REVIEW—Axial Compressor Stall

Phenomena
Stall in compressors can be associated with the initiation of several types of fluid
dynamic instabilities. These instabilities and the different phenomena, surge and
rotating stall, which result from them, are discussed in this paper. Assessment is
E. M. Greitzer made of the various methods of predicting the onset of compressor and/or com-
Associate Professor, pression system instability, such as empirical correlations, linearized stability
Department of Aeronautics and Astronautics, analyses, and numerical unsteady flow calculation procedures. Factors which affect
Massachusetts Institute of Technology,
the compressor stall point, in particular inlet flow distortion, are reviewed, and the
Cambridge, Mass. 02139
techniques which are used to predict the loss in stall margin due to these factors are
described. The influence of rotor casing treatment (grooves) on increasing com-
pressor flow range is examined. Compressor and compression system behavior
subsequent to the onset of stall is surveyed, with particular reference to the problem
of engine recovery from a stalled condition. The distinction between surge and
rotating stall is emphasized because of the very different consequences on
recoverability. The structure of the compressor flow field during rotating stall is
examined, and the prediction of compressor performance in rotating stall, including
stall' lunstall hysteresis, is described.

Introduction
In normal operation of a compressor, the flow is nominally
steady and axisymmetric, apart from the blade to blade
pressure variations and the small scale unsteadiness associated
with the moving pressure and velocity fields of the rotors (or STALL LINE CONSTANT RPM LINE
impellers). However, if the performance map of a compressor PRESSURE RPM= N 3
is plotted in the usual form shown in Fig. 1, as pressure ratio RISE
versus mass flow for different rotational speeds,1 a line can be
defined which is commonly referred to as the stall line (or
surge line) and which divides the map into two regions. To the
left of the line, the flow is no longer steady. Large oscillations
of the mass flow rate may occur (called surge) or severe self-
induced circumferential flow distortions may rotate around
the annulus (rotating stall), or a combination of both
phenomena may appear.
FLOW
Rotating stall induces large vibratory stresses in the blading Fig. 1 Compressor performance map
of compressors and is therefore often unacceptable for
structural reasons. In addition, there can be a large drop in
performance associated with this flow regime (efficiencies
below twenty per cent can be seen in the literature) [l] 2 so that The position of the stall line is a matter of great concern to
overall gas turbine engine cycles may not be self-sustained. In the designer, and considerable effort is frequently expended
an engine the greatly decreased mass flow through the system to ensure that there is sufficient margin between the stall line
can also cause turbine overtemperatures. Surge can be in- and the operating line. Even if this is done, there are still
tolerable from the point of view of system operation and can certain situations under which stall will occur, such as rotor
also lead to high blade and casing stress levels [2]. Thus, no speed transients, flow distortions, and nonsteady inlet and
matter which type of instability appears when the stall line is exit flow pulsations. Thus, the problem of recovery from a
crossed, the stall line generally represents a limit to the useful stall condition also becomes extremely important.
operation of the machine and is therefore to be avoided. Because of this the overall topic to be described can be
separated, at least conceptually, into two main parts: (1)
examination of the basic fluid mechanics associated with the
1 onset of the instabilities that lead to rotating stall and/or
Actually, corrected flow versus corrected speed. surge, including such aspects as the impact of inlet distortion,
2
Numbers in brackets denote references.
effects of downstream system components on compressor
Contributed by the Fluids Engineering Division for publication in the
JOURNAL OF FLUIDS ENGINEERING. Manuscript received by the Fluids stability, and stability enhancement, and (2) behavior sub-
Engineering Division, September 17, 1979. sequent to the onset of this initial instability including large

134/Vol. 102, JUNE 1980 Transactions of the ASME

Copyright
Downloaded From: [Link] on © 1980
06/30/2019 Terms of Use:by ASME
[Link]
 amplitude system transients, limit cycle surge oscillations,
structure of the rotating stall flow field, and the effect of
stall/unstall hysteresis on system recovery from rotating stall.
The choices of subjects to be covered in a survey of this type Direction of stal
will reflect, to some extent, the background of the author. The
present author's experience is mainly with aircraft engine propagation
turbomachinery. This review is therefore focused on the fluid
mechanics of axial compressor stall, although it should be
emphasized that many of the ideas (especially those con-
cerning system behavior) can be applied to centrifugal
compression systems as well.

Fluid Dynamic Instabilities in Compressors and

Compression Systems
We can start this section by describing in a bit more detail
the types of phenomena associated with instabilities in
compression systems which arise due to the presence of stall.
Looked at from the point of view of the individual diffusing
passages in the compressor, stall generally implies separation Compressor
of the flow from one or more of the passage walls. However, blade row
compressor blade rows consist of many of these diffusing Fig. 2 Physical mechanism for inception of rotating stall
passages in parallel, so that phenomena can occur which do
not happen with a single airfoil or diffusing passage. One of
the most striking of these is rotating stall. This is a flow pens, the flow can separate from the suction surface of the
regime in which one or more "stall cells" propagate around blade so that a substantial flow blockage occurs in the channel
the circumference of the compressor with a constant between B and C. This blockage causes a diversion of the inlet
rotational speed, usually between twenty and seventy percent flow away from blade B and towards C and A (as shown by
of the rotor speed. In the cells the blades are very severely the arrows), resulting in an increased angle of attack on C and
stalled. Typically there is negligible net through-flow, with a reduced angle of attack on A. Since C was on the verge of
areas of local reverse flow, in these regions. The cells can stall before, it will now tend to stall, whereas the reduced
range from covering only part of the span (either at the root angle of attack on A will inhibit its tendencies to stall. The
or at the tip) and being only a few blades in angular width, to stall will thus propagate along the blade row in the direction
covering the full span and extending over more than 180 shown, and under suitable conditions it can grow to a fully
degrees of the compressor annulus. It is this latter situation developed cell covering half the flow annulus or more. In this
which most commonly occurs in multistage compressors at fully developed regime, the flow at any local position is quite
speeds near design and which is most serious. The part span unsteady; however the annulus averaged mass flow is steady
stall of the front stages of multistage compressors at low with the stall cells serving only to redistribute this flow.
speed attracted a large amount of interest in earlier years, but The onset of rotating stall is thus associated with an in-
generally has much less severe consequences and is hence not stability which arises due to the stall of the compressor blade
of primary concern. passages. 3 For the overall compression system, this can be
The basic explanation of the mechanism associated with the regarded as a localized instability. However, a more global
onset of stall propagation was first given by Emmons [3] and system instability can also occur, leading to surge. In contrast
can be summarized as follows. Consider a row of axial to the behavior during rotating stall, the annulus averaged
compressor blades operating at a high angle of attack, such as mass flow and the system pressure rise during surge undergo
is shown in Fig. 2. Suppose that there is a nonuniformity in
the inlet flow such that a locally higher angle of attack is 3
The point of stall can, however, be affected by other closely coupled
produced on blade B which is enough to stall it. If this hap- components in the system as described below.

Nomenclature

a = speed of sound averaged axial velocity

PT total pressure
Ac = c o m p r e s s o r flow- parameter at full-span
AP pressure change across
through area stall cessation
compressor
b = blade chord </>« = flocal) value of <f> in
AP,stage pressure rise per stage
B = nondimensional stability unstalled region of
5PT total pressure distortion
parameter; compressor
inlet dynamic pressure
U Q circumferential coor-
v„ compressor mean radius
dinate
B= 7n^ . R rotor speed at midspan
2coLc 2a ' ACLC U P density
vP plenum volume
Cx = axial velocity pressure rise parameter;
w relative velocity
Cx = mean axial velocity relative flow angle
C'x = nonuniformity in axial p , " exit -*in!et

A stall cell blockage

velocity axial velocity parameter;
Lc = effective length of Helmholtz resonator
4> = Cx/U frequency;
compressor
m = mass flow design value of axial
N/Vd = rotor corrected speed velocity parameter co = a.
P = static pressure value of annulus

Journal of Fluids Engineering JUNE 1980, Vol. 102/135

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 Rotating stall Surge the blading, using the local flow conditions generated by use
Circumferentially nonuniform flow Axially oscillating flow of one of the many axisymmetric compressor flow field
calculations, to see whether any section would be operating
under too adverse a condition, while for a multistage com-
pressor the parameter might be applied only on a meanline or
averaged basis.
Several of the early correlations of this sort are described by
Horlock [5], but one of the well-known examples of this type
of approach, which is still much in use, is the work of Leiblein
[6]. He developed a parameter which he called the diffusion
factor (or D-factor). This was related to the adverse pressure
gradient to which the boundary layer on the suction surface of
the airfoil was subjected. The Z>-factor was defined by
Leiblein as
Frequency ~ 50 - 100 Hz Frequency ~ 3 - 1 0 Hz
Fig. 3 Modes of system response resulting from stall D=l - + "-
Wx 2aWx
where Wx is the inlet relative velocity, 1V2 is the exit relative
large amplitude oscillations. The frequencies of these velocity, AWg is the change in circumferential velocity
oscillations are generally at least an order of magnitude below component and cr is the solidity.
those associated with passage of a rotating stall cell and It was found that the total pressure loss correlates quite well
depend on the parameters of the entire system. In addition, with D, and, based on Leiblein's cascade results, one can see a
during the surge cycles the instantaneous mass flow through rather sharp rise in loss occur as D is increased past a value of
the compressor changes from values at which (in steady state roughly 0.6. This can therefore be taken as a very ap-
operation) the compressor would be free from stall, to values proximate criterion for the onset of stall in a cascade (see also
at which one would find rotating stall or totally reversed flow. [7] for further work using this approach). Although much of
Because of the low frequency of the oscillations, the com- the work done by Leiblein was based on two-dimensional
pressor can pass in and out of these flow regimes in an ap- cascades, the use of the/)-factor has been carried over to axial
proximately quasi-steady manner.4 as well as to centrifugal compressors [8]. Features such as the
The two types of instability are indicated schematically in differences between the flows in a cascade and the flows at the
Fig. 3. Sketches of the transient signatures that would be tips of axial compressor rotors, for example, are
given by typical high response instrumentation, such as a hot "recognized" by noting that different limiting values of the
wire at the compressor inlet for the rotating stall situation and Z)-factor are used for the rotor tips than for other sections.
a pressure probe in the combustor (or other volume down- The correlations based on ZMactor, as well as other im-
stream of the compressor) for the surge cycles, are also in- proved correlations for the stall point, have been intensively
dicated in the figure. Rough magnitudes of the time scales investigated by the aircraft engine companies. Virtually all of
associated with the different phenomena (in an aircraft gas this information, however, is held as proprietary and there is
turbine context) are shown as well. It is to be emphasized that very little in the open literature, particularly as regards
although the illustration indicates the phenomena only in this multistage compressors and transonic fans. Several features
particular context, these instabilities are inherent in pumping that are significant, however, are that the "limiting" D-
systems involving all types of turbomachines. factor, or other loading parameter, (i.e., the value at onset of
The two phenomena (surge and rotating stall) are seen to be stall) tends to increase as the aspect ratio and/or the non-
quite distinct. However, they are not unrelated, since often dimensional tip clearance6 decreases. An example of the first
the occurrence of the local instability (associated with the of these trends, as presented by Smith [9], is given in Fig. 4
onset of rotating stall) can trigger the more global type of which shows nondimensional stage pressure rise versus flow
system instability (leading to surge). Thus one has to consider coefficient from two compressors with aspect ratios of 1.96
the possibilities of either type of instability and develop and 5. The two compressors each have four stages and the
methods for their prediction. We will examine the techniques same non-dimensional tip clearance6. The lower aspect ratio
which have been applied to prediction of rotating stall first. compressor has a higher pressure rise per stage at the stall
point, and stalls at a lower flow rate. Similar results have been
Prediction of the Onset of Rotating Stall found by Fligg [10]. Effects of tip clearance are also
illustrated in the above mentioned paper by Smith [9] as well
Correlations for Stall Inception. As noted, a key problem as in [11] and [12].
for the designer is the prediction of the point at which stall There are several other factors such as Reynolds number,
occurs. This problem has been attacked by many investigators tolerances and deterioration etc., that can also have an effect
at several quite different levels of approach. The most em- on the stall point. These are also accounted for in practice by
pirical are the correlations that have been developed for stall using correlations, and they will not be discussed here save for
onset. The basic concept is to find a parameter (or the remarks that there can be a substantial loss in stall margin
parameters) which correlates the onset of stall5 for a number as the blade Reynolds number drops below roughly 100,000.
of different blade geometries, compressor designs, etc. In a
design procedure for a low hub-tip ratio fan, for example, the Although the use of correlations such as .D-factor may
parameter could be applied at different span locations along appear to be an overly simple approach, it is one that is, at
present, in common usage for multistage axial compressors. A
recent example of such an approach can be found in [13].
4
More theoretical approaches have not yet led to the definition
In view of this behavior E. S. Taylor [4] has paraphrased P. T. Barnum to of stall point with additional precision, but this is not to say
describe the distinction between surge and rotating stall, respectively: "You can
operate a compressor to stall all of the blades some of the time, or to stall some that the correlative procedures are all that is needed to put this
of the blades all of the time.''
5
Defined here as the condition at which the steady axisymmetric flow
becomes unstable. Tip clearance/staggered spacing.

1 3 6 / V o l . 102, J U N E 1980 Transactions of the ASME

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 Stoll Stability boundary
o AR = 1.96
flow (zero slope point)
- AR = I.96 • AR = 5

- V^ Stall o
flow o
AR = 5 * TS
o Unstable
i i 1 i
0.4 0.6 0.8 Compressor
(cx/u; performance
Fig. 4 Performance of high and low aspect ratio compressors (data of curve
Smith [9])

problem to rest. Indeed, a comment in an overview of this

topic by a representative of one of the engine companies states *
flatly "we still cannot predict accurately the onset of engine Fig. 5 Onset of compressor flow instability (rotating stall inception
instability" [14]. (However, it should be recognized that point)
requirements for stall prediction in aircraft gas turbine
engines may be quite stringent compared to those in other time of an initially prescribed small perturbation (for times
areas of axial compressor usage.) until the disturbances become too large for the linearized
Although it is important to mention these correlations, and theory to apply).
to give several relevant references, the main emphasis in this These calculations yield results in terms of the critical slope
review is on the fluid dynamic phenomena associated with of a function such as the mean compressor performance
compressor stall, and the correlative procedures do not really curve, or cascade loss curve, etc. at which the axisymmetric
provide insight into this aspect. Therefore we now consider flow becomes unstable. However, these slopes are extremely
approaches to the stall prediction problem, which are tied difficult to obtain accurately, especially for the cases of
more closely to the flow instabilities occurring in the com- greatest interest—multistage compressors or transonic
pressor. These can be divided into two types: linearized (small fans—and this is a major reason for the relatively little usage
disturbance) stability analyses for predicting the onset of of these methods by designers. In other words, if one wishes
rotating stall, and nonlinear treatments which follow the to use one of the instability calculations as a predictive tool,
growth of small perturbations to a fully developed (finite one must be able to predict the slopes of the compressor
amplitude) state. constant speed characteristic at off-design conditions, and, at
present, this cannot be done with an adequate degree of
Stability Analyses for Rotating Stall Onset. The first of the precision.
linearized stability analyses was by Emmons [3], who in- Putting aside this difficulty, there is still the question of
vestigated the conditions under which a small amplitude, how well existing stability criteria describe the onset of
circumferentially nonuniform, two-dimensional flow per- rotating stall. As an example let us consider one of the best
turbation in a cascade would grow. Since then there have been known of the (two-dimensional) criteria, due initially to
many extensions of this investigation, of which two recent Dunham [20], which states that rotating stall inception will
examples are the papers by Nenni and Ludwig [15] and by occur at the peak (zero slope point) of the exit static pressure
Fabri [16]. An excellent summary of the earlier work is given minus inlet total pressure compressor characteristic. This is
by Emmons, Rockett, and Kronauer [17] and a com- illustrated schematically in Fig. 5. In this figure the horizontal
prehensive bibliography up to 1967 has been compiled by axis is axial velocity parameter, </>, ( = CX/U), and the vertical
Fabri [18]. A useful overall introduction to the topic is given axis is nondimensional pressure rise, \pTS = ([-Pexit -
byStenning[19]. (PT)inXa]/plP).
In these analyses a small amplitude perturbation is This criterion, which has also been derived by other in-
superimposed on a given mean operating condition of the vestigators, has been applied with some success and does
compressor or cascade. The linearized equations of motion appear to furnish a rough "rule of thumb." However,
are solved to yield the forms of the flow perturbations in counter examples in which it does not hold can readily be
regions upstream and downstream of the compressor blade found. As illustration of this, Fig. 6 shows data from a
row (cascade). The wavelength of the flow nonuniformities is representative sample of low speed multistage compressors.8
taken to be much larger than the blade pitch (as is ex- Curves V, VI, and VII do appear to show approximately
perimentally found to be the case) so that an actuator disk7 zero slope (within the accuracy of the data), but curves I-IV
model of the cascade can be used. Suitable matching con- have a negative (i.e., stable) slope right up to the stall points.
ditions are applied across the cascade to link the flow This situation, where there does not appear to be a zero slope
quantities upstream and downstream. In general the con- region of the compressor characteristic, is even more apparent
ditions used (for the two-dimensional case) have been con- in high speed multistage compressor data. (The author has not
tinuity of mass across the cascade, an inlet/exit flow angle seen any multistage data in which the converse is true—i.e., in
relation, and an inlet angle/total pressure loss relation, which operation significantly on the positive slope part of the
although some investigators have found it more convenient to characteristic is found.)
work in terms of circulation and shed vorticity [15] rather Having said this, however, it should be emphasized that in
than total pressure. From this procedure one can determine the opinion of this author these types of stability calculations
the eigenvalues of the system of equations which define the can still be useful. This is true on several counts, not the least
stability of the flow field, or one can examine the growth in of which is that they provide an overall physical un-
7
Defined as a representation of a blade row as a plane across which the mass 8
flow is continuous but the total pressure, circumferential velocity, pressure, etc. Three Stage Compressors: I, II, V, VII, [21]; IV [22]; VI [23]; Four Stage
can be discontinuous. compressor III, [24].

Journal of Fluids Engineering JUNE 1980, Vol. 102/137

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 tcx
(Dimensionless time,"!",
2TTR
1.0
T=0 T = 0.l

Cx
-1.0
1.0
T=0.2
Ci
Cx
1.0
1.0
7=0 A T= 1.2

-1.0
1.0
T=0.6
Fig. 6 Multistage compressor performance (low speed rigs);
references identified in footnote 8 Cx
derstanding of the breakdown of the axisymmetric flow in 0 27T
axial pumps or compressors. 9
An important advantage of the stability analyses is that
(Circumferential P o s i t i o n )
they can be used to show effects that correlations do not Fig. 7 Growth of small disturbance (calculations of Takata and
predict. One of these effects is the coupling, or interaction, Nagano [25])
between blade rows, between different compressor spools, or
between a compressor and components downstream in the
compression system. Such coupling is much stronger with uniform flow. However, they do seem to offer potential for
asymmetric flows than in the axisymmetric situation. This assessing the effect of inlet distortion on stall inception, as is
feature, which is central to the understanding of discussed below.) An example of the results of this type of
nonaxisymmetric flows in compressors, occurs because of the calculation for an isolated rotor is given in Fig. 7 taken from
much larger length scales of the flow nonuniformities than in Takata [25]. The figures show the normalized axial velocity
the axisymmetric situation. Specifically, it is the "single nonuniformity, C'x/Cx, versus circumferential position, 8, at
lobed" (i.e., one per circumference) flow perturbation which different nondimensional times, T, = (time • Cx)/2irR. It
appears to be of most practical interest in multistage com- can be seen that the initial sinusoidal perturbation grows into
pressors. For disturbances of this type the relevant length over a large amplitude disturbance which propagates round the
which adjoining rows can be considered to interact is of the circumference.
order of the radius of the machine. Thus, adjacent blade Calculations of this type can show some of the features of
rows, or even the two spools of a gas turbine engine, can be rotating stall such as the hysteresis in the stall/unstall process.
considered to be closely coupled as far as circumferentially Other aspects, however, do not appear to be modeled very
nonuniform flow is concerned, and, as emphasized by well, and a general comment on these nonlinear models is
Dunham [20], it is not correct to analyze the stability of one that, in spite of their complexity, many of the central features
row in a compressor without considering the influence of the of the fully developed stalled flow may not yet be represented.
other rows. This coupling can be taken into account by the For example, the present calculation procedures are either
stability analyses and although, as stated, they may not be two-dimensional, assume that the flow occurs along
quantitatively precise, they can still give valuable guidance as axisymmetric stream surfaces so that flows perpendicular to
to when row-row, spool-spool, etc., interactions will be im- these surfaces are supposed zero, or only include three-
portant. dimensionality in a potential flow representation. Ex-
periments, however, indicate that strong radial flows can
Nonlinear Investigations of Rotating Stall Onset and occur. In addition, the models of unsteady blade row per-
Growth. Although the linearized analyses have, in the past, formance that are used are quite rudimentary. They are
been used to try to predict some of the features of fully essentially one-dimensional and rely heavily on extrapolation
developed rotating stall, it has become clear that in this flow of the loss and turning characteristics (to the negative flow
regime the stall cells are definitely not small perturbations and regime according to the experimental measurements). Fur-
linear analysis is inapplicable. Thus this type of investigation ther, the calculations are carried out with a constant mass
is only useful for the problem of stall inception, and to follow flow, whereas in the actual situation the mass flow will change
the subsequent development of the rotating stall, one must use (along a throttle line) as the flow develops from unstalled to
a nonlinear model. This has been done by Takata and Nagano fully developed rotating stall. In addition, the steepness of the
[25], as well as by Orner [26]. These models use time marching downstream throttle characteristics can also have significant
techniques to determine whether a small disturbance will grow effect [27]. In view of the path dependent behavior shown by
or decay by calculating the development of the flow to some some features of rotating stall (i.e., the hysteresis) it may be
eventual steady-state solution. This could consist of a flow that these latter aspects should also be taken into account.
with a large amplitude disturbance propagating around the Because of the above, the author feels that while it is useful
compressor, which is taken to be indicative of compressor to pursue calculations of this type, they should be closely
operation in rotating stall. compared with experimental data to assess the effects of the
So far these calculations have only been applied to an assumptions. For example, one result of these calculations
isolated rotor or to a single stage. (It should be noted that which is not satisfactorily explained is that for an isolated
these procedures can give no more information than the rotor the number of cells that appears from the calculation is
linearized stability analyses in regard to stall inception in a equal to the number of lobes in the initial perturbation,

138/Vol. 102, JUNE 1980 Transactions of the ASME

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 STATIC INSTABILITY DYNAMIC INSTABILITY
h-
PIPE.

^ >
\
COMPRESSOR PLENUM THROTTLE
(ACTUATOR DISK)
Fig. 9 Equivalent compression system for stability analysis

The basic analysis of the dynamic system stability was also

UNSTABLE IF SLOPE OF COMPRESSOR EVEN IF STATICALLY STABLE by Emmons, et al. [3], who treated a lumped parameter
CHARACTERISTIC GREATER THAN SLOPE SYSTEM CAN BE DYNAMICALLY
OF THROTTLE LINE (POINT B)
system similar to that shown in Fig. 9. This consisted of a
UNSTABLE (POINT D)
compressor operating in a duct of small flow through area, a
Fig. 8 Types of compression system instability
plenum volume (where the flow had no appreciable kinetic
energy), and a downstream throttle. Basic criteria were
whereas tests with isolated rotors have shown that the number developed for the stability of this system. Since then other
of cells is rather a function of the overall mass flow through investigators have looked at extensions of this problem to
the stage [28]. predict the system instability point. Although there are dif-
ferences between the various analyses, the general result that
Overall Compression System Instability emerges is that the system will be unstable near the peak of the
(constant speed) pressure rise/mass flow characteristic, at
We have so far discussed the prediction procedures for the some slightly positively sloped point, which is generally well
inception of rotating stall in a compressor with uniform inlet before the simple static stability criterion is violated [19].
flow, and we can now examine the criteria for the onset of the
Although the calculation of this system stability point can
more global type of instability leading to surge. This is a
readily be carried out, once again we encounter a situation
system phenomenon and one in which, in contrast to rotating
where the critical parameters are the slopes of the steady-state
stall, the compressor appears to participate in an essentially
(uniform flow) compressor speed lines at the stall point and,
one-dimensional manner.
as stated, there is difficulty in predicting these.
We can sketch briefly the basic mechanism for compression
We have discussed the two types of instability, and it is
system instability. The simplest view of this, and one that is
useful to now relate them. The global (system) instability is a
found in many texts, can be derived with reference to a system
basically one-dimensional phenomenon, involving an overall,
consisting of a compressor and downstream throttle. The
annulus averaged, (in some sense) compressor performance
operating point of the "system" is the intersection of the
curve. For typical volumes, lengths, and throttle charac-
compressor and the throttle pressure-flow characteristic
teristics this must generally be slightly positively sloped for
curves. From considerations of the pressure rise through the
system instability to occur. We have also seen that the
compressor and the drop through the throttle it can be shown
axisymmetric flow through a compressor can be unstable to
that the system will become unstable when the slope of the
two- (or three-) dimensional infinitesimal disturbances, and
compressor pressure rise curve is steeper than the slope of the
that this local instability marks the inception of rotating stall.
throttle curve [29].
However, the onset of this rotating stall is very often
This static stability argument is illustrated on the left of associated with a precipitous drop in the overall ("one-
Fig. 8. For a small perturbation in mass flow (a decrease, dimensional") pressure rise mass flow curve of compressor
say), if the system is operating at point A a pressure imbalance performance. In other words, the inception of rotating stall
will arise to cause fluid accelerations that return the system to can lead to a situation where the instantaneous compressor
operation at the initial point. Point A is thus a stable operating point is on a steeply positively sloped part of the
operating point. At point B, however, where the throttle line characteristic, with a consequent violation of the dynamic
is tangent to the compressor characteristic, the pressure forces and/or the static instability criteria. In this sense, the onset of
associated with a small decrease in mass flow will cause the the local compressor instability can trigger the more global
system to depart further from the initial operating point, so compression system instability. Because of this, the com-
that point B is an unstable operating point. pressor designer does not generally differentiate between the
This criterion is, however, too simple to describe the real two modes as far as the applications of the stall correlations
phenomenon, since it only considers the static stability of the are concerned.
system. In fact it is generally the dynamic stability criteria
which are violated first.9 Thus as indicated on the right side of Multi-Element Stability Models. Up to now the discussion
the figure a compression system can be statically stable of compression system stability has been based on viewing the
(according to the foregoing slope criterion) and still exhibit compressor as a single element. However under some cir-
instability. cumstances this can be too simple, since there can be dif-
ferences between the mass flow perturbations at the front of
9 the compressor and those at the rear due to compressibility.
The terms dynamic and static instability can be made more quantitative by Thus, different investigators have extended the overall
the following illustration. Consider a simple second order system described by
the equation lumped parameter analysis and modeled the compressor in a
stage by stage (or even row by row) method, with separate
d2x dx volumes and actuator disks for each stage. Mass, momentum,
r— + 2a +/?x = 0 and energy balances are written for each of the stages.
dt1 dt
Examples of such investigations are references [30]-[34], and a
where a and |3 are constants of the system. If @ > a , the condition for in- discussion and comparison of some of the basic models can be
stability is simply a < 0, but, independently of a, instability will occur if /3 < found in [31].
0. It is usual to denote these two types of instability as dynamic and static
respectively. Static stability W > 0) is a necessary but not sufficient condition The determination of the instability point for these models
for dynamic stability. can be found either by solving for the eigenvalues or by

Journal of Fluids Engineering JUNE 1980, Vol. 102/139

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 (digital) simulation of the system transients. In addition these turbation and using a linearized approach. Such an analysis
models can explore the effects of pulsations that are imposed can give very useful physical insight as well as quantitative
on the compression system. However, as with the simpler information about the performance of the compressor in a
models, "the compressor stability boundary is strongly in- circumferentially nonuniform flow. In particular it is able to
fluenced by the stage characteristics. The overall results are show clearly the role of the various design parameters and
thus dependent on the accuracy of the stage characteristics operating conditions on distortion transfer, l0 to demonstrate
. . ." [34]. Hence the ability to predict the slopes of the steady- the manner in which the upstream velocity field due to the
state (uniform-flow) compressor speed lines at the stall point distortion is substantially modified by the compressor, and to
is a key item in the successful use of these techniques. motivate the basic length rules for scaling the overall flow
phenomena. Excellent examples of analyses of this type are
Inlet Distortion Effects on Stall Inception the papers by Plourde and Stenning [35], Dunham [20], or
Stenning [36], and these are recommended as an introduction
The discussion so far has been confined to stability to the field. (See also [37] for an extension of the basic ideas to
predictions with a uniform inlet flow, but critical situations compressible flow.)
for stall are often those in which an inlet nonuniformity or From the point of view of stability prediction, however, the
distortion is imposed on the compressor. For an aircraft analysis of inlet distortion as a small perturbation can give no
engine, for example, distortions can arise from inlet information as to the effect of inlet distortion on stability.
separations, armament firing, aircraft maneuvers, or take off This is because, in a linear system, the steady imposed
in cross wind. In any event, if we consider a compressor distortion and the self-excited propagating perturbation
operating in nonuniform flow at the same overall mass flow (rotating stall) do not interact (i.e., the interaction is second
rate as one in a uniform flow, it is clear that in the former case order and thus not included in a linear analysis). We will
some parts of the blading will be working at more unfavorable therefore concentrate on those approaches which do attempt
conditions than in the latter, and that some sections of the to assess the loss in stability due to inlet distortion. As in the
blading will be closer to the stall point. The overall effect of uniform inlet flow situation, the approaches again range from
distortion is therefore to degrade the overall performance of correlations to more basic analyses.
the machine and, in particular, to decrease the stall free range
of operation. Since the effects of distortion on compressor The most empirical approach makes use of so-called
stall onset can be substantial, it is useful to discuss the general "distortion indices" to describe the stability degradation to be
fluid mechanics of compressor operation with nonuniform expected from inlet distortion. The indices are based on the
flow. inlet total pressure distribution, and are a way to quantify the
effects of a given total pressure pattern. For a given engine (or
In the investigations that have been carried out on the effect compressor) a correlation will be built up by testing with
of inlet distortions, the flow nonuniformities are commonly different types of inlet distortion to give the maximum
divided into radially varying steady-state, circumferentially distortion that can be tolerated at a given corrected speed.
varying steady-state, and unsteady distortions. In reality the Once this is done the stability margin can be assessed in
distortions encountered are combinations of two or possibly various other situations by converting the total pressure
all three of these types, but significant progress has been made distortion in that case into an equivalent distortion index.
using the above simplifications since, in many situations, the
principal loss in stall margin can be regarded as due to one of Much effort has been spent on these correlations for
the three. For example, during changes of aircraft attitude the engines and they have been developed to take into account the
inlet distortion may vary significantly with time over a time fact that the distortions in general have a radial distribution to
scale of the order of tenths of seconds or longer, which is them as well as can contain a reasonable amount of un-
many times longer than the time for a fluid particle to move steadiness. Descriptions of the basic methodology are given
through the compressor. The distortion can therefore be by Hercock and Williams [38] and by Collins [39], which also
considered as if it were steady-state. In analyzing the response contains a list of other references on the subject. We will not
of an axial flow compressor to an inlet flow distortion, it is discuss these correlations much further although there is one
also essential to recognize the strong interaction that exists point which should be mentioned. There are situations in
between the compressor and the distorted flow field. Put which the inlet total pressure distortion can be varying with
another way, the compressor does not passively accept the time so that the instantaneous distortion factor undergoes
distortion but plays an active role in determining the velocity very rapid changes. It is found in these situations that
distribution that will occur at the compressor face, which is disturbances which have a time duration of less than roughly a
what the individual compressor airfoils actually respond to. rotor revolution have little effect on compressor stability [39].
Thus, for these higher frequency distortions, the unsteady
We will very briefly discuss the pure radial distortion first. response of the compressor appears to be beneficial as regards
Typically this is analyzed using the axisymmetric procedures the onset of instability. We will discuss this important point
that have been developed for a uniform inlet flow, i.e., the further below.
distortion is essentially viewed as an off-design situation. (In At a somewhat more fundamental level of assessing
doing this, one comment that can be made is that distortions compressor stability with inlet distortion are the so-called
encountered in practice can have substantially more severe parallel comperessor model and its extensions. Descriptions
gradients than do flows that arise solely from off-design of the basic ideas and assumption of this model are given by
operation. Hence, the calculations may have to be carried out Reid [40], Stenning [36], and Mokelke [41]. The approach is
using stations within the blade rows as well as at inlet and to view a compressor operating in a circumferentially
exit). The criteria that would then be applied to predict the nonuniform flow as two compressors operating "in parallel"
onset of stall are based on the correlations referred to earlier. with each performing at a condition corresponding to the
Therefore this type of distortion really does introduce any local value of mass flow. If the flow geometry downstream of
new phenomena from those mentioned previously, and we the compressor is a constant area annulus, and the flow angle
will not discuss it further. from the last stator is constant around the circumference, the
With circumferential distortion, however, the asymmetry two compressors can be taken to have the same exit static
of the flow does introduce a new element into the fluid pressure. In addition, because there is very small opportunity
dynamic analysis of the compressor behavior. There has been
a large amount of work on this topic. Much of it has been
carried out treating the distortion as a small amplitude per- The ratio of total pressure distortion at the exit to that at the inlet.

140/Vol. 102, JUNE 1980 Transactions of the ASME

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 ODOAkbPWAHPC
OOPWA1PC
OHGTE 118
COMPRESSOR Q ROLLSROVCE ^
CHARACTERISTIC SURGE
PRESSURE
RATIO
AVG - A
CHANGE
OPERATING h,
POINT T %

0 0.05 0.1 0.15 0.2 0.25 0.3

o MATCHING CONDITIONS
INLET PT DISTORTION AMPLITUDE
®BOUNDARY MAX - MIN/AVG
CONDITION Fig. 11 Sensitivity to 180 degree circumferential distortion [45]
I
Fig. 10 Circumferential distortion analysis calculation procedure for 0.18 I
180 degree distortion ® UNIFORM INLET
\ \ ^^"l
— 180° DISTORTION
0.14
for circumferential cross flows within the compressor, the n i
ROTOR
mass flow distribution at inlet and exit can be taken to be the J
LOSS 0.10
same — i.e., there is no mixing of the flow in the two *\l*»
"segments" of the compressor. The overall mass flow, the APT/Q 9 s
two boundary conditions of specified local inlet total pressure _ UNIFORM V \

X
0.06
distribution and constant exit static pressure, plus the idea INLET S T A L L - ! @9
that each segment is performing at some local point on the •\
0.02
uniform flow characteristic, are enough to define the local i i h T I 1 1

mass flow distribution at the compressor inlet and thus the II .0 20.0 22.0 24.0 26.0 28.0 30.0 32.0 34.0
distorted flow performance. A schematic of this procedure is ROTOR INLET AIR ANGLE - DEGREES
shown in Fig. 10 for a 180 degree square wave distortion in Fig. 12 Unsteady rotor response [45]
inlet total pressure. The model is not restricted to small
amplitude nonuniformities, so effects due to the curvature of
the compressor characteristic can be included. It should be response of the rotor can be considerably different than that
noted that these models are also capable of treating inlet total predicted by a quasi-steady analysis, especially near the stall
temperature distortions as well as total pressure [41], [42]. point. As one example of this, Fig. 12 shows the (relative)
In order to predict the stall point, however, some additional total pressure loss coefficient for a rotor operating in a 180
criterion must be invoked since the parallel compressor model degree distortion [45]. The horizontal axis is inlet flow angle
is steady-state. The simplest is that the onset of rotating stall (measured from the tangential direction, so that decreasing @x
will occur when the flow in the low velocity region (point L in means increasing angle of attack) and the vertical axis is the
Fig. 10) is reduced to a level at which the machine will stall in rotor total pressure loss coefficient. It can be seen that the loss
a uniform flow. This concept, plus the basic parallel com- coefficient exhibits an unsteady "loop" as a function of inlet
pressor ideas, leads one to the general design philosophy that angle as the rotor moves through the distortion. The values
steep speedlines are desired for distortion tolerant com- obtained in a uniform flow test of the rotor are also shown
pressors [43], [44]. and it can be seen that there are substantial differences bet-
A comparison of the predictions of this model with test ween the steady and unsteady performance.
data is shown in Fig. 11, which presents the loss in stall A further very important point which is illustrated by this
pressure ratio as a function of the amplitude of the total figure is that with distortion the rotor is operating transiently
pressure distortion, for 180 degree "square wave" distortions at a flow angle that is beyond the uniform flow stall point.
[45], The solid line is the theoretical prediction given by the That is, excursions to higher angles of attack (i.e., lower flow
basic parallel compressor theory. It can be seen that although rates) than could be tolerated with a uniform flow can be
the model gives reasonable qualitative trends, on a quan- tolerated locally (transiently) in the distorted case. This is in
titative basis the predictions leave considerable scope for direct contrast to the hypothesis used in the basic parallel
improvement. compressor model for defining the stall point, and the un-
For a distortion of smaller extent this discrepancy is in- steady response appears to be an important feature to include
creased. For these distortions the theory predicts that there in developing predictive methods for circumferential inlet
will be a greater loss in stall margin (compared to the uniform distortion.
flow case) than is observed with the 180 degree distortion. There are several different approaches to doing this. In the
(This is because the circumferential segment with the lowest first, the idea of a critical angle of distortion has been
axial velocity is assumed to be responsible for the stall, developed [38], [40], [44]. This is essentially an angular extent
whereas the stall pressure ratio and flow are calculated as the of distortion below which "further loss of stall pressure ratio
average.) Experimental results are in substantial disagreement for a given distortion intensity is negligible" [40]. Thus, the
with this, however, since it is found that distortions of small procedure is to base the inlet distortion intensity on the
circumferential extent have little effect on stall line. average total pressure in a sector, of the extent of the critical
The reason for this, and one of the major defects of the angle, surrounding the lowest pressure. This constitutes an
basic parallel compressor theory, appears to be associated extension of the basic parallel compressor stall line hypothesis
with the assumption that the performance of the compressor to account for unsteady response. However, it is necessary to
can be described by the quasi-steady application of the local evaluate the critical angle for each individual compressor, and
uniform flow performance curve. However, with a cir- quite different values have been reported in the literature [43].
cumferential distortion, the rotor passes through a spatially In an alternative approach [45], a one-dimensional model of
nonuniform flow field and thus sees an unsteady flow. The the unsteady rotor response is incorporated in a multisegment

Journal of Fluids Engineering JUNE 1980, Vol. 102/141

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 parallel compressor model (roughly thirty segments are used).
This model can have some segments operating at flows below
the uniform flow stall point.
Better prediction of stall than with the basic model can be
achieved using such approaches. As an example, the use of
one of these methods (the multisegment parallel compressor
analysis) for prediction of the distorted flow stall line in a two
spool turbofan engine is described by Braithwaite and Soeder
[42]. It should be emphasized, however, that adequate
definition of the uniform flow performance curve (including
stall point) is a prerequisite for successful use of any of these
methods.
The third class of studies of the problem of compressor
Nozzle Constant Diffuser
stability in a distorted flow is the numerical solution of the exit area exit exit
time dependent nonlinear inviscid equations of motion. These Fig. 13 Effect of downstream components on distortion transfer [23]
describe the flow field from a specified initial state to a final
"steady-state solution" similar to that described earlier with
the uniform inlet flow. As with the uniform flow situation the compressor downstream, as is usually the case in a modern
analyses can either predict a steady flow or a large amplitude multi-spool gas turbine engine, the effect of the coupling can
propagating disturbance, which is now modulated by the be expected to be much stronger.
presence of the inlet total pressure distortion. They can
therefore give an indication of the stability point of the Coupled Compressor-Diffuser Flow Instability and Dif-
distorted flow. So far these have been applied to single rows fuser Instability. The above discussion of the coupling be-
or stages and have not really been thoroughly assessed against tween a compressor and a downstream component furnishes a
experimental data. A good example of this approach is the useful introduction to another aspect of the onset of in-
work by Adamczyk [46], which has been extended to the stability in compression systems. Consider the compressor
compressible flow regime by Pandolfi and Colasurdo [47]. stability boundary for a nominally uniform inlet flow con-
Again, one of the main limitations is the modeling of the dition, i.e., the inception point of rotating stall. This occurs
dynamic blade row performance, especially at flows near and when conditions are such that a circumferentially nonuniform
below the uniform flow stall point. There is not space in this flow perturbation in the compressor annulus can grow. Based
review to discuss some of the various experimental and on the previous section, it might be expected that this point
theoretical approaches to investigating unsteady blade row could be affected by components downstream of the com-
response. However, several useful review articles on unsteady pressor, since the downstream boundary condition on the
flows have appeared recently and the interested reader is flow perturbation is altered. The quantitative extent of
referred to these [48-50]. (In addition, one novel approach to changes in stability boundary is of course dependent not only
obtaining data at flows below the stall flow is that described in on the nature of the downstream component, but on the
[51].) circumferential length scale of the flow perturbation, since
this also determines how closely the compressor and the
downstream component are coupled [52]. For many situations
Compressor-Component Coupling. As stated previously, if of practical interest the predominant mode of instability
one considers the behavior of compressors in asymmetric occurs with a one-lobed type of disturbance so that the
flow, the general conclusion that emerges is that for the type relevant axial distance within which there can be a strong
of flow nonuniformities which have the greatest effect on interaction between components is on the order of the mean
stability, namely those with a circumferential length scale of radius of the machine.''
the order of the circumference of the compressor (i.e., a "one This generally means that, in terms of the stability bound-
lobed" type of pattern with a strong first Fourier com- ary, the compressor is not isolated from the influence of
ponent), the compressor and the downstream components downstream components. More detailed analytic and ex-
must be viewed as closely coupled. Hence the distorted flow perimental studies of this phenomenon bear out this general
performance of the compressor can be quite dependent on the conclusion [53]. As noted before this has important ap-
downstream component. plication in analysis of two spool compression engines. It also
An example of this is given in Fig. 13, which shows both emphasizes again that one cannot analyze the stability of a
predictions and experimental data from several research single row in a multistage compressor by itself as a guide to
compressor experiments [23]. The data are from three dif- the compressor stability, but must include the other stages
ferent tests with an inlet circumferential total pressure both upstream and downstream.
distortion. In these tests the only thing that was altered was This type of interaction between the pumping element and
the exit section, one test being conducted with an annular exit the downstream components also can be strongly manifested
nozzle downstream of the compressor, another with a con- in centrifugal compressors. In these situations the diffusers
stant area annulus exit section, and the third with an annular (vaned or vaneless) can be not only the cause of premature
exit diffuser downstream of the compressor. The vertical axis compressor stall, but can themselves exhibit rotating stall [54,
shows the compressor distortion transfer, which is the ratio of 55], In modern high pressure ratio centrifugal compressors, in
the magnitude of the total pressure nonuniformity at exit to fact, the initiation of system surge has been postulated to be
that at inlet, for the three different exit sections. It can be seen linked to the instabilities in the diffuser [56].
that the presence of different downstream ducts has a sub-
stantial effect on the compressor performance in distorted A Digression on Unsteady Response. We have discussed the
flow. This is quite contrary to the usual experience in fact that the compressor blading responds in an unsteady
axisymmetric flow where, for a given flow rate, the per- manner, and that this departure from quasi-steady behavior
formance of an upstream component is only weakly affected appears to be quite significant near stall. However, there is a
by downstream conditions. Note that the results shown in the
figure were obtained with downstream components that are In other words, the coupling can only be neglected for axial separations
all relatively passive fluid mechanical devices. With another that are significantly greater than this.

142/Vol. 102, JUNE 1980 Transactions of the ASME

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 200 1—
O
a STALL LINE WITH SOLID CASING
A ^
150 -
Circumferential l^o PRESSURE
oaS 1.4
extent 100 RATIO
• Symbol Percent N/-/W
of stall cell 4 3 ! m/sec
A / O IOQ 1.2
(Degrees) a 96
50 A 91 335 m/sec
81
1.0
o 60 70 80 90 100 110 120 130 140
i i 1 l l
0 FLOW
Onset of i i
Fig. 16
Rotor revolutions Stall margin increase due to casing treatment [43]; flow in
Ibm/s
Fig. 14 Circumferential growth of stalled region (three staae
pressor) [57]
transient may therefore differ substantially from the steady
state performance. This facet of unsteady response, which
BLADE may be quite important in the application of one-dimensional
ROTATION compression system models, for example, is another of the
features of the stall process that are at present not well un-
BLADE derstood, and on which there is little data.

Stability Enhancement Using Rotor Casing Treatment

AIRFLOW We have described some of the various types of instability,
commented on the criteria for determining their onset, and
discussed the adverse effects of inlet distortion on stall in-
ception. However, another area which is of interest concerns
techniques for enhancing the stability margin of a tur-
bomachine. The most straightforward of these is to achieve
the needed stability margin by matching the compression
system below its peak efficiency point (in effect setting the
match point so that the compressor blading has incidences
and pressure rises far below the maximum). Although this
BLADE MOTION would provide an increase in airfoil incidence range between
the operating line and the stability limit, it would also lead to
decreased efficiency on the (down-rated) operating line, which
is generally unacceptable.
SECT A-A One solution to this problem is the use of so-called "rotor
Fig. 15 Axial skewed groove casing treatment casing treatment" to improve the stability of compressors.
This casing treatment consists of grooves or perforations over
the tips of the rotors in an axial compressor or located on the
larger scale aspect of the unsteadiness which is also associated shroud in a centrifugal machine. Numerous investigations of
with the onset of rotating stall. these types of configuration have been carried out under
The time scale that one would tend to associate with blade widely varying flow conditions (e.g., references [58-65]), and
unsteady response is on the order of bl W where b is the blade these have demonstrated that the range of usefulness of these
chord and W is the relative velocity. However let us consider casing configurations extends from compressor operation in
the time scale that characterizes the transformation of an basically incompressible flow (relative Mach numbers of
axisymmetric flow to the severely nonuniform one of rotating roughly 0.15) to the supersonic flow regime (relative Mach
stall. A representative time scale might now be the disturbance numbers of 1.5). A schematic of one of the more successful of
wavelength (e.g., the mean circumference of the machine) these casing configurations is shown in Fig. 15. These grooves
divided by the through flow velocity. This is the time scale are known as axial skewed grooves. A typical improvement in
used in Fig. 7 and is much longer than that based on blade stall line brought about by use of these grooves is shown in
unsteady response. A qualitative physical argument for this Fig. 16 for a transonic axial fan. It is to be noted that far
scaling can be made by noting that the change from unstalled larger improvements have been seen and that casing treatment
to stalled flow involves the shedding of blade circulation of has also been used to inhibit instability in centrifugal com-
the same sign over a significant circumferential extent of the pressors [66, 67].
flow annulus. The flow will only approach a "fully Although the basic mechanism of operation of the casing
developed" state when this shed vorticity has been convected treatment has not yet been fully elucidated, some important
downstream some distance on the order of the disturbance points have emerged. If we refer back to the description of the
wavelength. onset of rotating stall, it is apparent that for the axisymmetric
These rough considerations imply that for a given machine flow to become unstable the blade passages must be operating
the stall cell growth time might scale with axial velocity (or at a condition such that they have the capacity to generate
since Cx/U is approximately constant at stall, with rotor large blockages for small changes in inlet conditions. What is
speed) and circumference. This has been found to be the case found, however, is that the level of blockage in the rotor
in one of the few instances where data is available, as shown passages can be greatly decreased due to the presence of these
in Fig. 14. This response of the overall flow field is therefore grooves. As an example of this, Fig. 17 shows data taken just
likely to occur on a time scale an order of magnitude or more downstream of an axial compressor rotor [68], The measured
longer than that associated with the individual blades. The quantity is the nondimensional relative frame total pressure
compressor performance (pressure rise, torque) during a stall loss across the rotor, i.e., the total pressure loss measured (by

Journal of Fluids Engineering JUNE 1980, Vol. 102/143

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 WITH GROOVED CASING important, experiments carried out using closed groove
WITH SMOOTH CASING (CASING TREATMENT) configurations at low speeds (where there can be no average
radial flow into the casing over a sizeable fraction of the
circumference) have also found substantial increases in stall
margin. Thus while the "radial relief" idea may be important
in some aspects of the problem it does not seem to be one of
the most critical features.
In contrast to this passage average approach, it may be
more useful to examine the flow in the grooves on a smaller
scale, i.e., on a blade to blade basis. In particular consider the
axial skewed groove configuration shown in Fig. 15. Note that
Fig. 17 Rotor sxit total pressure loss contours near stall [88] there will in general be a higher static pressure over the rear
part of the grooves than over the front part. One would
therefore expect that fluid would be forced down into the
a probe rotating with the rotor) in a coordinate system fixed groove at the rear and up as a jet at the front end. Such flows
to the rotating blades. Data are shown for both solid wall and have been observed by Takata et al., who measured jet
casing treatment, at a mass flow rate near the stall limit of the velocities that were comparable with free stream velocities
solid wall configuration. In this instance the onset of rotating [58], In the frame of reference attached to the blade these jets
stall appeared to be associated with a large blockage created have very high velocities (in a direction toward the pressure
near the end wall of the blade passage. The contours are side of the blades). The flow in and out of the grooves is thus
increments of 0.05 in relative total pressure loss across the a potent device for the transfer of momentum between the
rotor divided by inlet dynamic pressure. It can be seen that moving wall and the low momentum flow in the boundary
with the casing treatment there is a much smaller region of layer, thus increasing its total pressure (in the relative frame)
low total pressure fluid and hence a greatly decreased and decreasing the blockage. In this connection it is relevant
blockage in the blade passage. One would therefore expect to note that data presented in [59] (from a compressor which
that the tendency towards rotating stall would also be greatly was run with an axial skewed groove casing treatment)
suppressed and the stable flow range of the machine extended showed local blade surface pressures near the rotor tip that
with the grooved casing, which indeed was found to be the were higher than the relative total pressure. These overall
case. considerations, however, have not yet been developed in a
These general ideas concerning the effect of the grooved quantitative manner, and substantially more investigation is
casing on the flow blockage in the rotor passage can in fact be needed of the interaction between the flow in the grooves and
taken somewhat further. As described previously, the basic that in the endwall region.
physical picture presented for the onset of rotating stall is one
that is dependent on the level of flow blockage in the blade Although casing treatment is a potent remedy in increasing
passages. In many instances, however, one can make a useful stall range, it should be emphasized that it is not a panacea. It
distinction between those situations in which the large has been found that those known treatments which have the
blockages are associated with the endwall boundary layer and most success in improving stall range generally have some
those in which the major part of the blockage is associated penalty in efficiency associated with them. This situation gives
with the blade boundary layers away from the endwall. These a strong motivation to understand the basic fluid mechanics
two situations are referred to as wall stall and blade stall of operation of casing treatment in order to enable one to
respectively. design optimum treatment configurations.
There are also situations in which casing treatment is
Based on observations of widely varying effects of casing unlikely to be successful. One has already been identified in
treatment on compressors exhibiting what appeared to be the discussion of wall stall and blade stall, and a more obvious
blade stall and those encountering wall stall, a series of ex- instance is when the stall occurs at the hub. In this situation,
periments were carried out using two different rotor builds, the flow shift towards the tip region, due to the casing
one designed so that a blade stall was encountered, the other treatment, can actually decrease the stall margin, since it will
designed for a wall stall [68]. The rotors were tested both with aggravate the hub stall condition. Finally, there can be
and without casing treatment. It was found that the casing mechanical and manufacturing difficulties associated with
treatment markedly improved the stall range for the wall stall casing treatment usage in production engines. Designers have
rotor (compared to the solid wall). In contrast, the blade stall therefore been reluctant to adopt casing treatment as a
rotor showed little change in stall point when the casing standard item.
treatment was used. These experimental findings are con-
sistent with the hypothesis that casing treatment is likely to be
effective only in a situation in which a wall stall exists, and Behavior Subsequent to Instability Initiation
thus furnish one general guideline for its use. We have, until this point, been concerned primarily with the
The experimental results mentioned have shown some of phenomena associated with the onset of instability. Another
the effects of the casing treatment on the overall flow field, very important problem is associated with the behavior
but as stated, the mechanism of operation is still unknown. subsequent to the onset of instability, where one of two modes
Different hypotheses have emerged but none is able to explain of system response, corresponding to surge or rotating stall,
all the experimental results satisfactorily. As one example, can be exhibited (see Fig. 3). For a system that goes directly
several investigations have invoked the concept of a radial into rotating stall, the transient behavior after the initial
flow through the porous casing, and calculations carried out instability consists of a rapid decrease in pressure rise and
including this idea in the basic models for predicting rotating flow, with eventual establishment of a new stable equilibrium
stall onset do show the possibility of a stabilizing influence of point (in rotating stall) at a greatly reduced flow and pressure
these radial flows [58], [65]. However, it should be noted that ratio. In surge, on the other hand, there is no new equilibrium
the analyses are based on the wavelength of the flow per- point and both the mass flow and pressure rise undergo large
turbations being much larger than the blade gap (i.e., treating amplitude oscillations.
the flow in the blade row on a passage averaged basis). While For an aircraft gas turbine engine, a key requirement is
such radial flows can occur in situations where there is a recovery from stalled conditions once instability occurs. In
plenum or external flow path, or where Compressibility is this connection, therefore, it is important to note that even

144/Vol. 102, JUNE 1980 Transactions of the ASME

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 1.0r

SURGE.
0.8 o—o—o-- ROTATING
-STALL-"
THEORY yr
0.6

0.4 LEGEND

< > V p • 34.9m"

D Vp > 22.8 m"

O Vp . 15.0 m"
0.2
A V P = 2.8 m' [NO SURGE FOUND A I THIS VOLUME}

0.2 0.4 0.6 0.2 0.4 0.6 0.8

Cx/U Cx/U
1000 2000 3000 4000
a) LARGE VOLUME b) SMALL VOLUME
CORRECTED SPEED ~ H//f LOW SPEED HIGH SPEED
Fig. 18 Surge/rotating stall boundary for different plenum volumes
[22] Fig. 19 Dependence of system response on B parameter [22]

compressor characteristic there is a critical value of B which

though surge may be associated with severe transient stresses determines whether the mode of instability will be surge or
due to the large amplitude flow variations, the engine does rotating stall. Systems with B above the critical value (e.g.,
operate in an unstalled condition over part of the surge cycle. speeds above a critical value) will exhibit surge oscillations,
Because of this, one is able to open bleed valves or make other while those having B lower than the critical (speeds below the
changes that will have an effect on bringing the system out of critical value) will undergo an initial transient to the (annulus
surge. In contrast to this, rotating stall, and in particular the averaged) steady flow and pressure rise associated with
full span, large extent rotating stall which is characteristic of a operation in rotating stall.
multistage compressor, can be very difficult to recover from, Experimental evaluation of this concept is shown in Fig. 18,
as will be discussed in detail below. Therefore it is the surge in which data is presented for a three stage compressor that
mode that is much more favorable for recovery in systems was run with several different downstream volumes [22]. The
such as aircraft gas turbine engines and is hence the more horizontal axis is corrected speed and the vertical axis is the B
desirable one. parameter. The open symbols mark the experimental values
On the other hand, for an industrial centrifugal compressor of B at which the change-over from rotating stall to surge
or pump, it may be that operation in rotating stall (if one must occurred for the different volumes tested. (The solid point
encounter some sort of instability at some point in the indicates the highest value of B that could be obtained with
operating range) can be tolerated, and that throttle movement the smallest volume used.) The regions in which one en-
is wide enough so that recovery is readily achieved. In this counters surge (B above the critical value) or rotating stall (B
situation it may be that the often violent oscillations that below critical) as the mode of instability are also indicated. It
occur during surge make this the less desirable of the two can be seen that the prediction of a constant B as the bound-
modes. Whatever the situation, however, it is apparent that ary between surge and rotating stall is well borne out,
an important question is whether a given pumping system will although the value at which the change occurs is somewhat
exhibit large amplitude oscillations of mass flow and pressure above that predicted by the theory. (It should be again em-
ratio (surge), or whether the system will operate in rotating phasized that the value of 0.8 is for this particular com-
stall where the annulus average mass flow and pressure ratio pressor.)
are essentially steady, but are greatly reduced from the pre- Although it is necessary to carry out the calculations for
stall values. any specific case, some motivation for this general result may
To answer this, one must analyze the nonlinear system be gained by the following qualitative physical arguments.
behavior. This was done by Greitzer [57] using the Helmholtz Notice that we can write B as
resonator type of compression system model introduced by
Emmons (for the linear case) and pictured in Fig. 9. The
analysis shows that for a given compression system i.e., B=
pU<jiLcAc
specified compressor characteristic, plenum volume, com-
pressor length, etc., there is an important nondimensional For a given compressor, the numerator, which is proportional
parameter on which the system response depends. This to the magnitude of the pressure difference across the duct
parameter is denoted as B: Cppienum--patmosPheric). represents the driving force for the
acceleration of the fluid in the duct. If we consider oscillations
B= U that are essentially sinusoidal in character, the order of
2oiLr
magnitude of the inertial forces that arise because of these
where u is the Helmholtz resonator frequency of the system, local fluid accelerations will be given by the product of the
Lc is an "effective length" of the compressor duct, and U is fluid density, p, the frequency of the oscillations, co, the
the rotor speed. If we insert the definition of o> we can define amplitude of the axial velocity fluctuation and the volume of
B in terms of geometric and physical system parameters as the fluid in the duct, LCAC. Hence if we assume that the axial
velocity fluctuation is a specified fraction of the mean wheel
speed the magnitude of these inertial forces will be propor-
U tional to the denominator. The ratio of the two forces
B=
2a v ArLr (pressure to inertial) is therefore proportional to B. Thus as B
is increased, for example by increasing the rotor speed, a fixed
where a is the speed of sound, Ac is the compressor flow- percent amplitude of the compressor axial velocity oscillations
will result in inertial forces that are relatively smaller and
through area and Vp is the plenum volume. For a given

Journal of Fluids Engineering JUNE 1980, Vol. 102/145

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 smaller compared to the driving force due to the pressure CASE I CASEH
differential. The capability to accelerate the fluid in the duct is NET ENERGY INPUT NET ENERGY DISSIPATION
thus increased as B increases. Hence as B becomes larger one COMPRESSOR
CHASACmtlSJIC
would expect greater excursions in axial velocity and thus a
AP
general trend toward surge rather than rotating stall, and this
is in accord with the experimental results.
The B parameter is useful not only in defining the boundary
between surge and rotating stall but also as a guide to the
scaling of the overall transient behavior of a compression
system. In other words, for the same value of B, a com-
pression system should exhibit the same transient behavior,
regardless of whether this value has been obtained using a
large volume and a low speed or a small volume and a high SAP
speed. Figure 19 shows a comparison of the measured
transient response for two systems at approximately the same
value of B, one with a large volume and low speed and one Sm x SAP
with a small volume and high speed, with the value of B high
enough such that surge occurred [22]. The figures show the
instantaneous annulus averaged axial flow parametr (Cx/U) Fig. 20 Physical mechanism for surge oscillations
versus nondimensional system pressure rise, i.e., the in-
stantaneous system operating point. The steady-state com-
pressor curves are also shown for reference. As can be seen were in steady operation at the mean flow rate. (In a similar
the two surge cycles show extremely similar behavior, em- fashion the net dissipation due to the throttle will also be
phasizing the influence of B as the relevant nondimensional higher than if the system were in steady operation.) When the
parameter for the phenomena under study. Although this net energy input over a cycle balances the dissipation, a
figure does not show overall flow reversal, it should be em- periodic oscillation can be maintained. In the case of an
phasized that with high pressure ratio multistage compressors, operating point on the negatively sloped region, as shown in
periods of reversed flow during the surge cycle are to be the right-hand side of the figure, the compressor actually puts
expected. in less mechanical energy over a cycle than in steady
operation, (since high mass flow is associated with low energy
We have said that the question of whether surge or rotating addition, and conversely) and no oscillations are possible. To
stall will occur is not only dependent on B but upon the summarize, (surge) oscillations are possible only when the
compressor characteristic. A basic attempt to take into ac- mechanical energy input from the compressor is greater
count one aspect of this, the number of stages (TV) leads to the during the oscillatory flow than during a mean (steady) flow,
use of NB as the relevant parameter for scaling rather than B. and this can occur only if the characteristic is somewhere
The motivation for this is that for N identical stages and a positively sloped so that high mass flow and high mechanical
given value of B the pressure rise will be approximately energy input per unit mass go together.
proportional to TV, so that the ratio of pressure to inertia
forces will increase as the number of stages. There is thus a We have discussed some of the physical features of surge,
strong trend toward lower values of B needed to encounter and we now turn to examination of the other consequence of
surge, rather than rotating stall, as the number of stages and compressor instability - rotating stall. We will look at some of
overall pressure rise of the machine is increased. the overall features of this regime, then discuss some of the
measurements on the structure of the flow field that have been
carried out, and then examine a basic method for predicting
Physical Mechanism for Compression System Oscillations. compressor performance in rotating stall - particularly as
We have described the necessity for certain system parameters regards the extent of the stall/unstall hysteresis loops.
to be of a specified magnitude in order to encounter surge.
One other aspect concerns the position of the compressor
operating point in order to have surge cycles. In particular, in General Features of Rotating Stall in Axial Com-
order for a surge cycle oscillation to occur, the compressor pressors
slope must be positive over some part of the surge cycle. The
physical mechanism associated with this can be seen by Flow Regimes. It is useful first to describe the flow regimes
considering a compression system undergoing oscillation encountered during compressor operation in rotating stall and
about a mean operating point. Since dissipation is occurring to relate them to the changes in compressor performance that
due to the presence of the throttle, there must be energy put can be expected. Let us consider a hypothetical compressor,
into the system to maintain this oscillation, and the only of, say, three stages, and examine the performance curves
source available to do this is the compressor. Thus let us plotted in the form of \j/TS versus </>, where i/TS = (exit static
examine the mass flow and pressure rise perturbation through pressure - inlet total pressure)//?^, and <f> = Cx/U, Two
the compressor. These are shown in Fig. 20 which presents the possibilities are shown in Fig. 21(a) and 21(b), which can be
perturbations in mass flow {m) and pressure rise (AP) through regarded as being representative of data from a number of
the compressor, plotted versus time over a period of one actual compressor tests. Figure 21(a) shows a compressor
cycle. The product of the two, 8m x 8Ap, which is propor- whose performance curve is either continuous or has only a
tional to the net excess (over the steady state value) of small discontinuity in pressure rise (a few percent) at the stall
production of mechanical energy is also shown. point, A. This behavior, where there is a very gradual drop in
In the case of a positive slope it is found that favorable delivery pressure (or often no drop) at the inception of stall, is
conditions for the energy addition occur since high mass flow associated with the compressor operating with one or more
rate and high rate of mechanical energy addition (in the form stall cells that do not cover the total height of the annulus.
of pressure rise) go together. As shown in the figure the This is known as part span stall, and an indication of a typical
product of the two is positive definite over the whole cycle. configuration is shown in the figure where we see two regions
Thus the net amount of mechanical energy that the com- of severely retarded flow, i.e., two stall cells, at the tip. It is of
pressor puts into the flow will be higher than if the system course not always true that the cells appear at the tip; they can

146/Vol. 102, JUNE 1980 Transactions of the ASME

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 4> 4> Fig. 22 Sketch of stall cell structure (drawn in absolute reference
a) Part span stall b) Full span stall frame)
Fig. 21 Compressor characteristics showing rotating stall regimes
of changes in the compressor output, since the former is
associated with a relatively minor deterioration of the
also be found at the hub, and the picture should just be compressor performance from the unstalled state, but the
regarded as giving one typical possibility. latter can cause severe reductions in performance. On a fluid
Part span stalls are also often seen during low speed mechanical basis the occurrence of these different types of
operation of multistage high pressure ratio compressors. behavior is due to operation in different regimes of rotating
Under these conditions, which can occur to the right of (i.e., stall. In order to predict the compressor performance sub-
below) the stall line shown in Fig. 1, there can be severe sequent to the onset of stall it is therefore important to be able
mismatching between the front and rear of the compressor, to predict which type of rotating stall, part span or full span,
and the overall characteristic may thus still be quite negatively will occur at the stall limit, since this is directly related to the
sloped. occurrence and size of the hysteresis region.
As the throttle is closed from the stall point, and the mass
flow through the compressor illustrated in Fig. 21(a) is further Structure of the Stalled Flow Field. There have been many
decreased, the performance curve can exhibit a large experimental investigations of the features of the flow fields
discontinuity where the pressure rise and mass flow jump to associated with compressor performance in rotating stall.
significantly reduced values; this occurs from point B to C These have included flow visualization [69], studies of the
on the figure. This jump is associated with a change in the three-dimensional velocity field [70, 71], as well as the use of
type of stall. At point C there is one single cell, occupying a instrumented blades to measure the time varying forces on the
sizable fraction of the annulus and extending over the full blades [72, 73].
annulus height. This regime is known as full span stall. Just as in the theoretical treatments of stall and stall in-
Further throttling causes this cell to increase in size with the ception, much of the detailed experimental work considers
delivery pressure remaining relatively constant from point C isolated rotors. However, it has become apparent that the
to D. As the mass flow approaches zero, the stall cell can grow constraints on the flow field in a multistage machine are quite
to fill the annulus so that the flow can become basically different from those in an isolated rotor. For this reason we
axisymmetric with the pressure rise often dropping off will concentrate here on one of the few investigations of
slightly. If we were to once again open the throttle, we would rotating stall in a multistage environment, which was by Day
find that the mass flow at which the compressor left the full and Cumpsty [21]. They used an ensemble averaging
span stall regime, point E, was different from that at which it technique which was triggered on each passage of the stall cell
entered. However, this hysteresis is usually negligible for the so that they could average the results of many revolutions of
onset and cessation of part span stall. the cell to provide detailed definition of stall cell properties.
If we examine Fig. 21(6) we see a somewhat different Their investigations were carried out on several different three
picture. The large discontinuity in pressure rise and flow now stage compressors having quite different design values of
occurs right at the stall limit (point A). The sharp drop in both axial velocity parameter, (Cx/U)*, so the effect of this
these quantities as the operating point jumps from point A to important parameter could be clearly seen.
point B in this figure is associated with the compressor going Their paper, which addressed the full span stall problem,
directly into single-cell, full span stall, as indicated shows that in most cases the flow in the compressor can be
schematically. Further throttling causes the stall cell area to divided into distinct areas of stalled and unstalled flow. To a
grow, although it may not reach 100 percent of the annulus reasonable approximation the flow in the unstalled area
area even at zero mass flow through the compressor, with the behaves as it would if there were no stall cell at all. In the stall
pressure rise being relatively constant from point B to cells the fluid velocities ahead of the rotors were near blade
"shutoff" at point C. If one opens the throttle it is found that speed and in the direction of rotation, while behind the rotor
there is a substantial hysteresis between the onset and the the velocities were much lower. The axial velocities in the cells
cessation of stall, in that the throttle area has to be increased were small compared with either blade speed or the unstalled
to a significantly larger value than that associated with stall axial velocity; their precise magnitude and direction, however,
onset, in order for the machine to unstall. It is in fact this varied from compressor to compressor. In addition it was
large hysteresis which is responsible for the difficulty in emphasized that the full span stall cell extends axially through
recovering from a stalled condition.12 the compressor. The basic reason for this is that the axial
The two types of behavior at the stall limit point shown in spacing between blades is quite small compared to the width
Fig. 21(a) and 21(b) have also been termed progressive and of the stall cell. Since a mass flow defect (which is what a stall
abrupt stall respectively. This aptly describes the stall in terms cell is) in the front of the compressor can only be "filled in"
by circumferential cross flows between the blade rows, the
close spacing means that this defect will extend straight
This hysteresis between stall and unstall has, in the past, been held through the compressor (this is the same concept that is used
responsible for the existence of surge cycle oscillations. This is not correct and
hysteresis in fact inhibits the occurrence of the surge cycles. in the modeling of circumferential distortion).

JUNE 1980, Vol. 102/147

Downloaded From: Journal of Fluids Engineering
[Link] on 06/30/2019 Terms of Use: [Link]
 LOW<ft* INTERMEDIATE (/>* HIGH

unstalled stall unstalled stall unstalled stall

flow, cell f lowN cell flow /cell
Average /
v z
f-- -r 1 Stalled {$), \ Installed
I I /side / / side
200- I I .i 200 (*u
0.11 N C- *
l i I s <o
100 .-I U..
«w 1 / Throttle line

i-5--"-! r-1 0 Unstalled region

a) Ahead of rotor Compressor flow field Performance map

Fig. 24 Compressor operation in rotating stall
1
i I i .-
•1
200 ! I 200 1 *• 1 200 1.0
! i I 100 Predicted
100 100
- k_J- L 0.8
\
0 I 0 J_ -o""oo
„ - ° t r - „ - -<VoOo
n—°- C
' °O"O" "5
-oo°o V " - "
0 o
0° 360° " 0° 360° 0° 360° ^cessation , J
< £ u ( a t 0 . 1 1 N) o0
Circumferential angle
b) Behind rotor
Fig. 23 Flow deflection measurements in three compressor builds [21]

Day and Cumpsty emphasize that the stall cell is very 0 10 20 30

different from the wake of a bluff body, as many previous
investigators have assumed. If this were the case, then in a Compressor
coordinate system fixed to the cell the edges of the cell would Fig. 25 Correlation for <j> at cessation of full span stall [27]; numbers
on horizontal axis refer to individual compressors
coincide with the steady streamlines in the unstalled flow
surrounding the wake. Such a description has been proposed
by Rannie [74], for example, and has also been used as a the full span, large extent type, because this is the type which
model by Fabri for an isolated rotor [18]. The measurements, has the most severe effects on compressor performance,
however, show clearly that were this to occur the cell would especially as far as the stall/unstall hysteresis. However, with
need to be helical, whereas it is essentially axial. The picture regard to rotating stall in fans and single stage compressors
that emerges, therefore, of the instantaneous streamlines as (where part span stall is likely to be encountered) there have
viewed in an absolute coordinate system is shown in Fig. 22. also been investigations of the part span stall regime [75],
There is, in fact, mass transport across the cell boundaries, so including one study of the stalled flow in a transonic com-
that fluid that was in the unstalled region (with a high axial pressor stage [76].
velocity) is violently decelerated as it enters the cell, whereas
particles which were in the cell are accelerated at the cell edges Prediction of Compressor Performance in Rotating Stall.
as they enter the unstalled flow. Therefore it should be em- The prediction of the compressor flow regimes in rotating
phasized that the "wake" models of stall cell flow cannot be stall is important for several reasons. One consideration can
applied to the single or multistage compressor situation. be that the speed and number of stall cells determine the
Some idea of the magnitude of the changes in flow angle forcing frequency for blade vibration and it is desirable to
that can occur is shown in Fig. 23 which shows the cir- design the blades such that no resonance occurs. However, a
cumferential distribution of flow angle (measured from axial) very much more pressing consideration is that of stall
ahead of and behind the first rotor in a three stage com- recovery, i.e., the prediction of overall performance including
pressor. The measurements were made at a radial location two the hysteresis loop.
thirds of the span from the hub. Data is shown from three In an approach to this problem Day, Greitzer, and Cump-
builds — a low (Cx/U)* compressor, one with an intermediate sty [27] developed a correlation to predict stalled flow per-
value, and one with a high value. It can be seen that the stall formance. The correlation is based on a heuristic fluid
cell flow is predominantly circumferential ahead of the rotor. dynamic model of the compressor flow field in rotating stall.
Although not shown, this pattern is repeated stage-by-stage The compressor is divided into a stalled and unstalled zone
through the machine. In addition, because of the high values analogous to the concept developed for inlet distortion. The
of swirl velocity, the centrifugal effects in the blade passages stall cell is modeled by a zone of zero flow, and in the un-
are very important. It is therefore suggested that analyses stalled part of the flow the compressor is assumed to operate
which do not model these effects may be more applicable to at a point on the unstalled performance curve. The situation is
prediction of rotating stall in stationary rather than rotating shown in Fig. 24 where we see the stalled "zone" of the
blade rows. compressor at S and the unstalled zone at the same pressure
Another important conclusion that was stressed in [21] is rise at U, with the mean operating point defined by the in-
that "although the flow in a single stage build shows tersection of the horizontal line from S to U and the throttle
similarities with that in a multistage configuration of the same curve at 0. This model is applied in conjunction with two
blading, the finer details of the flow are sufficiently different experimental observations: 1) (following an observation
to preclude the extrapolation of single stage data to multi- originally due to McKenzie [77]) the nondimensional inlet
stage configuration." The thesis by Day [24] discusses in total to exit static pressure rise per stage in rotating stall is
detail these differences between the rotating stall behavior of constant independent of the unstalled performance, {\j/TS =
isolated rotors (on which many previous investigations were 0.1 \N in stall, where N is the number of stages), 2) there is a
focussed) and multistage compressors. critical value of stall cell blockage, X, (i.e., extent of annulus
The emphasis in this discussion of rotating stall has been on covered by the cell) below which full span stall cannot exist,

148/Vol. 102, JUNE 1980 Transactions of the ASME

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 No hysteresis found Minimal hysteresis hub tip ratio machines is either wrong or else is a gross over-
<t>*=0.35 simplification" [21].
0.5
+.TS V Research Needs and Suggestions for Future Work
We can now review those areas which the author thinks are
fruitful to pursue. On the general topic of stall prediction in a
Hysteresis uniform flow, there appears to be little more to be gained
from purely two-dimensional linearized treatments, but three-
dimensional aspects of stall inception are subjects on which
work is needed. In particular, questions such as the impact of
wall stall versus blade stall on the stall point, as well as stall
onset in low hub/tip radius ratio compressors, have not been
really resolved. Approaches that are based on treating the
0.5 1.0 blade row or rows as a "black box," however, will still have
to contend with the capability of present axisymmetric flow
Fig. 26 Effect of 0* (design Cx/U) on stall/unstall hysteresis (three calculation procedures to predict the slopes of the uniform
stage compressors) [27] flow speedlines or loss characteristics near stall. Because of
this it may well be that a numerical treatment of the flow in
(Xcrit = 0.3). Using these, one can make predictions about the blade passage region would be of considerable use here in
whether a given compressor will exhibit full span or part span clarifying how the blockage changes with small changes in
stall as well as about the extent of the hysteresis loop. incidence angle near the stall point.
To understand the central idea of the model, consider a For distorted flows there is also potential for research. To
compressor operating in the full span stall regime. To recover, the present author, it seems that a disproportionate effort has
we must open the throttle. Referring to Fig. 24, this would gone into the relatively straightforward tasks of solving the
move point 0 to the right (points S and U remain fixed) as inviscid flow equations outside of the blade row, while the
more of the annulus becomes unstalled and the blockage simple models that are in use for the fluid dyamics in the blade
decreases. When the critical value of blockage is reached (at row are those that were presented by Emmons [3] and
0.3) full span stall can no longer exist and the compressor Stenning [69] approximately twenty-five years ago. In fact,
comes out of rotating stall. Since the blockage can be directly one still has to essentially sketch that part of the loss and
related to the ratio SO/SU, which is just </>/$unstaiiCd > by turning characteristics associated with large angles of attack.
Thus, the author feels that more effort should be spent
^/^unstalled = 1 ~ X, concentrating on the more difficult problem of understanding
we can obtain a formula for the value of <j> at stall cessation ifthe unsteady rotor response in the heavily loaded (near stall)
we know the level of the horizontal line SU, since this picks regime. Probably both experimental and numerical in-
out 0Unstaiied 0 n t n e (known) unstalled part of the compressor vestigations will be needed for this.
characteristic. As stated, from experimental observations this There is, in addition, a need to develop better stability
can be placed at a level of 0.1 IN where N is the number of criteria in the case of distorted inlet flows. As one example, it
stages. Since the critical value of blockage at which the full would be useful to attempt to formulate the linearized
span stall ceases is 0.3 we have, for stall cessation stability problem for a situation where the "mean flow" is a
steady finite inlet distortion. Such an analysis might give some
<t>a additional insight into the basic fluid mechanics of the onset
— =0.7 of instability with the inlet distortion.
[Link]/V
Although not discussed in the review, it should be noted
This correlation is shown in Fig. 25. Data are presented from that there are distortions that are characterized by large
thirty-four different single and (mostly) multistage low speed variations in inlet flow angles rather than by nonuniformities
compressors. It can be seen that the correlation furnishes a in total pressure. The inlet vortex (or ground vortex) is an
very useful method for predicting stall cessation. important example of this type of distortion, which can have a
Using this basic procedure, parametric studies can be significant effect on stability [78]. At present no theory exists
carried out to determine the effect of different design to adequately describe this type of inlet nonuniformity.
parameters on stall performance. As shown in [27], two On stability enhancement using casing treatment, there is
important parameters are (Cx/U)* (design value of axial considerable scope for useful work. As stated, the
velocity parameter) and number of stages. For a given design mechanism(s?) by which the treatment decreases the passage
value of Cx/U the larger the number of stages the larger the blockage is (are) still unclear and there is a need for both
size of the stall/unstall hysteresis. However, a more potent theoretical and experimental work on this topic.
effect is the value of (Cx/U)*\ the higher the value for a given There are also important areas which are associated with
number of stages the larger the size of the stall/unstall the behavior subsequent to the onset of the initial instability.
hysteresis loop. Experimental evidence to support this idea is More accurate models are needed to predict the local details
shown in Fig. 26 which presents data for four different three of the surge phenomenon in multistage compressors. This is
stage compressors with different values of (Cx/U)*. At the important from an aeroelastic standpoint as well as from
lowest value of (Cx/U)* (curve I) no hysteresis could be purely aerodynamic grounds.
found. As we examine curves II, III, and IV which have in- Prediction of the features of rotating stall in multistage
creasing values of (Cx/U)* we find an increase in the extent of compressors is also needed. We can discuss first the overall
hysteresis. performance characteristic. The data reported in [27] and the
The procedure described in [27] also addresses the question subsequent discussion by Harman [79] seem to bear out the
of whether a given machine will exhibit part span stall or full hypothesis that the inlet total to exit static pressure rise in
span stall at the stall onset. Only one aspect of this will be rotating stall is roughly constant per stage independent of
pointed out here—namely that the same machine can develop unstalled performance, but there is no understanding of why
either part span or full span stall and "the widely held view this occurs. Further, even though the pressure rise is in-
that part span and full span are the modes for low and high dependent of the unstalled performance, the torque does

Journal of Fluids Engineering JUNE 1980, Vol. 102/149

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 depend strongly on details of stall behavior related to com- Propagating Stall in Axial-Flow Compressors," ASME Transactions, Apr
1954, pp.463-471.
pressor design parameters, and this is not understood. In- 2 Mazzawy, R. S., "Surge Induced Structural Loads in Gas Turbines,"
vestigation of these "global" properties should be pursued ASME Paper 79-GT-91, 1979.
since they directly affect the hysteresis phenomenon and hence 3 Emmons, H. W., Pearson, C. E., and Grant, H. P., "Compressor Surge
the recoverability of the compressor. A further facet of this and Stall Propagation," ASME Transactions, Vol. 27, Apr. 1955, pp. 455-469.
4 Taylor, E. S., "Evolution of the Jet Engine," Astronautics and
topic is that for aircraft engines which have two or more Aeronautics, Vol. 8, 1970, pp. 64-72.
spools, it is extremely important to understand the in- 5 Horlock, J. H., Axial Flow Compressors, Chapter 3, "Two Dimensional
teractions between the high and low pressure compressors and Cascades, Experimental Work," Krieger Publishers, 1973.
their effect on stalled flow performance. 6 Leiblein, S., "Experimental Flow in Two Dimensional Cascades,"
Chapter VI in Aerodynamic Design of Axial Flow Compressors, NASA SP-36,
The central features of the stall cell structure also are not 1965.
well modeled at present. Work is being carried out on this 7 Leiblein, S., "Loss and Stall Analysis of Compressor Cascades," ASME
topic, but present models are still rudimentary, compared to Journal of Basic Engineering, Vol. 81, Sept. 1959, pp. 387-400.
the experimentally measured details of the flow. Here again 8 Rodgers, C , "A Diffusion Factor Correlation for Centrifugal Impeller
an important facet seems to be the complex three- Stalling," ASME Journal of Engineering for Power, Vol. 100, Oct. 1978, pp.
592-603.
dimensional, separated unsteady flow in the blade passages. 9 Smith, L. H., Jr., "Casing Boundary Layers in Multistage Axial Flow
The blade row models in use for the calculation procedures Compressors," in Flow Research on Blading, edited by L. S. Dzung, Elsevier
are essentially the same as those used for the inlet distortion Publishing Co., 1970.
problem, however they are now being extrapolated even 10 Fligg, J. A., "Tests of a Low Speed Three-Stage Axial Flow Compressor
at Aspect Ratios of One, Two and Four," AIAA Paper 66-613, A1AA Second
further (to the negative flow regime). Again, the author feels Propulsion Joint Specialist Conference, 1966.
that more attention should be paid to the blade row 11 Smith, L. H., "The Effect of Tip Clearance on the Peak Pressure Rise of
characteristics, particularly in view of Takata's comment that Axial Flow Fans and Compressors," ASME Symposium on Stall, 1958.
the nonlinearities in the equations of motion do not seem to 12 McNair, R. E., "Tip Clearance Effects on Stalling Pressure Rise in Axial
Flow Compressors," ASME Symposium on Compressor Stall, Surge, and
have an important role in determining the wave shape, System Response, 1960.
disturbance amplitude, etc., but rather that these aspects are 13 Howell, A. R., and Calvert, W. J., " A New Stage Stacking Technique for
believed to be determined chiefly through the nonlinear ef- Axial-Flow Compressor Performance Prediction," ASME Journal of
fects due to the blade row characteristic [25]. Engineering for Power, Vol. 100, Oct. 1978, pp. 698-703.
14 Mikolajczak, A. A., "Introduction" to Unsteady Flows in Jet Engines,
Work also is needed in the area of overall compression Project Squid/AFOSR/UARL Workshop at United Aircraft Research
system modeling, particularly for high pressure ratio Laboratories, ed. by F. O. Carta, 1974.
multistage compressors. Systems which include two streams 15 Nenni, J. P., and Ludwig, G. R., " A Theory to Predict the Inception of
and two spools (i.e., turbofans) are an important application Rotating Stall in Axial Flow Compressors," AIAA Paper 74-528, AIAA 7th
Fluid and Plasma Dynamics Conference, 1974.
of this work. In this context, it should be noted that some of 16 Fabri, J., "Growth of a Perturbation in an Axial Flow Compressor,"
the transients associated with surge in modern gas turbine ASME Paper 78-GT-30, 1978.
compressors can be on the same time scale as rotor rotation, 17 Emmons, H. W., Kronauer, R. E., and Rockett, J. A., "A Survey of
or particle flow-through time. Under such conditions the Stall Propagation — Experiment and Theory," ASME Journal of Basic
Engineering, Vol. 81, Sept. 1959, pp. 409-416.
compressor response can be far from a quasi-steady one. This 18 Fabri, J., "Rotating Stall in Axial Flow Compressors," Internal
is true for transients both from unstalled to negative flow as Aerodynamics (Turbomachinery), Institution of Mechanical Engineers, 1967.
well as from negative flow to unstalled. Thus (again) if one 19 Stenning, A. H., "Rotating Stall and Surge," Chapter 15 in Fluid
wants to do significantly better than present models, the Dynamics of Turbomachinery, ASME Lecture Course, Iowa State University,
1973.
effort should be in trying to understand the overall unsteady 20 Dunham, J., "Non Axisymmetric Flows in Axial Flow Compressors,"
performance of the blading. Mechanical Engineering Sciences Monograph No. 3, Institution of Mechanical
As a summation, it is interesting to refer to a survey paper Engineers, 1963.
on aircraft engines written by Hawthorne over twenty years 21 Day, I. J., and Cumpsty, N. A., "The Measurement and Interpretation
of Flow within Rotating Stall Cells in Axial Compressors," Journal of
ago [80]. He stated: "Stall propagation is undoubtedly a Mechanical Engineering Sciences, Vol. 20, 1978, pp. 101-114.
phenomenon of the greatest importance to the engine 22 Greitzer, E. M., "Surge and Rotating Stall in Axial Flow Compressors.
aerodyamics and is one in which our understanding is still in Part Ii: Experimental Results and Comparisons with Theory," ASME Journal
the preliminary stage." It should be apparent from this review of Engineering for Power, Vol. 98, Apr. 1978, pp. 199-217.
23 Greitzer, E. M., Mazzawy, R. S., and Fulkerson, D. A., "Flow Field
that the author agrees completely with the first part of this Coupling Between Compression System Components in Asymmetric Flow,"
statement. As to the latter part, although much has been ASME Journal ofEngineering for Power, Vol. 100, Jan. 1978, pp. 66-72.
achieved, it should also be apparent that understanding, if not 24 Day, I. J., "Axial Compressor Stall," Ph.D. thesis, Cambridge
quite preliminary, is still far from complete. University Engineering Department, 1976.
25 Takata, H., and Nagano, S., "Nonlinear Analysis of Rotating Stall,"
ASME Paper 72-GT-3,1972.
26 Orner, N., "Rotating Stall in Axial Flow Compressors," Von Karman
Institute Lecture Series, "Unsteady Flow in Turbomachines," 1979, Brussels,
Acknowledgment Belgium.
27 Day, I. J., Greitzer, E. M., and Cumpsty, N. A., "Prediction of Com-
In the preparation of this review the author has been helped pressor Performance in Rotating Stall," ASME Journal of Engineering for
by a number of people. In particular, however, he would like Power, Vol. 100, Jan. 1978, pp. 1-14.
to acknowledge Dr. N. A. Cumpsty (who found time to pause 28 Montgomery, S. R., and Braun, J. J., "Investigation of Rotating Stall in
at M. I. T. during his journeys between Cambridge University a Single Stage Axial Compressor," NACA Technical Note TN3823, 1957.
and Caltech) and Mr. R. S. Mazzawy of Pratt & Whitney 29 Hill, P. G., and Peterson, C. R., Mechanics and Thermodynamics of
Propulsion, Addison-Wesley Publishers, 1965.
Aircraft, both of whom suggested clarifications and im- 30 Tesch, W. A., Steenken, W. G., "J85 Clean Inlet Flow and Parallel
provements to the several versions of the paper. The com- Compressor Models," NASA Contractor Report CR-134978, Mar. 1976.
ments of Dr. C. S. Tan (of M. I. T,), and Messrs. J. P. 31 Corbett, A. G., and Elder, R. L., "Stability of an Axial Flow Compressor
Nikkanen and B. A. Robideau (of P & W) are also ap- with Steady Inlet Conditions," Journal of Mechanical Engineering Science,
Vol. 16, No. 6,1975.
preciated. This work was carried out under the sponsorship of 32 Corbett, A. G., and Elder, R. L., "Mathematical Modeling of Com-
NASA Lewis Research Center Grant NSG 3208. pressor Stability in Steady and Unsteady Flow Conditions," in Unsteady
Phenomena in Turbomachinery, AGARD CP-177, 1976.
33 Jansen, W., and Swarden, M, C , "Compressor Sensitivity to Transient
and Distorted Transient Flows,'' AIAA paper 71 -670,1971.
References 34 Willoh, R. G., and Seldner, K., "Multistage Compressor Simulation
Applied to the Prediction of Axial Flow Instabilities," NASA TMX-1880,
1 Iura, T., and Rannie, W. D., "Experimental Investigations of 1960.

150/Vol. 102, JUNE 1980 Transactions of the ASME

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]
 35 Plourde, G. A., and Stenning, A. A., "Attenuation of Circumferential 59 Prince, D. C , Jr., Wisler, D. C , and Hilvers, D. E., "Study of Casing
Inlet Distortion in Multistage Axial Compressors," Journal of Aircraft, Vol. 5, Treatment Stall Margin Improvement Phenomena," NASA CR-134552, Mar.
No. 3, 1968. 1974.
36 Stenning, A. A., "Inlet Distortion Effects in Axial Compressors," 60 Osborn, W. M., Lewis, G. W., Jr., and Heidelberg, L. J., "Effect of
Chapter 14 in Fluid Dynamics of Turbomachinery, ASME Lecture Course, Several Porous Casing Treatments on Stall Limit and on Overall Performance
Iowa State University, 1973. of an Axial-Flow Compressor Rotor," NASA TN D-6537, Nov. 1971.
37 Mokelke, A., "The Prediction of Steady, Circumferential Pressure and 61 Moore, R. D., Kovich, G., and Blade, R. J., "Effect of Casing Treatment
Temperature Distortions in Multistage Axial Flow Compressors," ASME on Overall and Blade-Element Performance of a Compressor Rotor," NASA
Paper 79-GT-184. TND-6538,Nov. 1971.
38 Hercock, R. G., and Williams, D. D., "Aerodynamic Response," in 62 Fabri, J., and Reboux, J., "Effect of Outer Casing Treatment on Stall
Distortion Induced Engine Instability, AGARD Lecture Series, No. 72, 1974. Margin of a Supersonic Rotating Cascade," ASME Paper 75-GT-95, ASME
39 Collins, T. P., "Engine Stability," Chapter 21, in The Aerother- Gas Turbine Conference, Mar. 1975.
modynamics of Aircraft Gas Turbine Engines, edited by G. Oates, AFAPL-TR- 63 Tesch, W. A., "Evaluation of Range and Distortion Tolerance for High
78-52, 1978. Mach Number Transonic Fan Stages, Task IV Stage Data and Performance
40 Reid, C , "The Response of Axial Flow Compressors to Intake Flow Report for Casing Treatment Investigations, Volume I , " NASA CR-72862,
Distortion," ASME Paper 69-GT-29, 1969 ASME Gas Turbine Conference. May 1971.
41 Mokelke, H., "Prediction Methods," in Distortion Induced Engine 64 Bailey, E. E., "Effect of Grooved Casing Treatment on the Flow Range
Instability, AGARD Lecture Series, No. 72, 1974. Capability of a Single-Stage Axial Flow Compressor," NASA TM X-2459, Jan.
42 Braithwaite, W. M., and Soeder, R. H., "Combined Pressure and 1972.
Temperature Distortion Effects on Internal Flow of a Turbofan Engineer," 65 Horlock, J. H., and Lakhwani, C. M., "Propagating Stall in Com-
Paper presented at AIAA/SAE, 15th Joint Propulsion Specialists Conference, pressors with Porous Walls," ASME Paper 75-GT-59, 1975.
1979. 66 Amann, C. A., Nordensen, G. E., and Skellenger, G. D., "Casing
43 Mikolajczak, A. A., and Pfeffer, A. M., "Methods to Increase Engine Modification for Increasing the Surge Margin of a Centrifugal Compressor in
Stability and Tolerance to Distortion," in Distortion Induced Engine In- an Automotive Turbine Engine," ASME Journal of Engineering for Power,
stability, AGARD Lecture Series, No. 72,1974. Vol. 97, July 1975, pp. 329-336.
44 Williams, D. D., and Yost, J. O., "Some Aspects of Inlet-Engine Flow 67 Jansen, W., Presentation at Panel on Centrifugal Compressor Efficiency
Cornpatability," Aeronautical Journal, Sept. 1973, pp. 483-492. Improvement, ASME Gas Turbine Conference, 1979.
45 Mazzawy, R. S., "Multiple Segment Parallel Compressor Model," 68 Greitzer, E. M., Nikkanen, J. P., Haddad, D. E., Mazzawy, R. S., and
ASME Journal ofEngineering for Power, Vol. 99, Apr. 1977, pp. 288-296. Joslyn, H. D., " A Fundamental Criterion for the Application of Rotor Casing
46 Adamczyk, J. J., "Unsteady Fluid Dynamic Response of an Isolated Treatment," ASME JOURNAL OF FLUIDS ENGINEERING, Vol. 101, June 1979, pp.
Rotor with Distorted Inflow," AIAA Paper 74-49, 1974. 237-244.
47 Pandolfi, M., and Colasurdo, "Numerical Investigations on the 69 Stenning, A. H., and Kriebel, "Stall Propagation in a Cascade of Air-
Generation and Development of Rotating Stalls," ASME Paper 78-WA/GT-5. foils," Trans. ASME, Vol. 80, May 1958, pp. 777-790.
48 McCroskey, W. J., "Some Current Research in Unsteady Fluid 70 Dunham, J., "Observations of Stall Cells in a Single Stage Compressor,"
Dynamics," ASME JOURNAL OF FLUIDS ENGINEERING, Vol. 99, Mar. 1977, pp. Aeronautical Research Council, C. P. 589, 1961.
8-38. 71 Tanaka, S., and Murata, S., "On the Partial Flow Rate Performance of
49 Platzer, M. F., "Unsteady Flows in Turbomachines — A Review of Axial Flow Compressor and Rotating Stall, First and Second Reports,"
Current Developments," from Unsteady Aerodyamics, AGARD CP-227, 1977. Bulletin oftheJ.S.M.E., Vol. 18, No. 117, Mar. 1975.
50 Telionis, D. P., "Review - Unsteady Boundary Layers, Separated and 72 Sexton, M. R., O'Brien, W. F., and Moses, H. L., "An On-Rotor In-
Attached," ASME JOURNAL OP FLUIDS ENGINEERING, Vol. 101, Mar. 1979, pp. vestigation of Rotating Stall in an Axial Flow Compressor," in Unsteady
29-43. Phenomena in Turbomachinery, AGARD CP-177, 1976.
51 Nagano, S., Machida, Y., and Takata, H., "Dynamic Performance of 73 Cossar, B. F. J., Moffat, W. C , and Peacock, R. E., "Compressor
Stalled Blade Rows," Japan Soc. of Mech. Eng. Paper #JSME 11, presented at Rotating Stall in Uniform and Non-Uniform Flow," unpublished report,
Tokyo Joint Int. Gas Turbine Conf., Tokyo, Japan, Oct. 1971. Cranfield Inst, of Technology, 1979.
52 Greitzer, E. M., and Griswold, H. R., "Compressor Diffuser Interaction 74 Rannie, W. D., "The Axial Compressor Stage," Part F in Aerodynamics
with Circumferential Flow Distortion," Journal Mechanical Engineering of Turbines and Compressors, edited by W. R. Hawthorne, Princeton
Science, Vol. 18, Feb. 1976, pp. 25-43. University Press, 1964.
53 Greitzer, E. M., "Coupled Compressor-Diffuser Flow Instability," J. 75 Breugelmans, "Flow Measurements in a Rotating Stall cell," from Von
Aircraft, Vol. 14, Mar. 1977, pp. 233-238. Karman Institute Lecture Series, "Unsteady Flow in Turbomachines,"
54 Jansen, W., "Rotating Stall in Radial Vaneless Diffusers," ASME Brussels, Belgium, 1979.
Journal of Basic Engineering, Vol. 86, Dec. 1964, pp. 750-758. 76 Bartlett, F., and Greitzer, E. M., "Stalled Flow Performance of a Single
55 Abdelhamid, A. N., and Bertrand, J., "Distinctions between Two Types Stage Transonic Compressor," M.I.T. Gas Turbine and Plasma Dynamics
of Self-Excited Gas Oscillations in Radial Vaneless Diffusers," ASME Paper Laboratory Report 144, Dec. 1978.
79-GT-58, 1979.
77 McKenzie, A., Private Communication.
56 Toyama, K., Runstadler, P. W., and Dean, R. C , "An Experimental
Study of Surge in Centrifugal Compressors," ASME JOURNAL OF FLUIDS 78 Motycka, D. L., "Ground Vortex - Limit to Engine/Reverser
ENGINEERING, Vol. 99, Mar. 1977, pp. 115-129. Operation," ASME Journal ofEngineering for Power, Vol. 98, Apr. 1976, pp.
57 Greitzer, E. M., "Surge and Rotating Stall in Axial Flow Compressors. 258-264.
Part I: Theoretical Compression System Model," ASME Journal of 79 Harman, J. W., "Discussion" (on the paper "Prediction of Compressor
Engineering for Power, Vol. 98, Apr. 1976, pp. 190-198. Performance in Rotating Stall,") ASME Journal of Engineering for Power,
58 Takata, H., and Tsukuda, Y., "Study on the Mechanism of Stall Margin Vol. 100, Jan. 1978, pp. 13-14.
Improvement of Casing Treatment," ASME Paper 75-GT-13, ASME Gas 80 Hawthorne, W. R., "Some Aerodynamic Problems of Aircraft Engines,"
Turbine Conference, Mar. 1975. Journal of Aeronautical Sciences, Vol. 24, Oct. 1957, pp. 713-730.

Journal of Fluids Engineering JUNE 1980, Vol. 102/151

Downloaded From: [Link] on 06/30/2019 Terms of Use: [Link]