AI-assisted prediction and optimization of micropollutants removal with forward osmosis membranes

This is the authors' abstract. We don't add key points for this paper.

Membrane technology is a simple, energy-saving, and high-performance separation process that can satisfy the high demand for water purification and separation of high-value products, such as dyes, pharmaceuticals, and therapeutical proteins. Despite all the recent progress in this field, serious issues remain unaddressed. High energy requirement and fouling in Reverse Osmosis (RO) membranes, concentration polarization, and reverse solute flux (RSF) for Forward Osmosis (FO) are among these. Recently, a number of Artificial Intelligence (AI) techniques, have been increasingly applied to optimize membrane performance by predicting and simulating the filtration process for a broad range of membranes and feed material. During this project, we try to harness the capabilities of different AI techniques namely Artificial Neural Networks (ANN) and Gradient Boosting Regressor (GBR) to first predict the performance of commercial FO membranes in removing various micropollutants with high accuracy and then use the best model available to develop a web-application accessible to public and researchers in order to estimate the water flux (J w ) and rejection rate (R%) they can obtain from a certain FO membrane removing a certain micropollutant without the need for costly and tedious experimental campaigns.