Coupling Machine Learning with Life-Cycle Assessment for Sustainable Electrocoagulation Treatment of Real Textile Dyeing Wastewater: A Multi-Objective Optimization Framework

Abstract

Electrocoagulation (EC) is increasingly adopted for the treatment of real textile dyeing wastewater, yet its process design is rarely guided by a joint assessment of treatment performance and environmental burden. This study presents an integrated machine-learning (ML) and life-cycle assessment (LCA) decision-support framework for aluminium-electrode EC treatment of real textile dyeing wastewater. A literature-calibrated, mechanistically consistent in-silico dataset of 500 operating conditions was generated by coupling Faraday's law of electrolysis with an Ohmic cell-voltage sub-model, spanning current density, electrolysis time, initial chemical oxygen demand (COD), pH, inter-electrode distance and NaCl dosage. Four ML algorithms (Random Forest, Gradient Boosting, an artificial neural network and support vector regression) were benchmarked as surrogate models of COD removal efficiency, specific energy consumption (SEC) and aluminium electrode consumption, achieving hold-out test R² of 0.986, 0.985 and 0.994, respectively. A surrogate-assisted NSGA-II multi-objective optimization identified operating conditions that increased COD removal from an 84.0% conventional-operation baseline to 90.8% while cutting SEC by 41.4% (6.48 to 3.80 kWh m⁻³). A comparative, ISO 14040/14044-compliant, ReCiPe 2016 Midpoint LCA (functional unit: 1 m³ treated wastewater) showed that the ML-optimized condition reduced global warming potential by 18.9% (7.06 to 5.73 kg CO2-eq m⁻³ — within 8% of an independently reported literature benchmark of 6.53 kg CO2-eq m⁻³), fossil resource scarcity by 18.1%, terrestrial acidification by 13.7% and freshwater eutrophication by 17.3%. Hotspot analysis identified aluminium electrode production as the dominant contributor (54–77%) to every impact category, ahead of electricity consumption. The framework demonstrates that ML-guided process optimization can simultaneously improve treatment performance and reduce environmental burden, and it identifies electrode-material substitution as the priority lever for further impact reduction.