Energetic Polymer Property Prediction using Machine Learning.

Energetic polymers are being actively studied by the public for potential use in modern propellant, explosive, and pyrotechnic material mixtures. Energetic polymers can be designed for use in a binder system as part of a formulation effort to improve safety and qualification of materials for end use. Beyond their mechanical properties, energetic mixtures benefit from the energetic polymer's ability to contribute to useful work during detonation. Inspired by recent advances in machine learning (ML) enabling efficient prediction of polymer properties, in this talk we will present an ML-based framework for predicting mechanical and physical properties of energetic polymers based on both off-the-shelf and bespoke physiochemical molecular descriptors. We curated an experimental dataset to train and validate the models explored. Enabling rapid polymer material screening, our models achieve predictive accuracy comparable to other leading efforts in ML for nonenergetic polymers.