Model Predictions and Experimental Results for the Rotordynamic Characteristics of Leakage Flows in Centrifugal Pumps

Adiei GlIinzburg is a Research Scientisl at the l nsfifll l de Machines Hydrau /iques et de Mecaniqlle des Fluides at the EcoLe Polytechn ique Federate de Lausallne, Switzerland. She IIo/eisa B.Sc. degree in AeronaMical Engineeringfrom the University a/the Witwatersrand and M.S. and Ph.D. degrees from the CaLifornia Institute a/ Technology. Christopher E. Brennen is a Professor of Mechanical Engineering and Deall of Stu dents at the Caiifomia Institute a/Technology j" Pasadena, California. He holds B.A., M.A., and D. Phil. degrees in Engineering Sciencefrom Oxford University. He has been honored by ASME. He has won the ASME Knapp A ward twice and is this year's recip ient a/the ASME Fluids Engineering A ward. He is a consultant and has over 120 technical publications. The role played by fluid forces in determining the rotordynamic stability and cha racteristics of a centrifugal pump is gaining increasing attent ion. The present research invest igates the contri butions to the roto rdynamic forces from the discharge-to-suction leakage flows between the front shroud of the rotating impel ler and the stationary pump casing. An experiment was designed to measure the rotordynamic shroud forces due to simulated leakage flows for different parameters sllch as nowrate, shroud clearance, face seal clearance, and eccentricity. The functiona l dependence on the ratio of whirl frequency to rotating frequency (termed the whirl ratio) is very similar to that measured in experiments and similar to that predicted by the theoretica l work of Childs fil. Childs' bulk flow model yielded some unusual results including peaks in the rotordynamic forcesal particular positive whirl ratios, a phenomenon which Childs tentatively described as a "' resonance" of the leakage flow. This unexpected phenomenon developed at small positive whirl ratios when the inlet swirl veloci ty ratio exceeds about 0.5. Childs points out that a typical swirl 41 veloci ty ratio at inlet (pump discharge) would be about 0.5 and may not, therefore, be large enough for the resonance to be manifest. To explore whether this effect occurs, an inlet guide vane was constructed which introduced a known amount of swirl into the flow upstream of the leakage flow inlet. A detai led comparison of model predictions with the present experimenta l program is presented. The experimental results showed no evidence of the "resonances," even at much larger swi rl inlet velocities than explored by Chi lds. INTRODUCTION The interaction of a cemri fugal pump impeller and the working flu id can cause various forces on the rotor. Some of these may cause self-excited whirl in which the axis of rotation of the impeller moves along a trajectory eccentric to the undeflected position. Tt is important to be ab le to predict these fluid-induced forces during the design phase. There is an ongoing effort dealing with an improvement in predicting the rotordynamic behavior of pumps (Frei, el al. [2], Pace, et al. [3], and Verhoeven {4]). This study has focused attention on one source of such whirl excitation, namely the forces acting on the shroud of an impeller due 10 the discharge-to-suction leakage flows external to the impeller. Rotordynamic forces imposed on a centrifugal pump by the fluid flow were firs t measured by Domm and Hergt [5], Hergt and Krieger [6) , Chamieh, et.1. [7), and Jery, et al. [8]. In the Rotor Force Test Facility (RFTF) at Ca ltech (Jery, et aI., [8]; Adkins, et a1. , [9]; Franz, et a1., [10]) known whirl motions over a full range of frequencies (subsynchronous, supersynchronous, and reverse whirl) arc superimposed on the normal motion of an impeller. This faci lity was also used for the present experiments. Fluid forces on a rotating centrifugaL impeller in a whirling motion have also been measured at the University of Tokyo, by Ohashi and Shoji {II). BoHeter, et a!. [12 ], also made an experimental determination of the hydrodynamic force matrices. It can be seen that there is an international attempt to understand these forces. The hydrodynamic force on a rotating shro ud or impeller (Figure 1) which is whirling can be expressed in the stationary laboratory fram e in linear form as: