Iron and aluminum based beaded sorbents for removing arsenic and fluorine from water: Application of machine learning for material selection

Abstract

In this study, the waste reuse of water purification plant sludge (WPS) and coal mine drainage sludge (CMDS) was carried out and synthesized into beaded adsorbents BWPS and BCMDS. The WPS is rich in aluminum, and the aluminum-based adsorbent has a good adsorption effect on the fluoride in water. CMDS is rich in iron and calcium, arsenic can be removed from water using this device. The experiments mainly probed the pollutant-removing effects of two beaded adsorbents on fluoride and arsenic in water. According to the Langmuir isotherm equation, the maximum adsorption capacities (Qmax) of F- on BWPS and BCMDS are 0.90 and 0.65 mg g−1, and the Qmax of As (V) are 9.87 and 14.88 mg g−1, respectively. With a range of pH 4 ∼ 10 in experiments, increasing pH decreased the pseudo-second-order rate (K2) of F- and As(V) adsorbed on the beaded adsorbents. The mechanism for removing F- by BWPS is physical adsorption, on BCMDS are ion exchange and precipitation, and for As(V) are physisorption and precipitation. In the desorption experiments, the results indicated that both adsorbents can be reused. In addition, it is combined with XGBoost and SHapley Additive exPlanations (SHAP), to predict the adsorption capacity. Data preprocessing and model training improved the prediction accuracy, resulting in a final RMSE of 0.429 and an average prediction accuracy of 91%. Through the evaluation index (EI) select the final adsorbent, the results showed that the choice of adsorbent was unchanged in general conditions and expert surveys, however, in some specific scenarios can be changed.