Predicting Separation and Transitional Flow in Turbine Blades at Low Reynolds Numbers—Part II: The Application to a Highly Separated Turbine Blade Cascade Geometry

Abstract

There has been a need for improved prediction methods for low pressure turbine (LPT) blades operating at low Reynolds numbers. This is known to occur when LPT blades are subjugated to high altitude operations causing a decrease in the inlet Reynolds number. Boundary layer separation is more likely to be present within the flowfield of the LPT stages due to increase in the region adverse pressure gradients on the blade suction surface. Accurate CFD predictions are needed in order to improve design methods and performance prediction of LPT stages operating at low Reynolds numbers. CFD models were created for the flow over two low pressure turbine blade designs using a new turbulent transitional flow model, originally developed by and (2004, “A New Model for Boundary Layer Transition Using a Single Point RANS Approach,” ASME J. Turbomach., 126(1), pp. 193–202). Part I of this study applied Walters and Leylek’s model to a cascade CFD model of a LPT blade airfoil with a light loading level. Flows were simulated over a Reynolds number range of 15,000–100,000 and predicted the laminar-to-turbulent transitional flow behavior adequately. It showed significant improvement in performance prediction compared to conventional RANS turbulence models. Part II of this paper presents the application of the prediction methodology developed in Part I to both two-dimensional and three-dimensional cascade models of a largely separated LPT blade geometry with a high blade loading level. Comparisons were made with available experimental cascade results on the prediction of the inlet Reynolds number effect on surface static pressure distribution, suction surface boundary layer behavior, and the wake total pressure loss coefficient. The kT-kL-ω transitional flow model accuracy was judged sufficient for an understanding of the flow behavior within the flow passage, and can identify when and where a separation event occurs. This model will provide the performance prediction needed for modeling of low Reynolds number effects on more complex geometries.