Hybrid RANS-LES Simulation of the MTU-T161 LPT Profile - Forcing LES in the Endwall Boundary Layer by Grid Refinement

Abstract

Seamless hybrid RANS-LES methods provide a practical and cost-effective approach to turbulence modeling as they bridge the gap between the precision of Large Eddy Simulations (LES) and the affordability of RANS-modeling. These methods employ RANS turbulence models to approximate small-scale turbulence in attached boundary layers, while resolving larger eddies in detached flow regions and in the free stream with LES capabilities. However, the accuracy in predicting flows in RANS regions is dependent on the model’s calibration, which can lead to unsatisfying predictions when used beyond their calibration scope, e.g., in secondary flow regions. This paper proposes a solution to this limitation by utilizing the grid dependency of the Improved Delayed Detached Eddy Simulations (IDDES) formulation: The aim is to reduce the effects of RANS modeling and induce scale-resolving capabilities in key areas by means of grid refinements. This approach effectively simulates a zonal hybrid method with a non-zonal basis model. The advantage of this approach, thereby, lies in the utilization of only one basis model. For this reason, a transfer of turbulence data between turbulence modeling RANS and turbulence resolving LES regions is avoided. The MTU-T161 low-pressure turbine cascade, at a Reynolds number of 90,000, serves as the test case for this study. The study focuses on the impact of grid refinements in the side wall region and expands upon previous work conducted in the midspan of the turbine. It demonstrates that the approach leads to a decrease in the effective hybrid length scale, thereby reducing the influence of the underlying turbulence model. Ultimately, improvements in the prediction of end wall boundary layers and wake losses are achieved.