Journal of Thermal Science Vol.26, No.4 (2017) 289296
DOI: 10.1007/s11630-017-0941-8
Article ID: 1003-2169(2017)04-0289-08
An Experimental Description of the Flow in a Centrifugal Compressor from Alternate Stall to Surge V. Moënne-Loccoz1, I. Trébinjac2, E. Benichou3, S. Goguey2, B. Paoletti2, and P. Laucher2 1. Safran Helicopter Engines Av. Joseph Szydlowsky, 64511 Bordes, France 2. Laboratoire de Mécanique des Fluides et d’Acoustique, UMR CNRS 5509, Ecole Centrale de Lyon, UCBLyon, INSA 36 av. Guy de Collongue, 69134 Ecully Cedex, France 3. FRATECH RP S.A.S, Seconded to BELGATECH Engineering Services SP Woluwe Gate Boulevard de la Woluwelaan, 2, 1150 Brussels, Belgium © Science Press and Institute of Engineering Thermophysics, CAS and Springer-Verlag Berlin Heidelberg 2017
The present paper gives the experimental results obtained in a centrifugal compressor stage designed and built by SAFRAN Helicopter Engines. The compressor is composed of inlet guide vanes, a backswept splittered unshrouded impeller, a splittered vaned radial diffuser and axial outlet guide vanes. Previous numerical simulations revealed a particular S-shape pressure rise characteristic at partial rotation speed and predicted an alternate flow pattern in the vaned radial diffuser at low mass flow rate. This alternate flow pattern involves two adjacent vane passages. One passage exhibits very low momentum and a low pressure recovery, whereas the adjacent passage has very high momentum in the passage inlet and diffuses efficiently. Experimental measurements confirm the S-shape of the pressure rise characteristic even if the stability limit experimentally occurs at higher mass flow than numerically predicted. At low mass flow the alternate stall pattern is confirmed thanks to the data obtained by high-frequency pressure sensors. As the compressor is throttled the path to instability has been registered and a first scenario of the surge inception is given. The compressor first experiences a steady alternate stall in the diffuser. As the mass flow decreases, the alternate stall amplifies and triggers the mild surge in the vaned diffuser. An unsteady behavior results from the interaction of the alternate stall and the mild surge. Finally, when the pressure gradient becomes too strong, the alternate stall blows away and the compressor enters into deep surge.
Keywords: centrifugal compressor, alternate stall, mild surge, deep surge
Introduction Compressors are limited in their operating range by instabilities occurring at low mass flow. As the mass flow is reduced, the pressure rise increases until the inertia of the flow can no longer overcome the pressure gradient. At this point, more or less severe instabilities can occur with more or less disastrous consequences for the compressor integrity. Received: April 2017
Classically, when reducing the mass flow, a compressor starts experiencing stall and then surge which are very different but linked phenomena [1]. Stall often refers to the setting of a circumferential non-uniform pattern rotating along the annulus at a speed of the same order as that of the rotor. The flow is no longer axisymmetric but contains regions of stalled cells spreading through the azimuthal direction. Depending on the number, the coherence and the size of the cells, the phenol-
V. Moënne-Loccoz: Doctorate www.springerlink.com
290
J. Therm. Sci., Vol.26, No.4, 2017
Nomenclature xstd Standard values
m std
x
Time-averaged values
ts stage
Mabs R P Pt p′ T Tt m
Absolute Mach number Impeller rotation speed Static pressure Stagnation pressure Pressure fluctuations Static temperature Stagnation temperature Mass flow
Rij a B Ac Lc U Vp fH
menon can be referred as part-span rotating stall or fullspan rotating stall. On the contrary, the surge process is often characterized by its axisymmetry and the oscillation in time of the overall annulus mass flow [2]. Depending on the violence of the process, surge can be referred as ’deep surge’ (destructive process with reversal of the mass flow) or ’mild surge’ (softer process with the operating point orbiting around the surge point). Yet this pattern is far from being universal and more complex paths to surge exist, especially in the centrifugal compressor field where stall does not automatically entail a decline in performance nor a first step toward surge. Whereas the instabilities can easily be displayed through post-treatment of high frequency measurements, predicting the onset and the links between those instabilities is still the trickiest part of the problem. At partial rotational speed, the centrifugal compressor described in this paper presents a very particular evolution towards surge involving instabilities with very few references. Previous numerical studies carried by Benichou and Trébinjac [3] in the same geometry revealed an alternate pattern evolving in the vaned radial diffuser with inter-blade passages occupied with a jet flow or stalled. This alternate pattern has been experimentally confirmed as described in this paper, simultaneously with mild surge. No other example of this behavior within aeronautical radial compressor is available in the literature. But similar alternate passages can nevertheless be found in hydraulic applications, either in the rotor or the stator on condition that the number of blades is even. Examples of this phenomenon are given by Acosta and Bowerman [4], Sano et al. [5], Pedersen et al. [6] and Braun [7]. The following results are the first experimental results obtained in this configuration with a special effort to detect the instabilities taking place in the compressor. First, the test case and the experimental setup are introduced. Then, the overall performance maps are showed. A review of the instabilities found in the compressor is
Corrected mass flow Total-to-static stage pressure ratio Cross-correlation between sensors i and j Speed of sound Dimensionless parameter Compressor flow area Compressor characteristic length Rotor outlet blade velocity Volume of compression system plenum Helmholtz resonator frequency
then exposed and a specific section is dedicated to the alternate stall phenomenon. Finally, a comprehensive description of the surge inception is given. Test Case The test case is a centrifugal compressor stage designed and built by SAFRAN Helicopter Engines within the framework of a research program. The compressor is composed of inlet guide vanes (IGV), a backswept splittered unshrouded impeller (IMP), a splittered vaned radial diffuser (RD) and axial outlet guide vanes (OGV). A meridional sketch of the stage is given in Figure 1.
Fig. 1
Meridional sketch of the compressor stage
Experimental Setup The compressor stage is mounted on a 1 MW test rig and is heavily instrumented. 130 steady sensors (temperature, pressure and vibration measurements) are used for the monitoring and the overall performance measuring. The rotational speed, the flow rate, the pressure and the temperature are measured at ±0.01%, ±0.5%, ±0.05% and ±0.5 K respectively. In order to capture the unsteady behaviour of the stage, 52 unsteady pressure sensors manufactured by Kulite Semiconductor Product, Inc. are installed at different locations on both the shroud and the hub of the compressor. The protection grids were removed in order to maximize the natural frequency of the probes allowing pressure measurements up to 150 kHz with an acquisition frequency of 500 kHz. In the following results, only the data from the sensors shown in Figure 2 are used. Their characteristics are given in Table 1.
V. Moënne-Loccoz et al.
An Experimental Description of the Flow in a Centrifugal Compressor from Alternate Stall to Surge
291
results are given below revealed that the no-converging zone corresponds to the mass flow range where mild surge occurs; that may explain why the numerical simulations could not properly converge in this flow range. The next section describes the instabilities evolving in the compressor before reaching surge.
Fig. 2 Table 1
Unsteady pressure sensors positions
Kulites unsteady pressure sensors characteristics Model
Position
XTE-190M-25PSIA XTE -7L-190M-100PSIA
IMP. Inlet (I1 to I6) RD (shroud) (D1 to D7)
Nat. frequency 240 kHz 380 kHz
Performance Map Figure 3 shows the performance map of the compressor. It depicts the total-to-static pressure ratio defined in Equation 1 as a function of the corrected mass flow defined in Equation 2. Both quantities are reduced by reference values. P ts (1) stage 1 Pt 0 m std m
Tt 0 Pt std Tt std Pt 0
(2)
Only one iso-speed line is studied in this paper corresponding to a partial rotational speed. The pressure rise characteristic presents an S-shape curvature (negative, positive then negative again) which is rather unusual for an aeronautical centrifugal compressor. Classically, centrifugal compressors show either a continuous negative slope or a negative slope at high mass flow followed with a plateau at lower mass flow. Some examples of S-shaped performance curves can be found in the literature. Díez et al. [8] presented numerical results in a compressor experiencing two stable operation ranges separated by an unstable zone. Zheng and Liu [9] presented experimental results in a compressor experiencing a two-regime-surge depending on the rotational speed: as the compressor mass flow is reduced, the compressor experiences mild surge before reaching another stable range and then finally goes into surge. Previous simulations performed by Benichou and Trébinjac [3] in the present test case showed comparable results to those obtained by Díez [8]. Numerical simulations were not able to converge over a large mass flow range (0.4 ≤ m m ref 0.6 ) but were able to converge at lower mass flow. The experiments which
Fig. 3
Stage pressure rise characteristic
Path to surge After reaching the Peak Efficiency conditions (PE in Fig. 3), if the compressor is further throttled, high-frequency rotating disturbances are observed at the impeller inlet (green zone in Fig. 3). They can be highlighted by the mean of cross-correlations. The cross-correlation between two signals P1 and P2 is formed from the ensemble average ‹› of the product of the two signals as shown in Eq 3 (where T is the total time of the signals). The cross-correlation is then normalized by the root square of the maximum of the auto-correlation of both signal P1 and P2 in order to obtain a coefficient between -1 and 1. 1 (3) R12 P1 t P2 t T When pressure measurements are realized at two distinct points with two sensors, the pressure signals will be correlated if they are delayed one from the other by a time τ, which corresponds to the travel time between these sensors. The time at which the two probes have maximum correlation gives the group speed of the travelling perturbation [10]. As such, if a coherent structure is steadily rotating at the impeller inlet, the pressure sensors will successively capture it, and the maximum correlation time will be proportional to the distance between the sensors. Figure 4(a) shows the cross-correlations between the pressure measurements at the sensor I6 (used as the reference) and all the other sensors (I1 to I6) located at the same axial position (Fig. 2). It shows proper peaks attesting the existence of correlations between the pressure measurement at I6 and the others. Figure 4(b) presents the time delay of the maximum of the cross-correlation
292
J. Therm. Sci., Vol.26, No.4, 2017
with the angular distance between the sensors. It confirms the existence of coherent structures rotating at 0.52 R in the impeller inlet. Other pressure measurements at operating points between deep surge and peak efficiency (0.48 ≤ m m ref 0.6 ) confirm the existence of such disturbances rotating at 50% to 60% of the rotor speed. Figure 5 presents the Power Spectral Density (PSD) of a pressure signal at I1 for two operating points: the dashed line corresponds to peak efficiency and the continuous line corresponds to a point within the zone of rotating disturbances occurrence. The characterization of the disturbances is quite difficult because the spectra of the signal experiencing the disturbances doesn’t reveal any specific peak but a bump composed of a large broadband of frequencies between a few hundred Hertz and the Blade Passing Frequency (BPF). It differs from classical rotating stall spectra as described by Dodds et al. [11] showing high discrete peaks corresponding to the rotating stall mode number and speed. In the present case, the extent of this frequency bump and the low energy included in these structures compared with the blade passing frequency exclude any rotating stall at the impeller inlet. Actually, these disturbances fit more to the
Fig. 4
Rotating disturbances in the Impeller: (a) Pressure inter-correlation with I6, (b) Maximum correlation time vs angular distance from I6
Fig. 5
Power Spectral Density at position I1
description of rotating instabilities proposed by Day [2] which rotate at around half the rotor speed and change in intensity and frequency (as the present case). However, the frequency spectrum observed at I1 does not correspond to the ones referenced while experiencing rotating instabilities as described by Young et al [12]. Indeed, the frequency spectrum presented by Young et al. [12] revealed a tighter bump located at approximately a third of the compressor BPF. If further throttled, the compressor starts to experience mild surge (orange zone in Fig. 3) which is characterized by local fluctuations of the mass flow without any reversal at the compressor scale. It is associated with high pressure fluctuations in the whole compressor at low frequency (a few dozens of Hertz). Figure 6 shows the unsteady pressure signals at sensors I1 to I6 as the compressor experiences mild surge. A correlation analysis (as performed in Fig. 4) of the low frequency filtered pressure signals did not lead to any evidence of any rotating structure. But the simultaneity of the pressure peaks suggests an axial propagation of the perturbation which is more characteristic of a surge phenomenon. Three distinct phases can be observed during a mild surge period. The oscillation phase (17 revs.), in which high frequency perturbations start and grow. Those perturbations correspond to the ones studied in Figure 4. They are rotating at around half of the rotor speed and show an equivalent frequency spectrum to the one in Figure 5. The breakdown phase (25 revs.) during which the compressor experiences high pressure level fluctuations. After a rise at a high value, the pressure suddenly drops below the initial one. A recovery period (11 revs.) in which the pressure in the impeller rises to its initial value without any high frequency fluctuations. The mild surge phenomenon remains until the compressor goes into deep surge. A discrete-time continuous wavelet transform of the unsteady pressure signals registered during throttling ramps from stable condition to surge is performed to give a more comprehensive representation of the path to instability and a characterization of the mild surge frequency evolution. The results are given in Figure 7. While throttling, the intensity and the frequency of the mild surge both slightly increase. The frequency grows from 10Hz to 12Hz before the stage enters into deep surge which frequency is around 6Hz. As stated by Fink et al. [13], mild surge should have a frequency of the order of the Helmholtz resonator frequency which represents the inverse of the characteristic time of the compression system up to the point it enters into deep surge. The Helmholtz frequency of the present stage given by the Equation 4 is around 16Hz which is very close to the mild surge frequency
V. Moënne-Loccoz et al.
An Experimental Description of the Flow in a Centrifugal Compressor from Alternate Stall to Surge
293
from low frequency deep surge to higher frequency mild surge. More recently, Zheng and Liu [9] reported a limit value Bcrit for which mild surge can appear with Helmholtz frequency, situated at the limit between the two precedent states. Further studies on the present test case at lower and higher rotational speed could confirm the existence of this critical value. In the present case, at the given rotational speed, B = 0.68. Vp U (5) B 2a Ac Lc
Fig. 6
Fig. 7
Mild surge period in the impeller at sensors I1 to I6
While throttling, the mild surge stays and grows until deep surge appears. Pressure measurements acquired at the sensor I1 (impeller inlet), D6 (radial diffuser inlet) and D7 (radial diffuser outlet), during a deep surge cycle are shown in Figure 8. The same phases as for mild surge (Fig. 6) can be distinguished for deep surge: the oscillation period, with a growth of the high frequency fluctuations, the breakdown phase with a decrease of the pressure at the diffuser outlet and high pressure fluctuations in the whole compressor and the recovery period, in which the compressor tries to get back to its stable state before entering surge again.
Wavelet Power Spectrum at impeller inlet (I1) during throttling ramps from stable condition to surge.
(10-12Hz). A lower frequency for the deep surge is relevant because it takes into account the emptying and filling times of the plenums. fH
a 2
Ac V p Lc
(4)
The relationship between surge frequencies and the geometrical system characteristics is far from being new. Greitzer [14] proposed a non-dimensional parameter B defined in Equation 5 that indicates the type of instabilities a compressor might experience. This parameter is representative of the ratio between pressure and inertial forces. For small value of B, a rotating stall phenomenon can sustain in the compressor. The flow inertia still manages to counter the pressure forces. While for high value of B, pressure forces are too strong and deep surge occurs directly without rotating stall. A critical value Bcrit seems to emerge at the limit of the two behaviors but this value is not universal and depends on the design of the compressor as Day [15] showed. Galindo et al. [16] analyzed the sensibility of the B parameter to the duct length Lc. By changing Lc, the compressor surge operation changed
Fig. 8
Unsteady pressure measurement during deep surge cycle
From all these results, it can thus be stated that the compressor first experiences rotating disturbances at the impeller inlet, then mild surge which finally evolves into deep surge. The next section aims at identifying the onset of the mild surge which is thought to be triggered in the radial vaned diffuser. Alternate stall The result of a previous numerical study conducted by Benichou and Trébinjac [3] on the present test case at very low mass flow ( m std m ref 0.4 ) are presented in Figure 9. Such a low mass flow operating point could not be experimentally reached because the compressor surges before. Figure 9 shows the time-averaged static pressure
294
J. Therm. Sci., Vol.26, No.4, 2017
(a) and the time-averaged absolute Mach number (b) maps in the radial diffuser at 95% of the blade height in two adjacent passages. The flow presents a particular pattern along the radial diffuser, with a stalled passage (channel 1) which experiences high pressure and low absolute Mach number at its inlet and a jet flow in the next passage (channel 2) characterized by high absolute Mach number and low pressure near the suction side. Equivalent results can be found in the experimental signals acquired in the compressor. The Figure 10 shows the evolution of the time-averaged pressure at sensors D0 to D6 in the diffuser inlet with the reduction of the mass flow. The slopes of the pressure curves are different depending on whether the probe belongs to channel 1 or 2 (square for channel 1 and triangles for channel 2). For m m ref 0.67 , while pressure signals at sensors D1 and D2 in channel 1 keep rising with the reduction of the mass flow, their equivalents in channel 2 (D4 and D5) start to level off and even to decrease ( m m ref 0.58 ). A 30 kPa difference can be measured between the signals at D5 and D1 at the lowest mass flow operating point attesting of an imbalance between the two channels. In
Fig. 9 Previous numerical results on alternate stall in radial diffuser at 95% blade height [3]: (a) Static pressure, (b) Absolute Mach number
Fig. 10
Average pressure in diffuser inlet at location D1 to D6
addition, for m m ref 0.67 , the decrease of the pressure at sensor D3 confirms the presence of a jet flow near the suction side of channel 2. This jet flow is not measured in channel 1, the pressure at D0 keeps rising with the reduction of the mass flow. Unsteady pressure signals presented in Figure 11 at the sensors D2 (red), D3 (blue) and D6 (purple) confirm this pattern. The plain lines give the unsteady pressure signals for an operating point experiencing mild surge while the dash lines give the time-averaged value of the pressure for the most throttled but stable operating point ( m m ref 0.6 ). Whereas the channel 1 is stalled (high pressure level at D2) the channel 2 lets flow (low pressure at D3). But, contrary to the results of the numerical simulations, this pattern reveals to be unsteady with the mild surge frequency. Thus during a mild surge period, the flow pattern periodically evolves in the diffuser. At high pressure peaks, the alternate pattern is strong, one channel is completely stalled (high pressure peaks at D2) and the other lets flow (high pressure drops in D3). Then the diffuser recovers, the pattern fades and the pressure levels tend to go back
Fig. 11
Fig. 12
Alternate pressure signal in the radial diffuser
Pressure in the diffuser during mild surge operation
V. Moënne-Loccoz et al.
An Experimental Description of the Flow in a Centrifugal Compressor from Alternate Stall to Surge
Fig. 13
295
Mild surge cycle
to the values of the stable operating point. Most of the time, the channels do not change roles: channel 1 remains the stalled channel and the high momentum flow stays in channel 2 for multiple mild surge periods. But randomly, the pattern reverts and the channels exchange states as happening at around 500 revolutions in Figure 11. Pressure measurements at D2 and D6 which are almost at the same circumferential position in their respective channel, really swap their roles. Unfortunately, no pressure sensor to an equivalent position of D3 is present in channel 1 to confirm an overall exchange of the diffuser channels alternate pattern. From alternate stall to mild surge The deep analysis of all the available unsteady pressure signals led us then to suggest that the alternate stall pattern is the triggering event of the mild surge as illustrated in Figure 13. As seen in Figure 10, the alternate state in the diffuser exists at a higher mass flow than the mild surge. It results from a small separation on the suction side of one over two diffuser vane [3] which could be caused by the rotating disturbances at the impeller inlet inducing a bad supply of the radial diffuser. As the mass flow is reduced, the pressure at the diffuser outlet increases, leading the separation to grow (step (a) in Fig. 13). At a certain point, the separation is large enough to induce a blockage of the entire channel and deviate the flow from channel 1 to channel 2 (step (b)). A reversed fluid material from channel 1 flows to channel 2 discharging the pressure at diffuser outlet (step (c)). With less pressure at diffuser outlet, the separation shrinks (step (d)). The diffuser channels tend to be better balanced, until the pressure at diffuser outlet increases again leading to another mild surge cycle. This scenario is supported by the pressure signals at the sensors D2, D3 and D7 during a mild surge cycle (Fig. 12). The four previous steps are easily identified: Step (a): The pressure at diffuser outlet (D7) starts to increase, so does the separation. When the se-
paration is large enough, around rotation 15, part of the fluid is deviated to channel 2 so the pressure at sensor D3 decreases and the pressure at position D2 increases. Step (b): When the pressure at diffuser outlet reaches its maximum, channel 1 is entirely blocked and the pressure levels at D2 and D3 reach a plateau. This step does not last a long time because the diffuser starts to discharge. Step (c): The diffuser starts to discharge as shown by the decrease in pressure at D7. The pressure at D2 remains at its maximum level indicating that channel 1 is still blocked. Step (d): As the diffuser pressure outlet decreases, the separation shrinks. At a certain point, the pressure at D2 starts to decrease indicating that more mass flow is passing through channel 1. On the contrary, pressure in D3 rises, indicating a diminution of mass flow in channel 2. Additional measurements are necessary to conclude on this scenario but it gives a first glimpse of what is happening in the diffuser during mild surge. This scenario could also be applied to deep surge except that the pressure at diffuser outlet is too strong for the separation to resist, the reversed flow would then be extended to the two channels and so to the entire compressor.
Conclusions From high-frequency pressure measurements, a comprehensive description of the path to surge of a centrifugal compressor has been proposed. As reducing the mass flow, the compressor first experiences rotating disturbances at impeller inlet travelling at around half of the rotor speed together with steady alternate stall in the vaned diffuser. If further throttled, the stage experiences mild surge which frequency is very close to the Helmholtz frequency. During mild surge, both rotating disturbances and alternate stall remain, the latter being then
296
highly unsteady. The alternate stall is thought to trigger the mild surge, which is relevant from the analysis of the pressure signals within the diffuser. However, the alternate stall onset remains unclear and its potential connection with the impeller inlet disturbances has to be further investigated.
Acknowledgements We would like to thank SAFRAN Helicopter Engines which supported this study. The research leading to these results has received funding from the European Union Seventh Framework Program (FP7) through the ENOVAL project under grant agreement n° 604999.
References [1] Greitzer, E. (1981). The stability of pumping systems – the 1980 freeman scholar lecture. Journal of Fluids Engineering, 103(2): 193–242. [2] Day, I. (2016). Stall, surge, and 75 years of research. Journal of Turbomachinery, 138(1): 011001–1–16. [3] Benichou, E. and Trébinjac, I., Numerical analysis of an alternate stall in a radial vaned diffuser. In ASME Turbo Expo 2016: Turbomachinery Technical Conference and Exposition (American Society of Mechanical Engineers, Seoul, South Korea, 2016). [4] Acosta, A.J. and Bowerman, R. (1957). An experimental study of centrifugal-pump impellers. Transactions of the ASME, 79(4): 1821–1839 [5] Sano, T., Nakamura, Y., Yoshida, Y., and Tsujimoto, Y. (2002). Alternate blade stall and rotating stall in a vaned diffuser. JSME international journal. Ser. B, Fluids and thermal engineering, 45(4): 810–819. [6] Pedersen, N., Larsen, P., S., and Jacobsen, C., B. (2003). Flow in a centrifugal pump impeller at design and offdesign conditions – part I: Partical image velocimetry (PIV) and laser doppler velocimetry (LDV) measurements. Journal of Fluids Engineering, 125(1): 61–72. [7] Braun, O. (2009). Part load flow in radial centrifugal pumps. Ph.D. thesis, Ecole Polytechnique Fédérale de
J. Therm. Sci., Vol.26, No.4, 2017 Lausanne. [8] Díez, N.G., Smeulers, J.P., Tapinassi, L., Del Greco, A.S., and Toni, L., Predictability of rotating stall and surge in a centrifugal compressor stage with dynamic simulations. In ASME Turbo Expo 2014: Turbine Technical Conference and Exposition, pages V02DT44A032– V02DT44A032 (American Society of Mechanical Engineers, Duesseldorf, Germany, 2014). [9] Zheng, X. and Liu, A. (2015). Phenomenon and mechanism of two-regime-surge in a centrifugal compressor. Journal of Turbomachinery, 137(8): 081007. [10] Allen, C., Blake, W., Dougherty, R., Lynch, D., Soderman, P., Underbrink, J., and Mueller, T., Aeroacoustic measurements (Springer Science & Business Media, 2013). [11] Dodds, J. and Vahdati, M., Rotating stall observations in a high speed compressor: Part 1 – experimental study. In ASME Turbo Expo 2014: Turbine Technical Conference and Exposition, pages V02DT44A011–V02DT44A011 (American Society of Mechanical Engineers, Duesseldorf, Germany, 2014). [12] Young, A., Day, I., and Pullan, G. (2013). Stall warning by blade pressure signature analysis. Journal of Turbomachinery, 135(1): 011033. [13] Fink, D.A., Cumpsty, N.A., and Greitzer, E.M., Surge dynamics in a free-spool centrifugal compressor system. In ASME 1991 International Gas Turbine and Aeroengine Congress and Exposition, pages 321–332 (American Society of Mechanical Engineers, Orlando, Florida, USA, 1991). [14] Greitzer, E.M. (1976). Surge and rotating stall in axial flow compressors – part I: Theoretical compression system model. Journal of Engineering for Power, 98(2): 190–198. [15] Day, I.J. (1994). Axial compressor performance during surge. Journal of propulsion and power, 10(3): 329–336. [16] Galindo, J., Serrano, J., Climent, H., and Tiseira, A. (2008). Experiments and modelling of surge in small centrifugal compressor for automotive engines. Experimental Thermal and Fluid Science, 32(3): 818–826.