Applied Mathematics and Computation 184 (2007) 704714

www.elsevier.com/locate/amc

Stochastic seismic response of Keban dam

by the nite element method
Mehmet Akkose, Suleyman Adanur, Alemdar Bayraktar *,
A. Aydn Dumanoglu
Karadeniz Technical University, Department of Civil Engineering, 61080 Trabzon, Turkey

Abstract
In this study, stochastic seismic response of a rock-ll dam is investigated by nite element method. The Keban dam
constructed in Elazg, Turkey is chosen as a numerical example. The interaction of the rock-ll dam with the reservoir is
neglected, but not the foundation rock. The properties of the dam materials were taken from the dam project and assumed
to be linearly elastic, homogenous and isotropic in the analysis.
A stationary and ergodicity assumption are is made for stochastic seismic analysis. The EW component of the Erzincan
Earthquake recorded on March 13, 1992, Erzincan, Turkey is chosen as a ground motion since it occurred nearby the dam
site. The component considered is applied to the dam in the horizontal direction.
The seismic response of the Keban dam subjected to the Erzincan Earthquake is also obtained by the deterministic
method. The results obtained from stochastic and deterministic analysis are compared to each other. It is seen that the
results obtained from the stochastic analysis are smaller than those from the deterministic analysis.
 2006 Elsevier Inc. All rights reserved.
Keywords: Stochastic response; Rock-ll dam; Finite element method; Seismic ground motion

1. Introduction
Rock-ll dams are constructed for various purposes such as irrigation, energy production, ood control
and recreation. A serious damage on these dams has been not recorded in the literature due to an earthquake
ground motion [1]. Accordingly, it can be said that rock-ll dams are highly resistant to seismic loads. The
satisfactory seismic behaviour of these dams is due to the capacity of the rock-ll body.
Ozkan [1] and Gazetas and Dakoulas [2] have presented comprehensive reviews on theoretical methods for
estimating the dynamic response and the performance of earth and rock-ll dams subjected to strong earthquake ground motions. Several factors such as liquefaction eects, non-linear material behaviour, and permanent deformations aect the dynamic response of earth and rock-ll dams during the earthquakes. Linear and

Corresponding author.
E-mail address: alemdar@ktu.edu.tr (A. Bayraktar).

0096-3003/$ - see front matter  2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.amc.2006.05.186

M. Akkose et al. / Applied Mathematics and Computation 184 (2007) 704714

705

non-linear earthquake responses of earth and rock-ll dams including these factors were carried out by many
researchers [36]. These studies were performed by using the deterministic methods. In recent years, the stochastic seismic responses of earth dams have also been investigated by only a limited number of researchers
[7,8]. However, it can be seen from the literature review that a few works on stochastic response of rock-ll
dams to ground motion have been studied. Therefore, the objective of this study is to determine the stochastic
seismic response of a rock-ll dam using the nite element method.
2. Stochastic analysis formulation
An accelerationtime history of ground motion recorded at one point is used as seismic input in the deterministic method. In the stochastic method, however, recorded ground motions appropriate to the site are characterized as statistically. Since the ground motion caused by seismic disturbance is random, the best way to
characterize the random excitation statistically is to employ a power density function and autocorrelation
function. Once, the power spectral density function or the autocorrelation function of the seismic input is
known, the cross power spectral density function can be determined easily.
In this study, only the nal expression for the cross power spectral density function will be given. Detailed
expressions for this function are explained elsewhere [911]. If a single ground acceleration record is used for
the input, cross power spectral density function, Sij (w), can be determined by using the equation of motion of
the system as [12,13]
S ij w S in w

N X
N
X
r1

wir wjs H ir wH js w;

s1

where w is the frequency, H(w) is the frequency response function, Sin(w) is the power spectral density function
of the ground motion, N is the number of modes which are considered to contribute to the response, wjr is the
contribution of the rth mode to Uj(t) displacement and * denotes the complex conjugate.
2.1. Spectral moments
Statistics related to the structural behaviour for a stationary process can be determined by using the zeroth,
the rst and the second spectral moments of the output process [11]. Spectral moments, which can be
expressed in terms of power spectral density function and frequency, may be calculated as follows [12]
Z 1
km;ij 2
wm S ij w dw; m 0; 1; 2;
2
0

where m = 0, 1, 2 is the zeroth, the rst and the second spectral moments, respectively. These parameters will
then be used while obtaining the mean of maximum value, variance and frequency of occurrence [1315].
2.2. Expected maximum value
The expected maximum value is considered to be the most important parameter in the stochastic analysis of
structures aected by seismic loads. In the stochastic analysis the expected maximum value (l) is the mean
value of all maximum values. The expected maximum value, which depends on the peak factor and the
root-mean-square response, can be expressed as
p
l p k0 :
3
Standard deviation of l is expressed as
p
r q k0 ;

where k0 is the zeroth spectral moment dened by Eq. (2), p and q are the peak factors, which are the functions
of the duration of the motion and the mean zero crossing rate, respectively [16].

706

M. Akkose et al. / Applied Mathematics and Computation 184 (2007) 704714

2.3. Occurrence frequency

Frequency of occurrence is described as the average number of times that the line (y(t) = 0) is crossed by the
response in a unit of time. For Gaussian process of zero average, the average number of times in the zero level
crossed by the process in a unit of time is expressed as
s
1 k2
t
:
5
p k0
Because the zero level is crossed two times for each cycle, frequency of occurrence for the response process
will be equal to t/2 as
s
t
1 k2
f0
;
6
2 2p k0
where k2 is the second spectral moment dened by Eq. (2).
3. Numerical application and discussions
In this study, the Keban dam constructed in Elazg, Turkey (Fig. 1) is chosen as a numerical example to
investigate the stochastic seismic response of a rock-ll dam by the nite element method. Fig. 2 shows the
vertical cross-section of the Keban dam. The nite element mesh of the dam is shown in Fig. 3. The Keban
dam is 163 m high from riverbed. The crest has a length of 1097 m. The main purpose of the dam is to regulate
river ow and supply energy. In the nite element mesh of the dam, there are 326 nodes and 286 quadrilateral
elements. The dam is treated as a plane strain problem. The interaction of the rock-ll dams with the reservoir
has generally neglected [17,18]. Therefore, the interaction with the reservoir is accordingly ignored, but not the
foundation rock.
Materials in the dam section can be grouped in three main categories: compacted rock-ll placed at various
lifts, the impervious clay core anked by transition lters and a concrete core at the bottom of the dam. The
properties of these materials taken from the dam project are as follows: For the compacted rock-ll, elasticity
modulus E = 1.632 1010 N/m2, mass density q = 2120.29 kg/m3, and Poissons ratio m = 0.36; for the impervious clay core, elasticity modulus E = 1.015 1010 N/m2, mass density q = 2089.70 kg/m3, and Poissons
ratio m = 0.45; for the concrete core, elasticity modulus E = 2 1010 N/m2, mass density q = 2446.48 kg/m3,
and Poissons ratio m = 0.15. The elasticity modulus, mass density and Poissons ratio of the foundation rock

Fig. 1. The Keban dam constructed in Elazg, Turkey.

M. Akkose et al. / Applied Mathematics and Computation 184 (2007) 704714

707

1 Compacted Rock-fill
2 Clay Core

1.755:1

1.667:1

3 Concrete Core
1

4 Foundation Rock

1
6:1

2.5:1

6:1

Fig. 2. Cross-section of the Keban dam.

1
Z

Line 1
Line 3

III

III

Line 2

Y
II

II
A

I
290.70 m

302.50 m

302.49 m

207.00 m

Fig. 3. Finite element mesh of the Keban dam.

are taken as 1.379 1010 N/m2, 2689.09 kg/m3, and 0.24, respectively. The materials used in this study are
assumed to be linearly elastic, homogenous and isotropic.
A stationary assumption where the statistical parameters are independent of time is made for the stochastic
analysis. Besides, while calculating the statistics to represent the random process, like ensemble averages, some
diculties are encountered. The ergodicity assumption is made to overcome these diculties and only one
earthquake record is used in this study. Although earthquake motions are not completely stationary, these
motions can be considered as stationary processes under certain conditions because of its analytical simplicity.
Perhaps an earthquake ground motion is not stationary along the whole motion, but it is an acceptable
approximation to consider the ground motion stationary along the durations where maximum structural
responses take place. A stationary model simplies the computations and gives satisfying results. The response
of structural systems to stationary excitations is of wide engineering interest [14,1922].
The EW component of the Erzincan Earthquake recorded on March 13, 1992, Erzincan, Turkey is chosen
as ground motion and given in Fig. 4 since it occurred nearby the dam site. The component considered is

Acceleration (m/s2 )

6.00
4.00
2.00
0.00
-2.00
-4.00
-6.00
0.0

3.0

6.0

9.0

12.0

15.0

18.0

21.0

Time (s)

Fig. 4. The EW component of the Erzincan Earthquake recorded on March 13, 1992.

M. Akkose et al. / Applied Mathematics and Computation 184 (2007) 704714

PSD Function (m2/s3)

708

0.04
0.03
0.02
0.01
0.00
0.0

10.0

20.0

30.0

40.0

50.0

60.0

Circular Frequency (rad/s)

Fig. 5. Power spectral density (PSD) function of the Erzincan Earthquake.

applied to the dam in the upstreamdownstream direction. The power spectral density (PSD) function of the
Erzincan Earthquake is determined with the Fourier transforms of the autocorrelation function as shown in
Fig. 5.
In this paper, the dynamic response of the Keban dam subjected to the Erzincan Earthquake is also
obtained by the deterministic method. The results obtained from stochastic and deterministic analysis are
compared to each other. The seismic responses of the Keban dam are obtained for a time interval of 0.00225 s.
3.1. Displacements
Mean of maximum values of displacements are calculated from stochastic seismic analysis. The absolute
maximum values of displacements are obtained from deterministic dynamic analysis. Horizontal displacements along the core of the rock-ll dam at the marked nodes on line 1 (Fig. 3) obtained from stochastic
and deterministic seismic analyses of the Keban dam are plotted in Fig. 6. Vertical displacements along the
horizontal length of the dam at the marked nodes on line 2 and line 3 (Fig. 3) obtained from stochastic
and deterministic seismic analyses of the dam are also plotted in Figs. 7 and 8, respectively. It is seen from
320
4.01cm

280

Height (m)

240
200
160
120
80

Deterministic
Stochastic

40
0
0.00

0.03

0.06

0.09

0.12

0.15

Displacement (m)

Fig. 6. Horizontal displacements along the height of the clay core at the marked nodes on line 1 of the rock-ll dam.

Displacement (m)

0.025
Deterministic
Stochastic

0.020
0.015
0.010
0.005
0.000
0

50

100

150

200

250

300

350

400

450

Horizontal Distance (m)

Fig. 7. Vertical displacements along the horizontal length at the marked nodes on line 2 of the rock-ll dam.

M. Akkose et al. / Applied Mathematics and Computation 184 (2007) 704714

709

Displacement (m)

0.020
Deterministic
Stochastic

0.016
0.012
0.008
0.004
0.000
0

20

40

60

80

100

120

140

160

Horizontal Distance (m)

Fig. 8. Vertical displacements along the horizontal length at the marked nodes on line 3 of the rock-ll dam.

Displacement (m)

0.15
0.10
0.05
0.00
-0.05
3.6s

-0.10

4.3s

-0.15
0.0

3.0

6.0

9.0

12.0

15.0

18.0

21.0

Time (s)

Fig. 9. Time-history of horizontal displacements at the crest point (node 1).

these gures that the expected maximum values of horizontal and vertical displacements obtained from stochastic dynamic analysis are smaller than the absolute maximum horizontal and vertical displacements
obtained from deterministic dynamic analysis. It is also seen from Figs. 7 and 8 that the vertical displacements
at the marked nodes on line 2 and line 3 decrease towards the core of the rock-ll dam.
The displacements obtained from the stochastic dynamic analysis can be veried by calculating the timehistory of the horizontal displacements at the crest point (node 1) of the rock-ll dam from the deterministic
dynamic analysis as shown in Fig. 9. Taking the average of 15 maximum horizontal displacements shown in
Fig. 9, the mean of maximum horizontal displacements can be calculated as 3.98 cm for the deterministic analysis. The expected maximum horizontal displacement obtained from stochastic dynamic analysis is 4.01 cm as
shown in Fig. 5. The maximum displacement obtained by averaging, which is 3.98 cm, is very close to 4.01 cm
obtained from the stochastic analysis. This shows the correctness of the displacements obtained from the stochastic analysis.
3.1.1. Frequencies of occurrence of displacements
The frequency of occurrence of horizontal displacements along the core of the rock-ll dam at the marked
nodes on line 1 of the rock-ll dam (Fig. 3) are calculated using Eq. (6) and depicted in Fig. 10. The values of
320

1.67
1.67
1.67
1.67
1.66
1.66
1.66
1.66
1.65
1.66
1.67
1.69
1.69
1.72
1.74

280

Height (m)

240
200
160
120
80
40
0
1.5

1.6

1.7

1.8

1.9

2.0

Occurence Frequency (Hz)

Fig. 10. Frequencies of occurrence of horizontal displacements along the core of the rock-ll dam.

710

M. Akkose et al. / Applied Mathematics and Computation 184 (2007) 704714

4.27725 - 3.68775 = 0.5895 s

Displacement (m)

0.15
0.10
0.05

t2=4.27725

0.00

t1=3.68775

-0.05
-0.10
-0.15
3.6

3.7

3.8

3.9

4.0

4.1

4.2

4.3

Time (s)

Fig. 11. Determination of value of occurrence period of the maximum horizontal displacement of the dam crest.

frequency of occurrence along the line 1 vary between 1.67 and 1.74. It can be seen by comparison of Figs. 6
and 10 that the frequency of occurrence decrease with increasing displacement values.
Frequency of occurrence of the displacements obtained from the stochastic seismic analysis can be veried.
For this purpose, a time interval between 3.6 and 4.3 s of the time-history of the crest displacements in Fig. 9
where maximum horizontal displacement is occurred is zoomed as shown in Fig. 11. Since the period is dened
as the time required to complete one cycle, it is taken the time interval between t1 = 3.68775 and t2 = 4.27725

320

320
Deterministic
Stochastic

280
240
200

Height (m)

240
355.37kN/m2

160
120

200
160
120

80

80

40

40

0
0

1000

2000

3000

4000

1000

2000

3000

yy (kN/m2)

zz (kN/m2)

(a) Horizontal Stresses

(b) Vertical stresses

320
Deterministic
Stochastic

280
240

Height (m)

Height (m)

Deterministic
Stochastic

280

200
160
120
80
40
0
0

1000

2000

3000

4000

yz (kN/m2)
(c) Shear stresses
Fig. 12. Horizontal, vertical, and shear stresses on section II of the rock-ll dam.

4000

M. Akkose et al. / Applied Mathematics and Computation 184 (2007) 704714

711

shown in Fig. 11 in order to compute the period of the maximum horizontal displacement. Dierence between
t1 and t2 is 0.5895 s. So, frequency of the maximum horizontal displacement can be calculated as 1.70 Hz from
the deterministic seismic analysis. The value of frequency of occurrence of the horizontal displacement at the
crest point (node 1) from stochastic dynamic analysis is obtained as 1.67 Hz. This also shows good accuracy of
the results obtained from the stochastic dynamic analysis.
3.2. Stresses
The stress components, which are obtained from stochastic and deterministic dynamic analyses, are also
compared with each other. The stress values are calculated at the middle points of the elements with a time
interval of 0.0025 s. Sections II, IIII, and IIIIII shown in Fig. 3 are taken into account for comparison
of the stress components, which are horizontal, vertical and shear stresses. The mentioned stress components
on sections II, IIII, and IIIIII of the rock-ll dam are compared in Figs. 1214, respectively. It can be seen
from these gures that the expected maximum values of all stress components for all sections are smaller than
the absolute maximum stresses obtained from deterministic dynamic analysis. It can also seen from Fig. 12
that the stresses decrease towards the dam crest.
To verify the stresses obtained from the stochastic dynamic analysis, time-history of the horizontal stresses
at the element A selected from the dam body is calculated from the deterministic dynamic analysis as shown in
Fig. 15. Taking the average of 19 maximum horizontal stresses shown in Fig. 15, the mean of maximum horizontal stresses can be calculated as 350.37 kN/m2 for deterministic analysis. The expected maximum horizontal stress obtained from stochastic dynamic analysis is 355.37 k N/m2 as shown in Fig. 12(a). The maximum
stress obtained by averaging, which is 350.37 kN/m2, is very close to 355.37 kN/m2 obtained from the stochastic analysis.

yy (kN/m2)

6000
Deterministic
Stochastic

5000
4000
3000
2000
1000
0
0

50

100

150

200

250

300

350

400

450

500

350

400

450

500

350

400

450

500

Horizontal Distance (m)

(a) Horizontal stresses

zz (kN/m2)

6000
Deterministic
Stochastic

5000
4000
3000
2000
1000
0
0

50

100

150

200

250

300

Horizontal Distance (m)

(b) Vertical stresses

yz (kN/m2)

6000
Deterministic
Stochastic

5000
4000
3000
2000
1000
0
0

50

100

150

200

250

300

Horizontal Distance (m)

(c) Shear stresses

Fig. 13. Horizontal, vertical, and shear stresses on section IIII of the rock-ll dam.

712

M. Akkose et al. / Applied Mathematics and Computation 184 (2007) 704714

yy (kN/m2)

2500
Deterministic
Stochastic

2000
1500
1000
500
0
20

40

60

80

100

120

140

160

180

200

160

180

200

160

180

200

Horizontal Distance (m)

(a) Horizontal stresses

zz (kN/m2)

2500
Deterministic
Stochastic

2000
1500
1000
500
0
20

40

60

80

100

120

140

Horizontal Distance (m)

(b) Vertical stresses

yz (kN/m2)

2500
Deterministic
Stochastic

2000
1500
1000
500
0
20

40

60

80

100

120

140

Horizontal Distance (m)

(c) Shear stresses

Fig. 14. Horizontal, vertical, and shear stresses on section IIIIII of the rock-ll dam.

yy (kN/m2)

1000
500
0
-500

3.7s

4.2s

-1000
0.0

3.0

6.0

9.0

12.0

15.0

18.0

21.0

Time (s)

Fig. 15. Time-history of horizontal stress of the element A.

3.2.1. Frequencies of occurrence of stresses

Values of frequencies of occurrence of horizontal stresses at section II of the rock-ll dam are depicted in
Fig. 16. It can be seen from Fig. 16 that the value of frequency of occurrence of stochastic horizontal stress at
the element A (Fig. 3) is obtained as 3.31 Hz. To verify the frequency of occurrence of the horizontal stress
obtained from the stochastic dynamic analysis, a time interval between 3.7 s and 4.2 s of the time-history
of horizontal stresses in Fig. 15 where maximum value is occurred is zoomed as shown in Fig. 17. It is taken
the time interval between t1 = 3.8205 and t2 = 4.1130 shown in Fig. 17 in order to compute the period of the
maximum horizontal stress. Dierence between t1 and t2 is 0.2925 s. So, frequency of the maximum horizontal
stress can be calculated as 3.42 Hz. These also show good accuracy of the results obtained from the stochastic
dynamic analysis.

M. Akkose et al. / Applied Mathematics and Computation 184 (2007) 704714

713

320

2.64
280

3.29
1.82

240

2.83

Height (m)

2.37
200

2.92
2.66

160

3.31
3.31

120

3.02
2.91
3.05
3.24
2.97
3.15

80
40
0
1.6

2.0

2.4

2.8

3.2

3.6

4.0

Occurence Frequency (Hz)

Fig. 16. Frequencies of occurrence of horizontal stresses on the section II of the rock-ll dam.

4.1130 - 3.8205 = 0.2925 s

yy (kN/m2)

1000
500
t2 =4.1130

t1=3.8205

-500
-1000
3.7

3.8

3.9

4.0

4.1

4.2

Time (s)

Fig. 17. Determination of occurrence period of the maximum horizontal stress at the element A.

4. Conclusions
Stochastic seismic response of a rock-ll dam by nite element method is investigated in this paper. The
project values of the Keban dam were considered in the analyses. Frequencies of occurrence of expected maximum values of displacements and stresses are compatible with the results of the deterministic seismic analysis.
It is observed that the displacement and stress results obtained from the deterministic dynamic analysis are
greater than the mean of maximum values obtained from the stochastic dynamic analysis. Because the mean
of maximum values obtained from the stochastic analysis is calculated by averaging all the maximum response
values, it should be expected that the absolute maximum values obtained from the deterministic dynamic analysis would be greater than the mean of maximum values.
The horizontal displacements from both analyses increase along the dam height. The vertical displacements
obtained along the horizontal length of the dam signicantly decrease towards the core of the rock-ll dam for
both analyses. The frequency of occurrences of horizontal displacements at the clay core decrease with increasing the values of displacements. The all stress components decrease towards the dam crest. The shear stresses
obtained along the clay core are greater than the horizontal and vertical stresses.
To generalize these results, solutions must be obtained using many earthquake inputs and dierent rock-ll
dam models.
References
[1] Y.M. Ozkan, A review of considerations on seismic safety of embankments and earth and rock-ll dams, Soil dynamics and
Earthquake Engineering 17 (1998) 439458.
[2] G. Gazetas, P. Dakoulas, Seismic analysis and design of rockll dams: state-of-the-art, Soil dynamics and Earthquake Engineering 11
(1992) 2761.

714

M. Akkose et al. / Applied Mathematics and Computation 184 (2007) 704714

[3] H.B. Seed, Considerations in the earthquake design of earth and rockll dams, Geotechnique 29 (3) (1979) 215263.
[4] F.I. Maksidi, H.B. Seed, Simplied procedure for estimating dam and embankment deformations, Journal of Geotechnical
Engineering Division, ASCE 104 (GT7) (1978) 849867.
[5] H.S. Sayed, A.M. Abdel-Ghaar, System identication of nonlinear seismic response of earth dams, in: Tenth World Conference on
Earthquake Engineering, Rotterdam, 1992, pp. 47334739.
[6] A.R. Khoei, A.R. Azami, S.M. Haeri, Implementation of plasticity based models in dynamic analysis of earth and rockll dams: a
comparison of PosterZienkiewicz and cap models, Computers and Geotechnics 31 (2004) 385410.
[7] R. Mellah, G. Auvinet, F. Masrouri, Stochastic nite element method applied to non-linear analysis of embankments, Probabilistic
Engineering Mechanics 15 (2000) 251259.
[8] M.T. Chen, R.S. Harichandran, Response of an earth dam to spatially varying earthquake ground motion, Journal of Engineering
Mechanics 127 (9) (2001) 932939.
[9] Y.K. Lin, Probabilistic Theory of Structural Dynamics, rst ed., McGraw-Hill, New York, 1967.
[10] C.Y. Yang, Random Vibration of Structures, rst ed., Wiley, New York, 1986.
[11] G.D. Manolis, P.K. Koliopoulos, Stochastic Structural Dynamics in Earthquake Engineering, WIT Press, Southampton, 2001.
[12] M.R. Button, A. Der Kiureghian, E.L. Wilson, STOCAL-User Information Manual, Report No. UCB-SESM/81-2, Department of
Civil Engineering, University of California, Berkeley, CA, 1981.
[13] A.A. Dumanoglu, R.T. Severn, Stochastic response of suspension bridges to earthquake forces, Earthquake Engineering and
Structural Dynamics 19 (1990) 133152.
[14] K. Soyluk, A.A. Dumanoglu, Comparison of asynchronous and stochastic dynamic responses of a cable stayed bridge, Engineering
Structures 22 (2000) 435445.
[15] C.D. Wung, A. Der Kiureghian, STOCAL-II: Computer-assisted learning system for stochastic dynamic analysis of structures, Part I
theory and development, Report No. UCB/SEMM-89/10, Department of Civil Engineering, University of California, Berkeley, CA,
1989.
[16] A. Der Kiureghian, Probabilistic modal combination for earthquake loading, in: Proceeding of Seventh World Conference on
Earthquake Engineering, Istanbul, 1980, pp. 729736.
[17] R. Priscu, A. Popovici, D. Stematiu, C. Stere, Earthquake Engineering for Large Dams, second ed., Editura Academiei, Bucuresti,
1985.
[18] A. Bayraktar, K. Haciefendioglu, M. Muvak, Asynchronous seismic analysis of concrete-faced rockll dams including dam-reservoir
interaction, Canadian Journal of Civil Engineering 32 (2005) 18.
[19] A. Bayraktar, E. Hancer, A.A. Dumanoglu, Comparison of stochastic and deterministic dynamic responses of gravity dam-reservoir
systems using uid nite elements, Finite Elements in Analysis and Design 41 (2005) 13651376.
[20] S.M. Allam, T.K. Datta, Analysis of cable-stayed bridges under multi-component random ground motion by response spectrum
method, Engineering Structures 22 (2000) 13671377.
[21] R.S. Harichandran, W. Wang, Response of indeterminate two-span beam to spatially varying seismic excitation, Earthquake
Engineering and Structural Dynamics 19 (1990) 173187.
[22] A. Zerva, Seismic loads predicted by spatial variability models, Structural Safety 11 (1992) 227243.