In order to improve turbomachinery performances it is necessary to correctly predict secondary and transitional flows inherent to this confined environment. Secondary flows can cause additional loss and operating domain reduction. For example, tip-leakage vortex flow or corner separation flow on high pressure compressor blades lead to total pressure loss limiting turbomachinery efficiency and favoring emergence of hazardous axial instability (surge).

Secondary flows often exhibit strong fluctuating behaviors and flow separations which make their prediction quite challenging, often found inaccurate with current turbulence modeling tools used during design process. One reason lies in the error on turbulent fluctuation predictions when they are entirely modeled by statistical averaging (RANS - Reynolds Average Navier-Stokes). However, more accurate methods such as Large Eddy Simulations (LES) are difficult to use at high Reynolds number due to excessive computation, treatment and storage costs.

A trade-off can be found in hybrid RANS/LES methods such as ZDES (Zonal Detached Eddy Simulation) method developed at ONERA. Through the main operating mode of ZDES method, boundary layers are treated by RANS modeling on their whole thickness to avoid near-wall excessive cost of LES method which is only used away of the walls. The interface between those two sub-methods is continuous. Its RANS sub-method is based on SA turbulence model.

As demonstrated by W. Riéra (Ph.D. thesis, Evaluation of the ZDES method on an axial compressor: analysis of the effects of upstream wake and throttle on the tip-leakage flow, Ecole Centrale de Lyon, 2014) the SA turbulence model used in the RANS sub-method of the current ZDES method makes difficult to predict turbomachinery flows at limit operating range because of its poor behavior near numerical surge line. A major reason is that RANS SA model tends to predict too massive or simply false flow separations, which is critical for such flows with corner flow separation. A suggested remedy is to re-based the ZDES method on a k-ω Menter turbulence model which present better behavior regarding progressive flow separation prediction.

Several ZDES k-ω Menter method formulations have been assessed on three academic cases (a mixing layer flow, a backward facing step flow and a circular cylinder flow at Re = 3900, AIAA Aviation 2017). One of these formulations has been selected based on its capability (i) to allow the emergence and the correct development of Kelvin-Helmholtz instabilities and (ii) to ensure that boundary layers are treated by RANS modeling in their entire thickness.

Alternatively to hybrid RANS/LES methods, upgraded URANS approaches are interesting tools for resolving a variable fraction of the larger scales of turbulent motions, allowing a better representation of unsteady flows than URANS methods. Among them, the Scale Adaptive Simulation (SAS) approach of has been implemented in ONERA's elsA solver with an improvement dealing with the development of instabilities within mixing layers.

This study focuses on the validation of the selected ZDES k-ω Menter formulation and the SAS approach on the same configuration as W. Riéra ZDES SA simulation i.e. a flow simulation of a realistic rotor of a high-pressure compressor with incoming stator wakes. Simulation is carried out on the same mesh and numerical framework as W. Riéra ones for appropriate comparison with their ZDES SA results and with experimental data.