27.6 Conclusion

Back

Commercial fault detection systems developed for helicopter gearboxes, generally as part of a Health and

Usage Monitoring System (HUMS)3, follow the traditional approach to fault detection and diagnosis. In

these systems, the individual vibration signals obtained from each accelerometer are processed first to

extract various features, and are then evaluated statistically or in comparison with established thresholds

to identify possible abnormalities. The abnormal features are then scaled and weighted to provide a

weighted average “figure of merit” that can be used for fault detection. Once a gearbox is identified as

abnormal, diagnosis is performed by evaluating the synchronized time averaged features associated with

individual shafts and gears — the signal average of a particular shaft contains the mounted gear meshing

characteristics which can be used to determine the gear condition. In HUMS, statistical analysis is usually

based on the mean and standard deviation of individual features, and thresholds are established based

either on experience or according to base-lines determined from initial data points at the beginning of

aircraft operation. It should be noted that the simplistic approach used by major HUMS producers for

abnormality identification and fault detection and diagnosis is often due to the stringent restrictions

imposed by the certification process4; otherwise, much development work has been conducted in the

recent years that can be included in commercial systems.

A truly integrated mechanical diagnostic system should have the following characteristics: (1) be able

to automatically correlate the information from various sensor suites including accelerometers, oil debris

sensors, acoustic emission (AE) transducers, and so on, and to consolidate the results into a reliable

solution; (2) be able to “predict” (prognose) a degrading mechanical condition based on trending of the

“features” or condition indices; (3) be compatible with the maintenance policies and operational

situations, (4) account for the fatigue life of the monitored components; (5) account for flight conditions

(such as transmission torque) that would affect the diagnostic results; (6) be based on open systems

architecture (both hardware and software) for future technology insertion; and (7) be lightweight

(including cabling and sensors) such that no significant extra weight will be added to the aircraft.

Oil analysis will undoubtedly play a more critical role in fault detection and diagnosis. One logical step

will be development of more reliable in-flight methods of chip detection based on optical technology and

image processing (Lukas and Yurko, 1996). The chip detectors of the future are expected to benefit from

the micro-electro-mechanical systems (MEMS) technology, making available microdetectors that can be

3A HUMS generally consists of structural usage monitoring, gearbox, engine, and rotor system fault detection and

diagnosis, rotor track and balance, and the associated data processing and maintenance logistics.

4No HUMS has been certified by the FAA yet because of the relative newness of the technology and lack of data to verify it.

Some HUM systems have been flying in the North Sea area, mostly manufactured by Teledyne-Stewart Hughes and GECBristows,

but they need to be developed further before they can be totally reliable.

TABLE 27.12 Summary Comparison of Rankings Obtained from the

Monitoring Effectiveness Values and from the Diagnostic Results for

Suites of All sizes

Suites of Match Exactly Mismatch by

1 2 3

7 7 1 0 0

6 17 11 0 0

5 39 17 0 0

4 48 22 0 0

3 38 18 0 0

2 17 11 0 0

1 8 0 0 0

Total 174 80 0 0

Fault Diagnosis of Helicopter Gearboxes 27-23

© 2005 by Taylor & Francis Group, LLC

located practically anywhere within the gearbox. Continued emphasis is also expected on postflight oil

analysis such as ferrography, thermography, ultrasonic analysis, and passive electric current analysis

(Saba 1996).

As for vibration monitoring, there will be continued need for algorithmic development to produce

new features for reliable identification of component faults. In this domain, there will be more focus on

acoustic monitoring and stress wave analysis. Accelerometers normally have a frequency response range

of a few Hz to over 50 kHz. In acoustic monitoring, a much higher bandwidth is considered —

microphones are currently available that have a flat response to over 100 kHz. This provides the

possibility of measuring a cleaner signal in higher frequency ranges, which may prove useful in detecting

certain bearing faults. Another advantage of microphones is that they do not need to be mounted on the

surface of the housing, therefore, unlike accelerometers, they are not affected by mounting resonance. For

stress wave monitoring, AE sensors are available that detect the stress waves generated by strain (elastic)

energy spontaneously released by materials when they undergo deformation. This type of energy, which

has a very broad bandwidth (DC to several MHz) due to the impulse nature of the strain release,

propagates away from a crack tip and throughout the structure body. For crack growth monitoring,

analysis is usually focused on the high frequency region (. 100 kHz) of the AE signal so that the effect of

low frequency noise caused by airframe, gearbox, and/or drive shaft vibration is minimized. Stress wave

monitoring through AE signals may, therefore, reveal the presence of cracks, as well as their location and

severity (Teller and Kwun, 1994).

Another important area of research and development in fault diagnosis is sensor technology. A generic

restriction in vibration monitoring is the limited number of mounting locations for accelerometers on

the housing. This restriction, however, is expected to be alleviated with the advancement of the MEMS

technology which will eventually make available miniaturized sensors that can be mounted practically

anywhere on a housing, and perhaps inside it. These sensors are expected to have the added capability of

processing the vibration signal and producing the features locally, so that the central processor can be

dedicated solely to integration of features for fault detection and diagnosis.

Навигация