The Fault Characteristic Extraction Method Of Railway Train Rolling Bearing Under Strong Background Noise

The axle box system of railway train is the core component of the train and plays an important role in the safe operation of the railway train.Due to the complicated structure of the axle box system,huge dynamic load and high running speed,the axle box bearing is also one of the parts that are easy to be damaged.When the bearing is working,the vibration signals generated often contain a large amount of external noise,and it is difficult to extract the impact component with fault information.Therefore,reducing the impact of environmental and system noise,extracting the impact components containing fault features from the background noise have become the key to solving the problem of fault diagnosis of axlebox bearings.In summary,the fault diagnosis research of train axle box bearings is of great value for saving resources and improving train safety performance.(1)The method is based on the difference of quality factor Q between harmonic signal and impulse signal,and the correlated kurtosis is introduced into Tunable QFactor Wavelet Transform(TQWT).TQWT is used to carry out sparse representation of high quality factor and low quality factor respectively.Then,Morphological Component Analysis(MCA)is used to separate the nonlinear signals.During the separation process,calculating the correlated kurtosis of each decomposition layer of low resonance components to achieve signal reconstruction for low resonance components.Envelope demodulation analysis is performed on the reconstructed low resonance component to realize the extraction of fault characteristics of the bearing.The results show that this method can effectively isolate the compound fault characteristics of rolling bearing,when the rotational speed is known.(2)The tunable Q-Factor Wavelet Transform method based on correlation kurtosis needs to calculate the period of fault impact as a known condition.If the speed information error is large,it will affect the effectiveness of the method.In order to compensate the influence of rotation speed error,this paper proposes a dynamic bayesian wavelet transform method based on negentropy and unscented kalman filter.This method does not require a period of fault shock as a known condition,but designs the equation of state with wavelet parameters that need to be optimized,and filters the negative entropy value of the square envelope of the signal to design the measurement equation.Based on the unscented Kalman filter method,The iterative solution of the unscented Kalman posterior probability is to achieve an optimal estimation of the complex Morlet wavelet parameters,and then the resonant frequency band containing the fault information can be adaptively determined.The results show that this method is also realize the separation of fault characteristics and strong background noise when the rotational speed is unknown.(3)Taking high-speed train bearing and railway freight car wheelset bearing as the research object,the above two diagnostic methods are tested and verified.The results show that the Tunable Q-Factor Wavelet Transform(TQWT),based on the difference of signal oscillation characteristics,can effectively extract the fault features;the dynamic bayesian wavelet transform method based on negative entropy and untracked kalman filter can effectively extract the faults of bearing outer ring,inner ring and rolling ring under strong background noise.