Machine Learning Models for Inverse Design of the Electrochemical Oxidation Process for Water Purification

This study reports the machine-learning-basedtarget-orientedframework for inverse design of the electrochemical oxidation processfor water purification. In this study, a machine learning(ML) framework is developedtowardtarget-oriented inverse design of the electrochemical oxidation (EO)process for water purification. The XGBoost model exhibited the bestperformances for prediction of reaction rate (k)based on training the data set relevant to pollutant characteristicsand reaction conditions, indicated by R (ext) (2) of 0.84 and RMSEext of 0.79. Based on 315data points collected from the literature, the current density, pollutantconcentration, and gap energy (E (gap)) wereidentified to be the most impactful parameters available for the inversedesign of the EO process. In particular, adding reaction conditionsas model input features allowed provision of more available informationand an increase in the sample size of the data set to improve themodel accuracy. The feature importance analysis was performed forrevealing the data pattern and feature interpretation by using Shapleyadditive explanations (SHAP). The ML-based inverse design for theEO process was generalized to a random case for tailoring the optimumconditions with phenol and 2,4-dichlorophenol (2,4-DCP) serving asmodel pollutants. The resulting predicted k valueswere close to the experimental k values by experimentalverification, accounting for the relative error lower than 5%. Thisstudy provides a paradigm shift from conventional trial-and-errormode to data-driven mode for advancing research and development ofthe EO process by a time-saving, labor-effective, and environmentallyfriendly target-oriented strategy, which makes electrochemical waterpurification more efficient, more economic, and more sustainable inthe context of global carbon peaking and carbon neutrality.

Automated summary

This study presents a machine-learning-based framework for the inverse design of the electrochemical oxidation process for water purification. The XGBoost model achieved excellent performance in predicting the reaction rate (k), with an R (ext) (2) of 0.84 and RMSEext of 0.79. The most influential parameters for inverse design were identified as current density, pollutant concentration, and gap energy (E (gap)). This research provides a paradigm shift towards a data-driven approach for advancing the research and development of electrochemical water purification.