| Summary: | This is concerned with an automated adaptive mesh design approach for Large Eddy Simulation (LES) of turbulent flows. Based on a dynamic moving mesh partial differential equation (MMPDE), a fixed number of grid points is redistributed according to statistical quantities of interest (QoI) selected to capture certain mean flow properties. Physically motivated LES-specific QoI, as the time-averaged gradient of streamwise velocity and the production rate, as well as more general QoI derived from the dual weighted residual method (DWRM) for time-averaged statistics are investigated for a flow over periodic hills with Re=10595$Re=10\,595$. Special emphasis is put on optimizing the grid adjustment phase. It is mainly determined by defining the length of the time interval for computing new time-averaged QoI, the number of time steps to solve the MMPDE, and the overall number of grid adjustment steps. The modifications proposed are nearly auto-adaptive with respect to the chosen QoI and lead to a reduction of the CPU time by more than one order of magnitude compared to the standard approach used so far. On the other hand, the DWRM not only provides suitable QoI to steer the grid movement, but also can be understood as a rigorous error analysis to assess the quality of numerical and subgrid modelling contributions of an LES. The numerical results compared to a highly resolved LES reference solution show the high potential of moving mesh methods to efficiently improve the resolution of turbulent flow features.
|
|---|