Investigating Discharge Coefficient of Slide Gate-Sill Combination Using Expert Soft Computing Models

The use of gate-sill combinations in recent years has been one of the new methods in increasing the hydraulic performance of gates, including the discharge coefficient (Cd). The present research aims to investigate the Cd of the gate with a sill in different dimensions in width and various positions relative to the gate using support vector machine (SVM) models, the K nearest neighbor (KNN) algorithm, and the artificial neural network (ANN) method using Statistica software. Out of 345 experimental data, 70% (241) were used for training and 30% (104) for testing. The best results are obtained when all dimensionless parameters (Atotal/B2, H0/B, Z/B, ε/B, and X/B) are used. The results of different kernels showed that RBF kernel has better results in predicting Cd compared to Polynomial, Linear, and Sigmoid kernels. The results of the statistical indexes of R, KGE, RMSE, and Mean RE% for the RBF kernel in the test phase are 0.955, 0.90, 0.0192 and 1.82%, respectively. In the KNN model, Manhattan distance measure has favorable results compared to other Euclidean, Euclidean Squared, and Chebychev criteria. The results showed that the ANN method has the best performance compared to SVM and KNN models with values of 0.984, 0.976, 0.0098, and 1.15%, respectively