Machinery fault diagnosis using advanced correlation filters

The purpose of machine condition monitoring is to determine the present health of machineries. Capturing the abnormal symptoms of machineries from vibration signatures involves the use of signal processing algorithms on measured vibrations. However, the commonly used method such as FFT based power s...

Full description

Bibliographic Details
Main Authors: Loo, C. K., Mastorakis, Nikos E.
Format: Conference or Workshop Item
Published: 2007
Subjects:
Online Access:http://shdl.mmu.edu.my/3233/
_version_ 1848790271159959552
author Loo, C. K.
Mastorakis, Nikos E.
author_facet Loo, C. K.
Mastorakis, Nikos E.
author_sort Loo, C. K.
building MMU Institutional Repository
collection Online Access
description The purpose of machine condition monitoring is to determine the present health of machineries. Capturing the abnormal symptoms of machineries from vibration signatures involves the use of signal processing algorithms on measured vibrations. However, the commonly used method such as FFT based power spectra assumes the the signal is ergodic and stationary. The FFT based method may produce unpredictable results especially in an industrial environment that subjected to random or periodic noise. Such effects can be detected, however, with second-order cyclostationary statistical method such as Degree of Cyclostationary (DCS). This paper discussed the implementation of machinery fault diagnosis using the Quad-Phase Unconstrained Optimal Tradeoff Synthetic Discriminant Function (QUOTSDF) and DCS features for fault classification. Quad-Phase Unconstrained Optimal Tradeoff Synthetic Discriminant Function (QUOTSDF) is used in this effort because of its ability to provide high discrimination while providing noise tolerance. The machine health condition is identified based on the comparison of acquired real-time vibration features with template features. The vibration data were collected from the Schenck Motor MM-61. Four machinery conditions are simulated by the motor, which are normal (no fault), bearing damage, machine imbalance, and foundation looseness. The Fast Fourier Transform (FFT) and the Degree of Cyclostationary (DCS) have been utilized for features extraction from the power spectrum of the vibration data. Fault diagnosis based on DCS features are shown to outperform FFT in accuracy.
first_indexed 2025-11-14T18:09:57Z
format Conference or Workshop Item
id mmu-3233
institution Multimedia University
institution_category Local University
last_indexed 2025-11-14T18:09:57Z
publishDate 2007
recordtype eprints
repository_type Digital Repository
spelling mmu-32332011-10-12T07:26:23Z http://shdl.mmu.edu.my/3233/ Machinery fault diagnosis using advanced correlation filters Loo, C. K. Mastorakis, Nikos E. T Technology (General) QA75.5-76.95 Electronic computers. Computer science The purpose of machine condition monitoring is to determine the present health of machineries. Capturing the abnormal symptoms of machineries from vibration signatures involves the use of signal processing algorithms on measured vibrations. However, the commonly used method such as FFT based power spectra assumes the the signal is ergodic and stationary. The FFT based method may produce unpredictable results especially in an industrial environment that subjected to random or periodic noise. Such effects can be detected, however, with second-order cyclostationary statistical method such as Degree of Cyclostationary (DCS). This paper discussed the implementation of machinery fault diagnosis using the Quad-Phase Unconstrained Optimal Tradeoff Synthetic Discriminant Function (QUOTSDF) and DCS features for fault classification. Quad-Phase Unconstrained Optimal Tradeoff Synthetic Discriminant Function (QUOTSDF) is used in this effort because of its ability to provide high discrimination while providing noise tolerance. The machine health condition is identified based on the comparison of acquired real-time vibration features with template features. The vibration data were collected from the Schenck Motor MM-61. Four machinery conditions are simulated by the motor, which are normal (no fault), bearing damage, machine imbalance, and foundation looseness. The Fast Fourier Transform (FFT) and the Degree of Cyclostationary (DCS) have been utilized for features extraction from the power spectrum of the vibration data. Fault diagnosis based on DCS features are shown to outperform FFT in accuracy. 2007-07 Conference or Workshop Item NonPeerReviewed Loo, C. K. and Mastorakis, Nikos E. (2007) Machinery fault diagnosis using advanced correlation filters. In: 11th WSEAS International Conference on Circuits, 23-25 JUL 2007, Agios Nikolaos, GREECE. http://apps.webofknowledge.com/full_record.do?product=WOS&search_mode=GeneralSearch&qid=1&SID=V2namOAJkanAEgD@AON&page=125&doc=1250
spellingShingle T Technology (General)
QA75.5-76.95 Electronic computers. Computer science
Loo, C. K.
Mastorakis, Nikos E.
Machinery fault diagnosis using advanced correlation filters
title Machinery fault diagnosis using advanced correlation filters
title_full Machinery fault diagnosis using advanced correlation filters
title_fullStr Machinery fault diagnosis using advanced correlation filters
title_full_unstemmed Machinery fault diagnosis using advanced correlation filters
title_short Machinery fault diagnosis using advanced correlation filters
title_sort machinery fault diagnosis using advanced correlation filters
topic T Technology (General)
QA75.5-76.95 Electronic computers. Computer science
url http://shdl.mmu.edu.my/3233/
http://shdl.mmu.edu.my/3233/