Imputation of missing gas permeability data for polymer membranes using machine learning

Polymer-based membranes have the potential for use in energy efficient gas separations. The successful exploitation of new materials requires accurate knowledge transport properties all gases interest. Open-source databases permeabilities are significant benefit to research community. Membrane Society Australasia (https://membrane-australasia.org/) hosts a database experimentally measured and reported polymer permeabilities. However, is incomplete, limiting its as tool. Here, missing values were imputed (filled) using machine learning (ML). ML model was validated against permeability measurements that not recorded database. Through imputing data, it possible re-analyse historical polymers look “missed” candidates with promising selectivity. In addition, systems limited experimental sparse features performed, we suggest once CO2 and/or O2 has been measured, most other selectivities, including those CO2/CH4 CO2/N2, can be quantitatively estimated. This early insight into system used at an initial stage rapidly identify worth further investigation.