Jakirlic, S., Kutej, L., Unterlechner, P., and Tropea, C., "Critical Assessment of Some Popular Scale-Resolving Turbulence Models for Vehicle Aerodynamics," SAE Int. J. Passeng. Cars - Mech. Syst. 10(1):235-250, 2017, doi:10.4271/2017-01-1532.
Some widely-used scale-resolving turbulence models are comparatively assessed in simulating the aerodynamic behavior of a full-scale AUDI-A1 car configuration. The presently considered hybrid RANS/LES (RANS – Reynolds-Averaged Navier-Stokes; LES – Large-Eddy Simulation) models include the well-known DDES (Delayed Detached-Eddy Simulation) scheme and two further variable-resolution formulations denoted by PANS (Partially-Averaged Navier-Stokes; Basara, 2011) and VLES (Very LES; Chang et al., 2014). Whereas the DDES method represents the originally proposed formulation based on the one-equation Spalart-Almaras model (Spalart et al. 2006), whose RANS/LES interface position is directly correlated to the underlying grid resolution, the other two models represent ‘true’ seamless formulations, providing a smooth transition from Unsteady RANS to LES in terms of a dynamic “resolution parameter” variation. The latter parameter is evaluated by contrasting the length scale related to the residual turbulence of both PANS and VLES methods to the grid spacing. The dynamics of residual motion in both methods is modelled by a four equation model (Hanjalic et al., 2004). All computations are performed by the OpenFOAM code. The PANS and VLES formulations, in conjunction with the ‘hybrid wall functions’ used for the wall treatment, were implemented by the authors. The car configuration considered accounts for mirrors, detailed underbody accommodating the exhaust system, as well as the rotating wheels including brake discs and rim details. The solution domain representing a regular hexahedron was meshed by two grids consisting approximately of 31 and 62 million cells in total. The results representation includes detailed time-dependent mean flow and turbulence fields, surface pressure distribution and the resulting drag and lift force coefficients. The results’ analysis is performed by discussing the models’ predictive capabilities along with the available experimental results.