Numeric and nonnumeric information input to predict adsorption amount, capacity and kinetics of tetracyclines by biochar via machine learning

Abstract

The detection of tetracyclines (TCs) adsorption performance by specific adsorbent is a time-consuming process. Thus, developing prediction framework based on existing data to quickly evaluate adsorption performance is necessary especially in the urgent scenario. Herein, we employ machine learning to deliver the accurate prediction of TCs adsorption performance by biochar via grouping numeric/nonnumeric information over physiochemical properties of biochar and environmental pressures. We find the TCs adsorption by biochar is a purely physical behavior where porous filling is the primary mechanism. Porous structure of biochar and environmental pressures co-determine the adsorption amount, and the synergism of physiochemical properties of biochar and environmental pressures determines adsorption capacity. Adsorption kinetics is subject to physical properties and environmental pressures. Chemical properties of biochar have a limited influence on adsorption performance and its main contribution is delivered via the interactions between C-contained or O-contained functional groups and TCs. Simplifying model only with reduced variables input can also accurately predict TCs adsorption performance by biochar. This strategy enables the portable prediction in the urgent scenario since expensive instruments and complex detections can be avoided. These findings provide a comprehensive understanding of the way of physiochemical characteristics and environmental pressures on adsorption performance, and offer useful tips in designing biochar-based adsorbents for TCs removal.