Hypersonic Trajectory Optimization Using a Data-Driven Surrogate Model for Convective Heat Transfer

Accurate and rapid prediction of aerothermodynamic loads remains an ongoing challenge for intelligent trajectory design and optimization for hypersonic vehicles. Existing approaches rely on engineering approximations to estimate aerothermal heating, which are valid only for a limited range of flow conditions and geometric shapes. To address this shortcoming, we introduce a data-driven surrogate model to predict the convective heat flux anywhere on the vehicle surface as a function of freestream conditions, vehicle attitude, and wall temperature. We couple this novel framework with a gradient-based trajectory optimization algorithm that leverages modern computing hardware and open-source deep learning libraries. Using this architecture, we solve for a series of active maneuvers which minimizes the peak temperature measured across a local region of interest on the HIFiRE-5 vehicle payload. We also calculate a glide phase trajectory satisfying aerothermodynamic constraints imposed at a critical surface location. The demonstration cases indicate that our framework can successfully enforce terminal constraints while achieving optimization objectives. The proposed convective heating surrogate model allows for rapid fluid-thermal interaction analysis of hypersonic trajectories. This framework can be used to determine optimal trajectories based on the design characteristics and limitations of individual vehicle components, enabling a more sophisticated approach to maximizing performance and ensuring vehicle survivability.