Ph.D. Proposal

Howon Lee

(Professor Juergen Rauleder)

"Development of a Rapid, Uncertainty-Aware Aerodynamic Evaluation and Design Tool using Machine Learning on Experimental Data"

Thursday, December 4

12:30 - 2:30 p.m.

Guggenheim 442

Abstract: 

Aerodynamic pressure, force, and moment coefficients are fundamental parameters in aerospace engineering, forming the cornerstone of aerodynamic design and analysis. The capability to efficiently evaluate these coefficients is essential to navigate the aerodynamic design space. Lower-fidelity methods, while computationally inexpensive, often fail to capture complex flow phenomena, whereas high-fidelity Computational Fluid Dynamics (CFD) simulations offer superior accuracy at the cost of substantial computational resources, limiting their practicality during early-stage design. This study introduces the Large Airfoil Model (LAM), a probabilistic machine learning (ML)-based aerodynamic evaluation model trained entirely on an extensive experimental database. The model predicts airfoil aerodynamic coefficients with accuracy comparable to CFD, while requiring significantly lower computational cost and user expertise. The model also provides uncertainty estimates, offering insight into its prediction confidence. Leveraging the LAM’s computational efficiency, a novel probabilistic inverse framework is developed, where both the target pressure distribution and the resulting airfoil design are characterized as statistical distributions. This uncertainty-aware approach captures the variability inherent in real-world aerodynamics, including manufacturing tolerances. This work also explores the incorporation of LAM predictions as prior distributions within a Gaussian Process framework. This LAM-prior enables the training of robust physics-driven aerodynamics models. The methodology is demonstrated across two distinct applications: the reconstruction of pressure distributions from limited sensor measurements, and a finite-wing aerodynamics prediction model trained on a sparse database.

Committee:
Dr. Juergen Rauleder (advisor), School of Aerospace Engineering
Dr. Marilyn Smith, School of Aerospace Engineering
Dr. Graeme Kennedy, School of Aerospace Engineering
Dr. Brian German, School of Aerospace Engineering
Dr. Andrew Wissink, US Army