Research On Fault Diagnosis Of Rolling Bearing Of High Speed Train Based On Parameter-optimized MOMEDA And Compressive Sensing

As the key ‘joint’ component of the high-speed train running gear,rolling bearing is of great significance to ensure the safe and stable operation of the train and the life safety of train personnel.Under the background of the increasing demand of technicians for the real-time running state of bearings and the rapid development of multi-sensor information fusion technology,the redundant data collected under the traditional Nyquist sampling theorem also shows an explosive growth trend.To this end,this paper proposes to use Compressive Sensing method for early fault diagnosis of rolling bearings.This method can make intelligent diagnosis of early fault bearings without disassembling the bearings,and can alleviate the huge pressure on related hardware caused by data acquisition,data storage,signal transmission and subsequent data processing through the implementation of Compressive Sensing software.The article is summarized as follows:(1)The paper summarizes the classification and basic structure of the rolling bearing,then summarizes the cause of rolling bearing vibration,the bearing’s common failure forms and corresponding reasons,then analyzes and deduces the frequency of bearing fault characteristics.On this basis,the algorithm used for fault diagnosis in this paper is determined through comparative analysis,which provides the necessary theoretical basis for the following algorithm application.(2)The sparsity of simulated vibration signals of rolling bearings in different sparse domains is analyzed,which provides the necessary theoretical premise for achieving compression sensing.At the same time,by comparing and analyzing the sparse expression ability of different sparse expression methods for bearing simulation signals,the K-SVD dictionary learning method is selected as the sparse expression method,which is expected to realize the sparse expression of signals and complete the establishment of classification dictionary for fault diagnosis.(3)In this paper,a preprocessing algorithm for noise reduction—MOMEDA is proposed to solve the problem of poor noise immunity of K-SVD dictionary learning method and the disadvantages of blind selection of basis functions in traditional signal processing methods such as short-time Fourier transform and wavelet transform.To solve the problem that the related parameters of MOMEDA need to be set manually before using this algorithm,a parameter adaptive MOMEDA algorithm is proposed,which is applied to the noise reduction pretreatment of bearing simulation signal and measured signal,the experimental results show that the proposed algorithm has a good noise reduction effect.(4)The three core problems of Compressed Sensing are analyzed theoretically,and a fault diagnosis algorithm based on Compressed Sensing theory is proposed.Firstly,the influence of K-SVD dictionary learning algorithm parameters on sparse expression ability of bearing signals is analyzed,and the determination method of dictionary learning related parameters is determined.Then,the method of determining the measurement matrix suitable for compressing bearing signal is studied,and the compression measurement of the original signal is completed by selecting the optimal measurement matrix.In addition,an improved J-SAMP algorithm is proposed,which solves the problems that the correlation function of the traditional SAMP algorithm can not distinguish atoms in the same direction with different length,and the termination conditions of iteration are not for noisy signals to a certain extent.Finally,a fault diagnosis process using compressed test signal and compressed dictionary directly is proposed,which solves the problem of strong subjectivity in feature selection in fault diagnosis.The proposed algorithm is used to diagnose the faults of the measured signals of bearings,and the good diagnosis results verify the effectiveness of the proposed algorithm.