363
International Journal of Fluid Machinery and Systems DOI: http://dx.doi.org/10.5293/IJFMS.2017.10.4.363
Vol. 10, No. 4, October-December 2017 ISSN (Online): 1882-9554
Original Paper
A Study on the Deep-Surge Frequencies in Various Conditions of
Axial Flow Compressors and Flow-paths
Nobuyuki Yamaguchi1
1Department of Mechanical Engineering, Meisei University (retired)
2-1-25, Akanedai, Aoba-ku, Yokohama City, 227-0066, Japan, [email protected]
Abstract
Frequencies of deep surges and their behaviors in axial flow compressors were surveyed numerically. Relative surge
frequencies, normalized by the basic acoustical resonance frequencies, are seen to tend to lower together with increases
in the stalling pressure ratios, i.e. increases in the number of stages and the compressor tip speed, and also together with
increases in the flow-path sectional area ratios. However, it appears difficult to express simply the general behaviors of
the relative frequencies affected by the various factors. In order to know the essential behaviors, a modified reduced
surge frequency is proposed, which is a dimensionless number comparing the mass flow filling and emptying the
plenum volume in surge and the mass flow provided by the compressor. The modified reduced surge frequencies are
found to have or approach a definite and nearly constant value in conditions of deep surges. The parameter suggests the
fundamental mechanism of deep surges and could be used to determine approximate frequencies of deep-surges in
various conditions of compressors and flow-paths.
Keywords: Fluid Machine, Axial Flow Compressor, Surge, Analytical Simulation, Frequency, Fluid Dynamics
1. Introduction
Information on the oscillation frequencies in surge phenomena in compressors and fans are very important, although the horribly
large amplitudes of the related flows and pressures are likely to attract attentions. The frequencies are ones of essential keys to get an
insight into the vibrational nature of the phenomena and also ones of aspects to take preventive measures into consideration.
Nevertheless, the information on the surge frequencies does not appear to have reached the stage of generalized one.
Greitzer [1, 2] has proposed a basic parameter named “B”, which is a kind of reduced frequency paying attention to the flow-path
resonance frequency approximated by a Helmholtz resonator. Although it has been shown by him that various phase of surge
phenomena ranging from deep surges to stall stagnations could occur in dependence on the magnitude of the B parameter, the detailed
situation of the surge frequencies have not been studied. In the subsequent many studies with respect to surge phenomena, details of the
behaviours of the surge frequencies have curiously not been paid attention to. Maybe, it could have been more important at the stage to
manage to reproduce analytically the surge phenomena as truly as possible (for example, Boyer and O’Brien [3]), with the frequencies
regarded as some specific and individual data belonging to the particular cases. In the circumstances, the general and comprehensive
nature of the surge frequencies has not been made clear.
In some recent analytical experiences of the author in deep surges in multi-stage axial compressors (Yamaguchi [4, 5]), the deep-
surge frequencies tended to change much for changes in the compressor speeds, where the consistent nature of the behaviour was
difficult to grasp. Large deviations from the near-resonant conditions appear in high-pressure-ratio compressors and multi-stage
machines, although the frequencies appear to be very near the flow-path resonance frequencies for low-pressure fans and compressors.
Yamaguchi [4] suggested a dimensionless parameter of volume-modified reduced surge frequency in relation with stall stagnation
boundaries, but no further detailed investigations into the frequencies in deep surges have been conducted subsequently.
Characteristic scenes with respect to the surge frequencies could be summed into the following three groups according to the
author’s present stage of analytical experiences.
The first is the ordinary deep-surge situation where the deep surge means the one whose every surge loop encircles or passes
by the initial steady-state stalling point in the flow-vs-pressure domain. They are common ones both in site and in analyses, but
the information on the frequencies are rather by bits and pieces, having been not yet put to order. The frequencies are often seen to
be much lower than the flow-path resonance frequencies.
The second is the situation of very low surge frequency in high-speed multi-stage compressors. In analytical results on multi-
stage compressors, very low deep-surge frequencies often occur where the surge period is elongated by intervention of small
pressure peaks between neighboring ordinary major pressure peaks (Yamaguchi [6]). The small peaks were found to occur as
Received February 14 2017;accepted for publication June 12 2017: Review conducted by Hideaki Tamaki. (Paper number O17006J)
Corresponding author: Nobuyuki Yamaguchi, [email protected]
364
incomplete recoveries resulting from temporary unstalling of rear-stages and its failure in the environment of whole stages in stall.
The situation is somewhat different from the ordinary deep-surge one described above.
The last is the situation of surge frequencies abruptly changeable in the immediate neighborhood of the stall stagnation
boundaries (Yamaguchi [8]). In the deep-surge zone very near the stagnation boundary, subharmonic deep surges often appear,
which could be considered to be a precursor of stall stagnation events. Stall stagnations here include not only decaying oscillations
converged onto some stalled point, but also mild surge loops outside of the steady-state stalling point. In the sense, in addition to
the zero frequency of the decayed condition, some frequencies other than zero could be present, particularly on the way toward the
converged conditions.
The above three groups are the main characteristic situations of surge frequencies distinguished in the author’s mind.
Particularly, the first two groups on the deep surges remain unclear, which the author considers necessary to make clearer. If a
reasonable formula for estimation of the deep-surge frequencies become available, it would contribute much in studying further
the essential dynamics of the deep surge phenomena, and also in planning the countermeasures to the flow oscillations in the
system. So, the present study surveyed deep-surge frequencies numerical-experimentally and devised a rule of thumb.
The last item with respect to the stagnation stall concerns has been already examined (Yamaguchi [8]) and the information in
the neighborhood of the stall stagnation boundaries will be only briefly described here.
The present study is based only on numerical results using a code for surge transient analysis, “SRGTRAN” developed by the
author (Yamaguchi [9]). The conclusions might be rather qualitative in the sense. The author wishes that the conclusions would be
substantiated experimentally in the future.
2. Outline of the Analyses
The contents described here are based on the results by use of a code SRGTRAN for the surge transient analysis and simulation,
developed by the author (Yamaguchi [9]).
Figure 1 is a schematic presentation of the flow-path system, consisted of a suction flow-path, a compressor, a delivery plenum,
and an exit valve path. The system has an open inlet and an exit terminated by a flow-adjusting valve.
The geometrical variables are as follows: Lc** and Lp: lengths of
the suction flow-path from the inlet to the compressor exit, and the
delivery plenum, respectively. Ac2 and Ap: sectional areas of the exit
of the last stage of the compressor for analysis, and delivery plenum,
respectively. At the same time, Ac2 and Lc** are the representative
sectional area and the length for normalization, respectively. The
plenum inlet is connected to the exit of the compressor immediately
downstream, thus the distance between both locations is assumed to
be very small.
The following dimensionless parameters are defined for features
of the flow-paths.
Sectional area ratio of the flow-path: AR=Ap/Ac2 (1)
Length ratio of the flow-path: LR=Lp/Lc** (2)
The suction flow-path and the delivery flow-path in the present
analyses are configured by fourteen control volumes (CVs), and
thirty-four CVs, respectively. The stages of the compressor are
represented by corresponding respective CVs. The suction flow-path
is given the same geometrical dimensions throughout the analyses.
Table 1 gives the main numerical figures and dimensions of the
compressors and the flow-paths for the analyses. The compressors are
designed for 11300rpm, having number of stages 1, 3, 5, and 9.
The stage characteristics of each stage is assumed to be the same,
and is shown in Fig. 2. Here,
Flow coefficient:
tst
stt
uA
Q (3)
Pressure coefficient: 2
1
1
21
)1(
t
stTp
Ptu
Tc
(4)
Temperature coefficient: 2
12
21
)(
t
TTp
Ttu
TTc (5)
Here, ut : representative tip speed of the compressor, Ast : average annulus area of the stage, πst : Stage total pressure ratio,
Qst : average volumetric flow in the stage, TT1 and TT2: total temperature at the stage inlet and the stage exit, respectively, CP:
specific heat at constant pressure, κ: ratio of heats. Subscript st stands for the concerned stage.
Fig. 2 Stage characteristics for each stage
Fig. 1 Schematic view of the flow-path system
365
The stage characteristics requires to cover the wide range of flow from the turbine action zone to the reversed flow zone. Figure
2 is made up on the basis of the author’s experience and literature survey results. It is the same as the one that the author has
employed in the previous studies (Yamaguchi [4-9]).
The following situations are analyzed by use of code SRGTRAN, whose results are surveyed paying attention to deep surges
and the surge frequencies fs0.
Compressor rpms: 12000, 10000, 7000, and 5000.
Flow-path geometries
Length ratios LR: 10, 1, and 0.3.
Area ratios AR: 1-10.
Table 1 includes also steady-state stalling pressure ratios PR, evaluated by use of a stage-stacking procedure for the compressor
performances in the initial stage of the SRGTRAN analyses. They are required for computations of concerned parameters, such as
given by Eqs. (7), (9) and (11), and are convenient also to see the overall tendency as in Fig. 4.
Table 1 Main dimensions and numerical figures for analyses
Name Comp19 Comp15 Comp13 Comp11
Stages 9 5 3 1
Inlet Annulus Area of Stage 1:
Ac1 (m2)
0.10314 0.10314 0.10314 0.10314
Exit Annulus Area of
the Last Stage: Ac2 (m2)
0.0975 0.07688 0.06057 0.04125
Reference Radius: r t1 (m) 0.254 0.254 0.254 0.254
Design rpm (rpm0) 11300 11300 11300 11300
Design ut1 (m/s) 300.6 300.6 300.6 300.6
Suction Temperature (K) 288.2 288.2 288.2 288.2
Suction Pressure (Pa) 101300 101300 101300 101300
Stalling PR
12000rpm 5.38 2.87 1.97 1.28
10000rpm 3.52 2.16 1.62 1.19
7000rpm 1.48 1.47 1.28 1.09
5000rpm 1.2 1.14 1.04
Suction Length: Lc** (m) 4.631 4.323 4.17 4.093
3. Global Tendency of Deep Surge Frequencies
In the study, length ratios LR and sectional area ratios AR for the delivery flow-paths were varied with the suction flow-path
geometry kept the same. In the specified flow-path configurations, the deep-surge frequencies were surveyed by use of
SRGTRAN for variation of number in stages and rotational speeds of the compressor.
First of all, a relative surge frequency is evaluated from normalization of a resulting deep-surge frequency fs0 relative to a
basic resonance frequency f1 in the flow-path system.
Relative deep-surge frequency fs0/f1 (6)
Here, the basic resonance frequency f1 is calculated for the acoustic resonance condition of a still air column in the flow-path
identical with the surge analysis with the boundary conditions of the open inlet and the closed exit. The pressure is supposed
uniform to be the atmospheric one throughout the path. The temperatures upstream and downstream of the compressor are
assumed to be equal respectively to the suction temperature and to the exit temperature for the compressor stalling condition. The
resonance frequency thus evaluated could naturally be a rather approximate one, since, in the actual surge situations, the pressure
and the temperature could significantly change locally and temporarily. But as a measure of the related frequencies it would be
difficult to replace by some other simple parameters.
Figure 3 shows some examples of the tendencies of the relative deep-surge frequencies against the relative compressor rpms
normalized by the design rpm (rpm0 of 11300 rpm, here). The data conditions are given in the explanatory notes in terms of the
stage numbers, flow-path area ratios AR and length ratios LR. For compressors of smaller numbers of stages and for the lower
speeds, the relative surge frequencies are seen to be near unity; namely, the deep-surge frequencies are near the system resonance
frequencies.
On the other hand, for compressors having more stages, the relative deep-surge frequencies are seen to tend to lower, and the
tendency becomes stronger for higher speeds.
Figure 4 shows the same data plotted against the compressor stalling pressure ratios, PR. For the values of PR near unity, the
relative deep-surge frequencies are near unity, but lower for increasing values of PR. For the value of PR greater than 3.5, the
relative surge frequencies are seen to fall more abruptly, approaching toward 0.1. The behavior of the relative deep-surge
frequencies could be considered more consistently in terms of the stalling pressure ratios in Fig.4 than in terms of the compressor
speed ratios in Fig. 3.
In Figs. 3 and 4, it is generally understood that, the relative deep-surge frequencies fs0/f1 tend to fall from unity toward lower
levels for higher stalling pressure ratios PR. The behavior appears to change in rather complicated manners affected both by the
366
stalling pressure ratios influenced by the compressor stage numbers and the speeds, and by the flow-path geometries, typified by
the sectional area ratios AR and the length ratios LR.
Fig. 5 Contours of relative deep-surge frequencies in relations of the sectional area ratios of flow-paths and the stalling pressure
ratios. (a) Nine-stage compressor, Comp19, (b) Five-stage compressor, Comp15, and (c) Single-stage compressor, Comp11.
(1) flow-path length ratio LR of 0.3, (2) LR of 1.0, and (3) LR of 10.0.
Figure 5 shows contour maps of the relative deep-surge frequencies fs0/f1 with the abscissa for the flow-path sectional area ratio
AR and the ordinate for the stalling pressure ratios PR. Figure 5(a), (b), and (c) are for compressors having stages 9, 5, and 1,
respectively. Figure 5(1), (2), and (3) are for flow-path sectional area ratios LR of 0.3, 1, and 10, respectively. They are contour
Fig. 3 Tendency of relative deep-surge frequencies
against variations in the compressor speeds
relative to the design one.
Fig. 4 Tendency of relative deep-surge frequencies
against the compressor stalling pressure ratios
367
maps corresponding to Fig. 4, in which only a few selected typical geometrical conditions are shown. In the contour maps in Fig.
5, the area ratios are given basically by about ten data points horizontally, except some failed-calculation conditions.
As a whole, the relative deep-surge frequencies tend to lower together with increasing flow-path area ratios AR and increasing
stalling pressure ratios PR where both ratios appear to affect the tendency in some manner of combination. Effects of the length
ratios are not clear in these figures.
The contour lines of the relative surge frequencies often show some local irregular changes in the behavior, particularly near
the borders. The causes are not understood at present. The situations around the boundaries are put aside for future study. The
present study pays attention mainly to the deep-surge regions.
4. Behaviors of Volume-modified Reduced Surge Frequencies
It is observed in the previous section that the global behaviors of the relative deep-surge frequencies fs0/f1 are affected in a
combined manner by the flow-path sectional area ratio AR and the stalling pressure ratio PR. The nature of the behaviors have not
yet been understood.
Here, let us pay attention to the behaviors of the volume-modified reduced surge frequency fRPVVs (Yamaguchi [4]). It is a
dimensionless parameter resulted from a ratio of the mass flow in the filling and emptying action during the surge in the delivery
volume to the mass flow supplied by the compressor. It is expressed as follows;
)1
1(1
0
2PRA
A
u
Lff
C
P
t
PsRPVVS (7)
As can be understood, it is constructed of a reduced frequency with respect to the surge frequency fs0, the delivery length Lp, and
the rotor tip speed ut, modified by the sectional area ratios AR, and a pressure-ratio-effect term.
Although it could have a physical validity as it is and shows relatively well-correlated behaviors of data (Yamaguchi [6]), some
further trials have yielded the following improved formula having a better correlation.
LR
LRff RPVVsRPVVsm
1 (8)
More concretely,
)1
1()(
1
**
0
2PRA
A
u
LLLff
C
P
t
PCPsRPVVsm
(9)
The right-hand side of the above formula suggests indirectly that the representative length should reflect not only the effect of the
delivery plenum length LP, but also that of the whole length of the flow-path (LC**+LP). It could be reasonable in consideration of
the nature of the surge phenomena as the flow dynamics in the whole flow-path system.
Thus the effective representative length of the concerned flow-path will be defined as follows;
)(**
PCPeff LLLL (10)
Hereafter, the modified reduced surge frequency fRPVVsm is paid attention to.
As another parameter, the following area-pressure ratio parameter, or modified area ratio, is employed.
)1
1(1
*
2PRA
AAPR
C
P (11)
It reflects the global effects of the levels of the stalling pressure ratios on the area ratios.
Figures 6-9 show behaviors of the relative deep-surge frequencies fs0/f1 and the modified reduced deep-surge frequencies
fRPVVsm against the modified sectional area ratios APR* for the compressors having single stage through nine stages, respectively.
The relative deep-surge frequencies fs0/f1 change much from near unity to roughly 0.2, affected by changes in the modified flow-
path area ratios and the length ratios. The behaviors of the relative surge frequencies appear difficult to describe simply.
On the other hand, the modified reduced surge frequencies fRPVVsm show a relatively well-correlated manner of behaviors,
tending to the value of nearly 0.1. The green-colored chain-line approximates the behavior.
For the nine-stage compressor, Comp19, however, scatters in the data are seen to increase in Fig. 9, where upper data points
and lower ones having the same marks show fs0/f1, and fRPVVsm, respectively. The differences in the behavior tendencies suggest
some other features of surge phenomena.
Thus, although the relative surge frequencies fs0/f1 are convenient to estimate the levels of the frequencies directly in
comparison of the system resonance frequencies, the modified reduced surge frequencies fRPVVsm are better to reflect the essential
nature of the deep surge phenomena. It could be said that the deep-surges occur so that the values of the modified reduced surge
frequency fRPVVsm tend to have or approach some definite value.
All the data points of fRPVVsm vs. APR* are gathered together in Fig. 10. The approximate tendency could be given
approximately by the following two straight lines shown as dotted lines.
368
Fig. 6 Behaviors of relative deep-surge frequencies fs0/f1
and modified reduced surge frequencies fRPVVsm against
modified sectional area ratios APR* for the single-stage
compressor, Comp11
Fig. 7 Behaviors of relative deep-surge frequencies fs0/f1
and modified reduced surge frequencies fRPVVsm against
modified sectional area ratios APR* for the three-stage
compressor, Comp13
Fig. 8 Behaviors of relative deep-surge frequencies fs0/f1
and modified reduced surge frequencies fRPVVsm against
modified sectional area ratios APR* for the five-stage
compressor, Comp15
Fig. 9 Behaviors of relative deep-surge frequencies f
s0/f1 and modified reduced surge frequencies fRPVVsm
against modified sectional area ratios APR* for the
nine-stage compressor, Comp19
Fig. 10 Tendency of the modified reduced deep-surge frequencies fRPVVsm against
the modified area ratios APR* for all compressors including Comp11 through to
Comp19 (superposition of all points in Figs. 6-9)
369
1.0RPVVsmf for APR* > 0.7 (12)
*APRfRPVVsm for APR* < 0.7 (13)
For sufficiently large values of APR*, the modified reduced frequencies fRPVVsm tend to have a definite value, and for smaller
values of APR*, the frequencies fRPVVsm tend to change proportionally to the square root of the APR*.
The effects of the flow-path area ratios and the stalling pressure ratios are put together into APR*, and the effect of the flow-
path length ratio LR is included in the effective representative length √{Lp(Lc**+Lp)}.
In addition to the above, the data points encircled by a dotted ellipse line in Fig. 10 indicate occurrence of deep surges having
further lower frequencies for the nine-stage compressor Comp19 at 12000rpm. As shown later, the situation suggests ultra-low
frequencies caused by interventions of incomplete surge recoveries in high-speed operations of high-pressure ratio compressors.
5. Tendency of Surge Frequencies
The behavior of the following reduced surge frequency is surveyed.
t
effs
RPsmu
Lff
0 (14)
The parameter fRPsm has an ordinary simple form of reduced frequency easier to understand. At the same time, the following
relation holds from the definition of fRPVVsm, Eq. (9);
*
0
APR
f
u
Lff RPVVsm
t
effs
RPsm (15)
Figures 11-14 show the behaviors of the reduced surge frequencies fRPsm against the modified area ratios APR*. They have
nearly the same tendencies, all of which are superposed in Fig. 15, where the data points form a relatively narrow band of
distribution. In correlation with Eqs. (12) and (13), the tendencies are approximately expressed as follows;
#2-1 Improved
Fig. 11 Behavior of reduced surge frequency fRPsm
against the modified area ratio APR* for the
single-stage compressor Comp11
Fig. 12 Behavior of reduced surge frequency fRPsm
against the modified area ratio APR* for the three-
stage compressor Comp13
Fig. 13 Behavior of reduced surge frequency fRPsm
against the modified area ratio APR* for the five-
stage compressor Comp15
Fig. 14 Behavior of reduced surge frequency fRPsm
against the modified area ratio APR* for the nine-
stage compressor Comp19
370
*1
APRfRPsm for sufficiently large values of APR* (16)
*1
APRfRPsm for sufficiently small values of APR* (17)
Figures 10 and 15 could be the most compact representation of the macroscopic behaviors of the deep-surge frequencies fs0, in
terms of the major parameters, such as the compressor stalling pressure ratios PR, and the flow-path geometries given by AR and
LR.
However, some deviations from the average behavior are seen for the data points in the shaded zone in Fig. 14 and encircled
by a dotted ellipse line in Fig. 15, which are for the nine-stage compressor Comp19 at 12000 rpm, indicating occurrences of
further lower surge frequencies. The situation suggests ultra-low frequencies caused by interventions of incomplete surge
recoveries in high-speed operations of high-pressure ratio compressors, as shown later.
6. Tendency of Surge Frequencies for Changing Compressor Speeds
In order to see more plainly the effects of the compressor speeds and the flow-path geometries, behaviors of the modified
reduced deep-surge frequencies fRPVVsm against the flow-path volume ratios VR are shown in Figures 16-19. The groups for the
same length ratio LR are seen to gather together. Here
VR=AR*LR (18)
A point of intersection of a vertical line for the specified value of VR on the abscissa with a curve for a wanted compressor rpm in
the data group for the specified LR value will give the value of fRPVVsm for the situation. For the specified values of VR and LR, the
value of AR is determined by Eq. (18). Thus the effect of compressor tip speeds or rpms on fRPVVsm is known. The design rpm for
the present study is 11300 rpm.
The tendency of the behaviors of fRPVVsm affected by compressor rpms thus obtained are seen to change relatively little for
conditions of a specified compressor and a specified geometry of the flow-path, in many situations. It could be said that
frequencies of deep surges are determined so as to give some definite value of the modified reduced surge frequencies fRPVVsm
during the speed changes in most situations.
However, for the single-stage compressor Comp11, the situation is unsettled yet where the level of the fRPVVsm values increases
with increases in AR, and at the same time, with increases in the compressor rpms. As a whole, it is in the range of 0.02-0.07, less
than 0.1of the ordinary level for other compressors.
For the nine-stage compressor Comp19, the level of the fRPVVsm values keeps nearly 0.1 for compressor speeds lower than the
design speed, but it drops to further lower level for a higher speed, 12000rpm, and for the flow-path length ratio LR of 0.3 and 1. It
corresponds to the situation in Fig. 9 where the relative surge frequencies fs0/f1 approach 0.1. The situation will be explained later
in relation with Figs. 21 and 22.
Fig. 15 Tendency of the reduced deep-surge frequencies fRPsm against the modified
area ratios APR* for the compressors having a single stage through to nine stages
(superposition of all points in Figs. 11-14)
371
7. Surge Loops
Some examples of surge loops will be shown here in relation with the variations in the frequencies.
7.1 Three-stage compressor, Comp13, LR=0.3 and AR=2
Figure 20 shows surge conditions for the three-stage compressor, Comp13, and the flow-path condition of LR of 0.3 and AR of
2, (1) the pressure oscillogram, (2) the mass flow oscillogram, and (3) surge loops in the pressure-mass flow domain at the exit of
the compressor last stage. The compressor speeds are (a) 12000 rpm, (b) 10000 rpm, (c) 7000 rpm, and (d) 5000 rpm. Numerical
figures 2, 15, and 16 in the explanatory notes in Fig. 20(a) and (b) stand for near the inlet of the suction flow-path, the inlet of the
first stage of the compressor and the exit of the last stage, respectively. The surge behavior at 5000 rpm in Fig. 20(d) shows loops
completely within the stalled zone, meaning the stall-stagnation condition. The surge behavior at 7000 rpm in Fig. 20(c) shows
loops passing just near the stalling point, meaning the conditions near the stall-stagnation boundary. The surge behaviors at
10000rpm in Fig. 20 (b) and 12000 rpm in Fig. 20 (a) are deep surges, respectively. The surge frequencies are seen to change little
in these environments.
Detailed situations with respect to the stall stagnations can be found in Yamaguchi [8]. Very near the stall stagnation
boundaries, subharmonic deep surges can often be observed. In the stalled zone after the stagnation, small loops of mild surges
can be kept as seen in Fig. 20(b) or can decay to a point in the zone.
7.2 Nine-stage compressor, Comp19, LR=1 and AR=3.5
Figure 21 shows surge conditions for the nine-stage compressor, Comp19, and the flow-path condition of LR of 1.0 and AR of
3.5, (1) the pressure oscillogram, (2) the mass flow oscillogram, and (3) surge loops in the pressure-mass flow domain at the exit
of the last stage of the compressor. The compressor speeds are (a) 12000 rpm, (b) 10000 rpm, and (c) 7000 rpm. All the cases are
in the deep-surge situations.
The distinctive feature of the situation is that, as seen in Fig. 21(a-1), small peak(s) exist between neighboring high pressure
peaks of ordinary pressure recovery and re-stalling, encircled by a small red dotted-line circle. The situation is observed at 12000
Fig. 16 Behaviors of the modified reduced surge
frequency fRPVVsm against the flow-path volume ratio
VR for the single-stage compressor Comp11
Fig. 17 Behaviors of the modified reduced surge
frequency fRPVVsm against the flow-path volume ratio
VR for the three-stage compressor Comp13
Fig. 18 Behaviors of the modified reduced surge
frequency fRPVVsm against the flow-path volume ratio
VR for the five-stage compressor Comp15
Fig. 19 Behaviors of the modified reduced surge
frequency fRPVVsm against the flow-path volume ra
tio VR for the nine-stage compressor Comp19
372
rpm, which is higher than the design rpm of the compressor. Also in the behaviors of the mass flows, corresponding small
recovery is observed, as can be seen in Fig. 21(a-2) and (a-3). It immediately returns to near-zero flow, having failed in achieving
complete recovery from the stalled conditions. The phenomenon might be named incomplete recovery.
The detailed situation of the incomplete recovery is shown in Fig. 22 for a complete surge period. Figure 22(a) and (b) show the
oscillograms of the pressure and the mass flow, respectively, where the respective tag numbers indicate the inlet (15) and the exit
(16) of the compressor, the plenum (29 and 39), and the flow-path exit (49). A small peak of pressure recovery is located around
time k of 110000, halfway between the neighboring major peaks of pressure while the mass flow shows an abrupt increase from a
near-zero negative level to a rather high positive level, followed by an abrupt return to the near-zero negative level previously
observed. Figure 22(c) shows the pressure-vs-mass flow loops at the compressor exit and in the plenum. The nearly-horizontal
movements of both trajectories in the bottom zone are corresponding to the small recovery. The movement appears to be induced
by relatively free oscillation of the flow in the plenum. The mass flow in the plenum is seen to be oscillatory also in Fig. 22(b).
Figure 22(d) shows the behaviors of the coefficients of flow and pressure for stages 1, 5 and 9. Before the incomplete recovery all
the flow coefficients are negative or in the reversed flow condition. In the incomplete recovery, the stage flow coefficients recover
abruptly from the negative or reversed condition to a positive large one. In the condition, the stage flow coefficient increases
stage-by-stage from a near-stall one of the first stage toward the very large one of the last stage reaching as far as the zone of
negative pressure rise or turbine action, as can be seen in the levels of the ninth-stage coefficients of flow and pressureφ9 andψ9.
The recovery fails abruptly because of the abrupt decrease in the flow coefficient down to negative or reversed one and resulting
failure in increasing the pressure ratios.
Similar phenomena have been observed analytically in multi-stage compressors for other conditions of compressor speeds
and flow-path geometries. In the previous study (Yamaguchi [6]), it is found that the small peaks are corresponding to
instantaneous recoveries and re-stalling of only some rear stages during the low-pressure phase in the deep surge cycles. The
intervention of the incomplete recoveries elongates the surge period time, which, in turn, lower the surge frequency. Resulting
values of fRPVVsm and fRPsm tend also to lower. More than one small peaks of the incomplete surge recoveries could appear also in
the same surge period, depending on the situation.
Fig. 20 Variations of surge behaviors for the three-stage compressor, Comp13 for changes in the compressor
speeds. LR=0.3 and AR=2. (a) 12000 rpm, (b)10000 rpm, (c) 7000 rpm, and (d) 5000 rpm. And, (1) pressure
oscillogram, (b) mass flow oscillogram, and (c) surge loop p vs. W at the compressor exit.
373
Fig. 22 Local conditions of Comp19 at 12000 rpm for the flo
w-path geometry of LR of 1 and AR of 3.5. Conditions near th
e incomplete recovery are expanded in time.
Mass
flow
Fig. 21 Variations of surge behaviors for the nine-stage compressor, Comp19 for changes in the compressor speeds.
LR=1 and AR=3.5. (a) 12000 rpm, (b)10000 rpm, and (c) 7000 rpm. And, (1) pressure oscillogram, (b) mass flow
oscillogram, and (c) surge loop p vs. W at the compressor exit. Red dotted-line circles in Fig. 20(a) indicate
intervening incomplete surge recovery phenomena.
Fig. 22 Local conditions of Comp19 at 12000 rpm for the flow-path geometry of LR of 1 and AR of 3.5.
Behaviors around the incomplete recovery are expanded in time.
Mass flow
374
Concerning the situation, it is supposed as a possibility that, although the deep-surge frequency should have been given
fundamentally by the fRPVVsm value of approximately 0.1, the deep-surge recoveries in the situations might have been degenerated
into alternate recoveries of complete one and incomplete one, one after another, which could apparently have brought about the
apparently lowered deep-surge frequency, thus reducing the apparent fRPVVsm value to roughly a half of 0.1, or 0.03-0.06. It might
be a result of delicate interactions among the free oscillations in the plenum, the compressor stage working conditions changing
stage-by-stage, and other origins. At the present stage, the degeneration argument is only a proposition, and the author would like
to continue to investigate the mechanisms of the lowered deep-surge frequencies further.
It is to be careful here also that, in the present study, the stage characteristics of the rear stages are assumed not to change in the
environment of stalled front stages. If the stalling of the front stages could have significant effect on the characteristics of the rear
stages, the situation might change. The incomplete recovery phenomena have been observed only analytically as far as the author
knows. The author would like to mention the phenomena only as an analytical possibility.
8. Summary on the Deep-Surge Frequencies From the above analytical observations, it could be said that the deep-surge frequencies fs0 are determined so as to give the
modified reduced surge frequency fRPVVsm basically a definite value, for example 0.1 or some near-by value. The surge frequencies
fs0 are generally lower significantly than the acoustical resonance frequency in the system. The work required for the surge action
of filling and emptying the delivery volume could have reduced the frequency. The condition could be named a deep surge with a
lowered frequency caused by the filling and emptying action of surge.
On the other hand, for low-pressure-ratio compressors having a small number of stages, the relative surge frequency fs0/f1 is
nearly unity, which situation could be named a resonant or near-resonant surge. In the situation, the values of fPVVsm tend to be
lower somewhat than 0.1.
The resonance frequency f1 is principally the fundamental one for surges, which could be suppressed gradually in the presence
of the convection effects in the surge actions of large amplitude for larger pressure ratios.
For high-speed operation of high-pressure multi-stage compressors, both fs0/f1 and fRPVVsm tend to drop to levels lower than
those expected for ordinary deep surges, resulting in the lowered levels of fs0/f1 of roughly 0.1 and fRPVVsm of roughly 0.03-0.06. It
could be named a deep surge with a very low frequency caused by both of the surge action and the intervention of incomplete
recoveries. As a possibility, the situation could have been brought about by degeneration of the deep surges for fRPVVsm of roughly
0.1 into surges of alternate deep one and incomplete one where the apparent fRPVVsm values are reduced to about a half of 0.1. The
phenomenon requires further study, since it is uncertain
at the present stage whether it occurs actually or it is
only an analytical possibility, and at the same time, the
cause has not been made clear yet.
9. Conditions at Stall-Stagnation
Boundaries
Here, frequency conditions at the stall stagnation
boundaries are examined along the line of thinking
described above. In the neighborhood of stall stagnation
boundaries, a variety of surge behaviors have been
observed. For example, in the stalled zone, continued
mild surges and/or convergence or decaying onto a point
on the stalled branch are observed, depending on the
situations. In the deep-surge zone, deep surges with
subharmonic frequencies could often occur very near
the stagnation boundaries (Yamaguchi [8]). The
subharmonic surges are events containing both a deep
surge loop and one or more mild-surge loops, both
having nearly the same surge period, thus resulting in a
global surge period of some integer multiple of the basic
surge period, i.e., a deep surge having the subharmonic
frequency. The occurrence of subharmonic surges could
be regarded to be a precursor of the stall stagnation.
On the basis of examinations and considerations on
such many numerical results, the possibility of stall
stagnations is considered to be deeply related with the
condition whether the initial infinitesimal oscillation
having an acoustical resonance frequency could be
amplified or damped in the particular situation. In
general, surge oscillations could develop through
amplification of the acoustical disturbances to large-
amplitude disturbances having another frequency under
influence of the non-linear nature of the phenomenon. In
Fig. 23 Behaviors of the modified reduced resonance
frequency fR1mstg and the flow-path sectional area-pressure
ratio APRstg for the stall stagnation boundaries against the
flow-path length ratio LR
375
the sense, the concerned frequency to be paid attention to in the stall stagnation event is not the surge frequency fs0 but the
acoustical resonance frequency f1.
Thus a parameter of volume-modified reduced resonance frequency with particular respect to the resonance frequency has
been proposed and it proves to give a relatively well-correlated manner of behavior of the stagnation boundaries (Yamaguchi[6,
8]). Figure 23 shows the behaviors of the flow-path sectional area-pressure ratio APRstg and the reduced resonance frequencies
fR1mstg, given by Eq. (20) below, against the flow-path length ratios LR for the abscissa. Here, the subscript stg stands for the
conditions at the stagnation boundaries. The reduced resonance frequency given below by Eq. (20) has been improved from that
proposed previously by Yamaguchi [6], i.e., Eq. (19). The data are for compressors having 1-9 stages, at 10000 rpm, and a single-
stage compressor at 15000-6000 rpms.
The previous paper (Yamaguchi [6]) has shown that the modified reduced resonance frequency fRPVV1 given below by Eq. (19)
can reduce the many numerical-experimental data at the stall stagnation boundaries into a rather well-correlated curve. The term
√(Mt*PR) is added to adjust the tendency of the numerical-experiment data.
PRMPRA
A
u
Lff
t
P
t
PRPVV
C
1)
11(
1
11
2
(19)
Furthermore, if the effective representative length of the flow-path Leff is employed in place of Lp in Eq. (19) similarly to the
derivation of Eq. (9), the correlation might become better. The following formula is proposed;
PRMPRA
A
u
LLLff
t
P
t
pcp
mstgR
C
1)
11(
)(
1
**
11
2
(20)
The stagnation boundary conditions given by the above reduced resonance frequency fR1mstg are shown in the lower part of Fig. 23.
The improvement has proved effective in reducing the dependence of the parameter on the flow-path length ratios.
The values of the improved reduced resonance frequency fR1mstg at the stall stagnation boundaries are given in average by the
following level of magnitude, though some scatters are present for variations in the length ratios and compressors.
fR1mstg ~ 0.07 (21)
With respect to occurrence of surges, when the value of fR1mstg given by the right-hand side of Eq. (20) is smaller than the
threshold value given approximately by Eq. (21), the stall will decay or stagnates. On the other hand, when the value of fR1mstg is
larger than the threshold value given by Eq. (21), the infinitesimal disturbance having the acoustic frequency f1 will be amplified
and transformed into a deep-surge oscillation having a frequency fs0 evaluated from Eqs. (12) and (13).
The area-pressure ratio has the following definition at stall stagnation boundary;
APRstg=AR*PR (22)
With respect to the flow-path geometrical conditions, minimum values of APRstg range roughly from 1.5 to 2.5 for LR of 1 to 10
in large, as shown in the upper part of Fig. 23.
The threshold condition given above by Eqs. (20) and (21) is similar in form to the following one in Greitzer’s B parameter
(Greitzer [1 and 2]).
B = 0.8 (23)
Here, the B parameter is given by the following formula.
mcc
m
uLfLf
uB
11
1
4
1
4 (24)
When the resonance frequency f1 is approximated by the Helmholtz’ resonator formula, then Eq. (24) is given as follows;
cc
ppm
cc
pm
LA
LA
a
u
LA
V
a
uB
22 (25)
Here, um: average peripheral speed of the rotor blades, a: speed of sound, and Lc: length of the suction flow-path.
Yamaguchi [6] evaluated the threshold values of the Greitzer’s B parameter for cases of analytical stagnation conditions over a
wide variety of compressors and flow-path geometries. The results has indicated that the threshold values of the B parameter
distribute over a wide range roughly from 0.2 for small LR values to 10 for large LR values, in which the B value of 0.8 stands for
some restricted area near the average conditions. In comparison of these results, Eqs. (20) and (21) could be said to predict the
stall stagnation condition more reasonably.
10. Conclusions
Behaviors of deep-surge frequencies over a relatively wide range of conditions of compressors and flow-path geometries were
surveyed. The following conclusions are obtained.
(1) When the stalling pressure ratios are near unity and the pressure-modified sectional area ratio APR* are relatively small, the
relative surge frequencies fs0/f1 tend to be near unity, namely the deep-surge frequencies are near the resonance one. The
surge could be named resonant surge or near-resonant surge.
(2) With increases in the stalling pressure-ratios PR and in area ratios AR, the relative deep-surge frequencies fs0/f1 decrease,
falling to as low as 0.2-0.1.
(3) A modified reduced surge frequency fRPVVsm is important in describing the deep-surge situations of higher stalling pressure
ratios. It is a dimensionless parameter evaluating the filling and emptying actions in surges. It contains the surge frequency
376
fs0, the flow-path sectional area ratio AR, the compressor tip speed ut, the representative flow-path length including the
delivery length Lp and the total length of the flow-path (Lp+Lc**), and a pressure-ratio-affected term. Although the relative
surge frequency indicates only the frequency level, the modified reduced surge frequency is a more essential parameter
related with the surge actions.
(4) Deep surges tend to have the parameter fRPVVsm of roughly 0.1 or tend to approach the level. It could be the general condition
determining the deep surge frequencies. The condition could be named a deep surge with a lowered frequency caused by the
filling and emptying action of surge.
(5) For situations of compressors having sufficiently many stages, the relative surge frequencies fs0/f1 tend to fall as low as 0.1,
and the modified reduced surge frequencies fRPVVsm tend to lower below 0.1. In the situation, small pressure peaks appear
between neighboring pressure recovery peaks of pressure oscillations, caused by incomplete surge recoveries of rear stages.
It could be named a deep surge with a very low frequency caused by both of the surge action and the intervention of
incomplete recoveries. As its possible cause, the situation could have been brought about by degeneration of the deep surges
into a sequence of alternately occurring deep one and incomplete one where the apparent fRPVVsm values tend to drop to about
a half of 0.1.
(6) When the stall stagnation boundary conditions are satisfied, the deep surges disappear and a stall stagnation occurs. The
improved condition for the stall stagnation is the modified reduced resonance frequency fR1mstg of nearly 0.07. Near the
boundary, surge frequencies are changeable. For example, subharmonic deep-surges could often occur above the boundary.
Below the boundary occur various surge phenomena, such as mild surges within the stalled region, decaying surge
oscillations onto some point in the stalled area, etc.
11. Postword
The study could have made clear to some extent the brief behaviors of the deep-surge frequencies, which have been overlooked
so far. The possibility of estimation of the rough order of magnitude of the surge frequencies will contribute much not only to
understanding of the surge phenomena itself, but also to consideration of preventive measures against the oscillation problems in
the actual sites.
However, the results have been based solely on the numerical-experimental procedures. Confirmations with practical data will
be required, since situations that have not been confirmed practically or experimentally could possibly exist. For more details,
further analyses and experiments will be wanted.
Nomenclature
a Sound of speed (m/s) Lp Length of the delivery flow-path (m)
Ac1 Annulus sectional area of the first stage inlet LR Length ratio of the flow-path
(m2) PR Stalling pressure ratio
Ac Sectional area of the suction-compressor flow- rpm Compressor speed (rpm)
path (m2) rpm0 Compressor design speed (rpm)
Ac2 Sectional area of the exit of the last stage f the Leff Effective representative length of the flow-
compressor (m2) path (m)
Ap Sectional area of the delivery plenum (m2) rt1 Tip radius of the first rotor of the compressor
APR Sectional area-pressure ratio of the flow-path (m)
APRstg APR at the stall stagnation boundary um Mean peripheral speed of rotor of the
AR Sectional area ratio of the flow-path compressor (m)
B Greitzer’s B parameter ut1 Tip peripheral speed of the first rotor of the
f1 Basic resonance frequency of the whole flow- compressor (m)
path (Hz) VR Volume ratio of the flow-path
fR Reduced frequency κ Ratio of specific heats
fs0 Deep-surge frequency (Hz) φt Stage flow coefficient
fs0/f1 Relative surge frequency ψPt Stage pressure coefficient
fR1mstg Modified reduced resonance frequency for the ψTt Stage temperature coefficient
stagnation boundary
fRPVV1 Reduced resonance frequency for the subscript
stagnation boundary stg Stall stagnation boundary
fRPVVs Volume-modified reduced deep-surge s Deep-surge condition, or surge,
frequency 1 Inlet to the compressor, or basic frequency
fRPVVsm Modified reduced deep-surge frequency 2 Exit from the compressor
Lc Length of the suction duct and the compressor c Compressor
(m) p Delivery plenum
Lc** Length of the suction duct and the compressor t Rotor tip
(m)
377
References
[1] Greitzer, E. M., 1976, “Surge and Rotating Stall in Axial Flow Compressors Part I-Theoretical Compression System Model,”
ASME, J. Engineering for Power, Vol.98, pp.190-198
[2] Greitzer, E. M., 1976, “Surge and Rotating Stall in Axial Flow Compressors Part II-Experimental Results and Comparison
with Theory,” ASME, J. Engineering for Power, Vol.98, pp.199-217
[3] Boyer, K. M., and O’Brien, W. F., 1989, “Model Predictions for Improved Recoverability of a Multistage Axial-Flow
Compressor,” AIAA-89-2687
[4] Yamaguchi, N., 2014, “Surge Phenomena Analytically Predicted in a Multi-stage Axial Flow Compressor in the Reduced-
speed Zone,” International Journal of Fluid Machinery and Systems, Vol. 7, No. 3, pp. 110-124
[5] Yamaguchi, N., 2014, “A Study on the Fundamental Surge Frequencies in Multi-Stage Axial Flow Compressor Systems,”
International Journal of Fluid Machinery and Systems, Vol. 7, No. 4, pp. 160-172
[6] Yamaguchi, N., 2016, “A Comparison of Surge Behaviors in Multi-Stage and Single-Stage Axial Flow Compressors,”
International Journal of Fluid Machinery and Systems, Vol. 9, No. 4, pp. 338-353
[7] Yamaguchi, N., 2013, “Analytical Study on Stall Stagnation Boundaries in Axial-Flow Compressor and Duct Systems,”
International Journal of Fluid Machinery and Systems, Vol. 6, No. 2, pp. 56-74
[8] Yamaguchi, N., 2016, “Analytical Surge Behaviors in Systems of a Single-stage Axial Flow Compressor and Flow-paths,”
International Journal of Fluid Machinery and Systems, Vol. 9, No. 1, pp. 1-16
[9] Yamaguchi, N., 2013, ”Development of a Simulation Method of Surge Transient Flow Phenomena in a Multistage Axial Flow
Compressor and Duct Systems,” International Journal of Fluid Machinery and Systems, Vol. 6, No. 4 , pp. 189-199