Discovery of an\nEnd-to-End Pattern for Contaminant-Oriented\nAdvanced Oxidation Processes Catalyzed by Biochar with Explainable\nMachine Learning

The utilization of biochar-catalyzed peroxymonosulfate\nin advanced\noxidation processes (BC-PMS AOPs) is widely acknowledged as an effective\nand economical method for mitigating emerging contaminants (ECs).\nEspecially, state-of-the-art machine learning (ML) technology has\nbeen employed to accurately predict the reaction rate constants of\nEC degradation in BC-PMS AOPs, primarily focusing on three aspects:\nperformance prediction, operating condition optimization, and mechanism\ninterpretation. However, its real application in specific degradation\noptimization targeting different ECs is seldom considered, hindering\nthe realization of contaminant-oriented BC-PMS AOPs. Herein, we propose\na hierarchical ML pipeline to achieve an end-to-end (E2E) pattern\nfor addressing this issue. First, the overall XGB model, trained with\nthe comprehensive data set, can perform well in predicting the reaction\nconstants of EC degradation in BC-PMS AOPs, additionally providing\nthe basis for further analysis of various ECs. Then, the submodels\ntrained with different EC clusters can offer specific strategies for\nthe selection of the optimum option for BC-PMS AOPs of specific ECs\nwith different HOMO–LUMO gaps, thus forming an E2E operating\npattern for BC-PMS AOPs. This study not only increases our understanding\nof contaminant-oriented optimization of AOPs but also successfully\nbridges the gap between ML model development and its environmental\napplication.