Development and Validation of an Isolated Unsteady Diffuser Computational Model for a Vaned Centrifugal Compressor

Abstract Recent advancements in computational technologies have greatly increased their ability to be applied to complex problems in aeroengine design. Computational Fluid Dynamics (CFD) is being applied to increasingly complicated systems, including centrifugal compressors, which are known for complex flow structures due to the compound curvature inherent in the flowpath. Modeling these flow fields is normally accomplished using mixing planes between the impeller and the diffuser, which are convenient due to their quick convergence and turnaround time needed for design iterations. However, the simplification of circumferentially uniform flow entering the diffuser of a centrifugal compressor may lead to incorrect design choices. In fact, the jet-wake characteristic of the impeller exit flow field leads to significantly unsteady nonuniform flow entering the diffuser. While incorporating unsteady effects allows for greater accuracy, it is frequently not practical to run full-stage, unsteady simulations due to the significantly increased computational cost, especially when conducting multiple design iterations. Thus, the motivation of this paper is to develop a simplified diffuser model that incorporates unsteadiness from the impeller flow field without requiring a full-stage unsteady CFD simulation. Steady and unsteady isolated-diffuser computational models were developed for a vaned centrifugal compressor, which reduced the computational cost of unsteady CFD simulations, when compared to full-stage unsteady models. The isolated unsteady diffuser model used unsteady inlet boundary conditions from a full-stage, unsteady model. The model was validated with laser Doppler velocimetry (LDV) test data gathered from the Centrifugal Stage for Aerodynamic Research (CSTAR) facility in the High-Speed Compressor Lab at Purdue University. Validation of the isolated unsteady diffuser model showed it consistently performed similarly to the full-stage unsteady model when predicting total pressure, absolute velocity magnitude, and incidence. Additionally, both models closely matched LDV test data. Incidence at the shroud-side diffuser leading edge was still overpredicted. These results show promising signs of the isolated unsteady diffuser model to solve preliminary concepts in parallel, with good accuracy, before evaluating best performing ones in a more costly and high-fidelity, full-stage unsteady CFD model.