Special aerospace seminar: Near-wall turbulence modeling for high Reynolds number wall-bounded flows

Abstract

Fine-scale modeling is the key to enabling scale-resolving simulations, e.g., large-eddy simulations (LES), of fluid flows at practically relevant Reynolds numbers and to bringing petaFLOP and exaFLOP machines to their full potential. This talk focuses on the modeling of the fine scales near a solid boundary, i.e., wall modeling, in the context LES. A data-based approach will be taken and use a feed-forward neural network to model the effect of the near-wall unresolved fine scales in a wall-modeled LES. The neural net is informed about the known physics and is trained using the available high-fidelity direct numerical simulation data. It will be shown that for a flow for which there is a reasonably good understanding, taking a physics-informed data-based approach leads to fine-scale models that generalize better than the conventional physics-based models, and for a flow for which there is not a good understanding, taking a data-based approach allows reasonably good results to be obtained.

About the speaker

Xiang Yang is an assistant professor in the Mechanical Engineering Department at Penn State University. He got his PhD in mechanical engineering from Johns Hopkins University in 2016 and was a postdoctoral researcher at the Center for Turbulence Research from 2016 to 2018, after which he joined Penn State and has been an assistant professor since.