| description abstract | This study assesses the accuracy of ansysfluent 19.2, a commonly employed general-purpose finite volume solver, in the context of wall-modeled large-eddy simulation for turbulent channel flow at a moderate Reynolds number, Reτ=2000. The sensitivity of the solution to variations in grid resolution, aspect ratio, grid arrangement (collocated versus staggered), and subgrid-scale (SGS) model is analyzed and contrasted to results from a corresponding direct numerical simulation (DNS) and a mixed pseudospectral and finite differences solver. Results indicate good convergence of first- and second-order statistics from the staggered grid setups as the grid is refined, whereas no clear trend is observed in cases with collocated grid setups. Velocity spectra show a lack of an apparent inertial range trend and rapid decay of energy density at high wavenumbers, with a spurious energy pile-up near the cutoff wavenumber indicating the presence of unphysical oscillations in the velocity fields. Grid refinement strengthens such oscillations in collocated grid setups and reduces them in staggered grid setups. Two-point streamwise velocity autocorrelation maps reveal an underprediction of turbulent structure size. In contrast, cross-stream autocorrelations agree with corresponding curves from direct numerical simulation, showing signatures of alternating high- and low-momentum streaks in the logarithmic layer. | |
