Abstract: Rolling element bearing fault feature is very weak in the incipient process and interfered by normal features. Therefore, it is not convenient to early fault classification. Frequency bands obtained by wavelet packet decomposition are investigated for rolling bearing vibration signal. A new fault diagnosis method is put forward based on wavelet packet-coordinate transformation(WP-CT) for feature enhancement. As every frequency band obtained by wavelet packet containing fault feature, principal component analysis(PCA) or independent component analysis(ICA) is used for every sub-band coordinate transformation. Then, a signal can be reconstructed for fault classification. This method can weaken the interference from normal signal information and intensify fault impact information. Thus, it can use all information from every sub-band and contribute to improve classification performance. Simulated signal and practical testified signal are used to testify the effectiveness of this method. It can be concluded that this new WP-CT method can enhance weak feature and reduce the interference from normal signal, which is very helpful for rolling bearing fault diagnosis technology development.

Key words: Coordinate transformation, Feature enhancement, Independent component analysis, Principal component analysis, Wavelet packet