Weirs play a crucial role as hydraulic structures in the regulation and control of water flow. This study investigates the relative energy dissipation in labyrinth weirs, examining various configurations, scales, and cycle types, using advanced computational models like Support Vector Machine (SVM), Random Forest (RF), and Artificial Neural Network (ANN). In SVM modelling, the results from different kernel functions reveal that the Radial Basis Function (RBF) kernel outperforms polynomial, linear, and sigmoid kernels in predicting relative energy dissipation. For the RBF kernel, the statistical metrics were found to be (R=0.907), (Mean RE%=1.38), (RMSE=0.0153), and (KGE=0.744) in test phase. where RE, Mean RE, RMSE and KGE represent the Relative Error, Mean Relative Error, Root Mean Square Error and Kling Gupta Efficiency, respectively. In contrast, in the ANN model, the multilayer perceptron (MLP) network showed higher accuracy than the RBF network, achieving 0.969, 0.73%, 0.007, and 0.968 for the same indicators. For the RF model, these values were recorded as 0.878, 1.78%, 0.0192, and 0.362, respectively. Comparative analysis indicates that the ANN model offers superior predictive performance over SVM and RF models. Additionally, non-linear polynomial regression equations, derived from dimensionless parameters, are proposed for estimating relative energy dissipation. Notably, single-cycle weirs exhibited the greatest energy dissipation among the configurations studied.
Abbaszadeh et al. (Wed,) studied this question.