Abstract Turbomachinery flows exhibit strong unsteadiness, characterized by periodic wake-related disturbances superimposed on background turbulence. The turbulence generation mechanisms in these flows involve multiple interacting scales, making accurate modeling challenging. Developing robust transition models capable of capturing the multi-modal nature of these processes is crucial for accurately predicting the overall performance of turbomachinery components. In our previous work (GT2024-128738) we employed a turbulent event recognition technique that was used on high-fidelity data of a compressor cascade under steady inflow conditions to classify individual flow features as “laminar” and “turbulent” depending on their role in transition. This allowed us to characterize features such as streaks and Kelvin-Helmholtz roll-ups in a manner that is consistent and data-driven. In this work we consider other high-fidelity scale-resolving datasets of compressor blades (NACA65) under unsteady inflow conditions. The objective of this work is to test whether “laminar” and “turbulent” classification is appropriate for this class of problems. For that purpose a multi scale analysis of the turbulence generation mechanisms is adopted. This is achieved by clustering flow scales using an unsupervised data-driven method. Following that, turbulence production for different scale clusters are examined. Data reported in the paper shows that a low dimensional multi-scale representation of turbulence generation mechanisms can be achieved for compressor flows. The approach discussed here has potential to extend the concept of two-scale laminar-turbulent kinetic energy models, especially with regards to incorporating the impact of incoming wakes into the RANS framework.
Dellacasagrande et al. (Mon,) studied this question.