Prediction of Cohesive Sediment Erosion Rate and Analyzing the Effective Parameters Using Artificial Neural Network

Transferring mechanic of cohesive sediments are different from non-cohesive sediments. For determining the erosion rate of non-cohesive sediments, physical parameters such as average diameter and density are used, such as average diameter and density. Due to the nature of the cohesive sediments, their erosion rates are determined interrelated with the shear stress of the bed with fixed coefficients related to the characteristics of each sediment. In this study, experimental results on the cohesive sediments of the Loire estuary of France has been used. After validating the results in Mike software, experimental data were developed to study the erosion of sediment with more data and different hydraulic conditions. In the following, due to the number of various parameters affecting the sediment erosion phenomenon, a neural network was used to analyze the data. The parameters used in the model include flow components, sediment and fluid characteristics. Due to the better performance of the neural network, these data were used for dimensionless data. The correlation coefficient and mean absolute error of data in the neural network were 0.98 and 0.0036, respectively, which indicated the proper performance of the network. Finally, after performing the sensitivity analysis, the and  parameters were introduced as the most effective parameters for increasing and decreasing erosion rates, respectively.