• Heterogeneous MIL-100(Fe)/persulfate Electro-Fenton enabled efficient tetracycline degradation. • RSM exhibited robust multivariate optimization with strong predictive reliability. • XGBoost surpassed Random Forest in capturing nonlinear and complex interactions. • NSGA-II identified Pareto-optimal trade-offs between removal efficiency and energy demand. • Hybrid integration of RSM and machine learning advanced pharmaceutical wastewater treatment design. In this study, a simultaneous application of RSM and machine learning models was employed to perform multi-objective optimization of the Electro-Fenton process, aiming to maximize tetracycline removal efficiency and minimize electrical energy consumption using the heterogeneous catalyst MIL-100(Fe) and persulfate. The experiments were conducted in a one-liter glass pilot reactor equipped with four graphite electrodes. The experiments were designed based on the CCD and the quadratic regression model of RSM demonstrated high predictive accuracy, with R 2 of 0.9756 and 0.9894 for tetracycline removal and energy consumption, respectively. To capture nonlinear relationships and improve predictive performance, two machine learning algorithms, Random Forest and XGBoost were implemented. Performance comparison results indicated that between the applied machine learning models XGBoost, with R² values of 0.78 for removal efficiency and 0.95 for energy consumption along with acceptable statistical errors, showed improved predictive capability relative to Random Forest, which yielded R 2 values of 0.73 for removal efficiency and 0.96 for energy consumption. For simultaneous optimization of removal efficiency and energy consumption, the NSGA-II algorithm was employed, providing a set of Pareto-optimal solutions for both XGBoost and Random Forest models. This precise and efficient approach provides a sustainable framework for the design of pharmaceutical wastewater treatment systems by integrating modeling techniques with machine learning.
Ezati et al. (Sun,) studied this question.