A modified SSG/LRR-omega Reynolds stress model for predicting bluff body aerodynamics

Abstract

This work focuses on a complex turbulent flow around a blunt body by predicting vortical structures, separation/re-attachment locations and velocity/turbulent kinetic energy profiles relying on a modified SSG/LRR-! Reynolds Stress Model (RSM). The investigated physical problem is in the centre of research interest, because the Reynolds-Averaged Navier-Stokes (RANS) turbulence models and Large Eddy Simulation (LES) approaches usually fail to reproduce the physically correct flow field. Due to the fact that there is lack of knowledge on the SSG/LRR- hybrid RSM closure model proposed by Cecora et al. [1], therefore this work has been devoted to further investigate the overall numerical behaviour of a modified hybrid RSM approach. The advantage of these RSM closure models is to take into account the anisotropy of Reynolds stresses caused by the streamline curvature, which has importance in maintaining the quasi non-diffusive nature of turbulent vortices. The numerical model implementation has been verified through the classical test case of a turbulent flow over a flat plate at zero pressure-gradient on a sequence of nested grids. A preliminary analysis performed on this benchmark problem showed that the hybrid model is capable of predicting the near-wall turbulence in the fully-developed boundary layer with a reasonable accuracy. It has also been observed that the SSG/LRR- adopts a free-stream independence feature of the specific dissipation rate from the Menters [2] k-! SST model, thus the turbulent quantities in the near-wall exhibit almost negligible sensitivity respect to the specific dissipation rate outside of the shear layer. In this work, a modified hybrid SSG/LRR-! RSM closure in conjunction with a simplified diffusion model has been proposed, which has been investigated through the classical cubic obstacle problem in a three-dimensional channel flow of Martinuzzi and Tropea [3]. An increased accuracy and demanding computational time have been observed by employing the modified SSG/LRR-! formulation, which could still be advantageous, because it unifies the favourable features of two distinct differential Reynolds stress models producing accurate results in the near-wall and the shear layer regions.