Subgrid-scale modeling
Most turbulent flows in engineering applications cannot be computed directly from the Navier-Stokes equations because the range of scales of motion is too large. Therefore, simulations of turbulent flow often have to resort to coarse-grained models of the scales for which numerical resolution is not available. In large-eddy simulation (LES), only the large eddies in a flow are resolved and the effect of the smaller scales is modeled. LES reduces the computational complexity of simulations of turbulent flow, and therefore much effort in computational fluid dynamics has been directed at the research of LES models.
We plan to continue developing subgrid-scale models for LES based on physical insight of turbulent flow, which can be extended to various flow configurations.
W. Rozema, H. J. Bae & R. Verstappen, "Local dynamic gradient Smagorinsky model for large-eddy simulation," Physical Review Fluids 7, 074604, 2022.
H. J. Bae & A. Lozano-Durán, "Towards exact subgrid-scale models for explicitly filtered large-eddy simulation of wall-bounded", Annual Research Briefs, Center for Turbulence Research, Stanford University, 207–214, 2017.
M. Abkar, H. J. Bae & P. Moin, "Minimum-dissipation scalar transport model for large-eddy simulation of turbulent flows", Physical Review Fluids, 1, 041701(R), 2016.
W. Rozema, H. J. Bae, P. Moin & R. Verstappen, "Minimum-dissipation models for large-eddy simulation", Physics of Fluids, 27, 085107, 2015.
Wall modeling
Even with a subgrid-scale model, the near-wall resolution required to accurately resolve the boundary layer in wall-bounded flows remains a pacing item in LES for high-Reynolds-number engineering applications. Choi & Moin (2012) estimated that the number of grid points necessary for a wall-resolved LES scales as Re^(13/7), where Re is the characteristic Reynolds number of the problem. By modeling the near-wall flow such that only the large-scale motions in the outer region of the boundary layer are resolved, the grid-point requirement for wall-modeled LES scales at most linearly with increasing Reynolds number. Therefore, wall-modeling stands as the most feasible approach compared with wall-resolved LES or DNS.
Our goal is to use physical insight of turbulent flow near the wall to develop dynamically robust wall-models that will adjust to provide the correct wall-shear stress for various flow configurations and geometries.
D. Zhou, M. P. Whitmore, K. P. Griffin & H. J. Bae "Large-eddy simulation of flow over Boeing Gaussian bump using multi-agent reinforcement learning wall model," AIAA Aviation, 2023.
H. J. Bae & P. Koumoutsakos, "Scientific multi-agent reinforcement learning for wall-models of turbulent flows," Nature Communications 13, 1443, 2022 (selected Editors' highlights).
H. J. Bae & A. Lozano-Durán "Effect of wall boundary conditions on wall-modeled large-eddy simulation in a finite-difference framework", Fluids 6, 112, 2021 (invited paper).
H. J. Bae, A. Lozano-Durán, S. T. Bose & P. Moin, "Dynamic slip wall model for large-eddy simulation", Journal of Fluid Mechanics, 856, 400–432, 2019.