BLADE AND SHAFT CRACK DETECTION USING TORSIONAL

BLADE AND SHAFT CRACK DETECTION USING
TORSIONAL VIBRATION MEASUREMENTS
PART 3: FIELD APPLICATION DEMONSTRATIONS
Ken Maynard 1 and Martin Trethewey 2
1
2
Applied Research Laboratory, The Pennsylvania State University
Dept. of Mechanical Engineering, The Pennsylvania State University
State College, PA 16801, USA
Abstract: The primary goal of the this paper is to summarize field demonstrations of the feasibility of
detecting changes in blade and shaft natural frequencies (such as those associated with a blade or shaft
crack) on operating machinery using non-contact, non- intrusive measurement methods. Laboratory
demonstration of feasibility and some special signal processing issues were addressed in Parts 1 and 2
[1, 2]. Part 3 primarily addresses the results of application of this non-intrusive torsional vibration
sensing to: a large wind tunnel fan; a jet engine high-pressure disk; a hydro station turbine; and to a
large coal- fired power plant induced-draft (ID) fan motors. During the operation of rotating equipment,
torsional natural frequencies are excited by turbulence, friction, and other random forces. Laboratory
testing was conducted to affirm the potential of this method for diagnostics and prognostics of blade and
shafting systems. Field installation at the NASA Ames National Full-Scale Aerodynamic Facility
(NFAC) reaffirmed the ability to detect both shaft and blade modes. Installation on a high-pressure
(HP) disk in a jet engine test cell at General Electric Aircraft Engines demonstrated that the fundamental
mode of the turbine blades was clearly visible during operation. Field installation at a hydro power
station demonstrated that the first few shaft natural frequencies were visible, and correlated well with
finite element results. Finally, field installation on the ID fan motors also showed the first few shaft
torsional modes. These field tests have resulted in high confidence in the feasibility of the application of
this technique for diagnosing and tracking shaft and blade cracks in operating machinery.
BACKGROUND
The detection of blade and shaft natural frequencies in the torsional domain requires that the signal
resulting from excitation of the rotating elements by turbulence and other random processes is
measurable. If measurable, these natural frequencies may be tracked to determine any shifting due to
cracking or other phenomena affecting torsional natural frequencies. Difficulties associated with
harvesting the potentially very small signals associated with blade and shaft vibration in the torsional
domain could render detection infeasible. Thus, transduction and data acquisition must be optimized for
dynamic range and signal to noise ratio [1, 4, 5]. An overview of the laboratory results may be found in
[1, 2].
The advantage of using shaft torsional natural frequency tracking over shaft lateral natural frequency
tracking for detecting cracks in direct-drive machine shafts is twofold:
•
A shift in natural frequency for a lateral mode may be caused by anything which changes
the boundary conditions between the rotating and stationary elements: seal rubs, changes in
bearing film stiffness due to small temperature changes, thermal growth, misalignment, etc.
So, if a shaft experiences a shift in lateral natural frequency, it would be difficult to
pinpoint the cause as a cracked shaft. However, none of these boundary conditions
influence the torsional natural frequencies. So, one may say that a shift in natural
frequency in a torsional mode of the shaft must involve changes in the rotating element
itself, such a crack, or perhaps a coupling degradation.
•
Similarly, finite element modeling of the rotor is simplified when analyzing fo r torsional
natural frequencies. The aforementioned boundary conditions, which are so difficult to
characterize in rotor translational modes, are near non-existent in the torsional domain for
many rotor systems. This means that characterization of the torsional rotordyanamics is
more straightforward, and therefore likely to better facilitate diagnostics.
Detection of the small torsional vibration signals associated with blade and shaft natural
frequencies is complicated by transducer imperfections and by machine speed changes. The use
of resampling methods has been shown to facilitate the detection of the sha ft natural frequencies
by: (1) correcting for torsional transduction difficulties [2, 4] resulting from harmonic tape
imperfections (printing error and overlap error); and (2) correcting errors as the machine
undergoes gradual speed fluctuation [2, 5, 6]. In addition, correction for more dramatic speed
changes was addressed in [6]. These corrections made laboratory testing quite feasible.
Transducer setup and methodology
The transducer used to detect the torsional vibration of the shaft included a shaft encoded with
black and white stripes, an infrared fiber optic probe, an analog incremental demodulator or
digital demodulator and an A/D converter. The implementation of the technique under
laboratory conditions was previously presented in [4, 5]. Figure 1 shows a sche matic of the
transducer system using analog demodulation.
Fiber optic cable
Fiber optic probe
Equally spaced black and
white stripe code
A-D converter
Power Supply
9.12
Analog
demodulator
Shaft experiencing
torsional vibration
Figure 1: Schematic of transducer setup for torsional vibration measurement
Figure 2 shows a spectrum of the torsional vibration of a desktop rotor, with the natural
frequency of the blade cluster and of the single detuned blade (“rogue blade”) evident [5, 6].
-40
-50
Cluster of
Blade
Modes
-60
Rogue
Blade
-70
First Shaft
Torsional
Mode
-80
-90
-100
0
50
100
150
200
250
300
350
Figure 2: Torsional Spectrum of Laboratory Rotor with One Detuned Blade
IMPLEMENTATION FOR BLADE CRACKING
NASA Ames National Full-Scale Aerodynamic Complex (NFAC) Fans
The NFAC facility is the largest wind tunnel in the world, able to test a full-scale 737 in the
largest section (see Figure 3). The fans used to drive the wind tunnel, shown in Figure 4, have
experienced some blade cracking in the past, and the repairs made continued inspection
unpractical. NASA decided to use modal impact testing of the blades to track the natural
frequencies of the blades to detect shifts that might be associated with cracking. However, it was
determined that the frequency might shift as much as ten percent simply by rotating the rotor
360o and retesting the same blade. Since this was on the order of the expected shift due to
cracks, other methods were considered. A feasibility study was conducted using torsional
vibration on an operating fan.
Figure 3: National Full-Scale Aerodynamic Complex (NFAC) at NASA Ames
Figure 4: NFAC Fans
The shaft was encoded with zebra tape, as shown in Figure 5, and testing was performed during
operation of the fans.
Figure 5: Zebra Tape Installation on the NFAC Fan Shaft
The results are shown in Figure 6. The first two peaks correspond to overall shaft torsional
frequencies, based on simple dynamic models developed by the vendor in the 70s. The third
peak, near 13.5 Hz, corresponds to the blade group. Of particular interest is the tight packing of
the fifteen blade modes, especially knowing that the impact testing on a single blade could vary
by 1 Hz or more during a single test session. We believe that the variable pitch mechanism
(VPM) does not provide a repeatable boundary support for the blades when the fan is not
operating, but provides uniform and repeatable results during operation, when the VPM is
preloaded.
0.0001
Volts
Shaft Torsional
Natural
Frequencies
Blade Natural
Frequencies
0.00001
0.000001
0
2
4
6
8
10
12
14
16
Frequency (Hz)
Figure 6: Averaged Torsional Spectrum of NFAC Fan
Subsequent to this testing, impact testing was performed on a blade with the oil pump operating
to attempt to allow torsional coupling. The resulting natural frequency was found to be about
14.6 Hz, as shown in Figure 7.
This testing demonstrated the ability of this measurement system to detect blade natural
frequencies during operation of the NFAC fans.
100
H(f)
14.609
10
1
0.1
0
10
20
30
40
50
60
Frequency (Hz)
Figure 7: Results of Modal Testing of Single Blade
Application to aging vehicular turbines
In vehicular turbines, the loss of turbine blades has resulted in accidents and fatalities. With the
aging of commercial and military fleets, fatigue cracking and failure of discs and blades becomes
a more pressing issue. For example, some of the engines currently used in about four thousand
737, A319, A320, A321, A340, and DC8-70 commercial aircraft are approaching twenty years of
age. Although these engines experienced little blade and disk cracking early in their lives,
cracking has been detected during regularly scheduled inspections on aging engines. There is
some concern that this cracking may accelerate as the engines age, and may require increased
inspections. The possibility of a crack precipitating and resulting in blade or disk failure in
between scheduled inspections increases with age, and an in situ system could provide the
user/maintainer with blade/disk health information.
Similarly, military aircraft engines are exhibiting symptoms of aging, which include blade and
disk failure. From 1982 to 1998, 55% of USAF engine caused mishaps were related to highcycle fatigue (HCF) of engine parts [11], predominately blades and disks at the blade attachment
[12]. In addition, about 87% of the risk management inspections were related to this HCF.
Thus, a large portion of the maintenance budget is associated with HCF of blades and disks at the
blade attachment.
Instrumentation of an aircraft jet turbine
The HP turbine of a commercial jet engine was instrumented to determine the feasibility of
detecting the blade natural frequencies during operation using torsional vibration measurement.
The testing was performed in a warm air test facility, under load at about 9400 RPM. Figure 8
shows the fully assembled commercial engine and the HP disk as installed in the warm air test
facility.
Figure 8: a) Fully Assembled Jet Engine; b) HP Disk in Warm Air Test Facility
Several individual blades were tested by the engine manufacturer to determine the natural
frequencies. These blades were fixtured to simulate the boundary conditions during operation.
Figure 9 shows a typical spectrum of the results of the modal testing, which is summarized for
the three specimen blades in Table 1.
Figure 9: Typical Spectrum of Individual Fixtured Blade
Blade F12
Blade C14
Blade B7
Average
Std. Dev.
Mode 1
2450
2413
2434
2432
19
Mode 2
3943
3853
3875
3890
47
Mode 3
6038
6091
5994
6041
49
Mode 4
8409
8469
8364
8414
53
Table 1: Results of Modal Testing of Sample Blades
The instrumentation included a 200-stripe zebra tape and fiber optic probes. In addition, the
output from a 60-tooth speed encoder was used as backup. The zebra tape did not survive the
test, and the speed signal from the 60-tooth gear was used.
Figure 10: a) Zebra Tape Installation; b) 60-Tooth Speed Gear
The use of the 60-tooth wheel limited the frequency range of the data to one- half the number of
teeth times the speed of the shaft, or about 4500 Hz. Figure 11 shows the torsional spectrum of
the HP turbine rotor. Note that torsional natural frequencies of the shafting system should
change very little with speed, and coupled blade and shaft torsional modes increase very slightly
with speed due to stiffening by axial force. So, we look for frequencies that are not shifting with
speed. Note tha t the first peak at about 2400 Hz is close to the first mode from the blade modal
testing.
Earlier experimental and analytical work, however, indicates that blade natural frequencies can
be significantly altered by coupling with shaft torsional modes [5,6]. It was shown that for a
small desktop rotor with much larger blade to rotor mass ratio the difference might be more than
30%, and may be higher or lower than rig impact testing. Although intuition might imply that
for relatively small blades coupling would be less of a factor, the effects of torsional coupling
should be clarified using a simple dynamic torsional model of the engine.
-10
9000 rpm, 60 tooth
-15
9450 rpm, 60 tooth
Blade F12, Mode 1
-20
Blade C14, Mode 1
-25
Blade B7, Mode 1
Gain
-30
-35
-40
-45
-50
-55
-60
2200
2400
2600
2800
3000
3200
3400
Frequency, Hz
3600
3800
4000
4200
Figure 11: Torsional Spectrum of HP Turbine
IMPLEMENTATION FOR SHAFT CRACKING
Application to a Hydro Turbine:
The hydro plant consists of five 3 MW electric turbine generators sets. The plant was originally
built in about 1910, but it has recently been redesigned to eliminate an underwater, wooden
(lignum vitae) bearing and improve efficiency. The layout of a unit is shown in Figure 12 and
Figure 13.
Figure 12: Hydro turbine-generator set
Figure 13: Disassembled Hydro-Turbine
In the last five years, three of the newly designed turbine rotors have experienced severe
cracking. Instrumentation and analysis was performed on one of the units that had not
experienced cracking to demonstrate the feasibility of detecting shaft natural frequencies. Figure
14 shows the optic probe, tachometer, and encoded tape placement.
Fiber Optic
Probe
Infrared
Tachometer
Encoded tape
Figure 14: Optic probe, encoded tape, and tachometer placement
The data was analyzed using the double resampling technique [3,6] to eliminate the adverse
effects of the presence of running speed and its harmonics on frequency identification. The
results of four test runs are shown in Figure 15. Note the peaks at about 16 Hz and 41 Hz. These
correspond well to the results from a simple finite element model (16 Hz and 40 Hz).
1.E-02
Run 4
Run 5
Run 7
Amplitude
Run 11
1.E-03
1.E-04
0
5
10
15
20
25
30
35
Frequency (Hz)
Figure 15: Torsional spectrum of hydro unit shaft motion
40
45
50
The frequencies below 5 Hz are somewhat enigmatic. Since the operating speed of the unit is
300 RPM, or 5 Hz, it was at first assumed that these frequencies correspond to fluid whirl, which
generally occurs at speeds between 0.42 and 0.48 times operating speed [10]. However, the shaft
lateral vibration data exhibited none of the signs of whirl. In addition, the three closely spaced
subsynchronous peaks were stable and repeatable from run to run, as seen in Figure 16. Such
stability and repeatability for three closely spaced frequencies does not correspond to the whirl
phenomenon. In addition, similar spectral components have since been observed on hydro units
at other sites. So, we hypothesize that these subsynchronous frequencies corresponds to the
“rigid body” torsional mode on torsional springs corresponding to the bearing film stiffness in
shear. Further investigation will be necessary to confirm this and to clarify the significance of
these spectral components.
1.E-02
Run 1
Run 2
Run 4
Run 5
Run 7
Run 9
Amplitude
Run 10
Run 11
1.E-03
1.E-04
0
1
2
3
4
5
Frequency (Hz)
Figure 16: Subsynchronous torsional spectrum for hydro unit shaft
Several issues arose during the on-site data acquisition and analysis. Figure 17 shows some of
the data of Figure 1 along with runs that had significant distortion due to tape errors. When the
tape was changed, or even the axial location of the transducer was changed on the same tape, the
spurious frequencies shifted. These spurious frequencies seem to be related to the encoded tape,
and often interfered with the identification of shaft natural frequencies.
1.E-01
Run 1
Run 2
Run 4
Run 5
Run 7
Run 9
1.E-02
Run 10
Amplitude
Run 11
1.E-03
1.E-04
0
5
10
15
20
25
30
35
40
45
50
Frequency (Hz)
Figure 17: Torsional spectra showing encoded tape error spectral content
Application to an Induced-Draft Fan Motor
The motors on the fossil- fired induced draft fan were constructed using rectangular cross section
webs from the shaft to the rotor coil supports. The square end on the web was then welded to the
circular shaft without machining to match the contours. The result has been a number of failures
of the motors due to failure of the web welds. Two of these motors were instrumented to detect
shaft natural frequencies and establish a baseline to track the changes that may be associa ted
with web weld failure. Figure 18 shows the fan motor.
Figure 18: ID fan motor: (a) Motor housing; (b) scaled with minivan
Installation of the tape was more difficult on the ID fan than on the hydro unit due to the shaft
size and the tight quarters. Figure 19 shows the installation of the transducer system.
Fiber Optic
Probe
Encoded tape
Infrared
Tachometer
Figure 19: Tape, fiber optic probe, and tachometer installation on ID fan
In addition, the butt joint misalignment of the ends of the tape appeared to be exacerbated by
thermal growth of the shaft. It was observed that a space between the ends appeared after heat
up of the unit. This underlap, in some cases, caused saturation and malfunction of the analog
demodulator. Figure 20 shows the results for one of the fans. The first mode appears to be about
10 Hz. Once again, it was observed that changing tapes or changing the shaft axial position of
the optical probe on the encode tape changed some of the spectral content above 20 Hz. It is
difficult to assess the remainder of the spectrum with high confidence due to the spectral content
of the tape. However, most likely the second and third modes are at about 16 Hz and 19 Hz.
1.E-02
Amplitude
Run 4
Run 5
1.E-03
1.E-04
0
5
10
15
20
25
30
35
40
Frequency (Hz)
Figure 20: ID Fan motor torsional spectrum
SUMMARY AND CONCLUSIONS
The techniques developed for detecting torsional natural frequencies in the laboratory were
implemented on a large wind tunnel fan which has experienced blade cracking, a jet engine
which has experience cracking in the disc near at blade root, a hydro-turbine shaft which has
experienced cracking, and a power plant induced draft fan motor shaft which has experienced
cracking. The goals of the implementation project were to demonstrate the feasibility of field
application, and to establish a baseline for each class of machine. The data acquired clearly
demonstrated the feasibility of field implementation, and established baseline natural frequencies
for specific blades and shafts.
However, interference from tape related spectral content was experienced. This interference was
not experienced in the laboratory due to shaft size, access, and environmental differences. It is
believed that this spectral content is associated not with tape printing error or overlap, but was
introduced by the installation. This may be overcome by using a more precisely fabricated
encoding device (such as the 60-tooth speed encoder at the jet engine test facility), or by
developing correction algorithms for the installation error.
Future work: Research to establish the size of detectable cracks in each class of machine is
needed. For instance, reactor coolant pumps (RCP) in nuclear power plants have experienced
severe cracking with no advance notice given by the crack detection instrumentation. Similarly,
although it is known that blade cracking in a jet engine would affect blade natural frequency, it is
unclear what effects are measurable in the torsional domain. In order to establish the torsional
vibration method as practical in field machines, it is important that we be able to estimate the
size of the crack at the detection threshold and the remaining useful life of the shaft/blade at that
detection threshold. In addition, correction of the installation errors must be accomplished to
remove ambiguity and make the technology widely accessible. This work is currently underway.
Acknowledgement : This work described herein was supported by the Southern Company
through the Cooperative Research Agreement Torsional Vibration and Shaft Twist Measurement
in Rotating Machinery (SCS Contract Number C-98-001172); NASA Ames Research Center
Grant Condition-Based Monitoring of Large-Scale Facilities, (Grant NAG 21182); by
Multidisciplinary University Research Initiative for Integrated Predictive Diagnostics (Grant
Number N00014-95-1-0461) sponsored by the Office of Naval Research; and by NASA
Glenn/GE Aircraft Engines Turbine Disk Crack Detection Using Torsional Vibration: A
Feasibility Study (GEAE Purchase Order 200-1X-14H45006). The content of the information
does not necessarily reflect the position or policy of the Government, and no official
endorsement should be inferred.
References:
1. Maynard, K. P., and Trethewey, M. W., Blade and Shaft Crack detection Using Torsional
Vibration Measurements Part 1: Feasibility Studies, Noise and Vibration worldwide, Volume
31, No. 11, December 2000, pp. 9-15.
2. Maynard, K. P., and Trethewey, M. W., Blade and Shaft Crack detection Using Torsional
Vibration Measurements Part 2: Resampling to Improve Effective Dynamic Range, Noise
and Vibration worldwide, Volume 32, No. 2, February 2001, pp. 23-26.
3. Vance, John M., Rotordynamics of Turbomachinery, John Wiley & Sons, New York, 1988,
pp. 377ff.
4. Maynard, K. P., and Trethewey, M., “On The Feasibility of Blade Crack Detection Through
Torsional Vibration Measurements,” Proceedings of the 53rd Meeting of the Society for
Machinery Failure Prevention Technology, Virginia Beach, Virginia, April 19-22, 1999, pp.
451-459.
5. Maynard, K. P.; Lebold, M.; Groover, C.; Trethewey, M., Application of Double Resampling
to Shaft Torsional Vibration Measurement for the Detection of Blade Natural Frequencies,
Proceedings of the 54th Meeting of the Society for Machinery Failure Prevention
Technology, Virginia Beach, VA, pp. 87-94.
6. Groover, Charles Leonard, “Signal Component Removal Applied to the Order Content in
Rotating Machinery,” Master of Science in Mechanical Engineering Thesis, Penn State
University, August 2000.
7. McDonald, D, and Gribler, M., “Digital Resampling: A Viable Alternative for Order Domain
Measurements of Rotating Machinery,” Proceedings of the 9th Annual International Modal
Analysis Conference, Part 2, April 15-18, 1991, Florence, Italy, pp. 1270-1275.
8. Potter, R., “A New Order Tracking Method for Rotating Machinery,” Sound and Vibration,
September 1990, pp. 30-35.
9. Hernandez, W., Paul, D., and Vosburgh, F, “On-Line Measurement and Tracking of Turbine
Torsional Vibration Resonances using a New Encoder Based Rotational Vibration Method
(RVM),” SAE Technical Paper 961306, Presented at the Aerospace Atlantic Conference,
Dayton, OH, May 22-23, 1996.
10. Sawyer, John W., Ed., Sawyer’s Turbomachinery Maintenance Handbook, 1st Ed., Vol. II,
Turbomachinery International Publications, 1980, p. 7-34ff.
11. Davenport, Otha B., “Operational Readiness and High Cycle Fatigue,” Turbine Engine
Symposium, Dayton Convention Center, 14-17 September 1998.
12. Davenport, Otha B., Senior Executive Service, director of engineering, Propulsion Product
Group, Aeronautical Systems Center, Air Force Materiel Command, Wright-Patterson Air
Force Base, personal conversation, 29 June 2001.