A machine learning approach for prediction of reverse solute flux in forward osmosis

This study introduces a novel approach using a supervised machine learning model to accurately predict reverse solute flux (RSF) in the forward osmosis process. This study employed feature engineering techniques to identify significant process parameters that influence RSF. Notably, the results demonstrate the high effectiveness of the Categorical boosting (CatBoost) machine learning algorithm in RSF prediction, achieving an R-square value of 0.94 and a root mean square error of 0.44 when comparing the actual and predicted data. Furthermore, the model underwent simulation using real experimental data, revealing a minimal percentage error ranging from 0 to 2 % compared to the experimental reverse solute flux. The result showcases the potential of machine learning to save valuable time typically spent on experimental data while offering accurate predictions of reverse solute flux. The implications are particularly valuable in various applications involving reverse salt flux, where precise predictions can be achieved solely based on input parameters related to the forward osmosis process, membrane water permeability, and salt permeability.