Unsteady Diffuser Flow in an Aeroengine Centrifugal Compressor
Rising fuel costs and growing environmental concerns have forced gas turbine engine manufacturers to place high value on reducing fuel burn. This trend has pushed compressor technology into new design spaces that are not represented by historical experience. Specifically, centrifugal compressor diffusers are trending toward higher pressure recovery and smaller diameters. The internal fluid dynamics in these new flow regimes are not well understood and additional study is necessary. This work outlines detailed experimental and numerical observations of the flow field through a vaned diffuser for aeroengine applications.
The experimental data consist of extensive Laser Doppler velocimetry measurements of the unsteady velocity field from the impeller trailing edge through the majority of the diffuser passage. These data were obtained non-intrusively and yielded all three components of the velocity vector field at approximately 2,000 geometric points. The correlation between fluctuations in the three velocity components were also observed at several key locations to determine the components of the local Reynolds stress tensor.
These data indicated a jet/wake profile at the impeller exit represented by a consistent velocity deficit region from hub to shroud adjacent to the suction surface of the passage. This region was more prevalent adjacent to the splitter blade. The unsteady fluctuations due to the propagation of the jet and wake through the diffuser passage persist to 40% downstream of the throat. A complex secondary flow field was also observed with large axial velocities and a passage-spanning vortex developing through the diffuser passage. The velocity data and total-pressure data indicated a region of flow separation developing along the pressure surface of the vane near the hub due to the unsteady propagation of the jet and wake flow through the diffuser. Although this region was stable in time, its development arose due to unsteady aspects of the flow. Finally, the strong interconnection between the jet and wake flow, unsteady fluctuations, secondary velocities, incidence, and flow separation was demonstrated.
Computationally, a “best-practice” methodology for the modelling of a centrifugal compressor was developed by a systematic analysis of various turbulence models and many modelling features. The SST and BSL-EARSM turbulence models with the inclusion of fillets, surface roughness, and non-adiabatic walls was determined to yield the best representation of the detailed flow development through the diffuser in steady (mixing-plane) simulations. The accurate modelling of fillets was determined to significantly impact the prediction of flow separation with the SST turbulence closure model. Additionally, the frozen rotor approach was shown to not accurately approximate the influence of unsteady effects on the flow development.
Unsteady simulations were also compared to the detailed experimental data through the diffuser. The BSL-EARSM turbulence model best matched the experimentally observed flow field due to the SST model’s prediction of flow separation in the shroud-pressure side corner of the passage. In general, lower levels of axial velocity were predicted numerically that resulted in less spanwise mixing between the endwall and freestream flows. Additionally, the turbulent kinetic energy levels in the computational results showed little streamwise variation through the vaneless and semi-vaneless space. The large variation observed experimentally indicated that the production and dissipation of turbulent kinetic energy through this region was not accurately predicted in the two turbulence models implemented for the unsteady simulations.