Recovery

The fan was recovered by rapidly opening the bypass throttle to max fbw. The point at which drop-out occurs is of interest as it helps define a suitable operating or working line for the fan to ensure recovery from rotating stall. Once the fan is operating in rotating stall, a much larger throttle opening is re­quired to unstall the blade than was in use at the time when stall first occurred; this effect is called hysteresis. Much work has been performed on low speed compressors to define the degree of this (Day & Cumpsty – Ref. 7), but for a supersonic fan it is not clear what determines the extent of the hysteresis.

To determine the point of recovery for the transonic fan the time-averaged casing pressures at instances within the rotating stall data set have been com­pared with the pneumatic casing pressure rise measured during the mapping of the unstalled characteristic. This comparison can be seen in Fig. 6. Due to the relatively long time the bypass valves take to open (seconds), when compared to the rotational speed (100 revolutions per second), discrete periods of time can be defined and are presented from the rotating stall data set. In rotating stall (top solid line on Fig. 6), a raised pressure at the leading edge is found compared to the steady operating points shown due to compressed air com­ing forward as the stall cell passes. The reduction in the overall pressure rise can also be seen relative to the peak trailing edge pressure at the near surge condition. The stall recovery point (second solid line on Fig. 6), represents data from the first 2 revolutions after the last evidence of rotating stall. They show a similar profile of pressure rise as seen with the fourth steady state point from max pressure rise, suggesting that the performance demanded from the fan must be reduced to this point or below for the fan to recover. The pressure distribution for the very end of the data set, one second after recovery is pre­sented (bottom solid line on Fig. 6). This shows the continued recovery from rotating stall and the operating point has dropped to below the working line.

The time histories of three of the gauges as the fan recovers from the stalled state are shown in Fig 7. These suggest that one stall cell drops out before the other: it has reduced strength at 0.54 seconds and is then missing at 0.56 seconds as illustrated on the upstream gauge.

McKenzie (Ref. 8) previously observed that during rotating stall the non­dimensional inlet stagnation to outlet static pressure rise is independent of the compressor blading and the unstalled pressure rise and operates at two fixed levels. For incompressible ft>w ф TS = (‘Pexit — Poinlet )/pU2 ~ 0.11N for full span rotating stall and ^TS = 0.17N for part span stall where N is the number of stages.

For a supersonic fan the loading parameter ^TS is effectively a loading pa­rameter or equivalent to Ah/U2 for compressible ft>w, and represents the actual work input by the compressor relative to the potential work available. Starting

from the thermodynamic relationship:

where Р/рp = C. If we assume the process is isentropic (reversible adiabatic) then

(1)

where p is the mean Kulite static pressure, Po is the inlet (freestream) total pressure, To is the inlet total temperature, and U is the inlet blade speed.

Ignoring losses in Eq. 1 and assuming the inlet total temperature and aver­age total pressure are constant during rotating stall, dh/U2 at 60, 95 & 100% speed can be plotted against non-dimensional fbw (Fig. 8). The static pressure is taken from the ring of Kulites between the fan and OGV and the inlet flow is calculated from the ring of Kulites ahead of the fan. After the fan drops into rotating stall it operates at an approximately constant value of 0.11 for 95 & 100% speeds and at 0.09 for the 60% speed. Day et al. (Ref. 9) have previously shown full span and part span rotating stall operates at different lev­els of blockage A where the effective blockage created by the stall cell can be calculated as defined by, фауетаде = фипаші(1 — A) where A is the blockage 0-1. When A > 0.3 full span rotating stall results and part span dominates for A < 0.3 . For 60% speed A = 0.44, 95% A = 0.18, and for 100% A = 0.13. The two observations of increased blockage and a lower dh/U2 for the 60% speed case both suggest the stall cell occupies a greater span of the fan at lower speeds than at high speed. Unfortunately inlet total pressure is not measured dynamically on the rig. However, inlet total pressure measured dynamically with a close-coupled Kulite on an engine test has shown that the fan operates in rotating stall with constant mean inlet total pressure. Based on this, Fig. 8 uses the rig value of free stream unstalled total pressure as a reference pressure as well as the measured blade speed U and the dynamic Kulite static pressure; To is assumed to be a constant and is a steady state measurement. In addition, steady state data have also been plotted to show the position of the primary characteristic. The calculated dynamic flow agrees reasonably well with the steady state measurements. Based on the dynamic measurements of the up­stream and downstream rings of Kulites the stalled fan at 60% speed drops out at the 4th steady state point from the max pressure rise point.

Agreement between the drop-out point inferred from the over tip unsteady data and the rings of upstream and downstream Kulites is very good, but both methods rely on the steady state data for reference.