Determining the Discharge Coefficient of Lateral Overflows Using Neural Network

Side weirs are flow diverting structures that are widely used in irrigation and drainage industry, flood control, sanitary engineering and urban sewage systems. So far, the discharge coefficient in this type of overflows has been determined experimentally using regression techniques, and for this reason, in this research, linear regression models and artificial neural network were used to predict the discharge coefficient of the lateral overflow, and their results are in agreement with each other and with Computational values were compared and the best model in this field was selected for prediction. The discharge coefficient of lateral spillways is based on several combinations of dimensionless independent variables including the ratio of length to width of the spillway (L/B), the ratio of the flow depth downstream of the spillway to the height of the spillway (Hd/P), the ratio of the length of the spillway to the height of the spillway (L/ P), the ratio of weir discharge flow to upstream flow (Qw/Qu) and the ratio of flow depth upstream of weir to weir width (Hu/B) were predicted. ANN7 model with inputs L/P, Qw/Qu, hu/B has the highest value of regression coefficient equal to 0.92 and RMSE and MAE error values equal to 0.23 and 0.15 performed the best prediction. Reg1 model with the values of regression coefficient, RMSE and MAE equal to 0.72, 0.17 and 0.11, respectively, it created the best results in prediction and was selected as the best linear regression model. The general results showed that ANN models compared to linear regression produces better results.