Instabilities identification based on a new centrifugal 3D impeller outflow model

Abstract

Previous research works have shown that the inflow boundary conditions have a significant effect on the behavior of diffusers in a centrifugal machine. To better understand the vaneless diffuser instability mechanism and save computing resources, several numerical works are planned to be conducted for the solely vaneless diffuser, excluding the rest of the centrifugal machine from the flow domain to simulate. Previous reduced-order models used either two-dimensional approaches that focused exclusively on the core-flow instability or three-dimensional models tested for a few inflow conditions. To obtain the more realistic diffuser inlet boundary conditions, a modeling method is here developed by fitting the diffuser inflow velocity derived from numerical simulations of the entire machine. The classic fitting methods used to approximate inflow profiles by algebraic polynomials or Gaussian functions are observed to introduce numerical artifacts that significantly affect the flow and therefore its stability. The multi-stage scale-matching fitting approach developed in this study is designed as a robust successive-order approximation of the inflow conditions. Our objective is to demonstrate its robust capability of taking into account the main physical features of the inlet velocity profiles, which, in turn, allows us to significantly improve the prediction of the flow instability occurring in the pump diffuser. Firstly, the RANS and URANS simulations of the entire machine are carried out by OpenFOAM using the SST turbulence model. The simulation results show that the RANS simulation is efficient in correctly capturing the diffuser inlet velocity profile except for developed stall conditions. The RANS simulations are carried out for cases with three different kinds of leakage configurations between the impeller and the diffuser. For each case, five flow rates were simulated to get the basic data for fitting the inlet profiles for a total of 125 simulations. The diffuser inlet velocity profiles are averaged in the azimuthal direction and fitted such to obtain an explicit function for the azimuthally-averaged velocity profile that varies with the flow rate Q. The fitting results are very close to the original data, and using our fits to predict the diffuser flow instabilities we show that our modeling approach compares well against the URANS simulations of the whole machine.