Scaling Test Data to Predict Performance

Abstract

In the aerodynamic design of centrifugal compressors, it is possible to scale test data to predict the geometry (impeller diameter and tip width) necessary for a stage to meet a certain performance. There is not, however, a standard industry scaling method. The important points on the performance curve to consider in the scaling process are the scaling point, the surge point and the choke point. Choosing a scaling point is not objective and generally impellers are tested as part of a multistage machine where one stage is controlling, so the true surge and choke points are rarely known; these details make it difficult to depend on scaling alone for impeller design. Thus, impellers are often designed with one-dimensional (1-D) wheel sizing programs that rely on inexact slip and blockage models. A method of scaling is introduced in this paper that allows a machine to be designed based solely on test data rather than a sizing program. This scaling method was verified by utilizing different sets of data to predict the geometry necessary to meet a stage performance curve; the results showed that when various data sets corresponding to different geometries were scaled to the same design point, the predicted geometries converged. The predicted geometry was also compared to the geometry generated by the 1-D wheel sizing program to verify that the method is reliable. This approach to scaling significantly improves the confidence in the predicted performance of new stages.