Generalized 5-DoF Model for Hypersonic Boost-Glide Vehicle Trajectory Predictions

Hypersonic boost glide (HBG) vehicles are strategically important due to difficult interception. Predicting the trajectory of an HBG vehicle, accounting for the aerodynamic, aeroelastic, and aeroheating effects is a computationally challenging task. On a fundamental level, several numerical difficulties are encountered because of the extremely small time steps required to obtain stable fluid solution. Further, a large gap in the time scales of fluid flow, structural elasticity, and thermal conduction in solid makes long-duration time-accurate aeroheating analysis computationally expensive. In our previous work, we addressed this shortcoming by developing a quasi-steady fluid-thermal interaction workflow (referred to as the high-fidelity model) that produced a 1600x speed up over time-accurate method without compromising the accuracy of aeroheating prediction. In yet another work, we integrated the quasi-steady workflow to predict long duration hypersonic boost glide trajectory considering 3-DoF motion of vehicle in 2D plane that included vehicle pitch and translation. In this paper, we extend the high-fidelity trajectory prediction framework in two aspects, the governing equation solution strategy is changed from earth fixed system to body fixed system. This latter approach enhances solution convergence by 2 times. Secondly, 5-DoF motion (excluding roll) is implemented. We investigated two different 5DoF approaches, earth-fixed model with linearized centrifugal force, and second a spherical geocentric Earth model. The latter is found to be more accurate than the former, although the former captures the overall qualitative pattern of the trajectory reasonably well. We have also developed a mid-fidelity model, wherein the fluid forces are replaced by a genetic programming (GP) based analytic model trained using the datasets generated using the high-fidelity model. The GP-based mid-fidelity model predicts trajectory in close agreement with the high-fidelity simulations, and provides results in seconds compared to days required by the latter. The accuracy of the GP model is limited by the underlying training datasets. For example, the model predictions deviate significantly from the high-fidelity model predictions for lower angle-of-attack case due to insufficient training datasets in those regions. Nonetheless, GP based mid-fidelity model is identified to be a viable approach for quick turn-around approximate analysis.