Physics-informed Machine Learning Using Low-fidelity Flowfields For Inverse Airfoil Shape Design

This article discusses the use of physics-informed neural networks (PINNs) as a scientific machine learning approach for modeling physical quantities. Specifically, a low-fidelity-influenced PINN (LF-PINN) is proposed as a surrogate aerodynamic analysis model for inverse airfoil shape design. The LF-PINN utilizes low-fidelity flowfields and physics residuals from the Navier-Stokes equations to improve accuracy and reduce computational costs. Results show that the LF-PINN is able to correct low-fidelity flowfield quantities and achieve target airfoil shapes with significantly reduced computational time.