Rolling Bearing Local Fault Detection During a Run-Up Test Using Wavelet-Filtered CEEMDAN Envelopes

In the context of the fast technological evolution and high reliance on machines, surveillance became critical and important. One of the most essential parts that need continuous inspection is bearings. For that purpose, many techniques have been developed on the base of lubricant analysis, infrared imaging and vibration analysis to make sure that maintenance operations are executed in the right time and guarantee a continuous production. Vibration analysis is often used for diagnosing constant speed machines faults, but rarely tested under non stationary conditions. In this work, the practicability of a hybrid vibration-based method, constructed on the base of the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and the Wavelet Multi-Resolution Analysis (WMRA), is tested for rolling bearing local fault detection under variable conditions, where the machine experiences a run-up test. The captured signals have been decomposed with CEEMDAN to extract the bearings vibration response, the extracted signals have been then filtered with WMRA trying to isolate the impulse train generated by bearing faults. Two criteria have been proposed in order to recognize the best modes and details: The bearing resonance frequency coverage and Kurtosis values. The results demonstrate that the presented hybrid approach has effectively highlighted the bearing faults in the non-stationary conditions, with both simulated and experimental signals.