Incipient Bearing Fault Signature Identification via Vibration Envelope Analysis and MUSIC

Rolling bearings represent the weakest element in low-voltage machines form a maintenance perspective. The detection of localized bearing faults in steady-state conditions is widely studied via vibration envelope analysis. However, the detection of these faults during the machine start-up presents increased challenges such as the accurate monitoring of time-evolving components. This paper introduces a time-frequency analysis tool to qualitatively evaluate localized bearing faults evolutions during an inverter-fed machine start-up. Transient vibration signal envelope is translated into the time-frequency domain by utilizing the Multiple Signal Classification (MUSIC) pseudo-spectrum. Furthermore, a thorough sensitivity analysis is performed to elucidate the strengths and weaknesses of the proposed methodology.