Influence of resolution and Reynolds number on large-eddy simulations of channel flow using relaxation filtering

Abstract

In this work, a large-eddy simulation (LES) approach based on relaxation filtering, val- idated for free shear flows in a number of previous studies, is applied to turbulent channel flows in order to assess its validity for wall-bounded flows. For this, simulations are per- formed using different grids, and for various Reynolds numbers. The reliability of the LES results is estimated based on both a priori and a posteriori analyses. In a first step, a channel flow at a friction Reynolds number Reof 300 is computed using grids with increasing resolution. For mesh spacings equal to and lower thanx + = 30, �y + = 1 at the wall and �z + = 10 in the streamwise, wall-normal and spanwise directions, respectively, the mean and fluctuating velocity profiles appear not to change significantly. For all grids including the coarsest ones, the large near-wall turbulent structures, whose typical size are given by integral length scales and by dominant components in velocity spectra, are found to be well discretized. However, the examination of the transfer functions in the wavenumber space of the dissipation mechanisms, namely molecular viscosity and relaxation filtering, shows that a mesh spacingx + � 30 is for instance necessary to ensure the physical relevance of the LES solutions, which is in agreement with the findings of the grid-convergence study. In a second step, channel flows at friction Reynolds numbers of 350, 600 and 960 are sim- ulated. The results are very similar to reference data from the literature. Furthermore, the Reynolds number effects in channel flows are well reproduced using the present LES approach.