Research On Fault Dianosis Of Automotive Transmission Bearings

Rolling element bearing is an important part of automobile transmissions and one of the main source of failure.It’s running condition directly affects the safety performance of the entire transmission.The operating conditions of transmission bearings is complex,and the fault signal has non-stationary,non-linear,low signal-tonoise ratio characteristics,which will increase the difficulty of fault feature extraction and affects the accuracy of fault diagnosis.Studying bearing fault diagnosis technology has important theoretical significance and application value for improving transmission performance and reducing maintenance cost.This article focuses on three important aspects of transmission bearing fault diagnosis: bearing fault diagnosis under constant speed condition,bearing fault diagnosis under variable speed condition,and early bearing fault feature extraction.When diagnosing bearings under constant speed condition,the spectral kurtosis has the advantage of rapid detection capability for transient impact signals.A rolling bearing fault diagnosis method based on the combination of fast kurtogram alororithm and resonance demodulation is used,but this method cannot be used to solve the problem of bearing fault diagnosis under variable speed condition.In order to solve this problem,a method based on Spline-kernelled chirplet transform and order spectrum analysis is used to realize variable speed bearing fault diagnosis.In order to further realize the bearing fault diagnosis decision,a method of early fault feature extraction of bearings based on the combination of manifold learning and s-k-means clustering is proposed.The above method has a good diagnostic effect verified by experimental data.The specific practices are as follows:(1)Rolling bearing fault diagnosis method based on the principle of resonance demodulation.First,fast kurtogram alororithm is used to process the bearing vibration signal to provide detailed parameters for the filter design,and the corresponding filtering process is performed,and then the filtered signal is subjected to envelope demodulation analysis to extract the characteristic frequency of the fault to complete the fault type identify.(2)Variable speed rolling bearing fault diagnosis method based on Splinekernelled chirplet transform,using Spline-kernelled chirplet transform’s timefrequency analysis method to obtain the vibration frequency of the faulty bearing to obtain the instantaneous frequency and then the instantaneous speed curve.Finally the fault diagnosis of the bearing under the condition of variable speed is realized by order analysis,.(3)Feature extraction of early fault of rolling bearing based on manifold learning and s-k-means clustering.First of all calculating the initial feature set of the original signal,and then obtain low-dimensional features through feature selection and manifold learning to reduce the dimensions.Finally,s-k-means clustering is used to process the fault feature matrix to automatically obtain the optimal number of clusters to accurately implement fault classification.