Mean bed shear stress estimation in a rough rectangular channel using a hybrid genetic algorithm based on an artificial neural network and genetic programming

Abstract

The determination of erosion and deposition patterns in channels requires detailed knowledge and estimation of the bed shear stress. In this investigation, the application of a Genetic Algorithm-based Artificial (GAA) neural network and Genetic Programming (GP) was presented to predict bed shear stress in a rectangular channel with rough boundaries. Several input combinations, fitness functions, and transfer functions were investigated to determine the best GAA model. Also, the effect of various GP operators on estimating bed shear stress was studied. A comparison between the GAA and GP techniques' abilities to predict bed shear stress was made and then investigated. The results revealed that the GAA model performs better in predicting the bed shear stress (RMSE = 0.0774), as compared to the GP model (RMSE = 0.0835).