On the Correlation Between Spanwise Inducer Incidence and Impeller Diffusion for Ruled Surface and Barreled Sweep-Bow Impeller Design at Inlet Guide Vane-Off-Design

Abstract

Abstract In the present paper, three centrifugal stages of high volume flow coefficient are compared to each-other regarding their aerodynamic performance in design point and off-design point conditions at different speed and inlet guide vane (IGV)-setting angle: two stages with full-blade design (no splitter blades) have been numerically designed with different design geometry methodology. One geometry is based on a classical ruling surface design with a linear leading edge, the second geometry based on a fully-three-dimensional surface including a blade bow at the trailing edge and a barreled sweep at the leading edge. According to impeller test rig measurements and computational fluid dynamics (CFD) calculation, the classical ruling surface designed impeller outperforms the more sophisticated centrifugal stage with fully 3D-blade at fully axially guided IGV-flow. In the contrary, at closing IGV-off-design setting angles, toward surge operation, the fully 3D-blade impeller performs with higher efficiency and steeper negative pressure slope. On the search of the geometrical causes for the different aerodynamic performance (especially at IGV-off-design conditions), focus is set on the analysis of IGV-flow-interaction with the inducer flow and impeller diffusion. The one-dimensional analysis of the spanwise flow at the impeller leading edge reveals that, compared with the ruling surface impeller, the fully 3D-blade performs with lower flow incidence losses in favor to IGV-off-design operation than at IGV-neutral position. The streamwise flow analysis confirms the improved flow incidence characteristics of the 3D-blade impeller due to reduction of aerodynamic blockage and entropy production in the vicinity of the impeller leading edge. Based on CFD calculations, a new correlation of secondary flow and flow incidence is proposed, to be used for one-dimensional modeling.