Machine learning-assisted fixed-bed column adsorption of chlorpyrifos using sustainable biomass-derived activated carbon

Abstract

Abstract Chlorpyrifos pollution in aquatic streams has become a serious environmental concern due to its high toxicity and adverse effects on ecosystems and human health. In this study, continuous fixed-bed column adsorption of chlorpyrifos was studied using activated carbon synthesised from Magnolia champaca leaf biomass (MCAC). Experiments were conducted at pH 2 by varying the flow rate (Q), bed height (Z), and influent chlorpyrifos concentration (C 0 ). Breakthrough curve analysis showed that C 0 = 25 mg/L, Z = 1 cm, and Q = 6 mL/min yielded optimal adsorption performance, achieving an adsorption capacity (q e ) of 105.81 mg/g. Among the conventional models, the Thomas, Adams-Bohart, and Yoon-Nelson models showed excellent agreement with high R 2 > 0.99. Furthermore, several machine learning (ML) models, including SVR, XGB, RF, GB, CB, LGBM, and ANN, were used to predict C t/ C 0 behaviour. Among these, the SVR model demonstrated superior predictive capability, with a test R 2 value of 0.9968 and low RMSE and MAE values of 0.0182 and 0.0122, respectively. Feature importance analysis and SHAP-based interpretation identified contact time as the most influential parameter governing breakthrough behaviour. Overall, this study demonstrates a robust approach for optimising adsorption-based wastewater treatment using fixed-bed modelling and interpretable ML.