Pelecanos L 2014 PHD Thesis PDF

SEISMIC RESPONSE AND
ANALYSIS OF EARTH DAMS
A thesis submitted in partial fulfillment of the requirements of

the Doctor of Philosophy (PhD) degree
and the Diploma of Imperial College (DIC).
LOIZOS PELECANOS
MEng(Hons) ACGI
Department of Civil & Environmental Engineering

Imperial College of Science, Technology & Medicine
London, United Kingdom

September 2013
SEISMIC RESPONSE AND ANALYSIS OF EARTH DAMS
This page is intentionally left blank.
ii
This thesis is dedicted to my beloved family,

my wife Katerina, my respected parents
Panayiotis and Irene, and my brothers
Fotis and Angelos.
iii
Declaration
The work presented in this thesis was carried out in the Department
of Civil & Environmental Engineering at Imperial College London from the
3rd of October 2009. It was jointly supervised by Dr Stavroula Kontoe and
Professor Lidija Zdravković and was funded by the Engineering and Physical
Sciences Research Council (EPSRC), UK.
This thesis is the result of my own work and any quotation from, or
description of the work of others is acknowledged herein by reference to the
sources, whether published or unpublished. This thesis is not the same as
any that I have submitted for any degree, diploma or other qualification at
any other university. No part of this thesis has been or is being concurrently
submitted for any such degree, diploma or other qualification. This document
is available online at www.imperial.ac.uk, it is less than 100,000 words long,
contains less than 300 figures and less than 20 tables.
The copyright of this thesis rests with the author and is made available
under a Creative Commons Attribution Non-Commercial No Derivatives li-
cence. Researchers are free to copy, distribute or transmit the thesis on the
condition that they attribute it, that they do not use it for commercial pur-
poses and that they do not alter, transform or build upon it. For any reuse
or redistribution, researchers must make clear to others the licence terms of
this work.
Loizos Pelecanos
London, September 2013
iv
Abstract
Many earth dams around the world are located in zones characterised
by moderate to high seismicity. Their seismic stability can be particularly
critical for the safety of the areas in the downstream side and therefore an in
depth understanding of their response during earthquakes is required. This
thesis describes a numerical study related to both the seismic response and
analysis of earth dams using the finite element method.
In the first part of the thesis, the effect of the upstream reservoir hy-
drodynamic pressures on the elastic seismic response of dams is explored.
Firstly, a methodology is proposed in which the reservoir domain is modelled
with finite elements focusing in particular on the accurate prediction of the
hydrodynamic pressures on the upstream dam face. Secondly, a parametric
study of dam-reservoir interaction is carried out to examine the effect of the
reservoir on the seismic response of dams.
The second part of the thesis is concerned with the nonlinear seismic
behaviour of earth dams. Firstly, a well-documented case study, the La
Villita dam in Mexico, is analysed in order to validate the numerical model
and a good agreement is obtained between the recorded and predicted data.
Subsequently, using as a reference the calibrated model, parametric studies
are performed in order to obtain a better insight into the dynamic response
and analysis of earth dams. The latter studies provide a means to assess the
effect of different modelling considerations on the seismic analysis of dams.
v
Acknowledgements
Firstly and most importantly, I would like to thank God for everything.
Not all the people have the opportunities I had.
Secondly, I would like to express my sincere gratitude to my two research
supervisors and academic mentors, Dr Stavroula Kontoe and Professor Lidija
Zdravković. I am indebted to them, for inspiring me since my undergradu-
ate years, for providing me the opportunity to do research, for introducing
me to the scholar family of Geotechnics and for wholeheartedly welcoming
me to their research group. I need to express my appreciation to them for
everything they have taught me, both academic and non-academic matters;
for their constant availability, for the excellent computing facilities they pro-
vided, for teaching me to be professional, consistent and patient in my work
and for giving me the freedom and support to investigate my own ideas. I
am hugely grateful to them for being extremely patient with me and I need
to apologise for any inconvenience I may have caused to them.
This study was funded by a Departmental Training Grant from the En-
gineering and Physical Sciences Research Council, UK. Their generous fi-
nancial help is highly appreciated and acknowledged. This acknowledgement
is of course extended to my supervisors who made the relevant necessary
arrangements.
My PhD thesis was examined by Professors David M. Potts and Panos
Dakoulas, two eminent academics and experts in their research fields, nu-
merical analysis and dynamics of dams, respectively. I would like to express
my great appreciation to them for their keen interest in my work and all the
kind advice, feedback and directions they provided to me.
I regard it a great honour and privilege to have been part of the Geotech-
vi
nics Research Group at Imperial College, considering its long history and
academic reputation. The University has provided exceptional research fa-
cilities and kind access to an enormous amount of library material. I would
like to thank all the members of the staff for all their contribution and es-
pecially Professor David M. Potts for his valuable knowledge and advice on
numerical analysis and for providing access to ICFEP. I also need to thank
my MPhil/PhD transfer examiner, Dr Catherine O’Sullivan for her interest
in my work and her advice on this project and my future career. Of course,
I need to express my acknowledgements to all the previous and current re-
search students with whom we spent the last years together. Special thanks
are due to my officemate Vasilis Avgerinos for our day-to-day cooperation,
for sharing all sorts of problems and for our endless discussions. Likewise,
thanks are also due to my colleague Bo Han for sharing various thoughts
about dams and dynamic analysis. Moreover, I need to thank Miss Evan-
gelia Skiada for her assistance with a number of analyses of Chapter 5 during
her MSc studies the summer of 2012.
The last years have been very enjoyable, due to my close friends in Lon-
don. I need to express my gratefulness to my brother Fotis and my friends
Michalis Trypiniotis, Dr Yiannis Vyrides, Melinos Averkiou and Michael
Yap. Special thanks are due to my friend Mohammed Latheef, for devel-
oping tailor-made scripts for automated run of the analyses presented in
Chapter 5.
I take also the opportunity to thank all the people who offered their
assistance in any way to complete this work; assistance that I am or I am
not aware of.
Finally, I would like to express my deepest gratitude to my beloved family
and the people closest to me, from whom I had to travel away and sacrifice
the invaluable time that could have been spent with them, in order to be
educated: my wife Katerina, my respected parents Panayiotis and Irene and
my brothers Fotis and Angelos. I hope I make you happy. May God Bless
you all.
vii
Contents
Abstract v
Acknowledgements vi
Contents viii
List of figures xiv
List of tables xxxvi
Nomenclature xxxvii
1 INTRODUCTION 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Scope of research . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Layout of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 SEISMIC PERFORMANCE OF DAMS 6

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Design & construction of dams . . . . . . . . . . . . . . . . . . 7
2.2.1 Earth dams . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.2 Concrete dams . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Design considerations . . . . . . . . . . . . . . . . . . . 15
2.2.4 Construction processes . . . . . . . . . . . . . . . . . . 17
2.3 Seismic behaviour of soil deposits . . . . . . . . . . . . . . . . 18
2.3.1 Earthquake generation & wave propagation . . . . . . 18
viii
2.3.2 The scenario of reservoir-triggered seismicity . . . . . . 20

2.3.3 Visco-elastic ground response . . . . . . . . . . . . . . 20
2.3.4 Nonlinear dynamic soil behaviour . . . . . . . . . . . . 23
2.4 Dynamic response of dams . . . . . . . . . . . . . . . . . . . . 25
2.4.1 Transverse vibrations . . . . . . . . . . . . . . . . . . . 25
2.4.2 Three-dimensional canyon geometry . . . . . . . . . . . 30
2.4.3 Inhomogeneous dam materials . . . . . . . . . . . . . . 33
2.4.4 Flexible dam foundation and canyon effects . . . . . . 37
2.4.5 Dam-reservoir interaction . . . . . . . . . . . . . . . . 39
2.5 Reservoir hydrodynamic pressures . . . . . . . . . . . . . . . . 46
2.5.1 Vertical upstream face . . . . . . . . . . . . . . . . . . 47
2.5.2 Inclined upstream face . . . . . . . . . . . . . . . . . . 51
2.5.3 Inclined reservoir base . . . . . . . . . . . . . . . . . . 55
2.5.4 Compressibility of water . . . . . . . . . . . . . . . . . 58
2.5.5 Seismic excitation . . . . . . . . . . . . . . . . . . . . . 60
2.5.6 Flexible dam . . . . . . . . . . . . . . . . . . . . . . . 61
2.5.7 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 63
2.6 Effects of earthquakes on dams . . . . . . . . . . . . . . . . . 63
2.6.1 Excessive displacements . . . . . . . . . . . . . . . . . 64
2.6.2 Liquefaction . . . . . . . . . . . . . . . . . . . . . . . . 67
2.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3 SEISMIC ANALYSIS OF DAMS 71

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.2 Pseudo-static method . . . . . . . . . . . . . . . . . . . . . . . 72
3.2.1 Principle of the method . . . . . . . . . . . . . . . . . 72
3.2.2 Comments on pseudo-static analysis of earth dams . . 74
3.3 Sliding block method . . . . . . . . . . . . . . . . . . . . . . . 74
3.3.2 Sliding block analysis of earth dams . . . . . . . . . . . 76
3.4 Shear beam method . . . . . . . . . . . . . . . . . . . . . . . . 77
3.4.2 Numerical shear beam analysis of earth dams . . . . . 81
ix
3.5 Dynamic finite element method . . . . . . . . . . . . . . . . . 83

3.5.2 Dynamic finite element analysis of earth dams . . . . . 86
3.5.3 Coupled hydro-mechanical analysis . . . . . . . . . . . 90
3.5.4 Modelling of reservoir-dam interaction . . . . . . . . . 94
3.6 Comparison of methods of analysis . . . . . . . . . . . . . . . 96
3.7 Dynamic analysis considerations . . . . . . . . . . . . . . . . . 99
3.7.1 Boundary conditions . . . . . . . . . . . . . . . . . . . 100
3.7.2 Location of the boundaries . . . . . . . . . . . . . . . . 101
3.7.3 Element size . . . . . . . . . . . . . . . . . . . . . . . . 103
3.7.4 Time integration . . . . . . . . . . . . . . . . . . . . . 103
3.7.5 Material constitutive models . . . . . . . . . . . . . . . 104
3.7.5.1 Cyclic Nonlinear Model . . . . . . . . . . . . 104
3.7.5.2 Elasto-plastic model . . . . . . . . . . . . . . 107
3.7.5.3 Rayleigh Damping . . . . . . . . . . . . . . . 108
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4 NUMERICAL MODELLING OF THE RESERVOIR DO-

MAIN 111
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.3 Ramp acceleration . . . . . . . . . . . . . . . . . . . . . . . . 113
4.3.1 Upstream boundary condition in the reservoir . . . . . 114
4.3.2 Reservoir-dam & reservoir-base interface . . . . . . . . 122
4.4 Harmonic acceleration . . . . . . . . . . . . . . . . . . . . . . 129
4.4.1 Frequency response . . . . . . . . . . . . . . . . . . . . 129
4.4.2 Effect of water compressibility . . . . . . . . . . . . . . 132
4.4.3 Effect of element size . . . . . . . . . . . . . . . . . . . 133
4.5 Random acceleration . . . . . . . . . . . . . . . . . . . . . . . 136
4.5.1 Sloped dam . . . . . . . . . . . . . . . . . . . . . . . . 137
4.5.2 Shear stiffness of water . . . . . . . . . . . . . . . . . . 149
4.5.3 Damping of water . . . . . . . . . . . . . . . . . . . . . 152
4.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
x
5 DYNAMIC RESERVOIR-DAM INTERACTION 157

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
5.2 Statement of the problem . . . . . . . . . . . . . . . . . . . . 158
5.3 Cantilever dam . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.3.1 Dam with empty reservoir . . . . . . . . . . . . . . . . 159
5.3.2 Dam with full reservoir . . . . . . . . . . . . . . . . . . 161
5.3.3 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 168
5.4 Earth dam . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.4.1 Dam with empty reservoir . . . . . . . . . . . . . . . . 170
5.4.2 Dam with full reservoir . . . . . . . . . . . . . . . . . . 172
5.4.3 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 177
5.5 Application with seismic loading . . . . . . . . . . . . . . . . . 177
5.5.1 Cantilever dam . . . . . . . . . . . . . . . . . . . . . . 178
5.5.2 Earth dam . . . . . . . . . . . . . . . . . . . . . . . . . 187
5.5.3 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 193
5.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
6 NUMERICAL ANALYSIS OF LA VILLITA DAM 196

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
6.2 Description of the dam . . . . . . . . . . . . . . . . . . . . . . 197
6.2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . 197
6.2.2 Construction sequence & seismic events . . . . . . . . . 197
6.2.3 Dam geometry & soil properties . . . . . . . . . . . . . 199
6.2.4 Earthquake instrumentation and data . . . . . . . . . . 201
6.3 Previous work on the dam . . . . . . . . . . . . . . . . . . . . 207
6.3.1 Numerical and experimental analyses . . . . . . . . . . 207
6.3.2 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 217
6.4 Description of the current FE model . . . . . . . . . . . . . . 220
6.4.1 Stages of analysis . . . . . . . . . . . . . . . . . . . . . 220
6.4.2 Model geometry & boundary conditions . . . . . . . . 220
6.4.3 Calibration of constitutive models . . . . . . . . . . . . 223
6.5 Static analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
6.5.1 Construction . . . . . . . . . . . . . . . . . . . . . . . 232
xi
6.5.2 Reservoir impoundment . . . . . . . . . . . . . . . . . 233

6.5.3 Consolidation . . . . . . . . . . . . . . . . . . . . . . . 237
6.5.4 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 237
6.6 Dynamic analysis . . . . . . . . . . . . . . . . . . . . . . . . . 238
6.6.1 Effect of canyon geometry . . . . . . . . . . . . . . . . 238
6.6.2 Analysis with new stiffness profile of the dam . . . . . 253
6.6.2.1 Dynamic response of the dam . . . . . . . . . 253
6.6.2.2 Dynamic soil behaviour . . . . . . . . . . . . 264
6.6.3 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 272
6.7 Obstacles in the analysis of the dam . . . . . . . . . . . . . . 275
6.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
7 PARAMETRIC SEISMIC ANALYSIS OF LA VILLITA DAM278

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
7.2 Previous earthquake history . . . . . . . . . . . . . . . . . . . 279
7.2.1 Scenarios considered . . . . . . . . . . . . . . . . . . . 279
7.2.2 Dynamic response of the dam . . . . . . . . . . . . . . 279
7.2.3 Dynamic soil behaviour . . . . . . . . . . . . . . . . . . 283
7.2.4 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 287
7.3 Uncoupled analysis . . . . . . . . . . . . . . . . . . . . . . . . 290
7.3.1 Description of analysis . . . . . . . . . . . . . . . . . . 290
7.3.3 Dynamic soil behaviour . . . . . . . . . . . . . . . . . . 291
7.3.4 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 296
7.4 Foundation-dam interaction . . . . . . . . . . . . . . . . . . . 299
7.4.1 Analysis of a dam-foundation system . . . . . . . . . . 299
7.4.2 Analysis of a dam built on a rigid base . . . . . . . . . 304
7.4.3 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 309
7.5 Reservoir-dam interaction . . . . . . . . . . . . . . . . . . . . 310
7.5.1 Description of analyses . . . . . . . . . . . . . . . . . . 310
7.5.3 Hydrodynamic pressures . . . . . . . . . . . . . . . . . 316
7.5.4 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 320
xii
7.6 Stiffness inhomogeneity . . . . . . . . . . . . . . . . . . . . . . 321

7.6.1 Description of analysis . . . . . . . . . . . . . . . . . . 321
7.6.3 Comments . . . . . . . . . . . . . . . . . . . . . . . . . 322
7.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
8 CONCLUSION 330
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
8.2 Summary of findings . . . . . . . . . . . . . . . . . . . . . . . 331
8.2.1 Numerical modelling of the reservoir domain . . . . . . 331
8.2.2 Dynamic reservoir-dam interaction . . . . . . . . . . . 332
8.2.3 Numerical analysis of La Villita dam . . . . . . . . . . 333
8.2.4 Parametric seismic analysis of La Villita dam . . . . . 335
8.3 Recommendations for further research . . . . . . . . . . . . . 336
8.3.1 Numerical modelling of the reservoir domain . . . . . . 337
8.3.2 Dynamic reservoir-dam interaction . . . . . . . . . . . 337
8.3.3 Numerical analysis of La Villita dam . . . . . . . . . . 339
8.3.4 Parametric seismic analysis of La Villita dam . . . . . 340
8.4 Comments on the methods of analysis . . . . . . . . . . . . . . 341
8.4.1 Discussion on numerical and analytical methods . . . . 342
8.4.2 Recommendations for use of the methods . . . . . . . . 344
REFERENCES 369
xiii
List of Figures
2.1 Two-dimensional sketch of a uniform earth dam . . . . . . . . 8

2.2 Two-dimensional sketch of a zoned earth dam . . . . . . . . . 8
2.3 The Kourris dam is the largest zoned earthfill dam in Cyprus
(Cyprus Public Works Department, 2010). . . . . . . . . . . . 9
2.4 Two-dimensional sketch of a concrete faced rockfill dam . . . . 9
2.5 The Kannaviou dam, built in 2005, is the first and until today
(2012) only Concrete-Faced Rockfill Dam in Cyprus (Cyprus
Public Works Department, 2010) . . . . . . . . . . . . . . . . 10
2.6 Two-dimensional sketch of a concrete gravity dam . . . . . . . 11
2.7 The Lympia dam, built in the late 1970’s, is an example of a
small gravity dam in Cyprus (Cyprus Public Works Depart-
ment, 2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.8 The Stwlan dam in Wales is a buttress dam (British Dam
Society, 2010) . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.9 The Baitings dam in Yorkshire, England is an example of an
arch dam (British Dam Society, 2010) . . . . . . . . . . . . . 14
2.10 The Willow Creek Dam in Oregon, USA is the first Roller
Compacted Concrete dam in the world (US Army Corps of
Engineers, 2010) . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.11 Normal faulting: the hanging wall has moved downwards with
respect to the footwall block (US Geological Survey, 2010) . . 19
2.12 Seismic body waves (Goree, 2010) . . . . . . . . . . . . . . . . 19
2.13 Uniform visco-elastic soil layer . . . . . . . . . . . . . . . . . . 21
xiv
2.14 Amplification Factor, |F (ω)| for a visco-elastic soil layer on a

rigid bedrock . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.15 Shear stress-strain relationship - Backbone curve and hystere-
sis loop (Ishihara, 1995) . . . . . . . . . . . . . . . . . . . . . 24
2.16 Relation between shear stiffness, G, damping, ξ and cyclic
shear strain γ, arising from experiments on Toyoura sand (Kokusho,
1980) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.17 Geometry of the dam considered by Ambraseys (1960a). . . . 26
2.18 Natural modes of vibration of a dam in the upstream-downstream
direction: (a) Modes n, i.e. for deformation in the z-direction
with respect to the y-axis and (b) Modes r, i.e. for deforma-
tion in the z-direction with respect to the x-axis. . . . . . . . 27
2.19 First two natural modes of vibration of a dam in the upstream-
downstream direction for a complete wedge (k 0 = 0) (Am-
braseys, 1960a). . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.20 Amplification Factor of a homogeneous earth dam. After Dak-
oulas and Gazetas (1985). . . . . . . . . . . . . . . . . . . . . 29
2.21 Comparison between natural frequencies from 2D and 3D anal-
yses of dams for different length to height, L/H ratios (Mejia
and Seed, 1983). . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.22 Effect of canyon geometry on the fundamental natural period,
T1 of a dam (Dakoulas and Gazetas, 1987). . . . . . . . . . . . 31
2.23 Midcrest amplification for dams under plane-strain conditions
and in a semi-cylindrical canyon (Dakoulas and Gazetas, 1986). 32
2.24 The inhomogeneous dam system studied by Dakoulas and
Gazetas (1985). . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.25 Effect of inhomogeneity on the displacement modal shapes of
dams (Dakoulas and Gazetas, 1985). . . . . . . . . . . . . . . 35
2.26 Effect of inhomogeneity on the natural periods of dams (Dak-
oulas and Gazetas, 1985). . . . . . . . . . . . . . . . . . . . . 35
2.27 Effect of inhomogeneity on the vertical variation of accelera-
tions in earth dams (Gazetas, 1987). . . . . . . . . . . . . . . 37
xv
2.28 Effect of the geometric characteristics of a homogeneous dam-

foundation system on the system’s fundamental period (Dak-
oulas and Gazetas, 1987). . . . . . . . . . . . . . . . . . . . . 38
2.29 The dam-foundation system considered by Dakoulas (1990) . . 39
2.30 Effect of geometric and material characteristics on the funda-
mental period of an inhomogeneous dam-foundation system
(Dakoulas, 1990). Te11 refers to the fundamental period of a
dam-foundation system, T11 refers to the fundamental period
of a dam founded on rigid rock, and m is the exponent of
inhomogeneity (Equation 2.16). . . . . . . . . . . . . . . . . . 40
2.31 Accelerations predicted at the crest of La Villita dam for dif-
ferent shear wave velocities, Vr of the rock canyon for (a) a
linear elastic (Papalou and Bielak, 2001) and (b) a nonlinear
system (Papalou and Bielak, 2004). . . . . . . . . . . . . . . . 41
2.32 Frequency response of the accelerations at the dam crest for
different load, dam and reservoir circular frequency ratios as
presented by Chopra (1968): (a) for 0 ≤ ωr /ωd ≤ 1.8 and (b)
for 2 ≤ ωr /ωd ≤ 5. . . . . . . . . . . . . . . . . . . . . . . . . 42
2.33 Geometry of the dam-reservoir systems considered by (a) Hall
and Chopra (1982b) and (b) Hall and Chopra (1982a) . . . . . 43
2.34 Frequency spectrum of the accelerations at the dam crest for
(a) a concrete gravity dam (Hall and Chopra, 1982b) and (b)
an earth dam (Hall and Chopra, 1982a) . . . . . . . . . . . . . 44
2.35 The dam-reservoir system considered by Westergaard (1933) . 48
2.36 Distribution of the width of water mass, b and concrete, b0
considered to move with water (Westergaard, 1933). . . . . . . 50
2.37 Hydrodynamic pressures on a dam face - values of the pressure
coefficient C. The solid line represents experimental results,
whereas the dashed line represents the empirical equation fit-
ted by Zangar (1952) . . . . . . . . . . . . . . . . . . . . . . . 53
2.38 Maximum and base coefficients of hydrodynamic pressures on
an inclined dam face C & Cm (Zangar, 1952) . . . . . . . . . . 53
xvi
2.39 The reservoir-dam system considered by Chwang and Housner

(1978) and Chwang (1978). . . . . . . . . . . . . . . . . . . . . 54
2.40 Values of the pressure coefficient C(y) for a dam with an in-
clined upstream face. The dashed line presents the results
from the work of Chwang and Housner (1978) whereas the
solid line corresponds to the work of Chwang (1978). . . . . . 54
2.41 The dam-reservoir system considered by Liu (1986) . . . . . . 56
2.42 Values of the pressure coefficient Cp (=P/ρao h) for reservoir
bottom angles (a) 0o and (b) 15o (Liu, 1986) (solid line). The
dashed line in (a) corresponds to the work of Zangar (1952)
and it is provided for comparison. . . . . . . . . . . . . . . . . 57
2.43 Values of the pressure coefficient Cp (=P/ρao h) for reservoir
bottom angles (a) 30o and (b) 45o (Liu, 1986). . . . . . . . . . 57
2.44 Frequency response of the total hydrodynamic force induced
by a compressible reservoir due to horizontal harmonic load-
ing. After Chopra (1967a) . . . . . . . . . . . . . . . . . . . . 59
2.45 The fluid-structure system considered by Lee and Tsai (1991) 61
2.46 Failure of the upstream slope of the Fategadh dam in India
during the 2001 Bhuj earthquake (Madabhushi and Haigh, 2001). 64
2.47 Major deformation patterns in an earth dam after succesive
earthquake cycles (Ambraseys, 1959b; Newmark, 1965). . . . . 65
2.48 Cracks on Fategadh earth dam in India during the 2001 Bhuj
earthquake (Madabhushi and Haigh, 2001). . . . . . . . . . . . 66
2.49 Failure of the Lower San Fernando dam due to liquefaction
in the upstream part of the rockfill (Cambridge University
Engineering Department, 2012). . . . . . . . . . . . . . . . . . 67
2.50 Failure and reconstructed cross section of the Lower San Fer-
nando dam during the 1971 San Fernando earthquake as pre-
sented by Seed et al. (1975). . . . . . . . . . . . . . . . . . . . 68
3.1 Illustration of the pseudo-static method of analysis in which

additional inertial forces kH W and kV W are included to ac-
count for seismic loading. . . . . . . . . . . . . . . . . . . . . . 73
xvii
3.2 Newmark’s sliding block method (Wilson and Keefer, 1985) . . 75

3.3 Shear beam approach for two-dimensional seismic analysis of
dams. After Gazetas (1987). . . . . . . . . . . . . . . . . . . . 79
3.4 Geometry of the system considered by Elgamal (1992): (a)
physical problem, (b) UD cross-section, (c) shear stresses act-
ing on an soil element and (d) geometric configuration of com-
putational model. . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.5 FE mesh used for the 2D nonlinear static and dynamic analysis
of Marana Capacciotti dam by Elia et al. (2010). . . . . . . . 86
3.6 The finite element mesh used by Clough and Chopra (1966)
(B = 450 ft, H = 300 ft). . . . . . . . . . . . . . . . . . . . . . 87
3.7 Shape of the first natural mode of vibration of a uniform,
homogeneous dam as obtained by SB (a) and FE (b-e) analyses
(Chopra, 1967a). . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.8 FE mesh used by Prevost et al. (1985) in the dynamic analysis
of Santa Felicia dam. . . . . . . . . . . . . . . . . . . . . . . . 89
3.9 Hysteretic soil behaviour was modelled by Prevost et al. (1985)
using multi-surface plasticity. . . . . . . . . . . . . . . . . . . 89
3.10 The FE mesh used by Elia et al. (2010): (a) the entire FE
mesh describing the problem and (b) detail of the FE mesh
close to the embankment. . . . . . . . . . . . . . . . . . . . . 93
3.11 Distribution of peak acceleration response with depth in an
earth dam (Gazetas, 1987) from FE (dashed line) and SB
(solid line) analyses. . . . . . . . . . . . . . . . . . . . . . . . 98
3.12 In numerical analysis of seismic soil-structure interaction prob-
lems, the response close to the lateral (artificial) boundaries
should be similar to the free-field response, so that the bound-
ary conditions do not interact with the response of the struc-
ture (Committe on Nuclear Structures and Materials, 1979). . 102
3.13 Backbone curve of the Logarithmic CNL model (After Taborda
(2011)). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
xviii
3.14 Illustration of the influence of parameters c and Gmin on the

damping introduced by the Logarithmic CNL model (After
Taborda (2011)). . . . . . . . . . . . . . . . . . . . . . . . . . 106
3.15 Yield surface of the Mohr-Coulomb model . . . . . . . . . . . 107
3.16 Illustration of Rayleigh damping and its mass and stiffness
components . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.1 Geometry of the dam-reservoir-foundation system: (1) dam,

(2) reservoir, (3) foundation soil, US dam face A-B, US reser-
voir boundary C-D and foundation boundary E-F-G-C. . . . . 113
4.2 Geometry of the dam-reservoir system used by Küçükarslan
et al. (2005). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.3 The ramp acceleration load applied to the dam-reservoir sys-
tem of Küçükarslan et al. (2005) . . . . . . . . . . . . . . . . . 114
4.4 The analytical solution for the hydrodynamic pressures at the
base of the dam due to the ramp acceleration using the solution
of Chopra (1967a). . . . . . . . . . . . . . . . . . . . . . . . . 115
4.5 The analytical solution for the distribution of the maximum
hydrodynamic pressure over the entire upstream face of the
dam due to the ramp acceleration using the solution of Chopra
(1967a). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.6 The hydrodynamic pressure time-history at the base of the
dam as obtained by Küçükarslan et al. (2005). . . . . . . . . . 116
4.7 The distribution of the peak hydrodynamic pressure on the
upstream face of the dam as obtained by Küçükarslan et al.
(2005). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
4.8 Geometry of the dam-reservoir system used for the ramp ac-
celeration loading. . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.9 Representation of the Sommerfeld radiation condition as sug-
gested by Parrinello and Borino (2007) . . . . . . . . . . . . . 118
4.10 Hydrodynamic pressure at the base of the dam by using the
“Free” BC for different lengths of the reservoir . . . . . . . . . 120
xix

“Viscous” BC for different lengths of the reservoir . . . . . . . 120
“Cone” BC for different lengths of the reservoir . . . . . . . . 121
“Viscous + hydrostatic stress” BC for different lengths of the
reservoir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
“Free” BC with interface elements for different lengths of the
reservoir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
“Viscous” BC with interface elements for different lengths of
the reservoir . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
“Cone” BC with interface elements for different lengths of the
reservoir . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
“Viscous + hydrostatic stress” BC with interface elements for
different lengths of the reservoir . . . . . . . . . . . . . . . . . 125
“Viscous” BC with interface elements for L/H = 3 . . . . . . 126
“Viscous” BC with interface elements for L/H = 5.33 . . . . . 126
“Viscous” BC with interface elements for L/H = 13.33 . . . . 128
4.23 Distribution of the peak hydrodynamic pressure on the up-
stream face of the dam by using the “Viscous” BC. Compari-
son between numerical and analytical (Chopra, 1967a) results. 128
xx
4.24 Frequency response of the hydrodynamic force induced by a

compressible reservoir due to horizontal harmonic loading, as
suggested by Chopra (1967a) . . . . . . . . . . . . . . . . . . . 130
ing. Comparison between numerical (Viscous BC) and ana-
lytical (Chopra, 1967a) results. . . . . . . . . . . . . . . . . . 131
ing. Comparison between numerical (Cone BC) and analytical
(Chopra, 1967a) results. . . . . . . . . . . . . . . . . . . . . . 131
by an incompressible reservoir due to horizontal harmonic
loading. Comparison between numerical and analytical (Chopra,
1967a; Westergaard, 1933) results. . . . . . . . . . . . . . . . . 133
4.28 In wave propagation problems a sufficient number of nodes
should be provided to accurately model the full wavelength. . 134
4.29 Eight-noded isoparametric quadrilateral elements have been
used in modelling the reservoir domain. . . . . . . . . . . . . . 134
4.30 Distribution of the peak hydrodynamic pressure on the dam
face for different element sizes . . . . . . . . . . . . . . . . . . 136
4.31 Accelerations recorded at the rock canyon of La Villita dam
in Mexico during the 15/11/1975 Mexico earthquake . . . . . 137
4.32 Geometry of the dam-reservoir system considered for the in-
vestigation of the effect of the slope . . . . . . . . . . . . . . . 138
4.33 Distribution of the peak hydrodynamic pressure on the face
of a stiff dam with B/H = 0 (i.e. vertical face) for various
lengths of the reservoir . . . . . . . . . . . . . . . . . . . . . . 140
of a stiff dam with B/H = 1/3 (i.e. sloped face) for various
xxi

of a stiff dam with B/H = 1/2 (i.e. sloped face) for various
4.36 Distribution of the peak hydrodynamic pressure on the face of
a stiff dam with B/H = 1 (i.e. sloped face) for various lengths
of the reservoir . . . . . . . . . . . . . . . . . . . . . . . . . . 141
of the reservoir . . . . . . . . . . . . . . . . . . . . . . . . . . 142
of the reservoir . . . . . . . . . . . . . . . . . . . . . . . . . . 142
of a flexible dam with B/H = 0 (i.e. vertical face) for various
a flexible dam with B/H = 1/3 (i.e. sloped face) for various
a flexible dam with B/H = 1/2 (i.e. sloped face) for various
of a flexible dam with B/H = 1 (i.e. sloped face) for various
a stiff dam for L/H = 15 for various values of the slope angle
(B/H) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
xxii

of a flexible dam for L/H = 15 for various values of the slope
angle (B/H) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
stiff and flexible vertical dams (B/H = 0) for L/H = 15 . . . 148
stiff and flexible sloped dams (B/H = 3) for L/H = 15 . . . . 148
4.49 Distribution of the peak hydrodynamic pressure on the vertical
face of a stiff dam for various values of the shear modulus of
the water, Gw . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
face of a stiff dam for various values of the shear modulus of
the water, Gw - detail . . . . . . . . . . . . . . . . . . . . . . . 151
4.51 Fourier Amplitude Spectrum of the acceleration record . . . . 153
face of a stiff dam for various values of the damping of the
water, ξw . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
face of a stiff dam for various values of the damping of the
water, ξw - detail . . . . . . . . . . . . . . . . . . . . . . . . . 154
5.1 Geometry of the dam-reservoir-foundation system: (1) Dam,

(2) Reservoir, (3) Foundation. . . . . . . . . . . . . . . . . . . 158
5.2 Geometry of the cantilever dam considered. . . . . . . . . . . 159
5.3 Amplification spectrum of the cantilever dam with an empty
reservoir. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
5.4 Amplification of acceleration, |F | at the crest of the cantilever
dam for various values of the frequency ratios ω/ωd and ωr /ωd . 163
5.5 Amplification spectrum of the cantilever dam with a full (ωr /ωd =
1) and an empty reservoir. . . . . . . . . . . . . . . . . . . . . 164
5.6 Amplification peaks for the cantilever dam with a full and an
empty reservoir. . . . . . . . . . . . . . . . . . . . . . . . . . . 164
xxiii
5.7 Amplification of acceleration, |F | for the cantilever dam with

respect to the ratio of the fundamental circular frequency of
the reservoir over that of the dam for the first mode of vibra-
tion, ωr /ωd . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.8 Amplification of acceleration, |F | for the cantilever dam with
respect to the ratio of the fundamental circular frequency of
the reservoir over that of the dam for the second mode of
vibration, ωr /ωd . . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.9 Geometry of the triangular earth dam considered. . . . . . . . 170
5.10 Amplification spectrum of the earth dam with an empty reser-
voir. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
5.11 Amplification of acceleration, |F | at the crest of the earth dam
for various values of the frequency ratios ω/ωd and ωr /ωd . . . 173
5.12 Amplification spectrum of the earth dam with a full (ωr /ωd =
1) and an empty reservoir. . . . . . . . . . . . . . . . . . . . . 174
5.13 Amplification peaks for the earth dam with a full and an empty
reservoir. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.14 Amplification of acceleration, |F | for the earth dam with re-
spect to the ratio of the fundamental circular frequency of the
reservoir over that of the dam, ωr /ωd , for the first mode of
vibration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
reservoir over that of the dam, ωr /ωd , for the second mode of
vibration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
reservoir over that of the dam, ωr /ωd , for the second mode of
vibration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
5.17 Acceleration time-history of the El Centro (1940) earthquake
record. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
5.18 Response spectrum of the El Centro (1940) earthquake record
for damping, ξ = 5%. . . . . . . . . . . . . . . . . . . . . . . . 179
xxiv
5.19 Cases of combinations of ω/ωd and ωr /ωd frequency ratios ex-

amined under seismic loading. . . . . . . . . . . . . . . . . . . 180
5.20 Acceleration time-history at the crest of the cantilever dam -
Case A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Case B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
Case C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Case D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Case E. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
5.25 Response spectrum (for damping ξ = 5%) of the accelerations
at the crest of the cantilever dam - Case A. . . . . . . . . . . . 183
at the crest of the cantilever dam - Case B. . . . . . . . . . . . 184
at the crest of the cantilever dam - Case C. . . . . . . . . . . . 184
at the crest of the cantilever dam - Case D. . . . . . . . . . . . 185
at the crest of the cantilever dam - Case E. . . . . . . . . . . . 185
5.30 Acceleration time-history at the crest of the earth dam - Case
A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
C. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
E. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
xxv

at the crest of the earth dam - Case A. . . . . . . . . . . . . . 190
at the crest of the earth dam - Case B. . . . . . . . . . . . . . 191
at the crest of the earth dam - Case C. . . . . . . . . . . . . . 191
at the crest of the earth dam - Case D. . . . . . . . . . . . . . 192
at the crest of the earth dam - Case E. . . . . . . . . . . . . . 192
6.1 Aerial photo of La Villita dam (Google Earth, 2010) with ap-
purtenant facilities: (1) Reservoir, (2) Embankment, (3) Hy-
droelectric power plant, (4) Spillway, (5) Downstream Balsas
river. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
6.2 Recorded displacements on some points on the dam body (El-
gamal, 1992). . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
6.3 Upstream-downstream cross-section of La Villita dam with
zones of different materials (After Elgamal (1992)). . . . . . . 200
6.4 Longitudinal cross-section of La Villita dam showing the geo-
logical profile of the site and the shape of the canyon (Elgamal,
1992). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
6.5 Plan view of La Villita dam and locations of accelerometers
(Elgamal et al., 1990). . . . . . . . . . . . . . . . . . . . . . . 202
6.6 Available bedrock records in the UD direction for EQ1-EQ3,
EQ5, EQ6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
6.7 Available crest records in the UD direction for EQ1-EQ6 . . . 204
6.8 Response spectra (damping, ξ = 5%) of the available acceler-
ation records in the UD direction for EQ1-EQ3 . . . . . . . . 205
6.9 Response spectra (damping, ξ = 5%) of the available acceler-
ation records in the UD direction for EQ4-EQ6 . . . . . . . . 206
6.10 The recorded UD motion on the crest of La Villita dam, during
EQ5 that shows an asymmetry in the peak values. . . . . . . . 207
xxvi
6.11 The experimental setup used by Elgamal et al. (1990) in the

prediction of permanent displacements of La Villita dam: Slid-
ing block mounted on inclined dynamic shaker. . . . . . . . . 208
6.12 The computed displacement of La Villita dam during EQ5 as
obtained by sliding block analysis of Elgamal et al. (1990). . . 209
6.13 The computed acceleration response of La Villita dam during
EQ5 as obtained by sliding block analysis of Elgamal et al.
(1990). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
6.14 5% damped response spectra of computed and recorded crest
accelerations during EQ5, obtained by Elgamal (1992). . . . . 211
6.15 The computed crest displacements as obtained from sliding-
block analyses from Elgamal (1992) . . . . . . . . . . . . . . . 211
6.16 The finite element mesh used by Gazetas and Uddin (1994) in
the analysis of La Villita dam with a pre-defined slip surface. . 214
6.17 Computed accelerations of two points on the crest of the dam:
(a) outside and (b) inside the failing mass, by Gazetas and
Uddin (1994). . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
6.18 Comparison of calculated (thin line) and final recorded (thick
line) permanent crest displacements of La Villita dam by Gaze-
tas and Uddin (1994). . . . . . . . . . . . . . . . . . . . . . . 216
6.19 The system analysed by Papalou and Bielak (2001, 2004): (a)
3D dam-canyon model, (b) cross-section of dam in the UD
direction, (c) cross-section of dam and canyon in the longitu-
dinal direction and (d) free body diagram of a shear wedge
element. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
6.20 The model used by Papalou and Bielak (2001, 2004) in the
analysis of La Villita dam. The different zones correspond to
different material properties (Gmax ). . . . . . . . . . . . . . . . 218
6.21 Calculated and recorded response of the midcrest of La Villita
dam, as obtained by linear and nonlinear analyses with vary-
ing Vsr of canyon rock by Papalou and Bielak (2001, 2004)
respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
6.22 2D Finite Element mesh used for the analysis of La Villita dam221
xxvii
6.23 Spatial variation (in the vertical direction) of shear stiffness,

Gmax . The grey solid line corresponds to the values used by
Elgamal (1992) (see also Table 6.3) and the black solid line
corresponds to the values used in this work. . . . . . . . . . . 222
6.24 Fourier Amplitude Spectra of EQ2 & EQ5 . . . . . . . . . . . 223
6.25 Calibration of the Logarithmic CNL model against the em-
pirical curves of Vucetic and Dobry (1991) for the clay core -
Stiffness degradation (G/Gmax − γ) . . . . . . . . . . . . . . . 227
6.26 Calibration of the Logarithmic CNL model against the em-
pirical curves of Vucetic and Dobry (1991) for the clay core -
Damping (ξ − γ) . . . . . . . . . . . . . . . . . . . . . . . . . 227
6.27 Calibration of the Logarithmic CNL model against the empir-
ical curves of Seed et al. (1986) for the sand filters - Stiffness
degradation (G/Gmax − γ) . . . . . . . . . . . . . . . . . . . . 228
6.28 Calibration of the Logarithmic CNL model against the empir-
ical curves of Seed et al. (1986) for the sand filters - Damping
(ξ − γ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
6.29 Calibration of the Logarithmic CNL model against the empiri-
cal curves of Rollins et al. (1998) for the rockfill and foundation
alluvium - Stiffness degradation (G/Gmax − γ) . . . . . . . . . 229
6.30 Calibration of the Logarithmic CNL model against the empiri-
cal curves of Rollins et al. (1998) for the rockfill and foundation
alluvium - Damping (ξ − γ) . . . . . . . . . . . . . . . . . . . 229
6.31 Comparison of the stiffness degradation curves (G/Gmax − γ)
predicted by the Logarithmic CNL model for the various ma-
terials (and the corresponding empirical curves used) with the
stiffness degradation used by Elgamal (1992). . . . . . . . . . 230
6.32 Stiffness degradation (G − γ) predicted by the Logarithmic
CNL model for the clay core . . . . . . . . . . . . . . . . . . . 230
CNL model for the sand filters . . . . . . . . . . . . . . . . . . 231
CNL model for the rockfill and foundation alluvium . . . . . . 231
xxviii
6.35 Contours of pore water pressure in the dam after the end of the
reservoir impoundment (positive values correspond to com-
pressive stresses). . . . . . . . . . . . . . . . . . . . . . . . . . 235
6.36 Flow net (contours of stream and potential functions) in the
clay core after the end of the reservoir impoundment. . . . . . 236
6.37 Calculated and measured crest settlement history of the dam
crest prior to the EQ events. . . . . . . . . . . . . . . . . . . . 237
6.38 End of static analysis: (a) Deformed mesh and (b) Vectors of
accumulated displacement of La Villita dam. . . . . . . . . . . 239
6.39 Location of the acceleration monitoring points of the upstream
corner of the crest (U), mid-crest (C) and downstream berm
(B) and location of application of the seismic excitation at the
rock (R) underneath the dam. . . . . . . . . . . . . . . . . . . 240
6.40 Acceleration time history at the crest of La Villita dam during
EQ2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
6.41 Response spectra (damping, ξ = 5%) at the crest of La Villita
dam during EQ2 . . . . . . . . . . . . . . . . . . . . . . . . . 241
EQ5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
dam during EQ5 . . . . . . . . . . . . . . . . . . . . . . . . . 242
6.44 New calibration of the Logarithmic CNL model against the
empirical curves of Vucetic and Dobry (1991) for the clay core
- Stiffness degradation (G/G∗max − γ) . . . . . . . . . . . . . . 245
empirical curves of Vucetic and Dobry (1991) for the clay core
- Damping (ξ − γ) . . . . . . . . . . . . . . . . . . . . . . . . 245
empirical curves of Seed et al. (1986) for the sand filters -
Stiffness degradation (G/G∗max − γ) . . . . . . . . . . . . . . . 246
empirical curves of Seed et al. (1986) for the sand filters -
Damping (ξ − γ) . . . . . . . . . . . . . . . . . . . . . . . . . 246
xxix

empirical curves of Rollins et al. (1998) for the rockfill and
foundation alluvium - Stiffness degradation (G/G∗max − γ) . . 247
empirical curves of Rollins et al. (1998) for the rockfill and
foundation alluvium - Damping (ξ − γ) . . . . . . . . . . . . . 247
6.50 Comparison of the stiffness degradation curves (G/G∗max − γ)
predicted by the Logarithmic CNL model (new calibration) for
the various materials (and the corresponding empirical curves
used) with the stiffness degradation used by Elgamal (1992). . 248
6.51 Stiffness degradation (G∗ −γ) predicted by the new calibration
of the Logarithmic CNL model for the clay core . . . . . . . . 248
of the Logarithmic CNL model for the sand filters . . . . . . . 249
of the Logarithmic CNL model for the rockfill and foundation
alluvium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
EQ2 for the new updated value of G∗max (z). . . . . . . . . . . 250
dam during EQ2 for the new updated value of G∗max (z). . . . . 250
EQ5 for the new updated value of G∗max (z). . . . . . . . . . . 251
dam during EQ5 for the new updated value of G∗max (z). . . . . 251
6.58 Acceleration time history at the downstream berm of La Vil-
lita dam during EQ2 for the new updated value of G∗max (z). . 254
6.59 Response spectra (damping, ξ = 5%) at the downstream berm
of La Villita dam during EQ2 for the new updated value of
G∗max (z). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
dam during EQ5. Comparison between the (a) recorded, (b)
calculated in this study and (c) calculated by Elgamal (1992). 255
xxx

EQ5. Comparison between filtered recorded and calculated
response. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
dam during EQ5. Comparison between filtered recorded and
calculated response. . . . . . . . . . . . . . . . . . . . . . . . . 257
6.63 End of the seismic analysis (after EQ5): (a) Accumulated (for
the whole static and dynamic analysis) deformed mesh and
(b) Vectors of accumulated displacement of La Villita dam. . . 259
6.64 End of the seismic analysis (after EQ5): (a) Sub-accumulated
(i.e. only for EQ5) deformed mesh and (b) Vectors of sub-
accumulated dispacement of La Villita dam. . . . . . . . . . . 260
6.65 End of the seismic analysis (after EQ5): Stress level, S, in La
Villita dam. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
6.66 Vertical displacement time history at the crest (Point C in
Figure 6.39) of La Villita dam during EQ2. . . . . . . . . . . . 262
6.67 Vertical displacement history at the crest (Point C in Figure
6.39) of La Villita dam during EQ5. . . . . . . . . . . . . . . . 262
6.68 Vertical displacement time history at the upstream corner of
the crest (Point U in Figure 6.39) of La Villita dam during EQ2.263
6.69 Vertical displacement history at the upstream corner of the
crest (Point U in Figure 6.39) of La Villita dam during EQ5. . 263
6.70 End of the seismic analysis (after EQ5): Accumulated total
deviatoric strain, Ed in La Villita dam. . . . . . . . . . . . . . 265
6.71 Sketch of the dam showing the locations of the monitoring
elements: upstream rockfill (UR), clay core (CC), downstream
rockfill (DR) and downstream alluvium (DA). . . . . . . . . . 266
6.72 Shear stress-strain response at the downstream alluvium (DA)
of La Villita dam during EQ5 . . . . . . . . . . . . . . . . . . 266
6.73 Shear stress-strain response in the clay core (CC) of La Villita
dam during EQ5 . . . . . . . . . . . . . . . . . . . . . . . . . 267
6.74 Shear stress-strain response in the upstream rockfill (UR) of
La Villita dam during EQ5 . . . . . . . . . . . . . . . . . . . . 267
xxxi
6.75 Shear stress-strain response in the downstream rockfill (DR)

6.76 Shear stress-strain response at the downstream alluvium (DA)
6.77 Shear strain time-history in the upstream rockfill (UR) of La
Villita dam during EQ5 . . . . . . . . . . . . . . . . . . . . . 269
6.78 Shear strain time-history in the downstream rockfill (DR) of
6.79 Stress paths in the upstream rockfill (UR) of La Villita dam . 270
6.80 Stress paths in the downstream rockfill (DR) of La Villita dam 270
7.1 Acceleration response at the crest of La Villita dam during

EQ5 - Comparison between Cases A and B . . . . . . . . . . . 280
7.2 Acceleration response at the crest of La Villita dam during
EQ5 - Comparison between Cases B and C . . . . . . . . . . . 280
dam during EQ5 - Comparison between Cases A and B . . . . 281
dam during EQ5 - Comparison between Cases B and C . . . . 281
7.5 Vertical displacements at the crest of La Villita dam during
EQ5 - Comparison between Cases A and B . . . . . . . . . . . 282
7.6 Vertical displacements at the crest of La Villita dam during
EQ5 - Comparison between Cases B and C . . . . . . . . . . . 282
7.7 Shear strain time-history in the downstream alluvium (DA) of
7.8 Shear strain time-history in the downstream alluvium (DA) of
7.9 Pore water pressure time-history in the downstream alluvium
(DA) of La Villita dam during EQ5 . . . . . . . . . . . . . . . 285
7.10 Pore water pressure time-history in the downstream alluvium
(DA) of La Villita dam during EQ5 . . . . . . . . . . . . . . . 285
7.11 Stress paths in the downstream alluvium (DA) of La Villita
dam during EQ5 - Scenario B . . . . . . . . . . . . . . . . . . 286
xxxii

dam during EQ5 - Scenario A . . . . . . . . . . . . . . . . . . 286
dam during EQ5 - Scenario C . . . . . . . . . . . . . . . . . . 287
7.14 Contours of stress level, S for (a) Case A - EQ5 without any
previous earthquake, and (b) Case C - EQ5 after a large earth-
quake, which is EQ5. . . . . . . . . . . . . . . . . . . . . . . . 288
7.15 Acceleration time-history at the crest of La Villita dam during
EQ2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
7.16 Response spectra (damping, ξ = 5%) the crest of La Villita
dam during EQ2. . . . . . . . . . . . . . . . . . . . . . . . . . 292
EQ5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
dam during EQ5. . . . . . . . . . . . . . . . . . . . . . . . . . 293
7.19 Vertical profile of the maximum values of horizontal displace-
ment in the core of La Villita dam during EQ2. . . . . . . . . 294
7.21 Vertical profile of the maximum values of horizontal accelera-
tion in the core of La Villita dam during EQ2. . . . . . . . . . 295
7.23 Shear stress-strain response in the clay core of La Villita dam
during EQ5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
7.24 Shear stress-strain response in the clay core of La Villita dam
during EQ5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
7.25 Shear strain time-history in the clay core of La Villita dam
during EQ5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
7.26 Pore water pressure time-history in the clay core of La Villita
dam during EQ5 . . . . . . . . . . . . . . . . . . . . . . . . . 298
xxxiii
7.27 Mesh of the dam showing the locations of the monitoring

points: Dam base (DB), Downstream dam toe (DDT) and
Free-field (FF). . . . . . . . . . . . . . . . . . . . . . . . . . . 300
7.28 Acceleration time-history response at various locations for a
dam-foundation system - EQ2 . . . . . . . . . . . . . . . . . . 301
7.29 Acceleration time-history response at various locations for a
dam-foundation system - EQ5 . . . . . . . . . . . . . . . . . . 302
7.30 Response spectra (damping, ξ = 5%) at various locations for
a dam-foundation system - EQ2 . . . . . . . . . . . . . . . . . 303
7.31 Response spectra (damping, ξ = 5%) at various locations for
a dam-foundation system - EQ5 . . . . . . . . . . . . . . . . . 303
7.32 Acceleration time-history response under various input mo-
tions for a dam on a rigid foundation - EQ2 . . . . . . . . . . 306
7.33 Acceleration time-history response under various input mo-
7.34 Response spectra (damping, ξ = 5%) under various input mo-
7.35 Response spectra (damping, ξ = 5%) under various input mo-
7.36 2D Finite Element mesh used for the analysis of La Villita dam311
EQ2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
dam during EQ2. . . . . . . . . . . . . . . . . . . . . . . . . . 312
EQ5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
dam during EQ5. . . . . . . . . . . . . . . . . . . . . . . . . . 313
xxxiv

7.45 Time-history of the total hydrodynamic force on the upstream
face of the dam during EQ2. . . . . . . . . . . . . . . . . . . . 317
stream face of the dam during EQ2. . . . . . . . . . . . . . . . 317
7.47 Time-history of the total hydrodynamic force on the upstream
face of the dam during EQ5. . . . . . . . . . . . . . . . . . . . 318
stream face of the dam during EQ5. . . . . . . . . . . . . . . . 318
7.49 Spatial variation (in the vertical direction) of maximum shear
stiffness, Gmax . The grey solid and black dashed lines corre-
spond to a linear and constant variation of Gmax in the dam
core respectively. . . . . . . . . . . . . . . . . . . . . . . . . . 321
EQ2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
dam during EQ2. . . . . . . . . . . . . . . . . . . . . . . . . . 323
EQ5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
dam during EQ5. . . . . . . . . . . . . . . . . . . . . . . . . . 324
xxxv
List of Tables
4.1 Geometry of the FE meshes considered for the ramp acceler-

ation loading . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.2 Size of the finite elements considered for the harmonic accel-
eration loading . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.3 Values of the angle of the dam slope considered. . . . . . . . . 138
4.4 Values of the length of the mesh considered. . . . . . . . . . . 139
4.5 Values of the Shear Modulus, Gw assigned to the reservoir
water elements . . . . . . . . . . . . . . . . . . . . . . . . . . 150
4.6 Values of the damping, ξw assigned to the reservoir water ele-
ments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
5.1 Cases of combinations of ω/ωd and ωr /ωd frequency ratios ex-

amined under seismic loading. . . . . . . . . . . . . . . . . . . 180
5.2 Material and geometric properties of the cantilever dam cases
considered under seismic loading. . . . . . . . . . . . . . . . . 186
5.3 Material and geometric properties of the earth dam cases con-
sidered under seismic loading. . . . . . . . . . . . . . . . . . . 187
6.1 Summary of earthquake events between 1975-1985 and accel-

erations in the upstream-downstream direction. . . . . . . . . 198
6.2 Summary of known material properties of La Villita dam . . . 201
6.3 Spatial variation of the maximum shear modulus, Gmax of the
dam and foundation material, as used by Elgamal (1992). . . . 201
6.4 Stages of FE analysis of La Villita dam . . . . . . . . . . . . . 221
6.5 Mohr-Coulomb model parameters for the soil materials. . . . . 224
xxxvi
6.6 Calibration parameters for the Logarithmic CNL model. . . . 225

6.7 New calibration parameters for the Logarithmic CNL model,
used for the dynamic analysis. . . . . . . . . . . . . . . . . . . 243
7.1 Values of the Rayleigh damping coefficients A & B. . . . . . . 311

7.2 Maximum hydrodynamic force on La Villita dam. . . . . . . . 319
xxxvii
Nomenclature
a acceleration
ao amplitude of harmonic acceleration load
am maximum value of random acceleration load
a(t) acceleration load time history
B breadth of dam slope face
c parameter of the cyclic nonlinear model
c0 cohesion
[C] damping matrix
Cp hydrodynamic pressure coefficient
d maximum element size
E Young’s modulus
Ed∗ three dimensional strain invariant
f frequency of vibration
F yield function
f1 fundamental frequency of the reservoir
Fdyn total hydrodynamic force
FH horizontal pseudo-static force
Fst total hydrostatic force
FV vertical pseudo-static force
F (ω) transfer function
FA Fourier Amplitude
g acceleration of gravity
g(θ) gradient of the yield function in the J − p0 plane
xxxviii
G shear modulus
Gmin , Gmax minimum and maximum values of the shear modulus
H height of the dam
I moment of inertia
Jo (·) Bessel function
J∗ three-dimensional stress invariant
JL parameter of the cyclic nonlinear model
[K] stiffness matrix
K permeability of materials
Kd bulk modulus of the dam materials
kH horizontal pseudo-static coefficient
Kw bulk modulus of water
kV vertical pseudo-static coefficient
L length
m mass
[M ] mass matrix
p0 mean effective stress
pf pore fluid pressure
Pst static water pressure
Pdyn hydrodynamic pressure
r distance of the boundary from the excitation source
Sa spectral acceleration
t time
T natural period of vibration
u displacement
u̇ velocity
ü acceleration
Un (·) shape of the nth mode of vibration
Vp p-wave velocity
Vs shear wave velocity
W width of the dam crest
x horizontal direction
y vertical direction
xxxix
z elevation
α parameter of the CH time-integration scheme
αf parameter of the CH time-integration scheme
αm parameter of the CH time-integration scheme
Γ(·) Gamma function
γ shear strain
γw unit weight of water
δ parameter of the CH time-integration scheme
θ Lode’s angle
λ wavelength
ν Poisson’s ratio
ξ damping
ρ mass density
ρ∞ spectral radius at infinity
σ normal stress
τ shear stress
φ angle of shearing resistance
χL parameter of the cyclic nonlinear model
ψ angle of dilation
ω circular frequency of harmonic load
ωd fundamental circular freqency of the dam
ωr fundamental circular freqency of the reservoir
Ω frequency ratio
xl
Chapter 1
INTRODUCTION
1.1 Background
In recent years, the need for services to the global community resulted in
the construction of various types of infrastructure around the world, such as
highways, bridges and dams. A considerable number of earth and rockfill
dams are built in areas of significant seismic activity and some of them have
experienced earthquakes with catastrophic results. These dams are in need
of seismic retrofit, which includes seismic analysis, re-design and upgrade
actions. Moreover, new dams proposed to be built in such areas need to
be designed to resist expected earthquake loads according to local seismic
hazard studies and achieve a certain level of seismic performance.
Additionally, dams built in areas of insignificant or low seismic activity
have up to now been designed without any consideration of seismic loads.
Though, the design life of dams is quite large and the safety of the dam
structure is quite critical for loss of life and socio-economic losses. Conse-
quently, the need of earthquake design even for these areas of low seismicity
has been outlined recently (e.g. introduction of Eurocode 8) and it depends
on the expected ground motions based on probabilistic hazard assessments
and the design life of structures.
The performance of dams during construction, impounding and opera-
tion can be evaluated reasonably well as the methods used for static analyses
1
provide results comparable to the observed measurements. However, the

response of dams during earthquake activity, although theoretically (using
simplified analytical models) has been widely studied, it is not very well
established as the theoretical models do not always match the observed be-
haviour. The seismic response of earth dams is admittedly rather complicated
and therefore advanced methods of dynamic analysis need to be employed in
order to capture the actual behavior of dams under seismic conditions. Such
methods and associated advanced constitutive models do exist nowadays, but
they need to be further developed and validated against known case studies,
so that reliable results can be obtained for further dam analysis and design.
1.2 Scope of research

The main aims of this research were (a) to review the current state of the
knowledge related to the seismic response of earth dams and the available
methods of analysis and (b) investigate a real case study of seismic dam
response in order to establish and further improve the current knowledge of
both seismic response and analysis of earth dams.
Therefore, the first objective was to review the available literature and
present the main information found related to the response and analysis of
earth dams. Various geometric and material constraints that affect the dy-
namic behaviour of dams were identified, along with the different patterns of
the hydrodynamic pressures from the reservoir and the possible consequences
of a seismic event. Besides, the relative merits of the available methods of
analysis, both analytical and numerical, were investigated and the numerical
methods were examined in more detail.
The second objective was to explore the effects of reservoir-dam interac-
tion on the seismic behaviour of dams. Firstly, a methodology was established
in order to discretise the reservoir domain with displacement-based finite el-
ements and therefore model the hydrodynamic pressures on dams. This was
followed by a study of the reservoir-interaction effects in both a cantilever
and a trapezoidal elastic dam.
The third objective was to investigate the nonlinear behaviour of earth
2
dams using numerical analysis. The La Villita earth dam in Mexico was cho-
sen as a case study and advanced nonlinear elasto-plastic finite element anal-
yses were performed. The calculated response was compared to the recorded
behaviour, exhibiting a good agreement. Finally, using the same dam, sev-
eral parametric studies were carried out to assess the effects of different mod-
elling approaches on the predicted seismic response of earth dams. Therefore,
comments were made regarding the applicability of the adopted modelling
techniques and constitutive assumptions.
1.3 Layout of thesis

The work presented in this thesis is organised in two parts. The first part
(Chapters 2 and 3) contains a review of the literature related to seismic
response and analysis of earth dams. The second part (Chapters 4 - 7) con-
tains new work carried out in this study which concentrates on (a) studies of
reservoir-dam interaction and (b) nonlinear elasto-plastic numerical analysis
of earth dams.
Chapter 2 presents an outline of the information found in the litera-
ture related to the seismic performance of dams. Firstly, a description of
the various types of dams is presented, followed by a description of a typical
construction sequence of earth dams. Secondly, the seismic behaviour of soil
deposits is explored focusing on wave propagation, elastic ground response
and nonlinear dynamic soil behaviour. Subsequently, the behaviour of dams
is investigated and their dynamic characteristics based on both material and
geometric constraints are discussed. This is followed by a discussion about
the hydrodynamic pressures on dams imposed by the upstream reservoir un-
der various conditions. Finally, the effects of earthquakes on dam structures
are evaluated and the consequences of a seismic event are detailed through
information from field investigations.
Chapter 3 describes the various methods used in seismic analysis of earth
dams. It consists of three parts: (a) an overview of the different methods
of seismic analysis of dams, (b) a more detailed description of the numerical
methods and (c) a discussion of several considerations regarding dynamic
3
finite element analysis which is employed in this thesis.

Chapter 4 describes work carried out on modelling the hydrodynamic
pressures on dams during earthquakes and general dynamic loads using finite
elements. The reservoir water was descritised under plane-strain conditions
using displacement-based eight-noded quadrilateral solid elements, which are
the same as the elements used to discretise the dam and the foundation soil.
Several aspects of the problem are discussed, such as the upstream reservoir
boundary condition and its distance from the dam, the reservoir-dam and
reservoir-base interface, element size, along with the compressibility, shear
stiffness and damping properties of the reservoir water.
Chapter 5 describes new work carried out in order to further investi-
gate reservoir-dam interaction and its influence on the dynamic acceleration
response of dams. Two types of dams were considered in order to evalu-
ate the influence of reservoir-dam interaction on the dynamic response of
dams: (a) a rectangular cantilever dam with a vertical upstream face and
(b) a trapezoidal wide earth dam with a sloped upstream face. The response
of both dams was investigated under both harmonic and random (seismic)
acceleration loads.
Chapter 6 includes numerical analysis of the seismic response of La Vil-
lita earth dam in Mexico employing the finite element method. Static and
dynamic nonlinear elasto-plastic analyses were performed and the calculated
response was compared to the recorded behaviour, exhibiting a good agree-
ment. Several issues were addressed which provided an insight into the real
behaviour of the dam and also provided valuable feedback and information
about considerations that need to be taken when performing seismic analysis
of earth dams using the finite element method.
Chapter 7 builds on the knowledge acquired from the previous chapter
and contains several parametric studies performed to assess the effects of dif-
ferent modelling approaches on the predicted seismic response of earth dams
using La Villita dam. The examined issues are (a) the effects of any previ-
ous seismic activity prior to the studied earthquake, (b) the relative merits
of coupled and uncoupled undrained dynamic analysis, (c) foundation-dam
interaction effects and modelling approaches, (d) nonlinear reservoir-dam
4
interaction and (e) stiffness inhomogeneity. Previous and where available re-
sults from the literature were also discussed and compared to the outcomes
of this study. Consequently, comments were made regarding the applicability
of the adopted modelling technique and constitutive assumptions.
Finally, Chapter 8 summarizes the findings of this study and describes
the various recommendations for further investigation. At the end, some
thoughts are discussed related to the relative advantages and use of analytical
and numerical methods of analysis.
5
Chapter 2
SEISMIC PERFORMANCE
OF DAMS
2.1 Introduction
This chapter describes issues related to the seismic performance of dams.
Information drawn from the literature is presented in five sections.
Firstly, a description of the various types of dams is presented in order
to obtain an insight into their materials and geometry and to identify any
vulnerable regions within the dam structure. This is complemented by a
description of a typical construction sequence of earth dams in order to un-
derstand the stress history within the dams and the loading conditions prior
to the seismic events. Secondly, the seismic behaviour of soil deposits is ex-
plored focusing on wave propagation, elastic ground response and nonlinear
dynamic soil behaviour. Subsequently, the behaviour of dams is investigated
and their dynamic characteristics based on both material and geometric con-
straints are discussed. This is followed by a discussion about the hydrody-
namic pressures on dams imposed by the upstream reservoir under various
conditions. Finally, building on the seismic behaviour of dams, the effects
of earthquakes on dam structures are evaluated and the consequences af-
ter a seismic event are detailed through information from post-earthquake
observations and field investigation reports.
6
2.2 Design & construction of dams

There are several types of dams constructed with different materials, geome-
tries and facilities. It is obvious that the choice of type and materials is
not random, but it depends on a number of parameters. Such parameters
are the engineering purpose, site conditions, material availability, ease and
price of construction, canyon geometry, seismicity etc. This section discusses
different types of dams and describes how these structures are designed and
constructed in engineering practice. The construction details are important,
as they directly affect the stress history of the structure and dictate the
material behaviour, especially for soils which behave in a highly nonlinear
elasto-plastic manner.
The different types of dams are mainly related to the material used for
their construction and its availability and price. Two broad categories may
be identified, which depend on the main material used to control (stop or
minimize) seepage. The first group includes the earth dams, which are mainly
constructed by soil compaction. The second group is related to concrete
dams, which are generally smaller in volume than earth dams.
2.2.1 Earth dams

Earth dams use geomaterials as the main construction material. These ma-
terials are placed in horizontal layers and they are then compacted in order
to increase their strength and reduce their deformation. The dams may be
uniform (i.e. consisting of a single material), zoned (i.e. consisting of sev-
eral zones of different materials), or covered with an impermeable upstream
membrane. Therefore, they generally fall in three categories: uniform earth,
zoned earth and upstream membrane rockfill dams.
The case of uniform dams is not very usual and it is the simplest. In
fact, not many dams are built with a single geomaterial nowadays, or at
least not many large dams. This is perhaps the first type of dams built
in the ancient times, when the lack of technical knowledge, construction
equipment and resources did not allow the design and construction of more
complicated structures. Uniform earth dams (Figure 2.1) are constructed as
7
a simple embankment of well compacted earth and they are generally very
wide (slopes 1:3) in the cross section and not very high (Fell et al., 2005).
Their major use today is for coffer dams, i.e. as temporary water retaining
structures, and they may consist of a low permeability material such as clay.
LOW PERMEABILITY
MATERIAL SUCH AS
CLAY
Figure 2.1: Two-dimensional sketch of a uniform earth dam
The second general type of earth dams, which is the most widely used,
is the zoned earth dam. These structures consist of a number of zones de-
pending on the soil used (Golzé, 1977). Generally, there are three main zones
(according to their function), starting from a central core which consists of a
low permeability material (such as clay) and its function is to minimize seep-
age of water through the dam section. Immediately after the central core
there is a thin zone of filters (which is mostly found at the downstream side
of the core). Filters consist of a coarser material, usually coarse aggregates,
similar to those used in concrete mixing, and act as a drainage path for the
water which has seeped through the core. This material is not very fine, in
order to allow some seepage, and also not very coarse so that erosion of the
core is prevented by stopping fine core material being washed through. The
final and outer zone is mainly constructed using a coarse material, which is
usually rockfill. Its function is to provide support and stability of the inner
seepage control system.
3 1. CLAY CORE
4 2. SAND FILTERS
1 3. ROCKFILL
2 4. RIP-RAP
Figure 2.2: Two-dimensional sketch of a zoned earth dam
Besides, many dams have an additional thin upstream zone consisting of

coarse rockfill or rip-rap which serves as a protection of the finer material
8
from erosion due to waves generated in the reservoir. Figure 2.2 shows a
sketch describing the geometry and basic features of a zoned earth dam,
whereas Figure 2.3 shows the Kourris dam which is the largest zoned earth
dam in Cyprus. Zoned earth dams are also called “clay core dams” because
of the core’s major function.
Figure 2.3: The Kourris dam is the largest zoned earthfill dam in Cyprus
(Cyprus Public Works Department, 2010).
Upstream membrane rockfill dams use a combination of earth and man-

made materials. The idea is to use a non-permeable thin surface (membrane)
on the upstream face of the dam, so that seepage is controlled before water
penetrates the dam body (Golzé, 1977). The two most commonly used mate-
rials for this membrane are asphalt and concrete. The latter case is nowadays
increasingly used, as it has been proved to be economically efficient and this
dam type is generally known as the Concrete-Faced Rockfill Dam (CFRD).
A sketch outlining the geometry of a CFRD is shown in Figure 2.4.
2
1 1. CONCRETE FACE
2. ROCKFILL
Figure 2.4: Two-dimensional sketch of a concrete faced rockfill dam
9
Figure 2.5: The Kannaviou dam, built in 2005, is the first and until today
(2012) only Concrete-Faced Rockfill Dam in Cyprus (Cyprus Public Works
Department, 2010)
An example of this type of dam is the Kannaviou dam in Cyprus which

is shown in Figure 2.5 (Cyprus Public Works Department, 2010). The figure
shows the upstream (close to the reservoir) concrete face and the downstream
rockfill face. For any material used for the upstream membrane however, the
rest of the dam body is constructed using rockfill, in order to provide support
and stability of the non permeable membrane. No fine material needs to be
used beyond the upstream membrane, as theoretically there is no seepage
through it and hence no further seepage control is required.
However, care needs to be exercised when designing CFRDs to with-
stand strong earthquakes (Gazetas and Dakoulas, 1992). Large transverse
and longitudinal vibrations of the dam body could result in severe damage
of the upstream membrane through cracks from high tensile stresses (Dak-
oulas, 2012b) or openings of face joints due to compressive slab-to-slab forces
(Dakoulas, 2012a).
10
2.2.2 Concrete dams

Concrete dams use concrete as the main body material, which is imperme-
able and controls seepage through the dam effectively. Because of this good
performance, less material needs to be used in concrete dams, and there-
fore these dams have generally smaller volume and thinner cross sections
(Fell et al., 2005). The use of concrete dams is usually preferred in narrow
and deep canyons where an earth dam would not demonstrate satisfactory
stability.
Moreover, because of concrete’s material properties and especially its
good performance in compression and (if reinforced) in tension, different
geometries may be used. However, concrete’s range of plasticity and duc-
tility and its continuous solid state, as opposed to soil’s particulate nature,
allows the creation and propagation of cracks.
Several types of dams belong to this category, related to the geometry
and construction method. The main types of concrete dams are Gravity,
Buttress , Arch dams and Roller Compacted Concrete (RCC) dams.
Concrete gravity dams have normally a large body made of concrete and
their stability is provided by their large volume (and hence mass and weight)
and base surface area, imposing a large stabilising force on the ground (U.S.
Army Corps of Engineers, 1995). Gravity dams are found in both wide and
narrow canyons. A sketch outlining a concrete dam is shown in Figure 2.6,
whereas an example of a gravity dam, the Lympia dam in Cyprus, is shown
in Figure 2.7 (Cyprus Public Works Department, 2010). It may be observed
from the latter figure that a concrete dam is much thinner and smaller in
volume than an embankment dam (Figure 2.3).
SOLID
CONCRETE
Figure 2.6: Two-dimensional sketch of a concrete gravity dam
Buttress dams are made from concrete or masonry and they are very sim-
11
Figure 2.7: The Lympia dam, built in the late 1970’s, is an example of a
small gravity dam in Cyprus (Cyprus Public Works Department, 2010)
Figure 2.8: The Stwlan dam in Wales is a buttress dam (British Dam Society,
2010)
12
ilar to the gravity dams. They have a watertight upstream side supported
by triangular shaped walls, called buttresses. The buttresses are spaced at
intervals on the downstream side. The main difference between gravity and
buttress dams is that they use less material because of the clear spaces be-
tween the buttresses. The buttresses act as stiffeners providing flexural sup-
port to the dam body and therefore the stability of the dam is not provided
entirely from its weight, but also from its strength. Their upstream face, is
not vertical but inclines about 25o to 45o from the vertical, so the resulting
force of the water on the upstream face inclines towards the foundation. The
Stwlan Dam in Wales shown in Figure 2.8 (British Dam Society, 2010) is
such an example. The figure shows the buttresses on the downstream side of
the dam which provide additional resistance to the dam body.
Concrete arch dams are found in very steep and narrow canyons where
space is not adequate to build a large embankment or gravity dam. These
dams are normally thick at the base and they become significantly thinner
when they reach the top of the crest. They are curved in the plan view, with
the convex side of the arch being on the upstream face (U.S. Army Corps of
Engineers, 1994). The reason for constructing dams with such geometry is to
take advantage of the concrete’s high compressive strength, as an arch loaded
on the convex side is able to carry great compressive stresses. Therefore, the
narrow canyon geometry, in which a small-span arch is built, provides great
resistance to the water load. An example of an arch dam is the Baitings dam
in Yorkshire shown in Figure 2.9 (British Dam Society, 2010). It may be
observed from the figure that the reservoir water is on the convex side of the
arch.
Roller Compacted Concrete (RCC) dams are a special type of gravity
dams built using the RCC construction method. Roller-Compacted Concrete
is a special blend of concrete that has the same ingredients as normal concrete
but in different mixing ratios (U.S. Army Corps of Engineers, 2000). Cement,
water and aggregates are the main constituents, as in the general concrete
mixing, but RCC is much drier and it is placed in a manner similar to paving.
The dam is built in successive horizontal layers resulting in a downstream
slope that makes a concrete staircase. Once a layer is placed, it can very
13
Figure 2.9: The Baitings dam in Yorkshire, England is an example of an arch

dam (British Dam Society, 2010)
Figure 2.10: The Willow Creek Dam in Oregon, USA is the first Roller
Compacted Concrete dam in the world (US Army Corps of Engineers, 2010)
14
quickly support large loads such as the equipment to place the next layer.
After RCC is deposited on the lift surface, small dozers typically spread it
in moderately thin layers of about 0.30m.
RCC materials offer a number of advantages to the construction of con-
crete dams. Firstly, it is more economic than normal gravity dams and an
RCC dam generally requires less material than a gravity dam of the same
height. Moreover, the quick curing and strengthening of the concrete result
in rapid construction as the successive layers of concrete can be constructed
in a very short time. On the other hand however, there is a number of disad-
vantages associated with RCC dams. The main disadvantage is the seepage
control, especially between two successive horizontal layers of concrete (con-
struction joints). For this reason, particular attention needs to be paid to
watertightness and the design of the joints within the dam body, especially
in the case of extreme loading such as earthquakes. The Willow Creek Dam
in the Morrow County of Oregon, US was the first RCC dam to be built
in the world (1979-1983) and it is shown in Figure 2.10 (US Army Corps of
Engineers, 2010).
The previous sections described the various types of earth and concrete
dams and their general features. However, as this work is related to the
performance of earth dams, the rest of this review and research is focused on
the case of earth dams.
2.2.3 Design considerations

Water retaining structures are large-scale engineering projects and their de-
sign is different for each case. Because of the large size and the remote
locations of these structures, a site specific study is usually required for each
project. Moreover, as these projects are not designed on a routine basis,
there are no specific design codes which standardize the design process, as it
is the case for example for residential buildings.
Therefore, each dam project needs special design considerations and per-
haps different construction procedures and equipment. This involves work
from different specialists from a range of technical fields, such as geotechnics,
15
geology, hydrology, concrete technology, seismology etc. However, the various

design considerations can be divided into two major categories, hydrological
and geotechnical (Golzé, 1977).
The decision for the construction of a dam follows the need for water
storage as part of a hydrological policy in a large geographical area. There-
fore, the first considerations are more related to issues concerned with the
hydrology of an area, which determine the function of the dam. Hydrological
design considerations for dams (Linsley et al., 1982) focus mainly on:
• Reservoir modeling - as a storage point of the hydrological network
• Water balance modeling - defining water demand
• Probable maximum precipitation - defining water supply
• Reservoir water quality - long term maintenance of the reservoir
The second group of considerations is more related to the dam structure

itself and its engineering performance (mainly stability and safety) through
its construction process and operation. For this reason, detailed site inves-
tigation needs to be carried out in order to obtain information about the
geology, hydrogeology and geotechnics of the area.
The geology of the area provides information about rock formations, tec-
tonic structures such as faults etc, so that engineers can decide on the exact
location of the embankment. This is also taken into consideration when it
comes to canyon rock strengthening (using anchorage etc), location of spill-
ways, diversion tunnel excavation etc.
Besides, a geotechnical investigation of the area can provide an insight
into the quality of the foundation material and any possible location of bor-
rowing pits for the material of the embankment body. Finally, hydrogeologi-
cal studies are also important in order to locate any possible aquitards (zones
of low permeability), permeability of the various geologic formations and then
decide on the grout curtain design, or any further reservoir protection (such
as “blocking” any near surface water canals).
16
2.2.4 Construction processes

The construction of earth dams is site-specific, because of the different site
conditions and needs for each project. It requires special construction pro-
cesses, material tests and, in some cases, special equipment and machinery.
The general procedure followed to construct an earthfill dam is outlined be-
low (Parmakkis, 2006).
a. River diversion
The first issue at a dam construction site is the river flow. Even in
cases of construction during dry summers, water stream from the river has
to be diverted from the site. This is achieved by excavating a diversion
tunnel for the stream to pass and by constructing a temporary cofferdam as
an artificial protective barrier to the construction site. In some cases, the
cofferdam may be part of the final embankment (such as the rockfill shells),
if it is constructed using appropriate material.
b. Foundation
Subsequently, the riverbed at the location under the embankment is
trenched to remove the local material which might not perform well as a
foundation layer. During this stage, if the underlying stratum does not con-
sist of an impermeable rock, a series of thin grouted piles are installed in a
row (or more) below the crest line of the dam. This group of piles, known as
the grout curtain, serve as a cutoff membrane which prevents (or controls)
seepage under the embankment. On top of the grout curtain, a block of
concrete is constructed which links the top of all these piles and it is called
grout cap. This also serves as the foundation of the embankment and rests
on the riverbed material.
c. Embankment
Once the concrete block is constructed, the material of the embankment
body needs to be placed. This is performed in several stages and the material
is placed in a number of thin (relative to the total height of the dam) layers.
Each layer is placed on top of the previous layer and it is compacted in order
to increase its density and strength. Special attention is paid to the core clay
which is the most important part, so that its water content is kept close to
17
the optimum value. For this reason, the core is wetted as it is constructed,
so that it does not lose its moisture.
d. Spillway
Once the embankment is finished, the spillway is constructed in order to
accommodate overflow of the water safely. The spillway is usually placed
next to the embankment on stiff rock abutments and its crest is lower than
the embankment’s crest to allow for some freeboard. The rock abutments
on the sides of the river canyon are often strengthened (stabilized) with
anchors as the excavation of rock (especially if using blast techniques) during
construction of the spillway may affect their stability.
e. Additional facilities
Finally, after construction of the main embankment and its operational
facilities, the dam structure is complete and additional features may be added
to it for secondary purposes. In cases where a dam serves as a link between
the two sides of the river (as this is the case for some long dams), a road
might have been included in the design to provide access. Construction of
this piece of infrastructure is the final part of the construction process even
though it is not directly related to the prime function of a dam.
Besides, if the dam is used for power generation, the hydroelectric facili-
ties are also constructed nearby. They need to be close to the embankment
and in a convenient place so that there is way for incoming water from up-
stream and discharge in the downstream. Moreover, there needs to be access
to the power plant and perhaps additional space for monitoring facilities.
2.3 Seismic behaviour of soil deposits

2.3.1 Earthquake generation & wave propagation
Generally an earthquake occurs when accumulated strain energy is released
after gradual deformation of lithosphere due to tectonic movements during
long periods of time and due to fault activity as shown in Figure 2.11 (US
Geological Survey, 2010).
Tectonic movements, as defined in the general theory of plate tectonics,
18
Figure 2.11: Normal faulting: the hanging wall has moved downwards with
respect to the footwall block (US Geological Survey, 2010)
are the result of movement of magma in the interior of the earth due to
temperature difference with the depth. This disturbance in the earth medium
propagates in a wave form in two mechanisms, compression and shear. These
mechanisms create a system of both longitudinal and transverse body waves
(also called pressure, P- and shear, S-waves respectively), as shown in Figure
2.12 (Goree, 2010). These body waves interact with each other and with
other boundaries and create more types of waves, generally called surface
waves, i.e. Rayleigh (1885) & Love (1927) waves (Kramer, 1996).
(a) P wave (b) S wave
Figure 2.12: Seismic body waves (Goree, 2010)
The motion of these waves causes the soil structure to deform and the soil
properties to change. Soil density tends to increase due to the relocation of
soil particles, a phenomenon called dynamic compaction (or densification). If
water is present, the strength can reduce significantly because of excess pore
19
water pressures and hence liquefaction may be observed. Liquefaction is an

important issue that needs to be appropriately addressed when dealing with
seismic soil behavior especially in dams and refers to loss of shear strength
due to generation of excess pore water pressures under undrained loading.
Seismic waves propagate from the source of the earthquake to the site
of the structure. It is important to note that these waves attenuate with
distance due to two different mechanisms, material and radiation damping.
The first is related to dynamic material behaviour and the second is related
to the geometrical spreading of the waves within the earth’s space.
2.3.2 The scenario of reservoir-triggered seismicity

The huge mass of water in a dam reservoir increases the pressure in the
underlying rock, which can trigger earthquakes by further stress increments
on the tectonic formations. Moreover, large stresses may re-activate old
inactive faults by destroying the bonds between the two parts of the fault
(Zhu, 1992; McCully, 1996).
A number of earthquakes have been attributed to the existence of large
dam reservoirs in the area. As an example, the 1967 (6.3 Magnitude) earth-
quake in Maharashtra, India, should be noted, which occurred close to the
Konya Dam reservoir (Chadha, 1995). Besides, the 1965 (6.1 Magnitude)
earthquake at Oroville, California, was attributed to seismicity induced by
a massive earthfill dam and reservoir recently constructed and filled closeby
(Beck, 1976; Morrison et al., 1976). Other studies related to reservoir-induced
seismicity include the Tarbela Reservoir in Pakistan (Jacob et al., 1979), the
Hsinfengkiang reservoir in China (Tsung-ho et al., 1976) and the Heron and
El Vado Reservoirs in New Mexico (El-hussain and Carpenter, 1990).
2.3.3 Visco-elastic ground response

Ideally a complete ground response analysis would model all the parts of
the earthquake phenomenon, i.e. source (fault), path (wave propagation
from source) and site (ground surface). However, because this is a rather
complicated task with inherent uncertainties, seismic hazard analyses are
20
conducted and then the site response problem concentrates on the response
of a soil deposit to the motion of the underlying bedrock.
The engineering significance of ground response is that the soil layer may
amplify or de-amplify the bedrock motion and even change its frequency
content. Moreover, surface topography (such as a triangular dam) and basin
effects (such as river canyons) are also important, because they affect the
dynamic response of the dam, as geometry affects the medium’s stiffness.
Several methodologies for ground response analysis were developed over time,
both analytical (Thompson, 1950; Ambraseys, 1959a; Seed and Idriss, 1969;
Gazetas, 1982b; Vrettos, 2013) and numerical (Haskell, 1953; Biot, 1963;
Kausel and Röesset, 1981; Schnabel et al., 1970; Phillips and Hashash, 2009).
Analytical methodologies adopt various simplifying assumptions and provide
valuable information about the dynamic response (such as natural periods
of vibration, amplification of the motion etc.) of soil layers (damped or
undamped) on top of the bedrock (rigid or not).
u
z uniform
visco-elastic vertically
H soil layer propagating
shear wave
rigid bedrock
Figure 2.13: Uniform visco-elastic soil layer
An analytical calculation of the amplification of the motion can be ob-

tained at the free surface of a soil stratum. Harmonic horizontal motion
of the rigid bedrock produces 1D vertical shear wave propagation in a uni-
form visco-elastic soil layer with a free surface (τ (z=0)=0) (Kramer, 1996)
as shown in Figure 2.13. The amplification factor can be obtained from the
modulus of the Transfer Function, |F (ω)|, described by Equation 2.1.
umax (z, t)
F (ω) = (2.1)
umax (H, t)
where ω is the circular frequency of the applied harmonic load, umax is
21
the maximum value of the horizontal displacement, u, z is the depth from

the ground surface, H is the height of the soil stratum and t is time. The
amplification function for a uniform damped soil layer, with free surface
(τ (z=0)=0) on rigid rock is given by Equation 2.2.
1
|F (ω)| = q (2.2)
cos2 (ωH/Vs ) + [ξ(ωH/Vs )]2
where Vs is the shear wave velocity of the soil, H is the height of the soil
deposit, ω is the circular frequency and ξ is damping ratio.
The frequency spectrum of the amplification (Equation 2.2) is shown
graphically in Figure 2.14 for three values of the damping ratio, ξ = 5%, 10%
and 20%. It is noticed that peak values of amplification of the response occur
at different values of the circular frequency, ωn , as described by Equation 2.3,
which correspond to the natural modes of vibration of the soil layer. It is
also shown that the higher modes of vibration are significantly affected by
the value of the damping ratio.
Amplification of a homogeneous soil layer

14
ξ=5%
12 ξ=10%
ξ=20%
Amplification Factor, |F(ω)| []
10
0
0 1 2 3 4 5 6
Frequency Ratio, ω/ω1 []
Figure 2.14: Amplification Factor, |F (ω)| for a visco-elastic soil layer on a

rigid bedrock
2n − 1 πVs
ωn = (2.3)
1 2H
22
The maximum value of amplification is given by Equation 2.4 and occurs

at circular frequency, ω1 , as given by Equation 2.5 which corresponds to the
first natural circular frequency of the soil layer.
1 2 0.637
|F (ω1 )| = |F |max = q = = (2.4)
cos2 (π/2) + [ξ(π/2)]2 πξ ξ
πVs
ω1 = (2.5)
2H
Therefore, the first three natural periods of vibration (i.e. associated with
the first, second and third natural modes) are given by Equation 2.6.
       

 T1 
 
 1/ω1 
 
 1 
 
 4 

4H H

 
 
 
 
 
 
 

T2 = 2π  1/ω2 = 1/3 = 1.333 (2.6)

 
  
  
 Vs  
 Vs
T3 1/ω3 1/5 0.8

 
 
 
 
 
   

2.3.4 Nonlinear dynamic soil behaviour

In general, seismic waves cause motion of the soil particles and changes to
the soil material properties. The deformation of the ground depends on the
material properties of the soil, mainly shear modulus, G and density, ρ and
also on the characteristics of the seismic waves propagating into the soil.
This is because the magnitude of the earthquake shock affects the amplitude
of vibrations and the induced shear strain γ on which stiffness degradation
is dependent (Das, 1993). Equation 2.7 defines the shear wave velocity (VS )
of the soil.
s
G
VS = (2.7)
ρ
Under seismic shearing, soil response can be highly nonlinear, especially
in the case of low plasticity granular soils such as sands and silts. Figure
2.15 (Ishihara, 1995) illustrates a typical shear stress-strain curve for cyclic
loading. The soil stress-strain curve is defined by an initial loading path (the
backbone curve, shown as a thick line, BOCA) and a hysteretic loop (line
23
ABEA) is formed during a loading and unloading cycle (Kramer, 1996).
Figure 2.15: Shear stress-strain relationship - Backbone curve and hysteresis

loop (Ishihara, 1995)
The slope of the curve at each point (i.e. for each value of shear strain, γ)
is the tangential shear modulus, Gtan , whereas the ratio of the current (max-
imum values in a cycle) shear stress to strain is the secant shear modulus,
Gsec (denoted as G in Figure 2.15). The area of a hysteresis loop represents
the amount of energy dissipated during that cycle. The damping ratio, ξ
(denoted as D in Figure 2.15) is defined as the ratio of the energy dissipated
over the maximum strain energy, and it is given by Equation 2.8.
1 ∆W
ξ= (2.8)
4π W
where ∆W is the area of the hysteresis loop, denoting the energy dissi-
pated in one cycle and W is the area of the triangle with sides τa (= τmax )
and γa (= γmax ) (Figure 2.15) corresponding to the peak energy during that
cycle (Kramer, 1996).
As shown in Figure 2.15, the values of shear modulus G and damping
ξ change with shear strain γ. According to their definition, G decreases
whereas ξ increases with increasing γ. These depend also on the confining
pressure, σ00 . Figure 2.16 describes the relation between G, ξ and γ, arising
from experiments on Toyoura sand (Kokusho, 1980).
24
(a) Stiffness, G/Gmax (b) Damping, ξ
Figure 2.16: Relation between shear stiffness, G, damping, ξ and cyclic shear
strain γ, arising from experiments on Toyoura sand (Kokusho, 1980)
2.4 Dynamic response of dams

The various types of dams were described in Section 2.2 and a brief descrip-
tion of the construction processes was presented. Apart from the layered
construction, typical engineering problems associated with earthfill dams are
the reservoir impounding, seepage flow through the embankment (and espe-
cially in the clay core) and consolidation with time. This section outlines an
additional engineering problem associated with dams: the response of dams
during earthquakes.
As explained in Section 2.3 an earthquake is a complicated phenomenon
which has to be treated as an additional type of load on engineering struc-
tures. However, its complexity comes from the fact that this load is applied
during a very short time interval and its magnitude and direction change
rapidly with time. Moreover, it can interact with the structure which may
affect the characteristics of the load in a dynamic manner. For this reason,
an insight into the dynamics of dams is essential to evaluate their response
during earthquakes.
2.4.1 Transverse vibrations

The vibrations of earth dams in the upstream-downstream direction were
initially investigated by Ambraseys (1960a) using the shear beam method
25
(see Section 3.4). The dam is represented by a truncated wedge as shown

in Figure 2.17. It is assumed to be homogeneous and symmetric with plane
inclined faces, thus deforming predominantly in shear (bending deformation
is considered negligible) and constructed in a rectangular canyon.
Figure 2.17: Geometry of the dam considered by Ambraseys (1960a).
Following the shear beam approach, the undamped circular frequency of

vibration, ωnr , for the (n, r) mode is given by Equation 2.9.
v
u !2
u
t 2 rπ Vs
ωnr = an+ (2.9)
µ H
where n refers to the number of the mode for deformation in the z-
direction with respect to the y-axis (see Figure 2.18 (a)), r refers to the
number of the mode for deformation in the z-direction with respect to the x-
axis (see Figure 2.18 (b)), µ = L/H is the ratio of the crest length, L, to the
height of the (full) wedge, H, as shown in Figure 2.17, Vs is the shear wave
velocity of the soil and an are constants related to the coefficient of trunca-
tion of the dam wedge, k 0 = h0 /H (where, h0 is the height of the truncated
wedge, as shown in Figure 2.17).
Ambraseys (1960a) listed the values of an for several values of the coeffi-
cient of truncation, k 0 , and the mode, n, in tabular form. For an untruncated
26
Natural Modes of Displacement Natural Modes of Displacement

1 1
First (n=1) Natural Mode
0.9 Second (n=2) Natural Mode 0.9
Third (n=3) Natural Mode
0.8 0.8
0.7 0.7
Normalised Length, z/H

Normalised Height, y/H
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2 First (r=1) Natural Mode

Second (r=2) Natural Mode
0.1 0.1 Third (r=3) Natural Mode
0 0
−0.5 0 0.5 1 −1 −0.5 0 0.5 1
Modal Displacement, Un(y) Modal Displacement, Ur(z)
(a) Modes n (b) Modes r
Figure 2.18: Natural modes of vibration of a dam in the upstream-

downstream direction: (a) Modes n, i.e. for deformation in the z-direction
with respect to the y-axis and (b) Modes r, i.e. for deformation in the z-
direction with respect to the x-axis.
wedge (k 0 = 0) and the first three natural modes of vibration (n = 1, 2, 3),

the values of an are: a1 = 2.4048, a2 = 5.5201 and a3 = 8.6537.
Therefore, considering a canyon for which L/H = 1, the first four natural
periods of transverse vibration, Tn,r (i.e. associated with the first, second,
third and fourth natural modes and the four largest values of the natural
period) of a triangular dam are given by Equation 2.10.
   




T1,1 







1.588 



   
T2,1 0.989 H

 
 
 

= (2.10)


 T1,2 




 0.934 

 Vs

 
 
 

T2,2 0.751

 
 
 

Figure 2.19 shows the first two natural modes of vibration of a dam in
the upstream-downstream direction for a complete wedge (k 0 = 0), i.e. for
(a) n = 1 and r = 1 and (b) n = 1 and r = 2 (refer also to Figure 2.18).
However, if an infinitely wide canyon (L/H → +∞) is considered, the un-
damped circular frequency of vibration of the nth mode is given by Equation
2.11.
27
(a) Mode (n,r) = (1,1) (b) Mode (n,r) = (2,1)
Figure 2.19: First two natural modes of vibration of a dam in the upstream-
downstream direction for a complete wedge (k 0 = 0) (Ambraseys, 1960a).
Vs
ω n = an (2.11)
H
Therefore, the first three natural periods of vibration (i.e. associated with
the first, second and third natural modes) are given by Equation 2.12.
   

 T1 
 
 2.613 

H

 
 
 

T2 = 1.138 (2.12)
 Vs

 
 
 

 T
 
  0.7261
 
3
Note that the natural periods of a homogeneous dam of height, H and

shear wave velocity, Vs are smaller than the corresponding periods of a ho-
mogeneous soil layer with the same properties (Equation 2.6). Therefore, a
homogeneous triangular dam exhibits a stiffer response than a uniform soil
layer of the same height and material properties (Vs ).
The displacement shapes of the natural modes of vibration, Un (y), (repre-
senting the normalised maximum horizontal displacement in the dam body,
under harmonic loading at steady-state conditions) are given by Equation
2.13. Figure 2.18 shows the corresponding maximum displacement shape of
the first three natural modes.
y

Un (y) = Jo an (2.13)
H
It is worth noting that according to the shear beam approach of Am-
braseys (1960a), the frequencies and modes of natural vibration do not de-
pend on the slope angle of the dam (Equations 2.12 and 2.13).
28
An estimate of the amplification of the earthquake motion at the crest

of a dam may be obtained from the Amplification Factor, |F (ω)|. This is
described by Equation 2.14 (Dakoulas and Gazetas, 1985) and it is shown
graphically in Figure 2.20 for three values of the damping ratio, ξ = 5%, 10%
and 20%. Note that for a damped (visco-elastic) dam, the shear wave velocity,
Vs should be replaced by the complex shear wave velocity, Vs∗ = Vs (1 + iξ)
√
(where, i = −1, is the imaginary unit). For the general expression referring
to the amplification of an inhomogeneous dam resting on a compliant rock
foundation, see Section 3.4 and Equation 3.10.
1
|F (ω)| = (2.14)
Jo ( ωH
Vs
)
Amplification of a homogeneous earth dam

18
ξ = 5%
16 ξ = 10%
ξ = 20%
14
Amplification Factor, |F(ω)| []
12
10
0
0 1 2 3 4 5 6 7
Frequency ratio, ω/ω []
1
Figure 2.20: Amplification Factor of a homogeneous earth dam. After Dak-

oulas and Gazetas (1985).
As it may be observed from Figure 2.20, the maximum amplification of

the base motion is associated with the fundamental mode of vibration (i.e. for
ω1 ) and its value reduces for higher values of the damping ratio. Besides, it
is shown that larger values of damping have a more pronounced effect on the
higher modes of vibration, which is also the case for a soil layer (see Section
2.3.3 and Figure 2.14). Moreover, the amplification for a homogeneous earth
dam (Equation 2.14) is generally larger than the amplification for a uniform
29
soil layer (Equation 2.2), as shown by Figures 2.14 and 2.20.

Similar studies were conducted for vertical (Gazetas, 1981b) and longi-
tudinal (Gazetas, 1981a) vibrations of earth dams and provided information
about natural modes of vibration. However, these are not discussed here as
vertical and longitudinal vibrations of dams are not considered in this thesis.
2.4.2 Three-dimensional canyon geometry

It has generally been shown that dams built in narrow deep canyons behave
in a stiffer manner than dams built in wider canyons. The stiffening effect
of a narrow canyon results in smaller natural periods. According to Gazetas
(1987), the assumption of two-dimensional plane strain conditions is valid
for infinitely long dams, where the stiffening effect of the canyon is absent or
negligible. The effect of the three-dimensional geometry of the dam-canyon
system was firstly studied in the 1950’s by Hatanaka (1952) and Ambraseys
(1960b). These authors concluded that canyon effects are negligible for dams
in rectangular canyons with length to height ratios, L/H > 4. Later in-depth
investigations include the work of Dakoulas and Gazetas (1986, 1987) using
the shear beam method and Mejia and Seed (1983) using the finite element
method.
Figure 2.21 shows a comparison between the natural frequencies of dams
computed from 2D and 3D finite element analyses of dams in triangular and
rectangular canyons by Mejia and Seed (1983). The results were expressed
as a function of the length to height ratio, L/H, of the dam. It is shown that
the ratio of the three-dimensional to the two-dimensional natural frequency
of the dam is increasing with decreasing L/H ratio and may reach a value
of 2.5 for triangular dams (i.e. the fundamental period can be as small as a
fourth of the period for an infinitely long dam).
The findings of a more comprehensive study by Dakoulas and Gazetas
(1987) are shown in Figure 2.22, which considered more canyon geometries.
The results were presented in the form of the fundamental period, T1 of a
dam in a narrow canyon over the period T∞ of a dam in an infinitely wide
canyon, as a function of the length over height ratio, L/H, of the dam. As
30
4
f3D /f 2D
1 2 3 4 5 6 7
L/H
Figure 2.21: Comparison between natural frequencies from 2D and 3D anal-
yses of dams for different length to height, L/H ratios (Mejia and Seed,
1983).
T1 H
0.5
T1
8
0
0 1 2 3 4 5 6
L/H
Figure 2.22: Effect of canyon geometry on the fundamental natural period,

T1 of a dam (Dakoulas and Gazetas, 1987).
31
it may be observed from the figure, the fundamental period of a dam in a

narrow canyon can be as small as a fifth of the corresponding dam in a wide
canyon for triangular dams with small ratios of L/H.
DAMPING 10%
MIDCREST AMPLIFICATION FUNCTION
SEMI-CYLINDRICAL
CANYON
PLANE STRAIN
SHEAR BEAM
DIMENSIONLESS FREQUENCY αo= ω H

Vs
Figure 2.23: Midcrest amplification for dams under plane-strain conditions

and in a semi-cylindrical canyon (Dakoulas and Gazetas, 1986).
In addition to the change in the fundamental period of the dam, the

canyon affects the acceleration response of the crest of the dam. Figure 2.23
(Dakoulas and Gazetas, 1986) shows the amplification of the midcrest of a
dam for plane strain conditions and in a semi-cylindrical canyon (for a more
generalised approach involving semi-elliptical canyons, see Dakoulas and Hsu
(1993)). It is shown that the amplification for dams in narrow canyons can be
significantly higher than for dams in wide canyons (modelled as plane shear
beam) especially for the higher modes of vibration. Besides, an increase in
the frequencies is noticed for all the natural modes of vibration in comparison
to the plane strain shear beam. This is supported by field evidence from the
El Infiernillo dam in Mexico (Elgamal and Gunturi, 1993) and the Zipingpu
dam in China (Guan, 2009). It should be noted that these effects were very
pronounced for the case of El Infiernillo dam (L/H = 2.4), where a crest
amplification of 5 times of the bedrock acceleration was observed during the
32
15/11/1975 Mexico earthquake and may be partly attributed to the very

narrow canyon geometry of the dam.
Considering the previous information from the literature, the stiffening ef-
fect of the three-dimensional geometry of the canyon should be taken into ac-
count when performing seismic analysis of dams. Therefore, two-dimensional
analysis may not be appropriate for dams built in very narrow canyons.
2.4.3 Inhomogeneous dam materials

Earth dams are usually constructed in layers which are then compacted as
described earlier in Section 2.2. This construction process results in non-
homogeneous dam materials, especially concerning the shear stiffness, Gmax
(i.e. spatial variation of the maximum value of G).
In a study of the Santa Felicia dam in California, Abdel-Ghaffar and
Scott (1979a,b) highlighted that in earth dams the shear wave velocity Vs
(and hence the shear modulus Gmax ) depends on the confining pressure. The
authors also reported that their observation was in agreement with earlier
experiments by Seed and Idriss (1970) who suggested that for clays, G is
0
proportional to the square root of the confining pressure, σm according to
0
Equation 2.15 (Note that G and σm are in psf).
0 1/2
G = 1000K2 (σm ) (2.15)
where K2 is a soil stiffness parameter, characteristic for each material and

determined experimentally.
Moreover, Abdel-Ghaffar and Koh (1981) expressed the variation of G
with respect to the depth from the dam crest normalised to the height of
the dam, (y/H)m , for various values of the exponent m. Using experimental
data from real dams in Japan, America and Europe, Gazetas (1981b) related
the confining pressure to the height of the dam using previous numerical
studies (Clough and Woodward, 1967; Poulos and Davis, 1974) and fitted
the relationship between the stiffness and the depth of the dam. Later, this
observation was quantified by Gazetas (1982a) who studied the observed
behaviour of Sannokai and Bouquet dams and suggested that G increases as
33
the 2/3 power of the depth from the crest in earthfill dams.
Figure 2.24: The inhomogeneous dam system studied by Dakoulas and Gaze-
tas (1985).
Dakoulas and Gazetas (1985) investigated the effect of inhomogeneity

of the dam materials on the dynamic response of dams. The shear beam
method (see Section 3.4) was employed and the system under study is shown
in Figure 2.24. The shear stiffness G varies with the depth according to
Equation 2.16
G(z/H) = Gb · (z/H)m (2.16)
where z and H are the vertical distance from the crest of the dam and
the height of the dam respectively as shown in Figure 2.24, Gb is the shear
modulus at the base of the dam and m is exponent of the stiffness which
dictates its variation with depth.
Figure 2.25 shows the displacement modal shapes for different values of
the inhomogeneity exponent parameter m (Equation 2.16). The stiffness is
varied from constant (m=0) to linear (m=1) with depth, i.e. from zero to
large inhomogeneity. It is shown that inhomogeneity has a significant effect
on the displacement modal shapes and results in a concentration of larger
displacements at the crest of the dam.
Figure 2.26 shows the ratio of the nth natural period, Tn (m) (correspond-
ing to the nth vibration mode) of an inhomogeneous dam with inhomogeneity
exponent parameter m, over the period of a homogeneous dam (m=0). The
stiffness is varied from constant (m=0) to linear (m=1) with depth, and re-
34
Figure 2.25: Effect of inhomogeneity on the displacement modal shapes of

dams (Dakoulas and Gazetas, 1985).
Figure 2.26: Effect of inhomogeneity on the natural periods of dams (Dak-

oulas and Gazetas, 1985).
35
sults are shown for two values of the truncation ratio, λ = h/H (Figure 2.24).
It is shown that inhomogeneity affects the periods corresponding to higher
modes of vibration (n = 2, 3, 4, ...) only for an untruncated dam (λ = 0, i.e.
for a complete triangle). However, in any case it does not affect significantly
the fundamental period of the dam T1 which seems to be constant for any
value of inhomogeneity parameter m and truncation ratio λ.
The effect of inhomogeneity on the variation of accelerations in a dam
was presented by Gazetas (1987) who showed that the accelerations in in-
homogeneous dams are higher close to the crest of the dam. That study
considered a 120m high uniform inhomogeneous dam under the Taft earth-
quake record scaled at 0.4g. Figure 2.27 (a) shows the variation of the peak
value of acceleration with the depth of the dam. It is clear that for highly
inhomogeneous (G ∼ z2/3 ) dams, higher accelerations are concentrated at
the top of the dam.
The previous conclusions were drawn from studies that considered linear
elastic behaviour of the dam materials. However, as reported by Gazetas
(1987), when nonlinear material behaviour is introduced, the effects of ma-
terial inhomogeneity are less pronounced as shown in Figure 2.27 (b) (Stara-
Gazetas, 1986). The figure shows the variation of the peak value of acceler-
ation in a highly nonlinear inelastic dam. That study (Stara-Gazetas, 1986)
considered a 120m uniform dam subjected to the 1952 Taft record. From the
latter figure, it is evident that the effects of material inhomogeneity diminish
and this could be due to the dissipation of energy through inelasticity.
Considering the previous studies from the literature, it may be concluded
that inhomogeneity affects the dynamic response of earth dams. This in-
fluence is mainly related to the modal shapes and the vertical variation of
accelerations. However, inhomogeneity has a smaller impact on the higher
natural periods of vibration and minimal impact on the fundamental period
of the dam. Finally, it was reported in the literature that the effects of inho-
mogeneity are less pronounced if nonlinear material behaviour is considered.
36
(a) Linear dam (b) Nonlinear dam
Figure 2.27: Effect of inhomogeneity on the vertical variation of accelerations

in earth dams (Gazetas, 1987).
2.4.4 Flexible dam foundation and canyon effects

Usually earth dams are built in canyons over an alluvial riverbed layer. The
presence of such a foundation layer affects the dynamic response of a dam
according to the relative properties of the two parts of the system, dam and
foundation, a phenomenon called dam-foundation interaction. In the general
context of soil-structure interaction, it is widely believed that the presence
of a soft base increases the fundamental period of a structure thus softening
the system (Wolf, 1985). As far as earth dams are concerned, analytical
and numerical studies have quantified the effect of a foundation layer on the
vibration characteristics of dams.
Chopra and Perumalswami (1971) showed that the response of a dam
founded on an elastic halfspace is damped because of dam-foundation in-
teraction effects. Idriss et al. (1974) suggested that the parameters that
affect dam-foundation interaction include the natural periods and the mate-
rial propeties of the dam and foundation layer and the lateral extent of the
dam. Later, in a study of the Santa Felicia dam in California, Abdel-Ghaffar
and Scott (1981) showed that energy is dissipated through wave radiation in
the dam foundation during vibration of dams. Deformable canyons of simple
shapes, such as semi-cylindrical (Dakoulas, 1993a,b) and semi-elliptical (Dak-
oulas and Hsu, 1993, 1995) were extensively examined analytically and it was
37
found that the amplitude of the dam response is affected by the impedance
between the dam and the canyon.
Shear beam analyses by Dakoulas and Gazetas (1987) investigated the
effect of the presence of a foundation layer on the fundamental period of a
dam. The study considered a uniform distribution of shear modulus G for
both the dam and the foundation layer. Figure 2.28 shows the ratio of the
period of a dam-layer system T f , over the period of a dam founded on rigid
1
bedrock T1 , for different dam length over height, L/H, and foundation-dam
height, h/H ratios.
Figure 2.28: Effect of the geometric characteristics of a homogeneous dam-

foundation system on the system’s fundamental period (Dakoulas and Gaze-
tas, 1987).
A more comprehensive study was carried out by Dakoulas (1990) who

considered different ratios of the stiffnesses of the foundation layer and an
inhomogeneous dam. The stiffness of the dam was assumed to be varying
with depth, as described by Equation 2.16 (Note that now the height of the
dam structure H becomes H1 as shown in Figure 2.29).
The system under study consisted of a triangular dam founded on a layer
of alluvium as shown in Figure 2.29. The results of this study were expressed
as the ratio of the period of the dam-foundation system over the period of a
dam founded on rock and they are shown graphically in Figure 2.30.
From both Figures 2.28 and 2.30, it may be observed that the funda-
mental period of a dam-foundation system is increasing as the height of the
foundation layer is increasing (as compared to the height of the dam). In
38
(a) (b)
Figure 2.29: The dam-foundation system considered by Dakoulas (1990)
addition, the effect of the presence of the layer is more pronounced for longer
dams and for softer foundation layers (as compared to the stiffness of the
dam).
The flexibility of the canyon rock was also examined by Papalou and
Bielak (2001, 2004) when studying the behaviour of La Villita dam in Mexico.
The dam was modelled as a shear beam, whereas the surrounding canyon was
discretised using finite elements. It was shown that the acceleration response
of the dam reduces as the stiffness of the rock canyon reduces because of
waves radiating away. Moreover, accelerations reduce further if nonlinear
behaviour is considered. Figure 2.31 shows the acceleration time histories
predicted at the crest of the dam for different shear wave velocities of the
rock, Vr , for (a) a linear elastic and (b) a nonlinear system.
It is therefore noticed that the presence of an underlying foundation allu-
vial layer tends to soften the dam-foundation system and hence increase its
fundamental period. In addition, the presence of a dam foundation has an
effect on the dam accelerations.
2.4.5 Dam-reservoir interaction

In static conditions the effect of the reservoir mainly involves the application
of a hydrostatic external pressure and the saturation of the upstream part
of earth dams. This saturation results in a reduction of the effective stresses
and hence the factor of safety, Fs , of the upstream slope changes. Therefore,
if the reduction of the Fs is not significant, there is no major risk for the
39
(a)
(b)
(c)
Figure 2.30: Effect of geometric and material characteristics on the fun-

damental period of an inhomogeneous dam-foundation system (Dakoulas,
1990). Te11 refers to the fundamental period of a dam-foundation system, T11
refers to the fundamental period of a dam founded on rigid rock, and m is
the exponent of inhomogeneity (Equation 2.16).
40
(a) (b)
Figure 2.31: Accelerations predicted at the crest of La Villita dam for dif-
ferent shear wave velocities, Vr of the rock canyon for (a) a linear elastic
(Papalou and Bielak, 2001) and (b) a nonlinear system (Papalou and Bielak,
2004).
safety of the dam.

This is also true for dynamic conditions; however, the presence of the
reservoir has additional effects on the dynamic response of the dam, because
of the hydrodynamic pressures induced by the vibration of the reservoir and
dam-reservoir interaction phenomena. The evaluation of the reservoir hy-
drodynamic pressures acting on the upstream face of a dam is discussed in
Section 2.5.
Dam-reservoir interaction was first studied by Chopra (1967b, 1968) who
adopted the equations describing the continuity of the reservoir fluid and
considered concrete gravity dams deforming in the fundamental mode under
harmonic loading of constant circular frequency, ω. The results were ex-
pressed graphically in terms of the amplification of accelerations at the dam
crest with respect to the frequencies of the harmonic load and of the funda-
mental frequencies of the elastic dam and the reservoir. Figure 2.32 shows
the amplification of accelerations at the dam crest, |F |, against the ratio
ω/ωd (the circular frequency of the load, ω, to the natural circular frequency
of the dam, ωd ) and the ratio Ωr = ωr /ωd (the natural circular frequency of
the reservoir, ωr to the natural circular frequency of the dam, ωd ).
41
NO RESERVOIR
0 2 4 6 ω
ωd
20 0.4
1
10
1.8
ωr
ωd
(a)
2
0 2 4 6 ω
20 ωd
10
4
5
ωr
ωd
(b)
Figure 2.32: Frequency response of the accelerations at the dam crest for
different load, dam and reservoir circular frequency ratios as presented by
Chopra (1968): (a) for 0 ≤ ωr /ωd ≤ 1.8 and (b) for 2 ≤ ωr /ωd ≤ 5.
42
Figure 2.32 shows that the presence of the reservoir affects significantly
the dynamic response of the dam. More specifically, the amplification of
the earthquake motion in the dam depends not only on the frequency of the
loading but also on the ratio of the natural circular frequency of the reservoir
over the natural circular frequency of the dam.
The effects of dam-reservoir interaction on the dynamic response of dams
were evaluated by Hall and Chopra (1980) with the finite element method
assuming elastic dam response and considering the compressibility of water
(Hall and Chopra, 1982c) (Figure 2.33). It was shown that the presence
of the reservoir increases the fundamental period, T of the dam-reservoir
system (i.e. it becomes more flexible). This is because the mass of the water
is added to the mass, m, of the dam without though adding any flexural or
shear stiffness.
(a) Concrete gravity dam
(b) Earth dam
Figure 2.33: Geometry of the dam-reservoir systems considered by (a) Hall

and Chopra (1982b) and (b) Hall and Chopra (1982a)
43
|F|
ω/ω
r
ωd
(a) Concrete gravity dam
|F|
ω/ωr
ωd
(b) Earth dam
Figure 2.34: Frequency spectrum of the accelerations at the dam crest for
(a) a concrete gravity dam (Hall and Chopra, 1982b) and (b) an earth dam
(Hall and Chopra, 1982a)
44
This is more pronounced in concrete dams (Hall and Chopra, 1982b) (and
more specifically thin concrete arch dams) which have a smaller volume (and
hence mass) than earthfill dams and usually a nearly vertical upstream face.
In contrast, the effects of reservoir-dam interaction on the dynamic response
of earth dams (mainly the fundamental period of the dam) are believed to
be less important (Hall and Chopra, 1982a). This is due to the fact that
(a) the volume (and hence the mass) of earth dams is very large compared
to the added mass from the water (see also Section 2.5) and (b) the sloped
upstream face of earthfill dams experiences smaller hydrodynamic pressures
from the reservoir than a vertical face.
Figure 2.34 (a) shows the frequency spectrum, (|F |), of the amplification
of accelerations at the dam crest for concrete dams (including damping, ξ =
5%), whereas Figure 2.34 (b) shows the corresponding spectrum for earth
dams (including damping, ξ = 10%). The horizontal axis shows the ratio
of the frequency of a harmonic load, ω over the fundamental frequency of
the reservoir, ωr . On the same axis, the values of the natural frequency
of vibration of the dam, ωd are also included. Both figures show that the
peaks in the acceleration response (which correspond to the natural modes of
vibration) occur at smaller frequencies for the case considering the reservoir
water (solid line) than the case ignoring the reservoir (dotted line). This
means that the natural frequencies of the system decrease and therefore the
system softens. Furthermore, it may be observed that the softening effect of
reservoir-dam interaction is larger in concrete dams than earth dams. More
specifically, the reduction in the fundamental frequency of the dam-reservoir
system is about 30% for concrete dams, whereas for earth dams it seems to
be less than 5%.
However, as far as the amplification is concerned, there is no clear trend
shown in these results. In Figure 2.34 (a) for concrete dams, the amplification
for the case considering the reservoir water is larger than the corresponding
amplification ignoring the water for the first natural mode, whereas it is
smaller for the higher modes of vibration. Similar observations hold for the
case of the earth dam (Figure 2.34 (b)).
It should be noted that the calculated first four natural circular frequen-
45
cies of the concrete dam (Hall and Chopra, 1982b) (ωd = ω1 , ω2 , ω3 and ω4 )
and the corresponding first and second calculated circular frequencies for the
earth dam (Hall and Chopra, 1982a) (ωd = ω1 and ω2 ) are also shown on
Figures 2.34 (a) & (b). It may be noticed that in the cases studied by Hall
and Chopra (1982b,a), the natural circular frequencies of the dam were not
at the proximity of the natural circular frequencies of the reservoir, ωr . It
would be therefore interesting to investigate what would be the response of
the dam in such cases where resonance is expected to occur.
Examples of dams for which the measured natural period with full reser-
voir was found to be larger than for partly full reservoir include the Pine Flat
dam in California, U. S. (Chopra and Gupta, 1978) and the Morrow Point
dam in Colorado, U. S. (Duron and Hall, 1988; Fenves et al., 1992; Tan and
Chopra, 1996). The Pine Flat dam is a 122m high concrete gravity dam
whose fundamental natural period was measured by forced vibration tests
to be 0.288 sec and 0.306 sec with the reservoir depth at 95m and 105m re-
spectively. The Morrow Point dam is a 142m high concrete arch dam whose
fundamental period was determined similarly by forced vibrations tests to
be 0.268 sec with the reservoir partially full and 0.303 sec with a full reser-
voir (Chopra, 1995). More information regarding reservoir-dam interaction
effects in embankment dams may be found in the work of Zhao et al. (1993)
and Guan and Moore (1997).
2.5 Reservoir hydrodynamic pressures

The static loading conditions imposed by the upstream reservoir include the
loading from the water impoundment over a period of time and the steady
seepage flow through the dam (see Vaughan (1994)). If there is no upstream
membrane, the effect of the water is mainly the wetting of the dam mate-
rial (and the associated reduction in the effective stress) and an additional
external load from the reservoir.
For dynamic conditions however, the water cannot be simply considered
as an additional external load on the dam slope. In fact, the presence of the
reservoir changes the modes of vibration of the dam, and the dam itself affects
46
the motion of the water. This is called dynamic fluid-structure interaction

and was discussed in Section 2.4.
The reservoir water imposes a dynamic load on the upstream dam face
which in some cases may significantly increase the value of the hydrostatic
pressure. Moreover, the water surface fluctuates because of waves generated
in the reservoir and this induces loads at higher parts of the dam, thus
reducing (temporarily) the freeboard.
Although shear waves (S-waves) cannot propagate through the reservoir
water, pressure waves (P-waves) can. Therefore, for P-waves the upstream
face of the dam is not a free end, but it is an interface of two materials of
different stiffness. This may lead to P-wave reflections and wave-trapping in
the dam body.
The early attempts to quantify the hydrodynamic pressures on dams con-
centrated on concrete (stiff with a vertical upstream face) dams and were
based on several simplifying assumptions. Those assumptions were essen-
tial for the development of analytical solutions and they were related to the
geometry of the dam, properties of the water and the characteristics of the
ground motion. These simple analytical solutions are still used today, both in
the design of dams and as a reference for comparison with numerical results.
2.5.1 Vertical upstream face

The first study on the hydrodynamic pressures on dams is the fundamental
work of Westergaard (1933) who derived an expression for the pressures on
a concrete (infinitely stiff with a vertical upstream face) dam. This work
was based on the theory of the dilation of elastic solids and considered the
motion of the water in the reservoir under the action of a harmonic horizontal
excitation. The geometry of the system is shown in Figure 2.35, in which the
dam is considered to be stiff (undeformable) with its upstream face vertical.
The horizontal and vertical movements, u and v of the water particles are
given by Equations 2.17 and 2.18 respectively.
∞
ac T 2 2πt X 1 −qn nπy
u=− 3
cos( ) e sin( ) (2.17)
π T n=1,3,5,... n 2h
47
h reservoir p(y) dam
po
Figure 2.35: The dam-reservoir system considered by Westergaard (1933)
∞
ac T 2 2πt X 1 −qn nπy
v= 3
cos( ) e cos( ) (2.18)
π T n=1,3,5,... ncn 2h
where,
s
16γw h2
cn = 1− (2.19)
n2 gKT 2
nπcn x
qn = (2.20)
2h
and ac and T are the coefficient and period of the horizontal vibrations,
γw is the unit weight of the water, x is the horizontal distance from the
upstream face of the dam, y is the vertical distance from the top of the
reservoir, h is the height of the dam (Figure 2.35), K is the compressibility
of the water and t is the time. Please note that according to Westergaard
(1933), the acceleration coefficient ac is a dimensionless parameter, equal to
the peak value of acceleration, ao , divided by the acceleration of gravity, g
(=9.81 m/s2 ).
The hydrodynamic pressure distribution on the dam face is given by
Equation 2.21 and its maximum value for y = h is given by Equation 2.22.
∞
8ac γw h X 1 nπy
p(y) = 2
sin( ) (2.21)
π n=1,3,5,... ncn 2h
48
∞ n−1
8ac γw h X (−1) 2
pmax = (2.22)
π 2 n=1,3,5,... ncn
For incompressible fluid, i.e. the bulk modulus of water Kw → +∞,
therefore cn → 1− and hence the maximum pressure is given by:
8ac γw h 1 1
pmax = 2
(1 − 2 + 2 − ...) = 0.743ac γw h (2.23)
π 3 5
The corresponding total hydrodynamic force over the upstream face of the
dam (by integrating the hydrodynamic pressure over the vertical distance)
for incompressible reservoir water is given by Equation 2.24.
16ac γw h2 1 1
Fmax = 2
(1 + 2 + 2 − ...) = 0.543ac γw h2 (2.24)
π 3 5
This hydrodynamic pressure is considered to be constant and added to
the pseudostatic calculations for the stability of the dam. Recognising the
inertial contribution of this pressure, Westergaard (1933) suggested that the
effect of the hydrodynamic pressures could also be simulated if an additional
mass of water was added to the mass of the dam, instead of adding an external
hydrodynamic stress. This is the so-called added mass concept. The width,
b of this water mass added to the dam follows the distribution given in
Equation 2.25. Additionally, if the water mass was replaced by an equivalent
concrete mass, the width, b0 of this mass would be given by Equation 2.26.
Both distributions are shown in Figure 2.36 (Westergaard, 1933).
7q
b= hy (2.25)
8
q
b0 = 0.38 hy (2.26)
In a complementary discussion of the paper of Westergaard (1933), von

Kármán (1933) used a ‘momentum-balance’ method to estimate the hydro-
dynamic pressures on a vertical rigid dam due to an incompressible reservoir.
His solution satisfied the continuity condition of fluid flow and the equilib-
rium between the water pressures and the mass forces and assumed that the
49
b'
h reservoir dam
0.38 h
7/8 h
Figure 2.36: Distribution of the width of water mass, b and concrete, b0

considered to move with water (Westergaard, 1933).
distribution of the hyrodynamic pressures is a quadrant of an ellipse. The

resulting approximate expressions for the maximum hydrodynamic pressure
at the bottom of the dam and the total hydrodynamic force acting on the
upstream face of the dam are given by Equations 2.27 and 2.28.
pmax = 0.707ac γw h (2.27)
Fmax = 0.555ac γw h2 (2.28)
When compared to the exact solutions of Westergaard (1933) (Equations

2.23 and 2.24), the relations of von Kármán (1933) slightly underestimate
the peak hydrodynamic pressure, but overerestimate the total hydrodynamic
force. However, the differences are small.
Newmark and Rosenblueth (1971) followed an approach similar to West-
ergaard (1933), but they used equations from fluid dynamics to describe
the motion of the water and adopted the following assumptions: the water
behaves as an inviscid fluid, no air is trapped by the water, only small dis-
placements are considered and the flow has small Reynolds number, Re (i.e.
no turbulence).
Considering the above-mentioned assumptions they employed the govern-
ing equation of the motion of the fluid (in terms of the velocity potential, φ)
50
as given by Equation 2.29.
∂ 2φ
Vp2 ∇2 φ = (2.29)
∂t2
The water pressure at any point p(x1 , x2 , t) (x1 is the vertical distance
from the bottom of the reservoir and x2 is the horizontal distance from the
upstream face of the dam) is given by Equation 2.30.
∞
(−1)n+1 −µn x2 µ 2 x1

p(x1 , x2 , t) = −2ac ω 2 ρh sin ωt
X
exp cos (2.30)
n=1 µ2n h h
where µn = (2n−1)π
2
, ρ is the density of water, ω is the circular frequency
of the excitation and h is the height of the reservoir.
The maximum value of the water pressure on the dam face (x2 = 0)
occurs at the bottom of the reservoir (x1 = 0) and it is given by Equation
2.31.
ẍo
pmax = 0.743ẍc ρh = 0.743 γw h (2.31)
g
where, ẍo is the maximum value of the horizontal acceleration.
Considering the three above-mentioned approaches (Westergaard, 1933;
von Kármán, 1933; Newmark and Rosenblueth, 1971) to estimate the hy-
drodynamic pressures on vertical rigid dams due to incompressible reservoir
under harmonic loading, it may be concluded that the maximum hydrody-
namic pressure occurs at the bottom of the reservoir and is given by Equation
2.32.
pmax = 0.743ac γw h (2.32)
2.5.2 Inclined upstream face

The first study of hydrodynamic pressures on dams with inclined upstream
face was the work of Zangar (Zangar, 1952; Zangar and Haefeli, 1952; Zangar,
1953). This work considered the water as incompressible and dealt with equa-
51
tions of fluid mechanics. Zangar (1952) noticed the relation of the resulting
equations with the equations describing electricity and then performed ex-
periments using an electric analog. He presented results obtained from both
the experiment and an analytical expression. The hydrodynamic pressures,
p on a dam face are given by Equation 2.33.
p(y) = C(y)ac γw h (2.33)
where ac is the horizontal acceleration coefficient and it is equal to the

maximum value of acceleration ao divided by the acceleration of gravity, g,
γw is the unit weight of water, h is the water depth and C is the pressure
coefficient obtained from Figure 2.37 (Zangar, 1952). The solid lines show the
experimental results (from the electric analog) whereas the dashed lines show
the approximate solution proposed by Zangar (1952) and given by Equations
2.34 and 2.35.
" s #
1 y y y y

C(y) = Cm 2− + 2− (2.34)
2 h h h h
θo
!
C(y = 0) = 0.743 1 − o (2.35)
90
where, y is the vertical distance from the free surface, h is the height of
the reservoir and Cm is the maximum value of C. The variation of C and
Cm with the value of the slope angle, θ, are shown in Figure 2.38.
Figure 2.37 shows that the distribution of the water pressure along the
vertical dam direction, normalised with respect to its magnitude at the base
of the dam has a similar shape to the one with a vertical upstream face (θ = 0
in Figure 2.37). However, the maximum value of the hydrodynamic pressure
is not always at the bottom of the reservoir, but depending on the inclination
of the upstream face, it is located within the bottom fifth of the dam.
Following the analytical and experimental work of Zangar (1952), an an-
alytical approach was followed by Chwang and Housner (1978) and Chwang
(1978) in order to calculate the hydrodynamic pressures on dams with in-
clined upstream face. Chwang and Housner (1978) adopted the momentum-
52
Figure 2.37: Hydrodynamic pressures on a dam face - values of the pressure

coefficient C. The solid line represents experimental results, whereas the
dashed line represents the empirical equation fitted by Zangar (1952)
Figure 2.38: Maximum and base coefficients of hydrodynamic pressures on

an inclined dam face C & Cm (Zangar, 1952)
53
s
h y
x θ=απ
Figure 2.39: The reservoir-dam system considered by Chwang and Housner

(1978) and Chwang (1978).
Figure 2.40: Values of the pressure coefficient C(y) for a dam with an inclined
upstream face. The dashed line presents the results from the work of Chwang
and Housner (1978) whereas the solid line corresponds to the work of Chwang
(1978).
54
balance method (von Kármán, 1933) whereas Chwang (1978) used two-
dimensional potential-flow theory, which he considered to give an exact solu-
tion. In both approaches, the dam was considered to be rigid and subjected
to horizontal forces. The geometry of the system used is shown in Figure
2.39.
For both approaches, the hydrodynamic pressures may be simply obtained
using Equation 2.36.
p(y) = Cp ao ρh = Cp ac γw h (2.36)
where, γw and ρ are the unit weight and mass density of water respectively,
ao is the maximum value of the horizontal harmonic acceleration, ac is the
acceleration coefficient (=ao /g), h is the height of the reservoir and Cp is the
pressure coefficient, which can be determined from Figure 2.40. The results
shown in dashed lines come from the work based on the momentum-balance
method (Chwang and Housner, 1978), whereas the results in solid lines come
from the work based on the potential-flow theory (Chwang, 1978).
From the previous studies, it may be concluded that the hydrodynamic
pressures on a dam with an inclined upstream face are generally smaller than
those on a dam with a vertical upstream face. Moreover, the maximum value
of the hydrodynamic pressure does not occur at the base of the dam (which
is the case for a dam with a vertical upstream face) but slightly higher.
2.5.3 Inclined reservoir base

Liu (1986) approached the problem of dam-reservoir interaction analytically
for a stiff dam with an inclined reservoir base. The procedure followed was
based on a two-dimensional potential flow theory and it was an extention of
the work of Chwang (1978). The system considered is shown in Figure 2.41.
The resulting equation for the hydrodynamic pressures on the upstream
face of the dam is given by Equation 2.37.
hρao 1
p(y) =
sin aπ sin βπ Γ(a)Γ(β)Γ(1 − a − β)
55
s
h y
x βπ θ=απ
Figure 2.41: The dam-reservoir system considered by Liu (1986)
 a 
y
 4[sin(a + γ)π cos βπ + sin(β−)π cos aπ] Z π/2
h (tan 2θ)2β θ 
dθ
 sin(a + β)π 0 (tan2 θ + hy )a+β sin θ cos θ 
" #
sin(a + γ)π cos(a + β)π + sin(β − γ)π h
−ρao − s(y) (2.37)
sin(a + β)π sin aπ
where, aπ, βπ and γπ are the angles of the upstream dam slope, reservoir
bottom and applied acceleration respectively, h is the height of the reservoir,
ρ is the mass density of the water, s(y) is the distance of any point on the
dam face from the bottom corner of the dam and Γ(·) is the Gamma function.
As with the work of Chwang and Housner (1978) and Chwang (1978),
the hydrodynamic pressures can be more practically obtained using Equation
2.38.
p(y) = Cp ao ρh = Cp ac γw h (2.38)
where, γw and ρ are the unit weight and mass density of water respec-
tively, ao is the maximum value of the horizontal harmonic acceleration, ac
is the acceleration coefficient (=ao /g), h is the height of the reservoir and
Cp is the pressure coefficient, which can be determined from Figures 2.42 &
2.43. In Figure 2.42, the results shown in dashed lines come from the exper-
imental work of Zangar (1952) whereas the results in solid lines come from
the theoretical work based on the potential-flow theory of Liu (1986).
Considering the study of Liu (1986), it may be concluded that the hydro-
dynamic pressures on a dam with an inclined reservoir base are even smaller
than the corresponding pressures on a dam with a horizontal reservoir base.
The maximum value of the hydrodynamic pressure does not occur at the base
of the dam, but slightly higher. The value of the hydrodynamic pressure at
56
(a) βπ = 0o (b) βπ = 15o
Figure 2.42: Values of the pressure coefficient Cp (=P/ρao h) for reservoir

bottom angles (a) 0o and (b) 15o (Liu, 1986) (solid line). The dashed line in
(a) corresponds to the work of Zangar (1952) and it is provided for compar-
ison.
(a) βπ = 30o (b) βπ = 45o
Figure 2.43: Values of the pressure coefficient Cp (=P/ρao h) for reservoir

bottom angles (a) 30o and (b) 45o (Liu, 1986).
57
the base of the dam can be extremely small, depending on the slope angle of
the reservoir base and the upstream dam face.
2.5.4 Compressibility of water

All the previous studies considered the water as being an incompressible
material. They all concentrate on the distribution and magnitude of the
hydrodynamic pressures and they agree on the maximum value of these on
the vertical upstream face of the dam at the bottom of the reservoir, which
was firstly obtained by Westergaard (1933) and is given by Equation 2.23.
Chopra (1967a) examined the hydrodynamic pressures resulting from a
compressible reservoir and concluded that the value of the maximum pressure
(at the base of the dam) depends on the ratio of the loading frequency to the
reservoir’s fundamental frequency of vibration. The estimated value of the
fundamental frequency of the reservoir (for a horizontal and rigid reservoir
base) is given by Equation 2.39.
s
1 K Vp
f1 = = (2.39)
4h ρ 4h
where, f1 is the fundamental frequency of the reservoir, h is the height
of the reservoir and K, ρ and Vp are the bulk modulus, mass density and
p-wave velocity of the reservoir water respectively.
Chopra (1967a) presented his results in the frequency domain, as the ratio
of the hydrodynamic force, Fdyn , multiplied by the gravitational acceleration,
g (=9.81 m/s2 ) over the hydrostatic force, Fst , against the ratio, Ω = ω/ω1
of the loading circular frequency, ω, to the first circular frequency, ω1 , of the
resrvoir. The relation between the ratio of the water pressures and the ratio
of the frequencies is given by Equation 2.40 and shown graphically in Figure
2.44.
Fdyn (Ω) 32 +∞
X 1
= 3 q (2.40)
Fst π g n=1 (2n − 1)2 (2n − 1)2 − Ω2
Figure 2.44 shows that the hydrodynamic pressures can be significantly
58
Frequency Response for Total Hydrodynamic force

4
Chopra, 1967
3.5
2.5
g Fdyn/Fst
1.5
0.5
0
0 1 2 3 4 5 6
Frequency Ratio, Ω = ω/ω1
Figure 2.44: Frequency response of the total hydrodynamic force induced by

a compressible reservoir due to horizontal harmonic loading. After Chopra
(1967a)
high for loading frequencies in the vicinity of the natural frequencies of the
reservoir, and specifically close to the first natural frequency. It should be
noted that if the water is incompressible, then the fundamental frequency of
the reservoir, ω1 becomes infinite and hence the ratio Ω = ω/ω1 → 0+ for any
loading frequency ω. The value of the hydrodynamic force corresponding to
this ratio based on Chopra (1967a) is given by Equation 2.41.
Fdyn (Ω = 0) 32 +∞
X 1
= 3 = 0.1106 (2.41)
Fst π g n=1 (2n − 1)3
The corresponding value for incompressible water (i.e. considering Equa-
tion 2.23) suggested by Westergaard (1933) is given by Equation 2.42.
0.543
Fdyn (Ω = 0) g
γw H 2
= 1 = 0.1107 (2.42)
Fst γ H2
2 w
Comparing Equations 2.41 and 2.42, it is concluded that indeed the so-
59
lutions of Chopra (1967a) and Westergaard (1933) agree when the water is
considered as incompressible. The work of Chopra (1967a) proved that the
general solution proposed by Westergaard (1933) (considering water com-
pressibility, i.e. Equation 2.22) is valid only for the cases where the excita-
tion frequency is smaller than the fundamental frequency, f1 (Equation 2.39)
of the reservoir. Finally, it proved that the added mass concept of Wester-
gaard (1933) is valid only when the water is considered to be incompressible,
(Kw → +∞, and therefore ω/ω1 → 0+ ).
2.5.5 Seismic excitation

The above-mentioned approaches considered the hydrodynamic pressures due
to a monochromatic harmonic excitation (i.e. a single frequency). Chopra
(1967a) extended the work of Westergaard (1933) to account for seismic
(multi-frequency) excitation. The response of the reservoir behind a ver-
tical rigid dam under a unit impulse was first examined. Then, by using
the convolution (Duhamel’s) integral and the principle of superposition the
hydrodynamic pressures resulting from a random excitation were obtained.
The computed hydrodynamic pressures for horizontal and vertical ground
motions are given by Equations 2.43 and 2.44 respectively.
x 4γw Vp +∞
X (−1)n−1 Z t
p (y, t) = cos (λn y) üg (τ )Jo [λn Vp (t − τ )]dτ (2.43)
πg n=1 2n − 1 0
y 4γw Vp +∞
X 1 Z t
p (y, t) = cos (λn y) v̈g (τ )sin[λn Vp (t − τ )]dτ (2.44)
πg n=1 2n − 1 0
where, Vp is the p-wave velocity in the water, λn = (2n − 1)π/2H is the

th
n wave length in the water, H is the height of the reservoir, Jo (·) is a Bessel
function of the first kind of order zero and üg and v̈g are the horizontal and
vertical ground acceleration respectively.
60
2.5.6 Flexible dam

All the previously mentioned approaches dealt with stiff undeformable dams
which do not interact with the reservoir during dynamic loading. While this
can be true for certain cases, such as large concrete gravity dams, it is not
the case for tall and slender arch dams and for earthfill dams which are less
stiff than concrete dams. The first approach to investigate reservoir-dam
interaction for flexible dams was the work of Bustamante et al. (1963). That
study set the ground for analytical studies of dam-reservoir interaction and
considered a dam which was assumed to deflect according to a prescribed
shape. The hydrodynamic forces associated with a prescribed deflection are
then assumed to be equal to the force required to cause that deflection.
Lee and Tsai (1991) extended the previous work and considered a flexible
dam structure with vertical upstream face which was subjected to ground
motion and it was allowed to deform in flexure. The flexible dam shown in
Figure 2.45 is assumed to behave as an elastic Euler-Bernoulli beam and the
corresponding equation of motion is given by Equation 2.45.
Structure
H Fluid
y
x
Figure 2.45: The fluid-structure system considered by Lee and Tsai (1991)
∂2 ∂ 2u ∂ 2u
!
EI 2 + m 2 = −müg − p(x = 0, y, t) (2.45)
∂y 2 ∂y ∂t
where, E is the Young’s modulus, I is the moment of inertia of the cross-
section about the axis of bending, m is the mass per unit length of the
structure, y is the vertical distance from the base of the dam, u is the hor-
izontal displacement of the dam structure, üg is the ground acceleration in
the upstream-downstream direction and p is the hydrodynamic pressure.
The displacement response of the dam structure, u, including the hydro-
dynamic effects can be expressed as a linear combination of modal shapes,
61
φn (y) and generalised coordinates Yn (t) and is given by Equation 2.46.
+∞
X
u(y, t) = φn (y)Yn (t) (2.46)
n=1
The resulting hydrodynamic pressures on the upstream face of the dam

are given by Equation 2.47.
4ρVp +∞
X (−1)k+1 Z t
p(y, t) = cos λk y üg (τ )Jo [λk Vp (t − τ )]dτ
π k=1 2k − 1 0
2ρVp +∞
X +∞
X Z H Z t
+ cos λk y φn (y) cos λk ydy Ÿn (τ )Jo [λk Vp (t − τ )]dτ (2.47)
H k=1 n=1 0 o
where, Vp is the p-wave velocity in the water, H is the height of the dam
and reservoir, üg is the ground acceleration in the upstream-downstream
direction, φn (y) is the nth vibration modal shape of the dam, Jo (·) is a Bessel
function of the first kind and zero order and λk is given by Equation 2.48.
(2k − 1)π
λk = (2.48)
2H
Generally, the flexibility of the dam is assumed to result in smaller val-
ues of hydrodynamic pressures, as part of the energy is absorbed by the
deformation of the dam structure (see also the effect of the reservoir on the
dam accelerations in Section 2.4). Therefore the introduction of flexibility
seems to be beneficial as it reduces the hydrodynamic loads. Centrifuge
studies comparing small-scale models of a stiff concrete dam and a flexible
aluminium dam (Saleh and Madabhushi, 2010) have confirmed this assump-
tion, as the measured hydrodynamic pressures were found to be smaller for
the flexible case.
However, it is the Author’s opinion that the hydrodynamic pressures on
flexible dams have not been investigated (analytically, numerically or ex-
perimentally) exhaustively. Reservoir-dam interaction is considered to be
beneficial, but cases in which this interaction might be detrimental have not
been clearly identified. These could include, for example, the case of reso-
62
nance between dam and reservoir. Therefore, further studies are required in
order to confirm these assumptions or identify any limitations.
2.5.7 Comments
This section described several methods for calculating the hydrodynamic
pressures on dams resulting from the upstream reservoir. The hydrodynamic
pressures in the simplest case, of a rigid dam with vertical upstream face and
an incompressible reservoir, reach a maximum at the base of the dam. The
magnitude of the pressures gets smaller for inclined dams and in those cases
the maximum value occurs higher than the base of the dam.
When the compressibility of the water is considered, the magnitude of the
total hydrodynamic force on the dam depends on the ratio of the frequency
of the load to the natural frequency of the reservoir. Furthermore, if the flex-
ibility of the dam structure is considered, the hydrodynamic pressures tend
to be smaller in general because of energy absorbed by the dam. However,
the latter case still has to be further investigated. Information about the
numerical modelling of hydrodynamic pressures follows in Chapter 4.
2.6 Effects of earthquakes on dams

The response of earth dams (i.e. the entire dam-reservoir-foundation system)
during earthquakes is mainly governed by (a) the dynamic-cyclic behaviour
of the ground, including the excess pore pressure generation within the soil
(Section 2.3), (b) the vibration of the dam body (Section 2.4) and the up-
stream hydrodynamic pressures from the reservoir (for evaluation of these
pressures, see Section 2.5). Generally, the effects of earthquakes on dams are
mainly related to excessive displacements due to instability (i.e. cases with-
out any liquefaction occuring in the dam or the foundation soil) and failure
due to liquefaction.
63
2.6.1 Excessive displacements

Earthquake shaking causes additional loads on the dam and increase of pore
water pressures due to cyclic loading. As a result, some localised or even
larger failures may occur. Examples of dams which suffered from excessive
displacements (without any evidence of liquefaction) are the Mahabad dam
(1978 Tabas earthquake) in Iran (Khoei et al., 2004), the Niwaikumine (1973
Hokkaido earthquake) and Makio dams (1984 Japan earthquake) in Japan
(Tani, 2000) and La Villita and El Infiernillo dams (a series of earthquakes
between 1975-1985) in Mexico (Reséndiz et al., 1982).
Field investigations (such as the Earthquake Engineering Research Insti-
tute (EERI), US and the Earthquake Engineering Field Investigation Team
(EEFIT), UK reports) have shown that slope instability is one of the most
common effects of earthquakes on earth dams. Slope failure is most com-
monly observed on the upstream face of the dam which is wet and fully
saturated and therefore expected to have lower effective stresses. Figure 2.46
(Madabhushi and Haigh, 2001) shows initiation of slope failure on the up-
stream face of the Fategadh dam in India during the 2001 Bhuj earthquake.
Figure 2.46: Failure of the upstream slope of the Fategadh dam in India
during the 2001 Bhuj earthquake (Madabhushi and Haigh, 2001).
Crest settlements may also be observed and they usually occur for two
64
reasons. Firstly, a large magnitude slope failure (such as flow failure) on both
faces of the dam may alter the dam’s shape in such a way that it becomes
wider at the base and shorter in height. This kind of failure, firstly described
by Ambraseys (1959b) may be observed in Figure 2.47 (Newmark, 1965),
where successive shocks create a series of slope failures and severly change
the geometry of the dam. Secondly, cyclic loading causes relocation of soil
particles, and in a large intensity earthquake, dynamic compaction may occur
in the absense of slope failures. These crest settlements are important and
care should be taken in order to avoid significant loss of freeboard (which
directly affects functionality and safety of the dam) and differential crest
settlements. Both of these events become even more crucial in cases where
a road or other facilites are built on top the dam crest.
Figure 2.47: Major deformation patterns in an earth dam after succesive

earthquake cycles (Ambraseys, 1959b; Newmark, 1965).
The development of cracks is unavoidable in cases of strong ground mo-

tion, as the soil is a limited tension material. In fact, the extremely small
tensile strength of soil comes from the soil structure (bonding and fabric),
and therefore the critical question is what level of earthquake can cause this
breakage of soil structure, and at which locations. Longitudinal cracks oc-
cur near the crest due to shear sliding deformations and large tensile strains
during lateral oscillations. Transverse cracks form near the abutments and
65
the central crest core because of tensile strains from longitudinal oscillations
or by different lateral response of adjacent material zones. Cracking is im-
portant as it may provide ways for leakage through the dam, erosion of the
internal dam material and finally local piping failures. Figure 2.48 shows an
example of cracking in dams during seismic events: wide and deep cracks on
the upstream face of Fategadh dam in India during the 2001 Bhuj earthquake
(Madabhushi and Haigh, 2001).
Figure 2.48: Cracks on Fategadh earth dam in India during the 2001 Bhuj
earthquake (Madabhushi and Haigh, 2001).
In the 19th Rankine Lecture, Seed (1979) summarised a number of possible

ways in which an earthquake may cause failure in an earth dam, before out-
lining methods of seismic analysis and suggesting remedial measures. These
effects of earthquakes on dams as suggested by Seed (1979) are listed below:
• Disruption of dam by major fault movement in foundation
• Loss of freeboard due to differential tectonic ground movements
• Slope failures induced by ground motions
• Loss of freeboard due to slope failures or soil compaction
• Sliding of dam on weak foundation materials
• Piping failure through cracks induced by ground motions
• Overtopping of dam due to seiches in reservoir
66
• Overtopping of dam due to slides or rockfalls into reservoir
• Failure of spillway or outlet works
2.6.2 Liquefaction
Liquefaction is another severe effect of earthquakes on dams. The most com-
mon type of liquefaction observed in earth dams is flow liquefaction which
can be extremely dangerous for the stability of the dam, as it causes large
scale failures. Flow liquefaction occurs when the shear stress for static equi-
librium is larger than the shear strength of the liquefied soil (Kramer, 1996).
It is mainly triggered by the development of excess pore water pressures in
contractive saturated cohesionless zones of the dam (such as rockfill).
Figure 2.49: Failure of the Lower San Fernando dam due to liquefaction in the
upstream part of the rockfill (Cambridge University Engineering Department,
2012).
The most famous case of dam failure due to liquefaction is the Lower
San Fernando dam during the 1971 San Fernando earthquake (Figure 2.49).
Figure 2.50 shows a back-analysis of the liquefaction failure by Seed et al.
(1975). The hatched region in the lower part of the upstream rockfill is the
liquefied material, which caused a global collapse of the whole dam structure.
67
Figure 2.50: Failure and reconstructed cross section of the Lower San Fer-
nando dam during the 1971 San Fernando earthquake as presented by Seed
et al. (1975).
Another classical example of a dam that suffered from liquefaction is

the Sheffield dam (1925 Santa Barbara earthquake) (Seed et al., 1969) in
California. More recent examples include the Chang, Shivlaka and Suvi
dams in India during the 2001 Bhuj earthquake (Singh et al., 2005) and
a considerable number of dams during the 2009 Wenchuan earthquake in
China.
2.7 Summary
This chapter summarizes the main information found in the literature re-
garding the seismic behaviour of dams. It includes details about the design
and construction of dams, a description of the seismic behaviour of soil and
the dynamic response of dams, the different patterns of the reservoir hydro-
dynamic pressures and concludes with the effects of earthquakes on dams.
The important information found from the literature are summarised below:
• Earth dams are massive structures which require site-specific design.
• The nonlinear cyclic soil behaviour results in a reduction of the stiffness
68
of the soil and an increase of the damping with the induced shear
strains. Reduced soil stiffness results in longer dam periods of vibration
and increased damping results in smaller values of accelerations in the
dam.
• The dynamic characteristics of earth dams are complicated due to their

three-dimensional geometry and their seismic response results from a
combination of vibrations in different directions.
• Narrow canyons are shown to reduce the periods of vibration of earth

dams due to their stiffening effect.
• The inhomogeneity of the dam materials affects the variation of the

accelerations within the dam body although this influence is less pro-
nounced for highly inelastic materials.
• Flexible dam foundations and surrounding canyons result in a softer

dam response.
• Reservoir-dam interaction increases the fundamental period of the dam

and results in a softer dam-reservoir system.
• The revervoir hydrodynamic pressures on the upstream face of a dam

are larger for stiff dams with a vertical face than for flexible dams with
an inclined face.
• The magnitude of the total reservoir hydrodynamic force on a dam

from a compressible reservoir depends on the relative magnitude of the
frequency of the load and the fundamental freqency of the reservoir.
However, this is not the case for an incompressible reservoir.
• The effects of earthquakes on dams can be separated in two categories:

(a) cases without liquefaction and (b) failure due to liquefaction.
• The cases without liquefaction occuring anywhere in the dam or foun-

dation, mainly include localised or major slope failures, cracking and
crest settlements.
69
• The cases which involve liquefaction can result in catastrophic failures

of a large part or the entire dam.
70
Chapter 3
SEISMIC ANALYSIS OF
DAMS
3.1 Introduction
A number of methods exist and are used today for the evaluation of the
seismic response of dams. Each method carries its own assumptions and
hence has its own limitations. It is therefore essential to be aware of the
limitations of the methods used so that useful analyses are performed.
The evolution of seismic analysis with time is interesting, as it is a valuable
indication of the advancement of science and shows how the main engineering
concerns change with time. The initial consideration of rigid body failure
mechanisms led to the development of simplified methods considering the
sliding mass as an undeformable structure. Field investigations however did
not confirm this assumption and thus more advanced methods considering
the dam as a deformable body were developed (Gazetas, 1987).
The various methods fall into several categories according to the main
objective of the analysis:
(a) assessment of the stability of the dam structure (such as the pseudo-
static method)
(b) calculation of the permanent displacements of the dam body (such as

the sliding block method)
71
(c) evaluation of the dynamic response of the whole dam (such as the shear
beam and finite element methods)
This chapter describes the various methods used in seismic analysis of

earth dams. It consists of three parts: (a) a description of the most important
methods of seismic analysis of dams (the pseudo-static, sliding block, shear
beam and finite element methods), (b) a discussion of the relative merits
of the different methods of analysis and finally (c) a discussion of several
considerations regarding dynamic finite element analysis which is employed
in this thesis.
3.2 Pseudo-static method

The very first methods of analysis were mainly concerned with the stability
of earth dams, considering the embankment as a rigid, non-deformable body.
The main assumption adopted is that the only possible mode of failure is
slope instability, in the form of a sliding mechanism.
3.2.1 Principle of the method

These early methods of analysis were drawn from the traditional methods
of static analysis of slope stability, mainly the Limit Equilibrium and Limit
Analysis techniques. The difference is the explicit introduction of the seismic
load as an additional horizontal and/or vertical inertial driving force acting
in the most critical direction. The seismic load is proportional to the weight
of the failing soil mass.
This is the so-called “pseudo-static” (PS) method of analysis. According
to Gazetas (1987), the first application of this approach to seismic slope
stability problems was attributed to Terzaghi. Figure 3.1 illustrates the PS
method: additional horizontal kH W , and vertical kV W , forces are included
to represent the inertial seismic loads.
The inertial forces are usually expressed as the weight of the soil multi-
plied by a seismic coefficient k. The seismic coefficient is defined as the ratio
of the seismic acceleration to the gravitational acceleration (aH /g). The value
72
kHW
kVW
W
S
N
Figure 3.1: Illustration of the pseudo-static method of analysis in which

additional inertial forces kH W and kV W are included to account for seismic
loading.
of the seismic coefficient k for which the slope fails is called critical seismic
coefficient and is denoted as kc . Equations 3.1 and 3.2 define the horizontal,
FH , and vertical, FV , inertial forces respectively.
aH W
FH = = kH W (3.1)
g
aV W
FV = = kV W (3.2)
g
where, aH and aV are the horizontal and vertical accelerations respectively
and kH and kV are the horizontal and vertical seismic coefficients.
The basic idea of this method is that equilibrium calculations are carried
out in a similar fashion to static slope stability problems, but with the addi-
tion of the earthquake inertial loads. For the example of the horizontal load,
the objective of the analysis is either of the following:
• given a dam geometry and a horizontal load (i.e. a specific earthquake

load, therefore a known aH ), to find the factor of safety, FS
• given a dam geometry, to find the critical value of horizontal accelera-

tion, ay , for which the slopes fails (i.e. FS = 1)
73
3.2.2 Comments on pseudo-static analysis of earth dams

Various different versions of this method were developed by a number of
researchers. The most widely known is perhaps the work of Sarma (1973,
1975) who developed a method which provides the critical acceleration for
which the slope fails. Later work (Sarma, 1979) included solutions for a
multi-wedge failure mechanism on inclined failure surfaces.
The PS method suffers from a number of unrealistic assumptions (Seed,
1973). It does not take into account the change of material properties during
earthquake loading which has been described earlier in Section 2.3.4, but in-
stead it considers the soil as a perfecly plastic material. Morevover, seismic
stability is considered as a static problem, simply with the inclusion of an
additional unidirectional inertial force. Therefore it ignores the dynamic re-
sponse of the dam and any associated resonance with the excitation. Finally,
a fundamental drawback of this method is the assumption that failure occurs
when the FS becomes less than 1 (i.e. aH > ay ). In reality the FS is less
than 1 for a very short duration (often not adequate to cause sliding) and its
value changes rapidly with time.
3.3 Sliding block method

Based on the last drawback of the stability methods of analysis (i.e. Pseudo-
static), obtaining the critical value of acceleration is not sufficient for the
seismic design of dams. Whenever the FS becomes less than 1, generalised
failure does not necessarily occur, but this results in permanent displacements
of the dam slope, as mentioned by Terzaghi (1950). Therefore, it is important
to calculate the permanent displacements induced by the earthquake and this
is the main objective of this class of methods of analysis.

A number of researchers studied this problem aiming to estimate the expected
magnitude of displacements for a given earthquake load. Early attempts in-
clude the work of Seed (1967), Ambraseys and Sarma (1967) and Ambraseys
74
(1960b, 1962), including the PhD research of the latter (Ambraseys, 1959b)
at Imperial College.
The most widely used method from this group is perhaps the “sliding
block analysis” method, proposed by Newmark (1965) in the fifth Rankine
Lecture. This method, known also as the Newmark’s analysis method, sug-
gests that whenever the acceleration becomes larger than the critical value, ac
(for static slope stability, i.e. FS = 1 and may be obtained using the pseudo-
static method of analysis), permanent displacements occur, according to the
difference of the applied and the critical acceleration. Consequently, by inte-
grating the acceleration time history with respect to time, the velocity time
history is obtained (Equation 3.3), and integrating this again, the displace-
ment time history is obtained (Equation 3.4).
Z t
u̇ = (a − ay )(t) dt ∀ a − ay > 0 (3.3)
0
Z tZ t
u= (a − ay )(t) dt dt ∀ a − ay > 0 (3.4)
0 0
Figure 3.2: Newmark’s sliding block method (Wilson and Keefer, 1985)
Figure 3.2 (Wilson and Keefer, 1985) illustrates this Newmark’s Sliding
Block analysis procedure. It is shown that the integrated difference of accel-
eration with respect to time yields the velocity and further, the integrated
75
velocity yields the permanent displacement. It is obvious that the calculated

permanent displacements are influenced by the duration, the amplitude and
the frequency content of the accelerations.
3.3.2 Sliding block analysis of earth dams

The expected permanent displacements due to different periodic wave forms
(rectangular, sinusoidal and triangular motions) were investigated by Sarma
(1975) and Yegian et al. (1991). Furthermore, the expected displacements
due to actual earthquake records were investigated among others by Makdisi
and Seed (1978) and Ambraseys and Menu (1988) who presented relations
(and associated graphs) between the permanent displacements and the ratio
of the yield to the maximum acceleration, ay /amax .
Moreover, Seed and Martin (1966) considering the deformability of the
dam structure, proposed a decoupled approach to calculate the permanent
displacements. Firstly, dynamic analysis of the embankment is carried out
to estimate the accelerations at the level of the slip surface. Consequently,
these accelerations are integrated using the sliding block analysis method
to estimate the permanent displacements. Later, Makdisi and Seed (1978)
accounted for the variation of the accelerations over the whole length of
the slip surface and presented graphs with the variation of the maximum
accelerations with respect to the height of the dam.
A comparison between the decoupled and coupled analysis procedures
was made by Chopra and Zhang (1991) who concluded that the decoupled
approach may be unconservative (predicting smaller displacements) for large
values of the acceleration ratio, ay /amax . Similar conclusions were later drawn
by Rathje and Bray (1999) who also reported that the decoupled analysis can
be significantly unconservative for various values of the ratio of the funda-
mental period of the earth dam structure to the dominant period of the
earthquake and they did not recommend its use.
A more recent and thorough study of the sliding block method was car-
ried out by Bray and Travasarou (2007) who proposed a new method to
estimate the earthquake-induced displacements. In their model, the spec-
76
tral acceleration of the input motion at a degraded period of the potential

sliding mass, Sa (1.5 · Ts ), was identified as the most efficient single ground
motion intensity measure and the authors proposed a relationship to derive
the seismic displacements, D. The latter were expressed as a function of the
yield coefficient of the dam slope, ky , the magnitude of the earthquake, Mw ,
and the spectral acceleration of the input motion at a degraded period of
the dam structure, Sa (1.5 · Ts ). Finally, they reported that their method was
validated through a number of case histories of earth dams and solid-waste
landfills.
The main advantage of the displacement methods of analysis is that an
indication of the expected values of displacements may be obtained. However,
although these methods are actually the next step after the stability methods
of analysis, they still carry the inherent limitations of the PS method. This
is because displacements are assumed to occur only when the earthquake
induced acceleration exceeds the acceleration required for slope instability,
aH > ay .
3.4 Shear beam method

The permanent displacements of the dam’s slope is perhaps one of the most
valuable information. However, the sliding block method described in the
previous section is still a static method and it ignores the dynamic behaviour
of the problem. Advanced methods of analysis take into account the dynamic
response of the dam system and consider both:
• the dynamic behaviour of the material, and
• the dynamic effects of the earthquake
The dynamic behaviour of the soil is briefly described in Section 2.3.4

which considers mainly the degradation of shear stiffness (G) and the in-
crease of damping (ξ) with the shear strain (γ). The dynamic effects of
77
the earthquake are mainly related to seismic wave propagation and hence
vibration mode interaction.
One of the first attempts to consider the dynamic effects of earthquakes is
the Shear beam (SB) (or shear wedge) approach, which is shown in Figure 3.3.
As stated by Gazetas (1987), the origins of this method have been attributed
to Mononobe et al. (1936), who first introduced the 1D shear beam model
for earth dams in the 1930’s.
Hatanaka (1952, 1955) showed that shearing deformations are predomi-
nant compared to the bending deformations and performed 2D shear beam
analyses in rectangular canyons. Ambraseys (1960a,b) extended the shear
beam model to account for a truncated wedge shape, rectangular canyon
and underlying elastic layer. Since then, a number of researchers extended
and used this method in various cases (Gazetas, 1987). Examples include
the work of Abdel-Ghaffar and Koh (1981) and Gazetas (1981a) for 2D lon-
gitudinal vibrations and Gazetas (1981b) for vertical vibrations.
The initial shear beam method (Mononobe et al., 1936; Hatanaka, 1952,
1955; Ambraseys, 1960a) assumes that:
• the dam deforms in simple shear and produces only horizontal displace-
ments (i.e. bending of the dam is ignored)
• the dam is idealised as a wedge rigidly connected to its base
• the shear stresses are uniform across horizontal planes
This method allows the calculation of the vibration characteristics, such

as natural modes of vibration and corresponding natural periods, Tn and
circular frequencies, ωn . Closed form solutions are available for idealized
geometries and material properties (Gazetas, 1982a).
Harmonic horizontal motion of the bedrock produces 1D vertical shear
wave propagation in the overlying uniform triangular dam. The dam deforms
in simple shear and horizontal soil layers (beams) displace horizontally as
shown in Figure 3.3. The dam is rigidly connected to the base (i.e. horizontal
displacement, u(z=H)=0) and its crest is free (i.e. shear stress, τ (z=0)=0).
Equilibrium is satisfied between the shear force (integration of shear stress,
78
z
H τ
dz
cτ dz
τ +cz
2B
Figure 3.3: Shear beam approach for two-dimensional seismic analysis of
dams. After Gazetas (1987).
τ along the shear surface) on both sides of the shear beam (top and bottom)
along with the inertial force.
An example of the relevant theory for an inhomogeneous dam is given
by Dakoulas and Gazetas (1985), who provided solutions for the case of
increasing shear modulus with depth, G(z)=Gb (z/H)m , where Gb =G(z=H)
is the value of the shear modulus at the base of the dam. The resulting
governing differential equation (for a single-material inhomogeneous dam)
refers to one-dimensional wave propagation and it is described by Equation
3.5.
" #
1 ∂ ∂u
ρ(ü + üg ) = G(z)z (3.5)
z ∂z ∂z
where ρ is the density of the material, G(z) the variation of the shear
modulus with depth, z the vertical distance from the crest of the dam and u
is the horizontal displacement, as defined in Figure 3.3.
The boundary conditions are zero displacement at the bottom boundary
[u(z=H) = 0] and zero shear stress at the top boundary [free end, therefore
τ (z=0) = 0]. The horizontal displacements at any point along the vertical
direction are given by Equation 3.6. This is in fact the solution of the wave
equation in two domains, space, z and time, t.
N
X
u(z, t) = ψn (z)In (t) (3.6)
n=1
79
where, ψn (z) describes the mode shape of the nth natural frequency (i.e.
the variation of horizontal displacement in the vertical direction) and is given
by Equation 3.7, whereas In (t) describes the response of a single-degree-of-
freedom (SDOF) structure of the nth mode frequency subjected to the ground
excitation under study (i.e. the variation of displacement in time).
−m/2 " 1−m/2 #
z z

ψn (z) = Jq βn (3.7)
H H
where, βn is the nth root of a period relation, Jq (·) is a Bessel function of
the first kind of order q=m/(2-m), given by Equation 3.8.
∞
(−1)k
q+2k
X x
Jq (x) = (3.8)
k=0 k! Γ(q + k + 1) 2
where, Γ(·) is the gamma function, given by Equation 3.9.
Z ∞
Γ(x) = e−x xn−1 dx (3.9)
0
The amplification of the base motion at the crest of the dam is obtained
from the modulus of the Transfer Function, F (ω) given by Equation 3.10.
q
ωH
2Vs (1+iξ) 1
F (ω) =
ωH

ωH
(3.10)
Γ(q + 1)Jq Vs (1+iξ)
iaz Jq+1 Vs (1+iξ)
√
where, i = −1 is the imaginary unit, az = ρρrsVVsrs is the impedance ratio at
the soil-rock (dam-foundation) interface, ρs and ρr are the mass densities of
the dam soil and foundation rock respectively and Vs and Vsr are the average
shear wave velocities of the dam soil and foundation rock respectively.
Although an attempt is made to capture the dynamic response of dams
and analytical solutions may be obtained, obviously the shear beam method
carries its own limitations which are directly related to its assumptions.
Therefore, its main limitations (of the initial version of the shear beam, such
as Ambraseys (1960a)) are:
• idealised dam geometries
• simple material behaviour
80
• 1D wave propagation
• assumptions regarding stress distribution within the dam
Further developments of the shear beam method followed later which

account for non-linear material behaviour and complicated geometries. These
advances attempted to overcome some of the above-mentioned limitations
and are described in Section 3.4.2.
3.4.2 Numerical shear beam analysis of earth dams

The first numerical approach associated with the seismic analysis of earth
dams was the shear beam method. This approach, presented in the previous
section is a simple method of dynamic analysis of earth dams which can
also provide closed-form solutions (see Section 2.4). However, the associated
closed-form solutions assume linear elastic or visco-elastic material behaviour
and therefore they pose restrictions on the applications of the method. If
material nonlinearity is considered, then the SB equations cannot be solved
analytically, and therefore numerical techniques have to be employed.
One-dimensional elasto-plastic analysis of earth dams was carried out by
Elgamal et al. (1985) who considered a dam as an one-dimensional (1D)
inhomogeneous (in terms of stiffness) hysteretic shear beam subjected to
horizontal base excitation. This method was applied on the case of Santa
Felicia earth dam predicting accelerations, stresses and strains at several
points of the dam. The hysteretic stress-strain behaviour was modelled using
elasto-plastic constitutive equations based on multi-surface kinematic plas-
ticity (Prevost, 1977). The smooth backbone curve was approximated by
linear segments along which the tangent modulus was constant. Therefore,
the degree of accuracy of the solution was dependent on the number of lin-
ear segments adopted. Each segment corresponded to a yield surface in the
three-dimensional stress space, thus defining a collection of nested surfaces.
The proposed non-linear approach was not compared to field data, but it
was compared to results of finite element analysis of the same dam. The
authors commented that this approach can provide a practical and inexpen-
81
sive (in terms of time and computational power) alternative analysis tool as
compared to the more elaborate FE analysis.
The previously-mentioned approach was further extended by Elgamal
et al. (1987b) to consider the longitudinal dimension of the dam, and there-
fore was able to account for the effect of the lateral canyon constraint on the
nonlinear seismic response of earth dams. This approach was applied on the
Santa Felicia dam (Elgamal et al., 1987a) and showed that the analyses tak-
ing into account the longitudinal dimension of the dam yield different results
than the analyses that ignore the longitudal dam direction. Therefore, it
highlighted the importance of modelling the canyon effect for narrower dams
(crest length over height ratio, L/H ≈ 4.7 for Santa Felicia dam).
A two-dimensional theory was developed by Yiagos and Prevost (1991b)
which also considered the two-phase elasto-plastic seismic response of earth
dams. This theory accounted for the presence of water in the dam material
by including the pore water equations in the saturated portion of the dam
and by adopting an effective stress multi-surface elasto-plastic constitutive
model. This theory was later applied on Santa Felicia and Long Valley dams
(Yiagos and Prevost, 1991a) as an elementary attempt to investigate the ef-
fect of coupled analysis on the seismic behaviour of dams. Their computed
crest accelerations and response spectra were compared to the field measure-
ments and exhibited a fair agreement. Although the authors did not directly
compare results of single- and two-phase analyses, they commented that tak-
ing into account the pore water pressure further enhances the simulation of
the dynamic response of a dam.
The shear beam method was also extended to include three-dimensional
(3D) analysis of dams. Abdel-Ghaffar and Koh (1982) developed another
method to analyse 3D isotropic linear elastic dams whose elastic moduli
varied spatially (inhomogeneous) and which were founded on rigid canyons.
The mathematical formulation of their approach was presented in detail and
was then applied on Santa Felicia dam. The computed modes and natural
frequencies of vibration were compared to measured results from full-scale
tests and earthquake response records with reasonable agreement. However,
the authors did not compare any results from 3D and 2D analysis to identify
82
their differences and highlight the importance of performing a full 3D seismic

analysis.
Moreover, nonlinear elasto-plastic material behaviour was introduced by
Abdel-Ghaffar and Elgamal (1987b) who considered the hysteretic response
of a 3D inhomogeneous uniform earth dam. The dam was assumed to sit
on a trapezoidal rigid canyon and vibrate in all three directions, upstream-
downstream, vertical and longitudinal. This method was also applied on
Santa Felicia dam (Abdel-Ghaffar and Elgamal, 1987a) and was compared to
finite element solutions with acceptable accuracy, emphasising its practicality
in analysing a large structure at a reasonable computational cost.
The method of Abdel-Ghaffar and Elgamal (1987b) was extended by Elga-
mal (1992) to include dam-alluvial foundation interaction and a more general
canyon geometry configuration. The geometry of the system considered is
shown in Figure 3.4.
Dynamic numerical analyses in time domain were performed with a multi-
yield surface kinematic model based on the theory of incremental plasticity
(Prevost, 1977) with the von Mises yield criterion. That constitutive model
was employed in order to simulate the nonlinear hysteretic response of soil
under dynamic loading. The numerical model was utilised in the seismic
analysis of La Villita earth dam in Mexico during two earthquake events and
the results were compared with the measured response. The numerical model
was able to reproduce the measured accelerations at the crest with a reason-
able accuracy, however, it was not able to predict the observed permanent
displacements which were attributed to a localised slope failure.
3.5 Dynamic finite element method

Although the shear beam approach takes into account the dynamic response
of a dam, it still carries its own limitations. More advanced numerical meth-
ods evolved which are able to handle more complicated boundary value prob-
lems. Such methods are the Finite Element (FE), Finite Difference (FD) and
Boundary Element (BE) approaches, which are able to overcome these re-
strictions. The FE is the most widely used method nowadays. It evolved in
83
Figure 3.4: Geometry of the system considered by Elgamal (1992): (a) physi-
cal problem, (b) UD cross-section, (c) shear stresses acting on an soil element
and (d) geometric configuration of computational model.
the late 1950’s and early 1960’s, and it was soon widely used in studying the
response of earth dams (Clough and Chopra, 1966; Chopra, 1967a; Chopra
et al., 1960).

The FE method treats a continuum as an assemblage of discrete elements
defined by nodal points at their boundaries. The dynamic finite element
method solves the equation of motion (Equation 3.11) for dynamic equilib-
rium with time. Solutions of the primary unknown (such as displacements
and pore pressures) are provided at these nodes and their value at any other
point in the whole domain can be obtained by using interpolation functions
(see Potts and Zdravković (1999)).
84
[M ]{ü} + [C]{u̇} + [K]{u} = {P (t)} (3.11)
where [M], [C] and [K] are the mass, damping and stiffness matrices re-
spectively with entries refering to the nodal points, whereas {ü}, {u̇}, {u}
and {P (t)} are the acceleration, velocity, displacement and external loading
vectors respectively. The stiffness matrix, [K] contains information about the
constitutive behaviour of the materials (see Potts and Zdravković (1999)).
Dynamic finite element analysis involves both spatial discretisation of the
problem into small regions (the finite elements) and temporal discretisation
according to the problem’s loading history. Then, appropriate boundary
conditions (both deformation and hydraulic) are applied and advanced con-
stitutive models may be assigned to different regions and for different time
periods. The acceleration time history is applied at a part of the FE mesh
(normally at the soil-bedrock interface, depending on the stiffness of the two
materials and the impedance contrast).
The advantage of this method compared to the previous methods is the
ability to:
• Handle complicated geometries
• Simulate sophisticated material behaviour
• Couple soil skeleton and pore fluid response
• Simulate reservoir hydrodynamic pressures (modelling the reservoir wa-

ter)
• Model the site more accurately (by advanced boundary conditions) and
include any nearby structures, such as spillway, drainage galleries etc.
• Simulate the structure’s previous (static) history (such as layered con-

struction, water impoundment, consolidation etc.)
An example of a FE mesh used for dynamic finite element analysis of

the seismic response of dams is shown in Figure 3.5, which was used by Elia
85
et al. (2010) for the analysis of the Marana Capacciotti dam in Italy. A more
detailed description of the FE method along with relevant recent examples
related to dams is given in Section 3.5.2.
Figure 3.5: FE mesh used for the 2D nonlinear static and dynamic analysis
of Marana Capacciotti dam by Elia et al. (2010).
Finally, it should be noted that the FE method is actually able to fulfill all
three analysis objectives described before: (a) it can provide an assessment
of the stability of the dam structure, through the adoption of constitutive
models that consider soil plasticity, (b) it can provide a reliable calculation
of the permanent displacements of the dam body, through a proper simulation
of degradation of soil stiffness and plastic yielding, and (c) it may offer an
evaluation of the dynamic response of the whole dam through the formulation
of the equation of motion.
3.5.2 Dynamic finite element analysis of earth dams

The first application of the finite element (FE) method in the dynamic anal-
ysis of earth dams was by Chopra (Chopra, 1966; Clough and Chopra, 1966)
who conducted linear elastic dynamic analyses of a homogeneous triangular
dam (mesh shown in Figure 3.6). They evaluated the stress state near the
faces of the dam in an attempt to realistically estimate the safety of these
surfaces to sliding. That work highlighted the limitations of the assumption
of uniform horizontal shear stress adopted by the shear beam method (Am-
braseys, 1960a). Although the horizontal shear stresses are uniform within
the dam body, they drop to zero at the free faces of the dam.
Moreover, Chopra (1967a) calculated the natural modes of vibration of
an earth dam and compared his results with the shear beam solutions of
Ambraseys (1960a). Figure 3.7 shows the shape of the first mode of vibration
86
Figure 3.6: The finite element mesh used by Clough and Chopra (1966) (B
= 450 ft, H = 300 ft).
of two triangular dams, with slopes H:V being 1.5:1 and 3:1, at the centreline
and at the face of the dam as obtained by FE analysis and compared to the
results of the SB solution (a single result as the SB solution is unique for
any slope angle and the same at the centerline and the face of the dam,
see Section 2.4). It is shown that, the FE results differ significantly between
themselves and with respect to the SB solution (see also Section 3.6). Besides,
the fundamental period calculated with the FE method is smaller than that
of the SB method and as expected is different for different slope angles.
Figure 3.7: Shape of the first natural mode of vibration of a uniform, homo-
geneous dam as obtained by SB (a) and FE (b-e) analyses (Chopra, 1967a).
87
Nonlinearity was later introduced by considering the behaviour as equivalent-

linear or fully elasto-plastic. The equivalent-linear approach (Schnabel et al.,
1970; Idriss et al., 1973; Idriss and Sun, 1992) essentially uses a constant value
of the shear modulus, G, and damping, ξ, compatible with the induced shear
strains, γ according to assumed relations describing strain dependent stiff-
ness degradation and damping variation (e.g. Vucetic and Dobry (1991)).
Examples of using the equivalent-linear approach include the work of Seed
et al. (1969) in the analysis of the failure of Sheffield dam under the 1925
Santa Barbara earthquake and of Vrymoed (1981) in the analysis of Oroville
dam.
Nonlinear elasto-plastic dam behaviour was firstly modelled with FEs by
Prevost et al. (1985), using their in-house software DYNAFLOW (Prevost,
1981). The case studied was the Santa Felicia dam and the FE mesh used is
shown in Figure 3.8. Their study considered the hysteretic response of soil
employing a multi-surface plasticity theory (Mroz, 1967), adopting an 11-
piece yield surface linearization to approximate the shear stress-strain curve
(Figure 3.9). Both 2D and 3D analyses were carried out and compared in
order to evaluate the significance of the canyon width. However, small differ-
ences were found between the 2D and 3D analyses and these were attributed
by the authors to the relatively long crest of Santa Felicia dam (for which
L/H = 389m/83m ≈ 4.7). It should be noted that this is in contrast to
the study of Elgamal et al. (1987b), discussed in Section 5.4.2, who used
the shear beam method on Santa Felicia dam and reported that their model
was able to highlight the importance of taking accound of the longitudinal
dimension of the dam.
Comparison between 2D and 3D elasto-plastic analyses were also carried
out by Griffiths and Prevost (1988) on the Long Valley dam in California in
which case the 3D results gave marginally better agreement with the mea-
sured values. Moreover, the failure of the FE mesh to capture the higher
frequencies in that study was thought to be partly due to the coarse mesh,
highlighting the importance of adopting appropriate element size and mesh
refinement.
Elastic-perfectly plastic analyses of Long Valley dam with viscous (Rayleigh)
88
Figure 3.8: FE mesh used by Prevost et al. (1985) in the dynamic analysis
of Santa Felicia dam.
Figure 3.9: Hysteretic soil behaviour was modelled by Prevost et al. (1985)
using multi-surface plasticity.
89
damping were performed by Woodward and Griffiths (1996). In the Rayleigh

(1885) damping formulation, the amount of damping introduced is frequency
dependent and it is equal to the target value only for a small range of frequen-
cies. Therefore, the response for the most of the frequencies, is either slightly
under-damped or significantly over-damped (see Section 3.7). In that study,
viscous damping was used to model the lack of hysteretic damping of the
simple elastic-perfectly plastic material behaviour, and a value of damping,
ξ > 9% was adopted to match the peak horizontal acceleration. Although
the response of the dam was well captured in the horizontal direction (since
Rayleigh damping was calibrated based on that), it was not well predicted
in the vertical direction and this was thought to be due to the frequency-
dependent calibration of Rayleigh damping. This showed a limitation of this
modelling technique, i.e. when the main frequency content in one direction
differs significantly from that of the other, and therefore the calibration does
not apply in both directions.
The FE method has also been used in conjuction with the Boundary El-
ement Method (BEM), another numerical technique which requires the dis-
cretisation of only the boundaries and not the whole domain of the problem.
This may result in less computational cost, because of the smaller number
of discretisation nodes, but it is restricted to linear analyses. This approach
has been widely used in studies of soil-structure interaction (Spyrakos and
Beskos, 1986; Von Estorff and Kausel, 1989), where the inelastic structure was
modelled with FEs, whereas the elastic ground was modelled with BEs. Sim-
ilar studies were conducted for dam-foundation interaction (Nahhas, 1987;
Abouseeda and Dakoulas, 1996, 1998), where the elasto-plastic dam was
modelled with FEs and its foundation was assumed to behave as an elastic
halfspace and was modelled with BEs.
3.5.3 Coupled hydro-mechanical analysis

More accurate predictions of the response of dams under seismic loading
require numerical analyses that are able to simulate the development and
dissipation of excess pore water pressures with time. These analyses need
90
fully coupled effective stress formulations following the Biot (1941) theory
for soil-fluid interaction (Zienkiewicz et al., 1999; Kontoe, 2006).
The first attempt to consider both the nonlinearity and the multi-phase
nature of the soil in a seismic analysis of earth dams was the work of Lacy and
Prevost (1987). That was an extention of earlier work (Prevost et al., 1985)
on Santa Felicia dam which considered the coupled dynamic equations for the
response of a two-phase soil system. The resulting maximum acceleration of
the two-phase model was found to be substantially lower than the maximum
of the one-phase model. This was attributed to the additional damping
introduced in the system from the introduction of pore fluid-soil skeleton
interaction.
Fully-coupled effective stress elasto-plastic analyses were carried out by
Muraleetharan et al. (2004) and were compared to centrifuge model tests of
two sand embankments. The constitutive model used in their code, DYSAC2
(Muraleetharan et al., 1988), was based on the bounding surface plasticity
framework (Dafalias and Popov, 1979) which allows plastic strains to occur
for stress points within the bounding surface. The numerical predictions of
accelerations, pore water pressures and deformations compared reasonably
well with the experimental results, although the peak values of accelerations
and pore pressures were underpredicted.
Moreover, Sica et al. (2008) performed coupled dynamic analyses of El
Infiernillo dam in Mexico using their in-house software GEFDYN (Aubry
et al., 1985), adopting an elasto-plastic model (Heujeux, 1985) characterised
by both isotropic and kinematic hardening. The numerical predictions were
compared to the measured response of the dam during the 14/3/1979, 19/9/1985
and 30/5/1990 Mexico seismic events. Their numerical model was calibrated
against the measured response and a good agreement was achieved between
the computed and measured accelerations and reponse spectra of the dam.
The calibration was then followed by a parametric study which aimed to
identify the influence of past loading history on the seismic response of earth
dams.
It was found that when the same seismic event was repeated after the first
event, the computed earthquake-induced crest settlements were smaller dur-
91
ing the second time. Moreover, it was found that the difference in the crest
settlements between two successive identical seismic events, was larger for
larger events. The authors commented that the observed difference in the
crest settlements (and therefore the influence of past loading history) was
due to strain hardening of the soil and therefore their study demonstrated
the effect of soil hardening due to past seismic history (a series of consecu-
tive earthquakes) on the computed loss of freeboard. Their conclusions were
drawn due to the adoption of the advanced elasto-plastic model that was
able to capture soil hardening. Therefore, this highlighted the advantages of
using appropriate constitutive models that take account of strain hardening.
However, it should be noted that their study did not directly compare results
from different numerical analyses that consider and ignore strain hardening
and hence it was assumed that the difference in the predicted crest settle-
ments was entirely due to the modelled strain hardening.
Fully coupled dynamic analyses of the Marana Capacciotti dam in Italy
were performed by Elia et al. (2010) employing an advanced elasto-plastic
constitutive model based on multi-surface plasticity. The model for struc-
tured soils, MSS (Kavvadas and Amorosi, 2000), uses two nested elliptic
surfaces in the stress space of the modified Cam-Clay type. This formulation
includes both isotropic and kinematic hardening and is able to describe decay
of stiffness, increase of hysteretic damping and build-up of pore water pres-
sures with shear strain. The FE mesh used is shown in Figure 3.10 and the
seismic behaviour of the dam was analysed using a number of input seismic
motions.
However, because the FE code they employed, DIANA-SWANDYNE II
(Chan, 1988, 1995) could not offer the option of dynamic absorbing bound-
aries, they had to move the lateral boundaries of the FE mesh (Figure 3.10
(a)) further away from the problem of interest (Figure 3.10 (b)). Moreover,
viscous boundaries along the left and right sides of the foundation layer were
simulated by means of two columns of elements characterised by a Rayleigh
damping equal to 25%, in order to minimise wave reflections during the seis-
mic action. This however, resulted in the adoption of a very wide mesh which
generally results in high computational cost.
92
(a) The entire FE mesh
(b) Detail of the FE mesh close to the embankment
Figure 3.10: The FE mesh used by Elia et al. (2010): (a) the entire FE
mesh describing the problem and (b) detail of the FE mesh close to the
embankment.
Their analyses showed significant plastic strain accumulation at the toes

of the dam and a large development of excess pore water pressures induced
because of the shaking. These excess pore water pressures were dissipated
during a consolidation stage following the earthquake and resulted in some
additional settlements. The predicted crest settlements for the large input
motions considered (corresponding to a large return period according to the
local seismic hazard studies) were found to be less than half of the freeboard
and it was therefore suggested that the seismic performance of the dam in a
future will be satisfactory.
It was commented that this work evaluated the significance of using cou-
pled formulation along with the adoption of advanced elasto-plastic models
in the analysis of earth dams to appropriately model any positive or negative
excess pore pressures. It was also reported that a consolidation period follow-
ing the seismic response can provide an indication of the resulting permanent
displacements due to dissipation of these excess pore pressures. However, it
93
should be noted that the numerical predictions of that study were not com-
pared to any field data, as such data did not exist and therefore this raises
questions about the reliability of those results. Nevertheless, the authors
commented that their study represents a class A prediction of the dynamic
response of the dam.
Finally, it should also be noted that, as discussed earlier, the authors
attributed the prediction of significant plasticity and additional consolida-
tion settlements to their adopted advanced kinematic hardening constitutive
model. But their study did not compare results from analyses considering
and ignoring kinematic hardening, and therefore kinematic hardening cannot
be identified as the sole reason for achieving those results. If such a compar-
ison was performed, it would better highlight the benefits of such advanced
constitutive models. Therefore, further investigation is required, particularly
using available measured field data, to confirm (or perhaps question) the
results of the study of Elia et al. (2010).
3.5.4 Modelling of reservoir-dam interaction

The hydrodynamic pressures on dams obtained from analytical methods were
described in Section 2.5. Those methods considered simple geometries and
generally rigid (or flexible with simple deformation modes) dams. However,
they should not be used in cases involving more complicated problems with
irregular geometries and deformable dams and canyons where reservoir-dam
interaction effects can become significant (see Section 2.4.5). This section
describes numerical methods employed to deal with hydrodynamic pressures
and reservoir-dam interaction in more complicated problems.
The reservoir is included in the analysis as a discretised domain employ-
ing usually the finite element (FE) or the boundary element (BE) method.
As far as the FEM is concerned, the reservoir domain may be discretised
using solid elements (the same as those used to discretise the dam domain,
i.e. Equation 3.11) or “fluid elements”, following either the Eulerian or La-
grangian approaches.
For the case of solid finite elements, Zienkiewicz et al. (1986) suggested
94
that the fluid should be treated as a ‘degenerate case’ of a soild in which

the shear modulus is made zero but the compressibility is retained (Wilson,
1975). Also, Wilson (1995) stated that a confined fluid can be considered
as a special case of a solid material, which has a very small shear modulus
compared to its compressibility modulus. This approach was also adopted
in other cases involving fluid-structure interaction problems. The work of
Dakoulas and Gazetas (2008) for example, on seismic FE analysis of caisson
quay walls, where the water was modelled as “a saturated, elastic sponge”.
In that study, the density and the bulk modulus of the water were assigned
to the reservoir elements, along with a small value for the shear modulus,
required to prevent numerical instability.
In the Eulerian FE approach (Küçükarslan et al., 2005), the variables
describing the fluid are the pressures, the velocities or the velocity potentials
of the fluid. The hydrodynamic pressure distribution in the reservoir is gov-
erned by the pressure wave equation, which for linear compressible inviscid
water is given by Equation 3.12.
1
∇2 φ(x, y, t) = φ̈(x, y, t) (3.12)
Vp2
where, Vp is the p-wave (acoustic) velocity of water and φ is the velocity
potential, which is related to the pressure, p and the velocity vector, {v}
through Equations 3.13 and 3.14.
{v} = ∇φ (3.13)
p = −ρφ̇ (3.14)
An absorbing boundary condition (BC) is usually specified on the trun-

cated upstream reservoir boundary. The most fundamental and most com-
monly used BC in fluid elements is the Sommerfeld (1912, 1949) radiation
condition given by Equation 3.15.
∂p ṗ
= (3.15)
∂n Vp
95
where, p is the water pressure, ṗ is the time derivative of water pressure,

n is the direction out of the boundary and Vp is the p-wave velocity of water.
Other commonly used BCs include the Sharan (1985) and Higdon (1991)
BCs. Examples of Eulerian FE description of the reservoir domain include
the work of Sharan (1985), Tsai and Lee (1991) and Küçükarslan et al. (2005).
In the Lagrangian FEM, the displacements are the unknown variables for
the fluid domain. The governing equation for the reservoir domain is given
by Equation 3.16 (Parrinello and Borino, 2007).
{p} = [Cf ]{} (3.16)
where, {p} is the vector of water pressures, [Cf ] is a matrix containing

constraint parameters and {} is a vector of strains. The Sommerfeld (1949)
BC is again a commonly used BC applied on the upstream reservoir bound-
ary. Examples of Lagrangian FE description of the reservoir domain include
the work of Wilson and Khalvati (1983) and Parrinello and Borino (2007).
Other approaches include the use of the BE Method coupled with the FE
method. As the BE method is restricted to linear behaviour, it is typically
used to model the reservoir domain, whereas the structure is modelled with
finite elements and can have different material behaviour (nonlinear elastic,
elasto-plastic etc). Von Estorff and Antes (1991) for example adopted a FEM-
BEM coupling procedure. In their study, the linear elastic dam structure
was modelled using finite elements, whereas the adjacent fluid was modelled
by boundary elements. A more comprehensive description of the different
numerical methods for modelling dam-reservoir interaction may be found by
Papazafeiropoulos et al. (2011).
3.6 Comparison of methods of analysis

The early methods of analysis, based on the pseudo-static approach have
been widely used to analyse real dams. However, they were combined with
the adoption of very conservative factors of safety in order to overcome the
severe assumptions of this approach. The PS approach cannot be used in
96
cases of dams with irregular geometries or if high variability in the dam

materials exist.
The sliding block method (Newmark, 1965) is still used nowadays for the
prediction of permanent displacements of dams, but it is similarly restricted
to simple loading and geometric conditions. One of its major disadvantages is
the inability to include realistic excess pore water pressure development and
hence predict displacements due to liquefaction. Moreover, the assumption
of a single sliding surface restricts movements along a single (or a finite
number) surface towards a certain direction. This cannot model a large
failure mechanism involving complicated interaction between various failure
surfaces.
The shear beam method has been used extensively in the seismic analysis
of dams, both in specific case studies (such as Frazier (1969) for Bouquet
Canyon dam, Petrofski et al. (1974) for Mavrovo dam and Elgamal (1992)
for La Villita dam) and in theoretical investigations of the dynamic behaviour
of dams during earthquakes. It had a significant contribution to the study of
dynamics of dams as the early analytical solutions for frequencies and modes
of vibration were based on this approach (see Section 3.4). Investigations
about the dynamic response and characteristics of dams including vertical
(Gazetas, 1981b) and longitudinal (Gazetas, 1981a; Abdel-Ghaffar and Koh,
1981) oscillations, 3D analysis effects (Ohmachi, 1981) and canyon geometry
effects (Dakoulas and Gazetas, 1986) were firstly approached using the shear
beam method. It is undoubtedly one of the most widely used and respected
methods of analysis of earth dams.
However, Gazetas (1982a) states that the SB method is only an ideal-
ization which violates the physical boundary condition of zero shear stress
at the two slopes of the dam. Besides, both tensile and compressive normal
stresses cannot be modelled by this method (as it accounts only for shear
wave propagation), although these forces do exist in dams due to multiple
wave reflections at the faces of the dam (Chopra et al., 1960; Clough and
Chopra, 1966). As stated earlier, the basic assumption of the original SB
method is that the soil particles are restricted to move parallel to the hori-
zontal plane (for a wave propagating upwards). According to Kramer (1996),
97
the SB method forces the entire elastic wave energy to produce shear defor-
mations, whereas the FE method can model the actual dam’s response in
both shear and dilatational deformations.
Tsiatas and Gazetas (1982) compared the results of 2D SB and plane-
strain FE analyses on a number of idealised dam cases. This comparative
study showed that the fundamental frequency of most dams was overesti-
mated by 5% by the SB method, whereas higher percentages have been ob-
served for frequencies corresponding to higher modes (as they involve a higher
contribution of bending deformation). Although the SB displacements can
be considered relatively accurate (within 10% of those resulting from FE
predictions), crest accelerations for flexible dams appeared to be up to 50%
higher than the corresponding FE predictions, due to the so-called ’whip-
lash effect’ (Gazetas, 1987). The latter phenomenon is illustrated in Figure
3.11 which compares the peak values of acceleration in an idealised triangu-
lar dam, subjected to a real earthquake acceleration record, as these result
from the FE (dashed line) and SB (solid line) methods. It was illustrated by
Gazetas (1987) that unrealistically high values of accelerations are obtained
close to the crest of the dam from the SB method (as compared to the more
accurate FE method).
Figure 3.11: Distribution of peak acceleration response with depth in an earth

dam (Gazetas, 1987) from FE (dashed line) and SB (solid line) analyses.
The above-mentioned observations provided an insight into the relative

merits and the suitability of the various methods of analysis. From what has
98
been found in the literature, the FE method seems to be the most powerful
tool, as it can overcome most of the assumptions of the other methods. In
addition, the SB method can also be used, particularly to provide analytical
relations of simplified problems in the frequency domain (such as periods
and modes of vibration) which could be used for comparison with the FE
predictions.
3.7 Dynamic analysis considerations

The analyses described in this thesis were carried out using the Finite El-
ement method. The in-house FE software ICFEP (Imperial College Finite
Element Program) was employed. The main features of this numerical code
are detailed by Potts and Zdravković (1999), whereas examples of its appli-
cation in engineering problems are discussed by Potts and Zdravković (2001).
However, the details of the dynamic capabilities of ICFEP are not in-
cluded in the previously mentioned books as these were developed subse-
quently. Therefore, as far as the dynamic part of the code is concerned, ad-
vances related to the formulation of the equations of motion are described by
Hardy (2003), time-integration schemes and advanced boundary conditions
are discussed by Kontoe (2006), whereas appropriate constitutive models for
dynamic analysis are presented by Taborda (2011). ICFEP was successfully
used before in the static analysis of dams and embankments (Dounias, 1987;
Potts et al., 1990; Kovacevic, 1994; Dounias et al., 1996; Day et al., 1998;
Zdravković et al., 2002; Lollino et al., 2005; Grammatikopoulou et al., 2007;
Kovacević et al., 2013) and in the dynamic analysis of tunnels, retaining
structures and slopes (Kontoe et al., 2008, 2011, 2012; Tripe et al., 2013).
This section describes the various issues that need to be taken into con-
sideration in order to carry out appropriate and useful dynamic numerical
analyses. Topics that need to be considered are boundary conditions, element
size, time integration schemes and material constitutive models.
99
3.7.1 Boundary conditions

Boundary conditions (BCs) need to be specified in a FE analysis because the-
oretically the equations describing the problem cannot be solved (integrated)
without specified BCs.
For common static problems, such as static loading of a foundation,
elementary BCs are used at the boundaries of the FE mesh (Potts and
Zdravković, 1999). Elementary BCs include the prescribed displacements
(Dirichlet BC - specifies the value of the unknown variable) and stresses
(Neumann BC - specifies the value of the derivative of the unknown vari-
able) (Wolf, 1985).
For dynamic analysis where wave propagation is modelled, special BCs
exist that overcome the problem of wave reflections at the boundaries. These
are called absorbing boundary conditions. Three dynamic BCs are employed
in this work: (a) Tied-degrees-of-freedom (TDOF) (Zienkiewicz et al., 1988;
Potts and Zdravković, 1999), (b) the Standard Viscous boundary (SV) of
Lysmer and Kuhlemeyer (1969) and (c) the Cone boundary of Kellezi (1998).
The latter two BCs were implemented in ICFEP by Kontoe (2006).
The TDOF BC essentially means that equal values of a degree of freedom
(displacement, pore pressure etc) are specified at two or more nodes, whilst
the magnitude is unknown. In dynamic analysis of geotechnical problems,
this BC is usually employed to tie the horizontal and vertical displacements
of nodes of the same elevation on the two lateral boundaries.
The Standard Viscous BC (Lysmer and Kuhlemeyer, 1969) is able to
perfectly absorb waves that propagate normal to the boundary due to the
application of a traction condition to a free artificial boundary so that any
reflected stresses are zero. This BC can be physically represented by a series
of normal and tangential dashpots on the FE boundary which for plane strain
conditions (Kontoe, 2006; Kontoe et al., 2009) are described by Equations
3.17 and 3.18 respectively.
∂u(t)
σ(t) = ρVp (3.17)
∂t
100
∂v(t)
τ (t) = ρVs (3.18)
∂t
where, σ(t), τ (t) are the normal and shear stresses on the boundary, u(t)
and v(t) are the normal and tangential displacements on the boundary, Vp and
Vs are the compression and shear wave velocities of the medium respectively
and t is the time.
The Cone BC (Kellezi, 1998, 2000) was developed to deal with vibration
problems on the surface of a half-space where the generated waves propagate
in an area that increases with depth. This BC could be considered as an
extension of the Standard Viscous BC and can be physically represented by
a series of normal and tangential sets of dashpots and springs on the FE
mesh boundary. Due to the inclusion of the additional springs, the cone
BC approximates the stiffness of the unbounded soil domain and therefore it
eliminates the permanent movement which occurs with the Viscous BC at low
frequencies (Kontoe, 2006; Kontoe et al., 2009). For plane strain conditions,
the dashpots are defined by Equations 3.17 and 3.18, whereas the springs are
defined by Equations 3.19 and 3.20
Vp2
σ(t) = ρ u(t) (3.19)
2r
Vs2
τ (t) = ρ v(t) (3.20)
2r
where, r is the distance from the boundary node to the source location.
3.7.2 Location of the boundaries

Concerning the location of the FE mesh boundaries, two issues need to be
considered. Clearly, one needs to decide where the bottom and lateral bound-
aries should be placed so that the solution of the problem is not affected by
the BCs.
Firstly, the location of the bottom boundary is affected by the location of
the bedrock and its properties, mainly its stiffness. Generally, if the bedrock
is much stiffer compared to the overlying soil, i.e. the impedance contrast is
101
Figure 3.12: In numerical analysis of seismic soil-structure interaction prob-

lems, the response close to the lateral (artificial) boundaries should be similar
to the free-field response, so that the boundary conditions do not interact
with the response of the structure (Committe on Nuclear Structures and
Materials, 1979).
high, then the bedrock may be assumed to be rigid and hence there is no need
to be discretised. In that case, the location of the bottom boundary is at the
interface of the soil and the bedrock. On the other hand, if the impedance
constrast is not relatively high, then a part of the bedrock should be modelled
too and special techniques may be used to reduce the computational domain,
such as the Domain Reduction Method (DRM) (Bielak et al., 2003; Kontoe,
2006; Kontoe et al., 2008).
As fas as the lateral boundaries are concerned, as illustrated by Figure
3.12, they should be placed at such distance so that the response close to
the boundaries is similar to the free-field response. For this reason, 1D site
response and 2D plane strain analyses should be performed and compared
102
in order to determine the distance at which the lateral boundaries should

be placed so that there is no interaction with the structure studied. This
means that the dynamic response at the lateral boundary of a boundary
value problem should be similar to the response of a soil layer without any
structure.
3.7.3 Element size

Another issue that needs to be taken into account when designing a FE
mesh is the size of the elements. The elements should be small enough to
provide FE solutions at an acceptable accuracy and large enough to avoid
excessive computational cost. For dynamic analysis in which wave propa-
gation is modelled, all the waves of frequencies of interest should be appro-
priately modelled. The size (length) of the element, ∆l, should be limited
to 1/10 (Kuhlemeyer and Lysmer, 1973) to 1/8 (Lysmer et al., 1975) of
the smallest wavelength, λmin , for 4-noded elements (with linear shape func-
tions), whereas for 8-noded isoparametric elements (with quadratic shape
functions), ∆l should be limited to 1/5 (Kuhlemeyer and Lysmer, 1973) to
1/4 of the smallest wavelength λmin (Bathe, 1996). The latter parameter is
the ratio of the smallest shear wave velocity, Vs , of the materials over the
highest frequency of the input wave to be modelled, fmax (Equation 3.21).
λmin λmin VSmin VSmin

∆l ≤ ∼ = ∼ (3.21)
5 4 5fmax 4fmax
3.7.4 Time integration

Time integration schemes are adopted in dynamic FE analyses to integrate
the equation of motion (Equation 3.11) with respect to time. The time-
integration scheme adopted for all the analyses in this work is the Generalised-
α algorithm of Chung and Hulbert (1993). This algorithm which is uncondi-
tionally stable, second order accurate with controllable numerical damping
was implemented in ICFEP by Kontoe (2006) and used in a number of case
studies. In all the dynamic analyses of this work, the default algorithm pa-
rameters were used, unless otherwise mentioned. The default parameters
103
correspond to spectral radius at infinity, ρ∞ = 9/11, and therefore are given

by Equations 3.22 - 3.25 (see also Kontoe (2006)).
2ρ∞ − 1
αm = = 0.35 (3.22)
ρ∞ + 1
ρ∞
αf = = 0.45 (3.23)
ρ∞ + 1
1
α = (1 − αm + αf )2 = 0.3025 (3.24)
4
1
δ= − αm + αf = 0.6 (3.25)
2
3.7.5 Material constitutive models

The FE equations describe the behaviour of the boundary value problem
(earth dam) according to the boundary conditions (earthquake excitation).
The behaviour of the material is described by the constitutive matrix, [D],
which is included in the stiffness matrix, [K], in Equation 3.11 (see also Potts
and Zdravković (1999); Kontoe (2006)). The behaviour of soil under cyclic
loading was briefly discussed in Section 2.3.4.
In this work, a Cyclic Nonlinear (CNL) model was adopted to simulate
the degradation of stiffness and increase of damping with shear strain. This
was coupled with an elasto-plastic failure criterion which describes the plastic
yielding behaviour of the soil.
3.7.5.1 Cyclic Nonlinear Model
In the analyses carried out in this thesis, the CNL model used was the
Logarithmic (Puzrin and Burland, 2000) model, implemented in ICFEP by
Taborda (2011). Equation 3.26 defines the backbobe curve described by the
Logarithmic model.
R #
E ∗ Gmax
"
∗
J = Ed∗ Gmax 1 − a ln (1 + d ) (3.26)
JL
104
where, J ∗ is the modified three-dimensional stress invariant, whereas Ed∗

is the modified three-dimensional strain invariant. The stress and strain
invariants, J and Ed , are defined by Equations 3.27 and 3.28 and can take
only positive values. In contrast, their modified counterparts, J ∗ and Ed∗ ,
used in the formulation are modified to assume both positive and negative
values as explained by Taborda (2011).
1 q
J = √ (σ10 − σ20 )2 + (σ20 − σ30 )2 + (σ30 − σ10 )2 (3.27)
6
2 q
Ed = √ (1 − 2 )2 + (2 − 3 )2 + (3 − 1 )2 (3.28)
6
where, σ10 , σ20 and σ30 are the principal effective stresses whereas 1 , 2 and
3 are the principal strains.
Parameters a, JL and R are model parameters described by Equations
3.29 - 3.31. It is worth to note that the maximum shear stiffness of the
soil, Gmax , is a parameter necessary for the calibration of the CNL model.
Therefore, the performance of the model depends on the spatial variation of
Gmax in a boundary value problem.
c(1 + χL ) ln (1 + χL )
R= (3.29)
χL (χL − 1)
χL − 1
a= (3.30)
χL [ln(1 + χL )]R
EdL
χL = Gmax (3.31)
JL
Figure 3.13 shows schemetically the backbone curve of the Logarithmic
model implemented in ICFEP. It is shown that the backbone curve of the
model follows a straight line after a value of the strain, EdL . The value of the
slope (Gimp ) of this straight line is determined by the value of the parameter
c. This has a direct impact on the induced damping by this model. It should
also be noted that there is an option in ICFEP to specify the minimum
value of the shear modulus, Gmin , independently of the value of parameter c
105
J*
Gmax
Gimp
JL
c JL
E dL E*d
Figure 3.13: Backbone curve of the Logarithmic CNL model (After Taborda
(2011)).
Damping, ξ
{
c=1
for AND
Gmin = 0
{
c<1
for OR
Gmin > 0
Shear strain, γ
Figure 3.14: Illustration of the influence of parameters c and Gmin on the
damping introduced by the Logarithmic CNL model (After Taborda (2011)).
106
specified. Then, the minimum value of the tangent shear modulus used by
the program will be the largest of the two, i.e. max{Gimp , Gmin } and this
value finally dictates the amount of damping introduced (see Figure 3.14).
More details about the formulation and implementation of these CNL models
in ICFEP can be found by Taborda (2011).
3.7.5.2 Elasto-plastic model
The previously described CNL model simulates the pre-yield elastic be-
haviour of the soil. It can be used though in conjuction with an elasto-plastic
model which can simulate the plastic yielding of the material. In this thesis,
the CNL model was coupled with a Mohr-Coulomb failure criterion. The
soil behaviour becomes plastic when the yield surface is engaged. The Mohr-
Coulomb yield surface is described by Equation 3.32 and shown in Figure
3.15.
c0
!
F =J− + p0 g(θ) = 0 (3.32)
tan φ0
where, J is the deviatoric stress (Equation 3.27), p0 is the mean effective
stress (Equation 3.33), θ is the Lode’s angle (Equation 3.34), φ0 is the angle
of shearing resistance and g(θ) defines the shape of the yield surface in the
deviatoric space (Equation 3.35).
J
g(θ)
c'
g(θ)
tanφ'
p'
Figure 3.15: Yield surface of the Mohr-Coulomb model
1
p0 = (σ10 + σ20 + σ30 ) (3.33)
3
107
σ 0 − σ30
" !#
1
θ = √ 2 20 −1 (3.34)
3 σ1 − σ30
sin φ0
g(θ) = φ0
(3.35)
cos θ + sin θ√sin
3
3.7.5.3 Rayleigh Damping
Viscous damping, ξ, is introduced using the Rayleigh (1885) damping for-

mulation (Woodward and Griffiths, 1996) and it is given by Equation 3.36.
A Bωi
ξi = + (3.36)
2ωi 2
where, coefficients A and B are called Rayleigh coefficients and they are
given by Equation 3.37:
   
 A  2ξt  ω1 ω2 
= (3.37)
 B  ω1 + ω2  1 
where, ξt is the target damping ratio and ω1 and ω2 are the circular
frequencies between which the damping ratio is approximately equal to ξt as
shown in Figure 3.16. Circular frequencies ω1 and ω2 are usualy taken as the
first and third circular frequencies of the system (Zerwer et al., 2002), but
this is strongly problem dependent and should be calibrated for each case
considered (Kontoe et al., 2011).
The values A and B are used to form the damping matrix, [C], of the
dynamic FE equations (Equation 3.11) as described in Equation 3.38:
[CE ] = A[ME ] + B[KE ] (3.38)
Where, [ME ] and [KE ] are the elastic mass and stiffness matrices respec-
tively, as they arise in the dynamic FE equations (Equation 3.11). More
information about Rayleigh damping in dams can be found in Woodward
and Griffiths (1996) and about geotechnical problems in general in Kontoe
et al. (2011).
108
Rayleigh
Rayleigh damping
Damping
10
8
Rayleigh damping
[%]
7
damping
6 Target damping
[%]
ξ5
Damping
t
Rayleigh
3
Mass damping
2
Stiffness damping
1
0
0 ω1 2 3 4 ω
5 6 7
1 Frequency,
Frequency, ω ω 2
Figure 3.16: Illustration of Rayleigh damping and its mass and stiffness
components
3.8 Summary
This chapter dealt with the seismic analysis of dams. Firstly, a review of the
various available methods of seismic analysis of dams was presented explain-
ing their assumptions and describing their potentials and limitations. These
methods fall into three main categories: (a) pseudo-static methods to assess
the stability of the earth dam (i.e. the dam slope), (b) sliding block analy-
sis methods which calculate the permanent displacements and (c) methods
evaluating the dynamic response of the whole dam structure, such as the
shear beam approach. However, advanced numerical techniques, such as the
finite element method are able to satisfy all the analysis objectives: stability,
displacements and dynamic response.
Afterwards, a description of the evolution of the numerical methods of
analysis applied to dams was presented. Numerical methods were formu-
lated in order to model the nonlinear material behaviour of the dam materi-
als, which cannot be calculated analytically. Various methods were developed
over the years and these range from the simple numerical shear beam method
up to the sophisticated nonlinear coupled dynamic analysis including consid-
eration of reservoir-dam interaction effects.
Finally, taking into consideration the plethora of information found in the
109
literature, several issues related to appropriate dynamic finite element anal-

ysis were discussed. Issues to be considered include the dynamic boundary
conditions, spatial discretisation of the FE domain, time integration schemes
and material constitutive models describing stiffness degradation and plastic
yielding.
110
Chapter 4
NUMERICAL MODELLING
OF THE RESERVOIR
DOMAIN
4.1 Introduction
The various methods and approaches available for modelling the hydrody-
namic pressures on dams and dynamic reservoir-dam interaction in general
were described in Section 3.5.4. However, there are still issues that need to be
investigated so that useful and reliable numerical analyses can be performed.
These issues are related to both geometric modelling, such as absorbing
boundary conditions and material modelling such as water compressibility
and shear stiffness.
This chapter describes work carried out on modelling the hydrodynamic
pressures on dams during earthquakes and general dynamic loads using finite
elements. The reservoir water is descritised under plane-strain conditions
using displacement-based eight-noded quadrilateral solid elements, which are
the same as the elements used to discretise the dam and the foundation soil.
111
4.2 Background
The hydrodynamic pressures on a rigid (undeformable) dam with a vertical
upstream face under harmonic loading were investigated in the fundamental
work of Chopra (1967a) (see also Sections 2.5.4 and 2.5.5). In this analyt-
ical approach, the pressures were expressed in the frequency domain using
transfer functions for various values of loading frequency. Morever, the re-
sults of Chopra (1967a) for incompressible water agree with Westergaard
(1933). Besides, it was shown that the pressure response due to an arbitrary
load (such as earthquake) may be obtained from the unit impulse response.
Therefore, the response under a number of impulses of different frequencies
is superimposed with the aid of the convolution integral. The horizontal hy-
drodynamic pressure, px , distribution on the dam face over time, t, under a
general acceleration load, üg (t), is given by Equation 4.1.
4γw Vp +∞
X (−1)n−1 Z t
px (y, t) = cos (λn y) üg (τ )Jo [λn Vp (t − τ )]dτ (4.1)
πg n=1 2n − 1 0
where, y is the vertical distance from the bottom of the dam, t is the
time, γw is the unit weight of water, Vp is the acoustic (p-wave) velocity of
water, λn is the nth wave length, üg (t) is the ground acceleration and J0 (·) is
a Bessel Function of the first kind of order 0.
The problem under study is shown graphically in Figure 4.1. It consists
of a dam (1) which retains a large volume of water in the reservoir (2) and
rests on the ground (3) which serves as the foundation of the dam. Under
seismic or general dynamic conditions, the reservoir induces hydrodynamic
pressures on the upstream (US) dam face, A-B.
As far as efficient and economic numerical modelling of these hydrody-
namic pressures is concerned, the US reservoir is truncated at some distance
from the dam, (C-D). Likewise, the foundation soil is truncated at some
distance from the dam and reservoir, E-F-G-C. On both soil and reservoir
boundaries, special absorbing boundary conditions (BCs) need to be applied
so that reflection of outgoing waves is avoided.
112
A D
1 2
B C
E
3
F G
Figure 4.1: Geometry of the dam-reservoir-foundation system: (1) dam, (2)

reservoir, (3) foundation soil, US dam face A-B, US reservoir boundary C-D
and foundation boundary E-F-G-C.
4.3 Ramp acceleration

Hydrodynamic pressures on dams were investigated by Küçükarslan et al.
(2005) who were particularly interested in the modelling of the reservoir
water. In this study, the reservoir was discretised with fluid finite elements
following the Eulerian approach (see Section 3.5.4). The geometry of the
dam-reservoir system used is shown in Figure 4.2. The left part of the mesh
is the stiff dam with a vertical face modelled with solid elements, whereas
the right part is the reservoir modelled with fluid elements.
Figure 4.2: Geometry of the dam-reservoir system used by Küçükarslan et al.

(2005).
A ramp acceleration as shown in Figure 4.3 (with maximum acceleration,

α = 1 m/s2 ) was used as the load and the numerical results were compared
with the analytical expression of Tsai et al. (1990). After some rearrangement
and substitutions, that analytical expression can be shown to be exactly the
113
same as Equation 4.1 of Chopra (1967a). The analytical solution for the
hydrodynamic pressure time history at the base of the dam (bottom left
corner of the reservoir in Figure 4.2) is shown in Figure 4.4, whereas the
distribution of the maximum hydrodynamic pressure over the entire upstream
face is shown in Figure 4.5.
Figure 4.3: The ramp acceleration load applied to the dam-reservoir system
of Küçükarslan et al. (2005)
The FE mesh used by Küçükarslan et al. (2005) modelled a reservoir

with a height, H, of 180m whereas three values were used for the length, L,
i.e. 540, 960 and 1260m. The applied boundary condition at the upstream
boundary was the Sommerfeld radiation condition (Sommerfeld, 1912, 1949)
(see also Section 2.5). The resulting hydrodynamic pressure time-history at
the base of the dam for the three lengths is shown in Figure 4.6, whereas the
distribution of the peak hydrodynamic pressure is shown in Figure 4.7.
4.3.1 Upstream boundary condition in the reservoir

Numerical analyses were carried out considering the dam-reservoir system
shown in Figure 4.8. Similar to Küçükarslan et al. (2005), the height of the
reservoir was H = 180m, whereas five values have been used for the length,
L of the reservoir, i.e. 540, 960, 1260, 1800 and 2400m, as listed in Table 4.1.
The thickness T of the stiff dam and foundation was 18m. Displacement-
based eight-noded isoparametric quadrilateral solid finite elements were used
114
Normalised Hydrodynamic Pressure at the base of the dam

1.4
Chopra, 1967
1.2
Normalised Hydrodynamic Pressure, Pdyn/ρ α H []
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6
Time, t [sec]
Figure 4.4: The analytical solution for the hydrodynamic pressures at the
base of the dam due to the ramp acceleration using the solution of Chopra
(1967a).
Normalised Hydrodynamic Pressure Distribution on the face of the dam

1
Chopra, 1967
0.9
0.8
0.7
Normalised Height, y/H []
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Figure 4.5: The analytical solution for the distribution of the maximum
hydrodynamic pressure over the entire upstream face of the dam due to the
ramp acceleration using the solution of Chopra (1967a).
115
Figure 4.6: The hydrodynamic pressure time-history at the base of the dam
as obtained by Küçükarslan et al. (2005).
Figure 4.7: The distribution of the peak hydrodynamic pressure on the up-
stream face of the dam as obtained by Küçükarslan et al. (2005).
116
to model the reservoir domain, the stiff dam and the foundation parts. The
maximum size of the elements was 4.5m which is equal to 1/40 of the reservoir
height.
Table 4.1: Geometry of the FE meshes considered for the ramp acceleration
loading
MESH Height, H Length, L L/H

[m] [m]
1 180 540 3
2 180 960 5.33
3 180 1260 7
4 180 1800 10
5 180 2400 13.33
The load was applied along the left (E-F in Figure 4.8) and the bottom (F-
G) boundary as prescribed values of acceleration in the horizontal direction.
The displacement in the vertical direction on both boundaries was prescribed
to be equal to zero. Different boundary conditions (BCs) were examined for
the upstream reservoir boundary (C-D), and these are:
• Free, i.e. zero displacement and stress in the vertical and horizontal
directions respectively
• Viscous (Lysmer and Kuhlemeyer, 1969), i.e. horizontal dashpots, nor-

mal to the boundary and zero displacement in the vertical direction.
• Cone (Kellezi, 1998), i.e. horizontal dashpots and springs, normal to

the boundary and zero displacement in the vertical direction.
• Viscous with hydrostatic stress distribution, i.e. horizontal dashpots

and prescribed values of hydrostatic horizontal stress throughout the
whole analysis and zero displacement in the vertical direction.
As far as the last BC is concerned, it should be noted that according

to Parrinello and Borino (2007), the Sommerfeld BC used as an absorbing
boundary of fluid elements (see Section 2.5) can be defined by Equation 4.2.
This equation, according to Parrinello and Borino (2007) “is applied by a
117
E
A D
reservoir
H
stiff material
B C
T F
G
T L
Figure 4.8: Geometry of the dam-reservoir system used for the ramp accel-
eration loading.
force distribution statically equivalent to the hydrostatic pressure pst and by

a layer of dampers with viscous parameter ρc” (where, c = Vp is the p-wave
velocity of the water). This boundary condition is schematically illustrated
in Figure 4.9.
p = ρcu̇x + pst (4.2)
Figure 4.9: Representation of the Sommerfeld radiation condition as sug-

gested by Parrinello and Borino (2007)
The water was assumed to behave as a linear elastic material with a bulk
modulus, Kw = 2.2 · 106 kPa (the exact value for water) and a nominal value
of shear modulus, Gw = 100 kPa. The reason for assigning a very small value
for the shear modulus is to avoid shear wave propagation in the water domain,
as physically the shear stresses carried by water are negligible. The dam and
foundation were modelled to behave as a “rigid” body (i.e. extremely stiff),
therefore the bulk modulus assigned is K = 108 · Kw = 2.2 · 1014 kPa. The
118
viscosity, CH (Equation 4.3) of the dashpots and the stiffness of the springs,
KH (Equation 4.4) were obtained from the elastic properties of the material
to which they were applied, i.e. water. No damping was assigned to the
reservoir or the dam domains.
CH = A(r)ρVp (4.3)
A(r) 2
KH = ρVp (4.4)
r
where, ρ is the mass density of water, r is the distance from the source of
the loading to the boundary and A(r) is the cross-sectional area of the cone.
The value of the distance parameter is chosen to be equal to the length
of the reservoir, i.e. r = L and A(r) was specified to be equal to the dis-
tance between the boundary nodes (see Section 3.7). More details about the
Viscous and Cone BCs as implemented in ICFEP, can be found in Kontoe
(2006).
Figure 4.10 shows the hydrodynamic pressure time-histories at the base
of the dam (Point B in Figure 4.8) for the Free BC (a) for all the five lengths
of the FE mesh along with the analytical solution. Likewise Figures 4.11,
4.12 and 4.13 show the corresponding time-histories of the Viscous (b), the
Cone (c) and the Viscous with hydrostatic pressure distribution (d) BCs
respectively. The hydrodynamic pressures are normalised with respect to
ρaH, where ρ is the mass density of water, a = 1 m/s2 is the maximum value
of the acceleration and H is the height of the reservoir.
Figures 4.10 - 4.13 show that none of the four models adequately re-
produced the hydrodynamic pressures at the base of the dam. The closest
results appear to be obtained from MESH3 (see Table 4.1), i.e. L = 1260 and
L/H = 7 with the Viscous and Cone boundary conditions. However, for all
four cases of the BCs considered, the first half cycle appears to be predicted
accurately and then depending on the distance of the lateral boundary, the
hydrodynamic pressures tend to increase with time for the shorter meshes
(L/H = 3 and 5.33) and to decrease for the longer meshes (L/H = 10 and
13.33).
119

6
Analytical
Numerical − L/H = 3
5
Numerical − L/H = 5.33
4 Numerical − L/H = 10
3
−1
−2
−3
−4
0 1 2 3 4 5 6
Time, t [sec]
Figure 4.10: Hydrodynamic pressure at the base of the dam by using the
“Free” BC for different lengths of the reservoir

6
Analytical
3
−1
−2
−3
0 1 2 3 4 5 6
Time, t [sec]
“Viscous” BC for different lengths of the reservoir
120

6
Analytical
5
3
−1
−2
−3
0 1 2 3 4 5 6
Time, t [sec]
“Cone” BC for different lengths of the reservoir

6
Analytical
5
3
−1
−2
−3
−4
0 1 2 3 4 5 6
Time, t [sec]
“Viscous + hydrostatic stress” BC for different lengths of the reservoir
121
Moreover, for the Free BC (Figure 4.10), the fluctuations in the hydrody-
namic pressure tend to have a larger amplitude than the analytical solution
(dashed line). For the Viscous and Cone BCs (Figures 4.11 and 4.12) the
amplitude of these fluctuations is similar and comparable to the amplitude of
the analytical solution. In fact, the behaviour of the latter two BCs is almost
identical. Finally, for the “Viscous + hydrostatic stress” BC (Figure 4.13),
one may observe some spurious peaks in the hydrodynamic pressures which
happen at different times for the different lengths of the mesh and which
decrease with time. Furthermore, it seems that the time at which these high
peaks occur increases with the length of the mesh and this can be attributed
to waves reflected back from the upstream reservoir boundary. It seems that
the specification of a constant (hydrostatic) pressure at the boundary results
in a reflection of the waves. It is therefore concluded that both the Free and
Viscous + hydrostatic pressure BCs are not appropriate for modelling the
upstream boundary of the reservoir.
4.3.2 Reservoir-dam & reservoir-base interface

To improve the comparison with the analytical solution, interface elements
(Potts and Zdravković, 1999; Day, 1990; Day and Potts, 1994) were intro-
duced between the reservoir domain and the dam and bottom boundaries
(along A-B-C in Figure 4.8). As a result, relative movement is allowed along
the water-dam and water-bottom interfaces. The values assigned for the nor-
mal and shear stiffness of the interface elements are KN = 108 kN/m and
KS = 1 kN/m respectively. The reason for using these values is to allow only
shear relative movement between the two material domains.
Figures 4.14 - 4.17 show the hydrodynamic pressure time histories at the
base of the dam with the new mesh for all four BCs and all five lengths.
It is shown that the introduction of the interface elements has significantly
improved the results, which for the cases of the Viscous and the Cone BCs
compare well with the analytical solution. Similar to what was mentioned
earlier for the case without the interface elements, the first half cycle is
predicted correctly for all the analyses. However, the introduction of the in-
122
terface elements eliminated the general increase and decrease of the pressures
with time, as compared to the analytical solution.
As before, for the case of the Free BC (Figure 4.14), it is clear that the
fluctuations of the numerical results are significantly larger than the analyt-
ical solution (dashed line). A better agreement appears for the cases of the
Viscous (Figure 4.15) and Cone (Figure 4.16) BCs for which the amplitudes
of the pressures are very close to the analytical solution. Finally, the results
for the “Viscous + hydrostatic stress” BC (Figure 4.17) still show spurious
peaks in the pressure, although the values tend to get closer to the analytical
solution with time.
The effect of the mesh length for the Viscous BC is shown in more detail
in Figures 4.18 - 4.22. Clearly, the comparison of the numerical results with
the analytical solution is improved. Moreover, the numerical response starts
to deviate from the analytical one earlier (i.e. at smaller values of the time,
t) for the shorter meshes. The differences are mainly a slight increase in the
amplitude and a small shortening in the period of the fluctuations. This is
expected as any waves reflected at the boundary require less time to travel
back to the dam if the distance between the dam and the boundary is small.
Note that the results for the cases of the Cone BC plot on top of the results
of the Viscous BC.
Finally, Figure 4.23 shows the distribution of the peak hydrodynamic
pressure on the face of the dam for all 5 lengths of the mesh and how these
compare with the analytical result. It is apparent that the distributions from
all the analyses are extremely similar as there are no visible differences. This
confirms that the model used (Viscous or Cone BC with interface elements)
is adequate to describe the hydrodynamic pressures on a rigid dam with a
vertical upstream face. Again, note that the results for the cases of the Cone
BC plot on top of the results of the Viscous BC.
Considering the previous results (especially Figures 4.15 and 4.16), it may
be concluded that the best comparison with the analytical solution can be
obtained using a FE mesh with L/H = 13.33 (Figure 4.22). Of course, the
ramp acceleration may not be the most suitable load to be used to determine
an appropriate distance of the BC. However, the study carried out so far
123

1.4
Analytical
1.2
1
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6
Time, t [sec]
“Free” BC with interface elements for different lengths of the reservoir

1.4
Analytical
1.2
1
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6
Time, t [sec]
“Viscous” BC with interface elements for different lengths of the reservoir
124

1.4
Analytical
1.2
1
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6
Time, t [sec]
“Cone” BC with interface elements for different lengths of the reservoir

4
Analytical
3 Numerical − L/H = 5.33
/ρ α H []
2 Numerical − L/H = 13.33
dyn
Normalised Hydrodynamic Pressure, P
−1
−2
−3
−4
0 1 2 3 4 5 6
Time, t [sec]
Figure 4.17: Hydrodynamic pressure at the base of the dam by using the “Vis-
cous + hydrostatic stress” BC with interface elements for different lengths
of the reservoir
125

1.4
Analytical
1.2
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6
Time, t [sec]
“Viscous” BC with interface elements for L/H = 3

1.4
Analytical
1.2
/ρ α H []
dyn
1
Normalised Hydrodynamic Pressure, P
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6
Time, t [sec]
“Viscous” BC with interface elements for L/H = 5.33
126

1.4
Analytical
1.2
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6
Time, t [sec]

1.4
Analytical
1.2
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6
Time, t [sec]
127

1.4
Analytical
1.2
0.8
0.6
0.4
0.2
0
0 1 2 3 4 5 6
Time, t [sec]
“Viscous” BC with interface elements for L/H = 13.33
Distribution of the Peak Hydrodynamic Pressure

1
Analytical
0.9 Numerical − L/H = 3
0.8 Numerical − L/H = 7
0.7 Numerical − L/H = 13.33
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Figure 4.23: Distribution of the peak hydrodynamic pressure on the upstream

face of the dam by using the “Viscous” BC. Comparison between numerical
and analytical (Chopra, 1967a) results.
128
suggests that both the Viscous and the Cone BCs can perform quite well
as absorbing BCs for the reservoir. In addition, interface elements with
appropriate values for the shear and normal stiffnesses should be introduced
at the interface between the reservoir and the solid domains.
4.4 Harmonic acceleration

As mentioned earlier, the hydrodynamic pressures (and hence the resulting
forces) on dams due to a compressible reservoir for different load frequencies
of a harmonic load were investigated by Chopra (1967a). The results were
presented in terms of a frequency spectrum of the forces as shown in Figure
4.24. The hydrodynamic effects were expressed as the ratio of hydrodynamic
over hydrostatic force, Fdyn /Fst for different values of the frequency ratio,
Ω = ω/ω1 . The parameter ω refers to the circular frequency of the load and
ω1 to the first natural circular frequency of the reservoir, which is given by
Equation 4.5 (Chopra, 1967a).
2π πVp
ω1 = = (4.5)
T1 2H
where, T1 = 4H/Vp is the fundamental frequency of the reservoir, Vp is
the acoustic (p-wave) wave velocity of the water and H is the height of the
reservoir.
The hydrodynamic force, Fdyn is obtained from the integration of the
normal pressures on the US face of the dam. The hydrostatic force, Fst =
1
2
ρgH 2 , where ρ is the mass density of the water, g is the acceleration of
gravity and H is the height of the reservoir.
4.4.1 Frequency response

Following the analytical work of Chopra (1967a) on the frequency response of
the hydrodynamic pressures, a number of analyses were performed with the
harmonic acceleration described by Equation 4.6, applied in the horizontal
direction along the boundary E-F-G in Figure 4.8. Different values of circular
129
Frequency Response for Total Hydrodynamic Force

4
Chopra (1967)
3.5
2.5
st
/F
dyn
2
gF
1.5
0.5
0
0 1 2 3 4 5 6
Figure 4.24: Frequency response of the hydrodynamic force induced by a

compressible reservoir due to horizontal harmonic loading, as suggested by
Chopra (1967a)
frequency, ω were considered, for constant amplitude, a0 = 1 m/s2 and for

40 cycles.
a(t) = a0 cos ωt (4.6)
where, a is the acceleration, a0 is the amplitude of the harmonic load, ω

is the circular frequency of the harmonic loading and t is time.
Considering the findings from the previous investigation with the ramp
loading, the analyses were carried out using the previously used MESH5
(see Table 4.1) with interface elements placed at the reservoir-dam and the
reservoir-base interfaces. As far as the upstream reservoir BC is concerned,
both the Viscous and the Cone BCs were used in order to check their per-
formance over a wide range of frequencies. The results are shown in Figures
4.25 and 4.26 for the Viscous and Cone BCs respectively, along with the
analytical result. The total hydrodynamic force was obtained by integrating
the hydrodynamic pressures on the dam face at steady state conditions. The
values of the hydrodynamic force for the natural frequencies of the reservoir
130

4
Numerical − Compressible (ICFEP) − Viscous BC
Analytical − Compressible (Chopra, 1967)
3.5
2.5
st
/F
dyn
2
gF
1.5
0.5
0
0 1 2 3 4 5 6
Figure 4.25: Frequency response of the total hydrodynamic force induced

by a compressible reservoir due to horizontal harmonic loading. Comparison
between numerical (Viscous BC) and analytical (Chopra, 1967a) results.

4
Numerical − Compressible (ICFEP) − Cone BC
3.5
2.5
g Fdyn/Fst
1.5
0.5
0
0 1 2 3 4 5 6
Figure 4.26: Frequency response of the total hydrodynamic force induced

by a compressible reservoir due to horizontal harmonic loading. Comparison
between numerical (Cone BC) and analytical (Chopra, 1967a) results.
131
(ω/ω1 = 1, 3, 5, ...) are not included as the response of the reservoir was
transient and the hydrodynamic pressures did not reach a steady state.
As it may be observed from both Figures 4.25 and 4.26, a generally good
agreement is obtained for the whole spectrum of frequencies, for both BCs.
In fact, their performance is almost identical. This suggests that both the
Viscous and the Cone BCs can be applied on the upstream boundary of the
reservoir in order to model the reservoir’s response due to a harmonic load
of a wide range of frequencies.
4.4.2 Effect of water compressibility

As discussed in Section 2.5, the analytical solution of Westergaard (1933)
(Equation 2.42) assumed that water was incompressible and that the magni-
tude of the resulting hydrodynamic pressures was constant for any frequency
of loading. In order to examine the behaviour of the numerical model when
water compressibility is not considered, a set of analyses with different load-
ing frequencies was performed, treating the reservoir water as incompressible.
The FE mesh used is the same as the one used in Section 4.3 and its
geometry is shown in Figure 4.8. The water compressibility, Kw is taken as
10000 times higher than the real value (Kw = 2.2 · 106 kPa), i.e. a value of
Kw = 2.2 · 1011 kPa is used in order to make in incompressible. The Viscous
BC was applied on the upstream reservoir boundary. The results are shown
in Figure 4.27 and are compared with the analytical solutions of Chopra
(1967a) for compressible water and Westergaard (1933) for incompressible
water.
As it may be observed from Figure 4.27, if water compressibility is not
considered and instead water is treated as incompressible, the numerical re-
sults agree with the analytical solution of Westergaard (1933). As a re-
sult, the amplification of the hydrodynamic pressures for loading frequencies
close to the fundamental frequencies of the reservoir cannot be predicted,
as proved by Chopra (1967a). Moreover, the hydrodynamic pressures are
overestimated for higher loading frequencies. Therefore, a realistic water
compressibility should be taken into account in order not to underestimate
132

4
Numerical − Incompressible (ICFEP) − Viscous BC
3.5 Analytical − Incompressible (Westergaard, 1933)
2.5
st
/F
dyn
2
gF
1.5
0.5
0
0 1 2 3 4 5 6
Frequency Ratio, Ω = ω/ω1 compressible
Figure 4.27: Frequency response of the total hydrodynamic force induced by

an incompressible reservoir due to horizontal harmonic loading. Compari-
son between numerical and analytical (Chopra, 1967a; Westergaard, 1933)
results.
the hydrodynamic pressures in the low frequency range that occur due to
resonance between the loading and the reservoir water, or overestimate the
hydrodynamic pressures for high frequency loads.
4.4.3 Effect of element size

A further investigation was carried out in order to examine the effect of the
size of the finite elements on the hydrodynamic pressures. When modelling
dynamic problems, care needs to be taken in order to appropriately model
wave propagation. Therefore a sufficient number of nodes should be provided
to model the shortest wavelength (see Figure 4.28). A very fine mesh could
in one hand provide a sufficient number of nodes, on the other hand however,
it would increase the computational cost. Therefore, the aim is to identify
the largest element size that provides accurate results.
Studies in dynamic soil-structure interaction (Kuhlemeyer and Lysmer,
1973; Lysmer et al., 1975) provided guidelines on the size of the finite elements
133
Figure 4.28: In wave propagation problems a sufficient number of nodes

should be provided to accurately model the full wavelength.
used to model the soil domain so that wave propagation can be adequately
modelled. These studies expressed the size of the element in terms of the
wavelength, as the full wavelength has to be appropriately modelled. For
this investigation, a FE mesh of height H=90m and length L=900m (i.e.
L/H = 10) was employed. Eight-noded isoparametric quadrilateral elements
(see also Potts and Zdravković (1999)) were used as shown in Figure 4.29.
Figure 4.29: Eight-noded isoparametric quadrilateral elements have been

used in modelling the reservoir domain.
The element size, d explored in this investigation was expressed in terms

of the wavelength. For a monochromatic load (i.e. for a single frequency,
f = ω/2π), the wavelength λ of a p-wave (with velocity Vp ) is given by:
Vp
λ= (4.7)
f
The fundamental frequency of vibration of a reservoir of height H is given
by (Chopra, 1967a):
1 Vp
f1 = = (4.8)
T1 4H
134
Therefore, the wavelength λ for a given monochromatic load of frequency

f is given by:
4H 4H
λ= = (4.9)
f /f1 ω/ω1
Therefore, for a harmonic load with frequency such that ω/ω1 = f /f1 = 4,
the wavelength is given by:
ω 4H

λ =4 = =H (4.10)
ω1 4
A harmonic load was applied with frequency ratio, ω/ω1 = 4 and the
range of element sizes d explored is listed in Table 4.2. The reservoir was
considered compressible and the Viscous BC was applied on the upstream
reservoir boundary.
Table 4.2: Size of the finite elements considered for the harmonic acceleration
loading
CASE Element size, d H/d = λ/d

[m]
1 90 1
2 45 2
3 30 3
4 22.5 4
5 18 5
6 15 6
7 11.25 8
8 9 10
9 7.5 12
10 6 15
11 4.5 20
Figure 4.30 shows the distribution of the peak hydrodynamic pressure on

the face of the dam for different ratios of the wavelength over the element
size, λ/d. The hydrodynamic pressures, Pdyn (y) (at any distance y from
the bottom of the reservoir) are normalised with respect to the maximum
hydrodynamic pressure Pdyn (0), which occurs at the base of the dam, y = 0
(i.e. bottom of the reservoir). The results from the finest mesh (λ/d = 20)
135
Peak Hydrodynamic Pressure Distribution

1
0.9
0.8
λ/d = 5
0.7
λ/d = 1
0.6 λ/d = 2 λ/d = 4
λ/d = 3
0.5 λ/d = 4
λ/d = 1 λ/d = 3
λ/d = 5
0.4 λ/d = 2
λ/d = 6
0.3 λ/d = 8
λ/d = 10
0.2 λ/d = 12
λ/d = 15
0.1
λ/d = 20
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Normalised Hydrodynamic Pressure, Pdyn(y)/Pdyn(0) []
Figure 4.30: Distribution of the peak hydrodynamic pressure on the dam

face for different element sizes
are considered to be the most accurate because a large number of nodes has
been provided to model the wavelength.
As it may be observed from Figure 4.30, the peak hydrodynamic pressure
obtained using elements of size d equal or smaller than a fifth of the wave-
length, λ, deviate significantly from the most accurate prediction (λ/d = 20),
whereas the results from λ/d = 6 and 8, are reasonably close. It is therefore
suggested that for modelling dynamic problems involving water using two-
dimensional eight-noded solid elements, the elements representing the water
domain should be smaller or equal to a fifth of the wavelength.
4.5 Random acceleration

The final part of the investigation considered the behaviour of the reservoir
under random acceleration, i.e. under seismic loading. The bedrock acceler-
ation record of the 15th of November 1975 Mexico earthquake, as recorded
at the rock canyon of La Villita dam (PGA = 0.04g), was used as the input
motion. The processed (filtered and baseline corrected) record is shown in
136
Figure 4.31.
Acceleration record of the 15/11/1975 Mexico earthquake

0.05
15/11/1975 Mexico earthquake
0.04
0.03
0.02
Acceleration, a, [g]
0.01
−0.01
−0.02
−0.03
−0.04
−0.05
0 1 2 3 4 5 6
Time, t [sec]
Figure 4.31: Accelerations recorded at the rock canyon of La Villita dam in

Mexico during the 15/11/1975 Mexico earthquake
4.5.1 Sloped dam

The investigations so far were restricted to stiff dams with a vertical upstream
face. However, earth dams are flexible compared to concrete gravity dams
and have a sloped upstream face. This section investigates the effect of the
slope and the flexibility of the dam on the distance of the upstream reservoir
boundary, i.e. the length of the reservoir domain of the FE mesh.
Analyses were performed with the acceleration record shown in Figure
4.31 for different values of the slope angle and the length, L of the reser-
voir. As before, the acceleration was applied at the bottom of the FE mesh
in the horizontal direction. The Viscous BC was applied on the US reser-
voir boundary and interface elements were placed at the reservoir-dam and
reservoir-foundation interfaces. The slope angle was expressed as the ratio of
the breadth over the height of the slope, B/H as shown in Figure 4.32. For
all the analyses, the height, H of the reservoir was equal to 60m, the width
of the crest, W was 18m and the thickness T of the stiff foundation was 18m.
The values for the various slopes considered are listed in Table 4.3, whereas
137
the corresponding values for the lengths of the reservoir are listed in Table
4.4. The maximum element size in all the analyses was 5m which is equal to
1/12 of the height of the reservoir. For the flexible dam analyses, material
properties were assigned to the dam that correspond to a shear wave velocity,
Vs = 300 m/s, which is considered to be a typical value for earth materials.
These properties are: the shear modulus, Gd = 180000 kPa, the unit weight,
γd = 20 kN/m3 and the Poisson’s ratio, νd = 0.3.
H
T
W B L
Figure 4.32: Geometry of the dam-reservoir system considered for the inves-
tigation of the effect of the slope
Table 4.3: Values of the angle of the dam slope considered.
MESH Slope
B/H
A 0
B 1/3
C 1/2
D 1
E 2
F 3
Figures 4.33 to 4.38 show the distribution of the peak hydrodynamic pres-
sure on the upstream face of a stiff dam for all 6 slope angles and all 10 mesh
lengths considered. The hydrodynamic pressure values are normalised with
respect to ρam H, where ρ is the mass density of water, am = 0.04g is the
maximum value of the applied acceleration and H is the height of the reser-
voir. Likewise, Figures 4.39 to 4.44 show the corresponding distribution of
the peak hydrodynamic pressure for a flexible dam. The results from the
longer reservoir (i.e. the mesh corresponding to L/H = 15, according to
Table 4.4) are considered to be the most accurate because the longer the
138
Table 4.4: Values of the length of the mesh considered.
MESH Length
L/H
1 0.5
2 1
3 2
4 3
5 4
6 5
7 8
8 10
9 12
10 15
distance of the upstream boundary, the smaller the effect on the hydrody-
namic pressures on the dam. Considering the extreme cases, the following
observations can be made:
For a stiff dam with a vertical upstream face, i.e. B/H = 0 (Figure 4.33),
the results for the peak hydrodynamic pressure on the dam for L/H = 0.5,
1, 2, 3, 4, 5 differ from the most accurate solution for L/H = 15. The
results for L/H > 5 plot very close to the most accurate solution. Therefore,
when modelling the pressures on a stiff dam with a vertical upstream face
the upstream boundary (and hence the relevant BC) should be placed at a
distance of more than 5 times the height of the reservoir.
For a stiff dam with a gentle sloped upstream face, like B/H = 3 (Figure
4.38), the results for the peak hydrodynamic pressure on the dam for L/H =
0.5, 1, 2, 3 differ from the most accurate solution for L/H = 15. The results
for L/H > 3 plot be very close to the most accurate solution. Therefore,
when modelling the pressures on a stiff dam with a sloped upstream face
distance of more than 3 times the height of the reservoir.
For a flexible dam with a vertical upstream face, i.e. B/H = 0 (Figure
4.39), the results for the peak hydrodynamic pressure on the dam for L/H =
0.5, 1, 2 differ from the most accurate solution for L/H = 15. The results
for L/H > 2 seem to be very close to the most accurate solution. Therefore,
139
Peak Hydrodynamic Pressure Distribution − Stiff dam: B/H = 0

1
L/H = 0.5
0.9 L/H = 1
L/H = 2
0.8 L/H = 3
L/H = 4
0.7 L/H = 5
L/H = 8
0.6 L/H = 10
L/H = 12
0.5 L/H = 15
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Normalised Hydrodynamic Pressure, Pdyn/ρamH []
Figure 4.33: Distribution of the peak hydrodynamic pressure on the face of a

stiff dam with B/H = 0 (i.e. vertical face) for various lengths of the reservoir
Peak Hydrodynamic Pressure Distribution − Stiff dam: B/H = 1/3

1
L/H = 0.5
0.9 L/H = 1
L/H = 2
0.8 L/H = 3
L/H = 4
0.7 L/H = 5
L/H = 8
0.6 L/H = 10
L/H = 12
0.5 L/H = 15
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Figure 4.34: Distribution of the peak hydrodynamic pressure on the face of

a stiff dam with B/H = 1/3 (i.e. sloped face) for various lengths of the
reservoir
140
Peak Hydrodynamic Pressure Distribution − Stiff dam: B/H = 1/2

1
L/H = 0.5
0.9 L/H = 1
L/H = 2
0.8 L/H = 3
L/H = 4
0.7 L/H = 5
L/H = 8
0.6 L/H = 10
L/H = 12
0.5 L/H = 15
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2

a stiff dam with B/H = 1/2 (i.e. sloped face) for various lengths of the
reservoir

1
L/H = 0.5
0.9 L/H = 1
L/H = 2
0.8 L/H = 3
L/H = 4
0.7 L/H = 5
L/H = 8
0.6 L/H = 10
L/H = 12
0.5 L/H = 15
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

stiff dam with B/H = 1 (i.e. sloped face) for various lengths of the reservoir
141

1
L/H = 0.5
0.9 L/H = 1
L/H = 2
0.8 L/H = 3
L/H = 4
0.7 L/H = 5
L/H = 8
0.6 L/H = 10
L/H = 12
0.5 L/H = 15
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6


1
L/H = 0.5
0.9 L/H = 1
L/H = 2
0.8 L/H = 3
L/H = 4
0.7 L/H = 5
L/H = 8
0.6 L/H = 10
L/H = 12
0.5 L/H = 15
0.4
0.3
0.2
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.6

142
Peak Hydrodynamic Pressure Distribution − Flexible dam: B/H = 0

1
L/H = 0.5
0.9 L/H = 1
L/H = 2
0.8 L/H = 3
L/H = 4
0.7 L/H = 5
L/H = 8
0.6 L/H = 10
L/H = 12
0.5 L/H = 15
0.4
0.3
0.2
0.1
0
−0.1 0 0.1 0.2 0.3 0.4 0.5 0.6

a flexible dam with B/H = 0 (i.e. vertical face) for various lengths of the
reservoir
Peak Hydrodynamic Pressure Distribution − Flexible dam: B/H = 1/3

1
L/H = 0.5
0.9 L/H = 1
L/H = 2
0.8 L/H = 3
L/H = 4
0.7 L/H = 5
L/H = 8
0.6 L/H = 10
L/H = 12
0.5 L/H = 15
0.4
0.3
0.2
0.1
0
−0.1 0 0.1 0.2 0.3 0.4 0.5 0.6

a flexible dam with B/H = 1/3 (i.e. sloped face) for various lengths of the
reservoir
143
Peak Hydrodynamic Pressure Distribution − Flexible dam: B/H = 1/2

1
L/H = 0.5
0.9 L/H = 1
L/H = 2
0.8 L/H = 3
L/H = 4
0.7 L/H = 5
L/H = 8
0.6 L/H = 10
L/H = 12
0.5 L/H = 15
0.4
0.3
0.2
0.1
0
−0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

a flexible dam with B/H = 1/2 (i.e. sloped face) for various lengths of the
reservoir

1
0.9
0.8
0.7
0.6 L/H = 0.5

L/H = 1
0.5 L/H = 2
L/H = 3
0.4
L/H = 4
0.3 L/H = 5
L/H = 8
0.2 L/H = 10
L/H = 12
0.1 L/H = 15
0
−0.1 0 0.1 0.2 0.3 0.4 0.5 0.6

a flexible dam with B/H = 1 (i.e. sloped face) for various lengths of the
reservoir
144

1
0.9
0.8
0.7
0.6 L/H = 0.5

L/H = 1
0.5 L/H = 2
L/H = 3
0.4
L/H = 4
0.3 L/H = 5
L/H = 8
0.2 L/H = 10
L/H = 12
0.1 L/H = 15
0
−0.1 0 0.1 0.2 0.3 0.4 0.5 0.6

reservoir

1
0.9
0.8
0.7
0.6 L/H = 0.5

L/H = 1
0.5 L/H = 2
L/H = 3
0.4
L/H = 4
0.3 L/H = 5
L/H = 8
0.2 L/H = 10
L/H = 12
0.1 L/H = 15
0
−0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 1.2

reservoir
145
when modelling the pressures on a flexible dam with a vertical upstream face
distance of more than 2-3 times the height of the reservoir.
For a flexible dam with a gentle sloped upstream face, like B/H = 3
(Figure 4.44), the results for the peak hydrodynamic pressure on the dam
for L/H = 0.5 and 1 differ from the most accurate solution for L/H =
15. The results for L/H > 1 seem to be very close to the most accurate
solution. Therefore, when modelling the pressures on a stiff dam with a
sloped upstream face the upstream boundary (and hence the relevant BC)
should be placed at a distance of more than 2 times the height of the reservoir.
Considering the above-mentioned observations it is concluded that for
a sloped and flexible dam, which is generally the case for earth dams, the
distance of the absorbing BC may be reduced down to twice or three times
the height of the reservoir, i.e. L/H = 2 - 3.
Combining the results for different slope angles (B/H), it may be ob-
served that for the case of the stiff dam (Figures 4.33 - 4.38), the distribution
of the hydrodynamic pressures over the height of the dam has a similar shape
for all slope angles. This may be observed from Figure 4.45 which includes
the distribution of the peak hydrodynamic pressure on the face of the stiff
dam for all values of B/H considered and for L/H = 15. Moreover, it is
shown that the value of the hydrodynamic pressures reduces with increas-
ing slope angle (B/H), and this confirms the observations of Zangar (1952)
who first investigated the hydrodynamic pressures on sloped dams (see also
Section 2.5).
As far as the case of the flexible dam is concerned (Figures 4.39 - 4.44),
the computed hydrodynamic pressures generally tend to be smaller than the
corresponding pressures for the stiff dam, but not in all the cases. This can
be observed in Figure 4.46 which compares the distribution of the peak hy-
drodynamic pressure on the face of the flexible dam for all values of B/H
considered and for L/H = 15. This is confirmed by centrifuge experiments of
model dams by Saleh and Madabhushi (2010) who observed smaller hydro-
dynamic pressures on a flexible model dam. Besides, the distributions of the
peak hydrodynamic pressures on the dam for different slope angles (B/H) do
146
Peak Hydrodynamic Pressure Distribution − Stiff dam: L/H = 15

1
B/H = 0
0.9 B/H = 1/3
B/H = 1/2
0.8 B/H = 1
B/H = 2
0.7 B/H = 3
0.6
0.5
0.4
0.3
0.2
0.1
0
−0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

a stiff dam for L/H = 15 for various values of the slope angle (B/H)
Peak Hydrodynamic Pressure Distribution − Flexible dam: L/H = 15

1
B/H = 0
0.9 B/H = 1/3
B/H = 1/2
0.8 B/H = 1
B/H = 2
0.7 B/H = 3
0.6
0.5
0.4
0.3
0.2
0.1
0
−0.5 0 0.5 1

a flexible dam for L/H = 15 for various values of the slope angle (B/H)
147
Peak Hydrodynamic Pressure Distribution − B/H = 0, L/H = 15

1
Stiff
0.9 Flexible
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
−0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

stiff and flexible vertical dams (B/H = 0) for L/H = 15
Peak Hydrodynamic Pressure Distribution − B/H = 3, L/H = 15

1
Stiff
0.9 Flexible
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
−0.5 0 0.5 1

stiff and flexible sloped dams (B/H = 3) for L/H = 15
148
not have the same shape. The negative values of the hydrodynamic pressure
imply that the deformation of the dam is dragging the reservoir and there-
fore reduces the total water (hydrostatic + hydrodynamic) pressure (which
is still positive, i.e. compressive). When comparing the values of the pres-
sures for different slope angles, in contrast to the stiff dam, the pressures do
not appear to reduce for increasing slope angles. Figure 4.47 compares the
distributions of the peak hydrodynamic pressures for both stiff and flexible
vertical dams (B/H = 0). Likewise, 4.48 compares the corresponding dis-
tributions for sloped dams (B/H = 3). It is shown that the hydrodynamic
pressures on flexible dams may be smaller or larger than those on stiff dams.
The latter observation may be attributed to the flexibility of the dam
which is a new variable in the study. It is believed that the response of
the reservoir (and hence the hydrodynamic pressures) are affected by the
response of the flexible dam, which is in return affected by the reservoir. This
is the so called reservoir-dam interaction phenomenon (see Section 2.4.5). In
order to better understand and explain the mechanisms of this phenomenon
one has to consider the natural modes of vibration of a dam (see Section 2.4)
and how these are compared to the modes of the reservoir. This of course
requires further study which is presented later in Chapter 5.
4.5.2 Shear stiffness of water

The shear stresses in the water are negligible for Civil Engineering applica-
tions and therefore are considered to be zero. However, a nominal value for
the shear modulus of a material is usually adopted in a numerical analysis
to avoid numerical instability. The critical question is how large can the
value of the shear modulus assigned to water may be, so that the results are
not affected significantly. An unrealistically large value of the shear modulus
would allow propagation of shear waves in the reservoir which does not occur
in reality.
A small parametric study was conducted to investigate the effect of the
value of the shear modulus on the hydrodynamic pressures and identify this
threshold value of the shear modulus of water, Gw . The same model with
149
the interface elements and the Viscous BC for a stiff dam with a vertical
upstream face under the acceleration load shown in Figure 4.31 was analysed
with different values of Gw as listed in Table 4.5.
Table 4.5: Values of the Shear Modulus, Gw assigned to the reservoir water
elements
CASE Shear Modulus % Bulk Modulus

Gw %Kw
[kPa]
1 220000 10
2 110000 5
3 22000 1
4 11000 0.5
5 2200 0.1
6 1100 0.05
7 220 0.01
8 110 0.005
9 22 0.001
10 11 0.0005
11 2.2 0.0001
12 1.1 0.00005
Figure 4.49 shows the distribution of the peak hydrodynamic pressure on

the upstream face of the dam for all 12 analyses with different values of Gw ,
whereas Figure 4.50 shows a detail of the previous figure. It is considered that
the results from the smallest value of the shear modulus (case 12, i.e. Gw =
0.00005%Kw ) are the most accurate. It may be observed that the pressure
distributions arising from the analyses with Gw = 10, 5, 1 and 0.5 %Kw
differ considerably from the most accurate solution (Gw = 0.00005%Kw ) by
predicting significantly smaller pressures at the bottom of the reservoir and
non-zero values at the top of the reservoir. The latter is believed to be due
to the unrealistically high value of shear stiffness assigned to water. On the
other hand, for Gw < 0.1%Kw the results compare very well to the most
accurate solution.
Based on the above observations, it is recommended that the value of the
shear modulus, Gw assigned to water should be kept below the 0.1% of the
150
Peak Hydrodynamic Pressure Distribution − Effect of shear stiffness

1
0.9
Gw = 10% Kw
0.8 Gw = 5% Kw
Gw = 1% Kw
0.7
Gw = 0.5% Kw
0.6
Gw = 0.1% Kw
0.5 Gw = 0.05% Kw
G = 0.01% K
0.4 w w
Gw = 0.005% Kw
0.3
Gw = 0.001% Kw
0.2 G = 0.0005% K
w w
Gw = 0.0001% Kw
0.1
G = 0.00005% K
w w
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Figure 4.49: Distribution of the peak hydrodynamic pressure on the vertical

face of a stiff dam for various values of the shear modulus of the water, Gw
Peak Hydrodynamic Pressure Distribution − Effect of shear stiffness

0.5
0.45 G = 10% K
w w
0.4 Gw = 5% Kw
Gw = 1% Kw
0.35
Gw = 0.5% Kw
0.3 Gw = 0.1% Kw
0.25 Gw = 0.05% Kw
G = 0.01% K
0.2 w w
G = 0.005% K
w w
0.15 Gw = 0.001% Kw
0.1 Gw = 0.0005% Kw
Gw = 0.0001% Kw
0.05
G = 0.00005% K
w w
0
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Normalised Hydrodynamic Pressure, P /ρa H []
dyn m

face of a stiff dam for various values of the shear modulus of the water, Gw
- detail
151
value of the bulk modulus, Kw .
4.5.3 Damping of water

In reality, water has a specific value of viscosity (dynamic viscosity of water,
µw = 8.9 · 10−4 Pa · s at a temperature of 25 o C) which damps the dynamic
response and therefore any possible resonance does not lead to infinite values
of displacements and pressures. In numerical modelling, a finite value for
damping should be used in order to avoid any unrealistic values originating
from spurious resonance.
Similar to the case of the shear stiffness, the aim was to identify an
appropriate value of the damping assigned to water so that the response
is not affected significantly in cases without any resonance. The analyses
described in the previous section were repeated with different values of ξw as
listed in Table 4.6. To introduce damping, the Rayleigh damping formulation
was employed and the values listed in Table 4.6 correspond to the target
damping, ξt (see Section 3.7).
The circular frequencies, ω1 and ω2 , between which the induced damping
is close to the value of the target damping, ξt , were chosen so that they
include the dominant frequency of the load. Figure 4.51 shows the Fourier
Amplitude Spectrum of the acceleration record used in the analysis. This
figure shows that the two dominant frequencies of the record are fd = 2 and
6 Hz, and therefore the dominant circular frequencies are ωd = 12.57 and
37.7 rad/s. For this reason, the Rayleigh damping frequencies (see Section
3.7) used are ω1 = 9.7 rad/sec (= ω1r /4) and ω2 = 194.1 rad/sec (= ω3r ).
The circular frequencies ω1r and ω3r correspond to the first and the third
fundamental modes of vibration of the reservoir respectively.
Figure 4.52 shows the distribution of the peak hydrodynamic pressure
on the upstream face of the dam for all 8 analyses with different values of
damping ξw considered, whereas Figure 4.53 shows a detail of the previous
figure. It is considered that the results from the smallest value of the damping
(case 8, i.e. ξw = 0.1%) are the most appropriate as the damping introduced
is minimal. The pressure distributions arising from the analyses with ξw =
152
Table 4.6: Values of the damping, ξw assigned to the reservoir water elements
CASE Damping
ξw
[%]
1 20
2 10
3 5
4 2
5 1
6 0.5
7 0.2
8 0.1
Fourier Amplitude Spectrum − 15/11/1975 Mexico earthquake

1.6
1.4
1.2
Fourier Amplitude
0.8
0.6
0.4
0.2
0
−1 0 1
10 10 10
Frequency, f [Hz]
Figure 4.51: Fourier Amplitude Spectrum of the acceleration record
153
Peak Hydrodynamic Pressure Distribution − Effect of damping

1
ξ = 20%
0.9 ξ = 10%
ξ = 5%
0.8
ξ = 2%
ξ = 1%
0.7
ξ = 0.5%
0.6 ξ = 0.2%
ξ = 0.1%
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

face of a stiff dam for various values of the damping of the water, ξw
Peak Hydrodynamic Pressure Distribution − Effect of damping

0.5
ξ = 20%
0.45 ξ = 10%
ξ = 5%
0.4
ξ = 2%
ξ = 1%
0.35
ξ = 0.5%
0.3 ξ = 0.2%
ξ = 0.1%
0.25
0.2
0.15
0.1
0.05
0
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2

face of a stiff dam for various values of the damping of the water, ξw - detail
154
20, 10, 5 and 2% differ considerably from the most accurate solution (ξw =
0.1%), as they result in smaller pressures. This is not surprising as the high
values of ξw damp significantly the response. Additionally, for ξw < 1%
the results compare very well with the accurate solution. It is therefore
recommended that the value of damping, ξw assigned to water should be
kept close to 1%.
4.6 Conclusions
This chapter describes the investigation carried out in order to numerically
model the reservoir hydrodynamic pressures on dams. The main aim was
to establish an approach/methodology to appropriately model the reservoir
with displacement-based solid elements in FE analysis. The same approach
may be used in analyses of other waterfront structures such as quay walls.
The main conclusions of this study may be summarised as follows:
• As far as the upstream absorbing boundary of the reservoir is con-

cerned, the Viscous and Cone Boundary Conditions can be employed
as they perform satisfactorily and they have almost identical response.
Both the BCs perform well for a range of loading frequencies and their
response is almost identical.
• At the dam-reservoir and foundation-reservoir interfaces, interface ele-

ments should be used in order to allow relative movement between the
water and the solid material. Appropriate values for the normal and
shear stiffnesses of these elements should be used.
• The reservoir water should not be treated as incompressible in order

not to underestimate resonant amplifications occuring under loading
frequencies at the vicinity of the fundamental frequencies of the reser-
voir and not to overestimate the pressures for higher frequency loads.
• The size of the reservoir elements should be smaller than a fifth of the
acoustic (p-wave) wavelength, λ of the water.
155
• The distance, L of the upstream reservoir boundary from the upstream

face of the dam may be reduced for a sloped and flexible dam (which
is the case for earth dams) down to two to three times the height, H,
of the reservoir: L = 2 - 3 H.
• The value of the shear stiffness assigned to water, Gw should be smaller

than 0.1% of the real value of the bulk modulus of the water, Kw .
• The viscous (Rayleigh) damping assigned to water, ξw should be kept

close to 1% of the critical. Also Rayleigh damping circular frequencies,
ω1 and ω2 should include the dominant frequency of the load.
156
Chapter 5
DYNAMIC
RESERVOIR-DAM
INTERACTION
5.1 Introduction
The dynamic behaviour of a dam with a full reservoir is different from the
behaviour of a dam with an empty reservoir. This is because the motion of the
reservoir affects the motion of the dam and vice-versa. This phenomenon is
called dynamic reservoir-dam interaction and it could be catastrophic in cases
of resonance, i.e. when the two domains (dam and reservoir) are vibrating
in phase.
Previous studies of reservoir-dam interaction, such as Chopra (1967b,
1968) and Hall and Chopra (1980, 1982a,b,c) (see Section 2.4.5 for a more
detailed discussion) showed that the effects of reservoir-dam interaction are
more pronounced in concrete dams than in earth dams. The effects of this in-
teraction are mainly concentrated on (a) the fundamental period of vibration
of the reservoir-dam system and (b) the magnitude of the dynamic response
of the dam. It is believed that reservoir-dam interaction firstly causes the
reservoir-dam system to soften, i.e. its fundamental period elongates, and
secondly it alters the dam’s response, i.e. it amplifies or de-amplifies the seis-
157
mic motion and results in larger or smaller accelerations at the dam crest.
This chapter describes work carried out as part of this research project
to further investigate reservoir-dam interaction and its influence on the dy-
namic acceleration response of dams. The main aim is to examine the effects
of reservoir-dam interaction for different dynamic characteristics of the load,
the dam and the reservoir. In this study, two types of dams are considered
for this purpose. Firstly, a rectangular cantilever dam is investigated which
has a vertical upstream face (and a small mass because of its slender ge-
ometry). Secondly, a trapezoidal wide earth dam is considered which has
a sloped upstream face and a larger mass because of its wide trapezoidal
geometry. The response of both dams is investigated under harmonic and
random (seismic) acceleration loads. In all the performed analyses, both the
dam and the reservoir rest on a stiff undeformable ground and therefore the
effects of reservoir-dam-foundation interaction are not considered.
5.2 Statement of the problem

The problem under study is shown in Figure 5.1. A dam (1) rests on the
ground (here, a stiff and undeformable foundation) (3) and retains a large
amount of water in the reservoir (2). Under seismic (or general dynamic)
conditions, an acceleration motion from the ground (such as at point A)
causes movement of both the dam structure and the reservoir.
B
1 2
A
3
Figure 5.1: Geometry of the dam-reservoir-foundation system: (1) Dam, (2)
Reservoir, (3) Foundation.
The critical question is how the motion of each domain (dam structure
and reservoir) will be affected by the motion of the neighbouring domain, i.e.
what is the effect of reservoir-dam interaction, especially on the accelerations
158
of the dam crest (point B). Therefore, in this investigation, acceleration input
motions were applied at the dam base (A) and the accelerations at the crest
of the dam (B) were monitored and examined for different cases, with a full
and an empty reservoir.
5.3 Cantilever dam

The first case considered a simple rectangular cantilever dam. The geometry
of the cantilever dam is shown in Figure 5.2. The region A-B-C-D represents
the upstream reservoir, A-B-F-E represents the cantilver dam and F-G-H-C
represents the foundation. The height of the dam was taken as, H=60m,
the width of the dam was W =18m, the length of the reservoir, L=300m and
the thickness of the foundation, T =18m. The reservoir length to height ratio
was taken as, L/H = 5 according to the suggestions of Chapter 4 for stiff
dams with a vertical upstream face.
E A D
Reservoir
H Flexible dam
Stiff foundation
F B C
T G H
W L
Figure 5.2: Geometry of the cantilever dam considered.
5.3.1 Dam with empty reservoir

Initially, a cantilever dam with an empty reservoir was analysed. Therefore,
the reservoir domain (A-B-C-D) was not included in the analysis.
The material properties of the dam were as follows: elastic modulus, E
= 35000000 kPa, Poisson’s ratio, ν=0.3 and mass density, ρ=2500 kg/m3 .
159
Damping, ξ, of the Rayleigh type (see Section 3.7.5) was specified in the
dam with a value of 5%. The values for the two natural circular frequencies
of Rayleigh damping, ω1 and ω2 , were taken as equal to the fundamental
circular frequency of the dam, ωd , and the circular frequency of the harmonic
load, ω (Equation 5.1). The foundation was modelled to behave as rigid
(i.e. extremely stiff) and a high value of the bulk modulus was assigned,
K = 108 · Kw = 2.2 · 1014 kPa (where, Kw = 2.2 · 106 kPa is the bulk modulus
of water).
The applied boundary conditions include: zero displacements specified in
the vertical direction along the bottom boundary (G-H), whereas prescribed
values of harmonic acceleration were specified in the horizontal direction,
as given by Equation 5.1. Different values of circular frequency, ω, were
considered, for constant amplitude a0 = 1 m/s2 and for 40 cycles, in order
to reach a steady-state response.
a(t) = a0 cos ωt (5.1)
where, a is the input acceleration, a0 is the amplitude of the harmonic

load, ω is the circular frequency of the harmonic loading and t is the time.
Figure 5.3 shows the amplification spectrum, i.e. the amplification of the
accelerations at the dam crest, |F |, with respect to the ratio of the circular
frequency of the harmonic load to the fundamental circular frequency of the
dam, ω/ωd . The amplification, |F |, is defined as the ratio of the amplitude
of crest acceleration over the amplitude of the input (at the base, G-H)
acceleration. It is shown that the amplification becomes maximum at the
values of ω/ωd = 1 and 4.7, which correspond to the first and second natural
modes of vibration of the dam respectively.
It should be noted that the analytical value of the fundamental circular
frequency of the cantilever dam given by Equation 5.2 (Chopra, 1995), based
on Euler-Bernoulli beam theory, was found to be larger (by 1.16 times) than
the calculated value from the FE analysis.
s
3.516 EI
ωd = (5.2)
H2 m/H
160
where, H is the height of the cantilever dam, E is the elastic modulus, I

is the moment of inertia and m is the mass of the cantilever dam.
This could be attributed to the stocky geometry (H/W =3.333) of the
examined cantilever in which shear effects are significant and which the Euler-
Bernoulli theory does not take into account (Timoshenko and Goodier, 1934).
A better approximation could be obtained by using the Timoshenko (1921,
1922) beam formulation which is an advancement of the Euler-Bernoulli beam
that takes into account shear deformation and rotational inertia effects and
hence it results in less stiff response, i.e. smaller natural frequencies (Chopra,
1995; Han et al., 1999).
Amplification Spectrum − Cantilever Dam

16
Empty reservoir
14
12
Amplification, |F|
10
0
0 1 2 3 4 5 6 7 8
Frequency ratio, ω/ωd
Figure 5.3: Amplification spectrum of the cantilever dam with an empty

reservoir.
5.3.2 Dam with full reservoir

Subsequently, a cantilever dam with a full reservoir was analysed, including
the reservoir domain (A-B-C-D) in the FE mesh.
The reservoir water was assumed to behave as a linear material with a
bulk modulus, Kw = 2.2 · 106 kPa (the exact value for water) and a nom-
inal value of shear modulus, Gw = 100 kPa (see Chapter 4). Again, the
161
foundation was modelled to behave as rigid, as before.

The material properties of the dam were the same as before, apart from
the value of the elastic modulus which was altered to provide some specific
values of the ratio of the fundamental circular frequency of the reservoir over
the fundamental circular frequency of the dam, ωr /ωd , over the range of 0.25
to 4. The fundamental circular frequency of the reservoir, ωr is given by
Equation 5.3 (Chopra, 1967b) (where, Vp = 1483 m/s is the p-wave velocity
of water).
Vp
ωr = 0.25 (5.3)
H
Interface elements were placed along the interface of the dam and the
reservoir (A-B) and the interface of the reservoir and the foundation (B-
C). The values assigned for the normal and shear stiffness of the interface
elements were KN = 108 kN/m and KS = 1 kN/m respectively, as discussed
in Chapter 4.
The same boundary conditions as in the case of the dam with an empty
reservoir were applied on the bottom boundary (G-H). The horizontal bound-
ary condition applied on the upstream reservoir boundary (C-D) was the
standard viscous boundary (see Section 3.7) and the values of the viscosities
of the dashpots were obtained from the elastic properties of the material ad-
jacent to the boundary to which the dashpots were applied, i.e. water (see
also Chapter 4). Finally, zero displacements were specified in the vertical
direction on the upstream reservoir boundary.
Figure 5.4 shows the amplification of the accelerations at the dam crest,
|F | with respect to the ratio of the circular frequency of the harmonic load
to the fundamental circular frequency of the dam, ω/ωd , for various values
of the ratio of the fundamental circular frequency of the reservoir to the
fundamental circular frequency of the dam, ωr /ωd . On the same figure, the
corresponding amplification of a dam with an empty reservoir (Figure 5.3) is
plotted with a dashed line for comparison.
The amplification spectrum (|F | − ω/ωd ) follows a similar trend for all
the values of the examined ωr /ωd ratio. Namely, there are generally two peak
162
ωr/ωd
2
0
78
6
ω/ωd
5
4
2 3
1
0
5
0
25
20
15
10
Amplification, |F|
Figure 5.4: Amplification of acceleration, |F | at the crest of the cantilever

dam for various values of the frequency ratios ω/ωd and ωr /ωd .
163
Amplification Spectrum − Cantilever Dam

25
Full reservoir − ωr/ωd=1
Empty reservoir
20
Amplification, |F|
15
10
0
0 1 2 3 4 5 6 7 8
Figure 5.5: Amplification spectrum of the cantilever dam with a full (ωr /ωd =
1) and an empty reservoir.
Amplification Peaks − Cantilever Dam

4
3.5
First Second
mode mode
3
Frequency ratio, ωr/ωd
2.5
1.5
Full reservoir
1
Empty reservoir
0.5
0
0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Figure 5.6: Amplification peaks for the cantilever dam with a full and an
empty reservoir.
164
Amplification − Cantilever Dam − First mode

25
20
Amplification, |F|
15
10
5 Full reservoir
Empty reservoir
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Figure 5.7: Amplification of acceleration, |F | for the cantilever dam with

respect to the ratio of the fundamental circular frequency of the reservoir
over that of the dam for the first mode of vibration, ωr /ωd .
Amplification − Cantilever Dam − Second mode

16
14 Full reservoir
Empty reservoir
12
Amplification, |F|
10
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Figure 5.8: Amplification of acceleration, |F | for the cantilever dam with

respect to the ratio of the fundamental circular frequency of the reservoir
over that of the dam for the second mode of vibration, ωr /ωd .
165
values of the amplification as observed in Figure 5.3 for the empty reservoir
case. However, the magnitude of the peak values of the amplification and the
value of the ω/ωd ratio at which they occur is different for different values of
the ωr /ωd ratio.
Moreover, it is shown that the maximum value of amplification occurs in
the region where ω/ωd ≈ ωr /ωd ≈ 1, i.e. where ω ≈ ωd ≈ ωr . This is due
to reasonance between the harmonic load, the dam and the reservoir. There
are also large values of amplification for ω/ωd ≈ ωr /ωd (shown diagonally in
the figure), i.e. where ω ≈ ωr . The value of amplification for the latter case
(ω ≈ ωr ) can be larger than the amplification corresponding to the second
mode of vibration (i.e. close to ω/ωd ≈ 4.7). However, it seems that in some
cases (e.g. close to ω/ωd ≈ ω/ωr ≈ 3) there is a combined effect of (a) ω ≈
ωr and (b) the second natural mode of vibration of the dam.
Figure 5.5 shows a comparison of the amplification spectrum (|F | - ω/ωd )
for a dam with an empty reservoir (dashed line) and a dam with a full
reservoir (solid line), for which ωr /ωd = 1, i.e. the fundamental circular
frequency of the reservoir, ωr , is equal to the fundamental circular frequency
of the dam, ωd , namely where high amplification is expected. It may be
observed that the reservoir-dam interaction results in higher ampification for
the first natural mode of vibration, but smaller amplification for the second
natural mode. Moreover, the maximum values of amplification (for both the
first and second natural modes of vibration) occur at smaller values of the
frequency ratio ω/ωd . The harmonic load (ω) resonates with the reservoir-
dam system (RDS) at a smaller frequency than that of the fundamental
frequency of a dam, ωd , with an empty reservoir. This implies that the
fundamental frequency of vibration of the dam-reservoir system is smaller
(and hence its fundamental period is larger) than that of a dam with an
empty reservoir.
Figure 5.6 shows the value of the ratio ω/ωd for which the maximum
amplification occurs (for both of the first two natural modes), for various
values of the ratio ωr /ωd . On the same figure, the corresponding values of
ω/ωd for an empty reservoir are included for comparison. It is shown that for
both of the first two natural modes of vibration, and for all ωr /ωd ratios, the
166
maximum amplification occurs at an equal or lower value of the ω/ωd ratio

at resonance for the case of empty reservoir, as observed earlier in Figure 5.4.
This means that for most of the values of the ωr /ωd ratio, both the first and
second natural periods of a dam-reservoir system are larger than those of a
dam with an empty reservoir. Moreover, the variation in the fundamental
periods of the RDS depends on the ratio ωr /ωd . As far as the first natural
mode is concerned, it is shown that the deviation from the empty reservoir
case is larger close to ωr /ωd ≈ 1, i.e. when the fundamental period of the
dam and the reservoir are close. In addition, regarding the second natural
mode, the period of the RDS increases close to ωr /ωd ≈ 2 ∼ 3, i.e. close
to the second frequency of vibration of the reservoir (which is 3 times its
first natural frequency (Chopra, 1967a)). Finally, it is shown that maximum
values of amplification occur for the cases of ωr /ωd ≈ ω/ωd as observed
earlier in Figure 5.4. It should be noted however that the peak values of
amplification, |F | at ω/ωd ≈ 2 ∼ 4 could be a combination of the effects of
(a) ω/ωd ≈ ωr /ωd (i.e. resonance between the load and the reservoir) and
(b) the second natural mode of vibration of the dam.
Finally, Figures 5.7 and 5.8 show the maximum value of amplification with
respect to the ωr /ωd ratio for the first and second natural modes of vibration
respectively. The corresponding values of amplification for an empty dam are
shown also on the same figure with dashed lines for comparison. Regarding
the first natural mode (Figure 5.7), it is shown that the maximum value of
amplification of a dam-reservoir system (i.e. a dam with a full reservoir) is
larger than that of a dam with an empty reservoir, for values of the ωr /ωd
ratio larger than approximately 1. This means that a dam with a full reservoir
experiences higher amplification of the first natural mode than a dam with
an empty reservoir if the fundamental frequency of the reservoir is larger
than that of the dam (and smaller amplification for ω/ωd < 1). Similar
observations can be made regarding the second natural mode (Figure 5.8).
The amplification of a dam with a full reservoir is larger than that of a dam
with an empty reservoir for ωr /ωd larger than approximately 3, namely when
the second natural period of the reservoir is larger than the fundamental
period of the dam.
167
5.3.3 Comments
For the examined case of a cantilever dam, the effects of dam-reservoir inter-
action are mainly related to changes of the fundamental period of the dam
and the amplification of its response. It is found that a dam with a full reser-
voir has a larger fundamental period, Td (i.e. the presence of the reservoir
increases the fundamental period of the dam and the latter exhibits a softer
response). This may be attributed to the added mass, m, of the reservoir
which vibrates with the dam, while it does notqprovide any additional (shear
or bending) stiffness, k to the dam (Td = 2 π m/k).
The amplification of the input acceleration can be either higher or lower
than the amplification in a dam with an empty reservoir. This depends on
the relative magnitude of the circular frequency of the load, ω, and the funda-
mental circular frequencies of the dam and reservoir, ωd and ωr respectively.
The effects of reservoir-dam interaction are more pronounced close to the
frequency ratios ω/ωd ≈ ωr /ωd ≈ 1, i.e. when there is resonance between
the load, the dam and the reservoir. Besides, high values of amplification
are observed for ω/ωd ≈ ωr /ωd , i.e. when the circular frequency of the load
is close to the fundamental circular frequency of the reservoir (ω ≈ ωr ), i.e.
when there is resonance between the load and the reservoir.
The studied case of reservoir-dam interaction effects on the dynamic re-
sponse of a cantilever dam structure could be relevant to thin slender arch
dams which have a similar cross-sectional geometry. Moreover, this study
could also provide some insight into the dynamic behaviour of canal locks,
which again have a slender geometry.
Finally, in comparing the results of this study with those of Chopra (1968)
(Figure 2.32), the following points should be noted:
• Chopra’s 3D figure shows generally only one peak value in the F −ω/ωd
curves, whereas this work (Figures 5.2 and 5.4) shows three peak values
which correspond to the first three natural modes of vibration of the
dam. This is because the work of Chopra (1968) considered a dam
deforming in its fundamental mode only.
• Chopra’s 3D figure shows another peak value of amplification which
168
corresponds to higher values of ω/ωd for increasing ωr /ωd and this is

in agreement to the results of this work (Figure 5.4) which exhibits the
same.
• A number of negative peaks (localised minima) appear in Chopra’s

figure, which form straight lines (i.e. a linear relationship between
ω/ωd and ωr /ωd ). Such negative peaks were not observed in this work.
This may be attributed to the fact that the points considered in this
study may not be adequately dense to show any possible negative peaks
in the amplification.
• From the 3D figure of Chopra it is not possible to observe clearly

whether the high peaks are moving with respect to ω/ωd for differ-
ent values of ωr /ωd . Therefore, direct comparison regarding the period
softening of the reservoir-dam system cannot be made. Such a compar-
ison would be possible should a plot similar to Figure 5.6 was available.
• The magnitude of amplification cannot be directly compared, as the

results of Chopra (1968) refer to 10% damping, whereas the corre-
sponding value used in this work was 5%. However, the results of
both studies are in comparative agreement, as the amplification in this
work (around 20, see Figure 5.7) is higher than that of Chopra (1968)
(around 15, see Figure 2.32).
5.4 Earth dam

The second case considered a simple trapezoidal earth dam. The geometry
of the examined earth dam is shown in Figure 5.9. The region A-B-C-D
represents the upstream reservoir, A-B-F-E represents the earth dam and
F-G-H-C represents the foundation. The height of the dam was taken as
H=60m, the width of the crest of the dam W =18m, the breadth of the dam
shoulders B=180m, the length of the reservoir L=300m and the thickness of
the foundation T =18m. Note that the slope ratio, B/H = 1/3. The L/H
169
ratio was taken as equal to 5 according to the suggestions of Chapter 4 for

trapezoidal earth dams.
E A D
H Reservoir
Flexible
dam B C
F
T G Stiff foundation H
W B L
Figure 5.9: Geometry of the triangular earth dam considered.
5.4.1 Dam with empty reservoir

Firstly, a trapezoidal earth dam with an empty reservoir was analysed.
Therefore, the reservoir domain (A-B-C-D) was not included in the anal-
ysis.
The material properties of the dam were as follows: shear wave veloc-
ity, Vs = 300 m/s, Poisson’s ratio, ν=0.3 and mass density, ρ=2000 kg/m3 .
Damping, ξ of the Rayleigh type (see Section 3.7.5) was specified in the dam
with a value of 5%. The values for the two circular frequencies, ω1 and
ω2 were taken as equal to the fundamental circular frequency of the dam,
ωd and the circular frequency of the harmonic load, ω (Equation 5.1). The
foundation was again modelled to behave as rigid, as before.
The boundary conditions were the same as the cantilever dam case:
zero displacements were specified in the vertical direction along the bottom
boundary (G-H), whereas prescribed values of harmonic acceleration were
specified in the horizontal direction as described by Equation 5.1. Again,
different values of circular frequency, ω were considered, for constant am-
plitude, a0 = 1 m/s2 and for 40 cycles, in order to reach a steady-state
response.
Figure 5.10 shows the amplification spectrum for the trapezoidal earth
dam, i.e. the amplification of the accelerations at the dam crest (Point
B), with respect to the ratio of the circular frequency of the harmonic load
to the fundamental circular frequency of the dam, ω/ωd . It is shown that
170
the amplification becomes maximum at ω/ωd = 1, 2 and 3.9, values which

correspond to the first, second and third natural modes of vibration of the
dam respectively.
It should be noted that the value of the fundamental circular frequency of
the dam calculated using the analytical shear beam (SB) relation (Equation
5.4) (Ambraseys, 1960a) (see Section 2.4) was found to be larger (by 1.21
times) than the corresponding value obtained from the finite element (FE)
analysis. This is in agreement with earlier work of Chopra (1966, 1967a) and
Tsiatas and Gazetas (1982) who observed that SB analysis predicts stiffer
response (i.e. smaller fundamental period of vibration) than FE analysis,
since the SB method considers only shear response and neglects bending
deformations (see also discussion in Section 3.6).
Amplification Spectrum − Earth Dam

18
Empty reservoir
16
14
12
Amplification, |F|
10
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Figure 5.10: Amplification spectrum of the earth dam with an empty reser-
voir.
2π 2π πVs πVs
ωd = = = 0.765 (5.4)
Td 2.613 H H
171
5.4.2 Dam with full reservoir

Subsequently, a dam with a full reservoir was analysed, including the reservoir
domain (A-B-C-D) in the FE mesh.
As with the cantilever dam case, the reservoir water was assumed to be-
have as a linear material and the foundation was modelled to behave as rigid.
The material properties (mass density, ρ, Poisson’s ratio, ν, and Rayleigh
damping, ξ) of the dam were the same as before, apart from the shear wave
velocity, Vs , which was altered to obtain specific values of the ratio of the
fundamental circular frequency of the reservoir over the fundamental circular
frequency of the dam, ωr /ωd , within a range of 0.25 to 4. Again, interface
elements were placed at the interface of the dam and the reservoir (A-B) and
the interface of the reservoir and the foundation (B-C). Finally, the same
boundary conditions were applied on the bottom (F-G) and the upstream
reservoir (C-D) boundaries as with the previous case of the cantilever dam.
Figure 5.11 shows the amplification of acceleration, |F |, at the dam crest
with respect to the ratio of the circular frequency of the harmonic load to
the fundamental circular frequency of the dam, ω/ωd for various values of
the ratio of the fundamental circular frequency of the reservoir to the fun-
damental circular frequency of the dam, ωr /ωd . The amplification of a dam
with an empty reservoir is also plotted with a dashed line for comparison.
Similar to the case of the cantilever dam, the amplification spectrum
(|F | − ω/ωd ) follows a similar trend for all the values of the ωr /ωd ratio.
Namely, for each ωr /ωd ratio three main peaks can be identified in the am-
plification response which correspond to the three first modes of vibration
of the dam. In contrast to the case of the cantilever dam, the magnitude
of the peak values of the amplification and the value of the ω/ωd ratio at
which they occur does not seem to depend significantly on the ωr /ωd ratio.
Overall, there are very small differences between the amplification spectra
for different values of the ωr /ωd ratio. This suggests that the presence of
the reservoir does not affect significantly the amplification (and hence the
dynamic response) of the dam, for any value of the ωr /ωd ratio.
Figure 5.12 shows a comparison of the amplification spectrum (|F | - ω/ωd )
172
ωr/ωd
3
2
1
0
4.5
4
3.5
d
ω/ω
2 3
2.5
0.5 1.5
1
0
8
0
18
16
14
12
10
Amplification, |F|
Figure 5.11: Amplification of acceleration, |F | at the crest of the earth dam

for various values of the frequency ratios ω/ωd and ωr /ωd .
173
Amplification Spectrum − Earth Dam

18
Full reservoir − ωr/ωd=1
16 Empty reservoir
14
12
Amplification, |F|
10
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Figure 5.12: Amplification spectrum of the earth dam with a full (ωr /ωd = 1)
and an empty reservoir.
Amplification Peaks − Earth Dam

4
3.5
First Second Third
mode mode mode
3
2.5
1.5
Full reservoir
1
Empty reservoir
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Figure 5.13: Amplification peaks for the earth dam with a full and an empty
reservoir.
174
Amplification − Earth Dam − First mode
16
14
12
Amplification, |F|
10
4 Full reservoir
Empty reservoir
2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Figure 5.14: Amplification of acceleration, |F | for the earth dam with respect
to the ratio of the fundamental circular frequency of the reservoir over that
of the dam, ωr /ωd , for the first mode of vibration.
Amplification − Earth Dam − Second mode

8
6
Amplification, |F|
2
Full reservoir
Empty reservoir
1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
of the dam, ωr /ωd , for the second mode of vibration.
175
Amplification − Earth Dam − Third mode

5
Full reservoir
4.5 Empty reservoir
3.5
Amplification, |F|
2.5
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
of the dam, ωr /ωd , for the second mode of vibration.
for a dam with an empty reservoir (dashed line) and a dam with a full
reservoir (solid line), for which ωr /ωd = 1. In contrast to the cantilever dam
case (Figure 5.4), the interaction does not have a significant impact on the
amplification spectrum. The amplification values are very similar for the two
cases (full and empty reservoir) and their maxima occur for almost identical
values of the ω/ωd ratio. This means that in contrast to the case of the
cantilever dam, the presence of the reservoir has a minimal impact on both
the amplification and the fundamental period of the dam.
Figure 5.13 shows the value of the ω/ωd ratio for which the maximum
amplification occurs (for the first three natural modes), for various values of
the ωr /ωd ratio. In the same figure, the values of ω/ωd for an empty reservoir
are included for comparison. It is shown that generally for all three natural
modes of vibration, and for all ωr /ωd ratios, the maximum amplification
occurs very close to the value of the ω/ωd ratio of the case of the empty
reservoir, which again suggests that the presence of the reservoir does not
affect the natural periods of the dam significantly.
Finally, Figures 5.14, 5.15 and 5.16 show the maximum amplification with
176
respect to the ωr /ωd ratio for the first, second and third natural modes of
vibration respectively. The amplification response for an empty reservoir is
also shown on the same figure with dashed lines for comparison. Referring to
all three modes of vibration, the amplification for the case of a full reservoir
seems to be very close to the case of an empty reservoir. However, some
small differences can still be observed which show that the case of the full
reservoir generally results in slightly smaller amplification values than those
of the empty reservoir.
5.4.3 Comments
For the examined case of a trapezoidal earth dam, the effects of dam-reservoir
interaction were found to be insignificant. In contrast to the previous case
of a cantilever dam, changes of both the fundamental period of the dam and
the amplification of its response were still observed but they were very small.
This is believed to be due to (a) the sloped upstream face and (b) the large
volume of a trapezoidal earth dam. As far as the former reason is concerned,
it is long known (Zangar, 1953) that the hydrodynamic pressures induced
on a sloped upstream face are smaller than those on a vertical face (see also
Section 2.5). Moreover, regarding the second reason, the inertial effects from
the additional mass from the reservoir are small compared to the inertia of a
large earth dam. This is due to the large volume (and hence mass) of earth
dams and therefore the influence of the reservoir on the dynamic response of
the dam seems to be minimal.
A further study including trapezoidal dams with more gentle slopes could
examine whether the above-mentioned conclusions apply to other dam slopes.
That could perhaps identify for which values of the slope angle the interaction
effects become significant.
5.5 Application with seismic loading

The investigation presented above was restricted to a harmonic load shaking
a reservoir-dam system. However, since in reality dams are subjected to
177
transient shaking, this section is concerned with seismic loading of reservoir-

dam systems. The horizontal acceleration record of the El Centro (Imperial
Valley, 1940) earthquake was used as the input motion. The acceleration
time history of this motion is shown in Figure 5.17, and the corresponding
5% damped response spectrum is shown in Figure 5.18. As it is shown in
the latter figure, there are two values of the period for which high values of
spectral acceleration, Sa occur (i.e. 0.27s and 0.52s). The period for which
the maximum value of Sa occurs is considered to be the dominant period of
the seismic load and it is TEQ = 0.52s.
Following the previous investigation for harmonic loading, both a can-
tilever and a trapezoidal earth dam were again considered. For each of the
dam types, five cases were examined for various combinations of the fre-
quency ratios ω/ωd and ωr /ωd , as listed in Table 5.1 and shown graphically
in Figure 5.19. These combinations were chosen based on the observations
of the previous sections concerning harmonic loading, i.e. large amplification
of accelerations is expected for Cases A-D, whereas smaller amplification of
the accelerations is expected for Case E.
5.5.1 Cantilever dam

For the case of the cantilever dam, the FE mesh used earlier (its geometry is
shown in Figure 5.2) was again employed. The imposed boundary conditions
were the same as those described in Section 5.3. All five cases (Table 5.2)
were analysed twice, with a full and an empty reservoir.
In all cases, the material properties of the cantilever dam were: Poisson’s
ratio, ν = 0.3, mass density, ρ = 2500 kg/m3 and Rayleigh damping, ξ =
5%. The elastic modulus, E, was varied for each case as listed in Table 5.2,
to achieve the combinations of frequencies listed in Table 5.1. The reservoir
water was again modelled as a linear material with the same values for the
bulk and shear modulus as before (Section 5.3).
Figures 5.20 - 5.24 show the acceleration time-histories at the crest of the
dam (for both full and empty reservoirs), whereas Figures 5.25 - 5.29 show
the corresponding 5% damped response spectra. It may be observed that
178
Acceleration time−history of input motion

0.4
0.3
0.2
Acceleration, a [g]
0.1
−0.1
−0.2
−0.3
0 10 20 30 40 50 60
Time, t [sec]
Figure 5.17: Acceleration time-history of the El Centro (1940) earthquake

record.
Response spectrum of input motion

1
0.9
0.8
Spectral Acceleration, Sa [g]
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
Figure 5.18: Response spectrum of the El Centro (1940) earthquake record

for damping, ξ = 5%.
179
Table 5.1: Cases of combinations of ω/ωd and ωr /ωd frequency ratios exam-
ined under seismic loading.
CASE ω/ωd ωr /ωd

A 1 1
B 3 3
C 1 3
D 1/3 1
E 1/3 1/3
Frequency ratio combinations

4
A
3.5 B
C
D
3 E
2.5
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Figure 5.19: Cases of combinations of ω/ωd and ωr /ωd frequency ratios ex-
amined under seismic loading.
180
Accelerations − Cantilever dam − Case A

1.5
Full reservoir
Empty reservoir
1
0.5
Acceleration, a [g]
−0.5
−1
−1.5
−2
0 10 20 30 40 50 60
Time, t [sec]
Figure 5.20: Acceleration time-history at the crest of the cantilever dam -

Case A.
Accelerations − Cantilever dam − Case B

1
Full reservoir
0.8 Empty reservoir
0.6
0.4
Acceleration, a [g]
0.2
−0.2
−0.4
−0.6
−0.8
0 10 20 30 40 50 60
Time, t [sec]

Case B.
181
Accelerations − Cantilever dam − Case C

2.5
Full reservoir
2 Empty reservoir
1.5
1
Acceleration, a [g]
0.5
−0.5
−1
−1.5
−2
−2.5
0 10 20 30 40 50 60
Time, t [sec]

Case C.
Accelerations − Cantilever dam − Case D

2
Full reservoir
1.5 Empty reservoir
0.5
Acceleration, a [g]
−0.5
−1
−1.5
−2
−2.5
0 10 20 30 40 50 60
Time, t [sec]

Case D.
182
Accelerations − Cantilever dam − Case E

1.5
Full reservoir
Empty reservoir
1
0.5
Acceleration, a [g]
−0.5
−1
−1.5
0 10 20 30 40 50 60
Time, t [sec]

Case E.
Response Spectrum − Cantilever dam − Case A

9
Full reservoir
8 Empty reservoir
7
Spectral acceleration, Sa [g]
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
Figure 5.25: Response spectrum (for damping ξ = 5%) of the accelerations

at the crest of the cantilever dam - Case A.
183
Response Spectrum − Cantilever dam − Case B

5
Full reservoir
4.5 Empty reservoir
4
3.5
2.5
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

at the crest of the cantilever dam - Case B.
Response Spectrum − Cantilever dam − Case C

14
Full reservoir
Empty reservoir
12
10
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

at the crest of the cantilever dam - Case C.
184
Response Spectrum − Cantilever dam − Case D

14
Full reservoir
Empty reservoir
12
10
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

at the crest of the cantilever dam - Case D.
Response Spectrum − Cantilever dam − Case E

8
Full reservoir
7 Empty reservoir
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

at the crest of the cantilever dam - Case E.
185
Table 5.2: Material and geometric properties of the cantilever dam cases
considered under seismic loading.
CASE Height, H Thickness, t Elastic modulus, E

[m] [m] [kPa]
A 180 18 1.2 · 10 9
B 180 18 1.4 · 10 8
C 60 6 1.4 · 10 8
D 60 6 1.2 · 10 9
E 180 18 1.2 · 10 10
for the dam with a full reservoir and for all 5 cases (A, B, C, D and E), the
peak value of amplification occurs at a value of the ω/ωd ratio smaller than
the corresponding for the dam with an empty reservoir (see Figures 5.25 -
5.29). This suggests that the fundamental period of the dam is larger when
reservoir-dam interaction is considered. The latter conclusion is in agreement
with the earlier observation in Section 5.3 for harmonic load, i.e. the load
resonates with the dam-reservoir system at a smaller value of natural circular
frequency, therefore the fundamental period of the dam-reservoir system (i.e.
dam with a full reservoir) is larger than that of a dam with an empty reservoir.
Moreover, the value of amplification of accelerations was larger for cases
A-D with a full reservoir than those with an empty reservoir (Figures 5.25
- 5.28). This is in agreement to the previous investigation using harmonic
loading (see Figure 5.4), where higher values of amplification were observed
for the frequency combinations of cases A-D for the dam with a full reservoir.
It can be observed from Figure 5.4, that the amplification with a full reservoir
was higher than with an empty reservoir for the following combinations of
(ω/ωd , ωr /ωd ) = (1, 1), (3, 3), (1, 3), (1/3, 1), which correspond to cases
A-D.
Besides, for case E (i.e. for ω/ωd = ωr /ωd = 1/3), the amplification of
accelerations was found to be smaller for a dam with a full reservoir than that
for a dam with an empty reservoir (Figure 5.29). This is again in agreement
with the previous investigation regarding harmonic loading (see Figure 5.4),
where the amplification for ω/ωd = ωr /ωd = 1/3 was higher in a dam with a
full reservoir than a dam with an empty reservoir.
186
5.5.2 Earth dam

For the case of the earth dam, the FE mesh used earlier and shown in Figure
5.9 is employed. The boundary conditions imposed were described in Section
5.3. Again, all five cases (Table 5.1) were analysed twice, with a full and
empty reservoir.
Table 5.3: Material and geometric properties of the earth dam cases consid-
ered under seismic loading.
CASE Height, H Shear wave velocity, Vs

[m] [m/s]
A 900 4500
B 900 1500
C 300 1500
D 300 4500
E 900 13500
In all the cases, the material properties of the earth dam were: Poisson’s
ratio, ν = 0.3, mass density, ρ = 2000 kg/m3 and Rayleigh damping, ξ = 5%.
The height, H and shear wave velocity, Vs of the dam were changed in each
case as listed in Table 5.3, to achieve the combinations of frequencies listed
in Table 5.1, but in all cases keeping the value of the p-wave velocity of the
reservoir constant, Vp = 1483 m/s. The reservoir water was again modelled
as a linear material with the same values for the bulk and shear modulus as
before (see Section 5.3).
Figures 5.30 - 5.34 show the acceleration time-histories at the crest of the
dam (for both full and empty reservoirs), whereas Figures 5.35 - 5.39 show
the corresponding 5% damped response spectra. It may be observed from
all figures that the observations for the cantilever dam case do not apply to
the earth dam. Although some minor differences between the full and empty
reservoir cases may be noticed in the acceleration time-histories (Figures 5.30
- 5.34), no major differences exist in the response spectra (Figures 5.35 - 5.39).
This means that both the frequency content and the value of amplification
were not significantly affected by the presence of the reservoir, for all cases
A-E, i.e. for various combinations of the frequency ratios. The previous
187
Accelerations − Earth dam − Case A

2
Full reservoir
Empty reservoir
1.5
1
Acceleration, a [g]
0.5
−0.5
−1
−1.5
0 10 20 30 40 50 60
Time, t [sec]
Figure 5.30: Acceleration time-history at the crest of the earth dam - Case
A.
Accelerations − Earth dam − Case B

0.6
Full reservoir
Empty reservoir
0.4
0.2
Acceleration, a [g]
−0.2
−0.4
−0.6
−0.8
0 10 20 30 40 50 60
Time, t [sec]
B.
188
Accelerations − Earth dam − Case C

2
Full reservoir
Empty reservoir
1.5
1
Acceleration, a [g]
0.5
−0.5
−1
−1.5
0 10 20 30 40 50 60
Time, t [sec]
C.
Accelerations − Earth dam − Case D

1.5
Full reservoir
Empty reservoir
1
0.5
Acceleration, a [g]
−0.5
−1
−1.5
0 10 20 30 40 50 60
Time, t [sec]
D.
189
Accelerations − Earth dam − Case E

0.6
Full reservoir
Empty reservoir
0.4
0.2
Acceleration, a [g]
−0.2
−0.4
−0.6
−0.8
0 10 20 30 40 50 60
Time, t [sec]
E.
Response Spectrum − Earth dam − Case A

8
Full reservoir
7 Empty reservoir
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

at the crest of the earth dam - Case A.
190
Response Spectrum − Earth dam − Case B

3.5
Full reservoir
Empty reservoir
3
2.5
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

at the crest of the earth dam - Case B.
Response Spectrum − Earth dam − Case C

8
Full reservoir
7 Empty reservoir
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

at the crest of the earth dam - Case C.
191
Response Spectrum − Earth dam − Case D

8
Full reservoir
7 Empty reservoir
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

at the crest of the earth dam - Case D.
Response Spectrum − Earth dam − Case E

3
Full reservoir
Empty reservoir
2.5
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

at the crest of the earth dam - Case E.
192
conclusion of the negligible influence of the reservoir on the dynamic response

of earth dams under harmonic loading was therefore confirmed for seismic
load as well.
5.5.3 Comments
This section considered the effects of RDI on cantilever and earth dams under
seismic loading. A real acceleration time-history was chosen and applied on
both types of dams in a series of dynamic analyses in the time domain. The
cases were chosen in terms of the frequency ratios ω/ωd and ωr /ωd so that
a direct comparison can be made with the results of the earlier harmonic
studies. The value of ω, was related to TEQ which was called the dominant
period of the load, and it was determined as the circular frequency for which
the highest value of spectral acceleration, Sa occurs (see Figure 5.18).
However, as it may be observed from Figure 5.18, there are two peaks in
the spectral acceleration values. The second peak, at T = 0.52s is the one
chosen as dominant period, and the other one, at T = 0.27s corresponds to
a slightly smaller value of Sa . It is expected to observe resonance at both
periods (0.27s and 0.52s) if they are close to the fundamental frequencies of
the dam and the reservoir. The chosen cases A-E (Table 5.1) were based on
the peak at T = 0.52s and indeed the results for the seismic loading (Section
5.5) are in agreement to the results for the harmonic loading (Sections 5.3 and
5.4). The second peak value of Sa did not affect the results of this section,
as no resonance was expected for the cases chosen.
5.6 Conclusions
This chapter describes the investigation carried out to assess the effects of an
upstream reservoir on the dynamic behaviour of dams. The main aim was to
examine whether the presence of the reservoir alters the response of a dam
and to identify the cases in which the dynamic reservoir-dam interaction may
have detrimental effects. Both rectangular cantilever and trapezoidal homo-
geneous earth dams were considered under harmonic and seismic loading.
193
The main conclusions of this study may be summarised as follows:
• The effects of reservoir-dam interaction (RDI) are mainly related to

(a) change of the fundamental period of the dam and (b) change in the
amplification of accelerations within the dam structure.
• The effects of RDI, in terms of both amplification of accelerations and

change in the fundamental period, are more pronounced for cantilever
dams. This could have an application on various water retaining struc-
tures with a slender cross-sectional geometry, such as arch dams, canal
locks etc.
• Generally, the presence of the reservoir “softens” the response of the

dam, i.e. it results in a larger fundamental period, Td of the dam. This
may be attributed to the added mass, m (see also Westergaard (1933))
of the reservoir which vibrates with the dam. In contrast, the reservoir
does not provide
q
any additional (shear or bending) stifness, k to the
dam (Td = 2π m/k).
• The amplification of the input acceleration in a dam with a reservoir

can be either higher or lower than the amplification in a dam without a
reservoir. This depends on the relative magnitude of the dominant cir-
cular frequency of the load, ω, and the fundamental circular frequencies
of the dam and reservoir, ωd and ωr respectively.
• The amplification of acceleration for a dam with a full reservoir is larger

close to the frequency ratios, ω/ωd ≈ 1, ωr /ωd ≈ 1 and ω/ωr ≈ 1, i.e.
when there is resonance between any two of the load, dam or reservoir.
The maximum amplification occurs when all three circular frequencies
are equal, ω ≈ ωd ≈ ωr (considering any softening due to RDI).
• The effects or RDI (both softening and amplification) were found to be

insignificant for earth dams. This may be attributed to (a) the sloped
upstream face and (b) the large volume of an earth dam. The hydro-
dynamic pressures induced on a sloped upstream dam face are smaller
than those on a vertical dam face (Zangar, 1952). Moreover, the inertial
194
effects from the additional mass from the reservoir are small compared
to the inertia of a large earth dam. However, the amplification in an
earth dam with a full reservoir seems to be slightly smaller than the
corresponding amplification for a dam with an empty reservoir. This
could suggest that the presence of the reservoir may damp the dynamic
response of earth dams (through the interactive oscillation of two bodies
with different vibrational characteristics, such as fundamental period).
• The calculated values for the fundamental period for both dams (can-
tilever and trapezoidal earth) obtained from this study were found
to be slightly different than the values obtained by using analytical
relations from the literature (Euler-Bernoulli beam theory and shear
beam). This is in agreement with earlier observations of previous re-
searchers for both types of dams. The differences were attributed to the
simplifying assumptions of the analytical methods, as the FE method
(employed in this study) is able to overcome their limitations (as it
considers both shear and bending deformations).
195
Chapter 6
NUMERICAL ANALYSIS OF
LA VILLITA DAM
6.1 Introduction
This chapter describes numerical analysis of the seismic response of a real
earth dam. The finite element method was employed to perform static and
time-domain dynamic nonlinear elasto-plastic analyses of La Villita earth
dam in Mexico. La Villita dam is considered to be a reasonably well-
documented case because of available recorded data, material properties and
because of an existing previous research that investigated its response.
The objective of the current work is to reproduce the observed behaviour
of the dam and compare the numerical predictions with the recorded re-
sponse. This was performed in order to validate the numerical approach
adopted for modelling earthfill dams in seismic conditions. In achieving this,
several issues related to numerical analysis were addressed which provided
an insight into the real behaviour of the dam and also provided valuable
feedback and information about considerations that need to be taken when
performing seismic analysis of earth dams using the finite element method.
196
6.2 Description of the dam

6.2.1 Background
La Villita dam (Presa José Marı́a Morelos) is an earthfill dam located in the
Guerrero area in Mexico on the Balsas river. It is one of the large dams of
the Comision Federal de Electricidad of Mexico and it is part of the irrigation
system of a large area in the Guerrero district.
Figure 6.1: Aerial photo of La Villita dam (Google Earth, 2010) with appur-
tenant facilities: (1) Reservoir, (2) Embankment, (3) Hydroelectric power
plant, (4) Spillway, (5) Downstream Balsas river.
The dam’s history states that between 1975 and 1985 it experienced 6
major earthquakes. Although no failure was observed, displacements of some
magnitude were recorded (Reséndiz et al., 1982; Gonzalez-Valencia, 1987;
Elgamal, 1992). It is considered to be a well documented case of seismic dam
response, because measured displacements and acceleration time histories
are available for several locations on the dam together with properties of the
materials that constitute dam’s body and foundations. Figure 6.1 shows an
aerial photo of the dam with the additional facilities.
6.2.2 Construction sequence & seismic events

According to Elgamal (1992), the construction of the whole dam body was
carried out between 1964 and 1968. The embankment however was built in
197
approximately one year, during 1967. Water was impounded in the reservoir
after its construction, i.e. in early 1969 and it operated normally for 7 years
until the first major earthquake event in October 1975.
Before the construction of the dam, the river stream was diverted so that
the construction site was kept dry. Then, a series of holes were drilled and
grouted under the proposed axis of the embankment in order to create the
grout curtain whose function is to control seepage under the dam. After that,
the dam body was constructed in a number of successive thin soil layers which
were compacted upon placement.
Impoundment of the water in the reservoir was carried out over about
half a year, until the reservoir reached its highest level, at approximately
54m. Following the impoundment, the dam was in constant operation and
no rapid drawdowns were reported.
Table 6.1: Summary of earthquake events between 1975-1985 and accelera-

tions in the upstream-downstream direction.
No Date Ms Epicentral Max. Max.

Distance bedrock crest
accel. accel.
[km] [g] [g]
1 11 October 1975 4.5 52 0.07 0.36
2 15 November 1975 5.9 10 0.04 0.21
3 14 March 1979 7.6 121 0.02 0.40
4 25 October 1981 7.3 31 0.09 0.43
5 19 November 1985 8.1 58 0.12 0.76
6 21 November 1985 7.5 61 0.04 0.21
The first significant seismic event occurred in 1975, and then a series of
strong earthquakes took place in the area close to the dam in the following
ten years. Table 6.1 summarizes the details of 6 significant earthquake events
from information obtained from Sociedad Mexicana de Ingenieria Sismica
(2000). Although not all the acceleration time histories could be obtained,
the values in the table were completed using information from Elgamal et al.
(1990) and Elgamal (1992).
As it may be seen from Table 6.1, the earthquakes between 1975 and 1985
198
resulted in high values of PGA at the crest. This implies that they could
have caused serious damage to the dam body.
However, the dam performed well without any major damage and the con-
sequences of these events were only some permanent displacements, which left
the dam still operational. No evidence of liquefaction was observed, neither
within the embankment or nearby. Figure 6.2 shows recorded displacements
with time at several points on the dam body. Further information about
its response during the 1985 earthquakes can be found in Gonzalez-Valencia
(1987).
Figure 6.2: Recorded displacements on some points on the dam body (Elga-
mal, 1992).
6.2.3 Dam geometry & soil properties

La Villita is a 60m high zoned earthfill dam, with a slightly curved crest
about 420m long, and founded on an alluvium layer of varying thickness,
which is about 70m on average. The dam cross-section is composed of a
central clay core of very low permeability, with filters, transitions and outer
rockfill shells. Alluvial deposits beneath the clay core were grouted below
the dam and there is a 0.6m thick concrete cut-off wall to control seepage
199
through the alluvium below the dam. Figure 6.3 shows a typical transversal
(upstream-downstream) cross-section, Figure 6.4 shows a longitudinal cross-
section (Elgamal, 1992) outlining the geological profile of the site, whereas
Figure 6.5 (Elgamal et al., 1990) shows a plan view of the dam.
Figure 6.3: Upstream-downstream cross-section of La Villita dam with zones

of different materials (After Elgamal (1992)).
Figure 6.4: Longitudinal cross-section of La Villita dam showing the geolog-

ical profile of the site and the shape of the canyon (Elgamal, 1992).
From the materials that make the dam, the clay and the alluvium seem to
be the less permeable and expected to behave in an undrained manner. The
rest of the materials (sand filters and rockfill shells) are reasonably coarse
and are expected to behave in a drained manner. Moreover, in-situ explosion
tests showed that the alluvium stratum underneath the dam is not considered
to be prone to liquefaction, as reported by Elgamal (1992).
A number of tests was carried out in order to characterise the materials
undertaken by the Comision Federal de Electricidad (CFE) of Mexico (Comi-
200
Table 6.2: Summary of known material properties of La Villita dam
No Material Density Poisson’s Cohesion Angle

Ratio of
shearing
ρ ν c´ φ0
[kg/m3 ] [kPa] [deg]
1 Clay core 2000 0.49 5 25
2 Sand filters 2180 0.33 0 35
3 Inner Rockfill 2080 0.33 5 45
4 Outer Rockfill 2080 0.33 5 45
5 Alluvium 2080 0.33 5 35
sion Federal de Electricidad, 1976, 1979, 1980, 1987) and the Universidad
Nacional Autónoma de México (UNAM) (Comision Federal de Electricidad
and Univercidad Nacional Autónoma de México, 1976). Table 6.2 provides
a summary of obtained material properties of La Villita dam, although no
detailed information was found about the types of tests carried out. Elgamal
(1992), who performed shear beam analysis of La Villita dam (see Section
6.3), provided information about the spatial variation of the shear modulus,
G of the dam and foundation according to Table 6.3.
Table 6.3: Spatial variation of the maximum shear modulus, Gmax of the
dam and foundation material, as used by Elgamal (1992).
Elevation Part Shear Modulus, Gmax

[m] [kPa]
40-60 Dam Embankment 162935
-70-0 Dam Foundation 193985
6.2.4 Earthquake instrumentation and data

Accelerometers were installed on the dam soon after the end of the construc-
tion. There are three accelerometers on La Villita dam, two being located on
the dam body close to the centre cross-section and one located on bedrock
201
at the right bank. The two accelerograms on the embankment are on the
downstream slope of the dam, one close to the crest and one at the down-
stream berm. Figure 6.5 (Elgamal et al., 1990) shows the location of these
accelerometers on a plan view of the dam.
Figure 6.5: Plan view of La Villita dam and locations of accelerometers

(Elgamal et al., 1990).
La Villita dam and the accelerometers on it belong to the Comision Fed-

eral de Electricidad. The records used in this work were obtained from
Sociedad Mexicana de Ingenieria Sismica (2000) in electronic form (CD).
Accelerograms were recorded at all three points and in all three global
(orthogonal) directions. However, because of instrument malfunction, not all
of these records are available or useful. It is stated by Elgamal (1992) that
only the acceleration records of earthquakes EQ2 (15 November 1975) and
EQ5 (19 November 1985) are useful (see Table 6.1), because the available
records for the rest of the seismic events are incomplete. Moreover, the
direction of the instruments is convenient, as these were installed to provide
records in the upstream-downstream (UD), longitudinal (L) and vertical (V)
directions, with respect to the dam.
Although not all records (i.e. for all 6 events, at all 3 locations, in all 3
directions) could be obtained, still the available records provide an invaluable
tool to investigate the behaviour of the structure during earthquakes and to
validate the applied numerical modelling approach.
All the available earthquake records were processed using the software
SeismoSignal (Antoniou and Pinho, 2004). A linear baseline correction and
202
0.05 EQ1
−0.05
0 1 2 3 4 5 6
0.05
EQ2
−0.05
0 1 2 3 4 5 6
0.02
EQ3
0.01
−0.01
−0.02
0 1 2 3 4 5 6
0.2
EQ5
0.1
−0.1
−0.2
0 10 20 30 40 50 60 70
0.04
EQ6
0.02
−0.02
−0.04
0 2 4 6 8 10 12
Time, t [sec]
Figure 6.6: Available bedrock records in the UD direction for EQ1-EQ3,

EQ5, EQ6
203
0.5
EQ1
−0.5
0 5 10 15
0.5
EQ2
−0.5
0 2 4 6 8 10 12 14 16 18
0.5
EQ3
−0.5
0 5 10 15 20 25 30 35
0.5
EQ4
−0.5
10 5 10 15 20 25 30
EQ5
0.5
−0.5
0 10 20 30 40 50 60 70 80
0.5
EQ6
−0.5
0 10 20 30 40 50 60
Time, t [sec]
Figure 6.7: Available crest records in the UD direction for EQ1-EQ6
204
1
EQ1 − Bedrock
EQ1 − Base
0.8
EQ1 − Crest
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
0.7
EQ2 − Bedrock
Spectral Acceleration, Sa, [g]
0.6 EQ2 − Base

EQ2 − Crest
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
1.4
EQ3 − Bedrock
1.2 EQ3 − Base
EQ3 − Crest
1
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
Figure 6.8: Response spectra (damping, ξ = 5%) of the available acceleration

records in the UD direction for EQ1-EQ3
205
1.4
EQ4 − Base
1.2 EQ4 − Crest
0.8
0.6
0.4
0.2
0
20 0.5 1 1.5 2 2.5 3 3.5 4
EQ5 − Bedrock
Spectral Acceleration, Sa, [g]
EQ5 − Crest
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
0.7
EQ6 − Bedrock
0.6 EQ6 − Crest
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
Figure 6.9: Response spectra (damping, ξ = 5%) of the available acceleration

records in the UD direction for EQ4-EQ6
206
a 4th order Bandpass Butterworth filtering were applied (frequency range:

0.1 ∼ 25 Hz). Figures 6.6 and 6.7 show the (processed) available bedrock
and crest accelerations respectively (5 records out of 6 events for the bedrock
accelerations, with the exemption of EQ4 which could not be obtained) in the
UD direction. Figures 6.8 - 6.9 show the response spectra of the available (of
the bedrock, base and crest) UD acceleration records for all six earthquake
events
6.3 Previous work on the dam

A number of researchers investigated the case of La Villita dam, concentrat-
ing on its response and performance during earthquakes. Both experimental
and analytical/numerical methods were employed in order to approach the
problem.
6.3.1 Numerical and experimental analyses

The work of Elgamal et al. (1990) was motivated by the available acceler-
ation records on the crest of the dam, and especially in the UD direction
during EQ5 (Figure 6.10) which exhibits an asymmetric response. Their in-
vestigation suggested a localised sliding block failure of the dam slope close
to the instrument and tried to obtain an understanding of stick-slip type
deformations.
La Villita Crest Response − EQ5
0.8
EQ5 − La Villita crest
0.6
0.4
0.2
−0.2
−0.4
0 10 20 30 40 50 60 70 80
Time, t [sec]
Figure 6.10: The recorded UD motion on the crest of La Villita dam, during
EQ5 that shows an asymmetry in the peak values.
207
Figure 6.11: The experimental setup used by Elgamal et al. (1990) in the pre-
diction of permanent displacements of La Villita dam: Sliding block mounted
on inclined dynamic shaker.
In their research, the crest acceleration record for the 19th of September
1985 earthquake (EQ5) was compared to the acceleration response of a block
sliding on an inclined plane. The response of the sliding block was exam-
ined using the experimental setup shown in Figure 6.11, to investigate the
observed acceleration response asymmetry. Information about the material
properties was obtained from reports of the CFE (Comision Federal de Elec-
tricidad, 1976, 1979, 1980), whereas the acceleration records were obtained
from UNAM.
A series of sliding block analyses were performed by shaking the appa-
ratus and changing the sliding slope and the friction angle until the total
displacement matched the observed one, which was about 20cm for EQ5.
The final computed displacement time history is shown in Figure 6.12, for
a sliding plane of θ = 22o . Their aim was to investigate whether a sliding
displacement of about 20cm would be associated with an asymmetric ac-
celeration response of the sliding block. And indeed, the accelerations of
208
Figure 6.12: The computed displacement of La Villita dam during EQ5 as

obtained by sliding block analysis of Elgamal et al. (1990).
the sliding block, shown in Figure 6.13 were asymmetric, i.e. the positive
(downwards sliding direction) peak accelerations had higher values than the
negative peak accelerations.
From the sliding block analyses, it was found that the strong motion
records not only indicated localised deformations but also revealed the mag-
nitude of the associated yield acceleration, namely, the value of acceleration
at which sliding occurred. The asymmetry of the positive-negative accelera-
tion peaks was attributed to a sliding block type of failure.
Following this, a three-dimensional numerical model was developed by
Elgamal (1992) based on the shear beam approach (see Section 3.4). The
numerical model was utilised in the analysis of La Villita dam, during the
15th November 1975 (EQ2) and 19th September 1985 (EQ5) earthquakes.
Initial stresses within the dam (prior to the EQs) were obtained by applying
static self weight in the whole numerical model, including the dam, thus
not modelling layered construction. The SB analysis employed a nonlinear
plasticity model (Prevost, 1977) which took account of degradation of shear
stiffness. As before, the material properties were obtained from reports of
the Comision Federal de Electricidad (1976, 1979, 1980) (CFE), whereas the
acceleration records were obtained from UNAM.
Accelerations and displacements were computed and compared to those
209
Figure 6.13: The computed acceleration response of La Villita dam during

EQ5 as obtained by sliding block analysis of Elgamal et al. (1990).
observed (not shown here for brevity). The calculated peak accelerations
were very similar to the recorded ones for the longitudinal and vertical direc-
tions, whereas this was not the case for the upstream-downstream direction
where the calculated values were smaller than the recorded. Furthermore,
as shown in Figure 6.14, their model predicted the dynamic characteristics
(fundamental period) reasonably well and there was a fair agreement between
the calculated and recorded response spectra. However, it was noted that the
recorded acceleration spectrum was broader in frequency content and lower
in amplitude when compared to the computed values.
The difference between the observed and predicted displacements was
attributed to the occurrence of seismically induced deformations due to lo-
calised yielding. Therefore, the calculated accelerations from the SB analy-
sis were used as input in a sliding block analysis, in order to estimate the
permanent displacements of the dam. However, the computed permanent
deformations from the sliding block analysis for EQ5 (≈ 5 cm) (Figure 6.15)
were found much smaller than the observed deformations (≈ 20 cm).
The upstream-downstream earthquake response of La Villita dam during
EQ2 and EQ5 was analysed by Succarieh et al. (1993). Firstly, a 1D nonlin-
ear elasto-plastic hysteretic inhomogeneous SB model was used to represent
210
Figure 6.14: 5% damped response spectra of computed and recorded crest

accelerations during EQ5, obtained by Elgamal (1992).
Figure 6.15: The computed crest displacements as obtained from sliding-

block analyses from Elgamal (1992)
211
the dam structure. Then, an elastic-perfectly plastic sliding block model

(Newmark, 1965) was developed to analyse the observed localised stick-slip
crest deformations. The analysis procedure dictated that the calculated crest
acceleration from the SB analysis was used as an input to the sliding block
model in order to calculate block displacement, similar to Elgamal (1992).
It was reported that their sliding block model was able to capture the
asymmetric response characteristics associated with the evolution of stick-
slip displacements and observed in the recorded crest accelerations of La
Villita dam. They also examined the effect of different values of the yield
acceleration and reported that higher values of yield acceleration result in
smaller values of displacements. Therefore, they commented that there is
a possibility of further permanent movements in future seismic events, but
these are likely to be smaller as the value of the yield acceleration will increase
with successive earthquakes due to hardening of the soil.
The researchers concluded that the asymmetric response of La Villita
dam crest could be attributed to overshoot acceleration (spikes) occurring at
the end of each slip phase and the inertial response of the accelerometer to
the stick-slip input motion.
A “unified” procedure was developed by Gazetas and Uddin (1994) in
order to compute the dynamic response of embankment dams. In that ap-
proach, a 2D dam mesh contained a pre-defined potentially-sliding interface,
was used to calculate the dynamic response, providing information about dis-
placements in a single analysis step. That was considered to be an improve-
ment over the previous approaches (such as Elgamal (1992) and Succarieh
et al. (1993)) that involved two steps: (a) conducting a SB analysis to deter-
mine the accelerations at the level of the sliding surface and (b) using those
accelerations in a Newmark’s sliding block model to evaluate the expected
permanent slope displacements.
A pseudostatic analysis was firstly performed in order to determine the
most critical failure surface and the associated yield acceleration. A FE mesh
was created which included a pre-defined slip surface (its location was already
determined from the pseudostatic analysis) and interface elements along that
chosen curve. Then, the static weight was applied using the gravity turn-on
212
procedure and subsequently dynamic analysis was performed.

That procedure was applied to La Villita dam (Gazetas and Uddin, 1994;
Uddin, 1997) using the finite element code ADINA (Bathe, 1992) in order to
investigate its response during EQ5. Interface elements were used to model
the slip surface and two-dimensional equivalent-linear visco-elastic analyses
were performed. The slip surface was activated and allowed relative shear
movement between the nodes of the pre-defined failure surface part and the
rest of the non-failed dam, according to the constitutive law of the interface
elements. These allowed relative shear movement only when the value of the
induced shear stress was larger than the value of the yield stress required
to cause movement along the slip surface. The FE mesh used is shown in
Figure 6.16.
The computed fundamental frequency of the numerical model was firstly
matched very closely the observed first natural frequency. The computed
accelerations of two points on the crest of the dam (one inside and one outside
the sliding mass) were monitored and are shown in Figure 6.17. It may be
observed that the point inside the failing mass shows an asymmetry between
the positive and negative peaks, whereas the point outside does not. These
results proved that stick-slip response of a soil mass may cause asymmetric
acceleration response in the direction of the sliding. Therefore, that study
confirmed the previous assumption of Elgamal et al. (1990) and Elgamal
(1992) that the observed asymmetry in the recorded crest accelerations might
be due to a localised slope failure.
The magnitude of the computed crest displacements (of the failing mass)
compared quite well with those recorded. The final calculated permanent
displacement (32cm) was reasonably close to the recorded one (≈ 25cm) as
it may be observed from Figure 6.18.
A more advanced method of seismic analysis of dams was developed by
Papalou and Bielak (2001) in which elastic models were used to consider the
deformability of the surrounding medium and effects of spatial variation of
the seismic excitation. The dam was idealised and formulated as a shear
beam of inhomogeneous elastic material and the surrounding medium was
modelled as an elastic halfspace. That was an extended version of the earlier
213
Figure 6.16: The finite element mesh used by Gazetas and Uddin (1994) in
the analysis of La Villita dam with a pre-defined slip surface.
(a) (b)
Figure 6.17: Computed accelerations of two points on the crest of the dam:
(a) outside and (b) inside the failing mass, by Gazetas and Uddin (1994).
214
shear beam approach used by Dakoulas (1993b) (see also Section 2.4.2) for
semi-cylindrical canyons in the sense that the surrounding canyon is included
in the computational model too. Figure 6.19 (Papalou and Bielak, 2001)
presents the components of the modelling system.
The method was applied on the case of La Villita dam, but considering
its behavior as being purely elastic. The main aim of that study was to
investigate dam-canyon interaction effects. A geometric model of the dam,
the canyon and the surrounding medium was considered and the excitation
consisted of SH waves at arbitrary angles of incidence. A cross-section of
the model used in the analysis is shown in Figure 6.20 (Papalou and Bielak,
2001) and the different layers correspond to different material properties
(shear modulus, G), similar to the model of Elgamal (1992) (see also Section
6.2.3 and Table 6.3). A parametric study was undertaken to examine the
effects of a wide range of the shear wave velocity of the surrounding bedrock
on the acceleration response of the dam.
It was found that dam-canyon interaction affects greatly the seismic re-
sponse of the dam as the dynamic coupling of the dam and the surround-
ing deformable canyon resulted in smaller strains than the case with a rigid
canyon due to the energy radiated back to the canyon. Finally, a rigid canyon
yielded higher values of accelerations and therefore the authors suggested
that including dam-canyon interaction effects in seismic design of dams would
prevent unnecessary conservatism.
The procedure developed by Papalou and Bielak (2001) was later ex-
tended (Papalou and Bielak, 2004) to account for nonlinear material be-
haviour and was applied again to the case of La Villita dam. In that investi-
gation, material nonlinearity was modelled using multi-yield surface plasticity
theory (Prevost, 1977), and the seismic response in the upstream-downstream
direction was investigated. Similar to the previous paper, the material prop-
erties were taken from Elgamal (1992) and the shear wave velocity of the
rock canyon was varied in a parametric study.
Then, it was suggested that even with the inclusion of material nonlin-
earity, the effects of dam-canyon interaction were still significant. In that
case however, they were more pronounced within the interior of the dam (i.e.
215
Figure 6.18: Comparison of calculated (thin line) and final recorded (thick
line) permanent crest displacements of La Villita dam by Gazetas and Uddin
(1994).
Figure 6.19: The system analysed by Papalou and Bielak (2001, 2004): (a)
3D dam-canyon model, (b) cross-section of dam in the UD direction, (c)
cross-section of dam and canyon in the longitudinal direction and (d) free
body diagram of a shear wedge element.
216
away from the dam-canyon boundary) where shear strains were larger. Con-
sidering the results in the frequency domain, it was found that at mid-crest,
nonlinearity increased the response at high frequencies, whereas in the inte-
rior of the dam the response was significantly reduced at low and mid-range
frequencies.
The calculated peak accelerations in both of the analyses of Papalou and
Bielak (2001, 2004) (linear elastic and nonlinear elasto-plastic respectively)
were significantly smaller than the recorded values. Figure 6.21 shows a
comparison of the calculated and recorded accelerations at the midcrest.
Clearly, the elastic analyses yielded higher values of peak accelerations, which
however were still much smaller than the recorded. As an example, the
recorded peak acceleration was 1.7 times higher than that obtained from the
linear analysis for the shear wave velocity of the rock, Vsr = 4000 m/s (see
Figure 6.21).
Following the earlier explanation of Elgamal (1992), Papalou and Bielak
(2004) commented that the spikes in the crest record of La Villita dam were
due to a stick-slip displacement mechanism in the zone under the accelerom-
eter. Those spikes appeared at the end of each slip phase, as a consequence
of a sudden change in the inertial load. In addition, it was also reported
that those spikes could not be represented by the coarse mesh used in the
analyses because of their very high frequencies.
6.3.2 Comments
The case of La Villita dam is of significant research interest because of the
available data: material properties, geometry, earthquake records, and mea-
sured displacements. These reasons make it a well-documented case which
attracted the attention of various researchers. However, considering the pre-
viously reviewed information from the literature, it seems that there is still
a number of issues that need to be addressed.
Section 3.7 discussed several issues that need to be taken into account
in the dynamic analysis of earth dams. But, from the work found in the
literature, it may be noticed that those issues were not fully taken into con-
217
Figure 6.20: The model used by Papalou and Bielak (2001, 2004) in the anal-
ysis of La Villita dam. The different zones correspond to different material
properties (Gmax ).
(a) Linear (b) Nonlinear
Figure 6.21: Calculated and recorded response of the midcrest of La Villita

dam, as obtained by linear and nonlinear analyses with varying Vsr of canyon
rock by Papalou and Bielak (2001, 2004) respectively.
218
sideration.
Firstly, the entire history of the dam (construction, operation and sub-
sequent earthquake events) was not simulated in order to obtain the corre-
sponding stress states just before the earthquake. This could provide a more
realistic insight into the stress states which is essential in order to appro-
priately consider soil plasticity (and hence stability and deformations of the
dam structure). Secondly, coupling of the soil skeleton and pore water (cou-
pled consolidation analyses) were not performed, which are useful in order to
capture the steady-state seepage through the dam and realistically consider
the effect of the pore water and of the reservoir.
Moreover, with the exemption of Papalou and Bielak (2004) who provided
a minimum of nine nodes per wavelength, issues related to the FE mesh dis-
cretisation and size of the elements were not strictly taken into consideration.
For example, a rather crude mesh (Figure 6.16) was used in the numerical
analyses of Gazetas and Uddin (1994). A fine mesh could allow the prop-
agation of high frequency wave components and therefore would be able to
predict better any possible high frequency peaks. Besides, the reservoir water
was not modelled and therefore hydrodynamic forces from the reservoir were
not simulated, not even as an additional external static force. Inclusion of
the reservoir could provide an insight into reservoir-dam interaction effects
and into any potential amplifications of the seismic response of dams.
Concerning the accelerations at the crest of La Villita dam, the reason
behind the record’s peculiar asymmetry was not fully understood. Most
researchers tend to agree that the spikes in the record were due to a localised
slip failure on the dam slope. This is strongly supported by the work of
Gazetas and Uddin (1994) who compared the response inside and outside a
potential failing mass. Moreover, the inability to predict such high-frequency
spikes was attributed to the geometry of the FE mesh used in the analyses
(Papalou and Bielak, 2004).
As far as the permanent displacements of the crest are concerned, earlier
work (Elgamal et al., 1990; Elgamal, 1992) suggested that a sliding block
analysis was required in order to capture their magnitudes. However, later
work (Gazetas and Uddin, 1994) provided better estimates (the calculated
219
displacements were 1.45 times higher than the recorded for EQ5) of the per-
manent displacements but it still requires further improvement. An elasto-
plastic analysis, taking account of soil plastic yielding could perhaps provide
a better insight into the actual seismic behaviour of La Villita dam (i.e.
without the need of a pre-defined failure surface).
6.4 Description of the current FE model

Numerical analysis of La Villita dam was carried out in order to explore its
performance during earthquakes. Two-dimensional (2D), static and dynamic,
coupled (mechanical and hydraulic) finite element analyses with the Imperial
College Finite Element Program (ICFEP) (Potts and Zdravković, 1999) were
performed in an attempt to investigate the seismic behavior of the dam.
6.4.1 Stages of analysis

Prior to the dynamic analysis, the entire history of the dam needs to be mod-
elled in order to compute the stress state at the time before the earthquake.
For this reason, a static analysis was performed before the dynamic analysis
(seismic events).
Looking at La Villita dam’s history, as described earlier in Section 6.2.2,
it is imperative that several important events need to be simulated. These
are the layered construction of the dam, the water (reservoir) impoundment
and the consolidation phases. The construction sequence of the dam and the
seismic events are described in Table 6.4. It should be noted that only EQ2
(15/11/1975) and EQ5 (19/9/1985) were simulated because the available
records for the other earthquake events (EQ1, EQ3, EQ4 and EQ6) were not
useful for numerical analysis as noted by Elgamal (1992) (see also Section
6.2.4).
6.4.2 Model geometry & boundary conditions

The FE model needs to adequately simulate the geometry and properties of
the physical problem. Several issues need to be taken into account, such as
220
Table 6.4: Stages of FE analysis of La Villita dam
NO STAGE DURATION TIME

1 Generate initial stresses of level ground – early 1967
2 Layered construction of embankment 1 year 1967
3 Consolidation phase 0 1 year 1968
4 Water impounding 6 months early 1969
5 Consolidation phase 1 6.5 years late 1969 - 1975
6 EQ event 2 – 15/11/1975
7 Consolidation phase 2 10 years 1975 – 1985
8 EQ event 5 – 19/9/1985
the location of the lateral boundaries, the conditions at those boundaries,

the size of the finite elements used, regions of refinement etc.
The mesh created for La Villita dam is shown in Figure 6.22 and it consists
of 1503 8-noded isoparametric quadrilateral elements. Each element has 20
degrees of freedom, displacements in the horizontal and vertical directions, u
and v at all 8 nodes, and pore water pressure, pf at the 4 corner nodes. The
material regions are shown in Figure 6.3.
342m
60m
A B
70m
462m
Figure 6.22: 2D Finite Element mesh used for the analysis of La Villita dam
The value of the maximum shear stiffness, Gmax used by Elgamal (1992) is
listed in Table 6.3. In this work, the value of Gmax was taken to vary linearly
with depth in the dam embankment, according to Equation 6.1 (where, z is
the elevation, i.e. 0 at the base of the dam and 60m at the crest of the dam)
and shown in Figure 6.23, i.e. varying from 140000 kPa at the crest of the
221
dam to 260000 kPa to the base of the dam. The corresponding value of Gmax
for the foundation alluvium was assumed constant and equal to 200000 kPa.
Gmax (z) = 260000 − 20000 · z [kP a] (6.1)
Variation of maximum shear stiffness in the dam, Gmax

60
Elgamal, 1992
Model
50
40
Elevation, z [m]
30
20
10
0
0 50000 100000 150000 200000 250000 300000
Maximum Shear Modulus, Gmax [kPa]
Figure 6.23: Spatial variation (in the vertical direction) of shear stiffness,
Gmax . The grey solid line corresponds to the values used by Elgamal (1992)
(see also Table 6.3) and the black solid line corresponds to the values used
in this work.
The size of the elements was discussed in Section 3.7 and according to
Equation 3.21, it depends on the maximum significant frequency of the load,
fmax (≈ 10 Hz, see Figure 6.24) and the smallest considered shear wave
velocity, VSmin (≈ 265 m/s). This means that, ∆l = 1/5 ∼ 1/4 · 265/10 =
5.3 ∼ 6.625 m, and therefore, the maximum size of the element used was 6
m.
The bottom boundary of the mesh was placed at the interface of the
bedrock and the foundation alluvium, as the bedrock was considered to be
significantly stiffer than the foundation alluvium (Papalou and Bielak, 2001,
2004), and therefore it does not need to be modelled (see also Section 3.7).
222
Fourier Amplitude Spectrum − EQ2 Fourier Amplitude Spectrum − EQ5

0.014 0.16
EQ2 EQ5
0.012 0.14
0.12
0.01
Fourier Amplitude, FA [g]
Fourier Amplitude, FA [g]

0.1
0.008
0.08
0.006
0.06
0.004
0.04
0.002 0.02
0 0
−2 −1 0 1 2 −2 −1 0 1 2
10 10 10 10 10 10 10 10 10 10
Frequency, f [Hz] Frequency, f [Hz]
(a) EQ2 (b) EQ5
Figure 6.24: Fourier Amplitude Spectra of EQ2 & EQ5
The lateral boundaries were placed at such distance from the toe of the
dam (distance A-B in Figure 6.22) so that the response close to the bound-
aries is similar to the free-field response (see Section 3.7). For this reason, 1D
column (site response) and 2D plane strain analyses (see Section 6.6) were
performed and compared to determine the appropriate distance (from the toe
of the dam) at which the lateral boundaries should be placed so that there
is no interaction with the structure studied. The required distance from the
toe of the dam was found to be 60 m.
6.4.3 Calibration of constitutive models

The materials of the dam were discussed in Section 6.2.3 and their properties
were summarized in Table 6.2 and are also shown in Figure 6.3. A nonlin-
ear elasto-plastic model was adopted to simulate the soil material (clay core,
sand filters, rockfill shells and alluvium foundation) behaviour, which is able
to take account of plastic yielding (Mohr-Coulomb failure criterion) and pre-
yield degradation of stiffness with strain (Logarithmic cyclic nonlinear model)
(see Section 3.7.5). The parameters of the Mohr-Coulomb model are sum-
marised in Table 6.5. The concrete block and the grout curtain were assumed
to be linear elastic materials as the developed strain levels are considered to
223
be small to induce plasticity in the concrete.
Table 6.5: Mohr-Coulomb model parameters for the soil materials.
No Material Cohesion Angle Angle

of of
shearing dilation
c´ φ0 ψ0
[kPa] [deg] [deg]
1 Clay core 5 25 0
2 Sand filters 0 35 0
3 Inner Rockfill 5 45 0
4 Outer Rockfill 5 45 0
5 Alluvium 5 35 17.5
As there were no site-specific tests available to determine the dynamic

properties of the dam materials (i.e. variation with induced strain), the
Logarithmic CNL model was calibrated against empirical curves of stiffness
degradation, G/Gmax , and damping, ξ, with shear strain, γ.
For the calibration of the Logarithmic CNL model, the maximum value
of the shear modulus, Gmax is also a model parameter which affects the
stiffness degradation and therefore needs to be specified in the calibration
process (see Section 3.7.5). As mentioned earlier, the value of the maximum
shear stiffness, Gmax was assumed to vary linearly with depth in the dam
embankment (see Equation 6.1 and Figure 6.23) and its average value (at
mid-height) is Gmax = 200000 kPa (which is also the corresponding value for
the foundation alluvium), therefore, the Logarithmic model was calibrated
using this value for Gmax . Furthermore, as mentioned earlier in Section
3.7.5, there is an option in ICFEP to specify the minimum value of the shear
modulus, Gmin (which affects the variation of damping, ξ). The value for that
parameter was taken as 10% of Gmax because it provided a good fit for the
variation of damping (a value of 10% of Gmax for Gmin was also suggested by
Taborda (2011)). A single value is specified for this parameter, and therefore
for the embankment domain (in which there is a linear variation of Gmax ),
Gmin was taken as 10% of the average value of Gmax , i.e. Gmin = 0.1 · 200000
kPa = 20000 kPa.
224
For the calibration of the CNL model, the materials of the dam were
divided in three categories: (a) clay core, (b) sand filters and (c) rockfill and
alluvium. The clay core was considered to be a plastic material and the CNL
model was calibrated against the curves of Vucetic and Dobry (1991) for
Ip = 30%. The sand filters were considered to be of low Plasticity Index Ip ,
and therefore they were calibrated against empirical curves for such material,
which in that case, were the curves of Seed et al. (1986). Finally, the rockfill
shells and the alluvial foundation were calibrated against the curves of Rollins
et al. (1998) for rockfill materials.
Table 6.6 summarizes the results of the calibration process for the Log-
arithmic CNL and for all materials. The results, both in terms of shear
modulus (G/Gmax − γ) and damping (ξ − γ) variation are shown graphically
in Figures 6.25 - 6.30 for all the dam materials. It may be observed that
a fair match was achieved between the curves predicted by the CNL model
(solid black line, for the middle of the dam, i.e. for Gmax = 200000 kPa)
and the empirical ones from the literaure (dashed black line), for both shear
stiffness degradation and damping.
Table 6.6: Calibration parameters for the Logarithmic CNL model.
Logarithmic CNL model parameters

Material Gmax EdL JL c Gmin
[kPa] [] [kPa] [] [kPa]
Clay core 200000 0.0019 185 0.4 20000
Sand filters 200000 0.0014 65 1.0 20000
Rockfill & alluvium 200000 0.0004 45 0.3 20000
As reported earlier (Equation 6.1 and Figure 6.23), the value of the max-
imum shear stiffness, Gmax in the embankment varied linearly with depth
(140000-260000 kPa). However, one set of calibration parameters (EdL , c,
JL and Gmin ) was specified for the whole embankment domain, which corre-
sponds to a calibration for Gmax = 200000 kPa, which is the average value.
Therefore, the materials at the top and bottom of the embankment which
have smaller (140000 ∼ 200000 kPa) and larger (200000 ∼ 260000 kPa) shear
stiffness respectively, exhibit a different variation of G/Gmax −γ and ξ −γ for
225
the specified model parameters (EdL , c, JL and Gmin , listed in Table 6.6), as
presented in Figures 6.25 - 6.30. The solid grey line corresponds to the top of
the embankment (smallest value of Gmax = 140000 kPa) and the dashed grey
line corresponds to the bottom of the embankment (largest value of Gmax =
260000 kPa).
It should be noted that the calibration of the Logarithimc model allowed
either (a) an excellent match between the model and the empirical curves for
the variation of shear stiffness (G/Gmax − γ) but at the same time, a poor
match for the damping curves (ξ − γ) or (b) the opposite, i.e. an excellent
calibration for the damping curves (ξ − γ) and a poor calibration for the
shear stiffness curves (G/Gmax − γ). In the first case (a), i.e. if the shear
stiffness curves (G/Gmax − γ) were matched, the CNL model would predict
higher damping than the empirical curves for large values of the shear strain,
γ. That would result in an overdamped response of the dam structure. In
the second case (b), i.e. if the damping curves (ξ − γ) were matched, the
CNL would predict lower values of the shear stiffness than the empirical
curves. That would then result in a softer response of the dam, with a
longer fundamental period and perhaps smaller amplification in the higher
frequencies range. It was therefore decided to calibrate the CNL on both
G/Gmax − γ and ξ − γ curves, obtaining an optimum match for both sets of
curves.
Figure 6.31 shows a comparison between the shear stiffness degradation
(G/Gmax − γ) predicted by the Logarithmic CNL model and that used by
Elgamal (1992). It may be observed that there is a fair agreement between
the two cases. The smaller values of shear stiffness could be attributed to
the calibration against both stiffness and damping, as discussed above. The
larger difference for the case of the sand filters should not significantly affect
the dynamic characteristics of the dam, as the sand filters are just a very
small part of the whole embankment.
Figures 6.25, 6.27 and 6.29 show that the normalised shear stiffness degra-
dation, G/Gmax − γ, varies with the elevation within the dam (i.e. points
at top, middle and bottom). It is shown that the values of G at the bottom
of the dam (which are the largest values, due to the adopted spatially inho-
226
0.9
0.8
Shear Modulus, G/Gmax []
0.7
0.6
0.5
0.4
0.3
0.2 Model − Top

Model − Middle
0.1 Model − Bottom
Vucetic & Dobry, 1991
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
Shear strain, γ [%]
Figure 6.25: Calibration of the Logarithmic CNL model against the empirical
curves of Vucetic and Dobry (1991) for the clay core - Stiffness degradation
(G/Gmax − γ)
30
Model − Top
25 Model − Middle
Model − Bottom
20
Damping, ξ [%]
15
10
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
curves of Vucetic and Dobry (1991) for the clay core - Damping (ξ − γ)
227
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Model − Top
0.2
Model − Middle
Seed et al., 1986
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
Figure 6.27: Calibration of the Logarithmic CNL model against the empir-
ical curves of Seed et al. (1986) for the sand filters - Stiffness degradation
(G/Gmax − γ)
30
Model − Top
25 Model − Middle
Model − Bottom
Seed et al., 1986
20
Damping, ξ [%]
15
10
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
curves of Seed et al. (1986) for the sand filters - Damping (ξ − γ)
228
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Model − Top
0.2
Model − Middle
Rollins et al., 1998
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
curves of Rollins et al. (1998) for the rockfill and foundation alluvium -
Stiffness degradation (G/Gmax − γ)
30
Model − Top
25 Model − Middle
Model − Bottom
20
Damping, ξ [%]
15
10
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
curves of Rollins et al. (1998) for the rockfill and foundation alluvium -
Damping (ξ − γ)
229
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Elgamal (1992)
0.2
Model (Vucetic & Dobry, 1991) − Middle
0.1 Model (Seed et al., 1986) − Middle
Model (Rollins et al., 1998) − Middle
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
Figure 6.31: Comparison of the stiffness degradation curves (G/Gmax − γ)

predicted by the Logarithmic CNL model for the various materials (and the
corresponding empirical curves used) with the stiffness degradation used by
Elgamal (1992).
300000
Model − Top
Model − Middle
Model − Bottom
250000
Shear Modulus, G [kPa]
200000
150000
100000
50000
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
Figure 6.32: Stiffness degradation (G−γ) predicted by the Logarithmic CNL

model for the clay core
230
300000
Model − Top
Model − Middle
Model − Bottom
250000
200000
150000
100000
50000
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10

model for the sand filters
300000
Model − Top
Model − Middle
Model − Bottom
250000
200000
150000
100000
50000
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10

model for the rockfill and foundation alluvium
231
mogeneous stiffness profile) degrade more with the induced shear strain, γ.
However, as shown in Figures 6.32 - 6.34, which present the absolute values
of shear modulus, G, the variation of the shear modulus within the dam,
occurs for the strains, up to γ = 0.04 ∼ 0.2 %, where it reaches the assigned
minimum stiffness value (Gmin ). Therefore, the stiffness inhomogeneity holds
for small values of the induced shear strain. Such small strains were reported
in this work (see Section 6.6).
6.5 Static analysis

Following the history of the dam as explained in Section 6.2.2, all the stages
prior to the earthquake events were simulated, i.e. layered construction,
reservoir impoundment and consolidation, as described by Table 6.4. A
coupled consolidation formulation was employed for the entire (static and
dynamic) analysis of the dam, treating the clay core and the foundation al-
luvium as consolidating materials, whereas the remaining materials (rockfill
and sand filters) were considered to be drained. The values of the permeabil-
ities used for the clay core and the foundation alluvium were Kclay = 10−10
m/s and Kalluvium = 10−7 m/s respectively, which are considered representa-
tive for such types of materials.
6.5.1 Construction
After the establishment of initial stress conditions, i.e. level ground (with the
water level being at 2 m depth), the embankment was constructed. Firstly, a
concrete block which serves as a foundation of the clay core was constructed
along with the placement of the grout curtain underneath. Then the em-
bankment was built in 10 successive layers of 6 m thickness. The clay core
was considered to have suction due to its compacted nature, and a value
of 50 kPa was specified, i.e. pf = −50kP a (positive values correspond to
compressive stresses, whereas negative values correspond to tensile stresses).
This stage lasted for a year (1967) as data about the dam’s history suggested
(Table 6.4). Each layer was constructed over a number of increments in order
232
to achieve numerical stability.

The deformation boundary conditions (BC) (shown graphically in Fig-
ure 6.22) along the bottom boundary were full fixity (i.e. the value of the
horizontal and vertical displacements is prescribed to be equal to zero), and
horizontal fixity along the lateral boundaries (i.e. prescribed horizontal dis-
placement and prescribed vertical stress being equal to zero). The hydraulic
BC along the bottom boundary was prescribed as zero flow, and along the
lateral boundaries, as no change in pore water pressure, ∆pf = 0, in order
to maintain the initial hydrostatic conditions. The hydraulic BCs on the
boundaries of the core were prescribed as follows: no flow on the bottom and
top boundaries, and precipitation BC on the two lateral sides of the core.
The precipitation BC (Potts and Zdravković, 1999) is an advanced BC
which allows a prescription of dual hydraulic conditions. It may act as a pre-
scribed flow in one direction, or as a prescribed value of pore water pressure.
On both US and DS core sides, it was specified that if the water pressure is
more tensile in the core than on the core boundary, there is no flow of water
in the core from the outside (which is dry, therefore, pf = 0). On the other
hand, if the pore pressures in the core are more compressive (pf > 0) than
those on the core boundary, then the pore water pressure value on the bound-
ary is prescribed to be equal to zero. In that case, there is going to be flow of
water from within the core, towards rockfill, with such a flow rate, that the
value of the pore water pressure, pf on the core boundary will be equal to
zero. The reason for the use of this advanced BC is to avoid the unrealistic
inflow of water in the core when there is suction during its construction.
6.5.2 Reservoir impoundment

Water was impounded in the reservoir after the end of the construction. Con-
struction of the whole dam project finished at the end of 1968 (i.e. one year
after the construction of the embankment), therefore water was impounded
in early 1969. Consequently, a whole year was modelled as pure consolida-
tion (i.e. 1968) as essentially no other construction process was applied to the
embankment (only dissipation of excess pore water pressures, i.e. consolida-
233
tion). During that stage, all the boundary conditions used before remained
the same, and only time elapsed.
Following the one-year consolidation phase, water was impounded in the
reservoir. That was modelled over a total duration of 6 months. Water level
was raised in a single layer which required 10 increments. Therefore, the wa-
ter in the reservoir was modelled as an additional external boundary stress on
the upstream face of the dam up to a height of 54 m (the higher level of the
reservoir), thus allowing 6 m for freeboard. This modelled the hydrostatic
pressure from the reservoir. Besides, an additional boundary stress was ap-
plied on the upstream riverbed alluvium equal to the maximum hydrostatic
value. At the same time of the application of the external boundary stress,
the pore pressure in the elements of the upstream rockfill and sand filters was
prescribed to be in equilibrium with the externally applied boundary stress
(i.e. hydrostatic). The approach followed to model water impoundment is
similar to that used by Kovacevic (1994) for the static analysis of Roadford
dam in the UK. A similar procedure was followed by Sica et al. (2008) for
the static and dynamic analysis of El Infiernillo dam in Mexico. They noted
that the water filling was simulated by applying a time dependent hydro-
static distribution of pore water pressure following the height of the reservoir
water and a normal stress distribution equal to the pore water pressures was
added to the wet (upstream) boundary of the core.
The deformation BCs were the same as before (i.e. during the consolida-
tion stage). The hydraulic BCs on the boundary of the mesh were also the
same except for the upstream lateral boundary of the mesh on which the pre-
scribed pore water pressures increased according to the pore pressure change
due to the water level rise. Moreover, the hydraulic BC on the upstream face
of the core was no longer precipitation. It had prescribed values of the pore
water pressure according to the elemental pore pressures prescribed within
in the upstream rockfill (i.e. a hydrostatic variation). This allowed a water
seepage through the core according to the value of the permeability, Kclay
(10−10 m/s) of the clay.
Figure 6.35 shows the pore water pressure distribution after the end of
the water impoundment, whereas Figure 6.36 shows the flownet in the core.
234
Figure 6.35: Contours of pore water pressure in the dam after the end of the
reservoir impoundment (positive values correspond to compressive stresses).
235
Figure 6.36: Flow net (contours of stream and potential functions) in the
clay core after the end of the reservoir impoundment.
It may be observed from the first figure that there are pore pressures in the
upstream rockfill as a result of the water impoundment. This distribution is
linear in the vertical direction as the water is expected to penetrate quickly in
the coarse rockfill and hydrostatic conditions are established. Moreover, the
contours of pore pressure drop in the clay core which means that the pressure
reduces in the downstream side of the core, compared to the hydrostatic
values on the upstream side. There is still some suction in the upper part
of the clay core, which indicates that a part of the core is still not fully
saturated. Besides, comparing the two parts of the foundation alluvium, the
upstream part had significantly higher values of water pressure which are a
result of the reservoir water. Finally, as far as the flow net in the core is
concerned, flow lines and equipotential lines were clearly formed indicating
seepage from the upstream to the downstream side of the clay core.
236
6.5.3 Consolidation
After the reservoir impoundment was finished, i.e. in mid 1969, the dam
started to operate normally until late 1975, when two major seismic events
occurred, EQ1 and EQ2. The time period between the end of the reservoir
impoundment and the first two seismic events is about 6.5 years, and it was
modelled again as a pure consolidation analysis stage. For this stage of the
analysis, the BCs (both deformation and hydraulic) were the same as in the
previous consolidation stage.
6.5.4 Comments
The static part of the analysis was performed in order to obtain the correct
stress states prior to the earthquake and ensure that appropriate plasticity
is introduced in the model. This issue is discussed in more detail after the
description of the dynamic analysis, in Section 6.6.
Crest settlement of La Villita dam
0
Recorded
Calculated
Settlement, d [cm]
−5
−10
−15
1968 1969 1970 1971 1972 1973 1974 1975 1976
Time, t [year]
Figure 6.37: Calculated and measured crest settlement history of the dam
crest prior to the EQ events.
Figure 6.37 shows the calculated and recorded crest settlement history of
the dam crest prior to the EQ events which exhibit a good agreement. Figures
6.38 (a) and (b) show the deformed mesh and the vectors of accumulated
displacement respectively at the end of the static part of the analysis. It is
shown that the dam had some minor deformations resulting from the static
stresses (construction, reservoir impoundment and consolidation), but no
237
large movements or failure were observed. As shown in the latter figure, the
maximum value of displacement is 0.258m and this is represented by the grey
thick vector, in the core of the dam.
6.6 Dynamic analysis

The next stage after the static analysis was the seismic analysis of the dam.
This was performed with dynamic FE analyses in the time domain, as ex-
plained in Section 3.7. The time step used was ∆t = 0.02s. The bound-
ary conditions adopted were as follows: the earthquake load was applied as
prescribed values of acceleration in the horizontal direction at the bottom
boundary, while the displacements in the vertical direction were specified to
be equal to zero. The BC on the lateral boundaries was the TDOF (see
Section 3.7), by tying nodes of the same elevation on both lateral boundaries
to have the same values of horizontal and vertical displacements. According
to Table 6.4, EQ2 (1975) was analysed first, then a time of pure consolida-
tion (10 years) was statically modelled and finally EQ5 (1985) was analysed.
The time between the two seismic events was modelled as discussed earlier
in Section 6.5.
6.6.1 Effect of canyon geometry

Figures 6.40 and 6.42 show the calculated acceleration time histories at the
crest of the dam (Point C in Figure 6.39) for EQ2 and EQ5 respectively.
As far as EQ2 is concerned (Figure 6.40), it should be noted that only part
of the bedrock acceleration record was available (which was used as input
in the analysis), whereas the full crest acceleration record was available.
That is why the calculated crest acceleration time-history is shorter than
the corresponding recorded one. Figures 6.41 and 6.43 show the correspond-
ing response spectra. It may be observed from the former two figures that
the calculated accelerations were found to be significantly smaller than the
recorded ones. A similar trend is observed in the corresponding response
spectra (Figures 6.41 and 6.43), where the calculated spectral acceleration,
238
(a) (b)
Figure 6.38: End of static analysis: (a) Deformed mesh and (b) Vectors of
accumulated displacement of La Villita dam.
239
U C
B
Figure 6.39: Location of the acceleration monitoring points of the upstream

corner of the crest (U), mid-crest (C) and downstream berm (B) and location
of application of the seismic excitation at the rock (R) underneath the dam.
Sa values are generally smaller than the corresponding recorded ones.

Interestingly, the calculated values of the spectral acceleration, Sa for
EQ5 (Figure 6.43) were found to be smaller than the recorded values for
small values of the period, T < 1.3 s, and larger for large values of the
period, T > 1.3 s. This implies that higher accelerations were observed for
larger values of the fundamental period, which would occur if a softer (i.e.
with a larger fundamental period) system was considered. Therefore, this
shows that the calculated response of the dam was softer.
As discussed earlier in Section 2.4.2, dams built in narrow canyons exhibit
a stiffer response than dams built in an infinitely wide canyon. In this case,
La Villita, is a 60m high dam founded on a 70m alluvial deposit and its
crest length is about 420m. Therefore, the canyon length over height, L/H
ratio (where H is the total height of the dam and alluvium, Htot ) is given by
Equation 6.2.
L 420
L/H = L/Htot = = = 3.23 (6.2)
Hdam + Halluvium 60 + 70
This implies that canyon effects could be important and a 2D analysis
would be inappropriate as the real problem is stiffer than the corresponding
2D plane strain model. Such a 2D analysis would be suitable for a wide
canyon (L/H > 4), see also Mejia and Seed (1983); Dakoulas and Gazetas
(1987) and relevant discussion in Section 2.4.2. In order to overcome this
soft response of a 2D analysis, the stiffening effect of the canyon geometry
was taken into account by increasing the material stiffness of the dam. A
240
Acceleration response at the crest of La Villita dam − EQ2

0.25
Recorded
0.2 Calculated
0.15
Acceleration, a [g]
0.1
0.05
−0.05
−0.1
−0.15
0 2 4 6 8 10 12 14 16 18
Time, t [sec]
Figure 6.40: Acceleration time history at the crest of La Villita dam during
EQ2
Response spectra at the crest of La Villita dam − EQ2

0.7
Recorded
0.6 Calculated
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
Figure 6.41: Response spectra (damping, ξ = 5%) at the crest of La Villita

dam during EQ2
241

1
Recorded
Calculated
0.8
0.6
Acceleration, a [g]
0.4
0.2
−0.2
−0.4
0 10 20 30 40 50 60 70 80
Time, t [sec]
EQ5

1.6
Recorded
1.4 Calculated
Spectral Acceleration, S [g]
1.2
a
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

dam during EQ5
242
Table 6.7: New calibration parameters for the Logarithmic CNL model, used
for the dynamic analysis.
Logarithmic CNL model parameters

Material G∗max EdL∗
JL∗ c∗ G∗min
[kPa] [] [kPa] [] [kPa]
Clay core 700000 0.0005 220 0.4 70000
Sand filters 700000 0.0004 100 0.5 70000
Rockfill & alluvium 700000 0.0009 280 0.4 70000
parametric study was carried out to determine the ratio of the new updated
shear stiffness over the initial stiffness, which would provide the best match
between the calculated and recorded response spectra. It was found that the
shape of the response spectrum was improved (and therefore the prediction
of the fundamental period of the dam was improved as shown later) if the
shear modulus, Gmax (z) was increased by 3.5 times.
The new shear stiffness profile, G∗max (z) was (similarly to the initial pro-
file, Gmax (z)) taken to vary linearly with depth in the dam embankment,
according to Equation 6.3 (where, z is the elevation, i.e. 0 at the base and
60m at the crest of the dam). This was varying from 420000 kPa at the crest
of the dam to 780000 kPa to the base of the dam (see also Equation 6.1 for
the initial variation of Gmax (z)). The corresponding value of G∗max for the
foundation alluvium was assumed constant and equal to 700000 kPa.
G∗max (z) = 3.5 · Gmax (z) = 780000 − 70000 · z [kP a] (6.3)
As mentioned earlier, the performance of the Logarithmic CNL (i.e. vari-

ation of G with shear strain, γ) depends on the maximum value of the shear
stiffness, Gmax . Therefore, for the new updated values of the shear modulus,
G∗max , a new calibration was performed. The results of the new updated
calibration (see also earlier discussion in Section 6.4.3) are listed in Table 6.7
and shown graphically in Figures 6.44 - 6.50. From these figures, it may be
observed that a good agreement was achieved between the CNL model pre-
diction and the empirical curves. The absolute values of G∗max are shown in
Figures 6.51 - 6.53. It should be again noted that the average value of G∗max
243
= 700000 kPa was used in the calibration process (see also Section 6.4.3).
Figures 6.54 and 6.56 show the calculated acceleration time histories at
the crest of the dam (Point C in Figure 6.39) for EQ2 and EQ5 respectively
with the new updated values of the G∗max . Likewise, Figures 6.55 and 6.57
show the corresponding response spectra for the two earthquakes respectively.
It may be observed from the former two figures that the new calculated
accelerations were found to be in better agreement with the recorded values.
Moreover, good agreement is also observed in the corresponding response
spectra (Figures 6.55 and 6.57) where the calculated spectral acceleration,
Sa values were closer to the corresponding recorded ones. The previously
observed small and high values of amplification for smaller and larger values
of the period respectively vanished and the calculated spectral accelerations
exhibit large amplifications for the smaller values of the fundamental period.
This shows that a better match of the fundamental period of the dam was
achieved by increasing the material stiffness (shear modulus, G∗max ).
The stiffening effect of narrow canyons was discussed earlier in Chapter
2 and Section 2.4.2. The studies of Mejia and Seed (1983) and Dakoulas
and Gazetas (1987) have quantified this stiffening effect and expressed it as
a function of the canyon length over height, L/H ratio. For La Villita dam,
this L/H ratio is given by Equation 6.2.
According to the study of Dakoulas and Gazetas (1987) (Figure 2.22),
the ratio of the fundamental period of vibration of a dam built in a narrow
canyon, Tn , over that of a dam built in an infinitely wide canyon, Tw , for
L/H = 3.23 and for various shapes of the canyon is given by Equation 6.4.
Tn
= 0.6 ∼ 0.95 (6.4)
Tw
In the present study, the updated value of the shear modulus was taken
as G∗ = 3.5 · G. The shear wave velocity, Vs is given by Equation 6.5 (where,
ρ is the mass density of the material).
s
G
Vs = (6.5)
ρ
244
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Model − Top
0.2 Model − Middle
Model − Bottom
0.1 Vucetic & Dobry, 1991
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
Figure 6.44: New calibration of the Logarithmic CNL model against the
empirical curves of Vucetic and Dobry (1991) for the clay core - Stiffness
degradation (G/G∗max − γ)
30
Model − Top
25 Model − Middle
Model − Bottom
20
Damping, ξ [%]
15
10
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
empirical curves of Vucetic and Dobry (1991) for the clay core - Damping
(ξ − γ)
245
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Model − Top
0.2 Model − Middle
Model − Bottom
0.1 Seed et al., 1986
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
Figure 6.46: New calibration of the Logarithmic CNL model against the em-
pirical curves of Seed et al. (1986) for the sand filters - Stiffness degradation
(G/G∗max − γ)
30
Model − Top
25 Model − Middle
Model − Bottom
Seed et al., 1986
20
Damping, ξ [%]
15
10
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
empirical curves of Seed et al. (1986) for the sand filters - Damping (ξ − γ)
246
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Model − Top
0.2
Model − Middle
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
pirical curves of Rollins et al. (1998) for the rockfill and foundation alluvium
- Stiffness degradation (G/G∗max − γ)
30
Model − Top
25 Model − Middle
Model − Bottom
20
Damping, ξ [%]
15
10
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
pirical curves of Rollins et al. (1998) for the rockfill and foundation alluvium
- Damping (ξ − γ)
247
0.9
0.8
0.7
0.6
0.5
0.4
0.3
Elgamal (1992)
0.2
Model (Vucetic & Dobry, 1991) − Middle
0.1 Model (Seed et al., 1986) − Middle
Model (Rollins et al., 1998) − Middle
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
Figure 6.50: Comparison of the stiffness degradation curves (G/G∗max − γ)

predicted by the Logarithmic CNL model (new calibration) for the various
materials (and the corresponding empirical curves used) with the stiffness
degradation used by Elgamal (1992).
1000000
Model − Top
900000 Model − Middle
Model − Bottom
800000
700000
600000
500000
400000
300000
200000
100000
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10
Figure 6.51: Stiffness degradation (G∗ − γ) predicted by the new calibration

of the Logarithmic CNL model for the clay core
248
1000000
Model − Top
Model − Bottom
800000
700000
600000
500000
400000
300000
200000
100000
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10

of the Logarithmic CNL model for the sand filters
1000000
Model − Top
Model − Bottom
800000
700000
600000
500000
400000
300000
200000
100000
0
−4 −3 −2 −1 0 1
10 10 10 10 10 10

of the Logarithmic CNL model for the rockfill and foundation alluvium
249

0.25
Recorded
0.2 Calculated
0.15
0.1
Acceleration, a [g]
0.05
−0.05
−0.1
−0.15
−0.2
0 2 4 6 8 10 12 14 16 18
Time, t [sec]
EQ2 for the new updated value of G∗max (z).

0.7
Recorded
Calculated
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

dam during EQ2 for the new updated value of G∗max (z).
250

1
Recorded
Calculated
0.8
0.6
Acceleration, a [g]
0.4
0.2
−0.2
−0.4
0 10 20 30 40 50 60 70 80
Time, t [sec]
EQ5 for the new updated value of G∗max (z).

1.6
Recorded
1.4 Calculated
1.2
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

251
Therefore, the updated value of the shear wave velocity, Vs∗ is given by
Equation 6.6.
√
Vs∗ = 3.5Vs (6.6)
The fundamental period of vibration, T , is inversely proportional to the

shear wave velocity (see Section 2.4 and Equations 2.10 and 2.12). Therefore,
the new updated value of the fundamental period of vibration, T ∗ is given
by Equation 6.7.
1
T∗ = √ · T = 0.54T (6.7)
3.5
Therefore, the ratio T ∗ /T = 0.54 is close to the ratio found in the liter-
ature and suggested by Dakoulas and Gazetas (1987) (Tn /Tw = 0.6 ∼ 0.95,
for various shapes of the canyon, see Figure 2.22). This observation confirms
that the calculated stiffening of the narrow canyon is in agreement with ear-
lier work from the literature. However, the stiffening observed in the present
study was found to be slightly larger, i.e. the ratio of T ∗ /T is smaller than
the ratio suggested in the literature. This difference could be attributed to
the fact that the previous theoretical studies assumed the following:
• an idealised geometry of earth dam and canyon geometries, and
• a linear soil material behaviour
As far as the former reason is concerned, it should be noted that the

results from the literature (Dakoulas and Gazetas, 1987) assumed an ide-
alised dam of a triangular cross-section built in a canyon of various shapes.
However, the geometry of La Villita dam (Figure 6.3) consists of a 60m (ap-
proximately) triangular in the cross-section embankment dam and a 70m
uniform foundation alluvium. Moreover, a triangular dam and a uniform
soil layer system (of overall height Htot ) have different vibrational charac-
teristics than a corresponding dam with H = Htot , and a dam exhibits a
stiffer response, i.e. smaller fundamental period (see Sections 2.3.3 & 2.4.1
and Equations 2.6 & 2.12). Therefore, the combination of a triangular dam
252
resting on a uniform soil layer is expected to be stiffer (i.e. have a smaller

fundamental period) than just a uniform triangular dam. This was already
examined by Dakoulas and Gazetas (1987) and Dakoulas (1990) who showed
that a dam-foundation system has a larger fundamental period than a uni-
form dam of the same height. Those studies also showed that the stiffening
effect of the canyon is more pronounced for larger foundation to dam heights
(see also relevant discussion in Section 2.4.4).
As far as the latter reason is concerned, the nonlinear soil behaviour
that was considered in this work, takes into consideration the degradation of
stiffness and damping with shear strain. Smaller stiffness results in a larger
period for the dam-foundation system and therefore a different T ∗ /T ratio
than an analysis considering linear soil behaviour.
In conclusion, it should be commented that the calculated stiffening of
the narrow canyon was in a good agreement with the stiffening suggested by
earlier studies found in the literature. The small difference in the value of
stiffening (as expressed by the ratios T ∗ /T and Tn /Tw ) can be qualitatively
justified.
6.6.2 Analysis with new stiffness profile of the dam

6.6.2.1 Dynamic response of the dam
As mentioned above, the dynamic analyses of La Villita dam were repeated

with the updated values of the maximum shear stiffness, G∗max . The acceler-
ation response at the crest of the dam for EQ2 and EQ5 seismic events was
shown in Figures 6.54 and 6.56 respectively, whereas the associated response
spectra were shown in Figures 6.55 and 6.57 and they have all been discussed
earlier.
Figures 6.58 and 6.59 show the acceleration response and the associated
response spetrum of the downstream berm during EQ2. It may be observed,
that similar to the response at the crest, a very good agreement is achieved
between the recorded and calculated response at the downstream berm.
Figure 6.60 shows a comparison between the calculated response spec-
tra at the crest of the dam for EQ5 and those obtained by Elgamal (1992),
253
Acceleration response at the downstream berm of La Villita dam − EQ2

0.1
Recorded
0.08 Calculated
0.06
0.04
Acceleration, a [g]
0.02
−0.02
−0.04
−0.06
−0.08
−0.1
0 2 4 6 8 10 12 14 16
Time, t [sec]
Figure 6.58: Acceleration time history at the downstream berm of La Villita

Response spectra at the downstream berm of La Villita dam − EQ2

0.35
Recorded
0.3 Calculated
0.25
0.2
0.15
0.1
0.05
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
Figure 6.59: Response spectra (damping, ξ = 5%) at the downstream berm

of La Villita dam during EQ2 for the new updated value of G∗max (z).
254

2.5
Recorded
Calculated − ICFEP
Calculated − Elgamal (1992)
2
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

dam during EQ5. Comparison between the (a) recorded, (b) calculated in
this study and (c) calculated by Elgamal (1992).
along with the recorded response spectra. It is evident that the predicted
response in this work is in better agreement with the recorded one than the
response calculated by Elgamal (1992). The response spectrum of Elgamal
(1992), obtained from a 3D shear beam analysis (see Section 6.3), has a
narrower frequency content and higher amplifications at the significant fre-
quencies, whereas the broader frequency content of the spectrum from the
present study matches better the low frequency spectral ordinates. However,
it should be noted that none of the two calculated response spectra (i.e. from
this study and from Elgamal (1992)) match the spectral ordinates for higher
frequencies (smaller periods).
Those high accelerations for high frequencies in the response spectrum of
the recorded motion are believed to be originated from the high peak values of
acceleration observed in the recorded acceleration time-history at the crest of
the dam for EQ5. These high peaks are believed to be due to a localised slope
failure, as suggested by previous researchers (see also Section 6.3), which
was not predicted by the FE analysis in this study. However, if the high
frequencies are filtered from the recorded accelerations at the crest, a better
255
agreement is obtained between the recorded and the calculated response.

Figure 6.61 shows a comparison between the filtered recorded accelerations
at the crest for EQ5 and those calculated in this work, whereas Figure 6.62
shows the associated response spectra. The filtering was performed using
the software Seismosignal (Antoniou and Pinho, 2004), adopting a 4th order
Bandpass Butterworth filter for frequencies higher than 4 Hz (i.e. periods
smaller than 0.25s). It is shown that the peak values of recorded accelerations
vanish along with the observed asymmetry in the record leading to a better
agreement between the calculated and recorded accelerations time-histories.
Therefore, after the filtering, the predicted response spectrum compares also
very well with the one of the recorded motion.
It should be noted that the absence of any high peaks and any asymmetry
in the calculated accelerations at the crest for EQ5 is believed to be due to
the fact that no localised failure was predicted close to monitoring location
at the crest (point C in Figure 6.39).
Figures 6.63 (a) and (b) show the accumulated deformed mesh and vec-
tors of accumulated displacement respectively at the end of EQ5. It may be
observed that after the seismic event of EQ5 the dam is still operational (the
induced deformations are much smaller than the freeboard) and no major
movements occur (that would have altered the shape of the dam signifi-
cantly). No failure is indicated in the dam after the seismic events. On the
latter figure, the value and orientation of maximum displacement (which has
a value of 0.265m) is represented by a grey vector and it is located within
the clay core.
Figures 6.64 (a) and (b) show the sub-accumulated deformed mesh and
vectors of accumulated displacement respectively for the duration of EQ5.
The latter two figures refer to the deformations that occurred only due to
and during the EQ5 seismic event (i.e. excluding the deformations from
the previous static and EQ2 parts). It is shown that the deformations that
occurred during the EQ5 seismic event were concentrated on the upstream
part of the dam in the rockfill. No major failure is indicated, but some
slope movements. However, the magnitude of these deformations is still very
small and not comparable to the recorded vertical settlement of the crest
256

0.5
Recorded − Filtered
0.4 Calculated
0.3
Acceleration, a [g]
0.2
0.1
−0.1
−0.2
−0.3
0 10 20 30 40 50 60 70 80
Time, t [sec]
EQ5. Comparison between filtered recorded and calculated response.

1.5
Recorded − Filtered
Calculated
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
Figure 6.62: Response spectra (damping, ξ = 5%) at the crest of La Vil-

lita dam during EQ5. Comparison between filtered recorded and calculated
response.
257
during EQ5 (around 30cm). On the latter figure, the value and orientation
of maximum displacement (which has a value of 0.036m) is represented by
a grey vector and it is located at the upstream dam slope. This means that
the deformations resulting from the earthquake events are very small.
Figure 6.65 shows the contours of stress level, S, in the dam during EQ5.
The stress level, S is defined as the ratio of the current deviatoric stress, Jc ,
over the value of the deviatoric stress at yield conditions, Jy , at the same
value of the mean efective stress, p0 . Therefore, S, takes values from 0 to 1,
and shows how close the stress state of the soil is to yielding. It is shown
that the values of stress level go up to 0.8 (contour B) at some places within
the upstream rockfill, the downstream alluvium and the downstream dam
slope. However, S is not very close to 1 and therefore the soil in the dam
is generally found not to be at the yielding stress state after the end of the
earthquake.
Figures 6.66 and 6.67 show the calculated vertical displacements at the
crest (point C in Figure 6.39) during the EQ2 and EQ5 seismic events. It may
be observed that the magnitude of the calculated settlements is extremely
small compared to the recorded values for both EQs (around 2cm and 30cm
for EQ2 and EQ5 respectively). Furthermore, it should be noted that accord-
ing to the global deformations of the dam during EQ5 (Figures 6.63 (a) and
(b)), the positive values of settlements for EQ5 are originated from the fact
that uneven displacements occur on the crest of the dam, i.e. the upstream
part of the crest experiences negative (i.e. downwards) whereas the down-
stream part of the crest experiences positive (i.e. upwards) displacements.
However, note that all crest displacements are extremely small (i.e. < 2cm).
Figures 6.68 and 6.69 show the calculated vertical displacements at the
upstream corner of the crest (point U in Figure 6.39). As discussed earlier,
according to Figures 6.63 (a) and (b), the main deformations in the dam were
concentrated in the upstream rockfill and include the upstream corner of the
crest (point U in figure 6.39). It is shown that the upstream part of the crest
experiences higher values of settlement, perhaps due to those movements of
the upstream rockfill, but still their magnitude is very small compared to the
recorded values. It should also be noted that the recorded values refer to the
258
(a) (b)
Figure 6.63: End of the seismic analysis (after EQ5): (a) Accumulated (for
the whole static and dynamic analysis) deformed mesh and (b) Vectors of
accumulated displacement of La Villita dam.
259
(a) (b)
Figure 6.64: End of the seismic analysis (after EQ5): (a) Sub-accumulated
(i.e. only for EQ5) deformed mesh and (b) Vectors of sub-accumulated dis-
pacement of La Villita dam.
260
BA
C
A
B
A
B
B
A
A
B
A
A
B
C
B
D
A
Figure 6.65: End of the seismic analysis (after EQ5): Stress level, S, in La
Villita dam.
261
Settlement of the crest of La Villita dam − EQ2

0.03
Calculated
0.02
0.01
Settlement, d [cm]
−0.01
−0.02
−0.03
−0.04
0 1 2 3 4 5 6 7
Time, t [sec]
Figure 6.66: Vertical displacement time history at the crest (Point C in

Figure 6.39) of La Villita dam during EQ2.
Settlement of the crest of La Villita dam − EQ5

0.3
Calculated
0.25
0.2
0.15
Settlement, d [cm]
0.1
0.05
−0.05
−0.1
−0.15
−0.2
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 6.67: Vertical displacement history at the crest (Point C in Figure

6.39) of La Villita dam during EQ5.
262
Settlement of the upstream corner of the crest of La Villita dam − EQ2

0.04
Calculated
0.02
0
Settlement, d [cm]
−0.02
−0.04
−0.06
−0.08
−0.1
−0.12
0 1 2 3 4 5 6 7
Time, t [sec]
Figure 6.68: Vertical displacement time history at the upstream corner of

the crest (Point U in Figure 6.39) of La Villita dam during EQ2.
Settlement of the upstream corner of the crest of La Villita dam − EQ5

0.2
Calculated
0
−0.2
−0.4
Settlement, d [cm]
−0.6
−0.8
−1
−1.2
−1.4
−1.6
−1.8
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 6.69: Vertical displacement history at the upstream corner of the crest
(Point U in Figure 6.39) of La Villita dam during EQ5.
263
downstream side of the crest, where the monitoring instrument was installed.
6.6.2.2 Dynamic soil behaviour
A better understanding of the seismic response of the dam can be obtained

by observing the behaviour of the soil during the seismic events. Figure 6.70
shows the contours of accumulated deviatoric strain in the dam during EQ5
(the values on the plot correspond to natural strain and not percentage). It
may be observed that high values of accumulated deviatoric strain occur in
the upstream part of the dam and especially, close to the upstream face and
on the upstream face of the clay core (up to γ = 0.005). Besides, smaller
values of accumulated deviatoric strain occur at the downstream dam slope
(up to γ = 0.001) This is in agreement with Figure 6.63 which shows the
deformations of the dam after the earthquake, and where movements of the
upstream rockfill are observed. However, the soil in the dam is generally
found not to be at the yielding stress state after the end of the earthquake.
A further insight may be obtained by looking at the behaviour of indi-
vidual points in the dam. Figure 6.71 shows the location of the monitoring
points, which is the first integration (Gauss) point of each of the following
elements: in the upstream rockfill (UR), clay core (CC), downstream rockfill
(DR) and downstream alluvium (DA).
Figures 6.72 - 6.75 show the shear stress-strain, τ −γ (referring to τxy −γxy )
response in the upstream (UR) and downstream rockfill (DR), clay core (CC)
and downstream alluvium (DA) for EQ5 (see Figure 6.71 for the locations). It
may be observed that the behaviour in the clay core (Figure 6.73) was found
to be generally linear; whereas the behaviour in the rockfill (Figures 6.74 &
6.75) was found to be slightly nonlinear and significantly nonlinear in the
alluvium (Figure 6.72). This is in agreement with the model calibration and
the induced shear strains. The CNL model for the clay core was calibrated
against the curves of Vucetic and Dobry (1991) (Figure 6.44), whereas for the
rockfill and the alluvium, the curves of Rollins et al. (1998) (Figures 6.48)
were used (see Section 6.6.1). The latter curves (Rollins et al., 1998) exhibit
more nonlinear response than the former curves (Vucetic and Dobry, 1991),
264
B
A
A
B
D
C
B
A
B
C
A
D
B
C
A
B
A
C
D
B
A
E
F
Figure 6.70: End of the seismic analysis (after EQ5): Accumulated total
deviatoric strain, Ed in La Villita dam.
265
CC
UR DR
DA
Figure 6.71: Sketch of the dam showing the locations of the monitoring
elements: upstream rockfill (UR), clay core (CC), downstream rockfill (DR)
and downstream alluvium (DA).
Shear stress−strain response at the downstream alluvium of La Villita dam during EQ5
250
200
150
100
Shear stress, τ [kPa]
50
−50
−100
−150
−200
−250
−0.04 −0.02 0 0.02 0.04 0.06 0.08 0.1
Figure 6.72: Shear stress-strain response at the downstream alluvium (DA)

of La Villita dam during EQ5
266
Shear stress−strain response in the clay core of La Villita dam during EQ5
150
100
50
−50
−100
−0.015 −0.01 −0.005 0 0.005 0.01 0.015 0.02 0.025
Figure 6.73: Shear stress-strain response in the clay core (CC) of La Villita
dam during EQ5
Shear stress−strain response in the upstream rockfill of La Villita dam during EQ5
150
100
50
−50
−100
−0.02 −0.01 0 0.01 0.02 0.03 0.04
Figure 6.74: Shear stress-strain response in the upstream rockfill (UR) of La

Villita dam during EQ5
267
Shear stress−strain response in the downstream rockfill of La Villita dam during EQ5
100
50
−50
−100
−150
−0.04 −0.03 −0.02 −0.01 0 0.01 0.02 0.03
Figure 6.75: Shear stress-strain response in the downstream rockfill (DR) of

La Villita dam during EQ5
Shear stress−strain response at the downstream alluvium of La Villita dam during EQ2
80
60
40
20
−20
−40
−60
−0.015 −0.01 −0.005 0 0.005 0.01 0.015
Figure 6.76: Shear stress-strain response at the downstream alluvium (DA)

of La Villita dam during EQ2
268
Shear strain time−history in the upstream rockfill of La Villita dam during EQ5
0.04
0.03
0.02
0.01
−0.01
−0.02
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 6.77: Shear strain time-history in the upstream rockfill (UR) of La

Shear strain time−history in the downstream rockfill of La Villita dam during EQ5
0.03
0.02
0.01
−0.01
−0.02
−0.03
−0.04
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 6.78: Shear strain time-history in the downstream rockfill (DR) of La

269
Stress paths in the upstream rockfill of La Villita dam

350
Mohr−Coulomb Yield Function
300 EQ5
EQ2
Consolidation between EQ2 & EQ5
Deviatoric stress, J [kPa]
250
Start of Reservoir Impounding
Static part of analysis
200
150
100
50
0
0 50 100 150 200 250 300 350 400
Mean effective stress, p‘ [kPa]
Figure 6.79: Stress paths in the upstream rockfill (UR) of La Villita dam
Stress paths in the downstream rockfill of La Villita dam

450
400 EQ5
EQ2
350
300 Start of Reservoir Impounding

250
200
150
100
50
0
0 100 200 300 400 500 600
Figure 6.80: Stress paths in the downstream rockfill (DR) of La Villita dam
270
i.e. the shear stiffness, G(γ) degrades more.

The soil response for EQ2 was generally less nonlinear, due to the smaller
intensity and duration of EQ2 as compared to EQ5. As an example, the shear
stress-strain response in the downstream alluvium (DA) (which was found
to exhibit significant nonlinearity during EQ5) is shown in Figure 6.76. It is
shown that the significantly smaller shear strains induced are not adequate
to cause nonlinearity in the soil behaviour and therefore the response is very
close to being linear.
The difference in the τ − γ response of the upstream and downstream
rockfill (Figures 6.74 and 6.75), could be explained by looking at the time-
history of the shear strain, γ(t). Figures 6.77 and 6.78 show the shear strain
time-histories in the upstream and downstream rockfill respectively. It may
be observed that a value of a “residual” permanent shear strain of around
0.01% was developed during EQ5, from the first 20s.
The last observation can be better understood by looking at the stress
paths at several points of the soil. Figures 6.79 and 6.80 show the stress paths,
J − p0 (see Equations 3.27 and 3.33 in Section 3.7.5 for their definition), for
elements in the upstream (UR) and downstream (DR) rockfill respectively.
The figures include the stress paths for the whole analysis, i.e. the static
and dynamic (both EQ2 and EQ5) parts. On the same figures, the Mohr-
Coulomb yield surface (YS) is also plotted. It should be noted that the
Mohr Coulomb YS is not constant, but changes with the Lode’s angle, θ
(see Section 3.7.5 and Equations 3.32 - 3.35). In Figures 6.79 and 6.80, the
YS plotted corresponds to the point that the YS was first engaged, i.e. the
first time that plastic strain developed at that point. However, it should
be commented that the YS was found not to change considerably during the
analysis. It is shown that for the upstream rockfill, the stress path approaches
the yield surface and travels along it at around J = 100 kPa and p0 = 120
kPa. This means that at that instant plasticity is introduced, which leads to
permanent values of shear strain, as shown in Figure 6.77.
Plasticity is not introduced in the downstream rockfill as the stress paths
are far away from the yield surface. It should be noted that the reservoir
impoundment causes a change in the direction of the stress path, as it results
271
in smaller values of the mean effective stress, p0 , and brings the stress path
closer to the yield surface. That is why more plasticity is introduced in the
upstream rockfill.
This seems to be a drawback of the constitutive models adopted in this
study, which introduce plasticity only when the yield surface is engaged. In
contrast, kinematic hardening models (such as Grammatikopoulou (2004))
are able to introduce plasticity much earlier, as they have a much smaller
elastic area which is more realistic. This is in agreement with earlier obser-
vations of Kontoe et al. (2011) who compared constitutive assumptions of
different levels of complexity. They commented that the prediction of plastic
strains by the constitutive assumptions used in their study was dependent
on the proximity of the initial stress state of the soil (i.e. at the beginning
of the earthquake) to the yield surface. This highlights the importance of
appropriate modelling of the construction sequence before the earthquake.
Finally, considering the results of this analysis, the constitutive assump-
tions used in this study seem to be incapable of taking account of the effects
of the previous stress history (i.e. a sequence of seismic events) and the re-
sulting development of plasticity properly. This issue is addressed in more
detail in the next chapter and particularly in Section 7.2.
6.6.3 Comments
From the dynamic analysis of La Villita dam, it was found that the three-
dimensional geometry of the surrounding canyon has an influence on the
dynamic characteristics (fundamental period and amplification of accelera-
tions) of the dam. As the current work did not involve 3D analysis, the
stiffening effect in a two-dimensional analysis was modelled by increasing
the material stiffness in order to explicitly take account of the additional
geometric stiffness from the canyon.
The dynamic response of the dam was well captured as evidenced by
the comparison of the accelerations at the crest and the resulting response
spectra. The calculated response spectrum for EQ5 at the crest of the dam
from this work was compared to that of Elgamal (1992) who performed three-
272
dimensional shear beam analysis of La Villita dam and was found to have a
better agreement with the recorded response spectrum.
However, the high peak values of acceleration observed in the recorded
acceleration and already attributed by previous researchers to a localised fail-
ure close to the monitoring instrument were not captured. Such a localised
failure was not predicted, from the inspection of the plots of the deformed
mesh and vectors of displacement after the earthquake. Nevertheless, when
the high frequencies (originating from the localised slip failure) from the
recorded accelerations were filtered, an excellent agreement was obtained be-
tween the calculated and recorded accelerations and response spectra. This
shows that the dynamic response of the dam was well captured and the fre-
quency content of the resulting calculated accelerograph at the crest matched
that of the recorded accelerograph.
The calculated vertical displacements at the crest of the dam were not
comparable to the recorded values. As mentioned earlier, a localised slip
failure that would result into high values of settlements was not predicted
and therefore no high values of settlements were observed. Minor deforma-
tion in the upstream rockfill showed some higher values of settlement at the
US corner of the crest, but still not large enough to be comparable to the
recorded crest settlement values. It should also be noted that both the ob-
served asymmetry in the recorded crest accelerations and the location of the
monitoring instrument suggest a localised failure on the downstream side of
the crest.
It is suggested that more research is required to be carried out in order
to further investigate the pattern of the recorded settlements. It is believed
that the failure of this study to predict the recorded displacements could be
due to:
• the possible existence of a discontinuity in the embankment, and
• the constitutive modelling assumptions adopted in this work.
Regarding the first possible reason, presumably some minor localised fail-
ure at the position close to the monitoring instrument may had happened
273
before which may had created a local discontinuity. This could have origi-
nated possibly from a previous seismic event or even from some construction
processes. Besides, it should be reminded that there is a public road on top
of the crest of La Villita dam and perhaps heavy vehicle traffic may have
caused some movements in the soil. If such a discontinuity exists, it will
form a weak zone in the dam and will be sensitive to seismic loads. As far
as modelling is concerned, the adopted procedure in this work assumed a
continuous soil profile, since no weak zones were introduced with interface
elements, such as those of Gazetas and Uddin (1994).
Regarding the second possible reason, perhaps the constitutive modelling
assumptions adopted in this work, while adequate to capture the dynamic
response of the dam (accelerations and fundamental period); were not the
most appropriate for predicting permanent movements. As observed from the
stress paths (Figures 6.79 and 6.80), plasticity (and permanent strains) were
introduced only when the yield surface was engaged, whereas the strains were
recoverable for the rest of the analysis. No pre-yield plasticity was introduced
although this is more consistent with the real behaviour of soils. Such a mod-
elling concept would introduce more strains and hence more movements (i.e.
larger settlements) should it was adopted. Moreover, kinematic hardening
and softening (i.e. change of the yield surface) would perhaps indicate some
more plasticity.
Although the yield function adopted (Mohr-Coulomb) was allowed to
change with the Lode’s angle, θ (i.e. the relative magnitude of the principal
stresses, σ1 , σ2 and σ3 , see Section 3.7.5 and Equations 3.32 - 3.35), this
change was perhaps still not sufficient. Better results could perhaps be ob-
tained should a more advanced constitutive model of the Critical State type
(Schofield and Wroth, 1968), such as the Modified Cam Clay (Roscoe and
Burland, 1968) was employed, as proved by earlier studies on embankments
(Zdravković et al., 2002).
A further improvement could also be the adoption of a kinematic harden-
ing (also known as “bubble”) model (KHM), such as that of Grammatikopoulou
(2004) which allows introduction of pre-yield plasticity and thus it pro-
vides a better prediction of deformations (Grammatikopoulou et al., 2007,
274
2008). This advanced model has already been applied in the static (Gram-
matikopoulou et al., 2008) and seismic analysis of tunnels (Kontoe et al.,
2009, 2011). Those studies showed that this KHM is able to provide bet-
ter predictions of strains and displacements (as compared to recorded values
from the field) than the constitutive approach used in this study, i.e. a small-
strain stiffness degradation model (such as the Logarithmic CNL (Taborda,
2011) or the Jardine et al. (1986) small-strain model) combined with a simple
failure criterion (Mohr-Coulomb or Modified Cam Clay).
Finally, a parametric analysis with different yield strengths of the rockfill
(i.e. different values of the cohesion, c0 and the angle of shearing resistance,
φ0 ) could investigate the dependency of the magnitude of displacements on
the yield strength of the rockfill.
6.7 Obstacles in the analysis of the dam

As mentioned earlier, La Villita dam is considered a reasonably well-documented
case study because of available data (material properties and field measure-
ments) and because of the existence of various researchers who analysed its
response. However, the amount and quality of available data is limited and
not sufficient for a full advanced numerical analysis.
Firstly, there were no specialist (laboratory or field) tests for the advanced
characterisation of the soil materials. Although some elastic (ρ, ν, Gmax ) and
strength (c0 and φ0 ) properties of the materials were available, no information
was found regarding the compressibility of the materials (such as κ and λ)
or the stiffness degradation properties (G − γ). The constitutive models used
in the analysis did not require data regarding the compressibility of the soil,
but they required information about stiffness degradation. For this reason,
the models were calibrated against published empirical stiffness degradation
curves (Vucetic and Dobry, 1991; Seed et al., 1986; Rollins et al., 1998) which
are considered representative for those types of materials.
In addition, information regarding the permeability of the materials,
which is useful for a coupled hydro-mechanical analysis, was not available.
Therefore, representative values of permeability of the clay core (10−10 m/s)
275
and the alluvium (10−7 m/s) were adopted and used in the analysis. How-
ever, because of the lack of relevant data for the rockfill shoulders, these parts
of the dam were assumed to have fully drained materials during the entire
analysis.
Moreover, available field measurements were very limited. Out of the
six earthquake that were monitored between 1975-1985 (see Table 6.1), only
two input motions (EQ2 and EQ5) are useful for numerical analysis (as the
rest were incomplete). Also, for those two seismic events, the full input
bedrock motion of EQ2 was not available and also the acceleration record
at the downstream berm for EQ5 was not available at all. For this reason,
only two seismic events were used in the analysis and the remaining events
(EQ1, EQ3, EQ4 and EQ6) were not included in the analysis. An ideal back
analysis of the response of the dam would require the consideration of all
the earthquakes in order to better assess the effect of loading history on the
seismic response of earth dams.
6.8 Conclusions
This chapter describes the work carried out related to the numerical analysis
of the static and dynamic behaviour of La Villita dam. This dam is consid-
ered to be a well-documented case study because of available information,
i.e. material properties and monitored response.
The dam was analysed in both static and dynamic conditions, considering
the nonlinear elasto-plastic soil behaviour. Two-dimensional plane-strain
coupled static and dynamic finite element analyses were performed.
The (dynamic) stiffening effect of the 3D narrow canyon geometry was
taken into consideration in a 2D dynamic analysis by increasing the material
stiffness of the dam. The computed accelerations and associated response
spectra were found to be in good agreement with the corresponding avail-
able recorded values. Also, a better prediction of the response spectra was
achieved than that of Elgamal (1992).
However, the high peak values of acceleration observed in the recorded
acceleration (which were attributed to a localised failure) were not captured
276
as the current FE did not predict any localised failure close to the dam crest
and the calculated crest settlements were much smaller than the recorded
values. A possible pre-existing discontinuity in the downstream embankment
crest soil profile would form a weak zone which would be very sensitive to
any seismic loads. Such a discontinuity was not artificially modelled in this
work (through the introduction of interface elements in the soil profile).
The soil in the upstream rockfill experienced higher values of shear strains
and larger deformations than that in the downstream rockfill. This was due
to the reservoir impoundment during which the stress paths were directed
closer the yield surface, because of the reduction in the mean effective stress,
p0 . Finally, it is believed that an improved prediction of the deformations of
the dam may be obtained by using more advanced constitutive models (such
as kinematic hardening) able to predict plasticity more realistically.
277
Chapter 7
PARAMETRIC SEISMIC
ANALYSIS OF LA VILLITA
DAM
7.1 Introduction
This chapter contains several parametric studies performed to assess the
effects of different modelling approaches on the predicted seismic response of
earth dams. The already studied in Chapter 6 case of La Villita dam is used
as the basis for this investigation.
The examined issues are (a) the effects of any previous seismic activ-
ity prior to the studied earthquake, (b) the relative merits of coupled and
uncoupled dynamic analysis, (c) foundation-dam interaction and modelling
approaches, (d) nonlinear reservoir-dam interaction and (e) stiffness inho-
mogeneity. Previous results from the literature are also discussed (where
available) and compared to the outcomes of this study. Finally, comments
are made regarding the applicability of the adopted modelling technique and
constitutive assumptions.
278
7.2 Previous earthquake history

7.2.1 Scenarios considered
The aim of this section is to assess the effect of any previous seismic activity
prior to the examined earthquake. The analysis performed in Chapter 6
consisted of two earthquakes, a small, in terms of both intensity and duration
(EQ2) and a large (EQ5) seismic event. Three scenarios are considered here,
which start from the same static analysis file and are listed below:
• A - EQ5 without any previous earthquake
• B - EQ5 after a small earthquake, which is EQ2 (see Chapter 6)
• C - EQ5 after a large earthquake, which is EQ5
7.2.2 Dynamic response of the dam

Figures 7.1 and 7.2 compare the acceleration at the crest of the dam of
scenario B with scenarios A and C respectively, while Figures 7.3 and 7.4 show
the corresponding response spectra. It is shown that no major difference is
observed on the accelerations or the response spectra of the crest for the three
different scenarios considered. This implies that in the scenarios considered,
the previous (seismic) loading history did not have any significant effect on
the dynamic response of La Villita dam.
Figures 7.5 and 7.6 compare the vertical displacements at the crest of the
dam of scenario B with scenarios A and C respectively. It is shown that mi-
nor differences occur in the displacements for the three scenarios considered,
i.e. the previous stress history has some effect on the calculated crest dis-
placements. It seems that Scenario C, i.e. with two strong earthquakes leads
to the largest crest vertical displacements, whereas Scenario A, i.e. the case
without any previous earthquake leads to the smallest crest displacements.
However, it should be noted that both the magnitude of the displacements
and their differences are very small (fractions of a centimetre).
279

0.4
B − EQ5 after EQ2
A − EQ5 without previous EQ
0.3
0.2
Acceleration, a [g]
0.1
−0.1
−0.2
−0.3
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.1: Acceleration response at the crest of La Villita dam during EQ5
- Comparison between Cases A and B

0.4
B − EQ5 after EQ2
0.3 C − EQ5 after EQ5
0.2
Acceleration, a [g]
0.1
−0.1
−0.2
−0.3
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.2: Acceleration response at the crest of La Villita dam during EQ5
- Comparison between Cases B and C
280

1.5
B − EQ5 after EQ2
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

dam during EQ5 - Comparison between Cases A and B

1.5
B − EQ5 after EQ2
C − EQ5 after EQ5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

dam during EQ5 - Comparison between Cases B and C
281
Vertical displacements at the crest of La Villita dam − EQ5

0.3
0.2
0.1
Displacement, d [cm]
−0.1
−0.2 B − EQ5 after EQ2

−0.3
−0.4
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.5: Vertical displacements at the crest of La Villita dam during EQ5
- Comparison between Cases A and B
Vertical displacements at the crest of La Villita dam − EQ5

0.5
0.4
0.3
Displacement, d [cm]
0.2
0.1
0 B − EQ5 after EQ2

C − EQ5 after EQ5
−0.1
−0.2
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.6: Vertical displacements at the crest of La Villita dam during EQ5
- Comparison between Cases B and C
282
7.2.3 Dynamic soil behaviour

A further insight into the dynamic behaviour of the dam can be obtained by
examining the dynamic soil behaviour. Figures 7.7 and 7.8 show the shear
stress-strain time histories at the downstream alluvium of La Villita dam
(first integration point of the element DA shown in Figure 6.71) for Scenar-
ios A and C respectively and their comparison with Scenario B. Likewise,
Figures 7.9 and 7.10 show the pore water pressure time-history in the down-
stream alluvium of the dam for the previously-mentioned scenarios. It may
be observed that the differences are extremely small, i.e. the previous load-
ing history had an insignificant effect on the induced shear strains and pore
water pressures. In both cases the values were slightly higher for Scenario C.
Moreover, Figures 7.11 - 7.13 show the stress paths, J − p0 in the DA for
the three scenarios considered. On the same figures, the Mohr-Coulomb yield
surface (YS) is also plotted. It should be noted that the Mohr Coulomb YS
is not constant, but changes with the Lode’s angle, θ (see Section 3.7.5 and
Equations 3.32 - 3.35). In those figures, the YS is plotted at inclination g(θ)
which corresponds to the magnitude of θ at which the YS was first engaged,
i.e. the first time that plastic strain developed in the stress path. However,
it should also be noted that the slope of the YS was found not to change
considerably during the analysis.
As expected, the stress paths for the static part of the analysis are the
same (as all three scenarios start from the same static analysis file). It is
shown that for all three scenarios considered, once the stress paths reached
the yield surface, the following stress states travelled along the YS until a
reverse in the load was engaged after which the stress path travelled back into
the elastic region below the YS. This shows that in this case, the previous
earthquake history did not have any significant effect on the stress paths (i.e.
for example, they have not moved considerably away from the yield surface)
and that is why the response was extremely similar.
Figures 7.14 (a) and (b) show the contours of stress level, S, within
the dam for Scenarios A and C. The corresponding plot for Scenario B was
shown in Figure 6.65 (b) in Chapter 6. The parameter S was defined earlier
283
Shear strain time−history in the downstream alluvium of La Villita dam during EQ5
0.1
0.08
0.06
0.04
0.02

−0.04
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.7: Shear strain time-history in the downstream alluvium (DA) of

Shear strain time−history in the downstream alluvium of La Villita dam during EQ5
0.1
0.08
0.06
0.04
0.02

C − EQ5 after EQ5
−0.04
−0.06
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.8: Shear strain time-history in the downstream alluvium (DA) of

284
Pore water pressure time−history in the downstream alluvium of La Villita dam during EQ5
100
B − EQ5 after EQ2
80 A − EQ5 without previous EQ
60
Pore water pressure, u [kPa]
40
20
−20
−40
−60
−80
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.9: Pore water pressure time-history in the downstream alluvium

(DA) of La Villita dam during EQ5
Pore water pressure time−history in the downstream alluvium of La Villita dam during EQ5
100
B − EQ5 after EQ2
80 C − EQ5 after EQ5
60
40
20
−20
−40
−60
−80
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.10: Pore water pressure time-history in the downstream alluvium

(DA) of La Villita dam during EQ5
285
Stress paths in the downstream alluvium of La Villita dam

700
600 EQ5 − second seismic event
EQ2 − first seismic event
500
400
300
200
Scenario AB
100
0
0 100 200 300 400 500 600 700 800 900
Figure 7.11: Stress paths in the downstream alluvium (DA) of La Villita dam
during EQ5 - Scenario B

700
600 EQ5 − seismic event
500
400
300
200
100 A
Scenario B
0
0 100 200 300 400 500 600 700 800 900
during EQ5 - Scenario A
286

700
600 EQ5 − second seismic event
EQ5 − first seismic event
Consolidation between seismic events
500
400
300
200
100 Scenario C
0
0 100 200 300 400 500 600 700 800 900
during EQ5 - Scenario C
and it shows how close the stress state of the soil is from the yield condition.
Comparing all three plots, it should be mentioned that only minor differences
exist for the three scenarios considered in this study. However, one could
notice that slightly larger values of S (up to 0.9) are observed in the upstream
rockfill for Scenario C (Figure 7.14 (b)) but they are in a very localised area
around the downstream berm. This was expected as the latter scenario
involved higher values of acceleration and therefore more engagement of the
yield surface.
7.2.4 Comments
It was shown that the effects of the previous loading history were gener-
ally found to be very small in the case of La Villita dam. For the cases
considered, no difference was observed in the crest accelerations and the as-
sociated response spectra. However, minor differences were observed in the
crest settlements, the induced shear strains and the pore water pressures. It
is believed that this happened for three reasons: (a) none of the scenarios
considered a major event which would have induced some significant failure
287
B C
C
B
A
A
B
A
A
B
A
A
A
A
B
A
A
B
B
C
B
A
A
B
A
A
A
A
A
B
B
C
D
A
B
C
D
B
E
A
(a) Case A (b) Case C
Figure 7.14: Contours of stress level, S for (a) Case A - EQ5 without any
previous earthquake, and (b) Case C - EQ5 after a large earthquake, which
is EQ5.
288
in the dam, prior to EQ5, (b) the nature of the materials of La Villita dam
and (c) the constitutive assumptions used in this study.
As far as the first reason is concerned, should an extremely large seismic
event (i.e. larger than EQ5) have been modelled prior to EQ5, the higher
values of plasticity would have induced large deformations and altered the
geometry of the dam significantly. In that case, the wave propagation would
have resulted in a different dynamic response of the dam structure. However,
none of the three scenarios considered such an extremely large seismic event,
and therefore the analysis of EQ5 in all three scenarios started from similar
conditions. Also, a large displacement analysis would be required to capture
the effects of a significantly altered dam geometry.
As far as the second issue is concerned, it should be noted that the clay
core constitutes of only a small part of the dam, whereas the rest consists of
sand filters and rockfill shells which were modelled (and expected to behave)
as drained materials. Therefore, only the clay core and the alluvial foun-
dation were considered as consolidating materials and were expected to be
more affected by the previous stress history. The effects of the previous stress
history would be more highlighted if all the parts of the dam were modelled
as consolidating materials (with appropriate values of permeability).
Moreover, as far as the last issue is concerned, the adopted CNL model
with a simple Mohr-Coulomb failure criterion is perhaps not adequate in
modelling rigorously the previous stress history. This was also discussed
earlier in Chapter 6, and was found inadequate to capture the recorded crest
settlements of La Villita dam. The inability of the adopted constitutive
assumptions to model pre-yield plasticity could be the main reason for the
minor differences between the three scenarios considered here. Should a
kinematic hardening model be adopted, more differences might be observed,
as plasticity would be introduced earlier in the analysis, and not only during
the limited occasions that the stress paths approached the Mohr-Coulomb
yield surface.
289
7.3 Uncoupled analysis

The aim of this section is to examine the effects of using a coupled formulation
for performing dynamic analysis of dams. The dynamic analysis of La Villita
dam described in Chapter 6 and the further two analyses described earlier in
Section 7.2 adopted a coupled formulation of the equation of motion for the
soil skeleton and Darcy’s law for the behaviour of the pore fluid (see Potts
and Zdravković (1999) for details about coupled analysis in soil mechanics).
More details about coupled formulation in dynamic analysis can be found in
Kontoe (2006) and Zienkiewicz et al. (1999).
A coupled hydro-mechanical formulation which accommodates modelling
of consolidation allows a better prediction of pore water pressures during a
dynamic analysis as the dissipation of the water is affected by the appropri-
ately specified permeability of the material. Here, an examination was car-
ried out to assess whether this coupled formulation provided different results
than an uncoupled formulation. In this case, the clay core and foundation
alluvium were modelled as undrained materials.
7.3.1 Description of analysis

In this section another dynamic analysis was performed in which the be-
haviour of the clay core and the alluvial foundation (both of which were
treated earlier as consolidating materials) were modelled as undrained ma-
terials. This analysis was uncoupled (in contrast to the previous coupled
analyses that considered the coupling between the solid and liquid phase of
the soil material) and therefore did not take account of water seepage through
the soil. The undrained behaviour of the clay and alluvium was achieved by
prescribing the bulk modulus of the pore fluid equal to Kf = 2.2 · 106 kPa,
which is the real value of the bulk modulus of water. The rest of the ma-
terials (sand filters and rockfill shells) were treated as drained materials, as
before, i.e. their pore fluid bulk modulus was set equal to zero. Therefore,
the difference between the coupled and the uncoupled analyses is that in the
former case, water seepage within the clay core and the foundation alluvium
is affected by the specified permeability of those materials.
290
The investigation was concerned with both EQ2 and EQ5 seismic events.
Therefore, two separate uncoupled analyses were carried out, one for EQ2
and one for EQ5 and they both started immediately after the static analysis.
The static analysis was the same as the one used for the previous analyses,
i.e. Chapter 6 and Section 7.2. Therefore, the uncoupled analysis for EQ2
(described in this section) is compared with the coupled analysis of EQ2
described in Chapter 6. The uncoupled analysis for EQ5 (described in this
section) is comparable to the analysis of Scenario A of Section 7.2, as they
both analyse EQ5 without any previous earthquake.

Figures 7.15 and 7.17 show the acceleration time-histories at the crest of
the dam for EQ2 and EQ5 respectively whereas Figures 7.16 and 7.18 show
the corresponding response spectra for both seismic events. As it may be
observed from the figures, there is barely any difference in the crest accelera-
tions and the associated response spectra between the coupled and uncoupled
analyses for both earthquakes.
Figures 7.19 and 7.20 show the vertical profiles of the maximum values
of horizontal displacement in the core of the dam for EQ2 and EQ5 respec-
tively, whereas Figures 7.21 and 7.22 show the corresponding profiles of the
maximum values of horizontal acceleration for both seismic events. It may
be observed from these figures that there is only a slight difference between
the two analyses, with the coupled analysis predicting slightly larger values
of horizontal accelerations and displacements. However, these differences are
negligible (less than 4%).
7.3.3 Dynamic soil behaviour

A better understanding of the different analysis approaches may be obtained
from the behaviour of the soil materials. Figures 7.23 and 7.24 show the
shear stress-strain, τ − γ curves in the clay core (first integration point of
element CC in Figure 6.71) of the dam during EQ5 for the coupled and the
291

0.15
Coupled Analysis
Uncoupled Analysis
0.1
0.05
Acceleration, a [g]
−0.05
−0.1
−0.15
−0.2
0 1 2 3 4 5 6 7
Time, t [sec]
Figure 7.15: Acceleration time-history at the crest of La Villita dam during

EQ2.

0.7
Coupled Analysis
Uncoupled Analysis
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
Figure 7.16: Response spectra (damping, ξ = 5%) the crest of La Villita dam
during EQ2.
292

0.4
Coupled Analysis
Uncoupled Analysis
0.3
0.2
Acceleration, a [g]
0.1
−0.1
−0.2
−0.3
0 10 20 30 40 50 60 70
Time, t [sec]

EQ5.

1.5
Coupled Analysis
Uncoupled Analysis
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
during EQ5.
293
Profile of the maximum horizontal displacements − EQ2

60
40 Coupled Analysis
Uncoupled Analysis
20
Elevation, z [m]
−20
−40
−60
−80
1 2 3 4 5 6 7 8 9 10
Displacement, d [m] −3
x 10
Figure 7.19: Vertical profile of the maximum values of horizontal displace-

ment in the core of La Villita dam during EQ2.

60
40
20
Elevation, z [m]
−20
−40
−60 Coupled Analysis

Uncoupled Analysis
−80
0.13 0.135 0.14 0.145 0.15 0.155 0.16 0.165
Displacement, d [m]

294
Profile of the maximum horizontal accelerations − EQ2

60
40
20
Elevation, z [m]
−20
−40
−60 Coupled Analysis

Uncoupled Analysis
−80
0.04 0.06 0.08 0.1 0.12 0.14 0.16
Acceleration, a [g]
Figure 7.21: Vertical profile of the maximum values of horizontal acceleration

in the core of La Villita dam during EQ2.

60
Coupled Analysis
40 Uncoupled Analysis
20
Elevation, z [m]
−20
−40
−60
−80
0.1 0.15 0.2 0.25 0.3 0.35
Acceleration, a [g]

295
uncoupled analyses respectively. It may be observed that the stress-strain

response in both analyses is very similar.
Looking further at the shear strain time-histories, shown in Figure 7.25
for both approaches. Minor differences are observed, particularly at the end
of the earthquake, with the uncoupled analysis exhibiting slightly higher
values of shear strain. However, it should be noted that the differences
are very small. What shows some difference however, is the induced pore
pressures, with time-histories as shown in Figure 7.26. It is shown that the
uncoupled analysis predicted higher fluctuation of the values of pore water
pressure although the average value seems to be the same. Therefore, these
differences are not considered to be very big. This might be due to the fact
that the permeabilities of the clay core and the alluvium are very small and
therefore the coupled analysis simulated undrained behaviour for those two
materials, i.e. what was simulated in the uncoupled analysis.
7.3.4 Comments
The dynamic response of the dam, if coupled or uncoupled analysis is per-
formed, was found to be almost identical, as evidenced by the calculated
accelerations at the dam crest, the associated response spectra and displace-
ments. Also, the dynamic behaviour of the soil was found to be very similar,
considering the shear strains (shear-stress strain response and shear stress
time-history). However, small differences were observed in the calculated
pore water pressures, where the uncoupled analysis predicted higher fluctu-
ations of the pore pressure.
It was shown that although an uncoupled dynamic analysis may predict a
very similar seismic response of an earth dam, it may however have different
pore water pressures than a coupled dynamic analysis. The fact that the dy-
namic response of the dam is very similar could be attributed to two reasons:
(a) the consolidating materials (i.e. those for which a coupled formulation
was adopted in the coupled analysis, namely clay core and foundation al-
luvium, and which are expected to show some difference if a coupled or an
uncoupled undrained analysis is performed) consisted only a small part of the
296
200
Uncoupled analysis
150
100
50
−50
−0.01 −0.005 0 0.005 0.01 0.015 0.02 0.025 0.03
Figure 7.23: Shear stress-strain response in the clay core of La Villita dam
during EQ5
200
Coupled analysis
150
100
50
−50
−100
−0.01 −0.005 0 0.005 0.01 0.015 0.02 0.025 0.03
Figure 7.24: Shear stress-strain response in the clay core of La Villita dam
during EQ5
297
Shear strain time−history in the clay core of La Villita dam during EQ5
0.03
0.025
0.02
0.015
0.01
0.005
−0.005 Coupled analysis

Uncoupled analysis
−0.01
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.25: Shear strain time-history in the clay core of La Villita dam
during EQ5
Pore water pressure time−history in the clay core of La Villita dam during EQ5
100
Coupled analysis
50 Uncoupled analysis
−50
−100
−150
−200
−250
−300
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.26: Pore water pressure time-history in the clay core of La Villita
dam during EQ5
298
dam geometry, and (b) the values of the permeability of the consolidating
materials adopted in this study were not adequately large to allow signif-
icant seepage and highlight the differences of a coupled and an uncoupled
undrained analysis.
Regarding the first reason, the differences between the two analyses (cou-
pled and uncoupled) would be more visible if all the dam materials were
modelled as consolidating in the coupled analysis. In that case, appropriate
values of permeability (e.g. for rockfill) should be adopted so that seepage
and dissipation of water pressures are correctly modelled.
As far as the second reason is concerned, larger differences (at least for the
pore water pressures) would be expected if the permeability of the materials
was larger and therefore allowed for some seepage in the dam and changes in
the water pressures. Therefore, it is suggested that in order to fully examine
the differences of performing a coupled and an uncoupled analysis, a dam
consisting entirely of consolidating materials should be considered and more
permeable materials (such as rockfill shoulders) should be included in the
coupled formulation.
7.4 Foundation-dam interaction

This section examines the soil-structure interaction (SSI) effects between a
dam and its foundation, i.e. foundation-dam interaction, using La Villita
dam as a case study. Previous studies reported in the literature related
to SSI effects in dams were discussed in Section 2.4.4, and they identified
as the main effect, the “softening” of the response of the dam structure,
i.e. elongation of its fundamental period and change in the amplification of
accelerations.
7.4.1 Analysis of a dam-foundation system

The dynamic analysis of La Villita dam described earlier in Section 6.6 is
further investigated. The acceleration response at three points having the
same elevation are compared. As shown in Figure 7.27, the points of interest
299
were: (a) free-field (FF), (b) downstream dam toe (DDT) and (c) dam base
(DB).
DB
DDT FF
Figure 7.27: Mesh of the dam showing the locations of the monitoring points:
Dam base (DB), Downstream dam toe (DDT) and Free-field (FF).
Figures 7.28 and 7.29 show the acceleration time-histories at the three
monitoring points (FF, DDT and DB) for EQ2 and EQ5 respectively, whereas
Figures 7.30 and 7.31 show the associated response spectra for the two earth-
quakes. From the figures of the accelerations, it may be observed that the
records at the FF (for both earthquakes) exhibit a higher frequency content,
whereas the records at the DB show larger periods of vibration.
This can be more clearly observed in the plots of the response spectra.
For higher periods (such as T=0.7s), the DB record has the highest spectral
ordinates whereas the FF record has smaller ordinates. At smaller values
of the fundamental period (such as T=0.2s), the opposite is observed, i.e.
the DB record has the smallest and the FF record has the largest spectral
ordinates. In both cases, i.e. at T = 0.2 and 0.7s, the DDT has spectral
ordinates between the two other records.
This means that for the DB record results in higher acceleration values at
longer fundamental periods, whereas the FF record at shorter fundamental
periods. Therefore, the DB record has a high-frequency content as compared
to the FF record which exhibits a low-frequency content. Practically, this
means that the presence of the dam results in a softer soil layer response
than the free-field soil layer response. On one hand, the presence of a large
structure (such as a massive earth dam) on top of a uniform soil layer would
provide some additional geometric stiffness to the soil layer, as the upper
300
Acceleration response − EQ2

0.1
EQ2 − Free−field (FF)
Acceleration, a [g]
0.05
−0.05
−0.1
0 1 2 3 4 5 6 7
Time, t [sec]
0.1
EQ2 − Dam base (DB)
Acceleration, a [g]
0.05
−0.05
−0.1
0 1 2 3 4 5 6 7
Time, t [sec]
0.1
Acceleration, a [g]
0.05
−0.05
EQ2 − Downstream dam toe (DDT)
−0.1
0 1 2 3 4 5 6 7
Time, t [sec]
Figure 7.28: Acceleration time-history response at various locations for a

dam-foundation system - EQ2
301

0.2
Acceleration, a [g]
0.1
−0.1
EQ5 − Free−field (FF)
−0.2
0 10 20 30 40 50 60 70
Time, t [sec]
0.3
EQ5 − Dam base (DB)
0.2
Acceleration, a [g]
0.1
−0.1
−0.2
0 10 20 30 40 50 60 70
Time, t [sec]
0.2
Acceleration, a [g]
0.1
−0.1
EQ5 − Downstream dam toe (DDT)
−0.2
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.29: Acceleration time-history response at various locations for a

302
Response spectra − EQ2

0.5
Free−field (FF)
0.45 Dam base (DB)
Downstream dam toe (DDT)
0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0 0.5 1 1.5 2 2.5 3
Period, T [sec]
Figure 7.30: Response spectra (damping, ξ = 5%) at various locations for a


1
Free−field (FF)
0.9 Dam base (DB)
Downstream dam toe (DDT)
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3
Period, T [sec]
Figure 7.31: Response spectra (damping, ξ = 5%) at various locations for a

303
boundary is no longer free to move (i.e. shear stresses being equal to zero,
τ = 0), but there are some stresses from the base of the dam. On the other
hand, an oscillator with a larger mass exhibits a softer response, i.e. it has
a larger fundamental
q
period, T as the latter is proportional to the mass, m
(T = 2π m/k, where, k is the stiffness of the oscillator). This observed
softening in the response is in agreement with earlier theoretical studies of
the shear response of a uniform elastic soil layer which was found to be stiffer
when a surface loading was applied (Ambraseys, 1959b).
7.4.2 Analysis of a dam built on a rigid base

Further analyses were carried out for a dam built on a stiff base. The dy-
namic analysis described in Chapter 6 and discussed in the previous section is
repeated for La Villita dam for different input accelerations. The aim of this
section is to examine whether a dynamic analysis of a dam can be performed
without fully considering the presence of the foundation layer.
The same FE mesh was used (Figure 6.22), but as the dam is considered
to have a rigid base, the stiffness of its foundation was increased and was
taken as G = 700000 · 106 kPa, in order to behave as an extremely stiff
base. The dynamic analyses followed the same static analysis as described
in Section 6.5 (i.e. with a non-stiff alluvial foundation) and the material
properties (stiffness) of the foundation were changed prior to the dynamic
analysis. This allowed direct comparison between the results of the analyses
in this section and the previous analyses, as they all start from the same
stress states.
Two scenarios were considered. Both of them consist of a dam resting
on a rigid base (on an infinitely stiff foundation), but have different input
accelerations. Both EQ2 and EQ5 seismic events were considered. The
input accelerations were taken from the dam-foundation system analysis, i.e.
Chapter 6 (see previous section) and are as follows:
• A - “Bedrock” input acceleration (BRI)
• B - “Free-field” input acceleration (FFI)
304
For each of the two seismic events, EQ2 and EQ5, the first input motion,
“Bedrock”, was the same as the one used in Chapter 6 and shown in Figure
6.6, for both EQ2 and EQ5, i.e. the real acceleration record monitored at
the dam site. The second input motion, “Free-field” is the one calculated
from the uncoupled analysis described in Section 7.3 and discussed in Section
7.4.1. It is shown in Figures 7.28 and 7.29 for EQ2 and EQ5 respectively.
Figures 7.32 and 7.33 show the acceleration time histories at the crest
of the dam for the two scenarios considered, for EQ2 and EQ5 respectively.
Figures 7.34 and 7.35 show the associated response spectra for the two earth-
quakes, along with that from the original analysis of the dam-foundation
system discussed in Section 7.4.1.
From the acceleration time-histories at the crest of the dam (Figures 7.32
and 7.33), it is shown that the dam accelerations for the FFI (Case A) (for
both earthquakes) exhibit higher values of acceleration and higher frequency
content, whereas the dam accelerations for the BRI (Case B) show smaller
values of acceleration and larger periods of vibration.
The latter observation is also confirmed from the graphs of the response
spectra. For EQ5 (Figure 7.35), at higher values of the period (such as
T=0.7s), the BRI record (black solid line) has higher spectral ordinates
whereas the FFI record (grey solid line) has smaller ordinates. At smaller
values of the fundamental period (such as T=0.4s), the opposite is observed,
i.e. the BRI record has smaller spectral ordinates than the FFI. This was
expected as the input motion used in the BRI analysis was richer in lower
frequencies than the input motion used in the FFI analysis, whereas the lat-
ter was richer in higher frequencies than the former, as shown in the previous
section. It should however be noted however that this trend is not very clear
for EQ2.
Moreover, on the same graphs, the response spectra at the crest of the
dam, of the BRI analysis (BR input motion) (black solid line) and the original
analysis of a dam-foundation system, DFS (grey dashed lined) are presented.
It is shown that the BRI analysis with rigid foundation resulted in higher
spectral amplifications for higher frequencies, than the original analysis on
the soil layer, whereas the latter resulted in higher amplification for smaller
305

0.15
0.1
0.05
Acceleration, a [g]
−0.05
−0.1
−0.15 EQ2 − Dam on rigid foundation: Bedrock Input (BRI)
−0.2
0 1 2 3 4 5 6 7
Time, t [sec]

0.4
0.3
0.2
Acceleration, a [g]
0.1
−0.1
−0.2
−0.3
−0.4 EQ2 − Dam on rigid foundation: Free−field Input (FFI)

−0.5
0 1 2 3 4 5 6 7
Time, t [sec]
Figure 7.32: Acceleration time-history response under various input motions

for a dam on a rigid foundation - EQ2
306

0.8
0.6
0.4
Acceleration, a [g]
0.2
−0.2
−0.4
−0.6
EQ5 − Dam on rigid foundation: Bedrock Input (BRI)
−0.8
0 10 20 30 40 50 60 70
Time, t [sec]

1
0.5
Acceleration, a [g]
−0.5
−1
EQ5 − Dam on rigid foundation: Free−field Input (FFI)
−1.5
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.33: Acceleration time-history response under various input motions

for a dam on a rigid foundation - EQ5
307

2
Dam on rigid foundation: Bedrock Input (BRI)
1.8 Dam on rigid foundation: Free−field Input (FFI)
Dam−foundation system: Bedrock Input
1.6
1.4
1.2
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3
Period, T [sec]
Figure 7.34: Response spectra (damping, ξ = 5%) under various input mo-
tions for a dam on a rigid foundation - EQ2

8
Dam on rigid foundation: Bedrock Input (BRI)
7 Dam on rigid foundation: Free−field Input (FFI)
Dam−foundation system: Bedrock
0
0 0.5 1 1.5 2 2.5 3
Period, T [sec]
Figure 7.35: Response spectra (damping, ξ = 5%) under various input mo-
tions for a dam on a rigid foundation - EQ5
308
frequencies. This was expected, as the same input record was used in the
analysis of two different systems (a dam on a rigid base and a dam-foundation
system) which had different fundamental periods and therefore amplification
of different frequencies.
7.4.3 Comments
This section examined dam-foundation interaction effects. Considering a
dam-foundation system, it was shown that the presence of a large embank-
ment dam on top of a uniform soil layer (foundation) changes the dynamic
characteristics of the layer and results in different values of amplification
and for different values of the loading frequency. It was found that smaller
amplification at the top of the soil layer occurs closer to the dam, and the
accelerations tend to contain smaller frequencies (i.e. the soil layer exhibits
a softer response).
Moreover, it was found that significantly different amplifications and fre-
quency contents are obtained if a dam is analysed on a rigid base. This was
true for a number of input motions examined, i.e. for using (a) the same
input motion (bedrock) and (b) the free-field motion (of the soil layer). This
raises issues regarding the validity of ignoring the foundation soil layer and
the use of a “de-coupled” dynamic analyses of earth dams.
Namely, considering Case A, it is incorrect to use the bedrock acceler-
ation at the base of a dam founded on a soft foundation soil layer, as this
would ignore the influence of the foundation layer, i.e. softening of the re-
sponse of the dam structure and change of the dynamic characteristics of
the earthquake (amplification and frequency content). Moreover, regarding
Case B, it is again inappropriate to consider the dam and the foundation
soil layer as separate independent domains in a de-coupled analysis, i.e. to
perform a site response analysis of the soil layer and then use the calculated
acceleration time-history at the top of the layer as input in a dynamic anal-
ysis of a dam founded on a rigid base. Therefore, it is concluded that a
dam-foundation system should be analysed together in order to capture the
effects of dam-foundation interaction appropriately and care should be taken
309
if a “de-coupled” analysis is performed.
7.5 Reservoir-dam interaction

The effects of reservoir-dam interaction (RDI) on elastic cantilever and trape-
zoidal dams founded on a rigid base were investigated in Chapter 5. This
section investigates the RDI effects on the nonlinear elasto-plastic La Villita
dam which is founded on a compliant soil foundation.
7.5.1 Description of analyses

The uncoupled dynamic analysis described in Section 7.3 was repeated again
for both EQ2 and EQ5 to investigate reservoir-dam interaction effects. The
same static procedure described earlier in Section 6.5 was followed apart from
the reservoir impounding part. The upstream reservoir domain was modelled
with solid finite elements and the impounding was simulated by constructing
these elements over 10 increments (i.e. activating their weight). The FE
mesh used is shown in Figure 7.36.
The approach described in Chapter 4 for modelling the reservoir domain
was followed. The behaviour of the reservoir was assumed to be linear elastic
and interface elements were placed along the interface between the reservoir
and the dam and the reservoir and the foundation alluvium. The material
properties assigned for the reservoir were bulk modulus, Kw = 2.2 · 10 6 kPa
and shear modulus, Gw = 100 kPa. The values for the shear and normal
stiffnesses of the interface elements were, Ks = 1 kN/m and KN = 108
kN/m. Damping of the Rayleigh type was specified in the reservoir domain
corresponding to a target damping, ξt = 1%. The two Rayleigh damping
circular frequencies, ω1 and ω2 were taken as the first and the third natural
circular frequencies of the reservoir and they are listed in Table 7.1 along with
the Rayleigh damping coefficients A and B. The boundary condition at the
upstream reservoir boundary during the static analysis was the same as the
other parts of the lateral boundaries, i.e. zero displacement in the horizontal
and zero stress in the vertical direction. The boundary condition on that
310
342m
60m
54m
70m
582m
Figure 7.36: 2D Finite Element mesh used for the analysis of La Villita dam
Table 7.1: Values of the Rayleigh damping coefficients A & B.
Target Circular Circular Rayleigh Rayleigh

Damping Frequency Frequency Coefficient Coefficient
ω1 ω2 A B
[%] [rad/sec] [rad/sec] [] []
1 43.13872 8.627744 0.718979 0.001932
boundary during the dynamic analysis was the Viscous boundary condition
(see Sections 3.7 and 4.3) and the values of the viscosities were obtained from
the elastic properties of the material to which they were applied, i.e. water
(see Section 4.3). The hydraulic boundary conditions were the same as those
described in Chapter 6.

Figures 7.37 and 7.39 show the acceleration time-histories at the crest of
the dam for EQ2 and EQ5 respectively, whereas Figures 7.38 and 7.40 show
the associated response spectra. The results of Chapter 6 (i.e. those that
considered the reservoir water pressure as a “boundary stress”, BS) are also
included in the same graphs. Moreover, Figures 7.41 and 7.42 show the
vertical variation of the maximum values of horizontal displacement in the
core for EQ2 and EQ5, whereas Figures 7.43 and 7.44 show the corresponding
variation of the maximum values of horizontal acceleration in the core.
It may be observed that minor differences exist between the analyses with
a hydrostatic boundary stress and those with reservoir hydrodynamic pres-
sures. For both EQ2 and EQ5, the analyses with the reservoir resulted in
311

0.15
BS
Reservoir
0.1
0.05
Acceleration, a [g]
−0.05
−0.1
−0.15
−0.2
0 1 2 3 4 5 6 7
Time, t [sec]

EQ2.

0.7
BS
Reservoir
0.6
Spectral Acceleration, S [g]
0.5
a
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

dam during EQ2.
312

0.4
BS
Reservoir
0.3
0.2
Acceleration, a [g]
0.1
−0.1
−0.2
−0.3
0 10 20 30 40 50 60 70
Time, t [sec]

EQ5.

1.5
BS
Reservoir
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]

dam during EQ5.
313

60
BS
40 Reservoir
20
Elevation, z [m]
−20
−40
−60
−80
1 2 3 4 5 6 7 8 9 10
x 10


60
BS
40 Reservoir
20
Elevation, z [m]
−20
−40
−60
−80
0.13 0.135 0.14 0.145 0.15 0.155 0.16 0.165 0.17 0.175
Displacement, d [m]

314

60
BS
40 Reservoir
20
Elevation, z [m]
−20
−40
−60
−80
0.04 0.06 0.08 0.1 0.12 0.14 0.16
Acceleration, a [g]


60
BS
40 Reservoir
20
Elevation, z [m]
−20
−40
−60
−80
0.1 0.15 0.2 0.25 0.3 0.35
Acceleration, a [g]

315
slightly smaller values of the accelerations at the dam crest as shown for the
acceleration time-histories and the associated response spectra. Moreover,
smaller values were calculated for the horizontal accelerations and displace-
ments in the dam core for the analyses considering the reservoir. Finally,
it should be noted that the effects of reservoir-dam interaction were more
pronounced for EQ5 which had higher intensity and resulted in higher values
of accelerations in the dam structure.
It may be concluded that the presence of the reservoir “damped” the
response of the dam and resulted in slightly smaller values of accelerations.
The effects of reservoir-dam interaction were investigated in Chapter 5 for
linear elastic dams founded on a rigid foundation. A direct comparison of
this study with the results of Chapter 5 cannot be made as (a) Chapter 5
considered dams on a rigid base, whereas La Villita dam is founded on a
soil layer and (b) the fundamental period of the dam cannot be accurately
estimated as nonlinear behaviour was taken into consideration. However, the
results of this section agree qualitatively with those of Chapter 5 showing that
RDI effects are insignificant in earth dams.
7.5.3 Hydrodynamic pressures

Figures 7.45 and 7.47 show the time-histories of the total hydrodynamic force
on the upstream face of the dam during EQ2 and EQ5 respectively, whereas
Figures 7.46 and 7.48 show the distribution of the peak hydrodynamic pres-
sure on the upstream face of the dam during EQ2 and EQ5 respectively. It
may be observed that significantly higher values of the hydrodynamic pres-
sure (and hence the total hydrodynamic force on the upstream face) occur
for EQ5 which is stronger than EQ2 and therefore induces higher values of
acceleration in the dam structure.
The absolute values of the maximum total hydrodynamic force (hydrody-
namic pressures on the upstream face integrated with respect to the length
of the face) on La Villita dam are listed in Table 7.2 along with the fraction
with respect to the hydrostatic value, calculated using Equation 7.1.
316
Total hydrodynamic force on the upstream dam face during EQ2

800
Total Hydrodynamic Force − EQ2
Maximum value
600
Total Hydrodynamic Force, Fdyn [kN]
400
200
−200
−400
0 1 2 3 4 5 6 7
Time, t [sec]
Figure 7.45: Time-history of the total hydrodynamic force on the upstream

face of the dam during EQ2.
Peak hydrodynamic pressure on the upstream dam face during EQ2

60
Peak Hydrodynamic Pressure − Calculated
Peak Hydrodynamic Pressure − Zangar
50
40
Elevation, z [m]
30
20
10
0
−5 0 5 10 15 20 25
Peak Hydrodynamic Pressure, Pdyn [kPa]

317
Total hydrodynamic force on the upstream dam face during EQ5

2000
Total Hydrodynamic Force − EQ5
Maximum value
1500
Total Hydrodynamic Force, Fdyn [kN]
1000
500
−500
−1000
−1500
0 10 20 30 40 50 60 70
Time, t [sec]
Figure 7.47: Time-history of the total hydrodynamic force on the upstream

Peak hydrodynamic pressure on the upstream dam face during EQ5

60
Peak Hydrodynamic Pressure − Calculated
Peak Hydrodynamic Pressure − Zangar
50
40
Elevation, z [m]
30
20
10
0
−10 0 10 20 30 40 50 60 70
Peak Hydrodynamic Pressure, Pdyn [kPa]

318
Table 7.2: Maximum hydrodynamic force on La Villita dam.
Earthquake Fst Fdyn Fdyn /Fst

[kN] [kN] [%]
EQ2 14303 617.71 4.32
EQ5 14303 1604.99 11.22
1
Fst = · γw · h2 = 0.5 · 9.81 · 542 = 14303kN (7.1)
2
It is shown that the maximum total hydrodynamic force for EQ5 was
larger than that for EQ2. This was expected, as EQ5 induced higher values
of acceleration in the dam structure and the reservoir domain and this is in
agreement to earlier theoretical studies (Westergaard, 1933; Zangar, 1952;
Chopra, 1968; Liu, 1986) which suggested a linear dependence of the max-
imum value of the hydrodynamic pressures on the maximum value of the
induced accelerations (see Section 2.5). Moreover, it may be observed that
the maximum value of the hydrodynamic pressure does not occur at the base
of the dam, but at some elevation. This is in agreement to the earlier obser-
vation of Zangar (1952) for dams with an inclined upstream face (see Section
2.5 and Figure 2.37).
Figures 7.46 and 7.48 show also a comparison of the calculated results
with analytical relations found in the literature. The grey line shows the
hydrodynamic pressures using the relations of Zangar (1952). The latter re-
lations were developed to estimate the peak hydrodynamic pressures on a
dam with a sloped upstream face from an incompressible reservoir (see rele-
vant discussion in Section 4.4). It is shown that the hydrodynamic pressures
calculated from this study were found to be larger than those obtained from
the relations of Zangar (1952). However, it should be noted that the study
of Zangar (1952) neglected the compressibility of the reservoir water (by as-
suming incompressible water). The significance of water compressibility was
highlighted by Chopra (1968) who showed that higher values of the hydro-
dynamic pressure may be expected in cases of resonance between the dam
and the reservoir.
319
Considering the comparison of the hydrodynamic pressures calculated in

this study and those obtained using relations from the literature (Zangar,
1952) (Figures 7.46 and 7.48), it may be observed that this study yielded
higher values of hydrodynamic pressures.
This may be attributed to the following reasons:
• The results of Zangar (1952) were based on dams built on a rigid base,
whereas La Villita dam is built on a compliant soil layer foundation.
• The results of Zangar (1952) referred to rigid incompressible dams and

therefore neglected any effects of reservoir-dam interaction, whereas La
Villita dam is (and was analysed as) a deformable earth dam (see also
Section 4.5).
• The results of Zangar (1952) assumed an incompressible reservoir, whereas

the upstream reservoir was modelled as compressible (see also Section
4.4).
7.5.4 Comments
This section investigated the effects of reservoir-dam interaction on the dy-
namic response of a nonlinear earth dam built on a compliant base. This
is an extension of the work described in Chapter 5 which investigated the
effects of reservoir-dam interaction in visco-elastic dams founded on a rigid
base.
Minor differences in the dynamic response of the dam were found for
a dam with and without an upstream reservoir. Slightly smaller values of
acceleration were found for the reservoir-dam system considered and it was
concluded that the presence of the reservoir “damped” the response of the
dam. The hydrodynamic pressures calculated were found to be larger for
the larger seismic event (EQ5) and were found to be generally larger than
the values obtained using relations from the literature. This difference was
attributed to the simplifying assumptions made by the theoretical relations
found in the literature.
320
7.6 Stiffness inhomogeneity

This section aims to investigate the effect of stiffness inhomogeneity on the
seismic response of earth dams. Earlier theoretical studies on visco-elastic
dams were performed by Dakoulas and Gazetas (1985) and Gazetas (1987)
and were discussed in Section 2.4.3. It was reported that inhomogeneity
affects the modal shapes of vibration of the dam and the vertical variation of
accelerations. However, it was suggested that the effects of inhomogeneity are
less pronounced if nonlinear material behaviour is considered (Stara-Gazetas,
1986).
7.6.1 Description of analysis
Variation of maximum shear stiffness in the dam, G

max
60
G(y) = linear
40 G(y) = constant
20
Elevation, z [m]
−20
−40
−60
−80
0 100000 200000 300000 400000 500000 600000 700000 800000 900000 1000000
Maximum Shear Modulus, Gmax [kPa]
Figure 7.49: Spatial variation (in the vertical direction) of maximum shear
stiffness, Gmax . The grey solid and black dashed lines correspond to a linear
and constant variation of Gmax in the dam core respectively.
The analyses of La Villita dam performed in Chapter 6 and Sections 7.2 -

7.5 assumed an inhomogeneous (in terms of shear stiffness) dam and adopted
321
a linear variation of maximum shear modulus with depth, Gmax (z). In order
to examine the effects of stiffness inhomogeneity in a non-linear analysis, an
additional analysis was carried out in which the maximum shear modulus of
the dam was taken as constant as shown in Figure 7.49. This value of Gmax
was chosen such that it is equal to the average value of the previously-used
linear variation, so that the fundamental period of vibration of the dam was
not significantly affected. It should be noted that the value of Gmax is also
used in the calibration of the CNL model (see Section 3.7.5).

Figures 7.50 and 7.52 show the acceleration time-histories at the crest of the
dam for the two analyses and for EQ2 and EQ5 respectively, whereas Fig-
ures 7.51 and 7.53 show the associated response spectra. Moreover, Figures
7.54 and 7.55 show the maximum horizontal displacements profiles in the
dam core for EQ2 and EQ5 respectively and Figures 7.56 and 7.57 show the
corresponding maximum horizontal accelerations profiles for the two earth-
quakes.
As it may be observed, there are small differences between the accelera-
tions at the crest of the dam for the two cases considered, homogeneous and
inhomogeneous. In fact, the inhomogeneous case results in slightly higher val-
ues of acceleration and displacement at the dam crest and the embankment,
and this is also confirmed from the response spectra at the crest. This is in
agreement with the earlier observations of Dakoulas and Gazetas (1985) and
Gazetas (1987) who reported that stiffness inhomogeneity results in higher
values of accelerations at the dam crest, but this concentration of higher
accelerations vanishes for highly nonlinear dams (see Section 2.4.3). In this
case the effects of stiffness inhomogeneity in a nonlinear analysis are shown
not to be very significant.
7.6.3 Comments
This section investigated the effects of stiffness inhomogeneity in the dynamic
response of an earth dam. Earlier visco-elastic studies reported in the liter-
322

0.15
G(z) = linear
G(z) = constant
0.1
0.05
Acceleration, a [g]
−0.05
−0.1
−0.15
−0.2
0 1 2 3 4 5 6 7
Time, t [sec]

EQ2.

0.7
G(z) = linear
G(z) = constant
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
during EQ2.
323

0.4
G(z) = linear
G(z) = constant
0.3
0.2
Acceleration, a [g]
0.1
−0.1
−0.2
−0.3
0 10 20 30 40 50 60 70
Time, t [sec]

EQ5.

1.5
G(z) = linear
G(z) = constant
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Period, T [sec]
during EQ5.
324

60
G(z) = linear
40 G(z) = constant
20
Elevation, z [m]
−20
−40
−60
−80
1 2 3 4 5 6 7 8 9 10
x 10


60
G(z) = linear
40 G(z) = constant
20
Elevation, z [m]
−20
−40
−60
−80
0.13 0.135 0.14 0.145 0.15 0.155 0.16 0.165
Displacement, d [m]

325

60
G(z) = linear
40 G(z) = constant
20
Elevation, z [m]
−20
−40
−60
−80
0.04 0.06 0.08 0.1 0.12 0.14 0.16
Acceleration, a [g]


60
G(z) = linear
40 G(z) = constant
20
Elevation, z [m]
−20
−40
−60
−80
0.1 0.15 0.2 0.25 0.3 0.35
Acceleration, a [g]

326
ature showed that an inhomogeneous stiffness profile results in higher values

of acceleration close to the crest of the dam (Dakoulas and Gazetas, 1985),
but the latter effect tends to vanish for highly nonlinear materials (Gazetas,
1987).
This nonlinear study is in agreement with those observations. It may
be concluded that there are some differences in the dynamic response of
the homogeneous and inhomogeneous nonlinear La Villita dam cases, with
the latter exhibiting slightly larger accelerations. However, the differences
between the two cases, homogeneous and inhomogeneous were found to be
extremely small. Perhaps inhomogeneity could be neglected in this example
and an equivalent homogeneous stiffness profile be adopted (corresponding to
the same fundamental period of the dam). This should perhaps be adopted in
cases for which calibration of the constitutive models for an inhomogeneous
stiffness profile may be cumbersome or inaccurate, such as the logarithmic
CNL model (Taborda, 2011) employed in this study (see Section 6.4.3 and
Figures 6.25 - 6.30) or constitutive models incorporating stress dependent
stiffness (G = f (p0 )). However, more linear and nonlinear dynamic analyses
are required to generalise this observation.
Finally, it should be noted that the distribution of shear stiffness in the
examined case may not be considered as a “proper” inhomogeneous case as
the values do not drop to zero (or close to zero) at the crest of the dam,
as investigated by Dakoulas and Gazetas (1985) (see Section 2.4.3). This
may be one of the reasons that the observed differences were insignificant in
this study. Therefore, it is suggested that further studies are carried out in
which different distributions of shear stiffness are adopted to fully examine
the effects of stiffness inhomogeneity.
7.7 Conclusions
This chapter dealt with various issues related to numerical analysis of the
seismic response of earth dams. The La Villita earth dam which was analysed
in Chapter 6, was used as an example for further investigation.
Several issues were considered related to the analysis approach followed
327
to analyse earth dams. Issues considered were (a) the effect of previous
earthquake history, (b) the effects of coupled and uncoupled undrained for-
mulation, (c) foundation-dam interaction, (d) reservoir-dam interaction and
(e) the effects of stiffness inhomogeneity. The main findings of this study
may be summarised as follows:
• The effects of the previous loading history were generally found to be

very small in the case of La Villita dam. For the cases considered,
no difference was observed in the crest accelerations and the associated
response spectra. However, minor differences were observed in the crest
settlements, the induced shear strains and the pore water pressures.
• The dynamic response of the dam, if coupled or uncoupled analysis is

performed, was found to be almost identical, as evidenced by the cal-
culated accelerations at the dam crest, the associated response spectra
and displacements. Also, the dynamic behaviour of the soil was found
to be very similar, considering the shear strains (shear-stress strain re-
sponse and shear stress time-history). However, some small differences
were observed in the calculated pore water pressures, between the cou-
pled and uncoupled analyses and this was attributed to the very small
permeability assigned to the consolidating materials.
• The presence of a large embankment dam on top of a uniform soil

layer (foundation) changes the dynamic characteristics of the layer and
results in various values of amplification and for different values of
the loading frequency. It was found that smaller amplification at the
top of the soil layer occurs close to the dam, and the accelerations
tend to contain smaller frequencies (i.e. the soil layer exhibits a softer
response).
• Moreover, it was found that significantly different amplifications and

frequency contents are obtained if a dam is analysed on a rigid base.
This was found to be true for a number of input motions examined.
The latter observation raises issues regarding the validity of ignoring
the presence of the foundation soil layer and the use of a “de-coupled”
328
dynamic analysis of earth dams, i.e. conducting a site response analysis

and feeding the results in a separate dynamic analysis of a dam founded
on a rigid base.
• Minor differences in the dynamic response of the dam were found for
a dam with and without an upstream reservoir (i.e. considering or
neglecting the reservoir hydrodynamic pressures). Slightly smaller val-
ues of acceleration were found for the reservoir-dam system considered
and it was concluded that the presence of the reservoir “damps” the
dynamic response of the dam.
• The hydrodynamic pressures on the upstream face of the dam cal-

culated were larger for the larger seismic event and were found to be
larger than the values obtained using relations from the literature. This
difference was attributed to the simplifying assumptions made by the
theoretical relations found in the literature.
• The effects of stiffness inhomogeneity on the dynamic response of the

nonlinear La Villita dam were found to be very small. The analysis of
the inhomogeneous dam yielded slightly larger values of horizontal ac-
celeration and displacement. Perhaps stiffness inhomogeneity could be
neglected in such cases and an equivalent homogeneous stiffness profile
be adopted in cases for which calibration of the constitutive models for
an inhomogeneous stiffness profile may be rather inconvenient. Besides,
care should be exercised so that the equivalent homogeneous stiffness
profile corresponds to the same fundamental period of the dam. How-
ever, more nonlinear dynamic analyses are required in order to establish
the potential insignificant influence of stiffness inhomogeneity on the
dynamic response of nonlinear dams.
329
Chapter 8
CONCLUSION
8.1 Introduction
As described in the beginning of this thesis, in Chapter 1, the main aims of
this research were (a) to review the current state of the knowledge related to
the seismic response of earth dams and the available methods of analysis and
(b) investigate a real case study of seismic dam response in order to establish
and further improve the current knowledge of both seismic response and
analysis of earth dams.
The three objectives of the research, as those were described in the Intro-
duction of the thesis (Section 1.2), were presented in the previous chapters.
The first objective, which was to review the available literature and present
the main information found related to the response and analysis of earth
dams was presented in Chapters 2 and 3. The second objective, which was
to establish a methodology to model reservoir hydrodynamic pressures and
explore the effects of reservoir-dam interaction on the seismic behaviour of
dams, was presented in Chapters 4 and 5. Finally, the third objective, which
was to investigate the nonlinear behaviour of earth dams using advanced
elasto-plastic finite element analysis and assess the effects of different mod-
elling approaches, was presented in Chapters 6 and 7.
This chapter summarizes the findings of this research and recognising
its limitations, it describes the various recommendations for further inves-
330
tigation. At the end, some thoughts are discussed related to the relative
advantages and use of analytical and numerical methods of analysis.
8.2 Summary of findings

8.2.1 Numerical modelling of the reservoir domain
The main conclusions of the study related to appropriate numerical mod-
elling of the reservoir domain and evaluation of the hydrodynamic pressures
(Chapter 4), are the following:
• The Viscous and Cone Boundary Conditions can be employed on the

upstream absorbing boundary of the reservoir. Both the BCs perform
well for a range of loading frequencies and their response is almost
identical.
• At the dam-reservoir and foundation-reservoir interfaces, interface ele-

ments should be used in order to allow relative movement between the
water and the solid material. Appropriate values for the normal and
shear stiffnesses of these elements should be used.
• The reservoir water should not be treated as incompressible in order

not to underestimate resonant amplifications occurring under loading
frequencies at the vicinity of the fundamental frequencies of the reser-
voir and not to overestimate the pressures for higher frequency loads.
• The size of the reservoir elements should be smaller than a fifth of the
acoustic (p-wave) wavelength, λ of the water.
• The distance, L of the upstream reservoir boundary from the upstream

face of the dam may be reduced for a sloped and flexible dam (which
is the case for earth dams) down to two to three times the height, H,
of the reservoir: L = 2 - 3 H.
• The value of the shear stiffness assigned to water, Gw should be smaller

than 0.1% of the real value of the bulk modulus of the water, Kw .
331
• The viscous (Rayleigh) damping assigned to water, ξw should be kept

close to 1% of the critical. Also Rayleigh damping circular frequencies,
ω1 and ω2 should include the dominant frequency of the load.
8.2.2 Dynamic reservoir-dam interaction

The main conclusions of the study related to the effects of reservoir-dam
interaction on the dynamic response of dams (Chapter 5), are the following:
• The effects of reservoir-dam interaction (RDI) are mainly related to (a)

a change of the fundamental period of the dam and (b) a change in the
amplification of accelerations within the dam structure.
• The effects of RDI, in terms of both amplification of accelerations and

change in the fundamental period, are more pronounced for cantilever
dams.
• Generally, the presence of the reservoir “softens” the response of the

dam, i.e. it results in a larger fundamental period, Td of the dam. This
may be attributed to the added mass, m (see also Westergaard (1933))
of the reservoir which vibrates with the dam. In contrast, the reservoir
does not provide
q
any additional (shear or bending) stiffness, k to the
dam (Td = 2π m/k).
• The amplification of the input acceleration in a dam with a reservoir

can be either higher or lower than the amplification in a dam without a
reservoir. This depends on the relative magnitude of the dominant cir-
cular frequency of the load, ω and the fundamental circular frequencies
of the dam and reservoir, ωd and ωr respectively.
• The amplification of acceleration for a dam with a full reservoir is larger

close to the frequency ratios, ω/ωd ≈ 1, ωr /ωd ≈ 1 and ω/ωr ≈ 1, i.e.
when there is resonance between any two of the load, dam or reservoir.
The maximum amplification occurs when all three circular frequencies
are equal, ω ≈ ωd ≈ ωr (considering any softening due to RDI).
332
• The effects of RDI (both softening and amplification) were found to

be insignificant for earth dams. This may be attributed to (a) the
sloped upstream face and (b) the large volume of an earth dam. The
hydrodynamic pressures induced on a sloped upstream dam face are
smaller than those on a vertical dam face (Zangar, 1952). Moreover,
the inertial effects from the additional mass from the reservoir are small
compared to the inertia of a large earth dam. However, the amplifica-
tion in an earth dam with a full reservoir seems to be slightly smaller
than the corresponding amplification for a dam with an empty reser-
voir. This could suggest that the presence of the reservoir might damp
the dynamic response of earth dams.
8.2.3 Numerical analysis of La Villita dam

The main conclusions of the numerical analysis of the seismic response of La
Villita dam (Chapter 6) are the following:
• The stiffening effect of the 3D narrow canyon geometry already inves-

tigated by previous researchers was taken into consideration in a 2D
analysis by increasing the material stiffness of the dam. The amount
of this additional stiffness was compared to earlier values from the lit-
erature and was found to be reasonable.
• The computed accelerations and associated response spectra were com-

pared to the corresponding available recorded values and were found
to be in fair agreement.
• The calculated response spectrum for EQ5 at the crest of the dam was
compared to that of Elgamal (1992), who performed 3D shear beam
analysis, and was found to be in a better agreement with the recorded
response spectrum. Therefore, an improved prediction of the frequency
content of the acceleration was obtained in this work.
• The high peak values of acceleration observed in the recorded accelera-

tion and already attributed by previous researchers to a localised failure
333
close to the monitoring instrument were not captured. However, when

the high frequencies (originating from the localised slip failure) in the
recorded accelerations were filtered, an excellent agreement was ob-
tained between the calculated and recorded accelerations and response
spectra.
• The calculated vertical displacements at the crest of the dam were not
close to the recorded values. A localised slip failure that would result
into high values of settlements was not predicted and therefore no high
values of settlements were observed.
• The soil in the upstream rockfill experienced higher values of shear

strains than that in the downstream rockfill. This was due to the
reservoir impoundment during which the stress paths were directed
towards the yield surface, because of the reduction in the mean effective
stress, p0 .
• A possible discontinuity in the downstream embankment crest soil pro-

file would form a weak zone which would be very sensitive to any seismic
loads. Such a discontinuity was not artificially modelled in this work
(e.g. through the introduction of interface elements in the soil profile).
• The constitutive modelling approach adopted in this work (i.e. a CNL

model with a Mohr-Coulomb failure criterion), while adequate to cap-
ture the dynamic response of the dam (accelerations and fundamental
period); was not the most appropriate for predicting permanent move-
ments. The formulation of the CNL model adopted is able to predict
only elastic strains. Plasticity (and associated permanent strains) was
introduced only when the yield surface was engaged, whereas the strains
were recoverable for the rest of the analysis. This raises questions about
the applicability of such constitutive assumptions in examining the ef-
fects of previous stress (seismic) history (see also Section 7.2).
334
8.2.4 Parametric seismic analysis of La Villita dam

The main findings of the parametric study of La Villita dam (Chapter 7) are
the following:
• The effects of the previous loading history were generally found to be

very small in the case of La Villita dam. For the cases considered,
no difference was observed in the crest accelerations and the associated
response spectra. However, minor differences were observed in the crest
settlements, the induced shear strains and the pore water pressures.
This could be partly attributed to the constitutive assumptions adopted
in this study which might not be the most appropriate for predicting
permanent strains accurately for stress states within the yield surface.
• The dynamic response of the dam, if coupled or uncoupled undrained

analysis is performed, was found to be very similar, as evidenced by the
calculated accelerations at the dam crest, the associated response spec-
tra and displacements. Also, the dynamic behaviour of the soil was
found to be very similar, considering the shear strains (shear-stress
strain response and shear stress time-history). However, some small
differences were observed in the calculated pore water pressures, be-
tween the coupled and uncoupled analyses and this was attributed to
the very small permeability assigned to the consolidating materials.
• The presence of a large embankment dam on top of a uniform soil

layer (foundation) changes the dynamic characteristics of the layer and
results in different values of amplification and for different values of the
loading frequency. It was found that smaller amplification at the top
of the soil layer occur close to the dam, and the accelerations tend to
contain smaller frequencies (i.e. a softer response).
• Moreover, it was found that significantly different amplifications and

frequency contents are obtained if a dam is analysed on a rigid base
(i.e. ignoring the underlying foundation). This was found to be true
for a number of input motions examined. The latter observation raises
335
issues regarding the validity of ignoring the presence of the foundation

soil layer and the use of a “de-coupled” dynamic analysis of earth dams,
i.e. conducting a site response analysis and then feed the results into
a separate dynamic analysis of the dam founded on a rigid base.
• Minor differences in the dynamic response of the dam were found for a
dam with and without an upstream reservoir. Slightly smaller values
of acceleration were found for the reservoir-dam system considered and
it was concluded that the presence of the reservoir damps the dynamic
response of the dam.
• The hydrodynamic pressures on the upstream face of the dam cal-

culated were larger for the larger seismic event and were found to be
larger than the values obtained using relations from the literature. This
difference was attributed to the severe simplifications made by the the-
oretical relations found in the literature.
• The effects of stiffness nonlinearity on the dynamic response of a non-

linear dam were generally found to be very small. The analysis of the
inhomogeneous dam yielded slightly larger values of horizontal accel-
eration and displacement. Perhaps stiffness inhomogeneity could be
neglected and an equivalent homogeneous stiffness profile be adopted
in cases for which calibration of the constitutive models for an inho-
mogeneous stiffness profile may be rather inconvenient. However, care
should be exercised so that the equivalent homogeneous stiffness profile
corresponds to the same fundamental period of the dam.
8.3 Recommendations for further research

Considering the findings of this research summarised in the previous section,
the lessons learnt during the course of this study and recognising its limi-
tations, directions for further investigation were identified. Therefore, it is
suggested that the following issues are further investigated in the future in
order to extend and perhaps complement the current study.
336
8.3.1 Numerical modelling of the reservoir domain

As far as numerical modelling of the reservoir domain is concerned, it is
suggested that the following issues are further investigated:
• The study should be extended to include a compliant base (soil layer

under the dam and reservoir) in order to investigate the effect of any
underlying soil foundation on the modelling considerations, such as the
distance of the upstream BC from the dam.
• The study should be extended to consider material nonlinearity of the

dam structure. This would explore if material nonlinearity affects the
numerical modelling considerations.
8.3.2 Dynamic reservoir-dam interaction

Considering the work related to the dynamic reservoir-dam interaction ef-
fects, it is suggested that the following issues are further investigated:
• This study considered a dam with an upstream reservoir founded on

a rigid base. It should be extended to include a compliant base (of
finite or infinite depth) in order to investigate reservoir-dam-foundation
interaction effects.
• This study assumed visco-elastic dam behaviour. It should be extended

to consider material nonlinearity of the dam structure to investigate
how the results of this study are affected by nonlinear material be-
haviour. This is very important as in reality no dams exhibit linear
behaviour. Therefore, the results of this study should be viewed as the
first step and be used as the basis to examine nonlinear reservoir-dam
interaction.
• Two-dimensional plane-strain conditions were assumed. Three-dimensional

analysis considering the surrounding canyon’s geometry should be un-
dertaken in order to examine the effects of the reservoir on the stiffening
effect of a narrow canyon. This could also be viewed as an extension of
337
the early work on canyon effects by Mejia and Seed (1983) and Dakoulas
and Gazetas (1987) (see Section 2.4.2).
• The insignificant effects of RDI for earth dams were attributed (par-
tially) to the large volume of those dams, which makes the inertial
effects from the additional mass from the reservoir small compared to
the inertia of a large earth dam. Therefore, one possible direction for
further research would be to examine the amplification of accelerations
with respect to the inertial ratio (additional inertia due to the reservoir
hydrodynamic pressures over the inertia of the dam). The inertia due
to the reservoir hydrodynamic pressures is related to the magnitude of
those pressures due to the added mass effect. Therefore, examining RDI
effects with respect to the inertia ratio using numerical analyses would
be rather cumbersome, because the hydrodynamic pressures need to be
obtained for each case. It is consequently suggested that an analytical
(simplified and hence approximate) approach should be followed first
which would allow an easy analytical estimation of the hydrodynamic
pressures. It is the Author’s opinion and suggestion that the work of
Lee and Tsai (1991) (see Section 2.5.6) could be the starting point.
• The differences in the calculated values of the fundamental periods

(from this study and other analytical approaches) were attributed to
the simplifying assumptions of the analytical methods (Euler-Bernoulli
beam theory and shear beam theory), as they consider only one and
ignore the other of: (a) shear and (b) bending deformations. A bet-
ter analytical approach to estimate the fundamental period (and other
vibrational characteristics) of both types of dams could be the Timo-
shenko (1921, 1922) beam theory (which takes account of both shear
and bending deformations). This was already examined and proved
by previous researchers to improve the predictions for the vibrations
of cantilever dams (Chopra, 1995; Han et al., 1999). It is suggested
that the same approach (Timoshenko (1921, 1922) beam formulation)
could be used to develop analytical relations for vibrations of elas-
tic earth dams (by modelling them as cantilever dams with varying
338
cross-sectional area) which eliminate the drawbacks of the shear beam

method (see Chopra (1966); Tsiatas and Gazetas (1982) and Section
5.3). The Author is not aware of any published analytical relation of
a Timoshenko (1921, 1922) beam with a linearly varying cross-section.
If such an analytical relation is developed, its accuracy could still of
course be assessed by comparisons against the rigorous finite element
method.
8.3.3 Numerical analysis of La Villita dam

Considering the work carried out for the numerical analysis of the seismic
response of La Villita dam, it is suggested that the following aspects are
examined in the future:
• A full three-dimensional seismic analysis of La Villita dam adopting

the initial values of shear stiffness should be performed and compared
to the two-dimensional analysis described in this chapter. This will
enable further insight into the effects of dam-canyon interaction: (a) it
will confirm whether the “amount” of stiffening adopted in this study
was appropriate, and (b) it will examine how properties (such as shear
strains, displacements etc.) other than the accelerations and the fun-
damental period change with the adoption of an “updated” stiffness
profile.
• Different constitutive models of a higher level of complexity should be

used in order to better investigate the permanent deformations of La
Villita dam. Those models should be able to predict pre-yield plastic-
ity so that permanent strains can be modelled during different times of
loading and unloading. Such models could be the kinematic hardening
model of Grammatikopoulou (2004) for the clay core and the alluvium
and the bounding surface model of Papadimitriou and Bouckovalas
(2002) for the sand filters and the rockfill shells. Both of them have al-
ready been appropriately formulated and implemented in ICFEP. How-
ever, it should be noted that these two constitutive models (and other
339
models of similar complexity) require a large number of parameters to

be appropriately calibrated (such as compression parameters: ν1 , λ and
κ) and therefore their calibration would be cumbersome for this case
study because of the limited availability of material properties of La
Villita dam (see also relevant discussion in Section 6.7).
• Although not all (and not the whole duration of) the earthquake time
histories are available, a series of successive earthquake events should
be analysed in order to examine the effect even of the smaller previous
earthquakes. Those earthquakes may perhaps have a contribution in
the accumulation of permanent displacements of the dam.
• A parametric analysis with different yield strengths of the rockfill (i.e.

different values of the cohesion, c0 and the angle of shearing resistance,
φ0 ) could evaluate how the yield strength of the rockfill affects the
magnitude of the permanent displacements. This could perhaps allow
a back-analysis of the properties of the rockfill.
8.3.4 Parametric seismic analysis of La Villita dam

Finally, regarding the work performed related the parametric seismic analyses
of La Villita dam, it is suggested that the following issues are examined in
the future:
• As discussed in the previous section, different constitutive models of a
higher level of complexity, able to predict pre-yield plasticity should be
used in order to highlight the effects of previous stress history in the
dynamic analysis of an earth dam.
• A dam consisting of consolidating materials in its entirety (in contrast

to La Villita dam, of which the consolidating core consisted only a
small part of the embankment’s geometry) should be analysed in order
to better appreciate the effects of previous stress history and coupled
formulation. Also, parts of the dam with high permeability (such as
rockfill shoulders) should be modelled as consolidating materials in cou-
pled formulation.
340
• To complete the investigation of foundation-dam interaction effects, an

additional examination should be made: the “de-coupled” analysis of
the dam founded on a rigid base should be extended to include the
effects of a compliant base without discretising the entire foundation
domain. This might perhaps be achieved by using equivalent springs
at the base of the dam with appropriate stiffness or by using boundary
elements to model the foundation. This could then be compared with
the results of the full dam-foundation analysis (Chapter 6) in order to
completely confirm that de-coupled seismic analysis of earth dams is
not appropriate or hopefully show some more directions to improve the
de-coupled analysis approach.
• The nonlinear reservoir-dam interaction analysis of Section 7.5 should

be repeated with the dam resting on a rigid foundation. That would
enable direct comparison of the predicted hydrodynamic pressures on
the upstream dam face with results from the literature.
• The comparison of homogeneous and inhomogeneous dams should be

repeated for a dam built in a very narrow canyon. Dam built in narrow
canyons exhibit significant additional geometric stiffness and higher
amplifications of accelerations in cases of resonance (see Section 2.4.2
and Figure 2.23). This is considered to be a case where any difference in
the accelerations at the crest of the dam should be visible and therefore
able to highlight better the effects (and hence the necessity of taking
account) of stiffness inhomogeneity. Finally, cases with higher degree
of inhomogeneity should also be considered in order to identify cases
for which stiffness inhomogeneity can be neglected.
8.4 Comments on the methods of analysis

The first objective of this research (see Section 1.2) was to review the avail-
able methods of seismic analysis of earth dams. Chapter 3 summarized the
results of that review and discussed the relative merits of the various methods
of analysis. Numerical methods, such as the finite element method, which
341
was employed in this work, were identified as the most rigorous and reliable
methods of analysis.
8.4.1 Discussion on numerical and analytical methods

Indeed FE analysis is a sophisticated method of analysis which is able to
overcome the limitations of analytical methods (see Section 3.6). It was
commented in Chapter 3 that this numerical method is able to provide in-
formation about the stability, deformations and dynamic characteristics of
earth dams. This observation was already reported in the literature by other
researchers.
Its superiority and rigour was also highlighted in a number of occasions
in this work, such as (a) prediction of the fundamental period of vibration
of cantilever and earth dams, as compared to the Euler-Bernoulli beam the-
ory and shear beam theory (see Chapter 5), (b) prediction of the dynamic
response of La Villita dam, as compared to the work of Elgamal (1992) who
adopted the shear beam approach (see Section 6.6.2) and (c) estimation of
the nonlinear stiffening effect of narrow canyon (see Section 6.6.1).
Additionally, it should be noted that analytical solutions are not always
available for all problems. Also, the available analytical solutions adopt
several simplifying assumptions to approach a problem and in some cases
they provide highly complicated functions which are not very practical for
use in routine dam analysis and design.
However, two potential pitfalls/limitations of advanced numerical analysis
should be recognised:
Firstly, the use of advanced sophisticated FE analysis should be accom-
panied by some knowledge of the method’s potentials and limitations. For
example, advanced FE analysis was employed in this work, which was able
to capture very well the dynamic response of La Villita dam, in terms of
accelerations and response spectra (see Section 6.6.2). However, the adopted
constitutive assumptions failed to capture the seismically induced permanent
deformations (see Section 6.6.2) and identify the effects of previous stress his-
tory (see Section 7.2).
342
It should always be remembered that this powerful method provides an-

swers to the question posed, and therefore the correct question (such as con-
stitutive behaviour) should be posed. The employment of the FE method
will not always provide the desired answers, unless the user is able to adopt
the appropriate numerical and constitutive assumptions. Pitfalls when using
numerical methods were elegantly highlighted earlier by Potts (2003) in the
42nd Rankine Lecture. Potts (2003) suggested that for useful geotechnical
numerical analysis to be performed, users should be fully aware of the theory
behind soil mechanics and numerical analysis and appreciate the limitations
of the numerical and constitutive models used.
Secondly, the numerical nature of the FE method does not allow the
development of analytical closed-form relations which would allow a quick
and (at least) qualitative estimate of the effects of different parameters. The
FE method can provide the numerical answer for a certain case considered
each time. Moreover, there are cases where a series of parametric analyses
would be required to obtain the solution with the FE method. In contrast,
analytical methods (such as the shear beam method, see Section 3.4) are able
to provide closed-form solutions (for a simplified problem, of course) which
enable direct estimation of the effects of various parameters. Such analytical
solutions were reported in this thesis and examples of them are the following:
• Hydrodynamic pressures on dams: Equation 2.40 was developed ana-

lytically by Chopra (1968), whereas the same problem was approached
parametrically by the FE method (see Section 4.4 and Figure 4.25).
• Fundamental period of vibration of earth dams: Equations 2.9 and 3.10

were developed analytically by Ambraseys (1960a) and Dakoulas and
Gazetas (1985) respectively, using the shear beam method, whereas the
same problem was approached parametrically by the FE method (see
Section 5.4 and Figure 5.10).
• Stiffening effect of a narrow canyon: Figure 2.23 was developed analyt-

ically by Dakoulas and Gazetas (1987) using the shear beam method,
whereas in this study the stiffening effect (amount of increase in the
343
shear stiffness) was examined parametrically by the FE method (see

Section 6.6.1).
It should be noted that regarding the last example, the fundamental

period of an earth dam in two dimensions or in a three-dimensional canyon
can be obtained using the FE method without the need of a parametric
study, through a modal analysis in the frequency domain (Chopra, 1966,
1995). The numerical solution however will be valid for the specific case
considered only. Nevertheless, this would not apply to the second example
of dam-reservoir interaction due to the use of time-dependent dashpots at
the upstream reservoir boundary, which therefore still require a time-domain
analysis.
8.4.2 Recommendations for use of the methods

Undoubtedly, the numerical FE method is more rigorous than analytical
methods (such as the shear beam method), and the previously-mentioned
examples do not question the accuracy of the FE method. Instead, they
refer to cases where the solution of a problem required a (rather time- and
resource-demanding) parametric study.
It is the Author’s opinion that both approaches, closed-form (if and where
they exist) analytical (such as the shear beam) and rigorous numerical (such
as the FE) should be used together to analyse a problem. It is suggested
that firstly an available closed-form analytical solution may be used to pro-
vide some (at least qualitative) insight into the relation between different
parameters (such as the stiffening effect of narrow canyons, see Figure 2.23).
Then a rigorous FE analysis should be performed to provide the accurate so-
lution of the problem (such as the additional stiffening of a narrow canyon, see
Section 6.6.1). Finally, it is recommended that, if possible, both approaches
should be used in conjunction and treated as a complementary couple.
344
References
Abdel-Ghaffar, A. M. and Elgamal, A.-W. M. (1987a), ‘Elasto-plastic seismic

response of 3D earth dams: application’, Journal of Geotechnical Engineer-
ing 113(11), 1309–1325.
Abdel-Ghaffar, A. M. and Elgamal, A.-W. M. (1987b), ‘Elasto-plastic seismic

response of 3D earth dams: theory’, Journal of Geotechnical Engineering
113(11), 1293–1308.
Abdel-Ghaffar, A. M. and Koh, A. S. (1981), ‘Longitudinal vibration of non-

homogeneous earth dams’, Earthquake Engineering & Structural Dynamics
9(3), 279–305.
Abdel-Ghaffar, A. M. and Koh, A. S. (1982), ‘Three-dimensional dynamic

analysis of nonhomogeneous earth dams’, Soil Dynamics & Earthquake
Engineering 1(3), 136–144.
Abdel-Ghaffar, A. M. and Scott, R. F. (1979a), ‘Analysis of earth dam re-

sponse to earthquakes’, Journal of the Geotechnical Engineering Division,
ASCE 105(GT12), 1379–1404.
Abdel-Ghaffar, A. M. and Scott, R. F. (1979b), ‘Shear moduli and damping

factors of earth dam’, Journal of the Geotechnical Engineering Division,
ASCE 105(GT12), 1405–1426.
Abdel-Ghaffar, A. M. and Scott, R. F. (1981), ‘Vibration tests of a full-

scale earth dam’, Journal of the Geotechnical Engineering Division, ASCE
107(GT3), 241–269.
345
Abouseeda, H. and Dakoulas, P. (1996), ‘Response of earth dams to P and

SV waves using coupled FE-BE formulation’, Earthquake Engineering and
Structural Dynamics 25, 1177–1194.
Abouseeda, H. and Dakoulas, P. (1998), ‘Non-linear dynamic earth dam-

foundation interaction using a FE-BE formulation’, Earthquake Engineer-
ing and Structural Dynamics 27, 917–936.
Ambraseys, N. N. (1959a), ‘A note on the response of an elastic overburden

of varying rigidity to an arbitrary ground motion’, Bulletin of the Seismo-
logical Society of America 49(3), 211–220.
Ambraseys, N. N. (1959b), The seismic stability of earth dams, PhD thesis,

Imperial College, University of London.
Ambraseys, N. N. (1960a), ‘On the shear response of a two-dimensional trun-

cated wedge subjected to arbitrary disturbance’, Bulletin of the Seismo-
logical Society of America 50(1), 45–56.
Ambraseys, N. N. (1960b), The seismic stability of earth dams, in ‘Proceed-

ings of the 2nd World Conference on Earthquake Engineering’, Tokyo.
Ambraseys, N. N. (1962), The seismic stability analysis of earth dams, in

‘2nd Symposium on Earthquake Engineering’, Roorkee, India.
Ambraseys, N. N. and Menu, J. M. (1988), ‘Earthquake-induced ground dis-

placements’, Earthquake Engineering & Structural Dynamics 16(7), 985–
1006.
Ambraseys, N. N. and Sarma, S. K. (1967), ‘The response of earth dams to

strong earthquakes’, Géotechnique 17(2), 181–213.
Antoniou, S. and Pinho, R. (2004), SeismoSignal: A computer program

for signal processing of strong-motion data, Technical Report 4.0.0, Seis-
moSoft.
Aubry, D., Chouvet, D., Modaressi, H. and Modaressi, A. (1985), GEFDYN5:

Logiciel d’ analyse du comportement statique et dynamique des sols par
346
éléments finis avec prise en compte du couplage sol-eau-air., Technical

Report Rapport scientifique, Ecole Central Paris.
Bathe, K. J. (1992), ADINA – A Finite Element Program for Automatic

Dynamic Incremental Nonlinear Analysis, Technical report, ADINA R&D
Inc.
Bathe, K. J. (1996), Finite Element Procedures, Prentice Hall, New Jersey.
Beck, J. L. (1976), ‘Weight-induced Stresses and the Recent Seismicity at

Lake Oroville, California’, Bulletin of the Seismological Society of America
66, 1121–1131.
Bielak, J., Loukakis, K., Hisada, Y. and Yoshimura, C. (2003), ‘Domain Re-
duction Method for Three-Dimensional Earthquake Modelling in Localized
Regions. Part I: Theory’, Bulletin of the Seismological Society of America
93(2), 817–824.
Biot, M. A. (1941), ‘General theory of propagation of three-dimensional con-

solidation’, Journal of Applied Physics 12(2), 155–164.
Biot, M. A. (1963), ‘Continuum dynamics of elastic plates and multilayered

solids’, Journal of Mathematics and Mechanics 12(6), 793–810.
Bray, J. D. and Travasarou, T. (2007), ‘Simplified procedure for estimating

earthquake-induced deviatoric slope displacements’, Journal of Geotechni-
cal and Geoenvironmental Engineering, ASCE 133(4), 381–392.
British Dam Society (2010), ‘Types of dams’, Online, available at

[Last accessed: 3 November 2010]: http://www.britishdams.org/about
dams/default.htm.
Bustamante, J. I., Rosenblueth, E., Herrera, I. and Flores, A. (1963), ‘Presión

hidrodinámica en presas y depósitos’, Boletin Sociedad Mexicana de Inge-
nieria Sismica 1(2).
Cambridge University Engineering Department (2012), ‘Lat-

eral spreading failure of Lower San Fernando dam’, Online,
347
available at [Last accessed: 11 July 2012]: http://www-

civ.eng.cam.ac.uk/geotechn ew/research/earthquakes/lateral/.
Chadha, R. K. (1995), ‘Role of Dykes in Induced Seismicity at Bhatsa Reser-

voir, Maharashtra, India’, Pure & Applied Geophysics 145(1), 155–165.
Chan, A. H. C. (1988), A unified finite element solution to static and dynamic

problems of geomechanics, PhD thesis, University College of Swansea.
Chan, A. H. C. (1995), User manual for diana-swandyne ii, Technical report,

School of Engineering, University of Birmingham, UK.
Chopra, A. K. (1966), Earthquake effects on dams, PhD thesis, University

of California, Berkeley.
Chopra, A. K. (1967a), ‘Earthquake response of earth dams’, Journal of the

Soil Mechanics and Foundations Division, ASCE 93(SM2), 1399–1412.
Chopra, A. K. (1967b), ‘Reservoir-dam interaction during earthquakes’, Bul-

letin of the Seismological Society of America 57(4), 675–687.
Chopra, A. K. (1968), ‘Earthquake Behavior of Reservoir-Dam Systems’,

Journal of the Engineering Mechanics Division, ASCE 94(6), 1475–1500.
Chopra, A. K. (1995), Dynamics of structures: Theory and application to

earthquake engineering, Prentice Hall, New Jersey.
Chopra, A. K., Dibaj, M., Clough, R. W., Penzien, J. and Seed, H. B.

(1960), Earthquake analysis of earth dams, in ‘Proceedings of the 4th
World Conference on Earthquake Engineering’, Santiago.
Chopra, A. K. and Gupta, S. (1978), ‘Hydrodynamic and foundation inter-

action effects in frequency response functions for concrete gravity dams’,
Earthquake Engineering and Structural Dynamics 10(1), 89–106.
Chopra, A. K. and Perumalswami, P. R. (1971), ‘Dynamics of earth dams

with foundation interactions’, Journal of the Engineering Mechanics Divi-
sion, ASCE 97(1), 181–191.
348
Chopra, A. K. and Zhang, L. (1991), ‘Earthquake-induced base sliding

of concrete gravity dams’, Journal of Structural Engineering, ASCE
117(12), 3698–3719.
Chung, J. and Hulbert, G. M. (1993), ‘A time integration algorithm for

structural dynamics with improved numerical dissipation: the generalised-
α method’, Journal of Applied Mechanics 60, 371–375.
Chwang, A. T. (1978), ‘Hydrodynamic pressures on sloping dams during

earthquakes. Part 2. Exact theory’, Journal of Fluid Mechanics 87(2), 343–
348.
Chwang, A. T. and Housner, G. W. (1978), ‘Hydrodynamic pressures on

sloping dams during earthquakes. Part 1. Momentum method’, Journal of
Fluid Mechanics 87(2), 335–341.
Clough, R. W. and Chopra, A. K. (1966), ‘Earthquake stress analysis in earth

dams’, Journal of Engineering Mechanics, ASCE 92(EM2), 1399–1412.
Clough, R. W. and Woodward, R. J. (1967), ‘Analysis of embankment

stresses and deformations’, Journal of the Soil Mechanics and Foundations
Division, ASCE 93(SM4), 529–549.
Comision Federal de Electricidad (1976), Behavior of dams built in Mexico,

Technical Report on the XII International Congress on Large Dams Vol.
1, Comision Federal de Electricidad.
Comision Federal de Electricidad (1979), Performance of El Infiernillo and

La Villita dams including the earthquake of March 14, 1979, Technical
Report 47, Comision Federal de Electricidad.
Comision Federal de Electricidad (1980), Behavior of dams built in Mexico

1974-1984, Technical Report on the XV International Congress on Large
Dams Vol. 2, Lausanne, Switzerland, Comision Federal de Electricidad.
Comision Federal de Electricidad (1987), Effects of the September 1985 earth-

quakes on dams built on the Balsas River, Technical Report 15, Comision
Federal de Electricidad.
349
Comision Federal de Electricidad and Univercidad Nacional Autónoma de

México (1976), Behavior of Dams Built in Mexico, Technical Report Vol.
1, Contribution to the 12th International Congress on Large Dams, Mexico
City, Mexico, Comision Federal de Electricidad & Univercidad Nacional
Autónoma de México.
Committe on Nuclear Structures and Materials (1979), Analyses of Soil-

Structure Interaction Effects for Nuclear Power Plants, Technical Report
Ad-Hoc Committee Report, American Society of Civil Engineers, New
York.
Cyprus Public Works Department (2010), ‘Dams of cyprus’,

Online, available at [Last accessed: 3 November 2010]:
http://www.moa.gov.cy/moa/wdd/Wdd.nsf/dams-gr/dams-
gr?OpenDocument.
Dafalias, Y. F. and Popov, E. P. (1979), ‘A model for nonlinearly hardeing

materials for complex loading’, Acta Mechanica 21(3), 173–192.
Dakoulas, P. (1990), ‘Nonlinear response of dams founded on alluvial deposits

in narrow canyons’, Soil Dynamics and Earthquake Engineering 9(6), 301–
312.
Dakoulas, P. (1993a), ‘Earth dam-canyon interaction effects for obliquely

incident sh waves’, Journal of Geotechnical Engineering,ASCE
119(11), 1696–1716.
Dakoulas, P. (1993b), ‘Response of earth dams in semi-cylindrical canyons to

oblique sh-waves’, Journal of Engineering Mechanics,ASCE 119(1), 74–90.
Dakoulas, P. (2012a), ‘Longitudinal vibrations of tall concrete faced rock-

fill dams in narrow canyons’, Soil Dynamics and Earthquake Engineering
41, 44–58.
Dakoulas, P. (2012b), ‘Nonlinear seismic response of tall concrete-faced rock-

fill dams in narrow canyons’, Soil Dynamics and Earthquake Engineering
34(1), 11–24.
350
Dakoulas, P. and Gazetas, G. (1985), ‘A class of inhomogeneous shear mod-

els for seismic response of dams and embankments’, Soil Dynamics and
Earthquake Engineering 4(4), 166–182.
Dakoulas, P. and Gazetas, G. (1986), ‘Seismic lateral vibration of embank-

ment dams in semi-cylindrical valleys’, Earthquake Engineering and Struc-
tural Dynamics 13(1), 19–40.
Dakoulas, P. and Gazetas, G. (1987), ‘Vibration characteristics of dams in

narrow canyons’, Journal of Geotechnical Engineering 113(8), 899–904.
Dakoulas, P. and Gazetas, G. (2008), ‘Insight into seismic earth and water
pressures against caisson quay walls’, Géotechnique 58(2), 95–111.
Dakoulas, P. and Hsu, C. (1993), ‘Lateral response of dams in semi-elliptical

rigid canyons’, Soil Dynamics and Earthquake Engineering 12(8), 497–507.
Dakoulas, P. and Hsu, C. (1995), ‘Response of dams in semielliptical

canyons to oblique sh waves’, Journal of Engineering Mechanics, ASCE
121(3), 379–391.
Das, B. M. (1993), Principles of Soil Dynamics, PWS-KENT, Boston.
Day, R. A. (1990), Finite element analysis of sheet pile retaining walls, PhD
thesis, Imperial College, University of London.
Day, R. A., Hight, D. W. and Potts, D. M. (1998), ‘Finite element analysis of

construction stability of thika dam’, Computers & Geotechnics 23(4), 205–
219.
Day, R. A. and Potts, D. M. (1994), ‘Zero thickness interface elements -

numerical stability and application’, International Journal for Numerical
and Analytical Methods in Geomechanics 18(10), 689–708.
Dounias, G. T. (1987), Progressive failure in embankment dams, PhD thesis,

Imperial College, University of London.
351
Dounias, G. T., Potts, D. M. and Vaughan, P. R. (1996), ‘Analysis of progres-

sive failure and cracking in old british dams’, Géotechnique 46(4), 621–640.
Duron, Z. H. and Hall, J. F. (1988), ‘Experimental and finite element studies

of the forced vibration response of morrow point dam’, Earthquake Engi-
neering and Structural Dynamics 16(7), 1021–1039.
El-hussain, I. and Carpenter, P. J. (1990), ‘Reservoir Induced Seismicity

Near Heron and El Vado Reservoirs, Northern New Mexico’, Bulletin of
the Association of Engineering Geologists 27, 51–59.
Elgamal, A. W. (1992), ‘Three-dimensional seismic analysis of La Villita

dam’, Journal of Geotechnical Engineering 118(12), 1937–1958.
Elgamal, A. W., Abdel-Ghaffar, A. M. and Prevost, J. H. (1985), ‘Elasto-

plastic Earthquake Shear Response of One-Dimensional Earth Dam Mod-
els’, Earthquake Engineering & Structural Dynamics 13(5), 617–633.
Elgamal, A. W., Abdel-Ghaffar, A. M. and Prevost, J. H. (1987a), ‘2D Elasto-

plastic Seismic Shear Response of Earth Dams: Applications’, Journal of
Engineering Mechanics 113(5), 702–719.
Elgamal, A. W., Abdel-Ghaffar, A. M. and Prevost, J. H. (1987b), ‘2D Elasto-

plastic Seismic Shear Response of Earth Dams: Theory’, Journal of Engi-
neering Mechanics 113(5), 689–701.
Elgamal, A. W. and Gunturi, R. V. (1993), ‘Dynamic behaviour and seis-

mic response of El Infiernillo dam’, Earthquake Engineering and Structural
Engineering 22(8), 665–684.
Elgamal, A. W., Scott, R. F., Succarieh, M. F. and Yan, L. (1990), ‘La Villita
dam response during five earthquakes including permanent deformation’,
Journal of Geotechnical Engineering 116(10), 1443–1462.
Elia, G., Amorosi, A., Chan, C. and Kavvadas, M. J. (2010), ‘Fully coupled
dynamic analysis of an earth dam’, Géotechnique 61(7), 549–563.
352
Fell, A. R., MacGregor, P., Stapledon, D. and Bell, G. (2005), Geotechnical

Engineering of Dams, A. A. Balkema, London.
Fenves, G. L., Mojtahedi, S. and Reimer, R. B. (1992), ‘Effect of contrac-

tion joints on earthquake response of an arch dam’, Journal of Structural
Engineering, ASCE 118(4), 1039–1055.
Frazier, G. A. (1969), Vibration characteristics of three dimensional solids

with application to earth dams, PhD thesis, Montana State University.
Gazetas, G. (1981a), ‘Longitudinal vibrations of embankment dams’, Journal

of Geotechnical Engineering, ASCE 107(1), 21–40.
Gazetas, G. (1981b), ‘Vertical oscillations of earth and rockfill dams: Analysis

and field observation’, Soils & Foundations 21(4), 56–68.
Gazetas, G. (1982a), ‘Shear vibrations of vertically inhomogeneous dams’,

International Journal for Numerical and Analytical Methods in Geome-
chanics 6(1), 219–241.
Gazetas, G. (1982b), ‘Vibrational characteristics of soil deposits with vari-

able wave velocity.’, International Journal for Numerical and Analytical
Methods in Geomechanics 6(1), 1–20.
Gazetas, G. (1987), ‘Seismic response of earth dams: some recent develop-

ments’, Soil Dynamics and Earthquake Engineering 6(1), 3–47.
Gazetas, G. and Dakoulas, P. (1992), ‘Seismic analysis and design of rock-

fill dams: state-of-the-art’, Soil Dynamics and Earthquake Engineering
11, 27–61.
Gazetas, G. and Uddin, N. (1994), ‘Permanent deformatin on preexisting slid-

ing surfaces in dams’, Jounral of Geotechnical Engineering 120(11), 2041–
2061.
Golzé, A. R. (1977), Handbook of Dam Engineering, Van Nostrand Reinhold,

New York.
353
Gonzalez-Valencia, F. (1987), Earthquake response of La Villita dam, in

M. A. Cassaro and E. Martinez-Romero, eds, ‘The Mexico earthquakes -
1985: Factors involved and lessons learned’, ASCE, pp. 134–147.
Goree, J. (2010), ‘Seismic waves’, Online, available at [Last accessed: 28 Oc-

tober 2010]: http://www.physics.uiowa.edu/adventure/spr-2006/feb-18-
06.html.
Grammatikopoulou, A. (2004), Development, implementation and applica-

tion of kinematic hardening models for overconsolidated clays, PhD thesis,
Imperial College London.
Grammatikopoulou, A., Zdravković, L. and Potts, D. M. (2007), ‘The effect

of the yield and plastic potential deviatoric surfaces on the failure height
of an embankment’, Géotechnique 57(10), 795–806.
Grammatikopoulou, A., Zdravković, L. and Potts, D. M. (2008), ‘The in-

fluence of previous stress history and stress path direction on the surface
settlement trough induced by tunnelling’, Géotechnique 58(4), 269–281.
Griffiths, D. V. and Prevost, J. H. (1988), ‘Two- and three-dimensional

dynamic finite element analyses of the long valley dam’, Géotechnique
38(3), 367–388.
Guan, F. and Moore, I. D. (1997), ‘New techniques for modelling reservoir-

dam and foundation-dam interaction’, Soil Dynamic and Earthquake En-
gineering 16(4), 285–293.
Guan, Z. (2009), ‘Investigation of the 5.12 Wenchuan Earthquake damages

to Zipingpu Water Control Project and an assessmet of its safety state’,
Science in China, Series E-Tech Science 52(4), 820–834.
Hall, J. F. and Chopra, A. K. (1980), Dynamic response of embankment,

concrete-gravity and arch dams including hydrodynamic interaction, Tech-
nical Report EERC-80/39, Earthquake Engineering Research Centre, Uni-
versity of California, Berkeley.
354
Hall, J. F. and Chopra, A. K. (1982a), ‘Hydrodynamic effects in earthquake

response of embankment dams’, Journal of the Geotechnical Engineering
Division, ASCE 108(GT4), 591–597.
Hall, J. F. and Chopra, A. K. (1982b), ‘Hydrodynamic effects in the dynamic

response of concrete gravity dams’, Earthquake Engineering & Structural
Dynamics 10(2), 333–345.
Hall, J. F. and Chopra, A. K. (1982c), ‘Two-dimensional dynamic analysis of

concrete gravity and embankment dams including hydrodynamic effects’,
Earthquake Engineering & Structural Dynamics 10(2), 305–302.
Han, S. M., Benaroya, H. and Wei, T. (1999), ‘Dynamics of transversely

vibrating beams using four engineering theories’, Journal of Sound and
Vibration 255(5), 935–988.
Hardy, S. (2003), The implementation and application of dynamic finite ele-

ment analysis to geotechnical problems, PhD thesis, Imperial College Lon-
don.
Haskell, N. A. (1953), ‘The dispersion of surface waves on multilayered me-

dia’, Bulletin of the Seismological Society of America 4(1), 17–34.
Hatanaka, M. (1952), ‘Three-dimensional consideration on the vibration

of earth dams’, Journal of the Japanese Society of Civil Engineers
37(10), 305–302.
Hatanaka, M. (1955), Fundamental consideration on the earthquake resis-

tant properties of the earth dam, Technical Report Bulletin 11, Disaster
Prevention Research Institute, Kyoto University.
Heujeux, J. C. (1985), Une loi de comportement pour le chargement cy-

clique des sols, in V. Davidovici, ed., ‘Genie parasismique’, Presses ENCP,
pp. 287–302.
Higdon, R. L. (1991), ‘Absorbing boundary conditions for elastic waves’,

Geophysics 56(2), 231–241.
355
Idriss, I. M., J., L., Hwang, R. and Seed, H. B. (1973), QUAD-4: A computer
program for evaluating the seismic response of soil structures by variable
damping finite element procedures, Technical Report Report EERC 73-16,
Earthquake Engineering Research Center, University of California, Berke-
ley.
Idriss, I. M., M., M. J. and Bolton, H. S. (1974), ‘Earth dam-foundation

interaction during earthquakes’, Earthquake Engineering & Structural Dy-
namics 2(4), 313–323.
Idriss, I. M. and Sun, J. I. (1992), SHAKE91: A computer program for con-

ducting equivalent-linear seismic response analyses of horizontally layered
soil deposits, Technical report, University of California, Davis.
Ishihara, K. (1995), Soil Behaviour in Earthquake Geotechnics, Clarendon

Press, Oxford.
Jacob, K. H., Pennington, W. D., Armbruster, J., Seeber, L. and Farhatulla,

S. (1979), ‘Tarbela Reservoir, Pakistan: A Region with Compressional
Tectonics with Reduced Seismicity upon Initial Reservoir Filling’, Bulletin
of the Seismological Society of America 69(4), 1175–1192.
Jardine, R. J., Potts, D. M., Fourie, A. B. and Burland, J. B. (1986), ‘Studies

of the influence of nonlinear stress-strain characteristics in soil-structure
interaction’, Géotechnique 36(3), 377–396.
Kausel, E. and Röesset, J. M. (1981), ‘Stiffness matrices for layered soils’,

Bulletin of the Seismological Society of America 71(6), 1743–1761.
Kavvadas, M. and Amorosi, A. (2000), ‘A constitutive model for structured

soils’, Géotechnique 50(3), 263–273.
Küçükarslan, S., Coşkun, S. B. and Taşkin, B. (2005), ‘Transient analysis of

dam-reservoir interaction including the reservoir bottom effects’, Journal
of Fluids and Structures 20, 1073–1084.
Kellezi, L. (1998), Dynamic soil-structure interaction transmitting boundary

for transient analysis, PhD thesis, Technical University of Denmark.
356
Kellezi, L. (2000), ‘Local transmitting boundaries for transient elastic anal-

ysis’, Soil Dynamics & Earthquake Engineering 19(7), 533–547.
Khoei, A. R., Azami, A. R. and Haeri, S. M. (2004), ‘Implementation of

plasticity based models in dynamic analysis of earth and rockfill dams: A
comparison of Pastor-Zienkiewicz and cap models’, Computers & Geotech-
nics 31(5), 385–410.
Kokusho, T. (1980), ‘Cyclic triaxial test of dynamic soil properties for wide
strain range’, Soils and Foundations 20(2), 45–60.
Kontoe, S. (2006), Development of time integration schemes and advanced

boundary conditions for dynamic geotechnical analysis, PhD thesis, Impe-
rial College London.
Kontoe, S., Zdravković, L. and Potts, D. M. (2009), ‘An assessment of the do-
main reduction method as an advanced boundary condition and some pit-
falls in the use of conventional absorbing boundaries’, International Jour-
nal for Numerical and Analytical Methods in Geomechanics 33(3), 309–
330.
Kontoe, S., Zdravković, L., Potts, D. M. and Menkiti, C. (2008), ‘Case study
on seismic tunnel response’, Canadian Geotechnical Journal 45(12), 1743–
1764.
Kontoe, S., Zdravković, L., Potts, D. M. and Menkiti, C. (2011), ‘On the
relative merits of simple and advanced constitutive models in dynamic
analysis of tunnels’, Géotechnique 61(10), 815–829.
Kontoe, S., Zdravković, L., Potts, D. M. and Menkiti, C. (2012), ‘Seismic

response and interaction of complex soil retaining systems’, Computers &
Geotechnics 39, 17–26.
Kovacevic, N. (1994), Numerical analysis of rockfill dams, cut slopes and

road embankments, PhD thesis, Imperial College London.
357
Kovacević, N., Hight, D. W., Potts, D. M. and Carter, I. C. (2013), ‘Finite-

element analysis of the failure and reconstruction of the main dam em-
bankment at abberton reservoir, essex, uk’, Géotechnique 63(9), 753–767.
Kramer, S. L. (1996), Geotechnical Earthquake Engineering, Prentice Hall,

New Jersey.
Kuhlemeyer, R. L. and Lysmer, J. (1973), ‘Finite element method accuracy

for wave propagation problems’, Journal of Soil Mechanics and Foundation
Division, ASCE 99(5), 421–427.
Lacy, S. and Prevost, J. H. (1987), ‘Nonlinear seismic response analysis of

earth dams’, Soil Dynamics and Earthquake Engineering 6(1), 48–63.
Lee, G. C. and Tsai, C. (1991), ‘Time domain analyses of dam-reservoir

system. part i: exact solution.’, Journal of Engineering Mechanics, ASCE
117(9), 1990–2006.
Linsley, R. K., Kohler, M. A. and Paulhus, J. L. H. (1982), Hydrology for

Engineers, McGraw-Hill, New York.
Liu, P. L. F. (1986), ‘Hydrodynamic pressures on rigid dams during earth-

quakes’, Journal of Fluid Mechanics 165, 131–145.
Lollino, P., Cotecchia, F., Zdravković, L. and Potts, D. M. (2005), ‘Numerical

analysis and monitoring of pappadai dam’, Canadian GEotechnical Journal
42, 1631–1643.
Love, A. E. H. (1927), The Mathematical Theory of Elasticity, University

Press, Cambridge.
Lysmer, J. and Kuhlemeyer, R. L. (1969), ‘Finite dynamic model for infinite

media’, Journal of the Engineering Mechanics Division, ASCE 95(4), 859–
877.
Lysmer, J., Udaka, T., Tsai, C. F. and Seed, H. B. (1975), FLUSH: A com-
puter program for approximate 3-D analysis of soil-structure interaction
358
problems, Technical Report EERC 75–30, Earthquake Engineering Re-

search Centre, University of California, Berkeley.
Madabhushi, S. P. G. and Haigh, S. K. (2001), The Bhuj, India Earthquake

of 26th January 2001, Technical Report A Field Report by EEFIT, Society
for Earthquake and Civil Engineering Dynamics, Institution of Structural
Engineers, London.
Makdisi, F. I. and Seed, H. B. (1978), ‘Simplified procedure for estimat-

ing dam and embankment earthquake-induced deformations’, Journal of
Geotechnical Engineering, ASCE 104(7), 849–867.
McCully, P. (1996), Silenced Rivers: The Ecology and Politics of Large Dams,
Zed Books, London.
Mejia, L. H. and Seed, H. B. (1983), ‘Comparison of 2-D and 3-D dy-

namic analyses of earth dams’, Journal of Geotechnical Engineering, ASCE
109(11), 1383–1398.
Mononobe, N., Takata, A. and Matumura, M. (1936), ‘Seismic stability of the

earth dam’, Transactions of the 2nd Congress on Large Dams, Washington
D. C. 4, 435–444.
Morrison, P. W., Stump, B. W. and Urhammer, R. (1976), ‘The Oroville

Earthquake Sequence of August 1975’, Bulletin of the Seismological Society
of America 66(4), 1065–1083.
Mroz, Z. (1967), ‘On the description of anisotropic work hardening’, Journal

of Mechanics and Physical Solids 15(3), 163–175.
Muraleetharan, K. K., Deshpande, S. and Adalier, K. (2004), ‘Dynamic de-

formations in sand embankemnts: centrifuge modeling and blind, fully
coupled analyses’, Canadian Geotechnical Journal 41(1), 48–69.
Muraleetharan, K. K., Mish, K. D., Yogachandran, C. and Arulanandan,

K. (1988), DYSAC2: dynamic soil analysis code for 2-dimensional prob-
lems., Technical Report Computer code, Department of Civil Engineering,
University of California, Davis.
359
Nahhas, T. M. (1987), Dynamics of earth dams, PhD thesis, University of

Southern California, Los Angeles.
Newmark, N. M. (1965), ‘Effects of earthquakes on dams and embankments’,

Géotechnique 15(2), 139–160.
Newmark, N. M. and Rosenblueth, E. (1971), Fundamentals of Earthquake

Engineering, Prentice-Hall Inc., Enlewood Cliffs, New Jersey.
Ohmachi, T. (1981), Analysis of dynamic shear strain distributed in 3-

dimensional earth dam models, in ‘Proceedings of the International Con-
ference on Recent Advances in Geotechnical Earthquake Engineering and
Soil Dynamics’, St. Louis.
Papadimitriou, A. and Bouckovalas, G. D. (2002), ‘Plasticity model for sand

under small and large cyclic strains: a multiaxial formulation’, Soil Dy-
namics and Earthquake Engineering 22(3), 191–204.
Papalou, A. and Bielak, J. (2001), ‘Seismic elastic response of earth dams

with canyon interaction’, Journal of Geotechnical and Geoenvironmental
Engineering 127(5), 446–453.
Papalou, A. and Bielak, J. (2004), ‘Nonlinear seismic response of earth dams

with canyon interaction’, Journal of Geotechnical and Geoenvironmental
Engineering 130(1), 103–110.
Papazafeiropoulos, G., Tsompanakis, Y. and Psarropoulos, P. N. (2011),

Dynamic interaction of concrete dam-reservoir-foundation: analytical and
numerical solutions, in M. Papadrakakis, M. Fragiadakis and N. D. La-
garos, eds, ‘Computational Methods in Earthquake Engineering’, Springer
Science, New York, pp. 455–488.
Parmakkis, A. (2006), Personal Communication.
Parrinello, F. and Borino, G. (2007), ‘Lagrangian finite element modelling of

dam-fluid interaction: Accurate absorbing boundary conditions’, Comput-
ers & Structures 85, 932–943.
360
Petrofski, J., Paskalov, T. and Jurukovski, D. (1974), ‘Dynamic fullscale test

of an earthfill dam’, Géotechnique 24(2), 193–206.
Phillips, C. and Hashash, Y. M. A. (2009), ‘Damping formulation for nonlin-

ear 1d site response analyses’, Soil Dynamics and Earthquake Engineering
29(7), 1143–1158.
Potts, D. M. (2003), ‘Numerical analysis: A virtual dream or practical real-

ity?’, Géotechnique 53(6), 535–573.
Potts, D. M., Dounias, G. T. and Vaughan, P. R. (1990), ‘Finite element

analysis of progressive failure of carsington embankment’, Géotechnique
40(1), 79–101.
Potts, D. M. and Zdravković, L. (1999), Finite element analysis in geotech-

nical engineering: Theory, Thomas Telford, London.
Potts, D. M. and Zdravković, L. (2001), Finite element analysis in geotech-

nical engineering: Application, Thomas Telford, London.
Poulos, H. G. and Davis, E. H. (1974), Elastic solutions for Soil and Rock
Mechanics, John Wiley and Sons, New York.
Prevost, J. H. (1977), ‘Mathematical modelling of monotonic and cyclic

undrained clay behavior’, International Journal of Numerical and Ana-
lytical Methods in Geomechanics 1(2), 195–216.
Prevost, J. H. (1981), DYNAFLOW: A Nonlinear Transient Finite Element

Analysis Program, Technical Report Report 81-SM-1, Princeton Univer-
sity.
Prevost, J. H., Abdel-Ghaffar, A. M. and Lacy, S. J. (1985), ‘Nonlinear

Dynamic Analyses of an Earth Dam’, Journal of Geotechnical Engineering
111(7), 882–897.
Puzrin, A. M. and Burland, J. B. (2000), ‘Kinematic hardening plasticity

formulation of small strain behaviour of soils’, International Journal of
Numerical & Analytical Methods in Geomechanics 24, 753–781.
361
Rathje, E. M. and Bray, J. D. (1999), ‘An examination of simplified

earthquake-induced displacement procedures for earth structures’, Cana-
dian Geotechnical Journal 36, 72–87.
Rayleigh, L. (1885), ‘On waves propagated along the plane surface of an

elastic solid’, Proceedings of the London Mathematical Society 17(1), 4–11.
Reséndiz, D., Romo, M. P. and Moreno, E. (1982), ‘El Infiernillo and La

Villita dams: seismic behavior’, Journal of the Geotechnical Engineering
Division, ASCE 108(GT1), 109–131.
Rollins, K., Evans, M., Diehl, N. and Daily, W. (1998), ‘Shear modulus
and damping relations for gravel.’, J. Geotechnical and Geoenvironmental
Engineering, ASCE 124(5), 396–405.
Roscoe, K. H. and Burland, J. B. (1968), On the generalized stress-strain

behavior of wet clay, in J. Heyman and F. A. Leckie, eds, ‘Engineering
Plasticity’, University Press, Cambridge, pp. 535–609.
Saleh, S. and Madabhushi, S. P. G. (2010), ‘Hydrodynamic pressures behind

flexible and rigid dams’, Dams & Reservoirs 20(2), 73–82.
Sarma, S. K. (1973), ‘Stability analysis of embankments and slopes’,

Sarma, S. K. (1975), ‘Seismic stability of earth dams and embankments’,

Sarma, S. K. (1979), ‘Stability analysis of embankments and slopes’, Journal

of the Geotechnical Division, ASCE 105(GT5), 1511–1524.
Schnabel, P., Lysmer, J. and Seed, H. B. (1970), SHAKE: A computer pro-

gram for the earthquake response for horizontally layered sites, Technical
Report EERC 72–2, Earthquake Engineering Research Centre, University
of California, Berkeley.
Schofield, A. N. and Wroth, P. C. (1968), Critical State Soil Mechanics,

McGraw Hill, London.
362
Seed, H. B. (1967), ‘Earthquake-resistant design of earth dams’, Canadian

Geotechnical Journal 4(1), 1–27.
Seed, H. B. (1973), Stability of earth and rockfil dams duing earthquakes,

in Hirschfeld and H. Poulos, eds, ‘Embankment - Dam Engineering’, John
Wiley.
Seed, H. B. (1979), ‘Considerations in the earthquake-resistant design of

earth and rockfill dams’, Géotechnique 29(3), 215–263.
Seed, H. B. and Idriss, I. M. (1969), ‘The influence of ground conditions on

ground motions during earthquakes’, Journal of the Soil Mechanics and
Foundations Division, ASCE 95(1), 93–137.
Seed, H. B. and Idriss, I. M. (1970), Soil moduli and damping factors for
dynamic response analyses, Technical Report EERC 70–10, Earthquake
Engineering Research Centre, University of California, Berkeley.
Seed, H. B., Lee, K. L. and Idriss, I. M. (1969), ‘Analysis of Sheffield dam fail-
ure’, Journal of the Soil Mechanics and Foundation Engineering Division,
ASCE 95(SM6), 1453–1490.
Seed, H. B., Makdisi, F. I., Idriss, I. M. and Lee, K. L. (1975), ‘The Slides
in the San Fernando Dams During the Earthquake of February 9, 1971’,
Journal of the Geotechnical Engineering Division 101(7), 651–688.
Seed, H. B. and Martin, G. R. (1966), ‘The seismic coefficient in earth dam

design’, Journal of the Soil Mechanics and Foundations Division, ASCE
92(SM3), 25–58.
Seed, H. B., Wong, R. T., Idriss, I. M. and Tokimatsu, K. (1986), ‘Moduli

and damping factors for dynamic analyses of cohesionless soils’, Journal of
Geotechnical Engineering 112(11), 1016–1032.
Sharan, S. K. (1985), ‘Finite element analysis of unbounded and incom-

pressible fluid domains’, International Journal of Numerical Methods in
Engineering 21(9), 1659–1669.
363
Sica, S., Pagano, L. and Modaressi, A. (2008), ‘Influence of past loading

history on the seismic response of earth dams’, Computers & Geotechnics
35(2008), 61–85.
Singh, R., Roy, D. and Jain, S. K. (2005), ‘Analysis of earth dams affected
by the 2001 bhuj earthquake’, Engineering Geology 80(3-4), 282–291.
Sociedad Mexicana de Ingenieria Sismica (2000), ‘Mexican strong motion

database’, CD.
Sommerfeld, A. J. W. (1912), ‘Die greensche funktion der schwingungsgle-

ichung’, Jahresbericht der Deutschen Mathematiker-Vereinigung 21, 309–
353.
Sommerfeld, A. J. W. (1949), Partial Differential Equations in Physics, Aca-

demic Press, New York.
Spyrakos, C. C. and Beskos, D. E. (1986), ‘Dynamic response of flexible

strip foundations by boundary and finite elements.’, Soil Dynamics and
Earthquake Engineering 5, 84–96.
Stara-Gazetas, E. (1986), Method for inelastic response analysis of earth

dams, PhD thesis, Rensselaer Polytechnic Institute.
Succarieh, M. F., Elgamal, A. W. and Yan, L. (1993), ‘Observed and

predicted earthquake response of La Villita Dam’, Engineering Geology
34(93), 11–26.
Taborda, D. M. G. (2011), Development of constitutive models for applica-

tion in soil dynamics, PhD thesis, Imperial College London.
Tan, H. and Chopra, A. K. (1996), ‘Dam-foundation rock interaction effects

in earthquake response of arch dams’, Earthquake Engineering and Struc-
tural Dynamics 122(5), 528–538.
Tani, S. (2000), ‘Behavior of large fill dams during earthquake and earthquake
damage’, Soil Dynamics and Earthquake Engineering 20(1-4), 223–229.
364
Terzaghi, K. (1950), Mechanics of landslides, Technical Report Engineering

Geology Volume, The Geological Survey of America.
Thompson, W. T. (1950), ‘Transmission of elastic waves through a stratified

solid medium.’, Journal of Applied Physics 21(2), 89–93.
Timoshenko, S. P. (1921), ‘On the correction for shear of the differential

equation for transverse vibrations of bars of uniform cross-section.’, Philo-
sophical Magazine 41, 741–746.
Timoshenko, S. P. (1922), ‘On the transverse vibrations of bars of uniform

cross-section.’, Philosophical Magazine 43, 125–131.
Timoshenko, S. P. and Goodier, J. N. (1934), Theory of Elasticity, Mc-Graw

Hill, New York.
Tripe, R., Kontoe, S. and Wong, T. K. C. (2013), ‘Slope topography effects

on ground motion in the presence of deep soil layers.’, Soil Dynamics and
Earthquake Engineering pp. –.
Tsai, C. S. and Lee, G. C. (1991), ‘Time domain analyses of of dam-

reservoir system. II: Substructure method’, Journal of Engineering Me-
chanics 117(9), 2007–2026.
Tsai, C. S., Lee, G. C. and Ketter, R. L. (1990), ‘A semi-analytical method for

time-domain analyses of dam-reservoir interactions’, International Journal
for Numerical Methods in Engineering 29, 913–933.
Tsiatas, G. and Gazetas, G. (1982), ‘Plane-strain and shear-beam free vibra-

tion of earth dams’, Soil Dynamics and Earthquake Engineering 1(4), 150–
160.
Tsung-ho, H., Hsueh-hai, L. Tu-hsin, H. and Cheng-jung, Y. (1976), ‘Strong-

motion observation of water-induced earthquakes at Hsinfengkiang reser-
voir in China’, Engineering Geology 10(2–4), 315–330.
365
Uddin, N. (1997), ‘A single-step procedure for estimating seismically-induced

displacements in earth structures’, Computers & Structures 64(5/6), 1175–
1182.
U.S. Army Corps of Engineers (1994), Engineering and design - arch dam
design, Technical Report EM 1110-2-2201, Department of the Army.
U.S. Army Corps of Engineers (1995), Engineering and design - gravity dam
design, Technical Report EM 1110-2-2200, Department of the Army.
U.S. Army Corps of Engineers (2000), Engineering and design - roller-

compacted concrete, Technical Report EM 1110-2-2006, Department of
the Army.
US Army Corps of Engineers (2010), ‘Willow creek lake’,

http://www.nww.usace.army.mil/html/pub/pi/flood/willow.htm.
US Geological Survey (2010), ‘The science of earthquakes’,

http://earthquake.usgs.gov/learn/kids/eqscience.php.
Vaughan, P. R. (1994), ‘Assumption, prediction and reality in geotechnical

engineering’, Géotechnique 44(4), 573–609.
Von Estorff, O. and Antes, H. (1991), ‘On FEM-BEM coupling for fluid-
structure interaction analyses in the time domain’, International Journal
for Numerical Methods in Engineering 31(6), 1151–1168.
Von Estorff, O. and Kausel, E. (1989), ‘Coupling of boundary and finite

elemenet for soil-structure interaction problems’, Earthquake Engineering
and Structural Dynamics 18, 1065–1075.
von Kármán, T. (1933), ‘Discussion on water pressures on dams during earth-

quakes’, Transactions, ASCE 98(1), 434–436.
366
Vrettos, C. (2013), ‘Dynamic response of soil deposits to vertical SH waves

for different rigidity depth-gradients’, Soil Dynamics and Earthquake En-
gineering 47, 41–50.
Vrymoed, J. (1981), ‘Dynamic FEM model of Oroville dam’, Journal of the

Geotechnical Engineering Division, ASCE 107(GT8), 1057–1077.
Vucetic, M. and Dobry, R. (1991), ‘Effect of soil plasticity on cyclic response’,

Journal of Geotechnical Engineering 117(1), 87–107.
Westergaard, H. M. (1933), ‘Water pressures on dams during earthquakes’,

Transactions, ASCE 98(1), 418–433.
Wilson, E. L. (1975), Finite element for foundation, joints and fluid, in ‘Con-
ference on Numerical Methods in Soil and Rock mechanics’, Karlsruhe.
Wilson, E. L. (1995), Three Dimensional Static and Dynamic Analysis Of

Structures, Computers & Structures Inc. - University of California, Berke-
ley, Berkeley.
Wilson, E. L. and Khalvati, M. (1983), ‘Finite elements for the dynamic anal-
ysis of fluid-solid systems’, International Journal for Numerical Methods
in Engineering 19(11), 1657–1668.
Wilson, R. C. and Keefer, D. K. (1985), Predicting areal limits of earthquake-

induced landslides, in J. I. Ziony, ed., ‘Evaluating earthquake hazards in
the Los Angeles region - An earth sciences perspective’, USGS Professional
Paper 1360, pp. 317–345.
Wolf, J. P. (1985), Dynamic soil-structure interaction, Prentice Hall, New

Jersey.
Woodward, P. K. and Griffiths, D. V. (1996), ‘Influence of viscous damping

in the dynamic analysis of an earth dam using simple constitutive models’,
Computers & Geotechnics 19(3), 245–263.
367
Yegian, M. K., Marciano, E. and Ghahraman, V. G. (1991), ‘Earthquake-

induced permanent deformations: a probabilistic approach’, Journal of
Geotechnical Engineering, ASCE 117(1), 35–50.
Yiagos, A. N. and Prevost, J. H. (1991a), ‘Two-phase elasto-plastic seismic

response of earth dams: applications’, Soil Dynamics & Earthquake Engi-
neering 10(7), 371–381.
Yiagos, A. N. and Prevost, J. H. (1991b), ‘Two-phase elasto-plastic seismic

response of earth dams: theory’, Soil Dynamics & Earthquake Engineering
10(7), 357–370.
Zangar, C. N. (1952), Hydrodynamic pressures on dams due to horizontal

earthquake effects, Technical Report EM 11, US Department of the Inte-
rior, Bureau of Reclamation.
Zangar, C. N. (1953), ‘Hydrodynamic pressures on dams due to horizontal

earthquakes’, Proceedings of the Society of Experimental Stress Analysis
10, 93–102.
Zangar, C. N. and Haefeli, R. J. (1952), ‘Electric analog indicates effect of

horizontal earthquake shock on dams’, Civil Engineering 22, 278–279.
Zdravković, L., Potts, D. M. and Hight, D. W. (2002), ‘The effect of strength

anisotropy on the behaviour of embankments on soft ground’, Géotechnique
52(6), 447–457.
Zerwer, A., Cascante, G. and Hutchinson, J. (2002), ‘Parameter estimation

in finite element simulations of rayleigh waves’, Journal of Geotechnical
and Geoenvironmental Engineering, ASCE 128(3), 250–261.
Zhao, C., Valliappan, S. and Tabatabaie, J. (1993), ‘Effect of impervious

members and reservoir bottom sediment on dynamic response of embank-
ment dams’, Soil Dynamics and Earthquake Engineering 12(4), 199–208.
Zhu, P. D. (1992), ‘Characteristics of Small Earthquakes and Stress Field:

A Study of Reservoir Induced Seismicity in Quebec’, Soil Dynamics &
Earthquake Engineering 113(11), 193–202.
368
Zienkiewicz, O. C., Bicanic, N. and Q., S. F. (1988), Earthquake input defi-

nition and the transmitting boundary condition, Technical Report Report
C/R/583/87, University College Swansea, Wales, UK.
Zienkiewicz, O. C., Chan, A. H. C., Pastor, M., Schefler, B. A. and Shiomi, T.

(1999), Computational Geomechanics with Special Reference to Earthquake
Engineering, John Wiley & Sons, London.
Zienkiewicz, O. C., Clough, R. W. and B., S. H. (1986), Earthquake anal-

ysis procedures for dams - state of the art, Technical Report Bulletin 52,
International Committee on Large Dams.
369

Pelecanos L 2014 PHD Thesis PDF

Uploaded by

Copyright:

Available Formats

Pelecanos L 2014 PHD Thesis PDF

Uploaded by

Document Information

Original Title

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

Pelecanos L 2014 PHD Thesis PDF

Uploaded by

Copyright:

Available Formats

SEISMIC RESPONSE AND

ANALYSIS OF EARTH DAMS

A thesis submitted in partial fulfillment of the requirements of

Department of Civil & Environmental Engineering

London, United Kingdom

This page is intentionally left blank.

This thesis is dedicted to my beloved family,

London, September 2013

List of figures xiv

List of tables xxxvi

2 SEISMIC PERFORMANCE OF DAMS 6

2.3.2 The scenario of reservoir-triggered seismicity . . . . . . 20

3 SEISMIC ANALYSIS OF DAMS 71

3.5 Dynamic finite element method . . . . . . . . . . . . . . . . . 83

4 NUMERICAL MODELLING OF THE RESERVOIR DO-

5 DYNAMIC RESERVOIR-DAM INTERACTION 157

6 NUMERICAL ANALYSIS OF LA VILLITA DAM 196

6.5.2 Reservoir impoundment . . . . . . . . . . . . . . . . . 233

7 PARAMETRIC SEISMIC ANALYSIS OF LA VILLITA DAM278

7.6 Stiffness inhomogeneity . . . . . . . . . . . . . . . . . . . . . . 321

2.1 Two-dimensional sketch of a uniform earth dam . . . . . . . . 8

2.14 Amplification Factor, |F (ω)| for a visco-elastic soil layer on a

2.28 Effect of the geometric characteristics of a homogeneous dam-

2.39 The reservoir-dam system considered by Chwang and Housner

3.1 Illustration of the pseudo-static method of analysis in which

3.2 Newmark’s sliding block method (Wilson and Keefer, 1985) . . 75

3.14 Illustration of the influence of parameters c and Gmin on the

4.1 Geometry of the dam-reservoir-foundation system: (1) dam,

4.11 Hydrodynamic pressure at the base of the dam by using the

4.24 Frequency response of the hydrodynamic force induced by a

4.35 Distribution of the peak hydrodynamic pressure on the face

4.46 Distribution of the peak hydrodynamic pressure on the face

5.1 Geometry of the dam-reservoir-foundation system: (1) Dam,

5.7 Amplification of acceleration, |F | for the cantilever dam with

5.19 Cases of combinations of ω/ωd and ωr /ωd frequency ratios ex-

5.35 Response spectrum (for damping ξ = 5%) of the accelerations

6.11 The experimental setup used by Elgamal et al. (1990) in the

6.23 Spatial variation (in the vertical direction) of shear stiffness,

6.48 New calibration of the Logarithmic CNL model against the

6.61 Acceleration time history at the crest of La Villita dam during

6.75 Shear stress-strain response in the downstream rockfill (DR)

7.1 Acceleration response at the crest of La Villita dam during

7.12 Stress paths in the downstream alluvium (DA) of La Villita

7.27 Mesh of the dam showing the locations of the monitoring

7.43 Vertical profile of the maximum values of horizontal accelera-

4.1 Geometry of the FE meshes considered for the ramp acceler-

5.1 Cases of combinations of ω/ωd and ωr /ωd frequency ratios ex-

6.1 Summary of earthquake events between 1975-1985 and accel-

6.6 Calibration parameters for the Logarithmic CNL model. . . . 225

7.1 Values of the Rayleigh damping coefficients A & B. . . . . . . 311

provide results comparable to the observed measurements. However, the

1.2 Scope of research

1.3 Layout of thesis

finite element analysis which is employed in this thesis.

2.2 Design & construction of dams

2.2.1 Earth dams

Figure 2.1: Two-dimensional sketch of a uniform earth dam

Figure 2.2: Two-dimensional sketch of a zoned earth dam

Besides, many dams have an additional thin upstream zone consisting of