Evaluation of utilizing neuro-fuzzy inference system and artificial neural network for estimation of suspended sediment load Sefidrud River

In the present research a comparison of estimation has been made for suspended sediment load in storage dams using methods of Artificial Neural Network (ANN) and Artificial Neural Network Fuzzy Inference System (ANFIS). The daily average discharges of Sefidroud River (at north of Iran) were used as the model inputs and its sediments concentration in time series as the outputs by utilizing MATLAB program. Out of 229 available data, 182 were applied as training data and the rest (47 data) as testing data. The inputs and outputs of sediment concentration had positive procedures. Eighty percent (80%) of the available data have been used for the training while the remaining used for testing network. The results showed that the concentration prediction of the suspended sediment loads obtained from the ANFIS were closer to the available sediment concentrations. The regression coefficient was ninety percent (90%) for the ANFIS while that for the ANN was eighty three percent (83%). Therefore, it could be concluded that as for concentration prediction of the suspended sediment loads the ANFIS is more efficient than the ANN.