Abstract: Advanced aircraft design puts forward higher requirements for the boundary layer separation simulation capability of the CFD method. The traditional Reynolds Averaged Navier-Stokes equations (RANS) vortex viscosity model has a low reliability of separation flow prediction due to theoretical defects in modeling and structural level. Because the Reynolds stress model does not model the Reynolds stress and its component relationships, it has natural theoretical advantages in turbulent non-equilibrium, rotation, and Reynolds stress anisotropy. To verify and confirm the prediction ability of the Reynolds stress model for typical aviation separation flows, based on the SSG/LRR-g model, cases of the NACA4412 airfoil high angle of attack separation, M6 wing transonic separation and F6 wing body junction zone separation flow are analyzed. The adaptability of the Reynolds stress model to the prediction of typical aviation separation flow such as reverse pressure gradient, shock-induced separation, and secondary flow separation is discussed. A comparison of the simulation result with that of the k-ω Shear Stress Transport (SST) model finds that the Reynolds stress model significantly improves the simulation accuracy for key features such as separation bubble size, velocity distribution, Reynolds stress distribution, and shock wave position compared to the eddy viscosity model, which basically verifies the advantage of the Reynolds stress model. The model can obtain the correct flow characteristics in the corner flow of the wing body junction area and the three-dimensional strong shock induced separation. However, the SST model basically fails in this type of flow, showing that the Reynolds stress model is more turbulent in the typical aviation separated flow.It is also found that the Bradshaw hypothesis contained in the k-ω SST model severely affects the accuracy of the model's prediction when the three-dimensional shock induced separation is strong, which is the main reason for the deviation of the prediction results from the experiment. In addition, the Reynolds stress model and the possible improvement directions of the eddy viscosity model are proposed based on the calculation results and model structure.

Key words: turbulence model, separated flow, Reynolds stress model, eddy viscosity model, shock induced separation, secondary flow