Optimal Large -Eddy Simulation of Turbulent Channel Flow

112 p.Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 2000.An optimal LES formulation is presented, in which the subgrid model is chosen such that it minimizes the rms error of the instantaneous time-derivative of the filtered velocity. The statistics required to compute the optimal subgrid model are computed from realizations of a direct numerical simulation (DNS) of a plane channel at a friction Reynolds number of 587. Due to the limited number of realizations, estimates of the model term are global only in the streamwise and spanwise directions and local in the wall-normal direction. Measurements of the a priori error in the estimated model term suggest that the subgrid is predominantly stochastic. This is true regardless of whether the model term is estimated directly, or the estimation is carried out in terms of the subgrid stress. Furthermore, the a priori error is equally large whether or not a wall-normal filter is applied. A large-eddy simulation using a model term based on a direct estimate of the model term underpredicts the mean velocity profile, while grossly overpredicting root mean-square velocity fluctuations. Detailed analysis of the subgrid energy transfer reveals, that two individual contributions have to be estimated separately. These contributions are the subgrid dissipation and subgrid transport. Subgrid dissipation is a local phenomenon and can be approximated accurately using a local estimate. Subgrid transport on the other hand is a global phenomenon that a local estimate cannot capture. With this in mind, the estimation procedure is revised to generate separate estimates of subgrid dissipation and subgrid transport, which are then used to construct an estimate of the subgrid force. An optimal LES based on individual estimates of subgrid dissipation and subgrid stress is shown to accurately predict second-order statistics.U of I OnlyRestricted to the U of I community idenfinitely during batch ingest of legacy ETD