Bearing Fault Detection and Classification Using ANC-Based Filtered Vibration Signal

The defective bearing in a rotating machine may affect its performance and hence reduce its efficiency. So the monitoring of bearing health and its fault diagnosis is essential. A vibration signature is one of the measuring parameters for fault detection. However, this vibration signature may get corrupted with noise. As a result this noise must be removed from the actual vibration signature before its analysis to detect and diagnose the fault. ANC (adaptive noise control)-based filtering techniques are used for this noise removal and hence to improve the SNR (signal-to-noise ratio). In our study an experimental setup is developed and then the proposed work is executed in three stages. In the first stage the vibration signatures are acquired and then ANC is implemented to remove the background noise. In the second stage the time (statistical) and the frequency analysis of the filtered vibration signals are done to detect the fault. In the third stage the statistical parameters of the vibration signatures are used for the classification of the fault present in the bearing using random forest and J48 classifiers.