Abstract: In view of the random noise interferes with the rolling bearing fault characteristic signal extraction this problem, a new method of fault feature extraction with a kind of filtering noise reduction based on singular value decomposition (SVD) and local mean decomposition (LMD) is put forward. The phase space reconstruction of Hankel matrix and singular value decomposition method is used to process the original vibration signal. The de-noising signal is decomposed by using local mean decomposition to make the multi-component modulation signal decomposed into a series of production function (PF). Combining with the technology of resonant demodulation, the production function component is analyzed and the fault characteristic frequency is extracted. Numerical simulation and practical bearing failure data analysis are done. The results show that the proposed method can be used to effectively improve the local mean decomposition, distinguish the typical fault of measured signal of rolling bearing, and improve the effect of the rolling bearing fault diagnosis.

Key words: fault feature extraction, local mean decomposition, singular value decomposition, rolling bearing