associated with the initial elastic response of the soil material, the
plastic shear strain at peak conditions,
res
the residual value of the friction angle, and
p
the ul-
timate dilation angle.
The FE modeling methodology employed herein has been extensively validated through
qualitative comparisons with numerous published experimental data (Cole & Lade, 1984) and
earlier case histories (Slemmons, 1957; Brune & Allen, 1967; Taylor et al., 1985), semi-
quantitative comparisons with case histories from the 1999 earthquakes of Kocaeli and Turkey
(Anastasopoulos & Gazetas, 2007; Faccioli et al., 2008), and most importantly through
quantitative blind predictions of centrifuge model tests (Anastasopoulos et al. 2008).
5 PERFORMANCE OF PILED FOUNDATIONS
Although piled foundations are used to protect the superstructure by minimizing the settlements
and the dynamic (shaking-induced) rotations, their performance to concentrated deformation is
not always beneficial. Evidence from recent earthquakes has implicated the piles for the ob-
served structural damage. The previously discussed failure of several piles of the Bolu Viaduct
in the second 1999 Turkey earthquake is definitely one such case. Another such example is the
damage of the pile-supported Attatrk Stadium in Denizerler during the earlier 1999 Kocaeli
earthquake.
This section discusses the key findings of the parametric analysis at the local foundation level
for the case of piled foundations. The detailed presentation of all analysis results is out of the
381
(a) (b)
(d) (c)
s = 5 m s = 7 m
s = 8 m s = 10 m
Free field rupture
emergence
Free field rupture
emergence
Free field rupture
emergence
Free field rupture
emergence
R
1
R
2
(a) (b)
(d) (c)
s = 5 m s = 7 m
s = 8 m s = 10 m
Free field rupture
emergence
Free field rupture
emergence
Free field rupture
emergence
Free field rupture
emergence
R
1
R
2
scope of this paper. Hence, we focus on characteristic results that provide insight to the govern-
ing interaction mechanisms.

5.1 Small 2 4 pile group
We first analyze the response of the small 2 4 pile group, which is adequate for the typical
overpass bridge. Layer 1 is assumed to be the idealized dense sand and Layer 2 the idealized
rock-type material the case of end-bearing piles. The vertical superstructure load transmitted
onto the group is equal to V = 2500 kN, typical of an overpass bridge. The role of the
superstructure is modeled in a simplified manner, as illustrated in Figure 2.
Figure 6 portrays the response of the pile group subjected to h = 0.05 m normal faulting at
distance s = 5, 7, 8, and 10 m, in the form of FE deformed mesh with superimposed plastic
strain. The selection of such a small imposed bedrock dislocation is deliberate: to demonstrate
clearly the sensitivity of pile foundations to this type of loading. As seen in Figure 6a, for s = 5
m (i.e. the unperturbed fault rupture would outcrop near the right edge of the pilecap), the pile
group forces the dislocation to divert towards the hanging wall (to the left side). As a result,
the foundation is not subjected to substantial deformation: bending moments do not exceed 600
kNm (with a heavy reinforcement ratio, the ultimate design capacity M
ult
of the d = 1.0 m
piles could reach 3000 kNm), but the pile group remains practically intact. Observe also that the
pilecap is not subjected to any measurable displacement or rotation.

Figure 6. FE deformed mesh with superimposed plastic strain contours for the 2 x 4 pile group, subjected
to h = 0.05 m normal faulting (deformation scale factor = 40) : (a) s = 5 m, (b) s = 7 m, (c) s = 8 m, and
(d) s = 10 m.

382
Moving the fault rupture at s = 7 m (i.e. in the free-field, the fault would have emerged 2 m to
the right of the footwall edge of the pilecap), the response of the group dramatically worsens
(Figure 6b). Now, a rather distinct bifurcation of the dislocation takes place, leading to devel-
opment of two separate ruptures: the first one, R1, is diverted by the first row of piles towards
the hanging wall (left); the second, R2, outcrops between the two rows of piles. As a result, the
front row of piles (left) is being pulled outwards (to the left) and downwards by the moving
hanging wall, while the back row (right) cannot follow as it lies on the footwall. This imposed
differential displacement of the two pile rows, in combination with the kinematic restraints of
the pilecap, leads to development of rather large bending moments of the order of 1500 kNm.
Furthermore, pilecap displacement and rotation is now discernible.
The distress of the pile group is largest for s = 8 m (Figure 6c). In contrast to the previous
case, the dislocation now just misses the tip of the front pile row. Paradoxically, this small de-
tail worsens the response of the soil-foundation system to a rather large extent. While in the
previous case (s = 7 m), the rupture experienced bifurcation and diffusion as it interacted with
first pile row, now such stress-relieving phenomena cannot develop: the rupture is left free to
develop to its full extent between the two pile rows. Hence, the two rows suffer the largest dif-
ferential displacement, which leads to bending moments M of the order of 2800 kNm (i.e. M
almost reaches M
ult
) and measurable displacement and rotation of the pilecap. It would be inter-
esting to think of this case in reality: while for an observer at the ground surface the rupture
would appear to have missed the foundation, due to its non-vertical propagation path it would
have intersected with the two pile rows inflicting substantial pile distress and rotation of the su-
perstructure.
Moving the fault rupture to s = 10 m (Figure 6d) leads to less stressing of the piles. Now, the
propagating rupture plane intersects the tip of the back row of piles, undergoes substantial diffu-
sion and limited local bifurcation. Both pile rows are lying on the hanging wall, being subjected
to almost the same horizontal and vertical displacements. As a result, they are not subjected to
substantial differential displacements, and consequently the tectonically-induced M does not ex-
ceed a mere 300 kNm. The rotation of the pilecap, which is directly related to the differential
displacement between the pile rows, is insignificant. In stark contrast, since the whole pile
group is moving along with the hanging wall, horizontal and vertical displacements at the pier
base are substantial.
Figure 7 synopsizes our results, giving emphasis to the effect of the location s of fault out-
cropping to: (a) the horizontal x and vertical y displacement at the pier base, (b) the rotation
at the pier base, (c) the maximum and minimum bending moments M of the piles, and (d) the
maximum and minimum axial forces N of the piles. One would have expected that x and y
are in general increasing with s : as the location of the rupture moves to the right, the pile group
tends to be more on the hanging wall, being subjected to larger displacements. However, x and
y at the pier base are also related to the rotation of the pilecap. The latter is directly related to
the differential displacement between the two rows of piles, and is thus maximum for s = 8 m.
As a result, y at the pier base is also largest for the same rupture location. On the other hand,
x is not affected to the same extent by , being maximum for s = 10 m. Being the direct result
of the differential displacement between the two pile rows, the stressing of the piles (exhibited
through M and N) is also largest for s = 8 m (i.e. when the rupture outcrops exactly between the
two rows).
In summary, it has been shown that a rather minor bedrock offset (h = 0.05 m in the case ex-
amined herein) is enough for typical end-bearing piles to reach their ultimate structural capacity,
even in case of very heavily reinforced piles.

5.2 Large 3 3 pile group
This section deals with the large 33 pile group, which would be a reasonable solution for the
7-span viaduct. Since the response of end-bearing piles has already been shown to be rather
problematic, we now focus on the response of floating piles (i.e. Layer 2 being the same with
Layer 1). The soil is assumed to be the idealized dense or loose sand. The vertical superstructure
load transmitted onto the pile group is assumed equal to 7000 kN, typical for the 7-span viaduct.

383
-8
-6
-4
-2
0
x
y
-6000
-4000
-2000
0
2000
4000
4 6 8 10 12
-3000
-2000
-1000
0
1000
0
0.2
0.4
0.6
0.8
4 6 8 10 12
(
d
e
g
)
(
c
m
)
M

(
k
N
m
)
N

(
k
N
)
s (m) s (m)
max
min
tension
compressi on
(a)
(b)
(c)
(d)
-8
-6
-4
-2
0
x
y
-6000
-4000
-2000
0
2000
4000
4 6 8 10 12
-3000
-2000
-1000
0
1000
0
0.2
0.4
0.6
0.8
4 6 8 10 12
(
d
e
g
)
(
c
m
)
M

(
k
N
m
)
N

(
k
N
)
s (m) s (m)
max
min
tension
compressi on
(a)
(b)
(c)
(d)

Figure 7. Summary of analysis results for the 2 x 4 pile group, h = 0.05 m. Effect of the location s of fault
outcropping to : (a) horizontal x and vertical displacement y at the pier base ; (b) rotation at the pier
base ; (c) maximum and minimum bending moment M of the piles ; and (d) maximum and minimum axial
force N of the piles.

To illustrate the effect of soil resilience, we compare the response of the pile group in dense
and loose sand. Figure 8 depicts the evolution of M with the increase of imposed bedrock offset
h for normal faulting at distance s = 11 m (representing the worst-case scenario). In dense sand,
all piles are subjected to substantial bending, with the hanging wall side piles (pile 1) being
stressed the most and the footwall side piles (pile 3) the least. Soil resilience is clearly beneficial
in terms of pile stressing : in loose sand the maximum M is a little more than merely one third of
that of dense sand. Naturally, such large bending moments would exceed the ultimate capacity,
M
ult
, of the d = 1.5 m piles: with a very heavy reinforcement ratio of the order of 4%, M
ult

would be of the order of 8000 kNm. This means that in dense and stiff soil the hanging wall side
piles (pile 1) would be the first to fail, at h = 0.35 m, followed by the medium row (pile 2), at h
= 0.52 m, and finally the footwall side row (pile 3), at h = 1.61 m. In stark contrast, in loose
sand only the first row (pile 1) would fail, and for substantially larger imposed deformation: h =
0.93 m. Note also that while at the early stages of deformation (h < 0.3 m) the stressing of the
three pile rows is qualitatively similar to the case of dense sand (pile 1 is stressed the most; pile
3 the less), the increase of the imposed deformation leads to a mechanism change : the footwall
side piles (3) experience more stressing than the medium row (2).
The demonstrated beneficial role of soil resilience is triple :
(a) In terms of quasi-elastic behavior : Before the soil surrounding the piles starts to yield, the
decrease of soil stiffness leads to an increase of the relative pile stiffness, facilitating pile resis-
tance to the imposed deformation ;
(b) In terms of plastic behavior : The decrease of soil strength (
p
= 45
in dense sand ;
p
=
32
in loose sand) speeds up soil failure at the pile-soil interface, allowing the piles to sustain
larger imposed deformation before reaching structural failure. Stress relieving phenomena, such
as fault rupture diversion and bifurcation, and diffusion of plastic deformation are also facili-
tated; and
(c) In terms of post failure behavior : After the soil at the pile-soil interface has reached fail-
ure, the decrease of soil dilatancy (
p
= 18
in dense sand;
p
= 3

in loose sand) leads to a sub-
stantial decline of the rate of increase of pile stressing with the imposed deformation h. Observe
in Figure 8 that while in dense sand the evolution of pile stressing with h exhibits a hardening-
like behavior, in loose sand it resembles an elastic-perfectly plastic one.
384
0 0.5 1 1.5 2
0
5000
10000
15000
20000
25000
30000
0 0.5 1 1.5 2
Dense sand
Loose sand
Pile 1
Pile 3
Pile 5
Pile 5
Pile 1
Pile 3
1 3 5
0
3
6
9
12
15
0 5000 10000 15000 20000 25000 30000
h = 0.1 m
h = 0.8 m
h = 2.0 m
0 5000 10000 15000 20000 25000 30000
h = 0.1 m
h = 0.8 m
h = 2.0 m
Dense sand Loose sand
(
k
N
m
)
D
e
p
t
h

(
m
)
M (kNm) M (kNm)
h (m) h (m)
Pile 1 Pile 1
0.35 0.52 0.93 1.61
0 0.5 1 1.5 2
0
5000
10000
15000
20000
25000
30000
0 0.5 1 1.5 2
Dense sand
Loose sand
Pile 1
Pile 3
Pile 5
Pile 5
Pile 1
Pile 3
1 3 5
0
3
6
9
12
15
0 5000 10000 15000 20000 25000 30000
h = 0.1 m
h = 0.8 m
h = 2.0 m
0 5000 10000 15000 20000 25000 30000
h = 0.1 m
h = 0.8 m
h = 2.0 m
(
k
N
m
)
D
e
p
t
h

(
m
)
M (kNm) M (kNm)
h (m) h (m)
Pile 1 Pile 1
0.35 0.52 0.93 1.61

Figure 8. The beneficial role of soil resilience for the large 3 x 3 pile group, subjected to normal faulting
at s = 11 m : evolution of pile bending moments M with the increase of imposed bedrock offset h (ranging
from 0.1 to 2.0 m) for idealized dense and loose sand.

All three reasons are also responsible for the superior performance of floating piles (this
group) compared to end bearing piles (24 pile group) : while only 5 cm of bedrock offset were
enough for all of the d = 1.0 m end-bearing piles to reach structural failure, in dense sand the
first row of the d = 1.5 m piles can sustain 35 cm before reaching failure ; the other two rows
have even larger safety margins (0.52 m and 1.61 m). Analyses (not shown herein) of the 33
pile group with Layer 2 being changed to the idealized rock-type material confirm this conclu-
sion. In fact, due to the disproportional increase of pile stiffness compared to M
ult
(increasing d
from 1.0 m to 1.5 m leads to a 500% stiffness increase, compared to a 260% increase of M
ult
),
the d = 1.5 m piles have even smaller safety margins if they are of the end-bearing type.

5.3 The Solution : hinged pile-to-cap connection
In all cases examined, the largest pile bending moments occur at the connection with the pile-
cap. This location is therefore the first candidate for plastic hinging. Preventing such failure by
introducing a-priori a hinged piletocap connection is rather intuitive. Furthermore, dynamic
analyses (strong ground shaking, not faulting related) of pile groups have shown that the type of
pile-to-cap connection greatly influences the performance of the foundation, with the hinged
connection leading to substantially less pile distress, albeit at the cost of larger cap displace-
ments and rotations compared to the traditional fixed connection (Tazoh et al., 2002; Gerolymos
et al., 2008). The idea of a devise that allows such a connection was introduced in Japan after
the devastating 1995 Kobe earthquake. One such devise, designed and manufactured by Shi-
mizu Co. & Kubota Co., is schematically illustrated in Figure 9 (after Tazoh et al., 2002).
385
Pile
Pilecap
Pile
Pilecap
0
1000
2000
3000
4000
5000
0 0.5 1 1.5 2
(
k
N
m
)
h (m)
(a)
Pile 1
Pile 3
Pile 5
1 3 5
0
3
6
9
12
15
-2000 -1000 0 1000 2000 3000 4000 5000
Pile 1
Pile 5
D
e
p
t
h

(
m
)
M (kNm)
(b)
0
1000
2000
3000
4000
5000
0 0.5 1 1.5 2
(
k
N
m
)
h (m)
(a)
Pile 1
Pile 3
Pile 5
1 3 5
0
3
6
9
12
15
-2000 -1000 0 1000 2000 3000 4000 5000
Pile 1
Pile 5
D
e
p
t
h

(
m
)
M (kNm)
(b)

Figure 9. Schematic of a hinged pile to pilecap connection, following the concept introduced by Shimizu
Co. & Kubota Co. (after Tazoh et al., 2002).

The cast-iron devise consists of a spherical table, a spherical lid, and a high-strength bolt at
the center to allow transmission of tensile forces. Shear forces are transmitted through the high-
friction contact surface (tablelid), while almost free rotation is allowed. The performance of
this devise has been validated through real scale testing, and it has been applied in practice.
To qualitatively verify the effectiveness of such a solution for the problem investigated
herein, the same pile groups were re-analyzed with hinged pile-to-cap connections. Typical re-
sults are shown in Figure 10 for the case of the 3 x 3 pile group subjected to normal faulting at
s = 9 m through idealized dense sand. Evidently, due to activation of the rotational degree of
freedom the distress of all piles is drastically reduced. Even for h = 2 m (a rather large offset,
typical of M > 7 earthquakes), M does not exceed 5000 kNm in any of the piles (Figure 10a)
substantially lower than the largest possible ultimate capacity (M
ult
= 8000 kNm) of the d = 1.5
m piles. As depicted in Figure 10b, due to the hinged piletocap connection the piles are now
behaving like simply supported vertical beams, with M being maximum at almost the mid-
height in the case of the hanging wall side row (pile 1), and at a shallower depth at the footwall
side row (pile 5). The difference is clearly due to the more intense soil plastification at the front
row of piles, which is more directly affected by the faulting-induced deformation.

Figure 10. The 3 x 3 pile group equipped with hinged piletopilecap connections, subjected to normal
faulting at s = 11 m through idealized dense sand : (a) evolution of pile bending moments M with the in-
crease of imposed bedrock offset h ; (b) distribution of pile bending moments with depth.
386
(a) (b)
(c) (d)
s = 1 m s = 5 m
s = 9 m s = 13 m
Free field rupture
emergence
Free field rupture
emergence
Free field rupture
emergence
Free field rupture
emergence
(a) (b)
(c) (d)
s = 1 m s = 5 m
s = 9 m s = 13 m
Free field rupture
emergence
Free field rupture
emergence
Free field rupture
emergence
Free field rupture
emergence
6 PERFORMANCE OF CAISSON FOUNDATIONS
The seismic performance of caisson foundations is in general considered advantageous (e.g.
Gerolymos & Gazetas, 2006), especially when subjected to large imposed deformation. The
Kobe Ohashi (Port Island) and the Nishinomiya-Ko bridge in Kobe (Japan) partially owe their
survival in the 1995 earthquake to their massive caisson foundations, which intercepted a sub-
stantial portion of the liquefactioninduced lateral spreading (Hanlong et al., 1997; Anas-
tasopoulos et al., 2001). In terms of faultinginduced deformation, the Banco Central de Nica-
ragua constitutes one of the earliest (and one of very few) such case histories. When the strike-
slip fault rupture of the 1972 Ms 6.3 Managua earthquake attempted to cross the Bank, thanks
to the existence of a rigid reinforced-concrete caisson (the Banks underground vault), it was di-
verted leaving the building totally unscathed (Niccum et al., 1976).
In this section, we briefly discuss characteristic results of the parametric analysis that was
conducted. We focus on the large 10 x 10 x 15 m caisson. Since caissons are commonly used
as floating foundations, the soil is assumed homogenous, consisting of dense or loose sand. The
caisson is assumed fully bonded to the bearing soil a rather conservative idealization.
Figure 11 illustrates the response of the caisson in idealized dense sand subjected to h = 2 m
normal faulting at s = 1, 5, 9, and 13 m, in terms of FE deformed mesh with superimposed plas-
tic strain. As depicted in Figure 11a, for s = 1 m, the caisson diverts the dislocation towards the
hanging wall (to the left), forming a distinct scarp at its left edge. Similarly to the case of the
piled foundations, the caisson does not experience any measurable rotation or displacement.
Moving the rupture to the middle of the foundation, s = 5 m (Figure 11b), leads to a more in-
tense diversion of the rupture path: the fault now emerges vertically along the sidewall of the
caisson. The latter experiences a rotation = 1
o
and measurable vertical and horizontal dis-
placements : y = 0.065 m and x = 0.28 m. Note also the formation of a secondary antithetic
rupture zone, which starts propagating to the left of the main rupture at a dip angle of about 60
o
.
Reaching the surface, in combination with the main rupture (diverted to the left edge of the
caisson), it generates a gravity graben : a feature purely related to the kinematic constraints
imposed by the rigid caisson.

Figure 11. FE deformed mesh with superimposed plastic strain contours for the large 10 x 10 x 15 m cais-
son, subjected to h = 2 m normal faulting (deformation scale factor = 1) through idealized dense sand :
(a) s = 1 m, (b) s = 5 m, (c) s = 9 m, and (d) s = 13 m.
387
(
d
e
g
)
(
m
)
s (m)
-3
-2.5
-2
-1.5
-1
-0.5
0
y
x
0
2
4
6
8
10
0 2 4 6 8 10 12 14
(
d
e
g
)
(
m
)
s (m)
-3
-2.5
-2
-1.5
-1
-0.5
0
y
x
0
2
4
6
8
10
0 2 4 6 8 10 12 14
(
d
e
g
)
(
m
)
s (m)
-3
-2.5
-2
-1.5
-1
-0.5
0
y
x
0
2
4
6
8
10
0 2 4 6 8 10 12 14
(
d
e
g
)
(
m
)
s (m)
-3
-2.5
-2
-1.5
-1
-0.5
0
y
x
0
2
4
6
8
10
0 2 4 6 8 10 12 14
For s = 9 m (Figure 11c), although the imposed deformation is diffused substantially, the
caisson is subjected to rather intense rotation = 8
o
and substantial vertical and horizontal dis-
placements: y = 0.69 m and x = 2.19 m. Coulomb-type active conditions are likely to form at
the back (footwall side) of the caisson. Finally, for s = 13 m (Figure 11d), the rupture path just
intersects with the base corner of the caisson, being defracted towards the footwall (to the
right), and finally emerging at the ground surface 8 m to the right of the footwall edge of the
caisson: i.e., about 5 m to the right of its free-field outcrop location. The caisson essentially fol-
lows the movement of the hanging wall, experiencing an appreciable rotation = 3
o
, combined
with displacements y x 1.65 m.
Figure 12 summarizes the results of such analysis, giving emphasis to the effect of the loca-
tion s of fault outcropping and soil resilience. As for the piled foundations, x and y are in
general increasing with s (as the fault rupture moves to the right, the caisson tends to be more on
the hanging wall, thereby suffering larger displacements). However, x and y at the pier base
are also related to , which is largest for s = 9 m. As a result, x is also maximum for the same
rupture location. In contrast, y is not affected to the same extent by , being maximum for s =
13 m. Note that this is opposite to what was observed for the pile group, where x was the one
unaffected by . This is attributable to the different load transmission mechanisms of the two
systems. While in the case of piled foundations the rotation of the pilecap is imposed through
the piles, atop the caisson is solely due to the imposed differential displacement at its base. In
stark contrast to the flexible pile group, the rigid caisson converts all of its base rotation to hori-
zontal displacement at the top (i.e. at the pier base).
Finally, soil resilience is detrimental for the displacements (x and y) and the rotation () at
the pier base. The same conclusion is also valid for piled foundations, although our emphasis
was on pile distress, for which exactly the opposite was shown (i.e. soil resilience to be benefi-
cial).

Figure 12. Synopsis of analysis results for the 10 x 10 x 15 caisson, h = 2 m. Effect of the location s of
fault outcropping and soil resilience to the horizontal x and vertical displacement y, and the rotation at
the base of the pier.

388
-300
-150
0
150
300
-60
-50
-40
-30
-20
-10
0
-600 0 600 -600 0 600 -600 0 600 -600 0 600 -600 0 600 -600 0 600
M (MNm)
y

(
m
)
P1 P2 P3 P4 P5 P6
P1
P2
P3 P4
P5
P6
M

(
M
N
m
)
-300
-150
0
150
300
-2
-1.5
-1
-0.5
0
P1
P2
P3 P4
P5
P6
y
(
m
)
-2
-1.5
-1
-0.5
0
P1
P2
P3 P4
P5
P6
P1
P2
P3 P4
P5
P6
y
(
m
)
M

(
M
N
m
)
-60
-50
-40
-30
-20
-10
0
-600 0 600 -600 0 600 -600 0 600 -600 0 600 -600 0 600 -600 0 600
M (MNm)
y

(
m
)
P1 P2 P3 P4 P5 P6
(a)
(b)
h = 0
h = 2.0 m
h = 0
h = 2.0 m
-300
-150
0
150
300
-60
-50
-40
-30
-20
-10
0
-600 0 600 -600 0 600 -600 0 600 -600 0 600 -600 0 600 -600 0 600
M (MNm)
y

(
m
)
P1 P2 P3 P4 P5 P6
P1
P2
P3 P4
P5
P6
M

(
M
N
m
)
-300
-150
0
150
300
-2
-1.5
-1
-0.5
0
P1
P2
P3 P4
P5
P6
y
(
m
)
-2
-1.5
-1
-0.5
0
P1
P2
P3 P4
P5
P6
P1
P2
P3 P4
P5
P6
y
(
m
)
M

(
M
N
m
)
-60
-50
-40
-30
-20
-10
0
-600 0 600 -600 0 600 -600 0 600 -600 0 600 -600 0 600 -600 0 600
M (MNm)
y

(
m
)
P1 P2 P3 P4 P5 P6
(a)
(b)
h = 0
h = 2.0 m
h = 0
h = 2.0 m
h = 0
h = 2.0 m
h = 0
h = 2.0 m
7 FAULTING-INDUCED DISTRESS OF THE SUPERSTRUCTURE
Using the output of the first step (x, y, and at the base of the pier), in Step 2 (global level)
we analyze the response of the detailed model of the superstructure. We focus the typical 7-span
viaduct, which is of greater importance, and compare the two extreme cases : (a) continuous
deck monolithically connected to piers, and (c) 7 simply supported spans on elastomeric (seis-
mic isolation) bearings.
The comparison is portrayed in Figure 13 (h = 2 m normal faulting at x = 150 m, i.e. under
pier P3), in terms of deck vertical displacements y and bending moments M along deck and
piers. In the first case (continuous deck monolithically connected to piers), the imposed tectonic
deformation generates large stressing (Figure 13a). The tectonically-induced deck bending mo-
ments (black line) are an order of magnitude larger than their static (h = 0) values (in grey).
Such stressing could not possibly be undertaken by any reasonable pre-stressed concrete box
section leading to failure. The stressing of the piers is also unacceptably intense.

Figure 13. 7-span viaduct subjected to h = 2 m normal faulting at x = 150 m (i.e. at pier P3). Deck verti-
cal displacement y and bending moments M of deck and piers for : (a) continuous deck monolithically
connected to piers, and (b) 7 simply supported spans on elastomeric bearings.
389
The performance of the second alternative (7 simply supported spans on elastomeric bear-
ings) is definitely favorable (Figure 13b). In stark contrast to the statically indeterminate alter-
native, the imposed tectonic deformation does not cause any stressing of either the deck or the
piers. The simply supported decks are only subject to rigid block type rotation and differential
displacements. With adequate seating (to avoid deck falling), this alternative would survive
even such a large tectonic deformation. Admittedly, the results shown herein refer to an extreme
dislocation deliberately to illuminate vividly the differences in response.
8 CONCLUSIONS
This paper has presented a general methodology for the design of bridges against large tectonic
deformation. The problem is decoupled in two analysis steps : the first (local level) dealing with
the response of a single bridge pier subjected to fault rupture deformation; the second (global
level) dealing with the detailed model of the superstructure. At the local level emphasis is given
to fault rupture soil-foundation-structure interaction (FR-SFSI), with the superstructure modeled
in a simplified manner to capture its kinematic constraints. The output of this local level analy-
sis is treated as the input for the global analysis.
The main conclusions are as follows :
1. The design of bridges against tectonic deformation is quite feasible with proper design. The
method of analysis presented herein may form the basis for future Code provisions and re-
quirements on the subject.
2. In all cases investigated herein, the rupture path is strongly affected by the presence of the
foundation. The emerging fault rupture is not only diverted, but is also subject to bifurcation
and diffusion.
3. Piled foundations are in general quite vulnerable to faulting-induced deformation. End bear-
ing piles cannot sustain even moderate bedrock offsets. Floating piles show better perform-
ance, which depends on soil resilience. The latter is in general beneficial in terms of pile
stressing, but not necessarily for the inflicted displacements and rotation at the base of the
pier. A hinged piletocap connection may provide substantial stress relief, allowing a float-
ing piled foundation to sustain larger imposed fault offsets, even of the order of a meter.
4. Rigid massive caisson foundations are clearly advantageous. The faulting-induced deforma-
tion will force the caisson to move and rotate as a rigid body, resulting in vertical and hori-
zontal displacement and rotation at the pier base.
5. The location of fault outcropping plays a major role. For both piled and caisson foundations,
displacements and rotation at pier base are not maximum for the same location of fault out-
crop. Since the exact fault location would never be known precisely a-priori, its location
relative to the foundation has to be parametrically investigated in design.
6. Continuous, statically indeterminate, superstructure systems are in general disadvantageous
(the deck is forced to follow the imposed differential displacements). Statically determinate
systems (such as multiple separate simply supported decks), allowing relative displacement
and rotation without stressing are quite favourable.
7. In all cases, special care should be taken to avoid fall of the deck due to excessive relative
displacements. Ample seating and adequate restraining devices, such as stoppers, are a ne-
cessity.
ACKNOWLEDGEMENT
This work was funded by OSE (the Greek Railway Organization), as part of the research project
Railway Bridges on Active Seismic Faults.

390

REFERENCES
Anastasopoulos I., Gerolymos N., Gazetas G. (2001), Possible collapse reasons of an access span of the
Nishinomiya-ko bridge : Kobe 1995, Proceedings of the 4
th
Hellenic Conference on Geotechnical En-
gineering, Athens, Vol. 2, pp. 8390.
Anastasopoulos I., & Gazetas G. (2007a), Foundation-Structure Systems over a Rupturing Normal Fault :
Part I. Observations after the Kocaeli 1999 Earthquake, Bulletin of Earthquake Engineering, 5 (3), pp.
253275.
Anastasopoulos I., & Gazetas G. (2007b), Behaviour of StructureFoundation Systems over a Rupturing
Normal Fault : Part II. Analysis of the Kocaeli Case Histories, Bulletin of Earthquake Engineering, 5
(3), pp. 277301.
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, 133 (8), pp. 943958.
Anastasopoulos I., Gazetas G., Bransby M.F., Davies M.C.R., and El Nahas A. (2008), Normal Fault
Rupture Interaction with Strip Foundations, Journal of Geotechnical and Geoenvironmental Engi-
neering, ASCE, 134 (8)
Berrill, J.B. (1983), Two-dimensional analysis of the effect of fault rupture on buildings with shallow
foundations, Soil Dynamics and Earthquake Engineering, 2 (3), pp. 156-160.
Bransby, M.F., Davies, M.C.R., and El Nahas, A. (2008a), Centrifuge modelling of normal fault-
foundation interaction, Bulletin of Earthquake Engineering, Special Issue : Integrated approach to
fault rupture- and soil-foundation interaction, 6 (4) (in press).
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., Seed, R.B., Cluff, L.S., and Seed, H.B. (1994a), Earthquake Fault Rupture Propagation
through Soil, Journal of Geotechnical Engineering, ASCE, 120 (3), pp. 543561.
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, 120 (3), pp. 562-580.
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 Ur-
ban Facilities, University of Tokyo Press, pp. 55-79.
Brune, J.N., and Allen, C.R. (1967), A low-stress-drop, low magnitude earthquake with surface faulting.
The Imperial, California, Earthquake of March 4, 1966, Bulletin of the Seismological Society of Amer-
ica, 57, pp. 501-514.
Cole, D.A. Jr., and Lade, P.V. (1984), Influence Zones in Alluvium Over Dip-Slip Faults, Journal of
Geotechnical Engineering, ASCE, 110 (5), pp. 599-615.
Duncan, J.M., and Lefebvre, G., (1973), Earth Pressure on Structures Due to Fault Movement, Journal of
Soil Mechanics and Foundation Engineering, ASCE, Vol. 99, pp. 11531163.
Erdik M., (2001), Report on 1999 Kocaeli and Dzce (Turkey) Earthquakes, Structural Control for Civil
and Infrastructure Engineering, Ed. by, F. Casciati, G. Magonette, World Scientific.
Faccioli, E., Anastasopoulos, I., Callerio, A., and Gazetas, G. (2008), Case histories of faultfoundation
interaction, Bulletin of Earthquake Engineering, Special Issue : Integrated approach to fault rupture-
and soil-foundation interaction, 6 (4) (in press).
Gerolymos N., and Gazetas G. (2006a), Winkler model for lateral response of rigid caisson foundations
in linear soil, Soil Dynamics and Earthquake Engineering, Vol. 26, No. 5, pp. 347361.
Gerolymos N., and Gazetas G. (2006b), Development of Winkler model for static and dynamic response
of caisson foundations with soil and interface nonlinearities, Soil Dynamics and Earthquake Engi-
neering, Vol. 26, No. 5, pp. 363376.
Gerolymos N., Giannakou A., Anastasopoulos I., and Gazetas G. (2008), Evidence of Beneficial Role of
Inclined Piles : Observations and Numerical Results, Bulletin of Earthquake Engineering, Special Is-
sue : Integrated approach to fault rupture- and soil-foundation interaction, 6 (4).
Hanlong L., Susamu I., Ichii K. (1997), Evaluation of deformation to the pneumatic caisson foundations
of the Kobe Ohashi bridge, Report of the Port and Harbor Research Institute, Japan.
Hwang, H.Y. (2000), Taiwan Chi-Chi Earthquake 9.21.99. Birds eye view of Cher-Lung-Pu Fault, Fly-
ing Tiger Cultural Publ., Taipei, Taiwan, pp. 150.
Kawashima, K. (2001), Damage of Bridges Resulting from Fault Rupture in The 1999 Kocaeli and
Duzce, Turkey Earthquakes and The 1999 Chi-Chi, Taiwan Earthquake, Workshop on Seismic Fault-
Induced FailuresPossible Remedies for Damage to Urban Facilities, University of Tokyo Press, pp.
391
171-190.
Niccum, M.R., Cluff, L.S., Chamoro, F., and Wylie, L. (1976), Banco Central de Nicaragua : A case his-
tory of a high-rise building that survived surface fault rupture, in Humphrey, C.B., ed., Engineering
Geology and Soils Engineering Symposium, No. 14, Idaho Transportation Department, Division of
Highways, pp. 133-144.
Pamuk, A., Kalkanb, E., Linga, H.I. (2005), Structural and geotechnical impacts of surface rupture on
highway structures during recent earthquakes in Turkey, Soil Dynamics and Earthquake Engineering,
Vol. 25, pp. 581-589.
Slemmons, D.B. (1957), Geological Effects of the Dixie Valley-Fairview Peak, Nevada, Earthquakes of
December 16, 1954, Bul. of the Seism. Soc. of America, 47 (4), pp. 353375.
Taylor, C.L., Cline, K.M. Page, W.D., and Schwartz, D.P. (1985), The Borah Peak, Idaho earthquake of
October 28, 1983 Surface Faulting and Other Phenomena, Earthquake Spectra, 2 (1), pp. 2349.
Tazoh T., Ohtsuki A., Aoki T., Mano H., Isoda K., Iwamoto T., Arakawa T., Ishihara T., and Ookawa M.
(2002), A new pile-head device for decreasing construction costs and increasing the seismic perform-
ance of pile foundations, and its application to structures, Proc. 12
th
European Conference on Earth-
quake Engineering, Elsevier Science Ltd. Paper No. 720.
Ulusay, R., Aydan, O., Hamada, M. (2002), The behaviour of structures built on active fault zones: Ex-
amples from the recent earthquakes of Turkey, Structural Engineering & Earthquake Engineering,
JSCE, 19 (2), pp. 149167.
Youd, T. L. (1989), Ground Failure Damage to Buildings During Earthquakes, Foundation Engineer-
ingCurrent Principles and Practices, Vol. 1, pp. 758-770. New York: ASCE.
Youd, T. L., Bardet, J-P, and Bray, J.D. (2000), Kocaeli, Turkey, Earthquake of August 17, 1999 Recon-
naissance Report, Earthquake Spectra, Suppl. A to Vol. 16, pp. 456.
392
1 INTRODUCTION
Seismic design of structures recognises that highly inelastic material response is un-avoidable
under the strongest possible earthquake. Ductility levels of the order of 3 or more are usually
allowed to develop under strong seismic shaking, implying that the strength of a number of
critical bearing elements is fully mobilized. In the prevailing structural terminology, plastic
hinging is allowed as long as the overall stability is maintained.
In marked contrast, a crucial goal of current practice in seismic foundation design, particu-
larly as entrenched in the respective codes [e.g. EC8], is to avoid the mobilisation of strength
in the foundation. In structural terminology : no plastic hinging is al-lowed in the foundation
soil. In simple geotechnical terms, the designer has to ensure that the below-ground (difficult to
inspect) support system will not even reach a number of thresholds that would statically imply
failure. Thus, mobilisation of the bearing-capacity failure mechanism, foundation sliding and
uplifting, or any relevant combina-tion is prohibited. To make sure that such mechanisms will
not develop, overstrength factors plus (explicit and implicit) factors of safety larger than 1 are
introduced against each of those failure modes. This way, the engineer is certain that founda-
tion performance will be satisfactory and that there will be no need to inspect and/or repair after
strong earthquake shaking.
However, a growing body of evidence suggests that soilfoundation plastic yielding under
seismic excitation may be advantageous, and should be seriously considered in analysis and per-
haps allowed in design [Pecker, 1998; Martin & Lam, 2000; FEMA-356, 2000; Kutter et al.,
2003, Apostolou & Gazetas, 2005; Kawashima et al., 2007]. The urgent need to explicitly con-
sider the possibility of the foundation system to go beyond failure thresholds, and the poten-
tial usefulness of doing so, emerge from :
Towards a Reversal of Seismic Capacity Design. Part A : Analy-
sis of Bridge PierFoundation System

I. Anastasopoulos, M. Loli, N. Gerolymos, M. Apostolou, G. Gazetas

ABSTRACT: The paper illuminates a new seismic design philosophy, which takes advantage of
soil failure to protect the superstructure. A reversal of conventional capacity design is in-
troduced, through intentional under-designing of the foundation. A simple but realistic bridge is
used as an illustrative example of the effectiveness of the new philosophy. Two alternatives are
compared : one in compliance with conventional capacity design, with over-designed founda-
tion so that the plastic hinge develops in the superstructure ; and one with under-designed
foundation, so that the plastic hinge may occur in the soil. The seismic performance of the
two alternatives is investigated through nonlinear dynamic time history analysis, using a variety
of seismic excitations. It is shown that the performance of both alternatives is totally acceptable
for moderate seismic motions. For large intensity seismic motions, the performance of the new
scheme is proven advantageous, not only avoiding collapse but hardly suffering any inelastic
structural deformation. The penalty to pay is a substantial foundation settlement.

393
(a) The large (often huge) effective ground acceleration, A, and velocity, V, levels recorded in
several earthquakes in the last 20 years. Few examples : 1994 M
s
6.8 Northridge : A = 0.98 g,
V = 140 cm/s ; 1995 M
JMA
7.2 Kobe : A = 0.85 g, V = 120 cm/s ; 1986 M
s
5.6 San Salvador :
A = 0.75 g, V = 84 cm/s ; 2007 M
JMA
6.9 Niigata : A =1.20 g, V = 100cm/s. With the corre-
spondingly large accelerations in the (aboveground) structure (spectral SA values well in ex-
cess of 1 g) from such ground motions, preventing plastic hinging in the foundation system
is a formidable task. And in fact, it may not even be desirable: enormous ductility demands
might be imposed to the structure if soilfoundation yielding would not take place to effec-
tively limit the transmitted accelerations. Several present-day criticallyimportant structures on
relatively loose soil could not have been de-signed against severe ground motions if plastic
hinging of some sort could not be allowed to take place in the foundation. [Example : the 90
m in diameter, 80 m tall foot-ings of the Rion-Antirrion bridge piers in Greece: allowing sliding
under certain seismic conditions was a beneficial necessity ; Pecker, 2003].
(b) In seismically retrofitting a structure, allowing for foundationsoil yielding is the only ra-
tional alternative. Because increasing the structural capacity of some elements, or introducing
some stiff elements, would then imply that the forces transmitted onto the foundation will be in-
creased, to the point that it might not be technically or economically feasible to undertake them
elastically. Thus, the new American retrofit design guidelines (FEMA 356) explicitly permit
some forms of inelastic deformation in the foundation. A simple hypothetical example referring
to an existing threebay multistory building frame which is to be retrofitted with a singlebay
concrete shear wall was presented by Martin & Lam [2000]. Such a wall, being much stiffer
than the col-umns of the frame, will carry most of the inertia-driven shear force and will thus
trans-mit a disproportionately large horizontal force and overturning moment onto the founda-
tion compared with its respective vertical force. If uplifting, sliding, and mobilisation of bearing
capacity failure mechanisms in the foundation were spuriously ignored, or conversely correctly
taken into account, would lead to dramatically different results. With beyondthreshold action
in the foundation the shear wall sheds off some of the load onto the columns of the frame,
which must then be properly reinforced ; the opposite is true when such action (beyond the
thresholds) is disallowed. The Engineer therefore should be able to compute the consequences
of plastic hinging in the foundation before deciding whether such hinging must be accepted,
modified or avoided (through foundation changes).
(c) Compatibility with structural design is another reason for the soilstructure interaction
analyst to compute the collapse lateral load of the foundation system, as well as the complete
loaddisplacement or loadrotation response to progressively increasing load up to collapse. In-
deed, in SOA structural engineering use is made of pushover analysis, which in order to be
complete requires the development of such information from the foundation analyst. In addition
to the above theoretical arguments, there is a growing need for estimating the collapse mo-
tion: insurance coverage of major construction facilities is often based on estimated losses un-
der the worst possible (as opposed to probable) earthquake scenario.
(d) The current trend in structural earthquake engineering calls for a philosophical change :
from strength-based design (involving force considerations) to performance-based design (in-
volving displacement considerations) [Pauley, 2002; Priestley, 2000]. Geotechnical earthquake
engineering has also been slowly moving towards performancebased design : gravity retaining
structures are indeed allowed to slide.
The time is therefore ripe for soilfoundationstructure interaction (SFSI) to also move from
imposing safe limits on forces and moments acting on the foundation (aiming at avoiding
pseudo-static failure) to performancebased design in which all possible conventional fail-
ure mechanisms are allowed to develop, to the extent that maximum and permanent displace-
ments and rotations are kept within acceptable limits. This paper introduces a new design phi-
losophy, beyond conventional capacity design, in which superstructure plastic hinging is
replaced by soilfoundation failure (see sketch of Figure 1) : i.e. soil failure is used as a shield
for the superstructure (exactly the opposite of conventional capacity design). A simple but real-
istic typical highway bridge is used as an illustrative example of the effectiveness of the new
seismic design philosophy.

394
Plastic hinging at
the superstructue
B
Avoidance of
foundation failure
B
Plastic hinging at
the foundation
Avoidance of
structural failure
Conventional Capacity Design New Design Philosophy
Plastic hinging at
the superstructue
B
Avoidance of
foundation failure
B
Plastic hinging at
the foundation
Avoidance of
structural failure

Figure 1. Conventional capacity design compared to the new design philosophy.
2 PROBLEM DEFINITION AND ANALYSIS METHODOLOGY
As depicted in Figure 2, we consider a typical highway bridge excited in the transverse direc-
tion. A deck of mass m = 1200 Mgr is monolithically connected to a reinforced concrete pier of
diameter D = 3 m and height H = 12 m. The bridge chosen for analysis is similar to the Hanshin
Expressway Fukae bridge, which collapsed spectacularly in the Kobe 1995 earthquake [Seible
et al., 1995; Iwasaki et al., 1995; Park, 1996]. The bridge is designed in accordance to EC8
[2000] and the Greek Seismic Code [EAK 2000] for a design acceleration A = 0.24 g, consider-
ing a (ductility-based) behavior factor q = 2. With an elastic (fixed-base) vibration period T =
0.48 sec and design spectral acceleration SA = 0.3 g, to undertake the resulting design bending
moment M
D
43 MNm, a longitudinal reinforcement of 100 d
bL
= 32 mm bars (10032)
is required, combined with d
bw
=13 mm hoops spaced at 8 cm.
The pier is founded through a square foundation of width B on an idealized homogeneous 25
m deep stiff clay layer, of undrained shear strength S
u
= 150 kPa (representative soil conditions
for which a surface foundation would be a realistic solution). Two different foundation widths
are considered to represent the two alternative design ap-proaches. A larger foundation, B = 11
m, is designed in compliance with conventional capacity design, applying an overstrength factor
Rd
= 1.4 to ensure that the plastic hinge will develop in the superstructure (base of pier).
Taking account of maximum allowable uplift (eccentricity e = M / V B/3, where V is the ver-
tical load), the resulting safety factors for static and seismic loading are FSV = 5.6 and FSE =
2.0, respec-tively. A smaller, under-designed, B = 7 m foundation is considered in the spirit of
the new design philosophy. Its static safety factor FSV = 2.8, but is designed applying an un-
derstrength factor 1/
Rd
= 1/1.4 0.7 for seismic loading. Thus, the resulting safety factor for
seismic loading is lower than 1.0 (FSE 0.7). In fact, as it will be shown below, the under-
designed foundation will not allow the design seismic action to develop. Hence, FSE does not
really have a physical meaning in this case ; it is just an apparent temporary factor of safety.
The analysis is conducted assuming plane-strain soil conditions, taking account of material
(in the soil and the superstructure) and geometric (due to uplifting and P- effects) nonlineari-
ties. The pier is modeled with nonlinear beam elements, while the deck is represented by a mass
element. Soil and foundation are modeled with quadrilateral continuum elements, nonlinear for
the former and elastic for the latter. The foundation is connected to the soil with special contact
elements, allowing for realistic simulation of possible detachment and sliding at the soil
foundation interface representing with realism the tensionless and frictionless interface. The
mass of the footing and of the pier are also taken into account.

395
Pier : Nonlinear beam elements
(d = 3m, H =12 m)
Soilfoundation interface :
contact elements (uplifting)
Deck :
mass element
(m = 1200 Mg)
25 m
Seismic excitation :
29 real records covering a wide range of
earthquake scenarios
Foundation :
elastic quadratic elements
(conventional design : B = 11 m ;
new design concept : B = 7 m)
Soil :
Nonlinear quadratic
plane strain elements
(S
u
= 150 kPa)
Pier : Nonlinear beam elements
(d = 3m, H =12 m)
Soilfoundation interface :
contact elements (uplifting)
Deck :
mass element
(m = 1200 Mg)
25 m
Seismic excitation :
29 real records covering a wide range of
earthquake scenarios
Foundation :
elastic quadratic elements
(conventional design : B = 11 m ;
new design concept : B = 7 m)
Soil :
Nonlinear quadratic
plane strain elements
(S
u
= 150 kPa)

Figure 2. Methodology of finite element modelling.

2.1 Constitutive Modeling of Soil
A nonlinear constitutive model with Von Mises failure criterion, nonlinear kinematic hardening
and associative plastic flow rule is employed. According to the Von Mises failure criterion, the
evolution of stresses is described by the relation :
= + (1)
where :
is the value of stress at zero plastic strain, assumed to remain constant. The parameter
is the backstress that defines the kinematic evolution of the yield surface in the stress space.
The evolution of the kinematic component of the yield stress is de-scribed by the following law:
pl pl
) (
1

=
0
C (2)
where :
pl
is the plastic strain rate, C the initial kinematic hardening modulus

(
y y
C E = = ) and a parameter that determines the rate at which the kinematic hardening
decreases with increasing plastic deformation.
The maximum yield stress (at saturation) is :
0 y

+ =
C
(3)
According to the Von Mises yielding criterion this ultimate stress is :
u
S 3
y
= (4)
From equations (3) and (4) we have :
0
3
C
=
u
S
(5)
Model parameters are calibrated to fit published G curves of the literature, following the
procedure described in Gerolymos et al. [2005]. Figure 3a illustrates the validation of the kine-
396
G

/

G
0
(%)
10
1
10
1.0
0.5
10
2
10
3
10
4
1
0

FE Model FE Model
Ishibashi & Zhang [1993]

M

(
M
N
m
)
Curvature (1/m)
0.06 0.08
60
40
0.04 0.02 0
0
FE Model FE Model
Section analysis Section analysis
d = 3 m
10032
S400
20
C30
(c)
C C
u u
G

/

G
0
(%)
10
1
10
1.0
0.5
10
2
10
3
10
4
1
0

FE Model FE Model

G

/

G
0
(%)
10
1
10
1.0
0.5
10
2
10
3
10
4
1
0

FE Model FE Model
FE Model FE Model

M

(
M
N
m
)
Curvature (1/m)
0.06 0.08
60
40
0.04 0.02 0
0
FE Model FE Model
d = 3 m
10032
S400
20
C30
(c)
C C
u u
M

(
M
N
m
)
Curvature (1/m)
0.06 0.08
60
40
0.04 0.02 0
0
FE Model FE Model
FE Model FE Model
d = 3 m
10032
S400
20
C30
(c)
C C
u u
matic hardening model (through simple shear finite element analysis) against published G
curves by Ishibashi and Zhang [1993].
2.2 Constitutive Modeling of Reinforced Concrete
The same model is calibrated to match the response of the reinforced concrete pier in the macro-
scopic momentcurvature level. The reinforcement of the pier circular section (D = 3 m) is cal-
culated according to the provisions of the Greek Code for Reinforced Concrete [, 2000]
for columns with large capacity demands in accordance with the capacity design provisions. The
moment curvature relationship is derived from static concrete section analysis employing the
USC-RC software, which uses the Mander model [Mander et al., 1984] to simulate the stress
strain relationship of confined concrete.
The bending moment of a circular section is by definition related to the normal stresses
with the following expression :
drd sin r 2
0
d/2
0
2

= M (6)
For the maximum yield stress
y
this relationship gives :
d sin
3
2 M
0
2 /
0
3
y y

=

d r
(7)
which yields:
3
y y
1
6
M d = (8)
And so, the maximum yield stress can be expressed as :
3
y
6
d
My
=
(9)
The initial kinematic hardening modulus C is equal to the modulus of elasticity E.
To simulate the softening behavior of the reinforced concrete section after ultimate capacity
is reached, a user subroutine is encoded in the ABAQUS finite element code. Figure 3b depicts
the results of model calibration for the pier against momentcurvature relation of the reinforced
concrete section calculated through section analysis utilizing the USC_RC software [Esmaeily-
Gh & Xiao, 2002], which uses the Mander model [Mander et al., 1984] for confined concrete.
As for soil, model parameters are calibrated using the aforementioned methodology of Geroly-
mos et al. [2005].

Figure 3. (a) Model calibration for soil (stiff clay, S
u
= 150 kPa) against published G- curves by Ishi-
bashi & Zhang [1993] ; (b) Model calibration for the reinforced concrete pier against moment-curvature
response calculated using reinforced concrete section analysis (USC-RC).

397
first yield first yield
C C
n n
M

(
M
N
m
)
Curvature (1/m)
0.02
60
0.01 0
0
M

(
M
N
m
)
(rad)
0.3
40
20
0.2 0.1 0
0
10
(a) (b)
40
20
C C
u u
C
u
C
n
= = 16.6
= = 5.6
30

y y

u u
= = 42
u
H
y
H
=
first yield first yield
C C
n n
M

(
M
N
m
)
Curvature (1/m)
0.02
60
0.01 0
0
M

(
M
N
m
)
(rad)
0.3
40
20
0.2 0.1 0
0
10
(a) (b)
40
20
C C
u u
C
u
C
n
= = 16.6
= = 5.6
C
u
C
n
= = 16.6
= = 5.6
= = 5.6
= = 5.6
30

y y

u u
= = 42
u
H
y
H
=
= = 42
u
H
y
H
=
3 PUSHOVER RESPONSE OF THE TWO ALTERNATIVES
Before proceeding with the dynamic time history analysis of the two alternatives, we investigate
their response in terms of monotonic loading through simulation of the static pushover test.
Displacement controlled horizontal loading is applied at the top of the pier (deck). Figure 4a il-
lustrates the results of the static pushover analysis of the conventionally designed system, in
terms of moment-curvature relation at the base of the pier. The curvature ductility capacity
of
the reinforced concrete section is equal to 16.6 (applying a standard bilinear approximation),
and the displacement ductility capacity of the pier is computed as follows [Priestley et al., 1996]
:
3
y
6
d
My
=
(10)
where : M
u
the ultimate and M
n
the yield bending moment of the reinforced concrete section
(corresponding to c
n
in the moment curvature diagram), H the height of the pier, and L
p
the
length of the plastic hinge :
0.08 0.022 0.044
p ye bl ye bl
L L f d f d = + (11)
where : f
ye
and d
bl
the design yield strength (in MPa) and the diameter of the longitudinal rein-
forcement in the region of the plastic hinge. This results in a displacement ductility capacity of
the conventionally designed system
= 5.6.
Figure 4b depicts the monotonic response of the alternative design according to the new phi-
losophy. Since the behavior of the pier is elastic, the ductility of the system is now associated
with foundation rotation due to bearing capacity failure. This renders the conventional defini-
tion of curvature ductility not applicable. Thus, an equivalent dis-placement ductility capacity
is defined, based on foundation rotation :
u u u
y y y
H
= = = (12)
where :
u
is the ultimate (critical for overturning) foundation rotation, and
y
the yield rota-
tion. This results in a displacement ductility capacity of the new concept (B = 7 m)
= 42.2,
which is almost an order of magnitude larger than the capacity of the conven-tionally designed
system (B = 11 m).

Figure 4. (a) Pushover analysis of the conventionally designed system : the curvature ductility capacity
is equal to 16.6 (using a bilinear approximation for the moment-curvature relation of the pier), yield-
ing displacement ductility capacity
= 5.6 ; (b) Pushover analysis of the new design concept. Since

ductility is now associated with foundation rotation due to mobilization of the bearing capacity failure
mechanism, a new definition of
is introduced, based on foundation rotation ; the estimated capacity
= 42 is almost an order of magnitude larger (compared to conventional design).

398
TCU052_ew
TCU052_ns
TCU068_ew
TCU068_ns
Duzce Bolu090
Imperial Valley #4140
JMA000 JMA090 Takarazuka
Takatori000 Takatori090
Sakarya
Jensen292 Rinaldi228 Rinaldi318
Lefkada2003 Yarimca330
Pyrgos
MNSA
Aegion Kalamata
Lefkada1973
Jensen022
Yarimca060 Duzce Bolu000
Sylmar090 Sylmar360
Lucerne
a (g)
t (s)
0.5
0
0 10
0 1 2 3 4
3
2
1
0
T (sec)
SA (g)
Design Spectrum
Kalamata (1986)
Takatori000 (Kobe)
TCU052_ew
TCU052_ns
TCU068_ew
TCU068_ns
Duzce Bolu090
JMA000 JMA090 Takarazuka
Takatori000 Takatori090
Sakarya
Jensen292 Rinaldi228 Rinaldi318
Lefkada2003 Yarimca330
Pyrgos
MNSA
Aegion Kalamata
Lefkada1973
Jensen022
Yarimca060 Duzce Bolu000
Sylmar090 Sylmar360
Lucerne
a (g)
t (s)
0.5
0
0 10
a (g)
t (s)
0.5
0
0 10
0 1 2 3 4
3
2
1
0
T (sec)
SA (g)
0 1 2 3 4
3
2
1
0
T (sec)
SA (g)
Design Spectrum
Kalamata (1986)
Takatori000 (Kobe)
Design Spectrum
Kalamata (1986)
Takatori000 (Kobe)
4 TIME-HISTORY DYNAMIC ANALYSIS OF THE TWO ALTERNATIVES
We now investigate the seismic performance of the two alternatives through nonlinear dynamic
time history analysis. An ensemble of 29 real accelerograms is used as seismic excitation. The
latter is applied at bedrock level. As illustrated in Figure 5, the selected records cover a wide
range of seismic motions, ranging from medium intensity (e.g. Kalamata, Pyrgos, Aegion) to
relatively stronger (e.g. Lefkada-2003, Imperial Valley), and to very strong accelerograms char-
acterized by forward-rupture directivity effects, or large number of significant cycles, or fling-
step effects (e.g. Takatori, JMA, TCU). In terms of spectral accelerations (SA), many of the
considered accelerograms exceed (by far, in many cases) the design spectrum of the bridge.
In the following sections, we compare the response of the two alternatives for : (i) moderate
intensity seismic motions not exceeding the design limits (at least not substantially), and (ii)
large intensity seismic motions that substantially exceed the design limits. In the first case, the
objective is to determine the serviceability of the bridge after such a moderate intensity earth-
quake. In the latter case, the main objective is safety (i.e. avoidance of collapse in an almost
improbable event). Bearing in mind that the spectral acceleration SA of a motion is not al-
ways the most crucial parameter of nonlinear response, the characterization of the seismic mo-
tions is conducted on the basis of spectral displacements SD, following the logic of displace-
ment-based design [e.g. Bertero, 1996; Tassios, 1998; Priestley, 2000 ; Faccioli et al., 2001].

Figure 5. Real earthquake records used for analysis of the two alternatives, along with their elastic spectra
and the design spectrum of the investigated bridge.
399
0 0.01 0.02 0.01 0.02
80
40
40
80
0
0 0.01 0.02 0.01 0.02
80
40
40
80
0
M

(
M
N
m
)
M

(
M
N
m
)
(rad) (rad)
0 0.01 0.02 0.01 0.02
0.1
0.05
0.05
0
(rad)
w

(
m
)
0 0.01 0.02 0.01 0.02
0.1
0.05
0.05
0
(rad)
w

(
m
)
(a1) (a2)
(b1) (b2)
Small
foundation
settlement
Larger
foundation
settlement
Practically
elastic
foundation
response
Inelastic
foundation
response
static
pushover
0 0.01 0.02 0.01 0.02
80
40
40
80
0
0 0.01 0.02 0.01 0.02
80
40
40
80
0
0 0.01 0.02 0.01 0.02
80
40
40
80
0
0 0.01 0.02 0.01 0.02
80
40
40
80
0
M

(
M
N
m
)
M

(
M
N
m
)
(rad) (rad)
0 0.01 0.02 0.01 0.02
0.1
0.05
0.05
0
(rad)
w

(
m
)
0 0.01 0.02 0.01 0.02
0.1
0.05
0.05
0
(rad)
w

(
m
)
(a1) (a2)
(b1) (b2)
Small
foundation
settlement
Larger
foundation
settlement
Practically
elastic
foundation
response
Inelastic
foundation
response
static
pushover
4.1 Performance in Moderate Intensity Seismic Motions
A comparison of the performance of the two design alternatives subjected to a moderate inten-
sity earthquake is illustrated in Figures 6 and 7. The excitation accelerogram is from the 1986
M
s
6.0 Kalamata (Greece) earthquake. At a fault distance of 5 km from the city center, the
earthquake caused substantial structural damage to a variety of building structures. With Modi-
fied Mercalli Intensity (MMI) levels reaching or exceeding VIII, almost 60% of the buildings
had to be retrofitted after the earthquake [Gazetas et al., 1990]. It is emphasized that the affected
building stock had been designed and constructed according to older seismic codes, practically
without any capacity design considerations. Evidently, the same degree of damage should not be
expected for modern structures. In terms of SA (Figure 5), the record exceeds the design spec-
trum by a factor of almost 2 for periods T ranging from 0.2 to 0.6 sec ; for the longer periods
that are of more relevance for inelastic systems, it is within the design SA.
In Figure 6a the comparison is portrayed in terms of the foundation experienced moment
rotation (M). As expected, while the response of the conventionally designed foundation is
practically elastic (Figure 6a1), the under-designed foundation (new design philosophy) experi-
ences some inelasticity (Figure 6a2). In Figure 6b the comparison is in terms of foundation set-
tlementrotation (w). The conventionally designed system is subjected to limited settlement
w 2 cm (Figure 6b1). In marked contrast, the new concept (Figure 6b2) experiences larger but
quite tolerable dynamic settlement : w 4 cm.
Figure 7a illustrates the momentcurvature response at the base of the pier for the conven-
tionally designed system. Some inelasticity takes place (i.e. minor structural damage), but the
curvature ductility is tolerable : the demand is almost an order of magnitude lower than the ca-
pacity of the reinforced concrete section. In the case of the new design philosophy, thanks to
foundation yielding the response of the pier (not shown herein) is purely elastic.

Figure 6. Comparison of the response of the two alternatives subjected to a moderate intensity seismic
motion (Kalamata, 1986), within the design limits. (a1 and a2) Overturning moment versus rotation (M
) for the two foundations. While the conventional design entails practically elastic response of the foun-
dationsoil system, the new design scheme experiences substantial inelastic action. (b1 and b2) Settle-
mentrotation (w) response for the two foundations. Thanks to its large foundation and pier yielding,
the conventionally designed system experiences limited settlement. In contrast, the smaller foundation
(new concept) experiences larger cumulative settlement, which is still quite tolerable.

400
C

r
(
m
)
(
m
)
t (sec)
0
0.1
0
0.2
0.1
5 10 15
t (sec)
0
0.1
0
0.2
0.1
5 10 15 20 25
(b1) (b2)
Slightly larger, but
tolerable, deck drift
Limited
deck drift
0.02 0.01 0.02
80
40
40
80
0
0.01 0
M

(
M
N
m
)
curvature (1/m)
(a1)
Slightly inelastic,
but tolerable, pier
response
static pushover
0.02 0.01 0.02
80
40
40
80
0
0.01 0
M

(
M
N
m
)
curvature (1/m)
(a2)
Purely elastic
response
static pushover
C

r
C

r
(
m
)
(
m
)
t (sec)
0
0.1
0
0.2
0.1
5 10 15
t (sec)
0
0.1
0
0.2
0.1
5 10 15 20 25
(b1) (b2)
Slightly larger, but
tolerable, deck drift
Limited
deck drift
0.02 0.01 0.02
80
40
40
80
0
0.01 0
M

(
M
N
m
)
curvature (1/m)
(a1)
Slightly inelastic,
but tolerable, pier
response
static pushover
0.02 0.01 0.02
80
40
40
80
0
0.01 0
M

(
M
N
m
)
curvature (1/m)
(a2)
Purely elastic
response
static pushover
The time histories of deck horizontal displacement, i.e. the drift , for the two alternatives are
compared in Figure 7b. As graphically illustrated in the adjacent sketch notation, the drift has
two components [see also Priestley et al., 1996] : (i) the flexural drift
C
, i.e. the structural
displacement due to flexural distortion of the pier column, and (ii) the rocking drift
r
= H,
i.e. the displacement due to rocking motion of the foundation. This way, the contribution of pier
flexural distortion and foundation rotation to the final result of interest (i.e. the total drift ) can
be inferred. As might have been expected, while for the conventional design (over-designed
foundation) is mainly due to pier distortion C (Figure 7b1), exactly the opposite is observed
for the under-designed foundation of the new design philosophy : is mainly due to foundation
rotation r (Figure 7b2). Nevertheless, despite the differences in the mechanism leading to its
development (pier distortion versus or foundation rotation), the total drift is quite similar :
maximum and residual is slightly larger for the new concept, but quite tolerable.

Figure 7. Comparison of the response of the two alternatives subjected to a moderate intensity seismic
motion (Kalamata, 1986), within the design limits. (a1 and a2) Bending momentcurvature response at
the base of the pier. In the conventionally designed system some inelasticity develops, but the ductility
demand is totally tolerable. The response of the pier of the new concept is purely elastic. (b1 and b2)
Time histories of deck drift (horizontal displacement). While for the conventional design is mainly
due to flexural pier distortion
C
, for the new design concept the drift is mainly due to foundation rota-
tion
r
. The residual drift is slightly larger in the new design scheme, but quite tolerable.

4.2 Performance in Large Intensity Seismic Motions
We now compare the response of the two alternatives for a large intensity seismic motion, sub-
stantially exceeding the design limits (Figures 8 and 9) : the Takatori accelerogram of the 1995
MJMA 7.2 Kobe earthquake. With a direct economic loss of more than $100 billion [EERI,
1995], the Kobe earthquake needs no introduction. Constituting the greatest earthquake disaster
in Japan since the 1823 Ms 8 Kanto earthquake, it is simply considered as one of the most dev-
401
(
m
)
t (sec)
0
2
1
1
0
5 10 15 20 25
(
m
)
t (sec)
0
2
1
1
0
5 10 15 20 25
0.25 0.15 0.25
80
40
40
80
0
0.15 0.05 0.05
M

(
M
N
m
)
curvature (1/m)
(a1) (a2)
(b1)
(c1) (c2)
Survival with
limited deck drift
Collapse
Collapse
Ultimate
ductility capacity
static pushover
curvature (1/m)
(b2)
Purely elastic
response
static pushover
0.25 0.15 0.25
80
40
40
80
0
0.15 0.05 0.05
M

(
M
N
m
)
curvature (1/m)
(
m
)
t (sec)
0
2
1
1
0
5 10 15 20 25
(
m
)
t (sec)
0
2
1
1
0
5 10 15 20 25
(
m
)
t (sec)
0
2
1
1
0
5 10 15 20 25
(
m
)
t (sec)
0
2
1
1
0
5 10 15 20 25
0.25 0.15 0.25
80
40
40
80
0
0.15 0.05 0.05
M

(
M
N
m
)
curvature (1/m)
(a1) (a2)
(b1)
(c1) (c2)
Survival with
limited deck drift
Collapse
Collapse
Ultimate
ductility capacity
static pushover
curvature (1/m)
(b2)
Purely elastic
response
static pushover
0.25 0.15 0.25
80
40
40
80
0
0.15 0.05 0.05
M

(
M
N
m
)
curvature (1/m)
astating earthquakes of modern times. Of special interest is the damage inflicted to the bridges
of Hanshin Expressway, which ranged from collapse to severe damage [e.g. Seible et al., 1995].
As aforementioned, the bridge chosen for our analysis is very similar to the Fukae section of
Hanshin Expressway, 630 m of which collapsed during the earthquake of 1995 [Iwasaki et al.,
1995; Park, 1996]. It is therefore logical to consider this as a reasonably realistic example of an
above the limits earthquake. In particular, the Takatori record [Fukushima et al., 2000] consti-
tutes one of the worst seismic motions ever recorded : PGA = 0.70 g, PGV = 169 cm/s, bearing
the mark of forward rupture directivity. Compare its response spectrum to the design SA
(Figure 5) to notice how much larger it is throughout the whole range of periods.
Figure 8a compares the response of the two alternatives, in terms of deformed mesh with su-
perimposed plastic strain. In the conventionally designed system (Figure 8a1) there is very little
inelastic action in the soil ; the red regions of large plastic deformation are seen only under the
severely battered edges of the rocking foundation but without extending below the founda-
tion. Plastic hinging forms at the base of the pier, leading to a rather intense accumulation of
curvature (deformation scale factor = 2). In stark contrast, with the new design scheme (Figure
8a2) the plastic hinge takes the form of mobilization of the bearing capacity failure mecha-
nisms in the underlying soil, leaving the superstructure totally intact. Notice that the red regions
of large plastic shearing are of great extent, covering both half-widths of the foundation and in-
dicating alternating mobilization of the bearing capacity failure mechanism.

Figure 8. Comparison of the response of the two alternatives subjected to a large intensity seismic motion
(Takatori, 1995), exceeding the design limits. (a1 and a2) Deformed mesh with superimposed plastic
strain, showing the location of the plastic hinge. (b1 and b2) Bending momentcurvature response at
the pier base. Subjected to ductility demand far exceeding the design, the conventionally designed pier
would collapse. With the new design philosophy, the pier remains elastic. (c1 and c2) Time histories of
deck drift . With its response dominated by pier flexural failure, the conventionally designed system
collapses. The maximum drift of the new concept is large (mainly due to foundation rotation), but it sur-
vives with insignificant residual drift.
402
0 0.03 0.06 0.03 0.06
80
40
40
80
0
0 0.03 0.06 0.03 0.06
80
40
40
80
0
M

(
M
N
m
)
M

(
M
N
m
)
(rad) (rad)
0 0.03 0.06 0.03 0.06
0.3
0.2
0
0.1
(rad)
w

(
m
)
0 0.03 0.06 0.03 0.06
0.3
0.2
0
0.1
(rad)
w

(
m
)
(a1) (a2)
(b1) (b2)
Small
foundation
settlement
Increased
foundation
settlement
Practically
elastic
foundation
response
Strongly
inelastic
foundation
response
static
pushover
0 0.03 0.06 0.03 0.06
80
40
40
80
0
0 0.03 0.06 0.03 0.06
80
40
40
80
0
M

(
M
N
m
)
M

(
M
N
m
)
(rad) (rad)
0 0.03 0.06 0.03 0.06
80
40
40
80
0
0 0.03 0.06 0.03 0.06
80
40
40
80
0
0 0.03 0.06 0.03 0.06
80
40
40
80
0
0 0.03 0.06 0.03 0.06
80
40
40
80
0
M

(
M
N
m
)
M

(
M
N
m
)
(rad) (rad)
0 0.03 0.06 0.03 0.06
0.3
0.2
0
0.1
(rad)
w

(
m
)
0 0.03 0.06 0.03 0.06
0.3
0.2
0
0.1
(rad)
w

(
m
)
0 0.03 0.06 0.03 0.06
0.3
0.2
0
0.1
(rad)
w

(
m
)
0 0.03 0.06 0.03 0.06
0.3
0.2
0
0.1
(rad)
w

(
m
)
(a1) (a2)
(b1) (b2)
Small
foundation
settlement
Increased
foundation
settlement
Practically
elastic
foundation
response
Strongly
inelastic
foundation
response
static
pushover
As seen in Figure 8b, the pier of the conventional system suffers a curvature ductility exceed-
ing the design limit by almost one order of magnitude clearly a case of collapse. This is fur-
ther confirmed by the time history of deck drift (Figure 8c1). In marked contrast, the system
designed according to the new philosophy easily survives (Figure 8c2). It experiences substan-
tial maximum deck drift (about 40 cm), almost exclusively due to foundation rotation
r
. Nev-
ertheless, the residual foundation rotation leads to a tolerable 7 cm deck horizontal displacement
at the end of the earthquake.
The momentrotation (M) response of the two foundations is depicted in Figure 9a. Re-
specting its design principles, the conventional B = 11 m foundationsoil system remains practi-
cally elastic (Figure 9b1) ; the causes are now evident : (i) the rocking stiffness of the founda-
tion, being proportional to B
3
, is large and leads to small stresses in the soil ; and (ii) pier failure
effectively limits the loading transmitted onto the foundation. Exactly the opposite is observed
for the under-designed (B = 7 m) foundation, the response of which is strongly inelastic (Figure
9b2) : mobilization of bearing capacity failure acts as a safety valve or a fuse for the super-
structure.
But despite such excessive soil plastification, not only the structure does not collapse, but the
residual (permanent) rotation is rather limited (as already attested by the residual deck drift).
Under static conditions, the development of this rotational mechanism on either side of the
foundation would have lead to toppling of the structure. However, dynamically, each side of
the rotational mechanism deforms plastically for a very short period of time (momentarily),
producing limited inelastic rotation which is partially cancelled by the ensuing deformation on
the opposite side. Obviously, exactly the same applies to structural plastic hinging in conven-
tional design. The main difference between the two alternatives lies in the mechanism of energy
dissipation, and the related displacement ductility margins.
However, energy dissipation is not attainable at zero cost : in our case the cost is the increase
of foundation settlement. Figure 9b compares the settlementrotation (w) response for the two
alternatives. While the practically elastic response of the conventional (over-designed) founda-
tion leads to a minor 7 cm settlement (Fig. 9b1), the under-designed foundation of the new phi-
losophy experiences an increased accumulated 24 cm settlement (Fig. 6b2). Although such set-
tlement is certainly not negligible, it can be considered as a small price to pay to avoid collapse
under such a tremendous ground shaking.

Figure 9. Comparison of the response of the two alternatives subjected to a large intensity seismic motion
(Takatori, 1995), exceeding the design limits. (a1 and a2) Overturning momentrotation (M) response
of the two foundations. While the response of the conventionally designed foundation remains practically
elastic, the response of the new concept is strongly inelastic. (b1 and b2) Foundation settlementrotation
(w) response. Again, while the settlement of the conventional system is minor, the new design experi-
ences a large (24 cm) settlement : a small price to pay to avoid collapse.
403
5 DISCUSSION AND CONCLUSIONS
The overall performance (for all 29 seismic excitations) of the two design alternatives is com-
plied and synopsized in Figure 10. We present key performance indicators with respect to peak
ground acceleration a
E
of the seismic excitation (at bedrock).
Figure 10a compares the ratio of displacement ductility demand over ductility capacity
demand
/
capacity
, for the two alternatives. For the conventional design (Figure 10a1), we also in-
dicate the likely damage level according to Response Limit States of Priestley et al. [1996]. In
accordance with conventional design principles, the damage to the conventional system is
within the serviceability limits only in moderate not exceeding the design limits earthquake
motions (e.g. Kalamata, Aegion, MNSA). In stronger motions (e.g. Yarimca, TCU-068, Rinaldi-
318), it falls within damage control or (barely) survival. Finally, for even stronger clearly ex-
ceeding the design limits earthquake shaking (e.g. Takatori-000, TCU-068, Jensen-022) fail-
ure is unavoidable. In fact, in some cases the ductility demand is an order of magnitude larger
than capacity. In refreshing contrast, the unconservative system designed according to the
new philosophy never comes close to its displacement ductility capacity (Figure 10a2) :
demand
/
capacity
is systematically lower than 0.25 for all seismic motions. Evidently, the new design con-
cept appears to provide much larger safety margins.
The performance of the new design concept is also slightly superior in terms of residual deck
drift (Fig. 10b), especially for large intensity earthquakes. The conventional design is superior
in terms residual only for small earthquakes, in which both superstructure and foundation re-
main completely elastic. Figure 10c compares the settlement w of the two alternatives after the
end of the earthquake. Evidently, the new design scheme is subject to larger settlement for all
seismic motions : w is roughly 3 times larger than for the conventionally designed system. How-
ever, even in the worst-case scenarios, w barely exceeds 0.2 m.
In conclusion :
For moderate intensity seismic motions not exceeding the design limits, the performance of
both alternatives is totally acceptable : both of them would be utilizable right after the earth-
quake, with only minor repair required. Sustaining limited structural damage (in the form of
minor flexural cracking), the conventionally designed system would be easily repairable. On
the other hand, the system designed according to the new philosophy would not sustain any
structural damage, but would be subjected to slightly increased but absolutely tolerable
deck drift and settlement.
For large intensity seismic motions that clearly exceed the design limits, the performance of
the system designed according to the new design philosophy is quite advantageous: while the
conventional system may collapse (as was the case with the Fukae bridge in Kobe), or at least
sustain severe (non-repairable) structural damage, the new design would survive with the
damage being in the form of increased settlements. Whether the bridge would be repairable
after such an earthquake depends on how settlement tolerant the design of its superstructure
is. In any case, preservation of human life through avoidance of collapse is the main design
objective against this type of extreme loading, and although it might be early to over-
generalize, the new design philosophy seems to have a potential for significantly larger safety
margins.
ACKNOWLEDGEMENT
th
Research Council Programme Ideas, Support for Frontier Research Advanced Grant. Con-
tract number ERC-2008-AdG 228254-DARE.

404
1 0.2 0
0
1
10
0.8 0.6 0.4
d
e
m
a
n
d
/

a
v
a
i
l
a
b
l
e
a
E
(g)
1 0.2 0
0
0.2
0.4
0.8 0.6 0.4
d
e
m
a
n
d
/

a
v
a
i
l
a
b
l
e
a
E
(g)
No damage Serviceability limit Damage control Survival Failure Limit States :
1 0.2 0
0
1
2
0.8 0.6 0.4
r
e
s
i
d
u
a
l
(
m
)
a
E
(g)
1.5
0.5
1 0.2 0
0
1
2
0.8 0.6 0.4
r
e
s
i
d
u
a
l
(
m
)
a
E
(g)
1.5
0.5
1 0.2 0
0
0.2
0.3
0.8 0.6 0.4
w
(
m
)
a
E
(g)
0.1
1 0.2 0
0
0.2
0.3
0.8 0.6 0.4
w
(
m
)
a
E
(g)
0.1
JMA000 JMA090 Takarazuka Takatori000
Takatori090 TCU052_ew TCU052_ns TCU068_ew
TCU068_ns Yarimca060 Yarimca330 Sakarya
Duze Bolu000 Duze Bolu090 Jensen022 Jensen292
Rinaldi228 Rinaldi318 Sylmar090 Sylmar360
Imp. Val. #04140 Imp. Val. #04230 Lefkada1973 Lefkada2003
MNSA Aegion Pyrgos Kalamata
Lucerne
(a
2
)
(b
1
) (b
2
)
(c
1
) (c
2
)
(a
1
)
1 0.2 0
0
1
10
0.8 0.6 0.4
d
e
m
a
n
d
/

a
v
a
i
l
a
b
l
e
a
E
(g)
1 0.2 0
0
0.2
0.4
0.8 0.6 0.4
d
e
m
a
n
d
/

a
v
a
i
l
a
b
l
e
a
E
(g)
No damage Serviceability limit Damage control Survival Failure Limit States :
1 0.2 0
0
1
2
0.8 0.6 0.4
r
e
s
i
d
u
a
l
(
m
)
a
E
(g)
1.5
0.5
1 0.2 0
0
1
2
0.8 0.6 0.4
r
e
s
i
d
u
a
l
(
m
)
a
E
(g)
1.5
0.5
1 0.2 0
0
0.2
0.3
0.8 0.6 0.4
w
(
m
)
a
E
(g)
0.1
1 0.2 0
0
0.2
0.3
0.8 0.6 0.4
w
(
m
)
a
E
(g)
0.1
1 0.2 0
0
0.2
0.3
0.8 0.6 0.4
w
(
m
)
a
E
(g)
0.1
1 0.2 0
0
0.2
0.3
0.8 0.6 0.4
w
(
m
)
a
E
(g)
0.1
Lucerne
Lucerne
(a
2
)
(b
1
) (b
2
)
(c
1
) (c
2
)
(a
1
)

Figure 10. Synopsis of the response of the two alternatives with respect to peak ground acceleration a
E
.
(a
1
and a
2
) Ratio of displacement ductility demand over ductility capacity. For the conventional design,
we also indicate the damage level with reference to Response Limit States [Priestley et al., 1996]: while
for earthquakes not exceeding the design limits the bridge would survive with some damage (ranging
from the serviceability to the survival limit state), it would probably collapse for several earthquakes
that exceed the design. In some cases, the ductility demand is an order of magnitude larger than ductility
capacity. (b
1
and b
2
) Residual deck drift . For earthquakes not exceeding the design, the residual of
the two systems is comparable. The new concept is clearly advantageous for earthquakes that exceed the
design limits. (c
1
and c
2
) Settlement w after the end of the earthquake. The new concept does suffer from
larger settlement. However, only in the veryworst-case scenarios, does w barely exceed 0.2 m. Whether
and under which conditions such a w can be tolerable will depend on the serviceability limits of the
superstructure. In any case, the new design concept may provide larger safety limits, trading-off struc-
tural damage (or collapse) with increased settlement.

405
REFERENCES
Apostolou, M., Gazetas G. (2005), Rocking of Foundations under Strong Shaking : Mobilisation of Bear-
ing Capacity and Displacement Demands, Proc. 1
st
GreeceJapan Workshop, Seismic Design, Obser-
vation, Retrofit, Athens 11-12 October, pp. 131-140.
Bertero V. (1996), State of the art report on: design criteria. Proc. of 11
th
World Conference on Earth-
quake Engineering, Acapulco, Mexico. Oxford: Pergamon.
Earthquake Engineering Research Institute (1995), The Hyogo-Ken Nanbu earthquake, January 17,
1995. Preliminary EERI Reconnaissance Report.
EAK (2000), Greek Seismic Code, Organization of Seismic Planning and Protection, Athens (in Greek).
EC8 (2000), Design Provisions for Earthquake Resistance of Structures, Part 5 : Foundations, retaining
structures and geotechnical aspects, prEN, 1998-5 European Committee for Standardization, Brus-
sels.
(2000), Greek Code for Reinforced Concrete, Organization of Seismic Planning and Protection,
Athens (in Greek).
Esmaeily-Gh. A., Xiao Y. (2002), Seismic Behavior of Bridge Columns Subjected to Various Loading
Patterns, PEER Report 2002/15, Pacific Earthquake Engineering Research Center, College of Engi-
neering, University of California, Berkeley.
Faccioli, E., Paolucci, R., and Vivero, G. (2001), Investigation of seismic soil-footing interaction by large
scale cyclic tests and analytical models, Proc., 4th Int. Conf. Recent Advances in Geotechnical Earth-
quake Engineering and Soil Dynamics.
FEMA 356 (2000), Prestandard and Commentary for the Seismic Rehabilitation of Buildings.
Fukushima, Y., Irikura, K., Uetake, T., and Matsumoto H. (2000), Characteristics of observed peak am-
plitude for strong ground motion from the 1995 Hyogoken Nanbu (Kobe) earthquake, Bulletin of the
Seismological Society of America, Vol. 90, pp. 545565.
Gazetas, G., Dakoulas, P., and Papageorgiou, A. S. (1990), Local soil and source mechanism effects in
the 1986 Kalamata (Greece) earthquake, Earthquake Engineering & Structural Dynamics, 19, pp.
431456.
Gerolymos, N., Gazetas, G., Tazoh, T., (2005), Seismic Response of Yielding Pile in Non-Linear Soil,
st
Greece Japan Workshop, Seismic Design, Observation, Retrofit, pp. 25 36,
Athens 11-12 October.
Ishibashi,I. and Zhang, X. (1993), Unified dynamic shear moduli and damping ratios of sand and clay,
Soils and Foundations, Vol. 33(1), pp. 12-191.
Iwasaki T., chm, et al. (1995), Report on Highway bridge damage caused by the Hyogoken Nanbu Earth-
quake of 1995, Committee on Highway Bridge Damage, Japan.
Kawashima K., Nagai T., Sakellaraki D. (2007), Rocking Seismic Isolation of Bridges Supported by
Spread Foundations, Proc. 2
nd
Japan-Greece Workshop on Seismic Design, Observation, and Retrofit
of Foundations, April 3-4, Tokyo, Japan, pp. 254265.
Kutter BL, Martin G, Hutchinson TC, Harden C, Gajan S, Phalen JD. (2003), Status report on study of
modeling of nonlinear cyclic load-deformation behavior of shallow foundations, PEER Workshop,
University of California, Davis, March 2003.
Mander J.B., Priestley M.J.N., Park R., (1988), Theoretical Stress Strain Model for Confined Concrete,
ASCE Journal of Structural Journal, Vol. 114, No. 8, pp. 1804-1825.
Martin, G., R., and Lam, I. P. (2000). Earthquake Resistant Design of Foundations : Retrofit of Existing
Foundations, Proc. GeoEng 2000 Conference, Melbourne.
Park, R. (1996), An Analysis of the failure of the 630 m elevated expressway in Great Hanshin Earth-
quake, Bulletin of the New Zealand National Society for Earthquake Engineering, 29, 2
Pecker, A. (1998), Capacity Design Principles for Shallow Foundations in Seismic Areas, Proc. 11
th

European Conference on Earthquake Engineering, A.A. Balkema Publishing.
Pecker, A. (2003), Aseismic foundation design process, lessons learned from two major projects: the
Vasco de Gama and the Rion Antirion bridges, ACI International Conference on Seismic Bridge De-
sign and Retrofit, La Jolla.
Priestley, M.J.N., Seible, F. and Calvi, G.M. (1996), Seismic Design and Retrofit of Bridges, John Wiley
and Sons, New York.
Priestley, M.J.N. (2000), Performance based seismic design, Proc. 12
th
World Conference on Earthquake
Engineering (12WCEE), Auckland, New Zealand, Paper No. 2831.
Seible, F., Priestley, M.J.N. and MacRae, G. (1995), The Kobe earthquake of January 17, 1995; initial
impressions from a quick reconnaissance, Structural Systems Research Report-95/03, University of
California, San Diego, 1995.
Tassios, T.P. (1998), Seismic design: state of practice, Proceedings of 11
th
European Conference on
Earthquake Engineering, Rotterdam: AA Balkema, pp. 255267.
406
1 INTRODUCTION
It has been several decades since the realization that the increase of the strength of a structure
does not always enhance safety. Accepting that failure of some structural members cannot
always be avoided during strong seismic shaking, earthquake engineering research focused on
ensuring : (a) that structural members can sustain dynamic loads that exceed their strength
without collapsingductility design, (b) that failure is guided to members that are less
important for the overall integrity of the structure (i.e. beams instead of columns), and (c) that
failure is in the form of non-brittle mechanisms (bending instead of shear failure) capacity
design [Park & Paulay, 1976].
With capacity design principles mainly referring to the superstructure, the effect of soil and
foundation is usually underestimated. In the words of Priestley [2000] the incorporation of
foundation compliance effects into force-based design is generally carried out inadequately, if
at all. Even when foundation compliance is taken into account, little care is given to the
nonlinearity of soil and foundation. In fact, current practice in seismic foundation design (e.g.
EC8), attempts to avoid at all costs the mobilization of strength in the foundation. In
structural terminology : no plastic hinging is allowed in the foundationsoil system. In simple
geotechnical terms, the designer must ensure that the foundation system will not even reach a
number of thresholds that would conventionally imply failure. Thus, the following states are
prohibited :
(a) mobilization of the bearing-capacity failure mechanisms under cyclically-uplifting
shallow foundations ;
(b) sliding at the soilfooting interface or excessive uplifting of a shallow foundation ;
(c) passive and shear failure along the sides and base of an embedded foundation ;
Towards a Reversal of Seismic Capacity Design : Part B.
Shaking-Table Testing of Bridge PierFoundation System
I. Anastasopoulos, T. Georgarakos, V. Drosos, S. Giannakos, G. Gazetas
ABSTRACT : This paper investigates experimentally the effectiveness of a new seismic design
philosophy, in which soil failure is utilized to protect the superstructure. A physical model of
a simple bridge pier is used as an example. Two alternatives are considered: one in compliance
with conventional capacity design, with over-designed foundation so that the plastic hinge
will develop at the bridge pier ; and one following the new philosophy, with under-designed
foundation, moving the plastic hinging will develop in the soil or the soil foundation interface.
The seismic performance of the two alternatives is investigated through 1-G shaking table test-
ing using real records and synthetic motions as base excitation. It is shown that the performance
of the new design concept can be advantageous : in cases where the conventionally designed
system collapses, the new concept can survive the seismic motion with the damage being lim-
ited to residual deck drift and increased settlement.
407
Overstrength factors and factors of safety larger than 1 are introduced against each of the
above failure modes, as in static design.
Although such a restriction may appear reasonable (since the inspection and rehabilitation of
foundation damage after a strong earthquake is not easy), neglecting such phenomena prohibits
the exploitation of strongly non-linear energy dissipating mechanisms in defence of the
superstructure in case of occurrence of ground motions larger than design. A growing body of
evidence suggests that soilfoundation plastic yielding under seismic excitation is not only
unavoidable, but may even be beneficial [Paolucci, 1997; Pecker, 1998; Martin & Lam, 2000;
Kutter et al. 2001; Faccioli et al., 2001; Gazetas et al., 2003; Gajan et al., 2005; Apostolou &
Gazetas,2005; Kawashima et al., 2007; Gajan & Kutter, 2008; Chatzigogos et al., 2009].
The need for such a reversal of current seismic design stems from :
The uncertainty of predicting the maximum credible earthquake and determining the
characteristics of the corresponding seismic motion (PGA, PGV, frequency content,
duration, details). For example, the notorious 1995 M
w
7.2 Kobe earthquake was generated
by an unknown fault, generating PGAs of up to 0.85 g, compared to 0.3 g of the design code
[e.g. Gazetas et al., 2005]. In fact, in each new earthquake larger PGAs are recorded. A
recent example is the long-awaited 2004 M
w
6.0 Parkfield earthquake, where the maximum
recorded PGA at close proximity to the seismogenic fault reached 1.8 g, accompanied by
PGVs of the order of 100 cm/s [Shakal et al., 2006]. Such observations lead to the
conclusion that the probability of occurrence of such large near-fault PGAs can be
substantial. Thus, the challenge of defining upper bounds on earthquake ground motions
[Bommer et al., 2004] can be seen from a different perspective. Therefore, it is considered
logical to accept that the risk of occurrence of seismic ground motions larger than assumed
in design will always be substantial. It is therefore important to develop new design methods
that will allow structures to withstand earthquakes larger than assumed in design without
collapsing or sustaining un-reparable damage.
The necessity of developing economically efficient and environmentally-friendly earthquake
protection solutions. The era of global economic crisis urgently calls for a drastic reappraisal
of our way of thinking. Seismic safety and protection of human life is and must remain
the first priority. However, a typical structure will have to withstand a strong earthquake
only once or twice in its life. Hence, economy and respect to the environment should also
play a role in the design process. So, instead of building larger and stronger (more
expensive) foundations to make sure that strong seismic shaking will manage to get to the
superstructure (i.e. conventional capacity design), and then reinforce the superstructure so
that it may withstand the earthquake without collapsing (making it also more expensive and
consuming more-and-more material resources), why not do exactly the opposite :
intentionally under-design the foundations to act as safety valves, limiting the acceleration
transmitted onto the superstructure. This way, we may achieve economy in the foundation
and the superstructure, without undermining safety. In fact, as it will be shown in the sequel,
due to the substantially larger ductility capacity of soil failure mechanisms compared to
structural yielding, the new design philosophy may provide increased safety margins.
This paper investigates experimentally the potential effectiveness of a new seismic design
philosophy, in which yielding of the soilfoundation system is utilized to protect the
superstructure i.e. exactly the opposite of conventional capacity design. The difference
between conventional design and the new philosophy is schematically illustrated in Figure 1. A
simple but realistic physical model of a bridge was constructed and tested in the shaking table of
the Laboratory of Soil Mechanics (of NTUA). The results presented herein can be seen as a first
experimental proof of the potential advantages of the new concept. To become applicable in
practice, the new design philosophy will have to be extensively verified analytically and
experimentally (shaking table and centrifuge testing), something which is the scope of the EU-
funded project DARE (SoilFoundationStructure Systems Beyond Conventional Seismic
Failure Thresholds).

408
Plastic hinging at
the superstructue
B
Avoidance of
foundation failure
B
Plastic hinging
at the foundation
Avoidance of
structural failure
Plastic hinging at
the superstructue
B
Avoidance of
foundation failure
B
Plastic hinging
at the foundation
Avoidance of
structural failure
h = 12 m
15 m
M
deck
=1200 Mg
d =3 m
D
r
= 85%
Dense Sand
B = 11 m B = 7 m
h = 12 m
15 m
M
deck
=1200 Mg
d =3 m
D
r
= 85%
Dense Sand
B = 11 m B = 7 m

Figure 1. Schematic illustration of the difference between conventional capacity design (plastic hinging
in the superstructure) with the new design philosophy (plastic hinging at the foundation soil).

2 ALTERNATIVE DESIGN PROTOTYPES
As depicted in Figure 2, we consider a typical highway bridge excited in the transverse
direction. A deck of mass m = 1200 Mgr is monolithically connected to a reinforced concrete
pier of diameter d = 3 m and height H = 12 m. The bridge chosen for shaking table testing is the
same with the one analyzed with finite elements in Anastasopoulos et al. [2009], and is
intentionally quite similar to the Hanshin Expressway Fukae bridge, which collapsed
spectacularly in the Kobe 1995 earthquake [Seible et al., 1995; Iwasaki et al., 1995; Park, 1996].
The bridge prototype is designed in accordance to EC8 [2000] and the Greek Seismic Code
[EAK 2000] for design acceleration A = 0.24 g, considering a (ductility-based) behavior factor q
= 1.5. With an elastic fixed-base vibration period T = 0.48 sec, the design spectral acceleration
is SA = 0.4 g.
The pier is founded through a square foundation of width B on an idealized 15 m deep dense
sand layer, of relative density D
r
= 85%. Two different foundation widths are considered to
represent the two alternative design approaches : a larger foundation, B = 11 m, designed in
compliance with conventional capacity design, applying overstrength factor
Rd
= 1.4 to ensure
that the plastic hinge will develop in the superstructure (base of pier) ; and a smaller, under-
designed, B = 7 m foundation, in the spirit of the new design philosophy, applying an
understrength factor 1/
Rd
= 1/1.4 0.7 for seismic loading. Thus, the resulting safety factor
for seismic loading is lower than 1.0 (FSE 0.7). In fact, as it will be shown below, the
underdesigned foundation will not allow the design seismic action to develop. Hence, FSE does
not really have a physical meaning in this case ; it is just an apparent temporary factor of safety.

Figure 2. Considered prototype alternatives.
409
B =11 m
d =3 m 12 m
m
2 m
R
e
i
n
f
o
r
c
e
d

c
o
n
c
r
e
t
e
Model Prototype
0.53 sec 0.076 sec T
86286 kN/m 33.2 kN/m K
600 Mg 4.8 kg m
Prototype Model
0.53 sec 0.076 sec T
86286 kN/m 33.2 kN/m K
600 Mg 4.8 kg m
Prototype Model
6.3 cm 6.3 cm
7 cm
B =22 cm
steel
aluminum
7

c
m
7

c
m
1
1

c
m
5 mm
2 cm
24 cm
m
Artificial
Plastic
hinge
1:50 Scale

3 PHYSICAL MODEL CONFIGURATION
Physical models of the two bridge alternatives were constructed and tested at the Laboratory of
Soil Mechanics of the National Technical University of Athens (NTUA), utilizing a recently
installed ANCO R51 shaking table. The table, 1.3 m x 1.3 m in dimensions, is capable of
shaking specimens of up to 2 tons at accelerations of up to 1.6 g. Synthetic accelerograms, as
well as real earthquake records can be simulated. The actuator is equipped with a servo-valve,
controlled by an analog inner-loop control system and a digital outer-loop controller ; it is
capable of producing a stroke of 75 mm.
At this point, it should be noted that the stress field in the backfill soil cannot be correctly
reproduced in reducedscale shaking table modeling, and this is the main advantage of
centrifuge testing. However, a centrifuge is substantially more expensive both to acquire and to
operate, and is not yet available in NTUA. Shaking table testing can be seen as a valid option,
provided that the results are interpreted carefully, with due consideration of scale effects and the
stress dependent behaviour of soil.
Taking account of the capacity of the shaking table, a N = 50 scale factor was selected for the
experiments The selection of model materials was conducted taking account of scaling laws
[Gibson, 1997], so that the simulation is as realistic as possible for the given prototype. The
bridge piers were constructed using commercially available steel and aluminium plates, as
schematically illustrated in Figure 3 (for the conventionally designed alternative).

Figure 3. Basic dimensions and details of the physical model, compared to the corresponding prototype
(conventionally designed alternative ; half bridge modelled).

At small scale, it is practically impossible to model stiffness correctly (in consistency with
the scaling laws) and achieve the desired (scaled) ultimate bending moment capacity of the pier
at the same time. Hence, an artificial plastic hinge was custom designed and constructed, and
placed at the base of the pier of the conventionally designed alternative. As schematically
illustrated in Figure 4a, the ultimate bending moment M
ult
of the plastic hinge can be calibrated
through adjustment of the torque applied at the nut-bolt assembly. To achieve repeatability, two
the Teflon washers were added between the bolts and the central steel plate. The calibration of
the assembly was performed through static pushover testing, using the experimental
configuration of Figure 4b.

410
Artificial
PlasticHinge
Actuator
LoadCell
DAQ
Artificial
PlasticHinge
Actuator
LoadCell
DAQ
Teflon
washer
Nut-bolt
assembly
Teflon
washer
Nut-bolt
assembly
Foundation Foundation
Pier Pier
(a)
(b)
Artificial
PlasticHinge
Actuator
LoadCell
DAQ
Artificial
PlasticHinge
Actuator
LoadCell
DAQ
Teflon
washer
Nut-bolt
assembly
Teflon
washer
Nut-bolt
assembly
Foundation Foundation
Pier Pier
(a)
(b)
Sand container
A
d
j
u
s
t
a
b
l
e

h
e
i
g
h
t
A
d
ju
s
ta
b
le

v
e
lo
c
ity
Conventional Conventional
Design Design
New New
Philosophy Philosophy
Shaking Table Shaking Table
Laminar Box Laminar Box (a)
(b)
Sand container
A
d
j
u
s
t
a
b
l
e

h
e
i
g
h
t
A
d
ju
s
ta
b
le

v
e
lo
c
ity
Conventional Conventional
Design Design
New New
Philosophy Philosophy
Shaking Table Shaking Table
Laminar Box Laminar Box (a)
(b)

Figure 4. (a) Schematic illustration of the artificial plastic hinge, (b) experimental static pushover setup
used for calibration of the ultimate bending moment M
ult
of the plastic hinge.

The physical models were placed inside a transparent laminar box, custom designed and
constructed in NTUA (Figure 5a). The foundation soil consisted of dry Longstone (M34) sand,
a very fine industrially produced uniform quartz sand with D
50
= 0.15 mm, uniformity
coefficient D
60
/D
10
= 1.42, e
max
= 0.995, e
min
= 0.614, and G
s
= 2.64. The shaking table models
were prepared by raining the sand from a specific height with controllable mass flow rate
(which controls the density of the sand), using a custom raining system, designed and
constructed in NTUA (Figure 5b). For the maximum raining velocity and with the current
configuration (width of sand container opening), for Longstone sand the system is capable of
achieving relative densities D
r
ranging from about 10% to 85%. The tests were conducted at the
maximum density (i.e. D
r
= 85%).
Tests were conducted with two different configurations : (i) both bridge physical models
placed inside the laminar box (Figure 5a), and (ii) each model was placed and tested separately.
The purpose of the first configuration was to demonstrate the differences between the two
design alternatives under exactly the same conditions. The second comfiguration was used to
measure horizontal and vertical displacements of the pier and its foundation. In that case, four
accelerometers and four wire displacement transducers were installed, as illustrated in Figure 6.
Two accelerometers were placed inside the sand specimen during construction at predetermined
positions, to measure soil response at the free-field and under the foundation ; an additional
accelerometer was installed at deck level to measure the response of the bridge. Two wire
displacement transducers were installed in the horizontal direction to measure the drift of the
structure, and two in the vertical direction to measure foundation rotation.

Figure 5. (a) Experimental configuration : the two physical models placed inside the transparent laminar
box ; (b) the electronically controlled sand raining system used for preparation of the models.
411
10
a(g)
egion 1995
0.5
0
0.5
0 6 2 4 8
Kalamata 1986
0.5
0
0.5
0 20 5 10 15
a(g)
0.8
0.4
0
0.4
0.8
0 15 30 5 10 20 25
Lefkada 2003
t(sec)
1
0.5
0
0.5
1
0 20 5 10 15
Rinaldi (Northridge) 1994
t(sec)
Sin1Hz,PGA=0.4g
0 30 50 10 20 40 60
0.8
0.4
0
0.4
0.8
Sin2Hz,PGA=0.4g
0.8
0.4
0
0.4
0.8
0 15 25 5 10
20 30
a(g)
10
a(g)
egion 1995
0.5
0
0.5
0 6 2 4 8
a(g)
egion 1995
0.5
0
0.5
0 6 2 4 8
Kalamata 1986
0.5
0
0.5
0 20 5 10 15
a(g)
0.8
0.4
0
0.4
0.8
0 15 30 5 10 20 25
Lefkada 2003
0.8
0.4
0
0.4
0.8
0 15 30 5 10 20 25
Lefkada 2003
t(sec)
1
0.5
0
0.5
1
0 20 5 10 15
Rinaldi (Northridge) 1994
t(sec)
Sin1Hz,PGA=0.4g
0 30 50 10 20 40 60
0.8
0.4
0
0.4
0.8
Sin2Hz,PGA=0.4g
0.8
0.4
0
0.4
0.8
0 15 25 5 10
20 30
a(g)
WT-68
WT-69
WT-71 WT-70
S-700
S-702
D-108
S-708 laminates laminates
Rigid frame
Base accelerometer
Accelerometer
Wire-DT
WT-68
WT-69
WT-71 WT-70
S-700
S-702
D-108
S-708 laminates laminates
Rigid frame
Base accelerometer
Accelerometer
Wire-DT
Base accelerometer Base accelerometer
Accelerometer Accelerometer
Wire-DT Wire-DT

Figure 6. Sketch showing the instrumentation of the shaking table tests using the second configuration.
4 SEISMIC EXCITATION
As depicted in Figure 7, the shaking table tests were conducted using four real records and two
artificial accelerograms as seismic excitations. More specifically, we selected the Aegion 1995
and Kalamata 1986 records as representative of moderate intensity earthquakes, Lefkada 2003
as representative of moderate intensity earthquakes but with a large number of strong-motion
cycles, and Rinaldi (Northridge 1994) as representative of large magnitude earthquakes. The
two artificial 30-cycle sinusoidal motions were used to investigate the performance of the two
alternatives in extreme events. Flingtype pulses were also utilized as seismic excitation, but the
results are not shown here due to space limitations.

Figure 7. Real and artificial accelerograms utilized as seismic excitation in the experiments.
412
-0.3
-0.2
-0.1
0
0.1
0 5 10 15 20 25 30
-0.3
-0.2
-0.1
0
0.1
0 5 10 15 20 25 30
a(g)
Conventional Design NewDesignPhilosophy NewDesignPhilosophy
0.8
0.4
0
0.4
0.8
t(sec)
0.8
0.4
0
0.4
0.8
t(sec)
(m)
Deck
Deck
-0.08
-0.04
0
0.04
0 5 10 15 20 25 30
w
-0.08
-0.04
0
0.04
0 5 10 15 20 25 30
t(sec)
w(m)
c

r
-0.3
-0.2
-0.1
0
0.1
0 5 10 15 20 25 30
-0.3
-0.2
-0.1
0
0.1
0 5 10 15 20 25 30
a(g)
0.8
0.4
0
0.4
0.8
t(sec)
0.8
0.4
0
0.4
0.8
t(sec)
(m)
Deck
Deck
-0.08
-0.04
0
0.04
0 5 10 15 20 25 30
ww
-0.08
-0.04
0
0.04
0 5 10 15 20 25 30
t(sec)
w(m)
c

r
5 PERFORMANCE IN MODERATE INTENSITY SEISMIC SHAKING
We first compare the performance of the two alternatives subjected to a seismic excitation of
moderate intensity. As such an example, we choose the record of the 2003 Ms 6.4 Lefkada
(Greece) earthquake [Gazetas et al 2004, 2005, Karakostas et al., 2004].
The comparison is portrayed in Figure 8 in terms of deck acceleration a, deck drift (i.e.
horizontal displacement), and foundation settlement w. All results are shown in prototype scale.
A first conclusion is that the conventionally designed system experiences larger deck
acceleration : approximately 0.4 g instead of roughly 0.3 g. In the first case, the acceleration is
limited by the ultimate bending moment of the plastic hinge ; in the latter case by the ultimate
capacity of the under-designed foundation.
As schematically illustrated in the sketch notation, the drift has two components [see also
Priestley et al., 1996] : (i) the flexural drift
C
, i.e. the structural displacement due to flexural
distortion of the pier column, and (ii) the rocking drift
r
= H, i.e. the displacement due to
rocking motion of the foundation. This way, the contribution of pier flexural distortion and
foundation rotation to the final result of interest (i.e. the total drift ) can be inferred. As shown
in the figure, while for the conventional design is mainly due to pier distortion
C
, the
opposite can be observed for the under-designed foundation : is mainly due to foundation
rotation
r
. Despite the differences in the mechanism leading to the development of (pier
distortion versus foundation rotation), the total drift is practically the same, and tolerable in both
cases.
In terms of foundation settlement w, the conventionally designed system is subjected to
limited settlement w 1.5 cm ; the new concept experiences larger, but tolerable dynamic
settlement : w 6 cm.

Figure 8. Comparison of the response of the two systems subjected to the Lefkada 2003 seismic excitation:
acceleration time histories at deck level, deck drift time histories (total, flexural, and rotational), and
foundation settlement time histories.
413
-1
-0.5
0
0.5
1
0 5 10 15 20
-1
-0.5
0
0.5
1
0 5 10 15 20
a(g)
t(sec)
1
0.5
0
0.5
1
0 20 5 10 15
1
0.5
0
0.5
1
Deck
Deck
(m)
t(sec)
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0 5 10 15 20
w
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0 5 10 15 20
w(m)
-1
-0.5
0
0.5
1
0 5 10 15 20
-1
-0.5
0
0.5
1
0 5 10 15 20
a(g)
t(sec)
1
0.5
0
0.5
1
0 20 5 10 15
1
0.5
0
0.5
1
Deck
Deck
(m)
t(sec)
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0 5 10 15 20
ww
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
0.08
0 5 10 15 20
w(m)
6 PERFORMANCE IN LARGE INTENSITY SEISMIC SHAKING
We now use the Rinaldi record of the 1994 Northridge M
s
6.8 earthquake [Trifunac et al., 1998]
as an example of large intensity seismic shaking. With PGA = 0.79 g and PGV = 164 cm/s, and
a very strong forward rupture directivity pulse, this record can be seen to constitute a very
severe seismic motion.
Figure 9 synopsizes the comparison in terms of deck acceleration a, deck drift , and founda-
tion settlement w. As for the previous case, the conventionally designed system experiences lar-
ger deck acceleration compared to the new design concept.
In terms of deck drift, as for the Lefkada 2003 seismic excitation, is almost purely related
to flexural pier drift
C
for the conventional design. The opposite is observed for the new design
concept : is mainly due to foundation rotation
r
. In contrast to the previous case, the
conventionally designed bridge is subjected to a maximum drift of about 65 cm (i.e. more than
5% of the pier height) and a residual drift of roughly 35 cm (i.e. about 3% of the pier height). In
reality, such flexural distortion would probably imply collapse or very serious damage of the
reinforced concrete pier. The reason why no collapse was observed in the experiment is none
other than the ductility capacity of the artificial plastic hinge, which is substantially larger than
the capacity of a reinforced concrete section.
Markedly different is the performance of the new design concept. The maximum drift is not
exceeding 35 cm (i.e. roughly 3% of the pier height), and the residual drift is limited to less than
13 cm (i.e. about 1% of the pier height). Most importantly, this drift is not associated with pier
structural damage but with inelasticity of the soil supporting the foundation. Naturally, the
increase of settlement is the price to pay : w 3.5 cm, instead of less than 2 cm of the
conventionally designed system.

Figure 9. Comparison of the response of the two systems subjected to the Rinaldi (Northridge) seismic
excitation : acceleration time histories at deck level, deck drift time histories (total, flexural, and rotational),
and foundation settlement time histories.
414
-0.16
-0.12
-0.08
-0.04
0
0.04
0 5 10 15 20 25 30
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 5 10 15 20 25 30
a(g)
(m)
w(m)
0.8
0.4
0
0.4
0.8
0.8
0.4
0
0.4
0.8
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 5 10 15 20 25 30
t(sec)
Deck
Deck
w
t(sec)
-0.16
-0.12
-0.08
-0.04
0
0.04
0 5 10 15 20 25 30
-0.16
-0.12
-0.08
-0.04
0
0.04
0 5 10 15 20 25 30
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 5 10 15 20 25 30
a(g)
(m)
w(m)
0.8
0.4
0
0.4
0.8
0.8
0.4
0
0.4
0.8
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 5 10 15 20 25 30
t(sec)
Deck
Deck
ww
t(sec)
-0.16
-0.12
-0.08
-0.04
0
0.04
0 5 10 15 20 25 30
7 PERFORMANCE UNDER EXTREME SHAKING
We now compare the response of the two systems subjected to the artificial 30-cycle sinusoidal
motions, aiming to investigate their performance in extreme events.
7.1 30-cycle sinus at 2 Hz
Figure 10 compares the response of the two alternatives in terms of deck acceleration and drift,
and foundation settlement.
The comparative performance of the two alternatives is qualitatively similar, with their main
differences being more pronounced due to the large number of strong motion cycles. In
particular, observe the accumulation of deck drift for both alternatives. In the case of the
conventionally designed system, is mostly due to flexural pier drift
C
, but some rotational
drift
r
is also accumulated. As for the previous cases, exactly the opposite is observed for the
new design concept. The conventionally designed bridge is subjected to a residual drift of about
90 cm (i.e. about 7.5% of the pier height). Obviously, in reality such flexural distortion would
imply collapse of the bridge, and (as for Rinaldi) the survival in the experiment is only due to
the unrealistically large ductility capacity of the artificial plastic hinge.
The accumulation of drift in the new design concept is much smaller : the residual drift does
not exceed 20 cm (i.e. less than 2% of the pier height). Exactly the opposite conclusion can be
drawn for the settlement : w 14 cm for the new concept versus 8 cm of the conventionally
designed system.

Figure 10. Comparison of the response of the two systems subjected to 30 cycle 2 Hz sinusoidal seismic
excitation of PGA = 0.4 g : acceleration time histories at deck level, deck drift time histories (total,
flexural, and rotational), and foundation settlement time histories.
415
1
0.5
0
0.5
1
collapse collapse
1
0.5
0
0.5
1
-8
-6
-4
-2
0
2
4
6
8
0 10 20 30 40 50 60
collapse collapse
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 10 20 30 40 50 60
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 10 20 30 40 50 60
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 5 10 15 20 25 30 35 40
a(g)
(m)
w(m)
t(sec)
Deck
Deck
w
t(sec)
collapse collapse
1
0.5
0
0.5
1
collapse collapse
1
0.5
0
0.5
1
1
0.5
0
0.5
1
-8
-6
-4
-2
0
2
4
6
8
0 10 20 30 40 50 60
collapse collapse
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 10 20 30 40 50 60
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 10 20 30 40 50 60
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
0
0.2
0 5 10 15 20 25 30 35 40
a(g)
(m)
w(m)
t(sec)
Deck
Deck
ww
t(sec)
collapse collapse
7.2 30-cycle sinus at 1 Hz
We now compare the response of the two alternatives subjected to a small frequency (f = 1 Hz)
seismic excitation. It is noted that such an excitation cannot be claimed to be representative of
any real earthquake. However, it is useful to demonstrate the substantially larger safety margins
of the system designed according to the new philosophy.
As shown in Figure 11, while the conventional system collapses at t = 23 sec (i.e. after 10
strong-motion cycles), the bridge with the under-designed foundation (new concept) survives
this tremendous seismic motion at the price of residual deck drift 50 cm and settlement
w 18 cm.
Before collapse, the performance of the two alternatives is qualitatively similar to the
previous cases : lager deck acceleration and predominantly associated with flexural pier drift
C
for the conventional system ; larger settlement w and mostly due to rotational drift
C
for
the new concept.

Figure 11. Comparison of the response of the two systems subjected to 30 cycle 1 Hz sinusoidal seismic
excitation of PGA = 0.4 g : acceleration time histories at deck level, deck drift time histories (total,
flexural, and rotational), and foundation settlement time histories.

8 CONCLUSIONS
The purpose of this paper was to investigate experimentally the effectiveness of a new seismic
design philosophy, in which soil failure is utilized to protect the superstructure. The seismic
performance of the new concept (involving an under-designed foundation) has been compared
to a conventionally designed bridge (with over-designed foundation).
416
A simple but realistic physical model of a bridge was constructed and tested in the shaking
table of the Laboratory of Soil Mechanics (of NTUA). The results presented herein can be seen
as a first experimental proof of the potential advantages of this new concept. An artificial plastic
hinge was custom designed and constructed, to model the nonlinear response of the bridge pier
at small scale.
For moderate intensity seismic motions, the performance of both alternatives is totally
acceptable, and both would be utilizable right after such an earthquake. The conventionally
designed bridge would probably sustain limited structural damage (minor flexural cracking),
and would be easily repairable. On the other hand, the new concept would be subjected to
slightly increased but tolerable deck drift and settlement, but would remain structurally
unscathed.
The advantage of the new seismic design philosophy becomes clear for large intensity
seismic motions, clearly exceeding the limits of the design. In such cases, while the convention-
ally designed system is driven to collapse, the new concept may survive the seismic motion with
the damage being contained in the form of deck drift and settlement.
The key conclusions of this work are in line with the results of the analytical work (referring
to the same bridge, under different soil conditions) presented in Anastasopoulos et al. [2009].
ACKNOWLEDGEMENT
th
Framework research project funded through the
European Research Councils Programme Ideas, Support for Frontier Research
Advanced Grant. Contract number ERC-2008-AdG 228254-DARE.
REFERENCES
Anastasopoulos, I., Loli M., Gazetas G. (2009), Towards a Reversal of Seismic Capacity Design I :
Analysis of Bridge Pier-Foundation System, Proc. 3
rd
GreeceJapan Workshop, Seismic Design,
Observation, Retrofit, 22-23 September, Santorini, Greece (companion paper)
Apostolou, M., Gazetas G. (2005), Rocking of Foundations under Strong Shaking : Mobilisation of
Bearing Capacity and Displacement Demands, Proc. 1
st
GreeceJapan Workshop, Seismic Design,
Observation, Retrofit, Athens 11-12 October, pp. 131-140.
Bertero V. (1996), State of the art report on: design criteria. Proc. of 11
th
World Conference on
Earthquake Engineering, Acapulco, Mexico. Oxford: Pergamon.
Earthquake Engineering Research Institute (1995), The Hyogo-Ken Nanbu earthquake, January 17, 1995.
Preliminary EERI Reconnaissance Report.
EAK (2000), Greek Seismic Code, Organization of Seismic Planning and Protection, Athens (in Greek).
EC8 (2000), Design Provisions for Earthquake Resistance of Structures, Part 5 : Foundations, retaining
structures and geotechnical aspects, prEN, 1998-5 European Committee for Standardization,
Brussels.
(2000), Greek Code for Reinforced Concrete, Organization of Seismic Planning and Protection,
Athens (in Greek).
Esmaeily-Gh. A., Xiao Y. (2002), Seismic Behavior of Bridge Columns Subjected to Various Loading
Patterns, PEER Report 2002/15, Pacific Earthquake Engineering Research Center, College of
Engineering, University of California, Berkeley.
Faccioli, E., Paolucci, R., and Vivero, G. (2001), Investigation of seismic soil-footing interaction by large
scale cyclic tests and analytical models, Proc., 4th Int. Conf. Recent Advances in Geotechnical
Earthquake Engineering and Soil Dynamics.
FEMA 356 (2000), Prestandard and Commentary for the Seismic Rehabilitation of Buildings.
Fukushima, Y., Irikura, K., Uetake, T., and Matsumoto H. (2000), Characteristics of observed peak
amplitude for strong ground motion from the 1995 Hyogoken Nanbu (Kobe) earthquake, Bulletin of
the Seismological Society of America, Vol. 90, pp. 545565.
Gazetas, G., Dakoulas, P., and Papageorgiou, A. S. (1990), Local soil and source mechanism effects in
the 1986 Kalamata (Greece) earthquake, Earthquake Engineering & Structural Dynamics, 19, pp.
431456.
Gazetas G., Anastasopoulos I., and Dakoulas P. (2005), Failure of Harbor Quaywalls in the Lefkada 2003
earthquake, Proceedings of Geotechnical Earthquake Engineering Satellite Conference
417
Performance Based Design in Earthquake Geotechnical Engineering : Concepts and Research,
Osaka, Japan, pp. 62-69.
Gazetas G., (2004), Geotechnical aspects of the Ms 6.4 Lefkada Island, Greece 2003 Earthquake :
Preliminary assessment, Proceedings of Fifth International Conference on Case Histories in
Geotechnical Engineering, New York, April 13-17 (in cd-rom)
Gazetas G., (2001), SSI issues in two European projects and in a recent earthquake, Proceedings of the
2nd Japan-U.S. Workshop on Soil-Structure Interaction, Tsukuba, Japan, 7-9 March.
Gerolymos, N., Gazetas, G., Tazoh, T., (2005), Seismic Response of Yielding Pile in Non-Linear Soil,
st
Greece Japan Workshop, Seismic Design, Observation, Retrofit, pp. 25 36,
Athens 11-12 October.
Ishibashi,I. and Zhang, X. (1993), Unified dynamic shear moduli and damping ratios of sand and clay,
Soils and Foundations, Vol. 33(1), pp. 12-191.
Iwasaki T., chm, et al. (1995), Report on Highway bridge damage caused by the Hyogoken Nanbu
Earthquake of 1995, Committee on Highway Bridge Damage, Japan.
Karakostas, V.G., Papadimitriou E.E., and Papazachos C.B. (2004), Properties of the 2003 Lefkada,
Ionian Islands, Greece, earthquake seismic sequence and seismicity triggering, Bulletin of the
Seismological Society of America, 94, pp. 1976 1981.
Kawashima K., Nagai T., Sakellaraki D. (2007), Rocking Seismic Isolation of Bridges Supported by
Spread Foundations, Proc. 2
nd
Japan-Greece Workshop on Seismic Design, Observation, and Retrofit
of Foundations, April 3-4, Tokyo, Japan, pp. 254265.
Kutter BL, Martin G, Hutchinson TC, Harden C, Gajan S, Phalen JD. (2003), Status report on study of
modeling of nonlinear cyclic load-deformation behavior of shallow foundations, PEER Workshop,
University of California, Davis, March 2003.
Mander J.B., Priestley M.J.N., Park R., (1988), Theoretical Stress Strain Model for Confined Concrete,
ASCE Journal of Structural Journal, Vol. 114, No. 8, pp. 1804-1825.
Martin, G., R., and Lam, I. P. (2000). Earthquake Resistant Design of Foundations : Retrofit of Existing
Foundations, Proc. GeoEng 2000 Conference, Melbourne.
Park, R. (1996), An Analysis of the failure of the 630 m elevated expressway in Great Hanshin
Earthquake, Bulletin of the New Zealand National Society for Earthquake Engineering, 29, 2
Park R & Pauley T. (1976), Reinforced Concrete Structures. John Wiley : NY.
Pecker, A. (1998), Capacity Design Principles for Shallow Foundations in Seismic Areas, Proc. 11
th

European Conference on Earthquake Engineering, A.A. Balkema Publishing.
Pecker, A. (2003), Aseismic foundation design process, lessons learned from two major projects: the
Vasco de Gama and the Rion Antirion bridges, ACI International Conference on Seismic Bridge
Design and Retrofit, La Jolla.
Priestley, M.J.N., Seible, F. and Calvi, G.M. (1996), Seismic Design and Retrofit of Bridges, John Wiley
and Sons, New York.
Priestley, M.J.N. (2000), Performance based seismic design, Proc. 12
th
World Conference on Earthquake
Engineering (12WCEE), Auckland, New Zealand, Paper No. 2831.
Seible, F., Priestley, M.J.N. and MacRae, G. (1995), The Kobe earthquake of January 17, 1995; initial
impressions from a quick reconnaissance, Structural Systems Research Report-95/03, University of
California, San Diego, 1995.
Tassios, T.P. (1998), Seismic design: state of practice, Proceedings of 11
th
European Conference on
Earthquake Engineering, Rotterdam: AA Balkema, pp. 255267.
Trifunac, M.D., Todorovska, M.I., and Lee V.W. (1998), The Rinaldi strong motion accelerogram of the
Northridge, California, earthquake of 17 January, 1994, Earthquake Spectra, 14(1), pp. 225-239.

418
1 INTRODUCTION
Strong seismic shaking constitutes the indirect result of the faulting (slip) process, and is typi-
cally experienced in an earthquake. The direct result of the faulting process, the quasi-static
fault offset, is less likely to be experienced in reality since the rupture does not always reach the
ground surface. Even when it does outcrop, it will only affect structures along or near the fault
trace. It is therefore understandable why earthquake engineering has mainly focused on ration-
ally facing strong seismic shaking, while over-conservative prohibitions were the rule for pro-
tecting against fault rupture [EC8, 1994].
However, evidence from recent large magnitude earthquakes such as the Kocaeli, Dzce-
Bolu, Chi-Chi, Wenchuan indicates that structures can survive large fault offsets (even of the
order of meters) if properly designed. An example of marginal survival from the Chi-Chi of a
2-story building resting on flexible foundation is illustrated in Figure 1. Although not particu-
larly well engineered, this building sustained about 3 m of reverse faulting (admittedly with a
smooth expression on the free field) with major damage but not collapse. Experimental and ana-
lytical studies have also confirmed that design against a direct hit by a rupturing fault is feasible
[Duncan & Lefebvre, 1973; Berrill, 1983; Anastasopoulos & Gazetas, 2007b;].
The distress of a foundationstructure system has been shown to depend on the interplay be-
tween the upward propagating rupture, the deforming soil, and the foundationstructure system.
This interdependence is called Fault RuptureSoilFoundationStructure Interaction, FR
SFSI, [Anastasopoulos & Gazetas, 2007a; 2007b].
Simplified Method for the Design of Raft Foundations against a
Direct Hit by Thrust Faulting

I. Anastasopoulos

ABSTRACT: This paper studies the response of raft foundations subjected to thrust faulting. A
parametric study is conducted utilizing a finite-element methodology, which has been exten-
sively validated against centrifuge model tests. It is shown that the stressing of the foundation,
and consequently of the superstructure, stems mainly from loss of soil support. It is shown that,
depending on the geometry, loss of support takes place either under the edges or under the mid-
dle of the foundation, generating hogging or sagging deformation, respectively. Even when the
rupture emerges beyond the structure, the raft may still experience substantial distress. Exploit-
ing the numerical results, a simplified method is developed for the design of foundation systems
against thrust faulting. The method calls for a conventional static analysis of a slab on Winkler
support, and simulates the fault rupture by removing the Winkler springs from equivalent areas
of loss of support ; the latter is estimated with the help of (simplified) design charts, further
facilitating the use of the method in practice.

419
Faultingimposedsmooth
grounddeformation
Faultingimposedsmooth
grounddeformation

Figure 1. Marginal survival of a 2-story building resting on a flexible foundation in the Chi-Chi, Taiwan,
1999 Earthquake.

In the absence of a structure (i.e. under freefield conditions), the rupture path and the magni-
tude of the surface fault scarp are functions of the fault style (normal, thrust, strike-slip) the
magnitude of the fault offset at bedrock, and the nature of the overlying soil cover (thikness, soil
stiffness). In the presence of a structure, the path of the fault rupture and the deformation of the
ground surface may be substantially modified. With rigid, continuous and, especially, heavily
loaded foundations, even complete diversion of the fault rupture is possible [Berrill, 1983; Dun-
can & Lefebvre, 1973; Anastasopoulos & Gazetas, 2007]. In addition, a heavy superstructure on
soft soil will push down the foundation, thereby compressing and flattening the faulting
induced asperities.
Recent integrated research efforts combining field studies, centrifuge experiments, and nu-
merical analysis under the auspices of a European Research Program [BEE 2008 Special Publi-
cation] have culminated in the development of a validated methodology for analysis of founda-
tionstructure systems against surface fault rupture. Our paper applies that methodology to
study the response of slab foundations subjected to thrust faulting.
2 SETTING OF THE PROBLEM AND ANALYSIS METHODOLOGY
We consider a soil deposit of depth H subjected to thrust faulting, which at the bedrock has a
vertical component of magnitude h and a dip angle = 60
o
(Figure 2). Assuming 2-D plane
strain conditions, the analysis is conducted in two steps. In the first step, we analyze the propa-
gation of the fault rupture in the freefield (i.e., in the absence of the structure). In the second
step, a raft foundation of width B with surcharge load q is positioned on top of the soil model at
distance s from the freefield fault outcrop (measured from the hanging-wall edge of the foun-
dation, as shown in the figure), and the FR-SFSI analysis is conducted.
The analysis is conducted with finite element modeling. To avoid parasitic boundary effects,
the total width of the model is B = 4H. As described in detail in Anastasopoulos et al. [2007;
2008; 2009], we adopt an elastoplastic constitutive model with Mohr-Coulomb failure criterion
and isotropic strain softening. The latter is achieved by reducing the mobilized friction
mob
and
dilation angle
mob
with increasing octahedral plastic shear strain :

420
s
q
B
Hanging wall
Footwall
h
E
z
s
q
B
Hanging wall
Footwall
h
E
z
s
q
B
q
B
Hanging wall
Footwall
h
E
z

Figure 2. Problem definition and model dimensions : Raft foundation of width B with surcharge load q,
subjected to thrust faulting at distance s .

1 for 0 <
0 for
P
p res P P P oct
p oct p oct f P P
f f
mob mob
P P
res res oct f

; ,

;
; ,

=

=

(1)
where :
p
and
p
the peak mobilized friction and dilation angles ;
res
and
res
= 0 their residual
values ; and
P
f
the octahedral plastic shear strain at the end of softening. Scale effects [Muir
Wood, 2002] are taken into account through an approximate simplified scaling method. Pre-
yielding behaviour is assumed to be elastic, with secant modulus G
s
=
y
/
y
linearly increasing
with depth. The procedure for constitutive model parameter calibration, based on direct shear
testing, can be found in Anastasopoulos et al. [2007; 2009].
The slab foundation is modeled with elastic beam elements, connected to the soil through
special contact elements. Being infinitely stiff in compression but tensionless, these elements al-
low for a realistic simulation of foundation detachment from the bearing soil ; their response in
shear obeys Coulombs friction law, allowing for slippage at the foundationsoil interface.
The finite element methodology employed herein has been extensively validated through
qualitative and semi-quantitative comparisons with experimental data, case histories and, most
importantly, through quantitative genuine (Class A) predictions of centrifuge model tests
[Anastasopoulos et al., 2007; 2008; 2009]. Figure 3 reproduces one such prediction, referring to
a rigid B = 10 m foundation with q = 90 kPa, subjected to thrust faulting through H = 15 m Fon-
tainebleau sand [Gaudin, 2002], at s = 9.2 m. The comparison is judged as quite satisfactory, es-
pecially in view of the fact that some of the discrepancies are attributed to a small leakage of
sand towards the glass window of the centrifuge container and hence are an experimental inac-
curacy. [Anastasopoulos et al., 2009].
3 SUMMARY OF RESULTS OF PARAMETRIC ANALYSIS
To derive deeper insights on the interaction of a thrust fault rupture with (2-D) slab foundations,
a parametric study is undertaken. Although the analysis is conducted for H = 20 m, the key re-
sults and conclusions are of more general validity. According to the principles of dimensional
analysis [e.g. Muir Wood, 2004], and as earlier suggested by Cole & Lade [1984] and Bray
[1990], the deformation field can be normalized with soil thickness H : i.e. the bedrock offset h
and the vertical displacement y can be written in non-dimensional form as h/H and y/H. Al-
though such normalization is not strictly accurate (due to scale effects), it constitutes a reason-
able approximation from an engineering point-of-view [Anastasopoulos et al., 2007]. Accord-
ingly, the surcharge load q and the foundation bending moment M are expressed in non-
dimensional form as q/gB and M/qB
2
, respectively [see also Berrill, 1983].
421

Figure 3. Class A prediction of centrifuge model test rigid B = 10 m foundation with surcharge load q
= 90 kPa, subjected to thrust = 60
faulting through H = 15 m Fontainebleau sand deposit, at distance s

= 9.2 m : (a) vertical displacement y at the soil surface (for h = 2.5 m), and (b) evolution of foundation
rotation with bedrock offset h.

The dip angle is taken equal to 60
o
(a typical value for dip-slip faults), and the foundation is
assumed practically rigid (EI = 108 kNm
2
, where E is the Youngs modulus of the foundation
and I its moment of inertia). The following parameters are parametrically investigated :
(a) The width of the foundation : B = 10 m (narrow) and 20 m (wide foundation), and in non-
dimensional form : B/H = 0.5 and 1.0.
(b) The location s of the structure relative to the free-field fault outcrop (measured from its
hanging wall edge) : s/B = 0.1, 0.3, 0.5, 0.7, 0.9, 1.1, 1.3, 1.5 and 2.1.
(c) The weight of the superstructure, i.e. the uniformly distributed surcharge load q acting on
the foundation : q = 10, 20, 40, 60, 80, and 100 kPa, and in non dimensionless form :
q/gB = 0.05, 0.1, 0.2, 0.3, 0.4, and 0.5 for the narrow B/H = 0.5 foundation and q/gB =
0.025, 0.05, 0.1, 0.15, 0.2, and 0.25 for the wide B/H = 1.0 foundation.
(d) The resilience of the foundation soil : idealized dense (
p
= 45
,
res
= 30
,
p
= 18
y
= 0.015) and loose sand (
p
= 32
,
res
= 30
,
p
= 3,
y
= 0.030) [see also Anastasopou-
los et al., 2007].
In the following sections, the key results of the parametric analysis are presented and discussed.

3.1 The Effect of the Location of the Fault Rupture
To highlight the effect of the location s/B of the outcropping fault rupture relative to the hang-
ing-wall side (left) edge of the structure, we focus on the response of the narrow B/H = 0.5 (i.e.
B = 10 m) foundation subjected to surcharge load q/gB = 0.1 (i.e. q = 20 kPa), positioned at s/B
= 0.1, 0.5, 0.9, 1.1, 1.5, and 2.1. The comparison for the first three locations (corresponding to
the fault rupture emerging within the limits of the foundation) is shown in Figure 4 in terms of
deformed mesh with superimposed plastic shear strain contours (for bedrock offset h/H = 10%),
evolution with h/H (from 0 to 10%) of non-dimensional contact pressure p/q, and normalized
foundation bending moment M/qB
2
.
As depicted in Figure 4a, for s/B = 0.1 the rupture is clearly diverted towards the hanging-
wall, i.e. to the left of the foundation. Despite this diversion, the foundation experiences signifi-
cant stressing. Initially, for h/H = 0, the foundation is in full contact with the bearing soil : p/q is
more or less equal to 1 throughout its width. For h/H = 1.5%, although the rupture has not yet
outcropped, some change in the distribution of p/q can be observed. Increasing the bedrock off-
set to h/H = 3.5%, the rupture outcrops at the hanging-wall edge of the foundation, generating
loss of contact at its center (x/B = 0.4 to 0.8) : i.e. it is supported only at the its two edges, from
x/B = 0 to 0.4 and from x/B = 0.8 to 1.0. This means that the foundation tends to behave as a
simply supported beam on elastic end supports. Further increase of h/H to 5% and 10% does
not seem to alter things significantly. These changes in p/q are responsible for the development
of the faultinginduced bending moments M/qB
2
. Initially, for h/H = 1.5%, M/qB
2
is almost two
times larger than its initial value M
o
(which is the foundation bending moment for h = 0, i.e. be-
-0.5
0
0.5
1
1.5
2
2.5
3
-30 -25 -20 -15 -10 -5 0 5
Centrifuge Test
Class "A" Prediction
Horizontal distance x (m)
y

(
m
)
(a)
0
0.5
1
1.5
2
0 0.5 1 1.5 2 2.5
Centrifuge Test
h (m)
(
d
e
g
)
(b)
-0.5
0
0.5
1
1.5
2
2.5
3
-30 -25 -20 -15 -10 -5 0 5
Centrifuge Test
Horizontal distance x (m)
y

(
m
)
(a)
0
0.5
1
1.5
2
0 0.5 1 1.5 2 2.5
Centrifuge Test
h (m)
(
d
e
g
)
(b)
0
0.5
1
1.5
2
0 0.5 1 1.5 2 2.5
Centrifuge Test
h (m)
(
d
e
g
)
(b)
h (m)
(
d
e
g
)
(b)
422
-0.02
0
0.02
0.04
0.06
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
-4
-3
-2
-1
0
1
0
0.02
0.04
0.06
0.08
0.1
0 0.2 0.4 0.6 0.8 1
p
/
q
2 h/H = 1.5% 3 h/H = 3.5% 1 h/H = 0 4 h/H = 5% 5 h/H = 10%
1
2
3 4 5
x/B
1
2
3
4
5
(a) s/B = 0.1 (b) s/B = 0.5 (c) s/B = 0.9
1
2
3
4
5
1
2
3
4
5
5
4
3
1
2
1
2
3
4
5
M
/
q
B
2
q/gB = 0.1
x/B
q/gB = 0.1
h/H = 10%
q/gB = 0.1
x/B
h/H = 10% h/H = 10%
M
o
M
o
M
o
-0.09
-0.06
-0.03
0
0.03
0.06
p
o
p
o p
o
-0.02
0
0.02
0.04
0.06
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
-4
-3
-2
-1
0
1
0
0.02
0.04
0.06
0.08
0.1
0 0.2 0.4 0.6 0.8 1
p
/
q
2 h/H = 1.5% 3 h/H = 3.5% 1 h/H = 0 4 h/H = 5% 5 h/H = 10% 2 h/H = 1.5% 2 h/H = 1.5% 3 h/H = 3.5% 3 h/H = 3.5% 1 h/H = 0 1 h/H = 0 4 h/H = 5% 4 h/H = 5% 5 h/H = 10% 5 h/H = 10%
1
2
3 4 5
x/B
1
2
3
4
5
(a) s/B = 0.1 (b) s/B = 0.5 (c) s/B = 0.9
1
2
3
4
5
1
2
3
4
5
5
4
3
1
2
1
2
3
4
5
M
/
q
B
2
q/gB = 0.1
x/B
q/gB = 0.1
h/H = 10%
q/gB = 0.1
h/H = 10%
q/gB = 0.1
x/B
h/H = 10% h/H = 10%
M
o
M
o
M
o
-0.09
-0.06
-0.03
0
0.03
0.06
p
o
p
o p
o
fore application of the imposed offset). Further increase of h/H to 3.5% leads to further stressing
of the foundation, with M/qB
2
being almost 5 times larger than M
o
. Then, in accordance with the
evolution of p/q any additional imposed displacement does not generate any additional stress-
ing.
Moving the fault rupture emergence to the middle of the foundation (i.e. s/B = 0.5), the re-
sponse becomes substantially different. As illustrated in Figure 4b, the rupture now cannot be
diverted, but some portion of plastic deformation is diffused underneath the foundation. As for
s/B = 0.1, although the change in the distribution of p/q is not that dramatic for h/H = 1.5%,
M/qB
2
becomes almost 3 times larger than M. Then, for h/H = 3.5%, the rupture outcrops be-
neath the foundation generating loss of support at the left edges, from x/B = 0 to 0.15, and near
the middle to the right, from x/B = 0.6 to 0.9. The unsupported span at the left essentially acts as
a cantilever on elastic supports, while the one at the middle-right as a simply supported beam.
The outcome of this change in the support conditions is a reversal of the stressing, with M/qB
2

reaching 0.02, or roughly M
o
(i.e. hogging instead of sagging deformation). Interestingly, for
h/H = 10% the foundation regains contact at its left edge, practically canceling the cantilever
type left span and reducing the hogging M/qB
2
to 0.01. Since the cantilever at the left was act-
ing as a counter-balance for the middleright simply supportedtype span, the sagging bending
moment is increased substantially to almost 2.5M
o
.
Moving the rupture close to the footwall-side (right) edge of the foundation (s/B = 0.9) makes
things quite simpler : neither diversion nor diffusion are observed (Fig. 4c). Initially for h/H =
1.5%, although the changes in p/q are not that pronounced, M/qB
2
is almost doubled ( 2M
o
).
Then, for h/H = 3.5%, the rupture outcrops under the right edge of the foundation, leading to
loss of support at both edges. As a result, both edge spans act as cantilevers leading to reversal
of the stressing (hogging instead of sagging deformation), with M/qB
2
reaching 0.08 (i.e.
roughly 4 M
o
). Further increase of h/H does not lead to any remarkable change.

Figure 4. The effect of fault rupture location s. Narrow B/H = 0.5 foundation, with surcharge load q/gB =
0.1 : (a) s/B = 0.1, (b) s/B = 0.5, and (c) s/B = 0.9. From top to bottom : deformed mesh with superimposed
plastic strain contours (for h/H = 10%), normalized contact pressures p/q , and normalized foundation
bending moment M/qB
2
.
423
0 0.2 0.4 0.6 0.8 1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0 0.2 0.4 0.6 0.8 1
1
2
3
4
5
p
/
q
x/B
(a) s/B = 1.1 (b) s/B = 1.5 (c) s/B = 2.1
M
/
q
B
2
q/gB = 0.1 q/gB = 0.1
x/B
q/gB = 0.1
x/B
-3
-2
-1
0
1
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
0 0.2 0.4 0.6 0.8 1
1
2
3
4
5
1
2
3
4
5
h/H = 10% h/H = 10% h/H = 10%
2 h/H = 1.5% 3 h/H = 3.5% 1 h/H = 0 4 h/H = 5% 5 h/H = 10%
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
0 0.2 0.4 0.6 0.8 1
1
2
3
4
5
p
/
q
x/B
(a) s/B = 1.1 (b) s/B = 1.5 (c) s/B = 2.1
M
/
q
B
2
q/gB = 0.1 q/gB = 0.1
x/B
q/gB = 0.1
x/B
-3
-2
-1
0
1
1
2
3
4
5
-3
-2
-1
0
1
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
1
2
3
4
5
1
2
3
4
5
h/H = 10% h/H = 10% h/H = 10%
2 h/H = 1.5% 3 h/H = 3.5% 1 h/H = 0 4 h/H = 5% 5 h/H = 10% 2 h/H = 1.5% 2 h/H = 1.5% 3 h/H = 3.5% 3 h/H = 3.5% 1 h/H = 0 1 h/H = 0 4 h/H = 5% 4 h/H = 5% 5 h/H = 10% 5 h/H = 10%

Since in many thrustfault earthquakes (e.g., Chi-Chi, Taiwan) serious structural damage has
also been reported in the hanging-wall segment (i.e. at some distance from the fault scarp), we
also investigate cases in which the fault rupture outcrops outside the width of the structure :
s/B = 1.1, 1.5 and 2.1 (Figure 5).
As depicted in Figure 5a, for s/B = 1.1 the rupture outcrops close to the footwall-side (right)
edge of the foundation and the whole response is very similar to the s/B = 0.9 case. For h/H =
1.5%, p/q is marginally altered and M/qB
2
remains almost equal to M
o
. Then, for h/H = 3.5%,
the rupture outcrops close to the footwall-side (right) edge of the foundation, and loss of support
is observed at both edges. The width of the unsupported spans increases with the imposed bed-
rock offset, and for h/H = 10% loss of support is observed from x/B = 0 to 0.3, and from 0.7 to
1.0. With both edges acting as cantilevers, the stressing is reversed and M/qB
2
is only margin-
ally lower than for the s/B = 0.9 case.
Moving the fault rupture further away (to the right) leads to a decrease of the unsupported
spans, and of the distress of the foundation. For s/B = 1.5 (Figure 5b), loss of support is ob-
served from x/B = 0 to 0.15 and from 0.85 to 1.0, leading to a hogging M/qB
2
3M
o
(instead of
roughly 4M
o
of the previous cases).
Even for s/B = 2.1 (Figure 5c), although the loss of support is quite limited (from x/B = 0 to
0.05 and from 0.95 to 1.0), the distress of the foundation is still non-negligible : M/qB
2

1.5M
o
. Most importantly, the foundation is still subjected to hogging deformation, which is ex-
actly the opposite of what it is normally designed for (i.e. sagging). This means, that even if the
rupture does not cross the structure (in this particular case, it missed the foundation for 12 m),
without proper design structural damage is quite likely : all reinforcements will be at the oppo-
site side (e.g. for the foundation slab at the bottom instead of the top).

Figure 5. The effect of fault rupture location s. Narrow B/H = 0.5 foundation, with surcharge load q/gB =
0.1 : (a) s/B = 1.1, (b) s/B = 1.5, and (c) s/B = 2.1. From top to bottom : deformed mesh with superimposed
plastic strain contours (for h/H = 10%), normalized contact pressures p/q , and normalized foundation
bending moment M/qB
2
.

424
h/H = 10% h/H = 10%
h/H = 10% h/H = 10%
h/H = 10%
h/H = 10%
(a) s/B = 0.1 (b) s/B = 0.5 (c) s/B = 0.9
h/H = 10%
q/gB = 0.4
q/gB = 0.1
q/gB = 0.4 q/gB = 0.4
h/H = 10% h/H = 10%
q/gB = 0.1 q/gB = 0.1
q/gB = 0.2 q/gB = 0.2 q/gB = 0.2
h/H = 10% h/H = 10%
h/H = 10% h/H = 10%
h/H = 10%
h/H = 10%
(a) s/B = 0.1 (b) s/B = 0.5 (c) s/B = 0.9
h/H = 10%
q/gB = 0.4
q/gB = 0.1
q/gB = 0.4 q/gB = 0.4
h/H = 10% h/H = 10%
q/gB = 0.1 q/gB = 0.1
q/gB = 0.2 q/gB = 0.2 q/gB = 0.2

3.2 The Effect of the Dead Load of the Superstructure
Focusing on the effect of the dead load of the superstructure, we compare the response of the
same B/H = 0.5 foundation subjected to surcharge load q/gB = 0.1, 0.2, and 0.4 (i.e. q = 20, 40,
and 80 kPa, respectively), positioned at s/B = 0.1, 0.5, and 0.9.
Figure 6 summarizes the comparison in terms of deformed mesh with superimposed plastic
shear strain contours for h/H = 10%. For s/B = 0.1 (i.e. the rupture outcrops close to the hanging
wall side left edge of the foundation), the increase of q/gB leads to a more pronounced di-
version of the rupture path and a reduction of foundation rotation (Figure 6a). The effect of
the superstructure weight on the normalized settlement profile y/H and the rotation is
shown in Figures 7a and 7b, respectively (for s/B = 0.1). While with q/gB = 0.05 the founda-
tion is detached from the soil from its centre to the right edge, a surcharge load q/gB = 0.1 is
enough for the foundation to maintain full contact (Figure 7a). At the same time, is signifi-
cantly decreased (Figure 7b) : from 2
o
(h/H = 10%) for q/gB = 0.05, to 0.4
o
for q/gB = 0.4.
Notice also that while increases almost linearly with h/H for small bedrock offsets (h
2.5%), it is hardly affected by further increase of h/H not only for the lightly loaded foundations
but also for the heavy ones. The effect of q/gB on p/q and M/qB
2
is depicted in Figures 7d and
7d, respectively. The increase of q/gB decreases the unsupported span at the middle of the
foundation: while with q/gB = 0.05 the loss of support is observed from x/B = 0.1 to 0.9, with
q/gB 0.2 full contact is maintained (Figure 7c). As a result, the maximum M/qB
2
is reduced
markedly with the increase of q/gB (Figure 7d).
Moving the rupture to the middle of the foundation, s/B = 0.5, the interaction becomes a little
more complicated (Figure 6b). While a light superstructure (q/gB = 0.1) does not really affect
the rupture path, increasing q/gB to 0.2 leads to diffusion of plastic deformation and some di-
version towards the footwall (to the right). With further increase of q/gB to 0.4, bifurcation can
be observed, with the main rupture being diverted to the hanging wall edge of the foundation,
reducing appreciably.

Figure 6. The effect of superstructure weight, expressed through the surcharge load q/gB ; narrow B/H
= 0.5 foundation : (a) s/B = 0.1, (b) s/B = 0.5, and (c) s/B = 0.9. From top to bottom, deformed mesh with
425
Horizontal distance x/B
y
/
H
(
%
)
q/gB = 0.05
q/gB = 0.10
q/gB = 0.20
q/gB = 0.40
q
B/H = 0.5
(a)
12.5
10
7.5
5
2.5
0
-1.0 -0.5 0 0.5 1.0 1.5
0
0.5
1
1.5
2
2.5
h/H (%)
(
d
e
g
)
(b)
0 2.5 5 10 7.5
-6
-5
-4
-3
-2
-1
0
1
0 0.2 0.4 0.6 0.8 1
x/B
p
/
q
(c)
q/gB = 0.05
q/gB = 0.10
q/gB = 0.20
q/gB = 0.40
0
0.03
0.06
0.09
0.12
0 0.2 0.4 0.6 0.8 1
(d)
M
/
q
B
2
x/B
y
/
H
(
%
)
q/gB = 0.05
q/gB = 0.10
q/gB = 0.20
q/gB = 0.40
q
B/H = 0.5
q
B/H = 0.5
(a)
12.5
10
7.5
5
2.5
0
-1.0 -0.5 0 0.5 1.0 1.5
0
0.5
1
1.5
2
2.5
h/H (%)
(
d
e
g
)
(b)
0 2.5 5 10 7.5
0
0.5
1
1.5
2
2.5
h/H (%)
(
d
e
g
)
(b)
0 2.5 5 10 7.5
-6
-5
-4
-3
-2
-1
0
1
0 0.2 0.4 0.6 0.8 1
x/B
p
/
q
(c)
q/gB = 0.05
q/gB = 0.10
q/gB = 0.20
q/gB = 0.40
-6
-5
-4
-3
-2
-1
0
1
0 0.2 0.4 0.6 0.8 1
x/B
p
/
q
(c)
q/gB = 0.05
q/gB = 0.10
q/gB = 0.20
q/gB = 0.40
q/gB = 0.05
q/gB = 0.10
q/gB = 0.20
q/gB = 0.40
0
0.03
0.06
0.09
0.12
0 0.2 0.4 0.6 0.8 1
(d)
M
/
q
B
2
x/B
0
0.03
0.06
0.09
0.12
0 0.2 0.4 0.6 0.8 1
(d)
M
/
q
B
2
0
0.03
0.06
0.09
0.12
0 0.2 0.4 0.6 0.8 1
(d)
M
/
q
B
2
x/B
shear strain contours for q/gB = 0.1, 0.2, and 0.4.

Figure 7. The effect of superstructure weight, expressed through the surcharge load q/gB, on : (a) normal-
ized vertical displacement y/H at the soil surface (for h/H = 10%) ; (b) foundation rotation with respect
to bedrock offset h/H ; (c) normalized contact pressure p/q ; and (d) normalized bending moment M/qB
2
for
h/H = 10% (narrow B/H = 0.5 foundation subjected to thrust faulting at s/B = 0.1).

For s/B = 0.9 (i.e. moving the rupture close to the footwall side right edge of the foundation),
the increase of q/gB leads to diversion of the rupture path towards the footwall and substantial
reduction of the areas of loss of support (Figure 6c). As for the previous cases, the increase of
q/gB leads to reduction of the unsupported spans of the foundation, leading to a substantial de-
crease of foundation distress (not shown here) : M/qB
2
decreases from roughly 5M
o
for
q/gB = 0.05 to 2M
o
for q/gB = 0.40, always remaining negative (i.e. hogging deformation).

3.3 The Effect of Soil Stiffness
To draw a picture on the effect of soil stiffness, we compare the response of the B/H = 0.5 foun-
dation with q/gB = 0.1 (i.e. q = 20 kPa), subjected to thrust faulting through idealized dense or
loose sand.
Figure 8 summarizes the results for fault rupture outcropping near the hanging wall (left)
edge of the foundation : s/B = 0.1. Although no clear difference can be observed in the y/H
profile (Figures 12a and b), in terms of p/q (Figure 12c) and M/qB
2
(Figure 12d) the disparity is
remarkable : while in dense sand the foundation experiences rather pronounced loss of contact
at the middle (x/B = 0.35 to 0.75), in loose sand contact is maintained throughout its whole
length. As a result the faultinginduced stressing is reduced substantially with the increase of
soil resilience : the maximum M/qB
2
is reduced from 4M
o
in dense sand to almost M
o
in the case
of loose sand (i.e. almost no stressing).
Moving the outcropping fault rupture towards the middle of the foundation (i.e. s/B = 0.3 and
0.5, not shown herein), although the interaction mechanism is different, in terms of foundation
distress the trends are quite similar. In general, the increase of soil resilience leads to a decrease
of foundation distress and to substantial limitation of the unsupported central span.

426
-4
-3
-2
-1
0
1
0 0.2 0.4 0.6 0.8 1
0
0.02
0.04
0.06
0.08
0.1
0 0.2 0.4 0.6 0.8 1
q/gB = 0.1
y
/
H
(
%
)
p
/
q
x/B
y
/
H
(
%
)
M
/
q
B
2
q/gB = 0.1
(a)
(b)
(c)
(d)
Dense sand
Loose sand
15
10
5
0
-1.0 0 1.0 2.0 3.0
-1.0 0 1.0 2.0 3.0
15
10
5
0
-4
-3
-2
-1
0
1
0 0.2 0.4 0.6 0.8 1
0
0.02
0.04
0.06
0.08
0.1
0 0.2 0.4 0.6 0.8 1
q/gB = 0.1
y
/
H
(
%
)
p
/
q
x/B
y
/
H
(
%
)
M
/
q
B
2
q/gB = 0.1
(a)
(b)
(c)
(d)
Dense sand
Loose sand
Dense sand
Loose sand
15
10
5
0
-1.0 0 1.0 2.0 3.0 -1.0 0 1.0 2.0 3.0
-1.0 0 1.0 2.0 3.0 -1.0 0 1.0 2.0 3.0
15
10
5
0

Figure 8. The effect of soil stiffness : (a) normalized vertical displacement y/H at the soil surface for
dense sand, and (b) for loose sand ; (c) normalized contact pressures p/q, and (d) normalized bending mo-
ment M/qB
2
; narrow B/H = 0.5 foundation with surcharge load q/gB = 0.1 subjected to thrust faulting at
s/B = 0.1 (all results plotted for h/H = 10%).

In Figure 9 we synopsize the results for s/B = 0.9 (i.e. the fault rupture emerging close to
footwall right edge of the foundation). Now, the difference in response is observable even in
the y/H profile (Figures 9a and 9b). In dense sand, loss of support is observed at both edges,
from x/B = 0 to 0.25 and from 0.95 to 1.0 (Figure 9c). In loose sand, the cantilever at the hang-
ing wall (left) side of the foundation is smaller (x/B = 0 to 0.15) ; full contact is maintained at
the right. As a result, the faultinginduced M/qB
2
(Figure 9d) is again reduced substantially with
the increase of soil resilience : from 4M
o
in dense to about 2M
o
in loose sand. It should be no-
ticed, however, that the foundation is subjected to hogging deformation in both cases (i.e. the
reverse of static loading).
4 DEVELOPMENT OF SIMPLIFIED DESIGN METHOD
4.1 The Concept
As already discussed, depending on the location of the outcropping fault rupture relative to the
structure, detachment may take place either at the middle or at the ends of the foundation. In the
first case, the unsupported span will behave as a simply supported beam on elastic supports, im-
posing sagging deformation onto the foundationstructure system ; in the latter case, the un-
supported end spans will act as cantilevers on a central elastic support, producing hogging de-
formation. Since the stressing of the foundation is directly related to this loss of support,
exploiting the results of our analysis the design of a foundationstructure system against fault-
ing can become very simple : just remove the Winkler-type springs (i.e. the supports typically
used in structural design software) from the area(s) of loss of support due to soil detachment.

427
q/gB = 0.1
(b)
-0.1
-0.08
-0.06
-0.04
-0.02
0
0 0.2 0.4 0.6 0.8 1
-5
-4
-3
-2
-1
0
1
0 0.2 0.4 0.6 0.8 1
p
/
q
x/B
M
/
q
B
2
q/gB = 0.1
(a) (c)
(d)
Dense sand
Loose sand
y
/
H
(
%
)
y
/
H
(
%
)
-1.0 0 1.0 2.0 3.0
-1.0 0 1.0 2.0 3.0
15
10
5
0
15
10
5
0
Dense sand
Loose sand
q/gB = 0.1
(b)
-0.1
-0.08
-0.06
-0.04
-0.02
0
0 0.2 0.4 0.6 0.8 1
-5
-4
-3
-2
-1
0
1
0 0.2 0.4 0.6 0.8 1
p
/
q
x/B
M
/
q
B
2
q/gB = 0.1
(a) (c)
(d)
Dense sand
Loose sand
y
/
H
(
%
)
y
/
H
(
%
)
-1.0 0 1.0 2.0 3.0
-1.0 0 1.0 2.0 3.0
15
10
5
0
15
10
5
0
q/gB = 0.1
(b)
-0.1
-0.08
-0.06
-0.04
-0.02
0
0 0.2 0.4 0.6 0.8 1
-5
-4
-3
-2
-1
0
1
0 0.2 0.4 0.6 0.8 1
p
/
q
x/B
M
/
q
B
2
q/gB = 0.1
(a) (c)
(d)
Dense sand
Loose sand
Dense sand
Loose sand
y
/
H
(
%
)
y
/
H
(
%
)
-1.0 0 1.0 2.0 3.0 -1.0 0 1.0 2.0 3.0
-1.0 0 1.0 2.0 3.0 -1.0 0 1.0 2.0 3.0
15
10
5
0
15
10
5
0
Dense sand
Loose sand

Figure 9. The effect of soil stiffness : (a) normalized vertical displacement y/H at the soil surface for
dense sand, and (b) for loose sand ; (c) normalized contact pressures p/q, and (d) normalized bending mo-
ment M/qB
2
; narrow B/H = 0.5 foundation with surcharge load q/gB = 0.1 subjected to thrust faulting at
s/B = 0.9 (all results plotted for h/H = 10%).

An example of this simplified design procedure is outlined in Figure 10. We compare the
stressing of the foundation in terms of M/qB
2
as computed through FE analysis, to the simplified
procedure of Winkler spring removal at the area of loss of support, for a narrow B/H = 0.5 foun-
dation with q/gB = 0.1 subjected to faulting at s/B = 0.1 through dense sand (Figure 10a), and
for the same foundation but with a heavier superstructure load q/gB = 0.2 subjected to faulting
at s/B = 0.7 through loose sand (Figure 10b). While in the first case (in which soil detachment is
prevailing), the comparison is quite good, in the latter case (in which no detachment is ob-
served) the simplified method grossly underestimates the distress of the foundation. Thus, it be-
comes clear that a change in the distribution of p/q is enough to produce considerable changes
in M/qB
2
, even without any uplifting taking place.
Hence, to render the simplified design method applicable for all cases, it is necessary to
slightly modify its main concept. For this purpose, we define equivalent area(s) of loss of sup-
port, as schematically illustrated in Figure 11. In the p/q diagram (at the top), we compare the
faultinginduced (for h/H = 10%) foundation contact pressures (black line) to the static (i.e. h/H
= 0, grey line). The grey-shaded areas correspond to the areas of the foundation where p/q is re-
duced due to the imposed tectonic deformation. Notice that this reduction does not lead to any
soil detachment (p/q 0). A clearer picture can be drawn through the (pq)/q diagram (at the
middle) : loss of support takes place when (pq)/q becomes negative (see grey-shaded areas). If
detachment had taken place, (pq)/q would become equal to 1. The same will happen if we
remove spring supports, which is the key objective of the simplified design method. The equiva-
lent area(s) of loss of support are black-shaded in the figure. We explain their logic through the
bottom diagram, which focuses on the first (from left to right) such area.

428
Simplified method
FE Analysis
q/gB = 0.1
Loss of support
q/gB = 0.2
No loss of support
Simplified method
FE Analysis
Simplified method
FE Analysis
p
/
q
M
/
q
B
2
0
-1
-2
-3
-4
0
-1
-2
-3
-4
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
0
0.02
0.04
0.06
0.08
0.1
-0.01
0
0.01
0.02
0.03
0.04
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
(a) (b)
x/B x/B
Simplified method
FE Analysis
q/gB = 0.1
Loss of support
q/gB = 0.2
No loss of support
Simplified method
FE Analysis
Simplified method
FE Analysis
Simplified method
FE Analysis
Simplified method
FE Analysis
p
/
q
M
/
q
B
2
0
-1
-2
-3
-4
0
-1
-2
-3
-4
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
0
0.02
0.04
0.06
0.08
0.1
-0.01
0
0.01
0.02
0.03
0.04
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0
(a) (b)
x/B x/B

Figure 10. Example comparison of the simplified method (removal of Winkler springs at the area of loss of
support) with FE analysis results in terms of non-dimensional foundation bending moment M/qB
2
: (a)
B/H = 0.5 foundation with q/gB = 0.1 subjected to faulting at s/B = 0.1 through dense sand ; (b)
B/H = 0.5 foundation with q/gB = 0.2 subjected to faulting at s/B = 0.7 through loose sand. While in the
first case (in which soil detachment is prevailing) the comparison is quite good, in the latter case (in which
no loss of support is observed) the simplified method grossly underestimates the distress of the foundation.

The area where loss of support is observed spans from x/B = = 0 to (grey-shaded). Assum-
ing equivalency in terms of area, we define the length L of the equivalent area of loss of support
(black-shaded) as follows :
1
0
( )
x
p x q
L dx
q
(2)
Since the distribution of (pq)/q is responsible for the development of M/qB
2
, by removing the
spring supports from this area (of length L), we can achieve approximately the same stressing.
Naturally, since the distribution will not be identical, the solution is approximate.

4.2 Effectiveness of the Simplified Method
The procedure described above was applied to all the cases investigated herein. Making use of
the distribution of (pq)/q as computed through FE analysis, for each case we estimate (as de-
scribed above) the equivalent area(s) of loss of support (i.e. the equivalent lengths L). Then, we
conduct a simplified beamonWinkler spring analysis, removing the spring supports from the
equivalent area(s) of loss of support. The whole procedure was applied for bedrock offset h/H =
5% and 10%. In Figures 12 and 13 we compare the simplified method with FE analysis results,
in terms of M/qB
2
, for dense and loose sand, respectively. The comparison is shown for h/H =
10%, for both the narrow B/H = 0.5 and the wide B/H = 1.0 foundation subjected to a variety of
superstructure loads q, and for s/B = 0.1, 0.5, and 0.9. In all cases (and in the ones not shown
here), the simplified procedure compares quite well with the rigorous FE analysis results.
429
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
-4
-3
-2
-1
0
1
p
/
q
x / B
(
p
q
)

/

q
Equivalent
loss of support
Faultinginduced
h/H = 10%
Static, h/H = 0
unloading
unloading
-1
-0.5
0
0.5
1
0 0.05 0.1 0.15 0.2
L
x
1
x = x / B
1
0
( )
x
p x q
L d x
q
=

(
p
q
)

/

q
-1
-0.5
0
0.5
1
0 0.2 0.4 0.6 0.8 1
-4
-3
-2
-1
0
1
p
/
q
x / B
(
p
q
)

/

q
Equivalent
loss of support
Faultinginduced
h/H = 10%
Static, h/H = 0
unloading
unloading
-1
-0.5
0
0.5
1
0 0.05 0.1 0.15 0.2
L
x
1
x = x / B x = x / B
1
0
( )
x
p x q
L d x
q
=

(
p
q
)

/

q

Figure 11. Example sketch illustrating the main concept of the improved simplified method.

As already discussed, the stressing of the foundation (and consequently of the superstructure)
depends largely on its position relative to the outcropping fault rupture. However, the exact lo-
cation of a fault rupture at the ground surface cannot be predicted with accuracy, even if the lo-
cation of the fault at bedrock is well known [see also Anastasopoulos & Gazetas, 2007a] : only
rarely does the fault trace follow precisely pre-existing outcrops ; soil non-uniformity compli-
cates the pattern of fault emergence. Hence, the design of a foundationstructure system must be
performed on the basis of design envelopes of internal forces, with respect to the location of the
fault rupture s/B.
430
-0.12
-0.1
-0.08
-0.06
-0.04
-0.02
0
0 0.2 0.4 0.6 0.8 1
0
0.02
0.04
0.06
0.08
0.1
0.12
x / B
M
/
q
B
2
-0.02
0
0.02
0.04
0.06
M
/
q
B
2
M
/
q
B
2
(a) s/B = 0.1
(b) s/B = 0.5
(c) s/B = 0.9
0
0.02
0.04
0.06
0.08
0.1
0.12
-0.04
-0.02
0
0.02
0.04
-0.1
-0.08
-0.06
-0.04
-0.02
0
0 0.2 0.4 0.6 0.8 1
x / B
q = 10 kPa q = 20 kPa q = 40 kPa q = 80 kPa Finite Element Analysis :
q = 10 kPa q = 20 kPa q = 40 kPa q = 80 kPa Simplified Method :
B/H = 0.5 B/H = 1.0
-0.12
-0.1
-0.08
-0.06
-0.04
-0.02
0
0 0.2 0.4 0.6 0.8 1
0
0.02
0.04
0.06
0.08
0.1
0.12
x / B
M
/
q
B
2
-0.02
0
0.02
0.04
0.06
M
/
q
B
2
M
/
q
B
2
(a) s/B = 0.1
(b) s/B = 0.5
(c) s/B = 0.9
0
0.02
0.04
0.06
0.08
0.1
0.12
-0.04
-0.02
0
0.02
0.04
-0.1
-0.08
-0.06
-0.04
-0.02
0
0 0.2 0.4 0.6 0.8 1
x / B
q = 10 kPa q = 10 kPa q = 20 kPa q = 20 kPa q = 40 kPa q = 40 kPa q = 80 kPa q = 80 kPa Finite Element Analysis :
B/H = 0.5 B/H = 1.0

Figure 12. Comparison of the simplified method with FE analysis results for dense sand : normalized bend-
ing moment M/qB
2
for imposed bedrock offset h/H = 10%, for narrow B/H = 0.5 and wide B/H = 1.0 foun-
dation : (a) s/B = 0.1, (b) s/B = 0.5, and (c) s/B = 0.9.

Figure 14 summarizes the results of the comparison (between simplified method and FE re-
sults) in terms of normalized bending moment M/qB
2
envelopes for dense and loose sand, for
the case of the narrow B/H = 0.5 foundation subjected to h/H = 10% bedrock offset, and for
three different normalized superstructure weights : q/gB = 0.1, 0.2, and 0.4. As it would be ex-
pected, for light superstructure loads (q/gB = 0.1), where uplifting is prevalent, the comparison
is excellent for both sand types. But even for heavier superstructure loads (q/gB = 0.4), and
even for loose sand (i.e. cases in which the stressing is not related to uplifting), the simplified
procedure provides very reasonable and slightly conservative results.

431
M
/
q
B
2
M
/
q
B
2
M
/
q
B
2
(a) s/B = 0.1
(b) s/B = 0.5
(c) s/B = 0.9
B/H = 0.5 B/H = 1.0
0
0.01
0.02
0.03
0.04
0.05
0.06
0 0.2 0.4 0.6 0.8 1
x / B
-0.01
0
0.01
0.02
0.03
0.04
0.05
0.06
-0.06
-0.04
-0.02
0
0.02
0.04
0
0.02
0.04
0.06
0.08
0 0.2 0.4 0.6 0.8 1
x / B
-0.02
0
0.02
0.04
0.06
-0.08
-0.06
-0.04
-0.02
0
M
/
q
B
2
M
/
q
B
2
M
/
q
B
2
(a) s/B = 0.1
(b) s/B = 0.5
(c) s/B = 0.9
q = 10 kPa q = 10 kPa q = 20 kPa q = 20 kPa q = 40 kPa q = 40 kPa q = 80 kPa q = 80 kPa Finite Element Analysis :
B/H = 0.5 B/H = 1.0
0
0.01
0.02
0.03
0.04
0.05
0.06
0 0.2 0.4 0.6 0.8 1
x / B
-0.01
0
0.01
0.02
0.03
0.04
0.05
0.06
-0.06
-0.04
-0.02
0
0.02
0.04
0
0.02
0.04
0.06
0.08
0 0.2 0.4 0.6 0.8 1
x / B
-0.02
0
0.02
0.04
0.06
-0.08
-0.06
-0.04
-0.02
0

Figure 13. Comparison of the simplified method with FE analysis results for loose sand : normalized bending
moment M/qB
2
for imposed bedrock offset h/H = 10%, for narrow B/H = 0.5 and wide B/H = 1.0 foundation :
(a) s/B = 0.1, (b) s/B = 0.5, and (c) s/B = 0.9.

4.3 Simplified Design Charts
The design procedure can be synopsized in simplified design charts, such as the ones of Figure
15 (referring to the narrow B/H = 0.5 foundation subjected to h/H = 10 % bedrock offset). Such
diagrams illustrate the normalized effective foundation width (in grey), along with the corre-
sponding equivalent areas of loss of support (in white), with respect to the surcharge load q/gB,
soil compliance (dense or loose sand), and fault rupture location s/B. In all cases, the increase of
q/gB leads to a suppression of the equivalent areas of loss of support (and, hence, of founda-
tionstructure distress). Soil compliance has a similar effect : b/B is larger in loose sand.

432
-0.1
-0.05
0
0.05
0.1
M
/
q
B
2
-0.1
-0.05
0
0.05
0.1
M
/
q
B
2
-0.1
-0.05
0
0.05
0.1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
M
/
q
B
2
(a) q/gB = 0.1
(b) q/gB = 0.2
(c) q/gB = 0.4
x / B x / B
Static loading FE Analysis Simplified method
-0.1
-0.05
0
0.05
0.1
M
/
q
B
2
-0.1
-0.05
0
0.05
0.1
M
/
q
B
2
-0.1
-0.05
0
0.05
0.1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
M
/
q
B
2
(a) q/gB = 0.1
(b) q/gB = 0.2
(c) q/gB = 0.4
x / B x / B
Static loading FE Analysis Simplified method
-0.1
-0.05
0
0.05
0.1
M
/
q
B
2
-0.1
-0.05
0
0.05
0.1
M
/
q
B
2
-0.1
-0.05
0
0.05
0.1
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
M
/
q
B
2
(a) q/gB = 0.1
(b) q/gB = 0.2
(c) q/gB = 0.4
x / B x / B
Static loading FE Analysis Simplified method Static loading Static loading FE Analysis Simplified method Simplified method

Figure 14. Synopsis of analysis results comparison of simplified method with FE analysis : normalized
bending moment M/qB
2
envelopes for dense and loose sand, for narrow B/H = 0.5 foundation subjected to
h/H = 10% bedrock offset : (a) q/gB = 0.1, (b) q/gB = 0.2, and (c) q/gB = 0.4.

Figure 16 illustrates schematically the use of the simplified design charts (the example is
shown for B/H = 0.5, h/H = 10 %, dense sand, s/B = 0.1). If, for example, we want to design a
B = 10 m structure with q = 20 kPa (i.e. q/gB = 0.1) on an H = 20 m dense sand deposit (i.e.
B/H = 0.5), subjected to h = 2 m (i.e. h/H = 10%) thrust faulting at s = 1 m (i.e. s/B = 0.1), then
we would have to remove the support springs at the middle, x/B = 0.15 to 0.84, i.e. x = 1.5 m to
8.4 m (Figure 16a). If the same foundation was carrying a heavier superstructure load, q = 80
kPa (i.e. q/gB = 0.4), then we would have to remove some support springs at the left edge, x/B
= 0 to 0.04, some at the middle, x/B = 0.52 to 0.57, and some more at the right edge, x/B = 0.95
to 1 (Figure 16b).
Naturally, the same procedure would have to be undertaken for all possible fault rupture loca-
tions (s/B = 0.3, 0.5, 0.7, and 0.9), leading to different load combinations. Then, foundation and
superstructure can be designed on the basis of the resulting internal force envelope.
433
DENSE
x
/
B
LOOSE
(a)
s/B = 0.1
(b)
s/B = 0.3
1.0
0.8
0.6
0.4
0.2
0
x
/
B
x
/
B
1.0
0.8
0.6
0.4
0.2
0
b
/
B
(c)
s/B = 0.5
(d)
s/B = 0.7
x
/
B
q/gB
1.0
0.8
0.6
0.4
0.2
0
0 0.1 0.2 0.3 0.4 0.5
q/gB
0 0.1 0.2 0.3 0.4 0.5
(e)
s/B = 0.9
Simply supported
beamtype span
Simply supported
beamtype span
Simply supported
beamtype span
Cantilevertype span
Cantilevertype span
Cantilevertype span
Cantilevertype span
Cantilevertype span
S
im
p
ly
s
u
p
p
o
rte
d

b
e
a
m
ty
p
e
s
p
a
n
Simply supported
beamtype span
1.0
0.8
0.6
0.4
0.2
0
1.0
0.8
0.6
0.4
0.2
0
DENSE
x
/
B
LOOSE
(a)
s/B = 0.1
(b)
s/B = 0.3
1.0
0.8
0.6
0.4
0.2
0
x
/
B
x
/
B
1.0
0.8
0.6
0.4
0.2
0
b
/
B
(c)
s/B = 0.5
(d)
s/B = 0.7
x
/
B
q/gB
1.0
0.8
0.6
0.4
0.2
0
0 0.1 0.2 0.3 0.4 0.5
q/gB
0 0.1 0.2 0.3 0.4 0.5
(e)
s/B = 0.9
Simply supported
beamtype span
Simply supported
beamtype span
Simply supported
beamtype span
Cantilevertype span
Cantilevertype span
Cantilevertype span
Cantilevertype span
Cantilevertype span
S
im
p
ly
s
u
p
p
o
rte
d

b
e
a
m
ty
p
e
s
p
a
n
Simply supported
beamtype span
1.0
0.8
0.6
0.4
0.2
0
1.0
0.8
0.6
0.4
0.2
0

Figure 15. Simplified design charts for B/H = 0.5 foundation subjected to h/H = 10 % bedrock offset : nor-
malized effective foundation width (in grey) and corresponding equivalent areas of loss of support (in white)
with respect to normalized surcharge load q/gB and soil compliance, for : (a) s/B = 0.1, (b) s/B = 0.3,
(c) s/B = 0.5, (d) s/B = 0.7, and (e) s/B = 0.9.
5 CONCLUSIONS
In the present paper we applied a validated FE analysis methodology to study the response of
slab foundations subjected to thrust faulting. After conducting a detailed parametric analysis,
investigating the effect of key response parameters, a simplified design method is developed and
434
b
/
B
1.0
0.8
0.6
0.4
0.2
0
q/gB
0 0.1 0.2 0.3 0.4 0.5
q/gB = 0.1
q/gB = 0.4
(a) (b)
0.15 0.84
0.15
0.84
0.04
0.52
0.57
0.95
0.04 0.52 0.57 0.95
b
/
B
1.0
0.8
0.6
0.4
0.2
0
q/gB
0 0.1 0.2 0.3 0.4 0.5
b
/
B
1.0
0.8
0.6
0.4
0.2
0
q/gB
0 0.1 0.2 0.3 0.4 0.5
q/gB = 0.1
q/gB = 0.4
(a) (b)
0.15 0.84
0.15
0.84
0.04
0.52
0.57
0.95
0.04 0.52 0.57 0.95
validated against FE analysis results. The final outcome of the research presented herein is in
the form of simplified non-dimensional design charts for a variety of foundation-structure con-
ditions. The main conclusions and limitations of this work are as follows :
(a) The stressing of the foundation, and consequently of the superstructure, is a function of its
location relative to the fault rupture. Depending on the relative geometry loss of support
may take place either under the edges or under the middle. In the first case, the unsupported
spans act as cantilevers on a central elastic support, producing hogging deformation ; in
the latter case as a simply supported beam on elastic end supports, producing sagging de-
formation.
(b) The increase of the weight of the superstructure, as expressed through the surcharge load q,
in general leads to less stressing of the foundation (and consequently of the superstructure).
The surcharge load has a double role : (i) it changes the stress field underneath the founda-
tion, facilitating the diversion of the fault rupture ; and (ii) compressing the soil, it tends to
flatten the faultinginduced anomalies. But on the other hand, increased height of the
structure makes the unavoidable foundation rotation more dangerous.
(c) In all cases examined here, increasing soil compliance is favourable in terms of the stress-
ing of the foundation (and consequently of the superstructure). Foundationstructure sys-
tems on loose sand will be subjected to less sagging or hogging deformation (depending on
the location of fault crossing), and will thus experience far less stressing compared to dense
sand. In addition, the increase of soil compliance leads to increased diffusion of plastic de-
formation underneath the foundation, facilitating the aforementioned flattening of the fault-
inginduced anomalies (which are responsible for the distress).
(d) The stressing of the foundation, and consequently of the superstructure, stems mainly from
loss of support. Exploiting the results of the parametric FE analysis, a simplified design
procedure is developed for the design of foundationstructure systems against faulting. The
concept is quite simple : just remove the Winkler-type support springs form the area(s) of
loss of support and conduct a conventional static analysis with the dead load q of the super-
structure.
(e) Given that soil detachment is not always prevalent (as with a light superstructure on com-
pliant soil), we define equivalent area(s) of loss of support, through a simplified procedure
based on the distribution of (pq)/q. Since the latter is mainly responsible for the develop-
ment of foundation distress (M/qB
2
), removing the spring supports from those areas, ap-
proximately the same stressing can be attained.

Figure 16. Example illustration of the use of simplified design charts, for B/H = 0.5 foundation subjected to
h/H = 10 % faulting through dense sand, s/B = 0.1 : (a) q/gB = 0.1, and (b) q/gB = 0.4.
435

(f) The simplified design procedure has been applied to all cases investigated herein. Simpli-
fied beamonWinkler spring analyses were conducted, removing the spring supports from
the equivalent area(s) of loss of support. The comparison of the simplified procedure with
FE analysis results was shown to be quite acceptable.
(g) The final outcome of the simplified design method is in the form if simplified design
charts, illustrating the equivalent area(s) of loss of support, with respect to the surcharge
load q/gB, soil compliance, and the location of the fault rupture s/B. Since the stressing of
the foundation depends largely on its location relative to the emerging fault rupture, the de-
sign of a foundationstructure system must always be performed on the basis of design en-
velopes of internal forces, with respect to s/B.
(h) The simplified design charts presented here can be considered adequate for preliminary de-
sign purposes. For the final design of a structure against faulting, a more detailed analysis
should be conducted, taking account the details of the soil profile. The idealized sands used
for our analyses cannot possibly cover the whole range of soil materials. In addition, varia-
tions of foundation and/or superstructure stiffness (the rule, rather than the exception, in
practise) should also be duly taken into account. Finally, all results presented herein corre-
spond to dry soil conditions, and hence the potential presence of a water table has not been
investigated.
ACKNOWLEDGEMENT
This work was funded by OSE (the Greek Railway Organization), as part of the research project
Railway Bridges on Active Seismic Faults. Centrifuge testing and finite element analyses
formed part of the EU research project QUAKER, funded through the EU Fifth Framework
Programme, under contract number : EVG1-CT-2002-00064.
REFERENCES
Anastasopoulos I., & Gazetas G. (2007a), Foundation-Structure Systems over a Rupturing Normal Fault
: Part I. Observations after the Kocaeli 1999 Earthquake, Bulletin of Earthquake Engineering, Vol.
5, No. 3, pp. 253275.
Anastasopoulos I., & Gazetas G. (2007b), Behaviour of StructureFoundation Systems over a Rupturing
Normal Fault : Part II. Analysis of the Kocaeli Case Histories, Bulletin of Earthquake Engineering,
Vol. 5, No. 3. pp. 277301.
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, 133(8), pp. 943958.
Anastasopoulos, I., Callerio, A., Bransby, M.F., Davies, M.C.R., El. Nahas, A., Faccioli, E., Gazetas, G.,
Masella, A., Paolucci, R., Pecker, A., Rossignol, E. (2008), Numerical Analyses of Fault
Foundation Interaction, Bulletin of Earthquake Engineering, , Vol. 6, No. 4, pp. 645675.
Anastasopoulos I., Gazetas G., Bransby M.F., Davies M.C.R., and El Nahas A. (2009), Normal Fault
Rupture Interaction with Strip Foundations, Journal of Geotechnical and Geoenvironmental Engi-
neering, ASCE, Vol. 135, No. 3, pp. 359370.
BEE (2008) : Bulletin of Earthquake Engineering , Special Issue : Integrated approach to fault rupture-
and soil-foundation interaction, Vol. 6, No. 4.
Berill, J.B. (1983), Two-dimensional analysis of the effect of fault rupture on buildings with shallow
foundations, Soil Dynamics and Earthquake Engineering, Vol. 2, No. 3, pp. 156-160.
Bransby, M.F., Davies, M.C.R., and El Nahas, A. (2007a), Centrifuge modelling of normal fault-
fault rupture- and soil-foundation interaction, Vol. 6, No. 4, pp. 585605.
Bransby, M.F., Davies, M.C.R., and El Nahas, A. (2007b), Centrifuge modelling of reverse fault-
fault rupture- and soil-foundation interaction, Vol. 6, No. 4, pp. 607628.
Bray, J.D. (1990), The Effects of Tectonic Movements on Stresses and Deformations in Earth Embank-
ments, Ph.D Dissertation, University of California, Berkeley.
Bray, J.D., Seed, R.B., Cluff, L.S., and Seed, H.B. (1994a), Earthquake Fault Rupture Propagation
through Soil, Journal of Geotechnical Engineering, ASCE, Vol. 120, No.3, pp. 543-561.
436
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, 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, 110 (5), pp. 599-615.
Duncan, J.M., and Lefebvre, G., (1973), Earth Pressure on Structures Due to Fault Movement, Journal of
Soil Mechanics and Foundation Engineering, ASCE, Vol. 99, pp. 11531163.
Faccioli, E., Anastasopoulos, I., Callerio, A., and Gazetas, G. (2008), Case histories of faultfoundation
interaction, Bulletin of Earthquake Engineering, Special Issue : Integrated approach to fault rupture-
and soil-foundation interaction, Vol. 6, No. 4, pp. 557583.
Gaudin, C. (2002), Experimental and theoretical study of the behaviour of supporting walls: Validation
of design methods, Ph.D. Dissertation, Laboratoire Central des Ponts et Chausses, Nantes, France.
Johansson J., Konagai K. (2006), Fault induced permanent ground deformationsan experimental com-
parison of wet and dry soil and implications for buried structures, Soil Dynamics & Earthquake Engi-
neering, Vol. 26, No. 1, pp. 4553.
Muir Wood, D. (2002), Some observations of volumetric instabilities in soils, International Journal of
Solids and Structures, Vol. 39, pp. 34293449.
Muir Wood, D. (2004), Geotechnical Modelling, by Spon Press , London & New York, Applied Geo-
technics, Vol. 1, ISBN 0-419-23730-5.
Ulusay, R., Aydan, O., Hamada, M. (2002), The behaviour of structures built on active fault zones: Ex-
amples from the recent earthquakes of Turkey, Structural Engineering & Earthquake Engineering,
JSCE, Vol. 19, No. 2, pp. 149167.
437
1 INTRODUCTION
The Messina Bridge, would have made the span 60% longer than the Akashi Kaikyo Bridge in
Japan (the currently largest suspension bridge, with a huge 1991 m central span). The bridge is
planned to connect Reggio Calabria to Messina, Sicily. The big depth of water in the straits
could only support the solution of a suspension bridge with a large central span. In the proposed
project plan, the main span of 3300 m and the 2 x 183 m side spans are supported with sus-
penders from overhead main cables. Figure 1 depicts the longitudinal crosssection of the
Bridge, along with the geologicalgeotechnical profile under the Bridge.

Figure 1. Longitudinal crosssection of the bridge, along with the geologicalgeotechnical profile (cour-
Preliminary SFSI Studies for the Messina Bridge Foundations

E. Stavropoulou, I. Anastasopoulos, G. Gazetas
Laboratory of Soil Mechanics, National Technical University, Athens, Greece

ABSTRACT: The Messina Bridge in Italy, when completed will be the longest suspension
bridge ever built, having a central span of 3300 m. In this paper, a generic study of the soil
pierfoundation interaction of the Calabrian-side pier of the preliminary Bridge design is exam-
ined. The massive concrete foundation of this pier is to be founded in a gravelly deposit, after
the latter has been subjected to improvement by jet grouting. A parametric elastic study is first
conducted to investigate the influence of the width and depth of soil improvement beneath the
foundation. Then, the response of an alternative less conservative foundation is investigated
and compared to the original design. Nonlinear features of material (soil) and geometry (up-
lifting, sliding, and second order effects), , as well as the flexibility of the tower, are taken into
account.

C : Coastal plain deposits, Holocene
E : Messina Gravels formation, Pleistocene
F : Continental Deposits, Lower Pleistocene upper Pliocene
Active Fault
G : Calcarenite and lightly cemented sands, Pliocene
H : Pezzo Conglomerate, Miocene
R : Cristaline base rock
Inactive Fault
200
100
0
-100
-200
-300
-400
-500
-600
-700
C : Coastal plain deposits, Holocene
E : Messina Gravels formation, Pleistocene
F : Continental Deposits, Lower Pleistocene upper Pliocene
Active Fault
G : Calcarenite and lightly cemented sands, Pliocene
H : Pezzo Conglomerate, Miocene
R : Cristaline base rock
Inactive Fault
200
100
0
-100
-200
-300
-400
-500
-600
-700
438
(a)
(b)
(c)
(a)
(b)
(c)
tesy of M. Jamiolkowski).
The height of the two offshore towers will be 383 m, in order to provide a minimum vertical
clearance for navigation of 65 m. The Calabrian tower (Figure 2), examined in this study, will
be founded on coastal sandy-gravelly deposits, which overlay Messina Gravel and Pezzo Con-
glomerate [Jamiolkowski & Lo Presti, 2003; Fiammenghi et al., 2009; Faccioli & Vanini,
2004]. (Coastal plane deposits had probably undergone liquefaction during the disasterous 1908
earthquake M
w
= 7+ which occurred in Messina).

Figure 2. Front and side elevations of the main tower of the Messina Bridge (courtesy of M. Jami-
olkowski).

The proposed tower foundation (Figure 3) will consist of two circular massive concrete foun-
dations (connected with a crossbeam), surrounded by diaphragm walls with tips at levels60
m and 50 m. The diameter of each circular foundation is 48 m. Jetgrouted sand is used below
and outside the diaphragm walls to improve the underlying soil. The immense dimensions and
the unique shape of the proposed foundation system were chosen so that the Bridge can with-
stand severe earthquake motion (Jamiolkowski) . In the framework of this study, the influence of
the width and depth of soil improvement beneath the foundation is investigated. Moreover, the
response of a smaller less conservative foundation is examined and compared to the proposed
design.

439
concrete
B = 50 m
H
emb
= 17.5 m
H
tot
= 50 m
V
s
= 500 m/s V
s
= 200 m/s H = 32.5 m
bedrock
concrete
B B B B
H
emb
= 17.5 m
V
s
= 200 m/s
V
s
= 500 m/s
-725
concrete
H
emb
= 17.5 m
V
s
= 500 m/s
V
s
= 200 m/s
-725 +725 -725 +725
H H H H
H = 32.5 m
(a)
(b)
(c)
concrete
B = 50 m
H
emb
= 17.5 m
H
tot
= 50 m
V
s
= 500 m/s V
s
= 200 m/s H = 32.5 m
bedrock
concrete
B B B B
H
emb
= 17.5 m
V
s
= 200 m/s
V
s
= 500 m/s
-725
concrete
H
emb
= 17.5 m
V
s
= 500 m/s
V
s
= 200 m/s
-725 +725 -725 +725
H H H H
H = 32.5 m
(a)
(b)
(c)
Figure 3. The Calabrian Tower foundation : (a) plan view, (b) transverse cross-section, and (c) longitudi-
nal cross-section (courtesy of M. Jamiolkowski).
2. HOW BENEFICIAL IS THE PROPOSED GROUND IMPROVEMENT, AND TO WHAT
EXTENT ?
The effect of the grouted area is certainly positive for static loading. However, under dynamic
excitation, the presence of improved, and hence stiffer, soil may perhaps lead to amplification
of the input motion, thus cancelling some of the benefits of increasing stiffness.
A parametric study is conducted to investigate the influence of the width and depth of the
grouted area under the tower foundation. An equivalent plane strain B = 50 m footing, embed-
ded at a depth of 17.5 m is considered to represent the Calabrian Tower foundation. The con-
figuration of the simplified layout studied herein is depicted in Figure 4a. The foundation is
supported on jetgrouted soil, which terminates at a depth of 50 m (bedrock). The shear wave
velocity, Vs, is assumed equal to 200 m/s and 500 m/s for the nonimproved and the jet
grouted sand, respectively. The lateral boundaries of the model are at distance equal to fourty
times the depth of the embedded foundation (40H
emb
), which constitutes the best trade off be-
tween simulation realism (i.e. avoidance of parasitic boundary effects) and computing time.
To evaluate the effect of the extent of ground improvement, seven scenarios of variable
grouting width are first examined : B/ B = 0, 1, 2, 3, 4, 8, , where B = 50 m the width of the
foundation.
Then, five scenarios of variable grouting depth are examined: with respect to the depth H of
the improved-soil area, we examine models with H/H = 0, 0.5 (i.e. improvement at an area al-
most equal to the H
emb
), 0.75 (improvement ending 8 m above the bedrock level), 0.88 (im-
provement ending 4 m above the bedrock level), and 1 (improvement ending at the bedrock
level). Figures 4b and 4c depict the schematic views of the scenarios mentioned above.

Figure 4. (a) Simplified layout used to simulate soilfoundation interaction concerning the Calabrian
Tower foundation, (b) evaluation of the influence of the width of the grouted area, and (c) evaluation of
440
H
tot
= 50 m
concrete
B = 50 m
V
s
= 200 m/s
-725 +725 -725 +725
concrete
V
s
= 500 m/s
50 m
concrete
V
s
= 200
m/s
50 m
H
tot
= 50 m
(a)
(b)
H
tot
= 50 m
concrete
B = 50 m
V
s
= 200 m/s
-725 +725 -725 +725
concrete
V
s
= 500 m/s
50 m
concrete
V
s
= 200
m/s
50 m
H
tot
= 50 m
(a)
(b)
the influence of the depth of the grouted area.

To derive deeper insights, three other idealized models are also considered: (i) a model of to-
tal width equal to the foundation, to ignore the kinematic response of the neighboring soil layer
(the boundaries coincide with the foundation edges); (ii) the same model, but with the bottom
soil layer improved; (iii) a concrete column resting on the bedrock, surrounded by nonjet
grouted soil (Figures 5a and 5b, respectively).
The aforementioned soilfoundation systems were subjected to dynamic excitation. Idealized
pulses (Ricker wavelets and Tsangtype pulses) and a real earthquake record (the JMA record
from Kobe 1995) were applied at bedrock level. Figure 6 illustrates the elastic response spectra
of the excitations used in the analyses.

Figure 5. Evaluation of the influence of the grouted area using idealized models : (a) a concrete column
resting on the bedrock, surrounded by nonjetgrouted soil ; (b) a model of total width equal to the founda-
tion, containing jetgrouted soil, and a model of total width equal to the foundation, containing non
improved soil.

441

0
0.5
1
1.5
2
2.5
3
0 0.5 1 1.5 2 2.5 3
Ricker 1 (T = 0.6 sec, a = 0.2 g)
Ricker 4 (T = 0.15 sec, a = 0.2 g)
Tsang 2 (a = 0.2 g)
JMA-000

Figure 6. Elastic response spectra of the excitations utilized in the analyses.
From the investigation of the idealized models it is derived that, interestingly, soil improve-
ment leads to a mechanism change in the response of the soilfoundationpier system. Without
improvement, the response is dominated by horizontal translation, resembling the behavior of a
shear beam. In contrast, with soil improvement the response of the system is dominated by
rocking.
The effect of the width of soil improvement is depicted in Figure 7, where the response of all
seven alternatives when subjected to JMA record is illustrated, in terms of horizontal accelera-
tion and displacement (both computed at the pier base), and rotation. It is revealed that rotation
is larger for the B/ B = 1 model and decreases dramatically when 2 B . Moreover, when
soil improvement is applied in an area twice the foundation width or wider, horizontal accelera-
tion and displacement reach relatively low values.

442

0
0.5
1
1.5
2
2.5
3
0 5 10 15 20 25 30
0
0.1
0.2
0.3
0.4
0.5
0 5 10 15 20 25 30
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
0 5 10 15 20 25 30
a a
h
o
r
h
o
r
(
g
)
(
g
)

x

(
m
)
x

(
m
)
B / B
(a)
(b)
(c)

( (
r
a
d
)
r
a
d
)

Figure 7. Synopsis of the effect of the width of the grouted area results for all seven alternatives sub-
jected to the JMA record : (a) peak acceleration at the pier base; (b) peak lateral displacement at the pier
base; and (c) peak rotation at the base of the pier, all as a function of B/

The response to the JMA record of the models in which depth of soil improvement is para-
metrically varied, is illustrated in Figure 8. Apparently, rocking phenomena are more intense
when soil improvement ends slightly above the bedrock, i.e. when the flexible caissonlike im-
proved soil column is floating above the bedrock. However, rotational response is not dominant
when the grouted area is extended at a depth almost equal to the depth of the embedded founda-
tion.
Based on the aforementioned, a combination, with jetgrouting to a depth below the foun-
dation base almost equal to the depth of the embedded foundation and a width twice that of the
foundation is considered optimum in terms of seismic performance, at least for the cases exam-
ined herein. The existence of the improved soil layer at an area of that size, leads to attenuation
of horizontal acceleration and displacement, without causing any substantial rotation of the
pier.

443

a a
h
o
r
h
o
r
(
g
)
(
g
)

x

(
m
)
x

(
m
)
(a) (b)
(c)
0
0.5
1
1.5
2
0 0.2 0.4 0.6 0.8 1
0
0.1
0.2
0.3
0.4
0.5
0 0.2 0.4 0.6 0.8 1

( (
r
a
d
)
r
a
d
)
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
0.0035
0 0.2 0.4 0.6 0.8 1
H / H
H / H
H / H

Figure 7. Synopsis of the effect of the depth of the grouted area results for all five alternatives subjected
to the JMA000 record : (a) peak acceleration at the pier base; (b) peak lateral displacement at the pier
base; and (c) peak rotation at the base of the pier.

2. FEASIBILITY OF A LESS CONSERVATIVE FOUNDATION SCHEME
Realistic modeling of soil nonlinearity is mandatory, especially when considering heavily
loaded foundation systems lying on relatively soft soils. Besides, the idea of taking advantage
of soil inelasticity in order to limit the structural distress during strong seismic shaking has
proven beneficial in a variety of cases. In the framework of this study, we compare the response
of the proposed foundation to a less conservative alternative.
A series of two dimensional finite element analyses is performed for the two foundation sys-
tems : B = 40 m and H
emb
= 10 m, B = 50 m and H
emb
= 17.5 m. A lumped mass structure with
an equivalent planestrain square footing is considered to represent the bridgepierfoundation
system. The action of the cables and the deck upon the bridge tower are replaced by an equiva-
lent vertical force of 1 GN acting on the top of the tower. The tower is simulated with linear
beam elements, with a distributed mass. The soil and the footing are simulated with quadrilat-
eral 2D plane strain elements, and the mass of the footing is also considered. The horizontal
spring, representing the horizontal constraint at the top of the tower due to the presence of the
suspension cables, ke is considered as follows:
el
c c
e
c c
e
L
A E
L
A E
k + = (1)
where A
c
, E
c
are the area and the Youngs modulus of the main cable, respectively, and L
e
, L
el

are the equivalent lengths of the cables (Younis and Gazetas, 1996). Figure 8 illustrates the
simplified layout utilized in the analyses, along with the quantities studied herein.
444

u
r
: rotational component
u
b
: struct. bending component
B
H
emb
V
s
= 200 m/s
-175 +175
H
pier
= 383m

u u
b b u u
r r
u u
tot tot
w w
dyn dyn
a a
E E
u
r
: rotational component
u
b
: struct. bending component
B
H
emb
V
s
= 200 m/s
-175 +175
H
pier
= 383m

u u
b b u u
r r
u u
tot tot
w w
dyn dyn
a a
E E

Figure 8. The configuration of the soilfoundationbridge pier system considered for the nonlinear
analyses.

1.1 Constitutive Modelling
Gap elements simulate the soilfooting interface, allowing for due consideration of sliding
and separation. The rigid bottom boundary, is placed at a depth of 50 m.
The response of the two systems to static and dynamic loading is investigated in terms of
nonlinear behavior of the soil and the superstructure. For all stages of analysis, P (second or-
der) effects are taken into consideration. Nonlinear soil behaviour is modeled with an elasto-
plastic constitutive relation with the Von Mises failure criterion, nonnlinear kinematic harden-
ing, and associative plastic flow rule. Specifically, a pseudostatic pushover analysis is first
conducted with a MohrCoulomb model, and then an equivalent S
u
is computed for the Von
Mises model (through trial and error method), to match the results. After multiple attempts, two
alternatives were calibrated: (i) an equivalent homogeneouskinematic hardening model, and,
(ii) an equivalent layered kinematic hardening model, with S
u
linearly increasing with depth. In
both cases, the matching was performed conservatively (i.e. the ultimate load of the cali-
brated Von Mises model was lower than the one predicted using the MohrCoulomb model).
The aforementioned soilfoundationbridge pier systems were also subjected to dynamic ex-
citation. Apart from idealized pulses (Ricker pulses and modified Tsang wavelets), and real
earthquake records, modified records aiming to match with the design spectrum of the Messina
Bridge were also utilized as bedrock seismic excitation to the nonlinear time history analyses.
The design spectrum of the Messina Bridge, is the result envelope of several spectra corre-
sponding to real or fitted time histories. The design earthquake of return period 2000 years is
considered, with peak ground acceleration in the free field equal to 0.64 g.
Two original records were modified: (i) the Aegion record, from an earthquake of moderate
intensity experienced in Greece in 1995, and (ii) the TCU 068ew record, from an earthquake
of high intensity occurred in Taiwan, 1999. The modification was done using the SPECPRO
2008 [Gerolymos, 2008]. Specifically, an iterative procedure is adopted in which an input mo-
tion is modified, as far as its frequency content is concerned.
445

-1
-0.5
0
0.5
1
0 5 10 15 20 20
0
0.25
0.5
0.75
1
1.25
1.5
0 1 2 3 4 5

-1
-0.5
0
0.5
1
0 5 10 15 20

a

(
g
)
a

(
g
)
20
0
0.25
0.5
0.75
1
1.25
1.5
0 1 2 3 4 5

T (sec) T (sec)
t (sec) t (sec)
fitted
original

S S
a a
(
g
)
(
g
)

T (sec) T (sec)
t (sec) t (sec)
(a)
(b)

-1
-0.5
0
0.5
1
0 5 10 15 20 20
0
0.25
0.5
0.75
1
1.25
1.5
0 1 2 3 4 5

-1
-0.5
0
0.5
1
0 5 10 15 20

a

(
g
)
a

(
g
)
20
0
0.25
0.5
0.75
1
1.25
1.5
0 1 2 3 4 5

T (sec) T (sec)
t (sec) t (sec)
fitted
original

S S
a a
(
g
)
(
g
)

T (sec) T (sec)
t (sec) t (sec)
(a)
(b)

Figure 9. Original accelerogram and modified motion fitted to the target design spectrum, along with their
elastic response spectra : (a) Aegion 1995, and (b) TCU 068ew record from Chi-Chi 1999.

The response of the two alternatives to the two spectrum-compatible motions is depicted in
Figure 10 in terms of horizontal acceleration (at the pier base), settlement, rotation (at the pier
base) and horizontal displacement at the deck level (horizontal drift). The amplitude of horizon-
tal displacement at the deck level is a matter of great importance for the response of bridges.
The total displacement is composed of a rotational component (u
r
), which is the result of the ro-
tation of the footing, and a structural bending component (u
b
), which is the result of pier bend-
ing.
It is revealed that that the dynamic behaviour of the less conservative foundation is congruent
with the concept studied herein: taking advantage of soil inelasticity in order to limit the struc-
tural distress during strong seismic events. Indeed, the acceleration transmitted to the super-
structure is effectively limited, for both excitation cases. However, the increased settlement of
the tower is the main (and severe) penalty to pay: settlement roughly reaches 55 cm subjected
to the Aegion motion. But notice that the conservatively chosen alternative foundation system
chosen, also leads to appreciable rotation, which causes increased horizontal drift at the deck
level..
3. TENTATIVE CONCLUSIONS
The main conclusions of this exploratory study can be summarized as follows :

The effect of grouting area is certainly positive for static problems. However, under dynamic
excitation, the presence of improved, and hence stiffer, soil may lead to amplification of the
input motion, worsening the seismic response of the soil-foundation system. Interestingly,
soil improvement also leads to a mechanism change in the response of the soilfoundation
pier system. Without improvement, the response is dominated by horizontal translation, re-
sembling the behavior of a shear beam. In contrast, with soil improvement the response of
the system is dominated by rocking. This phenomenon is more intense when the improve-
446
ment ends slightly above the bedrock, i.e. when the flexible caissonlike improved soil col-
umn is floating above the bedrock.

-1
-0.5
0
0.5
1
0 2 4 6 8 10
-1
-0.5
0
0.5
1
-0.6
-0.4
-0.2
0
-0.6
-0.4
-0.2
0
-0.01
-0.008
-0.006
-0.004
-0.002
0
0.002
0.004
-0.01
-0.008
-0.006
-0.004
-0.002
0
0.002
0.004
-6
-3
0
3
0 2 4 6 8 10
-6
-3
0
3
0 2 4 6 8 10
a

(
g
)
a

(
g
)
w w
b b
(
m
)
(
m
)

( (
r
a
d
)
r
a
d
)
u

(
m
)
u

(
m
)
t (sec) t (sec)
t (sec) t (sec)
Aegion
Aegion
Aegion
Aegion
TCU 068ew
TCU 068ew
TCU 068ew
TCU 068ew
B = 50 m, H
emb
= 17.5 m
B = 40 m, H
emb
= 10 m
B = 50 m, H
emb
= 17.5 m
B = 40 m, H
emb
= 10 m
B = 50 m, H
emb
= 17.5 m
B = 40 m, H
emb
= 10 m
B = 50 m, H
emb
= 17.5 m
B = 40 m, H
emb
= 10 m
B = 50 m, H
emb
= 17.5 m
B = 40 m, H
emb
= 10 m
B = 50 m, H
emb
= 17.5 m
B = 40 m, H
emb
= 10 m
B = 50 m, H
emb
= 17.5 m
B = 40 m, H
emb
= 10 m
B = 50 m, H
emb
= 17.5 m
B = 40 m, H
emb
= 10 m

Figure 10. Dynamic analysis of the two alternative foundations subjected to the modified Aegion and
TCU 068ew motions : acceleration at the pier base a, foundation settlement w
b
, rotation at the base of
the tower, and drift u at the top of the tower.

Rotational response is not dominant when the grouted area is extended at a depth almost
equal to the depth of the embedded foundation and at a width twice as much. Besides, the
existence of the improvedsoil layer leads to attenuation of horizontal acceleration and dis-
447
placement. Thus, this combination is considered optimum in terms of seismic performance,
at least for the cases examined herein.
Realistic modelling of soil inelasticity the structural distress during strong seismic shaking
is reduced. This concept is investigated in this study, comparing the response of the con-
servative foundation to an underdesigned alternative of smaller dimensions. It is shown
that the acceleration transmitted onto the Tower is effectively limited due to mobilization
of mechanisms. However, the increase of foundation settlement along with substantial in-
crease of rotation and horizontal displacement is a serious disadvantage.
ACKNOWLEDGEMENT
th
Research Council Programme Ideas, Support for Frontier Research Advanced Grant. Con-
tract number ERC-2008-AdG 228254-DARE. Professors M. Jamiolkowski and E. Faccioli
kindly provided two of their powerpoint presentations on the geotechnical aspects of the foun-
dations and the seismic hazard of the site. The authors would also like to thank Dr. Nikos Gero-
lymos for kindly providing the SPEC-PRO software for computing spectrum-compatible mo-
tions.
REFERENCES
ATI IMPREGILO (2005), Messina Strait Crossing Bridge, Tower foundations, Geotechnical Design Re-
port, by COWI A/S,
Apostolou, M. Gazetas, G. Garini, E. (2007) , Seismic response of slender rigid structures with founda-
tion uplifting, Soil Dynamics and Earthquake Engineering, Vol. 27, No. 7, pp. 642654.
Faccioli, E., Vanini, M. (2004), Ground motion simulation of the 1908 Messina-Reggio earthquake as a
tool for verifying the design earthquake of the Messina Straits Bridge, Proceedings of the 1st Interna-
tional Conference of Applied Geophysics for Engineering, Messina, 1315 October 2004, pp. 83.
Fiammenghi G., Jamiolkowski M., Van Impe W.F. (2009), Geotechnical engineering issues related to the
Messina Strait Crossing, Deep Foundations on Bored and Auger Piles, Van Impe and Van Impe (eds.),
Taylor & Francis Group, London, UK.
Gazetas, G. (2006), Seismic Design of Foundations and Soil-Structure Interaction, Proc. of the 1st
European Conference on Earthquake Engineering and Seismology, Geneva, Switzerland, September.
Gerolymos N. (2008), SPEC-PRO : A computer code for spectral matching of acceleration time histories.
Report, NTUA.
Jamiolkowski, M., and Lo Presti, D.C.F. (2003), Geotechnical characterization of Holocene and Pleisto-
cene Messina sand and gravel deposits. Characterisation and Engineering Properties of Natural Soils.
Lisse, The Netherlands, pp. 1087-1119.
Kirtas, E. Trevlopoulos, K. Rovithis, E. Pitilakis, K. (2007), Fundamental period of Sdof systems includ-
ing soilstructure interaction and soil improvement, Proceedings of 2nd JapanGreece workshop on
Seismic Design, Observation, and Retrofit of Foundations, Tokyo.
Kourkoulis, R. (2003), Static and Dynamic geotechnical analysis of the M3 pier of the RionAntirion
Bridge, Diploma Thesis, National Technical University of Athens.
Younis, C.J & Gazetas G. (1996), Dynamic response of a tower-pier system on viscoelastic foundation
with frictional interface, Computers & Structures, 18(7), pp. 546-557.

448
5.
Dams,Slopes,RetainingWalls
1 INTRODUCTION
Wenchuan earthquake of magnitude 8.0, struck Sichuan, China on May 12 2008, has taken a toll
of more than 69000 people. Many infrastructures were damage due to unexpected strong shak-
ing of the earthquake as well as landslides. This report gives an overview of the performance of
four high dams, some bridges and landslides in the affected area.

2 BEHAVOIR OF DAMS
Hundreds of dams close to the source area were damaged some degree during the earthquake,
but none collapsed to cause flood. Most of damaged dams are earth dam with a height less than
20m, some of them were in unsafe condition before the earthquake. 4 dams more than 100m in
height have suffered strong shaking.

Zipingpu concrete faced rockfill dam, 156m in height, is located 17km from the epicenter,
Wenchuan, but the fault distance is less than 10km. The design GPA was 0.26g for the dam and
the water level was 830m when earthquake occurred, about 47m below the normal water level.
Figure 1 shows the dam after the earthquake, and the damage was very limited. The acceleration
responses of the dam are given in Figure 2, the maximum peak is more than 800gal.

Maximum settlement of the dam of 81cm and horizontal deflection of 30cm towards down-
stream were measured after the earthquake. The main damage of the dam was on the concrete
face at El. 845m, as shown in Figure 3. More than 200m long dislocation appeared along the
horizontal construction joints, the maximum sliding was about 15cm, and some vertical con-
struction joints were damaged too. Damage to the downstream stone pitching slope protection
was obvious (Figure 4 and Figure 5) and the transmission facility (Figure 6), the sluice gate
building were severely damaged.

Field Observations of Wenchuan Earthquake of May 12 2008
Haibo WANG

Professor
China Institute of Water Resources and Hydropower Research, Beijing, CHINA
ABSTRACT: Wenchuan earthquake of magnitude 8.0, struck Sichuan, China on May 12 2008,
has taken a toll of more than 69000 people. Many infrastructures were damage due to the earth-
quake. This report gives an overview of the performance of four high dams, some bridges and
landslides in the affected area.
449

Figure 1 Zipingpu concrete faced rock dam after the earthquake

Figure 2 Seismic recorders on the right side of the dam near the top
450

Figure 3 sliding of the concrete face upstream at the construction joints of Level 845m

Figure 4 sliding of the downstream stone pitching
451

Figure 5 Distortion of step guards on downstream slop

Figure 6 Damage of the transmission facility
452
Shapai RCC arch dam (Figure 7 and Figure 8) of 132m in height was completed in June 2003. It
was designed with a ground acceleration of 0.14g. The distance from the dam site to the fault
plane of Wenchuan earthquake is about 20km. The water level during the earthquake was close
to the normal level, 1.5m below the top of the dam. There are two contraction joints and two in-
duced joints along the arch direction.

Except some cracks on the columns of the intake gate frame (Figure 10) and the hoistway on the
crest, no visible damage on the dam itself was found, although sever landslides near the dam site
can be seen, which blocked the roads for a long time. But the power plant, located about 3.5km
away, has been inundated due to the collapse of the partial penstock (Figure 11).

Figure 7 Downstream view of Shapai RCC arch dam

453

Figure 8 Shapai RCC dam in construction

Figure 9 Photo from helicopter of Shapai Arch dam after earthquake
454

Figure 10 Cracks on the columns of the intake gate frame

Figure 11 Power plant and water from the broken penstock

455
Baozhusi gravity concrete dam of maximum height of 130m was completed in 1998, it com-
poses of 27 monolith in a total length of 524.48m, as shown in Figure 12. The water level was
558.19m on May 11, 2008 and design water level is 588m.

No damage was found on the dam, except some cracks due to the opening of the contraction
joints between the monolith on the crest and inside the gallery (Figure 13 to 15). The horizontal
movement is 2.1mm towards upstream at the top of 17th monolith of the dam, 0.98mm towards
upstream at the base of the same monolith after the earthquake. Pore water pressure measured at
the dam base gives little change after earthquake, but some filters at the end of the drain pipe in
the gallery were blown up during the earthquake (Figure 16), demonstrating high transient pore
pressure during the earthquake.

As the orientation of the dam axis, NE393442 is close to that of Longmenshan fault, NE30
to 50, the strong shaking was almost parallel to the dam axis, and the intensity in stream direc-
tion was much weaker. An auxiliary steel frame weighted 33 ton on the dam top was moved
about 43cm in cross river direction and 5cm in downstream direction, as shown in Figure 17,
indicating the direction of the quake.

Figure 12 Birdview of Baozhusi gravity concrete dam
456

Figure 13 Downstream view of Baozhusi dam

Figure 14 Upstream view of Baozhusi dam
457

Figure 15 Crack at the contraction joint

Figure 16 Blown up drain pipe in the gallery
458

Figure 17 Movement of the auxiliary steel frame

Bikou loam core embankment dam of height 101.8m was completed in 1976, with a length of
297.36m on the crest, shown in Figure 18 and Figure 19. The water level was 691.41m when the
earthquake occurred, 12.59m lower than normal water level.
Maximum settlement was 24.2cm on the crest, and the horizontal motion was in upstream direc-
tion, generally, with the maximum displacement of 15.4cm on the top. The motion of down-
stream slope is towards downstream, and the maximum displacement of 7.1cm was occurred at
the EL.670m, near the middle of the dam in height. No other damage was reported, except the
cracks of the break made gate lifter building on the crest (Figure 20).

459

Figure 18 Bikou loam core embankment dam

Figure 19 Bikou dam from upstream
460

Figure 20 The gate lifter building

3 LANDSLIDE

Abundance of landslides and slope failures can be seen on the disaster zone. The landslides ob-
served were largely rock falls, shallow rock and debris slides triggered by the strong ground
shaking, Figure 21. Deep seated landslides involving failure through the rock mass also oc-
curred but were less common. Large scale landslides blocked roads and giant falling rocks de-
stroyed the bridges, house and other structures (Figure 22 and Figure 24), therefore cut the routs
to the towns and villages in the mountainous areas where help was needed. Secondary land-
slides continued to hinder the deployment of aid and reconstruction in the weeks due to the af-
tershocks. In Yingxiu town, almost 90% of the mountain around lost the original green (Figure
23 and Figure 24).
461

Figure 21 Destroyed roads due to landslides in Hanwang town

Figure 22 Houses destroyed by falling rocks
462

Figure 23 Landslides in Yingxiu town

Figure 24 Landslides in Yingxiu town

4 BRIDGE
Many bridges were damaged due to strong shaking, landslides and falling rocks. At the time of
field visit, some of the damaged bridges were temporarily strengthened to for rescue activities
(Figure 25 and Figure 26). As shown in Figure 27 and Figure 28, one span of Miaoziping Minji-
463
ang Bridge on Dujiangyan to Wenchuan highway has completely fallen. And Baihua Bridge of
State Road No 213 near Yingxiu town has been completely destroyed on curved section due to
ground motion (Figure 29). Severely damaged part of the bridge was removed soon after earth-
quake since a temporary route had been constructed just next to it. Yingxiu river side bridge
along Minjiang completely collapsed (Figure 30), while Yingxiu Minjiang Bridge several hun-
dreds meters away destroyed by landslide in one end (Figure 31) and the slabs shifted from the
pier in river direction for almost one meter (Figure 32).

Figure 25 Yuzixi Bridge in Yingxiu town
464

Figure 26 Shoujiang Bridge of the road from Dujiangyan to Yingxiu town

Figure 27 Miaoziping Minjiang Bridge on Dujiangyan to Wenchuan highway
465

Figure 28 Miaoziping Minjiang Bridge on Dujiangyan to Wenchuan highway

466

Figure 29 Baihua Bridge of State Road No 213 near Yingxiu town

Figure 30 Yingxiu river side bridge along Minjiang

467

Figure 31 Yingxiu Minjiang Bridge destroyed by landslide in one end

Figure 32 Yingxiu Minjiang Bridge permanent shifting

468
5 CONCLUSIONS
The ground motions recorded in the near-source region of Wenchuan earthquake were much
larger than ground motion used in the design for the area. The activity of Longmenshan fault
has been underestimated before the earthquake. Damage and loss of life were caused by not
only strong ground shaking but also landslides which made rescue activities much more difficult
for the mountainous area. From the observation of the performance of four well designed high
dams, the damage was limited, which show high capacity of the dams against earthquakes, but
more investigation are needed in order to make clear the real behaviors of the dams.

REFERENCES
Renqun WANG, 2009, Behavior of Shapai arch dam in Wenchuan Earthquake(in Chinese)
M. Free, 2008, The Wenchuan Earthquake Of May 12, 2008: Field Observations And Recommendations
For Reconstruction

469

1
1 INTRODUCTION
The strong earthquake (M = 7.8) that shook the Sichuan region, China, on the May 12, 2008,
caused significant cracks in the concrete slab of the 156 m-high Zipingpu dam and, at least for
several hours, some anxiety for the safety of the highly populated areas in the downstream re-
gion. The dam suffered significant damage during this very strong ground shaking, but it did not
fail. Soon after the earthquake, the damage was repaired and the dam became operational again.
Due to the high shear strength of compacted rockfill and the lack of water pressures, concrete
faced rockfill dams (CFRDs) such as Zipingpu are considered to behave better during earth-
quakes compared to conventional rockfill dams with an earth core (Wieland 2007, Gazetas &
Dakoulas 1992, Sherard & Cooke 1987). However, this assessment is based on a very limited
number of case histories of such dams. Moreover, there is presently very little rigorous analyti-
cal evidence about the performance of the concrete slab in CFRDs built in narrow canyons. The
Zipingpu Dam incident demonstrated very clearly the need to improve our basic understanding
of the slab performance of CFRDs during strong earthquakes.
The seismic performance of a CFRD depends on a large number of factors, such as the qual-
ity of compaction, the quality of rockfill, the dam geometry, the narrowness of the canyon, the
irregularity of the abutment surface, the flexibility of the canyon rock, the quality of joints be-
tween adjacent slab panels and along the plinth, the ground motion intensity, characteristics and
spatial variability, etc. (Sherard & Cooke 1987, Gazetas & Dakoulas 1987, 1992, Uddin & Ga-
zetas 1995, Dakoulas & Hsu 1995, Abouseeda & Dakoulas 1998, Wieland 1997).
To be able to study the effect of such factors on the structural integrity of the entire face slab,
a realistic simulation of the phased construction and loading due to reservoir impoundment is an
essential requirement. Indeed, the development of high static stresses in the slab, due to creep
settlements and reservoir filling, affects its capacity to sustain additional seismic stresses with-
Lateral and Longitudinal Seismic Vibrations of Concrete Face
Rockfill Dams

P. Dakoulas
University of Thessaly, Volos, Greece

ABSTRACT: A general methodology is presented for nonlinear 3D finite element analysis of
concrete faced rockfill dams based on an integrated simulation of the phased construction, creep
settlements, reservoir impoundment and seismic shaking of the dam, accounting for dynamic
settlements. The static analysis uses the nonlinear model by Duncan et al. (1970), whereas the
dynamic analysis uses a nonlinear hysteretic model which accounts for the initial dynamic stiff-
ness of the material and produces hysteresis loops in agreement with the experimental data on
the shear modulus and damping ratio variation with cyclic shear strain. A damage plasticity
model is used for the reinforced concrete slab. Apart from the overall behavior of the dam, the
method allows a detailed evaluation of (a) the tensile stresses within the concrete slab panels,
(b) the compressive stresses at the slab-to-slab vertical interfaces and (c) the opening of the
joints. It is applied to study the response of a 150 mhigh CFRD in a narrow canyon.

470

2
out significant cracking. Experience from the static performance of CFRDs shows that many
such dams have developed small or significant cracks in the concrete slab. The most common
factors for such crack development are (Marques Filho & Pinto 2005): (a) a low stiffness of the
rockfill material in the upstream and downstream zones due poor compaction, material type,
improper wetting, etc., resulting to significant rockfill settlements; (b) a significantly lower
stiffness in the downstream side compared to that in the upstream side, causing higher deflec-
tions in the upper part of the slab; and (c) the presence of local steep plinth inclinations and ir-
regular rock surface at the abutments, causing stress concentrations and convex curvature in the
slab.
The objective of this paper is to present a methodology for a realistic evaluation of the seis-
mic behavior of the dam and slab, based on an integrated, refined simulation of the phased con-
struction, reservoir filling and earthquake shaking. The multi-stage static analysis is based on
the Duncan et al. (1970) constitutive model, whereas the dynamic analysis uses a nonlinear hys-
teretic model which accounts for the initial dynamic stiffness of the rockfill. The dynamic
model produces hysteresis loops that are in agreement with the experimental data regarding the
variation of the shear modulus and damping ratio with cyclic shear strain.
The methodology is applied here to study the static and seismic performance of the concrete
slab of Messochora Dam, built in a narrow gorge of the Acheloos River in the north part of
Greece (Fig. 1). The dam has a height of 150 m at its maximum cross-section, a crest length of
350 m, and supports a 160 MW power station. The reservoir has a live storage volume of
228*10
6
m
3
. The dam was completed in 1995, but environmental terms have been reconsidered,
causing a delay to its first reservoir impoundment. The geometry of the maximum cross-section
of the dam and its material zones are shown in Fig. 2a. The upstream slope is 1:1.4, whereas the
downstream slope is 1:1.4 at the lower 110 m and 1:1.55 at the upper 40 m. Zones 3B and 3C
consist of rockfill obtained from quarries or necessary excavations, utilizing healthy or slightly
weathered limestone. The material was compacted in four passes by a 12 ton vibratory roller,
using a layer thickness equal to 1 m for zone 3B and 1.5 m for zone 3C. The average void ratio
of the rockfill after compaction is 0.38. The concrete slab was placed upon a 4 m thick layer of
well graded gravel (zone 2B), which due to its gradation and heavier compaction is quite stiffer
than zone 3B. A zoned fill was placed upon the slab up to a height of 55 m from the dam base as
an additional line of defense against seepage. Fig. 2b illustrates the discetization of the 3-D dam
geometry consisting of 60467 elements.

Figure 1. Messochora Concrete Face Rockfill Dam, Greece.
471

3

Figure 2. Messochora Dam: (a) Maximum cross-section with material zones (b) Perspective
view of the FE discretization of the dam and face slab, consisting of 23 concrete panels.

The numerical simulation is performed in eight stages:
Stage 1: The dam body is constructed in a sequence of 20 height increments.
Stage 2: The concrete slab is placed upon the upstream slope.
Stage 5: The reservoir water is raised to an el. of 55 m.
Stage 8: The canyon base is subjected to earthquake shaking

In the following, the methodology for the evaluation of the dam and slab performance is di-
vided in two parts: (a) phased construction and reservoir impoundment and (b) seismic analysis.
2 CONSTITUTIVE MODELS
The static analysis is based on the Duncan & Chang (1970) constitutive model, whereas the
dynamic analysis uses a nonlinear hysteretic model.

2.1 Duncan and Chang model
A realistic prediction of the dam performance during construction and loading requires the
use of a constitutive model that accounts for the dependency of the elastic moduli on both the
current stress state and the loading/unloading stress path. The model by Duncan et al. (1970)
considers a hyperbolic stress-strain relationship, which results into a tangent Youngs modulus
t
E given by:
472

4
2
1 3
3
3
( )(1 sin )
1
2 sin 2 cos
n
f
t a
a
R
E K p
p c

=

+

(1)
where
1 3
, = principal stresses, K ,
f
R , n =material constants, =friction angle, c =cohesion,
a
p =atmospheric pressure. The friction angle depends on the confining stress as follows

3
log
o
a
p

=

(2)
where
o
and = material constants. The unloading modulus
ur
E is given by

3
n
ur ur a
a
E K p
p

=

(3)
where
ur
K =material constant with a range of variation between 1.2 3
ur
K K K . Moreover,
the bulk modulus B is given by

3
m
b a
a
B K p
p

=

(4)
where
b
K , m =material constants. Limits to the bulk modulus may be defined by the relation-
ship ( ) / 3 1 2
t
B E v = . For a Poissons ratio v =0.49, the maximum value is equal
to 17
max t
B E = . For Poissons ratio at rest conditions, ( ) / 1
o o
v K K = + , the minimum value of
B becomes

min
1 2 sin
3 1 3 sin
t o t
o
E K E
B
K

+
= =

(5)
Following the formulation in the program SOILSTRUCT (Ebeling et al. 1992), the load-
ing/unloading criterion is based on the expression

1/ 4
1 3 3
1 3 f f
S
p

(6)
where
1 3
, = principal stresses and
1 3
,
f f
= principal stresses at failure. For
max
S S , the
loading tangent moduli
t
E are used, whereas for
max
S S < the much higher unloading tangent
moduli
ur
E are activated.

2.2 Hysteretic model
For the dynamic analysis, the shear modulus at small strains is given by
( )
2
0 0
(2.17 )
1000
1
r e
G A p
e
=
+
(7)
where
0
p =mean effective stress, e = void ratio and A, r =material constants (
0
G ,
0
p in kPa).
For monotonic loading, the hysteretic model assumes that the secant shear modulus is given by
473

5
0
/
1 exp( (log ) / )
s
e
a
G G
c b
=
+
(8)
where , , a b c = material constants. The equivalent shear strain
e
is given by

( ) ( ) ( )
2 2 2
2 2 2
11 22 22 33 33 11 12 23 31
2
3( )/2
3
e
= + + + + + (9)
For the special case of
11 22 33 23 13
0 = = = = = , eq. (9) yields that
12 e
= and
therefore the material constants , , a b c can be calibrated from experimental results on the varia-
tion of the shear modulus ratio and damping ratio versus the amplitude of cyclic shear strain.
The tangent shear modulus for monotonic loading is given by

0 2
exp( (log ) / )
/
1 exp( (log ) / ) (1 exp( (log ) / )) ln10
e
t
e e
a c b a
G G
c b b c b

= +
+ +
(10)
whereas for unloading and reloading, using the Masing criterion, it becomes

0 2
exp( (log(| | / 2) ) / )
/
1 exp( (log(| | / 2) ) / ) (1 exp( (log(| | / 2) ) / )) ln10
ref
t
ref ref
a c b
a
G G
c b b c b

= +
+ +
(11)
where = deformation vector and
ref
= the deformation vector from which unloading or re-
loading took place.
Fig. 3 presents the mean value and one standard deviation of the shear modulus ratio
0
/
s
G G
and damping ratio versus the cyclic shear strain for gravelly soils from a comprehensive
study based on results from 15 independent experimental investigations (Rollins et al. 1998).
Although there is some dependence of
0
/
s
G G and on the confining pressure, Rollins at al.
(1998) suggest that the use of the mean value will not cause significant errors for pressures
100 kPa. The figure also gives the values of
0
/
s
G G and predicted by the hysteretic model
for a = 1.014, b = -0.65 and c = -1.30. These model parameters will be used in the seismic
analysis for both gravel and rockfill in this study. The combined static-plus-dynamic model has
been implemented by the author into the FE code ABAQUS (2008) as a user defined material.
The validity of the model implementation was verified using laboratory test simulations
(Dakoulas 2007). Fig. 4 presents the imposed simple shear deformation on a soil element and
the stress-strain relationship derived from the hysteretic model using the program ABAQUS.

Table 1. Properties and model parameters for rockfill and gravel
_______________________________________________________________
Zone 3B 3C 3B
__________________________________________________
Density, , kg/m
3
2150 2150 2150
Void ratio, e 0.38 0.38 0.30
Poissons ratio, v 0.35 0.35 0.35
K 600 450 1200
ur
K 1500 1125 3000
b
K 150 112.5 300
n 0.45 0.45 0.45
m 0.22 0.22 0.22
f
R 0.59 0.59 0.59
0
51 51 51
9 9 9
A (best estimate) 16600 12450 33200
A (lower bound) 14300 10725 28600
A (upper bound) 19000 14250 38000
r 0.5 0.5 0.5
474

6

Figure 3. Dynamic properties of gravel from extensive experimental results (Rollins et al. 1998)
and hysteretic model predictions: (a) secant shear modulus ratio,
0
/
s
G G and (b) hysteretic
damping ratio, , versus cyclic shear strain.

Figure 4. Simulation of a simple shear test using the hysteretic model: (a) imposed shear strain
history (b) predicted stress-strain relationship.

Figure 5. Simulation of the stress-strain behavior of concrete in cyclic uniaxial loading using
the damage plasticity model: (a) tension test and (b) compression test (Lee & Fenves 1998)

475

7
2.3 Concrete damage plasticity model
The plastic-damage constitutive model for cyclic loading of concrete developed by Lee and
Fenves (1998) is used here for modeling the behavior of the reinforced concrete panels. The
rate-independent model takes in account the effects of strain softening, distinguishing between
the damage variables for tension and compression. A degradation model that represents the ef-
fects of damage on the elastic stiffness and the recovery of stiffness after crack closure is util-
ized. For the case of large crack opening displacements, the inelastic strain is stopped beyond a
critical value for the tensile damage variable, whereas subsequent deformation can be recovered
upon crack closure. Figure 5 plots experimental data and simulations of the stress-strain behav-
ior of concrete in cyclic uniaxial loading using the damage plasticity model from (a) a tension
test and (b) a compression test (Lee & Fenves 1998). It is noted that an extension of the rate-
independent plastic-damage model, not used in this study, allows for a rate-dependent viscoplas-
tic concrete behavior. More details about the plastic-damage constitutive model and its applica-
tion for the evaluation of the seismic response of Koyna dam in the 1967 Koyna earthquake is
given in the paper by Lee and Fenves (1998). In the following analysis the concrete slab is as-
sumed to have a density =2350, kg/m
3
, a compressive strength of 25 MPa, a tensile strength
of 3 MPa, an elastic Youngs modulus E = 21 GPa and Poissons ratio = 0.2.

Figure 6. (a) Comparison of measured settlements and computed settlements at the end of con-
struction and the end of creep versus distance from the crest. (b) Distribution of settlements
from numerical simulation at the end of creep.
3 SIMULATION OF PHASED CONSTRUCTION
Stage 1: The phased construction process is simulated by using 20 increments of the dam
height, each having a thickness of 7.5 m. A refined finite element mesh is utilized, in which the
dam body is divided into 54827 elements (Fig 2b). The numerical model represents accurately
the 3D geometry of the abutment surfaces. The cross section of the dam is divided in 3 main
zones: (a) the rockfill zone 3B, (b) the rockfill zone 3C and (c) the gravel zone 2B (Fig. 2a).
Since no direct measurements of the material stiffness are available, the initial values of the
material parameters are obtained first from published data for similar rockfill materials used in
similar dams (Duncan et al. 1980) and, subsequently, are calibrated utilizing available settle-
ment measurements taken during construction. Thus, the values of K and
ur
K in eqs (1) and
(3) are calibrated using settlement measurements along the height of the dam at the maximum
cross-section. The model parameters used for the rockfill and gravel materials are given in Table
1. In order to simulate, in an approximate way, the recorded continued settlements after con-
struction of the dam and to account for the imposed deformation on the slab, only 94% of the
476

8
gravity load is applied to the dam materials before slab construction, whereas the rest 6% is ap-
plied after placement of the slab and the fill, in Stage 4. Fig. 6a compares the distribution of
computed settlements at the middle cross-section of the dam at the end of Stage 1 and the corre-
sponding measured settlements by the instrument IDEL L+IS-1 (shown in Fig. 2a). At the end
of Stage 1, the maximum settlement is about 2.25 m and is located at about mid-height. The
simulation of Stage 1 is repeated 3 or 4 times and in each analysis the initial geometry of the
dam is modified such as, after application of 94% of gravity, its upstream surface forms a per-
fect plane. The slab is placed on this plane in Stage 2.
Stage 2: The concrete slab consists of 23 independent panels that are discretized with 5640
elements. The geometry of the numerical model of each panel is identical to its actual geometry.
The width of each panel is 15 m, except of the first and last ones where it is about half of this
value. The thickness of each panel is 0.30 m at crest and varies with the vertical distance from
the maximum water level, h, according to 0.30 0.003 w h = + . For h= 148 m, the thickness at
the base of the slab in the middle section is 0.744 m. The thickness of the slab is discetized us-
ing two solid finite elements. The other dimensions of the FE range between 2 to 3.75 m, so that
the aspect ratios are within reasonable limits. The concrete behavior is modeled using the dam-
age plasticity model. Reinforcement is placed at the mid-thickness of the slab, consisting of 25
mm diameter steel bars at 15 cm distance in the longitudinal (X) and lateral (Y) directions of the
concrete panels. Additional reinforcement is placed near the base and top surfaces of the slab
within a 35 m wide perimetric (parallel to the plinth) zone. Each panel has interface properties
between its base and the upstream slope gravel, as well as between the vertical walls of adjacent
slabs. The concrete-to-gravel friction coefficient is =0.70 and the concrete-to-concrete friction
coefficient is
c
=0.5. At Stage 2 the slab is loaded by its own weight and let to rest upon the
upstream slope. Support along the perimeter of the slab simulates the support of the plinth.
Stage 3: A fill is placed upon the slab top from el. 0 m to el. 55 m (Fig. 2a). The placement
of the fill is simulated as linearly distributed normal and shear stresses applied on the top sur-
face of each panel. The fill has a maximum thickness of 25 m at the toe of the slab.
Stage 4: Gravity in the dam body is increased from 94% to 100%, causing additional settle-
ments which at the crest are equal to 13 cm. This settlement is equal to the observed settlements
of Messochora dam caused by rockfill creep. Fig. 6b plots contours of the settlements develop-
ing within the dam body after the simulation of the creep settlements. The maximum settlement
at the end of creep is 2.37 m (Fig. 6a).
Stage 5: The reservoir is filled with water to an el. of 55 m. At the same time, the stresses
imposed by the fill at Stage 3 are reduced to account for buoyancy.
Stage 6: The reservoir is filled to an el. of 100 m. During this stage, the support along the pe-
rimeter allows all slabs to move towards the dam crest, as they usually tend to do, but it restricts
their outward movement beyond the plinth in any direction.
Stage 7: The reservoir is filled to the final el. of 148 m.
4 STATIC ANALYSIS
Fig. 7a plots the deflection
z
U of the slab after reservoir filling at el. 148 m. Fig. 7b plots the
distribution of the deflection
z
U along panel No. 13 (see also Fig. 7a) versus the distance from
the panel toe for Stages 3 to 7. After application of the creep settlements,
z
U at the crest is
about 10.5 cm. During water filling to el. 100 m, the maximum value of
z
U is 15 cm at about
30 m from the slab toe. After filling to the final level, the maximum deflection is 33 cm at a dis-
tance equal to 150 m from the toe.
Fig. 8a plots the movement of the various slab panels in the longitudinal (X) direction of the
local coordinate system (XYZ) after reservoir impoundment. During reservoir impoundment the
slab tends to move towards the dam crest, due to the deformation of the underlying dam mate-
rial. The maximum movement is 3.2 cm and is located near the slab toe. This opening of the
plinth joint is less than the 10 cm limit of the waterstops along the slab-plinth interface.
Fig. 8b plots the slab movement in the horizontal (Y) direction. Due to the underlying dam
material deformation, the slab panels tend to move towards the middle section of the dam, with
a maximum displacement 6 cm. Such movement causes compressive stresses in the vertical
walls of adjacent panels, as discussed below.
477

9

Figure 7. (a) Slab deflection U
z
after reservoir filling at elevation 148 m; (b) Deflection U
z
in
panel 13 (middle section) at various stages of loading versus distance from the slab toe.

Figure 8. Slab displacement (a) in X direction and (b) in y direction, after reservoir filling at
elevation 148 m.

Figure 9. Tensile and compressive stresses at the top surface of the slab after reservoir filling at
elevation 148 m: (a) Major principal stress and (b) Minor principal stress. The magnitude of the
major principal stress is influenced mostly by the normal stress
x
.
478

10
Fig. 9a plots the distribution of major principal stress,
1
, at the top surface of the slab. Note
that the maximum tensile values of
1
depend mostly on the contribution of the normal stress
x
. Such tensile stresses appear in an approximately 35 m-wide zone along the perimeter of the
lower 2/3 the slab and may reach the tensile strength of concrete (3 MPa). Therefore some
cracks are expected to develop in this region. High tensile stresses may develop due to an abrupt
change in abutment slope or a local anomaly on the underlying base rock surface, combined
with a relatively low stiffness of the rockfill in zones 3B and 3C. For cases of irregular founda-
tion geometry, Marques Filho & Pinto (2005) suggest limited reshaping of the foundation or
changes in the rockfill zoning to reduce the propensity to convex curvatures in localized spots of
the face. The presence of a tensile zone that is parallel and above the plinth, as shown in Fig.
9a, is in agreement with the reported 10 to 20 m-wide zone of cracking in Xing, It and Itapebi
dams (Marques Filho & Pinto 2005). Also, the direction of the major principal stress
1
sug-
gests a crack direction in agreement with the direction of observed cracks in these dams.
Fig. 9b plots the distribution of the minor principal stress
3
, which attains maximum com-
pressive values mostly due to the contribution of the normal stress
y
. In a region having a
center at about 1/5 of the dam height from the toe of the slab and diameter about 50 m, the nor-
mal stress
y
between adjacent panels reaches a maximum value of about -10 MPa, less than
the compressive concrete strength ( -25 MPa).
5 DYNAMIC ANALYSIS
Due to lack of in situ measurements, the best estimate and a range of variation for the shear
wave velocity is considered, based on experience from other dams. The shear wave velocity at
small strains is computed from
0 0
/
s
V G = , where
0
G = the shear modulus at small strains
from eq. (7). The values of the parameter A for the considered three cases are given in Table 1.
Fig. 10 plots the variation of
0 s
V versus depth from the crest at the maximum cross-section.
This study considers that the design earthquake has a magnitude about M = 6.5 and a peak
ground acceleration of 0.35g. It is assumed that it is caused by a normal fault with an epicentral
distance 10 km and depth 10 km. Several seismic excitations have been used based on his-
toric and artificial records. Due to lack of space, only one excitation yielding the most critical
response is presented here. Fig. 11 plots the acceleration history and response spectra, as well as
the Eurocode spectra for rock and rock-like formations with
30 s
V 800 m/s. The excitation is
exerted initially in the upstream-downstream direction, and subsequently in the longitudinal di-
rection. The canyon rock is taken to be rigid and the hydrodynamic affects are ignored in this
study.

Figure 10. Shear wave velocity at small strains
0 S
V versus depth from the crest at maximum
cross-section (adopted best estimate value and range of variation).
479

11
For the best estimate of
0 s
V , referred to as Case 1, Fig. 12 presents the horizontal displace-
ment and acceleration in the upstream-downstream direction at mid-crest. The peak displace-
ment is 0.16 m and the peak acceleration 1.54g. The peak accelerations for the range of
0 s
V
considered are between 1.5g and 1.63g. Such high-amplitude accelerations developing in the
near mid-crest area are caused by wave focusing effects associated with the narrow canyon
geometry, as shown by several previous studies (Mejia & Seed 1982, Dakoulas & Gazetas 1987,
Gazetas & Dakoulas 1992, Dakoulas 1993, 1995), as well as by the soil inhomogeneity seen in
Fig. 10 (see Gazetas 1981, 1987). Despite their very high magnitude, such peak accelerations,
acting during a very short period, do not affect the stability of the dam. Results using other exci-
tations based on historic records obtained from rock sites yield mid-crest displacement between
0.10 m to 0.16 m, peak shear strains between 0.1% to 0.15% and peak accelerations between 1g
to 1.7g.
It is interesting to note that earlier studies of the seismic response of the same dam (Gazetas
& Dakoulas 1992) based on a simplified closed-form analytical solution using a 3D shear beam
model by Dakoulas & Gazetas (1986) have predicted crest accelerations about 1.5g, in agree-
ment with the much more elaborate 3D nonlinear analysis presented in this paper. Moreover, re-
corded crest accelerations records at Zipingpu dam, having approximately the same height and
canyon narrowness, during May 12, 2008 earthquake, are also about 1.5g.

Figure 11. Seismic Excitation: (a) Acceleration time history (b) Acceleration spectra and Euro-
code spectra for soil type A (rock and rock-like formations with
30 s
V 800 m/s).

Figure 12.Case 1: (a) Hor. lateral displacement and (b) hor. lateral acceleration time history at
mid-crest.
480

12

Figure 13. Case 1: Major principal stress in the slab at time t = 2.98 s.

Figure 14. Case 1: Minor principal stress in the slab at time t = 5.24 s.

Fig. 13 plots the major principal stress
1
within the slab at a moment in which the slab is sub-
jected to maximum tension reaching the concrete tensile strength. This maximum tension occurs
during the movement of the slab in the downstream direction at t=2.98 s (point A in Fig. 12a).
During shaking, the high tensile stresses move from the perimeter towards to central region of
the slab. Additional analyses has shown that by increasing the rockfill stiffness (
0 s
V ), the mag-
nitude and extent of tensile stresses decrease. Fig. 14 plots the minor principal stress
3
at a
moment (t = 5.24 s, point B in Fig. 12b) in which the slab is subjected to maximum compres-
sion as the slab moves towards the upstream direction. The compressive stresses reach a maxi-
mum value of -18 MPa and are located in the central area at 2/3 of the dam height.
The results in Fig. 14 were computed without accounting for any compaction during shaking.
A strong ground shaking may induce dynamic settlements, such as in the 85m-high Cogoti dam
(Chile) during the 1943 Illapel earthquake (M=7.9), in which the crest settled by 0.4 m. Simi-
larly, the Zipingpu dam during the May 12, 2008 earthquake experienced about 0.7 m of crest
settlement. It is anticipated that a CFRD during a strong shaking may experience a maximum
settlement about 0.5 to 1 m (Wieland 2007). To examine the effect of such a dynamic settlement
on the slab stress state, an analysis referred to as Case 2 is conducted, assuming a crest settle-
ment of 0.5 m.
481

13

Figure 15. Case 2 - Lateral vibrations assuming 0.5 m dynamic settlement at mid-crest: (a) Slab
deflection and (b) Displacement of the slab in the y direction.

Figure 16. Case 2 - Lateral vibrations assuming 0.5 m dynamic settlement at mid-crest: Minor
principal stress in the slab at time t = 8.84 s.

Fig. 15a plots the deflection of the slab at the end of the earthquake, which at mid-crest in-
creases from 0.22 m before the shaking to 0.83 m after the shaking. Fig. 15b, plots the dis-
placement
y
U , which increases from 0.06 m before shaking (Fig. 6b) to 0.16 m after shaking,
due to the dynamic settlement of the rockfill. Finally, the displacements
x
U of the slabs remain
< 0.05 m at all times and, thus, no leaks are expected along the plinth joints during shaking.
Results on the major principal stress
1
within the slab show that the development of a dy-
namic settlement leads to smaller tensile stresses. This is explained by the fact that, as the dam
rockfill settles, it imposes compressional
x
stresses along the slab panels, by transferring
stresses through the base friction. In the tension zones of the slab, this leads to reduction of the
tensile stresses. In the compression zones, it leads to higher compressive stresses. Indeed, Fig.
16 plots minor principal stress
3
at a moment (t = 8.84 s) in which the slab is subjected to
maximum compression. The compressive stresses reach a maximum value of -23 MPa, ap-
proaching the compressive strength of concrete, and are located in the central area of the slab.
Additional analyses have shown that for dynamic settlements of 1 m, the compressive stresses in
this region may reach the concrete strength.
482

14

Figure 17. Case 3 - Longitudinal vibrations assuming 0.5 m dynamic settlement at mid-crest:
Minor principal stress in the slab at time t = 8.84 s.

Figure 18. Case 3 - Longitudinal vibrations assuming 0.5 m dynamic settlement at mid-crest:
Displacement of the slab in the y direction.

Finally, it is of interest to examine the effect of longitudinal vibrations on the compressive
stresses developing in the concrete slab. Due to the high frictional forces between the slab and
the underlying gravel, significant inertial forces are transferred from the dam body to the slab,
which is acting as a stiff diaphragm wall. To examine this effect on the behavior of the con-
crete slabs, a parametric investigation has been conducted. In this paper, the results of a single
analysis are presented, corresponding to the dam considered in Case 2, subjected to longitudinal
vibrations using the excitation in Fig. 11.
Fig. 17 plots the minor principal stress
3
at a moment (t = 8.84 s) in which the slab is sub-
jected to maximum compression. As the dam body is displaced to the left, significant compres-
sive stresses reaching the compressive strength of concrete develop in the concrete slab panels
at a point where there is a sharp change in the plinth inclination. It should be noted that these
values of
3
are higher than those developed when the same excitation was applied in the up-
stream-downstream direction. Thus, it appears that, due to the diaphragm-wall action, vibrations
483

15
in the longitudinal direction may be more critical in terms of compressive stresses than those in
the upstream down stream direction.
Fig. 18 plots the distribution of the lateral movement
y
U at t = 9.98 s during shaking. As
shown in the figure, the horizontal displacements are significant (about 0.28 m) and therefore
opening of the vertical joints is possible. However, due to the very small duration of the joint
opening, very little water leakage, if any, is expected to take place, unless shear failure caused
by the high compression takes place.
Finally, it is interesting to note that, on one hand, the narrow canyon has an important benefi-
cial effect in increasing the overall stiffness of the dam and reducing both the construction and
dynamic settlement of the rockfill, thereby improving slab performance. For Messochora dam,
for which the slab area ( ) A to the square of the height ( ) H ratio is
2
/ A H = 2.3, the narrow
canyon geometry has indeed a major effect in reducing settlements. On the other hand, the dif-
ferential (static-plus-dynamic) settlement in the longitudinal dam direction, caused by the 3D
narrow canyon geometry, leads to the development of high compressive
y
stresses in the slab
panels.
6 CONCLUSIONS
The presented methodology for nonlinear 3D finite element analysis of CFRDs allows an in-
tegrated simulation of the phased construction, creep settlements, reservoir impoundment, and
seismic shaking of the dam, considering dynamic settlements. In addition to the overall behavior
of the dam, the method allows a detailed evaluation of (a) the tensile stresses within the concrete
slab panels, (b) the compressive stresses at the slab-to-slab vertical interfaces and (c) the open-
ing of the joints.
The following conclusions can be drawn from this study:
1. A realistic simulation of the phased-construction and reservoir impoundment of a CFRD
can be achieved using reliable measurements of construction settlements in the middle sec-
tion (and perhaps other sections) of the dam and of creep settlements after construction of
the slab. For seismic analysis, measurements of the distribution of the in situ small-strain
shear wave velocity along the dam height are required.
2. For the 150 m-high CFRD examined, a 35 m-wide tension zone parallel to the plinth devel-
ops at the lower part of the slab during static loading. This pattern of tensile stress zone
seems to develop in dams that have a relatively low overall stiffness and an irregular abut-
ment geometry (e.g. Xing, It and Itapebi dams). The maximum tensile stress developing
within this zone may reach the tensile strength of concrete under static conditions and ex-
tend to larger areas in the upper central part of the slab during earthquake shaking, causing
tensile cracks.
3. During reservoir impoundment, the lateral horizontal movement of the slab panels causes
compressive stresses about -11 MPa. During seismic shaking, if no settlements occur in the
rockfill, these compressive stresses reach a value of -18 MPa, whereas for a dynamic set-
tlement of 0.5 m at mid-crest, the compressive stresses may reach a value of -23 MPa. If
higher rockfill settlements are assumed, the maximum compressive stresses may exceed the
compressive concrete strength.
4. The distribution of the horizontal lateral displacement of the slab panels reaches a value of
0.06 cm during impoundment and 0.16 m after the earthquake, assuming 0.5 m of crest set-
tlement. Due to the distribution of these movements, no water leaks are expected at the
joints between the slabs. Moreover, the openings at the plinth joints remain <0.05 m at all
times and, therefore, no leaks are expected along the plinth.
5. Due to the diaphragm-wall action of the concrete slab, CFRDs may experience higher com-
pressive stresses in the concrete panels when subjected to longitudinal vibrations than those
caused by upstream downstream vibrations of the same amplitude.
6. Parametric studies have shown that the most crucial parameter for ensuring an excellent
dam and slab performance under both static and dynamic loading conditions is a high qual-
ity compaction leading to high values of rockfill stiffness.
484

16
REFERENCES
ABAQUS. 2008. Users Manual, Version 6.8, Simulia, Providence, RI, USA.
Abouseeda, H. & Dakoulas, P. 1998. Nonlinear Earth Dam - Foundation Interaction Using a
BE-FE Method, J. of Earthquake Engineering and Structural Dynamics, Vol. 27(12), 917-
936.
Dakoulas, P. 1993. Response of Earth Dams in Semi-Cylindrical Valleys Subjected to Oblique
SH Waves, Journal of Engineering Mechanics, ASCE, 119(1): 74-90.
Dakoulas, P. 2007. Messochora CFR Dam: slab and dam performance during reservoir filling
based on 3D nonlinear analysis, Research Report, Dept. of Civil Engineering, University of
Thessaly, Volos, Greece.
Dakoulas, P. & Hsu .1995. Response of Earth Dams in Semi-Elliptical Canyons to Oblique SH
Waves, Journal of Engineering Mechanics, ASCE, 120(3): 379-391.
Dakoulas, P. & Gazetas, G. 1986. Seismic Shear Vibration of Embankment Dams in Semi-
Cylindrical Valleys, J. of Earthq. Engin. and Structural Dynamics, Vol. 14, No. 1, pp. 19-
40.
Dakoulas, P. & Gazetas, G. 1987. Vibration Characteristics of Dams in Narrow Canyons,
Journal of Geotechnical Engineering , ASCE, 113(8): 899-904.
Duncan J. M. & C. Y. Chang. 1970. Nonlinear analysis of stress and strain in soils, J. of Soil
Mech. & Found. Engineering, ASCE, 96(5): 1629-1653.
Duncan J. M., Byrne, P, Wong, K. & Mabry P. 1980. Strength, stress-strain and bulk modulus
parameters for finite element analyses of stresses and movements in soil masses, Report
UCB/GT/80-01, University of California, Berkeley, CA, USA.
Ebeling, R. M., Peters, R. F. & Clough G. 1992. Users guide for the incremental construction
soil-structure interaction program SOILSTRUCT, Waterways Experiment Station, US Corps
of Engineers, Technical Report ITL-90-6.
Gazetas G. 1981. A new dynamic model for earth dams evaluated through case histories. Soils
& Foundations, 21(1): 599-612.
Gazetas G. 1987. Seismic response of earth dams: some recent developments. Soil Dynamics &
Earthquake Engineering, 6(1): 1-48.
Gazetas, G. & Dakoulas, P. 1992. Seismic Analysis and Design of Rockfill Dams: State of the
Art, J. of Soil Dynamics and Earthquake Engineering, 11(1): 27-61.
Marques Filho, F. & Pinto N. de S. 2005. CFRD dam characteristics learned from experience,
Hydropower & Dams, Issue 1.
Mejia, L.H. & Seed, H.B. 1983. Comparison of 2D and 3D analyses of earth dams, Journal of
Geotechnical Engineering, ASCE, 109(11): 1383-1398.
Itasca Consulting Group (2005), FLAC, Users Manuals, Minneapolis.
Rollins, K., Evans, M., Diehl, N.B. & Daily, W.D. 1998. Shear modulus and damping relations
for gravel, J. Geotechn. and Geoenvironm. Engineering, ASCE, 124(5): 396-405
Sherard, J.L. & Cooke, J.B. 1987. Concrete face rockfill dams: Assessment. J. Geotechnical
Engineering, ASCE, 113(10): 1096-1112.
Uddin, N. & Gazetas, G. 1995. Dynamic response of concrete-faced rockfill dams to strong
seismic excitation, Journal of Geotechnical Engineering, 121(2): 185-197.
Wieland, M. 2007. Seismic Performance of CFRDs, Water Power and Dam Construction,
March.

485
Dynamic Crack Propagation Analysis of Concrete Gravity Dams
with Jointed Rock Foundation

H. Kimata
Civil Engineering Division, Shimizu Corporation, Japan
Y. Fujita
Nuclear Project Division, Shimizu Corporation, Japan
H. Horii
Department of Civil Engineering, University of Tokyo, Japan
M. Yazdani
Civil Engineering Department, Tarbiat Modares University, Iran

1 INTRODUCTION
Since a concrete gravity dam is a plain concrete structure, its ultimate state under large-scale
earthquakes is expected to be penetration failure resulting mainly from the occurrence and
propagation of cracks in the dam body (MLIT, 2005). Dynamic crack propagation analysis is
used to assess the seismic safety of concrete gravity dams in large-scale earthquakes. We
proposed a dynamic crack propagation analysis method with time-domain stiffness-proportional
damping to evaluate suitable crack propagation behavior in real concrete gravity dams (Kimata
et al., 2005).
Dynamic crack propagation analysis is performed assuming that the rock foundation is
generally continuous even though the actual rock foundation is discontinuous. In order to
precisely evaluate the seismic safety of concrete gravity dams in large-scale earthquakes, the
failure behavior of not only the dam body but also the rock foundation is investigated. To the
best of our knowledge, however, there are no existing studies on the interaction of the two
failure behaviors.
We performed dynamic crack propagation analysis on a full-scale dam model taking rock
discontinuity into consideration to clarify crack occurrence and propagation in the dam body
during a large-scale earthquake.
2 EVALUATION OF DAMPING CHARACTERISTICS DURING THE OCCURRENCE
AND PROPAGATION OF CRACKS IN THE DAM BODY
We proposed a dynamic crack propagation analysis method with time-domain stiffness-
proportional damping to evaluate suitable crack propagation behavior in real concrete gravity
dams (Kimata et al., 2005). The following provides an overview.
The equation of motion used to analyze dynamic crack propagation in concrete gravity dams
is shown in Eq. (1) below. Here, Rayleigh damping is generally used to deal with structural
damping. The Rayleigh damping matrix can be composed of a mass matrix and initial stiffness
matrix, as shown in Eq. (2).

ABSTRACT: Dynamic crack propagation analysis is used to assess the seismic safety of concrete
gravity dams in large-scale earthquakes, assuming a generally continuous rock foundation. We
performed dynamic crack propagation analysis on a full-scale dam model taking rock
discontinuity into consideration to clarify crack occurrence and propagation in the dam body
during a large-scale earthquake. The results show that crack propagation length is reduced when
considering discontinuous rock compared with continuous rock.
486

M [ ]
&& u + C [ ]
& u + K [ ]u = f (t ) { }
(1)
C [ ]= a M [ ]+ b K
0
[ ]
(2)
Here, [M] is the mass matrix, [C] is the damping matrix, [K
0
] is the initial stiffness matrix,
&& u

is the acceleration vector,
& u
is the velocity vector,
u
is the displacement vector, {f(t)} is the
external force vector, and the constants a and b are determined from the first and second circular
frequency and mode damping constant.
For a plain concrete structure, a past study in which the Rayleigh damping shown in Eq. (1)
was used to study the occurrence and propagation of cracks in dams (METI and JEPOC, 2001)
showed that cracks become more dispersed and less likely to propagate as the damping
increases, whereas cracks become localized and propagate more easily when the damping is
small. Essentially, when cracks occur in plain concrete structures, it is expected that tensile
stress will no longer propagate at the crack plane, unlike in reinforced concrete structures.
However, when the damping matrix in Eq. (2) is employed in the equation of motion, when the
crack opening velocity is orthogonal to a crack, it is accompanied by a damping force acting as
a resisting force that transmits the tensile stress so that the crack is more likely to disperse and
less likely to propagate (see Fig. 1).
Therefore, in order to represent the abovementioned characteristics as faithfully as possible,
we set up a damping matrix as described below.
(i) With regard to the Rayleigh damping used for dynamic crack propagation analysis, since
the existence of [M] results in increased damping at low frequencies and an inability for stress to
be suitably released in the regions adjacent to cracks, the effects of the damping force when
cracks occur are eliminated by ignoring a[M] in Eq. (2), as shown by the following formula:
C [ ]= b K
0
[ ]
(3)
(ii) Furthermore, [K
0
] is not treated as a matrix consisting of constant elements, but rather as a
matrix that varies as a function of time t as follows:
C(t ) [ ]= b K(t ) [ ]
(4)
Here, [C(t)] is the damping matrix dependent on time t, [K(t)] is the stiffness matrix at time t,
and b is a constant determined by the first circular frequency and mode damping constant. In
other words, by applying damping that varies according to the constantly varying stiffness [K(t)],
the stiffness becomes zero and no damping forces act on the crack plane when a crack fully
opens. In this way, the damping affects the transmission of tensile stress at crack surfaces so that
it coincides with the actual transmission when cracks occur and propagate.
In this study, we performed dynamic crack propagation analysis using time-domain stiffness-
proportional damping Eq. (4).

Fig. 1 Equilibrium of force in the crack plane
Crack
Damping force
Inertial force
u
[ ]{ } u K [ ]{ } u& C
[ ]{ } u& & M
( ) { } t f External force
Crack
Restoring force
487
3 DEFORMATION PROPERTIES OF ROCK JOINTS
Two-dimensional FEM is basically used in this study and all rock joints are modeled by
interface elements. Deformation properties of the interface element used as joints are as follows:
(i) The relationship between shear stress and shear strain is assumed to be elasto-perfectly
plastic hysteresis, as shown in Fig. 2(a). The shear strength
f
is defined by Mohr-Coulombs
failure criteria, Eq. (5). Tensile strength and shear stress at tensile normal stress are equal to zero,
as shown in Fig. 3.
tan m c
f

(5)
where c: cohesion, : internal friction angle.
(ii) The relationship between normal stress and normal strain is linear (elastic) on the
compression side and the rigidity is zero on the tension side, as seen in Fig. 2(b).
(iii) In the stress-strain relationship shown by Eq. (6), matrix [D] is expressed by Eq. (7) for
the elastic strain component and Eqs. (8a)(8c) for the plastic strain component. As seen in the
non-diagonal term in Eq. (8a), shear stress increment due to the confining pressure is taken into
consideration when the value of is positive in the plastic region. On the other hand, the
influence due to the roughness of the joints, namely dilatancy, is disregarded. An extremely
small value is given to S
v
in order to avoid numerical ill-conditioning when the value of the
diagonal component is zero.
[ ]
D
(6)
where
[ ]

E
G
D
0
0
(7)
for the elastic strain component, and

[ ]

v
v h
S
Sign S S
D
0
) ( tan

(8a)
) 10 1 / ( 0
4
= G S
h
(8b)
=
<
) 0 )( 10 1 / ( 0
) 0 (
4
E
E
S
v
(8c)
for the plastic strain component. where E: Youngs modulus, G: shear modulus, Sign(): +1
( 0), 1 ( 0).

(a) Shear stress vs. shear strain (b) Normal stress vs. normal strain
Fig. 2 Deformation properties of interface element used as joints

Elastic

0
Elastic
Elastic
Plastic
Plastic
f

Loading
Unloading
(Tension)
(Closing)
(Compression)

(Opening)
0
488

4 ANALYSIS CONDITIONS
4.1 Analysis model
We performed dynamic crack propagation analysis using a model of the large class of
concrete gravity dam in Japan (height: 100 m). The analysis model is shown in Fig. 4. The dam
was assumed to be subjected to a plane strain in the direction perpendicular to the dam axis
(upstream/downstream direction), and two-dimensional FEM was performed for dynamic crack
propagation analysis using the model comprising the dam body, rock foundation and reservoir
water.
To prevent snap-back in the dynamic crack propagation analysis, the dam body was divided
into a mesh of triangular elements based on an equilateral triangle with side length of 1.5 m
(crack band width = 1.5 sin 60 m).
The rock foundation was modeled over a region extending to about 10 times the width of the
dam at the base (760 m), and to a depth of twice the dam body height (200 m). The bottom
surface of the model was provided with dashpots simulating a semi-infinite boundary to allow
the dissipation of seismic waves to be taken into account. Free ground was provided at the sides
of the model with dashpots to transmit energy to the free ground, thereby allowing the
dissipation of seismic waves to be taken into consideration in a similar manner. The reservoir
water was treated as an additional mass determined from Westergaards formula (1933).

In Fig. 4, considering rock discontinuity near the dam body, 4-node quadrilateral elements
were used for both intact rock and joints. As for joints, the interface element was applied.
Many joints usually exist in a real rock foundation and there are frequently three conjugated
joint planes crossing perpendicularly to each other in plutonic rock such as granite and
sedimentary rock, which is a popular rock foundation for concrete gravity dams. Hence, a joint
system consisting of two joint sets crossing perpendicularly to each other was adopted on the
two-dimensional cross section of the foundation. Although the spacing of joints in real rock is
various and wide ranging, a uniform distribution of joints was assumed to have spacing of about
3.54 m. Regarding the spacing of relatively continuous joint surfaces having a length of over
2 m, the standard spacing of joint surfaces ranges from 2 to 10 m in the case of nearly fresh rock
including slightly weathered joint surfaces. In the case of fresh rock having no weathered joints,
the joint spacing can exceed 10 m according to the classification by Kikuchi et al. (1982).
Furthermore, since the progressive failure behavior of rock foundations could depend on joint
direction, we assumed joint directions to be
1
= 45,
2
= +45 in this study. Here, the positive
direction means anticlockwise from the X-coordinate, namely the upstream and downstream
direction.

Fig. 3 Shear strength
f
(Mohr-Coulombs failure criteria)

(Compression)
c
(Tension)
0

-c
tan m
f
489

4.2 Analytical constants
The main analytical constants of the model dam are shown in Table 1. These values were set
by referring to past studies.
We used the bilinear type of concrete tension-softening constitutive model (1/4 model)
(Rokugo, 1989)

to simulate the relationship between tensile stress and crack opening
displacement in the concrete of the dam body (see Fig. 5), and an origin-oriented hysteresis
characteristic was assumed during unloading. Also, with regard to the concrete compression
characteristics, we used the failure criterion formulated by Drucker and Prager (1952)

to take
into account nonlinearities, and we used a coefficient of 0.005 for the reduction of shear
stiffness after crack formation.
Intact rock is assumed to be an elastic body and the joints are assumed to be an elastoplastic
body with the aforementioned mechanical deformation. Strength parameters of joints such as
cohesion c and internal friction angle significantly affect the progressive failure behavior of
the rock foundation. The following parameters that were assumed in a previous study (Tada et
al., 2002) were used again in this study.
Fresh joints: c = 0.16 MPa and = 42 (JRC = 6, JCS = 98.1 MPa,
b
= 35)
(b) Dam body
(a) Model comprising dam body, rock foundation and reservoir water
Fig. 4 Analytical model
(c) Jointed rock

2
0
0
760
1
0
0
5
7
4
0
Dam body
Rock foundation
(Unit: m)
Reservoir water
Sediment Free ground
Jointed rock
X
Y
Joints
X
Y
1
= +45
2
= 45
Intact rock
490
where JRC is the coefficient of roughness, JCS is the strength of the wall face and
b
is the
residual friction angle incorporated in Bartons equation (1973), which is widely used for
estimating the shear strength of joints. The above values were assumed respectively in this study
and were transformed into c and values.
With regard to the damping constant, the first natural mode damping h
1
was assumed to be
7% (METI and JEPOC, 2001) for the dam body. The first and second natural mode damping h
1
,
h
2
were assumed to be 2% respectively for the rock foundation.

Table 1 Analytical constants

Constant Dam body Rock foundation
Youngs modulus E (GPa) 28 10
Poisson ratio 0.2 0.25
Density (kg/m
3
) 2300 2500
Compressive strength fc (MPa) 30.0
Tensile strength ft (MPa) 3.0
Tensile energy G
f
(N/m) 400
Cohesion c (MPa) 0.16 (Joints)
Friction angle () 42 (Joints)
First-order mode damping constant h
1
Second-order mode damping constant h
2

0.07

0.02
0.02

4.3 Input seismic motion
To simulate the response of a large-scale earthquake, the input seismic motion used in the
dynamic crack propagation analysis was created by using the target spectrum called Minimum
Acceleration Response Spectrum (see Fig. 6; MLIT, 2005). The response spectrum is provided
to consider minimum horizontal seismic motion in the design of dams, assuming the absence of
an active fault at the bedrock surface and the occurrence of a magnitude M6.5 earthquake near
the dam site.
The simulated seismic motion we used is shown in Fig. 6. Phase characteristics were decided
using the observation records at the Hitokura Dam during the Hyogo-ken Nanbu Earthquake of
1995 in Japan. This corresponds to the seismic motion at the bedrock surface, and in the
analysis we decided to use the input seismic motion brought down to the bottom of the model in
the horizontal and vertical direction.

Fig. 5 Constitutive model of concrete tension softening
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
0.0 0.2 0.4 0.6 0.8 1.0
Crack opening displacement w(mm)
T
e
n
s
i
l
e

s
t
r
e
s
s
(
M
P
a
)
f
t
= 3.0 MPa
f
1
= 0.75 MPa
w
1
= 0.10 mm
w
= 0.667 mm

= 5.13210
-4

491

5 ANALYSIS RESULTS
Prior to the dynamic crack propagation analysis, we conducted a static analysis to obtain the
initial stress taking into account the dead weight of the dam body and the rock foundation, the
static pressure of the reservoir water, and the uplift pressure on the bottom surface of the dam
body.
Relative displacement between the top and bottom of the dam body is shown in Fig. 7,
calculated using the results of the dynamic crack propagation analysis. Relative displacement
indicated that the top of the dam has residual deformation of about 4 mm on the downstream
side after excitation.
Figures 8 and 9 show deformation modes of the rock foundation just at the time of the
maximum dam body deformation on the upstream side and the downstream side, respectively, at
the times indicated in Fig. 7. In Fig. 9, the opening and sliding of rock on the downstream side
is largely caused at the joints in the
2
= 45 direction near the bottom edge of the dam body.
On the other hand, in Fig. 8, the opening and sliding of rock on the upstream side is smaller than
that mentioned for the rock on the downstream side, due to the restraint effect of the static
pressure of the reservoir water.
Figure 10 shows the crack propagation in the dam body exceeding the full crack opening
displacement w
c
. Figure 11 shows the same state, not taking rock discontinuity into
consideration. Focusing on the dam body, it can be seen in these two figures that a crack has
progressed at the lower surface of the dam body. The length of the crack path taking rock
discontinuity into consideration is much smaller than that not taking rock discontinuity into
(b) Acceleration response spectrum
Fig. 6 Simulated seismic motion
(a) Acceleration
-4.0
-2.0
0.0
2.0
4.0
0 5 10 15 20
Time(s)
A
c
c
e
l
e
r
a
t
i
o
n
(
m
/
s 2
)
Horizontal Vertical
h=0.05
0.0
2.0
4.0
6.0
8.0
10.0
0.01 0.1 1 10
Period(s)
A
c
c
e
l
e
r
a
t
i
o
n
(
m
/
s 2
)
Horizontal
Vertical
492
consideration, due to the dissipation of excitation energy resulting from the opening and sliding
that occurred in the jointed rock foundation.

Fig. 7 Relative displacement between the top and bottom of the dam body
Fig. 8 Deformation mode of the
rock foundation just at the time of
maximum dam body deformation
on the upstream side (t = 7.82 s)
Fig. 10 Crack propagation of dam
body taking rock discontinuity into
consideration.
Fig. 9 Deformation mode of the
rock foundation just at the time of
maximum dam body deformation
on the downstream side (t = 8.13 s)
Dam body
Downstream side
Jointed rock foundation
Dam body Upstream side
Jointed rock foundation
Fig. 11 Crack propagation of dam
body not taking rock discontinuity into
consideration.
-0.05
-0.03
0.00
0.03
0.05
0 5 10 15 20
Time(s)
R
e
l
a
t
i
v
e

D
i
s
p
l
a
c
e
m
e
n
t
(
m
)
t = 7.82 s
t = 8.13 s
493
6 CONCLUSIONS
We performed dynamic crack propagation analysis to evaluate the seismic safety of a
concrete gravity dam during a large-scale earthquake. By focusing on the fact that propagation
of cracks in the dam structure is affected by rock discontinuity failure, we then investigated the
relationship between rock discontinuity failure and crack propagation of the concrete dam body,
and we conducted dynamic crack propagation analysis on a full-scale dam model taking rock
discontinuity into consideration. Findings from the present study can be summarized as follows:
Prior to performing dynamic crack propagation analysis using a two-dimensional element
method, the joints of each rock element were provided with an elastoplastic function for
mechanical deformation. Strength parameters of joints such as cohesion c and internal friction
angle significantly affect the progressive failure behavior of the rock foundation. Using the
strength parameters of fresh joints, we conducted dynamic crack propagation analysis. As a
result, we found that the relative displacement between the top and bottom of the dam body has
residual deformation after excitation, because the joints rapidly repeated both the open-close
condition and sliding through the earthquake. While the excitation energy dissipated using the
dam model taking rock joint failure into consideration, the crack propagation length of the dam
body could be reduced compared with the model considering continuous rock.

In the future, we plan to evaluate the seismic safety of concrete gravity dams with varying
joint angle and strength of rock foundation by subjecting them to the same kind of analysis. We
also plan to investigate other issues such as the three-dimensional effects of concrete gravity
dams, the effects of water penetrating into cracks that form during an earthquake, the effects of
aftershocks, and earthquake-resistant strengthening methods.

REFERENCES
Ministry of Land, Infrastructure and Transport (MLIT), Japan. 2005. Guideline for the Seismic
Performance Evaluation of Dams against Large Earthquake (Draft) (In Japanese).
Kimata, H., Fujita, Y., Niimi, K., Hoshido, T., Kase, T., Hikawa, N. and Horii, H. 2005. Seismic
safety of concrete gravity dams based on dynamic crack propagation analysis during
large-scale earthquakes, ICOLD 73
rd
Annual Meeting, Proceedings of the Workshop on
Seismic Aspects of Dams.
Ministry of Economy, Trade and Industry (METI) and Japan Electric Power Civil Engineering
Association (JEPOC). 2001. Report on an investigation into improving the quake
resistance of dam designs, 2000 report on quake-proof measures for electrical power
facilities (quakeproof reliability proof testing of power generating facilities) (In Japanese).
Westergaard, H.M. 1933. Water pressures on dams during earthquakes, Trans. ASCE, Vol. 75,
pp. 418433.
Kikuchi, H., Saito, K. and Kusunoki, K. 1982. Geological synthetic evaluation on the stability
of rock foundation for dam, J. of Large Dams, combined No. 102, 103 (In Japanese).
Rokugo, K., Iwasa, M., Suzuki, T. and Koyanagi, W. 1989. Testing methods to determine tensile
strain softening curve and fracture energy of concrete, In Fracture Toughness and
Fracture Energy, Balkema, pp. 153163.
Drucker, D.C. and Prager, W. 1952. Soil mechanics and plastic analysis or limit design, Q. Appl.
Math., Vol. 10, No. 2, pp. 157165.
Tada, H., Ishizuka, Y., Horii, H., Uchida, Y., Kashiwayanagi, M. and Kimata, H. 2002. Stability
evaluation method of dam foundation during earthquake in consideration of rock joints, J.
JSCE, No. 715/III-60, pp. 201213 (In Japanese).

494
1 INTRODUCTION
1.1 Local site effects
Local site conditions can influence the characteristics of strong ground motion in various ways,
which are collectively referred to as site effects. These are related to the thickness and imped-
ance contrast between soil layers, the surface topography, and the subsurface topography in
terms of lateral discontinuities, basin edges, and inclined interfaces. While soil layer effects can
be predicted in a rather straightforward manner, surface and subsurface topography effects pre-
sent more complications, since they are tied to two-dimensional phenomena such as focusing
and defocusing of seismic energy, interference of direct and diffracted wave fields, and genera-
tion of surface waves leading to complex wave fields (Bard 1999).
1.2 Outline of study
In this study, a site is chosen which combines several of the aforementioned geomorphological
features. It is modeled based on existing knowledge of the geometry and properties of the geo-
logical structures. The numerical results are compared to data from earthquakes recorded at the
site. Analysis results are used to highlight the influence of the sites features, namely topogra-
phy and basin effects from complex non-horizontal layering.
2 DATA
2.1 Area studied and stations used
The area studied was the city of Aegion, which lies on the southern part of the Gulf of Corinth,
Greece (Fig. 1a), a highly seismic area comprising many WNW-trending, north-dipping active
normal faults. The city has been struck by significant earthquakes in the past, the strongest re-
Engineering aspects of site and topography effects at Aegion,
Greece

O.-J. Ktenidou, D. Raptakis & K. Pitilakis
Aristotle University of Thessaloniki, Thessaloniki, Greece
F.-J. Chvez-Garca
Instituto de Ingeniera, Universidad Nacional Autnoma de Mxico, Mxico DF, Mxico
ABSTRACT: In this study, a site is chosen (Aegion) which combines several geomorphological
features: surface topography and a basin with complex non-horizontal layering. It is modeled
based on existing knowledge of the geometry and properties of the geological structures, using
two finite difference schemes (SH and P-SV). The numerical results at certain control points of
the model are successfully compared to data from 19 earthquakes recorded at the site by a sur-
face station (DIM) and a downhole array (CORSSA), revealing SH amplification to be stronger
than SV amplification in the basin. Numerical results along the cross-section are used to high-
light the influence of the sites features on seismic ground motion. We focus on engineering as-
pects of the motion, such as differential displacement over the basin and topographic amplifica-
tion near the slope crest. Where possible, we compare results to design code provisions, which
in some cases do not suffice.

495
Aegion city - elevated area
Aegion Bay
CORSSA
escarpment
A
A
DIM
cent one being that of June 15, 1995 (Ms=6.2). The chosen region is marked on the one hand by
its characteristic surface topography, given that it is traversed by the Aegion fault, whose es-
carpment of roughly 90m divides the city into two levels: the lower Northern part lying on the
hanging wall and the upper Southern part lying on the foot wall (Fig. 1a). On the other hand,
another characteristic feature of the site relates to the loose soft deposits present downhill,
which form an open basin whose depth increases along with the sea depth.
After a series of field and laboratory tests, as well as borehole logging, geophysical prospect-
ing and microtremor measurements, Apostolidis et al. (2005) modeled the complex geology of
the site in sufficient detail and estimated the geotechnical and dynamic soil properties to a satis-
factory degree. The final model proposed by them can be seen in Figure 1b.
Two locations near this site are instrumented, marked in Figure 1 as CORSSA and DIM. The
Corinth Soft Soil Array (CORSSA - http://geo.civil.auth.gr/Staff/dep/pitilakis/CORSSA/) lies at
the lower part of the city, very near the coast. It is a vertical array consisting of four accelerome-
ters at depths of 14m, 31m, 57m and 178m as well as a surface accelerometer. The soil profile
there consists of soft loose marine materials overlying hard conglomerate. This is found at
160m, so the deepest accelerometer lies within the conglomerate and can be used as a reference
station. At the upper part of the city, a surface accelerometer is installed at the City Hall - Di-
marhio (DIM), which is often referred to as OTE, which is the name of its (nearby) location
prior to 1996. This station lies on a thin layer of deposits overlying the stiff conglomerate.
2.2 Dataset
The instrumental dataset used in this study is the same as the one used for the preliminary site
effect study in Ktenidou et al. (2008). 19 earthquakes were chosen, with local magnitudes rang-
ing between 3.1 and 5.3, which were recorded at both stations CORSSA and DIM. The records
were rotated with respect to the orientation of the fault (roughly E-W), which is also the orienta-
tion the deposit-conglomerate interface defining the basin at the lower part of town.

Figure 1. a. Location of the site under study and map of the city of Aegion, indicating fault, elevation dif-
ferences, and instrumented sites (adapted from Athanasopoulos et al. 1999). b. Geotechnical model of a
roughly N-S cross-section crossing the fault (adapted from Apostolidis et al. 2005).
CORSSA
1
7
8
m
A
A
DIM
496
3 INSTRUMENTAL RESULTS
3.1 Spectral ratios
One way to instrumentally study site effects in the frequency domain is to calculate spectral ra-
tios. The most usual are spectral ratios relative to a reference site, often called standard spectral
ratios (SSR, e.g. Borcherdt 1970) and horizontal-to-vertical spectral ratios (HVSR), where hori-
zontal component spectra are divided by the vertical component spectra recorded at the same
site (e.g. Lermo and Chvez-Garca 1993). In the interest of comparison with numerical results
(in which there is no incident vertical component), we will only deal with SSR in this study.
3.2 SSR at CORSSA
SSR was calculated for each of the stations in soil with respect to the motion at the deepest ac-
celerometer, the one at 178m which lies within the conglomerate (Ktenidou et al. 2008). This
was done for the two horizontal components, the transverse one, which is parallel to the slope,
and the radial one, which is perpendicular to it. The average results for the 19 events are shown
in Figure 2a for all stations.
It is easily observed that the t-component exhibits a clearly higher peak than the r-component
over the sites fundamental frequency at 0.9 Hz, which corresponds to the soil-conglomerate in-
terface at 160m. This holds true for all depths. At the surface, the amplification reaches as high
as 8 for the transverse component, while for the radial the peak is only barely distinguishable
with a value no more than 4. At larger depths, the amplification level does not diminish gradu-
ally because it is mainly due to the soil-conglomerate interface.
Taking into account the cross-section shown in Figure 1b, the t-component corresponds to the
direction perpendicular to the section, which would mean SH waves. The r-component, respec-
tively, is parallel to the section and thus corresponds to SV waves which also generate P waves
when reaching the basins lateral boundaries.

Figure 2. a. Standard spectral ratios (SSR) at all stations within soil (0m, 14m, 31m, 57m) with respect to
the station in rock (178m). Average for 19 recorded events is shown for the radial and transverse compo-
nent (Ktenidou et al. 2008). b. Numerical transfer functions derived from 2DFD modelling for SV and
SH incident motion, corresponding to r and t components of the instrumental results.
497
S
S
R
2
D
F
D

T
F
0.1 1 10
Frequency (Hz)
1
10
r
t
0.1 1 10
Frequency (Hz)
1
10
SV
SH

Figure 3. a. Standard spectral ratio (SSR) at the surface of DIM with respect to the CORSSA station
in rock (178m). Average for 19 recorded events is shown for the radial and transverse component (Kteni-
dou et al. 2008). b. Numerical results derived from 2DFD modelling for SV and SH incident motion.

Jongmans and Campillo (1993) studied the seismic response of a closed basin and found
higher spectral amplification in the SH than in the SV sense. Also, in Chvez-Garca & Faccioli
(2000), results in the response spectrum domain for numerical analysis of a sine-shaped valley
show greater amplification in the SH than the SV sense. This could be due to SH motion taking
place without restriction by the basins lateral boundaries and also to the fact that it needs not
lose energy in order to give rise to P motion, as happens in the SV case. Raptakis et al. (2004)
also mention that for two sites in the open basin they studied in Thessaloniki, the transverse
component was amplified more than the radial one at the fundamental frequency of SSR, and at-
tributed this finding to strong Love waves.
3.3 SSR at Dimarhio
In a similar way as before, SSR was also calculated here with respect to the motion at the deep-
est accelerometer of CORSSA (Ktenidou et al. 2008). This was done for the two horizontal
components. The average for the 19 events processed is shown in Figure 3a.
At this site, the two components do not seem to differ much. The fundamental peak is not as
well defined as for the CORSSA site, with maximum amplification between 4 and 10 over a
wide range of frequencies between 3-7 Hz. The sites fundamental frequency has been theoreti-
cally calculated through Kennets reflectivity method (Kennet 1983) to be 5.5 Hz, in the middle
of this range.
4 NUMERICAL MODELLING AND RESULTS
4.1 Simulation
The 2D numerical modelling was based on the geotechnical model of Apostolidis et al. (2005)
shown in Figure 1b, which accounts for the dipping layers at the hanging wall. One alteration
was that the deposit-conglomerate interface was smoothed so as not to introduce angular inter-
faces which are uncertain in the first place and may generate artificial complications. Another
alteration was that no more interfaces were considered beneath the conglomerate-deposit inter-
face, since their actual location is dubious and only assumptions can be made as to the deeper
soils dynamic properties. Finally, anelastic attenuation was ignored by assuming very high
quality factors. The geometry and soil properties of the model used are shown in Figure 4.
Two different 2D finite difference (2DFD) schemes were implemented for the dynamic nu-
merical analysis, one using vertically incident SV motion and allowing for P-SV interaction and
one using plane SH vertically incident waves. The first code was FLAC 2D v. 4 (Itasca Consult-
ants 2002) and the second one was the scheme introduced by Moczo (1989) and Moczo & Bard
(1993). For both schemes the discretisation was performed so as to allow frequencies up to 10
Hz to propagate without distortion through the grids and the lateral boundaries were placed so
as to avoid artificial reflection.
The input motion time-history was the same in both cases: a displacement pulse given by the
Gabor wavelet, which has been used by Moczo (1989) and is described by equation 1:
498
E
C
A
B
C
D
2000
DIM
CORSSA
d [kN/m
2
] v Vs [m/s]
A 18.5 0.3 180
B 19.5 0.3 265
C 19.5 0.3 440
D 20 0.3 540
E 21 0.2 1000

Figure 4. a. Geometry of the model used for the 2D finite difference modelling and dynamic analyses. b.
Soil properties assumed for the analysis (damping was ignored).

( )

= ) ( cos
) (
exp ) (
2
s
s
t t
t t
t S (1)
where
p
f = 2 ,
p s
f t = 45 . 0 ,
p
f is the predominant frequency, and the pulse is de-
fined in the interval [0. 2
s
t ]. In this case, we considered the parameters used by Makra et al.
(2001), namely: =1.5,
p
f =4 Hz, =0.4, and
s
t =0.169 s. This allows for a pulse whose am-
plitude spectrum is almost flat over most of the frequency range we are interested in, which is
roughly between 0.1-10 Hz.
4.2 Results at control points - Comparison to instrumental results
Before presenting full results, we first try to attempt to compare numerical and instrumental re-
sults. This we do by estimating transfer functions between the 2D synthetics of the receivers
whose location in the model corresponds to that of the accelerometric stations for which we
formed the SSA ratios in the previous section. For CORSSA, the numerical counterparts of the
empirical standard spectral ratios shown in Figure 2a are shown next to them in Figure 2b.
It is reasonable to compare the 2DFD SH results to the transverse component of the records
and the SV results to the radial component, since the former correspond to motion parallel to the
cross-section and the latter are perpendicular to it. It can be easily observed that the numerical
results reproduce some of the observations we made previously for the instrumental data. Am-
plification at the fundamental frequency is higher for SH motion than it is for SV at all depths,
just as the first SSR peak was higher for the transverse than the radial component. We already
speculated as to the reasons that might cause this. When comparing absolute level of amplifica-
tion as estimated instrumentally and numerically, it is found for both components and at all
depths very similar, though slightly higher in the latter case, which seems reasonable given that
the numerical calculations did not account for material damping.
In the case of surface station DIM, we follow the same procedure, forming the transfer func-
tion between the surface motion at DIM and the reference point at 178m beneath CORSSA.
The resulting transfer function is shown in Figure 3b next to the respective SSR. Both numeri-
cal schemes exhibit the fundament peak at 5 Hz, sharper and narrower than the wide peak de-
fined by SSR, but within it. The absolute amplitude of the peaks is roughly comparable but the
numerical results do not reproduce the empirical as closely.
4.3 Results across the profile
4.3.1 Time domain
One of the techniques usually implemented for depicting results in the time domain is to plot
synthetics from appropriately spaced receivers along the surface of the model, in the style of a
seismic section. Here, the results are displacement time histories. In the interest of brevity we
499
0
1
2
3
4
5
6
T
i
m
e

[
s
]
0
T
i
m
e

(
s
)
0 500 1000
Distance (m)
0
1
2
3
4
5
6
7
8
9
1
0
F
r
e
q
u
e
n
c
y

(
H
z
)
1
2
4
6
8
10
20
30
F
r
e
q
u
e
n
c
y

(
H
z
)
Distance (m)
only present the results of one of the numerical schemes implemented; we choose the SH one,
since it was already seen to be associated with larger amplification, at least within the basin. The
displacement traces, taken every 12m along the profile, are shown in Figure 5a. Following arri-
val of all direct and reflected S waves (around 1 s for DIM and 1.5 s at CORSSA) the laterally
propagating Love waves are clearly seen. These add to the duration of ground motion and also
amplify the fundamental spectral peak.
4.3.2 Frequency domain
The synthetic traces were divided in the frequency domain by the input motion so as to form
transfer functions in the more usual sense of the term (i.e., as opposed to the ratios with respect
to the motion at 178m below CORSSA). Uphill, the amplification is significant for frequencies
above 4 Hz, while high amplification is noted around 1 Hz near CORSSA and for lower fre-
quencies as the basin deepens.
5 ENGINEERING CONSIDERATIONS
5.1 Peak absolute and differential motion
Peak ground motion is of interest when considering site effects from an engineering point of
view, because in design it is often a single value prescribed by codes that is used to account for
site amplification. Even though a single value corresponding to zero period cannot reflect the
complex, frequency-dependent phenomena that take place when surface and lateral subsurface
discontinuities are present, it is interesting to observe the distribution of this peak value across
our profile with the intension of illustrating its variation. Based on Figure 5a, we plot the maxi-
mum value of each trace in the time domain across the length of the cross-section. This yields
Figure 6a, in which surface values are dimensionless because they correspond to unit input dis-
placement.

Figure 5. a. Synthetic displacement time histories along the surface of the profile derived from the 2DFD
SH scheme, ignoring material damping. The locations of the two stations are marked on the cross-section.
b. Transfer functions with respect to input motion for the synthetics of Figure 5a.
500
0
1
2
3
4
5
0 300 600 900 1200 1500
Distance [m]
P
e
a
k

d
i
s
p
l
a
c
e
m
e
n
t

SH
SV
0
0.1
0.2
0.3
0.4
0.5
0.6
0 300 600 900 1200 1500
Distance [m]
P
e
a
k

d
i
f
f
.

d
i
s
p
l
.

SH
SV

Figure 6. a. Peak values of absolute displacement along the profile surface, as calculated by the two
2DFD schemes (P-SV and SH), assuming unit displacement as input. b. Peak values of differential dis-
placement, calculated as the maxima of the spatial derivative of the traces in Figure 5a.

We point out that although the vertical component of motion is not plotted in Figure 6a, it is
present in the SV case due to the non-horizontal boundaries and further amplifies the ground
motion. Also, in the vicinity of the crest, although the 1D profile does not change significantly,
the peak ground motion is somewhat amplified. This is due to topographic amplification and
will be mentioned more extensively in the following section. Downhill, the motion amplifica-
tion in the time domain increases along with the sediment depth and is roughly stabilised at a
distance of 400m from the toe of the slope. On the whole, the amplification level is not ex-
tremely higher downhill than uphill, with a maximum peak ratio of around 1.5. Also, the two
2DFD schemes yield roughly similar results, which is rather straightforwardly explained by the
fact that the maximum value in the time domain for all traces corresponds to the direct S first ar-
rival (0.7-1.1 s in Fig. 5a).
Apart from translational motion, another consideration when dealing with lateral discontinui-
ties is differential motion, as it can inflict uneven constraints on structures. This is all the more
interesting to study given that design codes do not account for it as part of site effects, and that
not very many studies deal with it (one recent example being Raptakis et al. 2004b). For this
reason we also calculated the spatial derivative of the displacement traces shown in Figure 5a.
Figure 6b shows the maximum values of the derivatives of the traces, yielding the peak differen-
tial displacement along the surface of the cross section, which is dimensionless. One of the first
observations to be made is that the values uphill and downhill are very different. This is reason-
able, since there are no lateral discontinuities uphill and the little differential motion observed is
due to the relief. Downhill, however, the differential displacement increases modestly at the ba-
sin edge and dramatically (as much as 6-10 times) as the basin deepens and new soil layer inter-
faces are encountered, up to a distance of roughly 500-600m. Another observation is that the
two numerical techniques do not yield very similar results. Similarly to the comparison of SSR
ratios and transfer functions in Figure 2, the motion is stronger in the SH sense, indicating
stronger Love waves as opposed to Rayleigh waves. Finally, we should mention that these ef-
fects may be played up through the lack of anelastic attenuation in the calculations, but could
well be played down through the smoothing of the interfaces in the modelling.
5.2 Topographic amplification
5.2.1 Comparing to theory and literature
Topographic effects are site effects related to the surface relief. They have been reviewed in
Geli et al. (1988) and Bard (1999) and studied numerically in various studies (e.g. in the case of
Aegion: Athanasopoulos et al. 1999, Bouckovalas et al. 1999, Ktenidou et al. 2007).
The usual observation in the presence of a slope is amplification of ground motion at the crest
and deamplification at the toe over a specific range of frequencies related to the material stiff-
ness and slope geometry. Although soil effects render it somewhat harder to discern in the pre-
sent case, this is confirmed by the peak motion values for distances 400-700 m in Figure 6a,
though not very strongly. Athanasopoulos et al. (1999) used a P-SV finite element scheme to
501
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
-150 -125 -100 -75 -50 -25 0
Distance from crest [m]
F

=

P
e
a
k

(
x
)

/

P
e
a
k

(
f
f
)
SH
SV
AFPS90
EC8
(?)
0.1 1 10
Frequency (Hz)
0.5
1
1.5
2
2.5
T
A
F
0m
10m
20m
30m
40m
model Aegion and came up with an amplification of 3.9 in the time domain with respect to the
input amplitude for the site of OTE (near DIM), and a value of 1.4 at the toe. Our respective
values based on Figure 6a are 3.0 and 1.7. In the same figure, for the P-SV numerical scheme,
amplification in the time domain extends to a distance of 100 m behind the crest and reaches up
to 17% (compared to the respective 1D case that includes soil layering effects), while for the SH
scheme this distance is only 40 m and the increase is not higher than 12%. Though this differ-
ence is not great, it is in agreement with the observation mentioned by Geli et al. (1988) and
Bard (1999) that the motion perpendicular to the ridge axis is often larger.
Another usual observation is that this amplification behind the crest is uneven and varies with
the distance from it. Speculations have been expressed as to this phenomenon causing non-
uniform structural damage patterns, such as those observed in uphill Aegion following the great
earthquake of 1995 (Bouckovalas et al. 1999). The results of the present study do not indicate
very abrupt variation in peak values of either absolute or differential motion behind the crest, at
least not compared to the variation taking place downhill (Fig. 6). However, in Figure 5b and
for frequencies higher than 4 Hz, it is possible to see that the transfer functions differ depending
on the distance behind the crest.
It must be noted here that the height of the Aegion fault escarpment and its slope vary along
the fault axis as it traverses the town, so topography effects should be expected to vary as well.
5.2.2 Comparing to codes
Although in very generic terms, topography effects are taken into account in some design codes,
namely the French AFPS90 and Eurocode 8 (AFPS, 1995; ECN, 2003):
The French recommendations AFPS define a coefficient that is frequency independent and
has a maximum value of 1.40. The coefficient is defined depending on the slope inclination (I)
and its difference with the slope of the ground behind the crest (i). It is applied over a length b
from the crest of the slope and is linearly reduced to 1.0 over transition lengths a and c in
front of and behind the crest, respectively (see Bard 1999 for an illustration and details). The
conditions under which the topographic effect is taken into account are the following: slope
height H>10m, slope inclination I>22 and ground inclination behind the slope i< 0.33I.
Eurocode 8 also defines an amplification factor (ST) that is frequency independent and is
said to be used for important structures. The minimum slope inclination taken into account is
15 and the minimum height considered is H=30m. The factor can have values up to 1.40. The
length over which it should be applied behind the crest is not defined. It is mentioned that for
isolated cliffs and slopes, a value of ST > 1.2 should be used for sites near the top edge.

Figure 7. a. Peak motion values in the time domain, as calculated by the two 2DFD schemes (P-SV and
SH), compared to code prescriptions by AFPS and EC8. Values are normalised by the free field. b. To-
pographic amplification factor in the frequency domain, given by the spectral ratio of the surface motion
at various distances from the crest with respect to the free field motion.

502
In the case studied here, due to the relatively mild relief, the factors assume low values, i.e.
1.07 and 1.2 for AFPS and EC8 respectively. The comparison with time-domain results is
shown in Figure 7a, where peak motion values from Figure 6a have been normalised with re-
spect to the free field motion corresponding to the same soil profile, so that topography effects
are decoupled from soil effects. The AFPS envelope has been drawn exactly but the EC8 enve-
lope is only indicative, as it is up to the engineer to interpret what distance is considered near
to the crest. However, it can be seen that the AFPS envelope does not successfully envelop the
2DFD results, particularly those of the P-SV scheme. EC8, on the other hand, if applied for a
distance of 50 m from the crest, could achieve that. Either way, it should be noted that, at this
particular site, topographic amplification (of the order of 10%-15%) is not exceptionally high.
A parameter usually estimated as a measure of the topography effect is the Topographic Am-
plification Factor (TAF), somewhat loosely defined as the ratio of the Fourier spectrum of the
ground motion near the crest with respect to that of the far field or free field motion. In this
case, since the maximum amplification is not observed upon the crest but rather at a distance
roughly 40 m from it, we use the motion at various distances up to 40 m from the crest to form
the ratios. The results are shown in Figure 7b for the SV calculations, since it yielded larger
peak amplifications than SH. Amplification between 1.15-1.35 is observed near the crest over a
frequency of roughly 4.5 Hz. Since the model is not homogeneous it is difficult to judge the fre-
quency where topographic amplification is expected based on theory, but if we apply the rule of
thumb given by Geli et al. (1988), that amplification is expected when =L, where L is the
length of the slope and =Vs/f is the wavelength, then the results are roughly in agreement. For
this frequency, the spectral amplification values lie outside the envelopes provided by the code
provisions.
6 CONCLUSIONS

The complex site of Aegion is chosen because its geomorphological features include surface to-
pography and a basin with non-horizontal layering. On site, two recording stations are available,
a surface station uphill (DIM) and a vertical array downhill in the basin (CORSSA). We have a
dataset of 19 events recorded at both stations for which we calculate SSR ratios with respect to
the deepest downhole instrument. The site is then modelled based on existing knowledge of the
geometry and properties of its geological structures. We use two finite difference schemes for
dynamic analysis, one to model SH motion and one to model P-SV motion.
The numerical results at certain control points of the model are compared to results from the
instrumental data processing. SSR ratios from earthquake data are compared to numerical trans-
fer functions, and agree in that they reveal SH amplification to be stronger than SV amplifica-
tion in the basin. This corresponds to the transverse component (fault-parallel) being stronger
than the radial one (fault-normal) for the fundamental frequency and has been mentioned before
in instrumental and computational studies. Absolute spectral amplification values also agree be-
tween instrumental and numerical results at all depths for the CORSSA array.
Numerical results along the entire cross-section are then used to highlight the influence of the
sites features on seismic ground motion. In the time domain, the surface waves are seen, which
add to the duration of ground motion and render the wave field complex. Peak ground motion
does not seem to vary dramatically across the profile. The maximum amplification level in the
basin is not more than 1.5 or 2.5 times higher than at any other point uphill or downhill.
However, when calculating differential motion as opposed to translational, results are very
different. Uphill differential motion is not strong (though it is suspected by other researchers to
have caused structural damage in the past). Downhill, it increases modestly at the basin edge
and dramatically (as much as 6-10 times) as the basin deepens, up to a distance of roughly 500-
600 m from the toe. As seen before in the spectral domain, the motion is stronger in the SH
sense, indicating strong Love waves.
Uphill, topographic amplification in the time domain is observed up to 100 m from the crest
according to the P-SV scheme and only 40 m for the SH scheme. The respective maximum am-
plification factors are 1.17 and 1.12, compared to the respective 1D case that includes soil layer-
ing effects. Though this difference is not great, it is in agreement with previous observations
that the motion perpendicular to the ridge axis is often larger. Deamplification at the toe is
503
slight. Due to the relatively mild relief, the amplification factors prescribed by design codes as-
sume low values (1.07 and 1.2 for AFPS and EC8 respectively). The AFPS envelope does not
successfully envelop the numerical results (particularly those of the SV scheme), while EC8
could achieve that if applied up to the appropriate distance from the crest. In the frequency do-
main, for the SV calculations, TAF values between 1.15-1.35 are observed up to 50 m from the
crest over a frequency of roughly 4.5 Hz. For this frequency, the spectral amplification values
lie outside the envelopes provided by the code provisions.
7 ACKNOWLEDGEMENTS

The installation of CORSSA was funded by European research project CORSEIS (EVG1-
1999-00002) and its maintenance continues through the CORSSA agreement between Aristotle
University of Thessaloniki, Greece, National Kapodistrian University of Athens, Greece, and
Institut de Radioprotection et de Sret Nuclaire, France. The records at DIM were provided
by K. Makropoulos of the National Kapodistrian University of Athens. Signal processing bene-
fited significantly from SAC2008 (http://www.iris.edu/software/sac; Goldstein et al., 2003).
This research was partly conducted during the stay of OJK at Instituto de Ingeniera, UNAM.
FJCG thanks Coordinacin de la Investigacin Cientfica, UNAM, for support.
REFERENCES

AFPS (Association Franaise du Gnie Parasismique). 1995. Guidelines for seismic microzonation stud-
ies. AFPS/DRM, 45 pp.
Apostolidis P., Raptakis D, Pandi K., Manakou M., and Pitilakis K. 2005. Definition of subsoil structure
and preliminary ground response in Aigion city (Greece) using microtremor and earthquakes, Soil
Dyn. Earthq. Eng. 26: 922 - 940.
Athanasopoulos G.A., Pelekis P.C., and Leonidou E.A. 1999. Effects of surface topography on seismic
ground response in the Egion (Greece) 15 June 1995 earthquake. Soil Dyn. Earth. Eng. 18: 135-149.
Bard P.-Y. 1999 Local effects on strong ground motion: Physical basis and estimation methods in view of
microzoning studies. Proc. of Advanced Study Course in 'Seismotectonic and Microzonation tech-
niques in Earthquake engineering, Kefalonia, Greece.
Borcherdt, R.D. 1970. Effects of local geology on ground motion near San Francisco Bay, Bull. Seism.
Soc. Am. 60: 29-61.
Bouckovalas G., Gazetas G., Papadimitriou A. 1999, Geotechnical aspects of the Aegion (Greece) earth-
quake. 2
nd
Intern. Conference on Geotech. Earthq. Engineering, Lisbon.
CEN (European Committee for Standardisation). 2003. Eurocode 8: Design of structures for earthquake
resistance. Part 5: Foundations, retaining structures and geotechnical aspects, Draft No. 5 (prEN
1998-5). Brussels, Belgium.
Chvez-Garca F.J., and Faccioli E. 2000. Complex site effects and building codes: Making the leap. J.
Seismology 4: 23-40.
Geli L., Bard P.Y., and Jullien B. 1988. The effect of topography on earthquake ground motion: a review
and new results. Bull. Seism. Soc. Am. 78: 42-63.
Goldstein, P., D. Dodge, M. Firpo, Lee Minner. 2003. SAC2000: Signal processing and analysis tools for
seismologists and engineers, Invited contribution to The IASPEI International Handbook of Earth-
quake and Engineering Seismology. WHK Lee, H. Kanamori, P.C. Jennings, and C. Kisslinger (eds).
Academic Press, London.
ITASCA Consultants FLAC. 2002. Fast Lagrangian Analysis of Continua, S. A. Version 4.0.
Jongmans D., and Campillo M. 1993. The response of the Ubaye Valley (France) for incident SH and SV
waves: Comparison between measurements and modelling, Bull. Seismol. Soc. Am. 83(3): 907-924.
Kennet, B.L.N. 1983. Seismic wave propagation in stratified media. Cambridge University Press.
tenidou O.-J., Arnaouti S., Chvez-Garca F.-J.,Raptakis D., Pitilakis K., Diagourtas D., Makropoulos
K. Site and topography effects on ground motion - Instrumental results from Aegion. 3
rd
Panhellenic
Conference on Earthquake Engineering and Engineering Seismology, Athens, 57 November, 2008.
Ktenidou O.-J., Raptakis D., Apostolidis P.,Pitilakis K. Aspects of surface topography and site effects -
Experimental and numerical studies at Aegion, Greece. 4
th
International Conference on Earthquake
Geotechnical Engineering, Thessaloniki, Greece, Thessaloniki, 25-28 June 2007.
Lermo J., and Chvez-Garca F.-J. 1993. Site effect evaluation using spectral ratios with only one station,
Bull. Seismol. Soc. Am. 83(5): 1574-1594.
504
Makra K, Raptakis D, Chvez-Garca FJ, Pitilakis K. 2001. Site effects and design provisions: the case of
Euroseistest. Pure Appl. Geophys. 58: 234967.
Moczo P. 1989. Finite difference technique for SH waves in 2-D media using irregular grids: application
to the seismic response problem. Geophys. J. Int. 99: 3219.
Moczo P, Bard P-Y. 1993. Wave diffraction, amplification and differential motion near strong lateral dis-
continuities. Bull Seism Soc Am 83: 85106.
Raptakis D., Makra K., Anastasiadis A., and Pitilakis K. 2004a. Complex site effects in Thessaloniki
(Greece): I. Soil Structure and Comparison of Observations with 1D Analysis. Bull. Earthq. Eng. 2:
271-300.
Raptakis D., Makra K., Anastasiadis A., and Pitilakis K. 2004b. Complex site effects in Thessaloniki
(Greece): II. 2D SH modelling and engineering insights. Bull. Earthq. Eng. 2: 301-327.

505
i

ABSTRACT: This is the first in a series of two papers presenting a decoupled methodology for
the design of slope stabilising piles. The proposed methodology combines the widely accepted
analytical calculations to obtain the required stabilising force, with non-linear finite elements
analysis to obtain the ultimate lateral capacity of piles. The soil and pile constitutive model is
validated against experimental and observational data. The simplified numerical model results
are compared satisfactorily with 3D Finite Elements analyses and theoretical studies.
1. INTRODUCTION
Landslides affect not only the structures founded on the slope, but also these ones lying in the
area of soil mass deposition. Stabilisation measures most often refer to the strengthening of the
interface. One of the most effective methods of this category is interface nailing using piles.
Their use as a means of stabilization has been discussed, among others, by De Beer & Walleys,
1970; Ito and Matsui, 1975; Sommer, 1977; Fukuoka, 1977; D Appolonia et al. 1977; Wang et
al, 1979; Nethero, 1982; Yamada et al, 1971; Fukumoto, 1972, 1973; itazima and Kishi,
1967; Leussink and Wenz, 1969; De Beer and Wallays, 1972; Nicu et al., 1971; Marche &
Lacroix, 1972; eyman & Boersma, 1961; Heyman, 1965; De Beer et al., 1972; Tschebotarioff,
1973; to & Matsui, 1975; Hassiotis et al., 1997; Poulos, 1995; Chen & Poulos, 1993; Oakland
& Chameaou, 1984; Goh et al., 1997; Poulos and Chen, 1997.
Most analysis methods are decoupled, i.e. they neglect the potential modification of the
shape and position of the failure surface due to the very existence of piles. The pile is modeled
as a flexural beam connected with the surrounding soil through non-linear springs. The soil
displacement profile is transmitted to the beam through these non-linear springs. All analysis
techniques necessarily include a number of simplifying assumptions regarding the springs.
Moreover, the actual soil displacement profile is not straightforward to obtain; it must either be
speculated or obtained by some sort of analysis (e.g. by means of finite elements) or by field
measurements.
For this reason, this study attempts to propose an improved design methodology for slope
stabilising piles which will maintain the simplicity of the most widely used methods but will
simultaneously take advantage of rigorous finite elements calculation of the ultimate load of
the soil-pile system incorporating all the complex non-linear phenomena discussed in the
previous section (i.e. soil arching, soil-pile interaction and pile-pile interaction).

Piles for Stabilising Seismically Precarious Slopes.
Part A : Development and Validation

R. Kourkoulis, F. Gelagoti, I. Anastasopoulos, G. Gazetas
Soil Mechanics Laboratory, National Technical University Athens, Greece
506
2. PROPOSED METHODOLOGY
2.1. Fundamentals of the Proposed Procedure
The studied configuration is illustrated in Figure 1.A row of piles is embedded within the slope
which is prone to failure. The upper soil layer, termed the unstable soil, overlies the stable
soil layer. Inside the unstable layer lies the potential sliding interface. The main issue of the
study is the estimation of the enhancement to stability offered by the piles.

Figure 1. Problem definition; a row of piles embedded within a slope which is prone to failure

The general design procedure follows the well documented decoupled approach described by
Viggiani (1981), Hull (1993) and Poulos (1995, 1999). It consists of two main steps :

1. Evaluation of the total shear force needed to increase the existing safety factor for the
slope (based on an analysis without reinforcement) to the desired value.
2. Evaluation of the maximum shear force that each pile can provide to resist sliding of
the potentially unstable portion of the slope without exceeding a preset design
displacement limit; and selection of the type, number and most suitable location of the
piles.

In the first step, the driving (F) and the Resisting Forces (FR )along the slip surface is
calculated utilizing one of the widely used slope stability analysis techniques (e.g. Sarma,
Spencer, Bishop, Janbu). Most of these methods discretize the slope in slices and integrate the
forces along each slice to obtain the total driving and resisting forces.
If the actual safety factor SF is less than the target safety factor , F
1
, the piles must provide
and additional resistance AR, so that :

F
1
=
lR+AR
lP
D
(1)
Hence, the required stabilizing force per unit width of soil that is to be provided by the piles
may be calculated as:

AR = XF

(F
1
- SF) (2)
For step (2), a reasonable procedure is to analyse the pile against lateral soil movements
simulating the movement of a landsliding mass. Knowing the force per unit width that the piles,
their properties and configuration may be obtained with 3 dimensional non-linear finite element
analysis. Design charts can thus be developed to this end, for various soil and pile properties
and pile configurations. Such a chart plots the Resisting Shear Force for the given soil
conditions and landslide depth to decide upon the optimum pile configuration. Hence, the
tedious procedure of designing the pile reinforcement is actually simplified to solving the slope
stability equations.
The lateral capacity of the pile systems subjected to slope movements may be rigorously
assessed utilizing the Finite Element technique which provides the ability to model the whole 3
507
dimensional geometry of the slope and pile system examined. Despite its rigor though, the
analysis of the full model may be computationally ineffective because:
1. Pile loading, independently of the slope inclination and interface position is stemming
from the application of a uniform displacement profile along the pile length. The
reasonable validity of uniformity of the displacement distribution has been verified by
Kourkoulis (2009) and is proposed by Poulos (1999) as very much appealing to the
case of slope displacements.
2. Although the required pile resistance force is indeed a function of the slope geometry,
its calculation has already been incorporated in the slope stability analysis described in
Step (1). This decoupled approximation is totally realistic in case of pre-existing sliding
planes within the soil mass.
3. Moreover, even if the slope geometry remains the same but the position of the interface
(or of piles along the slope) has to be parametrically varied, a new model must be
constructed and analysed each time, at enormous computational effort.
Therefore, in this section a more versatile simplified Finite Element Model is proposed that
may be utilized for multiple parametric analyses. Despite the fact that the amount of induced
loading on the pile is indeed a function of the slope inclination and interface properties, the
ultimate load which is sought for at this stage is assumed to be a function of the interface
position and soil properties only. Therefore, the model suitable for computing the pile ultimate
load concentrates on the region around the pile. The validity of this demonstrated below..
3. NUMERICAL MODEL
3.1. Description
The model is schematically displayed on Fig. 2, where the slope geometry has been eliminated.
The position of the interface is defined by the depth of the unstable soil layer in the area around
the pile. Therefore, the interface should be lying at depth E
u
from the free surface of the model.
The new model focuses on the region which is mostly affected by the pile. Therefore, for a
pile of diameter D, the model dimension should be at least 10D, i.e 5D of soil behind and 5D of
soil in front of the pile as shown in Fig. 2 (e.g. Reese and Van Impe, 2001).
Unlike the model length and depth which are determined by the slope geometry, the model
width is a function of the piles spacing. The FE model represents a typical slice of the slope
stabilized with piles spaced at distance S, which is assumed to be repeated infinitely in the y-
direction. Consequently the width of the model is equal to 2S. [Pile center-to-center spacing: s,
distance of each pile from the nearest side 0.5 s, i.e. model total of 2s]. The displacement is
applied following a non-linear increase from zero to an ultimate value. The maximum value of
the imposed displacement must be sufficient to mobilize the full resistance of the pile-soil
system. A uniform displacement profile is applied in accord with several other researchers.

3.2. Soil and Pile Constitutive Modelling
The soil is assumed to obey an elastoplastic constitutive model with Mohr-Coulomb failure
criterion. The bottom stratum is considered to be rock and is therefore modelled as a stiff elastic
material. The overlying soil layers are assigned the properties of the soil being modelled in each
case. A row of hexahedral elements of reduced strength (residual strength parameters) models
the sliding interface.
As schematically illustrated in Fig. 3 the pile is modeled as a 3D beam element is
circumscribed by 8-noded hexahedral solid elements. The central beam element node is
connected with the circumferential solid element nodes at the same height through appropriate
kinematic restraints. The surrounding elements have zero rigidity : their presence only aims at
capturing the solid geometry of contact between soil and pile.
508
Both elastic and inelastic piles are modeled. For the latter case, the moment-curvature
relation, H- 0, of the pile cross-section is required, which initial depends (for reinforced-
concrete piles) on the amount of reinforcement. An example is given in Fig.3.

Figure 2. Schematic Representation of the proposed versatile model for assessing the Pile lateral
capacity.

Figure 3. Schematic Representation of the proposed versatile model for assessing pile lateral capacity.
3D8nodedhexahedral
solidElements,E=0
Imposedkinematic
Displacement
PileBeam
(B31inABAQUSlibrary)
7428
10/20
PileCrossSection C20/25,S400
0
2000
4000
6000
8000
10000
0 0.005 0.01 0.015 0.02
M (kNm)
Curvature
Elastic
behavior
Nonlinearbehavior
Uniformly imposed
displacement
Stable Soil
Unstable Soil
Interface
H
u
L
e
L
p
u
u
5D 5D
Piles of
diameterD
S / 2
S / 2
S
509
4. NU
4.1.
The n
Kour
prese
4.2. T
This
groun
was a
the sl
upslo
pile h
The d
5(a) a

Figure
instru
of the

UMERICAL
Validation of
numerical m
rkoulis (2009
ents the valid
The experim
unique 16-y
nd is compos
assigned at t
liding soil an
ope of the em
head was pu
displacement
and 5(b).
e 4. (a) Geolo
umentation ; an
e embankment
L MODEL V
f Soil and Pi
modeling of p
9) against exp
dation of the
ental site of
year field ex
sed of 5 m to
the pile head
nd the pile. H
mbankment.
lled back to
t time history
ogical Section
nd (c) pile hea
t (Frank and P
VALIDATIO
ile Constituti
piles subject
perimental d
numerical m
Salledes (Fra
xperiment ref
o 8 m thick
d in order to
Hence, the pil
Due to mov
its original
y at free fiel
n of the site
ad displaceme
Pouget, 2008)
ON
ive Behaviou
ted to soil la
data (Rosque
model utilizin
ank & Rouge
fers to a site
colluvial ma
achieve the
le head was
vements link
position fou
d (inclinome
at the positio
ent was set to

ur
ateral displac
et, 2004; Pras
ng field data b
et, 2008)
e with surfac
arls (Fig.4). A
e maximum r
held in place
ked to its anc
ur times in th
eters G2, G3
n of embankm
zero by a dea
cement has b
sad,1999). T
by Frank & P
ce inclinatio
A zero displa
relative disp
e by a deadm
choring syste
he 16 years o
and G5) are
ment B ; (b)
ad man anchor
been validat
The next para
Pouget, 2008
on of 7% , a
acement con
lacement be
man anchor lo
em (creep et
of the experi
e plotted in F
site plan view
r located at up
ted by
agraph
8.
and its
ndition
tween
ocated
tc) the
iment.
Figure
w and
pslope
510

Figure
and P

Figure
2008)

Figure
mesh

e 5(a). Displa
ouget, 2008).
e 5(b). Groun
).
e 6. Finite El
near the pile.
cements of th

d displacemen
lement Model

e natural grou
nts at the grou
l utilized for t
und with time

und surface (in
the simulation
(mean of incl
nclimometers
n of the Sall
linometers G2
G5 and G4), (
des case stud
2 and G3) (Fra
(Frank and Po
dy and detail
ank
ouget,
of the
511
4.3. Comparison with 3D Finite Element Model Results
The finite element model is depicted in Figure 6. In the original case, the interface properties
changed seasonally, resulting to soil movements and pile deformation. In the Finite Element
analysis presented here, the shear strength of the interface is assumed to be reducing until the
onset of failure. This is achieved through a user subroutine encoded in ABAQUS which defines
the strength reduction pattern. The deadman anchor force was modelled as an induced
concentrated force on the pile head. The data provided by the author have been utilized to model
the sequence of : (a) pile loading , (b) pile head deflection, and (c) anchor pull-back.

Figure 7. Finite Element Model utilized for the simulation of the Salldes case study and detail of the
mesh near the pile.

Figure 8. Pile displacements before (left figure) and after (right figure) the pulling-back operation of
November 1986 (upper figure), September 1992 (middle figure) and July 1995 (bottom figure). Black
line denotes the numerical analysis results while the field measurements of Frank & Pouget are denoted
with the gray markers.
0
5
10
15
20
25
30
0 0.2 0.4 0.6 0.8 1 1.2
analysis step time
F
r
e
e
-
f
i
l
e
d
s
o
i
l

d
i
s
p
l
a
c
e
m
e
n
t

(
c
m
)
Experimentaldata
Numerical
prediction
512
The Finite Element analysis was performed in several steps. An initial run (artificial
reduction of the shear strength along the interface) was performed (without the pile) to compute
the time history of free field soil displacement (Figure 7).
To model the seasonal variation of the slip surface strength, the strength reduction pattern has
been calibrated accordingly. The anchor force vs soil displacement can be converted to anchor
force time history (where the time is measured in analysis step time and not real time). This
loading history is applied on the pile head in the second run of the analysis.
During this second run, the analysis is repeated, but with the pile installed in its place. The
produced pile deflection line before and after each anchor pulling is plotted in Figure8 at
various stages of the anchor pull-back, compared with the Frank & Pouget (2008)
measurements. It is observed that the model accuracy in the prediction of the slope and pile
displacements is remarkable, enhancing the confidence in the proposed modelling technique.
5. VALIDATION OF THE SIMPLIFIED NUMERICAL MODEL
5.1. Validation Against Rigorous 3D FE Analyses
The proposed simplified procedure is validated via comparisons with the 3-D finite element
model. We assume a slope of the geometry and soil properties displayed in Fig. 9. The top layer
is assumed to be a relatively loose sandy soil with = 28
o
, c = S kPo, and = 2
o
. The
bottom layer is assumed to be soft rock with Su = 6uu kPo. A predefined sliding interface
(highlighted in blue in Fig. 9) is assumed of residual strength : = 16
o
, c = S kPo, and
= 1
o
.
5.2. Slope Stability Analysis : Calculation of Required Pile Shear Force
Utilizing the simplified Bishops method, the sum of the driving forces is XF
= 19S6 kNm,
while the sum of the resisting forces is calculated to be XF
R
= 1744 kNm. Hence the slope is
characterized as unstable. The analysis seeks for the extra resistance force that must be obtained
by the piles for the safety factor to become unity, i.e the pile force must be: RF = 192 kNm.

Figure 9. The Full 3D Finite Element model. The pre-defined slip surface is highlighted in blue.

=23
o
Predefined slip surface :=16
, c=3 kPa, =1
Stiff Rock
Su = 600 kPa
Soft sand
=28
, c=3 kPa , =2
20 m
50 m
Averageelementsize:0.8m
Symmetryconditionsassumedat
modellateralboundaries
513
5.3. Full 3 Finite Elements Analysis
An initial analysis of the 3D model is performed in free field conditions, (i.e. without any
piles). A snapshot of the deformed mesh at the end of the analysis is portrayed on Figure 10(a).
The unstable soil clearly slides along the slip surface. Failure is confirmed by the increasing
nodal velocity of elements on the sliding soil mass indicated on the time history plotted on
Figure10(b).
In the next step, the analysis is repeated with the presence of reinforced concrete piles of
diameter = 1.2 m spaced at S are used to stabilize the slope. For the example geometry
under consideration the depth of the unstable soil is E
u
= 8 m. The results of the fully non-
linear FE analysis are displayed in terms of horizontal displacement contours on Figure 11(a). It
is evident that the chosen pile configuration able to prevent slope failure. The time history of the
pile head displacement is portrayed in Figure 12. Note that the ultimate pile head displacement
is u
p
=2 cm.

5.4. Comparison between the 3D and the proposed simplified FE model
The simplified model utilized for the assessment of the piles lateral capacity for the case
examined is shown in Figure 13. The model is imposed to a horizontal displacement u
]]
. The
plot of the net resistance force against pile head displacement is displayed on Figure 14. For pile
head displacement u
p
= 2. cm (which corresponds to the pile deformation necessary to impede
the landslide calculated by the fully 3D FE analysis), the corresponding resistance force
calculated by this model is RF = 17S kNm, which is almost practically equal to the one
calculated through slope stability analysis.

Figure 10. Snapshot of the deformed mesh in the free-field case (upper figure) and plot of displacements
time history of 2 nodes on the ground surface (bottom figure).
NodeA
NodeB
0
0.05
0.1
0.15
0.2
0 0.5 1 1.5 2 2.5
NodeA
NodeB
t (sec)
u (m)
514
5.5. Validation Against Theoretical Studies

Poulos (1995) suggests that for the calculation of the ultimate lateral pressure developed on a
pile by flowing sand, the simplest approach is to utilize the formula proposed by Broms (1964),
in which:

P
u
= o K
p
o
o
(3)
K
p
is the Rankine passive pressure coefficient which is defined as
K
p
= ton
2
(4S +
q
2
) (4)
Where :
is the angle of internal friction of soil ;
o
o
is the effective overburden pressure ;and o is a coefficient ranging between 3 and 5.
For clayey soils, total stress approach is usually adopted, in which P
u
is related to the undrained
shear strength S
u
as follows :

P
u
= N
p
S
u
(5)
where N
p
is the lateral capacity factor.
The lateral pile capacity can be calculated by means of the finite elements method utilizing
the decoupled design approach presented in the previous sections. The plot of Resistance Force
developed per pile (and not per unit width) versus soil free-field displacement for the case of a
shallow landslide in sand stabilized by piles is displayed in Figure 15. for three cases of pile
spacings. The ultimate value (RF
uIt
) of the Resistance Force is clearly indicated by the curve
flattening (displacements keep increasing while the resistance force remains constant). The
ultimate lateral soil pressure calculated according to the Broms (1964) formula is:

P
u
= S ton
2
[4S +
q
2
o
i
o
(6)
And therefore the theoretically calculated ultimate pile resistance force is approximately:

RF = S ton
2
[4S +
q
2

1
2
y z
2
(7)
for z = 6m and D=1.2 RF is equal to 3590 kN

The value of the ultimate pile resistance force calculated by the numerical analysis for the
three cases is:
4: RF
uIt
= 3650 kN For pile spacing
S: RF
uIt
2: RF
uIt
It can be seen that the calculated values RF
uIt
for the case of closely spaced piles deviate
substantially from the Broms 1964 approach. This is due to the group interaction effect caused
by neighboring piles.

6. CONCLUSION
In this paper, an improved methodology has been presented and validated for the design of slope
stabilising piles. The calculation of the pile ultimate lateral capacity is achieved by means of 3
dimensional finite elements analyses. The comparison of numerical model predictions with
analytical, theoretical and field data reveals very good. The methodology presented herein is
used in the second (companion) paper to contact a detailed parametric analysis, and derive
deeper insight, on the performance of slope stabilizing pile.
515

Figure 11 (a). Contours of horizontal displacements when piles of 1.2m diameter are installed with 3D
pile to pile distance.

Figure 11 (b). Zoom-in in the pile area. Note that the model can capture the 3-d displacement pattern
around the piles.

Figure 12. Displacement time history at the head of the slope stabilizing pile
1.2mdiameter piles spacedat3D
6D
8 m
Depthuntilinterface: 8m
0
0.005
0.01
0.015
0.02
0.025
0 1 2 3 4 5 6 7
t (sec)
u
p
(m)
g
e
o
s
t
a
t
i
c
u
geost
=5mm
u
p
max
=2cm
516

Figure 13. The Simplified Finite Element Model utilized for the design of the stabilization piles of the
previously described case study.

Figure 14. Pile resistance Force against pile head displacement. Case study: Pile spacing at 3D, landslide
depth 8m, unstable soil properties = 28
, c=3 kPa.
Directionofimposed
displacementu
ff
Interface
=28
, c=3 kPa , =2
Unstable soil
Su = 600 kPa
Stable soil, elastic
x
y
z
0 1 2 3 4 5
0
100
200
300
400
500
600
700
u
p
: cm
R
F

:

k
N
/
m

2D
Type HOM, linear Pile Type HOM, linear Pile
atu
p
=2cm
u
p
(cm)
R
F

(
k
N
/
m
)
176kN/m
Pilespacing3D
700
517

Figure 15. Resistance Force Developed on each pile as a function of the free field displacement and pile
spacing.
ACKNOWLEDGEMENT
The research presented in this paper was financially supported by the Secretariat for Research
and Technology of Greece, under the auspices of PENED Programme with Contract number
03ED278.
REFERENCES
Bishop, A. W. (1955), The use of the slip circle in the stability analysis of slopes,
Geotechnique, 5, No. 1, pp. 7-17
Broms, .B., (1964),Lateral resistance of piles in cohesionless soils, J. Soil Mech. Foundations,
Div. ASCE, 90, pp. 123-156
Cai, F., and Ugai, K., (2000), Numerical analysis of the stability of a slope reinforced with
piles, Soils and Foundations, Japanese Geotechnical Society,40(1), pp. 7384.
Chen, LT, and Poulos, HG., (1993), Analysis of pile-soil interaction under lateral loading
using infinite and finite elements, Computers and Geotechnics ,15, pp. 189220.
D'Appolonia, E., Alperstein, R., and D'Appolonia, D.J., (1997)."Behaviour of colluvial slope",
J. Soil. Mech. And Found. Div., ASCE, 93, pp. 447-473
De Beer, E.E, and Wallays. M.,(1972),Forces induced in piles by unsymmetrical surcharges
on the soil round the piles, Conference on Soil Mechanics and Foundation Engineering;1,
pp. 32532.
Frank R., and Pouget P., (2008) Experimental pile subjected to long duration thrusts owing to a
moving slope" , Geotechnique, 58, No 8, pp. 645-658
Fukumoto, Y., (1972), Study on the behaviour of stabilization piles for landslides, Soils and
Foundations, 12(2), pp. 6173. (In Japanese.)
Fukuoka, M. (1977), "The effects of horizontal loads on piles due to landslides", Proc. 10
th

Spec. Session, 9
th
Int. Conf. on SMFE, Tokyo, pp. 27-42
Goh, ATC, The, CI, Wong, KS., (1997), Analysis of piles subjected to embankment induced
lateral soil movements Journal of Geotechnical and Geoenvironmental Engineering,
ASCE,123(4), pp. 31223.
Hassiotis S, Chameau JL, Gunaratne M., (1997), Design method for stabilization of slopes
with piles, Journal of Geotechnical and Geoenvironmental Engineering, ASCE ,123(4), pp.
31423.
0 10 20 30 40 50
0
1000
2000
3000
4000
u
ff
: cm

2D
3D
4D
4000
3000
2000
1000
0
0 10 20 30 40 50
u
ff
: cm
3100
2200
RF per pile (kN)
3650
518
Heyman, L., and Boersma, L., (1961), "Bending moment in piles due to lateral earth pressure"
Proc., 5
th
ICSMFE, Paris, Vo2, pp. 425-429
Hull TS (1993) Analysis of the stability of slopes with piles, 11
th
Southeast Asian
Geotechnical Conference, May, Singapore, pp. 639-643
Ito, T. and Matsui, T. (1975). Methods to estimate lateral force acting on stabilizing piles,
Soils and Foundations, Vol. 15, No. 4, pp. 43-60.
Janbu N. (1957), "Earth pressures and bearing capacity calculations by generalised procedure of
slices." Proc. 4
th
Int. Conf. Soil Mech. Found. Eng., 2, pp. 207-212.
Kitazima, S., and Kishi, S., (1967), "An effect of embedded pipes to increase resistance against
circular slides in soft clay foundation", Technical Note of Port and Harbour Research
Institute, No 29, pp. 63-94 (in Japanese)
Leussink, H., and Wenz, K. P., (1969), "Storage yard foundations on soft cohesive soils", Proc.,
7
th
ICSMEE, Mexico, Vo. 2, pp. 149-155
Nethero, M.F, (1982), Slide Control by Drilled Pier Walls. Application of Walls to Landslide
Control Problems R.B. Reeves, ed., ASCE, New York, pp. 19-29.
Nicu, N. D., Antes,D . R., and Kesslerr, S., (1971), Field measurements on instrumented piles
under an overpass abutment. High. Res. Board Rec, No. 354, Washington, D.C.
Oakland, M.W., Chameau, JLA., (1984), Finite-element analysis of drilled piers used for
slope stabilization, American Society for Testing and Materials, pp. 18293.
Poulos H.G. (1999) "Design of slope stabilizing Piles", Slope Stability Engrg, Yagi, Yamagami
and Jiang, Balkema, Rotterdam
Poulos HG, Chen LT (1997) "Pile response due to excavation-induced lateral soil movement
Journal of Geotechnical and Geoenvironmental Engineering, 23(2), pp. 949.
Poulos HG. (1995) Design of reinforcing piles to increase slope stability, Canadian
Geotechnical Journal ,32, pp. 80818.
Prasad Y., and Chari T.R. (1999) "Lateral Capacity of model rigid piles in cohesionless soils",
Soils and Foundations, 39,2, pp. 21-29
Reese L.C. and Van Impe W. F. (2001), "Single piles and pile groups under lateral loading",
A.A. Balkema/Rotterdam/Brookfield, Book pp.463.
Rosquet F., (2004) "Pieux sous charge laterale cyclique", These de Doctorat, LCPC et
Universite de Nantes
Sarma S.K., (1973), "Stability analysis of embankments and slopes", Geotechnique, 23, 3, pp.
423 433.
Sarma, S.K. & Tan, D., (2006), Determination of Critical Slip Surface in Slope Stability
Analysis, Geotechnique, 56, No. 8, pp. 539-550
Tschebotarioff, GP., (1973), Lateral pressure of clayey soils on structures" Proc, 8
th
ICSMFE
Specialty Session 5, Moscow ;4(3), pp. 22780.
Viggiani, C. (1981), Ultimate lateral load on piles used to stabilize landslides, Proc. 10
th
. Int.
Conf. Soil Mechanics and Foundation Engineering, Stockholm, Vol. 3, pp. 555-560.
Wang, M.C., Wu, A.H. and Scheessele, D.J., (1979), Stress and deformation in single piles due
to lateral movement of surrounding soils, Behavior of Deep Foundations, ASTM 670,
Raymond Lunggren, Ed., American Society for Testing and Materials, pp. 578-591.
519
i

ABSTRACT: The simplified numerical model formulated and validated in the companion paper is
utilized to perform parametric analyses of piles embedded in an unstable slope undergoing lateral soil
movement. Pile diameter and spacing, depth of pile embedment, soil layering and stiffness are the key
problem parameters investigated. Dimensionless Charts are developed for guiding the design of piles
for slope stabilization.

1. INTRODUCTION
The pile design methodology presented in the companion paper involves two steps: (a) perform a
simple slope stability analysis to obtain the force required to stabilize the slope and (b) use of design
charts providing the maximum shear force that may be offered by the piles in order to decide upon the
optimum pile configuration. The present paper deals with the second step, aiming at exploring the
effect of key problem parameters on the behavior of slope stabilizing piles.
2. PARAMETRIC ANALYSES
Both cohesive and non-cohesive soils have been utilized to model the unstable soil. The interface
depth from the surface (
), is varied parametrically, covering from a shallow (H

u
= 4m) to a very
deep (H
u
= 12 m) landslide.
A plethora of parametric analyses have been performed for the behaviour of slope stabilizing piles
nailing unstable soil layers of various depths and material properties. The factors examined are:
(a) Effect of Pile Spacing
(b) Inhomogeneity of the Unstable Soil
(c) Strength of the Stable Soil Layer
(d) Depth of Pile Embedment into the Stable Layer
(e) Pile Non-linearity

Piles for Stabilising Seismically Precarious Slopes.
Part B : Parametric Analysis and Design Charts

R. Kourkoulis, F.Gelagoti, I. Anastasopoulos, G. Gazetas
Soil Mechanics Laboratory, National Technical University Athens, Greece
520
2.1. Soil Arching between the Piles
Wang and Yen (1974) studied analytically the behavior of piles in a rigid-plastic infinite soil slope
with emphasis on arching effects, and concluded that a critical pile spacing exists in both sandy and
clayey slopes, beyond which practically no arching develops. In general, arching stems from the stress
transfer through the mobilization of shear strength. It is the transfer of stress from yielding parts of a
soil mass to adjoining non-yielding or less compliant parts.
Two cases of pile arrangement are compared: a dense pile spacing assumed to ensure arching and a
loose one where the piles are distant and soil can freely flow between them. It is assumed that loading
is imposed on the free field (i.e. far enough from the pile region) on the soil nodes. It is supposed that
after application of the load, the pile displacement has a value of u
, while the soil between the piles

displaces u
. Soil arching is measured with the ratio u
/u
. If this ratio ranges between 1 and at

most 2, the pile and neighboring soil displace almost equally and the piles are effective thanks to
arching. For much higher u
/u
ratios arching is not effective.

As an example Fig.1 displays two characteristic snapshots of the FE analyses. The unstable soil
layer is considered to be a sand with 28
, 2
, and c 3 kPa. The bottom soil layer is

assumed to be very hard soil (bordering on soft rock) with S
600 kPa. The interface properties are

16
, c 3 kPa, and 1
o
Its location is assumed at 4m depth from the ground surface. The top
figure plots the displacements contours on the ground surface for the case of piles of diameter
D 1.2m spaced center-to-center at 2D, i.e, 2.4 m. From the distribution of displacement contours
distribution it is evident that the soil between the piles has been restricted by the presence of the piles
hence displaced almost equally with thema clear manifestation of arching. On the contrary, in the
case of piles spaced at 7D (bottom figure), the intermediate soil has not been confined by the piles and
flows between them.
3D numerical parametric analyses have been performed to define the maximum pile spacing that
ensures sufficient degree of arching as a function of their diameter. The results are summarized in Fig.
2, which plots the dependence of the U
ip
/ U
p
ratio on U
p
/U
p
max
. It is apparent that spacings of 2, 3 and
4 times the pile diameter are able to provide soil arching. For spacings greater than 5 diameters soil
flows between the piles; such arrangements are therefore not applicable to slope stabilization and will
not further examined. Evidently, the most effective pile arrangement in terms of arching is the spacing
of 4 diameters. However, the 2D and 3D cases of will be examined since the piles attract smaller
forces in these cases.
2.2. Effect of Pile Spacing
Spacings of 2, 3 and 4 times the pile diameter have been investigated. While, increasing pile spacing
improves their effectiveness, it nevertheless reduces the total resistance force offered per unit width.
The total stiffness and capacity of the soil-pile system is a function of both pile spacing and height of
the unstable soil layer. In the case of the shallow landslide of
u
= 4m (Fig. 3), RF is nearly
independent of pile spacing. In the deep landslide (H
u
= 8m), (Fig. 4), the differences among the
different pile spacings are clear : since the soil-pile system is much more flexible, hence requiring a
substantially increased pile deformation for the same Resistance Force (RF). Note that these results
refer to elastic piles of diameter D = 1.2m. In case of the non-linear pile, the maximum realistic
moment that may be developed must not exceed the actual structural strength of the pile. Hence, it
must be pointed out that although the maximum value of RF developed (elastically) is nearly
independent of pile spacing for all landslide depths, the only acceptable ultimate RF values are those
which are achieved at acceptable levels of the bending moment.

521

Figure 1. Contours of horizontal displacements (a) of a dense pile configuration (distance between piles
2D) and (b) of a sparse pile configuration (pile distance 7D).

Figure 2. Comparison of the interpile displacements calculated for different pile spacings in sandy soil. It is
obvious that for spacings S > 5D, soil flows between the piles.
Imposed
Force
Imposed
Force
1
6
11
16
21
26
31
0 0.2 0.4 0.6 0.8 1 1.2
U
i
n
t
e
r
/
U
p
i
l
e
Up / Upmax
PILES ANA 2D
PILES ANA 4D
PILES ANA 7D
PILES ANA 5D
U
p
/ U
p
max
U
i
p
/

U
p
S = 2D
S = 4D
S = 7D
S = 5D
S = 7D
S = 5D
S 4D
u
i
p
/

u
p
u
p
/ u
p
max
522

Figure 3. Resistance Force (RF) offered by the pile vs Pile head displacement diagrams and RF vs maximum
Bending Moment Diagrams for various pile spacings in case of a shallow landslide of H
u
= 4m

Figure 4. Resistance Force offered by the pile vs Pile head displacement diagrams and RF vs maximum Bending
Moment Diagrams for various pile spacings in case of a deep landslide of H
u
= 8m.
1.2 m
H
u
= 4 m
Slope and Pile Characteristics
L
e
= H
u
Pile Arrangement
= 2D, 3D or 4D S
1.2 m
H
u
= 8 m
L
e
= H
u
Pile Arrangement
= 2D, 3D or 4D S
523
2.3. Effect of Soil Inhomogeneity
Both homogeneous and non-homogeneous soils in terms of elasticity modulus have been investigated.
For the case of non-homogeneous soils the modulus has been assumed to vary linearly but with the
same mean value as the homogeneous soil. For the cases examined, soil inhomogeneity only faintly
affects the results in the shallow landslide case of H
u
= 4 m which involves very small displacements.
2.4. Effect of Pile Non-linearity
Both linear and non-linear piles have been examined. For the case of non-linear piles, the considered
longitudinal reinforcement is the maximum amount allowed by the Greek Reinforced Concrete Code
(EKOS 2000) : 4% of the cross-sectional area. Pile non-linearity is introduced via its Moment
Curvature relationship. The ultimate Moment value for the case of a pile diameter of D = 1.2 m is
M
ult
= 7.2 MNm.
Discrepancies between the linear and non-linear pile are noticed only in flexible soil-pile systems,
i.e. either in sparsely spaced piles or with deep landslide. In the shallow landslide case (H
u
= 4m), the
system behaviour is practically elastic; hence the difference in the behaviour between the two pile
types is negligible (Fig. 5). On the contrary, the differences are conspicuous in the deep landslide case
(Fig. 6). The behaviour of the two types is the same as long as the pile remains below its yielding
point. Once the ultimate bending moment is reached, RF offered by the non-linear pile does not
increase further. However, up to that point the two pile types behave identically.

Figure 5. RF vs Pile head displacement diagrams and RF vs maximum Bending Moment Diagrams with
and without considering pile non-linearity in case of a shallow landslide of H
u
= 4m

1.2 m
H
u
= 4 m
L
e
= H
u
Pile Arrangement
= 2D, S
Non-linear Pile
524

Figure 6. RF vs Pile head displacement diagrams and RF vs maximum Bending Moment Diagrams with
and without considering pile non-linearity in case of a deep landslide of H
u
= 8m

2.5. Effect of the Strength of the Underlying Stable Ground
The strength and stiffness of the stable soil varied parametrically to model materials ranging from a
relatively soft sand of low strength to a very stiff rock. The soils examined are:

(a) loose silty sand : 28
, 2
, 3 , 16
(b) dense sand : 38
, 2
, 0 , 32
(c) soft rock : 45
, 5
, 50 , 1.2
(d) stiff rock : 45
, 5
, 100 , 4

The strength parameters of the stable layer have been chosen so that the ultimate lateral pile soil
pressure be compatible with the ultimate passive soil pressure of the unstable layer. The latter is
defined as
for cohesionless soils, as
for clayey soils of

undrained shear strength (Broms 1964)
The strength properties of each stable soil type examined correspond to the following values of the
ultimate lateral pile soil pressure
.

i.
for the loose sand

ii.
1.6
for the dense sand

iii.
for the soft rock

iv.
for the rock

The stable layer strength determines the fixity conditions of the pile below the interface. As
expected, the analysis reveals that the very soft stable layer is unable to provide adequate fixity
1.2 m
H
u
= 8 m
L
e
= H
u
Pile Arrangement
= 2D, S
Non-linear Pile
0 5 10 15 20 25 30 35 40
0
500
1000
1500
2000
u
p
: cm
R
F

:

k
N
/
m

linear
nonlinear
0 0.5 1 1.5 2
x 10
4
0
500
1000
1500
2000
max M : kNm
R
F

:

k
N
/
m

linear
nonlinear
525
conditions, thus enabling the rotation of the pile as a rigid body. Conversely, with a stiff stable layer,
the pile displacement is mainly attributed to its deformation and thus leads in development of
substantial bending moments. Fig. 7 reveals that a pile embedded in the low strength substratum
would not provide the same level of ultimate resistance as when embedded in a stiff stratum, unless it
is greatly displaced.

Figure 7 (a). The effect of the stiffness of the Stable Ground.Unstable Ground Characteristics: G=16 MPa,
=28
o
, c=1 kPa, H
u
= 6m. Pile Characteristics: D=1.2m, L
e
=H
S
, , S = 4D, Elastic pile.
= 4D S
Piles Arrangement
1.2 m
L
e
= H
s
Dense Sand
Pu
L
e
> 2 H
s
Soft Rock
Rock
or
or
6 m
526

Figure 7(b). Effect of the stiffness of the Stable Ground.Unstable Ground Characteristics: G=16 MPa,
=28
o
, c=1 kPa, H
u
= 6m. Pile Characteristics: D=1.2m, L
e
=H
S
, S=4D, Elastic pile. ()

2.6. Effect of the Depth of Pile Embedment (
)
The depth of the embedment of the pile into the stable ground influences the pile behavior, depending
on the strength of the soil and the thickness of the sliding soil that must be stabilized. The embedment
depth
is expressed as a function of the height
of the unstable block. The values examined are:

0.7
, H
u
, 1.2 H
w
, and 1.5 H
u

From the results (Fig.8 and Fig. 9) it appears that the required embedment increases with
decreasing stable soil strength. Insufficient embedment depth results in rigid body-type rotation of the
pile (Fig. 10). The optimum pile embedment depth will be this which ensures adequate pile fixity
while remaining economical.
= 4D S
Piles Arrangement
1.2 m
L
e
= H
s
Dense Sand
Pu
L
e
> 2 H
s
Soft Rock
Rock
or
or
6 m
527

Figure 8. Effect of Pile Embedment Length. Unstable Ground Characteristics: E=40 MPa, =28
o
, c=1 kPa,
H
u
= 6m. Stable Ground Characteristics: E=40 MPa, =28
o
, c=1 kPa. Pile Characteristics: D=1.2m, S=4D,
Elastic pile.

= 4D S
Piles Arrangement
1.2 m
L
e
= 0.7H
u
, H
u
,
1.2 H
u
or 1.5 H
u
Pu
6 m
Pu
528

Figure 9. Parametric Analysis Results investigating the effect of Pile Embedment Length. Unstable Ground
Rock with Characteristics: G=16 MPa, =28
o
, c=1 kPa, H
u
= 6m. Stable Ground Characteristics: E=3
GPa, =42
o
, c=50 kPa. Pile Characteristics: D=1.2m, S=4D, Elastic pile.

= 4D S
Piles Arrangement
1.2 m
L
e
= 0.7H
u
, H
u
,
1.2 H
u
or 1.5 H
u
Pu
6 m
3Pu
529

Figure 10. Snapshot of the FE analysis of pile subjected to lateral soil movement. The insufficient
embedment depth of the pile leads to its rigid-body-type rotation.

3. PRODUCTION OF DESIGN CHARTS FOR SINGLE PILES
Design charts have produced with non-linear finite elements analyses utilizing the versatile Numerical
Model presented and validated in the companion paper by Kourkoulis et al (2009). Fig. 11 is one such
set of charts.

Each graph in this figure portrays

a. Resistance Force per unit width (RF ) vs Imposed Displacement at Free Field (u
).
This plot actually provides the ultimate lateral resistance offered by the pile. The plateau of RF
defines the maximum calculated Resistance Force (RF) per unit width that may be offered by the
specific pile configuration. However, RF should always be considered in conjunction with the
required displacement.
b. Resistance Force per unit width (RF) vs Pile Head Displacement (u
)
c. Reaction Force per unit width (RF) vs Soil Displacement between Piles (u
)
The aim of this plot is clearly the investigation of the effect of arching. Although it has already
been discussed in the previous section that all the configurations and spacings examined provide
adequate arching, soil displacement between the piles (although small enough as to not signify failure)
is not independent of pile spacing. Therefore for reasons similar to those described above, it is
considered necessary that the soil displacement information should be available to the designer.
d. Reaction Force per width (RF) vs Pile maximum Bending Moment (maxM)
and
e. Reaction Force per unit width (RF) vs Pile maximum Shear Force (maxQ)
f. Distribution of Bending Moment vs depth at the moment of failure

x
y z
530

Figure 11. Example design chart
4. DIMENSIONAL ANALYSIS
It is understandable, that design charts such as these above cannot cover all possible soil profiles and
pile dimensions. Therefore, a dimensional analysis has been attempted, refering to concrete piles
embedded only in very stiff soil, and considering only the flow mode type of failure. This mode
creates the least damaging effect of the soil movement on the pile. Poulos (1999) suggests that efforts
should be made to promote this mode of behavior. The embedment length is assumed to be equal to
the unstable soil layer depth thus providing adequate fixity conditions.
According to -theorem (Langhaar 1951; Barenblatt 1996), the terms involved in the calculation of
the pile ultimate load may be combined to form 3 independent dimensionless variables. For the case
of non-cohesive soil, this study adopts the following correlation among independent variables:
(1)
531
The first term in this equation u

H
10
will be refered to as dimensionless pile displacement.

The term
may be considered as an index of the pile rigidity (assuming fixity at the

interface). Increasing the pile rigidity reduces pile displacements. The term 3

represents the dimensionless soil-pile system ultimate resistance.
The term apparently represents the piles distance as a function of the pile diameter. As
increases, the dimensionless ultimate resistance term increases, i.e group interaction effects are
less pronounced enhancing the efficiency of each pile, as the pile spacing increases.
The term
is defined as the pile slenderness ratio and depends on the height of the soil mass
contained between the piles. As the ratio increases, the same amount of will be achieved at a
higher displacement.
Two dimensionless Charts in Figures 12(a) and 12(b) materialize Eq.(1) for soil-pile system
ultimate resistance as a function of the dimensionless displacement for cohesionless soils. The
curves are plotted for two slenderness ratios:

3.2 representing a rigid soil-pile system and

5 which represents a flexible soil-pile system.

Figure 12 (a). Results of Dimensional Analysis for the case of single piles spaced at 2D embedded in sandy soil
(pile fixity is assumed just below the interface).

0 2 4 6 8 10 12
0
1
2
3
4
5
6
7
8

S=2D
H
u
/ D = 5
H
u
/ D = 3.2
2
3 L D Kp
Rf
4
3
D
H u
u p
S = 2D
532

Figure 12b. Results of Dimensional Analysis for the case of single piles spaced at 4D embedded in sandy soil
(pile fixity is assumed just below the interface).
ACKNOWLEDGEMENT
The research presented in this paper was financially supported by the Secretariat for Research and
Technology of Greece, under the auspices of PENED Programme with Contract number 03ED278.

REFERENCES
Barenblatt, G. I.(1996). Scaling, self-similarity, and intermediate asymptotics, Cambridge
University Press, Cambridge, U.K.
Broms, B., (1964), Lateral resistance of piles in cohesionless soils, J. Soil Mech. Foundations Div.
ASCE, 90, pp. 123-156
EKOS, (2000), Greek Reinforced Concrete Code, Ministry of Public Works, Athens, Greece
Kourkoulis, R., Gelagoti, F., Anastasopoulos, I., Gazetas, G., (2009), An improved method for the
design of Slope Stabilising Piles Part A: Development and Validation, Proc. 3rd Greece Japan
Workshop on Seismic Design, Observation and Retrofit of Foundations, Santorini.
Langhaar, H. L. (1951). Dimensional analysis and theory of models,Wiley, New York.
Poulos H.G. (1999) "Design of slope stabilizing Piles", Slope Stability Engrg, Yagi, Yamagami and
Jiang, Balkema, Rotterdam
Wang WL, Yen BC., (1974), Soil arching in slopes, Journal of the Geotechnical Engineering
Division, ASCE;100(No. GT1), pp. 6178.
0 5 10 15
0
1
2
3
4
5
6
7
8
3 4

S=4D
2
3 L D Kp
Rf
4
3
D
H u
u p
S = 4D
H
u
/ D = 5
H
u
/ D = 3.2
533

1
Reprinted from Proceedings of Int. Conf. on Performance Based Design in Earthquake Geotechnical Engineering,
IS-Tokyo 2009, Tsukuba City, Japan, June 15-18, 2009
1 INTRODUCTION
The dynamic response of even the simplest type of retaining wall is a complex soil-structure in-
teraction problem. Wall movements and dynamic earth pressures depend on the response of the
soil underlying the wall, the response of the backfill, the inertial and flexural response of the
wall itself, and the nature of input motions. The problem of seismically induced lateral earth
pressures on retaining structures has received significant attention from researchers over the
years. The pioneering work was performed in Japan following the Great Kanto Earthquake of
1923 by Okabe (1926) and Mononobe & Matsuo (1929). The method proposed by these authors
and currently known as the Mononobe-Okabe (M-O) method is based on the Coulombs theory
of static earth pressures and is today, with its derivatives, the most commonly used approach to
determine seismically induced lateral earth pressures. Later studies provided design methods
mostly based on analytical solutions or experimental programs. While many studies have been
conducted on the subject of seismic earth pressures over the last eighty years, to date, there
seems to be no general agreement on a seismic design method for retaining structures or
whether seismic provisions should be applied at all.
Given the importance of the seismically induced lateral earth pressures problem in the design
of retaining structures in seismically active areas, an experimental study was undertaken aimed
at improving our understanding of the seismic response of U-shaped cantilever walls retaining
dry medium dense sand deposits. Herein we present a brief review of relevant existing studies
On seismic response of retaining structures
1

N. Sitar & L. Al Atik
University of California, Berkeley, CA
ABSTRACT: Recent changes in codes and improved understanding of strong ground motions
have led to increased demands in the seismic design of retaining structures. Methods for evalu-
ating the seismically induced lateral earth pressures gradually evolved from the seminal Japa-
nese work performed in the 1920s. The resulting design procedures suggest large dynamic
loads indicating that older retaining structures may be significantly under-designed. However
field evidence from recent major earthquakes fails to show any significant problems with the
performance of retaining structures designed for static earth pressures only. Results of a series
of centrifuge experiments performed by the authors indicate that seismically induced lateral
earth pressures are significantly less than those estimated using the most current design meth-
ods. Specifically, the data show that the earth pressure distribution remains roughly triangular,
increasing with depth, and the maximum dynamic moments on the retaining structure are to a
large extent caused by the moment of inertia of the structures themselves. Most importantly,
there is no evidence for the development of a failure wedge postulated in the M-O method of
analysis and, hence, the basis for the continued use of the M-O method has to be re-examined in
light of the preponderance of field evidence and experimental data.

534
on dynamic earth pressures including evidence from recent earthquakes highlighting the field
performance of retaining walls. Observations from recently performed centrifuge tests are then
presented to elucidate the factors controlling the seismic performance of cantilever retaining
structures.
2 BACKGROUND
Since the pioneering work of Mononobe & Matsuo (1929) and analytical work of Okabe
(1926), researchers have developed a variety of analytical and numerical models to predict the
dynamic behavior of retaining walls or performed various types of experiments to study the
mechanisms behind the development of seismic earth pressures on retaining walls.
2.1 The Mononobe Okabe Method and its Derivatives
The M-O method developed by Okabe (1926) and Mononobe & Matsuo (1929) is an extension
of Coulombs static earth pressure theory to include the inertial forces due to the horizontal and
vertical backfill accelerations. The M-O method was developed for dry cohesionless backfill re-
tained by a gravity wall and is mainly based on the following assumptions (Seed & Whitman
1970):
1 The wall yields sufficiently to produce minimum active pressure and the soil is assumed to
satisfy the Mohr-Coulomb failure criterion;
2 When the minimum active pressure is attained, a soil wedge behind the wall is at the point of
incipient failure, and the maximum shear strength is mobilized along the potential sliding
surface; and
3 The soil wedge behaves as a rigid body, and accelerations are constant throughout the mass.
The theoretical underpinnings of the M-O method were developed by Okabe (1926). To vali-
date the analytical method developed by Okabe (1926), Mononobe & Matsuo (1929) carried out
experiments on dry, relatively loose sand in a rigid shaking table container in order to measure
dynamic earth pressures on retaining walls. The Mononobe & Matsuo (1929) experimental con-
figuration is presented in Figure 1. The experiments consisted of rigid base sand boxes with two
vertical doors hinged at their base and hydraulic pressure gauges at their tops to measure the
horizontal pressure exerted on the walls. The modeled walls were of 1.2 and 1.8 m height. The
sand boxes were set on rollers and horizontal simple harmonic motion was imparted by means
of a winch driven by an electric motor. Mononobe & Matsuo (1929) obtained experimental re-
sults consistent with the Okabe (1926) theoretical solution and their proposed seismic earth
pressure theory became known as the M-O method.
While these experiments were very meticulous and pioneering in their scope, their applicability
is limited by the fact that 1-g shaking table experiments on frictional material cannot be simply
scaled to taller structures because of the stress dependency of the material properties. Conse-
quently, Mononobe and Matsuos results are strictly correct only for the tested geometry and
material, i.e. walls up to 1.8 m height with relatively loose granular backfill.

Figure 1. Mononobe & Matsuo (1929) experiments setup
535
Results from various later experimental programs aimed at determining dynamic earth pres-
sures on retaining walls have been similarly reported in the literature based on 1-g shaking table
experiments. Results of such experiments generally suggested that the M-O method predicts rea-
sonably well the total resultant thrust but that its point of application should be higher than one
third the height of the wall above its base. However, as with the Mononobe & Matsuo (1929)
experiments, the accuracy and usefulness of the 1-g shaking table experiments are limited due to
the inability to replicate in-situ stress conditions, especially for granular backfills. More impor-
tantly, the observed amplification of ground motion and the observed increase in earth pressure
upwards is the direct result of the physical layout of the geometry of the shaking table box and
properties of the sand and not necessarily representative of field response.
2.2 Analytical Methods
Analytical solutions for the dynamic earth pressures problem can be divided into three broad
categories depending on the magnitude of the anticipated wall deflection. These categories in-
clude rigid-plastic, elastic, and elasto-plastic methods. Relatively large wall deflections are usu-
ally assumed for rigid plastic methods while very small deflections are assumed for elastic
methods. Elasto-plastic methods, appropriate for moderate wall deflections, are usually devel-
oped using finite element analysis.
Numerical modeling efforts have been applied to verify the seismic design methods in prac-
tice and to provide new insights into the problem. While elaborate finite element techniques and
constitutive models are available in the literature to obtain the soil pressure for design, simple
methods for quick prediction of the maximum soil pressure are rare. Moreover, while some of
the numerical studies reproduced experimental data quite successfully; independent predictions
of the performance of retaining walls are not available. Hence, the predictive capability of the
various approaches is not clear. It is important to note that analytical methods for computing
seismic earth pressures are usually based on idealized assumptions and simplifications that do
not necessarily represent the real retaining structures-backfill seismic behavior. Therefore, such
methods often result in overly conservative estimates of dynamic earth pressures.
2.3 Dynamic Centrifuge Testing
The basic principle behind centrifuge testing in geotechnical engineering is to create a stress
field in a model that simulates prototype conditions. This allows the investigation of phenomena
that otherwise would be possible only on full-scale prototypes. The major advantage of dynamic
centrifuge modeling is that scaling is relatively straight forward and correct strength and stiff-
ness can be readily reproduced for a variety of soils. Thus, in granular soils, for a reduced scale
model with dimensions 1/N of the prototype and a gravitational acceleration during spinning
that is N times the acceleration of gravity, the soil in the model will have same strength, stiff-
ness, stress, and strain as the prototype. Thorough discussions of centrifuge scaling laws are
given by Kutter (1995).
Dynamic centrifuge tests on model retaining walls with dry and saturated cohesionless back-
fills have been performed by Ortiz (1983), Bolton & Steedman (1985), Zeng (1990), Steedman
& Zeng (1991), Stadler (1996), and Dewoolkar et al. (2001). The majority of these dynamic
centrifuge experiments used sinusoidal input motions and pressure cells to measure earth pres-
sures on the walls. While most of these researchers observed a general agreement between the
maximum measured forces and the M-O predictions, the point of application of the dynamic
thrust remained uncertain. Moreover, Stadler (1996) observed that the incremental dynamic lat-
eral earth pressure profile ranges between triangular and rectangular and suggested that using a
reduced acceleration coefficient of 2070% of the original magnitude with the M-O method
provides good agreement with the measured forces.
Nakamura (2006) performed a series of dynamic centrifuge experiments to study the seismic
behavior of gravity retaining walls and investigate the accuracy of the M-O assumptions. His
study presented invaluable insights into the seismic behavior of the gravity wall-backfill system.
The configuration of the Nakamura (2006) centrifuge model is presented in Figure 2. Nakamura
(2006) studied the displacement, acceleration and earth pressures responses in order to under-
536
stand the seismic behavior of the wall/backfill system. His conclusions can be summarized as
follows:
1 Contrary to the M-O rigid wedge assumption, the part of the backfill that follows the dis-
placement of the retaining wall deforms plastically while sliding down;
2 While the M-O theory assumes that no phase difference occurs between the motion of the re-
taining wall and backfill, Nakamura (2006) experimentally observed that the acceleration is
transmitted instantaneously through the retaining wall and then transmitted into the backfill;
and
3 The M-O theory assumes that seismic earth pressures increase when the inertia force acts in
the active direction on the wall/backfill system. In reality, dynamic earth pressures and iner-
tia forces are not in phase. Dynamic earth pressure increment is around zero when the inertia
force is maximum and vice versa.
3 FIELD PERFORMANCE
The performance of retaining structures and basement walls during earthquakes greatly depends
on the presence of liquefaction-prone loose cohensionless backfills. Case histories from recent
major earthquakes show that retaining structures supporting loose, saturated, liquefiable, cohe-
sionless soils are quite vulnerable to strong seismic shaking. On the other hand, flexible retain-
ing walls supporting dry cohesionless sands or cohesive, clayey soils have performed particu-
larly well during earthquakes. It is important to note that many of these retaining structures were
not designed for seismic loading and others were designed for base accelerations not more than
20% of the peak accelerations that they actually experienced during the earthquake.
Clough & Fragaszy (1977) investigated the seismic performance of open channel floodway
structures in the Greater Los Angeles area during the 1971 San Fernando Earthquake. The
floodway structures studied consisted of open U-shaped channels with wall tops set flush to the
ground surface as shown in Figure 3. The backfill soil consisted of dry medium-dense sand with
an estimated friction angle of 35. The structures were designed for a conventional Rankine
static triangular earth pressure distribution, and no seismic provisions were applied in the de-
sign. The cantilever walls were damaged during the earthquake, with the typical mode of failure
as shown in Figure 3.
Clough & Fragaszy (1977) performed pseudo-static analyses and shear wave propagation
studies, and concluded that conventional factors of safety used in design of retaining structures
for static loadings provide a substantial strength reserve to resist seismic loadings. Peak accel-
erations of up to 0.5 g were sustained by the floodways with no damage even though no seismic
loads were explicitly considered in the design. The relationship between wall damage and
ground acceleration obtained by Clough & Fragaszy (1977) is shown in Figure 4.
537
During the 1994 magnitude 6.7 Northridge Earthquake, numerous temporary anchored
walls were subjected to acceleration levels in excess of 0.2 g and in some cases as large as 0.6 g.
Lew et al. (1995) described four such prestressed-anchored piled walls in the greater Los Ange-
les area with excavation depths of 15 to 25 m and supporting relatively stiff soils. The authors
reported that the measured deflections of these walls did not exceed 1 cm and that no significant
damage was observed.
During the 1995 magnitude 7 Kobe Earthquake in Japan, a wide variety of retaining struc-
tures most of them located along railway lines were put to test. Gravity-type retaining walls
such as masonry, unreinforced concrete and leaning type were heavily damaged. On the other
hand, reinforced-concrete walls experienced only limited damage. Koseki et al. (1998) pre-
sented preliminary evaluations of the internal and external stability of several damaged retaining
walls during the Kobe earthquake. The aim of their study was to improve the current design
procedures that are mostly based on the M-O theory. Koseki et al. (1998) concluded that a hori-
zontal acceleration coefficient based on a reduced value of the measured peak horizontal accel-
eration (60 to 100% of peak ground acceleration) is appropriate for use with the M-O method.
During the 1999 magnitude 7.6 Chi-Chi Earthquake in Taiwan, flexible reinforced-concrete
walls and reinforced-soil retaining walls performed relatively well. Similarly, Gazetas et al.
(2004) reported that during the 1999 magnitude 5.9 Athens Earthquake several metro stations
were being constructed. Although the retaining structure of the Kerameikos metro station was
not designed for seismic shaking, it was able to withstand nearly 0.5 g of peak ground accelera-
tion during the earthquake with no visible damage. Maximum wall displacements were esti-
mated to have been on the order of few centimeters.

Figure 2. Nakamura (2006) centrifuge model, horizontal shaking direction, dimensions in mm

Figure 3. Section through open channel floodway and typical mode of failure due to earthquake shaking
(Clough & Fragaszy 1977)
538

Figure 4. Relationship between channel damage and peak accelerations (Clough & Fragaszy 1977).

Most recently, observations of the seismic response of retaining structures during the great 5-
12 Sichuan Earthquake showed excellent performance of all types of retaining structures. Fig-
ure 5 shows a simple cobble and mortar retaining wall for an unfinished bridge abutment in
Hanwang, which experienced very significant shaking that caused extensive damage to other
types of structures in the vicinity. This type of retaining structure is extensively used through-
out the region and, as far as it could be ascertained, none of the structures were designed for the
severity of shaking that they experienced. Nevertheless, no evidence of significant damage was
observed during post-earthquake reconnaissance.

Figure 5. Unfinished cobble and concrete gravity retaining structure in Hanwang, Sichuan Province,
China, following the 5-12 Sichuan Earthquake

Overall, the case histories show that retaining structures perform quite well under seismic
loading, even if they were not specifically designed to handle dynamic loads. These observa-
tions clearly run counter to the accepted theories of seismically induced earth pressures.
4 OBSERVATIONS FROM THE DYNAMIC CENTRIFUGE EXPERIMENTS
4.1 Model Configuration
In our study we performed a series of centrifuge experiments on the 400g-ton dynamic centri-
fuge at the Center for Geotechnical Modeling at the University of California, Davis. The centri-
539
fuge has a radius of 9.1m, a maximum payload of 4,500Kg, and an available bucket area of 4m
2
.
The shaking table has a maximum payload mass of 2,700Kg and a maximum centrifugal accel-
eration of 80g. Additional technical specifications for the centrifuge and the shaking table are
available in the literature (Kutter et al., 1994, Kutter, 1995). The centrifugal acceleration used in
this experiment was 36g. All test results are presented in terms of prototype units unless other-
wise stated.
The models for these experiments were constructed in a rectangular flexible shear beam con-
tainer with internal dimensions of 1.65m long by 0.79m wide by approximately 0.58m deep.
The model container is designed such that its natural frequency is less than the initial natural
frequency of the soil in order to minimize boundary effects.
In prototype scale, the models consist of approximately 12.5m of dry medium dense sand
overlaid by two U-shaped retaining wall structures, stiff and flexible, spanning the width of the
container. Both structures support dry medium dense sand backfill. The first centrifuge experi-
ment was performed on a two-layer sand model with sand backfill and foundation having rela-
tive density of 61% and 73%, respectively. The second centrifuge experiment was performed
on a uniform density sand model with relative density of 72%. Dry pluviation was used to place
the sand in different layers underneath and behind the structures. Lead was added to the struc-
tures in small pieces of 1 in
2
each in order to match the masses of the prototype reinforced con-
crete structures. Figure 6 presents the configuration of the second centrifuge model.
4.2 Instrumentation
The centrifuge models were densely instrumented in order to collect accurate and reliable meas-
urements of accelerations, displacements, shear wave velocities, strains, bending moments and
earth pressures. Horizontal and vertical accelerations in the soil and on the structures were
measured using miniature ICP and MEMS accelerometers. Soil settlement and structures de-
flection and settlement were measured at different locations using a combination of spring
loaded LVDTs and linear potentiometers. Shear wave velocities in the soil underneath and be-
hind the structures were measured using piezo-ceramic bender elements and mini-shear air
hammers. The locations of accelerometers, bender elements, air hammers, and displacement
transducers for the second centrifuge experiment are shown in Figure 6.
Accurate measurement of lateral earth pressure distribution was the major goal of this study.
Lateral stress measurements in laboratory experiments are usually made using pressure cells.
Such measurements are not always reliable due to possible cell/soil reaction, the relative stiff-
ness of the cell with respect to surrounding soil and arching effects. Therefore, three different
sets of independent instruments were used in this study. The lateral earth pressures were directly
measured in the two centrifuge experiments using flexible tactile pressure Flexiforce sensors
manufactured by Tekscan. Lateral earth pressures on the south stiff and north flexible walls
were also calculated by double differentiating the bending moments measured by the strain
gages mounted on the model walls. Finally, direct measurements of the total bending moments
at the bases of the south stiff and north flexible walls were made using force-sensing bolts at the
wall-foundation joints.
4.3 Shaking Events
Multiple shaking events covering a wide range of predominant periods and peak ground accel-
erations were applied to each model in flight at a centrifugal acceleration of 36g. The shaking
was applied parallel to the long sides of the container. The shaking events consisted of step
waves, ground motions recorded at the Santa Cruz (SC) and the Saratoga West Valley College
(WVC) stations during the 1989 Loma Prieta Earthquake, ground motions recorded at the Port
Island (PI) and Takatori (TAK) stations during the 1995 Kobe Earthquake, and ground motions
recorded at the Yarmica (YPT) station during the 1999 Kocaeli Earthquake. The detailed de-
scription of the input ground motions can be found in Al Atik & Sitar (2007).
540
Figure 6. Model configuration for the second centrifuge experiment in profile view
4.4 Observations on the Seismic Response of the Retaining Wall-Backfill Systems
Data collected during the centrifuge experiments were used to study the seismic behavior of the
wall-backfill systems and to evaluate the hypothetical assumptions of the M-O theory. While
the M-O theory assumes that the seismic response occurs simultaneously and uniformly in the
backfill and the retaining walls, experimental results showed that the inertial force does not in
fact occur at the same time in the backfill and the walls. Moreover, accelerations are not uni-
form in the backfill or on the walls.
The M-O theory further assumes that dynamic earth pressures and inertial forces simultane-
ously take their maximums and stability analyses of retaining walls are conducted for such case.
Comparisons of the dynamic wall moments, dynamic earth pressures and inertial forces acting
on the walls suggest that when the inertial force is at its local maximum, the dynamic wall mo-
ment (due to dynamic earth pressures and wall inertia) is at its local maximum too but the dy-
namic earth pressure increment is at its local minimum or around zero. On the other hand, when
the dynamic earth pressure is at its local maximum, the inertial force and dynamic wall moment
reach their local minimum values or zero. Figure 7 presents a close-up comparison of the dy-
namic wall moments (due to dynamic earth pressure and wall inertia) interpreted from a strain
gage measurements to the dynamic earth pressures recorded by a Flexiforce sensor on the south
stiff and north flexible walls during a Loma Prieta, Santa Cruz shaking event from the second
centrifuge experiment. Figure 7 illustrates the out of phase relation between dynamic earth pres-
sures and dynamic wall moments on the stiff and the flexible walls.

Figure 7. Comparison of dynamic wall moments and dynamic earth pressures on the south stiff and north
flexible walls

541
4.5 Total Earth Pressure Distribution and Moment Generation
Figure 8 shows the maximum total earth pressure profiles recorded by the Flexiforce sensors
and interpreted from the strain gage measurements from a centrifuge experiment on a stiff and a
more flexible wall for the Loma Prieta-SC-2 and KocaeliYPT060-3 shaking events. Figure 8
also shows the total pressure distributions estimated by the M-O method using the measured
peak ground accelerations at the top of the soil in the free field. As shown in Figure 8, results
from the centrifuge experiments consistently demonstrate that the maximum dynamic earth
pressure increases with depth and can be reasonably approximated by a triangular distribution
analogous to that used to represent static earth pressures (Al Atik, 2008, Al Atik & Sitar,
2008a). The magnitude of seismic earth pressures depends on the magnitude and intensity of
shaking, the density of the backfill soil, and the flexibility of the retaining walls. As can be seen
from these results, the traditional M-O and the Seed & Whitman (1970) methods currently used
in practice provide overly conservative estimates of the maximum induced seismic earth pres-
sures (Al Atik, 2008, Al Atik & Sitar, 2008b).
While it is traditional in geotechnical practice to work in terms of earth pressures, it is the
moment distribution and the moment at the base of the wall that is of paramount interest from
structural design point of view. An important contribution to the overall dynamic wall moments
is the mass of the wall itself. While Richards & Elms (1979 and 1980) make a strong case for
the consideration of the inertial forces due to the mass of the retaining structure in the design of
gravity walls, cantilever walls have not received similar attention. Results from the centrifuge
experiments presented here show that the wall inertial moments contribution to the overall dy-
namic wall moments is substantial and should be accounted for separately. Moreover, wall iner-
tial moments are generally in phase with dynamic wall moments. This suggests that dynamic
wall moments are largely influenced by the inertia of the wall itself.
As shown in Figure 7, the moments generated by the earth pressure on the wall are out of
phase with the dynamic wall moments. As a result, the current trend to design retaining walls
for maximum dynamic earth pressures and maximum wall inertia is overly conservative and
does not reflect the true seismic performance of the backfill-wall systems. Since wall inertial
forces and dynamic earth pressures are not in phase, their cumulative effect results in reduced
overall moments acting on the walls.

Figure 8. Maximum total earth pressure profiles measured and estimated using the M-O method on the
south stiff and north flexible walls during Loma Prieta-SC-2 and Kocaeli-YPT060-3.

542
5 CONCLUSIONS
A review of the history of the development of methods for the estimation of seismically induced
lateral earth pressures suggests that the experimental basis for the current design methodology
based on the M-O method is not representative of the actual field response. This conclusion is
supported by observed excellent performance of various types of retaining structures in recent
earthquakes, which suggest that retaining structures underdesigned with respect to seismic
forces perform well under seismic loading with peak acceleration in excess of 0.5 g.
Similarly, the results from a series of centrifuge experiments designed to evaluate the magni-
tude and distribution of dynamic earth pressures on retaining walls show that the mechanism of
development of dynamic earth pressures is quite different from the simple limit equilibrium
wedge assumption inherent in the M-O method. Specifically, the earth pressure distribution in-
creases downward and the moments on the walls are to a large extent due to the contribution of
the moment of inertia of the walls themselves. In addition, the moments generated by wall iner-
tial forces and earth pressures are out of phase producing much lower overall moments than
would be predicted by the M-O method. Retaining walls should therefore be designed for these
reduced dynamic wall moments that include the combined effects of the inertial forces on the
wall and the backfill.
Finally, given the apparent shortcomings of the assumptions inherent in the current analysis
and design methods, there is a need for a complete and thorough re-examination of the entire
methodology and philosophy for estimation of seismic earth pressures in different soil condi-
tions and for different types of structures.
ACKNOWLEDGMENTS
The authors gratefully acknowledge the support and valuable input provided by Prof. Bruce
Kutter, Dr. Dan Wilson and all the staff at the Center for Geotechnical Modeling at the Univer-
sity of California, Davis. This research was supported by a grant from the San Francisco Bay
Area Rapid Transit (BART) and the Santa Clara Valley Transportation Authority (VTA) to the
Pacific Earthquake Engineering Research Center at UC Berkeley. The authors also received
valuable input from Mr. Ed Matsuda, Dr. Jose Vallenas at BART and Mr. James Chai at VTA.
Any opinions, findings, and conclusions or recommendations expressed in this paper are
those of the authors and do not necessarily reflect those of the funding agencies.
REFERENCES:
Al Atik, L. 2008. Experimental and analytical evaluation of seismic earth pressures on cantilever retain-
ing structures. PhD Dissertation, Department of Civil and Environmental Engineering, University of
California, Berkeley.
Al Atik, L. & Sitar, N. 2007. Development of improved procedures for seismic design of buried and par-
tially buried structures. PEER Report 2007/06, College of Engineering, University of California,
Berkeley.
Al Atik, L. & Sitar, N. 2008a. Dynamic centrifuge study of seismically induced lateral earth pressures on
retaining structures. Proc. of the 4
th
Decennial Geotechnical Earthquake Engineering and Soil Dy-
namics Conference.
Al Atik, L. & Sitar, N. 2008b. Development of improved procedures for seismic design of buried and
partially buried structures. PEER Report in preparation, College of Engineering, University of Cali-
fornia, Berkeley.
Bolton, M.D. & Steedman, R.S. 1985. The behavior of fixed cantilever walls subject to lateral loading.
Application of Centrifuge Modeling to Geotechnical Design, Craig (ed.). Rotterdam: Balkema.
Clough, G.W. & Fragaszy, R.F. 1977. A study of earth loadings on floodway retaining structures in the
1971 San Fernando Valley Earthquake. Proc. of the Sixth World Conference on Earthquake Engineer-
ing 3.
Dewoolkar, M.M., Ko, H. & Pak, R.Y.S. 2001. Seismic behavior of cantilever retaining walls with lique-
fiable backfills. J. of Geotechnical and Geoenvironmental Engineering, ASCE 127(5): 424-435.
543
Gazetas, G., Psarropoulos, P.N., Anastasopoulos, I. & Gerolymos, N. 2004. Seismic behavior of flexible
retaining systems subjected to short-duration moderately strong excitation. Soil Dynamics and Earth-
quake Engineering 24: 537-550.
Koseki, J., Tatsuoka, F., Munaf, Y., Tateyama, M. & Kojima, K. 1998. A Modified procedure to evaluate
active earth pressure at high seismic loads. Soils and Foundations, Special Issue on Geotechnical As-
pects of the January 17, 1996 Hyogoken-Nambu Earthquake, 2: 209-216.
Kutter, B.L., Idriss, I.M., Kohnke, T., Lakeland, J., Li, X.S., Sluis, W., Zeng, X., Tauscher, R.C., Goto,
Y. & Kubodera, I. 1994. Design of a large earthquake simulator at UC Davis. Centrifuge 94, Leung,
Lee, and Tan (eds.), 169-175. Rotterdam: Balkema.
Kutter, B.L. 1995. Recent advances in centrifuge modeling of seismic shaking. Proc., St. Louis, 2: 927-
941.
Lew, M., Simantob, E. & Hudson, M.E. 1995. Performance of shored earth retaining systems during the
January 17, 1994, Northridge Earthquake. Proc. of the Third International Conference on Recent Ad-
vances in Geotechnical Earthquake Engineering and Soil Dynamics, St Louis, Missouri, 3.
Mononobe, N. & Matsuo M. 1929. On the determination of earth pressures during earthquakes. Proc.
World Engineering Congress, 9: 179-187.
Nakamura, S. 2006. Reexamination of Mononobe-Okabe theory of gravity retaining walls using centri-
fuge model tests. Soils and Foundations 46 (2):135-146.
Okabe S. 1926. General theory of earth pressure. J. of the Japanese Society of Civil Engineers, Tokyo,
Japan 12 (1).
Ortiz, L.A., Scott, R.F. & Lee, J. 1983. Dynamic centrifuge testing of a cantilever retaining wall. Earth-
quake Engineering and Structural Dynamics 11: 251-268.
Richards, R., & Elms, D.G. 1979. Seismic behavior of gravity retaining walls. J. of the Geotechnical En-
gineering Division, ASCE, 105 (GT4): 44964.
Richards, R., & Elms, D.G. 1980. Seismic behavior of gravity retaining walls closure. J. of the Geo-
technical Engineering Division, ASCE, 106 (GT6): 737-738.
Seed, H.B. & Whitman, R.V. 1970. Design of earth retaining structures for dynamic loads. ASCE Spe-
cialty Conference, Lateral Stresses in the Ground and Design of Earth Retaining Structures, Cornell
Univ., Ithaca, New York: 103147.
Stadler, A.T. 1996. Dynamic centrifuge testing of cantilever retaining walls. PhD Dissertation, Univer-
sity of Colorado at Boulder.
Steedman, R.S. & Zeng, X. 1991. Centrifuge modeling of the effects of earthquakes on free cantilever
walls. Centrifuge91, Ko (ed.), Rotterdam: Balkema.
Zeng, X. 1990. Modeling behavior of quay walls in earthquakes. PhD Thesis, Cambridge University,
Cambridge, England.

544
1 THE NIKAWA LAND-SLIDE (1995)
1.1 Introduction
In addition to the devastation on all types of engineered structures, the Kobe earthquake trig-
gered nearly 400 landslides that also took place did not catch the attention of the casual observer
(Sassa et al. 1996). Most of them were of relatively small size, often associated with tensile
cracking, and of limited motionnot unexpectedly in view of the fact that the earthquake oc-
curred in the dry season. A conspicuous exception was the Nikawa rapid landslide one of
the most devastating landslides directly related to an earthquake (Figs 1 and 2). With a landslide
volume in the order of 110,000 m
3
(Sassa et al., 1996), moving in just a few seconds over a dis-
tance of more than 100 m, it destroyed 11 residential buildings causing 35 fatalities. In addition
of course to strong seismic shaking perhaps accentuated by topographic amplification (Kalou &
Gazetas, 2001), several deeper causes, such as sliding-surface liquefaction (Sassa, 1995) and
water-film generation (Kokusho, 2000), have been proposed to explain rapid runoff of the
slide.

The Nikawa (1995) and Higashi Takezawa (2004): Modelling
and Numerical Analysis

N. Gerolymos
ABSTRACT: Two earthquaketriggered violent landslides are studied. The Nikawa, activated
by the M
JMA
7.2, 1995 Kobe Earthquake, and the HigashiTakezawa, activated by the M
JMA
6.8,
2004 Niigata-ken Chuetsu earthquake, in Japan. Both landslides involved about 100 m dis-
placement of a large wedge from an originally rather mild slope. The surprisingly large and
rapid in nature runoff of the soil masses, motivated several researchers to interpret the sliding
process. There is still no consensus as to which were the actual causes of those two catastrophic
events. The goal of the paper is to study: (i) the landslide triggering and propagation, and (ii) the
mechanism of material softening inside the shear band responsible for the accelerating move-
ment of the two slides. To this, a model is utilised considering two mechanically coupled sub-
structures: (a) the accelerating deformable body of the slide, and (b) the rapidly deforming shear
band at the base of the slide. It combines features of an extended SavageHutter approach for
the sliding soil body, with (a) a MohrCoulomb failure criterion, (b) BoucWen hysteretic
stressstrain relationship, and (c) the Voellmys rheology for the deformation of the material
within the shear band, and exploits the concept of grain crushinginduced instability. The
method successfully interprets the studied landslides.
545

Fig. 1. The Nikawa landslide: aerial photos [Sassa et al, 1996]
1.2 Ground Shaking
The recorded Peak Ground Acceleration (PGA) exceeded 0.60 g at level ground very close to
Nikawa. The very small distance from the NE part of the causative fault can explain such a large
PGA. The combination of soil and topographic amplification could have played a major role, at
least in triggering the landslide. In such a case, basic freefield motion could have possibly been
amplified at the sliding mass. With such an excitation, the developing shear stresses could pos-
sibly lead to liquefaction of even marginally sensitive soil layers. However, even if such a trig-
gering-related mechanism were true the 140 m displacement could not be easily explained.

546

Horizontal distance (m)
D
e
p
t
h

(
m
)
80
60
40
20
20 40 60 80 100 120 140 160 180
Landslide mass
Landslide deposit
Sliding surface
(b)
(a)
D
e
p
t
h

(
m
)
80
60
40
20
20 40 60 80 100 120 140 160 180
Landslide mass
Landslide deposit
Sliding surface
(b)
(a) (a)

Fig 2. The Nikawa landslide : (a) plan view, and (b) cross section [Sassa et al, 1996]
1.3 Geometry, Geology, Groundwater and Soil Properties
The earthquake took place during a dry season, which followed a historicallydry 1994 sum-
mer. The limited amount of rainfall possibly played a major role in reducing the number of
landslides triggered by the earthquake. The most important landslides (such as Nikawa) were
associated with the socalled Osaka Group layers, that consisted of limnic and marine deposits
of sands and clays from Pliocene to Middle Pleistocene, with low permeability. Within these
fairly impermeable layers, pore water could have been preserved despite the dry season. There-
fore, while most landslides originated within un-saturated soil and hence were of moderate
magnitude engineers this was not the case with Nikawa. Engineers were surprised with the sig-
nificant distance and speed of the runoff, because: (a) the slope inclination barely exceeded 20
o
,
(b) the water table was not high (although there was evidence that it was above the sliding sur-
face for a significant length), (c) the soil along the sliding surface consisted of rather dense
coarse-grained sand to silty sand, material not readily susceptible to liquefaction. A typical grain
size distribution of this soil is plotted in Fig. 3.
547

0
0.2
0.4
0.6
0.8
1
0.001 0.01 0.1 1 10
particle diameter D (mm)
F
r
a
c
t
i
o
n

f
i
n
e
r

b
y

w
e
i
g
h
t

Fig 3. Grain size distribution for blue granitic sand found at the bottom of the Nikawa landslide mass
(Sassa et al. 1996). The initial potential for breakage B
p0
is equal to 0.72.

As depicted in the geotechnical cross-section of Fig. 4, the slope of the landslide mass did not
exceed 20
o
. The number of blows of the standard penetration test, N
SPT
, ranged from 10 (near
the surface) to 60 (refusal, in the Japanese scale). Secondary sedimentary layers and terrace lay-
ers were found to overly the Osaka group (granitic sand and clay). The bedrock granite was de-
tected at 25 m (borehole B9) to 35 m depth (borehole B6). The apparent residual friction angle
of the shear band material was measured to be
a
~ 8.5
(compared to the drained residual fric-

tion angle ~ 30
). Several witnesses asserted that water was flowing from the base of the land-
slide the very next day.
1.4 Possible Causes or Contributing Factors
The extent of the runoff along with its rapid nature, which left no time for response, led Sassa
et al. (1996) in developing a hypothesis which he called sliding-surface liquefaction. As de-
picted in Fig. 5, in the conventional liquefaction the strength loss is associated with pore pres-
sure buildup, due to the tendency of the soil to contract when subjected to shearing. Sliding sur-
face liquefaction is quite different: when the soil is subjected to shearing, and after a sliding
surface has been developed, the crushing of sand grains and the consequent increase in volume
of solids is the mechanism of pore pressure buildup, leading to a different type of liquefac-
tion. Sassa and coworkers developed a high-speed ring-shear apparatus to test soil specimens
from Nikawa, with shearing speeds in the order of 0.3 m/s. While in conventional liquefaction
the pore pressure buildup is rapid, these tests showed a gradual increase of pore pressure and a
subsequent drop of the apparent friction angle to about 8.5
o
, without any sign of liquefaction in
the sample. The grain crushing became evident from the grain size distribution along the shear
zone.
Some other alternative mechanisms can be invoked to explain the phenomenon. For instance,
the mechanism of gradual smoothing of the sliding surface was proposed by Kokusho [2000]
(not for Nikawa): it supposes that when a soil layer of significant thickness underneath the slid-
ing surface liquefies, and the soil directly on top is impermeable (both conditions might apply in
Nikawa), then the natural tendency of the liquefied layer to settle could produce a very slim
film of water, only a few centimeters or even milimiters in thickness. The consequence of the
appearance of this water film along the sliding surface could explain the extent of the runoff
(about 100 m). However, the nature of the soils does not support such a theory.

548

Terrace deposits Osaka group Initial Landslide mass
Landslide mass deposit Granite Secondary deposits
0 50 100 150 200
B6
B7
B8 B9
110
100
90
80
70
60
D
e
p
t
h

(
m
)
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
Initial Landslide mass
(before motion)
Landslide mass deposit
(after motion)
0 50 100 150 200
B6
B7
B8 B9
110
100
90
80
70
60
D
e
p
t
h

(
m
)
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
(before motion)
(after motion)
0 50 100 150 200
B6
B7
B8 B9
110
100
90
80
70
60
D
e
p
t
h

(
m
)
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
0 50 100 150 200
B6
B7
B8 B9
110
100
90
80
70
60
D
e
p
t
h

(
m
)
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
0 60
N
SPT
0 60 0 60
N
SPT
(before motion)
(after motion)

Fig 4. Geotechnical cross section, along with NSPT blows on boreholes that were conducted after the
slide [courtesy of Professor K. Sassa]
2 THE CASE STUDY OF HIGASHITAKEZAWA LANDSLIDE, JAPAN (2004)
2.1 Introduction
The devastating 2004 NiigataKen Chuetsu earthquake (M
w
6.8) triggered 374 landslides
more than 50 m wide (Sassa et al., 2005) 12 of which with volume larger than one million cubic
meters. Among these landslides, HigashiTakezawa was one of the largest. The landslide mass
filled a valley and stopped a river flow forming a large natural reservoir (Fig. 6). It is believed,
however, that the heavy rainfall during the last three days before the earthquake was a signifi-
cant contributor to the triggering of those landslides (Sassa, 2005; Tsukamoto and Ishihara,
2005).
549

(a) Conventional (mass) Liquefaction
Loose zone
Ground water path
Destruction of
the structure Normal Stress
S
h
e
a
r

S
t
r
e
s
s
(b) Sliding Surface Liquefaction
Before shear After shear
Normal Stress
S
h
e
a
r

S
t
r
e
s
s
Loose zone
Ground water path
Destruction of
S
h
e
a
r

S
t
r
e
s
s
Loose zone
Ground water path
Destruction of
S
h
e
a
r

S
t
r
e
s
s
Normal Stress
S
h
e
a
r

S
t
r
e
s
s
Normal Stress
S
h
e
a
r

S
t
r
e
s
s

Fig 5. (a) Conventional (mass) liquefaction, compared to (b) sliding surface liquefaction, in terms of the
mechanism and the associated stress paths. (R.F.L. : Residual Failure Line, P.F.L. : Peak Failure Line)
[Sassa, 1996].

The surprisingly large and rapid runoff of the soil mass motivated several researchers (Koku-
sho 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. It is pointed out, that 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).
In an effort to address the complex issue of triggering and postfailure travel distance of the
landslide, Tsukamoto and Ishihara (2005) performed a series of drained triaxial tests on partially
saturated sand samples taken from HigashiTakezawa. Motivated by the reasonable assumption
that rapid landslides would take place under conditions of little or practically no volume change,
they followed a test procedure which is as follows: First, the nonsaturated soil specimen is pre-
pared with the wet tamping method. Then, it is consolidated through the application of confin-
ing stress and is loaded axially following a straincontrolled procedure. The volume change of
the specimen is monitored during the phase of axial loading. Since the soil generally tends to
decrease its volume during the early stages of shearing, the cell pressure is appropriately re-
duced to keep the volume of the specimen unchanged. This reduction is maintained until the cell
pressure becomes equal to zero, and the test is continued until the axial strain reaches 10%.

550

Fig 6. The HigashiTakezawa landslide (source: Sassa et al, 2005)

Three tests were performed on specimens consolidated to a confining pressure
3
= 98 kPa
with relative density D
r
between 30 % to 35% and water content w = 10, 20 and 49 % (full satu-
ration, S
r
= 100%). The measured stress paths were found to be similar to those of conventional
(mass) liquefaction. However, the friction angles were calculated to be as large as 65
o
to 75
o

(with larger values corresponding to higher water contents), do not justify such a mechanism of
porepressure generation.
Figure 7 depicts the residual strength curve with respect to the water content. Interestingly the
residual strength, defined as half of the deviatoric stress at large strains, is a decreasing function
of water content. Indeed, there is a threshold limit of the water content, at w 25 %, below
which no reduction of the residual strength is observed. For water content between 25% and
49% (full saturation), the residual strength reduces sharply with increasing water content. Fi-
nally, for water content greater than that corresponding to full saturation, the residual strength
remains constant.
The drained triaxial test under conditions of no volume change provides valuable information
about the residual strength of unsaturated sand, which can be readily applied for the analysis of
rapid landslide. Nevertheless, it has the following drawback: The residual strength is not the
physical result of a porepressure generation mechanism. It is derived artificially through an
imposed reduction of the cell pressure under the condition of zero volumetric strain rate.
After a detailed field survey of the head scarp of the HigashiTakezawa landslide, Sassa et al.
(2005) concluded that the sliding surface could have been formed within either the weathered
(due to the existence of groundwater flow) top part of the outcropped siltstone layer, or the bot-
tom of the overlain sand layer which was probably a part of previously moved landslide mass.
Ring shear tests conducted by Sassa et al. (2005) on soil specimens taken from the sliding sur-
face revealed that the residual friction angle of the silt was by far smaller than that of the sand,
and close to the inclination angle of the slip plane. Neither of them could explain the rapid and
large displacement of the landslide. However, the sand was found to be susceptible to grain
crushing induced softening.

551
Water content w (%)
R
e
s
i
d
u
a
l

s
t
r
e
n
g
t
h

(
k
P
a
)
Transitional zone
Full saturation
w = 20 %
w = 49 %
20 40 60 80 0
0
10
20
30
Water content w (%)
R
e
s
i
d
u
a
l

s
t
r
e
n
g
t
h

(
k
P
a
)
Transitional zone
Full saturation
w = 20 %
w = 49 %
20 40 60 80 0
0
10
20
30

Fig 7. Idealized plot of residual strength of the HigashiTakezawa sand against water content (source:
Tsukamoto and Ishihara, 2005)
2.2 Earthquake Characteristics and Ground Shaking
At 17:56 on October 2004, the NiigataKen Chuetsu earthquake with a magnitude M
JMA
=
6.8, struck central Niigata-ken (Chuetsu area) and seriously damaged infrastructures of hilly and
mountainous areas including Kawagushi town, Ojiya city, Nagaoka city, and their environs. The
earthquake was an epicentral thrustfault earthquake with a hypocenter of about 10 Km depth
(Toyota et al., 2006). The main shock was followed by three major aftershocks with M
JMA
equal
or larger than 6.0 that occurred within one hour after. In addition, several other subsequent af-
tershocks continued for about two weeks (Koseki et al., 2006).
The actual seismic excitation of the landslide cannot be known with accuracy: there was no
nearby record. Certainly, parameters such as the geology, topography, site conditions and dis-
tance from the fault played a role. The nearest observation station NIG019 at Ojiya, around 10
km west of the HigashiTakezawa landslide and WNW 7 km from the epicenter of the main
shock, recorded acceleration time histories characterized by peak values of PGA = 1.3 g (Sassa
et al. 2005).
2.3 Geometry and Kinematics
The main body of the landslide is indicated in the plan of Fig. 8 deduced from an air borne la-
ser scanning survey carried out three days after the earthquake (Sassa et al., 2005). A cross
section of the landslide is also depicted in Fig. 8. The gentle slope inclination before the failure
suggested that the landslide was probably a reactivation of an earlier one. The landslide in-
volved a soil volume of about 1,200,000 m
3
(Kokusho and Ishizawa, 2005). The maximum di-
mensions in plan were about 300 m width and 250 m length, and the maximum thickness was
about 40 m (Kokusho and Ishizawa, 2005, Sassa et al., 2005). The landslide mass moved rap-
idly about 100 m, and hit the opposite bank of Imokawa river. One 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 by
the aforementioned authors to be approximately 20
o
.

552
E
a
rlie
r la
n
d
s
lid
e
S
ilt
s
t
o
n
e

la
y
e
r
Sand layer
Before failure
After failure
School
Imokawa River
0 100 200 300 400 500 600 700
0
100
300
E
l
e
v
a
t
i
o
n

(
m
)
A
A
200
E
a
rlie
r la
n
d
s
lid
e
S
ilt
s
t
o
n
e

la
y
e
r
Sand layer
Before failure
After failure
School
Imokawa River
0 100 200 300 400 500 600 700
0
100
300
E
l
e
v
a
t
i
o
n

(
m
)
A
A
A
A
200

Fig 8. The HigashiTakezawa landslide: (a) plan view, and (b) cross section (based on information by
Sassa et al., 2005)
2.4 Geology, Groundwater and Soil Properties
A schematic geological section of the landslide area is shown in Fig 8. The subsoil is a Neo-
gene 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. The groundwater flow over the siltstone layer, lead Sassa et al. (2005) to as-
sume the existence of a thin silt layer between the sandstone and the siltstone, due to the un-
avoidable 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 nearby Terrano landslide located also near river Immokawa. The Terrano landslide,
triggered by the NiigataKen Chuetsu earthquake, had the same subsoil and groundwater condi-
tions. The silt encountered at the head scarp of the Terrano landslide was weathered and soft.
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 occurred
for three-four days before the earthquake, but heavy rains had occurred the weeks before.

553

0
0.2
0.4
0.6
0.8
1
0.001 0.01 0.1 1 10
particle diameter D (mm)
P
e
r
c
e
n
t

f
i
n
e
r

b
y

w
e
i
g
h
t

Fig 9. Grain size distribution of the HigashiTakezawa sand (black line) and Terrano silt (gray line), after
Sassa et al. (2005)

The grain size distribution of the sand involved in the sliding surface of the Higashi
Takezawa landslide is illustrated in Fig. 9, along with that of the Terrano silt which is consid-
ered to be representative of the HigashiTakezawa silt. The strength properties of the soils un-
der consideration were obtained from consolidateddrained and undrained high speed ring shear
tests (Sassa et al., 2005). The undrained friction angle of the sand was found to be 37
o
, while the
residual friction angle of the Terrano silt was 24
o
. However, one peculiar aspect of the Higashi
Takezawa sand is its mechanical instability due to grain crushing. Indeed, the cyclic loading
test, resulting in an apparent residual friction angle of merely 3.3
o
, strongly suggesting that the
HigashiTakezawa sand is susceptible to grain crushinginduced instability.
3 THE MODEL : EQUATIONS AND PARAMETERS
3.1 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 10). 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 uniformlydistributed (depth
integrated) velocity along each column. At the base of the sliding mass we assume a shear band
of infinite length and of zero thickness d
b
= 0 (Fig 10), subjected to an acceleration time history
uniformly imposed along the entire length of the sliding base. 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 an Eulerian description of
motion, the following system of partial differential equations are obtained:
0
h h
h
t x x
u
u
c c c
+ + =
c c c
(1)
and

554

x
z
h(x,t)
(x,t)
(x)
x
o
dx
x
x
x
c
c
+
o
o
dx
g h cos
T
r
+ T
f
T
d
bedrock
g
(t)hdx
g
(t)
x
z
h(x,t)
(x,t)
(x)
x
o
dx
x
x
x
c
c
+
o
o
dx
g h cos
T
r
+ T
f
T
d
bedrock
g
(t)hdx
g
(t)
d
b
0
(t)
x
T
d
(t) T
r
(t) T
f
(t)
Shear band
'
n
(t)
z
d
b
0
(t)
x
T
d
(t) T
r
(t) T
f
(t)
Shear band
'
n
(t)
z
'
n
(t)
'
n
(t)

Fig 10. 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.

( )
g x
d r f
h
T T T h
x t x t
u o
u u
u
c c | | c c
+ = + +
|
c c c c
\ .
(2)
h is the depth in the z direction normal to the bed (Fig 10), is the depthaveraged velocity in
the x direction parallel to the base of the landslide,
g
is the seismic velocity imposed at the base
of the landslide parallel to the dip direction of the sliding surface. By inspecting Eq (2), it is in-
ferred that spatial variability of the seismic motion has been neglected, and the same accelera-
tion time history is imposed instantaneously along the entire length of the sliding surface. The
term on the righthand side of Eq (2) is an approximation of the inertial loading that acts on the
555
landslide mass. It is inferred that soil amplification due to the passage of seismic waves through
the landslide mass, is not considered by Eq (2). T
d
is the gravitational driving force acting on the
landslide mass
sin
d
T g h u = (3)
in which g is the gravitational acceleration and is the local base slope. T
r
is the resisting force
due to Coulomb friction at the bed influenced by bed curvature (Hutter and Koch, 1991; Gray et
al., 1999; Iverson and Delinger, 2001)
2
1
cos
r m
T
g x
u u
t
u
| | c
~ +
|
c
\ .
(4)
in which
m
is the cyclic shear resistance mobilized along the shear band. A detailed description
of this term will be provided below. T
f
is the turbulent resisting force which will also be dis-
cussed in the sequel.
In Eq (2),
x
o is the average (along the depth of the sliding mass) longitudinal normal stress
primarily due to the overburden load, and secondarily due to elongation or compression of the
soil mass in the x direction. This stress is assumed to be a combination of a lithostatic (depth
dependent) term and a strainrate dependent term
1 1
2 2
x s p d p
h g h g
x
u
o q q
c
=
c
(5)
in which g
p
is the corrected acceleration of gravity approximated by
2
cos
p
g g
x
u
u u
c
~ +
c
(6)
s
and
d
are the lithostatic (or the lateral earth pressure coefficient at rest) and viscous coeffi-
cients, respectively.
s
is dimensionless, while
d
has the dimensions of time. Eq (5) can be re-
written in the following equivalent form
1
2
x dyn p
K g h o = (7)
in which K
dyn
is termed dynamic lateral earth pressure coefficient determined as
x
K
d s dyn
c
c
=
u
q q (8)
The first term on the righthand side of Eq (5) governs the rheological behaviour of the slid-
ing mass. A typical range of values for the lithostatic coefficient
s
is between 0 and 1, with
larger values corresponding to deformation of fluidlike mixtures, and smaller ones to solid
like mixtures. Representative values of
s
for soils are between 0.5 and 0.65 (Sassa et al, 2006).
For the special case of
s
=
d
= 0, the sliding mass behaves as a rigid body and Eq (2) reduces
to the well known Newmark (1963) sliding block model. In the conventional shallow water
equations for flow of grainfluid mixture (e.g. Iverson and Delinger, 2001), K
dyn
is equal to
s

which in turn is a rigid function of the sign of the strain rate. This means that the longitudinal
effective normal stress moves rigidly from the active to the passive state and vice versa, without
the intervention of a transitional regime. Such a transitional behaviour is provided by the vis-
cous term on the righthand side of Eq (8).
The suggested term, plays also the role of a damping factor for the surface waves emanating
from the frontal edge of the landslide and propagating along its crest. Such waves could be gen-
erated due to (i) elongationcompression of the soil mass in the direction of flow; (ii) abrupt
variations of bed inclination; (iii) impact of the avalanche front (shock wave) on a stumbling
block or a retaining structure.
556
3.2 Equations for Frictional Behaviour of the Shear Band
A versatile onedimensional constitutive model is utilised to describe the shear stress
displacement 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 pro-
posal by Bouc (1971) and Wen (1976), the model was recently extended by Gerolymos and Ga-
zetas (2005) and applied to cyclic response of soils and earthquaketriggered rapid landslides
(Gerolymos et al., 2007; Gerolymos and Gazetas, 2007a; 2007b). It is used herein in conjunc-
tion with a MohrCoulomb friction law and Terzaghis effective stress principle.
The mobilized shear stress inside the shear band is expressed as:
m m
y
t x t
t t ,
u t
c c c
+ =
c c c
(9)
where
y
is the ultimate shear strength, which is a function of time. Notice that Eq (9) allows for
material convection along the shear band.The parameter is a hysteretic dimensionless quantity,
controlling the nonlinear response of the soil. It is governed by the following differential equa-
tion:
( ) ( )
{ }
1
1 1 sgn
n
y
d
b b
du u
,
, u, ( = +

(10)
in which u
y
, n and b, are parameters that control the shape of the shear stress versus displace-
ment curve. The parameter
y
is defined as:
( )
0 y n
p t o' = (11)
in which the friction coefficient, , is expressed in terms of the Coulomb friction angle
'
of
the soil in direct shear,
0 n
o'
is the initial effective normal stress, and p is the excess porewater
pressure, generated due to particle breakage.
The turbulent resisting force at the base of the slide is represented by a constitutive law of the
quadratic Chezy type
2
sgn( )
f
T
T g
u
u
= (12)
in which
is the turbulent (or hydraulic) friction coefficient relating the viscous shear stress
with the landslide velocity with dimensions of acceleration. Eq (12) in conjunction with the
Coulomb frictional law given by Eq (11) constitutes the Voellmys rheology (Voellmy, 1955).
What is new in this paper is that the pressuredependent term has been enhanced by the hyster-
etic parameter , given by Eqs (9) and (10), to account for non-linear cyclic response of the ma-
terial within the shear band. The Voellmys rheology has been frequently applied in the analysis
of debris flows and landslides (Hungr, 1995, Evans et al, 2001, Medina et al, 2008, Armento et
al, 2008) with satisfactory results.
Exploiting numerical results from systematic backanalyses of real landslides, Medina et al
(2008) and Armento et al (2008) showed that a typical range of values for
is between 25 m/s
2

and 400 m/s
2
. In general
varies significantly from avalanche to avalanche (Hutter and Greve,

1993) and its contribution to granular 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 viscous-type at large velocities.
Summarizing, there is no consensus as to what type of viscous frictional law is the most ap-
propriate for granular avalanche flow. Back-calculating the hydraulic friction coefficient by fit-
ting experimental results and/or field observations in the context of an assumed viscous consti-
tutive law, seems to be the most realistic approach to the problem.
557
3.3 Equations for Grain CrushingInduced PoreWater Pressure
The mechanism of porewater pressure generation due to particle breakage is assumed to be
governed by the following equation (Gerolymos and Gazetas, 2007a; 2007b):
( )
0
p
v p n
B
p p p
c B
t x z z t
u o
c
c c c c | |
' + =
|
c c c c c
\ .
(13)
in which B
p
is the current value of the breakage potential; c
v
and are the coefficients of con-
solidation and porepressurebreakage, respectively. Note that c
v
is a function of B
p
. In fact, c
v

decreases with decreasing particle size and thus with particle crushing evolution.
This expression is being simplified in the limit of undrained loading conditions, which is a
reasonable assumption when the shear band is deformed at a large velocity (rapid landslide). Pa-
rameter controls the ultimate value of the porewater pressure.

0.1 0.01 1
10
100
25
50
75
100
0
Particle diameter D (mm)
P
e
r
c
e
n
t

f
i
n
e
r

b
y

w
e
i
g
h
t
(
%
)
0.074
B
e
f
o
r
e

l
o
a
d
i
n
g
Silt
Evolution of grain size
distribution with loading
A
f
t
e
r

l
o
a
d
i
n
g
B
p0
C A
B
B
C
(ABC) = B
p0
(ABC) = B
pl
0.1 0.01 1
10
100
25
50
75
100
0
P
e
r
c
e
n
t

f
i
n
e
r

b
y

w
e
i
g
h
t
(
%
)
0.074
B
e
f
o
r
e

l
o
a
d
i
n
g
Silt
A
f
t
e
r

l
o
a
d
i
n
g
B
p0
0.1 0.01 1
10
100
25
50
75
100
0
P
e
r
c
e
n
t

f
i
n
e
r

b
y

w
e
i
g
h
t
(
%
)
0.074
B
e
f
o
r
e

l
o
a
d
i
n
g
Silt
A
f
t
e
r

l
o
a
d
i
n
g
B
p0
C A
B
B
C
(ABC) = B
p0
(ABC) = B
pl

Fig 11. Definition of the initial breakage potential B
p0
, after Hardin (1985)
3.4 Equations for Grain Crushing
The breakage potential B
p
is a measure of the evolution of the particle size distribution curve
with loading; it is indicative of the amount of grain crushing. A key assumption, motivated by
scant experimental evidence is that the evolution of B
p
with time is governed by the following
equation (Gerolymos and Gazetas, 2007a):
( )
p p
pl p
B B
B B
t x
u
c c
+ =
c c
(14)
in which is the coefficient of grain crushing; B
pl
is the final (after loading) breakage poten-
tial relating to B
p0
, the initial (before loading) value of the potential as defined by Hardin
(1985), according to:
0
1
b
p
pl n
B
B
S
=
+
(15)
The definition of B
p0
is schematically illustrated in Fig. 11. n
b
is the crushing number (Har-
din, 1985), S is the stress loading factor which is a function of the mobilized shear stresss
m
, the
effective normal stress
n
o ' , the crushing hardness h
c
,shape number of the particle n
s
, and the
initial void ratio e
0
of the particles mixture (Hardin, 1985). Calibration of the model parameters
is achieved through numerical simulations of undrained cyclic ring shear tests (Gerolymos and
Gazetas, 2007a; 2007b).
558
For a given shear stress time history, Eqns (9), (10), (13), and (14) form a system of highly
nonlinear partial differential equations with four unknowns: the excess porewater pressure p,
the breakage potential B
p
, the hysteretic parameter , and the velocity .

x
y
0
1
2
3 4
5
6
7
8
9
10
11
12
1
1 15
16
17
18 19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36 37
38
39
40
41
42
43 44
Siltstone
Sand
Marine
Sand
Imokawa
River
School
Assumed
weathered silt layer
Previous
Landslide
Assumed
Water table

Fig 12. Geometry and geological characteristics of the finite element model used in the slope stability
analysis of HigashiTakezawa

FS = 2.5 FS = 2.5
FS = 1.9 FS = 1.9
FS = 2.5 FS = 2.5
FS = 1.9 FS = 1.9

Fig 13. Deformed finite element mesh of the HigashiTakezawa landslide at the beginning of failure.
Failure surface: (a) within the sand layer, and (b) within an assumed thin silt layer at the top of the silt-
stone. The factors of safety were computed to be 2.5 and 1.9 for the first and the second case, respec-
tively
4 NUMERICAL ANALYSIS: RESULTS AND DISCUSSION
4.1 The Higashi Takezawa Landslide
First, a finite element analysis of the slope stability is carried out. The finite element model
(Fig 12) refers to the geological section of Fig 8. The analysis is performed using the code
PLAXIS (Brinkgreve and Vermeer, 1998). The soil behaviour is described by an elastic perfectly
plastic MohrCoulomb failure criterion with nonassociative plastic flow rule. Due to lack of
specific experimental results, estimation of Youngs modulus E, Poissons ratio , for all soils
559
involved in the analysis, and strength parameters
'
(effective internal friction angle) and (di-
latancy angle) for the marine sand, were based on engineering judgement. However, the actual
values of those parameters would only slightly affect the results. The dilatancy angle of the sand
and the silt is assumed to be zero, implying deformation with zero volume change during yield.
The groundwater level is estimated based on the field observations of Sassa et al. (2005). Two
cases are analyzed regarding the location of the failure surface:
(a) within the sand layer, and
(b) within the thin silt layer atop the siltstone.
The results of the analyses are shown in Fig 13 in terms of the deformed finite element mesh.
The factors of safety were computed to be 2.5 and 1.9 for the first and the second case, respec-
tively. Therefore, one could reasonably assume that the potential sliding surface during strong
earthquake loading would be formed within the silt layer.

-15
-10
-5
0
5
10
15
0 5 10 15 20
t (sec)
A
c
c
e
l
e
r
a
t
i
o
n

(
m

/

s
2
)
0
1
2
3
4
0 5 10 15 20
t (sec)
S
h
e
a
r

d
i
s
p
l
a
c
e
m
e
n
t

(
m
)
(a)
(b)

Fig. 14. (a) Input (Record NIG019 2004: PGA = 1.3 g) and response acceleration time histories at the
base of the landslide (black line), and at the soil wedge (gray line) for case b (sliding surface within an
assumed thin silt layer), and (b) computed time histories of relative shear displacements, for sliding sur-
face: (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)

With the developed model for seismic triggering and evolution of graincrushinginduced
landslide [Differential Equations: (1), (2), (9), (10), (13), and (14)] we analyse the case of Higa-
shiTakezawa . The seepage force is ignored, since the actual level of the water table during the
560
earthquake is not known. Three scenarios are studied regarding the potential location of the slid-
ing surface and the susceptibility of sand to grain crushing:

0
30
60
90
120
0 50 100 150 200 250 300 350 400
0
30
60
90
120
0 50 100 150 200 250 300 350 400
0
30
60
90
120
0 50 100 150 200 250 300 350 400
t = 5 s
0
30
60
90
120
0 50 100 150 200 250 300 350 400
0
30
60
90
120
0 50 100 150 200 250 300 350 400
0
30
60
90
120
0 50 100 150 200 250 300 350 400
t = 40 s
E
l
e
v
a
t
i
o
n

(
m
)
t = 34 s
t = 20 s
t = 25 s
t = 12.5 s
0
30
60
90
120
0 50 100 150 200 250 300 350 400
0
30
60
90
120
0 50 100 150 200 250 300 350 400
0
30
60
90
120
0 50 100 150 200 250 300 350 400
t = 5 s
0
30
60
90
120
0 50 100 150 200 250 300 350 400
0
30
60
90
120
0 50 100 150 200 250 300 350 400
0
30
60
90
120
0 50 100 150 200 250 300 350 400
t = 40 s
E
l
e
v
a
t
i
o
n

(
m
)
t = 34 s
t = 20 s
t = 25 s
t = 12.5 s

Fig 15. Snapshots of the computed evolution of the HigashiTakezawa landslide. The school is indicated
with the gray box. The initial ground surface is shown as dotted line

561
(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, 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.

0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
E
x
c
e
s
s

p
o
r
e

w
a
t
e
r

p
r
e
s
s
u
r
e

r
a
t
i
o

r
u
t = 5 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
t = 12.5 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
t = 20 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
t = 25 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
t = 34 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
t = 40 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
E
x
c
e
s
s

p
o
r
e

w
a
t
e
r

p
r
e
s
s
u
r
e

r
a
t
i
o

r
u
t = 5 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
t = 12.5 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
t = 20 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
t = 25 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
t = 34 s
0
0.3
0.6
0.9
1.2
0 50 100 150 200 250 300 350 400
t = 40 s

Fig 16. Snapshots of the computed excess porewater pressure ratio along the sliding surface for the Hi-
gashiTakezawa landslide. The school is indicated with the gray box

The results of the analysis for cases (a) and (b) are shown comparatively in Fig 14 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 displacements 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.
The results of the analysis for case (c) which allows for development of grain crushing
induced porewater pressures are presented in Fig 15 in terms of snapshots of the landslide evo-
lution, in Figs 16, 17, 18 and 20 for the distributions of velocity , excess porewater pressures,
dynamic lateral earth pressure coefficient K
dyn
, and breakage potential B
p
, respectively, at vari-
562
ous times along the sliding surface. Acceleration time histories at characteristic points of the
sliding surface are plotted in Fig 19. The following observations are worthy of note regarding
the response of the sliding mass:
At the early stages of the seismic motion, excess pore waterpressure due to particle crushing
is generated at the head of the mass and propagates rapidly towards its toe. In the following few
seconds the excess porewater pressure ratio rises up and reaches values larger than 0.9 along
the entire length of the sliding surface (t = 12.5 sec). At this time, sliding originates at the head
of the soil mass, and landsliding begins. It is very interesting that triggering occurs almost at the
end of seismic shaking, when the motion of the mass has passed its peak, and not during the
strong seismic excitation as one would expect. This implies that grain crushinginduced pore
pressure is a cumulative process (see Fig 20) and depends on the history of loading.

0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
V
e
l
o
c
i
t
y

(
m

/

s
)
t = 5 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
t = 12.5 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
t = 20 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
t = 25 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
t = 34 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
t = 40 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
V
e
l
o
c
i
t
y

(
m

/

s
)
t = 5 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
t = 12.5 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
t = 20 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
t = 25 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
t = 34 s
0
2
4
6
8
10
0 50 100 150 200 250 300 350 400
t = 40 s

Fig 17. Snapshots of the spatial distribution of velocity of the HigashiTakezawa landslide. The school is
indicated with the gray box

After its initiation, the landslide moves rapidly towards the riverbed, developing velocities
between 3 m/s and 6 m/s. Lower velocities concentrate on the rear of the landslide, while larger
values are mostly at the front which essentially governs the race of the entire landslide. At t =
25 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
8 m/s (at t 34 sec), and finally reaches the school at t 40 sec. Following the motion of the de-
563
tached 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.

0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
D
y
n
a
m
i
c

l
a
t
e
r
a
l

e
a
r
t
h

p
r
e
s
s
u
r
e

c
o
e
f
f
i
c
i
e
n
t

(
K
d
y
n
)
t = 5 s
t = 34 s
t = 40 s
K
a
= 0.25
t = 12.5 s
t = 25 s
t = 20 s
s
= 0.50
K
p
= 4
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
0 50 100 150 200 250 300 350 400
D
y
n
a
m
i
c

l
a
t
e
r
a
l

e
a
r
t
h

p
r
e
s
s
u
r
e

c
o
e
f
f
i
c
i
e
n
t

(
K
d
y
n
)
t = 5 s
t = 34 s
t = 40 s
K
a
= 0.25
t = 12.5 s
t = 25 s
t = 20 s
s
= 0.50
K
p
= 4

Fig 18 Snapshots of the computed dynamic lateral earth pressure coefficient along the sliding surface
for the HigashiTakezawa landslide

564

-15
-10
-5
0
5
10
15
0 5 10 15 20 25 30 35 40
0
30
60
90
120
0 50 100 150 200 250 300 350 400
A
B
C
D
-15
-10
-5
0
5
10
15
0 5 10 15 20 25 30 35 40
-15
-10
-5
0
5
10
15
0 5 10 15 20 25 30 35 40
-15
-10
-5
0
5
10
15
0 5 10 15 20 25 30 35 40
A
c
c
e
l
e
r
a
t
i
o
n

(
m

/

s
2
)
Excitation
C
B
A
+ uphill
downhill
X = 20 m
X = 200 m
X = 250 m
-15
-10
-5
0
5
10
15
0 5 10 15 20 25 30 35 40
D X = 330 m
Time (s)
-15
-10
-5
0
5
10
15
0 5 10 15 20 25 30 35 40
0
30
60
90
120
0 50 100 150 200 250 300 350 400
A
B
C
D
-15
-10
-5
0
5
10
15
0 5 10 15 20 25 30 35 40
-15
-10
-5
0
5
10
15
0 5 10 15 20 25 30 35 40
-15
-10
-5
0
5
10
15
0 5 10 15 20 25 30 35 40
A
c
c
e
l
e
r
a
t
i
o
n

(
m

/

s
2
)
Excitation
C
B
A
+ uphill
downhill
X = 20 m
X = 200 m
X = 250 m
-15
-10
-5
0
5
10
15
0 5 10 15 20 25 30 35 40
D X = 330 m
Time (s)

Fig 19 Acceleration time histories at selected point along the sliding surface of the HigashiTakezawa
landslide. Points C and D do not correspond to bedrock acceleration (excitation), but to response accel-
eration atop a 1 m thick soil layer due to weathering of the bedrock, which is not visible in the sketch.
565
It is seen that the calculated sliding process extended from the ruptured scarp in the source
zone to the deposition fan on the riverbed and near the school, is consistent with the field obser-
vations (Sassa et al., 2005). Clearly, there are four major stages in the run-out process: trigger-
ing (at t 12.5 sec); accelerating motion towards the riverbed (12.5 sec < t < 25 sec); separation
of the frontal part from the main body of the landslide (at t 25 sec); and deposition and decel-
eration ( t > 25 sec).
To further elucidate the kinematics of this catastrophic landslide, Fig 18 plots the evolution of
the dynamic lateral earth pressure coefficient, K
dyn
. At the early stages of the seismic loading (t
< 12.5 sec), K
dyn
is approximately equal to the lithostatic coefficient
s
except from the frontal
part of the landslide which is in the passive state. The soil mass is still at rest deforming elasti-
cally. After the activation of the sliding, K
dyn
reaches the limiting value of K
a
= 0.25 almost
along the entire length of the landslide, revealing that the avalanche body is expanding during
its course. An exception to this is the frontal edge of the sliding soil mass which essentially is in
the passive state ( K
p
= 4).

0.29
0.3
0.31
0.32
0.33
0.34
0.35
0 50 100 150 200 250 300 350 400
t = 5 s 12.5 s 20 s 25 s 34 s 40 s
B
p

(
m
m
)
0.29
0.3
0.31
0.32
0.33
0.34
0.35
0 50 100 150 200 250 300 350 400
t = 5 s 12.5 s 20 s 25 s 34 s 40 s
B
p

(
m
m
)

Fig 20. Distributions of the breakage potential B
p
along the sliding surface at various times.

At t > 34 sec, K
dyn
increases beyond the earth pressure coefficient at rest (
s
= 0.5) along the
riverbed. This signals the initiation of the deposition process. Indeed, the downhill bank of the
river restricts the motion of the depositing soil mass, which in turn blocks the flow of the incom-
ing trailing part of the slide. Being pushed from both sides, the depositing soil mass unavoidably
reaches the passive state.
A more lucid explanation of the landslide evolution is given from Fig 19, which plots the ac-
celeration time histories at characteristic points along the sliding surface, and compares them
with the excitation. At x = 20 m (point A) and x = 200 m (point B), the acceleration transmitted
into the overlying soil mass diminishes gradually, vanishing in the downhill direction. This
arises from the material softening due to rapid development of high excess porewater pressures
within the shear band. Obviously, gravity is the predominant driving force at t > 12.5 sec.
It is stressed that points C and D do not correspond to bedrock acceleration (excitation), but
to the response acceleration atop a 1.5 m thick soil layer due to weathering of the bedrock.
Therefore, yielding is allowed. This thin soil layer is not visible in the crosssection of the land-
slide area sketched in Fig 19.
By contrast to the gradual reduction of acceleration at points A and B, such reduction is not
observed at points C and D. The propagation of excess porewater pressure due to particle
crushing has not yet reached those distant points. The abrupt increase in acceleration at t 18
sec at point C, and at t 33 sec at point D, signals the arrival of the frontal edge of the land-

566

0
20
40
60
0 40 80 120 160 200
0
20
40
60
0 40 80 120 160 200
0
20
40
60
0 40 80 120 160 200
0
20
40
60
0 40 80 120 160 200
0
30
60
0 40 80 120 160 200
0
20
40
60
0 40 80 120 160 200
E
l
e
v
a
t
i
o
n

(
m
)
t = 10 s
t = 14 s
t = 21 s
t = 23 s
t = 25 s
t = 31 s
0
20
40
60
0 40 80 120 160 200
0
20
40
60
0 40 80 120 160 200
0
20
40
60
0 40 80 120 160 200
0
20
40
60
0 40 80 120 160 200
0
30
60
0 40 80 120 160 200
0
20
40
60
0 40 80 120 160 200
E
l
e
v
a
t
i
o
n

(
m
)
t = 10 s
t = 14 s
t = 21 s
t = 23 s
t = 25 s
t = 31 s

Fig 21. Snapshot of the computed evolution of the Nikawa landslide. The initial ground surface is shown
as dotted line

slide. Due to variations in the inclination of the basal topography, shock waves emanating from
the toe of the landslide, are propagating upwardly (e.g. Gray and Hutter, 1997, Tai et al, 2002)
resulting in oscillations in the response acceleration time histories. These oscillations diminish
rapidly to zero as the landslide is coming to rest. The large accelerations when the toe of the
landslide passes through each specific point (C or D) is because the velocity at these points
changes from 0 to a certain value in an extremely small period of time.
567

0
5
10
15
0 40 80 120 160 200
0
5
10
15
0 40 80 120 160 200
0
5
10
15
0 40 80 120 160 200
0
5
10
15
0 40 80 120 160 200
0
5
10
15
0 40 80 120 160 200
0
5
10
15
0 40 80 120 160 200
V
e
l
o
c
i
t
y

(
m

/

s
)
t = 10 s
t = 14 s
t = 21 s
t = 23 s
t = 25 s
t = 31 s
0
5
10
15
0 40 80 120 160 200
0
5
10
15
0 40 80 120 160 200
0
5
10
15
0 40 80 120 160 200
0
5
10
15
0 40 80 120 160 200
0
5
10
15
0 40 80 120 160 200
0
5
10
15
0 40 80 120 160 200
V
e
l
o
c
i
t
y

(
m

/

s
)
t = 10 s
t = 14 s
t = 21 s
t = 23 s
t = 25 s
t = 31 s

Fig 22. Snapshots of the spatial distribution of velocity of the Nikawa landslide

To get further insight into the mechanics behind this disastrous response, Fig 12 plots the
evolution of particle breakage potential B
p
. Notice that B
p
approaches a steady state value of
0.30 at t > 15 seconds; this is larger than the initial value of B
pl
(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 practically terminated. The effective normal stress is not adequate by large for further
breakage. However, the landslide is still accelerating due to the action of gravity.
568
4.2 The Nikawa Landslide
The developed model for landslide kinematics was also applied to analyse the case of Ni-
kawa. The response of the potentially sliding wedge is summarized in Figs 21 to 23. Specifi-
cally:

0
20
40
60
0 40 80 120 160 200
-10
-5
0
5
10
0 5 10 15 20 25 30
-10
-5
0
5
10
0 5 10 15 20 25 30
-10
-5
0
5
10
0 5 10 15 20 25 30
-10
-5
0
5
10
0 5 10 15 20 25 30
-10
-5
0
5
10
0 5 10 15 20 25 30
Time (s)
A
c
c
e
l
e
r
a
t
i
o
n

(
m

/

s
2
)
X = 30 m
Excitation
X = 81 m
X = 120 m
X = 159 m
D
C
B
A
+ uphill
downhill
A
B
C
D
0
20
40
60
0 40 80 120 160 200
-10
-5
0
5
10
0 5 10 15 20 25 30
-10
-5
0
5
10
0 5 10 15 20 25 30
-10
-5
0
5
10
0 5 10 15 20 25 30
-10
-5
0
5
10
0 5 10 15 20 25 30
-10
-5
0
5
10
0 5 10 15 20 25 30
Time (s)
A
c
c
e
l
e
r
a
t
i
o
n

(
m

/

s
2
)
X = 30 m
Excitation
X = 81 m
X = 120 m
X = 159 m
D
C
B
A
+ uphill
downhill
A
B
C
D

Fig 23. Acceleration time histories at selected point along the sliding surface of the Nikawa landslide.

569
A more lucid explanation of the landslide evolution is given in Fig. 23, which plots the accel-
eration time histories at characteristic points along the sliding surface, and compares them with
the excitation. At x = 30 m (point A) and x = 81 m (point B), the acceleration transmitted into
the overlying soil wedge diminishes gradually, vanishing in the downhill direction. This arises
from the material softening due to rapid development of high excess porewater pressures
within the shear band. Obviously, gravity is the predominant driving force at t > 14 sec.
By contrast to the gradual reduction of acceleration at points A and B, such reduction is not
observed at points C and D. The propagation of excess porewater pressure due to particle
crushing has not yet reached those distant points. The abrupt increase in acceleration at t 21
sec at point C, and at t 26 sec at point D, signals the arrival of the frontal edge of the landslide.
5 CONCLUSIONS
In this paper a model was presented for seismic triggering, evolution and deposition of a
landslide. The evolution of the landslide is modeled via an extended SavageHutter model cou-
pled with the MohrCoulomb sliding law for the frictional deformation of the material within
the shear band, and exploiting the concept of material softening due to grain crushing induced
pore water pressures. The capability of the model is investigated through prediction of two
earthquakeinduced catastrophic landslides. The HigashiTakezawa landslide in 2004, and the
Nikawa landslide in 1995.
For the case of HigashiTakezawa, three scenarios were analyzed regarding the location of
the sliding surface and the susceptibility of sand to grain crushing: (a) shear band within the
sand layer, with stable sand, (b) shear band within an assumed thin silt layer atop the siltstone,
with stable sand, and (c) shear band within the sand layer, but sand susceptible to grain crush-
ing. The residual displacement is calculated to be 0.65 m and 3.4 m for the first and second sce-
nario, respectively. The observed approximately 100 m displacement of the landslide, associated
to a shear velocity of 16 m/sec (20 sec after the triggering), is only reproduced with the third
scenario, despite the experimental fact that the residual friction angle of the silt [scenario (b)]
was 13
o
smaller than that of sand.
The following phases of the sliding process were shown to be reproduced with certain realism
by the proposed model for landslide kinematics:
- Material softening. Seismically induced shearing causes particle crushing inside the
shear band, which in turn results in pore pressure generation and decreases the effec-
tive normal stress.
- Landslide Triggering. This phase starts as soon as the pore water pressure inside the
shear band reaches and surpasses a critical value, the gravitational driving force act-
ing on the landslide mass dominates upon the resisting force due to hysteretic (Cou-
lomb) friction at the base of the slide, and landsliding begins. It was shown that trig-
gering occurs towards the end of seismic shaking, when the motion of the mass has
passed its peak, and not during the strong seismic excitation.
- Postfailure behavior. During this phase the landslide is accelerating towards the
deposition fan due to the action of gravity. Excess pore water pressure continuous to
increase but the rate of its generation diminishes gradually, because the extra pore
pressure cannot increase more than the initial effective stress, or the effective normal
stress has decreased to a certain value under which grain crushing does not further
take place.
- Deposition and Termination of the Landslide. Deceleration and finally termination of
the landslide occurs primarily due to the influence of negative inclination of the basal
topography. For example, in the case of Higashi Takezawa, the downhill bank of the
river restricted the motion of the depositing soil mass, which in turn blocked the flow
of the incoming trailing part of the slide. Being pushed from both sides, the deposit-
ing soil mass came to rest. The hydraulic resisting force at the base of the slide is also
a major stabilizing factor contributing positively to the deceleration of the landslide.
570
ACKNOWLEDGMENTS
The research for this paper was funded by the EU sixth Framework Program: Integrated R&D
Project of the EC Risk Mitigation of Earthquakes and Landslides, Research Project
LESSLOSS, Contract number: GOCE-CT-2003-505448.
REFERENCES
Armento M.C., Genevois R., Tesca P.R. [2008] Comparison of numerical models of two debris flows in
the Cortina d Ampezzo area, Dolomites, Italy, Landslides 5, 143-150.
Bouc R. [1971] Modele mathematique d hysteresis, Acustica, Vol. 21, pp.16-25
Committee of Earthquake Engineering [1996] The 1995 Hyogo-Ken Nambu earthquakeInvestigation
into damage to civil engineer structures, Japan Society of Civil Engineers.
Earthquake Engineering Research Institute [1995] The Hyogo-Ken Nambu earthquake, January 17,
1995, Preliminary EERI Reconnaissance Report.
Evans S.G., Hungr O., Clague, J.J. [2001] Dynamics of the 1984 rock avalanche and associated distal
debris flow on Mount Cayley, British Columbia, Canada; implications for the landslide hazard as-
sessment on dissected volcanoes, Eng. Geol. 61, 2951.
Gerolymos N, Gazetas G [2004] Constitutive Model for 1-D Cyclic Soil Behavior Applied to Seismic
Analysis of Layered Deposits, Soils and Foundations, Vol. 45, No 3, pp. 147159.
Gerolymos N., Gazetas G [2007a] Seismic Triggering, Evolution and Deposition of Massive Landslides:
The Case of HigashiTakezawa (2004), Proceedings of the 4th International Conference on Earth-
quake and Geotechnical Engineering (ICEGE), Thessaloniki, Greece, June 2528, (in Cd).
Gerolymos N and Gazetas G [2007b] A Model for Grain Crushing Induced Landslides Application to
Nikawa, Kobe 1995, Soil Dynamics and Earthquake Engineering, Vol. 27, pp. 803817.
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, Vol. 47, No. 1, pp. 11-25.
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, Vol. 455, pp. 1841-1874.
Hardin B [1985] Crushing of Soil Particles Journal of Geotechnical Engineering, Vol. 111, No. 10, pp.
1177-1192.
Hungr O. [1995] A model for the runout analysis of rapid flow slides, debris flows, and avalanches.
Can Geotech J 32(4), 610623.
Hutter K., Koch T. [1991] Motion of granular avalanche in an exponentially curved chute: experiments
and theoretical predictions,. Phil. Trans. R. Soc. London 1991, A 334, 93138.
Hutter K., Greve R. [1993] Twodimensional similarity solutions for finitemass granular avalanches
with Coulomb and viscoustype frictional resistance,J. Glac. 39, 357372.
Iverson R. M., Delinger R. P. [2001] Flow of variably fluidized granular masses across three-
dimensional terrain. 1. Coulomb mixture theory J. Geophys. Res. B, Vol. 106, pp. 537-552.
Japanese Geotechnical Society [1996] Geotechnical aspects of the January 17, 1995 Hyogoken-Nambu
earthquake, Soils and Foundations, Special Issue 1996.
Kallou P, and Gazetas G [2001] Dynamic Analysis of Nikawa Landslide, Proceedings of the 4th Hel-
lenic Conference on Geotechnical and Geoenvironmental Engineering, Athens, Vol. 2, pp. 171-178.
Kokusho T [2000] Mechanisms for water film generation and lateral flow in liquefied sand layer, Soils
and Foundations, Vol. 40, No. 5, pp. 99-111.
Kokusho T, Ishizawa T [2005] Energy approach to slope failures and a case study during 2004 Niigata
ken Chuetsu earthquake, Proceedings of Geotechnical Earthquake Engineering Satellite Conference
Osaka, Japan, ISSMGE, pp. 255-262.
Koseki J., Sasaki T., Wada N., Hida J., Endo M., Tsutsumi Y. [2006] Damage to earth structures for na-
tional highways by the 2004 Niigata-Ken Chuetsu earthquake, Soils and Foundations, Vol. 46, No.
6, pp. 739-750.
Medina V., Hurlimann M., Baterman A. [2008] Application of FLATModel, a 2D finite volume code to
debris flows in the northeastern part of the Iberian Peninsula, Landslides 5, 127-142.
Newmark, N M. [1965] Effect of earthquakes on dams and embankments Gotechnique, Vol.15, pp.
139160.
571
Pudasaini S P, Hutter K [2003] Rapid shear flows of dry granular masses down curved and twisted
channels, J. Fluid Mech., Vol. 495, pp. 193-208.
Sato T. [1996] Estimation of peak acceleration in the severely damaged area during the 1995 Hyogo-
Ken Nanbu earthquake, Soils and Foundations, Special Issue, pp. 35-44.
Sassa K, Fukuoka H, Wang F, Wang G. [2005] Dynamic properties of earthquake-induced large-scale
rapid landslides within past landslide masses, Landslide, Vol. 2, pp. 125-134.
Sassa K. [2005] Landslide disasters triggered by the 2004 Mid-Niigata Prefecture earthquake in Japan,
Landslides, Vol. 2, pp. 135-142.
Sassa K, Fukuoka H, Scarascia-Mugnozza G, Evans S. [1996] Earthquake induced Landslides: distribu-
tion, motion and mechanisms, Special Issue of Soils and Foundations, Japanese Geotechnical Soci-
ety, pp. 53-64.
Sassa K. [1995] Keynote lecture: Access to the dynamics of landslides during earthquakes by a new cy-
clic loading high-speed ring shear apparatus, Proceedings 6th International Symposium on Land-
slides 1992, In Landslides. Balkema, Vol. 3, pp. 1919-1939.
Sassa K. [1994] Development of a new cyclic loading ring shear apparatus to study earthquake
inducedlandslides, Report for Grant-in-Aid for Development Scientiffic Research by the Ministry on
Education, Science and Culture, Japan (Project No. 03556021), pp. 1-106.
Toyota H., Wang J., Kouichi N., Sakai N. [2006] Evaluation of natural slope failures induced by the
2004 Niigata-Ken Chuetsu earthquake, Soils and Foundations, Vol. 46, No. 6, pp. 727-738.
Tsukamoto Y, Ishihara K. [2005] Residual strength of soils involved in earthquakeinduced landslides,
Proceedings of Geotechnical Earthquake Engineering Satellite Conference, Osaka, Japan, 10 Sep-
tember 2005, Performance Based Design in Earthquake Geotechnical Engineering: Concepts and Re-
search, ISSMGE, pp. 117-123.
Voellmy A. [1955] ber die Zerstrungskraft von Lawinen, Schweiz Bauztg 73, 212285.
Wen Y.-K. [1976] Method for random vibration of hysteretic systems, Journal of Engineering Me-
chanics, ASCE, Vol. 102, pp. 249-263.
Zwinger T. [2000] Dynamik einer Trockenschneelawine auf beliebig geformten Berghngen. Ph. D. The-
sis, Vienna University of Technology (in German).

572
1 INTRODUCTION
Since soil shows nonlinear behavior even at small strain, it is essential to consider it in the
earthquake response analysis. In many constitutive models used for geo-material, stress-strain
relationship is composed of skeleton curve under virgin loading and hysteresis curve under
unloading and/or reloading process. On the other hand, strain dependent characteristics of soil
under repeated loading such as earthquake is expressed as dynamic deformation characteristics
or, so called, G- and h- relationships. Comparison of there two concept, it is usually recog-
nized that G- and h- relationships corresponds to skeleton and hysteresis curves, respectively.
As shown in this expression, both stiffness and damping characteristics have been believed to
be important factor. It looks obvious that larger damping suppress amplification of the earth-
quake motion toward the ground surface. However, the possibility appears during the 1995
Hyogoken-nambu (Kobe) earthquake that this simple concept may not be true at large earth-
quakes. Suetomi and Yoshida (1998) showed that deamplification of earthquake motion occurs
if there is soft soil such as Holocene clay from the earthquake response analysis at Kobe city. It
explained the damage ratio of wooded houses is smaller near the shore and how south (shore
side) boundary of the seismic belt zone with severe earthquake motion of seismic intensity I
JMA

is seven (maximum intensity) existed north of the one expected by theory. After that Suetomi et
al. (2000) confirmed that the peak acceleration at the ground surface has upper boundary associ-
ated by the shear strength of the weakest layer. It indicates that amplification of the ground may
be determined regardless to damping characteristics. It is, however, difficult to confirm it by
numerical analysis because of the following reason.
The hysteresis curve is obtained from the skeleton curve by applying the Masing's rule in
many constitutive models. It is convenient to handle or to develop computer codes because hys-
teresis no other information except the skeleton curve is enough in evaluating hysteresis curve.
On the other hand, the Masing's rule does not hold in the actual soil. It can be easy recognized,
Effect of hysteretic damping and stiffness at unloading on
response of ground during Earthquake
N. Yoshida
Tohoku Gakuin University, Japan

ABSTRACT: Although many stress-strain models used in the dynamic response analysis of
ground employ Masing rule to build hysteresis curves from a skeleton curve, actual damping
characteristics does not follow it. The difference between the hysteresis curves made by Masing
rule and those obtained in the laboratory test becomes significant at large strains. A rule to build
hysteresis curves that satisfy both damping characteristics and modulus at unloading point is
presented. In order to use this method, modulus at unloading point is obtained from the dynamic
deformation characteristics test result, and it is shown that this modulus can be expressed by a
hyperbolic model with respect to strain, with minimum modulus. Finally a case study is con-
ducted to examine the effect of hysteretic damping characteristics on the dynamic response. It is
shown that hysteretic damping characteristics affect the dynamic response significantly. More-
over, in contrary of common sense, earthquake motion at ground surface is larger as hysteretic
damping becomes larger.
573
for example, that damping characteristics by Hardin and Drnevich (1972) do not coincide with
the damping characteristics of hyperbolic model that they proposed as G- curve. In spite of this
clear fact, many computer codes used Masing's rule because there is no good hysteresis model.
It indicates that numerical analysis that change damping characteristics by keeping stiffness
characteristics unchanged is impossible.
In order to overcome this shortage, the author had proposed a new hysteresis rule by which
one can obtain damping characteristics that completely coincide with test result (Ishihara et al.,
1985). This concept has been employed in several models (Iai et al., 1999; Railway Technical
Research Institute, 1999). This model extended by employing piecewisely defined skeleton
curve such as piecewise linear or Lagrangean interpolation, by which one can obtain stress-
strain model that can fit G- and h- relationships perfectly (Yoshida et al., 1990).
The author then compared earthquake response of the ground by using both hyperbolic and
Hardin-Drnevich models (Takeshima et al., 2003). Note that both models have identical skele-
ton curve and different damping characteristics. It was shown that not only strain dependent
stiffness and damping characteristics but also stiffness at unloading may affect the response of
the ground. In this paper, a new model is proposed in which stiffness at unload can be consid-
ered and effect of both damping characteristics and stiffness at unload is examined.
2 MODELING OF HYSTERESIS CURVE
The concept to build hysteresis curve that has damping characteristics arbitrary specified is
shown in Figure 1 schematically (Ishihara et al., 1985), where and denote shear stiffness and
shear strain, respectively, and the subscript R denotes unloading. Here, hysteresis curve should
satisfy two requirements. The one is that if unload occurs at (
R

R
), then it should pass axial
symmetric point (-
R
-
R
). The other is that damping constant is to be equals to the specified
value. We consider a fictitious skeleton curve by which above two conditions is satisfied if
Masing's rule is employed to this fictitious skeleton curve. Since there is two conditions,
mathematical that has more than two parameters can be used for the fictitious skeleton curve.
The simplest model that satisfy this condition is a hyperbolic model expressed as
0
1 /
r
G

=
+
(1)
where G
0
is shear modulus at unload and
r
is a parameter. When we use this model, G
0
is re-
placed with elastic shear modulus G
max
, and
r
is evaluated from shear strength
f
as
/
r f max
G = . In the skeleton curve, the reference strain has physical meaning, but it does not
have physical meaning but a mere parameter.
Skeleton curve
Fictitious skeleton cuve
Hysteresis curve
R
-
R
-
R

Figure 1. Hysteresis curve that developed from fictitious skeleton curve.

Since the damping ratio varies from zero to 2/, the theoretically maximum value, this model
can be used under any damping ratio. In this sense, it is one of the most convenience and easy
handling model although it has shortage that G
0
is used as a parameter.
If one wants to add an additional condition that, for example, shear modulus at unload is
equal to elastic shear modulus, one needs a model that uses more than three parameters.
574
In this paper, we control shear modulus at unload in addition to strain dependent shear
modulus and damping ratio, we needs a three-parameters model at minimum. A Ramberg-
Osgood model is one of the relevant models, and frequently used model models that have been
used in practice. This model is expressed as
{ }
1
0
1
G

= + (2)
where and are parameters. If Masing's rule is applied, damping ratio yields
( )
0
1 /
max
h h G G = (3)
where h
max
is a maximum damping ratio. Note that damping ratio is same as that pro-
posed by Hardin & Drnevich (1972), which is the reason why this model has been
used in practice as well as hyperbolic model. The relationship between h
max
and model
parameters is expressed as
2 1
1
max
h

=
+
(4)
Since values of h, G and G
0
are known at the unloading point, h
max
is evaluated from Eq. (3).
The value of is evaluated by substituting h
max
into Eq. (4). Finally the value of is computed
from Eq. (2) by substituting these quantities. It is necessary to solve nonlinear equation in order
to calculate parameters for hyperbolic model is used for hysteresis curve, but they are sequen-
tially obtained in the Ramberg-Osgood model.
3 MODELING OF STIFFNESS AT UNLOAD
In many constitutive models, stiffness at unload is defined to be same as initial or elastic
modulus, but it may change. In order to grasp strain dependency of stiffness at unload, stiffness
at unload is read from stress-strain curve of the Toyoura sand with relative density D
r
of 50 and
80 % and Holocene clay sampled at Tokyo. Here, the stiffness is evaluated by using several
points after unloading. They are shown in Figure 2 as solid circles and G
0
with ordinary secant
modulus G (with hollow circle). The solid line in the figure is a hyperbolic model
1
1 /
max r
G
G
=
+
(5)
whose coefficient is chosen to fit the experiment. Here, it is noted that both Eq. (5) and Eq. (1) are identi-
cal. The evaluated values are summarized in
Table 1.
It is clearly seen that modulus at unload is not constant but depends on shear strain. These re-
lationships are modeled into another hyperbolic model as
0
0
1 /
1 /
min max min
max r max
G G G G
G G
= +
+
(6)
where G
min
is the minimum stiffness at unload, and
r0
is reference strain for stiffness at unload. Dotted
line in Figure 2 is a fitted curve and agreement with test result is as good as ordinary G- curves. The
evaluated values are shown in
Table 1. General tendencies of the curves are as follows:
1)
r0
becomes greater as
r
increases
2) G
min
/G
max
is larger for sand than that of clay
In the case of sand, two characteristics which are decrease of stiffness as excess porewater pres-
sure generates and increase of stiffness by cyclic mobility seems key factors affecting the stiff-
ness at unload. Since decrease of stiffness occurs at small strain where excess porewater genera-
tion is not expected occurs, another factor, possible a mere nonlinear behavior, also affects this
575
characteristics. Minimum stiffness is larger for sand with D
r
=80 % than that of sand with
D
r
=50 %, which seems to come from the difference of cyclic mobility.

1.2
1.0
0.8
0.6
0.4
0.2
0.0
S
h
e
a
r

m
o
d
u
l
u
s

r
a
t
i
o
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
Strain,
Test Model
G
G
0
Toyoura Sand
D
r
=50%
1.2
1.0
0.8
0.6
0.4
0.2
0.0
S
h
e
a
r

m
o
d
u
l
u
s

r
a
t
i
o
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
Strain,
Test Model
G
G
0
Toyoura Sand
D
r
=80%
1.2
1.0
0.8
0.6
0.4
0.2
0.0
S
h
e
a
r

m
o
d
u
l
u
s

r
a
t
i
o
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
Strain,
Test Model
G
G
0
Clay

Figure 2. Shear modulus and stiffness at unload as a function with respect to strain

Table 1 Evaluated model parameters
Material
r

r0
G
min
/G
max

Sand (D
r
=50%) 0.00025 0.0006 0.18
Sand (D
r
=80%) 0.0005 0.0015 0.35
Clay 0.0013 0.013 0.1
4 A CASE STUDY
The effects of hysteretic damping and stiffness at unload is examined through a numerical study. A
ground in Tokyo city area (Sato et al., 1998) shown in
Figure 3 is used in this case study.

35
35
V
s
(m/s)
Depth
(m)
SPT-N
value
0 50
234
70
261
259
135
221
228
234
221
231
221
400
10
15
5
0
Bed-
rock
Sand
Sand
Clay
1.75
1.95
1.75
2.00
1.95
2.10
1.75
(t/m
3
)
c
(kPa)

12.6
75.0

Figure 3. Soil profiles
4.1 Soil profiles and material
Dynamic deformation characteristics of this site are modeled into Hardin-Drnevich model by
Sato et al. (1998). Model parameters are summarized in Table 2, dynamic deformation charac-
teristics are shown in Figure 4 as H-D model (dashed line).
Three stress-strain models are used in this case study. All models have the same hyperbolic
skeleton curve expressed as Eq. (1), but hysteresis curves are differently defined.
1) Hyperbolic model: ordinary hyperbolic model whose hysteresis loop is made by applying
the Masing's rule. This model is referred as "Hyperbolic."
2) H-D model: damping characteristics is evaluated from Eq. (3). This model gives the same
strain dependent shear modulus and damping characteristics proposed by Hardin and
Drnevich (1972). Since the hyperbolic model is used for hyperbolic is used for the hyper-
576
bolic model, stiffness at unload cannot be controlled, but determined automatically. This
model is referred as "H-D"
3) Same as "H-D", but hysteresis curve is made by means of R-O model. Therefore, stiffness
at unload can also be controlled, and Eq. (6) is used to evaluate it. This model is referred as
"H-D /w E". Note that both "H-D" and "H-D /w E" gives the same damping characteristics
shown as H-D in Figure 4, and will be call H-D type model in the discussion, especially to
compare with hyperbolic model which shows much larger damping.
Damping characteristics of the hyperbolic model is also shown in Figure 4. As well known,
the hyperbolic model shows much larger damping at large strain than that of the H-D model, but
damping at small strain is nearly the same for both models. Figure 5 compares stress-strain
curve of three models for shear strain amplitude of 0.6 % and 4 %, respectively. Curves are
quite different between hyperbolic model and two H-D type models, but that those by two H-D
type models are similar to each other, which indicates that stiffness at unload may not be a key
factor.

Table 2. Model parameters
Material
r
h
max

r0
G
min
/G
max

Sand 8.6310
-4
0.22 0.002 0.4
Clay 1.4210
-3
0.22 0.013 0.1

1.2
1.0
0.8
0.6
0.4
0.2
0
S
h
e
a
r

m
o
d
u
l
u
s

r
a
t
i
o
,

G
/
G
m
a
x
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
Strain,
40
30
20
10
0
D
a
m
p
i
n
g

r
a
t
i
o
,

h

(
%
)
Sand
Hyperbolic
H-D
1.2
1.0
0.8
0.6
0.4
0.2
0
S
h
e
a
r

m
o
d
u
l
u
s

r
a
t
i
o
,

G
/
G
m
a
x
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
Strain,
40
30
20
10
0
D
a
m
p
i
n
g

r
a
t
i
o
,

h

(
%
)
Clay
Hyperbolic
H-D

Figure 4. Dynamic deformation characteristics of sand and clay.

-10
-5
0
5
10
S
t
r
e
s
s
,

(
k
P
a
)
-0.8 -0.4 0 0.4 0.8
Strain, (%)
Hyperbolic
H-D /w E
H-D
-15
-10
-5
0
5
10
15
S
t
r
e
s
s
,

(
k
P
a
)
-6 -4 -2 0 2 4 6
Strain, (%)
Hyperbolic
H-D /w E
H-D

Figure 5. Stress-strain curve of clay.

4.2 Earthquake motions
Two earthquake motions are chosen among the earthquake motions shown by Sato et al.
(1998), which is shown in Figure 6. They have limited number of large amplitude wave and
large number of large amplitude wave, and are called shock wave and vibration wave, respec-
tively. Since difference of damping characteristics appears at large strains. Therefore, the waves
are scaled so that peak acceleration becomes 8 m/s
2
at the outcrop base layer.
577

-4
-2
0
2
4
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
20 15 10 5 0
Time (sec.)
Shock type
-0.5
0.0
0.5
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
45 40 35 30 25 20 15 10 5 0
Time (sec.)
Vibration type

Figure 6. Earthquake motions to the engineering seismic base layer

4.3 Response under shock wave and discussion
Maximum response is shown in Figure 7. The maximum acceleration decreases rapidly at
GL-5.8 m, resulting in about 2 m/s
2
at the ground surface. Since the layer between GL-3.8 and
5.8 m (clay layer) shows large strains of several percent, shear stress in these layer reaches shear
strength, which can be confirmed through the chained line (shear strength) in Figure 7. If a layer
reaches shear strength, acceleration above this layer reaches limit acceleration (Suetomi et al.,
2000) because of the equilibrium condition of shear stress in this layer and inertia force above
this layer. There is slight difference between hyperbolic model and H-D type model in the layer
between GL-5.8 and 11.8 m; hyperbolic model shows larger acceleration. However, the differ-
ence is very small that we can say the response is identical in the engineering practice.

Depth
(m)
1.0
1.9
2.8
3.8
4.8
5.8
7.6
8.8
9.5
10.8
11.8
12.8
13.8
14.8
16.0
Soil
Type
Sand
Clay
Clay
Sand
Sand
Sand
Sand
Sand
Sand
Sand
Sand
Base
V
s
(m/s)
234
234
70
70
70
70
171
261
259
221
228
234
221
231
221
400
Unit
weight
(tf/m
3
)
1.75
1.95
1.75
1.75
1.75
1.75
1.75
2.00
2.00
1.95
1.95
1.95
1.95
1.95
1.95
2.10
Peak Acceleration
(m/sec
2
)
2 4 6
Peak Displacement
(cm)
5 10 15
Max. Stress
(kN/m
2
)
20 40 60 80
Max. Strain
(%)
2 4 6
Shock wave
Hyperbolic
H-D
H-D /w E
Shear strength

Figure 7. Maximum response under shock wave

Comparison of acceleration time history at the ground surface is shown in Figure 8. Although
there is less difference in peak response, there are significant differences between hyperbolic
model and H-D type model in the acceleration; hyperbolic model show large number of large
acceleration wave whereas H-F type model shows small amplitude waves after the peak accel-
eration.
578
As discussed by Suetomi et al. (2000) and in the preceding, acceleration at the ground surface
is strongly affected by the nonlinear behavior of the weakest layer. In order to see the differ-
ences more clear, stress-strain curve of the 6th layer where shear strength is the smallest and
maximum shear strain is the largest is shown in Figure 9 and time history of strain is shown in
Figure 10.

-2
0
2
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
20 15 10 5 0
Time (sec.)
Hyperbolic model
Shock type
-2
0
2
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
20 15 10 5 0
Time (sec.)
Hardin-Drnevich model
Shock type
-2
0
2
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
20 15 10 5 0
Time (sec.)
Hardin-Drnevich model /w E
Shock type

Figure 8. Acceleration time history at ground surface under shock wave

-10
-5
0
5
10
S
h
e
a
r

s
t
r
e
s
s
,

(
k
P
a
)
-5 0
Shear strain, (%)
Hyperbolic Shock
GL-4.8~5.8m
-10
-5
0
5
10
S
h
e
a
r

s
t
r
e
s
s
,

(
k
P
a
)
-5 0
Shear strain, (%)
Hardin-Drnevich /w E Shock
GL-4.8~5.8m
-10
-5
0
5
10
S
h
e
a
r

s
t
r
e
s
s
,

(
k
P
a
)
-5 0
Shear strain, (%)
Hardin-Drnevich Shock
GL-4.8~5.8m

Figure 9. Stress-strain curves at 6th layer under shock wave

-5
0
S
h
e
a
r

s
t
r
a
i
n

(
%
)
20 15 10 5 0
Time (sec.)
Shock type
Hyperblic
H-D /w E
H-D

Figure 10. Stress-strain curve under shock wave

Maximum strain of several percents occurs at about 7.5 s following a relatively large spike at
about 6.5 seconds with shear strain of about 1%. Shear strain is the largest in the hyperbolic
model in both positive and negative sides. This is against the common sense that hysteresis
curve with larger damping constant suppress the amplification resulting in smaller response (e.
g., Railway Technical Research Institute, 1999). If, let suppose, earthquake motion is a mere
pulse, the it is clear that only skeleton curve affects the maximum response and hysteresis loop
does not affect it because maximum response occurs on the skeleton curve. Therefore, differ-
ence comes from the hysteretic behavior.
Since hyperbolic model has large damping, stiffness after unload becomes larger than H-D
model, which can be seen clearly in Figure 5 and Figure 9. Therefore, shear stress increase more
rapidly than that of H-D type model, which indicates that hyperbolic model keeps nearly perfect
579
state longer, resulting in large strains. On the other hand, since the stiffness is smaller after
unload in the H-D type models, larger strain is required to cause the same shear stress as hyper-
bolic model. However, generating larger strain requires more time. If unload occurs before it,
shear stress at the cycle becomes smaller. This seems the reason why H-D type models show
smaller shear stress amplitude after maximum stress. This again explain why acceleration is
smaller in H-D type models than that in hyperbolic model.
Finally acceleration response spectrum is shown in Figure 11. Acceleration is nearly identical
in all models at period larger than 0.5 s, but that by hyperbolic model is much larger than that by
H-D type models. This frequency range is important in the engineering practice. Therefore ef-
fect of shape of hysteresis loop is important in the earthquake response. It is also emphasized
that model with larger damping ratio shows larger response acceleration.

15
10
5
0
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
0.1 1
Period (sec.)
Shock type 5% Damping
Hyperbolic
H-D
H-D /w E

Figure 11. Response spectrum under shock wave

Differences between two H-D type models, that have same stiffness and damping characteris-
tics but stiffness at unload is taken into account only in the proposed model in this paper, is not
large; the behavior is quite similar. Regardless of the models, shape of the hysteresis loop must
be in spindle shape. Therefore, if their area are same, then the possible shape is limited. Actually
the shapes of both models is quite similar as can be seen in Figure 5 and Figure 9. This is the
reason why two H-D type models show similar behavior.
4.4 Response under vibration wave and discussion
Same as shock wave case, maximum response is shown in Figure 12, acceleration time his-
tory at the ground surface is shown in Figure 13, stress-strain curves are shown in Figure 14,
time history of shear strain is shown in Figure 15, and acceleration response spectrum is shown
in Figure 16.

580
2 4 6 8
Shear strength
Depth
(m)
1.0
1.9
2.8
3.8
4.8
5.8
7.6
8.8
9.5
10.8
11.8
12.8
13.8
14.8
16.0
Soil
Type
Sand
Clay
Clay
Sand
Sand
Sand
Sand
Sand
Sand
Sand
Sand
Base
V
s
(m/s)
234
234
70
70
70
70
171
261
259
221
228
234
221
231
221
400
Unit
weight
(tf/m
3
)
1.75
1.95
1.75
1.75
1.75
1.75
1.75
2.00
2.00
1.95
1.95
1.95
1.95
1.95
1.95
2.10
Peak Acceleration
(m/sec
2
)
2 4 6
Peak Displacement
(cm)
5 10 15
Max. Stress
(kN/m
2
)
20 40 60 80
Max. Strain
(%)
Vibration type wave
Hyperbolic
H-D
H-D /w E

Figure 12. Maximum response under vibration wave
-2
0
2
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
45 40 35 30 25 20 15 10 5 0
Time (sec.)
Hyperbolic model
Vibration type
-2
0
2
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
45 40 35 30 25 20 15 10 5 0
Time (sec.)
Hardin-Drnevich model
Vibration type
-2
0
2
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
45 40 35 30 25 20 15 10 5 0
Time (sec.)
Hardin-Drnevich model /w E
Vibration type

Figure 13. Acceleration time history at ground surface under vibration wave

-10
-5
0
5
10
S
h
e
a
r

s
t
r
e
s
s
,

(
k
P
a
)
10 5 0 -5
Shear strain, (%)
Hyperbolic Vibration
GL-4.8~5.8m
-10
-5
0
5
10
S
h
e
a
r

s
t
r
e
s
s
,

(
k
P
a
)
-5 0 5
Shear strain, (%)
Hardin-Drnevich /w E Vibration
GL-4.8~5.8m
-10
-5
0
5
10
S
h
e
a
r

s
t
r
e
s
s
,

(
k
P
a
)
-5 0 5
Shear strain, (%)
Hardin-Drnevich Vibration
GL-4.8~5.8m

Figure 14. Stress-strain curve at 6th layer under vibration wave

581
-10
-5
0
5
10
S
h
e
a
r

s
t
r
a
i
n

(
%
)
45 40 35 30 25 20 15 10 5 0
Time (sec.)
Vibration type
Hyperblic
H-D /w E
H-D

Figure 15. Strain time history at 6th layer under vibration wave

15
10
5
0
A
c
c
e
l
e
r
a
t
i
o
n

(
m
/
s
2
)
0.1 1
Period (sec.)
Vibration type
5% Damping
Hyperbolic
H-D
H-D /w E

Figure 16. Response spectrum under vibration wave
Stiffness at unload is shown not to be same as initial elastic modulus but depend on shear strain.
Then a new constitutive model is proposed, which completely satisfy arbitrary given G- and h-
relationships as well as stiffness at unload. Comparisons are made between the conventional
hyperbolic model which shows much larger damping ratio than test and two H-D model type
models both of which have the same strain dependent stiffness and damping characteristics and
their damping characteristics follow the formula shown by Hardin and Drnevich (1972). In the
two H-D models, the one uses hyperbolic model for hysteresis curve, where as the other uses R-
O model in order to control stiffness at unload as well as strain dependent stiffness and damping
characteristics. The following conclusions are obtained.
1) Stiffness at unload decreases with strain, and it can be modeled by a hyperbolic equation
with minimum stiffness.
2) Difference of hysteresis loop does not affect the maximum response significantly under
large earthquakes, because response at the ground surface is controlled by the shear strength of
the weakest layer. On the other hand, they affect hysteretic behavior after unload. It affects re-
sponse spectrum in shorter period, but affected period is important frequency for many struc-
tures.
3) It has been believed that response is smaller when damping constant is larger, but it is not
true. If damping is larger, stiffness at unload becomes larger, which result in larger response at
short period, but this period range is important in the engineering practice.
4) Effect to consider stiffness at unload is not large if damping characteristics is same, proba-
bly because shape of hysteresis loop essentially similar if area of the hysteresis loop is the same.

In conclusion, we should distinguish the effect of damping and maximum surface response
which is strongly constrained by the shear strength at the weakest layer under large earthquakes.
We also should recognize that larger damping creates larger stiffness after unload.
582
REFERENCES
Hardin, B. O. & Drnevich, V. P: 1972. Shear modulus and damping in soils: design equations and curves.
Proc. of the American Society of civil engineers, 98 (SM7):.667-692
Ishihara, K., Yoshida, N. & Tsujino, S. 1985. Modelling of stress-strain relations of soils in cyclic load-
ing, Proc. 5th International Conference for Numerical Method in Geomechanics, 1: 373-380
Iai, S., Matsunaga, Y. & Kameoka, T. 1999. Strain space plasticity model for cyclic mobility, Soils and
Foundations. 32 (2): 1-15
Railway Technical Research Institute. 1999. Design standard of railway facilities and commentary, Maru-
zen, Tokyo
Yoshida, N., Tsujino, S. & Ishihara, K. 1990. Stress-strain model for nonlinear analysis of horizontally
layered deposit, Summaries of the Technical Papers of Annual Meeting of AIJ, Chugoku. B (Structure
I): 1639-1640 (in Japanese)
Takeshima, Y., Sawada, S., Fujii, N. & Yoshida, N. 2003. Effect of nonlinear modeling of hysteresis be-
havior on the dynamic response of ground, Proc., 58th Annual Conf. of the JSCE. I: 537-538 (in
Japanese)
Sato, M., Yasuda, S., Yoshida, N. & Masuda, T. 1998. Simplified method for estimating maximum shear
stress in the ground during earthquakes, J. of Geotechnical Engineering, Proc. JSCE. 610/III-45: 83-96
(in Japanese)
Suetomi, I., Sawada, S., Yoshida, N. and Toki, K. 2000. Relation between shear strength of soil and upper
limit of earthquake ground motion, J. of Structural Mechanics and Earthquake Engineering, Proc.
JSCE. 654/I-52: 195-206 (in Japanese)
Suetomi, I. and Yoshida, N. (1998): Nonlinear behavior of surface deposit during the 1995 Hyogoken-
nambu earthquake, Soils and Foundations, Special Issue on Geotechnical Aspects of the January 17
1995 Hyogoken-Nambu earthquake, No. 2, pp. 11-22

583
6.
Monuments
1 INTRODUCTION

Dafni Monastery, situated at approximately 10km from Athens on the way to Eleusis, is one of
the major Byzantine monuments in Greece (Millet 1899, Stikas 1962, Bouras 1998, Delinikolas
et al. 2003). The Katholikon (main Church) of the Monastery (Figures 1 and 3) has suffered se-
vere damages during the 7
th
September 1999 earthquake that affected the region of Attica.
Immediately after the earthquake, a multidisciplinary working group (Delinikolas et al. 1999)
was formed by the Hellenic Ministry of Culture (HMC) with the assignment to do the necessary
inspections, assess the nature and the significance of damages and elaborate, together with a
Scientific Committee set to this purpose (composed by Professors Ch. Bouras, T. P. Tassios, E.
Mariolakos and N. Zias) and all competent authorities of the Ministry, a strategic plan for the
protection, conservation and restoration of the monument, its mosaics included.
Due to the severe damages of the structure and the danger of eventual aftershocks, the decision
was taken for the application of emergency measures (Miltiadou-Fezans et al. 2003a). The aim
of those measures was (a) to reduce the danger of further deterioration of structural damage and
(b) to ensure accessibility and safe working conditions for all the scientific and technical per-
sonnel, thus enabling the execution of all the surveys and investigations, necessary for the de-
sign and implementation of the most adequate structural restoration interventions.
In parallel with the design and implementation of emergency measures, a first series of re-
search programs were initiated, in order to record the state of the monument immediately after
the earthquake. These programs were aiming at: (i) accurate survey of geometry (Georgopoulos
et al. 2003), (ii) preliminary survey of damages, assessment of mechanical properties of con-
struction materials, monitoring of the width evolution of main cracks, and preparation of a first
detailed computer model of the monument used for preliminary structural analyses (Vintzileou
Structural restoration process to protect a world heritage
monument after an earthquake disaster: the case of the Katholikon
of Dafni Monastery in Attica Greece

A. Miltiadou-Fezans, Dr Civil Engineer,
Hellenic Ministry of Culture
ABSTRACT: The Byzantine monastery of Dafni (world heritage list of UNESCO), is one of
the most important monuments of middle Byzantine period, famous worldwide for the mural
mosaics of its Katholikon (main church). Being constructed in a seismic area, the monastery
was affected throughout its history by a large number of earthquakes that caused structural
problems and damages to its mosaics. This paper focuses on the structural restoration process
adopted after the Athens earthquake disaster of September 1999 to protect this world heritage
monument. A synopsis of the historical and recent pathology of the main church and the emer-
gency measures undertaken by the Hellenic Ministry of Culture is first presented, followed by a
synthesis of the results of research programs, investigations and structural restoration studies.
Furthermore, a brief presentation is made of the first phase of interventions executed for the
consolidation of the masonry and the control of their efficiency using novel methodologies, non
destructive techniques and monitoring systems.

584
2002) and (iii) assessment of the physicochemical characteristics of the construction and point-
ing mortars (Papagianni 2002).
The above research works were carried out in close collaboration with the responsible multi-
disciplinary team nominated by the competent authorities of the HMC. On the basis of a syn-
thesis of the obtained results, adequately completed with detailed in situ observations and arc-
hive researches, the multidisciplinary group of the HMC has elaborated the structural
restoration study (Delinikolas et al. 2003, Miltiadou-Fezans et al. 2003b). Following the pro-
posals of this study, approved by the Scientific Committee and the Central Archaeological
Council, it was decided to implement the structural restoration design and interventions in two
phases.

Figure 1. N-W view of the Katholikon of Dafni Monastery

Given the importance of the monument, a step by step multidisciplinary approach was adopted,
both concerning the design and the implementation of the structural restoration interventions.
Such an approach gives the possibility to start the execution of a series of interventions, and in
the same time to perform the in situ and laboratory investigations that are necessary for the de-
sign of the next step. Furthermore, the assessment of the effect of the first step interventions to
the structural behaviour of the monument can be carried out and taken into account for the de-
sign of the works of the second step. The first phase of interventions (that is completed) com-
prises measures taken to repair and strengthen masonry elements. The interventions of the
second phase aim to improve the overall behaviour of the entire structure; their design is still in
progress.
Moreover, within the HMC, it was judged as absolutely necessary to support the two phases
of works with further research and investigations, comprising the design of adequate mortars
and grouts, the experimental estimation of mechanical characteristics of masonry before and af-
ter grouting, the seismic monitoring of the structure, the application of NDT for investigating
invisible parts of the monument and controlling the grouting effect and efficiency.
In this paper, a brief synthesis of all the aforementioned investigations, studies, emergency and
structural restoration works will be presented.
2 BRIEF DESCRIPTION OF THE KATHOLIKON OF DAFNI MONASTERY
The Byzantine monastery of Dafni comprises various buildings (laid out in a square plan), con-
structed over a long period of time, starting on the 11th century AD (Delinikolas et al. 2003).
Currently, most of the buildings are in ruins, with the exception of the Katholikon, part of the
internal range of cells, the cistern and the northern fortification walls.
The Katholikon belongs to the octagonal type and preserves large part of the original mural
mosaics. It comprises the main church, the sanctuary, the narthex and four chapels, which com-
plete its orthogonal plan. In the western part, only the perimeter walls of an exonarthex or por-
tico and those of a spiral stairway tower leading to the upper floor have survived (Figs. 1-2, 5).
The central part of the main church is cross-shaped in plan, the hemispherical dome rising over
585
its square core. The dome is 8,2m in diameter and 16,4m high, and rests on an almost cylindric-
al drum with 16 piers and 16 vaulted windows. The dome and its drum are carried by eight
pendentives and eight arches (four semicircular and four embodied in the squinches of the cor-
ners), forming an octagon and achieving in this way the transition from circle to square. Thus,
twelve piers (laying out in a square plan), provide support to the dome together with the groin
vaulted arms of the cross, situated in a higher level (Figs. 3-4). All the other parts of the monu-
ment are covered with byzantine groin vaults.

Figure 2. a) N-E view of Katholikon.

Figure 3. The central dome and its drum,
curried by eight pendentives and eight
arches.

As one can clearly distinguish in Figures 1 and 3, the exterior face of the vertical perimeter
walls are built according to the enclosed brick system (stones with bricks around them), follow-
ing two different masonry construction types. In fact, there is a lower zone, built with large di-
mension stones (often, reused material), their length placed horizontally or vertically, in order
to form crosses, and an upper zone constructed with ashlar masonry, using smaller and per-
fectly cut orthogonal stones (Delinikolas et al. 2003). Bricks are placed in all horizontal joints,
and in most of the vertical ones. The thickness of mortar joints is of 3cm approximately. Plain
brick masonry was used in the construction of all the windows and doors (Fig. 2). In the inter-
nal face of the perimeter walls, as well as in internal masonry elements in general, the enclosed
brick system is not followed systematically, whereas in various locations, a mixture of cut,
semi-cut and rubble stones are used together with bricks. Thus, the thickness of masonry joints
is varying. Finally the entire vaulted roof was constructed with plain brick masonry.
Both the lower and upper parts of the perimeter walls are constructed following the three-leaf
masonry type, with varying widths of the leaves, as will be more analytically presented below
(Section 9).
3 CONCEPT AND BRIEF PRESENTATION OF EMERGENCY INTERVENTIONS
The emergency interventions were designed taking into account specific demands deriving from
the importance of the monument and the necessity for implementation of final restoration works
without removing the supports and scaffoldings. Thus, they had to be reversible, easily assem-
bled and allowing for gradual disassembling in the interior, as well as adjustable to the de-
formed geometry of damaged elements. Moreover, any contact with the vulnerable mural mo-
saics was to be avoided.
To this end various alternative solutions were examined (Miltiadou-Fezans et al. 2003a).
Figure 4 shows schematically the retained one. In the NE corner of the building, three double-
framed steel raking shores was constructed, as in this area a pronounced tilting of the external
walls has been noticed and the telltales installed just after the earthquake, indicated further
opening of cracks and a tendency of the corner to detach (Fig. 5). In the interior, and in the ex-
586
onarthex, vertical steel props were built beneath the main arches, in order to provide vertical
support to their cracked structure (Fig. 6).

Figure 4: Plan of the Katholikon and schematic
presentation of emergency measures
Figure 5: Raking shores installed in the NE corner

Both the raking shores and the vertical props were self supported elements, which would be ac-
tivated only in case of further increase of the deformations of the structure. Thus, between the
metal framework and masonry walls, a 12cm full layer of wooden beams and wedges (together
with a 3mm soft packing) were inserted, to provide good contact with the masonry without
harming the surfaces, while allowing relative movement, unless the structure did start to move
towards them. Moreover, the upper part of the exonarthex walls and the NW and SW piers were
confined using steel plates and bars.

Figure 6. Interior and exonarthex vertical steel props Figure 7. Domes mergency measures

Regarding the drum of the dome, specially designed steel elements were constructed to brace
the windows and confine in two levels the masonry piers (Fig. 7), taking special care to assure
the in situ assemblage of all these structures without harming the mural mosaics. Furthermore,
the following measures have been undertaken: (a) the installation of adequate types of scaffold-
ings in the interior and exterior of the church, in order to offer safe working conditions for the
personnel, and (b) the removal of the tiles of the roof and the application of temporary water
isolation membranes just below them, to protect the cracked extrados of the vaulted structures,
and hence the mural mosaics from leaking water.
4 HISTORICAL PATHOLOGY AND PAST INTERVENTIONS
The Monastery is situated in a Neocene tectonic graben between the mountains Egaleo and Ko-
rydallos at the west side of Athens, 150m away from the E-W trending marginal fault between
the alpine Mesozoic limestone and the post-alpine deposits (Mariolakos et al. 2000). Located in
a tectonically active area, many intensive earthquakes damaged Dafni Monastery throughout
the centuries (from 11th c. to our era). Partial collapses and extended damage were provoked to
the structure and the mural mosaics, which led to major interventions. Detailed presentation of
587
the long history of the monument is reported in Delinikolas et al. (2003), whereas a concise
summary is given in this paragraph.
Two main phases of construction have been identified in the Katholikon of Dafni Monastery,
both dating back to the middle Byzantine period. Although the construction type of masonry is
similar in the two phases, the composition and final texture of pointing mortar used is different,
thus allowing for the two phases to be identified in situ. The first phase comprises the main
church and the narthex (Fig. 8). A reddish mortar containing crashed bricks was used for the
pointing of the joints of the monuments external faades. At a second phase, a portico or ex-
onarthex was added to the church (in its west part); masonry piers in the corners and marble io-
nic columns in the middle were used to support the arches of the added perimeter walls. Over
the portico and the west part of the church a first floor was also added. Finally a spiral staircase
tower was constructed in the NW corner of the narthex (Fig. 9). A whitish mortar with a yello-
wish patina was used for the pointing of the joints of the external faades.

Figure 8. W and S faade: 1
st
construction
phase. Drawings of Benouville modified by De-
linikolas (Millet 1899, Delinikolas et al. 2003).
Figure 9. W and S faade: 2
nd
construction phase.
Graphical representation of Stikas modified by Deli-
nikolas (Stikas 1962, Delinikolas et al. 2003).

Severe damage was provoked to the church during the 13
th
century, when Athens was occupied
by the Franks. The first floor and the three groined vaults of the exonarthex have probably col-
lapsed, provoking severe damage to the rest of the structure. The narthex, the dome and its sup-
porting system were as well severely damaged. Thus, extensive interventions were undertaken
by the Cistercian monks mainly in the west part of the monument (13
th
-15
th
c.). The upper part
of the west wall and the majority of arches, were reconstructed with curved stones following
the gothic style (Fig. 10). Furthermore the exonarthex was covered with a timber roof and bas-
tions were built on the top of the walls. The remains of the constructions of the first floor, the
narthex and the rest of the church were also locally repaired. These extended interventions in
the west part of the Katholikon can be considered as the third construction phase of the build-
ing.

Figure 10. Graphical representation of the
W faade of the portico as restored by the
Cistercians (Delinikolas et al. 2003).

Figure 11. West faade of the exonarthex before and after
Stikas restoration interventions (photos: Stikas 1962)

588
During Othoman occupation strong earthquakes provoked further deterioration of the damages
of the entire church. Cracks appeared on the vaulted roof of the crypt and the narthex, whereas
severe damage and partial collapse occurred to the exonarthex and the spiral staircase. Thus,
the arches of the exonarthex perimeter walls were filled with masonry, and extended alterations
took place in the west part of the monument. During that period some of the ionic columns were
taken by Lord Elgin to England.

(Level of entrance doors) (Level of lower windows) (Level of squinches)

Figure 12. Interventions to the monument: Plans

Figure 13. Interventions to the monument: S and N faade

Figure 14. Interventions to the monument: West faade of (a) the church and (b) the exonarthex

The structural condition of the monument was further deteriorated during the 19
th
and the 20
th

centuries, due to numerous earthquakes occurred and major interventions were applied for the
preservation of the monument. One should mention the 1889 and 1894 earthquakes (estimated
magnitudes 6.7 and 7.0 on the Richter scale), the 1914 earthquake (M 6.0R), the 1981 earth-
quake (M 6.7R) and finally the 1999 one (M 5.9R). After the 1889 earthquake, the heavily
damaged dome and its drum were removed and reconstructed. Three concentric iron rings (I-
589
beams) were inserted at the base of the drum of the dome, as identified during the recent resto-
ration works. Before removing the old dome, the mosaics were detached and, after their conser-
vation, partially re-attached to the monument by Italian Conservators. During these interven-
tions, the lime based mortar of the mosaics substrata was replaced with a hydraulic lime based
one. The 1894 strong earthquake caused further damages; even the recently reconstructed dome
and its drum were once again cracked. Extended interventions took place to the west part of the
monument, including the reconstructions of almost the entire narthex (Troump 1896). In the pe-
riod of 1897 to 1907, two stone buttresses both sides of the north entrance to the church were
constructed, while metallic trusses (fixed on new foundation) were installed both sides of the
south central piers, together with iron plates and bars inserted at three levels, in order to confine
them. In 1914, the suggestion is made for external confinement of the central cupola with an
iron ring. This solution, however, was applied 40 years later.
In that period of 1955-1960 the ottoman interventions were removed from the exonarthex
(Fig. 11) and the Cistercian construction phase was restored by Stikas (Stikas 1962). The col-
lapsed part was rebuilt and the missing ionic columns were replaced by masonry piers. A con-
crete beam was installed in the interior of the reconstructed parts, without affecting the byzan-
tine remains of the SW pier. The upper part of the SE chapel was also reconstructed, while
extended local re-pointing took place in various areas of the Katholikon, using mainly cement
based mortars.
Fortunately the non deteriorated old mortars were not removed. Thus, they survived to our
days in better condition than those used by Stikas, which presented extended decay and during
the recent restoration works were removed.
Figures 12 to 14, taken from Delinikolas et al. (2003), present the historical pathology and
past interventions on drawings. The reconstructed parts of the monument are shown with solid
colors. Yellow: 13
th
-15
th
c., orange: 1891, green: 1895, purple: 1897-1907, dark blue: 1955-60.
The 1981 earthquake (Alkyonides islands) caused numerous hair cracks to the building and se-
vere damages to the mosaics. Thus, the pathology observed in the monument after the 1999
earthquake was only partly due to this severe earthquake.
5 SURVEY OF DAMAGES AFTER THE 1999 EARTHQUAKE
The systematic survey of damages after the September 7
th
, 1999 earthquake, reported in Miltia-
dou-Fezans et al. 2003b and 2004, is summarized in this section. In Figure 15 some typical
drawings of cracks and deformations survey are presented, showing the severe damages ob-
served in the monument (both to its structural part and to the mosaics).
An extensive network of shear and bending cracks (ranging from hair cracks to those several
centimeters wide) has appeared on the walls and piers of the monument, whereas numerous old
cracks (due to previous earthquakes) increased in length and width. Severe structural disloca-
tion and outwards movement of the walls was recorded in the NE corner of the main church
(~14cm to the N and ~10 cm to the E). Significant out-of-plane displacement of the N and S
arms of the cross (~16 cm and 21 cm respectively), and of the free standing west wall of the ex-
onarthex were also recorded, due to further deterioration of previous deformations (~16 cm in
the corners and ~25 cm in the middle).
The damages were more extensive in the higher parts of the structure, especially in the
sanctuary, the arms of the cross and all the arches below the dome area. As shown in Figure 15,
the NE and NW small arches just below the squinches presented severe dislocation near their
crown, followed by out of plane deformations of the squinches themselves. Cracks appeared al-
so in all the groin vaults of the church. The structural condition of the dome (reconstructed at
1891 and damaged soon after its reconstruction at 1894), was assessed as extremely critical
immediately after the earthquake (Delinikolas et al. 1999, Miltiadou-Fezans et al. 2003a,
2003b). Horizontal cracks have appeared along the perimeter of the drum (both at its base and
top). In the piers of the drum that are situated perpendicular to the East-West direction, horizon-
tal cracks (due to out-of-plane bending) have opened at their top and bottom. In the piers that
are situated parallel to the E-W axis, diagonal or bi-diagonal (shear) cracks have appeared. In
the intermediate piers, mixed type of (less severe) cracks was observed.

590

(a) East-West section. View to South (b) East-West section. View to North

(c) North-South section. View to West (d) North-South section. View to East.
Figure 15. Typical presentation of damages

This pathological image does seem to confirm seismological data regarding the predominant di-
rection of the 1999 earthquake. Thanks to the external upper metal tie-rod, confining the he-
mispherical dome near the springing level, the occurrence of severe cracking of the domes
shell was prevented.
6 QUALITATIVE INTERPRETATION OF DAMAGES
As reported in Miltiadou-Fezans et al. 2004, it was observed that the number and the opening of
crack in the vertical elements of the Katholikon increase from the base to the top of the monu-
ment. It was therefore obvious that the monument exhibits the tendency to open from the base
to the top along both main directions. This deformed shape of the church, verified by the pho-
togrammetric survey (Georgopoulos et al. 2003), is confirmed also by the history of the monu-
ment. The South faade of the narthex reached in 1894 a total out-of-plane deformation larger
than 200mm and it was reconstructed (Figs. 13-14). The photograph in Figure 16 shows the re-
gion between the original and the reconstructed part of that faade, where one may clearly see
the permanent out-of-plane deformation of the original part.
However, even this reconstructed part of the monument presents today a total out-of plane
deformation of 90 mm (drawing in Fig. 16). This is another element proving that the feature we
observe now in the monument is an inherent characteristic (due to its initial construction
scheme and the extended alterations/interventions undertaken during its lifetime).
Thus, out of plane deformations were reported both for the perimeter walls and for all the
main arches bearing the dome in the central area of the monument, followed by a geometrical
deformation (and loss of initial shape) of the arches themselves. The lack of wooden or metallic
ties (typical structural elements for the Byzantine architecture) or other horizontal elements
591
connecting the vertical walls, pillars and piers should have played an important role towards
this pathology.

Figure 16. N-S transversal section in the narthex. The south wall, although reconstructed in 1896 (see pho-
to from A. Kambourakis collection), presents today a 90 mm out-of-plane deformation.

Moreover, the increase of crack openings with height was found more pronounced along the
transversal axis (N-S), than along the longitudinal one (E-W). Such a difference in behaviour is
usual in churches with an orthogonal plan and could be attributed to the larger number and sec-
tions of vertical elements available along the E-W axis in the main church (see Figure 12). This
behaviour was also noticed in the past, whereas previous interventions were applied with the
aim to alleviate this problem (external stone buttresses in the north, metallic trusses and con-
finement of piers in the south), as described in Section 4 (Figs 13-14). Although those correc-
tive measures were in the right direction (allowing the church to withstand the 20
th
century
earthquakes without local collapse), they were proven to be insufficient to prevent extensive
cracking of the monument.
The damages observed in the drum of the cupola may, therefore, be attributed to the (increas-
ing with height) tendency for out-of-plane deformations of the church. It should be reminded
here that the damages that made imperative the demolition and reconstruction of the cupola at
the end of the 19
th
century were of the same nature, as those observed now; this is proved by the
missing parts of the mosaics. As described in Section 3, most of the piers in the drum exhibited
out-of-plane deformations. Since the substructure on which the system of the cupola rests is de-
forming out-of-plane and the cupola itself (being very stiff) is practically non deforming, the
piers of the drum (being rather flexible out-of-plane) are called to follow the deformations of
the substructure.
Some final comments should be made here, regarding the foundation and foundation soil of
the monument and their effect on its behaviour. The recent pathology does not give signs of ma-
jor differential settlements that might have contributed to the damages of the monument. This is
also confirmed by geotechnical investigations conducted immediately after the 1999 earthquake
(O.T.M. 2000). On the other hand during the application of urgent measures local investiga-
tions revealed that rather shallow foundations (0.5m deep in the middle of South perimeter
wall, 1.25m deep in the N-E chapel) are provided to the perimeter walls.
Thus, in order to collect qualitative information on foundation elements and on eventual
empty underground spaces or local variations of the ground material characteristics, geophysi-
cal investigation was also undertaken, using mainly 3D seismic tomography (Polymenacos et al.
2005). This investigation suggested that, in the north-northwest part of the church and exonar-
thex, inferior ground material quality has to be expected, as low velocities were recorded.
Moreover, archaeological excavation in the region of the N-W chapel revealed that Troump had
probably noticed that the foundation were relatively shallow and has dictated their strengthen-
ing during the extended reconstruction of the upper part of the north wall in 1896. For this pur-
pose, a 2.0m deep wall was constructed under the NW perimeter wall of the church. To this end
extensive excavations had to be undertaken in that area, thus disturbing the foundation soil.
592
This fact could probably explain the lower velocities found with the seismic tomography. Nev-
ertheless, all the above information and eventual further investigations concerning the founda-
tion conditions of the church and exonatrhex have to be further examined during the design of
the interventions of the second phase.
7 NUMERICAL VERIFICATION OF THE PATHOLOGICAL IMAGE
One of the major steps of the assessment of an existing structure is the analytical reproduction
of its pathological image. In the case of the Katholikon, preliminary linear parameter analyses
were performed (Miltiadou-Fezans et al. 2004), as a means for selecting adequate emergency
interventions. Those analyses provided a quite satisfactory numerical verification of the patho-
logical image of the monument. For the preliminary analytical study, using the computer code
ACORD, the structure was modeled by shell elements (Fig. 17), whereas the mechanical prop-
erties of elements belonging to various parts of the structures were assumed on the basis of the
available data for the construction materials. Linear elastic analyses were performed for various
combinations of actions (self weight alone or combined with seismic action). Both static and
dynamic analyses were performed.
Figure 18 shows the calculated stresses for the inner face of shell elements, due to vertical
loads. Irrespectively of the accuracy of numerical values of stresses, one may clearly distin-
guish the vulnerabi-lity of the region of arches and domes (especially in the west part), even for
the self-weight of the monument alone. As expected, tensile stresses are developed in the apex
of several arches, in the groin vaults of the narthex, in the base of the cupola and that of its
drum, as well as in the four squinches. The obtained results are conform with the qualitative in-
terpretation of the pathology of the monument and they prove the inherent vulnerability of the
structural system, in which (a) a stiff central cupola is resting (through the drum) on four major
arches parallel to the two main axes of the church (see Figs. 3-4), as well as on four arches ob-
lique in respect with the longitudinal and the transverse axis, (b) the vertical, as well as the ho-
rizontal component of the self weight of the whole system of (intersecting) groin vaults ar-
ranged around the central dome, are transferred to rather flexible stone masonry piers, and (c)
there are not currently ties in the arches and vaults, or other elements to confine critical struc-
tural elements or link the various parts among them. Therefore, the tendency of the structural
system to deform laterally in its upper region is expected to be apparent even under self-weight
alone.

Figure 17. One of the finite elements models pre-
pared for parameter analyses

Figure 18. Inner face of shell elements. Tensile
stresses due to self-weight of the structure.

This working hypothesis seems to be confirmed by the deformed shape of the structure in-plan,
shown in Figure 19. One can even observe the more pronounced lateral deformation along the
N-S axis of the monument, as well as the out of plane deformation of the perimeter walls of the
593
narthex, which historically has been proved to be a vulnerable part of the structure, already re-
constructed in 1896 and deformed again as measured after the 1999 earthquake. On the same
Figure, the excessive out-of-plane deformation of the exonarhtex is also apparent.
Expectedly, this behaviour is deteriorating when a seismic event occurs. In Figure 20, the out of
phase movement of the east and west parts of the monument is shown. Such a movement can
explain the severe damages occurred to the drum of the dome, as well as to the arches and
vaults supporting the dome. In general, the analyses for loading combinations including the
seismic action have shown a critical concentration of tensile stresses in arches at various levels,
as well as in the piers of the drum. In addition, extensive damages in vertical elements (masonry
in the perimeter of the monument, as well as piers) were confirmed.

Figure 19. Deformed shape of
the monument in plan under
self-weight
Figure 20. Deformed shape of the structure looking from the N. Dy-
namic analysis; seismic action along the longitudinal axis.

Furthermore, in the framework of this preliminary work, the plots of principal tensile stresses
were compared with the respective drawings on which observed cracks were reported (Miltia-
dou-Fezans et al. 2004). This comparison proved to be quite satisfactory, as, in general, the ob-
served crack pattern (location and inclination of cracks) seems to be confirmed by the analytical
results in all regions of the monument.
Although this preliminary analytical work allowed for a better understanding of the structural
behaviour of this important monument, it was judged that a more accurate assessment of its
seismic behaviour was of primordial importance for the design of optimum interventions.
8 STRUCTURAL RESTORATION SCHEME
The pathological image of the monument (both the historical and the current ones), as well as
the fact that the Katholikon is built in a tectonically active area, suggest that in addition to in-
terventions that are necessary for the repair of observed damages, adequate measures should be
taken with the aim to alleviate the inherent vulnerability of the structure and, thus, to improve
its future behaviour in some critical regions. Due to the high values of the monument, its vulne-
rability and the fact that interventions should not drastically alter the initial structural system,
the decision was taken to investigate thoroughly its structural behaviour. To this end additional
data were necessary in order to avoid conservative assumptions (both for actions and resis-
tances). Such assumptions would lead to extensive interventions that might not be needed and
that would inevitably alter the architectural value of the monument.
On the basis of the aforementioned reasoning, it was decided to implement the structural res-
toration works in two phases, thus giving the possibility for these additional data to be col-
lected. The first phase of works comprises all those considered necessary to achieve the better
possible repair and strengthening of masonry elements (mainly stitching and deep re-pointing
where necessary, systematic grouting injections, local reconstructions, etc). The second phase
concerns the various strengthening interventions that will be designed and selected as optimum,
in order to improve the overall behaviour of the whole building (such as installation of ties, di-
aphragmatic structures in the extrados of the vaults and the exonarthex, etc). Due to the type
and extend of damages, it was judged as absolutely necessary the first phase of interventions to
be applied in priority, in order to avoid further deterioration of damages, which could lead to a
total disruption of continuity and even local collapses, in case of further seismic events.
594
Besides, during this first phase of works a better structural survey of invisible parts of the
monument (internal face of masonry elements, extrados of vaults) could be possible.
The implementation of the first phase of structural restoration interventions has been now
accomplished, together with the most of the research and investigations undertaken to support
both phases of works, briefly presented below. Preliminary proposals for the second phase of
interventions have also been approved, but their final design is still under elaboration.
9 DESIGN OF GROUTS AND INVESTIGATION OF MASONRY BEHAVIOUR BEFORE
AND AFTER GROUTING
9.1 Design of alternative grout compositions and tests on cylinders simulating the infill
material
The design of high injectability grouts was carried out on the basis of the methodology pro-
posed by Miltiadou & Tassios (2006). First, were taken into account the performance require-
ments deriving from the structural restoration study (Miltiadou et al. 2003b), and then the fol-
lowing target values were set for the basic mechanical properties of the grouted masonry:
tensile strength approximately double than that one of the masonry before grouting, and com-
pressive strength approximately equal to 3.0 MPa.
On the basis of the available literature (Vintzileou & Tassios 1995, Tassios 2004), it was es-
timated that the compressive strength of the grout at the age of six months should lie between
6MPa and 10MPa; a grout flexural strength of the order of 2 to 3MPa was required. In addition,
the physical-chemical properties of the raw materials should be selected in a way that the dura-
bility of the structure and its precious mosaics would not be jeopardized. Finally, the grouts
should have high injectability capacity, so that, under low pressure (~0.075 MPa), they enter
and fill fine voids and cracks, with a nominal minimum width (W
nom
) equal to two tenths of
millimeter. Two main categories of grouts could satisfy injectability, strength and durability re-
quirements: (i) ternary grouts composed of lime, pozzolan and a low cement content (30%) and
(ii) natural hydraulic lime based grouts.
Thus, various grout mixtures, belonging to the above two categories, were designed and
tested (in order to assess their physical, chemical and mechanical properties) at the laboratory
of the Directorate for Technical Research on Restoration of the Hellenic Ministry of Culture
(DTRR/HMC). The main results of the research are reported in Kalagri et al. (2007) and Mil-
tiadou-Fezans et al. (2007).

Table 1. Mechanical and injectability characteristics of the grouts selected to be injected in the wallettes
Grout properties Ternary (1) Nhl5-based grout (2)
T36 (sec)
Sand column 1.25/2.50 mm
(W
nom
~ .0.2 mm)
19 22.5
Bleeding % 2 3
App.viscosity t
d=4.7
(sec) 20.5 22
Compressive and flexural
strength (MPa)
Age (days)
28 90 180
f
gc
f
gt
f
gc
f
gt
f
gc
f
gt
Ternary 4.08 2.11 8.16 2.29 10.6 3.13
NHL5-based grout 2.82 2.47 4.50 2.52 6.36 3.87
(1) Ternary: 30% white Danish cement, 25% Lime (powder), 45% natural pozzolan Petr. (<45m), su-
perplasticer 1%, water 80% of the solid phase of the grout
(2) NHL5-based grout: 100% NHL5 (St Astier), superplasticer 1%, water 80% of the solid phase of the
grout

In order to determine injectability characteristics, the penetrability, fluidity and stability of the
suspensions were fully examined in various water/solids ratios, with or without superplastisizer.
The compositions presenting satisfactory injectability capacity were further tested to evaluate
their behavior to salt decay and estimate their mechanical characteristics (compressive and
flexural strength). Furthermore, six alternative grout formulations presenting similar injectabili-
595
ty were injected into twenty eight cylindrical specimens, simulating the infill material of three-
leaf stone masonry. The cylinders were then subjected to compression in different hardening
ages (Kalagri et al. 2007). After comparative evaluation of the results, two grout compositions
(a ternary grout and a NHL5-based grout) fulfilled simultaneously the injectability, the strength
and durability requirements (Table 1).Therefore, they were selected to be applied to six three-
leaf stone wallettes, simulating the masonry of the upper parts of the monument. It has to be
noted however that the hydraulic lime based grout presented a better sulphate resistance than
the ternary one (Miltiadou-Fezans et al. 2007).
9.2 Construction and testing of wallettes
In order to estimate in the best possible way the mechanical characteristics of the masonry be-
fore and after grouting, considerable effort was devoted to the identification of the construction
type of masonry. For this purpose, radar and endoscopy were applied in a systematic way. The
in-depth geometry of the perimeter stone masonry was rather accurately identified (Vintzileou
et al. 2004). As anticipated, both the lower and upper parts of the perimeter walls belong to the
three-leaf masonry type, with varying widths of the leaves. Although both regions were ex-
amined, the investigation was concentrated mainly on the upper (more vulnerable) zone of ma-
sonry. In Figure 21 characteristic examples of vertical sections of the lower and upper zone of
perimeter masonry are shown.

(a) (b)
Figure 21. Vertical sections of lower (a) and upper zone Figure 22. The wallettes prepared for grouting

On the basis of the materials characteristics (stones, bricks and mortars) determined in the La-
boratory on samples taken from the monument and the type of masonry determined in-situ, an
attempt was made to estimate masonrys compressive strength using empirical formulae availa-
ble in the literature. The calculated strength values were unacceptably scattered, thus making
unsafe or over-conservative the selection of any of the calculated compressive strength values
(Vintzileou 2002).
Thus, the decision was made to construct wallettes and to determine basic mechanical prop-
erties by testing them before and after grouting (Vintzileou et al. 2006, Miltiadou et al. 2006).
The geometry of the wallettes was chosen to simulate the upper and more vulnerable part of pe-
rimeter masonry. In order to avoid scale effects a scale 2:3 was selected. Six three-leaf stone
masonry wallettes were constructed using materials of similar characteristics as the in-situ ones.
Half of the wallettes were tested in compression and the other half in diagonal compression up
to their maximum resistance. After unloading, the wallettes were grouted using one of the se-
lected grouts of the Table 1 (Fig. 22). The injections were carried out using a specific method-
ology and the total grout consumption (calculated as the ratio of grout volume consumed for
each wallette per total volume of wallette) was of the order of ~10%. Subsequently, they were
tested again in compression or in diagonal compression to failure.
Both grouts applied to the specimens were able to achieve homogenisation of masonry by
filling cracks and voids of wallettes. Thus, the wallettes exhibited substantial improvement of
their behaviour, in terms of compressive strength, tensile strength and reduction of the separa-
596
tion between the three leaves of masonry, without substantial increase in their stiffness, and
proved to be efficient from the mechanical point of view. The main results are summarized in
Table 2.

Table 2. Basic mechanical properties of Wallettes before and after grouting
W f
w0
(MPa) f
ws
(MPa) f
ws
/ f
w0

v0
()
vs
() E
0
(GPa) E
s
(GPa) E
s
/ E
0

1 1.82 3.00 1.65 * -1.76 1.00 1.20 1.20
2 1.74 3.75 2.16 -1.6 -2.50 1.44 1.55 1.08
3 2.26 3.73 1.65 -2.25 -3.39 1.50 1.30 0.87
f
t,0
(MPa) f
t,s
(MPa) f
t,s
/f
t,0

4 0.1 0.21 2.10
5 0.1 0.33 3.30
6 0.1 0.22 2.20
W2 and W5 injected with the ternary grout;
W1, W3, W4, W6 injected with the hydraulic lime based grout

From the two alternative compositions, the natural hydraulic lime based grout was selected for
the application to the Katholikon of Dafni Monastery, due to the substantial (compressive and
tensile) strength enhancement of wallettes, the rather ductile behaviour under diagonal com-
pression (compared to that of masonry grouted with the ternary grout), and the better durability
properties (Miltiadou-Fezans et al. 2007) that contribute to the protection of mosaics and fres-
coes.
9.3 Optimum grout composition
In order to further improve the hydraulic lime based grout, the addition of fine natural pozzolan
(d
max
<45m) in various proportions was investigated. The addition of a small percentage of
pozzolan (10%) was decided, on the basis of additional data, deriving from porosity measure-
ments, salt durability tests and from in situ pilot trials.

Table 3. Optimum grout composition and injectability characteristics measured
in the laboratory and in situ at the first pilot preparation
Grout composition
NHL5 (St Astier) 90%
Pozzolan Petrotechniki 10%
Superplasticizer (1), (2) 1%
Water (1) 80%
Grout properties In lab In situ
T36 (sec) Sand column 1.25/2.50 mm
(W
nom
~ .0.2 mm)
19 - 22
Bleeding <1% 1%
Apparent viscosity - t
d=4.7
(sec) In lab In situ
0 min after mixing 21 22
60 min after mixing (agitated) 23 25
Apparent density (gr/cm
3
) In lab In situ
0 min after mixing 1.5050 1.4978
60 min after mixing (agitated) 1.4986 1.4870
(1) % of the solid phase of the grout
(2) superplasticizer based on polycarboxylic ether

597
The mix proportions and the injectability characteristics of the optimum grout composition are
presented in Table 3, whereas more detailed information is given in Miltiadou-Fezans et al.
(2007, 2008). This optimum grout composition was used for the repair of the monument. More-
over it was used for the repair of a large scale model of a Byzantine groin vaulted structure,
constructed and tested on a earthquake simulator, as briefly presented below.
9.4 Construction and testing of a large scale structure covered with a byzantine groin vault
As aforementioned in Section 2, byzantine groin-vaults are used for covering the most of the
parts of the Katholikon of Dafni Monastery, including the arms of the cross, where severe dam-
ages have been occurred. Thus, important effort has been undertaken for achieving a better
knowledge of the dynamic behaviour of such structures. An experimental research was carried
out by DTRR/HMC in collaboration with the Laboratory of Earthquake Engineering of NTUA.
A model of a byzantine groin-vaulted structure bearing locally mural mosaics was constructed.
The materials and the construction type used for its masonry walls were exactly the same with
those used for the construction of the wallettes, whereas bricks and mortar were used for the
construction of the arches and the groin vault (Fig. 23).

Figure 23. The groin vaulted structure on the seis-
mic simulator before grouting.
Figure 24. The groin vaulted structure after grout-
ing

It has to be noted that the groin vault was built without any formwork, following the traditional
way of byzantine masons, as reported by Delinikolas et al. (2003). The total dimensions of the
model are in plan 2.70m x 2.60m; the thickness and the height of the walls are 0.45m and
2.60m respectively, whereas the thickness of the vault in its centre is 0.20m. The total height of
the model is 2.85m. These dimensions were selected taking into account the limitations im-
posed by the capacity of the seismic simulator.
The model was tested after nine months from its construction by imposing seismic loads
gradually increasing, until rupture. Then local application of grouts and installation of ties to
the arches was performed and after a suitable period of time the tests were repeated, until rup-
ture. The model was then injected with the final grout composition (Table 3) to homogenize the
whole structure, following the same methodology with that used for the injection of the wal-
lettes and the injection of the monument itself (Miltiadou-Fezans et al. 2006, 2008). Tests were
again repeated until rupture. In Figure 24 the model after the grouting application is shown. As
expected, it was observed that the dynamic characteristics of the model were changed after the
application of grouting to the whole structure, and the model could suffer stronger base mo-
tions. The results of this experimentation are still under elaboration and are going to be pre-
sented in a separate paper. An attempt will be made to correlate these results with those of the
wallettes and the in situ seismic monitoring.
10 BRIEF PRESENTATION OF THE WORKS OF THE FIRST PHASE
As reported in the relevant study, (Mitiadou-Fezans et al. 2003b), the main objective of this
598
first phase of structural restoration interventions was (i) to improve in the best possible way the
mechanical behaviour of the masonry elements by reinstating their continuity (lost due to the
numerous cracks) and increasing their resistance (mainly to traction and shear), without side ef-
fects due to possible durability matters. This objective had to be achieved taking into account
the existence of mural mosaics, frescoes and old pointing mortars, that had to be preserved and
conserved in situ. In fact all uncolored joints of Figures 12-13 still preserve the old Byzantine
pointing mortars.

Figure 25 Old mortars of first and second construction phase and cracked mosaics
that have to be preserved in situ.

This first phase of masonry repair interventions comprised mainly the following works (Fig.
26): i) very careful removal of plasters and deteriorated pointing mortars applied during pre-
vious interventions, without harming the old ones, adjacent to or underneath them, ii) removal
of tiles and other covering and filling materials to reach the extrados of all vaulted structures,
iii) stitching of the most severe cracks, using long stones, bricks, or thin titanium plates, iv) few
local reconstructions necessary either for the repair of dislocated or collapsed parts or for the
restoration of past morphological alterations, v) deep re-pointing where necessary and prepara-
tion of the masonry for injection grouting, vi) implementation of injection grouting, vii) remov-
al of all injection tubes, (viii) in situ conservation of all deteriorated old mortars using frescoes
techniques, ix) execution of all necessary works to ensure the protection of the extrados from
rainwater.

Figure 26. Typical examples of the first phase structural restoration interventions (Recent mortars remov-
al, grouting preparation, stitching of cracks, local reconstruction)

In parallel with the works for the masonry repair, the competent Conservators have carried out
all the necessary works for the in situ conservation (including grouting) of severely damaged
mural mosaics, (Chryssopoulos et al. 2003). As anticipated, in most cases, the cracks on the
masonry affected also the mosaics beard on its internal face (Fig. 27).
From the above brief presentation of the works undertaken, it is ensued that the application of
adequate mortars and grout compositions, constitute a key parameter in order to ensure a suc-
cessful intervention, both for the masonry and the mosaics.
Thus, extended laboratory and in situ pilot tests have been undertaken for the formulation of the
adequate mortars compositions taking into account the characteristics of the existing materials,
599
the requirements set by the architectural and structural study and the worksite conditions. For
the majority of the re-pointing works adequate lime-pozzolan based mortars were used. For the
local reconstructions, deep re-pointing of extremely damaged critical areas and stitching with
thin titanium plates, hydraulic lime based mortars were applied. Detailed information concern-
ing the properties of existing mortars is given in Papagianni (2002), while all the results con-
cerning both the properties of existing materials and those designed and applied to the monu-
ment are reported in Anagnostopoulou & Miltiadou-Fezans (2007). In order to ensure the
protection and in situ conservation of old pointing mortars, on the external facades, the upper
central area just below the dome, and the groin vaults of the sanctuary, the deep re-pointing
works have been executed by experienced Conservators.

Figure 27. Typical examples of in situ mural mosaics conservation
works (including installation of fine tubes for grouting)

As anticipated special attention has been given to the design of the adequate grouts, not only
because grouting constitutes the main of the works undertaken that can substantially improve
the mechanical behaviour of the masonry, but also due to the fact that it is an invisible and irre-
versible intervention, which affects the masonry elements and all the mosaics, frescoes and old
mortars beard on them. Furthermore, as the proper design of a grout composition cannot ensure
on its own the successful execution of the grouting intervention, a specific application metho-
dology has been developed and applied. The most important aspects of this methodology are
reported in Miltiadou-Fezans et al. (2008). This application methodology permitted the imple-
mentation of injections to this important monument, bearing mosaics, frescoes and old mortars
that had to be preserved in situ, in a more rational and fully controlled way. The monitoring re-
sults and those obtained by sonic tomographies summarized below, gave important information
for the effect and the efficiency of the whole grouting intervention, and enhanced the validity of
the methodology applied.
During the whole project, all interventions have been documented in a detailed way (ade-
quate as- built drawings have been elaborated), together with all the new findings concerning
the materials, types of construction, past interventions and pathology of various invisible ele-
ments (i.e extrados of groin vaults). Thus the historical and recent pathology drawings have
been updated and all necessary information has been collected to be taken into account in the
study of the second phase interventions.
10 SEISMIC MONITORING SYSTEM
As aforementioned, in order to increase our understanding of the seismic structural response
and to decrease the uncertainty of the seismic action, the installation of an earthquake monitor-
ing system was decided. The system was installed on the central core of the monument at 2003,
after the implementation of the aforementioned emergency measures, and before the beginning
of restoration works.
Thus, when the hydraulic lime grouting application started (Miltiadou-Fezans et al. 2008), in
June 2006, the system was functioning and had already recorded two weak earthquakes. The
first took place before the beginning of masonry repair (26/9/2004) and the second during the
preparation of masonry for grouting by stitching of cracks and local deep re-pointing
600
(8/1/2006). After three and four months from the end of the first grouting period, during which
grouting was implemented up to the springing level of groin vaults, two more weak earthquakes
were recorded (5/11/2006 and 12/12/2006). The analyzed data, have shown that the monitoring
system installed can give important information for the overall behaviour of the structure, be-
fore during and after the implementation of various interventions. Detailed information is given
in Mouzakis et al. (2008).

Figure 28. Variation of the period of the first mode

As shown in Figure 28, during the earthquake of 08/01/2006 occurred after the urgent measures
and while preparing the masonry for grouting, the natural period of the first mode was in-
creased, and hence the stiffness of the structure was decreased. This is totally expected, as only
local deep re-pointing and stitching of some cracks had been applied at that moment, while the
emergency measures have not been designed to undertake seismic actions. Moreover an in-
crease of damping was recorded. These results may indicate a slight deterioration of damages.
On the contrary, as a result of the implementation of grouting injections up to the springing
level of groin vaults, the period and the damping of the first mode were decreased, along the
EW and the NS directions. This result was reached by analysing the data of both seismic
events, happened three and four months after the first grouting period. Thus, it was proved that
the effect of interventions and more specifically of grouting on the overall behaviour of the
structure can be evaluated by a seismic monitoring system. This promising finding led to the
decision to install in 2007 complementary instrumentation, with the aim to cover the whole
structure.
The extended seismic monitoring system has recently recorded two more earthquakes
(6/1/2008 and 30/1/2008), and the data are still under processing. These recent seismic events
took place after nine and ten months from the end of masonry repair works (28/4/2007), during
which grouting has been carried out to the entire structure below the level of the base of the
domes drum. Thus, it is expected that the seismic monitoring records will give further informa-
tion for the effect and efficiency of all the interventions undertaken until now.
11 APPLICATION OF NDT FOR MAPPING MOSAICS SUBSTRATA AND GROUTING
MONITORING
In 2005, an exhaustive high frequency ground-penetrating radar survey has been realized on the
fifty main mosaics of the Katholikon of Dafni Monastery, with the aim to investigate the possi-
bility of GPR application to locate doubtful zones (delaminations, changes of substrata mortar
or other buried heterogeneities, etc), related to the mosaics very near bearing-structure. As pre-
sented in Cte et al. (2004, 2008), a methodology has been developed for the execution of spe-
cific parallel radar profiles, the processing of data and the construction of gray-color scale maps
from chosen trenches of the structure, related to the amplitude of the echoes in the selected
trench. These gray-scale maps were correlated to a level of detachment or other heterogeneities
by comparison with manual-sonic maps and pathology ones realized by the Conservators of the
mosaics (see Fig. 29). It has been shown that the GPR maps, giving a more detailed and less
subjective qualitative evaluation of mosaics substrata, can be a very useful tool for the Conser-
vators in order to locate doubtful zones in a more refined way. Furthermore, the possibility of
PERIOD OF STRUCTURE
0.27
0.22
0.20
0.30
0.32 0.39
0.26 0.26
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
26/9/2004 8/1/2006 5/11/2006 12/12/2006
P
e
r
i
o
d

i
(
s
e
c
)
West - East South - North
601
using GPR procedures to monitor the movement of the grout behind the mosaics, in real time,
during injection was also investigated; detailed information is given in Cte et al. (2008).

Figure 29. Example of manual-sonic
and sub-surface radar maps.
Figure 30. Tomographical reconstruction in a grouted and in a
non grouted area.

Investigations have been also realized to control the grouting effect into the masonry walls, us-
ing seismic techniques, including some sonic 2D travel time tomographies. The objective here-
in was to propose a simple survey methodology which gives information for the mechanical
characterization of the materials inside the structure, before and after grouting. The processing
of the results after the completion of grouting is still in progress. In Figure 30 two indicative
tomographical reconstructions of a grouted and in a non grouted area are presented. The com-
parison of these two cases permits to conclude that the grouting survey of such masonry struc-
tures may be achieved by the observation of the velocities of their inner part (Cte et al. 2008).
12 CONCLUSIONS
The high values of the atholikon of Dafni Monastery and, hence, the need for accurate infor-
mation to serve the design of optimum structural interventions led to the adoption of a step by
step multidisciplinary approach, both concerning the design and the implementation of the
structural restoration interventions. This approach proved to be very efficient, as it gives the
possibility to perform the in situ and laboratory investigations that are necessary for the design
of the next step, as well as for the evaluation of the previous ones.
In this framework, a series of quite novel investigations were undertaken; the most important
results are summarized below.
1. The holistic design of hydraulic grouts, carried out on the basis of rational criteria, laboratory
and in situ pilot tests, led to the improved knowledge on the hydraulic lime based grouts and,
hence, to the selection of appropriate mixtures.
2. The experimental assessment of the mechanical characteristics of three-leaf stone masonry
(before and after grouting) contributed to better understanding of the behaviour of this type of
structures and proved that their repair with adequately designed highly injectable hydraulic
grouts (hydraulic lime based ones or ternary grouts) can be very efficient.
3. A hydraulic lime based grout was considered as optimum composition to be applied in situ,
on the basis of injectability, durability, strength and deformability characteristics.
4. Testing a scaled model of a byzantine groin-vaulted structure on an earthquake simulator (be-
fore and after grouting) provided valuable information regarding the structural behaviour of this
complex type of structures, as well as regarding the strengthening capacity of injections.
5. The developed methodology for grouting application gives the possibility for a more rational
and fully controlled implementation of injections, not only to ordinary structures, but also to
important monuments bearing mosaics and frescoes, that have to be preserved in situ.
6. The entire procedure for the design and application of adequate lime-pozzolan mortars en-
hanced our knowledge on this kind of mortars, traditionally used in the past.
7. The monitoring system, installed to record the seismic behaviour of the monument, provides
valuable data on the effect of interventions, as they are applied in steps to the monument. Thus,
the effects of interventions to various characteristics of the monument, as well as their effi-
ciency can be continuously assessed. It should also be noted that, a similar monitoring system
602
having been installed in another important monument (the Katholikon of Osios Loucas Monas-
tery), the comparative evaluation of data obtained in the two churches will allow a better under-
standing of the structural seismic response of middle-Byzantine churches of octagonal type.
8. The potential of NDT was further enhanced to serve the needs of investigating important
monuments. High frequency GPR combined with endoscopy proved to be efficient for the iden-
tification of the masonry construction type. High frequency GPR contributed also to locate un-
safe regions of mosaics substrata. Furthermore, sonic techniques, undertaken before and after
grouting, were proven very efficient in checking the effect and efficiency of grouting interven-
tion, whereas geophysical investigation by means of 3D seismic tomography provided valuable
information about the foundation elements.
The entire plan of investigations has supported the rational design and implementation of the
first phase of works, as well as the evaluation of their effect and efficiency.
The collected results are now being used for the design of the second phase interventions. A
synthesis of all data provided by the situ surveys during the first phase of interventions, the re-
sults of the monitoring system and those obtained by testing the wallettes and the groin vaulted
structure is under elaboration. Analytical work will be extended, with further development and
calibration of detailed numerical models. This/these calibrated model/models will then be used
to check the efficiency of various strengthening intervention techniques (installation of ties,
metallic confinements, diaphragmatic structures in the extrados of the vaults and the exonar-
thex, etc). Thus, the final design of the optimum interventions will be carried out and the re-
spective works will be implemented during the second and final phase of restoration works.
Last but not least, it is strongly believed that the experience and knowledge acquired by all
levels of involved stuff in the framework of the structural restoration of this important monu-
ment of World Heritage constitute a solid base for similar investigations, studies and structural
restoration interventions to other historical structures and monuments.
ACKNOWLEDGEMENTS
The contribution of Ass. Prof. E. Vintzileou for the critical reading of the text is gratefully ac-
knowledged. The close collaboration of the Scientific Committee with the personnel of the 1
st

Ephorate of Byzantine Antiquities, the Directorate for the Restoration of Byzantine and Post
Byzantine Monuments, the Directorate for Technical Research on Restoration and the Directo-
rate for the Conservation of Ancient and Modern Monuments has also to be mentioned.
The project was included in the Operational Program CULTURE. It was co-funded by the
European Regional Development Fund (ERDF-75%) and by National Funds (25%).
REFERENCES
Anagnostopoulou S., Miltiadou-Fezans A. 2007. Design of restoration mortars of the Katholikon of Dafni
Monastery. Internal report, DTRR, Hellenic Ministry of Culture (in Greek)
Bouras Ch. 1998. The Dafni Monastic Complex Reconsidered, S Studies in honor of Cyril Mango,
B. G. Teubner Stuttgart und Leipzig.
Chryssopoulos D., Anamaterou L., Georganis F. 2003. Documentation study for the mosaics of the Katho-
likon of Dafni Monastery after the 1999 earthquake. Hellenic Ministry of Culture (in Greek).
Cte Ph., Drobert X., Miltiadou-Fezans A., Delinikolas N. 2004. Mosaic-grouting monitoring by ground-
penetrating radar. Proc.IV Int. Sem. Struct. Anal. of Hist. Const., Padova, Vol. 1, pp 401-406.
Cte Ph., Drobert X., Miltiadou-Fezans A., Delinikolas N. Durand O., Alexandre J. Kalagri A., Savvidou
M, Chryssopoylos D., Anamaterou L, Georganis, F. 2008. Application of non-destructive techniques
at the Katholikon of Daphni Monastery for mapping the mosaics substrata and grouting monitoring.
Sub. to the 6
t
intern. conf. on Structural Analysis of Historical Constructions, Bath.
Delinikolas, N., Miltiadou-Fezans, A., Tsofopoulou, E., Minos, N., Chrissopoulos, D. 1999. Initial In-
spections, Assessment and Proposals for Protecting the Monastery of Dafni, Internal report. Hellenic
Ministry of Culture (in Greek).
Delinikolas N., Miltiadou-Fezans A., Chorafa E., Zaroyianni E. 2003. Study on restoration of the Katho-
likon of Dafni Monastery: Phase A - Architectural and historical Survey. Hellenic Ministry of Culture
(in Greek)
603
Georgopoulos A. 2003. Novel survey and digital photogrammetry methods for geometrical and architec-
tural digital survey of Byzantine monuments. The case Dafni Monastery. Research Report, NTUA and
Hellenic Ministry of Culture.
Kalagri A., MiltiadouFezans A., Vintzileou E. 2007. Design and evaluation of hydraulic lime grouts for
the strengthening of stone masonry historic structures. Proc. int. symp. on Studies on Historical Heri-
tage, Antalya, Turkey, 2007.
Mariolakos I., Fountoulis I., Andreadakis E. 2000. Engineering geological problems caused by human in-
terference on monuments that influence their seismic behavior. Compatible materials recommenda-
tions for the preservation of European cultural heritage. Ac. Civil Engineers of Greece.
Millet, G., 1899. Le Monastre de Daphni. Histoire, Architecture, Mosaiques, Monuments de lArt By-
zantine I, Paris.
Miltiadou-Fezans A., Tassios T.P., Delinicolas N., Chorafa E., Zarogianni E., Chandrinos I. 2003a.
Earthquake structural problems and urgent measures undertaken to support the Katholikon of Dafni
Monastery in Athens, Greece. Proc.8th inter. conf. STREMAH 2003, Halkidiki, Greece, 7-9 May.
Miltiadou-Fezans A., Delinikolas N., Chorafa E., Zaroyianni E. 2003b. Study on restoration of the Katho-
likon of Dafni Monastery, Phase A - Structural Survey, Analysis and remedial measures. Hellenic
Ministry of Culture (in Greek).
Miltiadou-Fezans A., Vintzileou E., Delinikolas N., Zaroyianni E., Chorafa E. 2004. Pathology of the
Dafni Monastery: survey, monitoring of cracks, interpretation and numerical verification. Proc. 4th in-
ter. Sem. on Structural Analysis of Historical Constructions, Vol. 2, Padova.
Miltiadou-Fezans A., Vintzileou, E., Papadopoulou, E., Kalagri, A. 2006. Mechanical properties of three-
leaf stone masonry after grouting. Proc. 5
th
inter. Sem. on Structural Analysis of Historical Construc-
tions, New Delhi, India.
Miltiadou-Fezans A., Tassios T.P. 2006. New rational criteria for the holistic design of hydraulic grouts.
1st Conf. of the Society for Research and Promotion of Scientific Restoration of Monuments (in
Greek)
Miltiadou-Fezans A., Kalagri A., Delinikolas N. 2007. Design of hydraulic grout and application metho-
dology for stone masonry structures bearing mosaics and mural paintings: the case of the Katholikon
of Dafni Monastery. Proc. int. symp. on Studies on Historical Heritage, Antalya, Turkey, 16-21 Sep-
tember, 2007.
Miltiadou-Fezans A., Kalagri A., Kakkinou S., Ziagou A., Delinikolas N., Zarogianni E., Chorafa E.
2008. Methodology for in situ application of hydraulic grouts on historic masonry structures. The case
of the Katholikon of Dafni Monastery. Submitted to the 6
th
intern. conf. on Structural Analysis of His-
torical Constructions, Bath, UK..
Mouzakis Ch., Miltiadou-Fezans A., Touliatios P., Delinikolas N., Dourakopoulos J. 2008. Earthquake
based condition monitoring of the Katholikon of Dafni Monastery. Submitted to the 6
th
intern. conf. on
Structural Analysis of Historical Constructions, Bath, UK..
O.T.M. 2000. Geothechnical Investigation of Dafni Monastery. Internal Report, Hellenic Ministry if Cul-
ture (in Greek).
Papagianni. I. 2002. Investigation of physicochemical characteristics of the historic mortars of the Katho-
likon of Dafni Monastery. Internal Report, AUTH and Hellenic Ministry of Culture (in Greek)
Polymenakos L., Papamarinopoulos S., Miltiadou A., Charkiolakis N 2005. Investigation of the founda-
tions of a Byzantine church by three-dimensional seismic tomography. Elsevier Journal of applied
geophysics 57, pp 81-93.
Stikas E. 1962. Restoration and consolidation of the exonarthex of Dafnis Katholikon, DChAE ser. 4, 3,
1962-1963.
Tassios T.P. 2004. Rehabilitation of three-leaf masonry. In Evoluzione nella sperimentazione per le
costruzioni, Seminario Internazionale 26 Settembre- 3Ottobre, CIAS.
Troump E. 1896. Quelques vielles glises byzantines de la Grce moderne, Marseille, 5-11, 11-15.
Vintzileou E., Tassios T.P. 1995. Three leaf stone masonry strengthened by injecting cement grouts.
Journal of Structural Engineering, ASCE, Vol. 121, No5
Vintzileou E. 2002. Methods for documenation of pathology and assessment of strengthening measures in
Byzantine churches. Application in Dafni Monastery. Research Report, NTUA and Hellenic Ministry
of Culture (in Greek).
Vintzileou E., Miltiadou-Fezans A., Palieraki V., Delinikolas N. 2004. The use of radar techniques and
endoscopy in investigating old masonry: the case of Dafni Monastery. Proc. of 4
th
intern. sem. on
Structural Analysis of Historical Constructions, Padova Vol. 2, pp. 351-360
Vintzileou E., Miltiadou-Fezans A., Vrouva A, Anagnostopoulou S. 2006. Mechanical properties of three-
leaf stone masonry. Proc.5
th
intern. Conf. on Structural Analysis of Historical Constructions, New
Delhi, India.
604
On the Dynamic Behavior of a Lightweight Isolator for Museum
Artifacts

Vlasis Koumousis
ABSTRACT: In earthquake prone areas, a low cost solution for the protection of museum arti-
facts is to seismically isolate unique pieces of great importance exhibited in existing museums
safely built or strengthened according to existing codes. The design of seismic isolation system
can be based on any available type of seismic isolator, but also on lightweight isolators specially
developed for museum object protection. The proposed lightweight system determines a nonlin-
ear response based on a cubic order restoring force and a quadratic damping force. This leads
into two uncoupled Duffing-Van der Pol type equations. For a specific artifact, the optimal pa-
rameters of the system that reduce the absolute acceleration, transmitted to the object, within
certain limits and do not exceed the given displacement capacity are determined. The behavior
of the system for a set of earthquake excitations is examined and the main results are presented.
1 INTRODUCTION
To avoid overturning and provide adequate stability and strength for a slender artifact, such as a
medium scale statue, or a big pottery piece, in the case of strong ground motion, one has either
to use conventional design methods aiming at increasing the strength, which in the case of
monuments and especially artifacts is prohibited or limited, or to rely on seismic isolation
Inman (2001), Naeim and Kelly (1999). This aims at reducing the seismic demand on the arti-
fact, which is more efficient as the structure remains intact and the seismic energy is absorbed at
a lower level, where the engineer can work in a reversible way to control the transmitted excita-
tion to the superstructure.
There are several applications of seismic isolation systems that protect artifacts against earth-
quakes all over the world, but mainly in Japan and United States. The Gates of Hell by Rodin
at the National Museum of Western Art in Tokyo, a Fountain of Neptune in Messina-Italy and
several applications at the J.P. Getty Museum in California, where the development of these
systems was initiated, are some well known applications of seismic isolation to museum objects.
In Greece a similar application of a seismic isolation system was installed for the statue of Her-
mes during the 2004 renovation of the new museum of Olympia. There exist also lightweight
isolators usually of sliding type that are used to protect artifacts of relatively small weights.
These type of isolators where developed initially for the isolation of consoles of main frame
computers, Agbabian (1991). At the J.P. Getty Museum the archeologists, conservators and
technicians were the first to develop these devises for museum artifacts in different variants.
Today there exist similar systems around the world and in Japan, Oiles, Shimizu Corporation
and Mitsubishi among others design and produce lightweight isolators.
2 LIGHTWEIGHT ISOLATORS
Lightweight systems, that are adequate for artifacts of the order of 10kN weight, are sliding sys-
tems with recentering capabilities with, or without additional damping devices. The most simple
one, designed by Shimizu Corporation and manufactured by Mitsubishi in Japan (Fig. 1), con-
sists of three square metal plates, the bottom one levelled and fixed at the floor and the other
two sliding in rails-linear guides in two perpendicular directions, accommodating any horizontal
605
movement. The restoring forces are provided by elastic springs acting in the direction of motion,
whereas damping devices may be added in each direction to enhance the overall behaviour. The
protected mass of the artifact is anchored at the top plate.

Figure 1: Shimizu Isolator type 1- manufactured by Mitsubishi
(courtesy of Dr. T. Tazo, Shimizu Corporation)

The system developed at the Getty Museum in the seventies, and underwent major technical
improvements since then, utilizes the same concept, but arranges the springs in a perpendicular
direction of that of motion of each plate (Fig.2), Stavridis (2006).

Figure 2: Getty Museum Isolator
Thus, the restoring force is provided by the projected component of the acting forces as in Fig.
3. The one end of the spring is attached to the lower plate and the other is always kept perpen-
dicular to the direction of motion and rolls on a linear ramp formed at the upper plate.
This system has no additional damping device, but is capable to absorb energy through the
friction and the small impacts at the tip of the ramp as this goes back and forth in a seismic
event.
Most of the available systems behave linearly with, or without a damper.
3 FORMULATION OF THE PROBLEM
3.1 Elastic Force
Consider a mass mattached to a carriage sliding along a linear guide with practically no friction.
606

Figure 3: Linear Ramp in the transverse direction (plan view)
Perpendicular to the direction of motion consider attached to the carriage a spring the outer
end of which slides on a general symmetric curve as in Fig. 4. z(x)

Fig. 4 Plan view of the spring moving along the curve

At a displaced position x, the spring is compressed by a force related linearly with the dis-
placement . Therefore the spring force developed is: z(x)

( )
sp sp
F K z x = (1)

where is the stiffness of the spring.
sp
k
This is equilibrated by the vertical component of the reaction N
s
which acts perpendicular to
the tangent line at the contact point, as the contact is considered frictionless, providing the re-
storing force along the direction of motion. The magnitude of the restoring force is given as:
r
F
r sp
F F tan k z(x) z (x ' = = ) (2)
where prime denotes derivative with respect of x.
3.2 Damping Force
In the same direction, i.e. perpendicular to the direction of motion, consider a viscous damper
with capacity as presented in Fig. 5. C
tan=dz/dx
z(x)
F
sp
+x
O
z

x
F
r

N
s

direction of motion
r

x
y
F
Fsp
Plate movement
607
The damping force is expressed as:
d
dz dz dx
F C z(x) C C C z (x) x
dt dx dt
' = = = =
(3)
where dot denotes derivative with respect to time.
Therefore the unbalanced force in the direction of motion due to viscous damping is given as:

Fig. 5. Plan view of the viscous damper moving along the curve
( )
2
c d d
F F tan F z (x) C z (x) x ' ' = = =
(4)
3.3. Equation of motion
Applying the dynamic equilibrium of all the forces acting on the mass m due to inertia, damping,
elasticity and the external dynamic force f ( , the equation of motion of the described single
degree of freedom system results as:
t)
( )
c r
2
sp
mx F F f (t)
or
mx C z (x) x k z(x) z (x) f (t)
+ + =
' ' + +
=
(5)
which is the general equation that depends on the selection of the sliding curve z(x).

3.4. 2nd order polynomial curve
A general second order symmetric curve can be described as:
2
z(x) x x = o + | + (6)
z (x) 2 x sgn(x) ' = o + |
(7)
3.5. Restoring Force
Substituting the above expressions one obtains the restoring force in the following form:
tan=dz/dx
z(x)
F
+x
x z
direction of motion
viscous
damper
F
c

d
N
O
608
( )
2
r sp
2 3 2 2
r sp
F k ( x x ) 2 x sgn(x)
or
F k (2 )x (3 sgn(x))x (2 )x sgn(x)
= o + | + o + |
(
= o + o| + o + | + |

(8)
which is a complete cubic order polynomial of the displacement x. For =0 the curve is a parab-
ola and is its value at x=0. In this case the restoring force is determined as:
2 3
r
F k (2 )x (2 )x
= o + o
(9)
Therefore, for an oscillator with no damping i.e. c=0, the above restoring force substituted
into the equation (5) leads to the well known Duffing equation with a linear and cubic term pro-
viding the restoring force.
2 3
k k
x 2 x 2 x
m m
+ o + o =
f (t)
m
(10)
The behavior of Duffing equation for the autonomous ( f (t) 0 = ) and the non-autonomous
SDOF system is one of the main topics on non-linear vibrations, Verhulst (2006),
Guckenheimer (1983), while a great number of published papers are devoted to the analyses of
this equation using different methods investigating its particular response.
(f (t) 0 = )
|
3.6. Damping Force
For the second order curve defined in relation (6), the damping force, for a viscous damping
device parallel to the spring, is expressed as:
|
2
c
F C 2 x sgn(x) x = o + | (12)
which is quadratic in x and provides always positive damping, while constant | offers the clas-
sical viscous damping term. This is interpreted as a variable damping which becomes at x=0
and increases quadratically with x for , >0.
2
|
Therefore, for a general symmetric quadratic curve the equation of motion becomes:
| |
2
2 3 2 2
sp
mx C 2 x sgn(x) x
k (2 )x (3 sgn(x))x (2 )x sgn(x) p(t)
+ o + | +
(
+ o + o| + o + | + | =

(13)
where it is observed that does not affect damping as expected, while damping is always posi-
tive and does not reveal the particular features of Van der Pol negative type of damping that
leads to limit cycle type of response.
This second order non-linear ordinary differential equation is of Duffing-Van der Pol type
that can be transformed into a state space form as:
( )
1 2
2
2
sp
2 3 2 2
x (t) x (t)
C
x (t) 2 x sgn(x) x
m
k
f (t)
(2 )x (3 sgn(x))x (2 )x sgn(x)
m m
=
= o + |
(
o + o| + o + | + | +

(14)
The above system can be easily solved numerically using Matlab, Maple, or Mathematica for
a given forcing function f and appropriate values of the parameters, together with initial con-
ditions for the displacement and velocity.
(t)
The above equation determines the acceleration of the system as a cubic function of the dis-
placement that constitutes the contribution of the restoring force and a quadratic function of the
609
displacement for damping force. In other words, when the displacement of the system increases
in the one or the other direction, the apparent stiffness is increasing faster in a cubic manner and
the damping force increases quadratically. Moreover, the parabolic shape of the sliding curve
affects both stiffness and damping, while (a 0) = | is supplying additional damping. For = 0,
the quadratic term in the restoring force drops in (14) and, in the absence of damping, the equa-
tions reduce to the so-called Duffing oscillator, whereas with the presence of a quadratic damp-
ing force the system responds following the so-called Duffing - Van der Pol non-linear
oscillator, where both have been extensively studied in the literature.
Notice also that when the plate passes from the equilibrium position at x=0 the damping force
decreases and becomes zero when b=0. This feature alters the out of phase interplay of the re-
storing and damping force that exists in the linear system and needs special attention. In addi-
tion, another configuration is possible that arranges the spring and damper to roll in different
curves in order to separate their influence.
The above system describes the motion of the one plate. A similar system is determined in
the perpendicular y-direction for the other plate to form the overall system of two uncoupled
oscillators in x and y directions.
For a given mass of the artifact, together with a set of stiffness and damping parameters of the
springs and dampers in both directions, as well as the values of the parameters of the corre-
sponding parabolas, the response of the system is determined for given earthquake excitations
along the x and y directions. The numerical solution is determined using a Runge-Kutta integra-
tor of 4-5 order implemented in a Matlab code specially developed for this purpose, Kefala
(2007).
Furthermore, one has to realize that specifications for designing such systems are not gener-
ally available and designers have to rely on relevant codes and specific investigations to deter-
mine the seismic hazard at a particular location and select the methods and limit states for a safe
design, as well as to provide experimental verification of the behaviour of such systems.
4 OPTIMAL DESIGN
For a particular application an optimization problem can be formulated as follows: Given an
artifact of mass m and the targeted displacement capacity of the system D , determine the pa-
rameters of the isolation system that safely survives a given set of scaled excitation records that
quantify the seismic hazard at the specific site. The displacement capacity is usually decided
with the museum officials and is a trade off between the good performance and space limita-
tions in the museum hall.
Due to the uncorrelated nature of the components of seismic events one seeks a symmetric
solution in both directions and thus the analysis in the one dimension suffices, depending on the
relative mass of the plates as compared to the mass of the artifact which alters the total mass for
the upper and lower plate, but can be scaled appropriately based on the parameterization of op-
timal solution described below.
The design variables are the spring stiffness constant, the damping constant for the viscous
damper in the direction perpendicular to the direction of motion and the parameters of the pa-
rabola i.e. in total five design variables.
, , , , , = o |
T
sp
(
x K C (15)
measured as: ( ) ( ) ( ) ( ) ( )
1 0 1 2
sp
K kN / m , C kN sec/ m , m , m (m ) and m kN sec / m
o | from
which, by virtue of relation (14) only the following four quantities are independent:

2
2
sp
K
C
, , ,
m m
( o
o |
o o
(

(
(16)
610
This allows for standardization of some optimal geometries for the quadratic curve depend-
ing on the set of the scaled excitation records and the displacement capacity of the system to
which stiffness and damping properties can be adjusted to the specific mass accordingly using
relation (16).
4.1. Optimization algorithm
The objective function of the problem corresponds to the minimization of the kinetic energy of
the system for all given excitations limiting the accelerations within acceptable bounds.The op-
timization problem can be stated mathematically as minimizing the kinetic energy of the
artifact:
( ) ( )
2
2
0 0 0
1
min min min ( ) min 1, 2,...,
2
| |
= ~ ~ =
|
\ .

} } }
N N
t t t
N
kin abs i
i
E mv t dt a t dt dt a t i N (17)
subjected to:
( ) ( )
2
max
sec , 1, 2,... s =
abs i
a t a m i N (17)
( )
i
x(t ) D m , i 1, 2,..., N s =
where is the maximum allowed transmitted acceleration, D is the given dis-
placement capacity and N is the total number of timesteps. In relation (17) the kinetic
energy is replaced with the similar function of the absolute sum of the acceleration
which computationally is preferable.
max
o
In this work the sequential quadratic programming method (SQP) is used to solve the
optimization problem defined above, Peppas (2008). The SQP method is implemented
in Matlab. The method is indirectly based on the Karush-Kuhn-Tucker conditions.
5 APPLICATION
For a statue of weight w=7.1214kN and a lightweight isolator with a displacement capacity of
D=0.35 -0.40 m, the following response can be observed for the optimal parameters of the isola-
tor presented in Table 1:
Table 1. Isolator Parameters
k
sp
(kN/m) a (1/m) b C (kNsec/m)
3,0450 0,5000 0,4131 2,5134
The seismic hazard analysis performed for the specific museum site determines a number of
excitations, with appropriate scaling factors, the average of which is compared to the design
spectrum of the region well magnified to account for the importance of the isolation design.
This corresponds to the specified return period of a strong excitation with a certain probability
of exceedance. For this application the following seven strong excitations were selected that
contain near and far field records scaled as in Table 2. From these spectra one can appreciate the
benefits of seismic isolation as for periods greater than 2 sec the corresponding spectral accel-
eration is reduced considerably.
611
0
5
10
15
20
25
30
0 1 2 3 4
T [sec]
a
b
s
o
l
u
t
e

a
c
c
e
l
e
r
a
t
i
o
n

[
m
/
s
e
c
2
]
Loma Prieta Corralitos
Loma Prieta Gilroy
Northridge Rinaldi
ChiChiTCU074
ChiChiTCU075
ChiChiTCU129
Loma Prieta Saratoga

Figure 6. Acceleration Spectra of Seven Scaled Excitations
Table 2. Response characteristic of the system for different earthquakes
Station Rinaldi Gilroy ChiChi129 ChiChi075 Corralitos ChiChi074 Saratoga
Scale Factor 0,60 3,70 1,20 0,60 2,30 1,00 1,60
Max. rel. displ. (m) 0,25 0,12 0,34 0,34 0,19 0,28 0,21
Max. rel. vel (m/s) 0,96 0,55 1,05 0,65 1,37 0,74 0,75
Max. abs. acc
(m/sec2) 1,24 0,41 1,37 1,05 0,90 1,10 0,74
PGA (m/sec
2
) 4,93 6,18 11,89 1,96 14,52 5,86 8,04
PGA reduction to 25,11% 6,57% 11,51% 53,61% 6,22% 18,71% 9,25%
In Table 2 the maximum relative displacement, relative velocity and the absolute acceleration
that is transmitted to the artifact are presented for the seven selected scaled excitations. A sub-
stantial reduction of the transmitted acceleration is observed. In Fig. 7 the time history of the
transmitted absolute acceleration is presented.
612
0 10 20 30 40 50 60 70 80 90
-1.5
-1
-0.5
0
0.5
1
1.5
time [sec]
a
b
s
o
l
u
t
e

a
c
c
e
l
e
r
a
t
i
o
n

[
m
/
s
e
c
2
]
Absolute Acceleration

Rinaldi
Gilroy
ChiChi129
ChiChi075
Corralitos
ChiChi074
Saratoga

Figure 7. Absolute acceleration at the artifact due to all earthquakes
0 10 20 30 40 50 60 70 80 90
-0,3
-0,2
-0,1
0
0,1
0,2
0,3
0,4
time [sec]
r
e
l
a
t
i
v
e

d
i
s
p
l
a
c
e
m
e
n
t

[
m
]
Relative Displacement

Rinaldi
Gilroy
ChiChi129
ChiChi075
Corralitos
ChiChi074
Saratoga

Figure 8. Relative displacement of the system for all excitations
In Fig. 8 the relative displacement is presented for the seven scaled excitations where it is ob-
served that, except Rinaldi and ChiChi excitations that reach a displacement of almost 0.35 m,
most of the excitations are kept below 0.20 m. This verifies that an isolator with displacement
capacity of 0.40 m is adequate to survive a fairly strong set of records in its lifetime with high
probability.
613
In existing museums sited in earthquake prone areas, Lightweight Type Isolators are relatively
low cost devices that can be used to isolate effectively showcases and single artifacts up to a
max mass of 1 tone. Most of the existing types utilize linear or bilinear type of stiffness and vis-
cous type of damping. The proposed cubic order Isolator enhances the existing capabilities by
changing the geometry, adding up to cubic order stiffness and up to quadratic order damping.
The cubic stiffness term acts as smooth barrier at large displacements and although it increases
the transmitted force and acceleration, restrains the relative displacement. The quadratic damp-
ing force increases also the damping at large displacements and acts in a similar way. These ad-
ditional features can be used effectively in the design of such systems to result into better
compromised solutions between minimum acceleration transferred to the superstructure and
maximum displacement capacity of the isolator.
REFERENCES

Agbabian M. S., Ginell W. S., Marsi S. F., Nigbor R. L., Evaluation of Earthquake Damage Mitigation
Methods for Museum Objects, Int. Conference Proceedings, May 1991.
Guckenheimer, J., Holmes, P., Nonlinear Oscillations, Dynamical Systems and Bifurcations of Vector
Fields, Springer-Verlag 1983.
Inman D. J., Engineering Vibration, 2nd Edition, Prentice-Hall, 2001
Kefala G., Seismic Isolation of Museum Artifacts, NTUA Diploma Thesis (in Greek), 2007.
Naeim F., Kelly J. M., Design of Seismic Isolated Structures: From Theory to Practice, John Willey &
Sons, 1999.
Peppas N., Optimal Design of Seismic Isolation Systems for Museum Artifacts, Diploma Thesis (in Greek),
2008.
Stavridis .., Schoettler M.J., Somerville P.G., Thio H.K., Podany J.C., Design of a Self Centering
Seismic Base Isolator for an Antiquity, Seismic Mitigation Conference, J.P. Getty Museum, 2006.
Verhulst, F., Nonlinear Differential Equations and Dynamical Systems, Spriger 2006.
614
1 INTRODUCTION
1.1 Sikkim: The regional setting
Sikkim, a small hilly state of India, is situated in the Eastern Himalayas. Geographically, it lies
in between Nepal in the west and Bhutan in the east. China surrounds the state from its north to
northeast. The Indian state of West Bengal borders it to its south (refer Figure 1). The latitudinal
and longitudinal limits of this hilly terrain are respectively 275'N to 209'N and 8759'E to
8856'E. The longest north-south distance is about 100 km and the east-west breadth ranges
within 60-70 km. Its total area is 7,299 km
2
. Surrounded by three international boundaries, this
strategically located state consists of four districts, viz. East Sikkim, West Sikkim, North Sik-
kim and South Sikkim.
Spanning Sikkims western borders are the Khangchendzonga and the Singalila Range, a
north-south spur of the Great Himalaya. The northern limit, which reaches out to the Tibetan
Plateau, is straddled by the Donkia Range while the eastern flank is bounded by the Chola
Range. The average steepness is about 45. Although the trend of Great Himalaya is to run
across in an east-west direction, the two ridges demarcating Sikkims eastern and western sides,
the Chola and the Singalila, follow a north-south pattern. Across the middle, another north-
south ridge of lesser elevation separates the Rangeet Valley from the Teesta Valley.
1.2 Monasteries: The religio-cultural heritage
Majority of the present population of Sikkim are of Nepali ethnic-national origin; but the native
Sikkimese, consist of the Bhutias and the Lepchas. The former migrated from the Kham district
of Tibet in the 14th century, and the later are believed to have migrated from the Far East in the
Lessons learnt from the Impact of Sikkim 2006 Earthquake on
Heritage Structures

P. Mukhopadhyay
Assistant Professor, Department of Architecture, Town & Regional Planning, Bengal Engineering and
Science University, Shibpur, Howrah, West Bengal, India
S. C. Dutta
Professor of Civil Engineering, School of Infrastructure, Indian Institute of Technology, Bhubaneswar,
Orissa, India (formerly Professor, Department of Civil Engineering, BESUS, Howrah, West Bengal,
India)
S. Bhattacharya
Lecturer in Dynamics Engineering, Department of Civil Engineering, University of Bristol, Queens
Building, Bristol, United Kingdom
ABSTRACT: Sikkim, a small Indian state in the Eastern Himalayas, was rocked by a moderate
earthquake of magnitude of 5.7 on 14.02.2006. The earthquake did not cause major casualties,
but resulted in considerable structural damages to property including few Buddhist monasteries,
which are part of the cultural heritage of the state. Damage surveys of few such century-old ma-
sonry structures were undertaken in March-April 2006. The masonry walls of the buildings
were found to have undergone shear cracks, generally propagated from the openings and pri-
marily created due to diagonal tension generated by the action of lateral seismic shear. The de-
corative plaster inside few monasteries had cracked and spilled at several locations, spoiling its
rich heritage. Inside the masonry walls, junctions of timber frames with the exterior walls had
opened up at places. The objective of the reconnaissance based damage survey was to assess the
nature and extent of the damages suffered, to identify the weaknesses of such structures and to
suggest strategy for preservation of these heritage structures in general.
615
8th century. Consequently, Buddhism is the religion professed by most of these native Sikki-
mese, who reside mostly in the northern and eastern reaches of the state.

Figure 1. Regional Setting of Sikkim
(Source: Asia Maps - Perry-Castaeda Map Collection)

The monasteries are establishments situated near Buddhist religious shrines, where the monks
are used to stay. These complexes, built of brick, include major and minor stups, and the resi-
dential quarters consisting of monks cells lined up around courtyards (Fletcher, 1999). These
historic structures thus constitute a significant segment of the religio-cultural heritage of Sik-
kim. About 200 monasteries or gomphs of Sikkim, belonging to the Nyingma and Kagyu or-
ders of Buddhism, influence the cultural heritage and lifestyle of the people, and also demon-
strate the ancient rituals in practice. The monasteries are adorned with lively frescoes of
Buddhist legends, rare silk and brocade Thangkas, which are painted or embroidered Buddhist
banner hung in a monastery altar and occasionally carried by monks in ceremonial processions.
Also preserved here, are ancient Tibetan manuscripts, exquisitely carved wood work and icons
of silver and gold. Beside the monasteries, the old Raj Bhavan at Gangtok, which is presently
used as the official residence of the Governor of Sikkim, is another heritage structure of great
significance. All these masonry structures are at least a century old. These can, therefore, be
designated as building structures of the Indian pre-codal era, that is, structures constructed be-
fore the publication of the first edition of the Indian code on Brick Works (IS 2212: 1991),
which was first published in 1962. Thus, it would be prudent to consider these heritage struc-
tures to be constructed by materials and methods, which were not engineered, but primarily tra-
ditional. Conservation and restoration of such structures in a permanent fashion through custo-
dial protection are internationally recognised moral and ethical obligations of any nation, the
doctrine of which has been encapsulated by the International Council on Monuments and Sites
(ICOMOS) in four charters, namely, the Venice Charter (1964), the Florence Charter (1981), the
Charter for the Conservation of Historic Towns (1987), and the International Charter for Ar-
chaeological Heritage Management (1990).
616
1.3 Sikkim Earthquake: Outcome of Himalayan Seismicity
The Himalayan Mountains were created by the collision of Indian plate with that of Asia which
began 60 million years ago (Yin & Harrison 2000). The range is presently characterized by high
rates of seismicity and deformation, and widespread exposure of post-collisional, high-grade
metamorphic rocks (Yeats et al. 1992). The slopes are highly susceptible to weathering and
prone to erosion, which along with intense rain, causes extensive soil erosion and heavy loss of
soil nutrients through leaching. These, combined with the high rates of seismicity, cause fre-
quent landslides, isolating the numerous small towns and villages. The Sikkim and adjoining re-
gion forming a part of the Himalayan mobile belt exhibit considerable seismic activity. Accord-
ing to the seismic zoning map of India (IS 1893 Part 1: 2002), this region is located in Zone IV
where maximum intensity of VIII is expected. Table 1 gives an idea about the recorded history
of earthquake of magnitude more than 5.0 in the Richter scale in Sikkim and the adjoining area
including this last moderate earthquake of magnitude 5.7 on 14th February 2006. A reconnais-
sance based damage survey was conducted from 26
th
March to 1
st
April to assess the nature and
extent of the damages suffered by different structures.

Table 1. List of moderate earthquakes in Sikkim and adjoining area.
_____________________________________________________________
Date Epicentre Magnitude
_________________
Latitude Longitude
__________________________________________________
30
th
August, 1964 2760N 8830E 5.2
12
th
January, 1965 2760N 8800E 5.9
21
st
August, 1972 2720N 8800E 5.1
19
th
November, 1980 2739N 8875E 6.0
5
th
April, 1982 2742N 8886E 5.1
27
th
September, 1988 2717N 8829E 6.7
25
th
September, 1996 2743N 8855E 5.0
2
nd
December, 2001 2715N 8817E 5.1
14
th
February, 2006 2735N 8836E 5.7
_______________________________________________________________

The quake, which lasted for 22 seconds, did not cause major casualties, but resulted in con-
siderable structural damages to property. Severe shaking was felt by the people in Gangtok; fur-
niture, books and other loose items were displaced from their positions. Hanging lights and fans
started swinging. The earthquake was followed by a number of aftershocks. The damage to rein-
forced concrete (RC) structures in government buildings was basically restricted to the non-
structural elements; nevertheless, the private or semi-private buildings were damaged from both
structural as well as non-structural view points. The damages of these buildings were mostly
found to be due to the lack of implementation of some fundamental codal provisions, and also
due to the use of poor quality materials. Besides the framed structured buildings, the real suffer-
ers of the earthquake were the houses of the common people built with semi-engineered and/or
non-engineered materials and methods of constructions. However, a few non-engineered build-
ings built with bamboo and/or wood were observed to have survived the earthquake. But unrein-
forced masonry structures, which are not as light as those constructed with bamboo and/or
wood, were damaged considerably by shaking of the quake. Had this been a shaking of intensity
of VIII, which this area is prone to suffer, these masonry structures would have hardly survived.
Among them, the century-old monasteries and such other buildings are worthy of mentioning,
as they represent heritage of the state; moreover these buildings, dating from the Indian pre-
codal era, suffered weathering for a very long time. This paper restricts itself to that portion of
the findings of the damage survey, which attempts to report the damage profile of these heritage
structures, to identify their weaknesses and to suggest strategy for preservation of these heritage
structures in general.
2 DAMAGE PROFILE
It was reported by the officials of the Government of Sikkim that out of the two most important
617
monasteries of Sikkim, the Enchey Monastery had suffered from the Earthquake, while the
Rumtek Monastery did not fall in the line of the earthquake ground motion. Consequently, the
damage survey team visited and documented the damages suffered by the earlier one. Labrang
Monastery, another damage suffering monastery of less fame, was also surveyed to see whether
the profile of the damages was of same nature or not. Apart from the religious structures, dam-
ages suffered by the Raj Bhavan, a century-old masonry structure of repute, were also surveyed
and documented.
2.1 Enchey Monastery
The Enchey Monastery, meaning the Solitary Temple, belongs to the Nyingma Order of
Buddhism. It is located on a hilltop, just above the township of Gangtok. This more than two-
century old monastery was originally the hermitage site of Lama Drupthob Karpo, a renowned
Tantric master. The monastery was rebuilt in its present form in 1910, imitating Chinese pago-
das, at the same site of the hermitage during the reign of Sikyong Tulku (refer Figure 2). The
walls of the monastery structure are about 500mm thick. Their internal surfaces are covered
with decorative plasters depicting images of Gods, Goddesses, and other religious symbols. In-
side the masonry walls of this two-storey structure, timber frame is provided. Such frame con-
sists of four timber columns connected by timber beams in both the principal directions. The
floor consists of wooden beams and planks, while the roof is made of Galvanised Iron (GI)
sheeting supported by timber trusses.

Figure 2. Enchey Monastery, a well known
heritage structure in Gangtok, Sikkim.

The masonry walls of this structure had undergone severe shear cracks. It was observed that
these cracks generally propagated from the window and door openings and were primarily
created due to diagonal tension generated by the action of lateral seismic shear. The photo-
graphs at Figure 3, taken from inside of the monastery, illustrate the issue where examples of
generation of diagonal shear cracks from the corners of a window and a door opening are
shown. It was further observed that there was absence of horizontal and vertical reinforcements
around such openings. The existing lintel bands were found to be of inadequate strength. The
photograph at Figure 4 further illustrates the above issue of diagonal shear crack through exter-
nal photographs of the monastery. Along with the diagonal shear cracks, vertical and horizontal
shear cracks had also developed in the walls and other structural elements of the monastery. The
photograph at Figure 5 demonstrates the example of a vertical shear crack occurring through the
capitals of a typical column of the Enchey Monastery. All the above figures are also witness to
the damage of the decorative plasters on the inner surfaces of the masonry walls in the form of
development of cracks, which had spilled at several locations primarily because of failure of the
masonry walls. This has spoilt rich heritage of the structure.

618
(a) Shear cracks from window openings, also (b) Shear cracks from door openings, further
damaging decorative wall plasters. damaging decorative ceiling plasters.

Figure 3. Propagation of shear cracks from the openings of Enchey Monastery, Sikkim in the 2006 Sik-
kim Earthquake.

Figure 4. Diagonal shear cracks developing through Figure 5. Vertical shear crack in the capital of a
the masonry walls of the Enchey Monastery from column of the Enchey Monastery developed in
corners of window openings as an impact of the the 2006 Sikkim Earthquake.
2006 Sikkim Earthquake.

In addition to the different types of shear cracks, analysis of some other damages reveals lack
of technical considerations and poor workmanship. The photograph at Figure 6 not only exposes
weakness of mortar joints, but also exposes arbitrary placement of RC bands without under-
standing the physics, making the zone around the band vulnerable. Again the photograph at
Figure 7 exposes an example of flexural failure of brick masonry structure within the Enchey
Monastery complex, which is circular in plan. The masonry structure had failed because of li-
mited capacity of mortar to resist the tension created by the meridional stress.

619

Figure 6. Damages due to 2006 Sikkim Earthquake
exposing poor quality of workmanship at the
Enchey Monastery.

Figure 7. Flexural failure of brick masonry within
the Enchey Monastery complex during the 2006
Sikkim Earthquake.

2.2 Labrang Monastery
Labrang Monastery, literally meaning Lamas Dwelling, also adheres to the Nyingma School
of Tibetan Buddhism; although it should not be confused with the more famous monastery of
the same name belonging to the Geluk school situated in Xiahe county of Gansu province at
China. The Indian monastery is situated in North Sikkim. It was commissioned in 1814 by
Gyalshe Rigzing Chembo, and was dedicated to the holy memory of Lastun Chembo of Kongpu
of Tibet, who is responsible for propagation of the Nyingma School in Sikkim. This old struc-
ture (refer Figure 8) is unique in the sense that unlike most monasteries of Sikkim, which were
razed by fires sometime or the other, this one still retains the original structure. Inside the mo-
nastery, the older four timber columns connected by timber beams in both principal directions
are reinforced with later introduced system of a number of steel trusses fixed with the older tim-
ber columns through gusset plates (refer Figure 9). The roof is made of GI sheeting while floor
consists of wooden beams and planks.

620

Figure 8. Labrang Monastery, North Sikkim

Figure 9. Internal arrangement of structural
support system, Labrang Monastery, North Sikkim

Likewise the Enchey Monastery, the propagation of diagonal shear cracks from corner of
door-window openings, vertical shear cracks through the masonry walls and junction of two
mutually perpendicular walls constitute the damage profile of this monastery as well. The two
photographs at Figure 10 depict such shear cracks in the Labrang Monastery. Again the photo-
graph at Figure 11 portrays the flexural failure of an isolated brick masonry, with circular plan,
present within the Labrang Monastery complex. A number of such structures had suffered heavy
damage due to poor capacity of mortar to withstand the tension generated due to meridional
stress.

621

Figure 10. Shear cracks produced in the brick masonry of the Labrang Monastery, North Sikkim due to
2006 Sikkim Earthquake.

Figure 11. Flexural damage of isolated masonry
structure having circular plan at the Labrang
Monastery, North Sikkim during 2006 Sikkim
Earthquake.

It was interesting to note that though the masonry structures in the monastery complex suf-
fered small to heavy damages due to the earthquake, a few traditional residences made of com-
bination of wood, bamboo and rattan on brick plinth with corrugated GI sheet roofing did not
suffer any damage whatsoever (Fig. 12). The natural characteristics of these traditional materials
of being light in weight, fibrous in morphology make them highly flexible. Buildings made with
such materials become inherently resistant to the lateral shock waves of the earthquake forces
622
because of their large deformation accommodating capability (Mukhopadhyay 2008). Conse-
quently they become a natural choice.

Figure 12: Close-up view of undamaged residential
buildings near the damaged Labrang Monastery
complex in North Sikkim.

2.3 Raj Bhavan, Gangtok
The present Raj Bhavan is an erstwhile Residency of the British Government, built in 1890 by
John Claude White, the first Political Officer of Sikkim, who was also a Civil Engineer, pre-
viously employed with the Public Works Department. In its previous designation as India House
or Baara Khoti, it had been rated as one of the Indias best Ambassadorial residences. How-
ever, the deadly Valentines Day earthquake of 14th February, 2006, the second big earthquake
faced by this magnificent complex more than a century after the 1897 one that ripped through
Gangtoks bowels, badly damaged the Raj Bhavan and rendered it structurally unstable (The
Raj Bhavan, Gangtok, 2009).
The walls as well as columns of this two-storied building are constructed with chisel dressed
stone and mud mortar. The columns of the building, connected by reinforced concrete beams,
are observed to be severely damaged. One such damaged column is shown in Figure 13. The
cracks clearly show that there is no reinforcement within the column section.

Figure 13: Damaged column in the Raj Bhavan
complex, deep peripheral crack shows that there is no
reinforcement in the column

623
The infill walls of many portions of the Raj Bhavan are found to be damaged showing di-
agonal and vertical shear cracks. Two such photographs are shown at Figure 14, where Figure
14a shows an example of vertical crack at a column and Figure 14b shows an example of di-
agonal crack in an infill wall. Again, propagation of cracks, from the corners of openings, is al-
so noticed at different doors and windows, two of whose examples are shown at Figure 15.

(a) Vertical crack of column (b) Diagonal crack of infill wall

Figure 14: Shear cracks in columns and walls

Figure 15: Cracks propagating from corner of door-window openings

3 CONSERVATION STRATEGY
3.1 The heritage context
The three heritage building complexes surveyed can be classified as vernacular structures which
have evolved using materials available locally. In vernacular architecture, improvements are
possible using modern materials to increase tensile resistance and bonding of structures, but a
blind use of the same can be disastrous; because nature of the building needs understanding.
Further, observation of the damage profile suffered by these structures during the 2006 Sikkim
Earthquake points towards a typical pattern. This pattern includes propagation of shear cracks
from the corners of openings, shear failure in the infill masonry walls and at the junctions of
624
walls with walls and those of floors with walls, failure of masonry joints due to poor quality of
mortars used, and, flexural failure of relatively slender masonry structures with circular plan. In
some cases, seismic damages have exposed poor workmanship that had taken place during some
earlier physical intervention. Yet again, the historic status of these structures calls for interven-
tion in them in accordance to the principles of conservation, which in this case, refers to preser-
vation, restoration and reconstruction. Again Article 3 of the Bura Charter declares Con-
servation is based on a respect for the existing fabric and should involve the least possible
physical intervention (ICOMOS 1981). Thus while formulating remedial measures, one needs
to reflect on the structural considerations as well as the relevant conservation issues. The rec-
ommendations suggested in the next paragraphs deliberate on both these aspects.
Before going into the recommendations, it is thought prudent to quote portions of Article 1 of
the Bura Charter, which defines the relevant conservation terminologies. Preservation means
maintaining the fabric of a place in its existing state and retarding deterioration. Restoration
means returning the existing fabric of a place to a known earlier state by removing accretions
or by reassembling existing components without the introduction of new material. Reconstruc-
tion means returning a place as nearly as possible to a known earlier state and is distinguished
by the introduction of materials (new or old) into the fabric ((ICOMOS 1981).
3.2 The necessity of restoration
Generally restoration is ought to be discouraged as in bulk of the cases restoration robs the ma-
jority of our ancient monuments of their true expression. In cases when the restorer succeeds in
deceiving common people by making them believe that the restored work is the original work,
he, in fact, falsifies historical evidence. And in cases when he is unsuccessful in fooling people,
he actually has tampered the original work. The true aim of conservation is to keep all artistical
and historical evidences intact or preserved. However, restoration by anastylosis, using original
material, is justified when supported by firm archaeological evidence and especially for cases
when the monument is in use. This is exceptionally true for the Sikkimese monasteries, which
unlike many of the world known Indian counterparts like the Nalanda Monastery, are very much
in operation. Particularly parts of these masonry structures, which are damaged or destroyed due
to the Earthquake, need to be restored for the safety of the Lamas and the people visiting them.
3.3 Reconstruction and restoration of brick masonry work
The size and texture of the bricks are to be investigated during repair of these old masonry
structures, and new bricks maintaining the same size and texture needs to be procured. If availa-
ble in sound condition, then the old types of bricks may also be gathered from some nearby old
abandoned building. The bricks are to be laid in the same bond using same mortar thickness ton-
ing it to the same colour as in the old work. However, due to less availability of old bricks,
modern bricks in the core may be used. During reconstruction or restoration work, mostly re-
cessed pointing of about 75mm or 125mm recess from the face of the wall is to be practiced.
Recessed pointing may be preferred over other popular pointing methods like flushed
pointing because of the following reason. It is noted that edges of stones or bricks become
rounded over time; and if the pointing is slightly recessed then the mortar does not spread over
the rounded edges. On the other hand, if pointing is kept flushed, a thin skin of mortar is spread.
This skin has no hold on the surface and in the course of time helps in accumulating moisture
behind it, which may eventually increase in size and cause deterioration of masonry.
3.4 Reconstruction and preservation of masonry suffering shear cracks
Cracks should be thoroughly cleaned with water and formed into V-shaped cuts at least 6mm
wide. The water of the cleaned cracks should then be allowed to soak in, and grouted with ce-
ment, lime and sand. Grouting can be done mechanically with the help of machines or manually
by gravity method. In special cases, polyester and epoxy resin grouts may be used, however
these are quite expensive. The penetrating power of these grouting materials is fairly high in
filling up fissures and fine cracks. However, the expenditure may be minimized by using them
after normal injections.
625
The repair of the cracks is to be followed by adequately providing horizontal and vertical
reinforcements around the door-window openings. In cases where a lintel already exists, the
weak horizontal lintel band should be replaced by RC lintel, extending them at least 400mm
beyond the opening to avoid future shear failures. The vertical reinforcements should be ade-
quately anchored into beam or slab to provide a monolithic behaviour with the walls. These ver-
tical and horizontal reinforcements would provide a confining action to the infill materials and
improve their performance in future earthquakes.
If it is found that the doors and windows are so placed as to cause a weakness in the wall,
long ties may be inserted in a concealed way. This will help in reducing stress concentration at
the corners. Further, the two mutually perpendicular walls forming a corner may be toothed with
each other as per guidelines suggested by IS 4326: 1993 and the International Association for
Earthquake Engineering (IAEE & NICEE: 2004). Cross walls and partitions must be similarly
securely attached to the main walls.
The columns should be reconstructed by concrete with adequate reinforcements. The joints of
such reinforced concrete members with beam should be adequately made with proper confining
reinforcements as suggested in IS 13920: 1993.
3.5 Restoration of the decorative plasterwork
The damaged or missing decorative plasterworks on the walls of the monasteries may be res-
tored with the help of lime mortar injected into the hollow cavities and by applying neat fillet of
lime mortar round the damaged edges, care being taken that the cavities and edges are first
washed with water. While carrying out plaster work, the thickness of original plaster, its compo-
sition and finish should be investigated properly. At times, the new plaster may be toned or
stained to match with the adjacent colour and/or texture. Before repairing such decorative plas-
ter and other decorative and ornamental works, extensive cement grouting should be done to
improve the local strength of the different cracked regions.
3.6 Reconstruction and preservation of flexural failures
It was also observed that some isolated round shaped masonry structures, present in numbers in
each of the monastery complexes, suffered flexural failures. The bulging out of these structures
may be prevented by preparing a mesh of vertical and horizontal reinforcements conforming to
their shape.
4 SUMMARY AND CONCLUSION
Buildings, which have survived the hazards of 100 years of usefulness, have good claim to be-
ing called historic. Such buildings appeal to our emotions and make us want to know more
about the people and culture that produced it. However, only a small fraction of the objects and
structures created in the past survives the ravage of time. These historic properties deteriorate
through the attack of natural and human agents. However, earthquake is one such destructive
natural force, which throughout human history has had an astonishing destructive effect on cul-
tural properties.
India possesses a rich and varied cultural heritage, spreading over history. Conservation of
such heritage is always a must responsibility of the state, more so if they represent a minority of
the population. The Buddhist monasteries and other age-old masonry structures of Sikkim
represent one such small minority group of India. Such century old masonry structures faced the
havoc of a moderate earthquake in 2006, which caused considerable damage to them. The gen-
eral profile of the damages surveyed after the 2006 Sikkim Earthquake on three heritage struc-
tures namely the Enchey Monastery, the Labrang Monastery and the Raj Bhavan are summa-
rised in the following paragraphs. Considering the fact that this geographic region is located in
Zone IV of the seismic zone map of India where maximum intensity of VIII is expected, the
findings of the damage survey become important to ensure safe performance of such structures
in future.
626
(a) The vulnerability of unreinforced masonry, due to its limited tensile and shear strength,
and its lack of ductility and energy-absorbing capacity is responsible for generation of horizon-
tal, vertical and diagonal shear cracks in the infill walls of the structures. The tensile load bear-
ing capacity may be increased through reinforcing the junctions between walls with walls, and
walls with floors.
(b) The propagation of the shear cracks are further aggravated when they encounter openings
in the form of doors and windows, where normally diagonal cracks are generated. The diagonal
cracks are primarily created due to diagonal tension generated by the action of lateral seismic
shear. The poor performance can be rectified by the introduction of lintel bands and vertical
reinforcements around the openings.
(c) The damage survey further shows that masonry structures with circular plan suffered from
meridional stresses, causing bulging out of the same. This can be retrofitted with the provision
of mesh of vertical and horizontal reinforcements along the surface.
(d) The vulnerability of the unreinforced brick masonry is further increased depending upon
the wall geometry, its aspect ratio, biaxial stress conditions etc., which needs further experimen-
tal and analytical investigation.
(e) The better performance of some of the local residences in the vicinity of the damaged mo-
nasteries can be attributed to their better damping capacity and plastic deformation ability due to
the lightweight and fibrous morphology of the materials used.
(f) During restoration work, enough care should be taken regarding the size and texture of
brick and mortar to be used to maintain the consistency between original and restored works.
(g) Recessed pointing is to be applied in the masonry work to decrease the probability of ac-
cumulation of moisture at the surface causing deterioration.
(h) Cracks should be properly cleaned and formed into V-shaped cuts of enough width fol-
lowed by appropriate grouting technique.
(i) Enough care should be taken for restoration of decorative plasterworks by use of lime
plaster of same thickness, toned or stained to match with the colour and/or texture of the origi-
nal one.
Learning from the past and feeling pride of ones own heritage is one of the premises of hu-
man civilization. However, modernisation and pursuit of new lifestyles have persistently impe-
riled national heritages. Moreover, historic structures throughout the world are being threatened
by atmospheric agents; natural disasters add to this ever-increasing menace. But, heritages of
the past should be passed on to the future generations. It is here where the importance of con-
servation of historic structures lies. The engineering and architectural community has to play an
immense role in this regard for ensuring proper application of technology for mankind. In fact,
the adequate technology, though available, sometime is not being implemented, even at the time
of restoration after an earthquake, due to the lack of specialists supervision. Steps should be
taken to translate the moral and ethical obligations put forward from time to time by different
international bodies like ICOMOS into mandatory rules. Conservation of heritage structures by
a team of technical experts is not to be kept at the discretion of nations, but to be made as bind-
ing; and then only the knowledge created can be transformed from laboratory to land, from
theory to reality in practice. In fact, the restoration techniques for the typical damages should be
clearly made available in the form of handy public documents. The present paper may prove
useful in collating the same.
REFERENCES
Bureau of Indian Standards. 1991. Indian Standard 2212: 1991, Brick Works Code of Practice (first re-
vision). New Delhi: BIS.
Bureau of Indian Standards. 2002. Indian Standard 1893 (Part 1): 2002, Criteria for earthquake resistant
design of structures, Part 1 General provisions and buildings (fifth revision). New Delhi: BIS.
Bureau of Indian Standards. 2002. Indian Standard 4326: 1993 (reaffirmed 1998), Earthquake resistant
design and construction of buildings Code of practice (second revision). New Delhi: BIS.
Bureau of Indian Standards. 2002. Indian Standard 13920: 1993 (reaffirmed 1998), Ductile detailing of
reinforced concrete structures subjected to seismic forces. New Delhi: BIS.
Fletcher, B. 1999. Indian subcontinent. In D. Cruickshank (ed.), Sir Banister Fletchers a history of ar-
chitecture: 754-755. New Delhi: CBS.
627
International Association for Earthquake Engineering & National Information Center of Earthquake En-
gineering. 2004. Guidelines for Earthquake Resistant non-engineered construction. Tokyo: IAEE &
Kanpur: NICEE.
International Council on Monuments and Sites. 1981. Bura Charter. Australia: ICOMOS.
Mukhopadhyay, P. 2008. Role of bamboo in seismic architecture. Indian Architect & Builder 22(2): 102-
106.
The Raj Bhavan, Gangtok, Sikkim, India. 2009. The Raj Bhavan: A brief history In Welcome to the offi-
cial website of Rajbhavan, Gangtok, Sikkim: <http://www.> (Jul. 28, 2009).
The University of Texas Libraries. 2009. Sikkim (Shaded Relief). In Asia Maps - Perry-Castaeda Map
Collection: <http://www.lib.utexas.edu/maps/asia.html> (Jul. 21, 2009).
Yeats, R.S., Nakata, T., Farah, A. et al. 1992. The Himalayan Frontal Fault System. Annales Tectonicae.
6 (suppl.): 8598.
Yin, A. & Harrison, T.M. 2000. Geologic evolution of the HimalayanTibet orogen. Annual Reviews in
Earth and Planetary Science. 28: 211280.
628

Seismic Response Analysis of Underground Monumental
Structures. The Catacombs of Kom El-Shoqafa, Alexandria,
Egypt.

K. Pitilakis
Professor, Department of Civil Engineering, Aristotle University of Thessaloniki, Greece.

S. Hemeda
Lecturer, Department of Restoration, Faculty of Archaeology, Cairo University, Egypt.

ABSTRACT: The paper presents the numerical analysis of the seismic response of the
underground monuments in Alexandria, Egypt and more precisely the catacombs of Kom El-
Shoqafa. The analysis of the static stability considering the degradation of the soft rock
properties due to weathering, aging and human activities and the behavior of these complex
monuments under extreme seismic loading, are the key factors for the efficient restoration and
retrofitting of these significant underground monumental structures.
The catacombs are essentially 2D/3D structures, which has to be modeled accordingly. In the
present study, we have modeled the complex catacomb in 2D and 3D, assuming an equivalent
plane strain approach and applying the Plaxis b.v. 8 code with different seismic scenarios,
corresponding to the seismtectonic features of Alexandria. Advanced soil-rock elastoplastic
modeling has been used. Extensive time domain parametric analysis were performed in order to
examine the response of the catacombs subjected to seismic motions with different amplitudes
of ground motion and different frequency content. (Kalamata, Greece, 1986, Erzincan, Turkey,
1992, Aqaba, Egypt, 1995). The analysis takes into account the complex behaviour of the
structure with the aim to determine the threshold PGA and the corresponding developed
stresses, which should remain lower than the actual strength of different elements composing
the catacombs.
We present in a first phase a comprehensible geotechnical survey undertaken in the
archaeological site, comprising geophysical ambient noise measurements (microtremors), in
order to define the physical, mechanical and dynamic properties of the soils and soft rock
materials within which the Catacombs are excavated. In the second stage, we present the main
results of the detailed static and seismic numerical analysis. The final goal of the numerical
analysis presented herein is to investigate the safety margins of the complex underground
monuments, against combined environmental and extreme seismic loads, in order to propose the
most efficient retrofitting technique to upgrade their stability.

629

1 INTRODUCTION

The aim of the analysis is to investigate the safety margins of the underground monuments,
under their present conditions, against unfavorable environmental (i.e. weathering and high
underground water table), utter lack of preservation, geotechnical and extreme seismic
conditions.
Underground structures safety analysis is performed using the finite element (FE) method.
The paper presents a comprehensive study for the underground monuments safety analysis. The
safety analysis includes not only a failure analysis but the effect of weathering specially the
underground water on the differential settlement has been investigated.
Commercial FE package Plaxis (Karstunen et al, 2006) was used for conducting stress, as
well as settlement analysis. PLAXIS is a finite element program specially developed for
numerical analysis of geotechnical and underground structures.
To compute the deformation of these underground monuments as realistically as possible, an
advanced nonlinear elastoplastic material model needs to be utilized in PLAXIS which is
capable of utilizing such advanced material models. MohrsCoulomb model is used for
deformation and consolidation analysis in this study. The consolidation analysis was performed
using PLAXIS that utilizes Biots (Biot, 1941) consolidation theory in 2D and the nonlinear
material behavior is taken into account as mentioned before.
Also in this paper, we attempt to construct and analyze a three-dimensional (3D) finite
element model (FEM) of the central rotunda in catacombs of Kom El-Shoqafa with its six
supporting rock piers excavated in sandy oolitic limestone deposit, using the FLAC 3D code.
For the seismic analysis, we have modeled the complex catacomb assuming an equivalent
plane strain approach and applying the Plaxis b.v. 8 with different seismic scenarios,
corresponding to the seismotectonic features of Alexandria. Advanced soil-rock elastoplastic
modeling has been used. Extensive time domain parametric analysis were performed in order to
examine the response of the catacombs subjected to seismic motions with different amplitudes
of ground motion and different frequency content. (Kalamata in Greece, 1986, Erzincan
inTurkey, 1992, Aqaba in Egypt, 1995). The analysis takes into account the complex behavior
of the structure with the aim to determine the threshold PGA and the corresponding developed
stresses, which should remain lower than the actual strength of different elements composing
the catacombs.

A-A/
10 M 8 6 4 2 0
(B)

B-B/
Figure 1. Plan view and the two main cross- sections of Catacombs of Kom El-Shoqafa.
0 2 4 6 8 1 0 m
(A)
(A
/
)
(B
/
)
630

Figure 2. Present state of catacombs of Kom El-Shoqafa..

Table 1. Material properties of rock and other construction material used for the stability analysis of
Catacombs of Kom El-Shoqafa.

Parameters Name
Rock
material
Brick
(old restoration)
Rock unit weight
unsat kN/m
3
18 17
Rock unit weight
sat kN/m
3
22 20
Youngs modulus E
ref
kN/m
2
2.270E+06 1.350E+06
Shear modulus G
ref
kN/m
2
8.867E+05 5.400E+05
Oedometer modulus E
oed
kN/m
2
2.902E+06 1.620E+06
Poissons ratio
(nu)
0.28 0.25
Cohesion c
ref
kN/m
2
500 450
Friction angle
O
35 31
Shear velocity Vs m/s 715 557.9
Longitudinal
wave velocity
Vp m/s
1293 96634
Uniaxial
compressive strength
UCS kN/m
2

2400 1100
Bending strength y kN/m
2
560 200
Shear strength
f
kN/m
2
364 -
Dilatancy (o) 1
o
0
631

2 PRELIMINARY 2D STATIC ANALYSIS

In the initial static analysis, the excavation is modeled by assuming non-linear soil / rock
behavior and the Mohr coulomb failure criterion. The following parameters are used: =36
o
,
c=500 kN/m
2
, E=2.270E+06 KN/m
2
, =0.28, Vs=715 m/sec for the rock material, and =31
o
,
c=500 kN/m
2
, E=1.350E+06 KN/m
2
, =0.25, Vs=550 m/sec for the modern brick support walls
and piers. Figure 3 presents the total displacements of the soft rock excavations of catacombs of
Kom El-Shoqafa, for two typical cross sections, 1and 2, for the first and second floor
respectively. The results from the preliminary static analysis indicated that the ground
displacements were small (of the order of few millimeters 1.13 mm); some rock pillars are
under relatively high compression stresses. The calculated peak effective principal compressive
stresses on supporting rock pillar 1 is 1.42 MPa.

First floor Second floor

Figure 3. Total vertical static displacements at the catacombs of Kom El-Shoqafa, the first and second
floor.

3 PRELIMINARY 3D STATIC ANALYSIS

An attempt has been made to construct a three-dimensional (3D) numerical model of the central
Rotunda in catacombs of Kom El-Shoqafa with its six supporting rock pillars excavated in
sandy oolitic limestone deposit. The objective of the 3D analyses was to evaluate the stress state
in the pillars taking into account the 3D geometry. The 3D effects issue is considered on a basic
engineering approach in the subsequent sections. The various simulations described herein were
conducted using the FLAC 3D code (Itasca, 2007).
The results from the 3D static analysis indicated that the ground displacements above the
rotunda (catacombs) were small; the maximum total (vertical) displacements is of the order of
2.6 mm and 3 mm in the whole domain and the peak horizontal displacements was 1.0 mm.
Some rock piers are under relatively high compression stresses. The calculated peak effective
principal vertical compressive stresses on supporting rock pillar 1 is 1.74*10
3
KN/m
2
and the
calculated peak effective principal tensile stress is 200 KN/m
2
. The factor of safety of the rock
pillar 1 is 1.47, where the acceptable safety factor is > 1.6. Also the overstress state is beyond
the elastic regime (Hemeda, 2008).

632

Figure 4. Contour of vertical effective stresses yy through the rotunda. The maximum effective
compressive stresses on pillar_1 = -1.74*10
3
kN/m
2
.

Figure 5. Results from the 3D analysis indicating that the maximum ground displacements above the
rotunda are small of the order of 2.6 mm, and the maximum horizontal displacements is only 1 mm.

4 SEISMIC RESPONSE ANALYSIS

In the present study, we have selected three reference earthquakes to be used as input motions at
the bedrock. (i) Aqaba, Egypt, 1995, (ii) Erzincan, Turkey, 1992 and (iii) Kalamata, Greece,
1986. The time histories (Figure 6) of these earthquakes representing different seismotectonic
settings and frequency content, were scaled to three peak ground acceleration values equal to
0.08g, 0.16g, and 0.24g respectively. The design acceleration in Alexandria according to the
Egypt seismic code is 0.08g.
Advanced and well focused computational tools can be efficiently applied to predict with
reasonable accuracy the seismic demands on these geometrically complex underground
monuments. In particular numerical modeling is a very powerful tool for identifying regions of
stress concentration where only non-invasive techniques are allowed. Detailed quantification of
633

induced stresses is also useful for understanding the direction of cracks propagation, to evaluate
the pathology of the structures and to quantify the seismic demands. The three selected
earthquakes have the following characteristics:

- Aqaba: 22/11/1995, M=7.1, Ml=6.2: Station Eilat. Distance (km):
Closest to fault rupture (R=93.8 km). ( El-Sayed et al, 2004)
- Erzincan: 13/3/1992, Mw=6.9, Ms=6.8.Station:95 Erzincan, Rrup=2km, Re=1km.
- Kalamata:13/9/1986.17:24:35, Ms=5.8 .Mw=5.9, Ml=5.5. Station:
Old Telecommunication Building, R=10 km

Aqaba,1995
-0.1
-0.05
0
0.05
0.1
0 10 20 30 40 50 60 70
Time {s}
a
c
c

(
g
)
E w component

Figure 6. Seismic excitations at the bedrock for the reference ground motions (acceleration-time history).
(a) Aqaba, 1995 (b) Erzincan 1992, and (c) Kalamata, 1986 earthquakes.
Erzinc
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0 5 10
a
c
c

(
g
)
an, 1992
15 20 25
time (sec)
EW component
Kalamata,1986
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0 5 10 15 20
time (sec)
a
c
c

(
g
)
E w component
634

-4
-3
-2
-1
0
1
2
3
0 3 6 9 12 15
Dynamic time {s}
a
c
c
e
l
e
r
a
t
i
o
n

{
m
/
s
2
}
Top of Catacombs
Bottom of Catacombs
Bottom of mesh
-0.024
-0.016
-0.008
0
0.008
0.016
0.024
0.032
0 3 6 9 12 15
Dynamic time {s}
D
i
s
p
l
a
c
e
m
e
n
t

{
m
}
Top of Catacombs
Bottom of Catacombs
Bottom of mesh
-2
-1
0
1
2
3
0 4 8 12 16 20
Dynamic time {s}
a
c
c
e
l
e
r
a
t
i
o
n

{
m
/
s
2
}
Top of Catacombs
Bottom of Catacombs
Bottom of mesh
-0.048
-0.032
-0.016
0
0.016
0.032
0 4 8 12 16 20
Dynamic time {s}
D
i
s
p
l
a
c
e
m
e
n
t

{
m
}
Top of Catacombs
Bottom of Catacombs
Bottom of mesh
-0.12
-0.08
-0.04
0
0.04
0.08
0.12
0 4 8 12 16 20 24 28 32 36 40
Dynamic time {s}
D
i
s
p
l
a
c
e
m
e
n
t

{
m
}
Top of Catacombs
Bottom of Catacombs
Bottom of mesh

-3
-2
-1
0
1
2
3
4
0 4 8 12 16 20 24 28 32 36 40
Dynamic time {s}
a
c
c
e
l
e
r
a
t
i
o
n

{
m
/
s
2
}
Top of Catacombs
Bottom of Catacombs
Bottom of mesh

Figure 7. Deformed meshes and peak total displacement, and acceleration, horizontal displacements-time histories, (a) Kalamata , (b) Erzincan , and (c) Aqaba eartquakes. Input
motion PGA = 0.24g. Catacombs of Kom El-Shoqafa, a typical cross section 1, first floor.
635

Figure 8. Effective vertical compressive stresses
/
YY
- time histories and shear stress versus shear strain
for the most critical rock pier 1 (Figure 9 at the right side of rotunda). (a) Kalamata (b) Erzincan (c)
Aqaba earthquakes. PGA value = 0.24g.
-4500
-4000
-3500
-3000
-2500
-2000
-1500
-1000
-500
0
0 5 10 15 20 25 30 35 40
Dynamic time {s}
y
y
{
}
Top of rock pier_1
Middle of rock pier_1
Base of rock pier_1
UCS of rock material
-4500
-4000
-3500
-3000
-2500
-2000
-1500
-1000
-500
0
0 3 6 9 12 15
Dynamic time {s}
-200
-100
0
100
200
-2.E-04 0.E+00 2.E-04 4.E-04 6.E-04
Shear strain %
S
h
e
a
r

s
t
r
e
s
s

s
i
g

/
-
x
y

{

K
N
/
m
2
}
Top of rock pier_1
Base of rock pier_1
y
y
{
}
Top of rock pier_1
Base of rock pier_1
E
f
f
e
c
t
i
v
e

s
t
r
e
s
s
e
s

y
y

{
k
N
/
m
2
}
(a)
-4500
-4000
-3500
-3000
-2500
-2000
-1500
-1000
-500
0
0 5 10 15 20
Dynamic time {s}
-200
-100
0
100
200
-3.E-04 -2.E-04 -1.E-04 0.E+00 1.E-04 2.E-04
Shear strain %
S
h
e
a
r

s
t
r
e
s
s

s
i
g

/
-
x
y

{

K
N
/
m
2
}
Top of rock pier_1
Base of rock pier_1
y
y
{
}
Top of rock pier_1
Base of rock pier_1
E
f
f
e
c
t
i
v
e

s
t
r
e
s
s
e
s

y
y

{
k
N
/
m
2
}
(b)
-200
-100
0
100
200
-1.E-03 -8.E-04 -4.E-04 0.E+00 4.E-04 8.E-04 1.E-03
Shear strain %
S
h
e
a
r

s
t
r
e
s
s

s
i
g

/
-
x
y

{

K
N
/
m
2
}
Top of rock pier_1
Base of rock pier_1
E
f
f
e
c
t
i
v
e

s
t
r
e
s
s
e
s

y
y

{
k
N
/
m
2
}
(c)
636

First floor second floor (the royal tomb)

Figure 9.The typical cross sections 1and 2, which have been used in the seismic analysis.

Figure 10. Differential horizontal and vertical displacements on the top and base of the rock piers in the
catacomb complex (a) Kalamata at time (t) =4.60 sec, (b) Erzincan at (t) = 7.09 sec, (c) Aqaba at time
(t) = 23.22 sec. PGA = 0.24g.
0
5
10
15
25
H
o
r
i
z
o
n
t
a
l

d
i
s
p
l
a
c
e
m
e
n
t

20
{
m
m
}
1 2 3 4
Number of rock pier
top of rock piers
B s ottomof rock pier
0
0.5
1
1.5
2
2.5
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

{
m
m
}
1 2 3 4
Number of rock pier
Top of rock pier
Bottomof rock pier
28
28.5
29.5
30
5
31
H
o
r
i
z
o
n
t
a
l

d
i
s
p
l
a
c
e
m
e
n
t

29
30.
{
m
m
}
1 2 3 4
Number of rock pier
Top of rock pier
Bottomof rock pier
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

{
m
m
}
1 2 3 4
Number of rock pier
Top of rock pier
Bottomof rock pier
74
75
76
80
81
82
83
H
o
r
i
z
o
n
t
a
l

d
i
s
p
l
a
c
e
m
e
n
t

{
m
m
}
1 2 3 4
Number of rock pier
77
78
79
top of rock piers
Bottomof rock piers
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

{
m
m
}
1 2 3 4
Number of rock pier
Top of rock pier
Bottomof rock pier
(b)
(c)
Pillar 1 Pillar 2 Pillar 3 Pillar 4
(a)
637

Figure 11. Deformed meshes, vertical displacement-time history, shear stresses-shear strain, (a) Kalamata , (b) Erzincan , and (c) Aqaba earthquakes. Input motion PGA =
0.24g. Catacombs of Kom El-Shoqafa, a typical cross section 2, second floor, the royal tomb.
-0.0005
0
0.0005
0.001
0.0015
0 3 6 9 12 15
Time{s}
Top of Catacombs
"Bottom of Catacombs"
"Bottom of mesh"
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

{
m
-80
-40
0
40
80
-2.E-04 -1.E-04 -5.E-05 0.E+00 5.E-05 1.E-04 2.E-04
S
h
e
a
r

s
t
r
e
s
s

s
i
g

/
-
x
y

{

K
N
/
m
2
}
Top of rock column_1
Middle of rock column_1
Base of rock column_1
}
-0.0005
0
0.0005
0.001
0.0015
0.002
0 4 8 12 16 20
Time {s}
Top of Catacombs
Bottom of Catacombs
Bottom of mesh
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

{
m
}
-0.001
0
0.001
0.002
0.003
0.004
0 5 10 15 20 25 30 35 40
Time {s}
Top of Catacombs
Bottom of Catacombs
Bottom of mesh
-80
-40
0
40
80
-1.E-04 -5.E-05 0.E+00 5.E-05 1.E-04 2.E-04
Shear strain %
S
h
e
a
r

s
t
r
e
s
s

s
i
g

/
-
x
y

{

K
N
/
m
2
}
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

{
m
}
-80
-40
0
40
80
-6.E-04 -4.E-04 -2.E-04 0.E+00 2.E-04 4.E-04
Shear strain %
S
h
e
a
r

s
t
r
e
s
s

s
i
g

/
-
x
y

{

K
N
/
m
2
}
638

0
02
04
06
08
0.1
0.08 0.16 0.24
PGA Value {g}
0.
0.
0.
0.
M
a
x
.

H
o
r
i
z
o
n
t
a
l

d
i
s
p
l
a
c
e
m
e
n
t

{
m
}

Aqaba RQ
Erzincan RQ
Kalamata RQ
0
0.001
0.002
0.003
0.004
0.005
0.006
0.08 0.16 0.24
PGA Value {g}
M
a
x
.

V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

{
m
}

Aqaba RQ
Erzincan RQ
Kalamata RQ
Figure 12. (a) Maximum horizontal displacements, (b) Maximum vertical displacements on the top of
Catacombs, for Aqaba, Erzincan, and Kalamata earthquakes, scaled to several values of PGA.

0
1000
2000
3000
4000
5000
M
a
x
.
V
e
r
t
i
c
a
l

e
f
f
e
c
t
i
v
e

s
t
r
e
s
s
e
s

s
i
g
/

y
y

{
K
N
/
m
2
}
0.08 0.16 0.24
PGA Value {g}
Aqaba RQ
Kalamata RQ
Erzincan RQ
UCS of rock material.experimental
0
100
200
300
400
500
0.08 0.16 0.24
PGA Value {g}
M
a
x
.

e
f
f
e
c
t
i
v
e

s
h
e
a
r

s
t
r
e
s
s

s
i
g
/

x
y

{
K
N
/
m
2
}
Aqaba RQ
Kalamata RQ
Erzincan RQ
Shear strength of rock .experimental
Figure 13. (a) Maximum vertical effective compressive stresses
/
yy, (b) Maximum effective shear
stresses
/
xy on the base of rock pier_1, for Aqaba, Erzincan, and Kalamata earthquakes, scaled to several
values of PGA.

Figures (7) through (13) are summarizing the main results of the analysis. In case of the Aqaba
earthquake, it is clear that a great part of seismic energy is dissipated to the upper parts of
catacombs (ground surface) even for small values of PGA (Peak Ground Acceleration).
Kalamata and Erzincan input motions give much lower displacements values. The maximum
horizontal displacement at the top of catacombs for Aqaba earthquake at PGA = 0.24g
earthquake scenario was u
x
=

7.95 cm, while the peak vertical effective principal stress was
4190 kN/m
2
. In the case of Erzincan and Kalamata earthquakes, the respective values were 3400
kN/m
2
and

3580 kN/m
2
. Moreover, the maximum horizontal displacement at the top of
catacombs was 2.34 cm and 2.01 cm for Erzincan and Kalamata earthquakes respectively. The
maximum vertical displacement at the top of catacombs, are 3 mm, 2.6 mm, and 5.3 mm for
Kalamata, Erzincan, and Aqaba earthquakes respectively.
Given the value of the static strength estimated in the laboratory (UCS=2.5 MPa), the
seismic analysis of the catacombs complex proved, that the supporting rock piers and columns,
which are the most vulnerable parts of the whole complex, are rather safe for PGA values lower
639

than 0.24g in case of the Kalamata and Erzincan earthquakes and PGA =0.12g for the Aqaba
seismic scenario (see Figure 13a, b).
For larger earthquakes, which are most likely to happen in the region, the seismic stability
of the catacombs is not satisfied and it is necessary to proceed to specific retrofitting works to
upgrade their seismic performance. The maximum differential horizontal displacements of the
top and the base of the rock piers are of the order of 3 to 5 mm (Figure 10). Considering that the
induced seismic ground deformations are better correlated with the intensity of damages in
underground structures, the seismic design of the catacombs must be based on these kinematic
forces.

5 POLYMER ANTI-SEISMIC PILING TO PROTECT THE CATACOMBS AGAINST
STRONG EARTHQUAKES

We employed a newly developed polymer seismic isolation method, which has been employed
successfully to retrofitting the Nakagawa underground station in Yokohama city, to protect the
underground structures of catacombs of Kom El-Shoqafa against strong earthquakes with
PGA>0.24g. The polymer seismic isolation method outlined in Figure 14 presents certain
advantages compared to other conventional anti-seismic methods (such as the steel jacket
method).This method is apriority suitable for the seismic protection of the underground
monuments, because the application of this technique is non-destructive and do not involve any
change in the original materials of the monuments; in addition we do not employ new materials
or constructions to the monumental initial structure.
The polymer seismic isolation method' involves the construction of polymer walls on both
sides of underground structure in order to reduce the seismic action transmitted from the
surrounding ground onto the structure. The stiffness of the polymer material should be about
1/10 to 1/100 that of the surrounding ground. This method is not intended to prevent or control
the seismic ground deformation itself, but to isolate structures from seismic forces transmitted
from the surrounding ground (Hemeda, 2008).

Figure 14. The polymer seismic isolation method to protect the catacombs against strong earthquakes,
(conceptual diagram).
Catacombs wall
640

-0.0005
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
0 3 6 9 12

-600
-400
-200
0
200
400
600
0 3 6 9 12 15
Dynamic time {sec}
E
f
f
e
c
t
i
v
e

s
h
e
a
r

s
t
r
e
s
s
e
s

/
x
y

{
k
N
/
m
2
}
Catacombs notequipped withpolymer
seismic isolator
Catacombs protected fromseismic
activity
Shear strengthof rock
material.experimental
Dynamic time {sec}
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t
s

{
m
}
Catacombs not equipped with
polymer seismic isolator
Catacombs protected from
seismic activity
15
(a)

-0.0005
0
0.0005
0.001
0.0015
0.002
0.0025
0.003
0 4 8 12 16
Dynamic time {sec}
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t
s

-600
-400
-200
0
200
400
600
0 4 8 12 16 20
Dynamic time {sec}
E
f
f
e
c
t
i
v
e

s
h
e
a
r

s
t
r
e
s
s
e
s

/
x
y

{
k
N
/
m
2
}
Catacombs notequipped withpolymer
seismic isolator
Catacombs protected fromseismic
activity
Shear strengthof rock
}
m
{
seismic activity
20
(b)

-600
-400
-200
0
200
400
600
0 10 20 30 40
Dynamic time {sec}
E
f
f
e
c
t
i
v
e

s
h
e
a
r

s
t
r
e
s
s
e
s

/
x
y

{
k
N
/
m
2
}
Catacombs not equipped
with polymer seismic isolator
seismic activity
Shear strength of rock

Figure 15. Vertical displacements, Effective shear stresses
/
xy
- time histories on the base of rock pier 1
before and after the installation of polymer anti-seismic slurry walls. (Figure 9 at the right side of
rotunda). (a) Kalamata (b) Erzincan (c) Aqaba earthquakes. PGA = 0.24g.

(c)
-0.002
0
0.002
0.004
0.006
0.008
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t
s

{
m
0 10 20 30 40
Dynamic time {sec}
}
seismic activity
641

0
100
200
300
400
500
0.08 0.16 0.24
PGA Value {g}
s
i
g
/

x
y

{
K
N
/
m
2
}
(a) (b)

Figure 16. Maximum effective shear stresses sig/xy on the base of rock pier_1, (a) Initial model (b) After
the installation of polymer anti-seismic piling. For Aqaba, Erzincan, and Kalamata earthquakes, scaled to
several values of PGA.

With the installation of polymer anti-seismic slurry walls, the shear forces on the rock pillars,
which is the most vulnerable parts inside the catacombs, reduced by up to 50% for the three
earthquakes scenarios, (see Figure 15). The effective shear stresses on the pillars are also
reduced considerably: in particular from 410 kN/m
2
to 250 kN/m
2
in the case of Kalamata
earthquake at PGA = 0.24g, from 400 kN/m
2
to 280 kN/m
2
in the case of Erzincan earthquake,
and from 430 kN/m
2
to 300 kN/m
2
in the cases of Aqaba earthquake. The decrease of the
computed acceleration values was also obvious in all the three earthquakes scenarios; for
example the horizontal acceleration on the top of the catacombs decreased from 3 m/s
2
to 2 m/s
2
in the case of Aqaba earthquake and from 2.6 m/s
2
to 2.3 m/s
2
in the case of Erzincan
earthquake. The horizontal displacements on the top of catacombs decreased from 24 mm to 20
mm in the case of Kalamata earthquake and from 25 mm to 21 mm in the case of Erzincan
earthquake, while in the case of Aqaba earthquake, always for PGA =0.24g the displacement
reduction is lower; from 80 mm to 78 mm. The vertical displacements at the top of catacombs
are decreased from 2.8 mm to 1.0 mm in the case of Kalamata earthquake, from 2.4 mm to 1.5
mm in the case of Erzincan earthquake, and from 6 mm to 2.5 mm in the case of Aqaba
earthquake.
From the above short presentation and discussion it is obvious, that the seismic stability of
the catacombs has been upgraded after the installation of the polymer anti-seismic slurry walls
in the perimeter of an underground monument. The relative deformations are reduced
considerably and the developed seismic shear forces on the sidewalls can be easily controlled
within acceptable safety margins even for major earthquakes.
It is expected that the above-mentioned polymer seismic isolation method may contribute
effectively to the improvement of the seismic safety margin of underground monuments,
without employing other retrofitting techniques, which may modify the architectural and
archeological principles of the preserved monuments.

0
100
200
300
400
500
0.08 0.16 0.24
PGA Value {g}
M
a
x
.

e
f
f
e
c
t
i
v
e

s
h
e
a
r

s
t
r
e
s
s

s
i
g
/

x
y

{
K
N
/
m
2
}
Aqaba RQ
Kalamata RQ
Erzincan RQ
M
a
x
.

e
f
f
e
c
t
i
v
e

s
h
e
a
r

s
t
r
e
s
s

Aqaba RQ
Kalamata RQ
Erzincan RQ
642

6 CONCLUSIONS

Considering all other affecting factors (aging, weathering, multiple geotechnical and seismic
instability problems) and the specific geometry of the complex it has been shown that the low
rock strength affects seriously the safety of the catacombs both under static and seismic loading
conditions.
The results from the 2D-3D static analysis indicated that the ground displacements above the
catacombs were small (the maximum total vertical displacements are of the order of 2.6 mm and
3 mm in the whole domain, and the peak horizontal displacements were 1.0 mm). Some rock
piers are exhibit relatively high compression stresses. The calculated peak effective principal
vertical compressive stress on the pillar 1 is -1.74*10
3
KN/m
2
. The calculated peak effective
principal tensile stress is 200 KN/m
2
and the factor of safety of the rock pillar 1 is 1.47, which is
not adequate, where the acceptable safety factor must be > 1.6. Also the overstress state is
beyond the elastic regime.
The seismic analysis of these underground monumental structures for three seismic scenarios
of different PGA values, proved that for PGA > 0.10g, which is a rather low value considering
the seismic activity and the past seismic history of the city, there are some critical supporting
parts of these catacombs structures (i.e. rock piers and columns) that are not safe, and in general,
the catacombs need considerable strengthening.
We also presented some preliminary results applying a polymer seismic isolation method
which has been proved to contribute effectively to reduce the induced inertial forces on this type
of underground monuments.

ACKNOWLEDGMENT

The second author wishes to express his deep acknowledgement to Dr Bakasis. I, Civil
Engineering Department, AUTH, Greece for his support in performing part of the numerical
analysis.

REFERENCES

Biot, M.A., 1941. General theory of three- dimensional consolidation, J Applied Physics 12, pp155164.
El-Sayed, A & Korrat, I & Hussein, M., 2004. Seismicity and seismic hazard in Alexandria (Egypt) and
its surrounding, pure appl.geophys.161 (2004), pp1003-1019.
Hemeda, S., 2008. An integrated approach for the pathology assessment and protection of underground
monuments in seismic regions. Application to some Greek-Roman monuments in Alexandria,
Egypt, PhD Thesis, Aristotle University of Thessaloniki, Greece.
ITASCA Consulting Group., 2007. Inc, 3 Dimensional distinct element code (3D DEC).
Karstunen, M & Wiltrafsky, C & Krenn, H & Scharinger, F., 2006. Modeling the behavior
of an embankment on soft clay with different constitutive models, vol 30, Issue 10, 25 August, pp 953
982.
PLAXIS Manual., 2002. Finite element code for soil and rock analysis published and distributed by AA
Balkema Publishers, Nederlands Comput. Geotech. 32(5):326339.

643
1 DIFFICULTIES IN THE ANALYSIS OF HISTORICAL STRUCTURES
The estimation of the seismic response of historic masonry structures is not an easy task. First,
because there are many uncertainties related to the structure itself. For instance, the geometry is
usually very complicated and difficult to be represented properly using classical plane finite
elements. Thus, a more rigorous analysis is required in many cases, based on three-dimensional
elements.
In addition, historical structures, especially ones located in seismic areas, are lightly or heavi-
ly damaged, even partially collapsed in some cases, due to the deterioration of the building ma-
terials, previous earthquakes, vandalism, etc. The implementation of the damage (e.g. cracks) in
the numerical model is not only difficult (the exact representation is practically impossible), but,
in many cases, infeasible, due to the limitations of the finite element codes in handling such im-
perfections.
Another difficulty in modeling historical buildings is related to the fact that they are usually
built in phases, with additions and/or alterations to the initial structural system applied in differ-
ent eras of their long life. Thus, various parts are built with different materials and different me-
thod of construction. In addition, the degree of connection between newer and older parts is
usually unknown.
Apart of the above-mentioned difficulties, related to the geometric representation of the real
structure in the numerical model, another source of uncertainties originates from the material
properties that should be used. Masonry parts of historical structures are usually made of vari-
ous materials (stones, ceramic bricks, mortar) with more or less unknown properties. In many
cases the walls consist of several leaves, each one of different type of construction. For the
modeling of these walls, the method of homogenization is usually applied, i.e. the walls are con-
Estimation of the seismic response of historical structures and
evaluation of interventions

I. N. Psycharis & H. P. Mouzakis
National Technical University, Athens, Greece
A. Miltiadou
Hellenic Ministry of Culture, Athens, Greece
E. P. Pavlopoulou & I. M. Taflampas
National Technical University, Athens, Greece
ABSTRACT: In this paper, the application of linear analysis for the estimation of the seismic
response of historical structures and the evaluation of interventions applied to them is presented
for the case study for a Byzantine church. The pros and cons of this analysis, compared to the
nonlinear one, are discussed. Despite its limitations, linear analysis can reveal significant infor-
mation on the earthquake response of the structure and the evaluation of strengthening interven-
tions, if the results are examined properly and the numerical model (geometry and the material
properties) is calibrated and verified. For this purpose, ambient vibration measurements and the
existing damage can be used. In certain cases, unfavorable scenarios of damage should be intro-
duced in the numerical model, in order to calculate the forces that will be developed at the
strengthening elements during an earthquake and will be used for their dimensioning.

644
sidered to be homogeneous with average properties, based on the ones of the individual mate-
rials. It is evident that this method introduces an approximation, since the determination of the
appropriate values is not an easy task, even if the properties of the individual materials were
known. In practice, due to numerous uncertainties concerning the construction details and the
real properties of the materials, values based on tests performed on samples of the structure (if
such tests are available), previous experience or the literature are used.
2 LINEAR VERSUS NONLINEAR ANALYSIS
The response of historical structures to strong seismic excitations is, in general, nonlinear, due
to the geometric discontinuities in their construction, the existing damage and the reduced resis-
tance caused by the deterioration of the materials. Therefore, for the calculation of the response
one should perform nonlinear time-history analysis (THA). However, despite the obvious ad-
vantages of this approach, there are serious drawbacks, not only because it requires sophisti-
cated computer codes of uncertain accuracy for complex geometries, but also because it is sensi-
tive to the parameters applied. For instance, the uncertainties concerning the geometric
assumptions that are implemented into the numerical models and the merely unknown nonlinear
properties of the materials might lead to results of doubtful accuracy. In general, the obtained
accuracy is not in accordance with the required computational effort and time.
It should be mentioned that, if detailed models are used in order to capture the actual geome-
try, nonlinear analysis requires large computational times that make it inapplicable for practical
purposes. In order to reduce the computational time of nonlinear THA, simplifications in the
geometry of the structure are, sometimes, applied; the results, however, might not be reliable,
because significant features of the response could be lost (Loureno 2007).
The nonlinear static (pushover) analysis is also used sometimes, but additional uncertainties
are introduced due to various reasons: historical structures do not respond monolithically, load-
ing patterns cannot be easily determined and the behavior of the masonry is brittle.
For these reasons, linear analysis is usually preferred, despite the fact that it cannot capture
the real response. Its advantage is that it can be applied using much simpler and more reliable
commercial codes and possible mistakes can be easily detected by critical examination of the re-
sults. Its main disadvantage is that it cannot predict accurately the real forces that will be devel-
oped at each member of the structure, since it cannot count for the redistribution of the stresses
after the initiation of the cracking at the masonry. For example, there is no limit to the magni-
tude of the stresses that are developed, which, thus, might exceed the strength of the masonry.
Also, the forces that are developed at the strengthening elements of an intervention (e.g. ties, ex-
ternal supports, etc) are underestimated, since the contribution of these elements to the resis-
tance of the structure starts after the reduction of the strength of the masonry and the increase of
the deformation, caused by cracks.
Despite these drawbacks, linear analysis can reveal useful information on the shape of the de-
formation and the regions where the resistance of the masonry is exceeded and cracks are ex-
pected to be developed. The effectiveness of proposed strengthening measures can be evaluated
by comparing the stresses that are developed at the masonry with and without the interventions.
The problem of the estimation of the real forces that are developed at the strengthening elements
of interventions can be solved if linear analysis is used in combination with pre-assumed crack
patterns implemented in the numerical model, consistent with unfavorable scenarios of damage
for the intervention under consideration.
Thus, linear analysis might be adequate for most applications if combined with proper engi-
neering judgment. It should be noted, however, that calibration of the parameters that are used
in the numerical model is necessary, due to the many uncertainties mentioned above. This can
be done by matching the dynamic properties with ambient vibration measurements, if such mea-
surements are available, and the verification of the existing damage. A case study of such an ap-
plication is presented in the following.
The reader can find more extensive discussion on the advantages and the limitations of vari-
ous methods of analysis in the papers by Loureno 2002, Penelis 2002 and Syrmakezis 2006.
645
3 CASE STUDY
3.1 Structure considered and numerical model
The estimation of the seismic response and the evaluation of the interventions through linear
analysis are demonstrated for the case study of the Byzantine church of Panhagia Krina in
Chios Island, Greece, dated to the second half of the 12
th
century.
The plan view and the N view of the church are shown in Figures 1 and 2. The church is
composed of three parts: the eastern comprises the main church, the middle is called the narthex
and the western is called the exonarthex, where the main entrance is placed. The bell tower is
located at the western wall of the exonarthex. The three parts are statically independent, except
of the two E columns of the longitudinal arches of the narthex that are connected to the W wall
of the main church at their lower part.

4.45 5.11 11.16
7
.
8
2
8
.1
1
4.40 5.08 11.31
8
.
0
9
Arches
Squinches
Pendentives

Figure 1. Plan view of the church (modified drawing by S. Vogiatzis and M. Paspati, Hellenic Ministry of
Culture).

Figure 2. N view of the church (drawing by S. Vogiatzis and M. Paspati, Hellenic Ministry of Culture).
Main church Narthex Exonarthex
646
The roof of the church consists of vaults and two domes, a large one, compared to the size of
the structure, that covers the central part of the main church and a smaller one over the narthex.
The roof of exonarthex does not exist today. The walls consist of three-leaf masonry with vary-
ing construction types. The outer and inner leaves are made of ceramic bricks and stones and the
middle one of stones and mortar. The vaults, the arches, the drums and the domes are made of
bricks or stones.
The numerical analysis was performed using the code ABAQUS and 3-dimensional, tetrahe-
dral finite elements. All the geometric details of the structure, including the imperfections in the
alignment of the walls, were introduced into the numerical model. The three parts were modeled
as statically independent, except of the partial connection between the main church and the nar-
thex mentioned above. However, pounding among adjacent parts was not considered, since it
would require non-linear analysis.
The masonry was assumed elastic with properties depending on the construction type at each
place. For the determination of the mechanical properties, the method of homogenization was
applied. Since measured data were not available, typical values were assumed, properly mod-
ified in order to account for the existing cracks. Details can be found in the report by Psycharis
et al 2006. According to an investigation by Palieraki et al 2007, timber ties were initially ar-
ranged within the masonry. At the main church, places with timber ties were detected at two le-
vels of the walls and at the base and the top of the dome. The wooden ties at each level were
twin and of cross section 0.100.10 m
2
. In the present condition of the monument, significant
parts of these ties are missing due to the deterioration of the wood, leaving a void at the maso-
nry. For this reason, a zone of 0.15 m height was assumed with reduced Youngs modulus at the
places of missing wooden ties. The existing steel and timber ties, which connect opposite walls,
were also included in the model.
The surrounding soil, in an area of 50.0030.00 m
2
was also incorporated in the model, in or-
der to capture the soil-structure interaction effects. According to available data from three bore-
holes (Edafos Consulting Engineers 1998), the surface layer (of varying depth) is considered as
soft, while the underneath soil is stiff marl. Only the top soft layer was included in the numeri-
cal model with thickness 1.50 m, which can be considered as the average depth of this layer and
which gave the best results concerning the dynamic characteristics of the system in a compari-
son with ambient vibration measurements (presented below). The soil was modeled as elastic
material with Youngs modulus E=25 MPa and Poissons ratio =0.25 according to the results
of the geotechnical investigation. The soil was considered without mass, in order to prevent the
development of stationary waves during the seismic motion. In this way, it behaved only as an
elastic support for the structure. The seismic motion was applied to the external boundaries of
the soil.
The general view of the model is shown in Figure 3, while in Figure 4 the longitudinal and
the transverse sections of the model are depicted. The maximum dimension of the finite ele-
ments was 0.30 m for the structure and 1.50 m for the soil. The walls were modeled with at least
two elements along their width.

Figure 3. NE view of the numerical model.
647

(a) (b)

Figure 4. Sections of the numerical model: (a) longitudinal looking south; (b) transverse looking east.
3.2 Calibration and verification of the numerical model
As mentioned above, there are many uncertainties concerning the assumptions embedded in the
numerical model (exact geometric representation, mechanical properties of the materials, etc). It
is essential, therefore, to calibrate and verify the model before any analysis for the prediction of
the seismic response of the structure is performed. In the case study that is presented here, two
approaches were applied for the verification of the material properties of the structure and the
calibration of the soil thickness adopted in the numerical model: In the first one, the dynamic
properties of the model (fundamental natural periods) were compared to ambient vibration mea-
surements. In the second, the existing crack pattern and the reported damage from previous
earthquakes were explained through the results for a strong earthquake motion.
For the first verification, ambient vibration measurements were performed at the real struc-
ture (Psycharis and Mouzakis 2006). Measurements were obtained at several places including
the top of the two domes, the top of the bell tower (exonarthex) and several places on the walls.
From the records obtained at each place, the predominant periods in each horizontal direction
were calculated, which were compared to the eigenperiods of the numerical model with a simi-
lar dominant motion. It should be noted that masonry historic structures are non-monolithic (es-
pecially the one under consideration that consists of structurally independent parts) and the
structural identification of their modes from ambient vibrations is practically impossible. The
predominant periods measured at each place depict the vibrations of the corresponding section
of the structure and, thus, different values were obtained at each place. For this reason, in order
to compare the ambient vibrations with the eigenmodes of the numerical model, one has to take
under consideration the predominant motion of each mode. The comparison of the periods is
shown in Table 1 and it is seen that the accuracy obtained is quite satisfactory.

Table 1. Comparison of the eigenperiods of the numerical model with the characteristic periods from am-
bient vibration measurements.

Mode

Numerical model Ambient vibrations

Predominant motion

Period (sec) Position/direction Period (sec)
1

Transverse motion of
main church

0.26 Dome of main church /
transverse direction
0.29

2

Longitudinal motion
of exonarthex

0.22 Exonarthex / longitudin-
al direction
0.26

3

Longitudinal motion
of main church

0.21 Dome of main church /
longitudinal direction
0.22

4

Transverse motion of
narthex

0.18 Dome of narthex / trans-
verse direction
0.16
648
In its present condition, the church is severely damaged, apparently as a result of previous
strong earthquakes that have hit the region. There is evidence (Kairou and Paschalides 1998,
Miltiadou et al 2003) that the church and especially the main dome suffered severe damage dur-
ing the 1389 earthquake (
w
=6.7). Extensive damage was also caused to the structure during
the strong earthquake of April 3, 1881 (
w
=6.5) that hit Chios and the nearby costs of Asia Mi-
nor (Kairou and Paschalides 1998): the dome of the main church collapsed; the whole roof, in-
cluding the vaults and the dome, the upper part of the S and N walls and the west wall of the
narthex also collapsed; at exonarthex, part of the bell tower fell. Severe cracking also occurred
to many other structural elements. After the earthquake, restoration of the church was performed
in 1884, during which the two domes and the roof of the narthex were restored or rebuilt. How-
ever, except of the placement of some steel ties, no other serious intervention was made to the
rest of the structure. As a result, extensive cracks still exist at the walls and the vaults of the
roof. More severe are those at the vaults and the E wall of the sanctum, the ones at the W wall
and the ones at the squinches and the pendentives of the main church.
For the second verification of the numerical model, the existing cracks at the walls and the
roof and the collapses that occurred during the 1881 earthquake were explained from the results
of a linear analysis under a strong earthquake base motion. The Roscoe record, obtained during
the Northridge, California 1994 earthquake, was used for this purpose. As explained in the fol-
lowing section, this record is compatible with earthquakes that can hit the area of the monument
with return period of 500 years, as seems to be the case of the 1881 earthquake.
For this base motion, the principal tensile stresses that were developed at the masonry for the
combined effect of gravity and earthquake loads were calculated and the regions, in which these
stresses significantly outrun the strength of the masonry, were compared with the existing crack
pattern of the church.
An example of the obtained results is shown in Figure 5 for the E wall of the main church
(sanctum). In Figure 5a, the existing cracks are shown, while in Figures 5b and c, the principal
tensile stresses are plotted for two different time instances. The areas with principal stresses
larger than 350 kPa are shown with grey color, a value that is significantly larger than the ten-
sile strength of the masonry (assumed to be around 100 kPa). This large value was selected in
order to narrow the areas of excess of the resistance and locate more accurately the places where
cracks should be developed. Comparison of the grey areas of Figures 5b and c with the crack
pattern of Figure 5a shows that the existing cracks are in agreement with the ones that would
occur during the earthquake.
Typical results concerning the vaulted roof are presented in Figure 6. Again, the grey areas
correspond to tensile stresses larger than 350 kPa and are in agreement with the cracks at the
roof. Note that cracks do not exist at the roof of the narthex, because this part of the church was
restored after the 1881 earthquake. In general, the areas of large principal tensile stresses were
in accordance with the existing cracks in all the walls, except for a vertical crack at the upper
part of the S wall of the main church, which could not be explained. This crack, however, could
be the result of secondary effects due to the redistribution of stresses after cracking and/or the
collapse of the dome of the main church during the 1881 earthquake.
The numerical results could also explain the collapse of the dome of the main church and the
collapse of the roof and the walls of the narthex that occurred during the 1881 earthquake. In the
case of the dome of the main church, the principal tensile stresses that were developed at the
most adverse instance of the motion during the Northridge-Roscoe earthquake are shown in
Figure 7. It is seen that high stresses are developed simultaneously at the upper part of almost all
the supporting piers of the dome. Therefore, cracks would be generated at these places, resulting
to a total loss of the shear strength at the base of the dome, which would collapse during such an
earthquake.
In the case of the narthex, it should be noted that the vaulted roof and the dome are supported
by two longitudinal arches, the E piers of which are connected to the W wall of the main church
at their base and this is the only connection between the narthex and the main church. The nu-
merical analysis showed that high tensile stresses would be developed simultaneously during the
earthquake at the base of the supporting piers of the longitudinal arches (Fig. 8). As a result,
shear failure would occur at these places and the arches would lose their capacity to bear the
weight of the roof. It seems, therefore, that the collapse of the narthex during the 1881 earth-
quake started from the two longitudinal arches and expanded to the walls and the roof.
649

(a)

(b)

(c)

Figure 5. Comparison of the existing crack pattern of the E wall (sanctum) of the main church
(drawing a) with the regions of principal tensile stresses larger than 350 kPa (grey areas in draw-
ings b and c) for two time instances during the Northridge-Roscoe earthquake.

(a) (b)

Figure 6. Comparison of the existing crack pattern of the roof (view from below,
drawing a) with the regions of principal tensile stresses larger than 350 kPa (grey
areas in drawing b) for one time instance during the Northridge-Roscoe earthquake.
Main church
Narthex
650

Figure 7. Longitudinal section of the main church and the narthex
(S view), showing principal stresses larger than 350 kPa (grey
areas) at the top of the supporting columns of the main cupola
(Northridge-Roscoe earthquake).

Figure 8. Transverse section of the narthex (W
view), showing principal stresses larger than 350
kPa (grey areas) at the eastern part of the longi-
tudinal supporting arches (Northridge-Roscoe
earthquake).
3.3 Selection of ground motions
For the case study presented here, it was decides to use linear time-history analysis (THA) for
the estimation of the seismic response of the structure. THA was preferred from dynamic modal
analysis, because, as mentioned above, historical structures, as the church under consideration,
do not behave monolithically during earthquakes and there is danger to lose some part of the re-
sponse by restricting the analysis to a limited number of modes. For the selection of the appro-
priate base excitations for the THA, a seismic hazard investigation was performed for the area
of the monument.
The seismic hazard investigation (Psycharis and Taflampas 2006) was performed taking un-
der consideration the earthquakes that occurred from 1904 to 1996 in a distance of 100 km from
the church with magnitude larger than 5.0 and focal distance less than 40 km. It should be noted
651
that Chios and the nearby cost of Asia Minor are located at the most seismic region of the Ae-
gean Sea, characterized by many active faults. The rupture of such a fault on April 3, 1881,
which crosses the S part of the island, caused the most recent devastating earthquake (M
w
=6.5).
Prior to this event, there is evidence that the church suffered severe damage during past earth-
quakes, mainly the one of 1389 (M
w
=6.7).
The results showed that the peak ground acceleration (pga) at the bedrock under the monu-
ment is expected to be 390 cm/sec
2
for seismic events with a return period of 500 years and 560
cm/sec
2
for earthquakes with a return period of 1000 years. Assuming alluvium deposits, the
corresponding values at the ground surface were calculated to be 260 and 370 cm/sec
2
, respec-
tively, using SHAKE software (Schnabel et al 1972). Note that, according to the geotechnical
investigation (Edafos Consulting Engineers 1998), the church is founded on weathered marl, the
depth of which varies from 1.50 to 2.70 m. Under this layer, the soil is characterized as rocky
marl, but with layers of clay. Thus, intermediate values of pga, between the ones mentioned
above for the rock and the alluvia, are expected at the ground surface.
According to the Greek Seismic Code, Chios belongs to the seismic zone II, in which the ef-
fective peak acceleration (epa) is 240 cm/sec
2
for seismic events with a return period of 475
years and 310 cm/sec
2
for a return period of around 1000 years (corresponds to an importance
factor of 1.3). These values should be increased by 25% due to the proximity to active faults;
thus, the design epa values should be 300 and 390 cm/sec
2
, respectively, which are slightly
larger than the ones expected for the ground surface according to the seismic hazard analysis.
Based on the results of the seismic hazard analysis, five seismic records were chosen from the
strong-motion databases Cosmos and European, the characteristics of which are shown in Table
2. The selected records were used unscaled, except of the Imperial Valley that was amplified by
25%, in order to obtain a better agreement with the results of the seismic hazard analysis. The
directions E-W and N-S that are shown in Table 2 indicate the orientation in which each com-
ponent was applied. Note that E-W is the longitudinal direction of the church and N-S the trans-
verse one (short). For the records for which the direction of the fault was known, the strike-
parallel component of each earthquake was applied to the E-W direction, because most of the
faults around Chios are oriented in this direction. For the others, the stronger component was
applied in E-W. The elastic spectra for 5% damping are given in Figure 9, together with the
spectrum of the Greek code for soil type B for comparison.

Table 2. Characteristics of the seismic motions used in the analyses.

Earthquake

Year

Position

Magnitude
(
w
)
Epicentral
distance
(km)
Maximum acceleration (cm/sec
2
)
E-W (long) N-S (short)

Vertical

Azores

1998

6.1 10 412 374

310
Imperial Valley 1979 Station 3 6.4-6.6 11-14 260 212 112
Northridge 1994 Arleta 6.7-6.8 8-10 337 302 541
Northridge 1994 Roscoe 6.7-6.8 11 430 262 279
Parkfield 1966 Temblor 6.2-6.4 11 264 340 129

The selected seismic events are compatible with the seismic motion expected at the site with a
return period of 500 years. This return period was chosen instead of the 1000-years one that is
common for historical structures for two reasons: first because it seems that strong earthquakes,
capable of causing significant damage to the structure, occur in the area approximately every
500 years (the two devastating earthquakes of 1389 and 1881, that caused collapses to the struc-
ture, are 492 years apart) and second, because the structure could not survive the seismic motion
corresponding to 1000-years return period without significant local collapses, even after the ap-
plication of several interventions.

652
0.00 0.40 0.80 1.20 1.60 2.00
Period, T (sec)
0
250
500
750
1000
1250
1500
S
p
e
c
t
r
a
l

a
c
c
e
l
e
r
a
t
i
o
n
,

S
a

(
c
m
/
s
e
c
2
)AZORES
IMPERIAL VALEY-STATION 3
NORTHRIDGE-ARLETA
NORTHRIDGE-ROSCOE
PARKFIELD-TEMBLOR
GREEK CODE

0.00 0.40 0.80 1.20 1.60 2.00
Period, T (sec)
0
250
500
750
1000
1250
1500
S
p
e
c
t
r
a
l

a
c
c
e
l
e
r
a
t
i
o
n
,

S
a

(
c
m
/
s
e
c
2
)

(a) (b)

Figure 9. Elastic spectra for 5% damping of the horizontal components of the seismic motions considered
in the analyses: (a) components applied in the E-W direction (long dimension of the structure); (b) com-
ponents applied in the N-S direction (short dimension of the structure).
3.4 Evaluation and dimensioning of interventions
Usually, interventions do not help supporting the gravity loads, since they are applied long after
construction, when the loads have already been undertaken by the masonry, being thus practi-
cally inactive. However, they are activated during earthquakes by the deformation of the struc-
ture and the redistribution of the stresses after the initiation of damage. Since linear analysis
cannot take under consideration this change in the performance of the structural elements of an
intervention that takes place during the seismic motion, the result should be viewed with caution
and proper engineering judgment.
The evaluation of the effectiveness of two interventions that have been proposed for the mo-
nument under consideration, using linear THA, is presented here: The first concerns the streng-
thening of the masonry walls by grouting and the second the fastening of the walls by steel ties.
The injection of grout in the walls aims to increase their strength, which has been decreased
due to the deterioration of the mortar and the degradation of the wooden ties that were placed
initially within the masonry, and fill the voids that exist where wood used to be. The results
showed that grouting enhances the ability of the structure to resist earthquakes, but it cannot en-
sure that severe damage will not occur in the case of a strong seismic event. This is shown in
Figure 10, in which the tensile principal stresses, which were developed at the N wall at the
most adverse instance during the Northridge-Roscoe earthquake, are shown: in Figure 10a, the
areas marked with grey color correspond to stresses larger than 100 kPa, the assumed present
strength of the masonry; in Figure 10b, grey color denotes the areas with stresses larger than
250 kPa, that corresponds to an upper bound of the increased strength after grouting. It is evi-
dent that the enhancement of the resistance of the walls significantly reduces the areas of poten-
tial damage; however, there are still large regions, where cracks can occur.

(a) (b)

Figure 10. Tensile principal stresses at N wall for the worst time instance during the Northridge-Roscoe
earthquake. Grey areas denote stresses larger than: (a) 100 kPa; (b) 250 kPa.
653
In order to further enhance the structure, it was proposed to fasten the walls of the main
church and the narthex with steel ties (Fig. 11), placed at three levels. The ties prevent the exist-
ing cracks from opening during the seismic motion and offer connection between the main
church and the narthex, which are currently statically and dynamically independent. Analyses
with these parts connected showed a significant reduction in the tensile stresses that are devel-
oped at the eastern part of the two longitudinal arches that support the roof of the narthex. In
Figure 12, the principal stresses that are developed at these arches are shown for the most ad-
verse instance during the Northridge-Roscoe seismic motion. Comparison of this plot with the
one depicted in Figure 8 for the structure without ties demonstrates the significant improvement
in the behavior, gained by the connection of the main church and the narthex: the grey areas, in-
dicating places with stresses larger than 350 kPa, practically disappeared. It is reminded that the
collapse of the narthex during the 1881 earthquake seems to have started from the failure of
these arches.
The above-mentioned examples showed that linear analysis can give valuable information
concerning the effectiveness of interventions by direct comparison of the response with and
without the enhancement. However, the dimensioning of the intervrntions structural elements,
as the steel ties in the second case, is not an easy task, because one should know the forces that
would be developed at them after severe cracking of the masonry, during the seismic motion. In
order to overcome this problem, linear analysis should be used in combination with the imple-
mentation of pre-assumed crack patterns in the numerical model, consistent with unfavorable
scenarios of damage for the intervention under consideration.

Figure 11. Proposed position of steel ties at the walls (plan view).

Figure 12. Tensile principal stresses (grey color
denotes stresses larger than 350 kPa) at the east-
ern part of the longitudinal supporting arches of
the narthex, under the assumption that the main
church and the narthex are connected (North-
ridge-Roscoe earthquake).
654
For the church that is examined here, this approach was applied by introducing longitudinal
and transverse cracks in the numerical model, modeling unfavorable scenarios of damage, in or-
der to take under consideration the redistribution of forces after cracking. More specifically, for
the calculation of the forces that would be developed at the transverse ties, longitudinal cracks
were assumed: (i) along the inner face of the N-S (longitudinal) walls, separating these walls
from the transverse ones, and (ii) at the middle of the church, as shown in Figure 13. Similarly,
for the calculation of the forces induced to the longitudinal ties, transverse cracks were as-
sumed: (i) along the inner faces of the E-W walls of the main church, separating these walls
from the longitudinal ones, and (ii) at the middle of the main church or at the middle of the
sanctum, as shown in Figures 14a and b, respectively.

Cracks at the transverse walls

Cracks at the transverse walls
Longitudinal
crack

Figure 13. Upward view of the model showing the consi-
dered longitudinal cracks for the calculation of the seismic
forces induced to the transverse ties of the walls.

Cracks at the longitudinal walls

Transverse crack


Transverse crack

Figure 17. Upward view of the model showing the considered trans-
verse cracks for the calculation of the seismic forces induced to the
longitudinal ties of the walls.
655
Along the preset lines of the cracks, the walls were disjointed and able to deform indepen-
dently of each other. Thus, the cracks could open during the seismic motion tensioning the ties.
No pounding or any other interaction between the adjacent parts of the walls was considered,
since this would require non-linear analysis.
The results of the analyses for all selected earthquake records showed that the tensile forces at
the ties were significantly increased compared to the structure without cracks. The difference
was not that large in the longitudinal direction, because in the original model the main church
and the narthex were statically independent, but it was very significant in the transverse direc-
tion. For example, the maximum force developed at the ties of the W wall of the narthex was
122 kN considering the longitudinal cracks shown in Fig. 13, while the corresponding force for
the uncracked structure did not exceed 20 kN.
4 SUMMARY AND CONCLUSIONS
The application of linear analysis for the estimation of the seismic response and the evaluation
of several interventions in the case study of a Byzantine church is presented in this paper.
Due to the complicated geometry of the structure, a detailed model was constructed using
three-dimensional finite elements. The parameters concerning the geometry and the materials
that were implemented in the numerical model were calibrated and verified with two approach-
es: (i) by matching the dynamic properties of the model with ambient vibration measurements at
the real structure and (ii) by explaining the existing damage and the reported collapses in pre-
vious earthquakes. For the latter, the existing crack pattern at the walls and the roof was com-
pared to the areas, where tensile stresses, significantly larger than the resistance of the masonry,
are expected to develop during strong earthquakes.
The effectiveness of two proposed interventions was evaluated by comparing the tensile
stresses that were developed at the masonry before and after the application of each interven-
tion. In order to calculate the forces that would be developed at the strengthening elements,
which are needed for their dimensioning, cracks compatible with unfavorable scenarios of dam-
age during future earthquakes were introduced to the numerical model.
The analysis showed that linear analysis, despite its limitations, can be used for the estimation
of the seismic behavior of historical monuments, if the numerical model is properly constructed
and calibrated and the results are viewed with critical engineering judgment. Time-history anal-
ysis is rather preferable compared to dynamic modal analysis, because, usually, historical struc-
tures do not respond monolithically to earthquakes. However, special care should be given to
the selection of the ground motions.
In the contrary, nonlinear analysis, although, theoretically, it is more accurate, might lead to
false conclusions due to its sensitivity to the parameters and the usually unknown properties of
the materials. Also, for structures with complex geometry, there is doubt concerning the accura-
cy obtained by many commonly used software that perform nonlinear analysis. In addition, the
extremely large computational time and effort required in such cases, makes non-linear analysis
practically inapplicable.
5 AKNOWLEDGEMENTS
The case study presented in this paper was conducted in the framework of a research pro-
gramme financed by the Directorate for Technical Research on Restoration (DITEA) of the Hel-
lenic Ministry of Culture, which is gratefully acknowledged. Special thanks are due to A. Kou-
mantos, architect with DITEA, for his help with the architectural details of the church and to L.
Karapitta, graduate student at NTUA, for her help with the numerical model.
6 REFERENCES
ABAQUS Inc. 2004. Abaqus Users Manual, U.S.A.
656
Consortium of Organizations of Strong-Motion Observations Systems (COSMOS), URL:
http://db.cosmos-eq.org/scripts/default.plx.
Edafos Consulting Engineers Ltd 1998. Church of Panhagia Krina in Chios, Geotechnical survey and
study of methods for the improvement of the foundation, DABMM, Hellenic Ministry of Culture,
Athens, Greece (in Greek).
European Strong-Motion Database, URL: http://smbase.itsak.gr.
Kairou, A. and Paschalides, G. 1998. Mapping, documentation and study of proposals for the church of
Panhagia Krina in Chios, DABMM, Hellenic Ministry of Culture, Athens, Greece (in Greek).
Loureno, P. 2002. Computations on historic masonry structures, Progress in Structural Engineering and
Materials 4, 301-319.
Loureno, P. B., Krakowiak, K. J., Fernandes, F. M. and Ramos, L. F. 2007. Failure analysis of Monas-
tery of Jernimos, Lisbon: How to learn from sophisticated numerical models, Engineering Failure
Analysis 14, 280-300.
Miltiadou, A., Zaroyianni, E., Chandrinos, I. 2002. Study of the restoration of the structural system of the
church of Panhagia Krina in Chios, DABMM, Hellenic Ministry of Culture, Athens, Greece (in
Greek).
Miltiadou, A., Delinicola, E., Konides, H., Zaroyianni, E. 2003. Study of support and restoration of the
church of Panhagia Krina in Chios, DABMM, Hellenic Ministry of Culture, Athens, Greece (in
Greek).
Palieraki, V., Vintzilaiou, E., Miltiadou-Fezans, A. 2007. The use of radar technique and boroscopy in
investigating historic masonry: Application of the techniques in Byzantine monuments in Greece, in
Proceedings of the International Symposium: Studies on Historical Heritage (SHH07), Sept. 17-21,
Antalya, Turkey.
Penelis, G. 2002. Structural restoration of historical buildings in seismic areas, Progress in Structural
Engineering and Materials 4, 64-73.
Psycharis, I., Mouzakis, H., Pavlopoulou, E. 2006. Investigation of the dynamic characteristics of the
church of Panhagia Krina in Chios before and after the interventions, Technical report of phase 1.3:
Numerical analysis justification of existing damage, Report No. NTUA/LEE-06/03, Laboratory for
Earthquake Engineering, National Technical University of Athens, Greece (in Greek).
Psycharis, I., Mouzakis, H. 2006. Investigation of the dynamic characteristics of the church of Panhagia
Krina in Chios before and after the interventions, Technical Report of phase 1.1: Ambient vibrations
measurements for the determination of the eigenperiods of the church in the present state, Report No.
NTUA/LEE-06/01, Laboratory for Earthquake Engineering, National Technical University of Athens
(in Greek).
Psycharis, I., Taflampas, I. 2006. Investigation of the dynamic characteristics of the church of Panhagia
Krina in Chios before and after the interventions, Technical Report of phase 1.2: Evaluation of the
seismic hazard, Report No. NTUA/LEE-06/02, Laboratory for Earthquake Engineering, National
Technical University of Athens (in Greek).
Schnabel, P. B., Lysmer, J. and Seed, H. B. 1972. SHAKE: a computer program for earthquake response
analysis of horizontally layered sites, Report EERC 72-12, Earthquake Engineering Research Center,
University of California, Berkeley.
Syrmakezis, C. 2006. Seismic protection of historical structures and monuments, Structural Control
and Health Monitoring 13, 958-979.

657
In Plane Response of old Masonry under Seismic Loads
T. Zimmermann
1
& A. Strauss
2

University of Natural Resources and Applied Life Sciences, Vienna,
Institute for Structural Engineering, Austria
ABSTRACT: Old masonry buildings represent a huge part in many European cities. The typol-
ogy of these buildings and structures varies from region to region, but all have in common to
represent historical importance and architectural heritage.
The paper addresses the problem of evaluation of mechanical properties and the determination
of the safety of these structures against earthquakes, which is a complex challenge. The seismic
impact and the behaviour of masonry under cyclic loads are of high interest for (a) a seismic as-
sessment of existing structures and (b) possibilities for strengthening or retrofit. Thereby bound-
ary conditions like foundation and soil properties have an influence on the seismic behaviour,
beside the material parameters itself.
The resistance of old masonry against earthquakes is normally a combination of experience and
modern engineering knowledge and is due to several uncertainties of mechanical parameters
based on experimental research. Since the introduction of Eurocode 8 a verification of old ma-
sonry structures is hardly possible, especially in case of plain masonry. A reason for this cir-
cumstance is the rough description of material behaviour, i.e. the response factor. A lot of re-
search and experimental work has been done in seismic behaviour of masonry structures in the
last decades, but nevertheless all these research studies are mainly focused on new brick mate-
rial. In this paper results of laboratory tests on old masonry specimens are presented. The objec-
tive of this paper is to present and discuss results regarding the stiffness and its degradation, the
energy dissipation as well as the hysteretic behaviour of plain masonry.
1 INTRODUCTION
Masonry has an inelastic, anisotropic and non-homogeneous material behaviour. In case of old
masonry, it incorporates of brick units and horizontally and vertically oriented mortar interfaces.
However, the material properties of bricks and mortar are different and consequently the behav-
iour of masonry is not elastic even in the range of small deformations and uncracked state re-
spectively. Since the introduction of Eurocode 8 in Austria at May 1
st
2009, buildings must cor-
respond with the requirements of this standard.
Masonry is a typical construction material to withstand compression, but with low shear re-
sistance. Because of this circumstance unreinforced masonry buildings became of high interest,
(a) to gather mechanical properties and their wide scatter, which is characteristic for old ma-
sonry and (b) to obtain appropriate tools for seismic assessment, analysis and retrofit methods.

Structural measures are a combination of both: conceptional design and design layout of the
bearing and non-bearing structure, especially in the case of retrofit. The prior objective is to
minimize the earthquake risk and the vulnerability of buildings, Bachmann 2002.
1
Thomas Zimmermann, Corresponding author. Tel.: +43-1-476 54/5265.
Mail: zi mmer mann. t homas@boku. ac. at
2
Alfred Strauss, Tel.: +43-1-476 54/5254. Mail: al f r ed. st r auss@boku. ac. at
658
According to Eurocode 8 the design seismic action is an earthquake with a return period of
475 years and a 10 % probability of exceedance in 50 years. Whereby the structure should with-
stand the action without local or global collapse, thus retaining its structural integrity and a re-
sidual load bearing capacity after the seismic events "No-collapse requirement" Eurocode 8.
The structure should also withstand a seismic action having a larger probability of occurrence
than the design seismic earthquake. Thereby no damage and/or no associated restriction in us-
ability should occur, where the costs of repair would be disproportionately high in comparison
with the costs of the structure itself "Damage limitation requirement" Eurocode 8.
Parameters regarding the seismic behaviour, i.e. energy dissipation, are important for the as-
sessment of existing structures and the design of retrofit. Masonry buildings can reduce seismic
actions due to non-linear material behaviour in such a way that the design to smaller forces is
sufficient than to forces which would occur by linear elastic behaviour.
Another point of interest is the actual situation of an existing structure. While new structures
remain in an elastic and uncracked state until the first earthquake event occurs, old structures
may have certain damage and crack state form prior events.
The main objective of the research presented in this paper is the seismic response of old ma-
sonry and associated with that stiffness degradation and energy dissipation. Further, mechanical
material properties obtained by experiments are compared to recommended characteristics of
Eurocode 6 and 8.
2 MECHANICAL PROPERTIES
In order to obtain mechanical properties of old masonry laboratory tests were performed.
Thereby basic material parameters, i.e. compressive strength, f
kn
and modulus of elasticity, E
were evaluated on simple compression tests. The specimens have dimensions of l = 59 cm, b =
14 cm and h = 37.5 cm. The tests have been performed in a 5000 kN testing machine, whereby
testing procedure followed up requirements according to EN 1052-1. This means the imposed
vertical load has been continuously increase in such a way that the failure in compression oc-
curred after 15 30 min after the beginning of testing. Table 1 shows the results of vertical
compression tests. Thereby an average value for compressive strength of f
kn
= 6.46 MPa and an
average value for modulus of elasticity of E = 2459 MPa were obtained.

Table 1: Basic material parameters, obtained by vertical compression tests
Specimen vertical loading
kN
compressive strength
MPa
modulus of elasticity
MPa
O-C/1 548 6.65 2200
O-C/2 498 5.86 2313
O-C/3 580 6.87 2863

The second test program focused on the shear behaviour. The specimens have been tested as
vertical cantilevers and have dimensions of l = 59 cm, b = 29 cm and h = 67.5 cm. The vertical
load was imposed by a hydraulic actuator, connected to a gas accumulator to provide a constant
vertical loading. This actuator has a capacity of 500 kN. Horizontal loads were applied by a
horizontal placed, two ways acting hydraulic actuator, with a capacity of 160 kN. It was con-
nected between a rigid steel frame and the top beam of the wall specimen. During testing, dis-
placement of the cantilever walls, induced forces (vertical and horizontal) and strains in ma-
sonry have been measured.
Specimens were constructed with two typical brick layouts (no. 1: stretching bound and no. 2:
heading bond) in order to evaluate the effect onto the shear behaviour.
During the shear tests typical failure states have been observed. A prevalent rocking type of
behaviour with horizontal cracks in the lower horizontal joints was typical for the initial stage.
With increased horizontal displacements cracks in bricks due to compression in the compressed
toes of the specimens appeared, as a result of rocking. The phenomena of rocking is due to (a)
the type of masonry unreinforced, (b) the level of vertical compression o
v
/f
kn
= 0.16 to 0.21
and (c) geometry of the walls h/l = 1.15.
659
Further on, diagonally oriented cracks appeared. These cracks developed through brick units
as well as along mortar interface. In addition, damage included crushing of bricks in the bottom
parts. Finally, at ultimate limit state the wall specimens failed in a shear mode. Thereby a di-
agonal crack developed from one corner at the bottom to the opposite corner at the top.
For both, compression test and shear test results numerical investigations are presented in
Zimmermann et.al. 2009. Thereby experimentally obtained data has been used to calibrate a
constitutive non linear FE Model and numerical simulations have been compared to com-
plex failure surface models according to Mann/Mller 1978 and Lu 2006.
In order to compare experimentally and calculated horizontal loads a shear strength formula-
tion according to Magnes & Calvi 1997, Tomazevic & Klemenc 1997 has been used to asses the
horizontal resistance:
1 +
o
=
tn
v kn
f b
A f
H (1)
where f
tn
= tensile strength of masonry, b = a parameter which takes into account the wall ge-
ometry h/l as well as the shear stress distribution, A = area of the horizontal cross section of the
wall and o
v
= vertical imposed compression stress. Equation 1 is affected by the parameter b
and according to Magnes & Calvi 1997 it can be assumed as: b = 1 for h/l 1, b = h/l for 1 < h/l
< 1.5 and b=1.5 for h/d 1.5.
Table 2 shows calculated and experimentally obtained horizontal forces H and their devia-
tions for different vertical loads V.

Table 2: Deviation between calculated and experimentally measured horizontal resistance for both
brick layouts
V
kN
o
v

MPa
f
tn
MPa
H
calc

kN
H
exp

kN
H
calc
/H
exp
-
Brick layout no. 1
175 1.02 0.12 55.10 59.75 0.08
200 1.17 0.13 61.14 64.10 0.05
225 1.32 0.15 69.75 73.61 0.05
Brick layout no. 2
175 1.02 0.12 55.10 68.50 0.20
200 1.17 0.13 61.14 61.21 0.00
225 1.32 0.15 69.75 76.95 0.09

The tensile strength in table 2 is assumed to be dependent of the imposed vertical compressive
strength. A linear correlation has been obtained by numerical simulations, Zimmermann et.al.
2009. The experimentally measured horizontal resistance is similar to both brick layouts. Hence,
average deviations A = H
calc
/H
exp
for the different compressive stress levels were calculated. o
v

= 1.02 MPa, A = 14 %; o
v
= 1.17 MPa, A = 2 %; o
v
= 1.32 MPa, A = 7 %.
3 STIFFNESS AND STIFFNESS DEGRADATION
Generally, masonry is an inelastic material. Elastic behaviour can be assumed only in the range
of small imposed deflections. For practical use in that state the effective, elastic stiffness, K
e
of
the wall is evaluated as the ratio between a horizontal load, H and the corresponding displace-
ment, d:
d
H
K
e
= (2)
In equation 2, H is equal to 0.67H
max
, the maximum attained horizontal force.
According to the theory of elasticity, the displacement d of a simple cantilever wall can be
calculated as follows, by taking into account bending and shear deformations:
660
A G
h H
I E
h H
d
k
+
o

=
3
(3)
where I = moment of interia of the horizontal cross section of the wall, k = shear coefficient (k
= 1.2 for rectangular section), o is a coefficient depending on boundary conditions (o = 12 when
the wall is fixed at both ends and o = 3 if it is fixed at the bottom and free to rotate on the top), h
= height of the wall specimen and G = shear modulus.
By calculation the elastic stiffness from an experimentally obtained resistance curve and de-
termining the modulus of elasticity via compression tests, shear modulus can be calculated by
introducing equation 2 into equation 3:
2
1 12
|
.
|
\
|

k
k
=
l
h
E
K
h
A
K
G
e
e
(4)
where l = length of the wall. For this definition of the shear modulus G, the compressive stress
o
v
in the wall is neglected. Generally, there is a slightly dependence between G and o
v
.
Since, experimentally obtained horizontal resistance for both brick layouts are in the same
range and since the brick layout of the walls for compression tests was not the same as either of
the two brick layouts for the shear tests, the shear modulus of masonry for both layout types has
been estimated as an average value of all tested specimens. Thereby G = 335 MPa and the re-
sulting ratio between shear modulus and modulus of elasticity, G/E = 0.14.
Compared to the proposed ratio by Eurocode 6 (G/E = 0.40) the obtained ratio is quiet
smaller. It can be concluded that the use of Eurocode 6 recommended G/E ratio leads to an un-
realistic distribution of seismic loads. For numerical simulations a G/E ratio in a range between
0.1 and 0.2 should be considered, which is also in accordance with obtained values by
Tomazevic 2008.
After cracking, typical limit states have been defined to represent the non-linear behaviour of
masonry. These limit states are the crack limit (Index cr), the maximum horizontal load (In-
dex max) and the ultimate limit state (Index u). In accordance with these limit states the increas-
ing degree of damage or damage state can be described by a damage index, I
d
, according to
Tomazevic & Klemenc 1997, Tomazevic et.al. 2006 or Tomazevic et.al. 1996.
The damage index can be divided in the following four types:
- I
d
=0.00: Actually this first state represents no damage because the specimen re-
mains in an elastic state without any cracking.
- I
d
=0.25: The first horizontal cracks occur in mortar joints, especially in the bottom
part of the wall. This index defines the elastic limit state.
- I
d
=0.50: Due to increasing horizontal deflection, the number of horizontal as well
as vertical cracks increases. The cracks appear along mortar joints but
also cracks pass thru brick units. Generally this crack pattern can be ob-
served at the applied maximum horizontal force.
- I
d
=1.00: Again, due to increasing horizontal deflection, the number and width of
cracks increases. Thereby also crushing of bricks in the compressed zone
can be observed. Finally there is a collapse of the specimen. This state
represents the ultimate limit state.

Stiffness degradation occurs, when the walls are subjected to repeated pseudo dynamic load
reversals. Figure 1 a shows the stiffness degradation. Thereby the stiffness ratio is plotted with
respect to the damage index. Stiffness degradation is in the same range for both brick layouts.
The shape of stiffness degradation can be described by the following simple exponential decay
function:
d
I
e
e
K
K
|
o = (5)
where o and | are parameters of stiffness degradation. As found out by curve fitting of experi-
mentally obtained degradation curves, parameter o =1.0 and | =4.1. At the occurrence of the
661
first cracks (I
d
= 0.25) and at the state of maximum horizontal force (I
d
= 0.50) the stiffness ra-
tio drops down dramatically. Between the last two damage indexes rather small stiffness degra-
dation takes place. In the fist damage states the plotted equation 5 in figure 1 a is comparable to
an average value of obtained experimental data. In the last damage states it is on the conserva-
tive side.

(a) K/K
e
and damage index I
d

(b) d/d
u
and damage index I
d

Figure 1 Correlation between (a) stiffness ratio K/K
e
and damage index I
d
and (b) deformation ratio
d/d
u
and damage index I
d

For a seismic assessment of new structures and structures without any damage a stiffness ratio
of 1.0 or slightly below can be assumed because they remain in an elastic state. On the other
side for a seismic assessment or a retrofit for old structures the stiffness ratio should be assumed
<< 1.0, since a crack state from earlier events can not be excluded.
Figure 1 b shows the dependence between damage index and horizontal displacement. The
displacement is normalized by the maximum attained displacement. Slightly larger displace-
ment amplitudes have been needed to reach characteristic damage states in the case of brick
layout no. 2 than in the case of brick layout no. 1.
4 ENERGY DISSIPATION AND HYSTERETIC BEHAVIOUR
Energy dissipation capacity has been evaluated on the basis of experimentally obtained hyster-
etic relationship between horizontal displacement and horizontal force. It represents one of the
important parameters of seismic resistance.
Cumulative input energy E
inp
is needed to deform the wall according to an imposed deforma-
tion. It has been defined as the cumulative work of the horizontal actuator, needed to cause a
given deformation status. In one cycle of loading (push and pull), AE
inp
is calculated as the sum
of the area under the hysteretic loop, see figure 2 a:
=
A =
j
i
i inp j inp
E E
1
, ,
;
( )
( )
}
+
=
=
= A
1
0
0
,
i
i
H
H
inp i inp
dE E (6)
where:
( )
=
,
, 0
d Hd
dE
inp

if
if ( )
( ) 0
0
>
s
d iHd
d Hd

and d(d) = differential displacement increment and H = applied horizontal force.
The amount of dissipated energy, AE
hyst
in one cycle of loading has been defined as the area
of the hysteretic loop between two consecutive maximum deformation states. The cumulative
dissipated energy E
hyst
has been defined as the sum of the dissipated energy of all loading cycles
i from the beginning until the end of testing, see figure 2 b:
=
A =
j
i
i hyst j hyst
E E
1
, ,
; ( )
( )
( )
}
+
=
=
= A
1
0
0
,
i
i
H
H
i hyst
d Hd E (7)
662

(a) Input energy

(b) Dissipated energy
Figure 2 Input energy and dissipated energy during one cycle of loading, from Tomazevic et.al. 1996

Figure 3 a represents the relationship between cumulative input energy and damage index and
figure 3 b represents the relationship between cumulative energy dissipation and damage index.
As can be seen, there is practically no difference in information for the input energy and only
slightly differences in higher damage states for the dissipated energy regarding both brick lay-
outs.

(a) Input energy E
inp
and damage index

(b) Dissipated energy E
hyst
and damage index
Figure 3 Correlation between (a) input energy E
inp
and damage index I
d
and (b) dissipated energy E
hyst

and damage index I
d

5 CONCLUSION
Typical failure mechanisms of masonry under seismic load can be summarized by the following
points:
- failure due to rocking: with an increasing horizontal deflection, horizontal joints in
the lower part of the specimen fail in tension and the shear load is carried by the ma-
sonry part under compression. The final failure occurs by overturning of the speci-
men and a local failure in compression in the compressed toes of the wall.
- failure due to sliding: again, cracks in horizontal joints develops and go along the in-
terface between brick and mortar. This failure type is typical for low vertical com-
pressive stress levels.
- failure in shear: the crack pattern consists of diagonally oriented cracks, which follow
along head and bed joints of pass through brick units. This failure depends on the
relative strength of mortar and bricks as well as the strength of the interface.
- failure in tension: due to combination of normal stress and shear stress a inclined
principal stress state arises and if tensile strength of bricks is exceeded, the brick
units fail in tension.
- failure in compression: mainly for high compressive stress levels, but it is also typical
for local failure mechanism in the compressed toes of the wall.

663
The elastic stiffness K
e
is for both layout types in the same range and stiffness degradation can
be described by a simple exponential decay function (equation 5).
Basic mechanical characteristics (modulus of elasticity and shear modulus) have been ob-
tained by experimental testing and verified by numerical simulations. The resulting ratio be-
tween shear modulus and modulus of elasticity, G/E = 0.14. Thus, the ratio is quiet smaller than
the proposed ratio by Eurocode 6 (G/E = 0.40). Hence, for numerical simulations a ratio be-
tween 0.1 0.2 should be considered.

Regarding energy dissipation the following points can be concluded:
- Input energy for brick layout no. 1 is for all damage states in the scattering range of
the input energy for brick layout no. 2. Thus, there is practically no difference be-
tween the two layouts.
- At least for the dissipated energy at higher damage states (I
d
= 0.8 to 1.0) a difference
between the two layouts can be ascertained. Whereby E
hyst
for brick layout no. 1 is
slightly higher than for layout no. 2.
- Due to specimen and testing uncertainties, there is practically no difference in infor-
mation for the energy dissipation.
6 ACKNOWLEDGEMENT
Research results discussed in this paper has been carried out within the European research pro-
ject SEISMID, supported and financed in cooperation with the Centre for Innovation and Tech-
nology (ZIT).
REFERENCES
Bachmann, H. 2002: Erdbebensicherung von Bauwerken. ISBN: 3-7643-6941-8. Birkhuserverlag, Basel
Boston Berlin.
Eurocode 6: Design of masonry structures, Part 1-1: General rules for reinforced and unreinforced ma-
sonry structures, 2006.
Eurocode 8: Design of structures for earthquake resistance, Part 1: General rules, seismic actions and
rules for buildings, 2005.
Lu, S. 2006: Evaluierung der Erdbebensicherheit von mageblichen Mauerwerkshochbauten fr das
Katastrophenmanagement. PhD, Institute of Building Construction and Technology, University of
Technology Vienna.
Magnes, G. & Calvi, G.M. 1997: In-Plane Seismic Response of Brick Masonry. Earthquake Engineering
and Structural Dynamics, 26:1091 1112.
Mann, W. & Mller, H. 1978: Schubtragfhigkeit von Mauerwerk. In Mauerwerk Kalender, ISBN:
3433007918. Ernst & Sohn, Berlin.
Tomazevic, M. 2008: Shear resistance of masonry walls and Eurocode 6: shear versus tensile strength of
masonry. Material and Structures, online.
Tomazevic, M. & Klemenc I. 1997: Seismic behaviour of confined Masonry Walls. Earthquake Engineer-
ing and Structural Dynamics, 26:1059 1071.
Tomazevic, M., Lutman, M. & Bosiljkov, V. 2006: Robustness of hollow clay masonry units and seismic
behaviour of masonry walls. Construction and Building Materials, 20:1028 1039.
Tomazevic, M., Lutman, M. & Petkovic, L. 1996: Seismic Behaviour of Masonry Walls: Experimental
Simulation. Journal of Structural Engineering, 122:1040 1047.
Zimmermann, T., Strauss, A. & Bergmeister, K. 2009: Numerical investigations of historic masonry walls
under normal and shear load. Construction and Building Materials, preprint.

664
1 INTRODUCTION

The temple of Zeus at Nemea was built around 330 B.C. and is notable for the unique slender
proportions of its exterior colonnade of Doric style which consists of 32 columns as shown in
the plan view of Figure 1. Among the 32 columns only one has survived (K-31) together with
two columns from the entrance of the Pronaos as shown in the 1982 photograph of Figure 2.
The isometric drawing of Figure 3 depicts the location of the three columns which remain stand-
ing from the ancient times.
For centuries the temple remained in ruins with its building elements partially embedded in
the ground (see Figure 4). This situation is most destructive for the integrity and overall condi-
tions of the ancient material since large surfaces are exposed to an adverse environment which
includes rain, snow, icing, roots, even microorganisms which have drastically eroded the origi-
nal material.

Figure 1. Plan view of the temple of Zeus at Nemea.

The reconstruction of the north-east corner of the temple of Zeus
at Nemea, Greece

N. Makris
Univeristy of Patras, Greece

ABSTRACT: The paper summarizes the reconstruction of the north-east corner of the Temple
of Zeus at Nemea which includes the crepida and the four multidrum columns together with
their capital. The various challenges, achievements and perspectives of this project are dis-
cussed.

665

Figure 2. View of the temple of Zeus at Nemea from south-west. Photograph of 1982.

Figure 3. Isometric drawing that shows the location of the three columns standing from the ancient times.

31 K
666

Figure 4. Aerial photograph of the temple of Zeus at Nemea on July 1977.
2 PILOT RECONSTRUCTION OF K-25 AND K-26
The preservation of the original building material (see Figure 4) is probably one of the most
convincing arguments for proceeding with the reconstruction of the Temple. The idea of recon-
structing the temple was first put forward in 1978 by Professor Stephen G. Miller from the Uni-
versity of California Berkeley. Soon after that he initiated a fundraising campaign to finance a
detailed re-synthesis study of the temple. During the period 1980-1982 Professor Frederic A.
Cooper from the University of Minnesota led the effort of measuring and recording approxi-
mately 1100 ancient building blocks and subsequently conducted the re-synthesis study for the
reconstruction of the temple. Following the work of Professor Cooper, the Greek Ministry of
Culture issued the permit for the pilot reconstruction of two columns, that of K-25 and K-26,
along the north side of the temple. The reconstruction of the two columns was successfully
completed in the summer of 2002 and as a result of this effort the temple of Zeus today consists
of five standing columns as the recent photograph of Figure 5 shows. The isometric drawing of
Figure 5 assists with the understanding of the morphology of the temple as it stands today. More
information of the history of the pilot reconstruction can be found in the article by Miller
(2000).

3 THE RECONSTRUCTION OF K-27, K-28, K-29 AND K-30
3.1 The shape of the temple in the near future
Following the successful reconstruction of K-25 and K-26, the reconstruction team of the
temple of Zeus at Nemea submitted via the American School of Classical Studies a reconstruc-
tion study for four additional columns in order to complete the north-east corner of the temple.
The re-synthesis study conducted by F. A. Cooper was carefully re-evaluated in order to

667

Figure 5. Left: The temple of Zeus at Nemea from north-west. The two front columns are those recon-
structed by the pilot reconstruction. The three ancient columns are distinguished in the background.
Right: Isometric drawing that shows the present layout with five columns standing.

confirm the precise location of the ancient building components in association with recent dis-
coveries of ancient material. For each building element (column drums, capitals, architraves and
frieze) it was prepared an individual file which describes its pathology together with the retrofit
proposal. Figure 6 shows two among the many detailed drawings that were submitted to support
the reconstruction proposal. The new reconstruction study of columns K-27, K-28, K-29 and
K-30 was approved by the Greek Ministry of Culture in March 2004. The isometric drawing of
Figure 7 shows schematically the morphology that the temple of Zeus at Nemea will assume
once this second reconstruction phase will be completed. The future morphology of the temple
(Figure 7) will improve appreciably the understanding of the architecture of the temple since its
current view creates some confusion to the visitor (see Figure 5). With the reconstruction of the
four columns of the north-east corner (K-27, K-28, K-29 and K-30), there is a closure of the ex-
terior colonnade starting from the two recently reconstructed columns, K-25 and K-26 to the
solely standing column from the ancient times, K-31. The closure of the north-east corner will
also offer to the visitor a sense of the interior space and magnitude of the monument.
25 K
26 K
668

Figure 6. Re-synthesis of architectural elements and reconstruction proposal of the north side of the tem-
ple (top) and of the east side (bottom). Drawings No. 18 and No. 22 from Katerina Skleres architectural
studies.
669

Figure 7. Isometric drawing that shows the future layout of the temple after the end of second phase of
reconstruction which is under way.
3.2 The reconstruction of the crepida
Before reconstructing the columns it is necessary to reconstruct the wings of the base of the
temple which consist of four layers of massive rectangular stones which support the gravity
loads of the monument. Figure 8 (left) shows the four layers to be reconstructed at the north-east
corner of the temple while Figure 8 (right) shows the major damage of the stylobate and of the
second layer in the vicinity of the support of column K-31. Figure 19 shows ancient stone
blocks from the base that needed to be pieced. In order to complete the section of the crepida
and the area where the four new columns are erected we prepared and installed 70 new stones
with dimensions that exceed (2.0 m)x(2.0 m)x(0.4 m) and weight that exceeds 3.5 tons. In addi-
tion to this work 15 ancient blocks need to be pieced together. It should be emphasized that the
reconstruction of the crepida has consumed a substantial fraction of our reconstruction effort
mostly because the rectangular blocks of the crepida were much more attractive to the subse-
quent occupants of the nearby area than the cylindrical column drums. The limestone that we
use for the preparation of the new stones is extracted from nearby area, very close to the ancient
quarries.

Figure 8. Left: The north-east corner of the crepida where the four layers which need to be reconstructed
prior to the erection of the columns are distinguished. Right: The poor condition of stylobate and the 2
nd

base layer at the vicinity of the support of column K-31 which remains standing from the ancient times.

25 K
26 K
27 K
28 K
29 K
30 K
31 K
670

Figure 9. Left: Two pieces of ancient stone of crepida that have been bonded. Right: Two frag-
ments of an ancient crepida block that have been retrofitted with new stone.

3.3 Extraction of new material
By the end of the third quarter of year 2005 we started the planning of the extraction of new
stone material from a nearby rocky site that was excavated in the past during the construction of
the Korinth Tripolis freeway. The original design of the freeway was altered and that site was
abandoned. Nevertheless, the original freeway excavation had revealed a healthy front of lime-
stone as shown in Figure 10. By October 2005 we started a major operation by drilling holes in
the limestone front in order to pass the cutting wire which created vertical sections on the rocky
volune at a 2m (6.0ft) spacing. Figure 10 shows nearly rectangular stone blocks which were cre-
ated from the wire cuts and subsequently were pushed away with steel flat jacks (steel pillows)
which were inflated hydraulically as shown in Figure 11.

Figure 10. Left: View of a healthy front of limestone showing the vertical sections generated with the cut-
ting wire. Right: Nearly rectangular limestone block which will provide the new material that we need for
the crepida and the columns.

671

Figure 11. Close view of an inexpensive (15 dollar) steel flat jack (steel pillow) which has been inflated
hydraulically to push away the cutted stone block.

Large masses of stone are transported at the periphery of the archeological site and they are
sliced with an electrical wire-saw at the desired dimensions. Figure 12 shows the sawing of a
large stone mass while Figure 13 shows new stone blocks that have been prepared for the recon-
struction of the stylobate.
3.4 Reconstruction of the columns
Along with the preparation of the crepida, significant progress has been accomplished with
the retrofit of the ancient drums and the production of new in order to reconstruct the columns.
The criteria that govern the state of practice of retrofitting emerge from the need to satisfy struc-
tural stability together with the need to achieve morphologic continuity. Numerical and experi-
mental studies have demonstrated that the small material loss at the base of the column drums
reduce substantially the dynamic stability of the columns (Psycharis et al. 2000). Further to the
issues of the structural stability, the loss of material and the visible discontinuities between suc-
cessive drums alter the distinctive elements of classical architecture. These two reasons (struc-
tural stability and morphologic continuity) make imperative the preparation and bonding of a
large number of fillings for the column drums, capitals and epistyles. Figure 14 shows photo-
graphs of new stone bonded on the eroded drums of the columns to be reconstructed. The filings
are worked out to the extent that they reach the original surface of the member (see Figure 15).
The edges of the flute are shaped to the extent that the material can sustain the fine workman-
ship of the stone carvers.

672

Figure 12. Wire-sawing of a large formless stone to produce new rectangular stone blocks.

Figure 13. Stack of new stones of stylobate that were extracted via wire-sawing of formless large stones.

673

Figure 14. New stone fillings that have been bonded with eroded drums of the columns to be recon-
structed.

Figure 15. Worked out fillings where their final shape reaches the original surface of the member.
3.5 Reconstruction of the epistyles and frieze
The approved reconstruction study includes the reconstruction of the epistyles and the frieze
above. Following a considerable effort we have identified the 10 epistyles (5 exterior and 5 inte-
rior) that will connect the top of the four columns to be reconstructed with the column standing
from the ancient times located to the east; and the two recently reconstructed columns along the
north side of the temple. The placement of the epistyles improves the seismic stability of the
solitary columns (Psycharis 2003). Two among the 10 epistyles have survived nearly intact
while the rest have been broken in two or more pieces which will be re-bonded and filled. Given
that the epistyles are subjected to combined bending and shear their structural capacity deserves
attention. A comprehensive analysis of the stresses and strains that develop within the retrofitted
epistyles was conducted by Makris and Psychogios (2004). Figure 16 shows a close view of the
broken epistyles, while Figure 17 shows schematic sections of a retrofitted epistyle. Figure 18
shows the distribution of normal stresses of the retrofitted epistyle under gravity loads.
3.6 The shape of the temple today
Figures 19 and 20 are recent photographs of the north-east view of the temple which show the
reconstruction progress during the last five years.
674

Figure 16. Close view of broken epistyles where the material loss is distinguished.

Figure 17. Ancient stone fragments, new stone fillings and reinforcement with titanium bars of epistyle
A.2 external.

Figure 18. Distribution of normal stresses along axis x of external epistyle A.2.

675

Figure 19. View of the north-east corner of the Temple of Zeus form the north showing the completed
crepida (stone levels , , , lighter color), together with the capitals atop K-29 and K-30.

Figure 20. View of the East side of the Temple.

676
4 REFERENCES
English

Miller, S. G. 2000. The temple of Nemean Zeus. A California Landmark. Chronicle of the University of
California, No. 4, Fall 2000 issue.
Psycharis, J. N., Papastamatiou, D. Y. & Alexandris, A. P. 2000. Parametric investigation of the stability
of classical columns under harmonic earthquake excitations, Earthquake Engineering & Structural
Dynamics, 29:1093-1109.
Psycharis, J. N., Lemos, J. V., Papastamatiou, D. Y., Zambas, C. & Papantonopoulos, C. 2003. Numeri-
cal study of the seismic behavior of a part of the Parthenon Pronaos, Earthquake Engineering & Struc-
tural Dynamics, 32:2063-2084

Greek

Makris, N., & Psychogios, T. 2004. Static and dynamic analysis of columns and part of the entablature
of the NE corner of the temple of Zeus at Nemea, American School of Classical Studies, Athens.
Zambas, K. 2004. The reconstruction of temple of Zeus at Nemea: The principles, the objectives and the
perspectives of the intervention, American School of Classical Studies, Athens.
677
Aiko FURUKAWA, Kyushu University
Junji KIYONO, Kyoto University
Kenzo TOKI, Ritsumeikan University
Dynamic Analysis of Rupture
Process of Masonry Buildings
using New Simulation Method
Background & Purpose
The purpose of this study is to develop an analysis method
for the seismic behavior of masonry structures, and to
investigate the effective reinforcement measures.
During the Iran Bam Earthquake,
about 49,000 buildings were
collapsed or damaged.
More than 43,200 people were
killed, and about 15,000 people
were injured.
About one third of the inhabitants
died mainly due to structural failure
of masonry buildings.
678
Development of Analysis Method
Existing Analysis Methods
Elastic Behavior
Failure Occurence
Collapse
Failure Propagation
FEM
(Continuum)
DEM
(Discontinuum)
(Lacks accuracy)
(accurate)
Analysis Parameters
Rupture Process
theoretically
determined
experimentally
determined
There are no analysis methods which can trace a whole
process from elastic to failure to collapse behavior.
679
Development of Analysis Method -DEECA-
DEECA stands for Discrete Element Elastic to Collapse Analysis.
It is a discrete element modeling with special treatment of springs.
Elastic Behavior
Failure Occurence
Collapse
Failure Propagation
DEECA
Analysis Parameters
Rupture Process
theoretically determined
DEECA
-Discrete Element Elastic to Collapse Analysis-
A structure is modeled as an assembly of rigid elements.
Structural deformation is modeled by overlapping of elements.
DEECA divides surfaces of the elements into segments, and a
spring and a dashpot are attached to each segment.
This enables determining spring constant theoretically based on
the material properties and the segment area.
Multiple springs & Multiple dashpots Segmentation of surface
Element
(brick)
A spring and
a dashpot
are attached
to each point.
680
DEECA
Forces in the normal and the shear directions (F
n
, F
s
) are
estimated theoretically from the relative displacements
between elements (u
n
, u
s
).
=
V
n
n
dV
n
n u E
F
) (
) 1 (
2
l

+
=
V
s
s
dV
n
n u E
F
) (
) 1 ( 2 l
G
dA
Fn
Fs
Element
(brick)
ft fm
c
=tan
DEECA
Normal stress
Shear stress
Elastic Behavior
Failure Criterion
Contact Detection
Compression failure
Shear failure
Tensile failure
Contact force is computed from
a contact spring and a dashpot.
Restoring force is computed
from restoring springs.
Failure occurrence is judged for each restoring spring.
A restoring spring is replaced with a contact spring and a dashpot.
681
DEECA
t x x x
t x x x
t t t t t
t t t t t
+ =
+ =

& & & &
&
Time integration
[ ] ( ) m f F x
x t x t
/ + = & &
Equation of motion
Contact force
Body force (gravity, earthquake)
mass
Time interval
The position of individual elements are obtained by solving
equations of motion step by step.
Drawbacks of DEM (Distinct Element Method)
Vertices and edges of elements are rounded.
Contact between elements is simply
classified into 4 patterns.
Contact area is not generally considered.
Vertex to Vertex Vertex to Edge Vertex to Surface
Spring constant cannot be determined theoretically.
Edge to Edge
682
Modeling Strategy of Masonry
Modeling of Wall
Bricks are called units. Mortar conjunction is called a joint.
Units are modeled with rigid elements.
J oint has zero thickness and modeled with springs.
Elasticity of bricks and mortar is included in springs.
Brick Mortar Unit Joint
Actual structure Modelling of masonry wall
683
Modeling of Failure
Tensile failure
(cracking in joint)
Shear failure
(sliding along joints
at low normal stress)
Compression failure
(splitting of bricks in tension
as a result of mortar dilatancy
at high normal stress)
ft fm
c
=tan
Normal stress
Shear stress
Compression failure
Shear failure
Tensile failure
Mohr Circular Cap Model
Verification of Developed Method
Elastic Behavior of Masonry Wall
- Simulation of Free Vibration
- Results by DEECA are compared with those by FEM
684
Analytical Model
z
y x
3m
0.2m
2.5m
Element 1
Adobe
bricks
Mass density(kg/m
3
) 1.810
3
Youngs
modulus(N/m
2
)
9.810
7
Poissons ratio 0.25
Mortar
Mass density(kg/m
3
) 1.810
3
Youngs
modulus(N/m
2
)
9.810
7
Poissons ratio 0.25
Structural damping
(mass-proportional damping)
5% at
1
st
mode
Material Properties
Brick
20cm x 10cm x 10cm
(segment size=0.125cm
2
)
Analytical Condition
Impact wave of 100gal is input in the out-of-plane direction (y).
Responses are computed by DEECA and the FEM.
Time interval: DEECA (dt=1.010
-4
s) explicit method
FEM (dt=1.010
-2
s) implicit method
Time (s)
0
A
c
c
e
l
e
r
a
t
i
o
n
(
g
a
l
)
100gal
Impact acceleration

z
y x
3m
0.2m
2.5m
Element 1
Responses computated
Input
direction
685
Results (FEM vs DEECA)
Response time histories computed using DEECA agrees well with
those using FEM.
Displacement
Velocity
Acceleration
DEECA can simulate the elastic behavior with high accuracy.
FEM DEECA
Seismic Analysis of Masonry Building
Ground motion from the 2003 Bam earthquake is input.
Effect of roof type (Flat roof vs Vault roof)
Effect of mortar strength (Weak vs Normal vs Strong)
Effect of thickness of the bearing wall (Thin vs Thick)
Effect of reinforcement with wood columns and beams
686
Flat Roof
Analytical Models
Vault Roof
5m
3m
Door 1.2m 2.0m
2.4m
Window 1.01.0m
0.2m
Bearing wall
20cm
1.5m
Roof thickness
0.2m
Nonbearing
wall 20cm
3m
Wood
Beam
Input Ground Motion
Accelerograms recorded at Bam station,
Iran, during the 2003 Bam earthquake
The direction of L component is N278E.
T component
(x direction)
L component
(y direction)
V component
(z direction)
x
y
z
687
Adobe
Bricks
Mass density(kg/m
3
) 1.810
3
Youngs
modulus(N/m
2
)
9.810
7
Poissons ratio 0.25
Mortar
Mass density(kg/m
3
) 1.810
3
Youngs
modulus(N/m
2
)
9.810
7
Poissons ratio 0.25
Wood
Beam
Mass density(kg/m
3
) 7.010
2
Youngs
modulus(N/m
2
)
6.310
8
Poissons ratio 0.3
Mass-proportional damping
5% at
1
st
mode
Material Properties
Tensile strength 110. MPa
Bond strength 9.0 MPa
Friction angle 0
Compressive
strength
45.0 MPa
Strength of Wooden Beam
Mortar Strength
1) Weak Mortar field test of collapsed buildings in Bam
2) Normal Mortar laboratory test of typical adobe bricks in Iran
3) Strong Mortar laboratory test of mortar used in Europe
Tensile test Shear strength
Field test in Bam for collapsed buildings
688
Varying Mortar Strength
Collapsed buildings
in Bam
(Field test)
Typical adobe
bricks in Iran
(laboratory test)
Mortar used
in Europe
(laboratory test)
Tensile strength 0.0046 MPa 0.127 MPa 1.42MPa
Bond strength 0.0029 MPa 0.0304 MPa 1.19MPa
Friction angle 32
no data (32
used)
36
Compressive
strength
no data
(0.49MPa used)
0.49MPa 26.6MPa
Ghannad, M.A. et al., 2006. Kiyono J . et al., 2004.
Is increasing the mortar strength effective for increasing
the seismic safety of buildings ?
Weak Mortar Normal Mortar
Marzahn, G., 1998.
Strong Mortar
Varying Failure Mechanism
The stability of the roof depends on two bearing walls.
The failure of the roof is avoided if the bearing walls are
supported somehow regardless of damage to nonbearing walls.
Bam Earthquake, Iran (2003) Friuli Earthquake, Italy (1976)
The nonbearing walls were demolished,
but the roof did not collapsed because
the bearing walls were supported by
the adjacent buildings.
The nonbearing walls were still standing,
even though the roof and bearing walls
were collapsed.
689
Analytical Cases
Thin Bearing
wall (20cm)
Thick Bearing
wall (40cm)
Weak
Mortar
Normal
Mortar
Reinforced with
wooden columns
and beams
Strong Mortar
Analytical Cases
Thin Bearing
wall (20cm)
Thick Bearing
wall (40cm)
Weak
Mortar
Normal
Mortar
Reinforced with
wooden columns
and beams
Strong Mortar
690
x
y
z
1
st
Natural Frequencies
Bearing walls
thickness
Roof
type
x y
20cm
Flat 6.64 Hz 6.05 Hz
Vault 3.91 Hz 3.13 Hz
40cm
Reinforced
20cm
1
st
natural frequency is obtained by
FFT of free vibration responses.
Models with flat roof has higher natural
frequencies since wood beams are
stiffer and lighter than adobe bricks.
x
y
z
x
y
z
Results of Simulation
will be shown by animations
691
Effects of Reinforcement (Vault roof)
Strong Mortar, 20cm
Thick bearing walls, 40cm
Columns & Beams, 20cm
Conclusions
A new analysis method, DEECA, is developed that can simulate elastic, failure
and collapse behaviors.
The elastic behavior of the masonry wall analyzed using DEECA agrees well
with that using the FEM.
The flat roof model is stronger than the vault roof model since the roof is lighter
and the roof reinforce the connection between bricks and walls.
The vault roof model is unstable since the roof moves the bearing walls in the
out-of-plane direction.
Failure mechanism depends on the strength of the bearing walls.
Increasing the mortar strength is effective to increase the seismic safety of the
both buildings.
Increasing the thickness of bearing walls and reinforcing with columns and
beams is only effective for the flat roof model.
692

3rd Greece Japan 2009 Proceedings

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

3rd Greece Japan 2009 Proceedings

Uploaded by

Copyright:

Available Formats

Editors

A study on the behavior of raked piles in seismically liquefiable

Earth pressure meter

approximately doubles the stiffnesses according to our FE results, or triples them

, and curvature (local),

, ductility demands and the maximum drift ra-

, and the maximum drift ratio

is defined as the maximum curvature

is the ratio of the maximum displace-

1) ratio is 3.3 instead of 3.7 in case of smaller pier

) demand for different parameters examined

) demand for different parameters examined

are small and so are the values of

. The maximum drift ratio

and decrease in the mean value of

increases for increased soil compliance.

are observed, when the maximum drift ra-

decreases as the soil

1) ratio has a higher value for a soft soil,

demand may refrain the local ductility demand

at levels lower than what cor-

demand, with consequent reduc-

correlations are proposed herein.

0 1000 2000 3000

0 500 1000 1500

0 1000 2000 3000

0 500 1000 1500

; perimeter length (in meter) of pile

0 5 400 15 10 800 1200

0 5 400 15 10 800 1200

0 5 400 15 10 800 1200

is proportional to the norm of the plastic component of stretching

Impact of Moderate Earthquakes in Post-Bhuj Era: Case Study of

More than M7.0

More than M7.0

ratio on maximum slippage with respect to

dy = 0.05 m dy = 0.05 m dy = 0.01 m dy = 0.01 m

' ' ' = + +

are prescribed functions. In eqns. (18),

of the beam-column of example 5.

, respectively. For the case of a

represent the lateral and rotational stiffness of the (elastomeric) bearings.

is the plastic strain rate, C the initial kinematic hardening modulus

= 5.6 ; (b) Pushover analysis of the new design concept. Since

is introduced, based on foundation rotation ; the estimated capacity

= 42 is almost an order of magnitude larger (compared to conventional design).

faulting through H = 15 m Fontainebleau sand deposit, at distance s

und with time

5.5. Validation Against Theoretical Studies

), is varied parametrically, covering from a shallow (H

, while the soil between the piles

. Soil arching is measured with the ratio u

. If this ratio ranges between 1 and at

ratios arching is not effective.

, and c 3 kPa. The bottom soil layer is

600 kPa. The interface properties are

for cohesionless soils, as

for clayey soils of

for the loose sand

for the dense sand

for the soft rock

for the rock

is expressed as a function of the height