Comparative analysis of response surface methodology and some artificial intelligence models in the prediction of methyl green degradation by Fenton process

Abstract

ABSTRACT Dyes rejected by various industries are one of the major hazardous pollutants to be quantified. It is therefore necessary to remove the dye before discharging it into the main water stream. In this study, the catalytic degradation of methyl green (MG) cationic dye by Fenton process ‘Fe2+/H2O2’ was modelled using four artificial intelligence (AI) models; artificial neural networks (ANN), support vector machine (SVM), random forest (RF) and gradient boosting regressor tree (GBRT) and compared with response surface methodology (RSM). The models were analysed by considering four factors that affect the degradation process: concentration of MG, concentration of Fe2+, concentration of H2O2 and temperature . The objective of this analysis was to quantify the accuracy of prediction of four types of AI models along with RSM model. Sensitivity analyses comprising correlation coefficient, mean square error (MSE) were employed to assess the adequacy of the proposed models. The mean square error (MSE) values corresponding to the validation set for MG degradation were 0.434, 0.232, 0.307, 0.177 and 0.223 while the respective coefficient of determination values were 0.9658, 0.95780 0.9440, 0.9677 and 0.9574 for the ANN, SVM, RF, GBRT and CCD models, respectively. According to these results, the AI and RSM gave a high accuracy in predicting the degradation of the methyl green dye. The four AI models have satisfactory goodness-of-fit, robustness and predictive ability compared to RSM model. The GBRT performs slightly better performance than the other models. Maximum degradation (De% > 98) was achieved at an initial dye concentration of 20 mg/L, a Fe2+ concentration of 7 mM, a hydrogen peroxide dose of 25 mM, and a temperature of 50°C. Overall, the applied statistical analysis of the results indicates that the four methods can be employed as an AI models for monitoring and prediction the degradation by Fenton process.