Optimization of horizontal drain dimensions in heterogeneous earth dams using Artificial Neural Network (ANN): A case study on Marvak dam

Abstract

Designing and optimizing the dimensions of the drainage system in order to keep the crust dry and prevent the increase in pore water pressure in the body of the earth dams is very important. On the other hand, finding the minimum factor of safety (FoS) and, consequently, reducing the cost of construction, will be possible by optimizing the dimensions of the drain. In this paper, Geostudio software is used to simulate Marvak dam with real dimensions and parameters to calculate its factor of safety. For different values, the parameters affecting the dam factor of safety are calculated, and then the maximum effect of each of them on the dam factor of safety is determined in the case of an equal condition. Then, these factors are classified in terms of costs and impact on the dam factor of safety and each of them is used to optimize dam construction costs or to provide the required factor of safety of the dam. In the following, a two-layer neural network will be considered in the Matlab software. The neural network is then trained on the basis of the data obtained for the dam barrier to predict the dam factor of safety. Then the trained neural network function is extracted and for producing dam factor of safety in the optimization problem. The objective of the optimization is to minimize the size of the dam drain so that the dam construction cost is optimized and a minimum factor of safety of the dam is provided. To solve this problem, an optimization of Mfile is tailored to the neural network function, which is calculated by executing the optimal values of the dam parameters.