Non-stationary Signal Analysis For Rolling Bearing Fault Diagnosis

Rolling bearings is widely used in industrial production, as a core part of mechanical equipment, its operating conditions directly affect the reliability and stability of the machinery and equipment. Therefore, the rolling bearing fault diagnosis for mechanical equipment operation maintenance is of great significance.The characteristics of the fault information extraction are the key to the bearing fault diagnosis. In this paper, for rolling bearing fault diagnosis, using the local mean decomposition(LMD)、morphological filtering and approximate entropy theory, from the perspective of signal denoising and signal sequence complexity, rolling bearing vibration signal extraction methods were studied experimentally and provided a new theoretical basis for signal extraction. The main research work is as follows:(1)For non-stationary characteristics of vibration signals of rolling bearings, and the difficult to extract the actual fault characteristic signal, researched the local mean decomposition method, this method can made complex non-stationary signal decomposed into a series of AM-FM functions, achieved the separation of modulation signals of different frequency components, and can effectively isolate the fault component, then applied to the actual rolling bearing vibration signal decomposition.(2)According to the serious actual bearing vibration signal noise, studied the morphological filtering algorithm in vibration signal processing, and research the adaptive morphological filtering method base on the signal extreme to determine the length of the structural elements, this method can maintain the appearance of the original signal, and greatest inhibition the impulse noise impact, using the local mean decomposition to extract the fault feature information from the filtered signal. Simulation results show that the morphological filter with the local mean decomposition can effectively extract the fault feature of signals.(3)Finally, from the perspective of the signal complexity, proposed approximate entropy multi-scale method base on LMD, to extract the fault feature of bearing, this method can effectively distinguish different types of bearing faults, and has stronger anti-interference ability to approximate entropy, obtaining more fault characteristic information, the simulation experiment and example analysis show that this method can be applied to the bearing fault feature, and determine the operating state of the bearing.