Research Information System
Ozone: Science and Engineering, 2026 (SCI-Expanded, Scopus)
Oxidation rate constants are critical parameters for the design and optimization of advanced oxidation processes (AOPs) employed to remove persistent organic pollutants from water. In this study, an interpretable, design-oriented modeling framework is developed to predict pseudo-first-order oxidation rate constants expressed as log10(k) for chlorophenols under photo-Fenton, Fenton, and photoperoxidation conditions, while direct UV photolysis is considered separately as a non-radical photochemical treatment. Experimental operating parameters, including pH, temperature, oxidant dosage, iron concentration, and radiation intensity, were integrated with Abraham solute descriptors to capture both process-level and molecular-level influences on oxidation kinetics. An Extreme Gradient Boosting (XGBoost) model was trained and interpreted using SHapley Additive exPlanations (SHAP). Model performance was evaluated using nested grouped cross-validation, with folds constructed to ensure prediction on previously unseen chlorophenol compounds. The optimized model achieved strong in-sample performance (training R2 = 0.94 ± 0.10) and moderate generalization to unseen compounds (test R2 = 0.41 ± 0.50; RMSE = 0.54 ± 0.27 on log10(k)). Model interpretation revealed that oxidation kinetics are predominantly governed by operating conditions, while molecular descriptors related to polarizability and hydrogen bonding interactions modulate pollutant-specific reactivity. Overall, the proposed framework provides mechanistically interpretable insights that support rational AOP selection and operating-condition optimization.