Modeling and Determination of Effective Parameters in Flow Roughness Coefficient in Alluvial Channels with Dun Bedforms Using Support Vector Regression

Determination of flow roughness coefficient in the channels and river hydraulics is necessary for calculating of the discharge, the velocity and the depth of flow. Calculating the exact values of this coefficient is complex and difficult due to the influence of various parameters on it. In this study, using support vector regression as one of the machine learning approaches, the flow roughness coefficient in alluvial channel with dune bedform is predicted for four experimental data series under three scenarios (modeling based on flow characteristics, flow and bedform characteristics and flow and sediment characteristics) and the rate of input parameters is investigated using different performance criteria. The obtained results show that the support vector regression approach has desired accuracy in predicting the roughness coefficient. Also, the flow Reynolds number parameter with the most impact was recognized as the most significant parameter in estimating the roughness coefficient in the erodible beds with dune bedforms.