Large eddy simulation (LES) and Reynolds-averaged Navier-Stokes (RANS) computations have been used for prediction of a spatially developing three-dimensional turbulent boundary layer over a bump swept at 45 deg with respect to the upstream flow. Subgrid-scale stresses in the LES were parameterized using the dynamic eddy viscosity model. Reynolds stresses in the RANS calculations were closed using the v2-f model and Spalart-Allmaras one-equation model. In the calculation, a zero-pressure gradient, statistically two-dimensional boundary layer at momentum thickness Reynolds number 3.8×103 is introduced one-half chord length upstream of the onset of curvature. The flow is statistically homogeneous along the coordinate parallel to the bump axis and is subject to combined perturbations in streamwise pressure gradient, spanwise pressure gradient, and surface curvature. The turning angle of the wall shear stress measured with respect to the upstream flow changes sign twice due to the alternating spanwise pressure gradient, with a maximum of more than 45 deg near the trailing edge. No-slip conditions were imposed on solid boundaries in RANS, whereas algebraic approximate boundary conditions were applied in the LES to model the near-wall flow. Mean wall shear stresses, necessary to close the approximate boundary conditions, were supplied from either experimental measurements or a separate RANS calculation. In general, the agreement between simulation and experiment in the present work is comparable to that previously obtained for the two-dimensional boundary layer at zero sweep angle. The mean flow in accurately predicted using both techniques, with some discrepancy occurring in prediction of the mean crossflow in the LES. Second-order statistics in the LES are in good agreement with measurements; RANS predictions of turbulence kinetic energy are slightly less accurate.
ASJC Scopus subject areas
- Aerospace Engineering