Comparison of harmonic removal techniques for computing envelope spectra from rolling element bearing vibrations

Fault detection and diagnostics of rolling element bearings are important and relatively mature techniquesused in current vibration-based condition monitoring. Despite this, since for many machines the bearingis often one of the most critical parts, improving the techniques are still interesting. The squared envelopespectrum is one of the most important tools for diagnosing roller element bearings. For a precise and correctestimation of the squared envelope spectrum, it is often important to remove harmonics not related to thebearing, as part of the signal processing prior to computing the envelope spectrum. In real-life measurements,harmonics can be caused by many kinds of sources, motors, gear boxes, and pumps are some examples ofthat. In the present paper, a recently proposed method for automated harmonic removal based on frequencyediting is examined. The results from this method are compared with the established method for harmonicremoval, the automated cepstral editing procedure. Differences between the methods and what effects theyhave on the computed squared envelope spectra are examined. The examination is conducted on a simulatedbearing signal with a fault on the outer race.