Application of intelligence models based on soft computing in investigating the discharge coefficient of the sluice gate under free-flow condition and symmetrical sill with the help of KNN, ANN, GEP and SVM models

The use of sills with the gates leads to a reduction in the height of the gate. The sills affect the flow and change quantities, especially the discharge coefficient. In the present research, the discharge coefficient of the sluice gate is examined for the first presented theoretical relationship in a non-suppressed sill state to measure its performance using soft computing methods. For the models, 70% of the data were used for the training and the rest for the testing phases. The results of statistical indicators showed that in all SVM, KNN, GEP, and ANN models, the model with all input parameters was recognized as the superior model. In the SVM model, the results of various kernels showed that the Radial Basis Function kernel has better results in predicting the discharge coefficient compared to the Polynomial, Linear, and Sigmoid kernels. The results of the correlation coefficient (R), Root Mean Square Error (RMSE), mean percentage Relative Error (MRE%), and Kling Gupta Efficiency (KGE) in the test stage for the SVM model were 0.96, 0.018, 0.90, and 1.92%, respectively. The neighbor coefficient (K) results showed that in the K equal 2, the RMSE and MRE had a lower value and were close to the experimental results. In addition, in the KNN model, among distance criteria measures (Manhattan, Euclidean, Euclidean Squared, and Chebychev), the Manhattan criteria have a higher accuracy in predicting the discharge coefficient than the others. In the testing phase, this model's results were 0.97, 0.016, 0.96, and 1.70%. In addition, the results for the GEP model were 0.98, 0.019, 0.85, and 2.28%, respectively. In the present research, the ANN method is more accurate compared to SVM, GEP, and KNN models, so, for the ANN model, the KGE was in the very good range.