A comprehensive study of various regressions and deep learning approaches for the prediction of friction factor in mobile bed channels

Abstract

A fundamental issue in the hydraulics of movable bed channels is the measurement of friction factor (λ), which represents the head loss because of hydraulic resistance. The execution of experiments in the laboratory hinders the predictability of λ over a short period of time. The major challenges that arise with traditional forecasting approaches are due to their subjective nature and reliance on various assumptions. Therefore, advanced machine learning (ML) and artificial intelligence approaches can be utilized to overcome this tedious task. Here, eight different ML techniques have been employed to predict the λ using eight different input features. To compare the performance of models, various error metrics have been assessed and compared. The graphical inferences from heatmap data visualization, Taylor diagram, sensitivity analysis, and parametric analysis with different input scenarios (ISs) have been carried out. Based on the outcome of the study, it has been observed that K Star in the IS1 with correlation coefficient (R2) value equal to 0.9716 followed by M5 Prime (0.9712) and Random Forest (0.9603) in IS2 and IS4, respectively, have provided better results as compared to the other ML models to predict λ in terms of least errors. HIGHLIGHTS The study employs ML algorithms to accurately predict friction factor (λ) in movable bed channels, comparing eight ML techniques.; The K Star model in input scenario 1 achieves the highest correlation coefficient (R2) value of 0.9716 for predicting λ.; The research findings guide engineers in selecting appropriate input variables and ML models to predict λ accurately.;

Keywords

• flow resistance
• friction factor
• machine learning
• open channels
• prediction