Fault diagnostics using statistical change detection in the bispectral domain

B. Eugene Parker, H. A. Ware, D. P. Wipf, W. R. Tompkins, B. R. Clark, E. C. Larson, H. Vincent Poor

Research output: Contribution to journalArticle

61 Scopus citations

Abstract

It is widely accepted that structural defects in rotating machinery components (e.g. bearings and gears) can be detected through monitoring of vibration and/or sound emissions. Traditional diagnostic vibration analysis attempts to match spectral lines with a priori-known defect frequencies that are characteristic of the affected machinery components. Emphasis herein is on use of bispectral-based statistical change detection algorithms for machinery health monitoring. The bispectrum, a third-order statistic, helps identify pairs of phase-related spectral components, which is useful for fault detection and isolation. In particular, the bispectrum helps sort through the clutter of usual (second-order) vibration spectra to extract useful information associated with the health of particular components. Seeded and non-seeded helicopter gearbox fault results (CH-46E and CH-47D, respectively) show that bispectral algorithms can detect faults at the level of an individual component (i.e. bearings or gears). Fault isolation is implicit with detection based on characteristic a priori-known defect frequencies. Important attributes of the bispectral SCD approach include: (1) it does not require a priori training data as is needed for traditional pattern-classifier-based approaches (and thereby avoids the significant time and cost investments necessary to obtain such data); (2) being based on higher-order moment-based energy detection, it makes no assumptions about the statistical model of the bispectral sequences that are generated; (3) it is operating-regime independent (i.e. works across different operating conditions, flight regimes, torque levels, etc., without knowledge of same); (4) it can be used to isolate faults to the level of specific machinery components (e.g. bearings and gears); and (5) it can be implemented using relatively inexpensive computer hardware, since only low-frequency vibrations need to be processed. The bispectral SCD algorithm thus represents a general methodology for the automated analysis of rotating machinery.

Original languageEnglish (US)
Pages (from-to)561-570
Number of pages10
JournalMechanical Systems and Signal Processing
Volume14
Issue number4
DOIs
StatePublished - Jul 2000

All Science Journal Classification (ASJC) codes

  • Control and Systems Engineering
  • Signal Processing
  • Civil and Structural Engineering
  • Aerospace Engineering
  • Mechanical Engineering
  • Computer Science Applications

Fingerprint Dive into the research topics of 'Fault diagnostics using statistical change detection in the bispectral domain'. Together they form a unique fingerprint.

  • Cite this

    Parker, B. E., Ware, H. A., Wipf, D. P., Tompkins, W. R., Clark, B. R., Larson, E. C., & Poor, H. V. (2000). Fault diagnostics using statistical change detection in the bispectral domain. Mechanical Systems and Signal Processing, 14(4), 561-570. https://doi.org/10.1006/mssp.2000.1299