Grid Effects for Wall-Modeled Large-Eddy Simulation of Flow over a Multi-Element Airfoil

Wall-modeled large-eddy simulations (WMLESs) of turbulent flow over a two-dimensional 30P30N multi-element airfoil at various angles of attack are performed using an unstructured finite-volume solver to document the sensitivity of the predictions to the computational grid. The stowed chord-based Reynolds number ([Formula: see text]) is 9 million, and the angle of attack ([Formula: see text]) is varied between 8 and [Formula: see text] with the expected stall angle around [Formula: see text]. We first analyze the predictions with baseline near-wall isotropic and mildly anisotropic grids. The predicted lift, wall pressure, near-wall velocity profiles, and the location of onset of resolved turbulence are sensitive to the near-wall grid anisotropy. The lift is better predicted at lower angles with the isotropic grid and at higher angles with the anisotropic grid. We then explore the sensitivities at lower angles of attack to independently vary the streamwise and wall-normal grid anisotropies to isolate their effects and to off-wall and near-wall grid refinement. The effects of streamwise grid refinement near the leading and trailing edges and off-body grid refinement away from the surface appear to have minimal effects on the predictions for this flow configuration. However, the wall-normal grid resolution in the vicinity of the surfaces has a dominant effect on the results. Our results suggest a monotonic trend for lift and other quantities based on the absolute value of the first wall-normal spacing, which coincides with the WMLES exchange location in our solver. Finally, for the same grid resolution away from the wall, varying the wall-normal spacing near the surface based on the angle of attack may lead to more optimal grid counts for WMLES.