Spectral domain-decomposition methods for LES of turbulence in wall-bounded flows.
Sangmo Kang · Deep Blue (University of Michigan) · 1996
In this study, we address various issues related to large-eddy simulation (LES) of transitional and turbulent flows in wall-bounded geometries. Large-eddy simulations of transitional and turbulent channel flow are performed to investigate the performance of various existing dynamic subgrid-scale models and filtering techniques. In addition, a grid-embedding strategy using nonconforming spectral domain-decomposition methods is developed to allow for clustering of grid points in the near-wall region. The dynamic subgrid-scale models investigated include the dynamic Smagorinsky model (DSM) of Germano et al. (1991) and the dynamic two-component model (DTM) of Ansari (1993) using a sharp Fourier cutoff filter, as well as the dynamic Smagorinsky model (DSM) of Germano et al. (1991); the dynamic mixed model (DMM) of Zang et al. (1993) and the dynamic two-parameter model (DTPM) of Salvetti et al. (1995) using a box filter. Results from DTM and DSM using a sharp Fourier cutoff filter show that DTM is more successful than DSM in capturing the statistics and the structure of the flow. DTM also has the advantage that the model is 'local' and unlike DSM does not require any averaging of the model constants in time or over the homogeneous flow directions. This makes DTM particularly attractive for LES of complex engineering flows. Results from DMM, DSM and DTPM using a box filter show DMM and DTPM to be superior to DSM, in agreement with earlier investigations. However, in all cases the predicted statistics are found to be inferior to those obtained using a sharp Fourier cutoff filter. In general, the use of a box filter with the above models results in an underprediction of the wall-shear stress, and turbulence statistics and structure in poor agreement with DNS data. One of the main unresolved issues in LES of wall-bounded flows is the requirement of high resolution in the near-wall region. Since most of the production of turbulence occurs in the near-wall region, accurate treatment of this region is critical to the success of LES. A grid-embedding strategy, using nonconforming spectral domain-decomposition methods, is developed to address this limitation. This method provides an efficient way of clustering grid points in the near-wall region. The spectral patching collocation method suggested by Orszag (1980) and developed by Zanolli (1987) and Israeli et al. (1994) are used in the grid-embedding. Results of LES using the grid-embedding strategy are found to be in good agreement with those obtained by LES using the global spectral method. Overall, the results show that grid-embedding strategies using the spectral domain-decomposition method provide an efficient method for resolving the near-wall region. These results of this study should enhance the utility of LES in the computation of complex flows of engineering interest.