Establishing structure–property linkages for wicking time predictions in porous polymeric membranes using a data-driven approach

Abstract

The goal of this study is to develop correlations between microstructure morphology and macroscopic material behavior, known as structure–property linkages. These correlations can be used to predict material behavior and enable virtual materials design efforts. In this work the structure–property linkages for the capillary-driven fluid transport through highly porous open-pored polymeric membranes are determined by a data-driven approach. To establish linkages, about 400 porous microstructures with different geometrical features are algorithmically generated and characterized in 3D, using fluid flow simulations and image analysis methods. The data processing pipeline for the generation and analysis of the microstructures is implemented by a generic workflow tool called KadiStudio, which is embedded in the research data infrastructure Kadi4mat. The data-driven analysis enables predictions about the propagation time of a fluid over definable distances when only the porosity and the ligament radius are known as microstructural properties. The generated knowledge can be utilized for an accelerated development of novel polymeric membranes with an optimized pore structure.