Autoregressive modelling for rolling element bearing fault diagnosis

H Al-Bugharbee, I Trendafilova

Research output: Contribution to journalConference Contribution

2 Citations (Scopus)
75 Downloads (Pure)


In this study, time series analysis and pattern recognition analysis are used effectively for the purposes of rolling bearing fault diagnosis. The main part of the suggested methodology is the autoregressive (AR) modelling of the measured vibration signals. This study suggests the use of a linear AR model applied to the signals after they are stationarized. The obtained coefficients of the AR model are further used to form pattern vectors which are in turn subjected to pattern recognition for differentiating among different faults and different fault sizes. This study explores the behavior of the AR coefficients and their changes with the introduction and the growth of different faults. The idea is to gain more understanding about the process of AR modelling for roller element bearing signatures and the relation of the coefficients to the vibratory behavior of the bearings and their condition.

Original languageEnglish
Article number012088
Number of pages9
JournalJournal of Physics: Conference Series
Issue number1
Publication statusPublished - 9 Jul 2015
Event11th International Conference on Damage Assessment of Structures - Ghent University, Ghent, Belgium
Duration: 24 Aug 201526 Aug 2015


  • bearings (machine parts)
  • damage detection
  • failure analysis
  • pattern recognition
  • roller bearings
  • time series analysis
  • vibration analysis

Cite this