Apparent Shear Stress and Its Coefficient in Asymmetric Compound Channels Using Gene Expression and Neural Network

Flow interaction over the interface between main channel and floodplains affects the overall discharge capacity and discharge distribution in compound open channels. Many investigators have attempted to empirically estimate flow interaction in terms of an apparent shear stress acting on the imaginary interface between the main channel and floodplain. However, past models are neither generalized for asymmetric channels nor applied to a wide range of data sets including field data, even though the apparent shear stress for asymmetric channels is found to be higher in comparison to symmetric channels for the same depth of flow. In this paper, using gene expression programming and a back propagation neural network, a generalized dimensionless formula is proposed for predicting percentage shear force and apparent shear stress on the vertical interface between the main channel and floodplain for asymmetric compound channels. The variation of bed characteristics and their dependability on the formula has been tested against a wide range of experimental and river data reported in the previous studies. Statistical analysis shows that the formulas produced in the curve fitting through gene expression and a feedforward back propagation neural network are very satisfactory and better than past models. The exceptionally high accuracy of the proposed models implies that they can be extended to use for a wide range of applications.