Prediction of oxygen transfer at modified Parshall flumes using regression models

To measure the flow rate in irrigation, sewer, and storm water systems, Parshall flumes (PFs) and modified Parshall flumes (MPFs) are commonly used. The dissolved oxygen (DO) in water is one of the important parameters and the primary indicator of the quality of water. In the present study, an attempt was made to develop, analyze, and compare three regression models, viz. artificial neural network (ANN), fuzzy logic (FL), and adaptive neuro-fuzzy inference system (ANFIS) models for predicting the oxygen transfer efficiency (E20) at PF and MPF using observed data in the laboratory. The predictive capability of these models was assessed by means of comparison with observed data through popular statistical indices like coefficient of determination (R2), bias, mean square error (MSE), root mean square error (RMSE), and Nash–Sutcliffe model efficiency (NSE). The results of ANN outperform the other discussed regression models and the existing predictive models. Sensitivity analysis depicts that discharge per width, q, of the channel and sill height, K, of MPF are the most influencing key factors for the prediction of E20. Furthermore, parametric study yielded that with increase of q and K, E20 gradually increases for this data set. © 2018, © 2018 Indian Society for Hydraulics.