The subgrid-scale estimation procedure in the physical space representation

The primary goal of this dissertation is to develop a new model for turbulence simulations and to implement it to simulate turbulent channel flow.; The subgrid scale estimation procedure for large eddy simulations (LES) is extended to the physical space representation. The procedure provides an estimate of the unfiltered velocity field appearing in the definition of the subgrid-scale stress tensor and consists of two steps. In the deconvolution step an approximate inversion of the filtering operation is performed. Subsequently, the nonlinear step is used to generate a range of subgrid-scales on a mesh two times smaller than the mesh employed for a discretization of the resolved quantities. The subgrid-scale estimation procedure is firstly developed by using one-dimensional top hat filters on uniform grids, and then it is generalized to nonuniform grids.; Another approach to simulations of the turbulent flows is obtained by evolving the estimated velocity through the truncated Navier-Stokes dynamics without the calculation of subgrid-scale stress tensor. An intermediate approach between the traditional LES and truncated Navier-Stokes dynamics is to combine them such that the subgrid-scale stress tensor needed in LES is computed from the estimated velocity advanced in parallel by the truncated Navier-Stokes equations.; The modeling approaches are evaluated by comparing results of large eddy simulations of turbulent channel flow with the corresponding results of direct numerical simulations, experiments, and other large eddy simulations.