Physics-Informed Neural Networks for Flow Around Airfoil

View Video Presentation: https://doi.org/10.2514/6.2022-0187.vid In this paper, surrogate models for fluid flows around airfoils for different angles of attack were developed using neural networks with physical constraints, known as physics-informed neural networks (PINN). Through fully connected layers, pressure and velocity fields were predicted and subsequently used to calculate the losses based on constraints set by boundary conditions as well as governing equations. The results obtained from the PINN are favorable when compared to computational fluid dynamics simulations. In addition, the PINN manages to compute the results up to 5 times faster than CFD.