Flows in turbomachines are difficult to analyze because of the time-varying geometries and inherently unsteady flow. Various experimental techniques exist to investigate the time-averaged and unsteady flow within turbomachines, but they can be expensive to use. Because of this, various analytical techniques have been used to supplement the knowledge gained from experimentation. As computer resources became available, a variety of computational techniques were also used to further supplement knowledge of turbomachinery flows. In these earlier works, various levels of approximation were applied to make turbomachinery flow computations tractable on the available computer resources. Unfortunately, these approximations also restricted the usefulness of the computational model and the information it generated. Only recently have two- and three-dimensional unsteady viscous flow computations been possible, however, these unsteady analyses have been considered impractical for routine design purposes because of their memory usage, run times and dependence on supercomputer technology. Improvements in computer technology are rapidly making these computations practical on a range of computers from supercomputers to single-user workstations.
Supercomputers are expensive to buy, maintain and upgrade. Because of the need to economize, they tend not to be replaced or upgraded until they are seriously overutilized. This leads to long job queues and slow turn-around times on jobs. Raw computer speed is irrelevant if jobs are unable to get through the system in a reasonable amount of time. To the researcher, the wall clock time is often more critical than the cpu time required for convergence.
Even on an unloaded supercomputer, job accounting procedures can limit the amount of cpu time available to an individual. Typically, an individual is allocated a certain amount of time or is charged for time used. In either case, supercomputer cpu usage has to be carefully budgeted and other sources of cpu time must be found. A reasonable compromise to these constraints has been provided by the latest generation of workstations. A dedicated workstation can provide wall clock time performance on the order of that of a heavily loaded supercomputer at a comparatively low cost to the researcher. This paper discusses the issues involved in implementing a two-dimensional unsteady viscous multistage turbomachinery code (STAGE-2) on workstations.
Results from STAGE-2 have been compared with experimental data for a single-stage turbine calculation and a þstage compressor configuration in Gundy-Burlet et al. (1989, 1990). In this study, STAGE-2 has been used to study the effect of axial gap on the unsteady flow in a þstage compressor. The axial gaps used in this study are 20%, 35% and 50% of the average axial chord in the compressor. The time-average and standard deviation of the pressure field are used to investigate steady and unsteady flow features. In addition, surface pressures and force polar plots are presented. Coarse grid results were obtained on workstations. Fine grid results were obtained on both supercomputers and workstations.