From Data to Membranes: Predicting Performance and Guiding Material Selection with Machine Learning

The integration of machine learning (ML) techniques into membrane research offerspromising possibilities for accelerating material discovery and enhancing processperformance prediction. In this talk, I will first provide a brief motivation for the use ofML in membrane development, followed by an overview of key ML concepts andcommon methods for regression and classification tasks.

Two case studies willillustrate the practical application of these tools. The first explores the use of ML topredict gas permeability in polymeric membranes and leverages these predictions toidentify promising new materials for gas separation.

The second applies ML models intwo different tasks: to identify and estimate the distribution of pores, and to forecastthe long-term performance of membranes used in membrane distillation processes,offering a data-driven approach to system monitoring and optimization. Theseexamples highlight how ML can complement experimental efforts and supportdecision-making in membrane research.