ライブラリ登録: Guest
TSFP DL Home アーカイブ 執行委員会

LARGE EDDY SIMULATION OF A COMPRESSIBLE MIXING LAYER WITH A TIME SELF-ADAPTIVE MULTILEVEL METHOD

Marc Terracol
ONERA, 29 av. de la Division Leclerc 92320 Chatillon, France.

Pierre Sagaut
Institut Jean Le Rond D'Alembert, UMR 7190 Université Pierre et Marie Curie 4 place Jussieu 75005 Paris, France; Laboratoire de Mecanique, Modelisation et Procedes Propres UMR CNRS 7340 Aix-Marseille Universite IMT La Jetee, Technopole de Chateau-Gombert 38, rue Frederic Joliot-Curie 13451 Marseille Cedex 13, France

Claude Basdevant
Universite Paris-Nord, LAGA UMR 7539, 99 av. J.B. Clement, 93430 Villetaneuse, France

要約

Large-Eddy Simulation (LES) allows to reduce the computational costs in the numerical simulation of turbulent flows in comparison with Direct Numerical Simulation (DNS). This reduction is obtained by a scale separation, the largest ones being directly resolved, while the smallest ones (subgrid scales) are modeled. Nevertheless, usual eddy-viscosity subgrid models have been developed in the framework of homogeneous isotropic turbulence, and are so not able to take into account in a proper way the presence of inhomogeneous subgrid scales, or backscatter. That is why LES still require the use of fine computational grids, and thus a large amount of CPU ressources. A multilevel method applied to LES is introduced here to reduce the CPU times. Flow variables are decomposed into several frequency bands, each band being associated to a computational grid in physical space. The high-frequency deterministic information from the finest levels can then directly be used on the coarse ones to get an accurate evaluation of the subgrid model, as in deconvolution-like approaches (Stoltz and Adams, 1999, Domaradzki and Yee, 2000). CPU time saving is obtained by performing the main part of the simulation on the coarse levels by freezing the smallest resolved scales (Quasi-Static approximation-Dubois et al., 1998), and performing an explicit reconstruction of these scales at certain times only. Such a strategy has previously been assessed on a quasi-steady plane channel flow configuration (Terracol et al., 2001a) by the use of a simple V-cycling strategy. Here, a particular two-level case of this method is considered for fully unsteady flows, in which a dynamic evaluation of the time during which the QS approximation remains valid is performed by a priori estimates of the small scales time variation. The method is assessed here on a fully unsteady time-developing compressible mixing layer.