An Early Fault Diagnosis Method For Rolling Bearing Based On Stochastic Resonance And Otsu-EWT

Early fault diagnosis of rolling bearings is the research content in the field of Prognostic and Health Management.It can identify faults before the early failure of equipment or abnormal performance,so that it is easy to arrange maintenance and repair,and has important research significance.This topic takes rolling bearing as the research object,focusing on the main line of early fault diagnosis,gradually develops the research of weak signal detection,signal decomposition and screening,and signal de-noising.The main research contents are as follows:Firstly,an adaptive stochastic resonance detection(SR)method based on quantum genetic algorithm(QGA)is proposed.Through the analysis of the model parameters influence on stochastic resonance,the necessity of parameter optimization is cleared,and quantum genetic algorithm is introduced to optimize the two parameters of the system,and is compared with the traditional stochastic resonance and single variable adaptive stochastic resonance.The effect of bivariate adaptive stochastic resonance is verified through simulation and open source data.Secondly,an empirical wavelet transform(EWT)signal decomposition method based on the Otsu is proposed.According to the problem of window function fixation and mode mixing in the traditional decomposition method,the empirical wavelet transform is used for signal decomposition,to overcome the original boundary problem by Otsu method and adaptive segment signal spectrum.Research on performance of dimensional and dimensionless index in the early stage of failure;select the impulse index as screening index of the AM-FM components.The effect of Otsu-EWT and evaluation index is verified through simulation and open source data.Thirdly,a signal de-noising method based on singular value difference spectrum(SVD)is proposed.In order to remove noise interference from AM-FM components,the singular value decomposition is used to AM-FM components selected and the largest mutation point is found,and the de-noising is realized by signal reconstruction.The effect of this method is more close to the original signal than traditional de-noising method through simulation and open source data.Finally,test platform is constructed and data is acquired.The vibration test rig of rolling bearing is set up and the required test data are collected.In this paper,double row angular contact ball bearing is used as the test object,and the fault types are respectively inner ring wear and outer ring fissures.This method is used for early fault according to the specific circumstances of signal.The test results show that this method can achieve the desired goal,and the effect is great.