Prediction of jump length, bed shear stress, and energy loss using machine learning models for a steady, free hydraulic jump on rough beds

Abstract

Determining or predicting the characteristics of hydraulic jumps over rough beds remains a complex issue in hydraulic engineering. While recent experimental and numerical simulation studies have enhanced understanding of this phenomenon, an efficient and accurate predictive approach is still required for the optimal design of stilling basins under complex flood flow conditions. This study proposes an advanced method for predicting the main hydraulic parameters of free and stable hydraulic jumps on right triangular prism rough beds, including jump length (L_j/y₁), bed shear stress (e), and energy loss (ΔE_L/E₁). Machine learning techniques, namely Decision Tree Regression (Fine Tree) and Support Vector Regression (SVR), were applied to model the nonlinear relationships between rough bed geometry and supercritical flow behaviour. The study results showed the models achieved high predictive accuracy (R² = 0,91 ¸ 0,99) with low RMSE error values, confirming their strong capability in predicting the complex hydraulic phenomena. The findings demonstrated that artificial intelligence (AI) can serve as a reliable and cost-effective alternative to physical model experiments and numerical simulations, offering a robust scientific basis for analysing and optimising the hydraulic performance and energy dissipation efficiency of hydraulic jumps over rough beds under complex hydraulic conditions.