A Review of Aerodynamically Induced Forces Acting On Centrifugal Compressors, and Resulting Vibration Characteristics of Rotors
A Review of Aerodynamically Induced Forces Acting On Centrifugal Compressors, and Resulting Vibration Characteristics of Rotors
A Review of Aerodynamically Induced Forces Acting On Centrifugal Compressors, and Resulting Vibration Characteristics of Rotors
listed in the bibliography. It has simply been condensed and where, in addition to the definitions above:
described in a manner that is most relevant to the particular c = Damping coefficient
problem of rotors. x⭈ = Velocity
Ccr = Critical damping
Free Vibration p = As defined below
Free Vibration Without Damping µ = As defined below
⭈⭈ ⫹ kx = 0
mx (1)
x = X sin(ωnt ⫹ Φ) (2) m
/
ω n2 = k m (3)
where:
x
m = Mass
⭈⭈
x = Acceleration
k = Spring coefficient Figure 2. Simple Spring-Mass-Damper System.
x = Position/amplitude
X = Maximum position/amplitude If critical damping is defined by Equation (5), and one investigates
ωn = Natural frequency only the case where damping is less than critical, then a solution is
t = Time generated as defined by Equation (6). This solution has a harmonic
Φ = Phase angle component (cos(pt + Φ)) with a natural frequency given by (pt,
c
where p = ω n2⫺µ2 and µ = ), and a decaying component (e⫺µt),
2m
with the rate of decay determined by the coefficient in the exponent
(µ). With any given initial position not at the equilibrium position
k of the system, the value of this exponent determines the rate at
which the vibration of the system will return to zero. The
coefficient over one period of vibration (µT, where T is one period)
is called the logarithmic decrement (log dec), and it is defined by
Equation (7).
m
δ = µT (7)
x = P cos(ω t ⫺ Φ) (10)
D
P= (11) Time
[1 ⫺ τ2)2 ⫹ (2ρτ)2] 1/2
where, in addition to the earlier definitions:
s(t) = Harmonic forcing function
ω = Forcing function frequency
F = Maximum of forcing function
P = Maximum of response amplitude
mode shape. These describe the way the rotor vibrates when at the
f(t) particular natural frequency in question. The first four of these
mode shapes for a typical beam style compressor would be as
depicted in Figure 7. The result is a rotor system that has multiple
natural frequencies, or principal modes. When undergoing free
vibration, the overall vibration of the system is always some
superimposed combination of these modes.
t' dt'
Figure 5. Force Due to Series of Impulses.
Free Vibration
Free Vibration Without Damping
If one has a multi degree of freedom system, such as a
compressor rotor, the above basic principles still apply. The system
is more complicated, but has the same basic components. For
calculation purposes, the mass is represented by the modeled rotor
stations, and the spring is represented by stiffness matrices
between mass stations on the rotor, and by stiffness coefficients at
the bearing locations. A typical rotor system model for an
undamped model of this type might look like Figure 6 without the
damping (C).
Figure 8. Cross-Coupling Coefficients.
forced vibration can therefore be separated from unforced Must increase flow
“instability” when evaluating vibration data most readily by E beyond onset line to
H eliminate phenomenon
viewing real time signals, to separate relatively steady or F
constantly increasing vibration from unsteady vibration at multiple Y
F.
frequencies. In real time, amplitudes and frequencies will S
constantly vary, with natural frequencies being the most prominent T
in most instances. Vibration levels will be determined by the P E
R C
stiffness and damping of the system, and the amplitude of the O R
forces applied. E
S E E
One can see from these results that unforced subsynchronous F.
S. S
vibration (instability) will be a relatively clean signal at the rotor
natural frequency. Forced subsynchronous vibration due to a I
harmonic force will be a clean single frequency that is often not at S
a rotor natural frequency. Forced subsynchronous vibration due to
a periodic force will be at a frequency and its multiples, again Onset line when
usually not at a rotor natural frequency (although the existence of decreasing flow
multiples increases the chance of resonance). Finally, forced
subsynchronous vibration due to arbitrary (impulse) forces will be
at multiple, and continuously varying frequencies and amplitudes, Flow
with maximum amplitudes usually occurring near rotor natural Figure 11. Hysteresis Zone.
frequencies. With this basis, it is useful to describe the causes of
some of the harmonic, periodic and arbitrary forces that can create compressor’s minimum stable flow rate as the “surge line” (Figure
such vibration signals. 12). In fact, the minimum flow rate shown on most as-tested
AERODYNAMIC FORCES performance maps is dictated by stall rather than surge. Stall or
rotating stall tends to be a precursor to true surge. That is, as the
Definitions inlet flow is reduced, the compressor will experience stall before it
Before further discussions on the aerodynamic forces, a encounters a full flow reversal or surge. In some compression
definition of four terms must be presented so that there is no systems, the flow increment between the onset of stall and the
confusion as to their interpretation in this document. Those terms onset of true surge is very small and it may be impossible to detect
are surge, stall, rotating stall, and hysteresis zone and they are the stall without encountering full surge. In other systems, there is
defined as follows: a significant difference in the flow rates at which stall and surge
occur.
• Surge—is a system phenomenon that is not only dependent on
the compressor but on all components of the process; i.e., piping,
valves, pressure vessels, volumes, etc. Surge is defined as an
operating condition at which full flow reversal occurs; i.e., flow
progresses backward through the compressor section or stage and
LINE
comes out the inlet. (Note, it is possible for one stage within a
compression section to surge without the entire section surging.) It
GE "
• Hysteresis zone—When reducing flow, a phenomenon will No further commentary will be offered on the system
appear at some flow rate. The phenomenon does not disappear phenomenon, surge. Rather, the discussion will center on stall and,
simply by moving back above that flow rate. Instead, the flow rate in particular, impeller and vaneless diffuser rotating stall.
has to be increased significantly beyond the onset rate to “wash
out” the stall cells (Figure 11). Stationary Stall
It is important to understand the distinction between surge and Before addressing rotating stall, a few comments are warranted
stall (or rotating stall). Quite often, turbomachinery engineers and on the stationary form. Stationary stall cells may or may not have
users misinterpret or misuse these terms, leading to confusion as to detrimental effects on compressor aerodynamic and mechanical
which of the phenomena are happening in a compressor. Much of performance. Much like its rotating counterpart, stationary stall
this confusion results from the common practice of labeling a causes nonuniform pressure fields or unbalanced forces that can
A REVIEW OF AERODYNAMICALLY INDUCED FORCES ACTING ON CENTRIFUGAL 269
COMPRESSORS, AND RESULTING VIBRATION CHARACTERISTICS OF ROTORS
influence the compressor rotor. Stationary stall cells can form in vibrations or contribute to such by prompting the formation of
vaned diffusers, return channels, guidevanes, near volute tongues, rotating stall. For example, if stationary stall cells form in a return
etc. The most common factor that promotes the formation of channel or guidevane passage, the stall may disturb the flowfield
stationary stalls is high levels of incidence (i.e., the difference such that rotating stall would initiate in the downstream impeller.
between the gas approach angle and the vane setting angle, Figure Therefore, stationary stall cannot be totally ignored as a possible
13). High incidence occurs as a compressor or compressor stage is contributor to rotor excitation. However, no further comments will
operated far from its nominal or design flow condition; that is, near be offered on the subject.
surge or in overload. The separation cells create pressure
disturbances that influence other upstream or downstream stage Rotating Stall
components. For example, the impeller shown in Figure 14 would Rotating stall is best described as a nonuniform circumferential
be affected by the nonuniform pressure field caused by the flow pressure field that rotates at a speed other than the compressor
separation from the diffuser vanes. The interaction of the impeller operating speed. The nonuniform pressure field might exert
with this nonuniform field may be evidenced by an increase in unbalanced forces on the compressor rotor, sometimes resulting in
vibration at blade-vane passing. Obviously, diffuser vane angles asynchronous vibrations. Since the rotational speed of the pressure
must be set accurately to ensure that high incidence and the field is most often lower than the rotor rotational speed, the
resulting separation will not arise within the compressor’s required vibration frequencies are subsynchronous.
operating map. As noted, the two most common forms are impeller and diffuser
rotating stall. Both may have significant effects on mechanical and
aerodynamic performance. In most compressors, rotating stall does
not appear except at lower flow rates; i.e., very near surge.
However, in some cases, they have been encountered very near
design flow. Further complicating matters, some forms of
“interaction stall” actually are more prevalent at the high capacity
end of the performance map.
Confounding the situation even further, other components can
produce forces that affect the rotor vibration signatures,
approximating stall characteristics; i.e., seals and bearings.
However, these rotor effects are typically not as sensitive to
compressor operating conditions. Their response frequency is
normally evident over the entire operating envelope and remains
relatively constant despite changes in flow rate or discharge
pressure. It is important to rule out such effects before attributing a
subsynchronous rotor vibration to rotating stall.
Impeller Rotating Stall
Numerous researchers have investigated impeller rotating stall
and there are equally numerous amounts of theories and opinions
as to its nature and influences on compressor aeromechanical
performance. Much of this work was drawn from or based upon
studies done in the axial compressor world. However, the two early
Figure 13. Stationary Stall Cell in a Return Channel. works that gained some notoriety in the centrifugal fraternity were
those of Abdelhamid (1980) and Frigne and Van den
Braembussche (1984). Their efforts sought to classify the various
forms of rotating stall that could occur in a centrifugal stage.
Abdelhamid suggested that impeller rotating stall could result
from flow perturbations at the impeller exit that would not allow
the flow to follow the blading. These perturbations may be a
consequence of disturbances within the impeller passages (i.e.,
separation cells, Figure 14) or strong interactions between the
impeller and the diffuser (i.e., the diffuser walls interfering with the
impeller exit area through misalignment of parts, etc.).
Frigne and Van den Braembussche also studied the characteris-
tics of rotating stall (impeller and diffuser) and published a series
of criteria that have been widely used to distinguish between the
various types. Their study identified two distinct forms of impeller
stall; abrupt and progressive. They also felt that the abrupt stall was
the result of strong interactions between the impeller and diffuser
while the progressive variety was more the consequence of the
impeller flowfield itself. Their publication identified the frequency
Figure 14. Sketch of Flow Separation in Impeller Due to range for the two types as follows: Abrupt–26 percent to 31 percent
Interaction with Separated Diffuser. of running speed; Progressive–67 percent to 82 percent of running
speed.
Stationary stall formation can also occur when flow encounters Other researchers have identified the probable causes for
very tight curvatures. If curvature radii are too small in a return progressive impeller rotating stall as:
bend, guidevane, on a return channel vane, etc., the flow separates • Flow separations near the impeller exit;
from the highly curved surface and a stall cell can form.
Misalignment of parts in a compressor flowpath can also cause the • Incidence angles at the impeller leading edge; or
flow to separate or stall. All these phenomena might induce rotor • Pressure disturbances caused by the impeller blade geometry.
270 PROCEEDINGS OF THE 29TH TURBOMACHINERY SYMPOSIUM
Efficiency
some investigators confused diffuser stall as impeller stall. In short,
though much has been published on the subject of impeller rotating
stall and though many claim to have complete knowledge on the
issue, the inconsistencies and contradictory nature of their Pressure Coeff.
additional work is required before anyone can claim a total Work Input
Coefficent
Presssure
understanding of this phenomenon.
Impeller Rotating Stall Characteristics
Compressors experiencing impeller rotating stall can exhibit
some or all of the following characteristics:
• The subsynchronous radial vibration frequency falls in the range
of from 50 percent to 80 percent of the compressor running speed.
Note: as suggested above, this range could be as large as from 26
percent to 100 percent of running, though the smaller range is more
widely accepted. FLOW
• The subsynchronous vibration frequency tracks with running Figure 15. Data from Compressor Stage Experiencing Impeller
speed. Stall.
• There is normally no hysteresis zone associated with impeller applying an excitation force to the rotor (Figure 17). Redesign of the
stall. That is, there will be a very distinct flow rate at which the impellers to an arbitrary bladed, full inducer configuration
problem will come and go. eliminated the stall and all related performance problems. These
• It may be possible to throttle through the indications of stall such results were reported by Sorokes and Welch (1991) and Sorokes
as subsynchronous vibration. That is, it may only occur over a (1993). In light of the problems with these impellers, new guidelines
limited flow band and disappear at lower flow rates. For example, to limit the amount of blade and cover turning were implemented.
when reducing flow rate, a subsynchronous vibration due to such a
stall may arise at some Q/N. The vibration persists with further
reduction in flow until a point is reached where it disappears.
• There may be a discontinuity or “droop” in the performance
curve associated with the onset of the subsynchronous vibration.
Unfortunately, as noted, there is no consensus on the character-
istics of impeller rotating stall. As such, it is difficult to provide
definitive guidelines for its identification. However, there is good
agreement on the various contributors to impeller stall; that is,
geometric configurations or flow profiles to avoid. All also agree
that detailed 2-D or 3-D analyses of the impeller and its associated
hardware can identify potential sources for impeller rotating stall.
However, how such stalls manifest themselves (i.e., characteristic
frequencies) is still the subject of much debate.
Examples of Rotating Stall Attributed to Impellers
In the past, high flow coefficient stages have been more
susceptible to impeller rotating stall. Some obsolete high flow
impellers have shown a drooping trend in the pressure coefficient
curve in the region between design and surge flow, especially when
applied at high tip Mach number, U2/Ao (Figure 15). This droop
was accompanied by an increase in subsynchronous vibration at
approximately 66 percent of the compressor running speed. Note
that this frequency falls within the guidelines for impeller stall Figure 16. Impeller CFD Analysis Showing Separation Zone.
published by Frigne and Van den Braembussche (1984). Also, the
general shape of the performance curve for these impellers agreed
with the trends observed by the two researchers.
One-dimensional analyses on these obsolete impellers showed no
obvious problems; i.e., relative velocity ratios, incidence levels, and
various other parameters gave no indications that these impellers
would suffer stall problems. Two-dimensional studies also yielded
satisfactory results, i.e., loading parameters and velocity
distributions fell within the generally accepted guidelines. However,
when 3-D flowfield analyses (computational fluid dynamics (CFD))
were performed, some clear shortcomings were observed.
CFD analyses showed that large separation zones were forming in
the impeller passages, prompted by high levels of turning both along
the shroud and on the blading (Figure 16). These separation zones
were coalescing into large wake regions at the impeller exit and Figure 17. Impeller Exit Mach Number Distribution.
A REVIEW OF AERODYNAMICALLY INDUCED FORCES ACTING ON CENTRIFUGAL 271
COMPRESSORS, AND RESULTING VIBRATION CHARACTERISTICS OF ROTORS
edge incidence (though within prior experience with this impeller) The stall may cause significant unbalanced forces on the rotor,
and the aerodynamic loading on the blading were contributing to especially when the vanes are closely coupled to the impellers
the high level of secondary flow. The impeller was redesigned to (Sorokes and Welch, 1992) (Figure 22). Consideration must be
address these two issues and the problem was resolved. given to the incidence swings that may occur as the end user
The reason for providing so much detail on this case is that the operates the compressor. If it is clear that one or more of the
original stage design adhered to all 1-D criteria existing at that time required operating conditions result in high incidence; i.e., in
for stall avoidance and yet a problem occurred. Further, the stall’s excess of 10 degrees; there is a high probability that flow
behavior was very unique and did not conform to any published separation and the associated stall will occur.
material regarding the characteristics of rotating stall in centrifugal
compressors. In short, this experience provided information on a
Diffuser Vanes
form of impeller rotating stall not previously encountered. Clearly,
more rigorous acceptance criteria were needed to properly assess Rotation
the potential for centrifugal impeller rotating stall. Such criteria
could not be based solely on simple 1-D calculations, though new
1-D guidelines were developed in light of this experience. The
more obvious solution that was implemented was to make more
Semi-Vaneless Space
effective (and frequent) use of 2-D or 3-D analyses.
Two more crucial observations must be made. First, as noted, the
subsynchronous vibration was only evident when the compressor
was run at full-load and full-pressure. Had only the class III test Impeller
been done, the stall would not have been detected until the
compressor was installed at the customer’s site. Resolution in the
field would have been far more difficult and time-consuming.
Clearly, there are significant advantages to performing full-load,
full-pressure testing prior to shipment of the compressor from the
vendor’s facilities.
Second, much of the research work done on rotating stall
(impeller, diffuser, etc.) has been done using low pressure test
vehicles; i.e., single-stage test rigs. As was seen in the above
example, it may not be possible to quantify the influences of stall
or even detect their presence during such low pressure testing.
Figure 22. Spacing Between Impeller and Diffuser Vanes.
Therefore, it would be more advantageous to conduct rotating stall
research using high pressure test vehicles.
It is difficult to characterize the radial vibration frequencies that
Diffuser Rotating Stall arise due to stall in the vaned portion of a diffuser. Unfortunately
(or fortunately), there are limited data available in which stall was
Though this section is intended to address the entire subject of attributed directly to the vaned diffuser. Some have reported
diffuser rotating stall, the primary focus will be on vaneless frequencies that are proportional to the number of impeller blades
diffusers or the vaneless portion of vaned diffusers; i.e., the portion or diffuser vanes; i.e., blade or vane passing. Others (nearly all
between the impeller and diffuser vanes (the “semi-vaneless associated with low solidity vaned diffusers, LSDs) have cited
space”) or between the diffuser vanes and the downstream subsynchronous frequencies in the same range as would be
component (return bend or volute). Some comments on the vaned expected for stall in a vaneless diffuser. In the latter case, it has
portion are offered but the majority of the rotating stall problems been speculated that the stall actually formed in the vaneless space
encountered in vaned diffuser stages arise in the vaneless portion between the impeller and diffuser due to either:
of those diffusers. Recall, problems associated with stationary stall
in vaned diffusers were noted previously. • Highly tangential vaneless space flow angles; or
Vaned Diffusers
• High levels of incidence leading to vortex shedding and stall cell
formation.
As noted, stall can occur in the vaned portion of a diffuser. The
primary cause is leading edge incidence, though poor passage area Examples of Stall Associated with Vaned Diffusers
distribution or vane design may also contribute. Regardless, the Despite building and testing a very large number of compressors
result will be flow separation from the vane surfaces or side walls. that employed LSDs and other vaned types (i.e., wedge, airfoil,
If the flow separation causes vortex shedding, this shedding may rib), the OEM has encountered very few cases of subsynchronous
promote the formation of rotating diffuser stall in the vaneless vibration or stall that were directly attributable to the diffuser. In
space upstream of the diffuser vanes. Regardless of whether the majority of these instances, the situation was rectified by
stationary or rotating, the disturbed flowfield can hinder the restaggering the diffuser vanes; i.e., adjusting the vane inlet angles
performance of downstream components, yielding higher losses to better match the impeller exit flow conditions. The frequency
and a reduction in operating range. The incidence or separation spectra resulting from the inappropriate setting angles showed both
effects can also cause disturbances in the flowfield upstream of the subsynchronous (12 percent to 25 percent of running speed) and
diffuser vanes (backflow) or may, in fact, be the consequence of supersynchronous (blade/vane passing) components.
some upstream disturbances; i.e., impeller secondary flows, etc. In There have been two cases in which restaggering the diffuser
fact, it may be possible to cause rotating stall in the vaned portion vanes did not eliminate the subsynchronous vibrations. In these
of a diffuser if the diffuser follows an impeller that is experiencing situations, the compressors were tested with one or more vaned
rotating stall. Since the stall cells exiting the impeller will be diffuser arrangements and, finally, with vaneless diffusers. When
rotational in nature, angular momentum will force the “cells” to the vanes were removed, the subsynchronous vibration was still
rotate through the vaned diffuser. Of course, as the “cells” impact apparent but at a reduced amplitude as compared to the test with
the diffuser vanes, they will likely be disturbed or even eradicated. the vanes installed. Consider the frequency spectra given in Figures
In short, the diffuser vanes may actually serve to “wash out” 23 and 24. The spectrum in Figure 23 was taken during a near
impeller stall cells. surge excursion with the LSDs installed. The spectrum shown in
A REVIEW OF AERODYNAMICALLY INDUCED FORCES ACTING ON CENTRIFUGAL 273
COMPRESSORS, AND RESULTING VIBRATION CHARACTERISTICS OF ROTORS
Figure 24 was taken at the same flow condition after the LSDs Vaneless Diffusers
were removed. Clearly, the amplitude was reduced when the LSD Some of the earliest research into vaneless diffuser rotating stall
vanes were not present, yet a response was still evident. In short, was conducted by Dr. Willem Jansen (1964). Jansen was
the LSD vanes were serving as “reflectors” that amplified the investigating stall in vaneless diffusers caused by localized nonuni-
effect of stall cells thought to be forming in the semi-vaneless formities in the radial gas velocity. He determined that these
space immediately outside the impeller exit. With the vanes in nonuniformities lead to pressure disturbances or “stall cells” that
place, these stall cells or pressure disturbances were being rotate circumferentially around the compressor, subjecting the
reflected between the impeller and the diffuser vanes, causing rotor to unbalanced pressure forces. Jansen’s work showed that the
unbalanced radial forces on the rotor. Without the vanes to serve as onset of diffuser rotating stall was most strongly influenced by the
“reflectors,” the blade - vane interactions were eliminated and the diffuser flow angle. Once the diffuser angle exceeded some critical
rotor vibrations reduced. angle, rotating stall occurred. Depending on the flow conditions
and the details of the diffuser geometry, single or multiple stall
cells (i.e., two, three, four, etc.) can form. The rotor vibration char-
acteristics will be a function of this number of cells and their
rotational speed relative to the compressor operating speed.
The flow angle in a diffuser is a function of operating conditions
as well as the details of the diffuser geometry. Jansen’s studies
showed that the critical angle for the onset of rotating stall is also
a function of the diffuser geometry and the gas Reynolds Number.
To aid designers in their efforts to avoid rotating stall, Jansen
developed a series of guidelines or correlations that showed how
the critical angle would change for various diffuser geometries and
Reynolds numbers (Figure 25).
of the Senoo methods. Interestingly, their work showed that frequency spectrum consisting of numerous harmonics of some
rotating stall could occur at higher flow (i.e., more radial flow basic frequency. The spectrum appears as would a “clipped square
angles) if there is excess area above the impeller exit. This situation wave.”
seems to be more critical in lower flow coefficient stages. Sorokes
(1994) supported this contention in his work based on CFD Examples of Stall Associated with
analyses of various diffuser entrance geometries. Vaneless Diffusers (or Vaneless Passages)
Despite all the published literature and ongoing research • Vaneless Diffuser—Rather than cite specific compressors that
(notably the Concepts ETI rotating stall consortium), like impeller experienced diffuser rotating stall, this section provides an
rotating stall, there remains no definitive set of criteria for rotating overview of the two most common reasons that the phenomenon
stall avoidance in vaneless diffusers. However, experience has arises in industrial turbomachinery. This section also describes
shown that conservative application of the Senoo criteria yields a how a stall can form in other vaneless portions of the flow passage
very high rate of success in avoiding diffuser stall. For example, and addresses two examples of such.
insuring that the diffuser flow angles do not come within three to Vaneless diffuser rotating stall can occur if designers do not
five degrees of the critical angle based on the Senoo criteria account for assembly tolerances within a compressor. Another
typically will ensure rotating stall avoidance in most cases (Figure common reason is improper calculation of the upstream impeller
26). Of course, it is crucial that the flow angles be assessed over the exit flow angle causing an inaccuracy in specification of the
entire required operating range for the compressor; i.e., from diffuser width necessary for stall avoidance. That is, the calculated
design flow back to the surge control line. impeller exit flow angle is more radial than in actuality, resulting
in insufficient diffuser pinch and a diffuser flow angle that is too
tangential.
Returning to the subject of assembly tolerances, it is common
practice to allow machining or assembly tolerances on the parts
(return channels, guidevanes, inlets, discharge volutes) that stack
together to form the compressor flowpath. When the unit is
operated, any gaps between components (created by the machining
or assembly tolerances) will be closed as the pressure builds on the
walls of the various flow passages. All walls will deflect away from
the point of highest pressure within the machine. In a compressor,
the highest static pressure occurs in the last stage diffuser.
Therefore, the walls of that diffuser will move apart as any
machining or assembly gaps present in the remaining components
are closed. Failure to account for the growth of this diffuser width
can have significant consequences.
Most past occurrences of diffuser rotating stall encountered by
the OEM were caused by the deflections described above. At
operating pressures, walls deflected, the last stage diffuser width
increased, the diffuser flow angle increased (became more
tangential), and rotating stall developed. The machines exhibited
low frequency subsynchronous radial vibrations in the range of 6
percent to 33 percent of the running speed, though the majority fell
in a 6 percent to 18 percent range. The excess vibration typically
arose as operators throttled the compressor toward the surge
Figure 26. Senoo Criteria for Stall Avoidance Showing control line and, once present, would remain until the compressor
Conservative Line. was moved to much higher flow rates. Again, since the vibrations
come and go at different flow rates, the phenomenon is said to have
Vaneless Diffuser Rotating Stall Characteristics a hysteresis zone.
In all but limited cases, the stall was eliminated by reducing the
Compressors experiencing subsynchronous vibration due to diffuser widths such that, at pressure, the deflections would not
diffuser rotating stall exhibit some or all of the following charac- result in flow angles that reached the critical level necessary to
teristics: instigate rotating stall (refer to Figure 27 for typically corrective
• The subsynchronous radial vibration frequency is in a range of measure). It is clear that manufacturing tolerances and material
deflection under pressure must be accounted for to ensure stall
6 percent to 33 percent of running speed.
avoidance.
• The vibration frequency is sensitive to flow rate and, after onset, In a small number of cases, rotating stall was eliminated by
will typically increase as the flow rate is reduced. installing low solidity vaned diffusers. Such diffusers delay the
• The vibration frequency tracks with speed as long as Q/N is held onset of rotating stall by influencing the growth of secondary flows
fairly constant. and boundary layers that promote stall inception. LSDs are a more
attractive option when dealing with low flow coefficient stages. In
• There is a hysteresis zone associated with the onset flow rate. such cases, the amount of vaneless diffuser pinch necessary to
That is, when reducing flow, the phenomenon will appear at some
flow rate. However, the vibration does not disappear simply by ensure proper flow angles often would produce unacceptable
moving back above that flow rate. Instead, operators have to efficiency loss. The LSD eliminates this concern by allowing stable
increase flow significantly beyond that rate to “wash out” the stall flow to be maintained with wider passages. Of course, in designing
cells. the vaned diffuser, one must avoid creating different forms of stall
caused by vane incidence effects. (Refer to Vaned Diffusers above.)
• There may be sudden jumps in frequency as the stall progresses • Return Bend—Another form of vaneless space rotating stall can
from one to two to three cells, etc.
occur in the return bend (the 180 degree bend between the diffuser
• If dynamic pressure probes are available, their output shows a exit and return channel entrance, Figure 28) as reported by
response in the 6 percent to 33 percent range. They also show a Sorokes, et al, (1994). Though not typically considered part of the
A REVIEW OF AERODYNAMICALLY INDUCED FORCES ACTING ON CENTRIFUGAL 275
COMPRESSORS, AND RESULTING VIBRATION CHARACTERISTICS OF ROTORS
Note that there are multiple harmonics (4.5 Hz, 9.0 Hz, 13.5 Hz,
etc.) in the dynamic pressure spectra in Figure 34(d). In this case,
the probe was not sensing many cells moving at different speeds.
Rather, the multiple peaks represent the FFTs rendition of the
typical “clipped square wave” appearance of a diffuser stall
waveform, which leads to multiples (harmonics) of the base
frequency (4.5 Hz).
Figure 32. Radial Vibration Spectrum During Impeller Stall. Also note that although the pressure signal in Figure 34(d)
indicates multiple frequencies present, on the vibration response in
Periodic Force Figure 34(c) only the primary and first multiple are evident, at the
same frequencies as in the pressure signal. This vibration response
Impeller and diffuser stall are both good examples of periodic
is larger at the lower frequencies due to high damping in the
forces. As noted previously, the pressure disturbances caused system, with relatively low stiffness. Such a result is evident from
during stall rotate at speeds lower than the rotor, resulting in solutions of simple systems, as shown in Equation (11) above. As
subsynchronous radial vibrations. Typical frequency spectra for indicated, when damping is a significant factor in the response
impeller and diffuser stall are given in Figures 33 and 34, equation, amplitude response for the same force will increase as
respectively. In each figure, the top trace (a) is the vibration probe the frequency is reduced. The high damping values that minimize
signal immediately prior to onset of the stall. The second trace (b), dangerous vibration response at higher frequencies allow greater
also taken just prior to the onset of stall, is from a dynamic pressure response at the low frequencies of this example. In fact, the highest
transducer in the diffuser immediately outside the impeller for the response is at the lowest frequency, while the forcing function is
impeller stall case or at the exit of the diffuser in the diffuser stall greater at a higher frequency.
case. The third trace (c) is from the vibration probe after onset of In the aero section introduction, emphasis was placed on the
stall and the fourth trace (d) is the dynamic pressure transducer difference between surge and stall. To illustrate the difference,
output after onset. consider Figure 35. The trace in Figure 35(a) is from a dynamic
Clearly, in the case of impeller stall (Figure 33), forced vibration pressure transducer in a diffuser passage when said diffuser is in a
is only occurring due to the stall at 133 Hz. Also note that the 1⫻ rotating stall mode. The trace in Figure 35(b) was taken while the
frequency (149 Hz) does not appear in the dynamic pressure probe same machine was in true surge. Note the extremely low
output. frequencies associated with the surge (approximately 1 Hz).
278 PROCEEDINGS OF THE 29TH TURBOMACHINERY SYMPOSIUM
(b) surge
Figure 35. Dynamic Pressure Spectra—Stall Versus Surge.
Figure 40 indicates the vibration spectrum at the time of the (b) 30 seconds after peak amplitude
highest amplitude at the first natural frequency, as well as the
spectrum two seconds before and two seconds after. Figure 41
indicates the spectrum at 10, 30, and 60 seconds after the peak
signal. These data show the transient nature, varying amplitude,
and varying frequencies typical of response to arbitrary forces. If
viewed in real time, such a spectrum would appear to be “dancing”
or “rumbling,” with amplitude peaks highest in the area of the rotor
first natural frequency (approximately 40 percent of running
speed). No “fixed spikes” exist at any frequency, and the various
peaks form and disappear in time, shifting in both amplitude and
frequency with no apparent correlation to any operating parameters (c) 60 seconds after peak amplitude
(i.e., flow rate, speeds, gas conditions, etc.). This is clear evidence
of a forced vibration, with the rotor acting in response to external Figure 41. Vibration Spectra—Arbitrary Aerodynamic Forces (10,
forces within the gas path. 30, and 60 Seconds After Peak Amplitude).
280 PROCEEDINGS OF THE 29TH TURBOMACHINERY SYMPOSIUM
CONCLUSIONS Senoo, Y., Kinoshita, Y., and Ishida, M., 1977, “Asymmetric Flow
in Vaneless Diffusers of Centrifugal Blowers,” Transactions
The forces acting within centrifugal compressors can cause a
ASME Journal of Fluids Engineering, 99, (1), pp. 104-114.
variety of responses in the rotor system. A general discussion of
compressor rotordynamic behavior was presented including Sorokes, J. M., 1993, “The Practical Application of CFD in the
discussions on free and forced vibrations. Distinctions were made Design of Industrial Centrifugal Impellers,” Proceedings of the
between significant subsynchronous vibration characteristics, such Twenty-Second Turbomachinery Symposium, Turbomachinery
as rotor instability, and aerodynamically forced vibration of Laboratory, Texas A&M University, College Station, Texas,
various types. Ways to distinguish between these various causes pp. 113-124.
using vibration data were also provided. Sorokes, J. M., 1994, “A CFD Assessment of Entrance Area
The various aerodynamic phenomena that contribute to Distributions in a Centrifugal Compressor Vaneless Diffuser,”
increased radial vibrations were discussed including impeller stall, ASME Paper No. 94-GT-90.
diffuser stall, and interaction “stall.” A general description of the
flow physics was provided as were the commonly held Sorokes, J. M. and Welch, J. P., 1991, “Centrifugal Compressor
identification criteria for each. Finally, various frequency spectra Performance Enhancement Through the Use of a Single Stage
and waveform plots were offered to illustrate how the various Development Rig,” Proceedings of the Twentieth
phenomena may appear in vibration or pressure pulsation data; i.e., Turbomachinery Symposium, Turbomachinery Laboratory,
on a spectrum analyzer or oscilloscope. Texas A&M University, College Station, Texas, pp. 101-112.
No single paper can adequately address the entire subject of Sorokes, J. M. and Welch, J. P., 1992, “Experimental Results on a
aerodynamic excitation of rotor systems. Still, it is hoped this Rotatable Low Solidity Vaned Diffuser,” ASME Paper No. 92-
document provides the reader with a good general overview of a GT-19.
very complex subject.
Sorokes, J. M., Kuzdzal, M. J., Sandberg, M. R., and Colby, G. M.,
REFERENCES 1994, “Recent Experiences in Full Load Full Pressure Shop
Testing of a High Pressure Gas Injection Centrifugal
Abdelhamid, A. N., 1980, “Analysis of Rotating Stall in Vaneless Compressor,” Proceedings of the Twenty-Third
Diffusers of Centrifugal Compressors,” ASME Paper No. 80- Turbomachinery Symposium, Turbomachinery Laboratory,
GT-184. Texas A&M University, College Station, Texas, pp. 3-18.
Abdelhamid, A. N., 1985, “Dynamic Response of a Centrifugal
Blower to Periodic Flow Fluctuations,” ASME Paper No. 85- BIBLIOGRAPHY
GT-195. Bonciani, L., Ferrara, P.L., and Timori, A., 1980, “Aero-Induced
Abdelhamid, A. N. and Betrand, J., 1980, “Distinction Between Vibrations in Centrifugal Compressors,” Proceedings of
Two Types of Self-Excited Gas Oscillations in Vaneless Radial Rotordynamic Instability Problems in High Performance
Diffusers,” Transactions ASME Journal of Turbomachinery, Turbomachinery, Texas A&M University, NASA CP 2133, pp.
109, (1), pp. 36-40. 85-94.
Abdelhamid, A. N., Colwill, W. H., and Barrows, J. F., 1979, Fulton, J. W., 1986, “Subsynchronous Vibration of Multistage
“Experimental Investigation of Unsteady Phenomena in Centrifugal Compressors Forced by Rotating Stall,”
Vaneless Radial Diffusers,” ASME Paper No. 78-GT-23, Proceedings of Rotordynamic Instability Problems in High
Transactions ASME Journal of Engineering for Power, 101, Performance Turbomachinery, Texas A&M University.
(1), pp. 52-60. Japikse, D., 1996, Centrifugal Compressor Design and
Frigne, P., Van den Braembussche, R., 1984, “Distinctions Performance, Concepts ETI, Inc., pp. 5/1-5/89.
Between Different Types of Impeller and Diffuser Rotating Kammer, H. and Rautenberg, M., 1985, “A Distinction Between
Stall in a Centrifugal Compressor with Vaneless Diffuser,” Different Types of Stall in Centrifugal Compressor Stage,”
ASME Paper No. 83-GT-61; Transactions ASME Journal of ASME Paper No. 85-GT-194.
Engineering Gas Turbine and Power, 106, (2), pp. 468-474.
Seidel, U., Chen, J., Jin, D., and Rautenberg, M., 1991,
Jansen, W., 1964, “Rotating Stall in a Radial Vaneless Diffuser,” “Experimental Investigation of Rotating Stall Behaviour
ASME Paper No. 64-FE-6, Transactions ASME Journal of Influenced by Varying Design and Operation Parameters of
Basic Engineering, pp. 750-758. Centrifugal Compressors,” Paper No. 91-Yokohama-IGTC-93.
Kobayashi, H., Nishida, H., Takagi, T., and Fukushima, Y., 1990, Sorokes, J. M., 1995, “Industrial Centrifugal Compressors—
“A Study on the Rotating Stall of Centrifugal Compressors,” Design Considerations,” ASME Paper No. 95-WA/PID-2.
(Second Report, Effect of Vaneless Diffuser Inlet Shape on
Rotating Stall) Transactions of JSME (B Edition), 56, (529), Thomson, W., 1965, Vibration Theory and Applications, Prentice-
pp. 98-103. Hall, Inc.
Nishida, H., Kobayashi, H., Takagi, T., and Fukushima, Y., 1988, Timoshenko, S., Young, D., and Weaver, Jr., W., 1974, Vibration
“A Study on the Rotating Stall of Centrifugal Compressors,” Problems in Engineering, John Wiley & Sons.
(First Report, Effect of Vaneless Diffuser Width on Rotating Volterra, E. and Zachmanoglou, E., 1965, Dynamics of Vibrations,
Stall), Transactions of JSME, 54, (499), pp. 589-594. Charles E. Merrill Books, Inc.
Senoo, Y. and Kinoshita, Y., 1977, “Influence of Inlet Flow
ACKNOWLEDGEMENTS
Conditions and Geometries of Centrifugal Vaneless Diffusers
on Critical Flow Angles for Reverse Flow,” Transactions The authors acknowledge the following individuals for their
ASME Journal of Fluids Engineering, pp. 98-103. help in generating the figures and other presentation material used
in this paper: Chuck Dunn, Jim Shufelt, Barry Clawson, Ed
Senoo, Y. and Kinoshita, Y., 1978, “Limits of Rotating Stall and
Thierman, and Dresser-Rand test stand personnel. We also thank
Stall in Vaneless Diffusers of Centrifugal Compressors,”
Dresser-Rand for allowing us to publish this document.
ASME Paper No. 78-GT-19.