Generative AI empowered inverse design of 3D energetic material structures for customizable combustion

Energetic materials typically lack precise control over energy release rates once ignited. Inverse design of their structures offers a promising solution by enabling regulation of burning surface evolution and energy release profiles, facilitating customizable combustion. Such a capability could significantly enhance energy management in aerospace and chemical industrial systems. While both optimization-driven and data-driven approaches have demonstrated their potential, their optimality, time efficiency, and feasibility cannot be guaranteed simultaneously. Here, we present a dual-model approach combining a deep eikonal auto-decoder and a denoising diffusion probabilistic model to generate novel energetic material structures while accurately representing their entire burning surface evolution. A key to success is a optimization-driven fine-tuning technique that substantially improves the optimality while maintaining high volume loading efficiency and other critical constraints. This combination of optimization-driven and data-driven approaches opens a pathway toward fully customizable combustion of energetic materials, capable of precisely matching complex target performance curves. Wentao Li and colleagues present a generative AI framework for inverse design of 3D energetic material structures that enables customizable control of combustion. The approach rapidly generates feasible structures matching target energy release profiles, with fine-tuning substantially improving both performance matching and volume loading.