Structure-Based Fault Diagnosis of Helicopter Gearboxes
Structure-Based Connectionist Network (SBCN)
A diagnostic method is introduced for helicopter gearboxes that uses knowledge of the gearbox structure and characteristics of the features of vibration to define the influences of faults on features. To define the structural influences, the gearbox is represented by a lumped-mass model so that the root mean square value of vibration from this model is used to assign the influences. The structural influences are then converted to fuzzy variables, to account for the approximate nature of the lumped-mass model, and used as the weights of a connectionist network. Diagnosis in this Structure-Based Connectionist Network (SBCN) is performed by propagating the abnormal features obtained from an unsupervised pattern classifier ( Single Category-Based Classifer (SCBC)) through the weights of SBCN to obtain fault possibility values for each component in the gearbox. In part I of this paper, the proposed diagnostic method is described, and in part II the performance of the proposed method is experimentally evaluated by applying it to two helicopter gearboxes: OH-58A and S-61 .
Schematic of the OH-58A helicopter gearbox.
Single Category-Based Classifier
A new unsupervised pattern classifier is introduced for on-line detection of abnormality in features of vibration that are used for fault diagnosis of helicopter gearboxes. This classifier, which is implemented by a connectionist network, classifies vibration features by assigning them a value in [0, 1] to reflect their degree of abnormality. Its salient feature is that it performs classification by considering only the normal values of the features, and as such, it does not require feature values associated with faulty cases to identify abnormality. In order to cope with noise and changes in the operating conditions, an adaptation algorithm is incorporated that continually updates the normal values of the features. The proposed classifier is tested using experimental vibration features obtained from an OH-58A main rotor gearbox. In order to evaluate the performance of this classifier, the abnormality-scaled features are integrated for detection of faults. The fault detection results obtained from this classifier are comparable to those obtained from the leading unsupervised neural networks: Kohonen's Feature Mapping and Adaptive Resonance Theory (ART2).
Schematic of the S-61 helicopter gearbox
Application of Structure-Based Connectionist Network (SBCN)
The generic fault diagnostic system for gearboxes, Structure-Based Connectionist Network (SBCN), which integrates features of pattern classification and deep expert systems to eliminate the need for supervised training was evaluated experimentally using vibration data from two helicopter gearboxes: OH58A and S-61. The diagnostic results indicate that SBCN is able to diagnose a large number of faults in OH-58A and S-61 gearboxes, which is remarkable considering that the results are obtained independent of any supervised training. Also, to validate the structural influences, they were compared with those obtained from experimental RMS values and the weights of a network similar to SBCN trained through supervised learning.
- V. B. Jammu, K. Danai, and D. G. Lewicki, "Unsupervised Pattern Classifier for Abnormality-Scaling of Vibration Features for Helicopter Gearbox Diagnosis", Machine Vibration, 1996, Volume 5, No. 3, pp. 154-162.
- V. B. Jammu, K. Danai, and D. G. Lewicki, "Structure-Based Connectionist Network for Fault Diagnosis of Helicopter Gearboxes", ASME Journal of Mechanical Design, Vol. 120, No. 1, pp. 100-105..
- V. B. Jammu, K. Danai, and D. G. Lewicki, "Experimental Evaluation of Structure-Based Connectionist Network for Fault Diagnosis of Helicopter Gearboxes", ASME Journal of Mechanical Design, Vol. 120, No. 1, pp. 106-112.
- K. Wang, D. Yang, K. Danai, and D. G. Lewicki, 1999, ``Model-Based Selection of Accelerometer Locations for Helicopter Gearbox Monitoring,'' J. of American Helicopter Society, Vol. 44, No. 4, pp. 269-275.