Support vector regression-based model for phenol adsorption in rotating packed bed adsorber

The excessive strength of phenol present in industrial wastewater is a major issue of concern to be looked upon. Among the pollutant removal techniques, a novel robust device, the rotating packed bed (RPB) adsorber, offers efficient adsorption of phenol due to its ability to magnify the mass transfer rate. In the present study, support vector regression (SVR) has been applied to predict adsorption of phenol on activated carbon in RPB by taking into account the independent parameters, namely, spray density, gravity factor, concentration, and contact time. The experimental data set of phenol adsorption sample has been randomized and normalized prior to constructing the models. The predictive ability of the SVR model has been compared with other data-driven models like artificial neural network (ANN) and multiple regression (MR) models. Both the SVR-based model and the ANN model have almost similar prediction efficacy; however, the ANN model was found to predict the outputs slightly better. The coefficient of determination (R²) and root mean square error (RMSE) values of test data set for the MR RPB adsorption model were found to be 0.934 and 0.149, while for the SVR and ANN-based models, these values were 0.996 and 0.045 and 0.998 and 0.027, respectively. Thus, it was concluded that the soft computing SVR and ANN models possessed tremendous potential to predict the adsorption process of RPB with remarkable accuracy and were greatly generalized.