Research On Diagnosis Of High-speed Railway Vehicle Axle Box Bearing Based On Variational Mode Decomposition And Morphologcial Filter

With the rapid development of high-speed railway in China,the requirements of monitoring and diagnosis of in-service railway vehicles are swiftly increasing.Axle box bearing is an extremely important component of vehicle,whose operating status directly affect the safety and stability of the vehicle.Therefore,the research on fault detection of axle box bearing is of tremendous engineering significance and application value.Comparing with the detecting technology based on temperature and acoustic signals,the vibration signal has more advantages in real-time analysis.Therefore,based on the analysis of the vibration characteristics of axle box bearing,this dissertation carried out certain researches on diagnosis methods of axle box bearing in the following aspects:For decomposing the resonance frequency bands of the faulty signal,the theory of variational mode decomposition(VMD)was introduced detailedly in this dissertation.By constructing a simple set of bearing fault simulated signal,the decomposition results of different parameters of VMD were comprehensively analyzed.Combining with the theory of VMD,the influences of each parameters of VMD were summarized.Considering the shortage of VMD that it requires a lot of posterior information to determine the parameters,an adaptive VMD called Scale-space guiding VMD was proposed in this dissertation.The proposed method estimated the number of resonance frequency bands in the analyzed signal spectrum,and the relationship of scale coefficient and rotation frequency was constructed according to the characteristics of bearing signal.The estimated number of resonance bands was used to guide VMD to decompose the analyzed signal,which greatly improves the adaptability of VMD method on the application of bearing signals.The effectiveness and adaptability of the proposed method were verified by a set of simulation and test rig signals.Subsequently,a modified scale-space representation which is more suitable for VMD was proposed to improve the over-segmentation issue of the original scale-space.The modified scale-space representation enhances the accuracy of estimation of the number of resonance bands.Moreover,the penalty factor and thus initial central frequencies of VMD were estimated through the result of the segmentation of frequency spectrum.Then the modified SVMD was proposed,based on the combination of modified scale-space representation and estimated parameters,to further improve the adaptability,decomposing accuracy and computational efficiency of SVMD.The advantages of MSVMD,comparing with SVMD,in decomposing carefulness and computational efficiency were testified by applying on simulation and test rig signals.In order to optimize the effect of bearing signal demodulation,a demodulation algorithm called PSO-MF was proposed by combining morphological filter(MF)and particle swarm optimization(PSO).The proposed method regarded the MF operations as the dimensions of PSO and applies kurtosis as the fitness value;then calculated the optimal parameters of structure element(SE)of various MF operations;finally the optimal MF operation and SE would be selected as the output of PSO-MF,which realizes the demodulation optimization of bearing signal.A set of simulation and test rig signals were processed by PSO-MF to validate the effectiveness and anti-interference of the proposed method.Aiming at the problem of redundant interference noise existing in the same resonance band in bearing signal,a periodic component extraction method called morphological filter autocorrelation outliers(MFAO)was proposed in this dissertation based on the cyclostationality of bearing signal.The proposed method applied PSO-MF to obtain the envelope of sub-signals obtained by MSVMD decomposing the analysed signal and exploited outlier detection to separate the periodic components and interference noise in the autocorrelation function of the envelope sub-signal.At last the effectiveness of the proposed method was verified by the simulation and test rig signals.