Predicting the side weir discharge coefficient using the optimized neural network by genetic algorithm

Side weir is one of the structures which are widely used in water engineering projects So study on the flow characteristics especially discharge coefficient (Cdsw ) of this type of weir is important. Several Empirical formulas proposed to calculate the Cdsw that they usually associated with significant errors. Thus, using mathematical methods based on artificial intelligence is inevitable. Artificial neural network (ANN) is a very useful data-modeling tool that is able to capture and represent complex input and outputs relationships. In this study, accuracy the some famous empirical formula was assessed and Borghei and Parvaneh was most accuracy formula with (R = 0.83). To increase the accuracy of production the Cdswmulti-layer neural network (MLP) has developed. This model, in the first, trained by a back propagation method such as levenberg_marquardt and its performance was determined. the error indexes in the train and testing process are equal to (Rtrain 2 = 0.93 and Rtest 2 = 0.96). To increase the accuracy of MLP model, instead of increasing the size of the network (increase in the number of neurons and layers), Genetic Algorithm was used to train this model (GANN)and obtained an optimized value for Bias and weights. The final results show that using the GA to training the MLP model cased to increase in accuracy .The performance of the optimized neural network (GANN) in training and testing is equal to (Rtrain 2 = 0.98 and Rtest 2 = 0.97).