circumferentially mounted on the casing wall
45
◦
apart from each other from
sensor 1 to sensor 7 in one cross section in front of the first rotor (sensor 1 for
the bottom plot and repeat at the very top) and the frequency range is chosen
for the appearance of the spiky disturbances. From these figures, the intermittent
character of the emergence of spikes and their annular propagation can be clearly
tracked. Fig. 1 (
a
) shows the period of early emergence of spikes several hundreds
rotor revolution times ahead of the final stall, during which two spikes appear
randomly in time, but disappear before returning to where they were born. As the
compressor is throttled closer to stall, the spikes appear more frequently and last
longer than before as shown in Fig. 1, (
b
). It is interesting to note that the spikes
were all born at the same location near sensor 1 where the structure asymmetry
was found since far more early time period (not shown here). Fig. 1, (
c
) depicts
how the stall is finally triggered. Among the four spikes born in this period, only
one of them finally develops into full stall in less than five rotor revolution times,
and the other three didn’t get the chance. The result of this kind reminds us to
look back the early proposition of Tryfonidis et al. and makes us possible now
to study the stalling process from its pre-stall stage rather than the stage of stall
inception. The further research should include the characterization of the behavior
of flow disturbances in the pre-stall stage and their possible link with the known
types of stall inception, and the exploration of such behavior with geometric non-
uniformities when the disturbances emerge and with blade flow mechanism when
the disturbances grow.
Blade Passage Flow Mechanism.
It is generally known that the formation of
stall cells is routed in the fluid mechanics behavior of blade flow. The model in
terms of the effect of blade flow on the formation of rotating stall was first given
by Emmons et al. [33] and was well received since then. The stall cell, according
to this model, is meant to a flow blockage caused by the flow separation in blade
passages, which in turn is caused by external disturbance such as the deviation of
incoming flow towards the increase of the angle of attack. In particular, the model
could give an explanation of why the stall cell becomes propagating around the
compressor annulus. However, the effect of tip leakage flow, reverse flow and the
three-dimensional migration of stall cells, exhibited in the further research, have
hinted much complex blade passage flow in the stalling process, which is beyond
the effect of incoming flow in the two-dimensional consideration of Ref. [33].
To study such internal flow mechanism of rotating stall requires high resolution
measurement system in experiment and huge computation power in numerical
computation, and therefore its development was hindered for a long time but
revitalized in recent years along with the advancement of experimental and
numerical techniques and the emerging research of active control. The initial
attempt was an experimental study of the transient internal pressure patterns on
the outer surface of revolution in centrifugal compressor impellers, using the
time-traced pressure data measured by a series of transducers along the flow path
on the casing wall of the compressor [34]. The method was later extended in an
experimental study of the transient characteristics of rotating stall in a transonic
axial compressor [35]. These investigations showed the details of redistribution
of blade loading as the flow changes from steady to unsteady and could locate
the time and length scales for the formation of stall cells. Similar approaches are
also implemented recently in Ref. [32, 36] in supplement with numerical study.
Nevertheless, the experimental results remain two-dimensional and are limited to
the pressure fields rather than the velocity ones. It seems that these restraints can
118 ISSN 0236-3941. Вестник МГТУ им. Н.Э. Баумана. Сер. “Машиностроение”. 2006. № 2
