A Highly Efficient Gene Expression Programming for Velocity Distribution at Compound Sewer Channel

Flow velocity distribution offers key information in open channels studies, and accurate evaluations of it prepare a wide range of applications in different fields such as sediment transport and power plant design. In this study, the aptitude of a new approach based on gene expression programming (GEP) is perused in velocity field predicting of the compound open channel. A simple GEP-based model is proposed for velocity distribution estimation in an entire cross section of the channel. The proposed model is trained in eight different hydraulic conditions, and the generalized ability of this model are evaluated in three different condition, which is no role in velocity distribution modelling. The results show that the proposed model predicts the velocity distribution precisely (MAPE=6.39 & RMSE =0.059). Also, the results indicated the model excellently estimated the dip phenomena, which is a complex feature in velocity distribution prediction. Furthermore, a partial derivative sensitivity analysis is presented to survey the effect of each effective parameters in the proposed model.