Vibration-based bearing fault detection on experimental wind turbine gearbox data

Cédric Peeters, Patrick Guillaume, and Jan Helsen
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmec_16_036.pdf2.79 MBJune 7, 2016 - 10:29pm

Rolling element bearing faults are one of the most common defects in rotating machinery. The detection of these faults has lately attracted an increasing amount of attention in industry. Detecting the faults in their incipient phase can prevent a more catastrophic breakdown of a machine and can save a company time and money. An often occurring problem in analyzing vibration measurements of a rotating machine is the presence of harmonic signal components originating from different machine parts. This paper focuses on separating the bearing fault signals from other masking signal content coming from elements like gears or rotating shafts. The separation is based on the assumption that signal components of gears or shafts are deterministic and appear as clear peaks in the frequency spectrum, whereas bearing signals are stochastic due to random jitter on their fundamental period and can be classified as cyclostationary. A technique that has recently gained more attention for separating these two types of signals is the cepstral editing procedure (CEP). This preprocessing method is investigated further in this paper as an automated procedure. The cepstral editing procedure makes use of the fact that deterministic components appear in the cepstrum as narrow peaks and removal of these cepstral peaks reduces the amplitude of the corresponding discrete components. The residual stochastic signal after cepstral editing should consist mainly out of bearing fault signals in certain frequency bands. To analyze the residual signal the squared envelope spectrum is examined. The issue of selecting the proper frequency band where to demodulate the signal necessary for enveloping is addressed through the use of spectral kurtosis. The performance of the developed methods will be validated on experimental data from the National Renewable Energy Laboratory (NREL). This data has been made available by NREL in context of the wind turbine gearbox condition monitoring round robin study. As such other research teams from around the world have analyzed this data as well and published their findings. The measurements were done on a wind turbine gearbox in a healthy and damaged condition and the setup is documented extensively, making it well suited for a performance comparison. A summary of the findings on the NREL data is discussed in this paper.

Publication Year: 
2016
Publication Volume: 
7
Publication Control Number: 
036
Page Count: 
12
Submission Keywords: 
Condition Based Monitoring
bearing fault detection
data preprocessing
Submission Topic Areas: 
Verification and validation
Submitted by: 
  
 
 
 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed