A comparison of second- and sixth-order methods for large eddy simulations (Q1598035)

The main purpose is the comparison of various finite difference discretizations of inviscid fluxes applied in the large eddy simulation of turbulent compressible flows. The schemes are a second-order mixed central-upwind scheme (advective upstream splitting method, AUSM) and higher-order compact Padé methods of up to sixth order in which a compact high-frequency filter is applied. Either the dynamical subgrid scale model of Germano and Lilly or no model was used for modeling the Reynolds stress. After a description of numerical methods, they are validated at a flow past a cylinder for Reynolds number 100. In this example, the differences between various methods are rather small. As the second example, turbulent channel flows are computed whose results agree well with results from the literature. The third example is the simulation of a planar turbulent jet. One of the turbulent channel flow solutions is taken as inflow. The results are compared to a reference solution computed with AUSM method on a fine grid without model for Reynolds stress. The influence of the discretization, the grid resolution and the model for Reynolds stress are studied by computing several statistical variables. It turns out that the results obtained with AUSM scheme and with higher-order compact finite difference schemes are rather similar. In all cases, the impact of the model for Reynolds stress on the distribution of statistical variables is negligible. The costs of using the higher-order compact schemes were about 150 than of the AUSM scheme.