A multi-frequency approach for the fault identification of bearings and gears based on frequency shifting and narrowband filtering

The vibration signal becomes impulsive due to the faults in bearings and gearboxes. A narrowband filter can detect an impulse-like event in a signal. The present work proposes a frequency-shifting narrowband filtering approach to extract the impulse information at multiple frequencies. Several narrowband-filtered signals from the vibration signal are obtained by frequency-shifting the signal and subsequently passing the frequency-shifted signals from a narrowband filter. The spectral mean and variance signals obtained from the envelopes at different frequencies capture the impulse information in the signal. The noise is uncorrelated across envelopes at different frequencies and gets nullified in the spectral signals. Thus, the proposed technique demonstrates a successful detection of the early stage and seeded faults in bearings from the IMS and CWRU datasets, respectively. The algorithm also reveals a clear distinction between healthy and damaged gearboxes through the analysis of NREL and in-house datasets, encompassing scuffing and chipped tooth faults, respectively. The present approach provides a robust condition-monitoring tool for bearings and gearboxes in industrial applications.