Separated Turbulent Boundary Layer Under Unsteady Adverse Pressure Gradients: DNS and RANS

Abstract

Predicitve capabilities of Reynolds Averaged Navier-Stokes (RANS) techniques have been assessed using SST k–ω model and Spalart-Allmaras model by comparing its results with direct numerical simulation (DNS) results. It has been shown that Spalart-Allmaras and SST k–ω model predict an earlier separation point and a bigger recirculation bubble as compared to the DNS result. Velocity profiles predicted by RANS for both models closely match with DNS results for the steady adverse pressure gradient case. However, the RANS fail to predict correct velocity profiles for unsteady adverse pressure gradients not only for inside the bubble but also after the reattachment zone. To provide the backgrounds for improving RANS models, these differences are explained with Reynolds stress and eddy viscosity which differ between the steady and unsteady adverse pressure gradient RANS cases.