Predicting Separation and Transitional Flow in Turbine Blades at Low Reynolds Numbers

Abstract

There is increasing interest in design methods and performance prediction for aircraft engine turbines operating at low Reynolds numbers. In this regime, boundary layer separation may be more likely to occur in the turbine flow passages. For accurate CFD predictions of the flow, correct modeling of laminar-turbulent boundary layer transition is essential to capture the details of the flow. To investigate possible improvements in model fidelity, CFD models were created for the flow over two low pressure turbine blade designs. A new three-equation eddy-viscosity type turbulent transitional flow model originally developed by Walters and Leylek was employed for the current RANS CFD calculations. Previous studies demonstrated the ability of this model to accurately predict separation and boundary layer transition characteristics of low Reynolds number flows. The present research tested the capability of CFD with the Walters and Leylek turbulent transitional flow model to predict the boundary layer behavior and performance of two different turbine cascade configurations. Flows over the Pack-B turbine blade airfoil and the midspan section of a typical low pressure turbine (TLPT) blade were simulated over a Reynolds number range of 15,000–100,000, and predictions were compared to experimental cascade results. The turbulent transitional flow model sensitivity to turbulent flow parameters was investigated and showed a strong dependence on free-stream turbulence intensity with a second order effect of turbulent length scale. Focusing on the calculation of the total pressure loss coefficients to judge performance, the CFD simulation incorporating Walters and Leylek’s turbulent transitional flow model produced adequate prediction of the Reynolds number performance for the TLPT blade cascade geometry. Furthermore, the correct qualitative flow response to separated shear was observed for the Pack-B blade airfoil. Significant improvements in performance predictions were shown over predictions of conventional RANS turbulence models that cannot adequately model boundary layer transition.