Applicability of Several Soft Computing Approaches in Modeling Oxygen Transfer Efficiency at Baffled Chutes

Abstract

The present study investigates the accuracy of five different data-driven techniques in estimating oxygen transfer efficiency in baffled chutes: feedforward neural network (FFNN), radial basis neural network (RBNN), generalized regression neural network (GRNN), adaptive neuro fuzzy inference system with subtractive clustering (ANFIS-SC), and adaptive neuro fuzzy inference system with fuzzy c-means clustering (ANFIS-FCM). Baffled apron chutes or drops are used on channel structures to dissipate the energy in the flow. A baffled chute design is effective both in energy dissipation and in aerating the flow and reducing nitrogen supersaturation. There is a close relationship between energy dissipation and oxygen transfer efficiency. This study aims to determine the aeration efficiency of baffled chutes with stepped (S), wedge (W), trapezoidal (T), and T-shaped (T-S) baffle blocks. The performances of the FFNN, RBNN, GRNN, ANFIS-SC, and ANFIS-FCM models are compared with those of multilinear and nonlinear regression models. Based on the comparisons, it was observed that all data-driven models could be successfully employed in modeling the aeration efficiency of S, W, and T-S baffle blocks from the available experimental data. Among data-driven models, the FFNN model was found to be the best. (C) 2017 American Society of Civil Engineers.