Coagulation-flocculation process optimization for industrial wastewater treatment: a multi-objective computational solution

Abstract

Coagulation-flocculation is widely applied in industrial wastewater treatment due to its effectiveness in removing heavy metals and ensuring water quality. However, traditional chemical dosage decisions rely on empirical methods or regression-based models that are time-consuming and not adaptive to real-time fluctuations. A hybrid computational framework is proposed that integrates Non-dominated Sorting Genetic Algorithm II (NSGA-II) and Artificial Neural Network (ANN) to optimize the coagulation-flocculation process under multiple conflicting objectives. The ANN model is trained using three algorithms, namely Levenberg–Marquardt (LM), Resilient Backpropagation (RP), and Scaled Conjugate Gradient (SCG), to predict key performance indicators, including treatment cost (USD m⁻³), effluent copper concentration (mg L⁻¹), and sludge level (ordinal scale 1–5). These predicted outputs are subsequently used by NSGA-II to derive Pareto-optimal solutions that reflect trade-offs among the competing objectives. Experimental results show that the proposed GA-ANN model with LM achieved the highest accuracy and computational efficiency, reducing optimization time from 1400 s to as low as 133 s. Compared to existing models, this approach enables faster, more accurate, and adaptive decision-making, making it highly suitable for real-time industrial applications. The proposed framework provides a novel, data-driven strategy for multi-objective optimization in wastewater treatment, contributing to more sustainable and cost-effective operations.