As the standard technique of the Computational Fluid Dynamics (CFD) industry for simulating turbulent flows, the traditional Reynolds-averaged Navier-Stokes (RANS) method offers efficiency but incurs significant modeling errors compared to high-fidelity simulations. This study uses physics-informed machine learning to predict a correction for these modeling errors based on local observations of the mean flow, an approach found to generalize across a family of periodic hill flows.
- Jonas Luther
- Patrick Jenny
