Application of artificial neural networks for predicting hydraulic jump characteristics on right triangular prism rough beds

Abstract

Hydraulic jump is a crucial hydraulic phenomenon in the design of energy dissipation structures. Until recently, studies on main hydraulic jump characteristics, such as the conjugate depth and the jump length, were based on theoretical and experimental studies. For hydraulic jumps with boundary conditions differing from traditional hydraulic jumps, the determination of these two characteristics has mainly relied on regression methods based on experimental results. Nowadays, the rapid advancement of computer science has introduced new approaches to solving regression problems, notably artificial neural networks, and other machine learning models. Among these, artificial neural networks stand out due to their superior effectiveness in classification and regression tasks. This study applies a three-layer artificial neural network, consisting of an input layer, a hidden layer, and an output layer, to predict two hydraulic jump characteristics on a rough bed with right-angled triangular prism elements, having the network structure only a single hidden layer, with the number of neurons varying from 3 to 8. The research results indicated that the neural network model with a single hidden layer containing 8 and 10 neurons provides highly accurate predictions for the subsequent depth and jump length, achieving high R² correlation coefficients of 0.992 and 0.912, respectively.