The process of extracting water from rivers by means of catchment structure is always associated with the deviation of sediments. The diversion of sediments in the form of bed load to the catchment channel, the water transfer facilities downstream of the catchment and the water supply plan face problems. The use of sediment control structures separating the wall in front of the catchment and at the same time the breakwater in front of the catchment reduces the incoming sediment and increases the catchment efficiency. In the current research, the effect of the separating wall and the breakwater structure in estimating the ratio of incoming sediment to the catchment has been evaluated by laboratory and data-mining and multiple regression methods. First, by performing dimensional analysis, the dimensionless ratios were extracted and the relationship between the variables and their value in the experiments was determined. Using XLSTAT and SPSS statistical software, equations for the relationship between independent and dependent variables were extracted from the step-by-step and inter method. After obtaining the equations, the relative error of each equation was calculated. Then the best equation with high R2 and low relative error was selected and proposed. In the next step, modeling was done with the methods of artificial neural networks (ANN), GEP and GMDH, and the best method was chosen to estimate the ratio of input sediment to the catchment. The results showed that the best performance of the ANN model with statistical indices, R2=0.99, MAD=0.004, RMSE=0.003 and MAPE=3.28, and GEP model in the next step, is the best performance in estimating the ratio of input sediment to the catchment. Also, data mining methods are more accurate than regression methods.