Vibration Monitoring of Rolling Element Bearings

“There are many various signal processing techniques
implemented for the vibration diagnosis of bearings
on rotating machines. They can be sorted into three
main categories:

– Frequency methods: Narrow-band spectral analysis
(FFT), envelope analysis (amplitude demodulation),
cepstrum analysis.

– Time methods or statistical methods: RMS value
(overall level over a given frequency band, usually foracceleration and in high frequencies for bearings: >
1 kHz), peak value, peak factor, kurtosis and manufacturers acceptance tests: defect factor DEF,
SEE technology, Gse, etc.

– Filtering methods applied to time signals: essentially high pass and band pass, denoising by spectral subtraction of all periodic components.
Frequency processing consists in detecting the presence of periodicities due to repeated shocks generated by the possible marking of an inner or outer raceway or by the scaling of rolling elements. Among the aforementioned techniques, envelope
analysis is the most advanced tool.
Warning! Envelope analysis can lead to gross
diagnosis errors, since the presence of a set of
harmonic lines does not necessarily reflect the
occurrence of periodic shocks. It is rather complex to implement this technique since it requires to know beforehand the frequency domain of interest for demodulation (resonance area). Restricted to
the search for bearing defects and/or
clearances, it is usually associated with:
narrow-band analysis (FFT) allowing for
the dynamic characterization of the line of
shafting. statistical processing, such as Kurtosis.
Cepstral analysis is used to easily extract and quantify sets of harmonic lines observed on a power spectrum: the richer the shock spectrum (as for the number of components and their emergence with respect to background noise), the higher the amplitude on the cepstrum for the fundamental line of this shock spectrum.
Another advantage is the relative lack of sensitivity to load variations and to modifications of transmission paths between excitation source (bearing) and transducer.


Kurtosis is a statistical indicator used to characterize the pulse character of a signal. It is a dimensionless parameter characterizing the flattening of the signal probability density…”

Continuing Reading:

Interested in monitoring rolling element bearings?

Learn about our Industrial Accelerometers