SINGLE PASSAGE DETACHED EDDY SIMULATION OF A CENTRIFUGAL COMPRESSOR STAGE USING THE TIME TRANSFORMATION METHOD

A subsonic rear stage centrifugal compressor (designed as the last compressor stage of an aero-engine following a multistage axial compressor) was simulated as a single passage using Detached Eddy Simulation (DES) and circumferential time inclination to enforce periodic boundary conditions according to the machine rotor-stator pitch ratio. The transient averaged statistics obtained with DES are compared to those of a standard steady mixing plane SST RANS simulation, an unsteady circumferential time-inclination SST URANS simulation and two-component non-intrusive Laser Doppler Velocimetry (LDV) measurements conducted in a centrifugal compressor test rig. The LDV and DES were carried out at the design point of the compressor. Significant discrepancies were found particularly in the unloading at the trailing edge of the impeller and the balancing of the diffusion throughout the stage, however the overall stage performance predictions were strikingly similar between the various turbulence modelling methods indicating that they are not particularly sensitive to the observed aerodynamic differences. The discrepancies observed do affect the ratio of loading between the impeller and diffuser, and could become exaggerated particularly at off-design conditions. when components are not as well matched. At design, the DES showed a 1.6% lower total-to-total pressure ratio in the impeller compared to RANS (1.4% compared to URANS), and 0.9% lower in stage total-to-total pressure ratio (0.2% compared to URANS). Trailing edge base pressure distributions show a larger deficit in the RANS wake in comparison to the DES, and pressure distributions show strong blade-to-blade variations in the steady RANS results in the near-trailing edge region, whereas the averaged DES results show a much faster diffusion of the blade-to-blade and spanwise gradients which was found to be in agreement with LDV velocity field measurements. The higher diffusion in the DES is due to higher Reynolds stresses predicted in this area compared to standard RANS.