DGZfP-Proceedings BB 90-CD Lecture 37 EWGAE 2004

MONITORING OF GAS TURBINE OPERATING PARAMETERS

USING ACOUSTIC EMISSION

R M Douglas, S Beugné, M D Jenkins, A K Frances¶, J A Steel, R L Reuben and P A Kew.

School of Engineering and Physical Sciences, Heriot-Watt University, Edinburgh, UK.

¶
Imes Ltd., Bridge of Don, Aberdeen, UK.

ABSTRACT
In this work, Acoustic Emission (AE) sensors were mounted on several parts of a laboratory-scale
gas turbine operating under various conditions, the object being to assess the value of AE for in-
service condition monitoring. The turbine unit comprised a gas generator (compressor and turbine
on a common shaft) and a free-power turbine for power extraction. AE was acquired from several
sensor positions on the external surfaces of the equipment over a range of gas generator running
speeds. Relationships between parameters derived from the acquired AE signals and the running
conditions are discussed. It is shown that the compressor impeller blade passing frequency is
discernible in the AE record, allowing shaft speed to be obtained, and presenting a significant blade
monitoring opportunity. Further studies permit a trend to be established between the energy
contained in the AE signal and the turbine running speed. In order to study the effects of damaged
rotor blades a fault was simulated in opposing blades of the free-power turbine and run again under
the previous conditions. Also, the effect of an additional AE source, occurring due to abnormal
operation in the gas generator area (likely rubbing), is shown to produce deviations from that
expected during normal operation. The findings suggest that many aspects of the machine
condition can be monitored.

INTRODUCTION
Gas turbines are a versatile, cost-effective source of electricity, mechanical power, and propulsion.
However, they also incur relatively high maintenance costs and require frequent in-service
inspection. Unscheduled downtime of a turbine can cause vast economic losses and costly time
delays. Therefore, the ability to detect abnormalities within turbines, such as blade tip degradation,
fouling or fracture, before catastrophic failure occurs is obviously an enormous advantage.
Established condition monitoring techniques for turbines generally involve vibration, oil-debris,
temperature and pressure monitoring. Whilst these are effective, they are indirect measurement
methods which often require significant fault progression before detection is possible. The use of
AE to provide earlier detection and classification of rotating and reciprocating machinery faults
than the aforementioned conventional techniques has been well documented [1, 2, 3] and there is no
reason to believe this will prove any different for turbines. Before such advantages can be realised,
however, it is necessary to understand how AE is generated during turbine operation and hence how
faults might manifest themselves in AE records.

Armor et al [4] have already reported on the potential use of AE for detection of turbine shaft
cracks, blade rubs and bearing deterioration within in-service steam turbines, all of which sources
are present in this type of machinery. Sato [5] has proposed a diagnostic system that can detect and
distinguish between several different fault types for a steam turbine, including rotor-stator rubbing
and bearing faults, from AE acquired at the bearing housings. This was further ratified by Board
[6] who applied stress wave analysis (essentially AE analysis) to gas turbine engines and reported
that differences in the AE, both in time- and frequency- domains, permitted detection of faults such
as in-service seal wear, blade rubbings and induced bearing faults. Bates and Webster [7] detailed a

389
DGZfP-Proceedings BB 90-CD Lecture 37 EWGAE 2004

successful application of AE whereby the wear of an abradable seal lining during the ‘run-in’ period
of a gas turbine engine for aircraft propulsion was monitored and evaluated. Mba et al have
investigated the use of AE for monitoring power generation steam turbines. They verified that a
source originating at the shaft-seal tip can be detected by an AE transducer mounted on the bearing
housing [8]. They also demonstrated that rubbing can be diagnosed and classified from in-service
turbine machinery through AE monitoring [9, 10]. Recently, they have observed a relationship
between AE activity and load on a steam turbine unit, and have indicated that a rotor passing
through its critical speed can be identified [11].

To date, no published work has related to the use of AE to monitor the combustion and fluid
dynamical aspects of turbine operation due principally to the difficulty and impracticability of
making observations of faults in large industrial units. In this work we investigate AE acquired
from several sensor positions on a vastly smaller scale turbine unit of the order of 4 kW maximum
output, although the same combustion, fluid and mechanical principles apply. One aim was to
examine if information relating to the normal blade operating conditions, not just abnormal blade
rubbing as most previous work has reported, was transmittable through AE. Further aims were to
study changes in the AE signature arising from different turbine rotor speeds and from operation
with abnormal, induced and non-induced, running conditions.

EXPERIMENTAL APPARATUS
The experimental rig was a laboratory scale Cussons P.9003 two-shaft gas turbine unit comprising a
centrifugal compressor and radial turbine on a common shaft, which, together with a propane-
fuelled combustion chamber, forms the gas generator. A further radial design, free-power turbine
extracts energy from the gas stream and is loaded by a dynamometer which absorbs all power
output. The turbine unit is shown in Figure 1 and is illustrated schematically in Figure 2. The AE
transducer used throughout these tests was a Physical Acoustics Micro 80D band pass filtered
between 0.1 and 1 MHz with data acquired via a CompuScope 512 data acquisition card.

Combustion
Chamber B

C Free-
Power
turbine

Gas
generator
turbine
Gas generator A
Compressor
Figure 1: Main features of gas turbine showing sensor positions A, B and C

390
DGZfP-Proceedings BB 90-CD Lecture 37 EWGAE 2004

A test run typically consisted of acquiring data over a range of incrementally increased gas
generator shaft speeds, at a sample rate of either 1 or 5 MHz (corresponding to a time period of
approximately 0.1 and 0.5 seconds respectively). The limiting factor on maximum attainable gas
generator shaft speed was the maximum fuel flow-rate available at a particular time.

To alleviate the problem of long-term degradation of the AE transducer by high-surface

temperatures, a waveguide was welded to the exhaust of the free-power turbine. This produced a
surface temperature at the sensor mounting plate of the waveguide well under the specified
maximum operating temperature of the AE transducer, in this case 177 oC.

Initial tests were performed using three sensor positions, identified in Figures 1 and 2, with normal
operating conditions over a range of operating speeds from 50,000 to 70,000 RPM at the gas
generator stage. Other, possibly more revealing, sensor positions, such as the turbine casings and
backplates, were discounted due to the high surface temperatures and inadvisability of welding on
waveguides. During these initial tests the gas generator shaft was fitted with a sinusoidal Hall-
effect timing signal for accurate measurement of shaft speed, and additionally, an estimate of
rotational speed could be made via a dial-gauge tachometer. Later tests used only the waveguide
sensor position with simultaneous acquisition of a shaft-encoding signal from the free-power
turbine which enabled the shaft speed to be accurately determined.

A: Compressor B: Combustion chamber lid

impeller backplate
C: Top of
waveguide

Figure 2: Schematic of turbine showing sensor positions A, B and C.

Blade damage was simulated in two opposing blades of the 10-blade free-power turbine impeller by
rounding the sharp corner off first to a radius of 4mm and then running the turbine again under the
previous conditions. Thereafter, 4mm incremental increases of the radius were applied, an example
of this induced damage being shown in Figure 3. The impeller itself had a maximum diameter of
approximately 90mm. The gas generator compressor impeller, shown in Figure 4, was of different
construction, consisting of 6 inner and 12 outer blades.

391
DGZfP-Proceedings BB 90-CD Lecture 37 EWGAE 2004

Rounded edge,
8mm radius

Figure 3: Free-power turbine impeller, Figure 4: Gas generator compressor

damaged blade shown. impeller.

A further fault examined was that of insufficient fuel supply. Here, the fuel supply was lowered
beyond that required for idle, at which point the combustion became sporadic and unstable. Any
system developed for the purpose of turbine monitoring would require a function whereby this fault
could be diagnosed and fuel settings adjusted accordingly.

RESULTS AND ANALYSIS

Over the course of the testing two different types of AE record were observed from the waveguide,
sensor position C, and these were:

i) An apparently consistent background noise level (see Figure 5a).

ii) AE consisting of a background level with higher amplitude random pulses (see
Figure 5b).

Figure 5a: Consistent background level Figure 5b: Additional events

Blade-passing frequency from AE

From the tests where the AE sensor was mounted on the combustion chamber lid, sensor position B,
it has been possible to identify the blade passing frequency of the gas generator compressor
impeller. Figure 6a shows the spectrum of the timing signal from the compressor/turbine shaft for
three different speed settings, estimated at 50,000, 55,000 and 60,000 RPM from the dial-gauge.
The actual shaft speeds are slightly less than indicated by the dial-gauge readings but this is
inconsequential to this analysis. Focusing in on the modulated region of the spectra from the
corresponding AE signals acquired from the combustion chamber lid, see Figure 6b, reveals a peak

392
DGZfP-Proceedings BB 90-CD Lecture 37 EWGAE 2004

for each record at around 12 times the shaft speeds identified in Figure 6a. The gas generator
compressor impeller consists of 12 outer blades and therefore it is not unreasonable to conclude that
the turbulent interaction of the airflow and each of the impeller blades is represented in the AE.
This feature was also evident, and somewhat clearer, from data acquired at the backplate of the gas-
generator compressor impeller, sensor position A. It should be noted that a plain bearing supports
the gas generator common shaft and that the AE records were of the type shown in Figure 5a.

Figure 6a: Frequency spectrum of gas Figure 6b: Corresponding spectrum of AE

generator shaft timing signal from combustion chamber lid

Effect of turbine speed on the AE energy

Figure 7 shows the results from 3 separate test runs in which the relationship between free-power
turbine shaft speed and AE energy per revolution of the free-power turbine was investigated (the
range of gas generator shaft speeds varied from 50,000 to 67,500 RPM). Data were acquired from
the waveguide, sensor position C, and were predominately of the type described in Figure 5a,
except for a few records which showed isolated bursts of the type seen in Figure 5b. These
occasional events account for the outlying points seen in Figure 7 and also a few additional points
that are not shown in Figure 7 up to a value of 150.

Figure 7: Increasing AE energy content with increasing shaft speed

393
DGZfP-Proceedings BB 90-CD Lecture 37 EWGAE 2004

It is clear that there is a relationship between the shaft speed/power output and the AE energy per
revolution of the free power turbine shaft. A similar relationship could have been constructed with
the energy per revolution of the gas generator shaft since, in these tests, the free-power turbine
speed was dependent upon the gas generator shaft speed. Indeed, the background AE energy may
be attributable to a number of sources from both turbines, notably fluid flow; exhaust gas and blade
turbulence, and also bearing-related sources. All these mechanical or fluid events would be
expected to increase in intensity with increasing speed/power output of the turbine unit and hence
the trend seen in Figure 6 for the background type AE records can be considered to be that produced
by normal operation of the turbine unit.

The results of a further test, test 4, are shown in Figure 8 together with the results previously given
in Figure 7, for tests 1, 2 and 3. The data in this case is presented as AE energy in the time-domain
per revolution of gas generator shaft against estimated gas generator speed. During this test, an
audible noise was heard to originate from the gas generator area and the AE acquired was of the
nature exhibited in Figure 5b. The results indicate a similar trend to that established in Figure 7 but
with increased AE energy levels. The precise mechanical or fluid origin of this increased AE is
unknown at present, but is almost certainly related to the audible noise from the gas generator area
and indicative of rubbing, either within the bearing or between the turbine or compressor impellers
and their respective housings.

Figure 8: Effect of further AE source from gas generator turbine unit,

indicative of abnormal running conditions

Detection of damaged blades in free-power turbine

Blade tip degradation was simulated by applying a radius to the originally sharp edges of two
opposing blades of the free-power turbine impeller (see Figure 3a). The results of running the
turbine over a range of gas generator speeds where 8 and 12 mm radii were applied are shown in
Figure 9, AE observed during these tests being of the type described in Figure 5a. The reduced
levels of AE energy associated with the larger fault radius may be due to the fact that with a larger
gas-leakage area, i.e. the 12mm radius, the gas-leakage velocities, assuming a constant flow-rate,
will be reduced and hence less AE generated.

394
DGZfP-Proceedings BB 90-CD Lecture 37 EWGAE 2004

Figure 9: Effect of simulated damage to blades of free-power turbine

Insufficient fuel supply / Unstable running

Figure 10 shows AE acquired from the backplate of the gas generator compressor during a period of
sporadic, unstable running, i.e. the fuel supply was reduced to below that required for consistent
idling. The increased AE levels and character of the events are indicative of impact-type events
which are most likely due to shock forces transmitted to the gas generator turbine unit through the
reverberations caused by the on/off nature of the combustion.

Figure 10: Insufficient fuel supply / unstable running

CONCLUSIONS
This work is at a preliminary stage, but has demonstrated that the operating parameters of a
laboratory-scale turbine unit can be monitored by AE. The identification of the gas generator
compressor blade-passing frequency via AE represents a significant monitoring opportunity.
Deviations from the expected frequency or a dispersion of the frequency over a wider band may be
an indication of individual blade malfunction. More revealing AE regarding the turbine impellers is
potentially available from the external surfaces of the turbine casings and backplates, although a
means needs to be found to interrogate these hot surfaces without damaging the sensor.

Data acquired from the waveguide permitted a trend to be identified whereby the turbine running
speed is represented in the energy content of the background AE. Moreover, the presence of
abnormal operation of the gas generator turbine unit is readily distinguishable in the AE from that

395
DGZfP-Proceedings BB 90-CD Lecture 37 EWGAE 2004

of normal operation. The effects of two degrees of induced fault within the free-power turbine have
also been shown.

Recent efforts by other authors have verified that similar characteristics to that found in this work
are also evident on full-scale power-generation turbines. Hence the results presented here,
particularly the new findings regarding blade-passing frequency, have the potential to be scaled up
to full size turbine units. Furthermore, this illustrates that a laboratory scale turbine unit, where
operating conditions are easily altered and simulated faults more readily introduced, has the
capability to assist in the development of an AE-based condition monitoring system for power-
generation and propulsion turbines.

REFERENCES
[1] Neill, G. D., Reuben, R. L., Sandford, P. M., Brown, E. R. and Steel, J. A. Detection of
incipient cavitation in pumps using acoustic emission. Proc. IMechE: Part E, vol. 211 (4),
1997, 267-277.
[2] Gill, J. D., Reuben, R. L., Steel, J. A., Scaife, M. W. and Asquith, J. A study of small HSDI
diesel engine fuel injection equipment faults using acoustic emission. Journal of Acoustic
Emission, vol. 18, 2000, 211-216.
[3] Li, C. J. and Li, S. Y. Acoustic emission analysis for bearing condition monitoring. Wear,
vol. 185, 1995, 67-74.
[4] Armor, A. F., Graham, L. J. and Frank, R. L. Acoustic emission monitoring of steam
turbines. Joint ASME/IEEE Power Generation Conference, Oct 4-8, 1981, St. Louis, USA.
[5] Sato, I, Rotating machinery diagnosis with acoustic emission techniques. Electrical
Engineering in Japan, vol. 100 (2), 1990, 115-127.
[6] Board, D. B. Stress wave analysis of turbine engine faults. Proc. IEEE Aerospace
Conference, vol. 6, 2000, 79-95.
[7] Bates, J. and Webster, J. Engine development application for monitoring turbine seals in a
Rolls-Royce RB211-524 G/H gas turbine, in The Acoustic Emission & Ultrasonics
Monitoring Handbook. Editor: Holroyd, T. J., Coxmoor, 1st Edition, 2000, 91-93.
[8] Mba, D. and Hall L. D. The transmission of acoustic emission across large-scale turbine
rotors. NDT&E International, vol. 35 (8), 2002, 529-539.
[9] Hall, L. D. and Mba, D. Diagnosis of continuous rotor-stator rubbing in large scale turbine
units using acoustic emissions. Ultrasonics, vol. 41 (9), 2004, 765-783.
[10] Hall, L. D. and Mba, D. Acoustic emissions diagnosis of rotor-stator rubs using the KS
statistic. Mechanical Systems and Signal Processing, vol 18 (4), 2004, 849-868.
[11] Zuluaga-Giraldo, C., Mba, D. and Smart, M. Acoustic emission during run-up and run-
down of a power generation turbine. Tribology International, vol. 37 (5), 2004, 415-422.

396