In this dissertation, turbulence processes and flow loss mechanisms in turbomachinery cascade flows are analyzed through the application of high-fidelity numerical methods. Detailed simulations are conducted of a highly loaded low-pressure turbine cascade (Re = 90,000) and a high-pressure compressor cascade (Re = 492,000) to identify and quantify the key physical mechanisms contributing to turbulence and flow losses. A particular focus is placed on the interaction between free-stream turbulence (FST), transition processes, and flow loss generation in cascade flows under off-design conditions. These two test cases were simulated in a three-dimensional configuration, including sidewalls, using high-fidelity methods. However, the investigation primarily focuses on the mid-span flow behavior. The numerical computations are validated against experimental measurements of blade pressure distributions and wake flow losses. Advanced statistical techniques are applied to characterize the dominant turbulence structures. Spectral and modal analyses reveal that in the low-pressure turbine (LPT) cascade, low-frequency structures are associated with transitional and separated flow, whereas high-frequency components correspond to the development of turbulent structures during reattachment. The high-pressure compressor (HPC) cascade exhibits dominant high-frequency structures within the boundary layer, driven by the strong interaction between FST and pressure-gradient-induced separation. Quadrant analysis provides additional insights into turbulent exchange processes within the boundary layer. In the boundary layer flow of both the LPT and HPC cascades, ejection and sweep events are the biggest contributors to turbulent exchange processes. In the LPT cascade, increased FST-intensity TI = 8% enhances boundary layer mixing, leading to a 20% shorter separation length and a 13% reduction in total flow losses. The main flow loss sources are found in the separated shear layer and reattachment area, where turbulent kinetic energy accumulates. Reynolds-averaged Navier-Stokes (RANS) simulations fail to capture the transition dynamics accurately and underestimate the integral wake flow losses by 32% compared to large eddy simulations (LES). In the HPC cascade, an increased FST-intensity (TI = 6%) induces an earlier transition of about 10% axial length, however, the total wake flow losses remain largely unchanged due to similar wake mixing characteristics across different FST-intensity cases. The most significant flow losses occur in the boundary layer near the suction side of the blade. The RANS model fails to capture the laminar separation accurately, overpredicting the integral wake flow losses by 55% relative to LES. The findings provide valuable insights for validating and improving low-order RANS turbulence models. By offering a detailed analysis of turbulent structure formation, this work contributes to enhancing turbomachinery flow modeling, ultimately leading to efficiency improvements and performance optimization of turbines and compressors. The study demonstrates that high-fidelity simulations can be used as virtual experiments to address deficiencies in existing models and complement targeted experimental investigations.
Nima Fard Afshar (Wed,) studied this question.