Research On The Fault Characteristic Separation Method Of Train Bearings Vibration Signal Under Strong Background Noise

Axle box bearing is one of the most important parts in the train's walking section.It is also one of parts most susceptible to damage due to the complicated structure,a high running speed and the bad working environment.While the trains in operation process,train axle box bearing is affected by the system and environment particularly severe due to the interferences like rail motivated,which leads the fault feather signal obtained is usually overwhelmed with strong background noise.Besides,the fault signal and noise signal aliasing mutually in the frequency band,the traditional filtering noise reduction method is difficult to the separation.How to separate out useful fault characteristic signal of train axle box bearing from strong background noise is an important content to condition monitoring and fault diagnosis.This paper mainly includes:(1)Considering the problem that the train bearing with a single point of fault under strong background noise is difficult to separate,this paper proposes an improved resonance demodulation method based on typical correlated kurtogram.Based on the ideas of typical kurtogram,the kurtosis was replaced by correlated kurtosis,and then the method located typical fault impact signal interested in the frequency range rapidly,relying on the optimal spectral correlated kurtosis.The analysis shows that the method reduces the interference of background noise and other nonimpact fault information,can effectively adaptive locate resonance frequency band and realizes the separation of fault characteristics and the strong background noise.(2)Considering that the train under strong background noise characteristics of complex composite bearing fault components are inseparable,this paper puts forward a kind of fault diagnosis of rolling bearing based on multipoint kurtosis spectrum and maximum correlated kurtosis deconvolution method,which overcomes the shortcoming that the traditional maximum correlated kurtosis deconvolution method needs to predict fault period.The sampling signal was processed by multipoint kurtosis spectrum method,through comparing the multipoint kurtosis of output signal under different period to modify the anticipated fault feature period.And then it put the accurate fault feature period into the algorithm to realize the extraction of compound fault.The analysis shows that even under the condition of unknown speed accurately,the method can still be effective to seperate the compound fault characteristic of rolling bearings.(3)This article takes wheelset axle box bearing as the research object,and the above two methods are analyzed.The experimental results show that: the improved resonance demodulation method based on typical correlated kurtogram shows a great superiority in both the accuracy and stability to the resonance demodulation method based on the typical kurtogram;The compound fault diagnosis of rolling bearing based on multipoint kurtosis spectrum and maximum correlated kurtosis deconvolution method expands the application range of the maximum correlated kurtosis deconvolution,and realises compound fault characteristic seperation.