ASSESSMENT OF RANS AND SBES METHODS FOR THE PREDICTION OF CORNER SEPARATION IN AXIAL FLOW COMPRESSORS

Abstract

Corner separation is a common three-dimensional (3D) separated flow presented at the junction of the blade suction surface and endwall within axial-flow compressors. One commonly used Reynolds-Averaged Navier-Stokes (RANS) turbulence model in turbomachinery Computational Fluid Dynamics (CFD) simulations, Menter's Shear Stress Transport (SST) model, usually struggles to predict corner separation and its impact on compressor performance with reasonable accuracy. To improve RANS modelling accuracy for corner separation, efforts have been focused on the modifications of SST model in order for more of the endwall flow physics to be captured, such as the increased turbulence non-equilibrium and anisotropy due to the generation of skew-induced streamwise vortices in the endwall secondary flow. In this paper, the effect of the non-equilibrium and anisotropic modifications on the prediction of corner separation is evaluated using the results of the SST model and its variants. A high-fidelity time-accurate turbulence statistics database constructed by a newly developed hybrid RANS-Large Eddy Simulation (RANS-LES) method, Stress-Blended Eddy Simulation (SBES), is then used to feedback on RANS modelling. The investigation of the scale resolved corner flow physics facilitates understanding the physical reasons underlying the increased modelling accuracy of corner separation, in the meantime it reveals the modelling deficiencies that still exist.