Assessment of grid anisotropy effects on LES models with different length scales

Abstract

The effect of the eddy viscosity length scale on the accuracy of large eddy simulation (LES) is investigated for several commonly used subgrid models. Multiple types of length scales are considered, both purely geometric length scales (including the cube-root of the cell volume and the maximum cell dimension) and length scales that depend on both the grid and the flow. The assessment is focused on how the LES prediction changes for increasingly anisotropic grids, in forced isotropic turbulence and channel flow at high Reynolds number. While some length scales perform better than others, the main conclusion is that no single length scale produces LES results that are completely independent of the grid anisotropy. Since every model and length scale tested produces accurate results on at least one grid and inaccurate results on at least one other grid, the study shows clearly the need to assess LES models on multiple grids with multiple different anisotropy ratios.