Optimized expressions to evaluate the flow discharge in main channels and floodplains using evolutionary computing and model classification

Abstract

ABSTRACTDue to momentum exchange between main channel and floodplains, the flow hydraulics in compound channels is taken into account as a comparatively complicated system. Most studies in this research fields are focused on the prediction of cross-sectional average velocity and total flow discharge. In quite a few situations, however, the subsection or individual flow discharges are imperative rather than the total discharge. For instance, in flood conditions and in the case of spill of water on the floodplains, the bed and suspended sediment loads are inevitably transported by the main channel flow discharge. In current investigation, using laboratory stage–discharge datasets from canals with compound channel sections, the individual flow discharges in the main channel and over floodplains are predicted applying gene-expression programming (GEP), model tree (MT) and evolutionary polynomial regression (EPR), and then compared with traditional divided channel methods. Results showed that the proposed soft computing methods have promising performance in the prediction of subsection flow discharges for both main channel and floodplains. EPR provided the flow discharge in main channel and floodplains with more efficient performance compared to the GEP and MT models. Furthermore, among the traditional methods, the diagonal and vertical divided channel methods with mean errors of 11% and 19% have the greatest and lowest precision in estimation of main channel discharge, respectively. Conversely, over the floodplains the vertical and horizontal divided channel methods estimated the flow discharge with mean errors of 6.8% and 247% as the best and worst models in terms of efficiency, respectively.