Predicting Endwall Corner Separation in Compressor Cascades Using RANS

Corner separations are a complex physical phenomenon that significantly affects the off-design operation of axial compressors, contributing to instability and limiting their operating range. At the same time, most Reynolds-averaged Navier–Stokes (RANS) models struggle to capture the interaction of endwall and airfoil boundary layers accurately, leading to incorrect predictions of aerodynamic performance, local blade loading, and instability mechanisms. This article evaluates the predictive capabilities of state-of-the-art, steady-state RANS modeling approaches for the aerodynamics of the linear compressor cascade of the École Centrale de Lyon. The evaluation is based on comprehensive studies of the computational fluid dynamics setup, focusing on the tripping process of the blade boundary layers, endwall treatment, and inflow boundary conditions. An optimized RANS setup is presented, featuring a meshed trip for the blade boundary layers, fully-turbulent endwalls, and a corrected inflow angle. This setup yields accurate predictions of static pressure distributions at various spanwise positions, including in direct proximity to the endwall, and a precise representation of the total-pressure losses induced by corner separations downstream of the cascade. Boosting the momentum of the fluid in endwall boundary layers, either by prescribing fully-turbulent endwall boundary layers or neglecting endwalls in the inflow conditions, improves the prediction of loaded operating points, enhances the blade loading, and makes the suction-sided boundary layer more resistant toward diffusion. A comparison with other RANS- and large eddy simulations-based results of this test case from the literature highlights the improved prediction capabilities of the proposed strategies. This study identifies the limitations of the RANS simulations concerning corner separation-induced vortex trajectories and the pitchwise spread of loss distributions downstream of the blade, providing a foundation for future improvements in the RANS modeling.