Evaluation of suspended sediment concentration using descent neural networks

Abstract

This study includes three methods having considerable differences with each other and with experimental observations, because the sediment measures have certain limits. The equations relating to sediment transport are used in estimating sediment load. In the present study, black box models, ANN (Artificial Neural Network) are used for the simulation of the suspended sediment load. Hence, models which give the lowest RMSE and highest R2 are considered to be the best model for this study. The lowest values of RMSE based on normalized data for Feed forward back propagation, Cascade forward back propagation and neural network fitting are 0.00873, 0.00834 and 0.01193 respectively. The corresponding values of R2 are 0.9304, 0.9713 and 0.9831 respectively for the cited MSE. The study shows Neural Network Fitting model is superior to the other models. However, a drawback of neural network fitting is that it produces few negative estimates, which is not at all tolerable in the field of estimation of sediment load and hence these models are not crowned in this study.