Wavelet Transform Based Feature Extraction for Fault Diagnosis of Rolling-element Bearing

摘要

Since rolling-element bearings are widely used in various mechanical and electrical systems, reliable bearing fault diagnosis technique is critically needed to prevent performance malfunction of the systems. In this study, the vibration signals of bearings with no faults, minor faults and serious faults are acquired to construct the experimental dataset. After that, Wavelet Packet Transform (WPT) method has been adopted to decompose the original signals and extract the corresponding energy features representing different fault stage of bearings. Finally, energy features are taken as input vectors to train Artificial Neural Network (ANN) for automatic fault identification of rolling-element bearings. The viability of the developed WPT based feature extraction technique is verified by tests, and the experimental results show that compared with conventional feature extraction method such as discrete wavelet transform, the ANN using WPT based features can achieve high classification accuracy with lower false positive and false negative.