Quantitative Visualization of the Flow in a Centrifugal Pump With Diffuser Vanes—II: Addressing Passage-Averaged and Large-Eddy Simulation Modeling Issues in Turbomachinery Flows

Abstract

The present paper addresses two basic modeling problems of the flow in turbomachines. For simulation of flows within multistage turbomachinery, unsteady Reynolds-averaged Navier–Stokes (RANS) of an entire series of blade rows is typically impractical. On the other hand, when performing RANS of each blade row separately one is faced with major difficulties in matching boundary conditions. A popular remedy is the “passage-averaged” approach. Unsteady effects caused by neighboring rows are averaged out over all blade orientations, but are accounted for through “deterministic” stresses, which must be modeled. To experimentally study modeling issues for deterministic stresses we use particle image velocimetry data of the flow in a centrifugal pump with a vaned diffuser that includes the flow in the impeller, the gap between the impeller and diffuser, between the diffuser vanes and within the volute downstream. The data have been presented in part A of this paper (Sinha and Katz, 1998, “Flow Structure and Turbulence of a Centrifugal Pump with a Vaned Diffuser,” Proceedings of the ASME Fluids Engineering Division, Washington, DC). Deterministic stresses are obtained from the difference between the phase-averaged and passage-averaged data, whereas the Reynolds stresses are determined from the difference between the instantaneous and phase averaged data. In agreement with previous findings, the deterministic stresses are larger than the Reynolds stresses in regions close to the interface between blade rows, and thus must be carefully accounted for in passage-averaged simulations. The Reynolds stresses are larger in regions located far from the transition region. The second series of issues involves modeling for large-eddy simulation. The measured subgrid stresses determined by spatially filtering the data are compared to eddy viscosity models and show significant discrepancies, especially in regions with separating shear layers. Backscatter of energy that persists during phase averaging is also observed. [S0098-2202(00)00901-9]