Master of Science (M.S.), University of Dayton, 2024, Aerospace Engineering
Traditional conceptual-level aerodynamic analysis is limited to empirical and/or inviscid
models due to considerations of computational cost and complexity. There is a distinct
desire to incorporate higher-fidelity analysis into the conceptual-design process as early as
possible. This work seeks to enable the use of high-fidelity data by developing and applying
multi-fidelity surrogate models that can efficiently predict the underlying response of a
system with high accuracy. To that end, a novel form of the multi-fidelity polynomial chaos
expansion (PCE) method is introduced, extending the surrogate modeling technique to
accept three distinct fidelities of input. The PCE implementation is evaluated for a series
of analytical test functions, showing excellent accuracy in creating multi-fidelity surrogate
models. Aerodynamic analysis of a generic hypersonic vehicle (GHV) is performed using three
codes of increasing fidelity: CBAERO (panel code), Cart3D (Euler), and FUN3D (RANS).
The multi-fidelity PCE technique is used to model the aerodynamic responses of the GHV
over a broad, five-dimensional input domain defined by Mach number, dynamic pressure,
angle of attack, and left and right control surface settings. Mono-, bi-, and tri-fidelity
PCE surrogates are generated and evaluated against a high-fidelity “truth” database to
assess the global error of the surrogates focusing on the prediction of lift, drag, and pitching
moment coefficients. Both monofidelity and multi-fidelity surrogates show excellent predictive
capabilities. Multi-fidelity PCE models show significant promise, generating aerodynamic
databases anchored to RANS fidelity at a fraction of the cost of direct evaluation.
Committee: Markus Rumpfkeil (Advisor); Jose Camberos (Committee Member); Timothy Eymann (Committee Member)
Subjects: Aerospace Engineering