International Journal of Civil Engineering

https://doi.org/10.1007/s40999-018-0325-9

RESEARCH PAPER

Assessment of Foundation Mass and Earthquake Input Mechanism

Effect on Dam–Reservoir–Foundation System Response
Mohsen Ghaemian1 · Ali Noorzad2,3 · Hamid Mohammadnezhad1

Received: 18 November 2017 / Accepted: 9 May 2018

© Iran University of Science and Technology 2018

Abstract
Concrete dams are one of the most important infrastructures in every country and the seismic safety assessment of them is
a major task in dam engineering field. Dam–foundation–reservoir system analysis is a complex interaction problem because
this system consists of three domains with different behaviors. For accurate analysis of this system, some important fac-
tors should be considered such as foundation mass and earthquake input mechanism. In this paper, the effect of foundation
mass and earthquake input mechanism on seismic response of concrete gravity dam is investigated. For this purpose, two
different methods are introduced for modeling of massed semi-infinite foundation in finite element method, namely free-
field boundary condition and domain reduction method (DRM). To verify the feasibility of proposed methods for seismic
analysis of dam–foundation–reservoir system, the displacement and stress outputs using proposed methods are compared
with EAGD-84 results. The obtained results indicate that both methods are accurate enough for finite element modeling of
massed foundation. Finally, Koyna concrete gravity dam is analyzed for rigid, massless and massed foundation cases using
DRM and it is concluded that the foundation has significant effect on dam response and the common massless foundation
approach overestimates the dam response.

Keywords Soil–structure interaction · Free-field boundary condition · Domain reduction method · Massed foundation ·
Boundary condition · Seismic wave propagation · Radiation damping

1 Introduction
Concrete dams are one of the most important infrastructures
in every country and because of their social and economic
effects the seismic safety assessment of them is a major task
in dam engineering field. Dam–foundation–reservoir sys-
tem analysis is a complex interaction problem because this
system consists of three domains with different behaviors.
For accurate analysis of this system, some important fac-
tors should be considered in each domain of the system.
Mohsen Ghaemian
ghaemian@sharif.edu
Ali Noorzad
a_noorzad@sbu.ac.ir
1
Department of Civil Engineering, Sharif University
of Technology, Tehran, Iran
2 coelastic half-space using finite element method has some
Faculty of Civil, Water and Environmental Engineering,
Shahid Beheshti University, Tehran, Iran
3 truncated in a finite domain and an appropriate artificial
Iranian Committee on Large Dams (IRCOLD), Tehran, Iran
concreteDane
The appropriate linear and nonlinear behavior of concrete,
geometric nonlinearity and appropriate damping mecha-
nism should be considered for proper modeling of the dam’s
structure. In foundation domain, foundation mass, material
and radiation damping, seismic wave propagation effect,
soil
earthquake input mechanism and absorbing boundary con-Foundation
ditions for representation of far field are effective factors.
Compressibility of the water, absorption of seismic wave by
the sedimentary materials at the bottom of the reservoir andreseroir
unbounded nature of reservoir domain should be considered
in fluid domain modeling. Most researchers have analyzed
the dam–foundation–reservoir system considering some of
these factors [1–4]. Based on previous studies, modeling
foundation properly and choosing the appropriate earthquake
input mechanism have major effects on responses obtained
from analyzing the system [5, 6].
or Referenced
The reasonable model for the foundation in dam–founda-
tion system is a viscoelastic half-space. Modeling of vis-
major challenges. The unbounded foundation should be

13
Vol.:(0123456789)
 International Journal of Civil Engineering

boundary condition should be applied at truncated boundary In 2D problems, a free-field element consists of a one-
to absorb the reflected waves. This procedure is not easy to dimensional element that simulates the vertically propagat-
implement. The reason for this particular problem is because ing plane wave. These elements are attached at the side of
proper artificial boundary in combination with an appropri- the model with viscous or viscous-spring boundary con-
ate earthquake input mechanism should be able to produce dition. Transformation of the free-field motion from free-
the free-field motion in absence of any structures. Plus, it field elements to main domain is performed by applying the
should be able to absorb the reflected waves emanating in equivalent forces based on the Eqs. (1) and (2).
different angles and frequencies. ( m )
𝜕u 𝜕uff
Considering challenges in massed foundation modeling, m
𝜎n = −𝜌VP − + 𝜎n0 , (1)
𝜕t 𝜕t
most researchers use massless foundation approach in
dam–foundation–reservoir analysis. Compelling number of ( )
studies show that the massless foundation approach is unre- 𝜕vm 𝜕vff
𝜏 m = −𝜌VS − + 𝜏 0. (2)
alistic. Tan and Chopra [7, 8], and Chopra [9] concluded that 𝜕t 𝜕t
in the massless foundation model the stresses and displace-
In these equations, the viscous boundary condition effect
ments of dam are overestimated. Earthquake input mecha-
is considered to absorb the reflected waves from structures
nism has significant effects in dam–foundation–reservoir
(Galavi et al. [13]). A model with attached free-field ele-
system. Leger et al. [6] investigated four common earth-
ments at side-truncated boundary is presented in Fig. 1.
quake input mechanisms and reported that some of them
are unrealistic. They also showed that input mechanism and
2.1 Dynamic Response of Dam–Foundation–
foundation type have considerable effects on dam response.
Reservoir System
In this paper, two different methods, namely FFBC and
DRM, which both could account for foundation mass (and
To present the capability of the explained method, the
in turn, wave propagation effects) are employed for seismic
dynamic response of Koyna dam–foundation–reservoir
analysis of a system of dam-massed foundation–reservoir.
system is analyzed with said method and EAGD-84. The
Results from both methods are compared against EAGD-
compressibility of water and semi-unbounded nature of res-
84 [10] program results which show good conformity. It is
ervoir are considered in the analysis. Material properties are
worth mentioning that EAGD-84 program, Developed by
shown in Table 1.
Chopra, is based on sub-structure approach for dynamic
Figure 2 shows the crest relative displacement time his-
analysis of concrete gravity dams which represents more
tory from proposed method and EAGD-84 software. Enve-
realistic approach for foundation and thus yields accept-
lope contours of maximum and minimum principal stresses
able results. Results approved by many researchers through
from proposed method and EAGD-84 software are presented
numerous studies. Finally, the results of massed, massless
in Figs. 3 and 4. The results indicate an acceptable conform-
and rigid foundation model using DRM, are compared and
ity between the FFBC method and EAGD-84 results. The
some conclusions are presented.
maximum value of dynamic principal stresses and displace-
ment are given in Table 2. These comparisons indicate that
the simulation considerations used in the proposed method
columnisthe
2 Free-Field Boundary Condition TheID soil
Ff Bc free field thatanons
element
to computetheequivalent
nodalforcesonthelateral
In most cases in practice, vertical propagation of earthquake boundary ateach
waves is a common assumption. Considering this, in a 2D nodeseparately
finite element model of a homogeneous medium, all nodes
at the base are excited with same input motions. At side- is
boundary nodes, each node, depending on its position, is
excited with the same input motion but out of phase. Thus,
for accurate simulation of wave propagation in a 2D founda-
tion, the right input motion should be calculated and then
applied at proper nodes separately. This requires a pre-pro-
cessing analysis for solving the wave propagation equation
and obtaining required excitation for each individual node.
A common replacement of such time-consuming process is
free-field element which was proposed by Zienkiewicz et al. Fig. 1 2D dam-semi infinite foundation–reservoir with attached free-
[11] and Wolf [12]. field elements (Itasca Consulting Group, Inc. [14])

13
International Journal of Civil Engineering

Table 1 Mechanical properties for mass concrete and foundation rock and water
Mass concrete Foundation rock Water
Modulus of Poison’s ratio Mass density Damping Modulus of Poison’s ratio Mass density Damping Bulk modulus Mass density
elasticity elasticity

30 GPa 0.2 2630 kg/m3 5% 30 GPa 0.2 2630 kg/m3 5% 2.07 GPa 1000 kg/m3

Fig. 2 Comparison of relative

crest displacement in dam–
foundation–reservoir case

Fig. 3 Envelope contours of

maximum principal stresses in
dam–foundation–reservoir case
a FFBC method (Pa), b EAGD-
84 (MPa)

are feasible for simulating the semi-unbounded compressible
reservoir and massed foundation.
3 Domain Reduction Method
The DRM is a two-step procedure that was developed for
seismic analysis of large-scale problems [15]. In the first
step, the free-field analysis is performed in a half-space
domain including seismic source. In this step, the effective
equivalent force on a one-element strip is calculated based
on the free-field response and reserved for the second step.
In the second step, the effective equivalent forces are applied
in a reduced domain including the local site and structure.
The same formulation in the original DRM paper [15] was
used in this study. In addition, the damping was added to
the original formulation for viscoelastic case. The general
framework of the dam-massed half-space foundation includ-
ing the seismic source is illustrated in Fig. 5.
The half-space foundation is limited to the Γ+ boundary
due to costs and time-consuming nature of numerical meth-
ods (Fig. 6). The total computational domain breaks into

13
 International Journal of Civil Engineering

Fig. 4 Envelope contours of

minimum principal stresses in
dam–foundation–reservoir case
a FFBC method (Pa), b EAGD-
84 (MPa)

Table 2 Comparison of maximum peak values obtained from the

FFBC method and EAGD-84

Max stress (MPa) FFBC 7.7

EAGD-84 7.3
Min stress (MPa) FFBC − 9.4
EAGD-84 − 9.5
Max displacement (cm) FFBC 7.4
EAGD-84 7.9

Fig. 6 Truncated seismic region

Fig. 5 Schematic of dam–foundation–fault system

and + domains by the virtual boundary of Γ (Fig. 7). In

Figs. 6 and 7, ui, ue and ub illustrate the nodal displacement
vectors in the internal domain of , external domain of +,
Fig. 7 Regions partitioned into two subdomains
and the boundary between them, respectively.
The dominant equation of motion for the entire domain
is Navier’s equation. The FE discretization of the Navier’s
equation in and + domains is as follows: [ Ω+ Ω+
]{ } [ Ω+ Ω+ ]{ }
Mbb Mbe ü b Cbb Cbe u̇ b
+
[ ]{ } [ Ω Ω ]{ }
+ + Ω+
Ω Ω ü e Ω+ u̇ e
Meb Mee Ceb Cee
MiiΩ Mib
Ω
ü i Cii Cib u̇ i [ Ω+ Ω+ ]{ } { } (4)
Ω Ω + Ω Ω Kbb Kbe ub −Pb
Mbi Mbb ü b Cbi Cbb u̇ b + = , in Ω+ ,
[ Ω Ω ]{ } { } (3) Keb Ω+
KeeΩ+ ue Pe
Kii Kib ui 0
+ Ω Ω = , in Ω,
Kbi Kbb ub Pb

13
International Journal of Civil Engineering

where M, C and K represent mass, damping and stiffness ue = u0e + we , (7)

matrices, respectively; and + superscripts define the
domain where the matrix belongs and i, e, and b indexes
related to the internal, external, and boundary nodes. A
simple auxiliary problem is solved first which includes the
free-field of the main model, while removing the structure
where we is the relative displacement with respect to u0e.
By substituting Eqs. (7) and (6) into Eqs. (4) and (3) and
add one of these equations to one another, the final equation
is achieved:

⎡ MiiΩ Ω
Mib 0 ⎤⎧ ü i ⎫ ⎡ CiiΩ Ω
Cib 0 ⎤⎧ u̇ i ⎫
+ + ⎪ ⎪
⎢ M Ω M Ω + M Ω M Ω ⎥⎨ ü b ⎬ + ⎢ CΩ CΩ + CΩ CΩ+ ⎥⎪
+ ⎪
be+ ⎥⎨ b ⎬
u̇
⎢ bi bb Ω+ bb be+ ⎥ ⎢ bi bb bb
⎦⎪ ẅ ⎪ ⎦⎪ ẇ ⎪
+
⎣ 0 Meb Ω
Mee
⎩ e⎭ ⎣
0 Ω
Ceb Ω
Cee
⎩ e⎭
(8)
⎡ KiiΩ KibΩ 0 ⎤⎧ ui ⎫ ⎧ 0 ⎫
+ ⎪ ⎪ ⎪ ⎪
+ ⎢ Kbi Kbb + Kbb Kbe ⎥⎨ ub ⎬ = ⎨ −Mbe ü e − Cbe u̇ e − Kbe ue ⎬.
Ω Ω Ω + Ω Ω + Ω + Ω +
0 0 0
⎢ ⎥
⎦⎪ w ⎪ ⎪ M Ω ü 0 + Ceb Ω 0 ⎪
+ + + + +
⎣ 0 Ω
Keb Ω
Kee
⎩ e ⎭ ⎩ eb b
Ω 0
u̇ b + Keb ub
⎭

and site conditions. Solving this problem is much simpler A key note in Eq. (6) is the replacement of the seismic
than solving the main problem. u0i , u0b , u0e and P0b , according fault force Pe, by the effective nodal forces, according to
to Fig. 8 that indicates the nodal displacements and bound- Eq. (9):
ary forces.
The equation of motion for + in the auxiliary problem ⎧ P eff ⎫ ⎧ 0 ⎫
is according to Eq. (5): ⎪ i eff ⎪ ⎪ + + Ω 0⎪
+
Peff = ⎨ Pb ⎬ = ⎨ −Mbe Ω 0 Ω 0
ü e − Cbe u̇ e − Kbe u e ⎬. (9)
[ ]{ } [ ]{ } ⎪ Pe eff ⎪ ⎪ M Ω ü 0 + CΩ u̇ 0 + K Ω+ u0 ⎪
+ +

Ω+
Mbb MbeΩ+ 0
ü b Cbb CbeΩ+
u̇ bΩ+ 0 ⎩ ⎭ ⎩ eb b eb b eb b ⎭
Ω+ Ω+ + Ω+ Ω+
Meb Mee ü 0e Ceb Cee u̇ 0e
[ Ω+ Ω+ ]{ 0 } { 0 } (5) The sub-matrices of M be, C be, K be, M eb, C eb and Keb,
Kbb Kbe ub −Pb
+ Ω+ Ω
Keb Kee
+ 0
ue
=
Pe
, in Ω+ . in Eq. (9) are mostly zero at all the points of + exclud-
ing a finite layer next to Γ. Thus, it is needed to calculate
effective equivalent forces in a finite layer in the auxiliary
The nodal force of Pe in terms of free-field variables at the
problem.
boundary Γ can be rewritten in Eq. (5) as follows:

Ω 0 + Ω 0 + Ω 0 + Ω 0 + Ω 0 + Ω 0 +
Pe = Meb ü b + Mee ü e + Ceb u̇ b + Cee u̇ e + Keb ub + Kee ue . 3.1 Dynamic Response of Dam–Foundation–
(6) Reservoir System
A good idea for simplifying the analysis is to use a change
of variable. Total displacement, ue can be written as sum of The assessment of the proposed method as a feasible
the free-field response and relative displacement according boundary condition in dam-massed foundation sys-
to Eq. (7): tem problem is investigated in this part. Results of the

Fig. 8 Auxiliary problem a entire auxiliary seismic region and b auxiliary seismic region partitioned into two subdomains [15]

13
 International Journal of Civil Engineering

proposed method are compared with EAGD-84 results.
The Koyna dam was selected for these analyses and water
level was selected to be 91.7 m. EAGD-84 (earthquake
analysis of concrete gravity dams) computer program
was developed by Fenves and Chopra [10]. This program
evaluates the linear response of concrete gravity dams
due to earthquake considering the effects of dam–foun-
dation–reservoir interaction and absorption of seismic
wave by the sedimentary materials at the bottom of the
reservoir. The concrete gravity dam is idealized as an
assemblage of planer, four-node finite elements. Constant
hysteretic damping is used to represent energy dissipation
mechanism in dam structure. The reservoir is idealized as
a fluid domain and the foundation is idealized as a semi-
infinite, homogeneous, isotropic, viscoelastic half-plane.
The Koyna record is selected for this case. The material
properties for F.E model and EAGD-84 models are given
in Table 3.
The crest relative displacement time history from the
proposed method and EAGD-84 program are presented in
Fig. 9. The envelope contours of maximum and minimum
principal stresses in dam body are presented in Figs. 10
and 11.

Table 3 Material properties

Concrete dam Rock foundation
Modulus of elasticity Poison’s ratio Mass density Damping Modulus of elasticity Poison’s ratio Mass density Damping

30 GPa 0.2 2630 kg/m3 5% 22.4 GPa 0.33 2643 kg/m3 5%

Fig. 9 compared the crest displacement in dam–foundation case between DRM and EAGD-84

Fig. 10 Envelope contours of

maximum principal stresses
in dam–foundation case a pro-
posed method (Pa), b EAGD-84
(MPa)

13
International Journal of Civil Engineering

Fig. 11 Envelope contours of

minimum principal stresses in
dam–foundation case a pro-
posed method (Pa), b EAGD-84
(MPa)

Table 5 Comparison of maximum peak values obtained from the

Table 4  Maximum value of response in F.E. model and EAGD-84 in massed, massless and rigid case
dam–foundation case
System Dam–founda-
Response Proposed model EAGD-84 tion–reservoir

Max stress (MPa) 3.2 3.5 Max stress (MPa)

Min stress (MPa) − 4.2 − 4.14  Massed 4.19
Max displacement (cm) 3.19 3.34  Massless 8.6
 Rigid 7.9
Min stress (MPa)
 Massed − 5.2
 Massless − 8.4
The maximum value of displacement and dynamic prin-
 Rigid − 6.45
cipal stresses are summarized in Table 4.
Max displacement (cm)
Figures 9, 10 and 11 prove that similar trend is observed
 Massed 4.6
for crest displacement time history as well as patterns for
 Massless 7.3
stress envelope contours. Table 4 gives a good agreement
 Rigid 4.7
between the values of two sets of obtained results.
The obtained results demonstrate good agreement
between proposed method and EAGD-84 program results.
These results indicate that the proposed method is a fea- The maximum value of the crest relative displacement
sible method for simulating massed semi-unbounded vis- in massless foundation case is larger than other models.
coelastic foundation in dam–foundation–reservoir system. In addition, in massed foundation case, the crest displace-
ment and maximum value for compressive and tensile
stresses have decreased by 37, 38 and 51%, respectively,
in comparison with the massless foundation model.
4 Comparison of Massed, Massless
and Rigid Foundation Model Results Using
DRM 5 Conclusions

In this section, Koyna concrete gravity dam is analyzed
for three cases, namely rigid, massless and massed foun-
dation using DRM. The maximum value of the dynamic
relative crest displacement and principal stresses in three
model cases are summarized in Table 5.
The soil–structure interaction has great effect on struc-
tural response under earthquake excitation. This effect is
fairly important in dam engineering field because of its
economic and social effect. Thus, finite element modeling
of semi-infinite foundation in dam–foundation–reservoir

13

system and considering its mass is of great importance.
In this study, two different methodologies have been
suggested to investigate this effect. Both methods were
verified through comparing their results with EAGD-84
Results. Then, DRM was employed for seismic analysis
of Koyna concrete gravity dam for three different cases of
rigid, massless and massed foundation. It is found that the
common massless foundation approach overestimate the
dam response. This overestimation is not acceptable from
the economical point of view. Thus, using these simple
and applicable suggested methods, the massed viscoelastic
semi-infinite foundation can be considered in analysis, car-
rying its true natural effects into consideration.
References
1. Ghaemian M, Ghobarah A (1998) Staggered solution schemes for
dam-reservoir interaction. J Fluids Struct 12:933–948
2. Ghaemian M, Ghobarah A (1999) Nonlinear seismic response of
concrete gravity dams with dam-reservoir interaction. Eng Struct
21:306–315
3. Chopra AK, Chakrabarti P (1981) Earthquake analysis of concrete
gravity dams including dam-water-foundation rock interaction.
Earthq Eng Struct Dyn 9:363–383
4. Fenves G, Chopra AK (1984) Earthquake analysis of concrete
gravity dams including reservoir bottom absorption and dam-
water-foundation rock interaction. Earthq Eng Struct Dyn
12:663–680
13
International Journal of Civil Engineering
5. Bayraktar A, Hancer E, Akköse M (2005) Influence of base-rock
characteristics on the stochastic dynamic response of dam–reser-
voir–foundation systems. Eng Struct 27:1498–1508
6. Léger P, Boughoufalah M (1989) Earthquake input mechanisms
for time-domain analysis of dam-foundation systems. Eng Struct
11:37–46
7. Tan H, Chopra AK (1995) Earthquake analysis of arch dams
including dam-water foundation rock interaction. Earthq Eng
Struct Dyn 24:1453–1474
8. Tan H, Chopra AK (1995) Dam-foundation rock interaction effects
in frequency-response functions of arch dams. Earthq Eng Struct
Dyn 24:1475–1489
9. Chopra AK (2008) Earthquake analysis of arch dams: factors to
be considered. In: Proceedings of the 14th World Conference on
Earthquake Engineering. Beijing
10. Fenves G, Chopra AK (1984) EAGD-84: A computer program
for earthquake response analysis of concrete gravity dams. Report
No. UCB/EERC-84/11. Earthquake Engineering Research Center,
University of California, Berkeley
11. Zienkiewicz OC, Bicanic N, Shen FQ (1989) Earthquake input
definition and the transmitting boundary conditions. Adv Comput
Nonlinear Mech 300:109–138
12. Wolf JP (1988), Soil-structure interaction analysis in time domain.
Prentice Hall, Upper Saddle River
13. Galavi V, Petalas A, Brinkgreve RBJ (2013) Finite element mod-
elling of seismic liquefaction in soils. Geotech Eng J SEAGS
GSSEA 44:55–64
14. Itasca Consulting Group, Inc (2004) FLAC (Fast Lagrangian
Analysis of Continua). Version 4. Minneapolis
15. Bielak J, Loukakis K, Hisada Y, Yoshimura C (2003) Domain
reduction method for three-dimensional earthquake mode-
ling in localized regions. Part I: theory. Bull Seismol Soc Am
93(2):817–824