jakbari@modares.ac.ir
**. Professor of Civil Eng. , Tarbiat Modares University, Tehran, Iran
mahmadi@modares.ac.ir
***. Assistant Professor of Civil Eng , Tarbiat Modares University , Tehran, Iran
hamid@modares.ac.ir
Keywords: Concrete Arch Dam, Shape Optimization, Cubic Spline, Design
ABSTRACT
This paper presents an applicable and practical computer program(ODPSAD1) and suggests a new
algorithm for geometry of concrete arch dams. In this program shape optimization of arch dams for
static and dynamic loads and enhancement of its seismic behaviour is carried out. abutment
excavation and dam body volumes have been considered as the objective function and in order to
increase the consistency with practical conditions, a great number of geometrical and behavioural
constraints have been included in the mathematical model. Three loading cases i.e., gravity loading,
static loading and dynamic loading have been accounted for analysis. In seismic loading, longitudinal
component of the OBE2 spectrum have been considered. For finite element analysis the base of
program FEAPpv3 is used and extended for this purpose .Sequential nonlinear approximation(SQP4)
procedure is used for the optimization analysis. Results show that the modification and optimization
of shape can be very versatile and professional leading to considerable reduction in the dam body
volume and costs in regular cases.
1. Introduction
In concrete arch dams, geometry of the structure has a great influence on the safety and economy of
design. Geometry is also an important factor in the stability of dam. Nowadays, a high number of arch
dams are constructed in seismic zones and despite their acceptable performance under dynamic
loading, there is still overmuch need to improve their behaviour and to increase their safety during
strong ground motion events. Generally, shape design of an arch dam is based mainly on the
experience of the designer, model tests and trial and error procedures. In trial and error procedures, an
initial scheme is selected or given and then analyzed. If it satisfies the demands of the design
specifications, the scheme is adopted. Otherwise, the shape of the dam is modified and reanalyzed.
This process continues until the whole demands of the design specifications are satisfied. To get a
better shape, the designer should select several alternative schemes with various patterns and modify
them such as the first scheme to obtain a number of feasible shapes. The best shape considering
economy of design, structural considerations, safety, etc. is selected as the final shape. The shape of
the dam obtained in this way is feasible but not necessarily optimal or even good. Moreover, the time
of design is rather long. To overcome these difficulties, special efforts have begun by researchers for
optimal shape design from the late 1960s.In the search of arch dams with optimal shape, early
1
research works (Fialho (1955) & Serafim (1966)) [1,2] dealt mainly with membranetype solutions
that ignored foundation elasticity and bending stresses, and considered a single, simple loading
condition (Hydrostatic pressure and the weight of concrete). These methods can provide only useful
starting points for more comprehensive studies. Later, Rajan, M.K.S. (1968) [3], Mohr, G.A. (1979)
[4] and Sharma, R.L. (1983) [5], considerably developed membrane shell theory based solutions.
Sharpe, R. (1969) [6] was the first one to formulate the optimization problem as a mathematical
programming problem. A similar method was also adopted by Rickeetts, R.E. & Zienkiewicz, O.C.
(1975) [7] who used finite element method for stress analysis and Sequential Linear Programming
(SLP) for the shape optimization of arch and buttress dams under static loads. Wassermann (198384)
[8,9] used a mathematical formulation for arch dam design, in which the arch dam and part of its
foundation were described by threedimensional 8node isoparametric hyperelements. Design
variables were the geometrical parameters of these elements and the objective function was the dam
volume. The resulting optimization problem was solved by SLP method. From other works in this
field, the works done by Rahim, A.S. (1983) [10], Samy, M.S. (1987) [11,12], Yao, T.M. & Choi,
K.K. (1989) [13], Guohua, L. & Shuyu, W. (1990) [14] can be mentioned. Fanelli, A., Fanelli, M. &
Salvaneschi, P. (1993) [15] used neural network approach to the definition of near optimal arch dam
shape. Nevertheless, one of the most important investigations in the field of shape optimization of
concrete arch dams carried out by Bofang, Z. et al. [16,17,18,19,20] in China since the middle 1970s.
The geometrical model used by Bofang, Z. is a continuous, rational and practical model and thus
could draw the interests of many researchers. It has been the basis of majority of later works in the
field of shape optimization of arch dams. From among the works done up to early 1990s in the field
of shape optimization of concrete arch dams, the effects of seismic dynamic loads in determination of
the optimum shape of arch dams has rarely been considered. Bofang, Z. et al., considered dynamic
loading in trial load method to optimize the shape of arch dams. Later on, Simes, L.M.C. (1994)
[21,22] continued his works using the finite element method for stress analysis. He accounted the
seismic dynamic loads considering damfoundation interaction and the reservoir effects. In recent
years, particular attention is paid to this subject and specially in China, a number of articles in this
field has been published in Chinese language. Last work is done by Tajalli & Ahmadi(2005) [23]
using Bofang formulation for parabolic arch dam. The finite element analysis is done by a
commercial program and optimization procedure is carried out with DOT software. In the latter
seismic loads , damfoundation reservoir interaction have been considered and method is applied
successfully in Shahid Rajaee Arch dam in Iran.
In this research we use new algorithm for modelling geometry of arch dams [24].As a high number of
design variables and large dimensions of shape optimization of concrete arch dams problem, make its
nature very complicated, preparing a suitable method for this problem is highly demanding. This
research employs Bofangs dam body geometrical expression but is quite improved in modelling the
DamFoundation (DF) system, dynamic load definition, stress constraints, numerical analysis method
and method of optimization. The main aim has been an attempt to develop the practical and
professional program for shape optimization of arch dams[25].
2. Dam Body Geometry
2.1.Coordinate Systems
Site geometry is described in the cartesian site coordinate system t0 , t1 , t 2 .This global coordinate
system is chosen at the beginning of the project and remains unchanged throughout the design
procedure. The axes t0 and t1 lie in a horizontal plane. The axis t0 points to the right side of the dam
body. The axis t2 points vertically upward. The geometry of the arch dam is described in the cartesian
dam coordinate system x0 , x1 , x2 , c whose origin and orientation change for each design with the
location of the dam and the orientation of the axis of the dam. Fig (1).
(1)
The thickness variation in the crown is specified by Tb , Tm , Tu , Tc design variables. The upper and
middle levels are specified relative to H c with the level ratios hu = .75, hm = .40 . The thickness is
interpolated with Splines over the height of the dam. The upstream face of the dam is divided into a
central zone of constant thickness, which will be denoted by Z C , and zones of variable thickness at
the left and right abutments, which will be denoted by Z L and Z R . The zones are shown in Fig.(3).
In zone Z C the thickness of the dam body in each elevation equals the thickness in the crown joint. In
the zones Z L and Z R , the thickness in each elevation varies parabolically from the thickness of Z C
to the abutment thickness. The start points of variable zone in 4 stations on the right and left are
defined with 8 design variables .i.e. tb , tm , tu , tc , sb , sm , su , sc .The thickness of the dam body at the
abutment is specified by the thicknesses Trb , Trm , Tru , Trc and Tlb , Tlm , Tlu , Tlc , which are design variables.
The thickness is interpolated with Splines over the height of the dam.
Consider the elevation of a dam body at level x2 as shown in Fig.(4). The upstream edge of the
elevation is a parabola whose vertex lies on the upstream edge of the crest crown cantilever profile.
The radius of curvature r of the parabola is specified at levels H b , H m , H u , H c by the Rb , Rm , Ru , Rc
design variables. r is interpolated between these levels with the following Spline
r = Spline( x2 , H i , Ri )
(2)
The equation of the parabola AB with vertex ( a0 , a1 ) and radius r is :
X 1 = a1 + ( x0 a0 ) 2 / 2r
(3)
With the above defined design variables a 3D shape of system is created .Typical shape of the system
is presented in Fig.(5).
Three basic loading cases have been considered in the process of shape optimization. Gravity load
(case 1) is applied to free cantilevers (even and odd cantilevers) once. This load is applied for three
stages grouting . in Dominant static loading (case 2) monolithic shell structure of dam body is
subjected to the action of hydrostatic pressure (with full reservoir) plus load case 1 as described
before.In dynamic seismic loading (case 3) loading case 2 is superposed by the effects of longitudinal
component of the Operating Basis Earthquake defined as a target spectrum.
4. Properties of the Mathematical Model
The mathematical model consists of damfoundation for loading case 1. In loading case 2, the
hydrostatic pressure is added to the damfoundation model. In loading case 3, the mathematical model
consists of damfoundation and reservoir. The foundation is weightless and massless in static and
dynamic loading cases respectively. The dam body is modelled by four layers of 8node brick
elements. The foundation have 4 layers of 8node brick elements that extends about 1.5 times the dam
body height into the rock mass. The governing equations are linear and the displacements are small.
Also initial condition for the dynamic loading is the dominant static response (due to loading case 2).
5. Objective Function
The objective function is the volume of the dam and foundation excavation, which expressed as
follows:
f ( x) = V1 ( x) + V2 ( x)
(4)
In which f (x) = the sum of volumes, V1 ( x) = volume of concrete in the dam body, V2 ( x) = volume of
foundation excavation and the x is the vector of design variables.
6. Constraints
In shape optimization of concrete arch dams, the three following types of constraints should satisfy
the demands of design and construction requirements:
1)Geometrical constraints 2) Stress constraints 3) Stability constraints
The constraints are shown by a set of g i ( x) 0 conditions
1.6. Geometrical Constraints
To facilitate construction, the maximum slope of overhang at the upstream and at the downstream
U .S .
D. S .
faces ( S max
, S max
) and the maximum overhang at the upstream face relative to the toe (Undercut)
should be limited to equations (5),(6) Fig.(7). For constructing facilities and having smooth
cantilevers over the height the start points of the variable thickness zone must satisfy the equations
(8),(9).Fig.(6)
(5)
D .S .
D.S .
S max
S alw
Undercut Undercut alw
cr , cl alw
br , bl alw
(6)
(7)
(8)
(9)
,
, Undercutalw and alw are the allowable absolute values of the aforementioned
parameters. According to the requirements of traffic and gaining access to the equipments located at
different points of the dam and so on, the minimum thickness of the crest is then decided. Sometimes
the maximum thickness of the base of the dam is also limited.
U .S
where S alw
D.S .
S alw
The main stress constraints for both faces of arch dam body, which is primarily subjected to a biaxial
stress state, result from the condition that the principal stresses are within the biaxial strength
envelope for unreinforced mass concrete. It could be constructed considering appropriate safety
factors. In general, the tensile stresses in either pure tension or combined tension/compression states
are important and control the design. In the static and dynamic loading conditions, stress constraints
are expressed by biaxial stress envelope [26]. Modified envelope curve for biaxial dynamic strength
of concrete is expressed as Shown in Fig.(8):In static and dynamic analysis a safety factor concept is
adopted. at any arbitrary stress point B(r , ) is defined as follows:
R
(10)
r
The latter value should not be less than the defined limit at any arbitrary stress point under the total
(static + dynamic) principal stresses.
S .F . =
Constraints ensuring the sliding stability of the dam may be expressed by at least one of the following
three conditions:
a) Constraint of coefficient of sliding stability (adequately defined for large dams abutments):
K i ( K i ) min
(11)
where K i is the coefficient of sliding stability at point i and ( K i ) min is the allowable minimum value
of Ki .
b) Constraint of the angle between the arch thrust and sound rock contour lines at each elevation of
the abutments (usually for small dams):
min
(12)
where is the angle between the arch thrust direction and the sound rock contour lines at each
elevation of the abutments and min is the allowable minimum value of .
The shape optimization problem of an arch dam can be expressed as a nonlinear constrained
optimization problem of the form:
Minimize : C = f(X)
Subject to : g j (X) 0 ,
(14)
(j = 1,2,..., q)
In which q is the total number of constraints .Because there is not any explicit formulation between
design variables and the above functions , these functions are approximated in x0 via Taylor
expansion series, so the nonlinear functions replaced with sequential approximated quadratic and
linear functions as equations (15),(16)
f ( x) = f ( x0 ) + ( xi x0i ) * f + .5 * ( xi x0i )T *H df *( xi x0i )
(15)
g j ( x) = g j ( x0 ) + ( xi x0i ) * g j ( x0i )
(16)
x i = x i / x 0i
(17)
f ( x) = f ( x0 ) + ( xi 1) * f + .5 * ( xi 1)T *H df *( xi 1)
(18)
g j ( x) = g j ( x0 ) + ( xi 1) * g j ( x0i )
(19)
8.Case Study
The proposed method is applied to shape optimization of the 130 mhigh concrete arch dam. The
canyon has trapezoidal shape with 100 m base and slopes 1:1 at left and right. The main
characteristics of the dam and the considerations related to its shape optimization process are
described at tables (1),(2),(3).Also seismic input load is illustrated in Fig(9).
Table.1. Characteristics of the Dam
Height(m)
f c (Mpa)
Ec (Gpa)
Er (Gpa)
130
35
22
C (kN/m3) C
24
.18 .22
D.S .
S alw
(degree)
20
Undercutalw (m)
alw (degree)
SFst
SFdy
6.5
70
pu
pm
pb
Tc
Tu
Tm
Tb
Rc
Ru
Rm
Rb
Trc
Tru
Trm
0.75
0.65
0.55
8.00
14.0
19.0
28.0
205
145
110
85
8.80
15.4
20.9
Trb
Tlc
Tlu
Tlm
Tlb
tc
tu
tm
tb
sc
su
sm
sb
30.8
8.80
15.4
20.9
30.8
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
Response Specturm
8.00
7.00
Sa(m/s2)
6.00
5.00
4.00
3.00
2.00
y = 2E05x 5  0.0015x 4 + 0.0457x 3  0.6588x 2 + 3.9044x  0.9004
1.00

5.0000
10.0000
15.0000
20.0000
25.0000
30.0000
W(rad/sec)
The optimization process of aforementioned arch dam according to the above formulation converged
after 12 iterations. Convergence rate . Notwithstanding the value of the objective function at the end
of the 11th optimization cycle is minimum, but due to a partial violation of stress constraints in this
cycle, results obtained from the 12th iteration is considered as the valid optimum solution of the
objective function Fig.(10). After performing the optimization process, dam volume has decreased by
37% in comparison with the initial design.
600000
550000
500000
450000
400000
350000
10
11
Iterations
12
13
In this paper a program(ODPSAD) is used to shape optimization of an arch dam . it suggests a new
algorithm for geometry of concrete arch dams. Abutment excavation and dam body volumes have
been considered as the objective function and in order to increase the consistency with practical
conditions, a great number of geometrical and behavioural constraints have been included in the
mathematical model. Three loading cases i.e., gravity loading, static loading and dynamic loading
have been accounted . Results show that the optimization of shape can reduce the dam body volume
37 %.
Acknowledgment
The authors would like to express their gratitude to Professor P.J. Pahl, professor emeritus of TUBerlin for his ideas and encouragements.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
Fialho, J.F.L. 1955. Leading Principles for the Design of Arch Dams A New Method of
Tracing and Dimensioning, LNEC, Lisbon, Portugal
Serafim, J.L. 1966. New Shapes for Arch Dams, Civil Engineering, ASCE, Vol. 36, No.2
Rajan, M.K.S. 1968. Shell Theory Approach for Optimization of Arch Dam Shapes,
Ph.D Thesis, University of California, Berkeley
Mohr, G.A. 1979. Design of Shell Shape Using Finite Elements, Computers &
Structures, Vol.10, No.5, pp.745749
Sharma, R.L. 1983. Optimal Configuration of Arch Dams, Ph.D Thesis, Indian Institute
of Technology, Kanpur
Sharpe, R. 1969. The Optimum Design of Arch Dams, Proc. Institution of Civil
Engineers (ICE), Paper No.7200s, Suppl Vol., pp.7398
Ricketts, R.E. & Zienkiewicz, O.C. 1975. Shape Optimisation of Concrete Dams,
Criteria and Assumptions for Numerical Analysis of Dams, Quadrant Press, Swansea,
London, UK, 1975, pp.11791206
Wassermann, K. 1983. Shape Optimization of Arch Dams Using Finite Element
Techniques, Journal of Structural Mechanics, Vol.9, pp.9397
Wassermann, K. 198384. Three Dimensional Shape Optimization of Arch Dams With
Prescribed Shape Functions, Journal of Structural Mechanics, Vol. 11, No.4, pp.465489
10. Rahim, A.S. 1983. Optimum Shape of An Arch Dam for Static Loads, MEng Thesis,
Asian Institute of Technology, Bangkok
11. Samy, M.S. 1987. Investigation of Optimum Shape of An Arch Dam for Static And
Dynamic Actions, MEng Thesis, Asian Institute of Technology, Bangkok
12. Samy, M.S. & Wieland, M. 1987. Shape Optimization of Arch Dams for Static and
Dynamic Loads. Proc. International Workshop on Arch Dams, Coimbra, April 59, 1987
13. Yao, T.M. & Choi, K.K. 1989. Shape Optimal Design of An Arch Dam, ASCE Journal
of Structural Engineering, Vol. 115, No.9, pp.24012405
14. Guohua, L. & Shuyu, W. 1990. Optimum Design of Concrete Arch Dam, Proc.
International Concrete Conference, Tehran, 1990, pp.444452
15. Fanelli, A., Fanelli, M., & Salvaneschi, P. 1993. A Neural Network Approach to the
Definition of Near Optimal Arch Dam Shape, Dam Engineering, Vol. IV, Issue 2,
pp.123140
16. Bofang, Z. 1987. Shape Optimization of Arch Dams, International Water Power & Dam
Construction, Vol. 39, No.3, pp.43 51
17. Bofang, Z. 1990. Optimum Design of Arch Dams, Dam Engineering, Vol. I, Issue 2,
pp.131145
18. Bofang, Z. 1991. Internal Force Expansion Method for Stress Reanalysis in Structural
Optimization, Communications in Applied Numerical Methods, Vol. 7, No.4, pp.295298
19. Bofang, Z., Rao, B., Jia, J. & Li, Y. 1992. Shape Optimization of Arch Dams for Static
and Dynamic Loads, ASCE Journal of Structural Engineering, Vol. 118, No.11, pp.29963015
20. Bofang, Z., Jia, J., Li, Y., Xu, S. & Wang, Z. 1993. Intelligent Optimal CAD (IOCAD)
for Arch Dams, Computing in Civil and Building Engineering, pp.10651072
21. Simes, L.M.C. & Lapa, J.A.M. 1994. Optimal Shape of Dams Subject to Earthquakes,
Proc. International Conference on Computational Structures Technology, Advances in
Structural Optimization, pp.119130
22. Simes, L.M.C. 1995. Shape Optimization of Dams for Static & Dynamic Loading, Proc.
International Course on Hydroelectric Power Plants, July, 1995, Coimbra, Portugal,
pp.171190
23. Tajalli,F & Ahmadi.M.T &Moharrami.H,(2005). Shape Optimization of Concrete Arch
Dam to Enhance Seismic Performance,M.S.c thesis of Civil .Eng. Tarbiat Modares
University, Tehran
24. Ahmadi, M.T. & Pahl, P.J. 2003. A Research Proposal: Design Optimization of Concrete
Arch Dam for Enhanced Seismic Response
25. Akbari.J.& Ahmadi.M.T.(ICCE 2006), A Fortran Objectoriented Program for Design of
Concrete Arch Dams,7th International Conference on Civil Engineering,Tehran ,Tarbiat
Modares University.
26. Kupfer, H.B. & Gerstle, K.H., 1973. Behavior of Concrete Under Biaxial Stresses,
ASCE Journal Of Engineering Mechanics Divisions, Vol. 99, pp.852866