Interpretable machine learning-driven modeling and optimization of a continuous-flow electrooxidation reactor for real wastewater treatment conditions
Process Safety and Environmental Protection, cilt.218, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 218
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.psep.2026.109453
- Dergi Adı: Process Safety and Environmental Protection
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Chemical Abstracts Core, Chimica, Compendex, INSPEC
- Anahtar Kelimeler: Continuous flow electrooxidation reactor, Conventional models, Feature importance analysis, Machine learning models, Municipal wastewater, Slaughterhouse wastewater
- Yıldız Teknik Üniversitesi Adresli: Evet
Özet
Continuous-flow electrooxidation (CFEO) is an environmentally sustainable and promising advanced technology for treating complex municipal and industrial wastewater. Optimizing this process remains challenging due to the strong nonlinear relationships between the operational parameters and wastewater characteristics. Our study presents a novel, integrated framework that combines machine learning (ML) models with conventional models to accurately predict, optimize, and systematically analyze CFEO performance in treating real municipal wastewater (MWW), slaughterhouse wastewater (SWW), and their combination (SWW/MWW). Under CFEO conditions, the organic matter oxidation rate was described by both first-order (R2 and Adj. R2 > 0.97) and heterogeneous (Kh reaching a maximum of 1.145 m/min) kinetic models. The CatBoost model achieved the highest predictive accuracy (R2 and Adj. R2 > 0.98) and lowest error rates in capturing the complex relationships between CFEO operating parameters and wastewater characteristics. The CatBoost model predicted a reduction in energy consumption and an increase in treatment efficiency. Feature importance analysis with grid-search-based hyperparameter optimization identified the EO time, anode material, and current density as the most influential parameters affecting CFEO efficiency. The marginal impact of the CatBoost model was further evaluated using PDP and ICE analysis. Robustness and fidelity analyses confirmed the reliability and interpretability of the CatBoost model. Our findings provide a scalable, predictive, and innovative design and optimization guide aimed at improving treatment performance, reducing energy consumption, and fostering a better understanding of the complex relationships among CFEO operational parameters in actual wastewater using ML models.