PREDICTION OF THE HIGH-REYNOLDS-NUMBER FLOW OVER A 2-DIMENSIONAL BUMP

Authors
Citation
Xh. Wu et Kd. Squires, PREDICTION OF THE HIGH-REYNOLDS-NUMBER FLOW OVER A 2-DIMENSIONAL BUMP, AIAA journal, 36(5), 1998, pp. 799-808
Citations number
39
Categorie Soggetti
Aerospace Engineering & Tecnology
Journal title
ISSN journal
00011452
Volume
36
Issue
5
Year of publication
1998
Pages
799 - 808
Database
ISI
SICI code
0001-1452(1998)36:5<799:POTHFO>2.0.ZU;2-K
Abstract
Large eddy simulation (LES) has been applied to prediction of the high -Reynolds-number boundary layer flowing over a bump. Approximate bound ary conditions were used in the LES to model the wall layer in which t he instantaneous wail stress is correlated with the velocity at the fi rst layer of grid points. Two formulations of the approximate boundary conditions were employed. In the first formulation, a constrained rel ation was used in which the mean wall shear stress is specified a prio ri either from experimental measurements or from a separate solution o f the Reynolds-averaged Navier-Stokes (RANS) equations. In the second formulation, an unconstrained relation, which is based on an instantan eous power-law velocity profile, was used. Calculations were also perf ormed in which the wall stress was computed directly, i.e., as if the near-wall how were resolved. In the region resolved by the LES, the st rong distortions of the mean flow, streamwise fluctuations, and turbul ent shear stress are reasonably well predicted, and the relatively rap id recovery downstream of the trailing edge is also captured. However neither LES nor RANS reproduces the plateau in skin friction measured upstream of the bump summit (attributed to early relaminarization in t he experiments). LES predictions of the mean flow and turbulence inten sities using the dynamic eddy viscosity model are relatively insensiti ve to the particular formulation of approximate boundary conditions. H owever, in calculations without a subgrid model, LES predictions exhib it large errors compared to experiments, and the quantitative levels o f such errors are also sensitive to the choice of mall-layer model.