Prediction of uranium adsorption capacity on biochar by machine learning methods

The effective separation of uranium is a challenge for the treatment of radioactive wastewater. In this study, four machine learning (ML) methods (linear regression, support vector regression, random forest, and multilayer perceptron artificial neural network) were applied to predict the adsorption capacity of uranium on biochar. The relative importance of physical and chemical properties of biochar was also analyzed. Independent adsorption experiments were conducted with four biochar to verify the ML model. After training and verification, the model obtained with two hidden layers perceptron artificial neural network performs best by comparing the values of R 2 and RMSE. The structural properties of biochar, such as specific surface area, are more important for the adsorption capacity of uranium than the chemical composition. ML modeling provides a new strategy for the design and tailoring of biochar for uranium adsorption, which can significantly reduce the experimental workload and the safety risks associated with radioactivity. • Machine learning methods were successfully applied to predict uranium adsorption on biochar. • The model obtained by multilayer perceptron with two hidden layers shows the best performance. • Machine learning models were verified by the independent adsorption experiments. • The physical properties of biochar are more important than the chemical properties for uranium adsorption.