It is important in development of turbomachinery to predict their performance precisely. Especially the prediction of multistage pump performance is one of the challenging problems because internal phenomena which relate to the performance are complicated. Therefore, in this research, we verified accuracy of Computational Fluid Dynamics (CFD) in predicting performance of a five-stage high-pressure volute pump by comparing predicted values by CFD with measurement data. We tried two methods to predict the pump performance. One is a computation with a complete pump model which includes all five stages and leakage passages. This method can be expected to represent total internal flow phenomena. The other method is totaling up the performance data from separate computations of 1st–2nd stages and series stages. This method is simpler than the former and involves less computational cost. As a result, it was clarified that all the methods could predict pump head at the best efficiency point to some extent, even by steady computation. However, no prediction can predict positive gradient in Q-H curve which was observed in measurement at low flow rate. Except for the unsteady complete pump model computation, efficiency and shaft power could not be predicted precisely. In addition, at high flow rate, unsteady computation of the complete pump model shows the best agreement in head. In the complete pump model computation at high flow rate, the series stage next to the long crossover has larger head because of the influence of it. Therefore, the separated model has difficulty in representing series stages’ performance. In order to predict performance at high flow rate, unsteady computations also including properly the influence of the long crossover properly are necessary. In addition, to predict performance at low flow rate, unsteady computation is necessary.

This content is only available via PDF.
You do not currently have access to this content.