Forced response predictions in modern centrifugal compressor design

Abstract

A computational interrogation of the time-averaged and time-unsteady flow fields of two centrifugal compressors of nearly identical design (the enhanced, which encountered aeromechanical difficulty, and production, which did not encounter any such difficulty) is undertaken in an effort to establish a causal link between impeller-diffuser interactions and the forced response behavior of the impeller blades. Through comparison of time- averaged flow variable and performance estimates with test rig data, the three- dimensional, unsteady, Reynolds-averaged Navier-Stokes flow solver (MSU Turbo) used in this interrogation is found to be adequate to the task of distinguishing the flow fields of the two centrifugal compressor designs. Thus, it is found that MSU Turbo can be a useful tool in comparing the unsteady flow fields in different centrifugal compressors. In addition, through comparisons of MSU Turbo/ ANSYS® estimates of strain with measured peak strain, MSU Turbo is also found to have the potential, as part of a CFD/ ANSYS® system, for serving as a predictive tool for forced response behavior in centrifugal compressors. Differences are found in the unsteady flow fields of the two compressors. The fluctuations over time of the unsteady blade loading on the enhanced impeller blades are greater than those on the production impeller blades. In the vaneless space, on each annular plane (from the impeller exit to the diffuser inlet), at a given spanwise location, the enhanced compressor has both a greater spatial variation in pressure and a higher average static pressure than the production compressor. At the diffuser inlet, there are differences in the time-averaged incidence angle distributions of the two compressors.

(cont.) Based on the observations delineated above, it is hypothesized that the differences in the time-averaged incidence angle distributions are the source of the differences in the pressure field that propagates upstream into the impeller passage, where these differences affect the unsteady blade loading. The differences in the unsteady blade loading then lead to the observed forced response behavior in the two designs.