Numerical Model and Computational Intelligence Approaches for Estimating Flow through Rockfill Dam

Abstract

A flood is a common natural disaster that causes enormous economic, social, and human losses. Over the years, a number of management approaches have been developed for lowering flood damage. A rock-fill dam is a suitable structure made of rocks for lowering the output hydrograph and controlling floods in watershed management. On the basis of experimental data, numerical method, artificial neural network (ANN), and neural network-genetic algorithm (NNGA) approaches were applied for predicting flow through trapezoidal and rectangular rock-fill dams. Input parameters for this prediction were selected on the basis of sensitivity analysis. According to the results of the sensitivity analysis, the heights of water in the upstream and downstream sides of the dams were considered as the inputs of the models. The results indicated that the application of a genetic algorithm for optimization of ANN parameters improved the flow estimates. The Delta-Bar-Delta algorithm presented a better performance compared with the other learning algorithms for ANN models. Meanwhile, the NNGA models trained with the Momentum learning algorithm gave the best flow estimates. In general, the used approaches performed well in estimating flow through rock-fill dam; however, the numerical method showed superiority over the other methods.