machine learning models for high-accuracy prediction of energy dissipation through gabion sills downstream of spillways

Objective: the objective of this paper is to develop and compare three mlms, svr, gep and ann- for the high-accuracy prediction of energy dissipation downstream of gabion sills in spillways. through dimensional analysis and sensitivity evaluation using the γ-test, the most influential hydraulic and geometric parameters are identified. the performance of each model is rigorously assessed using statistical metrics to determine their predictive reliability and accuracy, with the aim of identifying the most effective computational approach for optimizing energy dissipation in gabion-structured spillway systems. material and methods: experimental data from a lab flume was used. dimensional analysis identified key parameters affecting energy dissipation. sensitivity analysis via the gamma test selected the most influential inputs. these were used to train and compare three machine learning models: svr, gep, and ann.results and discussion: the gep model demonstrated superior performance, achieving the highest r² (0.936) and lowest errors (rmse=0.003) in predicting energy dissipation. it outperformed both ann and svr. sensitivity analysis identified four key hydraulic parameters as the most influential inputs. conclusions: the study conclusively found the gene expression programming (gep) model to be the most accurate and reliable for predicting energy dissipation over gabion sills, significantly outperforming ann and svr. the four key hydraulic parameters identified were crucial for model success, demonstrating the effectiveness of this machine learning approach.