Predicting Transition in Turbomachinery, Part II - Model Validation and Benchmarking

摘要

The ability to predict boundary layer transition locations accurately on turbomachinery airfoils is critical both to evaluate aerodynamic performance and to predict local heat-transfer coefficients with accuracy. In state-of-the-art Reynolds Averaged Navier-Stokes (RANS) simulations used to predict flowfields on turbomachinery airfoils, boundary layers are often assumed to be turbulent over the entire airfoil surface. Consequently, losses are not accurately predicted, particularly in the case of stalled airfoils. Here we report on an effort to include empirical transition models developed in Part I of this report in a RANS solver. To validate the new models, a two-dimensional design optimization was performed to obtain a pair of Low-Pressure Turbine (LPT) airfoils with the objective of increasing airfoil loading by 25%. Subsequent experimental testing of the resulting two new airfoils confirmed pre-test predictions of both high and low Reynolds number loss levels. In addition, the accuracy of the new transition modeling capability was benchmarked with a number of legacy cascade and LPT rig data sets. Good agreement between measured and predicted profile losses was found in both cascade and rig environments. However, use of the transition modeling capability has elucidated deficiencies in typical RANS simulations that are conducted to predict component performance. Efficiency-versus-span comparisons between rig data and multi-stage steady and time-accurate LPT simulation results indicate that loss levels in the endwall regions are significantly under-predicted. Possible causes for the under-predicted endwall losses are discussed as well as suggestions for future improvements that would make RANS-based transitional simulations more accurate.