Hayden Valerie Dean
(Advisor: Prof. Mavris)
"A Reduced Order Modeling Methodology For Quantifying Entry Capsule Aerodynamics in Modeling and Simulation"
Thursday, December 14
1:00 p.m. EST
Collaborative Visualization Environment (CoVE) Weber Space and Technology Building (SST II)
Entry, descent, and landing (EDL) is a critical aspect of planetary missions enabling payloads to land safely on planetary surfaces. EDL is most often achieved utilizing a blunt-body entry vehicle because they generate large amounts of drag to slow the vehicle down during entry and dissipate heat effectively. Engineers rely on advanced Modeling and Simulation (M&S) to ensure requirements are being met to ensure EDL mission success, leveraging Monte Carlo analysis to identify mission sensitivities resulting from uncertainties. The most impactful type of uncertainty in EDL M&S is aerodynamic uncertainty, which can impact the vehicle design, subsystem needs, and feasible trajectories. Currently, aerodynamic databases are used to incorporate aerodynamic data in EDL M&S. These databases can take decades to generate, costing valuable resources to develop, often in cases where the outer mold line of the vehicle is continually iterating. The state-of-the-art methodology for incorporating aerodynamic data into EDL M&S, Computational Fluid Dynamics (CFD)-in-the-loop flight simulation, is too computationally expensive to apply at scale. Therefore, there is an industry need to incorporate high-fidelity aerodynamic data into EDL M&S, without developing costly aerodynamic databases.
Blunt-body entry capsules display the most chaotic aerodynamic interactions in the low supersonic to transonic flight regime, displaying dynamic instabilities as the vehicle decelerates through the subsonic flight regime. Both physical and computational testing methods are used to quantify entry vehicle aerodynamics. Static aerodynamics are easy to quantify, whereas dynamic aerodynamics are much more complicated to decipher. Simplifying assumptions on dynamic coefficient quantification methods omit valuable data during test post-processing that is critical to understanding the aerodynamic nature of the vehicle in the low supersonic to transonic flight regime, and advanced methods continue to rely on data fit techniques to predict dynamic aerodynamic coefficients.
Surrogate modeling techniques can efficiently incorporate high-fidelity data into EDL M&S. Reduced Order Models (ROMs) can leverage high-fidelity aerodynamic data to predict a field of responses. Fundamentally, ROMs aim to find a low-dimensional representation of a high-fidelity Full Order Model (FOM). ROMs can be classified in a variety of ways, each method being driven by the type of data the ROM must be capable of representing. Applications of ROMs in aerodynamics focus on capturing the parametric variation of fluid flows, or the time-variation of fluid flows. Applications of both parametric and time-varying ROMs is an emerging research area, with few developments applying ROMs in a coupled simulation, like what is shown in querying aerodynamics in EDL M&S.
Key gaps were identified through literature, addressing the needs for developing a ROM capable of predicting high-dimensional, unsteady, parametric, low supersonic to transonic entry capsule data for EDL M&S. Research Questions and Experimentation seek to find a Dimensionality Reduction (DR) technique and advanced regression strategy for the ROM to achieve the necessary fidelity of predicting entry capsule aerodynamics on a test set of CFD pressure and shear surface data, and develop a sampling strategy for Free-Flight-CFD (FF-CFD) to develop a free-flight ROM. The demonstration of this research seeks to first establish the developed ROM methodology on FF-CFD data, and then benchmark the ROM methodology against trajectory simulations utilizing a standard aerodynamic database, and by state-of-the-art CFD-in-the-loop flight simulation. The developed ROM-in-the-loop flight simulation is expected to incorporate a higher-fidelity of aerodynamic information than what is possible with aerodynamic databases with a lower online computational cost than CFD-in-the-loop flight simulation.
- Prof. Dimitri Mavris – School of Aerospace Engineering (advisor)
- Prof. Lakshmi Sankar – School of Aerospace Engineering
- Prof. Graeme Kenndy – School of Mechanical Engineering
- Dr. Bradford Robertson – School of Aerospace Engineering
- Dr D. Bruce Owens – Flight Dynamics Branch, NASA Langley Research Center