A throughflow method for designing and analysing compressors has to be supplied with loss, deviation, and blockage estimates for every blade row. The earliest methods used empirical correlations for profile loss and deviation, together with an empirical blockage or “work done” factor, and empirical estimates of additional losses near the endwalls. Previous papers by the author have described how to replace the empirical blockage factor and endwall corrections by explicit calculations using a new mathematical model of the endwall phenomena. Those papers illustrated the application of the method near design conditions, using either design profile loss and deviation figures or computations by a viscous-inviscid interaction blade-to-blade method.
In order to estimate off-design performance rapidly over the whole operating range, some way of estimating off-design profile loss and deviation must be chosen. In this paper, the previously-derived design point loss and deviation figures are retained, and an empirical correlation due to Miller, Wasdell, and Wright is used to predict the changes in loss and deviation off-design. It is shown by means of sample three-dimensional Navier-Stokes computations that the endwall model remains applicable off-design.
The method has been tested against two low speed and two high speed compressors, one of each example having controlled-diffusion blading. The low speed compressor characteristic maps are predicted only approximately, but the predicted high speed compressor maps are good. It is widely believed that endwall flow separation can initiate stall or surge. As stall or surge was approached the shape factor of the annulus wall boundary layer at one location rose sharply, but no single stall-predicting value could be found.