Machine learning-assisted artificial neural network model for enhanced photocatalytic degradation of reactive dye textile wastewater effluent

This is the authors' abstract. We don't add key points for this paper.

The present study integrates statistical optimization and machine learning to model the photocatalytic degradation of three representative reactive dyes – Reactive Black 5 (RB5), Acid Blue 113 (AB113), and Acid Red 114 (AR114) – in simulated textile wastewater using a TiO 2 /UV shallow-pond reactor. The Box–Behnken design (BBD) was employed to investigate the effects of catalyst dosage, dye concentration, pH, A/V ratio, UV intensity, and irradiation time on degradation efficiency. The optimized conditions (TiO 2 dosage 1.0 g L −1 , dye concentration 20 mg L −1 , pH 4, irradiation time 90 min) achieved maximum degradation efficiencies of 95.2 %, 93.8 %, and 96.1 % for RB5, AB113, and AR114, respectively, with corresponding pseudo–first-order rate constants of 0.0246, 0.0205, and 0.0253 min −1 . The developed artificial neural network (ANN) model exhibited strong predictive accuracy (R 2 > 0.98, MSE < 0.002) and successfully captured the nonlinear interactions among process parameters. Overall, the study establishes a hybrid BBD–ANN framework as a powerful predictive and optimization tool for TiO 2 -based photocatalysis, providing quantitative insights into dye-specific degradation behavior and bridging experimental design with intelligent modeling for sustainable wastewater treatment.