Thermoplastic Matrix Composite Design for Cryotanks Using Multiscale Modeling and Bayesian Optimization

Designing lightweight, robust cryogenic storage tanks is critical for future launch vehicles, in-space propellant storage, and hydrogen powered aircraft. This work presents a multiscale modeling and Bayesian optimization framework for the design of thermoplastic matrix composite cryotanks. Molecular dynamics simulations are first used to determine temperature-dependent constituent properties for candidate thermoplastic matrices, which are homogenized to the lamina scale using NASA’s Multiscale Analysis Tool (NASMAT). These lamina properties, in combination with laminate family generation rules, are evaluated in HyperX structural optimization software to identify stacking sequences that meet all cryogenic load requirements. A Bayesian optimization framework is applied, with HyperX in the loop (via the HyperX API) to efficiently search across material and laminate design variables, yielding an optimized cryotank configuration with significant reductions in design cycle time compared to exhaustive search approaches.