Physics-Guided Airfoil Optimization Using Neural Network with Locally Converging Input (NNLCI)

Date:

Tuesday, November 25, 2025

Time:

10:00 a.m. – 12:00 p.m. ET

Location:

MK 325, Guggenheim Building, and Microsoft Teams (https://teams.microsoft.com/l/meetup-join/19%3ameeting_NzNiNjlmZDMtYTA2Zi00NzcwLThhY2YtNjUzZWU5ZDEzMGUy%40thread.v2/0?context=%7b%22Tid%22%3a%22482198bb-ae7b-4b25-8b7a-6d7f32faa083%22%2c%22Oid%22%3a%227575b221-afee-4070-8506-4cf2cd68f2cd%22%7d)

Tzu-Jung Lee

Machine Learning Ph.D. Candidate

Daniel Guggenheim School of Aerospace Engineering

Georgia Institute of Technology

Committee

Dr. Vigor Yang (Advisor) - Daniel Guggenheim School of Aerospace Engineering, Georgia Institute of Technology

Dr. Yingjie Liu - School of Mathematics, Georgia Institute of Technology

Dr. Yongxin Chen - Daniel Guggenheim School of Aerospace Engineering, Georgia Institute of Technology

Dr. Lakshmi Sankar - Daniel Guggenheim School of Aerospace Engineering, Georgia Institute of Technology

Dr. Robert Funk - Principal Research Engineer, Aerospace, Transportation and Advanced Systems Laboratory (ATAS), Georgia Tech Research Institute

Abstract

This dissertation develops a physics-guided framework for airfoil optimization using a Neural Network with Locally Converging Input (NNLCI) in a PARSEC parameter space. The NNLCI surrogate reconstructs high-fidelity flow fields from local multi-fidelity patches, while preserving shock structures and near-wall behavior that are critical for accurate lift and drag prediction. From the reconstructed fields, we recover surface pressure and viscous traction, and integrate them to obtain lift and drag, which are validated against high-fidelity CFD force integrals. The framework then performs constrained drag minimization at fixed lift using Differential Evolution, enforcing geometric feasibility and a lift constraint while relying mainly on low-fidelity evaluations with periodic high-fidelity checks.