Machine learning enabled catalytic wet peroxidation of levofloxacin bearing wastewater using Cu/MCM-41

Levofloxacin (LVOX) is a widely used antibiotic and persistent in environment that causes major health and environmental risks. The present study investigated the catalytic wet peroxidation (CWPO) of LVOX wastewater using different weight percentages of copper (0.5-5 wt%) on MCM-41. Among these, 1 % Cu/MCM-41 showed better catalytic activity for the removal of LVOX. The surface area and pore volume of the MCM-41 decreased upon copper loading onto the MCM-41 framework (Cu/MCM-41), from 712 m2/g to 605 m2/g and 0.987 mL/g to 0.816 mL/g, respectively. X-ray photoelectron spectroscopy (XPS) analysis confirmed the presence of Si 2p, O 1 s, and C 1 s at consistent binding energies across all samples. However, in Cu/MCM-41, an additional Cu 2p peak was detected at 933.3 eV, that indicates the successful incorporation of copper species on MCM-41 framework. The maximum LVOX removal was observed at 94 %, while mineralization was attained 64 % through CWPO process at optimized reaction conditions of pH 10, a catalyst dosage of 1 g/L, H2O2 of 13.7 mmol/L, LVOX initial concentration of 500 mg/L, temperature of 333 K, and residence time of 180 min. LVOX mineralization kinetics follows a pseudo-first-order reaction with an R2 of 0.99. The thermodynamic study revealed that the CWPO of LVOX is non-spontaneous and endothermic in nature. Machine learning (ML) models were deployed to analyze the experimental data, including Gaussian support vector machine (G-SVM), Fine Tree-Random Forest regression (FT RFR), LS-boost regression (LS-BR), and Artificial neural network (ANN). The G-SVM, FT-RFR, LS-BR, and ANN model showed an adequate prediction of the response, with absolute average deviation (AAD) of 0.462, 1.71, 2.401, and 2.917 and root mean squared error (RMSE) of 6.531, 7.700, 9.346 and 9.665, respectively. Among these models, G-SVM demonstrated the highest prediction accuracy, with the lowest RMSE and AAD compared to other fitted models.