Application Research Of Adaptive Signal Decomposition Method Using Optimized Scale Space Representation In Bearing Fault Diagnosis

With the proposal and deployment of "The Belt and Road",China's high-speed railway plays a decisive role in domestic transportation,and is increasingly concerned on the international stage.The train usually runs in vast territory,complicated environment with variable climate and complex road conditions,which brings in uncertainty external interference and poses threats to train's performance.Bearing,as one of the core components of the transmission device,should be paid attention to.The bearing has a high probability of failure when it runs in a relatively violate environment.Vibration analysis is the most widely used method in the field of fault diagnosis because of its economic and practical advantages.Gilles proposed the Empirical Wavelet Transform(EWT).The main idea of EWT is applying series of wavelet filters to decompose the original signal into sub ones according to the preset filtering frequency band.Compared with Empirical Mode Decomposition(EMD),the Meyer filters employed in EWT will not cause serious energy leakage so that it can extract the frequency information of the target band more accurately.Also,EWT is able to overcome the translation sensitivity of Discrete Wavelet Transform(DWT)and is more flexible than it.The choice of filtering band directly affects the decomposition result of EWT.This paper improves Scale Space algorithm introduced by Gilles in EWT for band allocation.The original Scale Space algorithm has high computational complexity and excessive redundancy in frequency division.In this paper,parameters and termination conditions of the original algorithm are changed according to the distribution of bearing vibration signal in spectrum.The enhanced algorithm can raise the calculation speed and optimize the result of frequency band allocation.However,the redundancy of frequency band allocation is still maintained in order to adjust the bandwidth through frequency band fusion.Before band fusion,the ideal sub-signal obtained by the enhanced Scale Space algorithm should be fund.Proper index is often used to get the best sub-signal.In this paper,the characteristics of several typical signals and their envelope in the time domain and frequency domain are studied.Then,kurtosis is applied in these signals' analysis.Afterwards,the characteristics and limitations of kurtosis index are found.Since the envelope contains the second order stationary features of bearings' vibration signal,the envelope spectrum is popular in bearing fault diagnosis.From the feature of envelope spectrum,an index called Characteristic Frequency Energy Proportion(CFEP)is proposed.The same index is used to search the optimal bandwidth in band fusion as well.Finally,get the useful information with Hilbert Transform.In the process of introducing the algorithm,the algorithm's principle is explained and the simulation signal is used to verify its feasibility.At the end of the paper,the measured bearing signal is analyzed and processed by the algorithm in the paper,and the feasibility of the algorithm in practical engineering is verified.The dynamic threshold method is used to highlight the different characteristic frequencies of the faults when multi-faults occur.