Research On Fault Diagnosis Method Of Axle Box Bearing Of EMU

With the continuous expansion of high-speed railway scale,higher requirements are put forward for the reliability and safety of each structural component of EMus.As one of the important components of the bogie of the EMU,the axle box bearing can convert the turning motion of the bogie wheelset into the train running along the track direction,and also plays the role of carrying the load of the train body.Under the coupling action of high-speed rotation and heavy load for a long time,the axle box bearing is easy to fail because of its complex load and harsh working environment.Once the axle box bearing failure,it will seriously affect the safety of the whole EMU.Therefore,it is of great practical significance to improve the accuracy of train axle box bearing fault diagnosis to ensure the safe operation of train.In this paper,the axle box bearing of the EMU is taken as the research object,and various signal analysis methods developed at present are studied.A fault diagnosis process based on Sparrow search algorithm(SSA)optimization of variational mode decomposition(VMD)and probabilistic neural network(PNN)is proposed to improve the efficiency and accuracy of fault diagnosis.The main work completed is as follows:(1)The structural composition,fault types and vibration mechanism of rolling bearings are analyzed,and the calculation formula of fault characteristic frequency of rolling bearings is derived,which lays the theoretical basis for the research of this paper.(2)The basic principles of EMD,LMD and VMD signal decomposition algorithms are introduced,and the advantages and disadvantages of the three algorithms are analyzed by comparing simulation experiments to determine the use of VMD algorithm for signal processing.Secondly,aiming at the problem that the decomposition number K and penalty factor are difficult to determine in VMD method,SSA algorithm is proposed to optimize the related parameters of VMD.The experimental results show that using SSA algorithm to optimize VMD can effectively avoid the mode aliasing problem.(3)In order to extract bearing fault information more comprehensively,the kurtosis coefficient is added to select the most relevant IMF component and calculate its corresponding multi-scale percolation entropy,which is used to construct fault feature vectors.Two neural network models,RBF and PNN,are established respectively for bearing fault diagnosis.Fault diagnosis rates of the two models are compared.PNN neural network with high accuracy is selected as the basic model to carry out the research.In this paper,genetic algorithm and sparrow search algorithm are used to optimize the smoothness coefficient of PNN neural network respectively,and the extracted fault feature vector is input into the optimized PNN neural network model to train the neural network.Through comparison experiment,the superiority of sparrow search algorithm to PNN smoothing parameter optimization is verified.(4)In order to verify the effectiveness of the proposed method,through the measured signal analysis of the axle box bearing of the EMU,the three aspects of vibration signal processing,fault feature extraction and fault pattern recognition were respectively studied,and compared with other fault diagnosis methods,the high fault diagnosis rate proved that the proposed method was effective.