Research On Fault Diagnosis Method Of Train Axle Box Bearing Based On Compression Sensing

As a key part of the train running part,the railway axle box system is also an important part that affects the running safety of the train.Therefore,monitoring condition of rolling bearing is of great importance.In recent years,the theory of compressed sensing(CS)provides a novel idea for the health monitoring of vibration data of rotating machinery,which is widely used in the field of vibration data processing.This paper studies the repair method of missing bearing vibration data based on the theory of compressed sensing,and puts forward the corresponding diagnosis and treatment method for the single fault and compound fault of bearing.The methods are verified by simulation and experiment The specific research contents are as follows:(1)In order to solve the problem of repairing the missing vibration data in the process of rolling bearing condition monitoring,a model of repairing the missing vibration data based on compression sensing is established.This model uses the same type of complete data to train the over complete dictionary through the improved K-singular value decomposition(KSVD).In the reconstruction stage,a data repairing method of variable step size sparse adaptive match tracking based on the improvement of selfadaptive threshold is proposed.The effectiveness of the proposed method is verified by repairing random and continuous defect data,through comparing,the proposed method improves the repair accuracy and operational efficiency.(2)In order to solve the problem of a single point fault diagnosis of rolling bearing,an envelope detection diagnostic method based on compression sensing is proposed.The fast spectral kurtosis method is utilized for signal de-noising.Fourier transform basis can be used as a sparse dictionary.A Gaussian random matrix is constructed as the measurement matrix,and the envelope feature of bearing signal is reconstructed by algorithm of CS to obtain the signal components that directly reflect the fault feature and the faulty feature frequency.Simulation and experimental data are used to verify the effectiveness of the proposed method.(3)In order to solve the problem that it is hard to extract the composite fault feature of rolling bearing.A compound fault diagnosis method based on maximum correlated kurtosis deconvolution(MCKD)and subspace pursuit(SP)algorithm is proposed.Firstly,the fault period is determined with prior knowledge,and the method of coarse and fine stages is proposed to search the length of the filter,taking the envelope entropy as the optimization index to achieve the optimal filtering effect.Then the envelope signal of the time-domain signal is reconstructed by the subspace pursuit algorithm.Finally,the proposed method is verified by the simulation and the test signal from the laboratory freight train running in the experiment platform.The fundamental and the second frequency of a single fault feature frequency are extracted from the composite fault signal.