Abstract: Weak fault feature extraction of rotating equipment is important for early fault warning. Continuous wavelet transform (CWT) can analyse signal in multiple scales, which can describe the local characteristics of the signal in different scales through translation. Therefore, it is convenient to the weak fault characteristic determination. As the reconstruction of different scale signals lead to different classification results, this research presents a new early fault diagnosis method based on CWT. CWT is carried on for signal decomposition. The optimal scale for signal analysis is selected based on wavelet entropy. Envelope spectrum analysis is applied to the reconstructed signal for feature extraction. The working condition can be classified based on characteristic frequency. Simulation and rolling element bearing signals are used to verify the effectiveness of this method. It can be concluded that this method is suitable for weak feature extraction based on this investigation.

Key words: continuous wavelet transform;signal reconstruction;wavelet entropy;early fault classification