Fault Diagnosis for Rolling Bearings in Non-Stationary Operating Conditions Caused by Torsional Vibrations

The present study aims to investigate the effects of torsional vibrations on the global vibrations of a rotating machine, from the point of view of bearing fault diagnosis. In this regard, a new numerical vibration signal model for faulty rolling bearings based on instantaneous angular positions of the rolling elements is proposed and validated against the experi-ments. This model then allows to introduce the desired time-varying rpm, and produce a corresponding bearing vibration signal subjected to the twisting oscillations (i.e. torsional vibrations). In addition to numerical studies, torsional vibrations are introduced in the controlled manner at a shaft-bearing system to perform experiments. For validation purposes, the numerically-generated signals are also compared with the captured experimental vibration signals. The results presented here confirm the reliability of methodology used in signal modelling for rolling bearings under torsional vibrations. More-over, both the experimental and the numerical signals show the complexities arising from torsional vibrations in inter-preting the diagnostic results in frequency domain analyses. Finally, to overcome this issue, and to deal with non-station-arities introduced by torsional vibrations, the well-known resampling-based order tracking method is used.