Can machine learning methods guide gas separation membranes fabrication?

Abstract

Transforming a vast array of candidate materials into membranes with suitable morphologies and improved molecular separation performance is an arduous and costly endeavor for membrane scientists. With the advancement made in artificial intelligence and machine-learning in recent years, it is timely to ask: can machine learning methods guide gas separation membranes Fabrication? The answer is “YES”, and this article explains the justifications for this answer by systematically reviewing and analyzing the up-to-date research efforts in the field. This work aimed to explore the potential of ML algorithms as an effective and cost-saving tool in guiding the experimental process of developing the next generation polymeric membranes, and in addressing the critical needs in the field. Findings demonstrate that training Heteropolymers instead of Homopolymers, synthesizing novel polymers by an inverse design approach, and using reliable datasets that are created under the same conditions, are the most crucial factors to achieve the design intent. A path from A to Z for anyone who intends to use ML algorithms in the membranes’ synthesis process is offered. The article concludes with a brief discussion on future development prospects and open issues that are yet to be addressed for ML‐driven polymeric‐based membranes design and optimization.