Predictive Insights into Arsenic Remediation: Advancing Electro and Chemical Coagulation Through Machine Learning Models

Abstract

Arsenic contamination in water sources remains a critical environmental and public health challenge, mainly due to the toxicity of its trivalent (As(III)) and pentavalent (As(V)) forms. This study compares advanced predictive modeling to enhance arsenic remediation, comparing electrocoagulation (EC) and chemical coagulation (CC) processes for their efficiency and cost-effectiveness. Higher As(III) removal rates were achieved using iron and aluminum electrodes in EC (up to 99 % in 5 min using Fe electrodes) compared to CC (up to 90 % using Fe(II) coagulant). The study's results highlight the operational advantages of EC, including a 40 % cost reduction due to lower chemical usage and sludge production. Machine learning models, including Support Vector Machines (SVM), Regression Trees, Random Forest, and Gradient Boosting, were developed to predict removal efficiencies under diverse operational conditions. SVM exhibited the highest predictive accuracy for As(III) removal in EC with Fe electrodes (MSE = 0.340, R2 = 0.954). At the same time, Regression Trees outperformed other models for As(V) removal in CC with Fe(III) coagulants (MSE = 0.371, R2 = 0.997). These techniques are highly effective in optimizing arsenic removal processes, allowing for precise regulation of treatment parameters and reducing dependence on trial-and-error methods. The findings highlight electrocoagulation with iron electrodes as a sustainable and cost-effective approach to arsenic remediation, particularly for As(III), while underscoring the transformative role of predictive modeling in water treatment. This study successfully integrates experimental insights with machine learning, driving improvements in the efficiency and adaptability of arsenic removal technologies.