Numerical investigation of the oxygen transfer rate in an open channel system

ABSTRACT This study explores the effectiveness of soft-computing techniques, such as M5P, random forest (RF), artificial neural network (ANN), Gaussian process (GP), and support vector machine (SVM), to predict the aeration efficiency (E20) and volumetric oxygen transfer (KLa20) in an open channel system. Experimental data were used to develop the models, and using three statistical indices, a Taylor diagram, and a Whisker box plot, the performance of each model was evaluated. The results indicate that these predictive models can effectively capture the complex nonlinear relationships inherent in aeration processes, with ANNs demonstrating the highest accuracy in predictions for the aforementioned outputs. The values of the correlation coefficient (CC) of 0.9207, mean absolute error (MAE) = 0.161, and root mean square error (RMSE) = 0.0229 for predicting E20 and CC = 0.9762, MAE = 0.0093, and RMSE = 0.0112 for KLa20 were obtained. The sensitivity analysis suggests the input parameter, hydraulic radius (HR), as the most influential factor in predicting oxygen transfer. The models also revealed insights into optimal operational ranges for maximizing aeration efficiency. This research underscores the potential of soft computing techniques as valuable tools for improving the management of aeration systems and promoting sustainable practices in wastewater treatment.