Superior decomposition of xenobiotic dyes and pharmaceutical contaminants in wastewater using response surface methodology, artificial intelligence and machine learning for optimisation of a novel three-dimensional electrochemical technology

Superior decomposition of xenobiotic dyes and pharmaceutical contaminants in wastewater using response surface methodology, artificial intelligence and machine learning for optimisation of a novel three-dimensional electrochemical technology

Rapid global urbanisation and industrialisation have led to the widespread production of emerging anthropogenic contaminants such as xenobiotic dyes and pharmaceutical pollutants discharged into our natural waters. These pollutants are recalcitrant to environmental degradation and often escape into water from industrial effluent systems. In this work, a novel graphite intercalation compound (GIC) particle electrode was used to investigate the adsorption of synthetic dye pollutant, Reactive Black 5 (RB5), using a threedimensional electrochemical reactor to decompose the anthropogenic dye pollutant. Various adsorption kinetics and isotherm models were used to characterise the adsorption phenomena of GIC and determine the viability of the sorption process. When coupled with electrochemical oxidation technology, remarkably high dye removal efficiency can be achieved, and GIC can be electrochemically regenerated. Optimisation studies were conducted using response surface methodology and ANOVA analysis to provide insight into the significance of selectivity reversal from the salting effect of xenobiotic textile dye on GIC adsorbent. Non-linear models were simulated using the kinetic data in the order: Elovich > Bangham > Pseudo-second order > Pseudo-first order. The Redlich-Peterson isotherm was calculated to have a dye-loading capacity of 0.7316 mg/g by non-linear regression analysis. A range of error function analyses were used to evaluate the accuracy and precision of regression models. The best dye removal efficiency achieved using three-dimensional electrochemical treatment was approximately 93% using a current density of 45.14 mA/cm2, whereas the highest total organic carbon (TOC) removal efficiency was 67%. Various advanced artificial intelligence (AI) and machine learning (ML) optimisation techniques were used to enhance the prediction efficiency of dye and total organic carbon (TOC) removal efficiencies. The AI/ML optimised decolourisation efficiencies were 99.30%, 96.63% and 99.14% using central composite design-novel progressive response surface methodology (CCD-NPRSM), hybrid artificial neural network-eXtreme boosting gradient (ANNXGBoost) ensemble, and classification and regression trees (CART), respectively. The prediction efficiency of optimised models ranked in the descending order of hybrid ANNXGBoost, CCD-NPRSM and CART. The ANOVA results revealed that hybrid ANNXGBoost ensemble yielded a mean square error (MSE) and coefficient of determination (R2) of 0.014 and 0.998, outperforming CCD-NPRSM and with MSE and R2 of 0.518 and 0.998. The overall result showed that the hybrid ANN-XGBoost approach is the most feasible technique for improving the prediction efficiency of RB5 dye wastewater decolourisation.

File reproduced in accordance with the copyright policy of the publisher/author/creator.