116
This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA
Overview
Download & View 116 as PDF for free.
More details
- Words: 4,383
- Pages: 6
H. Hbnen
Monitoring of Aerodynamic Load and Detection of Stall in Multistage Axial Compressors
H. E. Gallus Institut fur Strahlantriebe und Turboarbeitsmaschinen, Rheinisch-Westfalische Technische Hochschule Aachen, Aachen, Federal Republic of Germany
The unsteady flow in a single-stage axial flow compressor at different operating conditions has been investigated with hot-wire and hot-film probes to find out the influence of the aerodynamic compressor load on the periodic fluctuations. These results are compared with measurements in the last stages of a multistage highpressure compressor of a gas turbinefor normal operation and under stall conditions. From the patterns of the frequency spectra of the measuring signals a parameter for the detection of the approach to the stability line of a compressor is derived. A method for the on-line monitoring of the aerodynamic load is presented. Based on these results a monitoring system has been developed. First experiences with this system, applied to two multistage compressors, are reported.
Introduction The trend to higher power densities and the wish for increasing efficiencies in turbomachines requires an operation with decreasing distance to the stability line of a compressor. Therefore the knowledge of this limit is the basis for an optimal operation. With the progress in the application of measuring techniques for unsteady flows, various investigations on the topic of detection and prediction of rotating stall have been made. Most of them were carried out in research compressors with laboratory techniques. These investigations give a basis for the understanding of the flow effects near the stability line of a compressor. Detailed measurements in axial compressors with arrays of hot-wire sensors illustrate the phenomena at the inception of rotating stall (Day, 1993; Poensgen, 1991). The changes in unsteady flow field are evident and point to a possibility to use these effects for the prediction of compressor stability line. Paduano et al. (1993) presented a system for the active control of rotating stall. With an array of hot-wire sensors the unsteady flow field was measured over the circumference. The detection of rotating waves of axial velocity indicated the inception of rotating stall. Rotating waves of velocities and static wall pressure were also observed by Gamier et al. (1991) as an indicator for stall inception. Gallus and Honen (1986) showed that in regions of separated flow all periodic fluctuations are damped and covered by increased noise caused by random fluctuations. Inoue et al. (1991) investigated the flow in two isolated axial flow compressor rotors near stall. He showed that the collapse of the periodic fluctuations caused by a large stall region is an indicator for the approach to the stability line of a compressor.
Contributed by the International Gas Turbine Institute and presented at the 38th International Gas Turbine and Aeroengine Congress and Exposition, Cincinnati, Ohio, May 24-27, 1993. Manuscript received at ASME Headquarters February 12, 1993. Paper No. 93-GT-20. Associate Technical Editor: H. Lukas.
All these investigations demonstrate the success in detection of stall in axial compressors. Nevertheless, the prediction of the stability limit, especially in multistage compressors is still an urgent problem that has to be solved. Measurements in a Single-Stage Research Compressor Stator Fundamental investigations were provided in a one-stage subsonic axial flow compressor with a hub-to-tip ratio of 0.38. The rotor speed was variable and the flow rate could be changed by a throttling valve system at the compressor outlet. Further constructional details of the compressor can be found from Gallus et al. (1979). The experiments at different aerodynamic blade loads were carried out with hot-wire probes and hot-film glue-on probes. The hot-wire probes were traversed inside the stator between two stator blades from the hub to the casing. For a highly loaded operating point of the compressor the position and extension of the corner stall derived from measurements with different measuring techniques (Gallus and Honen, 1986) is shown in Fig. 1. A hot-wire probe traversed near the suction side trailing edge of the vanes measures undisturbed flow in the upper part of the blade channel. In the lower part the probe is positioned inside the separation zone. The distributions of the real time signals for these two regions and the corresponding frequency spectra show the influence of separated flows onto the periodic fluctuations (Fig. 2). In the frequency spectrum for undisturbed flow the peaks for the rotor blade frequency and the first harmonic causes by the rotor wakes passing the stator can be detected. For the measurements in the separated region the signal shows only strong random fluctuations. In the frequency spectrum for this case only a high noise level without periodic parts can be observed. In the midspan of one stator blade on the suction side a hotfilm glue-on probe with 11 sensors was mounted (Fig. 3). For two different operating points (
Journal of Turbomachinery
JANUARY 1995, Vol. 117/81
Copyright © 1995 by ASME Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
f, =0.73
or =150 mm
tfi = C m / U 2
hub Fig. 3
Fig. 1
Corner stall region in the single-stage axial compressor
Measuring blade with hot-film glue-on probes
technique the same effects can be observed as in the measurements with hot-wire probes. All sensors outside the separation region detect the periodic fluctuations from the rotor wakes. The sensors inside the separated regions only detect high random fluctuations. This is also visible in the frequency spectra. For the operating point
Measurements in a Multistage Compressor Measurements in a multistage axial compressor of a LM5000 gas turbine validate and utilize this phenomenon. For a longterm investigation of the unsteady pressure distribution in eight stages of the 14-stage high-pressure compressor piezoelectric pressure transducers were mounted in the casing between rotor and stator. The transducers were connected to a FFT-analyzer via a multiplexer (MUX) and to a multichannel magnetic tape recorder (Fig. 5). Fig. 2 Measuring results from a hot-wire probe for different radii The measuring equipment was automatically controlled by the compressor the separation lines on the surface are drawn. a PC computer, which also-picked up the measuring data from In the case of a flow rate of
Nomenclature a ADC c c / HPC n p pfre
= = = = = = = = =
weighting coefficient analog-to-digital converter velocity chord length frequency high-pressure compressor rotational speed pressure frequency amplitude of pressure fluctuations
8 2 / V o l . 117, JANUARY 1995
pfre = mean value of frequency spectrum PS = pressure side pval = calculated fluctuating coefficient r = radius SL = stall level SP = signal processor SS = suction side t = time u = circumferential velocity
VSV = variable stator vanes x = coordinate in chord direction
Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
RHC Spa**™, K m l A 00 Uart: O.OOOOOEtOOJ] ( Kwal 1 ] MOHE <= •
1/0=0.54
* x*c -0.58
f (/)
\
;tyy(**»V M Sv***! «/c -0.B6
POJJP « :>
f (3)
Tt
Oura
<- 1
Curs
1 ->
r i Hartfcopy F2 Plotter
o:r C
l*Wv>i**f*^*
frequeni 50.0000 naa_i : S.02844E-03
0.0 HUE_Zu63.rrT
;!MvWf>/,wV^V,txlc -°
Trigger 0 -1/16
Fig. 6 Frequency spectrum in the third stage for compressor operation near design point
:!S1)/^yVw^KW!»''-ii-97 P2/P1
40
Trigger t[s] 0.005 ' - i - J
Fig. 4(a) Real time signals and frequency spectra lor a moderately loaded operating point (^ = 0.73)
30 20 10 I I I t . , -M t * I
140
160
180
200
Fig. 7 Pressure ratio of the H PC for a period of three days before turbine shut down
0.005
5 f[kHi]
t [s] 0
Fig. 4(b) Real time signals and frequency spectra for a highly loaded operating point (
interface
FFTanalyzer
D
MUX
CE=£
T
8 amplifiers
Fig. 5
PC computer
\ tape recorder
host computer
Experimental setup for long-term measurements
Journal of Turbomachinery
the controlling computer of the gas turbine (host computer) were received. Changing operating conditions of the compressor influence the unsteady flow field in a blading that can be detected by the casing-mounted dynamic pressure sensors. The frequency spectra of these measured real time signals show different patterns. This effect was used for the automatic long-term investigation of this compressor. After the start of the gas turbine, the first frequency spectra of each sensor signal were stored as a reference pattern. When during operation differences occurred that exceeded a defined range, the magnetic tape was started by the computer and picked up the real time signals of all sensors. At the same time the current frequency spectra were stored as new reference patterns. By this procedure all changes in the unsteady flow field in the stages could be observed and saved. For normal compressor operation the frequency spectrum of each stage sensor shows the peak for rotor blade frequency of this stage (Fig. 6). But also the characteristic frequencies of the neighboring stages are visible (marked with numbers for the several stages). The decrease of the fluctuating amplitudes at higher frequencies is caused by the damping of the adapters in which the pressure sensors were mounted. The investigations were carried out over a time period of about four months at nearly constant operating conditions of the gas turbine. Figure 7 shows the pressure ratio of the compressor for a period of three days just around the occurrence of stall. The Xaxis shows the absolute number of measurement and the Y axis the pressure ratio of the whole compressure (p2 = outlet pressure, p\ = inlet pressure). At the end of this period a strong increase caused by a fault of the controlling system occurred. At the maximum pressure the surge control of the compressor became active and the gas turbine was shut down. JANUARY 1995, Vol. 117/83
Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
f(8l
f<7)
f(9,10) f i l l , 12,13) 11th stage 11th stage
'.00E-O3 _|
1
1
1
1
1
_,
1
,
,_
1.0CE-01
f(8) f(9,10) f(11.12,13)j
13th stage
1 3 t h stage
Fig. 8(a) Frequency spectra of the 11th and 13th stages for normal compressor operation Fig. 8(c) Frequency spectra of the 11th and 13th stages just after gas turbine shutdown
1.O0E4OO
f(9,10) L
(V)
1
f(8> /
III h n7Ki
f(11,12.13)
/
1/ VAJ - ^ yw K
1.00E-03 . ~
vlyA>
1 1 t h stage
— i — i — i — i — i — i — i — i — i —
1 3 t h stage
Fig. 8(b) Frequency spectra of the 11 th and 13th stages for compressor operation near the stability limit
Figures 8(«-c) show the frequency spectra of the 11th and 13th stage sensors for three different points of time around the shut down point. The patterns for normal operating conditions are demonstrated in Fig. 8(a). The peaks for the rotor blade frequency of the 11th to 13th stage with 76 rotor blades each are clearly visible. Also the 9th and 10th stage peaks (60 rotor blades each) stand out of the noise in the spectrum. Hereafter, the fault in the control system occurred and the operating point of the compressor drifted slowly toward the 84 / V o l . 117, JANUARY 1995
stability line. The spectra during these operating conditions show different patterns in the 11th and 13th stages (Fig. 8b). Whereas the periodic fluctuations in the 13th stage disappeared, in the 11th stage an increase of the characteristic frequency peak can be observed due to increased profile loads and wakes. Caused by the high pressure ratio of the compressor, the aerodynamic load of the 13th stage becomes so high, that lowenergy material from increased wakes or even separated profile boundary layers (approaching stall) is centrifugated from the rotor blades. It accumulates near the casing behind the rotor and forms here a casing stall region, which reduces the pressure rise in this stage. The pressure sensor mounted in the gap between rotor and stator is covered by this low-momentum material so that the information from the undisturbed flow outside of the stall is extinguished at the measuring position due to increased turbulence and noise. Due to the high aerodynamic load of the compressor and the reduced pressure rise in the last stages, the load of the upstream stages is increased. This causes a growth of the rotor wakes in these stages so that here the characteristic frequency peaks increase (see Fig. &(b), 11th stage). The multistage compressor runs under stable conditions although in the 13th stage the limit for stall as it was defined for a single stage (Inoue et al., 1991) is exceeded. After about a further 30 minutes the compressor load entered such a high level that surge occurred and the surge control system shut down the gas turbine. As the automatic measuring system had a cycle time of about 20 seconds from the start of the FFT analysis from the first to the last sensor the exact point of compressor surge was not recorded on the magnetic tape. The signals taken during this operation phase started the recorder a few moments after this event. Just after the shutdown the compressor load diminishes and the frequency information of several stages can be detected again in the 13th stage (Fig. 8c). Due to the deceleration of rotor speed, the characteristic frequencies occur at lower values. Transactions of the ASME
Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
pval
Alarm Ports
SP 1
SP 2
SP 3
Amplifiers
stall near the casing
time
Fig. 9 Change of amplitudes at characteristic frequencies with increasing blade load Fig. 10 Hardware setup for the on-line monitoring system
Design of an On-Line Monitoring System Based on these results, a monitoring system for detecting the approach of the stability line in a multistage compressor has been developed. The measurements showed that the peak level at the characteristic frequency of a stage is an indicator for the aerodynamic load. Up to the point of the collapse of the periodic fluctuations, the peak amplitude grows out of the normal noise level, which also rises with increasing aerodynamic load (Fig. 9). This behavior can be described in the following form:
very small compared to the increased noise level. Due to the lower pressure rise in this stage the loads of the upstream stages increase. Thus the amplitudes of the characteristic frequency peaks are enlarged. After the collapse of the observed frequency peak, the value of pval will decrease. Thus, although the compressor load is higher, a lower value for the stall level SL would be calculated with Eq. (2). In order to obtain an increasing trend with higher compressor load, the calculation of this parameter has to be switched. When in the 13th stage casing stall occurs, the maximum value of pval13 just before the collapse of the frequency Pfrestage-Pfre amplitude is stored and used for the further stall level calcuPostage — (1) lations. The same procedure will be applied if also in the next pfre upstream stage casing stall damps the periodic information at As the measuring results from the multistage compressor the sensor. When the operating conditions change back to demonstrated, the collapse of the periodic information at the unseparated flow caused by lower aerodynamic loads, the cal13th stage sensor is not the limit for the stable compressor culation switches back to the former way (Eq. (2)). operation. Therefore, additional information must be used to This algorithm was applied to the unsteady pressure measdefine another parameter for the observation. urements of the long-term investigation (Fig. 8). It is mentioned Measurements of unsteady wall pressures in three neigh- that in this case the stall level was only calculated with the boring stages provide information on the stall development measuring results of the 11th and 13th stages. For the last from the last to the upstream stages. Therefore, dynamic pres- measurements just before the shutdown (see Fig. 7) with insure transducers are located in the axial gaps between rotor creasing pressure ratio of the compressor, the stall level exceeds and stator of the 11th, 12th, and 13th stages. a value of 15. During the whole investigation period of four In order to get an informative value for the operator or an months at normal compressor operation, values up to 10 were operating and control system from the measuring results in calculated for the stall level. each stage, a new parameter was defined for the monitoring This procedure of calculating a monitoring parameter for of the compressor load. This value (called stall level SL) takes the compressor load described above was the foundation for into account the amplitudes of the frequency peaks in the three the design of an on-line monitoring system on the basis of a observed stages and the information about flow separation. PC computer with 80386-CPU and signal processor boards for Since with increasing compressor load the amplitudes in the parallel signal acquisition and enhancement (Fig. 10). The unstages grow, this parameter can be defined in the following steady pressure signals are Fourier analyzed by the signal proform: cessors. The CPU coordinates the tasks of the processors, calculates the stall level from the frequency spectra, and sends SL = an pvali3 + a12 pval,2 + «ii pvaln (2) the results to a host computer where the data are displayed on wherein au, o12, and ctn are weighting coefficients for the a monitor and stored for a trend analysis. influence of the stages on the stability limit of the compressor. The optimal choice of these coefficients depends upon the compressor operating conditions and geometry. They are still Experiences With the Monitoring System in Operation the subject of further investigation and are proprietary. This system was applied to the high-pressure compressors As previously mentioned (see Fig. 8) for the shutdown of of two gas turbines of the type LM5000. In order to get an the compressor, the 11th stage was not yet separated at the impression of the reliability of the monitoring system, the stall casing measuring position. Therefore, for this compressor, level is compared with the operating parameters of the comterms of peak values (pval) from further upstream stages were pressor measured by the normal control system of the gas not included. For other part-load operating points the station turbines. The trends of the speed of the high-pressure rotor of measurements and the terms of the stages to be included in and the pressure ratio of the HPC show a satisfying relation the stall level value (SL) must be selected correspondingly. to the observed stall level (Figs. 11« and b). When casing stall in the 13th stage occurs the periodic flucWith increasing pressure ratio, the stall level also grows up tuations at the sensor are damped so that the observed char- to values of about 7. After a while it decreases slowly although acteristic peak in the frequency spectrum disappears or becomes the pressure ratio keeps constant. The comparison with the Journal of Turbomachinery
JANUARY 1995, Vol. 117/85
Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
1991). The mixing of tip leakage vortex with low-momentum material centrifuged from the rotor blades play a nonnegligible role in the development of casing stall. This causes different limit values of stall level with changing geometries and operating conditions. After longer operation time of the monitoring system more measuring data for different operating conditions of the compressor will be available. With these experiences the model for the definition of the limit value can take into account more influence parameters.
09.09.1992 ( 7.00 h - 10.00 h )
80 120 time [min] Fig. 11(a) Stall level over a time period of three hours
200
09.09.1992 1 7.00 h - 10.00 h I
njty/<\l>Y~v) 2
pressure ratio
rh
21
I*
t
VSV angle 19
80
120
Summary and Conclusions Based on experimental results of unsteady flow in a single and a multistage compressor a monitoring system for stable compressor operation has been developed. By detection of aerodynamic load and casing stall in three neighboring end stages of a multistage high-pressure compressor, the approach to the stability line can be observed. Experiences with the monitoring system at two compressors demonstrate the reliability of the method. Changes in compressor load cause differences in calculated stall level. Additional measuring results from system operation will allow a more exact definition of various influence parameters. By the detection of casing stall in the end stages only a stall monitoring near-design speed of the compressor can be realized at the present state. The system will be able to obtain monitoring data for lower speeds if measurements also will be taken in the front stages. Furthermore, additional measurements in different compressors can help to obtain further information about the influence of various geometries and loads.
200
time [min]
Fig. 11(6) HPC pressure ratio and VSV angle over a time priod of three hours
Acknowledgments The authors wish to acknowledge the DOW Stade GmbH for the financing of the investigations leading to this research work. Especially, we wish to thank Mr. H. Walter and Mr. E. Mizera for their support in preparing the experiments and the helpful comments concerning the application of the monitoring system. The discussions about the measuring results during the test phase are gratefully acknowledged. A patent application that relates to the technology covered by this work has been filed by DOW Deutschland Inc.
VSV angle (Fig. lib) shows the reason for this behavior. At the same time when the stall level decreases, the VSV angle moves from an open position toward 0 deg so that the air flow through the compressor decreases. This lower aerodynamic load causes the decrease in stall level. In order to get information about the vicinity to the stability line of the compressor, a limit value for the stall level has to be defined. The results from the long-term investigation and the experience with the monitoring system until now allow an References J. J., Celestina, M. L., and Greitzer, E. M., 1993, "The Role of estimation of this limit. The measurements and trends have TipAdamczyk, Clearance in High-Speed Fan Stall," ASME JOURNAL OF TURBOMACHINERY, shown that the 13th stage of the investigated compressor can Vol. 115, pp. 28-39. operate stably with casing stall without the compressor exDay, I. J., 1993, "Stall Inception in Axial Flow Compressors," ASME JOURceeding the stability line. The stall region covers only a part NAL OF TURBOMACHINERY, Vol. 115, pp. 1-9. Gallus, H. E., Lambertz, J., and Wallmann, Th., 1979, "Blade-Row Interof the blade height so that stable compressor operation is still action in an Axial-Flow Subsonic Compressor Stage," ASME Paper No. 79possible. Assuming that with occurrence of casing stall also in GT-92. the 12th stage the stall region in the 13 th stage covers a large Gallus, H. E.,andH6nen, H., 1986, "Experimental Investigations of Airfoilpart of the span, this point would be the limit of stable com- and Endwall Boundary Layers in a Subsonic Compressor Stage," ASME Paper pressor operation. Thus, if the maximum of pval12 is known, No. 86-GT-143. V. H., Epstein, A. H., and Greitzer, E. M., 1991, "Rotating Waves the limit value of stall level can be calculated with Eq. (2). As as Gamier, a Stall Inception Indication in Axial Compressors," ASME JOURNAL OF the three observed stages have the same blade numbers and TURBOMACHINERY, Vol. 113, pp. 290-302. Inoue, M., Kuroumaru, M., Iwamoto, T., and Ando, Y., 1991, "Detection geometries in each rotor and each stator row, the patterns of periodic fluctuations should look quite similar. So, the max- of a Rotating Stall Precursor in Isolated Axial Flow Compressor Rotors," ASME OF TURBOMACHINERY, Vol. 113, pp. 281-289. imum value of pvalu will be in the same range as the maximum JOURNAL Paduano, J. D., Epstein, A. H., Valavani, L., Longley, J. P., Greitzer, E. of pvalo just before occurrence of casing stall in the 13th stage. M., and Guenette, G. R., 1993, "Active Control of Rotating Stall in a LowBy inserting these values into Eq. (2) the stall limit is calculated. Speed Axial Compressor," ASME JOURNAL OF TURBOMACHINERY, Vol. 115, pp. The value of the stall limit is influenced by many geometric 48-56. Poensgen, Ch., 1991, "Experimentelle Untersuchung der Stromung in einer parameters of the bladings. One important factor is the tip Unterschall-Axialverdichterstufe bei hoher Drosselung und im Rotating Stall," clearance of the rotor (Adamczyk et al., 1993; Inoue et al., Dissertation, RWTH Aachen, Federal Republic of Germany.
8 6 / V o l . 117, JANUARY 1995
Transactions of the ASME
Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
Monitoring of Aerodynamic Load and Detection of Stall in Multistage Axial Compressors
H. E. Gallus Institut fur Strahlantriebe und Turboarbeitsmaschinen, Rheinisch-Westfalische Technische Hochschule Aachen, Aachen, Federal Republic of Germany
The unsteady flow in a single-stage axial flow compressor at different operating conditions has been investigated with hot-wire and hot-film probes to find out the influence of the aerodynamic compressor load on the periodic fluctuations. These results are compared with measurements in the last stages of a multistage highpressure compressor of a gas turbinefor normal operation and under stall conditions. From the patterns of the frequency spectra of the measuring signals a parameter for the detection of the approach to the stability line of a compressor is derived. A method for the on-line monitoring of the aerodynamic load is presented. Based on these results a monitoring system has been developed. First experiences with this system, applied to two multistage compressors, are reported.
Introduction The trend to higher power densities and the wish for increasing efficiencies in turbomachines requires an operation with decreasing distance to the stability line of a compressor. Therefore the knowledge of this limit is the basis for an optimal operation. With the progress in the application of measuring techniques for unsteady flows, various investigations on the topic of detection and prediction of rotating stall have been made. Most of them were carried out in research compressors with laboratory techniques. These investigations give a basis for the understanding of the flow effects near the stability line of a compressor. Detailed measurements in axial compressors with arrays of hot-wire sensors illustrate the phenomena at the inception of rotating stall (Day, 1993; Poensgen, 1991). The changes in unsteady flow field are evident and point to a possibility to use these effects for the prediction of compressor stability line. Paduano et al. (1993) presented a system for the active control of rotating stall. With an array of hot-wire sensors the unsteady flow field was measured over the circumference. The detection of rotating waves of axial velocity indicated the inception of rotating stall. Rotating waves of velocities and static wall pressure were also observed by Gamier et al. (1991) as an indicator for stall inception. Gallus and Honen (1986) showed that in regions of separated flow all periodic fluctuations are damped and covered by increased noise caused by random fluctuations. Inoue et al. (1991) investigated the flow in two isolated axial flow compressor rotors near stall. He showed that the collapse of the periodic fluctuations caused by a large stall region is an indicator for the approach to the stability line of a compressor.
Contributed by the International Gas Turbine Institute and presented at the 38th International Gas Turbine and Aeroengine Congress and Exposition, Cincinnati, Ohio, May 24-27, 1993. Manuscript received at ASME Headquarters February 12, 1993. Paper No. 93-GT-20. Associate Technical Editor: H. Lukas.
All these investigations demonstrate the success in detection of stall in axial compressors. Nevertheless, the prediction of the stability limit, especially in multistage compressors is still an urgent problem that has to be solved. Measurements in a Single-Stage Research Compressor Stator Fundamental investigations were provided in a one-stage subsonic axial flow compressor with a hub-to-tip ratio of 0.38. The rotor speed was variable and the flow rate could be changed by a throttling valve system at the compressor outlet. Further constructional details of the compressor can be found from Gallus et al. (1979). The experiments at different aerodynamic blade loads were carried out with hot-wire probes and hot-film glue-on probes. The hot-wire probes were traversed inside the stator between two stator blades from the hub to the casing. For a highly loaded operating point of the compressor the position and extension of the corner stall derived from measurements with different measuring techniques (Gallus and Honen, 1986) is shown in Fig. 1. A hot-wire probe traversed near the suction side trailing edge of the vanes measures undisturbed flow in the upper part of the blade channel. In the lower part the probe is positioned inside the separation zone. The distributions of the real time signals for these two regions and the corresponding frequency spectra show the influence of separated flows onto the periodic fluctuations (Fig. 2). In the frequency spectrum for undisturbed flow the peaks for the rotor blade frequency and the first harmonic causes by the rotor wakes passing the stator can be detected. For the measurements in the separated region the signal shows only strong random fluctuations. In the frequency spectrum for this case only a high noise level without periodic parts can be observed. In the midspan of one stator blade on the suction side a hotfilm glue-on probe with 11 sensors was mounted (Fig. 3). For two different operating points (
Journal of Turbomachinery
JANUARY 1995, Vol. 117/81
Copyright © 1995 by ASME Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
f, =0.73
or =150 mm
tfi = C m / U 2
hub Fig. 3
Fig. 1
Corner stall region in the single-stage axial compressor
Measuring blade with hot-film glue-on probes
technique the same effects can be observed as in the measurements with hot-wire probes. All sensors outside the separation region detect the periodic fluctuations from the rotor wakes. The sensors inside the separated regions only detect high random fluctuations. This is also visible in the frequency spectra. For the operating point
Measurements in a Multistage Compressor Measurements in a multistage axial compressor of a LM5000 gas turbine validate and utilize this phenomenon. For a longterm investigation of the unsteady pressure distribution in eight stages of the 14-stage high-pressure compressor piezoelectric pressure transducers were mounted in the casing between rotor and stator. The transducers were connected to a FFT-analyzer via a multiplexer (MUX) and to a multichannel magnetic tape recorder (Fig. 5). Fig. 2 Measuring results from a hot-wire probe for different radii The measuring equipment was automatically controlled by the compressor the separation lines on the surface are drawn. a PC computer, which also-picked up the measuring data from In the case of a flow rate of
Nomenclature a ADC c c / HPC n p pfre
= = = = = = = = =
weighting coefficient analog-to-digital converter velocity chord length frequency high-pressure compressor rotational speed pressure frequency amplitude of pressure fluctuations
8 2 / V o l . 117, JANUARY 1995
pfre = mean value of frequency spectrum PS = pressure side pval = calculated fluctuating coefficient r = radius SL = stall level SP = signal processor SS = suction side t = time u = circumferential velocity
VSV = variable stator vanes x = coordinate in chord direction
Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
RHC Spa**™, K m l A 00 Uart: O.OOOOOEtOOJ] ( Kwal 1 ] MOHE <= •
1/0=0.54
* x*c -0.58
f (/)
\
;tyy(**»V M Sv***! «/c -0.B6
POJJP « :>
f (3)
Tt
Oura
<- 1
Curs
1 ->
r i Hartfcopy F2 Plotter
o:r C
l*Wv>i**f*^*
frequeni 50.0000 naa_i : S.02844E-03
0.0 HUE_Zu63.rrT
;!MvWf>/,wV^V,txlc -°
Trigger 0 -1/16
Fig. 6 Frequency spectrum in the third stage for compressor operation near design point
:!S1)/^yVw^KW!»''-ii-97 P2/P1
40
Trigger t[s] 0.005 ' - i - J
Fig. 4(a) Real time signals and frequency spectra lor a moderately loaded operating point (^ = 0.73)
30 20 10 I I I t . , -M t * I
140
160
180
200
Fig. 7 Pressure ratio of the H PC for a period of three days before turbine shut down
0.005
5 f[kHi]
t [s] 0
Fig. 4(b) Real time signals and frequency spectra for a highly loaded operating point (
interface
FFTanalyzer
D
MUX
CE=£
T
8 amplifiers
Fig. 5
PC computer
\ tape recorder
host computer
Experimental setup for long-term measurements
Journal of Turbomachinery
the controlling computer of the gas turbine (host computer) were received. Changing operating conditions of the compressor influence the unsteady flow field in a blading that can be detected by the casing-mounted dynamic pressure sensors. The frequency spectra of these measured real time signals show different patterns. This effect was used for the automatic long-term investigation of this compressor. After the start of the gas turbine, the first frequency spectra of each sensor signal were stored as a reference pattern. When during operation differences occurred that exceeded a defined range, the magnetic tape was started by the computer and picked up the real time signals of all sensors. At the same time the current frequency spectra were stored as new reference patterns. By this procedure all changes in the unsteady flow field in the stages could be observed and saved. For normal compressor operation the frequency spectrum of each stage sensor shows the peak for rotor blade frequency of this stage (Fig. 6). But also the characteristic frequencies of the neighboring stages are visible (marked with numbers for the several stages). The decrease of the fluctuating amplitudes at higher frequencies is caused by the damping of the adapters in which the pressure sensors were mounted. The investigations were carried out over a time period of about four months at nearly constant operating conditions of the gas turbine. Figure 7 shows the pressure ratio of the compressor for a period of three days just around the occurrence of stall. The Xaxis shows the absolute number of measurement and the Y axis the pressure ratio of the whole compressure (p2 = outlet pressure, p\ = inlet pressure). At the end of this period a strong increase caused by a fault of the controlling system occurred. At the maximum pressure the surge control of the compressor became active and the gas turbine was shut down. JANUARY 1995, Vol. 117/83
Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
f(8l
f<7)
f(9,10) f i l l , 12,13) 11th stage 11th stage
'.00E-O3 _|
1
1
1
1
1
_,
1
,
,_
1.0CE-01
f(8) f(9,10) f(11.12,13)j
13th stage
1 3 t h stage
Fig. 8(a) Frequency spectra of the 11th and 13th stages for normal compressor operation Fig. 8(c) Frequency spectra of the 11th and 13th stages just after gas turbine shutdown
1.O0E4OO
f(9,10) L
(V)
1
f(8> /
III h n7Ki
f(11,12.13)
/
1/ VAJ - ^ yw K
1.00E-03 . ~
vlyA>
1 1 t h stage
— i — i — i — i — i — i — i — i — i —
1 3 t h stage
Fig. 8(b) Frequency spectra of the 11 th and 13th stages for compressor operation near the stability limit
Figures 8(«-c) show the frequency spectra of the 11th and 13th stage sensors for three different points of time around the shut down point. The patterns for normal operating conditions are demonstrated in Fig. 8(a). The peaks for the rotor blade frequency of the 11th to 13th stage with 76 rotor blades each are clearly visible. Also the 9th and 10th stage peaks (60 rotor blades each) stand out of the noise in the spectrum. Hereafter, the fault in the control system occurred and the operating point of the compressor drifted slowly toward the 84 / V o l . 117, JANUARY 1995
stability line. The spectra during these operating conditions show different patterns in the 11th and 13th stages (Fig. 8b). Whereas the periodic fluctuations in the 13th stage disappeared, in the 11th stage an increase of the characteristic frequency peak can be observed due to increased profile loads and wakes. Caused by the high pressure ratio of the compressor, the aerodynamic load of the 13th stage becomes so high, that lowenergy material from increased wakes or even separated profile boundary layers (approaching stall) is centrifugated from the rotor blades. It accumulates near the casing behind the rotor and forms here a casing stall region, which reduces the pressure rise in this stage. The pressure sensor mounted in the gap between rotor and stator is covered by this low-momentum material so that the information from the undisturbed flow outside of the stall is extinguished at the measuring position due to increased turbulence and noise. Due to the high aerodynamic load of the compressor and the reduced pressure rise in the last stages, the load of the upstream stages is increased. This causes a growth of the rotor wakes in these stages so that here the characteristic frequency peaks increase (see Fig. &(b), 11th stage). The multistage compressor runs under stable conditions although in the 13th stage the limit for stall as it was defined for a single stage (Inoue et al., 1991) is exceeded. After about a further 30 minutes the compressor load entered such a high level that surge occurred and the surge control system shut down the gas turbine. As the automatic measuring system had a cycle time of about 20 seconds from the start of the FFT analysis from the first to the last sensor the exact point of compressor surge was not recorded on the magnetic tape. The signals taken during this operation phase started the recorder a few moments after this event. Just after the shutdown the compressor load diminishes and the frequency information of several stages can be detected again in the 13th stage (Fig. 8c). Due to the deceleration of rotor speed, the characteristic frequencies occur at lower values. Transactions of the ASME
Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
pval
Alarm Ports
SP 1
SP 2
SP 3
Amplifiers
stall near the casing
time
Fig. 9 Change of amplitudes at characteristic frequencies with increasing blade load Fig. 10 Hardware setup for the on-line monitoring system
Design of an On-Line Monitoring System Based on these results, a monitoring system for detecting the approach of the stability line in a multistage compressor has been developed. The measurements showed that the peak level at the characteristic frequency of a stage is an indicator for the aerodynamic load. Up to the point of the collapse of the periodic fluctuations, the peak amplitude grows out of the normal noise level, which also rises with increasing aerodynamic load (Fig. 9). This behavior can be described in the following form:
very small compared to the increased noise level. Due to the lower pressure rise in this stage the loads of the upstream stages increase. Thus the amplitudes of the characteristic frequency peaks are enlarged. After the collapse of the observed frequency peak, the value of pval will decrease. Thus, although the compressor load is higher, a lower value for the stall level SL would be calculated with Eq. (2). In order to obtain an increasing trend with higher compressor load, the calculation of this parameter has to be switched. When in the 13th stage casing stall occurs, the maximum value of pval13 just before the collapse of the frequency Pfrestage-Pfre amplitude is stored and used for the further stall level calcuPostage — (1) lations. The same procedure will be applied if also in the next pfre upstream stage casing stall damps the periodic information at As the measuring results from the multistage compressor the sensor. When the operating conditions change back to demonstrated, the collapse of the periodic information at the unseparated flow caused by lower aerodynamic loads, the cal13th stage sensor is not the limit for the stable compressor culation switches back to the former way (Eq. (2)). operation. Therefore, additional information must be used to This algorithm was applied to the unsteady pressure measdefine another parameter for the observation. urements of the long-term investigation (Fig. 8). It is mentioned Measurements of unsteady wall pressures in three neigh- that in this case the stall level was only calculated with the boring stages provide information on the stall development measuring results of the 11th and 13th stages. For the last from the last to the upstream stages. Therefore, dynamic pres- measurements just before the shutdown (see Fig. 7) with insure transducers are located in the axial gaps between rotor creasing pressure ratio of the compressor, the stall level exceeds and stator of the 11th, 12th, and 13th stages. a value of 15. During the whole investigation period of four In order to get an informative value for the operator or an months at normal compressor operation, values up to 10 were operating and control system from the measuring results in calculated for the stall level. each stage, a new parameter was defined for the monitoring This procedure of calculating a monitoring parameter for of the compressor load. This value (called stall level SL) takes the compressor load described above was the foundation for into account the amplitudes of the frequency peaks in the three the design of an on-line monitoring system on the basis of a observed stages and the information about flow separation. PC computer with 80386-CPU and signal processor boards for Since with increasing compressor load the amplitudes in the parallel signal acquisition and enhancement (Fig. 10). The unstages grow, this parameter can be defined in the following steady pressure signals are Fourier analyzed by the signal proform: cessors. The CPU coordinates the tasks of the processors, calculates the stall level from the frequency spectra, and sends SL = an pvali3 + a12 pval,2 + «ii pvaln (2) the results to a host computer where the data are displayed on wherein au, o12, and ctn are weighting coefficients for the a monitor and stored for a trend analysis. influence of the stages on the stability limit of the compressor. The optimal choice of these coefficients depends upon the compressor operating conditions and geometry. They are still Experiences With the Monitoring System in Operation the subject of further investigation and are proprietary. This system was applied to the high-pressure compressors As previously mentioned (see Fig. 8) for the shutdown of of two gas turbines of the type LM5000. In order to get an the compressor, the 11th stage was not yet separated at the impression of the reliability of the monitoring system, the stall casing measuring position. Therefore, for this compressor, level is compared with the operating parameters of the comterms of peak values (pval) from further upstream stages were pressor measured by the normal control system of the gas not included. For other part-load operating points the station turbines. The trends of the speed of the high-pressure rotor of measurements and the terms of the stages to be included in and the pressure ratio of the HPC show a satisfying relation the stall level value (SL) must be selected correspondingly. to the observed stall level (Figs. 11« and b). When casing stall in the 13th stage occurs the periodic flucWith increasing pressure ratio, the stall level also grows up tuations at the sensor are damped so that the observed char- to values of about 7. After a while it decreases slowly although acteristic peak in the frequency spectrum disappears or becomes the pressure ratio keeps constant. The comparison with the Journal of Turbomachinery
JANUARY 1995, Vol. 117/85
Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
1991). The mixing of tip leakage vortex with low-momentum material centrifuged from the rotor blades play a nonnegligible role in the development of casing stall. This causes different limit values of stall level with changing geometries and operating conditions. After longer operation time of the monitoring system more measuring data for different operating conditions of the compressor will be available. With these experiences the model for the definition of the limit value can take into account more influence parameters.
09.09.1992 ( 7.00 h - 10.00 h )
80 120 time [min] Fig. 11(a) Stall level over a time period of three hours
200
09.09.1992 1 7.00 h - 10.00 h I
njty/<\l>Y~v) 2
pressure ratio
rh
21
I*
t
VSV angle 19
80
120
Summary and Conclusions Based on experimental results of unsteady flow in a single and a multistage compressor a monitoring system for stable compressor operation has been developed. By detection of aerodynamic load and casing stall in three neighboring end stages of a multistage high-pressure compressor, the approach to the stability line can be observed. Experiences with the monitoring system at two compressors demonstrate the reliability of the method. Changes in compressor load cause differences in calculated stall level. Additional measuring results from system operation will allow a more exact definition of various influence parameters. By the detection of casing stall in the end stages only a stall monitoring near-design speed of the compressor can be realized at the present state. The system will be able to obtain monitoring data for lower speeds if measurements also will be taken in the front stages. Furthermore, additional measurements in different compressors can help to obtain further information about the influence of various geometries and loads.
200
time [min]
Fig. 11(6) HPC pressure ratio and VSV angle over a time priod of three hours
Acknowledgments The authors wish to acknowledge the DOW Stade GmbH for the financing of the investigations leading to this research work. Especially, we wish to thank Mr. H. Walter and Mr. E. Mizera for their support in preparing the experiments and the helpful comments concerning the application of the monitoring system. The discussions about the measuring results during the test phase are gratefully acknowledged. A patent application that relates to the technology covered by this work has been filed by DOW Deutschland Inc.
VSV angle (Fig. lib) shows the reason for this behavior. At the same time when the stall level decreases, the VSV angle moves from an open position toward 0 deg so that the air flow through the compressor decreases. This lower aerodynamic load causes the decrease in stall level. In order to get information about the vicinity to the stability line of the compressor, a limit value for the stall level has to be defined. The results from the long-term investigation and the experience with the monitoring system until now allow an References J. J., Celestina, M. L., and Greitzer, E. M., 1993, "The Role of estimation of this limit. The measurements and trends have TipAdamczyk, Clearance in High-Speed Fan Stall," ASME JOURNAL OF TURBOMACHINERY, shown that the 13th stage of the investigated compressor can Vol. 115, pp. 28-39. operate stably with casing stall without the compressor exDay, I. J., 1993, "Stall Inception in Axial Flow Compressors," ASME JOURceeding the stability line. The stall region covers only a part NAL OF TURBOMACHINERY, Vol. 115, pp. 1-9. Gallus, H. E., Lambertz, J., and Wallmann, Th., 1979, "Blade-Row Interof the blade height so that stable compressor operation is still action in an Axial-Flow Subsonic Compressor Stage," ASME Paper No. 79possible. Assuming that with occurrence of casing stall also in GT-92. the 12th stage the stall region in the 13 th stage covers a large Gallus, H. E.,andH6nen, H., 1986, "Experimental Investigations of Airfoilpart of the span, this point would be the limit of stable com- and Endwall Boundary Layers in a Subsonic Compressor Stage," ASME Paper pressor operation. Thus, if the maximum of pval12 is known, No. 86-GT-143. V. H., Epstein, A. H., and Greitzer, E. M., 1991, "Rotating Waves the limit value of stall level can be calculated with Eq. (2). As as Gamier, a Stall Inception Indication in Axial Compressors," ASME JOURNAL OF the three observed stages have the same blade numbers and TURBOMACHINERY, Vol. 113, pp. 290-302. Inoue, M., Kuroumaru, M., Iwamoto, T., and Ando, Y., 1991, "Detection geometries in each rotor and each stator row, the patterns of periodic fluctuations should look quite similar. So, the max- of a Rotating Stall Precursor in Isolated Axial Flow Compressor Rotors," ASME OF TURBOMACHINERY, Vol. 113, pp. 281-289. imum value of pvalu will be in the same range as the maximum JOURNAL Paduano, J. D., Epstein, A. H., Valavani, L., Longley, J. P., Greitzer, E. of pvalo just before occurrence of casing stall in the 13th stage. M., and Guenette, G. R., 1993, "Active Control of Rotating Stall in a LowBy inserting these values into Eq. (2) the stall limit is calculated. Speed Axial Compressor," ASME JOURNAL OF TURBOMACHINERY, Vol. 115, pp. The value of the stall limit is influenced by many geometric 48-56. Poensgen, Ch., 1991, "Experimentelle Untersuchung der Stromung in einer parameters of the bladings. One important factor is the tip Unterschall-Axialverdichterstufe bei hoher Drosselung und im Rotating Stall," clearance of the rotor (Adamczyk et al., 1993; Inoue et al., Dissertation, RWTH Aachen, Federal Republic of Germany.
8 6 / V o l . 117, JANUARY 1995
Transactions of the ASME
Downloaded 08 Mar 2009 to 194.225.236.227. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm
