Speed-spectral whitening for enhancing envelope analysis in speed varying conditions

Nowadays, the vibration analysis of rolling element bearing operating under varying-speed condition appears to be an incentive field of investigation. In this context, envelope analysis is among the most suited technique for extracting fault signatures. The most common method consists of order tracking and demodulating the signal around a spectral frequency band after an appropriate discrete/random separation technique. This accounts for speed variation by order tracking the signal implicitly, but not explicitly. Indeed, it has been inspected by the authors that bearings respond differently according to the operating regime. There may be modulations that jointly depend on the machine speed and on the frequency content of the response, leading to an amplification/attenuation of the bearing fault signature. The main contribution of this paper is to provide a preprocessing step called speed-spectral prewhitening to compensate the explicit contribution of the regime variation. Last, the potentiality of the proposed approach for envelope analysis is proved on actual data measured from an industrial bearing test-rig.