Abstract: To overcome the difficulty of early fault diagnosis for mechanical system, considering that wavelet transform correlation filter (WTCF) is good at filtering and permutation entropy algorithm is sensitive to weak changing in a signal, the two algorithms are integrated together, then a new conception named wavelet correlation permutation entropy (WCPE) is defined, and a new feature extraction method based on WCPE is proposed. The gathered vibration signal of equipment is processed by way of WTCF, and the high signal-to-noise wavelet coefficients are gotten. Then, the permutation entropy complexity of each layer wavelet coefficients are calculated to construct the WCPE feature vectors, which are employed to analyze the weak change of variation signal. The proposed method is verified with on the full lifetime datasets of a certain bearing, which proves that signal features extracted by WCPE method can not only express truly the bearing detailed condition changing from normal to fault, but also detect duly the early fault of bearing. Comparing with other methods for early fault diagnosis, such as wavelet entropy, wavelet correlation scale entropy, etc., the proposed method can advance the finding time of early fault obviously.

Key words: Early fault diagnosis, Roller bearing, Wavelet correlation permutation entropy