Comparative study of data-driven reduced-order modeling for atmospheric reentry

Accurate aerodynamic modeling is critical for atmospheric reentry simulations, particularly for design-for-demise analyses that require full trajectory propagation. This study investigates data-driven reduced-order models as efficient alternatives to high-fidelity computational fluid dynamics. Proper orthogonal decomposition (POD) and isometric feature mapping (ISOMAP) are considered in this work and applied to the analysis of the loads on the Automated Transfer Vehicle across representative continuum-regime reentry conditions. The two methods are integrated in the TITAN reentry simulation tool, and results are compared for surface pressure reconstruction and aerodynamic force and moments prediction. ISOMAP demonstrates better agreement with high-fidelity pressure fields, particularly in capturing localized nonlinear flow features. POD retains key global surface features with reduced model complexity. Both methods offer substantial reductions in computational cost while retaining key aerodynamic trends, supporting their use in system-level reentry analyses where efficiency and predictive fidelity are essential.