Fault Diagnosis Of Rolling Bearing Based On Wheels Of Low Speed Vibration Signal

Although China has a very large land area,but the country's resource distribution is uneven.There is a big difference between the development of different regions.Railway transport is a long-distance transport,the volume is very large,its role is difficult to be replaced.In addition,as of now China's "four vertical and four horizontal" high-speed rail backbone network has been basically completed in the inter-regional transport plays an important role.In rail transport,the most noteworthy must be security issues.At present,China's train and its key parts of the online identification,predictive diagnosis and monitoring technology has great room for development,and has great practical significance.Therefore,as an important part of the wheel wheel,this paper has carried on the characteristic signal analysis to the rolling bearing and has carried on the diagnosis research to its fault.Firstly,the structure of the rolling bearing and the failure mechanism of the vibration are analyzed,and the failure mode,the vibration type and the cause of the failure are discussed in a more comprehensive way.Then,the time domain characteristic parameter values and their FFT spectra and power spectra of different state bearings are analyzed.In this paper,four kinds of bearing vibration signals are filtered and processed.Then,the hilbert envelope analysis of the filtered signal is carried out.Finally,the shock value of the signal is calculated by SPM method,and it can judge whether it has failed.But the method can not effectively determine the location of the bearing failure.Based on the relationship between the alpha stable distribution parameter estimation and the bearing vibration signal of different states,a fault classification model combining the stability distribution parameter estimation with Bp neural network and the LSSVM is proposed respectively.First,each type of rolling bearing vibration signal is used as an alpha stable distribution parameter estimate.And then select the most sensitive and stable two sets of characteristic parameters as the fault feature quantity.The two parameters are input into the Bp neural network model and the LSSVM model for fault classification.The standard motor bearing vibration signal data and the measured vibration signal data on the wheel running test bench were analyzed by example.The results show that the two models can achieve accurate fault classification for standard motor bearing test data.The Bp neural network model is suitable for the fault identification capability of the train wheel bearings,but the LSSVM model can effectively classify the different fault locations of the rolling wheel bearings.