Modeling of photolytic degradation of sulfamethoxazole using boosted regression tree (BRT), artificial neural network (ANN) and response surface methodology (RSM); energy consumption and intermediates study

Abstract

This study explores the boosted regression trees (BRT), artificial neural network (ANN) and response surface methodology (RSM) to model and optimize the operational variables for the simulation of the Photolytic degradation of Sulfamethoxazole (SMX) and concurrent total organic carbon (TOC) removal, based on the experimental data set. Four candidate variables involving initial pH (2–11), initial SMX concentration (50–200 mg L−1), temperature (15–45 °C) and time (6–120 min) were considered for simultaneous optimization of SMX and TOC degradation. The result revealed that all the three models are statistically considerable as the values of R, R2, adj-R2 are >0.85, thus be deemed to work well in data fitting, prediction, and optimization, nevertheless, the values of R, R2, adj-R2, RMSE, MAE and AAD are far better for ANN and BRT than RSM method. The ∼100% SMX degradation conditions were found to be as follows: treatment time: 25 min, pH: 2.0, temperature: 35 °C and SMX concentration: 50 mg L−1, while the maximum possible removal of TOC under the given conditions was ∼25%. The percentage contribution (PC) of each variable was deduced by ANOVA analysis of proposed quadratic models which indicated that time and pH are important factors than temperature and SMX concentration. The photolytic intermediates and inorganic ions of SMX, were identified and a potential route of transformation was also proposed.