Estimation of apparent shear stress of asymmetric compound channels using neuro-fuzzy inference system

Abstract

Discharge calculations and conveyance estimations in rivers are of prime importance during high flow events. Conventional methods using single global resistance parameters are incapable of assessing discharge with high precision; thus incorporating apparent shear into these methods not only improves the results but also addresses the momentum transfer phenomena over sub-sections. Apparent shear stress modelling for asymmetric channels with different geometric and roughness parameters is found to be monotonic and tedious in linear regression. To improve its accuracy and efficiency, one of the modelling tools is to apply adaptive network-based fuzzy inference system (ANFIS). In this paper, apparent shear stress is trained, modeled and tested for a wide range of asymmetric channels including various experimental and field data. It is found that asymmetrical channels have large positive shear force acting at the vertical imaginary interface between the floodplain and main channel. The positive apparent shear force indicates that the slower floodplain flow retards the faster flow in the main channel. Therefore, considering correct shear stress in the discharge estimation will result in higher accuracy. The application of ANFIS in modelling the output variable of apparent shear stress is done using five geometrical and hydraulic parameters of varying range from small-scale experiments to real field measurements as the input variables. The coefficient of determination (R2) for the modeled apparent shear is 0.96, which is exceptionally high as comparison to past models. Furthermore, comparison of discharge calculation using modeled apparent shear stress and other conventional methods has shown that the results obtained using vertical interface shear stress have higher accuracy.