Geometric considerations in the capture of localized flow reversal in centrifugal compressor vaneless diffusers using steady state simulations. (November 2017)

• Record Type: Journal Article
• Authors: Clarke, Chris; Marechale, Russell; Engeda, Abraham; Cave, Michael
• Is Part Of: Proceedings of the Institution of Mechanical Engineers. Volume 231:Number 21(2017)
• Page Start: 3959
• Page End: 3973
• Publication Date: 2017-11
• Subjects: Centrifugal Compressor Stability -- Vaneless Diffuser Stability -- Computational Fluid Dynamics (CFD) -- Vaneless Diffuser -- Design Research; Mechanical engineering -- Periodicals; 621.05
• DOI: 10.1177/0954406216656213
• Languages: English
• ISSNs: 0954-4062

Abstract

A steady state simulation procedure is proposed to capture localized flow reversal inside of a centrifugal compressor vaneless diffuser. The procedure was performed on 12 compressor stages of varying geometry for speed lines of 13, 100, 19, 240, and 21, 870 r/min. The simulations were run for all points from choke to surge including the experimentally determined rotating stall onset point. The experimental data and geometry were provided by Solar Turbines Inc. San Diego, CA. It was found possible to capture localized flow reversal inside of a vaneless diffuser using a steady state simulation. The results showed that using a geometric parameter, comparing the diffuser width, b4, to the impeller blade pitch distance, dpitch, it could be determined whether or not a steady state simulation could capture localized flow reversal. For values of b4 /dpitch beneath 0.152 flow reversal could not be captured. But, for values of b4 /dpitch above 0.177 localized flow reversal was captured. For values between 0.152 and 0.177, no conclusions could be drawn. Where possible, experimental data were compared against the diffuser inlet and outlet numerical profiles and the meridional contour plot. These comparisons served to validate the approach used in this article. These validations showed that the procedure defined herein is accurate and trustworthy within a specific range of geometric and flow characteristics. There are two other conclusions. First, the b4 /dpitch parameter helps to define the type of flow breakdown. For b4 /dpitch below 0.152, the flow breaks down in the circumferential direction, but for values of b4 /dpitch above 0.177, the flow breaks down in the span-wise direction. Second, the simulations were able to capture instances of localized flow reversal before rotating stall onset. This concludes that localized flow reversal is not the determining factor in rotating stall onset as has been suggested by other investigators.