PropellantAI data

This study proposes an AI-driven workflow for the intelligent design of energetic molecules in solid propellants. By integrating molecular property prediction, SHAP-based feature attribution, and latent space generative models, the workflow enables both interpretable analysis and performance-oriented molecular optimization. The results demonstrate that the carbon–oxygen ratio plays a dominant role in determining energetic characteristics, leading to the introduction of a concise metric (<i>OB</i>_CO) for guiding design. Furthermore, Bayesian optimization in the latent space yields candidate molecules with improved specific impulse and synthetic feasibility. This work highlights the potential of combining explainable AI and generative modeling to accelerate solid propellant formulation development.