Application of Soft Computing Methods in the Analysis of Velocity Field in Dividing Channel

The simplest water diversion method in irrigation systems is using intakes. Measuring the mean velocity is one of the essential hydraulic parameters in increasing the efficiency of the intake. In this study, the mean velocity was predicted for different width ratios of an intake using ANN-MLP neural network model. In order to do that, the flow field within a 90-degree intake was first simulated three-dimensionally using ANSYS-CFX. The neural network used in this study includes 3 inputs; longitudinal coordinate (Y*), ratio of the branch channel to the main channel (wr), and mean velocity of the middle line of the channel cross section (V*line). V*lineis the average velocity in the vertical column of the branch channel, which has been measured by the ANSYS-CFX model after the validation. Comparison of the ANSYS-CFX results with the experimental ones indicated that this model, with mean Root Mean Squared Error (RMSE) of 0.013, has a proper accuracy in simulating the characteristics of the flow field within the intake. In addition, comparison of the obtained results from ANN-MLP model and the experimental results indicated that this model, with mean determination coefficient (R2) of 0.948, has a high performance in predicting the mean velocity of open channel intakes.