Machine Learning-Based Prediction of Polymer Chemical Resistance to Organic Solvents

Predicting the chemical resistance of polymers to organic solvents is a longstanding challenge in materials science, with significant implications for sustainable materials design and industrial applications. In this study, we present an interpretable data-driven framework that extends analysis of polymer chemical resistance beyond conventional solubility models. By systematically analyzing a large data set of polymer–solvent combinations, we observe associations between polymer crystallinity and density, as well as solvent polarity, and resistance outcomes, consistent with established theoretical models. Using a curated data set of 2231 polymer–solvent combinations and gradient boosting models trained on MD-derived descriptors, force-field kernel mean descriptors, and COSMO-RS χ parameters, the model achieved ROC-AUC values of 0.85 and 0.91 within the present data set and validation setting under polymer-level and solvent-cluster-level validation, respectively. These findings provide a foundation for exploratory screening and hypothesis generation for polymer chemical resistance, while acknowledging the scope of the underlying data set.