Keywords: optimization, genetic algorithm, neural network, double arch dam.

Optimizing large scale structures such as double arch dams is very important because of the high cost of construction. It is well known that the design of the shape of a double arch dam has great influence on its economy and safety [1]. Generally, double arch dams are designed by trial and error; that is, an initial scheme is selected and then it is analyzed. If the demands of the design specifications are satisfied, the scheme is adopted. Otherwise, the shape of the dam is modified and reanalyzed until the entire designer's demands are satisfied [2].

In the literature, for the optimum shape of a double arch dam, early research investigations have dealt mainly with the membrane-type solutions [3,4]. Shape optimization of arch dams has been developed in the past thirty-five years. The Titter method [5] and finite element method [6] have been used for stress analysis, and the sequential linear programming (SLP) method has been used to search for the optimum shape. Also, a geometrical model for the dam has been presented by Zhu [7]. This model is a continuous and practical model and therefore, it has been common among researchers.

For shape optimization of double arch dams, other methods have been also used such as linear programming, Lagrange multipliers method, etc. The genetic algorithm (GA) is the most common method among other methods in structural optimization. Since the GA considers the whole domain of the design variables during the optimization process, it is expected to find the global optimum instead of the local optimum. However the GA, takes a long time for optimization of large scale structures and this is the disadvantage of this method. Using intelligent methods such as neural networks instead of time consuming analysis can resolve this problem. Therefore, in this study, the GA has been used for the optimization of the concrete volume of double arch dams concerning the design constraints, and a radial basis neural network has been defined for calculating objective and penalty functions. Several programs have been developed in the MATLAB software environment for the optimization process and also the training of the neural network. The ANSYS parametric design language (APDL) has been used for modelling, analysis and calculation of the dam concrete volume. The above-mentioned programs have been applied to the Raeisali Delvari dam in Bushehr Province, Iran and the results have been compared with the characteristics of the existing dam.

This work indicates that the genetic algorithm gives better results because it considers the whole domain of the design variables. Combining an artificial neural network (ANN) instead of the time consuming analyses with the genetic algorithm increases the speed of the optimization process. The efficiency of the combined GA and ANN compared with the usable-feasible direction method and the quadratic approximation method has been considered in this research.

1

B. Zhu, B. Rao, J. Jia and Y. Li, "Shape Optimization of Arch Dam for Static and Dynamic Loads", Journal of Structural Engineering, 118(11), 2996-3015, 1992. doi:10.1061/(ASCE)0733-9445(1992)118:11(2996)

2

B. Zhu, "Optimum Design of Arc Dams", Dam engineering, 1(2), 131-145, 1990.

3

J.F.L. Fialho, "Leading Principles for the Design of Arch Dams-New Method of Tracing and Dimensioning", LNEC, Lisbon, Portugal, 1955.

4

J.L. Serafim, "New Shape for Arc Dams", Civ. Engrs., ASCE, 36(2), 1966.

5

R. Sharpe, "Optimum Design of Arc Dams", Proceeding., Institute of Civil Engineers, Supplementary Vol., 73-98, 1969.

6

R.E. Ricketts and O.C. Zienkiewicz, "Optimization of concrete Dams", Proc. Int. Symp. On Numerical Analysis of Dams, London, England, 1975.

7

B. Zhu, "Shape Optimization of Arc Dams", Water Power and Dam Construction, March, 1987.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £105 +P&P)